Browse the corpus

reater relative reductions in incidence are achieved at 10 years than for prevalence, with the relative impact on incidence being 101%–130% greater than the impact on prevalence for OST and HCNSP scale-up to 60% coverage (Supplementary Figure 2). In contrast, treatment alone has equal impact on incidence as prevalence. Scaling Up From Baseline With IFN-free DAAs Results simulating the likely impact of new IFN-free DAAs are shown in Figures 1B and 2D–F and suggest that approximately 30% fewer treatments are necessary than with peg-IFN + RBV to halve prevalence within 10 years. Increasing OST and HCNSP coverage to 40% may only require annual treatment rates of 7 (95% CrI, 6–10), 16 (95% CrI, 14–21), and 29 (95% CrI, 25–39) per 1000 PWID for 20%, 40%, or 60% baseline chronic prevalences, respectively, to achieve a halving of prevalence in 10 years.

By December 2010, 6.65 million human immunodeficiency virus (HIV)–infected adults and children in low- and middle-income countries were receiving antiretroviral therapy (ART) [1]; however, this was only 47% of those in need. Adults in these settings have an excess mortality risk during the first 2–3 months on ART compared with those in high-income settings, even after adjusting for important cofactors (including CD4 count, age, sex, and ART regimen), with more-similar mortality risks thereafter [2]. Although early mortality has declined in low-income settings over the last decade, this has mainly been driven by fewer severely immunocompromised individuals starting ART [3–5]; for example, in 7 South African programs [4], 12-month mortality declined from 9% to 6% during 2002–2007 as the median pre-ART CD4 count increased from 69 to 113 cells/μL. Nevertheless, early mortality remained approximately 6-fold higher in those initiating ART with CD4 <50 cells/μL vs >200 cells/μL even in the latest period [4].

in 7 South African programs [4], 12-month mortality declined from 9% to 6% during 2002–2007 as the median pre-ART CD4 count increased from 69 to 113 cells/μL. Nevertheless, early mortality remained approximately 6-fold higher in those initiating ART with CD4 <50 cells/μL vs >200 cells/μL even in the latest period [4]. Although 2010 World Health Organization (WHO) guidelines recommend ART at a CD4 count threshold of <350 cells/μL [6], patients continue to present with low CD4 counts or to refuse ART at higher CD4 counts [7]. If only 15% of the 7.5 million untreated HIV-infected persons in need [1] have CD4 <50 cells/μL and experience 7% early mortality, this corresponds to 78 750 deaths in this subgroup alone. However, few data inform how excess early mortality might be reduced [8]. While several adult studies have demonstrated substantial declines in mortality risk on ART [2, 9–13], time has been variably and arbitrarily categorized [14], preventing precise exploration of how mortality risk changes following ART initiation. Furthermore, few studies have addressed this question in children, and only 1 study has compared mortality risks with those of adults [15]. We therefore investigated how mortality risk changed over time in adults and children starting ART, using flexible statistical models for these risks, and explored possible reasons that might inform rational interventions. We used pooled data from 2 large African adult and pediatric trials of ART management strategies, in which mortality ascertainment was near complete.

over time in adults and children starting ART, using flexible statistical models for these risks, and explored possible reasons that might inform rational interventions. We used pooled data from 2 large African adult and pediatric trials of ART management strategies, in which mortality ascertainment was near complete. METHODS Mortality during the first year on ART was estimated in HIV-infected adults (aged 18–73 years) and children (aged 4 months to 15 years) enrolled in the Development of Antiretroviral Therapy in Africa (DART) [16] and Antiretroviral Research for Watoto (ARROW; www.arrowtrial.org) trials, respectively. Both trials recruited previously untreated individuals (except to prevent mother-to-child-transmission) from 3 centers in Uganda and 1 in Zimbabwe. In DART, adults had a CD4 count <200 cells/μL and symptomatic (WHO stage 2/3/4) HIV disease; children in ARROW met WHO 2006 pediatric guidelines for ART [17] due to immunosuppression (age <12 months: CD4% <25%; age 1 to <3 years: <20%; age 3 to <5 years: <15% [or CD4 <350 cells/μL]; age ≥5 years: <15% [or CD4 <200 cells/μL]) and/or WHO 3/4 stage disease.

cells/μL and symptomatic (WHO stage 2/3/4) HIV disease; children in ARROW met WHO 2006 pediatric guidelines for ART [17] due to immunosuppression (age <12 months: CD4% <25%; age 1 to <3 years: <20%; age 3 to <5 years: <15% [or CD4 <350 cells/μL]; age ≥5 years: <15% [or CD4 <200 cells/μL]) and/or WHO 3/4 stage disease. The primary randomized comparison in both trials was clinically driven monitoring (CDM) vs routine laboratory plus clinical monitoring (LCM) for toxicity (hematology/biochemistry) and efficacy (CD4 counts); no real-time HIV loads were assayed. As there was no evidence of mortality or disease progression differences between randomized groups during the first 2 years in DART [16], and no plausible mechanism for early differences in mortality other than toxicity-related deaths (see Results below), groups were pooled for analysis of 12-month mortality on ART; subsequent deaths and follow-up were censored at this time point. At enrollment in 2003–2004, DART participants received 3-drug ART (coformulated zidovudine/lamivudine plus tenofovir, abacavir, or nevirapine), open-label except for 600 participants randomized to abacavir vs nevirapine [18]. In 2007–2008, ARROW participants received open-label 3- or 4-drug ART (abacavir/lamivudine plus a nonnucleoside reverse transcriptase inhibitor [NNRTI] or zidovudine + NNRTI); participants starting 4 drugs reduced to 3 drugs at 36 weeks.

open-label except for 600 participants randomized to abacavir vs nevirapine [18]. In 2007–2008, ARROW participants received open-label 3- or 4-drug ART (abacavir/lamivudine plus a nonnucleoside reverse transcriptase inhibitor [NNRTI] or zidovudine + NNRTI); participants starting 4 drugs reduced to 3 drugs at 36 weeks. Participants in both studies saw a doctor and underwent full blood count, lymphocyte subset (CD4, CD8), and biochemistry testing (bilirubin, urea, creatinine, alanine aminotransferase, and aspartate aminotransferase) every 12 weeks; they were reviewed by a nurse using a standard symptom checklist every 4 weeks. All LCM results were returned to clinicians, whereas CDM results were only returned if requested for clinical reasons (not CD4) or grade 4 laboratory toxicity. For all participants, diagnostic investigations and other tests (except CD4/lymphocytes for CDM) could be requested, concomitant medications prescribed, and antiretrovirals substituted for adverse events. Switching for “failure” before 48 weeks was discouraged [17, 19]. In both trials, participants missing visits were contacted by phone or by fieldworker visit. All reported WHO stage 4 events and deaths were reviewed by an endpoint review committee (ERC) with independent chair and members, who assigned cause of death blinded to randomization.

FN + RBV to halve prevalence within 10 years. Increasing OST and HCNSP coverage to 40% may only require annual treatment rates of 7 (95% CrI, 6–10), 16 (95% CrI, 14–21), and 29 (95% CrI, 25–39) per 1000 PWID for 20%, 40%, or 60% baseline chronic prevalences, respectively, to achieve a halving of prevalence in 10 years. Scaling Up From 20% or 50% Baseline Coverage of OST and HCNSP With peg-IFN + RBV Figure 1C shows the required levels of intervention scale-up necessary for halving chronic prevalence within 10 years with 50% OST and HCNSP at baseline. For example, prevalence can be halved within 10 years by increasing OST and HCNSP coverage from 50% to 70% and annually treating 12 (95% CrI, 11–14), 27 (95% CrI, 24–34), and 48 (95% CrI, 42–59) per 1000 PWID for the 20%, 40%, or 60% chronic prevalence scenarios, respectively. Scaling up OST and HCNSP from already moderate or high coverage levels also leads to greater reductions in the number of antiviral HCV treatments required to achieve chronic HCV prevalence reductions as compared to scale-up from no OST or HCNSP at baseline. At 20% coverage of OST and HCNSP at baseline, achieving >40% prevalence reduction within 10 years always requires scale-up of antiviral treatment, whereas at 50% baseline coverage, treatment is always required to achieve >30% prevalence reductions at 10 years (Supplementary Figure 3).

before 48 weeks was discouraged [17, 19]. In both trials, participants missing visits were contacted by phone or by fieldworker visit. All reported WHO stage 4 events and deaths were reviewed by an endpoint review committee (ERC) with independent chair and members, who assigned cause of death blinded to randomization. To estimate a continuously varying death rate (hazard), we used flexible parametric models [20, 21] counting time from ART initiation to earliest of death, loss to follow-up, or 1 year (see Supplementary Methods). Pre-ART CD4 was categorized by absolute CD4 counts in participants aged ≥4 years and CD4 cell percentage (CD4%) in those aged <4 years, given similar predictive ability of CD4 in untreated children aged ≥4 years and adults [22, 23]. Prespecified categories were 0–49 cells/μL or 0%–4%; 50–99 cells/μL or 5%–9%; 100–149 cells/μL or 10%–14%; and ≥150 cells/μL or ≥15%. The impact of CD4/CD4%, age, sex, WHO stage, and cotrimoxazole at ART initiation was investigated in multivariable models. A second multivariable model also included pre-ART hemoglobin and body mass index (BMI) converted into BMI-for-age z scores following WHO guidelines [24, 25]. Norms at 19 years were used for adults (z score = 0 at BMI = 22.2 [men] and 21.4 [women]). Incidence of different causes of death were compared between adults and children using cause-specific hazards (competing risk subhazards [26] were similar, data not shown). Exact tests were used where no child deaths meant these could not be estimated. All analyses were performed using Stata software version 11.2.

n]). Incidence of different causes of death were compared between adults and children using cause-specific hazards (competing risk subhazards [26] were similar, data not shown). Exact tests were used where no child deaths meant these could not be estimated. All analyses were performed using Stata software version 11.2. RESULTS A total of 3316 adults (aged 18–73 years) and 1206 children (aged 4 months to 17 years) initiated ART in DART and ARROW, respectively, in the same Zimbabwean/Ugandan centers. As there were few older adolescents, participants aged 16–17 years (n = 7) were excluded from analysis. More adults than children had severe immunodeficiency (Table 1), likely reflecting earlier accrual (2003–2004 vs 2007–2008) and more stringent DART eligibility criteria of CD4 <200 cells/μL and symptomatic HIV disease. Of 428 children aged 4–15 years with CD4 ≥200 cells/μL, 46% had <350 cells/μL and 48% had a CD4% of <15%, reflecting some discrepancies between CD4/CD4% in children [23]; remaining children with neither (36%) initiated ART for WHO 3/4 events. Table 1. Characteristics at Antiretroviral Therapy Initiation

ic HIV disease. Of 428 children aged 4–15 years with CD4 ≥200 cells/μL, 46% had <350 cells/μL and 48% had a CD4% of <15%, reflecting some discrepancies between CD4/CD4% in children [23]; remaining children with neither (36%) initiated ART for WHO 3/4 events. Table 1. Characteristics at Antiretroviral Therapy Initiation Factor DART (N = 3316) ARROW 4–15 y (n = 738) ARROW 0–3 y (n = 461) Center Entebbe, Uganda 1020 (31%) 135 (18%) 52 (11%) IDI/PIDC, Uganda 300 (9%) 194 (26%) 118 (26%) JCRC, Uganda 997 (30%) 286 (39%) 114 (25%) Harare, Zimbabwe 999 (30%) 123 (17%) 177 (38%) Women/girls 2156 (65%) 369 (50%) 236 (51%) Age, y 36 (31–42) 8 (6–10) 1 (1–2) Pre-ART CD4, cells/μL 86 (31–139) 251 (95–398) 725 (471–1081) 0–49 (0%–4% if <4 y) 1109 (33%) 131 (18%) 27 (6%) 50–99 (5%–9% if <4 y) 785 (24%) 56 (8%) 87 (19%) 100–149 (10%–14% if <4 y) 759 (23%) 52 (7%) 128 (28%) 150–199 (15%–19% if <4 y) 663 (20%) 71 (10%) 98 (21%) ≥200 (≥20% if <4 y) 0a 428 (58%) 121 (26%) WHO stage 1/2 673 (20%) 214 (29%) 136 (30%) 3 1864 (56%) 444 (60%) 235 (51%) 4 779 (23%) 80 (11%) 90 (20%) Weight, kg 57 (50–64) 20 (17–26) 9 (7–11) BMI, kg/m2 21.1 (19.1–23.6) 15.0 (14.0–15.9) 15.6 (14.2–16.7) BMI-for-age z score (WHOb) −0.2 (−1.0 to 0.6) −0.8 (−1.5 to −0.1) −0.2 (−1.6 to 0.7) Hemoglobin, g/dL 11.4 (10.3–12.7) 11.1 (10.2–11.9) 9.9 (9.2–10.7) On cotrimoxazole prophylaxis 2048c (62%) 735c (99.6%) 460d (99.8%) First ART regimen ZDV/3TC/TDF 2469 (74%) 0 0 ZDV/3TC/ABC 300 (9%) 0 0 ZDV/3TC/NVP 547 (16%) 0 0 3TC/ABC/EFV 0 128 (17%) 11e (2%) 3TC/ABC/NVP 0 111 (15%) 143 (31%) ZDV/3TC/ABC/EFV 0 284 (38%) 19e (4%) ZDV/3TC/ABC/NV 0 215 (29%) 288 (62%) Data are presented as No. (%) or median (interquartile range). Weight and BMI not available pre-ART for 23 and 33 DART participants, respectively.

%) 0 0 ZDV/3TC/NVP 547 (16%) 0 0 3TC/ABC/EFV 0 128 (17%) 11e (2%) 3TC/ABC/NVP 0 111 (15%) 143 (31%) ZDV/3TC/ABC/EFV 0 284 (38%) 19e (4%) ZDV/3TC/ABC/NV 0 215 (29%) 288 (62%) Data are presented as No. (%) or median (interquartile range). Weight and BMI not available pre-ART for 23 and 33 DART participants, respectively. Abbreviations: 3TC, lamivudine; ABC, abacavir; ARROW, Antiretroviral Research for Watoto; ART, antiretroviral therapy; BMI, body mass index; DART, Development of Antiretroviral Therapy in Africa; EFV, efavirenz; IDI/PIDC, Infectious Diseases Institute/Pediatric Infectious Diseases Clinic (Mulago Hospital); JCRC, Joint Clinical Research Center; NVP, nevirapine; TDF, tenofovir; WHO, World Health Organization; ZDV, zidovudine. a All DART participants had CD4 <200 cells/μL at ART initiation as trial entry criterion. b For adults, calculated using WHO references [24], assuming age 19 years. c In addition, 3 (0.1%) adults and 3 (0.4%) older children were taking dapsone prophylaxis at ART initiation. d One child had grade 2 neutropenia at ARROW enrollment and initiated cotrimoxazole 4 months later. e No dosing available for EFV in children <3 years of age or <15 kg.

b For adults, calculated using WHO references [24], assuming age 19 years. c In addition, 3 (0.1%) adults and 3 (0.4%) older children were taking dapsone prophylaxis at ART initiation. d One child had grade 2 neutropenia at ARROW enrollment and initiated cotrimoxazole 4 months later. e No dosing available for EFV in children <3 years of age or <15 kg. Thirty-eight (1.1%) adults and 9 (0.8%) children had unknown vital status at 1 year. In contrast to many programs [27], pre-ART CD4 counts were similar between those lost to follow-up and those followed up at 1 year (adults: median 117 vs 88 cells/μL, P = .26 [vs 33 cells/μL among deaths in year 1]; children: 501 vs 368 cells/μL, P = .30 [vs 103 cells/μL among deaths in year 1]). Participants lost to follow-up were therefore censored at last clinic attendance (median 26 weeks [adults] and 20 weeks [children]). Eleven (0.3%) adults and 3 (0.3%) children switched to second-line ART (all but 1 adult >46 weeks).

r 1]; children: 501 vs 368 cells/μL, P = .30 [vs 103 cells/μL among deaths in year 1]). Participants lost to follow-up were therefore censored at last clinic attendance (median 26 weeks [adults] and 20 weeks [children]). Eleven (0.3%) adults and 3 (0.3%) children switched to second-line ART (all but 1 adult >46 weeks). Overall, 179 (5.4%) adults and 39 (3.3%) children died in the first year following initiation of ART. As expected, mortality varied strongly with pre-ART CD4 (P < .0001 for adults and older children), with no statistical evidence of variation in younger children (P = .70). Interestingly, mortality risk was similar in adults and older children (≥4 years) in the same CD4 strata (Figure 1). Risk was also similar in younger children (<4 years) in parallel CD4% strata, although power was low to detect genuine differences. Results were similar with a 5-year threshold [19] and in older children with CD4 <350 cells/μL (data not shown). Risk gradients were much stronger <100 cells/μL (or <10% for age <4 years); above this, differences were smaller with no evidence supporting important variation (P > .6). Subsequent analyses therefore pooled CD4 >100 cells/μL. Figure 1. Kaplan-Meier mortality 1 year after antiretroviral therapy (ART) initiation according to age and pre-ART CD4 count. Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; DART, Development of Antiretroviral Therapy in Africa.

sequent analyses therefore pooled CD4 >100 cells/μL. Figure 1. Kaplan-Meier mortality 1 year after antiretroviral therapy (ART) initiation according to age and pre-ART CD4 count. Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; DART, Development of Antiretroviral Therapy in Africa. Figure 2 shows how mortality risk varies day-by-day over the first year on ART. At all CD4/CD4%, and in adults and children, mortality risk increased from enrollment to a maximum between 30 and 50 days after ART initiation, declined rapidly to 180 days, then declined more slowly. In both adults and children, half and three-quarters of the deaths occurred in the first 3 and 6 months, respectively. The sharp initial risk increase is likely because of trial consent (excluding moribund patients); the earliest deaths occurred on day 8 (DART) and day 16 (ARROW). In sensitivity analyses, assuming the 1% of participants lost to follow-up had died, differences between groups were similar. Pooling data from adults and children, there was no evidence for a differential effect of pre-ART CD4 with age (heterogeneity P = .95 [0–49 cells/μL, 0%–4%]; P = .98 [50–99 cells/μL, 5%–9%]; and P = .15 [≥100 cells/μL, ≥10%]). Figure 2. Daily risk of death and survival through 1 year on antiretroviral therapy (ART) according to age and pre-ART CD4 count. Flexible parametric model [20, 21] on log-normal scale with 1 interior knot. Points show times when deaths occurred. Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; DART, Development of Antiretroviral Therapy in Africa; PY, person-years.

RT) according to age and pre-ART CD4 count. Flexible parametric model [20, 21] on log-normal scale with 1 interior knot. Points show times when deaths occurred. Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; DART, Development of Antiretroviral Therapy in Africa; PY, person-years. Adjusting for CD4/CD4%, mortality risks were lower in those with earlier pre-ART WHO stage or on cotrimoxazole at ART initiation (P < .0001), but there was no evidence of additional effects of age, sex, or center (P > .16; Supplementary Table 1). After adjusting for increased mortality risk with lower pre-ART hemoglobin and BMI (P < .0001), there was weak evidence of slightly increased mortality risk in older individuals on ART (P = .02). There was no evidence supporting a large contribution of ART (or other drug) toxicity to mortality on ART during the first 3 months or the first year (Table 2). Only 6% of adult and 3% of child year-1 deaths were ERC-adjudicated as primarily medication-related (4% and 3% primarily ART-related, respectively; Table 2). ART-related deaths were from septicemia (4 adults, 1 child), neutropenia without sepsis (2 adults), anemia (1 adult), and hepatitis (1 adult). Sixty percent of primarily medication-related deaths occurred in those with pre-ART CD4 <50 cells/μL, similar to deaths that were uncertainly medication-related (62%) and primarily HIV-related (68%). Table 2. Causes of Death in the First Year on Antiretroviral Therapy

adults), anemia (1 adult), and hepatitis (1 adult). Sixty percent of primarily medication-related deaths occurred in those with pre-ART CD4 <50 cells/μL, similar to deaths that were uncertainly medication-related (62%) and primarily HIV-related (68%). Table 2. Causes of Death in the First Year on Antiretroviral Therapy Median (IQR) d From ART Initiation to Death Deaths Within 3 mo of ART Initiation DART 18–73 y (n = 179) ARROW 4 mo to15 y (n = 39) Crude Cause-Specific (ARROW:DART) HR (95% CI), P Value Adjusteda Cause-Specific (ARROW:DART) HR (95% CI), P Value DARTb ARROWb DART (n = 90) ARROW (n = 20) Relationship to HIV and drugs Primarily HIV related 92 (51%) 28 (72%) .83 (.54–1.27), .39 1.34 (.86–2.09), .19 69 (32–160) 86 (42–153) 50 (56%) 15 (75%) Primarily medication related 10c (6%) 1d (3%) .27 (.04–2.14), .22 .43 (.05–3.54), .44 53 (25–75) 39 (29–47) 9 (10%) 0 Uncertain whether primarily HIV or medication related 24e (13%) 5f (13%) .57 (.22–1.49), .25 .91 (.34–2.47), .85 81(47–128) … 13 (14%) 5 (25%) Uncertain whether HIV related or not, but not medication related 1 (1%) 1 (3%) … … … … 1 (1%) 0 Uncertain whether medication related or not, but not HIV related 1e(1%) 0 … … … … 0 0 Unlikely to be HIV or medication related 17 (9%) 1 (3%) .16 (.02–1.20), .08 .16 (.02–1.26), .08 141 (86–216) … 6 (7%) 0 Relationship to HIV/medications could not be determined 34 (19%) 3 (8%) .24 (.07–.78), .02 .29 (.09–.97), .04 138 (72–216) 128 (107–205) 11 (12%) 0 Cause of death Septicemia/meningitisg 36 (20%) 14 (36%) 1.06 (.57–1.97), .84 1.50 (.78–2.86), .22 76 (39–136) 79 (51–126) 20 (22%) 9 (45%) Unknown causeh 33 (18%) 3 (8%) .25 (.08–.81), .02 .35 (.10–1.16), .09 117 (47–207) 128 (107–205) 14 (16%) 0 Extrapulmonary cryptococcosis 20i (11%) 0 … … 50 (28–109) … 14 (16%) 0 Other non-WHO stage 4 brain disease 16j (9%) 0 … … 88 (56–188) … 8 (9%) 0 Tuberculosis 14 (8%) 1 (3%) .19 (.03–1.48), .11 .29 (.04–2.32), .25 72 (36–143) … 8 (9%) 0 Pulmonary 9 (5%) 0 6 (7%) 0 Extrapulmonary 5 (3%) 1 (3%) 2 (2%) 0 Pneumoniag 10 (6%) 11 (28%) 3.01 (1.28–7.09), .01 4.72 (1.91–11.7), .001 34 (28–274) 41 (29–138) 6 (7%) 8 (40%) Other WHO 4 OIs (toxoplasmosis, PCP, CMV, cryptosporidiosis, isosporiasis) 8 (4%) 1 (3%) .34 (.04–2.74), .31 .67 (.08–5.66), .72 49 (40–88) … 6 (7%) 1 (5%) Wasting, diarrhea, gastrointestinal 6 (3%) 3 (8%) 1.36 (.34–5.45), .66 2.83 (.68–11.8), .15 171 (69–283) 100 (39–107) 3 (3%) 1 (5%) AIDS-defining malignancy (KS, lymphoma) 6 (3%) 0 … … 179 (119–206) … 0 0 Anemia, neutropenia, thromb

iasis) 8 (4%) 1 (3%) .34 (.04–2.74), .31 .67 (.08–5.66), .72 49 (40–88) … 6 (7%) 1 (5%) Wasting, diarrhea, gastrointestinal 6 (3%) 3 (8%) 1.36 (.34–5.45), .66 2.83 (.68–11.8), .15 171 (69–283) 100 (39–107) 3 (3%) 1 (5%) AIDS-defining malignancy (KS, lymphoma) 6 (3%) 0 … … 179 (119–206) … 0 0 Anemia, neutropenia, thromb ocytopenia without sepsis 6 (3%) 0 … … 78 (59–102) … 3 (3%) 0 Hepatic 6 (3%) 0 … … 212 (75–339) … 2 (2%) 0 Trauma, obstetric, suicide 6 (3%) 1 (3%) .45 (.05–3.76), .46 .30 (.04–2.48), .26 195 (181–216) … 1 (1%) 0 Other lung disease 3k (2%) 0 … … 95 (81–141) … 1 (1%) 0 Malaria 2 (1%) 1 (3%) … … … … 1 (1%) 0 Renal (non-AIDS) 2l (1%) 0 … … … … 0 0 Cerebrovascular disease 0 3 (8%) … … … 137 (17–186) 0 1 (3%) Other single causes in either trial 5m (2%) 1n (3%) … … … … 3 (3%) 0 Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; CI, confidence interval; CMV, cytomegalovirus; DART, Development of Antiretroviral Therapy in Africa; HIV, human immunodeficiency virus; HR, hazard ratio; IQR, interquartile range; KS, Kaposi sarcoma; OI, opportunistic infection; PCP, pneumocystis pneumonia; WHO, World Health Organization. a Adjusted for CD4/CD4% categories (0–49 cells/μL, 0%–4%; 50–99 cells/μL, 5–9%; ≥100 cells/μL, ≥10%). Sub–hazard ratios corresponding to the cumulative incidence [26] were similar (data not shown). b For groups with 3 or more deaths.

ocytopenia without sepsis 6 (3%) 0 … … 78 (59–102) … 3 (3%) 0 Hepatic 6 (3%) 0 … … 212 (75–339) … 2 (2%) 0 Trauma, obstetric, suicide 6 (3%) 1 (3%) .45 (.05–3.76), .46 .30 (.04–2.48), .26 195 (181–216) … 1 (1%) 0 Other lung disease 3k (2%) 0 … … 95 (81–141) … 1 (1%) 0 Malaria 2 (1%) 1 (3%) … … … … 1 (1%) 0 Renal (non-AIDS) 2l (1%) 0 … … … … 0 0 Cerebrovascular disease 0 3 (8%) … … … 137 (17–186) 0 1 (3%) Other single causes in either trial 5m (2%) 1n (3%) … … … … 3 (3%) 0 Abbreviations: ART, antiretroviral therapy; ARROW, Antiretroviral Research for Watoto; CI, confidence interval; CMV, cytomegalovirus; DART, Development of Antiretroviral Therapy in Africa; HIV, human immunodeficiency virus; HR, hazard ratio; IQR, interquartile range; KS, Kaposi sarcoma; OI, opportunistic infection; PCP, pneumocystis pneumonia; WHO, World Health Organization. a Adjusted for CD4/CD4% categories (0–49 cells/μL, 0%–4%; 50–99 cells/μL, 5–9%; ≥100 cells/μL, ≥10%). Sub–hazard ratios corresponding to the cumulative incidence [26] were similar (data not shown). b For groups with 3 or more deaths. c Eight primarily ART related: 7 zidovudine (1 alone, 2 + cotrimoxazole, 2 + tenofovir, 1 + sulphadiazine/pyrimethamine, 1 + sulphamethoxazole), 1 nevirapine; 2 primarily other medication related only: 1 rifampicin/isoniazid/ethambutol/pyrazinamide, 1 dapsone. d Primarily ART related: zidovudine + cloxacillin + ceftriaxone.

c Eight primarily ART related: 7 zidovudine (1 alone, 2 + cotrimoxazole, 2 + tenofovir, 1 + sulphadiazine/pyrimethamine, 1 + sulphamethoxazole), 1 nevirapine; 2 primarily other medication related only: 1 rifampicin/isoniazid/ethambutol/pyrazinamide, 1 dapsone. d Primarily ART related: zidovudine + cloxacillin + ceftriaxone. e Twenty-four uncertain whether primarily ART related (or HIV related): 21 zidovudine (12 alone, 3 + cotrimoxazole, 1 + tenofovir, 1 + tenofovir + amoxicillin + paracetamol, 1 + tenofovir + ciprofloxacin + diclofenac, 1 + tenofovir + dexamethasone + carbamazepine, 1 + tenofovir + rifampicin + isoniazid, 1 + tenofovir + rifampicin + isoniazid + ethambutol + pyrazinamide + ceftriaxone), 2 nevirapine, 1 stavudine; 1 uncertain whether primarily other medication related (or HIV related): 1 fluconazole. f Four uncertain whether primarily ART related (or HIV related): 4 zidovudine (2 alone, 2 + cotrimoxazole); 1 uncertain whether primarily other medication related (or HIV related): cotrimoxazole. g Organisms isolated from bacterial infections (blood cultures unless stated): DART: Streptococcus pneumoniae (3), Escherichia coli (4, plus 1 with urinary E. coli only), Staphylococcus aureus (2). ARROW: S. pneumoniae (3), S. aureus (1), Pseudomonas aeruginosa (1), Klebsiella pneumoniae + Enterococcus spp (1), plus 1 with urinary K. pneumoniae only and 1 with Salmonella spp isolated from stool. h Primary cause of death could not be determined (eg, because the patient died at home or presented very sick without time for diagnostic tests).

g Organisms isolated from bacterial infections (blood cultures unless stated): DART: Streptococcus pneumoniae (3), Escherichia coli (4, plus 1 with urinary E. coli only), Staphylococcus aureus (2). ARROW: S. pneumoniae (3), S. aureus (1), Pseudomonas aeruginosa (1), Klebsiella pneumoniae + Enterococcus spp (1), plus 1 with urinary K. pneumoniae only and 1 with Salmonella spp isolated from stool. h Primary cause of death could not be determined (eg, because the patient died at home or presented very sick without time for diagnostic tests). i Nineteen cryptococcal meningitis, 1 cryptococcemia. jParticipant fulfilled clinical criteria for at least 1 of cerebral toxoplasmosis, cryptococcal meningitis, tuberculosis meningitis, or progressive multifocal leukoencephalopathy, but no diagnostic tests were done and/or the patient failed to respond to first-line treatment and died without further investigations. k2 pulmonary embolus; 1 chronic obstructive pulmonary disease. l Glomerulonephritis, chronic renal failure in a patient with type 1 diabetes and hypertension. m Stevens-Johnson syndrome, diabetes, lactic acidosis, non-AIDS cancer (carcinomatosis), cardiomyopathy. n HIV encephalopathy.

jParticipant fulfilled clinical criteria for at least 1 of cerebral toxoplasmosis, cryptococcal meningitis, tuberculosis meningitis, or progressive multifocal leukoencephalopathy, but no diagnostic tests were done and/or the patient failed to respond to first-line treatment and died without further investigations. k2 pulmonary embolus; 1 chronic obstructive pulmonary disease. l Glomerulonephritis, chronic renal failure in a patient with type 1 diabetes and hypertension. m Stevens-Johnson syndrome, diabetes, lactic acidosis, non-AIDS cancer (carcinomatosis), cardiomyopathy. n HIV encephalopathy. A large range of causes of death was observed in adults and children, the most common being septicemia/meningitis, with pneumonia also a common child cause of death (Table 2). Causes of death in the first 3 months were similar to the first year overall. Organisms were identified in relatively few cases but were typical for the setting, including Streptococcus pneumoniae, Escherichia coli, Staphylococcus aureus, and Klebsiella pneumoniae. Cryptococcus accounted for 11% of adult deaths, but no child deaths (exact P = .03 vs noncryptococcal cause); tuberculosis accounted for 8% vs 3% of deaths, respectively (adjusted P = .25). The only strong evidence for differing incidence of death between adults and children was from pneumonia (4.72-fold higher risk in children, P = .001; 3.73-fold higher risk pooling all respiratory-related deaths, P = .002). Uncertain or unlikely relationship to HIV/drugs and uncertain cause of death tended to be reported less frequently in children (adjusted P = .04, .08, and .09, respectively). Of causes with ≥10 adult deaths, the earliest were from pneumonia (median 34 days on ART), then cryptococcal disease (median 50 days), tuberculosis (median 72 days), septicemia/meningitis (median 76 days), other severe brain disease (median 88 days), and unknown causes (median 117 days). Child deaths from pneumonia and septicemia/meningitis occurred at similar timescales (median 41 and 79 days, respectively).

hen cryptococcal disease (median 50 days), tuberculosis (median 72 days), septicemia/meningitis (median 76 days), other severe brain disease (median 88 days), and unknown causes (median 117 days). Child deaths from pneumonia and septicemia/meningitis occurred at similar timescales (median 41 and 79 days, respectively). One question is whether high early mortality risks merely reflected carryover effects of late presentation, with a delay in ART effectiveness in those with low pre-ART CD4 counts. We therefore estimated mortality risks over the first year after enrollment in adults with CD4 <200 cells/μL in the Entebbe cohort [28] and children in the 3Cs4kids cohort collaboration [29] who presented for care with low CD4 but did not receive ART, using the same CD4 strata and models as above (see Supplementary Methods for cohort details). Without ART, 40% (206/514) of adults and 9% (126/1377) of children died during the first year after enrollment; no adults, but 451 (33%) children, actually initiated ART during this first year and were censored. Figure 3 shows that high mortality risks were similar in the first 30 days after enrollment to early risks on ART. Without ART, risks remained high, particularly for adults, whereas those starting ART experienced rapid drops in risk. Figure 3. Daily risk of death and survival through 1 year before and on antiretroviral therapy (ART). Flexible parametric model [20, 21] on log-normal scale with 1 interior knot. Points show times when deaths occurred. Fewer than 40 children aged 1–3 years with 0%–4% pre-ART CD4 count were enrolled in the 3Cs4kids study (data not shown). Abbreviations: ART, antiretroviral therapy; PY, person-years.

viral therapy (ART). Flexible parametric model [20, 21] on log-normal scale with 1 interior knot. Points show times when deaths occurred. Fewer than 40 children aged 1–3 years with 0%–4% pre-ART CD4 count were enrolled in the 3Cs4kids study (data not shown). Abbreviations: ART, antiretroviral therapy; PY, person-years. DISCUSSION Although the substantial benefits of ART are clear, it remains uncertain how best to reduce excess early mortality in severely immunocompromised HIV-infected individuals initiating ART in low-income countries [2, 4]. Our main finding is that patterns of changing mortality risk in the first year on ART are indistinguishable in adults and children ≥4 years; both experience similarly high mortality risks in the first 3 months on ART if they have low pre-ART CD4 counts. Children aged <4 years with low CD4% are at similarly high risk for mortality. Nevertheless and importantly, risks on ART are lower than mortality risks without ART. While earlier HIV diagnosis and prompt ART initiation remain key goals, similarities between adults and children call for interventions to target reductions in early mortality in both; simplification and harmonization of management approaches across both groups would be an advantage in most resource-limited settings. Our findings contrast with the only comparison of early mortality in adults and children to date, which suggested somewhat lower overall risks over the first year on ART in 2- to 14-year-olds entering the Malawi national ART program [15]; however, data on CD4, which is the key determinant of early mortality, were not available.

ngs contrast with the only comparison of early mortality in adults and children to date, which suggested somewhat lower overall risks over the first year on ART in 2- to 14-year-olds entering the Malawi national ART program [15]; however, data on CD4, which is the key determinant of early mortality, were not available. In addition to comparing mortality risks in adults and children, we also examined potential mechanisms for increased early mortality to inform future interventions. Possible reasons for an early limited increase in mortality risk after ART initiation include ART toxicity, immune reconstitution inflammatory syndrome (IRIS), or time required for ART to become effective. Our study does not support a major role of ART toxicity; although a small number of deaths were adjudicated as primarily drug-related (mainly to zidovudine [anemia/neutropenia/sepsis] and nevirapine [hepatic failure]), participants initiating ART with low CD4 counts suffered disproportionately from both drug- and HIV-related deaths, suggesting that advanced immunodeficiency at ART initiation, rather than ART alone, played a role in toxicity-related deaths [30]. We did not directly assess the contribution of IRIS because these clinical criteria were not systematically recorded in 2003–2004, when the early deaths occurred in DART. It is pertinent that, even if IRIS was a mechanism, early mortality risks on ART were no greater than observed in similar groups without ART. This supports recent trial findings that, despite increased IRIS, early ART significantly reduced mortality in patients with advanced HIV disease and tuberculosis [31, 32]. The fact that mortality risks in the first weeks on ART were similar to those without ART might also suggest that pre-ART risks persist until ART reaches maximal effectiveness. If this were the case, then increasing ART potency (eg, by adding an integrase inhibitor to initial ART) might benefit patients with advanced immunodeficiency. Integrase inhibitors are obvious candidates for such induction-maintenance strategies as they achieve the most rapid viral load declines [33].

al effectiveness. If this were the case, then increasing ART potency (eg, by adding an integrase inhibitor to initial ART) might benefit patients with advanced immunodeficiency. Integrase inhibitors are obvious candidates for such induction-maintenance strategies as they achieve the most rapid viral load declines [33]. The diverse early causes of death observed in adults and children present a challenge in selecting interventions targeting specific causes. Although certain opportunistic infections (cryptococcus, tuberculosis) made clear contributions, individually their effects were modest, similar to findings from other studies in adults [34]. Although no child died from cryptococcal disease, this is probably owing to low numbers, as deaths have been reported in older children [35–37]. Our findings suggest that any augmented prophylaxis approach to reduce early morbidity/mortality needs to cover multiple organisms. Although fluconazole [38], isoniazid [39], and cotrimoxazole [40, 41] prophylaxes have important benefits, it remains unclear whether toxicity risks associated with simultaneous initiation with 3-drug (or 4-drug) ART would outweigh any potential advantages in patients presenting with very low CD4 counts. Aside from toxicity concerns, the considerably increased pill burden and potential for drug–drug interactions suggests that such strategies need formal evaluation.

associated with simultaneous initiation with 3-drug (or 4-drug) ART would outweigh any potential advantages in patients presenting with very low CD4 counts. Aside from toxicity concerns, the considerably increased pill burden and potential for drug–drug interactions suggests that such strategies need formal evaluation. Invasive bacterial infections (septicemia/meningitis/pneumonia) were the commonest early cause of death in adults and children. Similarly high contributions of bacterial infections to early mortality on ART were described in a comparison of adult cohorts in Brazil and the United States [42]. In contrast to tuberculosis and cryptococcal disease, little attention has been paid to this finding in the wider literature, although septicemia and pneumonia were relatively commonly reported as “other” cause of death in a recent meta-analysis of 1-year mortality in adults [34] and the importance of bacterial infections in child mortality has been noted [43]. Cotrimoxazole prophylaxis, received by all ARROW children and 62% DART adults at ART initiation, has major mortality benefits in HIV-infected persons [40, 41], presumed to occur through reduction of bacterial infections. Given high rates of cotrimoxaozole resistance, our data suggest that additional broad-spectrum antibiotic prophylaxis might also improve outcomes among adults and children presenting with low CD4 counts.

or mortality benefits in HIV-infected persons [40, 41], presumed to occur through reduction of bacterial infections. Given high rates of cotrimoxaozole resistance, our data suggest that additional broad-spectrum antibiotic prophylaxis might also improve outcomes among adults and children presenting with low CD4 counts. As previously reported [12, 34, 44], low BMI and hemoglobin were associated with higher early mortality in adults and children in our study. Of note, BMI associations were present even with mildly abnormal BMI, suggesting a potential role for additional nutritional supplementation in reducing early mortality following ART initiation [45]. This approach could improve drug absorption, known to be impaired in severe HIV disease, and possibly adherence, as many patients report feeling acute hunger after starting ART [46], probably reflecting a profound catabolic state induced by severe HIV infection, which abruptly reverses with ART. However, moderately sized trials in HIV-infected adults [47] and in patients with tuberculosis [48] starting treatment with severe malnutrition (BMI <18.5) have shown no significant effect of ready-to-eat and fortified soya foods on mortality, although early weight, BMI, and CD4 gains were observed. Therefore, the role of supplementation in reducing early mortality in those without severe malnutrition remains unclear.

tarting treatment with severe malnutrition (BMI <18.5) have shown no significant effect of ready-to-eat and fortified soya foods on mortality, although early weight, BMI, and CD4 gains were observed. Therefore, the role of supplementation in reducing early mortality in those without severe malnutrition remains unclear. Our trial data have 3 major advantages over previous studies for investigating early mortality. First, vital status was accurately and completely ascertained in contrast to most programs [27, 34], thus limiting the impact of mortality misclassification. Even the assumption that all those lost to follow-up in the first year had died (1.1% adults, 0.8% children) did not alter results. Second, causes of death were assigned by an independent committee, based on a structured narrative and without knowledge of the randomized group or CD4 (part of the randomizations). Third, we used innovative flexible parametric models to directly assess how mortality risk changed over time on ART. The alternative strategy, categorizing time on ART [2, 10–13, 15, 43, 49, 50], produces risk estimates that change abruptly, making them biologically implausible [14], and may poorly represent the underlying data. In recognition of this problem, 2 recent papers used a piecewise Weibull model (which nevertheless makes strong assumptions about changing risks) [9] or smoothed hazards from semi-parametric Cox models [51].

s that change abruptly, making them biologically implausible [14], and may poorly represent the underlying data. In recognition of this problem, 2 recent papers used a piecewise Weibull model (which nevertheless makes strong assumptions about changing risks) [9] or smoothed hazards from semi-parametric Cox models [51]. WHO guidelines advocate ART initiation in HIV-infected adults and children aged ≥5 years at a CD4 count of <350 cells/μL [6], but substantial numbers of individuals continue to present for care late in HIV infection [7, 52–54]. Given experiences in high-income countries [55, 56], late presentation will continue in ART programs for the foreseeable future. However, the consequences of such late presentation, in terms of early morbidity/mortality on ART, are far more severe in low-income settings. To maximize ART benefits, it is essential to identify which, if any, interventions could reduce high early death rates if given with ART. Recent experiences in a trial of hydroxychloroquine, which significantly increased HIV viral load in ART-naive patients despite being expected to reduce immune activation [57], demonstrate the importance of testing potential interventions in randomized controlled trials. Our findings do not identify any single plausible mechanism; rather, they suggest that anti-HIV, anti-infection (or enhanced prophylaxis), and anti-malnutrition/malabsorption may all be important potential approaches. The REALITY trial (ISRCTN43622374) plans to address these questions in adults and older children in a 2 × 2 × 2 factorial design from 2012. Our findings also demonstrate strong similarities in early mortality patterns between adults and children. Where the same health providers treat both age groups, as in most of sub-Saharan Africa, integrating research will likely provide the most relevant evidence base for future management.

factorial design from 2012. Our findings also demonstrate strong similarities in early mortality patterns between adults and children. Where the same health providers treat both age groups, as in most of sub-Saharan Africa, integrating research will likely provide the most relevant evidence base for future management. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://www.oxfordjournals.org/our_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments.  We thank all the participants and staff from all the centers participating in the Development of Antiretroviral Therapy in Africa (DART) and Antiretroviral Research for Watoto (ARROW) trials. DART Trial: MRC/UVRI Uganda Research Unit on AIDS, Entebbe, Uganda. H. Grosskurth, P. Munderi, G. Kabuye, D. Nsibambi, R. Kasirye, E. Zalwango, M. Nakazibwe, B. Kikaire, G. Nassuna, R. Massa, K. Fadhiru, M. Namyalo, A. Zalwango, L. Generous, P. Khauka, N. Rutikarayo, W. Nakahima, A. Mugisha, J. Todd, J. Levin, S. Muyingo, A. Ruberantwari, P. Kaleebu, D. Yirrell, N. Ndembi, F. Lyagoba, P. Hughes, M. Aber, A. Medina Lara, S. Foster, J. Amurwon, B. Nyanzi Wakholi, J. Whitworth1, K. Wangati1, B. Amuron1, D. Kajungu1, J. Nakiyingi1, W. Omony1, K. Fadhiru1, D. Nsibambi1, P. Khauka1.

ayo, W. Nakahima, A. Mugisha, J. Todd, J. Levin, S. Muyingo, A. Ruberantwari, P. Kaleebu, D. Yirrell, N. Ndembi, F. Lyagoba, P. Hughes, M. Aber, A. Medina Lara, S. Foster, J. Amurwon, B. Nyanzi Wakholi, J. Whitworth1, K. Wangati1, B. Amuron1, D. Kajungu1, J. Nakiyingi1, W. Omony1, K. Fadhiru1, D. Nsibambi1, P. Khauka1. Joint Clinical Research Centre, Kampala, Uganda. P. Mugyenyi, C. Kityo, F. Ssali, D. Tumukunde, T. Otim, J. Kabanda, H. Musana, J. Akao, H. Kyomugisha, A. Byamukama, J. Sabiiti, J. Komugyena, P. Wavamunno, S. Mukiibi, A. Drasiku, R. Byaruhanga, O. Labeja, P. Katundu, S. Tugume, P. Awio, A. Namazzi, G. T. Bakeinyaga, H. Katabira, D. Abaine, J. Tukamushaba, W. Anywar, W. Ojiambo, E. Angweng, S. Murungi, W. Haguma, S. Atwiine, J. Kigozi, L. Namale1, A. Mukose1, G. Mulindwa1, D. Atwiine1, A. Muhwezi1, E. Nimwesiga1, G. Barungi1, J. Takubwa1, S. Murungi1, D. Mwebesa1, G. Kagina1, M. Mulindwa1, F. Ahimbisibwe1, P. Mwesigwa1, S. Akuma1, C. Zawedde1, D. Nyiraguhirwa1, C. Tumusiime1, L. Bagaya1, W. Namara1, J. Kigozi1, J. Karungi1, R. Kankunda1, R. Enzama1.

L. Namale1, A. Mukose1, G. Mulindwa1, D. Atwiine1, A. Muhwezi1, E. Nimwesiga1, G. Barungi1, J. Takubwa1, S. Murungi1, D. Mwebesa1, G. Kagina1, M. Mulindwa1, F. Ahimbisibwe1, P. Mwesigwa1, S. Akuma1, C. Zawedde1, D. Nyiraguhirwa1, C. Tumusiime1, L. Bagaya1, W. Namara1, J. Kigozi1, J. Karungi1, R. Kankunda1, R. Enzama1. University of Zimbabwe, Harare, Zimbabwe. A. Latif1, J. Hakim, V. Robertson, A. Reid, E. Chidziva, R. Bulaya-Tembo, G. Musoro, F. Taziwa, C. Chimbetete, L. Chakonza, A. Mawora, C. Muvirimi, G. Tinago, P. Svovanapasis, M. Simango, O. Chirema, J. Machingura, S. Mutsai, M. Phiri, T. Bafana, M. Chirara, L. Muchabaiwa, M. Muzambi, J. Mutowo, T. Chivhunga, E. Chigwedere1, M. Pascoe1, C. Warambwa1, E. Zengeza1, F. Mapinge1, S. Makota1, A. Jamu1, N. Ngorima1, H. Chirairo1, S. Chitsungo1, J. Chimanzi1, C. Maweni1, R. Warara1, M. Matongo1, S. Mudzingwa1, M. Jangano1, K. Moyo1, L. Vere1, N. Mdege1, I. Machingura1. Infectious Diseases Institute (formerly the Academic Alliance), Makerere University, Mulago, Uganda. E. Katabira, A. Ronald, A. Kambungu, F. Lutwama, I. Mambule, A. Nanfuka, J. Walusimbi, E. Nabankema, R. Nalumenya, T. Namuli, R. Kulume, I. Namata, L. Nyachwo, A. Florence, A. Kusiima, E. Lubwama, R. Nairuba, F. Oketta, E. Buluma, R. Waita, H. Ojiambo, F. Sadik, J. Wanyama, P. Nabongo, J. Oyugi1, F. Sematala1, A. Muganzi1, C. Twijukye1, H. Byakwaga1. The AIDS Support Organisation (TASO), Mulago, Kampala, Uganda. R. Ochai, D. Muhweezi, A. Coutinho1, B. Etukoit1. Imperial College, London, UK. C. Gilks, K. Boocock, C. Puddephatt, C. Grundy, J. Bohannon, D. Winogron1.

Infectious Diseases Institute (formerly the Academic Alliance), Makerere University, Mulago, Uganda. E. Katabira, A. Ronald, A. Kambungu, F. Lutwama, I. Mambule, A. Nanfuka, J. Walusimbi, E. Nabankema, R. Nalumenya, T. Namuli, R. Kulume, I. Namata, L. Nyachwo, A. Florence, A. Kusiima, E. Lubwama, R. Nairuba, F. Oketta, E. Buluma, R. Waita, H. Ojiambo, F. Sadik, J. Wanyama, P. Nabongo, J. Oyugi1, F. Sematala1, A. Muganzi1, C. Twijukye1, H. Byakwaga1. The AIDS Support Organisation (TASO), Mulago, Kampala, Uganda. R. Ochai, D. Muhweezi, A. Coutinho1, B. Etukoit1. Imperial College, London, UK. C. Gilks, K. Boocock, C. Puddephatt, C. Grundy, J. Bohannon, D. Winogron1. MRC Clinical Trials Unit, London, UK. D. M. Gibb, A. Burke, D. Bray, A. Babiker, A. S. Walker, H. Wilkes, M. Rauchenberger, S. Sheehan, C. Spencer-Drake, K. Taylor, M. Spyer, A. Ferrier, B. Naidoo, D. Dunn, R. Goodall, J. H. Darbyshire, L. Peto1. Independent DART Trial Monitors. R. Nanfuka, C. Mufuka-Kapuya. Trial Steering Committee. I. Weller (Chair), A. Babiker (Trial Statistician), S. Bahendeka, M. Bassett, A. Chogo Wapakhabulo, J. H. Darbyshire, B. Gazzard, C. Gilks, H. Grosskurth, J. Hakim, A. Latif, C. Mapuchere, O. Mugurungi, P. Mugyenyi. Observers: C. Burke, S. Jones, C. Newland, G. Pearce, S. Rahim, J. Rooney, M. Smith, W. Snowden, J.-M. Steens. Data and Safety Monitoring Committee. A. Breckenridge (Chair), A. McLaren (previous Chair—deceased), C. Hill, J. Matenga, A. Pozniak, D. Serwadda. Endpoint Review Committee. T. Peto (Chair), A. Palfreeman, M. Borok, E. Katabira. ARROW Trial:

Trial Steering Committee. I. Weller (Chair), A. Babiker (Trial Statistician), S. Bahendeka, M. Bassett, A. Chogo Wapakhabulo, J. H. Darbyshire, B. Gazzard, C. Gilks, H. Grosskurth, J. Hakim, A. Latif, C. Mapuchere, O. Mugurungi, P. Mugyenyi. Observers: C. Burke, S. Jones, C. Newland, G. Pearce, S. Rahim, J. Rooney, M. Smith, W. Snowden, J.-M. Steens. Data and Safety Monitoring Committee. A. Breckenridge (Chair), A. McLaren (previous Chair—deceased), C. Hill, J. Matenga, A. Pozniak, D. Serwadda. Endpoint Review Committee. T. Peto (Chair), A. Palfreeman, M. Borok, E. Katabira. ARROW Trial: Joint Clinical Research Centre, Kampala, Uganda. P. Mugyenyi, V. Musiime, V. D. Afayo, E. Bagurukira, J. Bwomezi, J. Byaruhanga, P. Erimu, C. Karungi, H. Kizito, M. Mutumba, W. S. Namala, J. Namusanje, R. Nandugwa, T. K. Najjuko, E. Natukunda, M. Ndigendawani, S. O. Nsiyona, F. Odongo, K. Robinah, M. Ssenyonga, D. Sseremba, J. Tezikyabbiri, C. S. Tumusiime. MRC/UVRI Uganda Research Unit on AIDS, Entebbe, Uganda. P. Munderi, P. Nahirya-Ntege, M. Aber, F. N. Kaggwa, P. Kaleebu, R. Katuramu, J. H. Kyalimpa, J. Lutaakome, L. Matama, M. Musinguzi, G. Nabulime, A. Ruberantwari, R. Sebukyu, I. M. Ssekamatte, G. Tushabe, D. Wangi. Baylor-Uganda, Paediatric Infectious Disease Centre, Mulago Hospital, Uganda. A. Kekitiinwa, P. Musoke, S. Bakeera-Kitaka, R. Namuddu, P. Kasirye, J. K. Balungi, A. Babirye, J. Asello, S. Nakalanzi, N. C. Ssemambo, J. Nakafeero, J. N. Kairu, E. K. George, G. Musoba, J. Ssanyu, S. Ssenyonjo.

MRC/UVRI Uganda Research Unit on AIDS, Entebbe, Uganda. P. Munderi, P. Nahirya-Ntege, M. Aber, F. N. Kaggwa, P. Kaleebu, R. Katuramu, J. H. Kyalimpa, J. Lutaakome, L. Matama, M. Musinguzi, G. Nabulime, A. Ruberantwari, R. Sebukyu, I. M. Ssekamatte, G. Tushabe, D. Wangi. Baylor-Uganda, Paediatric Infectious Disease Centre, Mulago Hospital, Uganda. A. Kekitiinwa, P. Musoke, S. Bakeera-Kitaka, R. Namuddu, P. Kasirye, J. K. Balungi, A. Babirye, J. Asello, S. Nakalanzi, N. C. Ssemambo, J. Nakafeero, J. N. Kairu, E. K. George, G. Musoba, J. Ssanyu, S. Ssenyonjo. University of Zimbabwe, Harare, Zimbabwe. K. J. Nathoo, M. F. Bwakura-Dangarembizi, F. Mapinge, T. Mhute, T. Vhembo, R. Mandidewa, D. Nyoni, C. Katanda, G. C. Tinago, J. Bhiri, D. Muchabaiwa, S. Mudzingwa, M. M. Chipiti, M. Phiri, J. Steamer, C. C. Marozva, S. J. Maturure, L. Matanhike, S. Tsikirayi, L. Munetsi. Medical Research Council Clinical Trials Unit, London, UK. D. M. Gibb, M. J. Thomason, A. D. Cook, J. M. Crawley, A. A. Ferrier, B. Naidoo, M. J. Spyer, A. S. Walker, L. K. Kendall. Independent ARROW Trial Monitors. R. Nanfuka, I. Machuringa. Trial Steering Committee. I. Weller (Chair), E. Luyirika, H. Lyall, E. Malianga, C. Mwansambo, M. Nyathi, A. Wapakhabulo, D. M. Gibb, A. Kekitiinwa, P. Mugyenyi, P. Munderi, K. J. Nathoo. Observers: S. Kinn, M. MacNeil, M. Roberts, W. Snowden. Data and Safety Monitoring Committee. A. Breckenridge (Chair), C. Hill, J. Matenga, A. Pozniak, J. Tumwine. Endpoint Review Committee. G. Tudor-Williams (Chair), H. Barigye, H. A. Mujuru, G. Ndeezi. Entebbe Cohort:

Trial Steering Committee. I. Weller (Chair), E. Luyirika, H. Lyall, E. Malianga, C. Mwansambo, M. Nyathi, A. Wapakhabulo, D. M. Gibb, A. Kekitiinwa, P. Mugyenyi, P. Munderi, K. J. Nathoo. Observers: S. Kinn, M. MacNeil, M. Roberts, W. Snowden. Data and Safety Monitoring Committee. A. Breckenridge (Chair), C. Hill, J. Matenga, A. Pozniak, J. Tumwine. Endpoint Review Committee. G. Tudor-Williams (Chair), H. Barigye, H. A. Mujuru, G. Ndeezi. Entebbe Cohort: MRC Research Unit on AIDS/Uganda Virus Research Institute, Entebbe, Uganda. C. Watera, G. Miiro, S. Zawedde, J. Nakiyingi, D. Rutebarika, H. Grosskurth. 3Cs4kids: South Africa. H. Zar, B. Eley, P. Roux, M. Cotton, T. Meyers, H. Moultrie. Zambia. V. Mulenga, C. Chintu, C. Kankasa. Cote d'Ivoire. P. Msellati, P. Fassinou, N. Elenga. Malawi. S. Graham, J. Ellis, R. Weigel. Uganda. C. Giaquinto, M. Nanyonga, E. Morelli, B. Atai. Brazil. J. Pinto, C. Araújo, A. Carvalho, I. Carvalho, A. Diniz, F. Ferreira, V. Lobato, T. Sanchez. United Kingdom. T. Duong, D. Dunn, D. M. Gibb, C. Duff.

Zambia. V. Mulenga, C. Chintu, C. Kankasa. Cote d'Ivoire. P. Msellati, P. Fassinou, N. Elenga. Malawi. S. Graham, J. Ellis, R. Weigel. Uganda. C. Giaquinto, M. Nanyonga, E. Morelli, B. Atai. Brazil. J. Pinto, C. Araújo, A. Carvalho, I. Carvalho, A. Diniz, F. Ferreira, V. Lobato, T. Sanchez. United Kingdom. T. Duong, D. Dunn, D. M. Gibb, C. Duff. Author contributions. The DART trial was conducted by P. Mugyenyi, P. Munderi, J. H., E. K., and C. K.. The ARROW trial was conducted by K. N., A. K., P. Mugyenyi, and P. N.-N.; trials were coordinated in the United Kingdom by D. M. G., A. S. W., A. J. P., and C. F. G. In addition, A. S. W. conducted the analyses. A. S. W. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to interpretation of the data. A. S. W. wrote the first draft of the paper with D. M. G. and A. J. P. All authors revised the manuscript critically and approved the final version. Financial support. The DART trial was supported by the UK Medical Research Council (MRC; grant number G0600344); the UK Department for International Development (DFID); and the Rockefeller Foundation. GlaxoSmithKline, Gilead Sciences, and Boehringer-Ingelheim donated first-line drugs for DART, and Abbott Laboratories provided lopinavir/ritonavir (Kaletra/Aluvia) as part of the second-line regimen for DART. The ARROW trial was supported by the UK MRC (grant number G0300400) and the UK DFID. Drugs are provided by GlaxoSmithKline and the National ART Programmes of Uganda and Zimbabwe.

first-line drugs for DART, and Abbott Laboratories provided lopinavir/ritonavir (Kaletra/Aluvia) as part of the second-line regimen for DART. The ARROW trial was supported by the UK MRC (grant number G0300400) and the UK DFID. Drugs are provided by GlaxoSmithKline and the National ART Programmes of Uganda and Zimbabwe. Potential conflicts of interest. A. S. W. is a board member of Tibotec, and is on the speakers’ bureau of Gilead; D. M. G. is a board member of Tibotec. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. 1 Staff that left during the course of the DART trial.

The global burden of liver disease caused by hepatitis C virus (HCV) is increasing faster than other causes [1, 2]. In developed countries the majority of transmissions and cases are among people who inject drugs (PWID) [1]. In the UK, PWID acquire >90% of HCV infections [3]. HCV prevalence among PWID can vary 2–3 fold within countries, with chronic prevalences of 20% in some areas up to >60% in others [4]. High-coverage needle and syringe programs (HCNSP) and opiate substitution therapy (OST) are key primary interventions, and emerging evidence suggests they can greatly reduce an individual's HCV risk [5]. Modelling has suggested that these interventions alone may not always lead to substantial reductions in HCV prevalence [6]. However, HCNSP and OST have other benefits including reducing HIV transmission [7], drug-related deaths [8], and drug-related crime [9]. In addition, there may be circumstances where their scale-up could be sufficient for achieving large reductions in HCV prevalence. We have shown HCV treatment could have a primary role in prevention if delivered at sufficient levels to PWID [10–14] and can be more cost-effective than treating ex- or non-PWID because of prevented secondary infections [15]. We consider the impact of combining OST, HCNSP, and HCV treatment on HCV prevalence and incidence among PWID.

V treatment could have a primary role in prevention if delivered at sufficient levels to PWID [10–14] and can be more cost-effective than treating ex- or non-PWID because of prevented secondary infections [15]. We consider the impact of combining OST, HCNSP, and HCV treatment on HCV prevalence and incidence among PWID. METHODS Model Description We extended our dynamic, deterministic model of HCV transmission and treatment among PWID [10] to include movement of PWID through various intervention states (OST and HCNSP, defined as obtaining 1 or more sterile syringes from an NSP for each injection). The model schematic and equations can be found in the Supplementary material.

ur dynamic, deterministic model of HCV transmission and treatment among PWID [10] to include movement of PWID through various intervention states (OST and HCNSP, defined as obtaining 1 or more sterile syringes from an NSP for each injection). The model schematic and equations can be found in the Supplementary material. Briefly, all PWID are initially susceptible (Xj,k, where subscripts represent intervention coverage such that off/on HCNSP [j = 0 or 1, respectively] and off/on OST [k = 0 or 1, respectively]) and become HCV infected at a per-capita rate, λj,k, specific to that intervention state. A proportion (δ) of those acutely infected will spontaneously clear infection and be at risk of reinfection (Ej,k), while the remainder (1 − δ) proceed to chronic infection, Cj,k. Chronically infected PWID can be put on antiviral treatment (Tj,k) at a rate of Φ per 1000 PWID annually for a duration 1/ω, whereupon a proportion (α) attain sustained viral response (SVR) and move to the previously infected compartment, Ej,k, where they are at risk of reinfection. Those who do not attain SVR (1 − α) move to the treatment failure compartment, Fj,k, where we assume they cannot be retreated. PWID leave all stages through permanent cessation of drug use (μ1) or death due to drug- or nondrug-related causes (μ2). The model does not include an acute infection category due to their small probable contribution (<2%) to HCV transmission even if they have heightened viremia [14] and also ignores any immunity following treatment or spontaneous clearance. Immunity is neglected in part because the evidence is uncertain [16] but also because previous analyses suggested that incorporation of partial immunity has negligible impact [6, 10].

o HCV transmission even if they have heightened viremia [14] and also ignores any immunity following treatment or spontaneous clearance. Immunity is neglected in part because the evidence is uncertain [16] but also because previous analyses suggested that incorporation of partial immunity has negligible impact [6, 10]. PWID are tracked through 4 intervention states: no intervention, OST only, HCNSP only, and OST with HCNSP. All PWID initially enter the no-intervention state. We assume the per-capita recruitment rates to OST (β) and HCNSP (η) are independent of the current intervention state [6]. The rates of leaving OST and HCNSP are γ and κ, respectively. The forces of infection for each susceptible state were defined by the relative risk in that state, such that infectivity and susceptibility were altered by a factor Γ, Π, or Β if the PWID was on OST, HCNSP, or both, respectively. The chance of a PWID having a transmission event with any PWID from another risk state and infectious status was assumed to be proportional to the relative frequency of transmission events for PWID in that state. Due to rapid reductions in viral load while on antiviral treatment, we assume the transmission potential of those on treatment is scaled down by a factor depending on the SVR rate.

er risk state and infectious status was assumed to be proportional to the relative frequency of transmission events for PWID in that state. Due to rapid reductions in viral load while on antiviral treatment, we assume the transmission potential of those on treatment is scaled down by a factor depending on the SVR rate. Intervention Parameters Model parameters can be found in Table 1. Effect estimates for PWID on OST or HCNSP were taken from a pooled analysis of UK data [5, 6]. We use a pooled SVR rate for pegylated interferon (pegIFN) and ribavirin (RBV) from a meta-analysis of individuals who report actively injecting (median 61.4%, range 51.2%–71.6%) [17] and explore the impact of new IFN-free direct-acting antiviral (DAA) treatments, using SVR rates from phase II studies (median 90%, range 80%–100%) [18–20]. Table 1. Model Parameters and Sources

and ribavirin (RBV) from a meta-analysis of individuals who report actively injecting (median 61.4%, range 51.2%–71.6%) [17] and explore the impact of new IFN-free direct-acting antiviral (DAA) treatments, using SVR rates from phase II studies (median 90%, range 80%–100%) [18–20]. Table 1. Model Parameters and Sources Parameter Symbol Value(s) or Range Units References HCV chronic prevalencea Vary π to fit 20%, 40%, or 60% … … PWID population size Vary θ to fit 1000 … … Exit rate (cessation + death) μ1 + μ2 8.5% per year As in [6, 10], sensitivity analysis varied 5%–20% per year [21, 26–28] Recruitment rate on OST β (0%–55%) per month Varied to achieve a range of intervention coverages Recruitment rate on HCNSP η Set equal to recruitment rate on OST per month Varied to achieve a range of intervention coverages Duration on OST 12/γ 8 months [6, 8] Duration on HCNSP 12/κ 8 months [6] Few data, assumed the same as OST Proportion spontaneously clear δ 25% … [29] Annual PWID treatment rate Φ 0–100 per 1000 PWID Varied to achieve a range of intervention coverages PEG-IFN + RBV SVR α 61.4% (51.2%–71.6%) … [17]Sampled from a uniform distribution IFN-free DAA SVR α 90% (80%–100%) … [18–20]Sampled from a uniform distribution PEG-IFN + RBV duration 52/ω 24 weeks [30] IFN-free DAA duration 52/ω 12 (8–16) weeks [18–20] Sampled from a uniform distribution Relative risk for acquiring HCV on OST Γ 0.48 (0.17–1.33) … [5, 6] Sampled from a lognormal distribution Relative risk for acquiring HCV on HCNSP Π 0.50 (0.22–1.12) … [5, 6] Sampled from a lognormal distribution Relative risk for acquiring HCV on OST and HCNSP Γ 0.21 (0.08–0.52) … [5, 6] Sampled from a lognormal distribution. Abbreviations: DAAs, direct-acting antivirals; HCNSP, high-coverage needle and syringe programs, defined as receiving 1 or more sterile syringes from an NSP per injection per month; OST, opiate substation therapy; peg-IFN, pegylated interferon; PWID, people who inject drugs; RBV, ribavirin; SVR, sustained viral response.

breviations: DAAs, direct-acting antivirals; HCNSP, high-coverage needle and syringe programs, defined as receiving 1 or more sterile syringes from an NSP per injection per month; OST, opiate substation therapy; peg-IFN, pegylated interferon; PWID, people who inject drugs; RBV, ribavirin; SVR, sustained viral response. a Used to estimate the infection rate, π (vary π and fit to the hepatitis C virus chronic prevalence). Modeled Scenario Analysis We project the 10-year impact on HCV prevalence and incidence for various combinations of scale-up of antiviral treatment (from none at baseline), OST and HCNSP (from 0%, 20%, or 50% coverage of each at baseline) for 3 baseline HCV chronic prevalence settings (20%, 40%, and 60%). We explore the impact of treatment using peg-IFN + RBV or new IFN-free DAAs in combination with scale-up of OST and HCNSP to a maximum of 80% for each. To do this, we performed a multivariate uncertainty analysis by running the model with 1000 randomly sampled parameter values from the uncertainty distributions for the antiviral treatment SVR and the efficacy of OST and HCNSP on reducing HCV transmission risk (Table 1). We utilize the projections to determine combinations of antiviral treatment, OST and HCNSP scale-up that halve baseline chronic prevalence within 10 years. Contour maps show what prevalence and incidence reductions are achievable with various levels of scale-up of antiviral treatment, OST and HCNSP using median estimates for the efficacy of OST, HCNSP, and SVR.

binations of antiviral treatment, OST and HCNSP scale-up that halve baseline chronic prevalence within 10 years. Contour maps show what prevalence and incidence reductions are achievable with various levels of scale-up of antiviral treatment, OST and HCNSP using median estimates for the efficacy of OST, HCNSP, and SVR. HCV Treatment Delivered Within OST Programs Since HCV antiviral treatment may best be delivered to PWID alongside OST, we explore the impact of restricting treatment to only those on OST. We calculate the minimum coverage of OST and HCNSP required to achieve different relative prevalence reductions for 20%, 40%, and 60% HCV chronic prevalence scenarios with no interventions at baseline. Model projections assume median estimates for efficacy of OST, HCNSP, and SVR for peg-IFN + RBV and assume either all chronically infected PWID on OST are treated annually (limited by the HCV prevalence) or 5% of PWID on OST are treated annually.

d 60% HCV chronic prevalence scenarios with no interventions at baseline. Model projections assume median estimates for efficacy of OST, HCNSP, and SVR for peg-IFN + RBV and assume either all chronically infected PWID on OST are treated annually (limited by the HCV prevalence) or 5% of PWID on OST are treated annually. Sensitivity Analysis We perform a 1-way sensitivity analysis on the intervention combinations required to halve baseline chronic prevalence over 10 years for the 40% baseline prevalence scenario with no baseline OST or HCNSP. Model projections assume median estimates for efficacy of OST, HCNSP, and SVR for peg-IFN + RBV (Table 1) and explore the impact of including PWID risk heterogeneity and varying the exit rate (due to injecting cessation or death). For the risk heterogeneity sensitivity analysis, we simulate a high-risk population comprised of 50% of PWID (the remainder low risk), no turnover between high- and low-risk states, and increased HCV risk among the high-risk group of 2- or 6-fold that of the low-risk PWID. We also explore scenarios where the high- and low-risk groups mix proportionally or partially (50%) assortatively and the effect of assuming that a proportion (20%) of PWID never go on OST.

ver between high- and low-risk states, and increased HCV risk among the high-risk group of 2- or 6-fold that of the low-risk PWID. We also explore scenarios where the high- and low-risk groups mix proportionally or partially (50%) assortatively and the effect of assuming that a proportion (20%) of PWID never go on OST. RESULTS Scaling Up From Baseline With peg-IFN + RBV For a baseline chronic HCV prevalence of 20%, 40%, or 60%, Figure 1A shows that by combining interventions (involving peg-IFN + RBV), chronic prevalence can halve within 10 years. The model projections vary considerably (95% credible interval [CrI] deviates 6%–21% from median projections for treatment only, with increasing uncertainty including OST and HCNSP scale-up [24%–71% deviation with 60% scale-up of OST and HCNSP]). Scaling up OST and HCNSP will substantially decrease the treatment rate required to halve prevalence within 10 years (by 19%–27% or 39%–44%, respectively). Hence, if coverage of OST and HCNSP were both increased to 40%, then annually treating 10 (95% CrI, 8–14), 23 (95% CrI, 19–32), and 42 (95% CrI, 35–58) per 1000 PWID would halve prevalence over 10 years in the 20%, 40%, or 60% chronic HCV prevalence scenarios, respectively, as compared to treating 18 (95% CrI, 17–20), 38 (95% CrI, 36–42), and 68 (95% CrI, 64–83) per 1000 PWID annually with no OST or HCNSP coverage. Figure 1. Combinations of annual treatment rates per 1000 injectors and coverage of opiate substitution therapy (OST) and high-coverage needle and syringe programs (HCNSP) required to reduce prevalence by 50% within 10 years. Results shown for 3 baseline chronic prevalence settings (20%, 40%, and 60%). A and B, Assumes no intervention coverage at baseline with OST and HCNSP scale-up to 0%, 20%, 40%, or 60% of each and using pegylated interferon and ribavirin (peg-IFN + RBV) (A) and interferon (IFN)-free direct-acting antivirals (DAAs) (B). C, Assumes 50% coverage of OST and HCNSP at baseline with OST and HCNSP scale-up to 50%, 60%, 70%, or 80% of each using peg-IFN + RBV. The box-plots signify the uncertainty (middle line is the median, limits of the boxes are 25% and 75% percentiles and whiskers are 2.5% and 97.5% percentiles) in the impact projections due to uncertainty in the intervention effect estimates.

h OST and HCNSP scale-up to 50%, 60%, 70%, or 80% of each using peg-IFN + RBV. The box-plots signify the uncertainty (middle line is the median, limits of the boxes are 25% and 75% percentiles and whiskers are 2.5% and 97.5% percentiles) in the impact projections due to uncertainty in the intervention effect estimates. Contour maps of the relative prevalence reductions for various combinations of intervention scale-up over 10 years show that scale-up of OST and HCNSP reduces the required treatment rate necessary to achieve a given impact (Figure 2) and that HCV treatment is required to achieve >45% reduction in prevalence within 10 years. Figure 2. Contour maps of the relative reductions in prevalence (%) at 10 years with combinations of antiviral treatment (y-axis) and opiate substitution therapy/high-coverage needle and syringe program (OST and HCNSP) (x-axis) scale-up with no baseline coverage of OST, HCNSP, or treatment. Results shown for 3 baseline hepatitis C virus chronic prevalence settings (20%, 40%, and 60%) with pegylated interferon and ribavirin (pegIFN + RBV) (A–C) and IFN-free direct-acting antivirals (D–F). Projections used the median estimates for efficacy of OST, HCNSP, and peg-IFN + RBV from Table 1. For a given coverage of OST and HCNSP, greater relative reductions in incidence are achieved at 10 years than for prevalence, with the relative impact on incidence being 101%–130% greater than the impact on prevalence for OST and HCNSP scale-up to 60% coverage (Supplementary Figure 2). In contrast, treatment alone has equal impact on incidence as prevalence.

from no OST or HCNSP at baseline. At 20% coverage of OST and HCNSP at baseline, achieving >40% prevalence reduction within 10 years always requires scale-up of antiviral treatment, whereas at 50% baseline coverage, treatment is always required to achieve >30% prevalence reductions at 10 years (Supplementary Figure 3). Scaling Up HCV Treatment Through OST Programs Figure 3 shows the minimum coverage of OST and HCNSP required (no coverage at baseline) to achieve different relative reductions in prevalence at 10 years, while assuming either there are no limits to treatment capacity/uptake within OST (all infected PWID on OST are treated each year) or that 5% of PWID on OST are treated annually. If all infected PWID on OST can be treated annually, halving prevalence in 10 years requires 18%–24% coverage of OST and HCNSP, whereas higher coverage levels are required if only 5% of PWID on OST are annually treated (25%–58% OST and HCNSP). Figure 3. Minimum coverage of opiate substitution therapy/high-coverage needle and syringe programs (OST and HCNSP) required if antiviral treatment (pegylated interferon and ribavirin [peg-IFN + RBV]) is delivered alongside OST. Figures show the minimum coverage of OST and HCNSP required (y-axis) for various desired relative prevalence reductions at 10 years (x-axis) with the 20%, 40%, and 60% baseline hepatitis C virus (HCV) chronic prevalence settings if all infected people who inject drugs (PWID) on OST are treated annually, limited by HCV prevalence (A) or 5% of PWID on OST are treated annually (B). Projections used the median estimates for efficacy of OST, HCNSP, and peg-IFN + RBV from Table 1.

and 60% baseline hepatitis C virus (HCV) chronic prevalence settings if all infected people who inject drugs (PWID) on OST are treated annually, limited by HCV prevalence (A) or 5% of PWID on OST are treated annually (B). Projections used the median estimates for efficacy of OST, HCNSP, and peg-IFN + RBV from Table 1. Sensitivity Analysis At higher exit rates (ie, in populations with shorter durations of injecting) scaling up OST and HCNSP achieves more impact than at lower exit rates, whereas the opposite occurs for scaling up antiviral treatment, but less so (Supplementary Figure 4A and B). Therefore, to halve prevalence within 10 years, a strategy using treatment alone would require more treatments at a high exit rate (shorter injecting duration) than at a low exit rate (longer duration), but a strategy using just OST and HCNSP would require the opposite (Supplementary Figure 4A). To minimize the uncertainty around injecting duration, it is possible to choose an intervention combination that achieves the same impact regardless of exit rate (Supplementary Figure 4C).

at a low exit rate (longer duration), but a strategy using just OST and HCNSP would require the opposite (Supplementary Figure 4A). To minimize the uncertainty around injecting duration, it is possible to choose an intervention combination that achieves the same impact regardless of exit rate (Supplementary Figure 4C). The model projections are insensitive to the inclusion of a high-risk group, even if it comprised 50% of the population with no turnover between high and low risk (Supplementary Figure 4D). Only if the high-risk group has a 6-fold relative risk and mixes partially assortatively with no turnover does the required number of treatments increase by a noticeable degree (20%–35% for a given OST and HCNSP coverage). In contrast, if there is turnover between risk groups, then heterogeneity has little effect. Additionally, there is no difference in the required treatment rates if 20% of the population never go on OST.

does the required number of treatments increase by a noticeable degree (20%–35% for a given OST and HCNSP coverage). In contrast, if there is turnover between risk groups, then heterogeneity has little effect. Additionally, there is no difference in the required treatment rates if 20% of the population never go on OST. DISCUSSION We projected the impact of combining OST, HCNSP, and HCV antiviral treatment on HCV prevalence and incidence among PWID. Halving chronic HCV prevalence within 10 years is not possible using OST and HCNSP alone but is achievable in all prevalence settings when combined with current treatments (peg-IFN + RBV) and will be more achievable with new IFN-free DAAs. For a given coverage of OST and HCNSP, greater reductions in incidence are achieved at 10 years than for prevalence, whereas no difference is found with antiviral treatment. This is because OST and NSP directly reduces incidence, while treatment directly reduces prevalence by curing infections. In general, increasing coverage of OST and HCNSP by 20% from any level reduces the required number of treatments by about 30%. In settings with shorter average injecting durations (such as South East Asia), scale-up of OST and HCNSP may be preferable as there is little time for the benefits of antiviral treatment to accrue. Conversely, in areas with long injecting durations (such as Zurich [21]), OST and HCNSP impact will be much reduced, so treatment is critical for achieving substantial HCV reductions. Finally, heterogeneity in injecting risk has marginal impact in most realistic scenarios.

the benefits of antiviral treatment to accrue. Conversely, in areas with long injecting durations (such as Zurich [21]), OST and HCNSP impact will be much reduced, so treatment is critical for achieving substantial HCV reductions. Finally, heterogeneity in injecting risk has marginal impact in most realistic scenarios. Limitations These projections are based on a theoretical model with several limitations. First, there is uncertainty in the model parameters including efficacy estimats for OST and HCV, and HCV antiviral treatment SVR rates among PWID. Second, it remains to be demonstrated that the higher treatment rates projected in some of the scenarios can be achieved, although the new IFN-free DAA treatment should make scale-up easier to implement, if earlier trials suggesting shorter treatment duration, higher SVR and lower toxicity than current treatment regimes [20] prove to be true. Multivariate sensitivity analyses were used to explore the implications of parameter uncertainty, including wide sampling ranges around the SVR estimates to account for settings with different SVR rates or genotype distributions. In addition, complexities involved in scaling up each intervention were not considered here. Previous modeling analyses considered these issues for OST and HCNSP [6], but additional case-finding interventions may be required.

Limitations These projections are based on a theoretical model with several limitations. First, there is uncertainty in the model parameters including efficacy estimats for OST and HCV, and HCV antiviral treatment SVR rates among PWID. Second, it remains to be demonstrated that the higher treatment rates projected in some of the scenarios can be achieved, although the new IFN-free DAA treatment should make scale-up easier to implement, if earlier trials suggesting shorter treatment duration, higher SVR and lower toxicity than current treatment regimes [20] prove to be true. Multivariate sensitivity analyses were used to explore the implications of parameter uncertainty, including wide sampling ranges around the SVR estimates to account for settings with different SVR rates or genotype distributions. In addition, complexities involved in scaling up each intervention were not considered here. Previous modeling analyses considered these issues for OST and HCNSP [6], but additional case-finding interventions may be required. Third, we neglect the other benefits of OST and HCNSP, such as the impact on reducing HIV transmission [7], drug-related deaths [8], and drug-related crime [9]. These benefits would accrue in addition to any HCV benefits and so give added impetus to scaling up HCNSP and especially OST.

In addition, complexities involved in scaling up each intervention were not considered here. Previous modeling analyses considered these issues for OST and HCNSP [6], but additional case-finding interventions may be required. Third, we neglect the other benefits of OST and HCNSP, such as the impact on reducing HIV transmission [7], drug-related deaths [8], and drug-related crime [9]. These benefits would accrue in addition to any HCV benefits and so give added impetus to scaling up HCNSP and especially OST. Implications Overall, our work supports current recommendations on HCV prevention issued by the European Centre for Disease Prevention and Control that interventions be combined to achieve maximum impact [22], as well as previous modeling studies that have shown that scale-up of antiviral treatment [10–14, 23], OST and HCNSP [6] among PWID can reduce prevalence in a variety of chronic prevalence settings. However, the relative affordability of each strategy is a key question, particularly with the new DAAs. Current treatment with peg-IFN + RBV costs between $16 000 and $33 000 per full treatment course, whereas triple therapy with boceprevir and telaprevir costs approximately $30 000–$80 000 [24]. By contrast, annual costs of delivering OST have been estimated at $10–$15 per day ($3650–$5475 per year) and high-coverage NSP may cost ∼$500 per year [25]. Therefore, strategies that increase OST and HCNSP in order to minimize the number of antiviral HCV treatments required to prevent and reduce chronic HCV are likely to be an efficient use of resources. Future work should address the optimal combination prevention intervention strategy that addresses both cost effectiveness and affordability.

that increase OST and HCNSP in order to minimize the number of antiviral HCV treatments required to prevent and reduce chronic HCV are likely to be an efficient use of resources. Future work should address the optimal combination prevention intervention strategy that addresses both cost effectiveness and affordability. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

rdjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Notes Financial support. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. N. M.: This work is produced by N. M. under the terms of the postdoctoral research training fellowship issued by the National Institute for Health Research (NIHR). The views expressed in this publication are those of the author and not necessarily those of the National Health Service, the NIHR, the UK Department of Health, or the London School of Hygiene and Tropical Medicine. P. V.: Medical Research Council New Investigator Award G0801627. M. H.: NIHR School of Public Health, Nationally Integrated Quantitative Understanding of Addiction Harm, and support of the Centre for the Development and Evaluation of Complex Interventions for Public Health Improvement, a UK Clinical Research Collaboration (UKCRC) Public Health Research: Centre of Excellence. Funding from the British Heart Foundation, Cancer Research UK, Economic and Social Research Council (RES-590-28-0005), Medical Research Council, the Welsh Assembly Government and the Wellcome Trust (WT087640MA), under the auspices of the UKCRC is gratefully acknowledged.

C) Public Health Research: Centre of Excellence. Funding from the British Heart Foundation, Cancer Research UK, Economic and Social Research Council (RES-590-28-0005), Medical Research Council, the Welsh Assembly Government and the Wellcome Trust (WT087640MA), under the auspices of the UKCRC is gratefully acknowledged. Supplement sponsorship. This article was published as part of a supplement titled “Prevention and Management of Hepatitis C Virus Among People Who Inject Drugs: Moving the Agenda Forward,” sponsored by an unrestricted grant from the International Network on Hepatitis in Substance Users (INHSU), The Kirby Institute (University of New South Wales), Abbvie, Gilead Sciences, Janssen-Cilag, and Merck. Potential conflicts of interest. N. M. has received an honorarium for speaking at a conference sponsored by Janssen. S. H. has received honoraria for speaking at conferences sponsored by MSD, Janssen, Gilead, and Roche. D. G. is a member of advisory boards and undertakes consultancy for Merck and Janssen. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Dengue is the most important mosquito-borne viral disease affecting humans. Infection can be caused by any 1 of 4 dengue viruses (DENV-1 to DENV-4), transmitted by Aedes mosquitoes [1]. Currently about 100 million clinically apparent cases are estimated to occur each year [2], resulting in approximately 20 000 deaths [3].

Dengue is the most important mosquito-borne viral disease affecting humans. Infection can be caused by any 1 of 4 dengue viruses (DENV-1 to DENV-4), transmitted by Aedes mosquitoes [1]. Currently about 100 million clinically apparent cases are estimated to occur each year [2], resulting in approximately 20 000 deaths [3]. Infection with any serotype can cause a broad range of disease manifestations, from inapparent infection to severe and fatal disease [4]. The most notable complication is an unexplained vasculopathy that manifests as a transient increase in vascular permeability resulting in leakage of plasma from the circulation. Substantial plasma losses may occur, leading to the potentially fatal dengue shock syndrome (DSS). Although adults do experience shock, vascular leakage is generally more severe in young children [5], and in endemic areas DSS is seen primarily in the pediatric population. Thrombocytopenia and coagulation derangements also occur, and a variety of bleeding manifestations ranging from minor skin petechiae to major mucosal bleeding may be seen. Neither vaccines nor specific therapies are currently available; careful clinical observation and judicious use of intravenous fluid therapy, in particular urgent shock resuscitation for DSS, are the foundations for successful management [6]. Many individuals with DSS respond to resuscitation with crystalloid solutions, but patients with profound or protracted shock often require additional support with colloid solutions and are at risk of developing respiratory compromise due to ongoing plasma leakage. Mortality rates for DSS vary from <1% to >10% depending on the severity of the cases reported, the level of monitoring available, and the experience of the attending healthcare personnel [7–9].

additional support with colloid solutions and are at risk of developing respiratory compromise due to ongoing plasma leakage. Mortality rates for DSS vary from <1% to >10% depending on the severity of the cases reported, the level of monitoring available, and the experience of the attending healthcare personnel [7–9]. Despite the increasing burden of dengue globally, only a few small retrospective reports have described the clinical characteristics, management, or outcomes of DSS [8, 10, 11]. At the Hospital for Tropical Diseases (HTD) in Ho Chi Minh City, a prospective observational study aiming to enroll all children presenting with DSS was conducted from 1999 to 2009. Here we present data on >1700 cases collected during this 10-year period, providing the first comprehensive description of the clinical features of DSS in children.

al Diseases (HTD) in Ho Chi Minh City, a prospective observational study aiming to enroll all children presenting with DSS was conducted from 1999 to 2009. Here we present data on >1700 cases collected during this 10-year period, providing the first comprehensive description of the clinical features of DSS in children. METHODS Study Design and Participants At HTD, children aged <15 years with DSS are managed on the pediatric intensive care unit (PICU). In 1999, we commenced a prospective observational study in which children admitted with clinically diagnosed DSS—that is, a history consistent with dengue, with hemodynamic compromise (either narrowing of the pulse pressure or hypotension for age, with evidence of impaired perfusion) thought by the treating clinician to be due to vascular leakage and to require volume resuscitation—were eligible to participate. Patients transferred from other facilities for tertiary care after initial resuscitation were not eligible. A double-blind randomized controlled trial (RCT) that was conducted within the study period (1999–2004) comparing different fluid solutions for initial resuscitation has been reported previously [7]. Both studies were approved by the HTD ethical committee and the Oxford Tropical Research Ethics Committee.

not eligible. A double-blind randomized controlled trial (RCT) that was conducted within the study period (1999–2004) comparing different fluid solutions for initial resuscitation has been reported previously [7]. Both studies were approved by the HTD ethical committee and the Oxford Tropical Research Ethics Committee. Study Procedures and Data Collection Following written informed consent by a parent or guardian, trained study physicians enrolled participants immediately after diagnosis of shock. Demographic characteristics, clinical history, and examination findings were recorded on a structured case report form at enrollment and then daily until discharge or death, together with detailed information about all therapeutic interventions and supportive care required. Specimens for dengue diagnostics were obtained at enrollment and discharge. All survivors were asked to return for follow-up assessment 1 month later.

e report form at enrollment and then daily until discharge or death, together with detailed information about all therapeutic interventions and supportive care required. Specimens for dengue diagnostics were obtained at enrollment and discharge. All survivors were asked to return for follow-up assessment 1 month later. A core group of senior clinicians supervised patient care throughout the study period, following the management guidelines for pediatric DSS at HTD. The routine initial regimen consisted of 25 mL/kg Ringer's lactate solution over 2 hours, with colloid solutions (dextran or starch) reserved for children presenting with profound shock. However, during the RCT, patients were randomized to receive 1 of these 3 fluid solutions, at the same rate but blinded, for their initial resuscitation [7]. A standardized schedule of Ringer's lactate was then used for all patients, involving staged reductions at specific time intervals, aiming for maintenance fluid therapy after 8 hours. Patients who failed to stabilize within 2 hours, or who deteriorated during the mandatory 36- to 48-hour period of close observation, received 10–15 mL/kg infusions of rescue colloid plus inotropes, blood products, or other therapy at the discretion of the treating clinician.

or maintenance fluid therapy after 8 hours. Patients who failed to stabilize within 2 hours, or who deteriorated during the mandatory 36- to 48-hour period of close observation, received 10–15 mL/kg infusions of rescue colloid plus inotropes, blood products, or other therapy at the discretion of the treating clinician. Cardiovascular status was monitored at least hourly until stable for 24 hours, and subsequently every 4 hours. The capillary hematocrit was measured at baseline, approximately 2 and 6 hours after study entry, and then every 12 hours or in the event of cardiovascular deterioration. A complete blood count was performed once daily, whereas other laboratory tests (eg, liver/renal function) were checked on clinical grounds rather than according to a defined study schedule. Disease classification was performed using the World Health Organization (WHO) 1997 and 2009 criteria (Supplementary Appendix 1) [4, 12]. Dengue Diagnostics Dengue immunoglobulin M and immunoglobulin G (IgG) capture enzyme-linked immunosorbent assays were performed on paired enrollment and early convalescent specimens, together with reverse transcription polymerase chain reaction (RT-PCR) on the enrollment specimen [13–15]. Seroconversion and/or detection of DENV RNA in plasma defined laboratory-confirmed cases (Supplementary Appendix 2). A positive dengue-specific IgG on or before day 7 of illness defined a secondary infection, whereas 2 negative dengue-specific IgG results, at least 1 obtained after day 7, were required to define a primary infection.

and/or detection of DENV RNA in plasma defined laboratory-confirmed cases (Supplementary Appendix 2). A positive dengue-specific IgG on or before day 7 of illness defined a secondary infection, whereas 2 negative dengue-specific IgG results, at least 1 obtained after day 7, were required to define a primary infection. Statistical Analysis Continuous and categorical variables were summarized as median and interquartile range (IQR), or frequency and percentage, respectively. All analyses were performed with the statistical software R, version 2.15.0 [16]. RESULTS From 1999 to 2009, a total of 1810 of 1847 children (98%) admitted to PICU with clinical DSS participated in the study. In 19 cases, both RT-PCR and paired serology were negative, whereas in 72 cases the results were inconclusive; in the remaining 1719 cases (95%) dengue was confirmed, with the infecting serotype identified in 1209 of 1647 cases (73%) for which RT-PCR was performed. Among the confirmed dengue patients, 503 (29%) participated in the RCT, with the remainder enrolled in the observational study. Almost all cases came from the local catchment area, with <5% of cases transferred from elsewhere; however, 2 patients were enrolled in error, having already received parenteral fluids for shock resuscitation prior to transfer.

, 503 (29%) participated in the RCT, with the remainder enrolled in the observational study. Almost all cases came from the local catchment area, with <5% of cases transferred from elsewhere; however, 2 patients were enrolled in error, having already received parenteral fluids for shock resuscitation prior to transfer. Characteristics at Presentation With Shock Demographic information and selected clinical characteristics for all 1719 patients with confirmed dengue are described in Table 1. For most parameters, data were missing in <5% of cases. The median age was 10 years, varying by year of study from age 9 to 11 years. The median day of illness at shock was consistently 5 (IQR, 4–6) for each study year, although 62 patients (4%) overall presented on illness day 3. Table 1. Baseline Characteristics of the Study Participants at Enrollment

ing in <5% of cases. The median age was 10 years, varying by year of study from age 9 to 11 years. The median day of illness at shock was consistently 5 (IQR, 4–6) for each study year, although 62 patients (4%) overall presented on illness day 3. Table 1. Baseline Characteristics of the Study Participants at Enrollment Characteristic No. Median/Frequency IQR/% Demographic characteristics Age, y 1719 10 (7–12) Male sex 1719 902 (52) Referral status 1719 Directly from home 720 (42) From other wards in HTD 911 (53) From other healthcare centers 65 (4) Unknown 23 (1) Clinical and basic laboratory features Day of illness 1719 5 (4–6) Weight, kg 1719 27 (20–35) Temperature ≥38°C 1718 153 (9) Pulse rate per min, if measurablea 1393 120 (100–120) Systolic blood pressure, mm Hg, if measurablea 1596 90 (80–100) Pulse pressure, mm Hg, if measurablea 1576 20 (15–20) Hemorrhage 1719 None 493 (29) Skin only 1153 (67) Mucosal 73 (4) Abdominal tenderness 1714 1238 (72) Liver palpable 1696 1478 (87) Hematocrit, % 1696 49 (46–52) Platelet count, cells/µL 1695 41 000 (28 000–61 000) Aspartate aminotransferase, IU/L 1030 125 (80–206) DHF according to WHO 1997 criteriab 1642 939 (57) Dengue diagnostic tests RT-PCR performed 1647 DENV-1 675 (41) DENV-2 367 (22) DENV-3 48 (3) DENV-4 110 (7) Mixed 9 (1) Negative 438 (27) Immune status 1618 Primary 6 (<1) Secondary 1506 (93) Unclassifiable 106 (6) Continuous variables are summarized as median (interquartile range) and categorical variables as frequency (%).

e diagnostic tests RT-PCR performed 1647 DENV-1 675 (41) DENV-2 367 (22) DENV-3 48 (3) DENV-4 110 (7) Mixed 9 (1) Negative 438 (27) Immune status 1618 Primary 6 (<1) Secondary 1506 (93) Unclassifiable 106 (6) Continuous variables are summarized as median (interquartile range) and categorical variables as frequency (%). Abbreviations: DENV, dengue virus; DHF, dengue hemorrhagic fever; HTD, Hospital for Tropical Diseases; IQR, interquartile range; RT-PCR, reverse transcription polymerase chain reaction; WHO, World Health Organization. a All patients were assessed for these parameters, but we only report values for patients in whom the parameter could be measured. Note that in some patients a systolic pressure could be detected but the pulse was too rapid to count. b We used data available at the time of presentation with shock only. Tourniquet tests were not routinely performed. For each patient, the baseline hematocrit level was defined using local population values taken from an unpublished dataset including >1000 healthy Vietnamese children (37% if aged ≤10 years, 38.5% if female aged >10 years, 40% if male aged >10 years).

ion with shock only. Tourniquet tests were not routinely performed. For each patient, the baseline hematocrit level was defined using local population values taken from an unpublished dataset including >1000 healthy Vietnamese children (37% if aged ≤10 years, 38.5% if female aged >10 years, 40% if male aged >10 years). Commonly reported symptoms included lethargy (1490/1719 [87%]), vomiting (1199/1713 [70%]), and abdominal pain (932/1709 [55%]). Most children were afebrile, but 153 of 1718 (9%) still had an axillary temperature of ≥38°C at onset of shock, without a clear relationship to the day of illness (P = .09, Wilcoxon rank-sum test). In 123 of 1719 (7%), no blood pressure was measureable, whereas 417 of 1596 (26%) of the remainder exhibited hypotension for age, and 1568 of 1596 (98%) had a pulse pressure of ≤20 mm Hg. Respiratory distress (3/1718 [<1%]) and cyanosis due to profound shock (10/1714 [<1%]) were extremely uncommon. The liver was palpable in 1478 of 1696 (87%) of cases, with abdominal tenderness in 1238 of 1714 (72%), whereas a palpable spleen was extremely uncommon (only 5 cases documented). Almost one-third (493/1719 [29%]) of the patients had no bleeding. Among cases with bleeding, this amounted to skin petechiae or minor bruising in the majority, with mucosal hemorrhage noted in only 73 cases.

l tenderness in 1238 of 1714 (72%), whereas a palpable spleen was extremely uncommon (only 5 cases documented). Almost one-third (493/1719 [29%]) of the patients had no bleeding. Among cases with bleeding, this amounted to skin petechiae or minor bruising in the majority, with mucosal hemorrhage noted in only 73 cases. Progress in Hospital Because many patients in the RCT were randomized to a colloid for initial resuscitation, information on management and complications after enrollment is presented for the observational study and RCT groups separately (Table 2). Apart from the greater colloid usage, there was little difference between the 2 study groups other than a slightly higher proportion of minor skin bleeding observed in the RCT group. Considering the observational study only, most children recovered well with standard crystalloid resuscitation, although 547 of 1211 (45%) patients also received colloid therapy, 244 (45%) of them within the first 2 hours. Most children (328 [60%]) in this group received only a single colloid bolus, but up to 7 colloid infusions were needed for severe cases, with a median volume of 19 mL/kg (IQR, 13–25 mL/kg) of colloid given throughout hospitalization, on a background of 114 mL/kg (IQR, 99–129 mL/kg) total parenteral fluid therapy. Considering the whole patient cohort, additional cardiovascular support with inotropic drugs was required in 75 of 1719 patients (4%), and 513 of 1717 (30%) patients developed signs of fluid overload (overt pleural effusion or ascites) following resuscitation. Among these patients, 313 of 513 (61%) were treated with diuretic therapy for 1–2 days after hemodynamic stabilization. Table 2. Summary of Complications, Management, and Outcomes

patients (4%), and 513 of 1717 (30%) patients developed signs of fluid overload (overt pleural effusion or ascites) following resuscitation. Among these patients, 313 of 513 (61%) were treated with diuretic therapy for 1–2 days after hemodynamic stabilization. Table 2. Summary of Complications, Management, and Outcomes Observational Study (n = 1216) RCT (n = 503) All Patients (N = 1719) New bleeding 77 (6) 81 (16) 158 (9) None 1138 (94) 423 (84) 1561 (91) Skin only 39 (3) 59 (12) 98 (6) Mucosal 38 (3) 22 (4) 60 (3) Severe bleeding 20 (2) 11 (2) 31 (2) Transfusion 18 (1) 8 (2) 26 (2) Inotropic support 55 (5) 20 (4) 75 (4) Total parenteral fluid volume, mL/kga 114 (99–129) 125 (110–143) 116 (102–133) Received colloida 547 (45) 418 (83) 965 (56) Total colloid volumeb, mL/kg 19 (13–25) 25 (24–25) 25 (15–29) Clinical fluid overloada,c 340 (28) 173 (34) 513 (30) DHF according to WHO 1997 criteriaa,d 796 (66) 406 (81) 1202 (70) Death 7 (<1) 1 (<1) 8 (<1) Continuous variables are summarized as median (interquartile range) and categorical variables as frequency (%). Abbreviations: DHF, dengue hemorrhagic fever; RCT, randomized controlled trial; WHO, World Health Organization. a Up to 14 missing values. b Only includes patients who received a colloid infusion. c Clinically detectable pleural effusion or ascites. d Using all available acute and convalescent information.

Observational Study (n = 1216) RCT (n = 503) All Patients (N = 1719) New bleeding 77 (6) 81 (16) 158 (9) None 1138 (94) 423 (84) 1561 (91) Skin only 39 (3) 59 (12) 98 (6) Mucosal 38 (3) 22 (4) 60 (3) Severe bleeding 20 (2) 11 (2) 31 (2) Transfusion 18 (1) 8 (2) 26 (2) Inotropic support 55 (5) 20 (4) 75 (4) Total parenteral fluid volume, mL/kga 114 (99–129) 125 (110–143) 116 (102–133) Received colloida 547 (45) 418 (83) 965 (56) Total colloid volumeb, mL/kg 19 (13–25) 25 (24–25) 25 (15–29) Clinical fluid overloada,c 340 (28) 173 (34) 513 (30) DHF according to WHO 1997 criteriaa,d 796 (66) 406 (81) 1202 (70) Death 7 (<1) 1 (<1) 8 (<1) Continuous variables are summarized as median (interquartile range) and categorical variables as frequency (%). Abbreviations: DHF, dengue hemorrhagic fever; RCT, randomized controlled trial; WHO, World Health Organization. a Up to 14 missing values. b Only includes patients who received a colloid infusion. c Clinically detectable pleural effusion or ascites. d Using all available acute and convalescent information. After admission, 158 of 1719 (9%) children developed at least 1 new bleeding manifestation, among them 98 cases with skin bleeding only and 60 cases with mucosal bleeding. Considering all 126 patients with overt mucosal bleeding, gastrointestinal bleeding occurred most frequently (n = 61), compared to epistaxis (n = 36), gum bleeding (n = 22), or unusual vaginal bleeding (n = 21). The bleeding was clinically severe in 31 cases, 26 requiring transfusion (18 during active resuscitation, and 8 during the recovery phase), 4 resulting in compensated anemia at discharge, and 1 case involving a critical organ (spinal cord hemorrhage, confirmed by magnetic resonance imaging). Although most severe bleeding primarily involved the gastrointestinal tract (n = 15), 7 children had isolated severe skin bleeding, mainly at sites of invasive procedures, and 4 of 7 required transfusion. Platelet concentrates were not available during the study, but children with severe coagulopathy and active bleeding received fresh frozen plasma or other blood products at the discretion of the treating clinician.

d isolated severe skin bleeding, mainly at sites of invasive procedures, and 4 of 7 required transfusion. Platelet concentrates were not available during the study, but children with severe coagulopathy and active bleeding received fresh frozen plasma or other blood products at the discretion of the treating clinician. The evolution of hematocrit and platelet values is shown in Figure 1. The median maximum hematocrit was 50% (IQR, 47%–52%), documented at presentation in most cases (1484/1719 [86%]). Among patients with both enrollment and 1-month follow-up values, 755 of 832 (91%) had at least 20% hemoconcentration at enrollment. The hematocrit declined rapidly during the first 4 hours of resuscitation, later rising again in the majority of children. In contrast, the platelet nadir (median, 28 000 cells/µL [IQR, 19 000–40 000 cells/µL]) occurred most frequently 1 day after onset of shock (720/1718 [42%]). Although a transient drop in platelet count was seen in all cases, in 25 of 1718 cases the nadir remained >100 000 cells/µL. Coagulation profiles were performed infrequently and are not reported here, but the abnormalities observed were consistent with previous reports [17, 18]. Liver enzyme levels were checked in approximately 60% and were moderately elevated at shock, with aspartate aminotransferase levels consistently higher than alanine aminotransferase levels. Figure 1. Box plots describing changes in hematocrit (A) and platelet count (B) during the evolution of the illness. Hematocrit data are presented for the 24 hours following admission, whereas platelet data are presented daily for the first 4 days, together with the discharge day and follow-up values for both parameters. The numbers displayed below each box plot represent the number of patients included within that time interval. If multiple values were recorded during any time interval, we chose the highest hematocrit and the lowest platelet count, respectively, for that patient. The hematocrit graph excludes data from the 73 patients with dengue shock syndrome with mucosal bleeding at presentation.

patients included within that time interval. If multiple values were recorded during any time interval, we chose the highest hematocrit and the lowest platelet count, respectively, for that patient. The hematocrit graph excludes data from the 73 patients with dengue shock syndrome with mucosal bleeding at presentation. All patients would have fulfilled the 2009 WHO criteria for severe dengue, whereas only 939 of 1642 (57%) of the children with sufficient data to allow classification at enrollment would have been categorized as having dengue hemorrhagic fever (DHF). Using all available information from the acute illness and any follow-up visits, 1202 of 1705 (70%) of the patients eventually fulfilled the 4 criteria for DHF, with the remainder classified as having dengue fever by default. Outcome During the 10-year study, only 8 patients died (1 infant and 7 children; Table 3), although 2 additional DSS-associated deaths outside the study were identified from hospital records. In 3 of 8 cases, shock occurred early, on illness day 4. All 8 patients developed profound shock within the first 12 hours, requiring multiple colloid infusions plus inotropic support and with rapid development of significant fluid overload. The interval from admission to death was generally short (median, 34 hours [range, 11–87 hours]), although 1 child with multiorgan failure was taken home moribund after 4 days. Major bleeding requiring transfusion was apparent in 7 of 8 cases before death. Table 3. Selected Clinical and Laboratory Characteristics for the 8 Children Who Died

ssion to death was generally short (median, 34 hours [range, 11–87 hours]), although 1 child with multiorgan failure was taken home moribund after 4 days. Major bleeding requiring transfusion was apparent in 7 of 8 cases before death. Table 3. Selected Clinical and Laboratory Characteristics for the 8 Children Who Died Characteristic Patient 1 Patient 2 Patient 3 Patient 4 Patient 5 Patient 6 Patient 7 Patient 8 Demographic characteristics Age 11 mo 7 y 6 y 7 y 7 y 13 y 6 y 3 y Sex Male Male Female Female Male Female Male Male Year of enrollment 2000 2006 2007 2007 2007 2009 2009 2009 At presentation with shock Day of illness 5 4 5 5 5 6 4 4 Temperature, °C 39 37 38 39.2 37.5 37 37 37 Pulse rate per min 120 Rapid, weak 120 140 120 152 Rapid, weak Rapid, weak Systolic blood pressure, mm Hg 90 95 85 90 85 95 0 90 Pulse pressure, mm Hg 10 15 25 20 10 15 0 20 Bleeding Petechiae Petechiae Petechiae, GI − − Petechiae Petechiae − Abdominal tenderness + + + + + + − − Liver size, cm 4 2 3 2 2 3 1 − Hematocrit, % 41 NA NA 42 51 56 53 53 Platelet count, cells/µL 108 000 7000 35 000 38 400 NA 39 000 55 800 97 000 During hospitalization Maximum hematocrit, % 41 53 53 46 51 56 53 53 Minimum platelet count, cells/µL 108 000 7000 35 000 13 300 30 400 20 000 7000 17 700 New bleeding GI − − GI GI GI Epistaxis GI, epistaxis Transfusion + − + + + + + + Inotropic support + + + + + + + + Number of colloid boluses 2a 2 2 3 5 3 5 5 Total colloid volume, mL/kg 35.8 23.5 11.0 33.1 85.2 54.1 107.5 90.2 Clinical fluid overload + + + + + + + + Hours from admission to death 11 23 24 39 34 37 87 NAb Dengue diagnostic tests Serotype DENV-3 DENV-1 DENV-1 DENV-3 DENV-3 DENV-1 DENV-1 DENV-1 Immune status Possible primary Secondary Secondary Secondary Secondary Secondary Secondary Secondary Abbreviations: DENV, dengue virus; GI, gastrointestinal bleeding; +, yes; −, no; NA, not available.

death 11 23 24 39 34 37 87 NAb Dengue diagnostic tests Serotype DENV-3 DENV-1 DENV-1 DENV-3 DENV-3 DENV-1 DENV-1 DENV-1 Immune status Possible primary Secondary Secondary Secondary Secondary Secondary Secondary Secondary Abbreviations: DENV, dengue virus; GI, gastrointestinal bleeding; +, yes; −, no; NA, not available. a Case 1 was enrolled in the fluid randomized controlled trial and received the first bolus of colloid according to the trial randomization. b Case 8 was taken home 4 days after admission and is presumed to have died. The child was profoundly hypotensive with multiorgan failure at the time of discharge. Overt organ dysfunction was very uncommon. Other than in association with prolonged shock (Table 3, cases 7 and 8) no patient in the cohort had clinically significant hepatic, renal, or neurological compromise, except for the child with spinal cord hemorrhage and 1 other child with profound shock, liver failure, and coma. These 2 children both eventually made a full recovery with supportive care.

nged shock (Table 3, cases 7 and 8) no patient in the cohort had clinically significant hepatic, renal, or neurological compromise, except for the child with spinal cord hemorrhage and 1 other child with profound shock, liver failure, and coma. These 2 children both eventually made a full recovery with supportive care. Dengue Serotypes and Immune Status The relative abundance of dengue serotypes identified in the patient cohort over time is presented in Figure 2A. With increasingly sensitive diagnostics, successful serotype identification increased, from 30%–50% initially to >80% after 2007. In 1999, DENV-3 was the most common serotype seen, replaced by DENV-4 peaking in 2001, DENV-2 peaking in 2004, and finally by DENV-1 extending from 2005 to 2009. Almost all patients had an IgG response consistent with secondary infection, although 106 of 1719 (6%) of cases were unclassifiable. The pattern of serotypes observed was very similar to that seen among 1509 children with secondary dengue without shock enrolled into a separate observational study between 2001 and 2009 (Figure 2B; unpublished data). Figure 2. Serotype distributions over time for patients with dengue shock syndrome (A), and for children with secondary dengue who were hospitalized at the same facility but did not experience severe complications (B). The numbers below each bar are the total number of patients in whom a serotype was identified (upper line), and the total number of patients enrolled into the corresponding study (lower line). Abbreviations: DENV, dengue virus; PCR, polymerase chain reaction.

me facility but did not experience severe complications (B). The numbers below each bar are the total number of patients in whom a serotype was identified (upper line), and the total number of patients enrolled into the corresponding study (lower line). Abbreviations: DENV, dengue virus; PCR, polymerase chain reaction. Only 6 cases had clear primary infections, including 4 infants aged <18 months, and 2 children aged 7 and 12 years (Table 4). Immune status was suggestive of primary disease in 4 of the 5 other children aged <18 months (no information in 1 case), whereas all 157 children aged 18–60 months with classifiable immune status had secondary dengue. Infants may be underrepresented in the cohort, however, as many parents elect to take young children to 1 of 2 specialist pediatric hospitals nearby. All of the patients with definite primary cases recovered, although 2 infants required colloid infusions. However, one 11-month-old boy with indeterminate/possible primary dengue died with profound shock. Table 4. Selected Clinical and Laboratory Characteristics for the 6 Primary Dengue Cases

ediatric hospitals nearby. All of the patients with definite primary cases recovered, although 2 infants required colloid infusions. However, one 11-month-old boy with indeterminate/possible primary dengue died with profound shock. Table 4. Selected Clinical and Laboratory Characteristics for the 6 Primary Dengue Cases Characteristic Patient 1 Patient 2 Patient 3 Patient 4 Patient 5 Patient 6 Demographic characteristics Age 4 m 11 m 13 m 7 y 11 m 12 y Sex Male Female Male Male Female Male Year of enrollment 2008 2008 2008 2008 2009 2009 At presentation with shock Day of illness 5 5 5 5 5 6 Temperature, °C 37.5 37 37 37 37 37 Pulse rate per min 148 Rapid, weak 168 140 Rapid, weak 110 Systolic blood pressure, mm Hg 50 70 90 90 80 90 Pulse pressure, mm Hg 20 20 15 20 20 20 Bleeding Petechiae Petechiae Petechiae − Petechiae Petechiae Abdominal tenderness + + + + − − Liver size, cm 3 1 4 3 2 1 Hematocrit, % 36 47 50 56 45 46 Platelet count, cells/µL 40 000 19 000 77 000 81 400 37 900 66 000 During hospitalization Maximum hematocrit, % 36 47 50 56 45 47 Minimum platelet count, cells/µL 24 000 9000 45 800 61 000 28 400 53 000 New bleeding − − − − − − Number of colloid boluses 0 2 1 0 0 0 Total colloid volume, mL/kg 0 25.0 25.3 0 0 0 Clinical fluid overload − − + − − − Outcome Recovery Recovery Recovery Recovery Recovery Recovery Dengue diagnostic tests IgG by day of illness Day 5: (−) Day 5: (−) Day 5: (−) Day 5: (−) Day 5: (−) Day 6: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 9: (−) Serotype DENV-2 DENV-1 Negativea DENV-1 DENV-1 DENV-1 Abbreviations: +, yes; −, no; DENV, dengue virus; IgG, immunoglobulin G.

covery Recovery Dengue diagnostic tests IgG by day of illness Day 5: (−) Day 5: (−) Day 5: (−) Day 5: (−) Day 5: (−) Day 6: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 8: (−) Day 9: (−) Serotype DENV-2 DENV-1 Negativea DENV-1 DENV-1 DENV-1 Abbreviations: +, yes; −, no; DENV, dengue virus; IgG, immunoglobulin G. a Diagnosis based on positive dengue immunoglobulin M capture enzyme-linked immunosorbent assay on samples taken on days 5 and 8. DISCUSSION Here we present the first comprehensive description of the clinical features of DSS in children, using data gathered prospectively over 10 years on 1719 patients managed in a single Vietnamese institution. More than 95% of all children admitted with DSS during the study period were evaluated. Because prior shock resuscitation might confound the clinical picture, we focused on direct admissions only. A few cases were missed, including 2 children who died, but overall the results are representative of the clinical spectrum of patients with DSS admitted directly to a busy hospital in a hyperendemic region.

ed. Because prior shock resuscitation might confound the clinical picture, we focused on direct admissions only. A few cases were missed, including 2 children who died, but overall the results are representative of the clinical spectrum of patients with DSS admitted directly to a busy hospital in a hyperendemic region. Apart from infants aged <18 months, virtually all children had secondary dengue, in line with established concepts of pathogenesis [19, 20]. We observed DSS caused by all 4 dengue serotypes during the 10-year study; the pattern of serotype replacement over time was similar to that seen among children with secondary infections enrolled in a contemporaneous study of hospitalized dengue without severe manifestations, and also to the relative virus prevalence identified by passive surveillance in southern Vietnam during the same time period [21]. Thus, the viruses associated with DSS appear to be representative of the virus population affecting the wider community, with no evidence that a particular serotype contributes to a greater risk for shock. Notably, however, 3 of 8 deaths were associated with DENV-3, although the total number of DENV-3 infections identified was small. Because a number of interacting host and viral factors influence an individual's propensity to develop severe vascular leakage [19], only very detailed studies can establish whether particular viral characteristics do confer an increased risk for DSS or death.

the total number of DENV-3 infections identified was small. Because a number of interacting host and viral factors influence an individual's propensity to develop severe vascular leakage [19], only very detailed studies can establish whether particular viral characteristics do confer an increased risk for DSS or death. The clinical signs and symptoms documented were generally consistent with empirical descriptions of DSS [4]. However, 9% of all patients were still febrile at presentation. Increased permeability commences during the febrile phase, but shock develops only when leakage exceeds the capacity of the homeostatic compensatory mechanisms to maintain adequate plasma volume [22, 23], potentially compounded by functional cardiac impairment [24]. Although defervescence and shock are often temporally linked, it is important that clinicians managing suspected dengue cases understand that DSS can occur earlier. Identification of more reliable warning signs of likely deterioration would be useful both for individual case management and to facilitate effective use of limited healthcare resources.

k are often temporally linked, it is important that clinicians managing suspected dengue cases understand that DSS can occur earlier. Identification of more reliable warning signs of likely deterioration would be useful both for individual case management and to facilitate effective use of limited healthcare resources. In agreement with other studies [7, 25], a considerable number of DSS patients had no bleeding during the illness. Severe bleeding was uncommon and primarily from the gastrointestinal tract, although massive soft tissue bleeding necessitating transfusion occurred in 4 children. Again consistent with other studies [25, 26], almost one-third of cases did not achieve the WHO 1997 classification for DHF, mainly due to failure to fulfill the hemorrhage and/or plasma leakage criteria as thrombocytopenia was almost universal. If positive, a tourniquet test might have allowed classification as DHF rather than dengue fever in 357 additional cases, but several studies have demonstrated poor utility of the test in clinical practice, and it is infrequently performed in Vietnam [25, 27]. We did not perform radiological investigations to identify plasma leakage unless clinically indicated, reflecting real-world practice. Thus, diagnosis of leakage typically rested upon demonstration of hemoconcentration, yet hemoconcentration below the WHO threshold of 20% has been noted previously in DSS patients [11]. Because patients must be treated according to their actual clinical status, it is apparent that the 1997 WHO classification system is not suitable for individual case management in real time.

ation of hemoconcentration, yet hemoconcentration below the WHO threshold of 20% has been noted previously in DSS patients [11]. Because patients must be treated according to their actual clinical status, it is apparent that the 1997 WHO classification system is not suitable for individual case management in real time. The case fatality rate was extremely low (0.5%). Most patients recovered well with the standard crystalloid regimen or following a single colloid bolus, and requirement for additional colloid therapy, inotropic support, and/or blood products was infrequent. Prompt diagnosis and immediate admission to PICU with management coordinated by a highly experienced team undoubtedly contributed to this favorable outcome. In line with WHO principles [4], the unit operates a generally conservative fluid management policy after initial resuscitation, relying on frequent clinical assessments and regular ward-based hematocrit measurements to limit fluid administration to the minimum required, reducing the risk of fluid overload. However, our study focused on direct admissions, and it is clear that external referrals with prolonged shock or established fluid overload are more difficult to manage and have correspondingly higher mortality rates [8]. Only a very small number of patients with confirmed primary dengue were included in the cohort, and all recovered quickly without notable complications. However, 1 death occurred in a suspected primary case, underlining the view that primary dengue can result in severe and even fatal disease [28–30]. Given that immune status could not be defined in 6% of patients, some primary cases might have been missed, but the number is likely to be small.

ut notable complications. However, 1 death occurred in a suspected primary case, underlining the view that primary dengue can result in severe and even fatal disease [28–30]. Given that immune status could not be defined in 6% of patients, some primary cases might have been missed, but the number is likely to be small. In summary, we present a comprehensive clinical description of DSS in a large cohort of Vietnamese children. We demonstrate that with prompt intervention and assiduous clinical care by experienced staff, the outcome of this potentially fatal condition can be excellent. As the emerging dengue pandemic spreads to new geographical locations, it is important that this accumulated experience be translated into practical advice for clinicians newly exposed to this severe complication of a common disorder. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

ordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We thank all the clinical staff on the Pediatric Intensive Care Unit at the Hospital for Tropical Diseases for their assiduous care of the patients and for supporting the research program over many years. In addition, we thank the staff in the main HTD laboratories and the members of the dengue research group at the Oxford University Clinical Research Unit for their help with laboratory investigations and dengue diagnostics. We also acknowledge the senior management and administrative staff of both institutions for their longstanding support. Finally, we are grateful to all the participants and their families for generously agreeing to take part in this study. Phung Khanh Lam is registered for a PhD at the Open University UK. Financial support. The work was supported by the Wellcome Trust. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Caliendo on pages 374–5.) Human cytomegalovirus (CMV) infection causes significant morbidity and mortality in the posttransplant period of both solid organ transplant (SOT) and hematopoietic stem cell transplant (HSCT) recipients [1–5]. Seronegative recipients of transplants from seropositive donors are infected by CMV transmitted through the transplanted organ or inadvertently through CMV-positive blood products [1–5]. In CMV-seropositive recipients, reactivation of latent CMV occurs when CMV-specific immune control is impaired by immunosuppressive drugs and T-cell–depleting therapies [1–5]. The diagnosis of CMV replication and disease in SOT and HSCT recipients can be made using different laboratory methods, including histology, pp65 antigenemia, or CMV DNA by quantitative nucleic acid testing (QNAT) [1–5]. Culture methods of body fluids and tissue samples are generally slow and are not quantitative [4]. The pp65 antigenemia test is rapid (1 day time-to-result) but less sensitive than QNAT and often difficult to perform on severely neutropenic patients [1–7]. CMV QNAT, based most commonly on polymerase chain reaction (PCR), has largely replaced conventional methods owing to better overall performance, and clinical guidelines now recommend the use of these assays for CMV load monitoring in SOT and HSCT recipients to prevent or to manage CMV replication and disease [4, 5, 7, 8].

QNAT, based most commonly on polymerase chain reaction (PCR), has largely replaced conventional methods owing to better overall performance, and clinical guidelines now recommend the use of these assays for CMV load monitoring in SOT and HSCT recipients to prevent or to manage CMV replication and disease [4, 5, 7, 8]. One central issue that has emerged with the use of different CMV PCR tests is the significant interassay quantification variability, as demonstrated in multicenter studies with standardized panels [9, 10]. This lack of assay agreement complicates the management of individual patients who may have testing performed in different laboratories and it has hampered the establishment of broadly applicable quantitative cutoff values that can be used in clinical decision making, potentially negatively impacting the management and long-term outcome of patients at risk of the direct and indirect effects of CMV replication [1, 4, 5]. Therefore, in the clinical management of CMV after transplantation, there is a significant unmet medical need for the development of standardized nucleic acid tests that deliver comparable quantitative data across different laboratories.

patients at risk of the direct and indirect effects of CMV replication [1, 4, 5]. Therefore, in the clinical management of CMV after transplantation, there is a significant unmet medical need for the development of standardized nucleic acid tests that deliver comparable quantitative data across different laboratories. The first international standard for CMV QNAT has recently been established by the World Health Organization (WHO) Expert Committee on Biological Standardization [11]. This CMV standard should help to improve interassay agreement. However, assay-specific variability is still expected owing to underlying differences in test constituents, including varying nucleic acid extraction methods, target-specific amplification efficiencies, assay biochemistries, and operator-dependent variability. As recently suggested [12], these residual quantification disparities could be solved through the widespread availability of commercial PCR tests that encompass all assay steps (nucleic acid preparation, reaction setup, calibration, amplification, and detection) and demonstrate reliable interlaboratory quantification as defined by agreement and precision. Here, we report the results of a multicenter international study designed to determine the comparability of quantitative data and precision of a new, fully automated, Food and Drug Administration–approved CMV QNAT (COBAS AmpliPrep/COBAS TaqMan CMV Test [CAP/CTM CMV test]) using a blinded panel across diverse laboratories and to compare these 2 parameters among the 5 different assays currently used in these laboratories. Agreement between the CAP/CTM CMV test and the diverse in-house quantification assays was also defined using clinical plasma specimens from immunocompromised individuals.

sing a blinded panel across diverse laboratories and to compare these 2 parameters among the 5 different assays currently used in these laboratories. Agreement between the CAP/CTM CMV test and the diverse in-house quantification assays was also defined using clinical plasma specimens from immunocompromised individuals. METHODS CAP/CTM CMV Test Colinearity to the First WHO CMV International Standard The CAP/CTM CMV test (Roche Molecular Systems, Inc [RMS], Branchburg, New Jersey) analytical performance characteristics, including traceability to the first WHO CMV international standard for nucleic acid amplification techniques (NIBSC 09/162) were described previously [13]. In brief, the CAP/CTM CMV test uses primers and probes targeting a conserved region of the CMV genome (UL54, virus encoded DNA polymerase) and has a linear quantification range from 150 to 10 000 000 (2.18–7.0 log10) copies/mL representing 137 and 9 100 000 (2.14–6.96 log10) IU/mL, respectively (1 copy = 0.91 IU). Several standards and control specimens were used during the development of the test to achieve traceability to the first WHO CMV international standard as recommended by the Clinical and Laboratory Standards Institute guidelines [14]. The standards included the WHO CMV standard, RMS CMV secondary standard, RMS CMV secondary standard source material (CMV strain AD169), and RMS CMV calibration panel (Lambda CMA1.2). The standards, the calibration panel, and an independent CMV clinical specimen were tested at similar levels to determine whether colinearity to the WHO CMV standard was achieved. Assessment of colinearity was performed to demonstrate that the WHO standard was commutable at any given titer throughout the measuring range and to thereby ensure traceability. The concentration range tested for the WHO CMV standard was from 500 IU/mL to 50 000 IU/mL (2.70–5.70 log10 IU/mL), the RMS CMV secondary standard source material was tested from 500 IU/mL to 10 million IU/mL (2.70–7.00 log10 IU/mL), the RMS CMV calibration panel was tested from 523 to 9.3 million IU/mL (2.72–6.97 log10 IU/mL), and the independent CMV clinical specimen was tested from 500 IU/mL to 22 686 IU/mL (2.70–4.36 log10 IU/mL). The standard and control specimens were demonstrated to be colinearly distributed to the WHO material across the linear range of the CAP/CTM CMV test (Supplementary Figure 1).

lion IU/mL (2.72–6.97 log10 IU/mL), and the independent CMV clinical specimen was tested from 500 IU/mL to 22 686 IU/mL (2.70–4.36 log10 IU/mL). The standard and control specimens were demonstrated to be colinearly distributed to the WHO material across the linear range of the CAP/CTM CMV test (Supplementary Figure 1). CMV DNA Quantification Tests CMV DNA quantification with the CAP/CTM CMV test was compared to in-house tests of record at each of 5 academic centers including The John Hopkins Hospital (site 2, real-time PCR based on Artus reagents [Qiagen, Germantown, Maryland]), Hospital Universitario de la Princesa (site 3, Affigene real-time PCR test [Cepheid, Sunnyvale, California]) University of Basel (site 4, user-defined real-time PCR [UL111a gene target]), Stanford University (site 5, COBAS AMPLICOR MONITOR CMV test [RMS]), and Helsinki University Hospital (site 6, user-defined real-time PCR [pp65 gene target]). The analytical performance characteristics for commercial and laboratory-developed quantitative PCR assays have been described previously [15–19].

gene target]), Stanford University (site 5, COBAS AMPLICOR MONITOR CMV test [RMS]), and Helsinki University Hospital (site 6, user-defined real-time PCR [pp65 gene target]). The analytical performance characteristics for commercial and laboratory-developed quantitative PCR assays have been described previously [15–19]. Comparability and Reproducibility of the CAP/CTM CMV Test and 5 Quantitative PCR Assays Comparability and reproducibility of the CAP/CTM CMV test was studied in comparison with 3 assays based on commercial reagents, and 2 tests that use laboratory-developed primers and probes using a panel prepared from a well-characterized CMV cultured virus stock (strain AD-169, titer assigned by the COBAS AMPLICOR CMV MONITOR Test). The panel consisted of 6 dilutions; 150, 550, 2000, 20 000, 50 000, and 5 000 000 copies/mL (2.18, 2.74, 3.3, 4.3, 4.7, and 6.7 log10 copies/mL, respectively). These dilutions covered the dynamic range of the CAP/CTM CMV test and represent relevant clinical viral load thresholds [1–7, 13]. The prepared virus stock dilutions were further diluted in CMV-negative human ethylenediaminetetraacetic acid plasma. Each site tested 15 replicates of each panel member with CAP/CTM CMV and in-house tests except site 2 (15 replicates by CAP/CTM CMV/12 replicates of in-house test). The CMV DNA panel was prepared at RMS and shipped to the study sites labeled with coded sample identification numbers to ensure that the site study staff was blinded to the CMV DNA concentration of each panel member. These experiments were designed in accordance with guidelines for establishing analytical performance characteristics of QNATs [20].

d at RMS and shipped to the study sites labeled with coded sample identification numbers to ensure that the site study staff was blinded to the CMV DNA concentration of each panel member. These experiments were designed in accordance with guidelines for establishing analytical performance characteristics of QNATs [20]. Quantitative Agreement Between CAP/CTM CMV Test and 5 Quantitative PCR Assays Using Clinical Specimens From Immunocompromised Patients Agreement between CAP/CTM CMV test and the 5 quantitative PCR assays described above was investigated with plasma samples collected at the study sites from immunocompromised patients monitored for CMV replication and disease. Specimens were assayed only at the site that performed original testing. In addition, 403 samples from 135 HSCT recipients participating in the maribavir prophylaxis for prevention of CMV phase 3 trial (NCT00411645) were provided by ViroPharma, Inc, to the study sites for PCR testing [21]. Only patients with plasma samples with a volume >600 μL were included in this analysis, and due to volume requirements for in-house testing, site 2 did not quantify samples from the maribavir prophylaxis trial. Plasma samples from the maribavir prevention trial were randomly distributed to the other 4 study sites. Institutional review board approval was obtained at each institution for this study.

alysis, and due to volume requirements for in-house testing, site 2 did not quantify samples from the maribavir prophylaxis trial. Plasma samples from the maribavir prevention trial were randomly distributed to the other 4 study sites. Institutional review board approval was obtained at each institution for this study. Statistical Analysis The precision of log10-transformed valid test results within the linear range of each assay was estimated at each expected log10 CMV DNA concentration. The log-normal mean and log-normal coefficient of variation (%) and 95% confidence intervals (CIs) (including the lower and upper confidence limits) for total variance were calculated using the linear mixed effect model with site and day/run, and within-run as random effects. Deming regression analysis of the viral load results for each local assay vs the CAP/CTM CMV test was performed to evaluate the correlation between the assays overall and by study site. All the statistical analyses were performed using the statistical software SAS version 9.2.

Statistical Analysis The precision of log10-transformed valid test results within the linear range of each assay was estimated at each expected log10 CMV DNA concentration. The log-normal mean and log-normal coefficient of variation (%) and 95% confidence intervals (CIs) (including the lower and upper confidence limits) for total variance were calculated using the linear mixed effect model with site and day/run, and within-run as random effects. Deming regression analysis of the viral load results for each local assay vs the CAP/CTM CMV test was performed to evaluate the correlation between the assays overall and by study site. All the statistical analyses were performed using the statistical software SAS version 9.2. RESULTS Comparability and Reproducibility of Quantitative PCR Assays With a Standardized CMV DNA Panel To determine comparability of quantitative data obtained with CAP/CTM CMV test across laboratories vs participating centers' PCR assays, plasma panels spiked with CMV strain AD-169 from 2.18–6.7 log10 copies/mL were tested at the 5 study sites. The level of quantitative agreement was high for the CAP/CTM CMV test across the different laboratories as demonstrated by a smaller range of mean concentrations of panel members by site and narrower CIs of the combined data per panel member compared to in-house PCR test in clinical use at the study sites (Figure 1). For CAP/CTM CMV, the greatest quantitative variability was observed for the lowest concentration panel member (2.18 log10 copies/mL, the test's lower limit of quantification); all replicates were detected, but 62/75 (83%) could not be quantified. This panel member was also variably detected by each comparator PCR assay. The AMPLICOR assay (site 5) failed to detect 11/15 replicates (73%), whereas the Affigene CMV trender test (site 3) was able to quantify 15/15 replicates. Overall, of 72 valid comparator PCR assay results, 12 replicates were not detected, 22 were below the lower limit of quantification, and 38 (53%) were quantified (Supplementary Table 1). Figure 1. Comparability of quantitative data across laboratory sites. A dilution series of cytomegalovirus (CMV) AD-169 was prepared using cytomegalovirus-seronegative plasma. Geometric means of tested replicates are plotted. Numerical ranges indicate 95% confidence intervals of the means of each panel member at each site. A, COBAS AmpliPrep/COBAS TaqMan CMV data. Dashed line indicates assay lower limit of quantification. B, Data from 5 comparator polymerase chain reaction assays. Abbreviations: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test; CMV, cytomegalovirus.

% confidence intervals of the means of each panel member at each site. A, COBAS AmpliPrep/COBAS TaqMan CMV data. Dashed line indicates assay lower limit of quantification. B, Data from 5 comparator polymerase chain reaction assays. Abbreviations: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test; CMV, cytomegalovirus. Data from experiments with the spiked plasma panel were also used to compare the precision of the CAP/CTM CMV test to the in-house PCR assays used by the study sites. The standard deviation was <0.2 log10 copies/mL for most panel members across the different sites performing the CAP/CTM CMV test, with 2 exceptions (2.18 and 6.7 log10 copies/mL for site 2 and site 6 tests, respectively, Figure 2). Similar reproducibility was observed for the AMPLICOR test (site 5), CMV real-time PCR (UL111a PCR, site 4) and real-time PCR (pp65 PCR, site 6) in-house assays. The Artus CMV PCR (site 2) and Affigene CMV (site 3) trender tests demonstrated greater imprecision (standard deviations >0.2 log10 copies/mL for multiple panel members). Coefficients of variation demonstrated similar trends (Supplementary Figure 2). Figure 2. Reproducibility of quantitative data across laboratory sites. Standard deviations of replicates from cytomegalovirus (CMV) AD-169 dilution series are plotted. A, COBAS AmpliPrep/COBAS TaqMan CMV Test precision by study site. Quantifiable replicates at 2.18log10 copies/mL: site 2, n = 2; site 3, n = 0; site 4, n = 6; site 5, n = 1; site 6, n = 4. Standard deviations at 2.18log10 copies/mL were not calculated for sites 3 and 5. At 2.74log10 copies/mL, all replicates were quantifiable at all sites except site 6 (11/15 quantifiable). B, Precision for the 5 comparator polymerase chain reaction assays. Replicates of each panel member tested at site 2, n = 12; at sites 3–6, n = 15. Quantifiable replicates at 2.18log10 copies/mL: site 2, n = 2; site 3, n = 15; site 4, n = 11; site 5, n = 0; site 6, n = 10. Standard deviation at 2.18log10 copies/mL was not calculated for site 3. At 2.74log10 copies/mL, all replicates were quantifiable at all sites except site 4 (0/15 quantifiable) and site 5 (9/15 quantifiable). Abbreviation: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test.

ite 4, n = 11; site 5, n = 0; site 6, n = 10. Standard deviation at 2.18log10 copies/mL was not calculated for site 3. At 2.74log10 copies/mL, all replicates were quantifiable at all sites except site 4 (0/15 quantifiable) and site 5 (9/15 quantifiable). Abbreviation: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test. Quantitative Agreement Between CAP/CTM CMV Test and 5 Quantitative PCR Assays Using Clinical Specimens From Immunocompromised Patients Trends in quantitative disagreement between the CAP/CTM CMV test and study sites' PCR assays were defined by comparing viral loads obtained by the 2 tests on individual plasma samples from immuncompromised patients. HSCT recipients comprised 67% (267/396) of the patients studied; these individuals contributed 80% and 76% of the total number of samples and valid PCR test results, respectively (Table 1). Two patterns of disparity were observed (Figure 3). CAP/CTM CMV test yielded lower values than 3 in-house tests (sites 2, 4, and 6) throughout the measuring range. Bland-Altman analysis demonstrated this constant bias (least squares regression slope absolute value <.1, with nonsignificant P value). Additionally, for in-house assays at sites 3 and 5, the difference in quantification compared to the CAP/CTM CMV test varied throughout the measuring range. Bland-Altman analysis further demonstrated this proportional bias (least squares regression slope absolute value >.1, with significant P value, Figure 3). Table 1. Patient Populations and Number of Samples Used in the Reproducibility Comparison of COBAS AmpliPrep/COBAS TaqMan CMV Test and 5 Quantitative Polymerase Chain Reaction Assays

ysis further demonstrated this proportional bias (least squares regression slope absolute value >.1, with significant P value, Figure 3). Table 1. Patient Populations and Number of Samples Used in the Reproducibility Comparison of COBAS AmpliPrep/COBAS TaqMan CMV Test and 5 Quantitative Polymerase Chain Reaction Assays Patient Population Total No. of Samples Tested No. of Valid Test Resultsa Solid organ transplant recipients (n = 107) 107 71 Hematopoietic stem cell transplant recipients (n = 267) 531 286 HIV-infected patients (n = 9) 9 9 Other immunocompromised patients (n = 13)b 13 12 Abbreviation: HIV, human immunodeficiency virus. a Only samples with paired results within the linear range of the polymerase chain reaction assays were included in the comparison analysis. b Other immunocompromised patients included subjects diagnosed with hematologic malignancies (n = 9) or autoimmune diseases (n = 4) receiving immunosuppressive therapy.

Patient Population Total No. of Samples Tested No. of Valid Test Resultsa Solid organ transplant recipients (n = 107) 107 71 Hematopoietic stem cell transplant recipients (n = 267) 531 286 HIV-infected patients (n = 9) 9 9 Other immunocompromised patients (n = 13)b 13 12 Abbreviation: HIV, human immunodeficiency virus. a Only samples with paired results within the linear range of the polymerase chain reaction assays were included in the comparison analysis. b Other immunocompromised patients included subjects diagnosed with hematologic malignancies (n = 9) or autoimmune diseases (n = 4) receiving immunosuppressive therapy. Figure 3. Agreement in cytomegalovirus (CMV) DNA load measurement between the COBAS AmpliPrep/COBAS TaqMan CMV Test (CAP/CTM CMV) and 5 comparator polymerase chain reaction (PCR) assays in plasma samples from immunocompromised individuals. Left-hand panels, agreement plots with Deming regression lines; dashed line indicates 100% agreement level. Right-hand panels, Bland-Altman plots (difference in quantification between the CAP/CTM CMV test and comparator PCR assays vs mean of the 2 measurements). Solid line, least squares regression; dashed lines, mean differences of +0.5/0/−0.5 log10 copies/mL. Abbreviation: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test.

ight-hand panels, Bland-Altman plots (difference in quantification between the CAP/CTM CMV test and comparator PCR assays vs mean of the 2 measurements). Solid line, least squares regression; dashed lines, mean differences of +0.5/0/−0.5 log10 copies/mL. Abbreviation: CAP/CTM CMV, COBAS AmpliPrep/COBAS TaqMan CMV Test. DISCUSSION The data from this multicenter international study demonstrate that the CAP/CTM CMV test performs consistently across laboratories, from the perspective of quantitative agreement and reproducibility. Furthermore, the finding of constant and variable quantification differences among PCR assays currently used at the participating centers compared with the CAP/CTM CMV test underscores the challenges in achieving a general quantitative standardization. Constant bias between different assays likely reflects the use of different calibrators in assays that are otherwise functionally similar. In these instances, interassay agreement may improve with adoption of a calibrator based on the international standard; alternatively, a conversion factor can be applied to normalize data if the use of a different calibrator is not feasible. These fairly easy adjustments are unlikely to improve agreement between assays whose functionality is sufficiently different to result in variable quantification differences throughout the measuring range. For these assays (with variable quantification), assay traceability to the international standard alone will not adequately correct for these types of differences. Instead, these assays must also demonstrate colinearity to the international standard throughout the assay measuring range. Ideally, this approach (to calibrate and establish colinearity to the reference material, eg, the first WHO CMV international standard) should be used whenever samples are evaluated in patients who are monitored for CMV DNA as part of the management of SOT and HSCT recipients.

dard throughout the assay measuring range. Ideally, this approach (to calibrate and establish colinearity to the reference material, eg, the first WHO CMV international standard) should be used whenever samples are evaluated in patients who are monitored for CMV DNA as part of the management of SOT and HSCT recipients. Monitoring CMV DNA has become critical for the early identification of viral replication for preventing progression to disease in the posttransplant period, and for monitoring the response to antiviral treatment in patients with CMV replication and disease. In SOT, universal prophylaxis and preemptive treatment approaches are both used to prevent CMV disease [1, 2, 4, 5]. Although both approaches are currently viewed as equivalent, the advantages of prophylaxis are prevention of replication and disease in the immediate posttransplant period and elimination of the need for viral load monitoring during prophylaxis, particularly for the high-risk, CMV-seropositive donor/CMV-seronegative recipient SOT. However adverse events associated with prolonged antiviral drug administration have limited the utility of this approach in some patients. Also, viral load monitoring may still be useful in some patients in whom drug-resistant viruses are suspected to emerge when treated with lower antiviral doses in an attempt to mitigate drug side effects. Finally, the onset of late CMV disease has been observed in up to 29% of SOT recipients after prophylaxis cessation [1, 2, 4].

load monitoring may still be useful in some patients in whom drug-resistant viruses are suspected to emerge when treated with lower antiviral doses in an attempt to mitigate drug side effects. Finally, the onset of late CMV disease has been observed in up to 29% of SOT recipients after prophylaxis cessation [1, 2, 4]. Viral load monitoring is the key feature of preemptive therapy. In this alternate approach, antiviral therapy is initiated before the onset of CMV disease when viral load measurements reach a predictive threshold. Advantages of preemptive therapy include a smaller proportion of treated patients, shortened therapeutic duration, lowered costs associated with posttransplant medications, and reduced occurrence of drug toxicity (primarily bone marrow suppression). Avoidance of marrow suppression is the major rationale for the use of preemptive strategies in allogeneic HSCT in the preengraftment period. Disadvantages of the preemptive antiviral strategy include risk of disease prior to treatment initiation in individuals with fast replication, indirect effects of CMV replication (in the absence of disease) on allograft survival and mortality, and laboratory costs due to more frequent viral load monitoring. Perhaps one of the most significant drawbacks to this approach is the lack of generally established quantitative cutoffs that are predictive of CMV disease in recipients of different allografts due in large part to the lack of standardized quantification assays that perform comparably across laboratories.

d monitoring. Perhaps one of the most significant drawbacks to this approach is the lack of generally established quantitative cutoffs that are predictive of CMV disease in recipients of different allografts due in large part to the lack of standardized quantification assays that perform comparably across laboratories. The lack of standardized CMV load assays has also complicated CMV management in other ways. For individual patients, quantification disparities across laboratories dictate that a single laboratory should be used for viral load testing, so that results can be accurately interpreted. In addition, although higher viral loads have been shown to correlate with an increased risk of disease in SOT and HSCT patients [1, 4, 5], the lack of standardization has hampered the development of discrete, globally applicable, quantitative predictors of active disease, and other cutoffs that can be used to determine relapse risk and adequate treatment duration [12]. Currently, the burden of defining these cutoffs is placed on individual laboratories, and as a result, clinically relevant values still vary from center to center.

ly applicable, quantitative predictors of active disease, and other cutoffs that can be used to determine relapse risk and adequate treatment duration [12]. Currently, the burden of defining these cutoffs is placed on individual laboratories, and as a result, clinically relevant values still vary from center to center. The implementation of an international standard and the availability of commercial QNATs with broad interlaboratory agreement that are traceable and colinear to the first WHO CMV international standard represent a much-needed advancement. As demonstrated here for CAP/CTM CMV, precise, accurate, and standardized results should allow the design of multicenter studies to delineate testing algorithms, including quantitative cutoffs and testing frequencies that enhance clinical outcomes of CMV infections in HSCT and SOT patients. In turn, these data can be used as the basis for management guidelines that should significantly clarify decision making for clinicians and improve infection outcomes in at-risk patients. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://www.oxfordjournals.org/our_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

r_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. The authors acknowledge Dr Stephen Villano at ViroPharma, Inc, for providing samples from patients in the marabivir phase 3 trial (NCT00411645); Michael Forman at Johns Hopkins Hospital; Dr Alexis Dumoulin and the technicians at the Abteilung Infektionsdiagnostik, University of Basel; Dr Laura Mannonen, Dr Raisa Loginov, Dr Ilkka Helanterä, and the technicians at the Helsinki University Hospital; Elisea Lomas at the Hospital Universitario de la Princesa; and Dr Tri Do and Ula Cowen at Roche Molecular Systems, Inc, for assistance provided in the preparation of the study. Financial support. This work was supported by Roche Molecular Systems, Inc, Pleasanton, California. Potential conflicts of interest. H. H. H. has received speaker honoraria from Roche Diagnostics; S. A. and B. C. are employees of Roche Molecular Systems, Inc; R. A. V. was an employee of Roche Molecular Systems, Inc; A. V. has been a member of the Roche Diagnostics Scientific Advisory Board, has been a consultant for Qiagen, and has received clinical trial funding from Roche Molecular Systems and Qiagen. All other authors report no potential conflicts.

olecular Systems, Inc; R. A. V. was an employee of Roche Molecular Systems, Inc; A. V. has been a member of the Roche Diagnostics Scientific Advisory Board, has been a consultant for Qiagen, and has received clinical trial funding from Roche Molecular Systems and Qiagen. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Sawe and Lockman on pages 447–9.) Effective medications for the prevention of mother-to-child human immunodeficiency virus (HIV) transmission (PMTCT) can reduce perinatal HIV transmission to <2% in the absence of breastfeeding and to <5% by 6 months of age among breastfeeding infants [1–3]. As a result, the World Health Organization (WHO) has called for the “virtual elimination” of pediatric HIV [1–3]. Access to antiretroviral medications (ARVs) for PMTCT remains limited, however; only 59% of HIV-infected pregnant women received ARVs for PMTCT in 2010 [4]. As a result, nearly 400 000 new infant HIV infections occur annually, and HIV-infected women experience high postpartum morbidity and mortality [4–6]. In 2010, WHO issued revised guidelines for PMTCT [1]. The guidelines included a renewed emphasis on identification of pregnant, HIV-infected women with CD4 count ≤350 cells/µL or WHO stage 3–4 disease, who require lifelong 3-drug antiretroviral therapy (ART) for treatment of their own HIV infections and for PMTCT. For women with less-advanced disease, WHO recommends a country- or program-level choice between Option A (maternal zidovudine in pregnancy; infant nevirapine [NVP] throughout breastfeeding), and Option B (maternal 3-drug ARV regimens throughout pregnancy and breastfeeding, with interruption after weaning). Select programs are considering Option B+, in which maternal 3-drug regimens are initiated in pregnancy (regardless of maternal CD4) and continued throughout life, including throughout breastfeeding and subsequent pregnancies [3, 7].

ARV regimens throughout pregnancy and breastfeeding, with interruption after weaning). Select programs are considering Option B+, in which maternal 3-drug regimens are initiated in pregnancy (regardless of maternal CD4) and continued throughout life, including throughout breastfeeding and subsequent pregnancies [3, 7]. HIV prevalence in antenatal care (ANC) is estimated at 16% in Zimbabwe, leading to approximately 61 000 births per year to HIV-infected women [8, 9]. Through 2009, the Zimbabwe National PMTCT Program provided single-dose NVP (sdNVP) to all HIV-infected women, with ART for women identified clinically as ART eligible [8]. Like most countries in sub–Saharan Africa, Zimbabwe initially implemented the revised WHO guidelines with Option A (with antenatal coverage of 46% in 2010) and will soon be examining the feasibility of Options B and B+ [4]. We used validated computer models of HIV disease and PMTCT [10–12] to project the clinical outcomes and cost effectiveness of implementing WHO-recommended PMTCT regimens in Zimbabwe.

WHO guidelines with Option A (with antenatal coverage of 46% in 2010) and will soon be examining the feasibility of Options B and B+ [4]. We used validated computer models of HIV disease and PMTCT [10–12] to project the clinical outcomes and cost effectiveness of implementing WHO-recommended PMTCT regimens in Zimbabwe. METHODS Analytic Overview We used 3 validated, linked computer models for this analysis (Figure 1): (1) a model of a single pregnancy and delivery (the mother-to-child HIV transmission [MTCT] model [10]); (2) the Cost-effectiveness of Preventing AIDS Complications (CEPAC) model of HIV infection and mortality among breastfed infants (the CEPAC infant model [13, 14]); and (3) the CEPAC-International model of HIV disease progression among postpartum women (the CEPAC adult model [11, 12, 15]). Clinical outcomes of the linked models included infant HIV infection risk at weaning, maternal life expectancy (LE) from delivery, and infant LE from birth. Economic outcomes, from the healthcare system perspective, included ANC costs (through delivery), maternal HIV-related healthcare costs, and infant healthcare costs. Figure 1. Model structure. Three linked models were used for this analysis, as described in the Methods, as well as in the Supplementary Appendix and previous work [10, 14, 15]. The mother-to-child human immunodeficiency virus transmission model includes events during pregnancy and delivery (left panel; Supplementary Figure 1). The Cost-effectiveness of Preventing AIDS Complications (CEPAC) adult model includes events occurring among mothers after delivery (bottom right panel; Supplementary Figure 2A), and the CEPAC infant model includes events for infants after birth (top right panel; Supplementary Figure 2B). Linkages between the models allow a combined analysis in which each woman–infant pair is simulated together from the time of first presentation at antenatal care through pregnancy and delivery, and then each woman and infant are simulated separately throughout their lifetimes. Abbreviations: ANC, antenatal care; ART, 2-drug antiretroviral therapy; ARVs, antiretroviral medications; HIV, human immunodeficiency virus; OI, opportunistic infection; PMTCT, prevention of mother-to-child HIV transmission; sdNVP, single-dose nevirapine.

and infant are simulated separately throughout their lifetimes. Abbreviations: ANC, antenatal care; ART, 2-drug antiretroviral therapy; ARVs, antiretroviral medications; HIV, human immunodeficiency virus; OI, opportunistic infection; PMTCT, prevention of mother-to-child HIV transmission; sdNVP, single-dose nevirapine. Incremental cost-effectiveness ratios (ICERs), in US dollars per year of life saved (YLS), were calculated from combined projected lifetime healthcare costs (antenatal + maternal + infant) and combined projected life expectancy (maternal + infant) [16], discounted at 3% per year. We used 2 criteria to interpret cost-effectiveness. First, following WHO guidance, an intervention was considered cost-effective if its ICER compared with the next least-expensive alternative was <3 times the 2008 Zimbabwe per capita gross domestic product, or 3 × $400 = $1200 per YLS [17, 18]. Second, we compared results with the recently reported range of ICERs for ART-related interventions in developing countries ($550–$5200 per YLS) [19]. This work was approved by the Partners Healthcare Institutional Review Board, Boston, Massachusetts.

capita gross domestic product, or 3 × $400 = $1200 per YLS [17, 18]. Second, we compared results with the recently reported range of ICERs for ART-related interventions in developing countries ($550–$5200 per YLS) [19]. This work was approved by the Partners Healthcare Institutional Review Board, Boston, Massachusetts. Modeled Population, PMTCT Regimens, and Uptake of PMTCT Services The linked models were used to simulate a cohort of pregnant, HIV-infected women in Zimbabwe and their infants. We examined 5 PMTCT regimens: (1) no antenatal ARVs (comparator), (2) sdNVP, (3) WHO Option A, (4) WHO Option B, and (5) Option B+ (Supplementary Table 1). Women were modeled to present to care at 24–28 weeks’ gestation and to breastfeed for 18 months, based on Zimbabwean data [20, 21], with ARV prophylaxis (Options A, B, and B+ ) continued throughout breastfeeding.

RVs (comparator), (2) sdNVP, (3) WHO Option A, (4) WHO Option B, and (5) Option B+ (Supplementary Table 1). Women were modeled to present to care at 24–28 weeks’ gestation and to breastfeed for 18 months, based on Zimbabwean data [20, 21], with ARV prophylaxis (Options A, B, and B+ ) continued throughout breastfeeding. To demonstrate the impact of guideline-concordant care, all women in the base-case analyses were assumed to be identified as HIV-infected at their first ANC visit. With no ARVs, women received no antiretroviral medications during pregnancy. With sdNVP, women initiated ART in pregnancy if clinical assessment indicated WHO stage 3–4 disease; CD4 testing was not included, reflecting its limited availability in the sdNVP-based National PMTCT Program in 2009. With Options A and B, women received ART during pregnancy if eligible by either CD4 or clinical criteria, and with Option B+, all women received lifelong ART. With all modeled regimens, women who linked to postnatal HIV care were assumed to undergo clinical and CD4 assessment at 6 weeks postpartum and to initiate ART if eligible, regardless of antenatal regimen received. In the base case, we assumed 100% adherence to PMTCT regimens (initiated at 30 weeks’ gestational age), 100% linkage to postnatal care for mothers and infants, and 100% retention in care and ART availability for women and infants meeting WHO ART initiation criteria [22, 23]. In sensitivity analyses, to reflect real-world programs, we examined reduced access to antenatal and postnatal care.

at 30 weeks’ gestational age), 100% linkage to postnatal care for mothers and infants, and 100% retention in care and ART availability for women and infants meeting WHO ART initiation criteria [22, 23]. In sensitivity analyses, to reflect real-world programs, we examined reduced access to antenatal and postnatal care. Model Structure The 3 simulation models are described in detail in the Supplementary Appendix and in previous publications [10, 14, 15]. The models were linked so that each mother–infant pair was simulated together from the time of first presentation at ANC through delivery (the MTCT model), and then each woman and infant were simulated separately over their lifetimes after delivery (the CEPAC adult and infant models), as in Figure 1 and Supplementary Figures 1 and 2. Model Input Parameters Maternal Characteristics, Disease Progression, and ART Based on Zimbabwean data, mean age at first ANC visit was 24 years [21]; mean CD4 count was 451 cells/µL (36% of women with CD4 count ≤350 cells/µL) [24]. Because detailed data to inform monthly risks of opportunistic infections (OIs) and HIV-related death in the absence of ART were not available from Zimbabwe, we derived these data from a cohort in South Africa (Supplementary Table 2) [25]. Details of ART initiation and switching, as well as CD4 and HIV RNA changes on ART, are provided in the Supplementary Appendix.

f opportunistic infections (OIs) and HIV-related death in the absence of ART were not available from Zimbabwe, we derived these data from a cohort in South Africa (Supplementary Table 2) [25]. Details of ART initiation and switching, as well as CD4 and HIV RNA changes on ART, are provided in the Supplementary Appendix. MTCT Risks, Infant Mortality Rates, and Infant Life Expectancy Estimates Risks of MTCT during pregnancy and breastfeeding were calculated from PMTCT studies among breastfeeding populations in Africa, leading to estimates similar to those derived by the Joint United Nations Programme on HIV/AIDS (Supplementary Appendix) [10, 26]. Data and assumptions to inform infant mortality rates and LE values are shown in Table 1 and detailed in the Supplementary Appendix. Table 1. Selected Model Input Parameters

ions in Africa, leading to estimates similar to those derived by the Joint United Nations Programme on HIV/AIDS (Supplementary Appendix) [10, 26]. Data and assumptions to inform infant mortality rates and LE values are shown in Table 1 and detailed in the Supplementary Appendix. Table 1. Selected Model Input Parameters Variable Value Data Sources Clinical Model Input Parameters Baseline Maternal Cohort Characteristics Age, mean, y (SD) 24 (5) MOHCW [21] Mortality during pregnancy 0.7% MOHCW [8] Proportion ART eligiblea 36% ZVITAMBO trial [24] CD4 count, cells/µL (SD) Total cohort 451 (50) ZVITAMBO trial [24] ART-eligible women 275 (50) ZVITAMBO trial [24] Non-ART-eligible women 550 (50) ZVITAMBO trial [24] Uptake of PMTCT services and postnatal care PMTCT uptakeb 100% (sensitivity analyses: 56%, 80%, 95%) WHO [1] Sensitivity of clinical assessment of ART eligibility 36% MTCT-Plus Cohort [47] Probability of linking to pediatric HIV diagnosis, care, and ART 100% (sensitivity analysis: 36%) WHO/UNICEF [48] Probability of linking to postnatal maternal HIV-related care 100% (sensitivity analyses: 87% if ANC received, 43% if no ANC received) After ANC: Mean of published values [49–54] No ANC: assumption Loss to follow-up from postnatal maternal care 0% per year (sensitivity analyses: 16% [year 1]; 6% per year [years ≥2]) [30–32] Base Case Value (range for sensitivity analysis) Maternal HIV Status Mother-to-Child Transmission Risks PMTCT Regimen Received Intrauterine/intrapartum period (one-time risks) No ARVs sdNVP Antenatal ZDVc 3-Drug Regimen Data Sources ART eligible at conception 0.273 (0.199–0.322) 0.176 (0.082–0.264) 0.136 (0.091–0.157) 0.033 (0.011–0.041) [24, 55–69] Non-ART eligible at conception 0.175 (0.127–0.206) 0.073 (0.033–0.109) 0.036 (0.024–0.041) 0.01 (0.004–0.028) [24, 55–64] [66, 67, 69–71] Postnatal period (rate per 100 person-years among HIV-uninfected infants aged 4-6 weeks) No ARVs Extended Infant NVP 3-Drug Regimen Data Sources ART eligible 9.1 (EBF); 15.4 (MBF) (5.7–28.4) NA 4.0 (0–6.4) [24, 57, 59, 65, 67, 69–72] Non-ART eligible 2.9 (EBF); 4.8 (MBF) (1.8–8.8) 2.7 (1.4–3.7) 2.2 (0–6.4) [24, 52, 59, 67, 70–77] Infant Mortality and Life Expectancy Probability of live birth 95.7%–98.0% MOHCW [21] Relative increase in infant mortality if maternal death occurs 2-fold increase [78–81] Short-term mortality risks, % 1-year risk 2-year cumulative risk HIV-exposed, uninfected children 7.4 [82] 9.2 [82] HIV-infected children, no ART Intrauterine/intrapartum infection

cy Probability of live birth 95.7%–98.0% MOHCW [21] Relative increase in infant mortality if maternal death occurs 2-fold increase [78–81] Short-term mortality risks, % 1-year risk 2-year cumulative risk HIV-exposed, uninfected children 7.4 [82] 9.2 [82] HIV-infected children, no ART Intrauterine/intrapartum infection 51.0 [83] 65.0 [83] Postpartum infection 24.0 [83] 38.0 [83] HIV-infected children, on ART 9.5 [84] 12.0 [85] Life-expectancy estimates, y Base Case Value Range for Sensitivity Analyses HIV-exposed, uninfected children (from weaning) 50.0 (assumption) 43.0–67.0 [86, 87] HIV-infected children, no ART Intrauterine/intrapartum infection (from birth) 1.1 [83] 1.1–2.0 (assumption) Postpartum infection (from time of infection) 9.4 [83] 5.0–10.0 (assumption) HIV-infected children, on ART Intrauterine/intrapartum infection (from birth) 20.0 (assumption) 10.0–25.0 (assumption) Postpartum infection (from time of infection) 20.0 (assumption) 10.0–25.0 (assumption) Maternal Disease Progression Parameters Value Data Source Impact of antiretroviral therapy Efficacy, % HIV RNA suppression at 24 wk First-line ART, TDF/FTC + (NVP or EFV) Initiated during pregnancy 90% [88] Initiated postpartum, no sdNVP exposure 90% OCTANE trial [89] Difference: [90–92] Initiated postpartum, with sdNVP exposure 85% (difference assumed vs no sdNVP, 5% [88]) Second-line ART (ZDV/3TC/LPV/r) 72% [93] CD4 cell decline over 6 mo following ART interruption 139 cells/µL [36–38] Laboratory and medication costs 2008 US Dollars Data Sources Economic Model Input Parameters

89] Difference: [90–92] Initiated postpartum, with sdNVP exposure 85% (difference assumed vs no sdNVP, 5% [88]) Second-line ART (ZDV/3TC/LPV/r) 72% [93] CD4 cell decline over 6 mo following ART interruption 139 cells/µL [36–38] Laboratory and medication costs 2008 US Dollars Data Sources Economic Model Input Parameters CD4 assay, performed once in ANC for Options A, B, and B+ 9.42 [33] Full blood count, performed once in ANC for Options B and B+ 9.27 [94] Single-dose NVP, 1 maternal and 1 infant dose 0.06 Antenatal ZDV, Option Ac 7.67 per month [27] Antenatal TDF/FTC/NVP, Options B and B+, CD4 count ≤350 cells/µLc 12.12 per month [27] Antenatal TDF/FTC/EFV, Options B and B+, CD4 count >350 cells/µLc 16.50 per month [27] Postnatal maternal ART First-line TDF/FTC/NVP; TDF/FTC/EFV 12.12 per month; 16.50 per month [27] Second line, ZDV/3TC/LPV/r 45.36 per month [27] Pediatric ART, d4T/3TC/NVP 4.54 per month [27] Healthcare Resource Utilization and Costs Antenatal care 2008 US Dollars Data Sources Routine antenatal care, 4 visits 45.77 Average of: [95, 96] Delivery costs, healthcare facility 54.50 [96] Routine and urgent health care costs: Children No. of Inpatient Days per Year No. of Outpatient Visits per Year Total Cost per Monthd Data Sources HIV-infected children, on ART 2.14 6 3.32 [97] Intrauterine/intrapartum infection, no ART 18 6 16.48 [98] Postpartum infection, no ART, aged 0–18 mo 18 6 16.48 [98] Postpartum infection, no ART, aged >18 mo 11 6 10.67 [98] HIV-exposed, uninfected children, aged 0–18 mo 1 3.5 1.73 Assumptione HIV-exposed, uninfected infants aged >18 mo 0 1 0.26 Assumptione Terminal care, last month of life 5 0 49.80 Assumptione Routine and urgent health care costs: Mothers No. of Inpatient Days per Event No. of Outpatient Visits per Event Total Cost per Eventd Data Sources Care for acute opportunistic infections Cape Town AIDS Cohort [99] WHO stage 3–4 HIV disease, range by specific disease 1.3–2.9 2.7–3.4 21.88–39.36 Bacterial infection 2.8 2.4 32.28 Mild fungal infection 1.2 2.3 19.04 Tuberculosis 2.9 2.2 35.66 Terminal care, last month of life 2.39 0.77 26.18 Routine HIV care costs per month 1.22–7.18 (range by CD4) See Supplementary Table 2 for complete list of parameters.

e, range by specific disease 1.3–2.9 2.7–3.4 21.88–39.36 Bacterial infection 2.8 2.4 32.28 Mild fungal infection 1.2 2.3 19.04 Tuberculosis 2.9 2.2 35.66 Terminal care, last month of life 2.39 0.77 26.18 Routine HIV care costs per month 1.22–7.18 (range by CD4) See Supplementary Table 2 for complete list of parameters. Abbreviations: 3TC, lamivudine; ANC, antenatal care; ART, antiretroviral therapy; ARV, antiretroviral medications; d4T, stavudine; EBF, exclusive breastfeeding (in first 6 months of life, followed by MBF); EFV, efavirenz; FTC, emtricabine; HIV, human immunodeficiency virus; LPV/r, lopinavir/ritonavir; MACS, Multicenter AIDS Cohort Study; MBF, mixed breastfeeding; MOHCW, Zimbabwe Ministry of Health and Child Welfare; NA, not applicable; NVP, nevirapine; PTMCT, prevention of mother-to-child HIV transmission; SD, standard deviation; sdNVP, single-dose nevirapine; TDF, tenofovir; WHO, World Health Organization; ZDV, zidovudine. a ART eligibility was defined as CD4 count of ≤350 cells/µL or WHO stage 3–4 disease. b PMTCT uptake was defined as proportion of HIV-infected, pregnant women accessing PMTCT services by the time of delivery. See Supplementary Appendix text and Supplementary Table 2 for details. c Two months of antentatal drug are assumed in all regimens for the base-case analysis, based on median gestational age at booking in Zimbabwe of 30 weeks.

b PMTCT uptake was defined as proportion of HIV-infected, pregnant women accessing PMTCT services by the time of delivery. See Supplementary Appendix text and Supplementary Table 2 for details. c Two months of antentatal drug are assumed in all regimens for the base-case analysis, based on median gestational age at booking in Zimbabwe of 30 weeks. d Total care costs for mothers and infants were calculated by multiplying resource utilization (number of outpatient visits and inpatient days) by an average of WHO-CHOICE estimates of costs for these encounters in 7 sub–Saharan African countries [28]. See Supplementary Appendix for details. e See Supplementary Table 2 for description of assumptions of outpatient healthcare resource utilization. Cost Inputs Monthly medication costs were from the Clinton Healthcare Access Initiative [27]. Costs of clinical care were determined by estimating resource utilization (number of inpatient days and outpatient visits) for specified health conditions, then multiplying by the estimated costs of these healthcare encounters in Zimbabwe (Table 1 and Supplementary Appendix) [28]. For children aged >18 months, monthly utilization estimates (stratified by HIV and ART status) were multiplied by LE to estimate lifetime healthcare costs.

sits) for specified health conditions, then multiplying by the estimated costs of these healthcare encounters in Zimbabwe (Table 1 and Supplementary Appendix) [28]. For children aged >18 months, monthly utilization estimates (stratified by HIV and ART status) were multiplied by LE to estimate lifetime healthcare costs. Model Validation and Sensitivity Analyses Model-derived risks of MTCT, infant mortality, and postpartum maternal OIs were validated against published data, reported previously with extensive sensitivity analyses [10, 14]. For this study, we conducted univariate and multivariate sensitivity analyses on key PMTCT, pediatric, maternal, and cost parameters. Access to Care Parameters We examined the impact of reported rates of PMTCT uptake, defined as the proportion of HIV-infected women receiving PMTCT services and ARVs by delivery (56%, estimated for Zimbabwe in 2009; 80%, the 2009 WHO target goal; 90%, the 2011 WHO target goal; and 95%, reported in neighboring Botswana in 2011) [5, 8, 29]. We varied the availability of CD4 assays from 25% to 100% in Options A, B, and B+; when CD4 count was unavailable in Option A, women were assumed to initiate ART only for WHO stage 3–4 disease. We also examined the impact of reduced pediatric ART availability (36%, estimated for Zimbabwe in 2009) [5] and of reported rates of maternal loss to follow-up (LTFU) from postnatal HIV care (Table 1) [30–32].

when CD4 count was unavailable in Option A, women were assumed to initiate ART only for WHO stage 3–4 disease. We also examined the impact of reduced pediatric ART availability (36%, estimated for Zimbabwe in 2009) [5] and of reported rates of maternal loss to follow-up (LTFU) from postnatal HIV care (Table 1) [30–32]. Clinical Health Parameters We defined a lowest-MTCT risk scenario, using the lowest published risks (best reported effectiveness/efficacy) for each modeled regimen (Table 1); a highest-MTCT risk scenario, combining the highest published risks for each regimen; and a scenario assuming equal MTCT risks with Options A and B. We also used 4 assumptions about LE for HIV-exposed and HIV-infected infants: (1) a high pediatric LE scenario, using the upper bound estimates shown in Table 1, (2) a low pediatric LE scenario, using the lower bound estimates, (3) a largest difference scenario (lowest estimates for HIV-infected children; highest estimates for HIV-uninfected children), and (4) a smallest difference scenario (highest estimates for HIV-infected children; lowest estimates for uninfected children).

(2) a low pediatric LE scenario, using the lower bound estimates, (3) a largest difference scenario (lowest estimates for HIV-infected children; highest estimates for HIV-uninfected children), and (4) a smallest difference scenario (highest estimates for HIV-infected children; lowest estimates for uninfected children). Finally, we investigated potential maternal health impacts of Option B and B+ in 2 ways. First, we varied the efficacy of first-line ART when resumed after ART interruption, reflecting potential interruption-associated drug resistance. Next, we examined the impact of “treatment fatigue” for women who begin ART with CD4 count >350 cells/µL solely for PMTCT, modeled as (1) an increased risk of virologic failure >6 months after ART initiation or (2) a reduction in second-line ART efficacy. Cost Parameters Because estimated costs of healthcare in Zimbabwe are markedly lower than in surrounding countries [28], we repeated the analysis using costs from South Africa (Supplementary Table 2) [33]. In the base case, we conservatively assigned lifelong costs of NVP-based ART to HIV-infected infants; in sensitivity analyses, as an upper bound on pediatric ART costs, we assigned the costs of lifelong lopinavir/ritonavir-based ART to sdNVP-exposed, HIV-infected children. Finally, the nondrug costs of providing 3-drug ARV regimens instead of zidovudine alone (e.g., personnel, laboratory costs) have not been reported; we also examined the impact of such implementation costs in the antenatal period.

e costs of lifelong lopinavir/ritonavir-based ART to sdNVP-exposed, HIV-infected children. Finally, the nondrug costs of providing 3-drug ARV regimens instead of zidovudine alone (e.g., personnel, laboratory costs) have not been reported; we also examined the impact of such implementation costs in the antenatal period. RESULTS Base-Case Results Pediatric HIV Risk and LE Among infants born to HIV-infected women, projected 18-month HIV infection rates were 24.8% (no antenatal ARVs), 14.2% (sdNVP), 7.5% (Option A), and 5.7% (Options B and B+) (Table 2). The resulting projected undiscounted LE (including both HIV-infected and HIV-uninfected infants) ranged from 38.35 years (no antenatal ARVs) to 44.18 years (Options B and B+). Table 2. Base-Case Results: Projected Maternal and Pediatric Outcomes of the Zimbabwe National Prevention of Mother-to-Child HIV Transmission Program Pediatric Life Expectancy, Years From Birth Maternal Life Expectancy, Years From Delivery 18-Month Infant HIV Infection Risk Undiscounted Discounted Undiscounted Discounted Projected Clinical Outcomesa No antenatal ARVsb 24.8% 38.35 21.34 21.25 14.69 sdNVP 14.2% 41.30 22.45 20.94 14.53 Option A 7.5% 43.27 23.19 21.26 14.70 Option B 5.7% 44.18 23.59 21.30 14.74 Option B+ 5.7% 44.18 23.59 22.42 15.45 Antenatal Care Costs, Through Delivery Pediatric Lifetime Healthcare Costs, From Birth Maternal Lifetime HIV-Related Healthcare Costs, From Delivery Undiscounted Discounted Undiscounted Discounted Projected costs, 2008 US Dollarsa

Pediatric Life Expectancy, Years From Birth Maternal Life Expectancy, Years From Delivery 18-Month Infant HIV Infection Risk Undiscounted Discounted Undiscounted Discounted Projected Clinical Outcomesa No antenatal ARVsb 24.8% 38.35 21.34 21.25 14.69 sdNVP 14.2% 41.30 22.45 20.94 14.53 Option A 7.5% 43.27 23.19 21.26 14.70 Option B 5.7% 44.18 23.59 21.30 14.74 Option B+ 5.7% 44.18 23.59 22.42 15.45 Antenatal Care Costs, Through Delivery Pediatric Lifetime Healthcare Costs, From Birth Maternal Lifetime HIV-Related Healthcare Costs, From Delivery Undiscounted Discounted Undiscounted Discounted Projected costs, 2008 US Dollarsa No antenatal ARVsb 85 730 520 8490 5280 sdNVP 92 530 360 8460 5300 Option A 118 490 310 8500 5280 Option B 134 370 240 8450 5260 Option B+ 134 370 240 9820 6240 Abbreviations: ARVs, antiretroviral medications; HIV, human immunodeficiency virus; sdNVP, single-dose nevirapine. a Base-case projections assume 100% uptake of PMTCT services by the time of delivery, 100% linkage to HIV care during breastfeeding, no maternal loss to follow-up after delivery, and 100% availability of pediatric antiretroviral therapy (ART) for HIV-infected infants (see Methods). b No antenatal ARVs refers to receipt of no ARVs or antiretroviral therapy prior to delivery. In all modeled strategies, ART-eligible women who linked to HIV-related healthcare after delivery were assumed to receive ART for their own health in all strategies (Supplementary Table 1).

a Base-case projections assume 100% uptake of PMTCT services by the time of delivery, 100% linkage to HIV care during breastfeeding, no maternal loss to follow-up after delivery, and 100% availability of pediatric antiretroviral therapy (ART) for HIV-infected infants (see Methods). b No antenatal ARVs refers to receipt of no ARVs or antiretroviral therapy prior to delivery. In all modeled strategies, ART-eligible women who linked to HIV-related healthcare after delivery were assumed to receive ART for their own health in all strategies (Supplementary Table 1). Pediatric Costs PMTCT regimens that prevented more infant infections resulted in lower pediatric healthcare costs over time. After the early cost of infant NVP during breastfeeding, the pediatric healthcare costs of Option A became less than those of no antenatal ARVs by 4 years after delivery (Figure 2A). This finding persisted over longer horizons; undiscounted lifetime costs per infant ranged from $730 (no antenatal ARVs) to $370 (Options B and B+) (Table 2). Figure 2. Projected costs (in US dollars [USD]) over the first 5 years after delivery for modeled prevention of mother-to-child human immunodeficiency virus (HIV) transmission (PMTCT) regimens in Zimbabwe. A–D, Undiscounted costs are shown on the vertical axis, and time from delivery is shown on the horizontal access. A, Total healthcare costs for infants (with 100% pediatric antiretroviral therapy [ART] availability). The costs of daily infant nevirapine (NVP) prophylaxis (Option A) are included in pediatric healthcare costs. Because infant NVP is modeled as a pediatric cost, Option A is more expensive than the others during the first 18 months (while breastfeeding continues). PMTCT regimens that are more effective in preventing infant infections result in slower increases in costs (flatter slopes) as time progresses because pediatric HIV care costs are averted, and the pediatric care costs following Option A become less than those following no antenatal antiretroviral medications (ARVs) by 4 years after delivery (arrow). B, HIV-related healthcare costs for women after delivery (with 100% retention in care). The costs of maternal ART and 3-drug ARV prophylaxis (Options B and B+) are included in maternal HIV-related healthcare costs. Postnatal care costs are similar following the no antenatal ARVs, single-dose NVP (sdNVP), and Option A strategies: women enrolled in HIV-related care following all 3 of these strategies are assumed to begin ART when CD4 count falls to ≤350 cells/µL or stage 3–4 disease develops. Small cost differences result from assumptions regarding NNRTI resistance following sdNVP, but the slopes of these 3lines are similar. In Option B+, all women continue their 3-drug regimens.

following all 3 of these strategies are assumed to begin ART when CD4 count falls to ≤350 cells/µL or stage 3–4 disease develops. Small cost differences result from assumptions regarding NNRTI resistance following sdNVP, but the slopes of these 3lines are similar. In Option B+, all women continue their 3-drug regimens. In Option B, women who did not have advanced disease before pregnancy interrupt their ARVs but remain in care and re-initiate ART once CD4 count falls to ≤350 cells/µL or stage 3–4 disease develops. As a result, maternal costs after weaning are greater with Option B+ than with the other regimens, and costs for Option B (due to delayed ART use) are much lower after weaning (becoming less than the costs after Option A by 5 years after delivery) (arrow). Antenatal costs are not included in (A and B). C, HIV-infected women with CD4 count >350 cells/µL, post-delivery. ART costs for women not eligible for ART during pregnancy (CD4 count >350 cells/µL, no stage 3–4 disease), from the Cost-effectiveness of Preventing AIDS Complications (CEPAC) adult model. Three postnatal scenarios are shown: (1) initiate 3-drug ARVs in pregnancy and continue ARVs after weaning (as in Option B+); (2) initiate 3-drug ARVs in pregnancy and interrupt ARVs after weaning (Option B); and (3) do not initiate ARVs in pregnancy but remain in care and initiate ART when needed (CD4 count ≤350 cells/µL or stage 3–4 disease, as in the no antenatal ARVs, sdNVP, and Option A strategies). Interrupting ART at weaning saves money compared with continuing ART; however, this ART interruption may be associated with negative health impacts for HIV-infected mothers if retention in care is less than 100% (Table 2). D, Total cohort costs over the first 5 years after delivery. These include antenatal care costs (through delivery), maternal HIV-related healthcare costs, and pediatric healthcare costs. Option B becomes cost-saving compared with Option A within 4 years after delivery (arrow). Abbreviations: ART, antiretroviral therapy; ARV, antiretroviral medication; sdNVP, single-dose nevirapine.

se include antenatal care costs (through delivery), maternal HIV-related healthcare costs, and pediatric healthcare costs. Option B becomes cost-saving compared with Option A within 4 years after delivery (arrow). Abbreviations: ART, antiretroviral therapy; ARV, antiretroviral medication; sdNVP, single-dose nevirapine. Maternal LE Among HIV-infected women, projected undiscounted maternal LE from delivery was 21.25 years (no antenatal ARVs), 21.26 years (Option A), and 22.42 years (Option B+). Projected maternal LE was lowest in the sdNVP strategy (20.94 years, due to the modeled impact of nonnucleoside reverse transcriptase inhibitor resistance on subsequent first-line ART) and intermediate in the Option B strategy (21.30 years, reflecting benefits from ART during pregnancy and breastfeeding but interruption after weaning). Maternal Costs Although small differences in short-term maternal costs resulted from modeled drug resistance following sdNVP, 5-year costs were similar for no antenatal ARVs, sdNVP, and Option A (Figure 2B). Options B and B+, requiring 3-drug regimens during pregnancy and breastfeeding, conferred the greatest initial maternal healthcare costs. Option B conferred lower maternal costs than Option B+ after weaning because of deferred ART costs when women without advanced disease interrupted ART, and maternal costs with Option B were less than with Option A by 5 years after delivery (Figures 2B and 2C). Undiscounted lifetime maternal HIV-related costs per woman ranged from $8450 (Option B) to $9820 (Option B+ (Table 2).

r weaning because of deferred ART costs when women without advanced disease interrupted ART, and maternal costs with Option B were less than with Option A by 5 years after delivery (Figures 2B and 2C). Undiscounted lifetime maternal HIV-related costs per woman ranged from $8450 (Option B) to $9820 (Option B+ (Table 2). Cost-effectiveness Analysis Option B was projected to result in a discounted combined lifetime cost (ANC + mother + infant) of $5630 per mother–infant pair and a discounted combined LE (mother + infant) of 38.32 years (Table 3). Compared with Option B, the sdNVP, Option A, and no antenatal ARVs strategies all resulted in lower combined LE (36.03–37.89 years) at greater discounted lifetime costs ($5710–5880 per mother–infant pair) and were therefore “dominated.” Replacing Option B with Option B+ would increase costs ($6620 per mother–infant pair) and LE (39.04 years), with an ICER of $1370 per YLS. Considering total combined costs (ANC + mother + infant), Option B became cost saving compared with Option A by 4 years after delivery (Figure 2D; Supplementary Table 7). Table 3. Cost-effectiveness of World Health Organization 2010 Prevention of Mother-to-Child HIV Transmission Guidelines in Zimbabwe

of $1370 per YLS. Considering total combined costs (ANC + mother + infant), Option B became cost saving compared with Option A by 4 years after delivery (Figure 2D; Supplementary Table 7). Table 3. Cost-effectiveness of World Health Organization 2010 Prevention of Mother-to-Child HIV Transmission Guidelines in Zimbabwe Modeled Scenario and PMTCT Regimen Combined Costs per Mother–Infant Pair, Discounted, 2008 US Dollarsa Combined Life Expectancy per Mother–Infant Pair, Discounted, Years From Deliveryb ICER, US Dollars per YLS Base-Case Projectionsc Base-case projections (100% PMTCT uptake, retention in postnatal maternal care, pediatric ART availability) Option B 5630 38.32 Option A 5710 37.89 Dominatedd sdNVP 5760 36.97 Dominated No antenatal ARVs 5880 36.03 Dominated Option B+ 6620 39.04 1370 Sensitivity Analysese Access to care parameters: Reduced PMTCT uptake (56% of HIV-infected women receiving ARVs by delivery; 87% linkage to postnatal care) Option B 4930 35.69 Option A 4980 35.44 Dominated sdNVP 5000 34.92 Dominated No antenatal ARVs 5060 34.39 Dominated Option B+ 5600 36.18 1370 Increased maternal loss to follow-up after delivery (16% in year 1, 6% per year thereafter) Option B 3420 35.23 Option A 3560 34.90 Dominated sdNVP 3620 34.06 Dominated No antenatal ARVs 3730 33.05 Dominated Option B+ 3910 35.81 850 Reduced pediatric ART availability (36% of infected children; 2009 Zimbabwe estimate) Option B 5610 38.00 sdNVP 5670 35.96 Dominated Option A 5670 37.41 Dominated No antenatal ARVs 5690 34.06 Dominated Option B+ 6590 38.71 1370 Current access to care (56% PMTCT uptake, 87% linkage to postnatal maternal care, increased maternal LTFU, 36% pediatric ART  availability) Option B 3010 31.99 sdNVP 3090 30.94 Dominated Option A 3090 31.72 Dominated No antenatal ARVs 3100 29.83 Dominated Option B+ 3340 32.38 850 Clinical health parameters: “Treatment fatigue”: monthly risk of virologic failure after 6 mo on first-line NNRTI-based ART = 2.39% for women starting ART with  CD4 count >350 cells/µL (Options B/B+) (1.5 × base-case risk) Option B 5700 37.82 Option A 5710 37.89 190 sdNVP 5760 36.97 Dominated No antenatal ARVs 5880 36.03 Dominated Option B+ 6700 38.67 1260 Resource utilization parameters: South Africa healthcare costs Option B 14 040 38.33 Option A 14 260 37.89 Dominated sdNVP 14 730 36.97 Dominated Option B+ 15 070 39.05 1410 No antenatal ARVs 15 520 36.04 Dominated Additional $150 antenatal implementation cost for 3-drug regimens compared with ZDV alone Option A 576

60 Resource utilization parameters: South Africa healthcare costs Option B 14 040 38.33 Option A 14 260 37.89 Dominated sdNVP 14 730 36.97 Dominated Option B+ 15 070 39.05 1410 No antenatal ARVs 15 520 36.04 Dominated Additional $150 antenatal implementation cost for 3-drug regimens compared with ZDV alone Option A 576 0 37.89 Option B 5760 38.32 2 sdNVP 5770 36.97 Dominated No ARVs 5880 36.03 Dominated Option B+ 6750 39.04 1370 Abbreviations: ART, antiretroviral therapy; ARV, antiretroviral medications; HIV, human immunodeficiency virus; ICER, incremental cost-effectiveness ratio; LTFU, lost to follow-up; NNRTI, nonnucleoside reverse transcriptase inhibitor; PMTCT, prevention of mother-to-child transmission; sdNVP, single-dose nevirapine; YLS, year of life saved; ZDV, zidovudine. a Combined costs = PMTCT program costs + maternal lifetime HIV-related healthcare costs + infant lifetime healthcare cost (per mother–infant pair). b Combined life expectancy = maternal life expectancy from delivery + infant life expectancy from birth. c Base-case results. Base-case projections assume 100% uptake of PMTCT services by the time of delivery, 100% linkage to HIV care during breastfeeding, no maternal loss to follow-up after delivery, and 100% availability of pediatric ART for HIV-infected infants. d Dominated refers to an intervention that is more expensive and less effective than an alternative intervention.

c Base-case results. Base-case projections assume 100% uptake of PMTCT services by the time of delivery, 100% linkage to HIV care during breastfeeding, no maternal loss to follow-up after delivery, and 100% availability of pediatric ART for HIV-infected infants. d Dominated refers to an intervention that is more expensive and less effective than an alternative intervention. e Sensitivity analyses. Please see Supplementary Table 5 for additional details regarding all sensitivity analyses, including the distribution of costs and life expectancy between mothers and infants. Sensitivity Analyses Access-to-Care Parameters The finding that no antenatal ARVs, sdNVP, and Option A were more costly but less effective than Option B was robust with reduced uptake of PMTCT services or access to CD4 testing, as well as with current availability of pediatric ART, and the ICER of Option B+ compared with Option B in these scenarios remained $1370 per YLS (Table 3, Supplementary Table 5). With reported rates of LTFU from maternal postnatal HIV care, the ICER of Option B+ compared with Option B decreased to $850 per YLS. This ICER remained $850 per YLS when current overall access to care in Zimbabwe was simulated (PMTCT uptake, 56%; pediatric ART availability, 36%; maternal LTFU, 16% in year 1, 6% per year thereafter).

ates of LTFU from maternal postnatal HIV care, the ICER of Option B+ compared with Option B decreased to $850 per YLS. This ICER remained $850 per YLS when current overall access to care in Zimbabwe was simulated (PMTCT uptake, 56%; pediatric ART availability, 36%; maternal LTFU, 16% in year 1, 6% per year thereafter). Clinical Health Parameters Base-case policy conclusions were unchanged in all modeled pediatric LE and MTCT risk scenarios, including when MTCT risks were equal with Options A and B, as well as throughout a variety of “treatment fatigue” scenarios for women initiating 3-drug regimens with CD4 count >350 cells/µL (Supplementary Table 5). Results were sensitive, however, to the risk of virologic failure after 6 months on ART. When this risk was increased 1.5-fold from the base case (to >2.4% per month), Option B no longer dominated Option A; when it was increased 2-fold (to 3.2% per month), Option A dominated Option B (Table 3; Supplementary Table 5).

. Results were sensitive, however, to the risk of virologic failure after 6 months on ART. When this risk was increased 1.5-fold from the base case (to >2.4% per month), Option B no longer dominated Option A; when it was increased 2-fold (to 3.2% per month), Option A dominated Option B (Table 3; Supplementary Table 5). Cost Parameters Policy conclusions were unchanged when lifelong lopinavir/ritonavir costs were assigned to sdNVP-exposed, HIV-infected infants (Supplementary Table 5). In sensitivity analyses using South Africa healthcare costs, the ICER of Option B+ compared with Option B was $1410 per YLS (Table 3). The difference in antenatal implementation costs between 3-drug regimens and zidovudine alone needed to be ≥$150 per person to change the comparison between Options A and B (Table 3); at $150 per person, Option B was no longer cost saving but remained very cost-effective ($2 per YLS), compared with Option A. Even with implementation costs as high as $400 per person, the ICER of Option B compared with Option A remained <$400 per YLS (Supplementary Table 6).

mparison between Options A and B (Table 3); at $150 per person, Option B was no longer cost saving but remained very cost-effective ($2 per YLS), compared with Option A. Even with implementation costs as high as $400 per person, the ICER of Option B compared with Option A remained <$400 per YLS (Supplementary Table 6). DISCUSSION There are 4 key findings from this work. First, a strategy of providing no antenatal ARVs for PMTCT is more expensive and less effective over a lifetime horizon than strategies based on sdNVP, Option A, or Option B. This result, which occurs because the upfront costs of these PMTCT regimens are greatly outweighed by the downstream costs of caring for HIV-infected infants, lends strong economic support to the well-recognized clinical impact of expanding access to PMTCT programs, regardless of the specific drug regimen provided [5]. Second, in settings where 3-drug ARV regimens are not available for PMTCT [5, 34], replacing sdNVP with Option A benefits infants and mothers and saves money over a lifetime horizon.

pport to the well-recognized clinical impact of expanding access to PMTCT programs, regardless of the specific drug regimen provided [5]. Second, in settings where 3-drug ARV regimens are not available for PMTCT [5, 34], replacing sdNVP with Option A benefits infants and mothers and saves money over a lifetime horizon. Third, healthcare programs would decrease costs and improve outcomes further by implementing Option B instead of Option A. Although short-term drug costs are greater with Option B, the incorporation of healthcare costs for both mothers and infants leads Option B to cost less than Option A within 4 years after delivery, primarily because of averted pediatric HIV costs (Figure 2D). Notably, however, if women with high CD4 counts develop poor adherence after Option B (increasing the monthly risk of late virologic failure by ≥25%) (Supplementary Table 5) or if mothers are lost to follow-up after delivery, Option B leads to shorter projected maternal LE than Option A.

ediatric HIV costs (Figure 2D). Notably, however, if women with high CD4 counts develop poor adherence after Option B (increasing the monthly risk of late virologic failure by ≥25%) (Supplementary Table 5) or if mothers are lost to follow-up after delivery, Option B leads to shorter projected maternal LE than Option A. Finally, these results strongly support lifelong ART for all pregnant, HIV-infected women (Option B+) [3, 7]. The interruption of effective ART in Option B may have deleterious effects on maternal health. Randomized trial data comparing maternal health outcomes of Options B and B+ are anticipated soon [35]. In the interim, we assume a rapid rate of CD4 decline after ART interruption based on other trials [36–38], with an associated increased risk of OIs. As a result, Option B+ is projected to increase undiscounted maternal LE by 1.12 years compared with Option B (consistent with modeled impacts of other HIV-related interventions [12, 39]), with an ICER of $1370 per YLS. Although this ICER exceeds the 2008 gross domestic product–based threshold for cost-effectiveness in Zimbabwe ($1200 per YLS) [17, 18], it falls in the lower range of ICERs reported for ART-related interventions in developing countries ($550–$5200 per YLS) [19] and thus represents a return on investment comparable with many current HIV programs in Zimbabwe and other resource-limited settings.

for cost-effectiveness in Zimbabwe ($1200 per YLS) [17, 18], it falls in the lower range of ICERs reported for ART-related interventions in developing countries ($550–$5200 per YLS) [19] and thus represents a return on investment comparable with many current HIV programs in Zimbabwe and other resource-limited settings. Option B+ may represent an even better healthcare investment compared with Option B under specific conditions. First, ART interruption (Option B) may cause greater detriment to maternal health under real-world programmatic conditions than in our guideline-concordant simulations. When women are lost to follow-up after weaning, disease progression is unobserved and cannot lead to prompt ART reinitiation. Such disease progression is more rapid when ART was interrupted months before LTFU (Option B) than at the time of LTFU (Option B+) because of lower CD4 counts at LTFU in Option B. As a result, Option B leads to a projected discounted LE (11.64 years) even lower than no antenatal ARVs (11.71 years) [10], and Option B+ becomes more cost-effective compared with Option B ($850 per YLS). Second, analyses using cost data from South Africa (ICER, $1410 per YLS; 2008 gross domestic product, $5700) [18] suggest that Option B+ may be very cost-effective compared with Option B in higher-income settings where healthcare costs are greater. Third, this analysis excludes several additional benefits of Option B+ that may render it even more effective and cost-effective, including prevention of maternal tuberculosis (also reducing infection risk in infants) [40], HIV transmission to male partners [40], hepatitis B flares due to ARV interruption [7], and MTCT during subsequent pregnancies when women are already on ART at conception [7].

render it even more effective and cost-effective, including prevention of maternal tuberculosis (also reducing infection risk in infants) [40], HIV transmission to male partners [40], hepatitis B flares due to ARV interruption [7], and MTCT during subsequent pregnancies when women are already on ART at conception [7]. There are several limitations to this analysis. First, all models necessarily simplify complex processes; for example, assumptions about infant LE involved uncertainties about healthcare in the distant future. However, LE assumptions, cost assumptions, and projected clinical and economic results were similar to those previously reported [41, 42], and we tested the impact of biologic and operational assumptions in extensive sensitivity analyses [10, 14]. Except where noted, the impact on policy conclusions was minimal, primarily because assumptions were consistent across PMTCT strategies. Second, we excluded the potential impact of drug-related viral resistance in infants who become infected despite exposure to modeled ARV regimens, because of limited data about acquisition of such resistance [43, 44] and its impact on later ART effectiveness. If resistant HIV is a greater concern for infants who become infected while exposed to maternal ARVs through breastmilk than to extended NVP monoprophylaxis, the benefits of Options B and B+ vs Option A will be attenuated. Finally, our analysis assumed a healthcare system perspective. If a societal perspective were assumed, interventions that avert HIV infections in infants and prevent morbidity and mortality in women would be even more cost-effective, avoiding transportation costs and lost wages for medical care and permitting the productivity gains of healthy women and of children who will become healthy adults.

erspective were assumed, interventions that avert HIV infections in infants and prevent morbidity and mortality in women would be even more cost-effective, avoiding transportation costs and lost wages for medical care and permitting the productivity gains of healthy women and of children who will become healthy adults. As in other studies, we find that PMTCT programs based on sdNVP are cost saving, compared with no PMTCT interventions [45]. This is the first analysis to compare sdNVP and Options A, B, and B+ and to consider both short- and long-term maternal and infant outcomes after PMTCT [16, 41, 42, 45]. We find that, with guideline-concordant care, Option A is cost saving compared with sdNVP; Option B becomes more effective and less expensive than Option A within 4 years of delivery; and Option B+ offers additional clinical benefits and economic value comparable with other widely used HIV interventions. We anticipate that the clinical results of these analyses will be generalizable to many African settings where prolonged breastfeeding is the norm and that the base-case economic results may also be applicable in low-income African countries with healthcare costs similar to Zimbabwe. Although specific policies will depend on available resources as well as important considerations of fairness, feasibility, and priority populations [15, 46], PMTCT programs should move rapidly toward these more effective and economically efficient strategies.

frican countries with healthcare costs similar to Zimbabwe. Although specific policies will depend on available resources as well as important considerations of fairness, feasibility, and priority populations [15, 46], PMTCT programs should move rapidly toward these more effective and economically efficient strategies. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. The authors are indebted to Sue J. Goldie, Steven Sweet, and Stephen Resch for derivation of healthcare cost estimates for Zimbabwe appropriate to the recent period of hyperinflation. We also gratefully acknowledge Jennifer Chu, Katie Doherty, and Kathleen Kelly for assistance with modeling analyses and manuscript preparation, and Batsirai Chikwinya, Agnes Mahomva, Stanley Mashumba, Rumbidzai Mugwagwa, and Charity Zvandaziva for critical interpretation of early model results. We also thank the CEPAC-International team and investigators for their contribution. A. L. C. had full access to all of the data and results of this study and has final responsibility for the decision to submit for publication.

mbidzai Mugwagwa, and Charity Zvandaziva for critical interpretation of early model results. We also thank the CEPAC-International team and investigators for their contribution. A. L. C. had full access to all of the data and results of this study and has final responsibility for the decision to submit for publication. Financial support. This work was supported by the Elizabeth Glaser Pediatric AIDS Foundation; the National Institute of Allergy and Infectious Diseases and the National Institute of Child Health and Human Development, National Institutes of Health (K01 AI078754 to A. L. C.; K24 AI062476 to K. A. F.; R01 AI058736 to R. P. W., A. R., J.-E. P., K. A. F; IMPAACT network to R. P. W., J.-E. P., K. A. F.); and Harvard University Center for AIDS Research (to K. A. F., R. P. W.). The funders had no role in study design, interpretation of results, or decision to publish. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed

Paradoxical tuberculosis immune reconstitution inflammatory syndrome (IRIS) occurs in 8%–43% of human immunodeficiency virus (HIV)–infected patients receiving tuberculosis treatment after starting antiretroviral therapy (ART) [1–4]. Tuberculosis IRIS results from rapid restoration of Mycobacterium tuberculosis–specific immune responses, but its pathogenesis remains poorly understood [1, 5, 6]. Neurological tuberculosis IRIS occurs in a substantial proportion (12%) of tuberculosis IRIS cases and is the commonest cause of central nervous system (CNS) deterioration during the first year of ART in settings of high tuberculosis/HIV prevalence [7, 8]. Mortality is high (up to 30%) in those affected [8]. Manifestations of neurological tuberculosis IRIS include meningitis [7–11], intracranial tuberculomata [7, 8, 12–14], brain abscesses [12, 15], radiculomyelitis [7, 8, 11], and spinal epidural abscesses [7]. There are no prospective studies describing tuberculous meningitis (TBM) IRIS; only isolated cases [9–15] and 1 case series of neurological tuberculosis IRIS [8, 16] have been published. Although consensus now exists that ART should be started early (around 2 weeks) in HIV/tuberculosis-coinfected patients with severe immunosuppression, a potential exception is TBM because of the perceived risk of TBM-IRIS [17, 18]. However, the frequency and severity of this complication are not well documented and no means exist to predict the syndrome.

RT should be started early (around 2 weeks) in HIV/tuberculosis-coinfected patients with severe immunosuppression, a potential exception is TBM because of the perceived risk of TBM-IRIS [17, 18]. However, the frequency and severity of this complication are not well documented and no means exist to predict the syndrome. We therefore investigated clinical and laboratory findings in ART-naive HIV-infected patients who presented with TBM. We undertook serial cerebrospinal fluid (CSF) sampling in patients who did and did not develop TBM-IRIS. MATERIALS AND METHODS Setting This prospective, observational study was performed at GF Jooste Hospital, a public sector referral hospital in Cape Town. The hospital serves a low-income, high-density population in which the tuberculosis notification rate exceeds 1.5% per year with 70% of tuberculosis cases coinfected with HIV [19].

METHODS Setting This prospective, observational study was performed at GF Jooste Hospital, a public sector referral hospital in Cape Town. The hospital serves a low-income, high-density population in which the tuberculosis notification rate exceeds 1.5% per year with 70% of tuberculosis cases coinfected with HIV [19]. Participants ART-naive HIV-infected patients aged ≥18 years presenting with meningitis from March 2009 through October 2010 were screened for study inclusion. HIV infection was diagnosed using 2 rapid HIV antibody tests and confirmed by HIV load. Definite TBM was diagnosed when acid-fast bacilli were seen, or when M. tuberculosis was cultured from CSF. Probable TBM was diagnosed when a patient showed clinical, laboratory, and radiological features of TBM in the absence of other infective causes for presentation [20]. Paradoxical TBM-IRIS was diagnosed according to a published definition for tuberculosis IRIS modified for meningitis [7, 8]. The definition had 3 components: (1) TBM diagnosis before starting ART and improvement on tuberculosis treatment prior to ART initiation; (2) onset of TBM-IRIS manifestations (ie, new, recurrent, or worsening clinical features of TBM) within 3 months of ART initiation; and (3) exclusion of alternative causes for clinical deterioration.

d 3 components: (1) TBM diagnosis before starting ART and improvement on tuberculosis treatment prior to ART initiation; (2) onset of TBM-IRIS manifestations (ie, new, recurrent, or worsening clinical features of TBM) within 3 months of ART initiation; and (3) exclusion of alternative causes for clinical deterioration. Patients were ineligible if they had a contraindication to lumbar puncture, including unequal pressures between individual brain compartments on brain imaging, or severe TBM (ie, modified British Medical Research Council [BMRC] grade III disease severity) [21]. The University of Cape Town Research Ethics Committee approved the study and written informed consent was obtained from all patients or their relatives.

pressures between individual brain compartments on brain imaging, or severe TBM (ie, modified British Medical Research Council [BMRC] grade III disease severity) [21]. The University of Cape Town Research Ethics Committee approved the study and written informed consent was obtained from all patients or their relatives. Procedures Demographic data and history of tuberculosis disease, HIV infection, and other systemic illnesses and medications were recorded. Patients underwent general physical and neurological examination. Chest radiography, phlebotomy, and lumbar puncture were performed. In patients with suspected raised intracranial pressure or focal neurological deficits, brain computed tomography scanning was performed prior to lumbar puncture. CSF analysis included biochemistry, cytology, microbiology (including microscopy and culture for fungi and pyogenic bacteria), syphilis serology, HIV load, and Cryptococcus latex agglutination titer. Ziehl-Neelsen staining of sediment and M. tuberculosis culture was performed. If mycobacteria were cultured from CSF, tuberculosis polymerase chain reaction (PCR; Genotype MTBDRplus, Hain Lifesciences) was performed to determine susceptibility to rifampicin and isoniazid. CSF varicella zoster virus PCR was performed if the etiology was suspected. CSF was also stored at −80°C and analyzed for a range of inflammatory markers on the Bio-Plex platform (Bio-Rad Laboratories, Hercules, CA) using customized Milliplex kits (Millipore, St Charles, MO) according to the manufacturer's instructions.

SF varicella zoster virus PCR was performed if the etiology was suspected. CSF was also stored at −80°C and analyzed for a range of inflammatory markers on the Bio-Plex platform (Bio-Rad Laboratories, Hercules, CA) using customized Milliplex kits (Millipore, St Charles, MO) according to the manufacturer's instructions. At TBM diagnosis, patients started tuberculosis treatment according to national guidelines [22] and prednisone (1.5 mg/kg/day). After 2 weeks of treatment and prior to initiation of ART, patients were assessed for improvement on tuberculosis treatment. The initial ART regimen was stavudine, lamivudine, and efavirenz. Later during the study, tenofovir replaced stavudine according to revised national guidelines. CSF investigations performed at TBM diagnosis were repeated at the time of ART initiation, 2 weeks later, and at time of TBM-IRIS presentation and 2 weeks thereafter. Prednisone was reduced to 0.75 mg/kg/day 4 weeks after starting ART and discontinued 2 weeks thereafter, unless the patient developed TBM-IRIS. At TBM-IRIS presentation, investigations were performed to exclude alternative causes of deterioration. Prednisone was either recommenced or the dose increased. Patients were followed for the duration of tuberculosis treatment (9 months); routine visits were at 2 weeks, 4 weeks, 6 weeks, 12 weeks, 6 months, and 9 months after TBM diagnosis. Patients were seen more frequently during deterioration.

causes of deterioration. Prednisone was either recommenced or the dose increased. Patients were followed for the duration of tuberculosis treatment (9 months); routine visits were at 2 weeks, 4 weeks, 6 weeks, 12 weeks, 6 months, and 9 months after TBM diagnosis. Patients were seen more frequently during deterioration. Statistical Analysis Statistical analysis was performed using the GraphPad Prism version 5, R version 2.14.1, and StatXact version 9 software packages. Categorical variables were compared using χ2 or Fisher exact test. Continuous variables were compared between groups and time points within groups, using the Wilcoxon rank sum and Wilcoxon matched pairs tests, respectively. Adjusted relative risks (RRs) were evaluated using Cochran-Mantel-Haenszel tests and tests of homogeneity when considering categorical risk factors. Log-binomial models were fitted to continuous risk factors. Significance testing was done using 2-sided P values with P < .05 taken as significant.

rs tests, respectively. Adjusted relative risks (RRs) were evaluated using Cochran-Mantel-Haenszel tests and tests of homogeneity when considering categorical risk factors. Log-binomial models were fitted to continuous risk factors. Significance testing was done using 2-sided P values with P < .05 taken as significant. The predictive accuracy of CSF neutrophil counts at TBM diagnosis for TBM-IRIS was assessed using nonparametric area under the receiver operating characteristic curve (AUC). Additionally, a model to predict TBM-IRIS risk was developed from 5 prespecified cytokines measured in CSF at time of TBM diagnosis. Interleukin 6 (IL-6), interleukin 10, interleukin 12p40, interferon gamma (IFN-γ), and tumor necrosis factor alpha (TNF-α) were selected as candidate markers of TBM-IRIS based on previous studies [23, 24]. We prespecified the analysis for evaluating the multivariate cytokine model as follows. Significant cytokines comparing TBM-IRIS and non-TBM-IRIS using Wilcoxon rank sum tests were selected for a logistic regression model. Nonsignificant cytokines were dropped, resulting in a final model. The entire model building process was evaluated using leave-one-out cross-validation, the bootstrap method [25], and the permutation test to provide a cross-validated (nonparametric) estimate of the AUC, P values, and 95% confidence intervals (CIs). As a secondary analysis, we examined whether the addition of CSF neutrophils and/or lymphocyte counts would improve the model's predictive ability.

idation, the bootstrap method [25], and the permutation test to provide a cross-validated (nonparametric) estimate of the AUC, P values, and 95% confidence intervals (CIs). As a secondary analysis, we examined whether the addition of CSF neutrophils and/or lymphocyte counts would improve the model's predictive ability. RESULTS TBM Presentation The final diagnoses and reasons for exclusion of patients with meningitis are shown in Figure 1. Thirty-four patients were included in the final analysis; 15 (44%) were female and the median age was 33 years (interquartile range [IQR], 29–44). Sixteen patients (47%) developed TBM-IRIS (TBM-IRIS patients) whereas 18 did not (non-TBM-IRIS patients). Tables 1 and 2 show the demographic and baseline characteristics of these 2 groups. Baseline characteristics were similar between groups, although patients who developed TBM-IRIS had a longer duration of symptoms (median, 19 vs 9 days) and more frequent features of extrameningeal tuberculosis, such as chest symptoms (81% vs 44%) and chest radiographic abnormalities (81% vs 50%). Five TBM-IRIS patients (31%) and 3 non-TBM-IRIS patients (17%) developed features of extrameningeal tuberculosis IRIS. Table 1. Demographic and Baseline Characteristics of Patients Who Developed Tuberculous Meningitis Immune Reconstitution Inflammatory Syndrome and Those Who Did Not

ographic abnormalities (81% vs 50%). Five TBM-IRIS patients (31%) and 3 non-TBM-IRIS patients (17%) developed features of extrameningeal tuberculosis IRIS. Table 1. Demographic and Baseline Characteristics of Patients Who Developed Tuberculous Meningitis Immune Reconstitution Inflammatory Syndrome and Those Who Did Not TBM-IRIS Non-TBM-IRIS No. (%) No. (%) Female 9 (56) 6 (33) Age, y, median (IQR) 33 (30–46) 31 (25–41) Duration between tuberculosis treatment and ART, d, median (IQR)a 15 (14–16) 15 (14–20) Previous tuberculosis 5 (31) 6 (33) Neurological symptoms Duration of neurological symptoms, d, median (IQR) 19 (6–31) 9 (6–21) Nausea/vomiting 13 (81) 8 (44) Headache 16 (100) 14 (78) Visual disturbances 6 (38) 5 (28) Confusion 7 (44) 6 (33) Neck pain/stiffness 13 (81) 14 (78) Systemic symptoms Chest symptoms 13 (81) 8 (44) Night sweats 12 (75) 8 (44) Abdominal symptoms 9 (56) 5 (28) Weight loss 14 (88) 12 (67) Clinical findings Body mass index, median (IQR) 18.4 (17.2–23.3) 20.7 (19.0–23.3) BMRC disease grade 1b 7 (44) 11 (61) Focal neurological signsc 4 (25) 3 (17) Blood investigations Sodium, mmol/L, median (IQR) 123 (121–129) 130 (128–134) Hemoglobin, g/dL, median (IQR) 10.2 (8.5–11.7) 12.4 (9.3–13.4) Other investigations CXR abnormalities 13 (81) 9 (50) Features of extra CNS tuberculosisd 14 (88) 12 (67) Abdominal ultrasound, abnormal/number performed 5/5 (100) 6/8 (75) Brain CT abnormal/number performede 4/5 (80) 5/9 (56) Data are presented as No. (%) unless otherwise specified. Definitions for TBM-IRIS and non-TBM-IRIS are taken from [16] and [18], respectively.

50) Features of extra CNS tuberculosisd 14 (88) 12 (67) Abdominal ultrasound, abnormal/number performed 5/5 (100) 6/8 (75) Brain CT abnormal/number performede 4/5 (80) 5/9 (56) Data are presented as No. (%) unless otherwise specified. Definitions for TBM-IRIS and non-TBM-IRIS are taken from [16] and [18], respectively. Abbreviations: ART, antiretroviral therapy; BMRC, British Medical Research Council; CNS, central nervous system; CT, computed tomography; CXR, chest radiograph; IQR, interquartile range; TBM-IRIS, tuberculous meningitis immune reconstitution inflammatory syndrome. a ART regimens included stavudine (D4T), lamivudine (3TC), and efavirenz (EFV; n = 19); tenofovir, 3TC, and EFV (n = 9); zidovudine, 3TC, and EFV (n = 5); and D4T, 3TC, and lopinavir/ritonavir (n = 1). b BMRC grade I: Glasgow Coma Scale (GCS) score of 15 with no focal neurologic signs; grade II: GCS score of 11–14 or GCS of 15 with focal neurologic signs; grade III: GCS score of ≤10 [21]. c Including cranial nerve palsies (n = 4), hemiparesis (n = 1), cerebellar signs (n = 2). d Including number of patients with 1 or more of the following: chest radiograph or abdominal ultrasound features of tuberculosis and acid-fast bacilli seen or M. tuberculosis cultured from specimen other than cerebrospinal fluid. e Including features compatible with TBM: hydrocephalus, meningeal enhancement, tuberculoma, and infarct. Table 2. Serial Blood and Cerebrospinal Fluid Findings in Patients Who Developed Tuberculous Meningitis Immune Reconstitution Inflammatory Syndrome (TBM-IRIS) and Those Who Did Not (non-TBM-IRIS)

d Including number of patients with 1 or more of the following: chest radiograph or abdominal ultrasound features of tuberculosis and acid-fast bacilli seen or M. tuberculosis cultured from specimen other than cerebrospinal fluid. e Including features compatible with TBM: hydrocephalus, meningeal enhancement, tuberculoma, and infarct. Table 2. Serial Blood and Cerebrospinal Fluid Findings in Patients Who Developed Tuberculous Meningitis Immune Reconstitution Inflammatory Syndrome (TBM-IRIS) and Those Who Did Not (non-TBM-IRIS) TBM-IRIS Non-TBM-IRIS Median (IQR) Median (IQR) P Valuec Blood C-reactive protein, mg/L TBM diagnosis 45 (13–98) 25 (1–71) .22 ART start 6 (4–15) 8 (1–20) .64 2 wk after ART start 56 (20–105) 14 (6–63) .07 CD4 count, cells/μL TBM diagnosis 131 (52–169) 102 (69–278) .79 ART start 93 (65–158) 145 (64–231) .40 2 wk after ART start 158 (139–193) 178 (103–261) .68 HIV load, log10 TBM diagnosis 5.39 (4.75–6.16) 5.60 (4.76–5.72) .83 ART start 5.61 (5.26–6.26) 5.30 (4.90–6.04) .13 2 wk after ART start 3.15 (2.50–3.36) 2.71 (2.43–2.98) .13 Cerebrospinal fluid Protein, g/L TBM diagnosis 2.70 (1.80–5.11) 1.88 (1.28–3.31) .24 ART start 1.63 (1.22–2.67) 0.94 (0.82–1.63) .03 2 wk after ART start/IRIS presentationa 3.11 (1.99–22.83) 0.61 (0.37–1.76) <.001 2 wk after IRIS presentationb 2.62 (2.24–12.74) … … Glucose (CSF blood ratio) TBM diagnosis 0.24 (0.15–0.31) 0.47 (0.28–0.57) .01 ART start 0.51 (0.43–0.55) 0.48 (0.38–0.57) .63 2 wk after ART start/IRIS presentationa 0.39 (0.33–0.43) 0.53 (0.37–0.59) .05 2 wk after IRIS presentationb 0.38 (0.34–0.48) … … Neutrophil counts, ×106/L TBM diagnosis 50 (15–86) 3 (0–44) .02 ART start 1 (0–4) 0 (0–5) .32 2 wk after ART start/IRIS presentationa 42 (17–244) 0 (0–3) <.001 2 wk after IRIS presentationb 3 (0–12) … … Neutrophil proportion, % of cells per sample TBM diagnosis 36 (6–53) 0 (0–12) .009 ART start 1 (1–11) 0 (0–18) .61 2 wk after ART start/IRIS presentationa 9 (4–22) 1 (0–33) .35 Lymphocyte counts, ×106/L TBM diagnosis 218 (63–366) 93 (24–274) .36 ART start 110 (44–225) 41 (18–79) .02 2 wk after ART start/IRIS presentationa 177 (69–363) 25 (6–52) <.001 2 wk after IRIS presentationb 147 (98–405) … … HIV load, log10 TBM diagnosis 6.35 (5.80–6.73) 5.60 (4.81–6.54) .05 ART start 5.56 (5.16–5.95) 5.40 (4.93–5.82) .27 2 wk after ART start 3.00 (2.24–3.67) 2.55 (2.42–2.80) .21 Definitions for TBM-IRIS and non-TBM-IRIS are taken from [16] and [18], respectively.

(6–52) <.001 2 wk after IRIS presentationb 147 (98–405) … … HIV load, log10 TBM diagnosis 6.35 (5.80–6.73) 5.60 (4.81–6.54) .05 ART start 5.56 (5.16–5.95) 5.40 (4.93–5.82) .27 2 wk after ART start 3.00 (2.24–3.67) 2.55 (2.42–2.80) .21 Definitions for TBM-IRIS and non-TBM-IRIS are taken from [16] and [18], respectively. Abbreviations: ART, antiretroviral therapy; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; IQR, interquartile range; IRIS, immune reconstitution inflammatory syndrome; TBM, tuberculous meningitis. a For TBM-IRIS patients, results are reported for lumbar puncture performed at TBM-IRIS presentation, which occurred a median of 14 days after ART initiation. b In 5 TBM-IRIS patients, lumbar puncture was not performed 2 weeks after TBM-IRIS because of death (n = 1), contraindication to lumbar puncture (n = 3), and patient admitted to alternative facility (n = 1). c P < .05 was considered statistically significant.

a For TBM-IRIS patients, results are reported for lumbar puncture performed at TBM-IRIS presentation, which occurred a median of 14 days after ART initiation. b In 5 TBM-IRIS patients, lumbar puncture was not performed 2 weeks after TBM-IRIS because of death (n = 1), contraindication to lumbar puncture (n = 3), and patient admitted to alternative facility (n = 1). c P < .05 was considered statistically significant. Figure 1. Flow diagram of patients with features of meningitis (eg, headache, confusion, vomiting, and/or neck stiffness) screened for study inclusion. aPatients defaulted within 3 months of starting antiretroviral therapy (ART). bTime points of lumbar punctures include tuberculous meningitis (TBM) diagnosis, ART initiation, and 2 weeks after starting ART. cCerebrospinal fluid varicella-zoster virus polymerase chain reaction was performed in 1 patient with who had shingles at time of TBM presentation, which was positive. Abbreviations: ART; antiretroviral therapy, IRIS; immune reconstitution inflammatory syndrome; TB; tuberculosis; TBM, tuberculous meningitis.

rting ART. cCerebrospinal fluid varicella-zoster virus polymerase chain reaction was performed in 1 patient with who had shingles at time of TBM presentation, which was positive. Abbreviations: ART; antiretroviral therapy, IRIS; immune reconstitution inflammatory syndrome; TB; tuberculosis; TBM, tuberculous meningitis. TBM-IRIS Presentation Features of TBM-IRIS developed a median of 14 days (IQR, 4–20) after starting ART. Symptoms and signs included new or worsening headache (n = 12), confusion (n = 6), neck pain/stiffness (n = 11), generalized tonic-clonic seizures (n = 4), vomiting (n = 5), paraparesis (n = 3), myoclonic jerks (n = 1), dysconjugate eye movements (n = 1), and aphasia (n = 1). At time of TBM-IRIS presentation, 15 patients underwent brain imaging, including computed tomography (n = 14) or magnetic resonance imaging (n = 1). Imaging showed features of TBM in 14 of these patients. Magnetic resonance imaging of the spine was performed in 2 patients with paraparesis; both had features of radiculomyelitis (Supplementary Figure 1).

15 patients underwent brain imaging, including computed tomography (n = 14) or magnetic resonance imaging (n = 1). Imaging showed features of TBM in 14 of these patients. Magnetic resonance imaging of the spine was performed in 2 patients with paraparesis; both had features of radiculomyelitis (Supplementary Figure 1). Serial Blood and CSF Findings Baseline blood investigations were similar between TBM-IRIS and non-TBM-IRIS patients with the exception of serum sodium concentrations, which were lower in TBM-IRIS patients (median, 123 vs 130 mmol/L, P = .01, Table 1). Table 2 and Figure 2 show serial blood and CSF findings. There was a significant rise in CD4 counts between starting ART and 2 weeks later in both TBM-IRIS (median, 93–158 cells/μL, P = .009) and non-TBM-IRIS (median, 145–178 cells/μL, P = .04) patients. Between these time points, blood and CSF HIV loads decreased significantly (P < .001) in both groups. Figure 2. Serial cerebrospinal fluid (CSF) findings in patients who developed tuberculous meningitis immune reconstitution inflammatory syndrome (left), and those who did not (right), including protein concentrations (A), CSF to blood glucose ratios (B), neutrophil counts (C), and lymphocyte counts (D). Significant differences (P < .05) between time points within groups are indicated. Abbreviations: ART; antiretroviral therapy, CSF cerebrospinal fluid; IRIS; immune reconstitution inflammatory syndrome; TBM, tuberculous meningitis.

(A), CSF to blood glucose ratios (B), neutrophil counts (C), and lymphocyte counts (D). Significant differences (P < .05) between time points within groups are indicated. Abbreviations: ART; antiretroviral therapy, CSF cerebrospinal fluid; IRIS; immune reconstitution inflammatory syndrome; TBM, tuberculous meningitis. At TBM diagnosis, TBM-IRIS patients had higher CSF cell counts, in particular neutrophils (median, 50 vs 3 cells ×106/L, P = .02, Table 2). Similarly, neutrophil percentages from individual samples were higher in TBM-IRIS patients compared with non-TBM-IRIS patients (median, 36% vs 0%, P = .009). CSF to blood glucose ratios were lower in TBM-IRIS patients (median, 0.24 vs 0.47, P = .005). In both groups, CSF parameters initially improved on tuberculosis treatment (Table 2 and Figure 2). However, at TBM-IRIS presentation, TBM-IRIS patients showed findings of recurrent inflammation. In this group, lymphocyte and neutrophil counts at TBM-IRIS presentation were similar, and protein concentrations higher, compared with the same parameters at TBM diagnosis (median protein, 3.11 g/L at TBM-IRIS vs 2.70 g/L at TBM diagnosis, P = .007).

tation, TBM-IRIS patients showed findings of recurrent inflammation. In this group, lymphocyte and neutrophil counts at TBM-IRIS presentation were similar, and protein concentrations higher, compared with the same parameters at TBM diagnosis (median protein, 3.11 g/L at TBM-IRIS vs 2.70 g/L at TBM diagnosis, P = .007). Mycobacterium tuberculosis was cultured from the CSF of 15 TBM-IRIS patients (94%) and 6 non-TBM-IRIS patients (33%) at TBM diagnosis; the risk of developing TBM-IRIS if CSF M. tuberculosis culture was positive at this time point was 71.4% (15/21), compared with a risk of 7.7% (1/13), corresponding to an RR of 9.3 (95% CI, 1.4–62.2, P = .004). Additional analyses considered the RRs of culture positivity adjusting for the following known or potential risk factors for tuberculosis IRIS: baseline viral load (median, ≥330 000 copies/mL, equivalent to log10 = 5.52, vs <330 000), CD4 count (median, ≤137 cells/μL vs >137), evidence of disseminated disease in the form of an abnormal chest radiograph suggesting miliary disease, and duration of illness (median duration, ≥2 weeks vs <2 weeks). Mycobacterium tuberculosis culture positivity remained a significant risk factor after adjusting for these factors (Supplementary Table 1). There was no evidence of differences in the RRs according to these factors after considering culture status. Some TBM-IRIS patients remained culture positive after 2 weeks (n = 7) and 4 weeks (n = 2) of tuberculosis treatment. No non-TBM-IRIS patients were culture positive after starting tuberculosis treatment. All cultures were fully drug sensitive with the exception of one, which was monoresistant to isoniazid.

ture status. Some TBM-IRIS patients remained culture positive after 2 weeks (n = 7) and 4 weeks (n = 2) of tuberculosis treatment. No non-TBM-IRIS patients were culture positive after starting tuberculosis treatment. All cultures were fully drug sensitive with the exception of one, which was monoresistant to isoniazid. Analysis of Baseline CSF Neutrophil Count and Cytokine Concentrations to Predict TBM-IRIS Concentrations of the prespecified cytokines from which the model to predict TBM-IRIS was developed are shown in Supplementary Table 2 and Supplementary Figure 2. The final multivariate logistic regression model included IFN-γ and TNF-α and produced a cross-validated AUC of 0.91 (95% CI, .53–.99, P = .02), indicating high diagnostic accuracy when jointly considering these 2 cytokines to differentiate TBM-IRIS from non-TBM-IRIS at time of TBM diagnosis. The odds ratio (OR) for TNF-α was 1.85 (per 10 pg/mL, P = .006), indicating an 85% increase in the odds of IRIS for every 10 pg/mL increase in TNF-α (after adjusting for IFN-γ). The OR for IFN-γ was 0.64 (per 100 pg/mL, P = .01), indicating decreased odds of IRIS with increasing IFN-γ (after adjusting for TNF-α). Figure 3 provides a heatmap representation of the predicted probabilities of the resulting model with the observed values overlaid. Neutrophil counts produced an AUC of 0.72 (95% CI, .54–.90, P = .03), indicating modest discriminatory accuracy for TBM-IRIS. However, including CSF neutrophil and lymphocyte counts in the model did not improve its ability to predict TBM-IRIS. Figure 3. Predictive model for tuberculous meningitis immune reconstitution inflammatory syndrome (TBM-IRIS; patients depicted by gray triangles) and non-TBM-IRIS (patients depicted by black circles). Tumor necrosis factor–α and interferon-γ concentrations (pg/mL) are reported. Darker gray indicates greater probability of TBM-IRIS, while lighter gray indicates greater probability of not developing TBM-IRIS. Probabilities associated with shading are indicated by the legend. The middle line indicates 50% chance of TBM-IRIS, while the upper and lower gray lines indicate probabilities of 90% and 10%, respectively. Observe that when using the median line to classify patients as TBM-IRIS or non-TBM-IRIS, all but 2 TBM-IRIS and 1 non-TBM-IRIS patients are correctly classified. Several points in the lower left were moved marginally to the right so that all subjects are clearly identifiable. Abbreviations: IFN, interferon; TNF, tumor necrosis factor.

using the median line to classify patients as TBM-IRIS or non-TBM-IRIS, all but 2 TBM-IRIS and 1 non-TBM-IRIS patients are correctly classified. Several points in the lower left were moved marginally to the right so that all subjects are clearly identifiable. Abbreviations: IFN, interferon; TNF, tumor necrosis factor. Management and Outcome In 13 patients prescribed prednisone (0.75–1.5 mg/kg/day), the dose was increased at TBM-IRIS diagnosis. Prednisone was restarted at a dose of 1.5 mg/kg/day in the other 3 TBM-IRIS patients. The median total duration of corticosteroid treatment in TBM-IRIS patients was 109 days (IQR, 69–141) compared with 35 days (IQR, 20–43) in non-TBM-IRIS patients. ART was interrupted during TBM-IRIS in 1 patient because of brainstem involvement. This patient made a full recovery and had no recurrent symptoms after recommencement of ART under prednisone cover. At 9 months’ follow-up, all non-TBM-IRIS patients were alive (including 1 who had defaulted study follow-up but continued tuberculosis treatment from a primary care tuberculosis clinic), 2 had marked cognitive impairment (international HIV dementia scale <10) [26], and 1 patient had marked cognitive impairment with residual hemiparesis. Twelve IRIS patients (75%) were alive at 9 months’ follow-up, 2 patients showed marked cognitive impairment, 1 patient defaulted study follow-up but was alive, and 1 had marked cognitive impairment, residual hemiparesis, and hearing impairment. Death occurred in 4 (25%) TBM-IRIS patients at 33, 53, 60, and 118 days after TBM diagnosis and was related to TBM-IRIS in 2 patients. The 2 other deaths were due to a road traffic accident (n = 1) and unknown cause (n = 1). Kaplan-Meier survival analysis by TBM-IRIS vs non-IRIS (with a log-rank hypothesis test of the difference in survival between these 2) was nonsignificant.

and 118 days after TBM diagnosis and was related to TBM-IRIS in 2 patients. The 2 other deaths were due to a road traffic accident (n = 1) and unknown cause (n = 1). Kaplan-Meier survival analysis by TBM-IRIS vs non-IRIS (with a log-rank hypothesis test of the difference in survival between these 2) was nonsignificant. DISCUSSION This is the first prospective study of TBM-IRIS. In our cohort, tuberculosis IRIS presenting as TBM-IRIS (47%), as well as tuberculosis IRIS involving any organ system (56%), was more frequent than in previous studies [2–4, 27]. Extrapulmonary tuberculosis is a risk factor for tuberculosis IRIS [14, 28], and our cohort included only patients with TBM. A shorter interval between starting tuberculosis treatment and ART (which was 2 weeks in our study) similarly increases the risk of tuberculosis IRIS [2–4]. The high TBM-IRIS incidence we observed is striking considering that all TBM-IRIS patients were taking prednisone (1.5 mg/kg/day) at time of ART initiation and 13 of these patients (81%) were taking prednisone (0.75–1.5 mg/kg/day) at the time of developing TBM-IRIS. Adjunctive corticosteroids have been shown to reduce mortality in TBM and tuberculosis pericarditis, presumably by reducing pathological host immune responses [21, 29]. In paradoxical tuberculosis IRIS, the symptomatic benefit of corticosteroids was demonstrated in a randomized trial in which prednisone was compared to placebo [30]. For these reasons, we anticipated that prednisone would decrease the risk of tuberculosis IRIS.

y by reducing pathological host immune responses [21, 29]. In paradoxical tuberculosis IRIS, the symptomatic benefit of corticosteroids was demonstrated in a randomized trial in which prednisone was compared to placebo [30]. For these reasons, we anticipated that prednisone would decrease the risk of tuberculosis IRIS. TBM-IRIS was associated with a poor outcome; 2 patients (13%) died as a result of TBM-IRIS, all-cause mortality at 9 months was 25%, and 3 of 11 (27%) survivors examined at 9 months’ follow-up were severely disabled, compared with no deaths and 3 of 17 (18%) patients with severe morbidity in the non-TBM-IRIS group. Our findings are similar to a previous study performed in neurological tuberculosis IRIS [8]. The poor outcome in at least 44% of TBM-IRIS cases emphasizes the need to predict and prevent, and improve the treatment of, TBM-IRIS. Low serum sodium concentration is associated with death in HIV-associated TBM [31]. In our study, serum sodium concentration was lower in TBM-IRIS patients compared with non-TBM-IRIS patients at the time of TBM diagnosis. This may reflect the higher degree of tuberculosis dissemination observed in TBM-IRIS patients, which could have contributed to their risk of developing TBM-IRIS.

ssociated TBM [31]. In our study, serum sodium concentration was lower in TBM-IRIS patients compared with non-TBM-IRIS patients at the time of TBM diagnosis. This may reflect the higher degree of tuberculosis dissemination observed in TBM-IRIS patients, which could have contributed to their risk of developing TBM-IRIS. Our finding of an association between higher CSF neutrophils at TBM presentation and subsequent development of TBM-IRIS provides important and novel insight into the pathogenesis of tuberculosis IRIS. Not only were neutrophil counts higher in TBM-IRIS patients compared with non-TBM-IRIS patients, but neutrophil percentages for individual patients were similarly raised in TBM-IRIS. The neutrophil counts showed dynamic fluctuations over time in TBM-IRIS patients with a marked decrease on tuberculosis treatment, and a striking increase at TBM-IRIS onset. Similar changes in lymphocyte counts were not observed. Studies of tuberculosis IRIS have hitherto focused on the contribution of helper T-cell type 1 lymphocyte responses [32, 33]. However, a role for myeloid cells in tuberculosis IRIS is suggested by a case report of a patient who died from unmasking pulmonary tuberculosis IRIS; postmortem histological examination of diseased lung showed a marked macrophage infiltrate [34]. We have found cytokines of predominantly myeloid origin (IL-6 and TNF-α) to be consistently elevated in patients with tuberculosis IRIS, compared with those who did not develop IRIS [23]. Oliver et al [6] also reported an association between plasma cytokines (interleukin 18) and chemokines (CXCL-10) of the innate immune system and tuberculosis IRIS. In an animal model, immune reconstitution following transfer of mycobacteria-specific CD4 T cells to T-cell–deficient mice infected with Mycobacterium avium was associated with marked increases of both blood and lung CD11b cells (likely representing inflammatory monocytes and neutrophils) [35]. Our results suggest that neutrophils contribute to tuberculosis IRIS pathogenesis. The combination of high CSF TNF-α and low IFN-γ concentrations at the time of TBM diagnosis predicted TBM-IRIS in this cohort. Simmons et al [36] reported a negative correlation between CSF IFN-γ and mortality in HIV-infected patients with TBM. Conversely, a positive correlation was found between CSF IFN-γ and TNF-α concentrations and TBM disease severity by others [37].

ntrations at the time of TBM diagnosis predicted TBM-IRIS in this cohort. Simmons et al [36] reported a negative correlation between CSF IFN-γ and mortality in HIV-infected patients with TBM. Conversely, a positive correlation was found between CSF IFN-γ and TNF-α concentrations and TBM disease severity by others [37]. Several studies have shown an association between disseminated and extrapulmonary tuberculosis and subsequent tuberculosis IRIS [14, 28, 38, 39]. At TBM diagnosis, CSF M. tuberculosis culture positivity, which reflects mycobacterial antigen load, was a major risk factor for developing TBM-IRIS (RR = 9.3). Furthermore, 7 TBM-IRIS patients (44%) were persistently CSF M. tuberculosis culture positive after 2 weeks of tuberculosis treatment and 2 patients (13%) remained culture positive after 4 weeks of tuberculosis treatment. This strongly supports the inference that a high M. tuberculosis bacillary load at time of starting ART is a risk factor for tuberculosis IRIS [40]. The findings suggest it important to optimize tuberculosis treatment prior to starting ART in patients at high risk of developing TBM-IRIS.

eeks of tuberculosis treatment. This strongly supports the inference that a high M. tuberculosis bacillary load at time of starting ART is a risk factor for tuberculosis IRIS [40]. The findings suggest it important to optimize tuberculosis treatment prior to starting ART in patients at high risk of developing TBM-IRIS. We acknowledge several limitations. Because of the relatively small sample size, the study may not have been powered to detect further differences between IRIS and non-TBM-IRIS patients. Only patients with less severe disease (BMRC TBM grade 1 and 2) and those without contraindications to lumbar puncture were enrolled, resulting in the exclusion of a significant proportion of TBM patients presenting in our setting [41]; our results may therefore not be generalizable to ART-naive patients presenting with severe HIV-associated TBM. The model to predict TBM-IRIS needs further validation and exploration with independent data. In conclusion TBM-IRIS complicated the course of treatment of HIV-associated TBM in nearly half our patients, despite the use of adjunctive corticosteroid therapy. The manifestations were severe, fatal in 2 cases. The occurrence of TBM-IRIS associated with CSF M. tuberculosis culture positivity and a high neutrophil count at both baseline and at the time of TBM-IRIS. The baseline relationship between CSF TNF-α and IFN-γ predicted TBM-IRIS. These observations provide novel insight into the pathogenesis of this condition and provide rationale to individualize ART beyond 2 weeks in this devastating, partly iatrogenic, condition.

phil count at both baseline and at the time of TBM-IRIS. The baseline relationship between CSF TNF-α and IFN-γ predicted TBM-IRIS. These observations provide novel insight into the pathogenesis of this condition and provide rationale to individualize ART beyond 2 weeks in this devastating, partly iatrogenic, condition. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://www.oxfordjournals.org/our_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We thank the patients who participated in the study and Monica Magwayi for care provided to patients during the study. We also thank the Radiology Department at GF Jooste Hospital, in particular Dr Ashmitha Rajkumar and Dr Marisa Mezzabotta, who reported the brain imaging. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Supplementary Data Notes Acknowledgments. We thank the patients who participated in the study and Monica Magwayi for care provided to patients during the study. We also thank the Radiology Department at GF Jooste Hospital, in particular Dr Ashmitha Rajkumar and Dr Marisa Mezzabotta, who reported the brain imaging. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Financial support. This work was supported by the Carnegie Corporation Training Award and Discovery Foundation Academic Fellowship Award (S. M.); Perinatal HIV Research Unit, the US Agency for International Development, and the President's Emergency Plan for AIDS Relief (S. M., D. J. P., and C. S.); Wellcome Trust (S. M., R. J. W., and G. M., WT 097254, 081667, 084323, and 088316); Fogarty International Center South Africa TB/AIDS Training Award (G. M., D. J. P., and C. S., NIH/FIC U2R TW007373-01A1 and U2R TW007370-01A1); European Union Grant (R. J. W., SANTE/2005/105-061-102); Medical Research Council Grant (R. J. W., U.1175.02.002.00014.01). Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Eddens and Kolls, on pages 180–1.) Most humans experience their first contact with Pneumocystis (ie, primary infection) shortly after birth [1–4]. This infection is rarely diagnosed because it is asymptomatic or may present as a mild upper respiratory infection [4–6]. Autopsy reports of Pneumocystis in infants have been available for many years [7]. However, characterization of this infection has been hampered by the lack of a microbiological culture method for Pneumocystis, by the low sensitivity of any method used to diagnose Pneumocystis pneumonia in the immunocompromised to detect the smaller quantities of this fungus in immunocompetent individuals, and because Pneumocystis cysts do not stain with the standard hematoxylin-and-eosin stain routinely used in most autopsies. Recent autopsy studies describe the focal (patchy) histological distribution [8–10] and that this infection is more frequent between the ages of 2 and 5 months [4, 5, 9–11]. This age range coincides with the most frequent age for sudden unexpected infant death (SUID) and bronchiolitis [12, 13]. However, the coverage extent of this age overlap and whether it carries any pathogenic significance for Pneumocystis are unknown. Increasing evidence shows that Pneumocystis induces a potent immune response in young immunocompetent rodents [3, 14–17], including a strong gene activation of ClCa3, a member of the calcium-activated chloride channel family of genes expressed in the goblet airway epithelial cell that relates to mucus secretion [14]. Mucus is produced constitutively by goblet cells and binds virtually all particles that land in the airway epithelium as an essential component of the mucociliary clearance system aimed to clean the airways from inhaled particles. This system comprises secretory and ciliated cells, a periciliary liquid (PCL) layer where the cilia move to impulse the mucus, and the propelled overlaying mucus [18]. The heights of the PCL and of the mucus layers need fine tuning to secure airway patency while maintaining clearance efficiency [18–20].

led particles. This system comprises secretory and ciliated cells, a periciliary liquid (PCL) layer where the cilia move to impulse the mucus, and the propelled overlaying mucus [18]. The heights of the PCL and of the mucus layers need fine tuning to secure airway patency while maintaining clearance efficiency [18–20]. Excess PCL will raise the floating mucus layer, making it unreachable to cilia for propulsion, and accumulating mucus could occlude narrow, developing, and distal airways [18–21]. Mucus release is an airway defense reaction stimulated through nonspecific pathways by multiple airway offenders [19–21]. A Pneumocystis-related increase in mucus would then suggest a cofactor role for Pneumocystis in lung disease of the immunocompetent host that is nearly undetectable with current autopsy procedures. Therefore, we undertook this cross-sectional study to describe the prevalence, age distribution, and mucus-associated response to the primary infection by Pneumocystis in autopsied infant lungs. MATERIALS AND METHODS Ethics Review This study was approved by the Ethics Commissions of the North Metropolitan Area of Health, and of the University of Chile School of Medicine in Santiago.

Excess PCL will raise the floating mucus layer, making it unreachable to cilia for propulsion, and accumulating mucus could occlude narrow, developing, and distal airways [18–21]. Mucus release is an airway defense reaction stimulated through nonspecific pathways by multiple airway offenders [19–21]. A Pneumocystis-related increase in mucus would then suggest a cofactor role for Pneumocystis in lung disease of the immunocompetent host that is nearly undetectable with current autopsy procedures. Therefore, we undertook this cross-sectional study to describe the prevalence, age distribution, and mucus-associated response to the primary infection by Pneumocystis in autopsied infant lungs. MATERIALS AND METHODS Ethics Review This study was approved by the Ethics Commissions of the North Metropolitan Area of Health, and of the University of Chile School of Medicine in Santiago. Study Population and Data Collection The Servicio Médico Legal in Santiago is the coroners' office institution for the Metropolitan Area of Chile. A medico-legal autopsy is required for infants who have died in the community in Chile. Infant autopsies performed during calls of a thanatology specialist physician (M.G.) between 1 May 1999 and 6 July 2004 were selected for the study. Inclusion criteria were unexpected death at home, no hospital admission, no immunocompromising conditions, and normal macroscopic examination. The forensic protocol considered clinical history, macroscopic examination and dissection with histological sampling of major organs, plus laboratory tests including toxicology determinations. No bacterial or viral cultures were considered. Medical information including age, date of death, findings including lung histology report, and autopsy diagnoses were collected from the coroner's report prior to Pneumocystis analyses. Autopsy diagnoses were categorized for the purpose of this study as (1) unexplained death (no abnormal findings at autopsy, sudden infant death syndrome); (2) unexplained death with autopsy findings whose contributory role to death was uncertain; and (3) explained death, when a definitive cause of death was established. (Groups 1, 2, and 3 would correspond to SUID or sudden unexpected death in infancy [12, 13]).

bnormal findings at autopsy, sudden infant death syndrome); (2) unexplained death with autopsy findings whose contributory role to death was uncertain; and (3) explained death, when a definitive cause of death was established. (Groups 1, 2, and 3 would correspond to SUID or sudden unexpected death in infancy [12, 13]). Autopsy Samples The complete right lung was carefully removed, placed in a sterile plastic bag, and transported to the investigatoŕs laboratory in an ice-pack container after obtaining legally required samples using sterile equipment. Each lung was processed at arrival, one at a time; lobes were dissected inside a biosafety cabinet using new sterile equipment as described [22]. The pleura was carefully removed to access untouched tissue using separate sterile equipment. Small samples were obtained from deep lung tissue through 2-cm-deep multiple incisions in the decorticated surface of each lobe. Specimens were cut into small pieces and distributed for nested polymerase chain reaction (nPCR) and microscopy. Lobes were processed and analyzed separately.

sing separate sterile equipment. Small samples were obtained from deep lung tissue through 2-cm-deep multiple incisions in the decorticated surface of each lobe. Specimens were cut into small pieces and distributed for nested polymerase chain reaction (nPCR) and microscopy. Lobes were processed and analyzed separately. Samples for Pneumocystis Categorization DNA was extracted and purified from a median of 0.172 g (mean, 0.168 g [range, 0.099–0.226 g]) of pulmonary tissue using the QIAamp DNA Mini Kit (Qiagen, Valencia, California) monitoring for cross-contamination [22]. Pneumocystis jirovecii DNA was identified by nPCR using human β-globin internal controls [22]. Standard cleaning and sterilization procedures using DNA breaking fluids (DNA Away, VWR Scientific Products) were applied to the biosafety cabinet and hood units between each lung. Infants were categorized as Pneumocystis positive when the P. jirovecii DNA–specific 267 bp band was visualized in 1 or more specimens, and as Pneumocystis negative if no P. jirovecii DNA was documented in the 3 lobes. Pneumocystis-negative lobes were analyzed twice, starting from tissue.

Samples for Pneumocystis Categorization DNA was extracted and purified from a median of 0.172 g (mean, 0.168 g [range, 0.099–0.226 g]) of pulmonary tissue using the QIAamp DNA Mini Kit (Qiagen, Valencia, California) monitoring for cross-contamination [22]. Pneumocystis jirovecii DNA was identified by nPCR using human β-globin internal controls [22]. Standard cleaning and sterilization procedures using DNA breaking fluids (DNA Away, VWR Scientific Products) were applied to the biosafety cabinet and hood units between each lung. Infants were categorized as Pneumocystis positive when the P. jirovecii DNA–specific 267 bp band was visualized in 1 or more specimens, and as Pneumocystis negative if no P. jirovecii DNA was documented in the 3 lobes. Pneumocystis-negative lobes were analyzed twice, starting from tissue. Microscopy Analyses A median of 0.396 g (mean, 0.399 g [range, 0.319–0.498 g]) of lung tissue was homogenized by magnetic stirrer agitation in sterile phosphate-buffered saline (PBS) pH 7.2 at 4°C for 30 minutes, sterile gauze filtered, centrifuged at 2900g, 10 minutes at 4°C, and the pellet was reconstituted in 700 µL of sterile PBS pH 7.2. Five-microliter drops were used for microscopy slides. Forms of Pneumocystis were identified using immunofluorescence stain (MeriFluo Kit Biosciences, Cincinnati, Ohio) in the 128 infants. Each sample was analyzed separately and blinded to nPCR results. The 3 lobes per infant were analyzed in duplicate for each lobe.

. Five-microliter drops were used for microscopy slides. Forms of Pneumocystis were identified using immunofluorescence stain (MeriFluo Kit Biosciences, Cincinnati, Ohio) in the 128 infants. Each sample was analyzed separately and blinded to nPCR results. The 3 lobes per infant were analyzed in duplicate for each lobe. Additional Microscopy Methods The first 36 of the 128 infant samples were additionally studied using Gomori-Grocott methenamine silver and Rapid Giemsa (Diff-Quick) staining of lung section imprints. For either microscopy technique, infants were considered “positive” when typical Pneumocystis forms were identified and agreed on by 2 observers (R.B. and C.P. or S.L.V.) in 1 or more lobes and “negative” if the 3 lobes contained no Pneumocystis. Interpretation was performed blinded to the results obtained using other techniques. Microscopy reading took up to 45 minutes per patient. Samples for P. jirovecii and MUC5AC Quantifications Additional lung samples (1 g) were obtained from 59 infants comprising all 20 Pneumocystis-negative infants older than 28 days, and 39 Pneumocystis-positive infants of closest possible age. Samples were flash-frozen, pulverized in liquid nitrogen using a mortar and pestle, homogenized, and frozen until quantitative PCR (qPCR) and Western blot analysis.

obtained from 59 infants comprising all 20 Pneumocystis-negative infants older than 28 days, and 39 Pneumocystis-positive infants of closest possible age. Samples were flash-frozen, pulverized in liquid nitrogen using a mortar and pestle, homogenized, and frozen until quantitative PCR (qPCR) and Western blot analysis. Pneumocystis jirovecii Quantification DNA was extracted from a 0.4-g aliquot. The multicopy msg gene was selected as target using primers PC MSG Forward (5′-CAA AAA TAA CAY TSA CAT CAA CRA GG-3′) and PC MSG Reverse (5′-AAA TCA TGA ACG AAA TAA CCA TTG C-3′) generating a fragment of 156 bp [23] that was cloned in pGEM-T Easy vector (Promega), and used for generating a calibration curve (range of 1 × 101 to 1 × 106 copies/μL). Amplified product was detected using SYBR Green I (Quantace, Bioscan). Quantitative PCR was done in triplicate using the LightCycler 2.0 (Roche) with preincubation period of 10 minutes at 95°C and 46 cycles of 10 seconds at 95°C, 10 seconds at 53°C, and 20 seconds at 72°C each, ending with 7 minutes at 72°C. Each run included negative (ultrapure H2O) and positive (DNA from a patient with Pneumocystis pneumonia) controls and 3 different plasmid standards used in the calibration curve. The specificity of amplified products was verified by melting-curve analysis. Human β-globin gene was used as internal control and for normalization of results as described [5, 24].

O) and positive (DNA from a patient with Pneumocystis pneumonia) controls and 3 different plasmid standards used in the calibration curve. The specificity of amplified products was verified by melting-curve analysis. Human β-globin gene was used as internal control and for normalization of results as described [5, 24]. Mucin Determinations Each aliquot (0.6 g) and a gastric tissue sample (control) were disrupted using a Tissue Tearor (Biospec) in chilled RIPA-modified lysis buffer. Total protein was quantified in supernatant by Bradford (Bio-Rad). Thirty-microgram aliquots were subject to sodium dodecyl sulfate polyacrylamide gel electrophoresis (4% stacking and 8% resolving Tris-Glycine gels). Proteins were transferred to polyvinylidene difluoride membranes and blocked. Mouse anti-MUC5AC immunoglobulin G (IgG) antibody (1:500, 45M1, SCBT) and goat antimouse IgG horseradish peroxidase (HRP)–conjugated antibody (1:2000, SCBT) were used for MUC5AC detection. Membranes were stripped, blocked, and reprobed using standard antiactin antibodies (goat antiactin IgG, 1:2000, SCBT and donkey antigoat IgG HRP, 1:3000, SCBT). Enhanced chemiluminescence reagent was used for membrane development (Pierce ECL WB Substrate, Thermo Scientific). Films were analyzed with Image J software (National Institutes of Health).

ed, and reprobed using standard antiactin antibodies (goat antiactin IgG, 1:2000, SCBT and donkey antigoat IgG HRP, 1:3000, SCBT). Enhanced chemiluminescence reagent was used for membrane development (Pierce ECL WB Substrate, Thermo Scientific). Films were analyzed with Image J software (National Institutes of Health). Statistical Analysis GraphPad Prism 5 software (San Diego, California) was used to compare prevalence of Pneumocystis in explained vs unexplained deaths using χ2 with Yates’ correction, Pneumocystis (MSG copies) at sequential age intervals using analysis of variance, MUC5AC expression according to Pneumocystis presence using unpaired t test with Welch's correction, and to analyze the correlation between expression of MUC5AC and Pneumocystis MSG copies using Pearson test. A P value of <.05 was considered significant.

SG copies) at sequential age intervals using analysis of variance, MUC5AC expression according to Pneumocystis presence using unpaired t test with Welch's correction, and to analyze the correlation between expression of MUC5AC and Pneumocystis MSG copies using Pearson test. A P value of <.05 was considered significant. RESULTS Infants and Lung Sample Characteristics A total of 669 infants (aged 3 days to 12 months) underwent a legally required autopsy at Servicio Médico Legal during the enrollment period. M.G. conducted 134 infant autopsies that fulfilled entry criteria and in which the right lung was submitted for analysis. Six newborn infants (mean age, 14.8 days; median, 17 days; range, 2–22 days) were excluded because of recent hospitalization, and 128 infants with a median age of 2 months 29 days (mean, 3 months 11 days [range, 7 days to 11 months 27 days]), 70 (54.7%) male, were considered for this study. Infants were assigned to specific diagnostic categories after autopsy completion (Table 1). Complete right lungs were obtained in 111 infants, 2 lobes in 3 and, 1 lobe in 14, respectively. Table 1. Detection of Pneumocystis by Nested Polymerase Chain Reaction in Homogenized Lung-Tissue Autopsy Specimens of Different Pulmonary Lobes from 128 Infants Dying Suddenly and Unexpectedly in the Community

n (Table 1). Complete right lungs were obtained in 111 infants, 2 lobes in 3 and, 1 lobe in 14, respectively. Table 1. Detection of Pneumocystis by Nested Polymerase Chain Reaction in Homogenized Lung-Tissue Autopsy Specimens of Different Pulmonary Lobes from 128 Infants Dying Suddenly and Unexpectedly in the Community Pneumocystis DNA Contribution to Diagnosis—Any Lobec Autopsy Result No.a RUL RML or RLL Total Unexplained death 85 61 10 71 (83.5%) Unexplained death with mild autopsy findings 28 18 6 24 (85.7%) Nonspecific lung inflammation 15 Congenital malformation (compatible with life) 4 Metabolic defect (hypoglycemia) 1 Signs of infection (mild and outside the lung) 8 Explained death 15 6 4 10 (66.7%) Bronchopneumonia 4 Congenital malformation (cardiac or brain) 2 Traumatic death 2 Asphyxia (immersion or food) 2 Systemic signs of infection (DIVC, meningitis, other) 5 Total 128 85 (80.9%)b 20 (19.1%)b 105 (82.0%) Abbreviations: DIVC, disseminated intravascular coagulopathy; RLL, right lower lobe; RML, right middle lobe; RUL, right upper lobe. a Age: mean, 3 mo 11 d; median, 2 mo 29 d; range, 7 d to 11 mo 27 d. b Percentage relative to the 105 Pneumocystis DNA–positive infants to indicate that 80.9% of positives was detected by analyzing the RUL and 19.1% additional positives by analyzing the RML or RLL specimens. For the purpose of this study, infants were considered to be negative for Pneumocystis DNA after analysis of 2 samples in each lobe. c Prevalence of Pneumocystis DNA among unexplained vs explained deaths, P = .28.

b Percentage relative to the 105 Pneumocystis DNA–positive infants to indicate that 80.9% of positives was detected by analyzing the RUL and 19.1% additional positives by analyzing the RML or RLL specimens. For the purpose of this study, infants were considered to be negative for Pneumocystis DNA after analysis of 2 samples in each lobe. c Prevalence of Pneumocystis DNA among unexplained vs explained deaths, P = .28. Sensitivity of Diagnostic Techniques Pneumocystis jirovecii DNA was detected by nPCR in the first 36 infants studied; 34 (94.4%) of them tested positive by immunofluorescence microscopy, and 2 (5.6%) by single PCR in the same homogenized tissue aliquot. Diff-Quick and Gomori-Grocott methenamine silver stains detected Pneumocystis trophic forms in 18 (50.0%) and cyst forms in 11 (30.6%), of lung tissue imprints (Figures 1 and 2). Figure 1. Diagnosis of Pneumocystis in infant biopsy specimens requires sensitive techniques applied to homogenized tissue: Percentage of Pneumocystis detection as relative to nested polymerase chain reaction (n-PCR), of immunofluorescence microscopy (IF), and single-round PCR in homogenized lung tissue specimens of 36 infants. Results of microscopy readings using rapid Giemsa (Diff-Quick) and Gomori-Grocott methenamine silver (GMS) stains in imprints of cruent-cut-surface lung tissue adjacent to the sections analyzed by n-PCR and IF are also presented. Abbreviations: IF, immunofluorescence microscopy; GMS, Gomori-Grocott methenamine silver; n-PCR, nested polymerase chain reaction; PCR, polymerase chain reaction.

i-Grocott methenamine silver (GMS) stains in imprints of cruent-cut-surface lung tissue adjacent to the sections analyzed by n-PCR and IF are also presented. Abbreviations: IF, immunofluorescence microscopy; GMS, Gomori-Grocott methenamine silver; n-PCR, nested polymerase chain reaction; PCR, polymerase chain reaction. Figure 2. Detection of this highly focal Pneumocystis infection by microscopy examination in homogenized preparations or imprints from lung tissue specimens. Pneumocystis forms as visualized by microscopy using immunofluorescence stain in aliquots of homogenized lung biopsy specimens (F = ×400; C and I = ×1000), or by rapid Giemsa stain (Diff-Quick) in imprints from fresh lung infant autopsy specimen sections (A, D, and G = ×400; B, F, and H = ×1000). Arrows on each ×400 picture point to their ×1000 magnifications. Bar = 10μ.

immunofluorescence stain in aliquots of homogenized lung biopsy specimens (F = ×400; C and I = ×1000), or by rapid Giemsa stain (Diff-Quick) in imprints from fresh lung infant autopsy specimen sections (A, D, and G = ×400; B, F, and H = ×1000). Arrows on each ×400 picture point to their ×1000 magnifications. Bar = 10μ. DNA Amplification Nested PCR detected P. jirovecii DNA in 105 (82.0%) of the 128 infants: 60 (85.7%) of 70 male and 45 (77.6%) of 58 female infants. Pneumocystic jirovecii DNA was detected in 88 (79.3%) of 111 infants having their 3 lobes analyzed; of them, 35 (39.8%), 21 (23.9%), and 32 (36.3%) had detectable P. jirovecii DNA in 3, 2, or 1 lobes, respectively. The first analysis detected 80 (94%) of the 85 infants whose right upper lobe (RUL) was P. jirovecii DNA positive (Table 1). Pneumocystis jirovecii DNA was detected in 4 of 7 infants < 1 month of age (Figure 3). All amplification reactions of controls for contamination of DNA extraction and purification were negative. Figure 3. Pneumocystis jirovecii infection in autopsied infant lungs peaks at 3–5 months. Lung autopsy specimens from 128 infants dying in the community were analyzed for P. jirovecii using nested polymerase chain reaction (nPCR) and immunofluorescence microscopy (IF). P. jirovecii DNA was detected in 105 (82.0%), and Pneumocystis forms were confirmed by IF in 99 (94.2%) of those found positive for P. jirovecii DNA by nPCR and in 0 of 23 infants who tested negative. Each bar represents a minimum of 5 infants. Pneumocystis was additionally detected in 4 of 4, 2 of 2, 2 of 3, 2 of 3, 1 of 1, and 0 of 2 infants dying at 6, 7, 8, 9, 10, and 11 months of age, respectively.

IF in 99 (94.2%) of those found positive for P. jirovecii DNA by nPCR and in 0 of 23 infants who tested negative. Each bar represents a minimum of 5 infants. Pneumocystis was additionally detected in 4 of 4, 2 of 2, 2 of 3, 2 of 3, 1 of 1, and 0 of 2 infants dying at 6, 7, 8, 9, 10, and 11 months of age, respectively. Microscopy Analyses Lung homogenate specimens from the 128 infants were analyzed by immunofluorescence microscopy in addition to nPCR, and cystic plus smaller trophic Pneumocystis forms were detected in 99 (94.3%) of 105 infants testing positive by nPCR. Immunofluorescence was negative in all 23 infants who were Pneumocystis DNA negative by nPCR (Table 1; Figure 2). Pneumocystis Quantification Pneumocystis normalized counts (MSG copies per nanogram of human DNA) were higher between 2 and 5 months and declined thereafter (P = .7630) (Figure 4). Figure 4. Pneumocystis organisms burden increases up to 3–5 months of infant age and declines thereafter. Age progression of Pneumocystis organisms load in autopsy lung samples from 39 infants dying suddenly in the community is shown. Pneumocystis MSG quantitative polymerase chain reaction results were normalized to nanograms of human β-globin DNA for comparisons and expressed as the normalized mean of Pneumocystis MSG copies ± SD.

ession of Pneumocystis organisms load in autopsy lung samples from 39 infants dying suddenly in the community is shown. Pneumocystis MSG quantitative polymerase chain reaction results were normalized to nanograms of human β-globin DNA for comparisons and expressed as the normalized mean of Pneumocystis MSG copies ± SD. MUC5AC Determinations Normalized levels of MUC5AC were significantly increased (P = .0134) in association with the presence of Pneumocystis (Figure 5). This increase was consistent at all age intervals (data not shown), and independent of Pneumocystis burden (Pearson r = 0.0908; P = .5822). MUC5AC determination values were normalized by human actin protein expression, and Pneumocystis MSG determinations by human β-globin levels (mean ± SD). Figure 5. Mucus (MUC5AC) expression is increased by Pneumocystis presence and not influenced by organism load. Top: MUC5AC protein expression according to Pneumocystis status in lung tissue specimens from 39 P. jirovecii–positive and 20 P. jirovecii–negative infants (mean ± SD). Bottom: Correlation between normalized MUC5AC protein expression and normalized quantification values of P. jirovecii MSG in the same lung sample specimen for each infant (Pearson r = 0.0908, P = .5822). MUC5AC level values were normalized by human actin protein expression, and Pneumocystis MSG determinations by human β-globin levels (mean ± SD). Abbreviation: MUC5AC, mucus.

ression and normalized quantification values of P. jirovecii MSG in the same lung sample specimen for each infant (Pearson r = 0.0908, P = .5822). MUC5AC level values were normalized by human actin protein expression, and Pneumocystis MSG determinations by human β-globin levels (mean ± SD). Abbreviation: MUC5AC, mucus. DISCUSSION This study confirms Pneumocystis as the most prevalent microorganism in autopsied infant lungs identified to date, and that Pneumocystis presence associates to increased mucus (MUC5AC) expression, suggesting that it increases the mucociliary clearance workload and upregulates innate immune responses in the airway epithelium [19–21]. Pneumocystis cells and P. jirovecii–specific DNA were identified in the lungs of nearly all infants in this study using immunofluorescence microscopy and nPCR, respectively. This high prevalence is consistent with previous evidence that Pneumocystis is common in infant lungs [9, 10] including a study documenting Pneumocystis DNA by nPCR in all of 58 infants of undisclosed age [25]. Furthermore, the structural forms of the fungus were all recognized using Giemsa and GMS stains, suggesting active replication [26].

consistent with previous evidence that Pneumocystis is common in infant lungs [9, 10] including a study documenting Pneumocystis DNA by nPCR in all of 58 infants of undisclosed age [25]. Furthermore, the structural forms of the fungus were all recognized using Giemsa and GMS stains, suggesting active replication [26]. The comprehensive diagnostic approach utilized in this study, including examination of up to 6 fresh homogenized tissue samples per infant, increased the sensitivity of detection and underlines the focal distribution of Pneumocystis in the nonimmunocompromised host [8, 9]. This approach detects smaller burdens of Pneumocystis than present in immunocompromised patients with Pneumocystis pneumonia, where the fungus is readily diagnosable by microscopy or single-round PCR. Pneumocystis burden in these infant lungs, although mild, was greater than in immunocompetent adults where diagnosis additionally requires of tissue-concentration techniques [22]. In addition, results show that the age peak with approximately 90% of infants having detectable Pneumocystis, and the higher normalized burden of organisms, coincide at 2–5 months. This age predominance was suggested in previous studies [9–11] and matches the age of onset of severe Pneumocystis pneumonia in immunosuppressed or debilitated infants prior to anti-Pneumocystis prophylaxis [27, 28]. Importantly, young age is by itself a risk factor for Pneumocystis severity exemplified by the worse prognosis of HIV-related Pneumocystis pneumonia in infants whose mortality is 60% vs 10% in adults [27, 28].

mocystis pneumonia in immunosuppressed or debilitated infants prior to anti-Pneumocystis prophylaxis [27, 28]. Importantly, young age is by itself a risk factor for Pneumocystis severity exemplified by the worse prognosis of HIV-related Pneumocystis pneumonia in infants whose mortality is 60% vs 10% in adults [27, 28]. This study also documents that Pneumocystis is associated with increased mucus production. Mucus is a gel composed by water (97%) and solids including mucins (3%) [19, 20]. MUC5AC, the gel-forming mucin used as a marker of mucus in this study, is the predominant solid component of mucus in infant airways [29]. Increased normalized levels of MUC5AC have been similarly documented in association with many other well-recognized, less prevalent airway offenders like respiratory viruses, bacteria, acetyl choline, cytokines, prostaglandins, lipopolysaccharides, nitric oxide, and other potential activators of nonspecific airway signaling pathways as the ErbB receptor epidermal growth factor receptor (EGFR) [20, 21]. Additional airway offenders were not studied. MU5AC was consistently increased in Pneumocystis-positive infants at all age intervals, suggesting that Pneumocystis predisposes the host to augmented mucus responses during this age period [19–21], and was unaffected by Pneumocystis burden in agreement with the concept that pathogenesis for Pneumocystis is mostly host dependent [6, 27, 30].

increased in Pneumocystis-positive infants at all age intervals, suggesting that Pneumocystis predisposes the host to augmented mucus responses during this age period [19–21], and was unaffected by Pneumocystis burden in agreement with the concept that pathogenesis for Pneumocystis is mostly host dependent [6, 27, 30]. Pathogenically, mucins are heavily glycosylated proteins stored in packaging intracellular granules [31]. Their release in response to airway insults is followed by immediate mucin hydration leading to several hundred-fold intraluminal volume increase in milliseconds [19–21, 31]. This mechanism could represent a risk for narrow, developing infant airways because minor height volume changes in the airway surface liquid can lead to small airway closure in times as short as a breathing cycle [32]. The clinical outcome of increased mucus depends on several factors affecting clearance including airway surface tension, geometry, size, and effective cough [32, 33]. Infants have airways of small diameter, with greater elasticity and compliance, fewer collateral airway channels, and a reduced functional residual capacity, compared with older children or adults [34]. In addition, mucins in infants are more acidic that may reflect greater viscosity [29, 34]. The presence of Pneumocystis could therefore favor airway collapse suggested as a mechanism in current hypotheses for SUID [35, 36]. This may occur with few clinical manifestations until most of the peripheral airways are occluded [21]. Airway collapse would be challenging to diagnose at autopsy as it may immediately resolve with postmortem airway relaxation. In addition, gravitational orientation of the lungs and the release of transpulmonary pressure upon opening the thorax may mobilize airway secretions and further decrease autopsy evidence.

[21]. Airway collapse would be challenging to diagnose at autopsy as it may immediately resolve with postmortem airway relaxation. In addition, gravitational orientation of the lungs and the release of transpulmonary pressure upon opening the thorax may mobilize airway secretions and further decrease autopsy evidence. Pneumocystis is common in the general population at any age. Therefore, Pneumocystis-associated mucus increase may also be relevant for chronic respiratory diseases such as chronic obstructive pulmonary disease and cystic fibrosis in which the coexistence of mucus excess and Pneumocystis is described [19, 37, 38]. Other pathways increase mucin in addition to the EGFR in the ErbB family of receptors, and include tumor necrosis factor α, STAT6, interleukin 1β, interleukin 13, and NF-κB and may be activated by Pneumocystis [16, 17, 30, 39]. In addition, Pneumocystis may induce collateral sensitization to a nonspecific antigen in immunocompetent mice, increasing the number of CD45+CD11c+ antigen-presenting cells that explain an hyper-reactive response upon a later challenge [16]. An airway hyperreactive response can explain airway collapse as documented in sensitized mice [40]. This type of response may be relevant to SUID and infant bronchiolitis whose peak incidences coincide with the age peak of Pneumocystis [12, 13, 35, 41].

at explain an hyper-reactive response upon a later challenge [16]. An airway hyperreactive response can explain airway collapse as documented in sensitized mice [40]. This type of response may be relevant to SUID and infant bronchiolitis whose peak incidences coincide with the age peak of Pneumocystis [12, 13, 35, 41]. This autopsy study was conducted in sudden unexpected infant deaths. This is the most frequent form of death in apparently healthy, nonimmunocompromised infants [12]. Pneumocystis prevalence was not different in infants with unexplained vs explained deaths in this study, in agreement with a previous study documenting a similar incidence of Pneumocystis in infants with unexplained deaths vs in those of similar age dying of accidental causes, confirming that Pneumocystis is not sufficient to cause SUID [11]. The high prevalence of Pneumocystis in SUID, however, raises the possibility that Pneumocystis may be a “necessary but not sufficient” cause of SUID as coadjuvant to diverse nonspecific triggers acting on top of Pneumocystis.

dying of accidental causes, confirming that Pneumocystis is not sufficient to cause SUID [11]. The high prevalence of Pneumocystis in SUID, however, raises the possibility that Pneumocystis may be a “necessary but not sufficient” cause of SUID as coadjuvant to diverse nonspecific triggers acting on top of Pneumocystis. Pathology reports in this study showed that inflammation was absent or too mild to explain infant deaths through inflammatory mechanisms, as in previous autopsy series [9]. Autopsy signs of a mild respiratory infection that per se does not explain death are present in approximately half of SUID cases [12]. The lack of evident inflammation in these infants can be explained by death occurring before inflammation develops, or by other reasons including focality of the infection [14]. Animal models demonstrate that the sequence of events leading to lymphocytic response is well demarcated [14, 42], and delayed during low-burden infections such as this one, until Pneumocystis multiplies and is able to induce the transient inflammation that eliminates the pathogen in the immunocompetent host [30]. Airway collapse may be favored by increased mucus and could explain death in a proportion of these infants [35, 36], suggesting that prevention of Pneumocystis-associated mucus increase until the airway is more developed could reduce vulnerability to SUID and, eventually, to bronchiolitis.

Pathology reports in this study showed that inflammation was absent or too mild to explain infant deaths through inflammatory mechanisms, as in previous autopsy series [9]. Autopsy signs of a mild respiratory infection that per se does not explain death are present in approximately half of SUID cases [12]. The lack of evident inflammation in these infants can be explained by death occurring before inflammation develops, or by other reasons including focality of the infection [14]. Animal models demonstrate that the sequence of events leading to lymphocytic response is well demarcated [14, 42], and delayed during low-burden infections such as this one, until Pneumocystis multiplies and is able to induce the transient inflammation that eliminates the pathogen in the immunocompetent host [30]. Airway collapse may be favored by increased mucus and could explain death in a proportion of these infants [35, 36], suggesting that prevention of Pneumocystis-associated mucus increase until the airway is more developed could reduce vulnerability to SUID and, eventually, to bronchiolitis. Pneumocystis is the most prevalent microorganism in the lungs of small infants. Pneumocystis-associated mucus increase may also be relevant to older children or adults with respiratory conditions associated with Pneumocystis and increased mucus.

Airway collapse may be favored by increased mucus and could explain death in a proportion of these infants [35, 36], suggesting that prevention of Pneumocystis-associated mucus increase until the airway is more developed could reduce vulnerability to SUID and, eventually, to bronchiolitis. Pneumocystis is the most prevalent microorganism in the lungs of small infants. Pneumocystis-associated mucus increase may also be relevant to older children or adults with respiratory conditions associated with Pneumocystis and increased mucus. Notes Acknowledgments. We recognize the dedicated work of Isabella Eyzaguirre Valderas and Maria Antonieta Perez Concha as part of their DVM graduate thesis, and thank Drs Jaime Fergie and Mauricio Henríquez for critical review of the manuscript. We are especially grateful to Dr Walter T. Hughes for his constant advice and for his critical review of the manuscript.

ed work of Isabella Eyzaguirre Valderas and Maria Antonieta Perez Concha as part of their DVM graduate thesis, and thank Drs Jaime Fergie and Mauricio Henríquez for critical review of the manuscript. We are especially grateful to Dr Walter T. Hughes for his constant advice and for his critical review of the manuscript. Author contributions. S. L. V. was responsible for the hypothesis, literature search, and writing the manuscript; S. L. V., C. A. P., F. P., J.-F. A., R. B., M. C., I. D.-J., P. I., E.-M. A., and E. D.-C. designed the categorization of lung specimens part of the study; S. L. V., C. A. P., F. P., J.-F. A., and R. B. designed the mucus part of the study; M. G. performed the autopsies; C. A. P., M. G., F. P., J.-F. A., R. B., and S. L. V. performed the determinations; S. L. V., C. A. P., F. P., R. F. M., and P. I. analyzed and interpreted the data; R. F. M., C. A. P., F. P., J.-F. A., R. B., M. C., M. G., I. D.-J., P. I., E.-M. A., and E. D. C. critically revised the paper. S. L. V. is the guarantor of the study. Financial support. This work was supported by Fondo Nacional de Desarrollo Científico y Tecnológico (FONDECYT-Chile) (grant numbers 1060750 and 1100225 to S. L. V.). Collaboration between French and Chilean groups was supported by the French cooperation Ministry and by ECOS-CONICYT (travel grant C05S02 to E. D.-C. and S. L. V.). Collaboration with R. F. M. was supported by Visiting Professorship funds included in FONDECYT-Chile (grant number 1100225 to S. L. V.) and by travel funds from University College London (to R. F. M.).

s was supported by the French cooperation Ministry and by ECOS-CONICYT (travel grant C05S02 to E. D.-C. and S. L. V.). Collaboration with R. F. M. was supported by Visiting Professorship funds included in FONDECYT-Chile (grant number 1100225 to S. L. V.) and by travel funds from University College London (to R. F. M.). Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Japanese encephalitis (JE), a mosquito-borne flaviviral disease, is the leading cause of epidemic encephalitis worldwide, accounting for approximately 70 000 annual cases of clinical disease [1]. Because of the severity of this disease and lack of antivirals, vaccinations are recommended, not only for inhabitants of endemic areas but also for travelers at risk [2]. However, it may not be widely recognized that there are several genotypes of Japanese encephalitis virus (JEV) circulating in endemic areas, and their epidemiology is evolving. Japanese encephalitis viruses are divided into 5 genotypes (GI–GV). GI and GIII have mainly been isolated in temperate and epidemic areas, whereas GII and GIV have mostly been found in tropical, endemic regions [3]; GV has only been isolated 3 times [4]. There have been several reports on GI replacing GIII as the dominant genotype in numerous regions [3, 5, 6]. All JE vaccines currently available are based on viral strains belonging to a single genotype, GIII, even if this no longer constitutes the dominant JEV type in many areas. At present, there are hardly any human data on the efficacy of the current inactivated travelers’ vaccines against the circulating JEV genotypes other than GIII. The need to assess the cross-reactive potential of the GIII-derived vaccines has become widely recognized [5–7]. We looked into the vaccine-induced cross-protection against JEV genotypes I–IV after immunization with the 2 inactivated JE vaccines (Ixiaro or Japanese Encephalitis Vaccine GCC) currently given to travelers.

All JE vaccines currently available are based on viral strains belonging to a single genotype, GIII, even if this no longer constitutes the dominant JEV type in many areas. At present, there are hardly any human data on the efficacy of the current inactivated travelers’ vaccines against the circulating JEV genotypes other than GIII. The need to assess the cross-reactive potential of the GIII-derived vaccines has become widely recognized [5–7]. We looked into the vaccine-induced cross-protection against JEV genotypes I–IV after immunization with the 2 inactivated JE vaccines (Ixiaro or Japanese Encephalitis Vaccine GCC) currently given to travelers. METHODS Study Population The study population, previously JEV-naive adult volunteers, received a primary series with either a SA14-14-2–based (Ixiaro, Intercell AG, Vienna, Austria; n = 29) or Nakayama-based (Japanese Encephalitis Vaccine GCC, Green Cross Corp, South Korea; n = 12) inactivated GIII JE vaccine before traveling to a JE-endemic area in Asia. The same volunteers had been evaluated for their immune response against the 2 GIII vaccine strains in our previous study exploring the ability of a heterologous vaccine to boost JE immunity [8]. These vaccinees were now evaluated for the presence of neutralizing antibodies against 5 JEV strains representing genotypes I–IV. The research protocol was approved by the ethics committees supervising the investigational sites. The study (EudraCT: 2010-023300-27; ClinicalTrials.gov: NCT01386827) was registered in the databases required and performed in accordance with the principles outlined in the Declaration of Helsinki. All volunteers provided informed consent.

ol was approved by the ethics committees supervising the investigational sites. The study (EudraCT: 2010-023300-27; ClinicalTrials.gov: NCT01386827) was registered in the databases required and performed in accordance with the principles outlined in the Declaration of Helsinki. All volunteers provided informed consent. Eligibility criteria for the study population have been described in detail previously [8]. Moreover, to avoid cross-reactions due to preexisting antibodies, participants found to be seropositive for 1 or more of the JEV test strains before vaccination (n = 5) were excluded from the final analyses. Determination of the Neutralizing Antibody Responses Serum samples were collected before vaccination (day 0) and 4–8 weeks after the last vaccine dose. The serological analyses were carried out in a blinded manner. All serum samples were tested by the plaque-reduction neutralization test (PRNT) as previously described [8] using 5 JEV target strains representing genotypes I–IV: SM-1 (GI; isolated in Thailand in 2002), B 1034/8 (GII; isolated in Thailand in 1983), Nakayama (GIII; strain in Japanese Encephalitis Vaccine GCC), SA14-14-2 (GIII; strain in Ixiaro), and 9092 (GIV; isolated in Indonesia in 1981). A PRNT50 titer (the reciprocal of the serum dilution that reduced the virus plaque count by 50% as compared with the virus-only controls) of ≥10 was considered protective [9].

Nakayama (GIII; strain in Japanese Encephalitis Vaccine GCC), SA14-14-2 (GIII; strain in Ixiaro), and 9092 (GIV; isolated in Indonesia in 1981). A PRNT50 titer (the reciprocal of the serum dilution that reduced the virus plaque count by 50% as compared with the virus-only controls) of ≥10 was considered protective [9]. Statistical Analysis Statistical analysis was performed with the R 2.13.0 software (R Development Core Team 2011). The statistical significance of differences in seroconversion rates (SCRs) was assessed by 2-sided χ2 tests, and in levels of neutralizing antibodies by 2-sided Wilcoxon exact tests. P < .05 was considered significant. RESULTS Study Group Characteristics The background characteristics of the subjects have been described in detail previously [8]. The final study population comprised 22 female and 19 male travelers between the ages of 18 and 61 years (median age, 26.0 years). Most subjects were healthy and of Finnish or Swedish origin; 1 volunteer had asthma.

acteristics The background characteristics of the subjects have been described in detail previously [8]. The final study population comprised 22 female and 19 male travelers between the ages of 18 and 61 years (median age, 26.0 years). Most subjects were healthy and of Finnish or Swedish origin; 1 volunteer had asthma. Serological Analyses Figure 1 shows the individual PRNT50 titers of neutralizing antibodies against all target strains before and after a primary series with either of the 2 JE vaccines, and the SCRs and geometric mean titers (GMTs) attained for both vaccine groups against all 5 test strains. Figure 1. Japanese encephalitis (JE) vaccine–induced immune response in previously Japanese encephalitis virus (JEV)–naive adult travelers: PRNT50 titers against viral strains of different JEV genotypes are shown before and 4–8 weeks after a vaccination series with SA14-14-2–derived (Ixiaro; n = 29) or Nakayama-derived (Japanese Encephalitis Vaccine GCC; n = 12) vaccine. The gray lines indicate PRNT50 titer = 10. PRNT50 titers of ≥10 were considered protective. (PRNT50 titer is the reciprocal of the serum dilution that reduced the virus plaque count by 50% as compared with the virus-only controls). The seroconversion rates (SCRs) and geometric mean titers (GMTs) are given in each panel.

he gray lines indicate PRNT50 titer = 10. PRNT50 titers of ≥10 were considered protective. (PRNT50 titer is the reciprocal of the serum dilution that reduced the virus plaque count by 50% as compared with the virus-only controls). The seroconversion rates (SCRs) and geometric mean titers (GMTs) are given in each panel. Of the 29 travelers immunized with the SA14-14-2–based JE vaccine (Ixiaro), 93%–97% attained protective levels of neutralizing antibodies against the 5 JEV test strains representing genotypes I–IV. One of the subjects did not reach protective PRNT50 titers against any of the strains tested, and another had neutralizing antibodies to GI, GII, and homologous GIII strains (SA14-14-2) but not to heterologous GIII (Nakayama) or the GIV test strain. Among the 12 travelers who were immunized with the Nakayama-based vaccine (JE Vaccine GCC), the SCRs varied between 83% and 100%, depending on the test strain used. All subjects exhibited a response to the GIII strain homologous to the vaccine strain (Nakayama); 1 subject failed to respond to all other strains, and 1 showed a response to all test strains except the GIII (SA14-14-2). No significant differences were found in the SCRs between the 2 vaccine groups. In both groups, the highest PRNT50 titers were presented against the GII target strain. These proved significantly higher than those against the GI, the heterologous GIII, and the GIV strains. The second highest titers were found against the test strain homologous to each vaccine strain; no difference was seen between the homologous strains and GII.

T50 titers were presented against the GII target strain. These proved significantly higher than those against the GI, the heterologous GIII, and the GIV strains. The second highest titers were found against the test strain homologous to each vaccine strain; no difference was seen between the homologous strains and GII. DISCUSSION All JE vaccines currently available are based on JEV strains isolated >50 years ago, representing only a single genotype (GIII), which no longer constitutes the dominant JEV type in many areas [3, 5, 6]. The notable changes occurring in the dynamics of the genotype distribution call for studies on the cross-reactive capacity of the current vaccines against the various genotypes.

isolated >50 years ago, representing only a single genotype (GIII), which no longer constitutes the dominant JEV type in many areas [3, 5, 6]. The notable changes occurring in the dynamics of the genotype distribution call for studies on the cross-reactive capacity of the current vaccines against the various genotypes. Evaluation of neutralizing antibodies with PRNT assay is generally accepted as a surrogate measure for efficacious JEV immunity [9]. Despite the widespread use of JE vaccines and the circulation of heterologous genotypes, vaccine-induced neutralization capacity has mostly been assessed solely against the strain homologous to the vaccine strain [10–12]. It is well known, however, that the various strains and genotypes exhibit antigenic differences [13]. Cross-protection elicited by JE vaccines has mainly been addressed by JEV challenge studies in animals; these suggest variable levels of protection against challenge with heterologous JEV [14–16]. In humans, differences have been found in the vaccine-induced immune response to heterologous virus strains even within the same genotype (GIII) after primary immunization with the live-attenuated [17], and the inactivated mouse brain [8, 13, 17] or Vero cell–derived vaccines [8].

allenge with heterologous JEV [14–16]. In humans, differences have been found in the vaccine-induced immune response to heterologous virus strains even within the same genotype (GIII) after primary immunization with the live-attenuated [17], and the inactivated mouse brain [8, 13, 17] or Vero cell–derived vaccines [8]. Human data on immune response against strains of heterologous genotypes are scarce [18, 19]. To our knowledge, this is the first study to explore the cross-reactive potential of the inactivated JE vaccines against nonvaccine genotypes in travelers, and the first human study to address the cross-protection elicited by the new inactivated SA14-14-2–based vaccine, Ixiaro. Our study shows protective levels of cross-reactive neutralizing antibodies to genotypes I–IV in European travelers after a primary series with inactivated SA14-14-2–based and Nakayama-based vaccines, implying good cross-protective capacity for both of these preparations against all major genotypes currently circulating. The GV strain was not available for testing, yet, as long as GV remains such a rare cause of encephalitis, this genotype appears to be of minor clinical significance.

ed and Nakayama-based vaccines, implying good cross-protective capacity for both of these preparations against all major genotypes currently circulating. The GV strain was not available for testing, yet, as long as GV remains such a rare cause of encephalitis, this genotype appears to be of minor clinical significance. Our study also showed differences in the levels of neutralizing antibodies against various genotypes: interestingly, the most pronounced immune response was observed against a strain representing GII, which is heterologous to both vaccine strains. The second-highest titers were seen against the strains homologous to those in the vaccines, whereas responses to GI and GIV remained somewhat lower. Notably, the low cross-protection found against GI, the genotype emerging as the most prevalent genotype in many areas of Asia [3, 5, 6], calls for special attention in the future. Importantly, despite the differences between the genotypes, responses to all genotypes reached protective levels. The majority of all JE cases are encountered in areas where JE vaccines have been implemented in the national vaccination program [1]. A marked decrease has been seen in the JE incidence after the introduction of childhood vaccinations [1], which supports our findings on the cross-protective capacity of GIII-derived vaccines.

ority of all JE cases are encountered in areas where JE vaccines have been implemented in the national vaccination program [1]. A marked decrease has been seen in the JE incidence after the introduction of childhood vaccinations [1], which supports our findings on the cross-protective capacity of GIII-derived vaccines. The present study shows that the inactivated JE vaccines currently given to travelers provide significant protection against the most important JEV genotypes circulating in endemic countries at the moment; however, no data are available on the duration of this cross-protection. This should be addressed in future studies in nonendemic populations where natural boosters can be excluded. Special attention should be paid to the longevity of the cross-protective response to GI. The present study is the first human study to explore the immunogenicity of the new JE vaccine Ixiaro against heterologous genotypes, and the first to explore the cross-genotype immunogenicity of the inactivated JE vaccines in travelers. Our data show a significant cross-protective capacity against heterologous strains representing genotypes I–IV. This implies that, at present, both the inactivated JE vaccines given to travelers can be expected to confer protection against all major genotypes found in endemic areas. Notes Acknowledgments. The authors thank the personnel of the Travel Clinic, Aava Medical Centre, Postitalo, Helsinki, Finland, and Cityakuten/Wasavaccination, Sweden, for help in collecting blood samples and recruiting patients.

The present study is the first human study to explore the immunogenicity of the new JE vaccine Ixiaro against heterologous genotypes, and the first to explore the cross-genotype immunogenicity of the inactivated JE vaccines in travelers. Our data show a significant cross-protective capacity against heterologous strains representing genotypes I–IV. This implies that, at present, both the inactivated JE vaccines given to travelers can be expected to confer protection against all major genotypes found in endemic areas. Notes Acknowledgments. The authors thank the personnel of the Travel Clinic, Aava Medical Centre, Postitalo, Helsinki, Finland, and Cityakuten/Wasavaccination, Sweden, for help in collecting blood samples and recruiting patients. Financial support. This work was supported by Oskar Öflunds stiftelse, the Finnish governmental subsidy for health science research, and the Centre for Clinical Research, Sörmland County Council, Sweden (all to E. O. E.). Potential conflicts of interest. A. K. and L. R. have participated as members in an advisory board for and received honoraria from Novartis; L. L. and L. R. have received honoraria from Baxter; A. K. has acted as a consultant on vaccination immunology and received research funds from Crucell; A. K., L. L., J. R., and L. R. have received honoraria for lectures from Crucell, GlaxoSmithKline, Baxter, and Pfizer. All other authors report no potential conflicts of interest.

R. have received honoraria from Baxter; A. K. has acted as a consultant on vaccination immunology and received research funds from Crucell; A. K., L. L., J. R., and L. R. have received honoraria for lectures from Crucell, GlaxoSmithKline, Baxter, and Pfizer. All other authors report no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Tuberculosis treatment with short-course chemotherapy has 3 aims: rapid bactericidal activity, which is measured by sputum conversion; sterilizing activity, which is measured by relapse; and suppression of acquired drug resistance (ADR). The World Health Organization's (WHO) DOTS (directly observed therapy, short-course) program was developed to ensure success of this chemotherapy. DOTS has 5 components: political commitment by governments, improved laboratory services, a continuous supply of good-quality drugs, documentation of individual patients’ success and program progress toward set targets, and direct observation by a healthcare worker of each patient swallowing pills (ie, directly observed therapy [DOT]). Historical trends of the decline of multidrug-resistant tuberculosis rates with implementation of the program, especially the dramatic reports from New York City and other large cities, provided powerful examples of the success of the program [1–4]. DOT, the namesake and heart of the program, is the most expensive [5–7]. However, DOT is considered by the World Bank to be one of the “most cost-effective of all health interventions, and indispensable to preventing ADR and relapse” [8, 9]. The several studies that were pivotal to the adoption of the DOTS program were retrospective, or employed quasi-experimental designs, and often emphasized the benefit of program-defined treatment outcomes [1–4, 8–12]. They did not tease out the effect of DOT from other program components. In contrast, one meta-analysis of prospective studies found no major benefit of DOT compared to self-administered therapy (SAT) for program-defined outcomes such as “cure” and “completion of treatment” in both active and latent tuberculosis [13]. In another systematic review, there was also no significant benefit for the outcome of recurrence [14]. However, in some high-burden countries such as in South Africa, up to 77% of recurrence is due to new infection and not relapse [15, 16].

re” and “completion of treatment” in both active and latent tuberculosis [13]. In another systematic review, there was also no significant benefit for the outcome of recurrence [14]. However, in some high-burden countries such as in South Africa, up to 77% of recurrence is due to new infection and not relapse [15, 16]. Because DOT is now the accepted standard of care everywhere, performance of randomized controlled trials in which some patients are randomized to SAT or DOT or placebo pills to see if ADR emerges more easily would be unethical [17]. To address this limitation, we recently performed hollow-fiber studies in which various degrees of nonadherence were examined during both bactericidal and sterilizing effect [18]. Surprisingly, microbiologic failure occurred only when >60% of doses were missed, but no ADR was encountered. Thus, we hypothesize that DOT has no impact on rates of sputum conversion, ADR, or relapse in tuberculosis patients. To test that hypothesis, we performed a meta-analysis of prospective clinical studies that compared DOT to SAT and reported microbiologic outcomes. We were particularly interested in microbiologic outcomes as primary outcomes, as it is a standard tenant of infectious diseases therapeutics that the best evidence for eradication of pathogens or ADR, or relapse, is microbiologic demonstration [19], and not program factors such as “completion of therapy.”

ic outcomes. We were particularly interested in microbiologic outcomes as primary outcomes, as it is a standard tenant of infectious diseases therapeutics that the best evidence for eradication of pathogens or ADR, or relapse, is microbiologic demonstration [19], and not program factors such as “completion of therapy.” METHODS Definitions We used WHO definitions [20]. DOT refers to the practice of supervising tuberculosis patients swallowing all their pills over the entire course of treatment by trained health personnel who are accountable to tuberculosis control staff. SAT refers to unsupervised administration of prescribed antituberculosis drugs by patients. We defined partial DOT as the practice in which patients are on DOT for only portions of the therapy duration. Defaulting refers to missing a cumulative ≥2 months of doses after initially taking at least 1 month's worth of medication. Patients reported as lost to follow-up by randomized clinical trials were included in the defaulting category. Microbiologic failure refers to positive smear microscopy or culture at the fifth month or later on therapy. Patients who had their treatment changed for persistent bacteriologic positivity or because of radiologic and/or clinical deterioration, including those with “doubtful responses,” were classified as having failed treatment. ADR was defined as new or additional resistance to 1 or more of the first-line antituberculosis drugs among failures or relapses. Relapse was when a patient was declared cured but subsequently developed microbiologically proven disease [20]. Molecular genotyping of repeat isolates was not performed.

g failed treatment. ADR was defined as new or additional resistance to 1 or more of the first-line antituberculosis drugs among failures or relapses. Relapse was when a patient was declared cured but subsequently developed microbiologically proven disease [20]. Molecular genotyping of repeat isolates was not performed. Search Strategy We searched PubMed, Embase, ISI Web of Science, and the Cochrane Library for studies published between 1 January 1965 and 31 December 2012. There was no exclusion of articles by language. Bibliographies of original articles, key reviews, and consensus statements were also searched for additional relevant studies [8, 10, 13, 14]. The following Medical Subject Heading terms and strategy was used: directly observed therapy OR supervised therapy OR directly observed treatment strategy OR DOT OR DOTS AND self-administered therapy OR self-supervised therapy OR unsupervised therapy AND tuberculosis. In addition, we also searched for articles in the gray literature at Inside Conferences, clinicaltrials.gov, and Open Grey (System for Information on Grey Literature in Europe; http://www.opengrey.eu).

tegy OR DOT OR DOTS AND self-administered therapy OR self-supervised therapy OR unsupervised therapy AND tuberculosis. In addition, we also searched for articles in the gray literature at Inside Conferences, clinicaltrials.gov, and Open Grey (System for Information on Grey Literature in Europe; http://www.opengrey.eu). Study Selection Criteria Inclusion criteria were prospective studies in which patients were diagnosed by microscopic examination of sputum smear or culture and were separately assigned to either DOT or SAT, treatment using a short-course chemotherapy regimen that includes isoniazid, rifampin, and pyrazinamide and evidence of evaluation for microbiologic failure. Studies were limited to prospective data from observational studies or controlled trials with concurrent controls. We excluded retrospective studies to avoid selection and information biases, studies carried out in children, studies that used retreatment regimens, and treatment in patients with a prior history of tuberculosis. Data Extraction and Quality Assessment of Included Studies Study selection was done independently by the 2 investigators. Reviewer agreements were measured using the κ statistic. The quality of each trial was graded by use of validated scores [21]. Disagreements were resolved by consensus. Outcomes The primary outcome was microbiologic failure. The secondary outcomes were ADR, relapse, and default. Standards We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [22].

Data Extraction and Quality Assessment of Included Studies Study selection was done independently by the 2 investigators. Reviewer agreements were measured using the κ statistic. The quality of each trial was graded by use of validated scores [21]. Disagreements were resolved by consensus. Outcomes The primary outcome was microbiologic failure. The secondary outcomes were ADR, relapse, and default. Standards We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [22]. Data Analysis We quantified heterogeneity of effect using the I2 statistic [23, 24]. We calculated the incidence rate (IR) and 95% confidence intervals (CIs) for DOT or SAT, for each study for each of the outcomes based on the number of events reported in each original study. We also computed for a second effect size measure, which is the risk difference (RD). This was used because several cells had zero outcomes events, which makes it difficult to calculate relative risk (RR) without imputation of data or excluding studies. However, all 3 effect sizes were reported, with IR and RD considered the primary. To permit unbiased comparison of outcome, we employed an “intention to treat” strategy (ie, by original assigned treatment groups, irrespective of whether treatment was subsequently changed), except when not stated by the primary study, when we analyzed outcomes as all patients randomized [24]. We decided a priori to use the DerSimonian and Laird random methods to pool effect size across studies, as these methods would provide more conservative CIs [23, 25, 26]. Fixed-effects models were used to pool effect size if there was no significant heterogeneity (ie, I2 ≤ 50%); otherwise, mixed-effects models were used for I2 > 50%. We employed mixed-effects models, in which random-effects analyses were used to combine IR of groups within each study, using Comprehensive Meta-Analysis software (Biostat Inc, Englewood, New Jersey). Study-to-study variance (T2) was not pooled across studies; however, it was computed within groups and was not assumed to be the same for all groups. Publication bias and small study effects were systematically evaluated by visual inspection for funnel plot asymmetry and by use of the Egger test [23, 26].

New Jersey). Study-to-study variance (T2) was not pooled across studies; however, it was computed within groups and was not assumed to be the same for all groups. Publication bias and small study effects were systematically evaluated by visual inspection for funnel plot asymmetry and by use of the Egger test [23, 26]. Subgroup and Sensitivity Analysis First, we examined the effect of removing one study at a time on effect size for microbiologic failure, ADR, and relapse. Second, we examined the effect of study design (randomized controlled trials vs observational studies) on effect size. Third, we examined whether combining all patients classified as partial DOT with either DOT or SAT led to significant changes in effect size. Fourth, we examined the role of study locale (rural patients vs urban patients) on effect size. Fifth, we examined the effect of study quality score on effect size. Meta-Regression Analysis To further explore potential source of heterogeneity, we performed meta-regression analyses in which study design and study locale were simultaneously examined as covariates. Random-effects meta-regression was utilized; we expected some unexplained or “residual” heterogeneity. The weight for each trial was equal to the inverse of the sum of the within-trial variance and the residual between-trial variance, in order to correspond to a random-effects analysis. An iterative method providing restricted maximum likelihood estimates of regression parameters, their asymptotic variance, and the residual heterogeneity variance was performed in Stata version 12.

of the within-trial variance and the residual between-trial variance, in order to correspond to a random-effects analysis. An iterative method providing restricted maximum likelihood estimates of regression parameters, their asymptotic variance, and the residual heterogeneity variance was performed in Stata version 12. RESULTS Study Selection and Characteristics of Included Studies Ten of 129 initially identified studies (8%) met selection criteria [5, 27–35], as shown in Figure 1. The κ value was 0.92 for the inclusion of studies and 0.90 for the rating of trials on considered methodologic aspects. There were 5 randomized controlled trials and 5 observational studies. The characteristics of included studies are shown in Table 1, as is the quality score for each study, which demonstrates that all 10 were good-quality studies. The combined number of participants enrolled in the selected studies was 13 752. From these, 13 112 (95%) participants were assigned or randomized to intervention: 8774 (67%) to DOT, 630 (5%) to partial DOT, and 3708 (28%) to SAT. Thus, the proportion of patients who received partial DOT was small, and this group was excluded from further computation of effect size. Table 1. Characteristics of 10 Studies Selected for the Meta-Analysis

were assigned or randomized to intervention: 8774 (67%) to DOT, 630 (5%) to partial DOT, and 3708 (28%) to SAT. Thus, the proportion of patients who received partial DOT was small, and this group was excluded from further computation of effect size. Table 1. Characteristics of 10 Studies Selected for the Meta-Analysis Study Reference Place (Study Period) Type of Location Regimens Examineda Patients Assigned to Interventions Study Quality Patients Selected Intervention Procedures DOT SAT Randomized trials [5] Pakistan (1996–1998) Rural/urban 2HRZE7/6HE7 497 4 New, sputum positive, >15 y HCW at facility monitored 6×/wk; trained CM and FM monitored monthly during collection of antituberculosis drugs Twice-monthly review and to collect antituberculosis drugs [32] Cape Town, South Africa (1994–1995) Urban 2HRZ7/4HR7; 3HRZE7/6HRE7 216 4 New and retreatment, drug susceptible, >15 y HCW monitored DOT at clinic during working hours, 5×/wk for IP, then thrice weekly for CP for new patients Patient self-supervised, nurse reviewed adherence card weekly during clinic visit to obtain antituberculosis drugs [33] Cape Town, South Africa (1994–1995) Urban 2EHRZ7/6EH7; 2EHRZ2/4EHR2; 2HRZ2/4HR2 156 4 New and retreatment, drug susceptible, >15 y HCW at clinic and trained LHW. Patients on LHW supervision took meds several times/wk at LHW home Patient self-supervised, nurse reviewed adherence card weekly during clinic visit to obtain antituberculosis meds [34] Madras and Chennai, India Urban 2EHRZ7/6EH7; 2EHRZ2/4EHR; 2HRZ2/4HR2 1203 3 Sputum smear positive, >15 y. HCW at clinic at least once/wk Completely unsupervised, weekly drug collection during IP and twice monthly during CP [35] Thailand (1996–1997) Rural/urban 2HRZE7/4HR7 837 4 New, sputum positive, >15 y CM, FM, both trained and monitored twice/mo during IP and once/mo in CP; compliance monitored by use of treatment cards, pill counts and urine test for rifampin. HCW; monitored daily One-mo supply of drugs after diagnosis and after follow-up visits. No supervision Observational studies [27] Blackburn, UK (1988–2000) Urban 2HRZ3/4HR3; 3HRZE3/6HRE3 205 3 Sputum smear positive HCW, at clinic thrice weekly. DOT mandatory for noncompliant or at-risk patients Monthly review, random urine testing and pill counts: all received FDC [28] Southern Thailand (1999) Rural/urban 2HRZE7/4HR7; 2HRZE3/4HR3 411 4 New, smear positive DOT supervisors not stated; various levels of DOT examined.

smear positive HCW, at clinic thrice weekly. DOT mandatory for noncompliant or at-risk patients Monthly review, random urine testing and pill counts: all received FDC [28] Southern Thailand (1999) Rural/urban 2HRZE7/4HR7; 2HRZE3/4HR3 411 4 New, smear positive DOT supervisors not stated; various levels of DOT examined. Strict DOT referred to observers actually watching patients swallow all the drugs during the first 2 mo Not strict DOT, referred to as SAT [29] San Francisco, USA (1998–2000) Urban 2HRZE3/4HR3 372 3 Culture positive HCW at clinic, home, or workplace; enablers given: DOT mandatory for at-risk patients and noncompliance Monthly review [30] Bangkok, Thailand (2004–2005) Urban 2HRZE7/4HR7 1256 4 New, smear positive Center-based (HCW), family-based (FM), or hybrids of center/SAT based; or center + family DOT Patients who could not attend to be accommodated in center- or family based DOT, were assigned SAT [31] Thailand (2004–2006) Rural/urban 2HRZE7/4HR7; Other 8031 (only 7070 analyzed for end-of- treatment analysis) 4 New, smear positive, drug susceptible HCW observed ingest antituberculosis drugs at least 5×/week, trained FM observed patient ingest antituberculosis and record event Self-administered antituberculosis, reviewed monthly. Study [35] reported that 83 (7%) of the 1203 patients assigned to interventions were lost to follow-up, ie, defaulted; however, these defaulters were not reported according to their assigned interventions. Thus, this study was not included in Figure 2.

record event Self-administered antituberculosis, reviewed monthly. Study [35] reported that 83 (7%) of the 1203 patients assigned to interventions were lost to follow-up, ie, defaulted; however, these defaulters were not reported according to their assigned interventions. Thus, this study was not included in Figure 2. Abbreviations: CM, community member; CP, continuation phase; DOT, directly observed therapy; FDC, fixed-dose combination; FM, family member; HCW, healthcare worker; IP, intensive phase (first 2 months of tuberculosis therapy); LHW, lay health worker; SAT, self-administered therapy. a For regimens examined, letters indicate H, isoniazid; R, rifampin; Z, pyrazinamide; E, ethambutol. Subscript denotes number of days per week on therapy and regular script indicates number of months on the regimen. Figure 1. Summary of literature search and study selection for the meta-analysis. Abbreviations: DOT, directly observed therapy; HIV, human immunodeficiency virus; SAT, self-administered therapy; TB, tuberculosis.

a For regimens examined, letters indicate H, isoniazid; R, rifampin; Z, pyrazinamide; E, ethambutol. Subscript denotes number of days per week on therapy and regular script indicates number of months on the regimen. Figure 1. Summary of literature search and study selection for the meta-analysis. Abbreviations: DOT, directly observed therapy; HIV, human immunodeficiency virus; SAT, self-administered therapy; TB, tuberculosis. DOTS Program Performance Significant heterogeneity of effect was observed in the 9 of 10 studies that reported defaulting as an outcome (I2 = 68%; P = .02); therefore, mixed effects models were employed. Results are shown in Figure 2. SAT (n = 3192) had worse defaulting than DOT (n = 8269), based on pooled RD of −0.05 (95% CI, −.07 to −.04; Figure 2). The pooled IR was 19.4% (95% CI, 18.0%–21.0%) on SAT vs 8.8% (95% CI, 6.1%–9.5%) on DOT (Table 2). If we calculated RR by omitting studies with zero cells, the pooled RR was 0.48 (CI, .43–.54), confirming that whichever one of the 3 effect sizes was utilized, DOT was associated with lower defaulting rates compared to SAT. Table 2. Incidence Rates of Defaulting in Patients on Directly Observed Therapy vs Self-Administered Therapy

ted RR by omitting studies with zero cells, the pooled RR was 0.48 (CI, .43–.54), confirming that whichever one of the 3 effect sizes was utilized, DOT was associated with lower defaulting rates compared to SAT. Table 2. Incidence Rates of Defaulting in Patients on Directly Observed Therapy vs Self-Administered Therapy Study [Reference] DOT (95% CI) Relative Weight (%) SAT (95% CI) Relative Weight (%) Randomized controlled trial Kamolratanakul et al [35] 6.5 (4.5–9.3) 25 13.0 (10.1–16.6) 27 Zwarenstein et al [32] 14.4 (9.0–22.2) 24 8.6 (4.5–15.7) 23 Walley et al [5] 32.1 (25.4–39.6) 26 32.7 (25.9–40.3) 27 Zwarenstein et al [33] 20.5 (14.0–29.0) 25 25.0 (14.4–39.7) 23 Pooled IR estimate; REM 16.3 (7.4–32.4) 18.2 (9.4–32.2) Heterogeneity measure (I2) 75% 92% Observational cohort Okanurak et al [30] 5.3 (3.6–7.9) 28 3.0 (1.6–5.7) 22 Jasmer et al [29] 14.8 (9.9–21.4) 27 11.7 (8.1–16.6) 23 Ormerod et al [27] 2.1 (.1–25.9) 3 .3 (–4.1) 8 Anuwatnonthakate et al [31] 7.7 (7.1–8.4) 35 23.6 (21.5–25.9) 24 Pungrassami et al [28] 2.9 (.7–11.0) 8 6.4 (4.3–9.5) 23 Pooled IR estimate; REM 7.5 (4.9–11.3) 6.8 (2.6–16.5) Heterogeneity measure (I2) 95% 96% Overall mixed-effects analysis 8.8 (6.1–9.5) 19.4 (18.0–21.0) Abbreviations: CI, confidence interval; DOT, directly observed therapy; IR, incidence rate; REM, random-effects model; SAT, self-administered therapy.

8 6.4 (4.3–9.5) 23 Pooled IR estimate; REM 7.5 (4.9–11.3) 6.8 (2.6–16.5) Heterogeneity measure (I2) 95% 96% Overall mixed-effects analysis 8.8 (6.1–9.5) 19.4 (18.0–21.0) Abbreviations: CI, confidence interval; DOT, directly observed therapy; IR, incidence rate; REM, random-effects model; SAT, self-administered therapy. Figure 2. Pooled risk differences for defaulting in patients on directly observed therapy compared to self-administered therapy. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy. Effect Size for Microbiologic Outcomes For microbiologic failure, 10 studies randomized patients to either SAT (n = 3376) or DOT (n = 8625). The combined I2 was 0%, indicating no significant heterogeneity. Therefore, fixed-effects models were utilized. The pooled RD for patients on DOT vs SAT was 0.0 (CI, <−.01 to .01; Figure 3). The results held true regardless of whether only randomized controlled trials were considered or observational studies were added (Figure 3). No single study demonstrated a significantly higher risk with SAT compared to DOT. The IR was 1.5% (95% CI, 1.3%–1.8%) on DOT vs 1.7% (95% CI, 1.2%–2.2%) on SAT (Table 3). Moreover, the pooled RR for failure on DOT vs SAT was 1.20 (CI, .81–1.78). No significant small study effects or publication bias was observed based on the Egger test and funnel plot examination (Figure 4). Table 3. Incidence Rates of Microbiologic Outcomes in Patients on Directly Observed Therapy vs Self-Administered Therapy

r, the pooled RR for failure on DOT vs SAT was 1.20 (CI, .81–1.78). No significant small study effects or publication bias was observed based on the Egger test and funnel plot examination (Figure 4). Table 3. Incidence Rates of Microbiologic Outcomes in Patients on Directly Observed Therapy vs Self-Administered Therapy Microbiologic Measures/Study Design DOT (95% CI) Relative Weight (%) SAT (95% CI) Relative Weight (%) Microbiologic failure: Randomized controlled trial Kamolratanakul et al [35] 1.4 (.7–3.2) 29 .2 (–1.7) 15 Zwarenstein et al [32] 1.8 (.5–6.9) 12 1.9 (.5–7.3) 22 Walley et al [5] .3 (–4.6) 3 .3 (–4.7) 10 Zwarenstein et al [33] 3.6 (1.3–9.1) 21 4.5 (1.1–16.4) 21 Tuberculosis Research Centre [34] 3.1 (1.6–6.1) 35 3.6 (2.0–6.4) 32 Pooled IR estimate; REM 2.2 (1.3–3.7) 1.7 (.6–4.6) Heterogeneity measure (I2) 21% 61% Observational cohort Okanurak et al [30] 2.1 (1.1–4.0) 9 2.0 (.9–4.4) 22 Jasmer et al [29] .7 (.1–4.6) 1 1.8 (.7–4.7) 14 Ormerod et al [27] .3 (–4.5) 1 2.1 (.1–25.9) 2 Anuwatnonthakate et al [31] 1.4 (1.1–1.7) 88 .9 (.5–1.6) 47 Pungrassami et al [28] 1.5 (.2–9.7) 1 1.2 (.4–3.1) 15 Pooled IR estimate 1.4 (1.2–1.7) 1.3 (.9–1.8) Heterogeneity measure (I2) 0% 0% Overall mixed-effects analysis 1.5 (1.3–1.8) 1.7 (1.2–2.2) Relapse: Randomized controlled trial Tuberculosis Research Centre [34] 9.3 (6.3–13.7) 100 5.2 (3.1–8.4) 100 Observational cohort Jasmer et al [29] .3 (–5.1) 43 1.9 (.6–5.9) 70 Ormerod et al [27] 4.3 (.6–25.2) 57 .5 (.1–3.7) 30 Pooled IR estimate; REM 1.5(.1–15.7) 1.3 (.4–4.1) Heterogeneity measure (I2) 55% 21% Overall mixed-effects analysis 3.7 (.7–17.6) 2.3 (.7–7.2) Acquired drug resistance Tuberculosis Research Centre [34] 2.7 (1.3–5.6) 71 1.0 (.3–3.0) 60 Jasmer et al [29] .3 (–5.1) 29 .9 (.2–3.5) 40 Pooled IR estimate; REM 1.5 (.2–9.0) .9 (.4–2.3) Heterogeneity measure (I2) 52% 0% Overall mixed-effects analysis 1.5 (.2–9.0) .9 (.4–2.3) Abbreviations: CI, confidence interval; DOT, directly observed therapy; IR, incidence rate; REM, random-effects model; SAT, self-administered therapy.

9] .3 (–5.1) 29 .9 (.2–3.5) 40 Pooled IR estimate; REM 1.5 (.2–9.0) .9 (.4–2.3) Heterogeneity measure (I2) 52% 0% Overall mixed-effects analysis 1.5 (.2–9.0) .9 (.4–2.3) Abbreviations: CI, confidence interval; DOT, directly observed therapy; IR, incidence rate; REM, random-effects model; SAT, self-administered therapy. Figure 3. Pooled risk differences for microbiologic failure in patients on directly observed therapy compared to self-administered therapy. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy. Figure 4. Publication bias analysis and small study effects for microbiologic failure. Abbreviation: RD, risk difference. Three studies reported relapse [27, 29, 34]. The studies had significant heterogeneity (I2 = 68%); therefore, random-effects models were utilized. The pooled RD for relapse on SAT (n = 649) compared to DOT (n = 649) was 0.01 (95% CI, −.03 to .06; Figure 5); the IR was 3.7% (95% CI, .7%–17.6%) on DOT vs 2.3% (95% CI, .7%–7.2%) on SAT (Table 3). The pooled RR was 1.49 (95% CI, 0.31–7.19) for DOT compared to SAT. There was no significant publication bias or small study effects observed (Figure 6). Figure 5. Pooled risk difference for relapse on directly observed therapy compared to self-administered therapy. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy. Figure 6. Publication bias analysis and small study effects for relapse. Abbreviation: RD, risk difference.

Three studies reported relapse [27, 29, 34]. The studies had significant heterogeneity (I2 = 68%); therefore, random-effects models were utilized. The pooled RD for relapse on SAT (n = 649) compared to DOT (n = 649) was 0.01 (95% CI, −.03 to .06; Figure 5); the IR was 3.7% (95% CI, .7%–17.6%) on DOT vs 2.3% (95% CI, .7%–7.2%) on SAT (Table 3). The pooled RR was 1.49 (95% CI, 0.31–7.19) for DOT compared to SAT. There was no significant publication bias or small study effects observed (Figure 6). Figure 5. Pooled risk difference for relapse on directly observed therapy compared to self-administered therapy. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy. Figure 6. Publication bias analysis and small study effects for relapse. Abbreviation: RD, risk difference. The 2 ADR studies were heterogeneous (I2 = 69%). The pooled RD was 0.0 (95% CI, −.01 to .01) when DOT (n = 415) was compared to SAT (n = 532; Figure 7); the IR was 1.5% (95% CI, .2%–9.0%) for patients on DOT and 0.9% (95% CI, .40%–2.30%) for patients on SAT (Table 3). The RR of ADR on DOT vs SAT was 1.40 (95% CI, .20–9.98). Figure 7. Effect of directly observed therapy vs self-administered therapy on acquired drug resistance. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy.

ts on SAT (Table 3). The RR of ADR on DOT vs SAT was 1.40 (95% CI, .20–9.98). Figure 7. Effect of directly observed therapy vs self-administered therapy on acquired drug resistance. Abbreviations: CI, confidence interval; DOT, directly observed therapy; ID, identity; RD, risk difference; SAT, self-administered therapy. Subgroup and Sensitivity Analysis In subgroup analysis, microbiologic failure for rural/urban studies was significantly higher on DOT compared to SAT (P = .045). The pooled RD for studies performed in urban locales was 0.004 (95% CI, −.016 to .008), whereas the RD from rural/urban studies was 0.004 (95% CI, .00–.009). This suggested that rural patients were more likely to fail on DOT compared to SAT. However, there were no studies performed solely in rural areas. No significant changes in pooled RD were encountered when we systematically removed 1 study at a time in influence analysis (Supplementary Figure 1). Next, we examined whether combining all patients classified as partial DOT with either DOT or SAT, or grouped studies by country (hence program quality), or by study design, led to significant changes in conclusions. There was no significant change in effect size for microbiologic failure or ADR or relapse, for all (Supplementary Figures 2–4).

combining all patients classified as partial DOT with either DOT or SAT, or grouped studies by country (hence program quality), or by study design, led to significant changes in conclusions. There was no significant change in effect size for microbiologic failure or ADR or relapse, for all (Supplementary Figures 2–4). Meta-Regression For microbiologic failure, the percentage residual variation due to heterogeneity for a model comprising study design and study locale was 0% and the joint test for both covariates revealed a P = .34. The restricted maximum likelihood estimate for between-study T2 was 0. The RD for study design was 0.01 (95% CI, −.01 to .02), whereas that for study locale was −0.01 (95% CI, −.02 to .01). Thus the findings from the meta-regression demonstrate no other source of variation for the effect obtained, which suggests that there was no significant difference between SAT and DOT.

study T2 was 0. The RD for study design was 0.01 (95% CI, −.01 to .02), whereas that for study locale was −0.01 (95% CI, −.02 to .01). Thus the findings from the meta-regression demonstrate no other source of variation for the effect obtained, which suggests that there was no significant difference between SAT and DOT. DISCUSSION Well-documented decreases in ADR in several cities and countries have provided strong historical evidence of the success of DOTS, based on decreased defaulting rates [1–4, 8–12]. A prior analysis of Volmink and Garner, in a mixture of patients with latent and active tuberculosis, found that DOT was not superior to SAT for the program-defined outcomes of “completion of treatment” [13]. We found that defaulting rates were indeed reduced by DOT. However, despite the poorer defaulting rates on SAT, we found no difference in microbial failure, ADR, or relapse, between DOT and SAT, similar to our findings in our previously published in vitro hollow-fiber studies [18]. One possible potential explanation for the discrepancies with historical data is that those studies were retrospective, and those that were prospective employed quasi-experimental designs. In evidence-based medicine, the highest quality of scientific evidence comes from >1 properly randomized controlled trial, whereas the lowest quality is generally that of descriptive studies or opinions of authorities, whether or not there is consensus [36]. Notably, no single study demonstrated a significantly higher risk of microbiologic failure with SAT compared to DOT. We speculate that the DOTS program is associated with a large infusion of resources such as upgrade in expertise and a reliable supply of drugs, and that the regular contact with a patient further provides a higher level of support apart from direct supervision of therapy, which would lead to apparent improvement in outcomes in retrospective studies, independent of DOT.

h a large infusion of resources such as upgrade in expertise and a reliable supply of drugs, and that the regular contact with a patient further provides a higher level of support apart from direct supervision of therapy, which would lead to apparent improvement in outcomes in retrospective studies, independent of DOT. Our findings should not be read as questioning the entire DOTS program, but are limited to supervision of patients swallowing pills. Although the full program is often accompanied by an infusion of resources, the DOT component itself consumes an inordinate portion of that, which is a problem in resource-constrained settings [6]. This may explain the suggested association between rural residence and microbiologic failure. We speculate that economic constraints were the most likely driver accounting for this observation. It may be that requiring patients to frequently come and pick up their medicines or to be observed swallowing their pills could actually impose economic hardships in some parts of the world, leading to microbiologic failure. Moreover, in some high-burden countries, baseline adherence rates measured using validated methods are already >97% on SAT [37], and there may be no room for further improvement in adherence with DOT. We propose that, instead, a concerted effort should be made to shift the resources toward the other possible reasons for such failure, beyond DOTS, including pharmacokinetic/pharmacodynamics and pharmacokinetic and microbial variability [18, 38]. However, the nature of the data reported precluded us from investigating the role of such factors in the current meta-analysis.

tal cases, 8 had severe underlying diseases and developed complications after being infected with L. monocytogenes. All died of multiple severe complications within 30 days after the onset of infection. The fatal cases were more likely to have sepsis (n = 9), rapid onset of coma (n = 6), and multiorgan failure (n = 3). Healthcare-Associated Listeriosis Eleven (44%; 95% CI, 26.67%–62.93%) nonmaternal adult cases were healthcare-associated. The patients were admitted for treatment of rheumatologic diseases (n = 6), malignancy (n = 4), and malignancy with ulcerative colitis (n = 1). The admitting department and its location, timing of infection, and duration are illustrated in Figure 1. The onset of symptoms related to listeriosis occurred after a median of 20 days (range, 3–44 days) following admission. The mortality among healthcare-associated cases was 27.2% (95% CI, 9.74%–56.56%). Figure 1. Distribution of admission departments and calendar years for 11 healthcare-associated cases of listeriosis. Admission duration is shown in the blue lines in proportion to the time period, and the onset of symptoms consistent with listeriosis is indicated with yellow arrows The letters E and W represent the eastern and western campuses of Peking Union Medical College Hospital.

made to shift the resources toward the other possible reasons for such failure, beyond DOTS, including pharmacokinetic/pharmacodynamics and pharmacokinetic and microbial variability [18, 38]. However, the nature of the data reported precluded us from investigating the role of such factors in the current meta-analysis. There are several limitations to our analyses. First, the WHO definitions we used, particularly for the secondary outcome of “defaulting,” are subject to different interpretations. Second, various DOT supervisors and various forms of DOT were employed by the selected studies, whereas some of the studies did not explicitly state whether DOT was for the initial 2 months of therapy only or for the entire treatment duration. Hence, these data are subject to misclassification bias, which can lead to erroneous failure to reject the null hypothesis [39]. However, the influence and sensitivity analyses we performed did not reveal significant change in the pooled RD, suggesting that these findings are internally robust. Third, it has also been argued that the quality of DOTS programs has an impact on results of meta-analysis, and therefore analysis should be stratified by quality of program. However, we performed a stratified analysis by quality of DOTS program using country as a surrogate, and DOT was still no better than SAT. Fourth, differences in study design and the heterogeneity between studies could make our conclusions less reliable. As an example, it could be that less reliable patients were assigned to DOT whereas more reliable patients were assigned SAT in the observational studies, which would bias the results. However, analysis of randomized studies alone vs analysis that included observational studies did not alter the conclusions (Figure 3). Fifth, ADR and relapse studies were fewer and these were of different study design. The single randomized clinical trial revealed higher risk for relapse with SAT compared to DOT when RR was calculated (RR, 1.74 [95% CI, .93–3.26]); however, it did not achieve statistical significance. For the observational studies, the pooled RR was 1.13 (95% CI, .02–54.91). These results were partly due to zero cells and the imputation strategies inherent with using RR as effect size. That is why our primary effect sizes were RD and IR, which require no such imputation. The differences by study design vanished when those effect sizes were employed.

pooled RR was 1.13 (95% CI, .02–54.91). These results were partly due to zero cells and the imputation strategies inherent with using RR as effect size. That is why our primary effect sizes were RD and IR, which require no such imputation. The differences by study design vanished when those effect sizes were employed. Finally, an inherent limitation of meta-analyses is that some influential studies may be missed during the search, thereby biasing the studies. However, we excluded no studies by publication language, examined the Cochrane database and the gray literature, and performed a manual search of references in key publications, in order to minimize bias. In conclusion, our evidence-based medicine approach found that DOT was not superior to SAT in terms of microbiologic outcomes. Other causes of poor microbiologic outcomes should be sought in new studies. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Financial support. This study was funded by the National Institutes of Health (NIH; grant number R01AI079497). Disclaimer.

Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Financial support. This study was funded by the National Institutes of Health (NIH; grant number R01AI079497). Disclaimer. The NIH was not involved in the design and conduct of the study; collection, management, analysis, and interpretation of data; and preparation, review, or approval of the manuscript. Potential conflicts of interest. T. G. has been a consultant for Merrimack Pharmaceuticals and has received research grants from Pfizer, Merck, and Astellas Pharma for work on antifungal agents. J. G. P. reports no potential conflicts. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Gerding and Johnson on pages 1601–3.) The widespread emergence of hypervirulent polymerase chain reaction (PCR) ribotype 027/NAP1/BI/sequence type (ST) 1 [1] strains in the early 2000s [2, 3] substantially increased Clostridium difficile infection (CDI) incidence. PCR ribotype 027 has also been associated with more severe outcomes in most [2, 4, 5] but not all [6–9] studies. Outcome variation across non-027 strains has rarely been investigated, invariably with small numbers, although these now account for most new CDIs. One study [6] (n = 395) found significantly more complicated disease outcomes with PCR ribotypes 018 (ST 17 from [10]; n = 23) and 056 (ST 34/58 [10]; n = 6), whereas another [11] (n = 168) reported similar 30-day mortality in PCR ribotype-027 (n = 46) and 017 (ST 37 [10]; n = 57). Although PCR ribotype 078 (ST 11), common in livestock [12] and rising in incidence [6, 13], is denoted hypervirulent on the basis of increased toxin production [14] and individual case severity [15], supporting clinical data are few. Attributable mortality and severe diarrhea was similar in PCR ribotype 078 (n = 54) and 027 (n = 124) in 1 study (both greater than in 501 non-027/078 cases) [13], but PCR ribotype 078 (n = 31) was not associated with complicated CDI in another [6]. Although scores to predict CDI severity, complications, or recurrence have variably included biomarkers (eg, white blood count [WBC], C-reactive protein [CRP]) [16], no studies have investigated associations between CDI strains and biomarkers.

ibotype 078 (n = 31) was not associated with complicated CDI in another [6]. Although scores to predict CDI severity, complications, or recurrence have variably included biomarkers (eg, white blood count [WBC], C-reactive protein [CRP]) [16], no studies have investigated associations between CDI strains and biomarkers. We aimed therefore to investigate whether the genotype of C. difficile clinical isolates from multilocus sequence typing (MLST) was associated with mortality and severity biomarkers using a large population-based database of CDI cases and to explore associations between strain-specific effects on host biomarkers and mortality to provide insights into infection pathogenesis.

f C. difficile clinical isolates from multilocus sequence typing (MLST) was associated with mortality and severity biomarkers using a large population-based database of CDI cases and to explore associations between strain-specific effects on host biomarkers and mortality to provide insights into infection pathogenesis. METHODS Oxford University Hospitals (OUH) NHS Trust provides >90% of hospital care and all acute services in Oxfordshire (approximately 600 000 people). It includes 2 large acute teaching hospitals and 1 specialist orthopedic hospital in Oxford and 1 district hospital 35 miles north. The OUH microbiology laboratory tests all stool samples from the county, including those from other healthcare facilities/primary care. From 12 September 2006 to 21 May 2011, all unformed stools submitted for C. difficile toxin testing, positive by enzyme immunoassay (EIA) and with sufficient sample remaining, were routinely cultured and MLST typed [1]. During this period, infection control policy required all inpatients with diarrhea (≥3 unformed stools within 24 hours) to have samples sent for EIA testing and to initiate vancomycin treatment empirically, continuing for 14 days if CDI was confirmed. Additionally, from May 2007, all unformed samples from those aged ≥65 years were routinely EIA tested following UK policy.

all inpatients with diarrhea (≥3 unformed stools within 24 hours) to have samples sent for EIA testing and to initiate vancomycin treatment empirically, continuing for 14 days if CDI was confirmed. Additionally, from May 2007, all unformed samples from those aged ≥65 years were routinely EIA tested following UK policy. C. difficile MLST data were anonymously linked to OUH hospital admissions/discharges, mortality, and laboratory test results from the Infections in Oxfordshire Research Database (IORD) through 21 August 2011 [17]. Admissions to other much smaller regional (including psychiatric/community) hospitals were not included, although samples taken at these locations were identifiable. Rates were calculated using overnight stays defined by the UK KH03 occupancy statistic. IORD has Research Ethics Committee (09/H0606/85) and UK National Information Governance Board (5-07(a)/2009) approval as an anonymized database without individual informed consent. The primary outcome was 14-day mortality after EIA-based CDI detection in adults aged ≥18 years (excluding repeat EIA-positive cases within 14 days; censoring follow-up at 14 days). EIA-negative samples were included as controls (excluding repeat negatives within 14 days and any sample taken after or within 21 days before the first EIA positive). See Supplementary Material for details.

on in adults aged ≥18 years (excluding repeat EIA-positive cases within 14 days; censoring follow-up at 14 days). EIA-negative samples were included as controls (excluding repeat negatives within 14 days and any sample taken after or within 21 days before the first EIA positive). See Supplementary Material for details. The primary exposure was type of CDI, categorized by EIA/culture status or C. difficile phylogenetic clade from MLST [1]. CDI-associated MLST STs correlate reasonably closely with ribotype [18] and can be grouped by evolutionary relationships into clades [10]. These clades persist despite homologous recombination and have the same phylogenetic structure with MLST or whole-genome sequences [19], suggesting they may behave more similarly in humans. Adjusted mortality risks in each clade and STs with >20 cases were estimated using Cox models, with robust variance adjustment for multiple episodes per patient [20]. EIA-negative controls comprised the reference category so that risks reflected CDI-attributable mortality. Independent predictors were identified using backward selection with the Akaike information criterion [21], allowing nonlinear effects of continuous factors [22]. Exposures considered were demographics, sample characteristics, previous hospital exposure, and previous healthcare-associated infections (Table 1) (antibiotic exposure not available). The impact of clade on the 15 biomarkers available for >50% cases within −3 to +1 days of sample collection was estimated using normal regression on BoxCox-transformed values. Associations between biomarkers and 14-day mortality were estimated using Cox models with multiple imputation (see Supplementary Material). Table 1. Characteristics of Clostridium difficile Samples 12 September 2006–21 May 2011 and Relationship With 14-Day Mortality

using normal regression on BoxCox-transformed values. Associations between biomarkers and 14-day mortality were estimated using Cox models with multiple imputation (see Supplementary Material). Table 1. Characteristics of Clostridium difficile Samples 12 September 2006–21 May 2011 and Relationship With 14-Day Mortality Number (%) or Median (IQR) Unadjusted Univariable Model Adjusted Multivariable Modela Factor Levels (Effect in Cox Model) In EIA Negative Controls In EIA Positive Cases HR (95% CI) P HR (95% CI) P Type of test EIA negative 27 550 (100%) … 1.00 <.0001 1.00 <.0001 EIA positive/culture negative 571 (21%) 1.59 (1.19–2.12) 1.59 (.93–2.73) EIA positive/not cultured 281 (10%) 2.61 (1.89–3.61) 2.45 (1.62–3.70) Clade 1 1168 (43%) 2.23 (1.88–2.66) 2.32 (1.71–3.13) Clade 2 (027/ST 1) 560 (20%) 3.95 (3.26–4.79) 3.40 (2.45–4.68) Clade 3 (023) 73 (3%) 1.31 (.53–3.26) 1.65 (.62–4.36) Clade 4 (017/ST 37) 29 (1%) 2.74 (1.04–7.21) 2.65 (.99–7.13) Clade 5 (078/ST 11) 63 (2%) 5.17 (3.16–8.46) 5.37 (3.10–9.32) Demographics Sex Female (vs male) 15 682 (57%) 1566 (57%) 0.79 (.72–.86) <.0001 0.75 (.68–.82) <.0001 Age, years Per 10 years older 74 (63–83) 78 (67–85) 1.42 (1.37–1.47) <.0001 1.41b (1.36–1.47) <.0001 Sample characteristics Location where sample taken Inpatient 16 598 (60%) 1860 (68%) 1.00 <.0001 1.00 <.0001 Primary care 8108 (29%) 557 (20%) 0.14 (.12–.17) 0.06c (.03–.14) Outpatient/ER/day case 1395 (5%) 148 (5%) 0.35 (.27–.47) 0.98c (.35–2.78) Other hospital 1449 (5%) 180 (7%) 0.50 (.40–.63) 0.12c (.05–.30) If inpatient, speciality Surgical 6112 (37%) 549 (30%) 1.00 <.0001 1.00 <.0001 Medical 10 486 (63%) 1311 (70%) 1.91 (1.71–2.15) 1.64 (1.44–1.88) if EIA− 1.64 (.88–3.06) if EIA +, cult − 0.98 (.73–1.30) if EIA +, cult + (interaction P = .004) If inpatient, method Elective 3609 (22%) 363 (20%) 1.00 <.0001 1.00 .01 Emergency 12 989 (78%) 1497 (80%) 1.64 (1.43–1.88) <.0001 1.22 (1.04–1.43) If inpatient, days since admitted Nonlinear effectd 5 (2–12) 9 (2–22) <.0001 <.0001 (Days/10)−1 0.87 (.78–.97) 0.76d (.68–.84) ln(days/10)a(days/10)−1 1.00 (.95–1.04) 0.90d (.86–.94) Clinician requested EIA test when submitting sample No (mild diarrhea) (vs yes) 7895 (29%) 436 (16%) 0.48 (.42–.54) <.0001 0.69 (.51–.92) .01 Days since last negative EIA teste (For every day closer in the last 2 wk) … 4 (1–8) (if test in last 2 wk) 0.97e (.95–1.00) .02 0.96 (.94–.99) .007 Previous C.

) 0.90d (.86–.94) Clinician requested EIA test when submitting sample No (mild diarrhea) (vs yes) 7895 (29%) 436 (16%) 0.48 (.42–.54) <.0001 0.69 (.51–.92) .01 Days since last negative EIA teste (For every day closer in the last 2 wk) … 4 (1–8) (if test in last 2 wk) 0.97e (.95–1.00) .02 0.96 (.94–.99) .007 Previous C. difficile Yes (vs no) 0 (0%) 634 (23%) 0.99e (.78–1.26) .94 (p = 0.18) Previous hospital exposure (strictly before the current admission, if inpatient) Ever previously admitted to OUH Yes, for ≥1 admission >8 hours 19 570 (71%) 2253 (82%) 1.00 <.0001 1.00 .01 Yes, but only for <8 hour admissions 2462 (9%) 139 (5%) 0.55 (.45–.68) 0.93 (.71–1.21) Never 5518 (20%) 353 (13%) 0.63 (.55–.72) 1.30 (1.03–1.63) Previously admitted to GI ward Yes (vs no) 8484 (31%) 981 (36%) 0.95 (.86–1.05) .34 0.89 (.80–.99) .03 Dialysis/chemotherapy at OUH Yes (vs no) 3051 (11%) 332 (12%) 1.37 (1.21–1.56) <.0001 1.39 (1.21–1.60) <.0001 Number of previous admissions >8 hours (per 5 additional >8 hours admissions) 2 (1–4) 2 (1–5) 1.06f (.99–1.12) .08 0.92 (.84–1.00) .06 Previous hospital stay (hours) (Per doubling of total previous hours in hospital) 169 (8–656) 478 (77–1229) 1.11g (1.09–1.13) <.0001 1.02g (.99–1.06) .20 Days since last discharged (Per additional 6 mo since last OUH discharge) 285 (42 to >1096) 78 (22–640) 0.92 (.90–.95) <.0001 0.96 (.93–.98) .002 SHEA [35] classification HO-HCFA 11 628 (42%) 1373 (50%) 1.00 <.0001 (P = .93) CO-HCFA 3432 (12%) 604 (22%) 0.66 (.57–.76) Indeterminate 1892 (7%) 248 (9%) 0.54 (.45–.66) CO 10 598 (38%) 520 (19%) 0.30 (.26–.34) Abbreviations: CI, confidence interval; CO, community onset; CO-HCFA, community onset–health-care facility associated; cult, culture; EIA, enzyme immunoassay; ER, emergency room; GI, gastrointestinal; HO-HCFA, hospital onset–health-care facility associated; HR, hazard ratio; IQR, interquartile range; OUH, Oxford University Hospitals; SHEA, Society for Healthcare Epidemiology of America

or 11 healthcare-associated cases of listeriosis. Admission duration is shown in the blue lines in proportion to the time period, and the onset of symptoms consistent with listeriosis is indicated with yellow arrows The letters E and W represent the eastern and western campuses of Peking Union Medical College Hospital. These infections were first detected in 2006, and there were 1, 3, 3, 1, and 3 infections detected per year from 2006 to 2010, consecutively. These infections were scattered in 6 different wards, both in the eastern and western campuses of PUMCH. There were 3 cases each in the rheumatologic and hematologic wards and 2 cases in the general medicine ward. Only 2 cases appeared to be clustered in space and time. Nine of these 11 cases did not appear to be clustered. There was no consistent pattern (location, seasonality, and timing) that emerged for the 9 nonclustered cases. The source of their infection could not be determined.

, community onset–health-care facility associated; cult, culture; EIA, enzyme immunoassay; ER, emergency room; GI, gastrointestinal; HO-HCFA, hospital onset–health-care facility associated; HR, hazard ratio; IQR, interquartile range; OUH, Oxford University Hospitals; SHEA, Society for Healthcare Epidemiology of America a HR with opposite effect to unadjusted univariable models due to confounding are underlined. P values in italics show the nonsignificant effects of adding in factors not chosen by the Akaike information criterion selection. b Although mortality was lower after tests that had not been directly requested by the clinician, the increase in risk with age was significantly greater following these tests (per 10 years HR = 1.71; 95% CI, 1.48–1.98; interaction P = .009). For those aged <84.4 years, mortality risks were therefore greater after clinician-requested tests; fore those aged ≥84.4 years, mortality risks were greater after tests that had not originally been requested by the clinician. c Mortality reduced even further if EIA test is negative rather than positive (additional HR = 0.63; 95% CI, .43–.94; P = .02). d Significant nonlinearity, with greatest risk of death on day of admission, then dropping sharply, and then gradually rising. e Univariable model also adjusts for positive vs negative EIA test. f Univariable model also adjusts for ever vs never previously admitted.

c Mortality reduced even further if EIA test is negative rather than positive (additional HR = 0.63; 95% CI, .43–.94; P = .02). d Significant nonlinearity, with greatest risk of death on day of admission, then dropping sharply, and then gradually rising. e Univariable model also adjusts for positive vs negative EIA test. f Univariable model also adjusts for ever vs never previously admitted. g Effects significantly (P < .0001) stronger if samples taken in primary care (HR = 1.25; 95% CI, 1.16–1.36 per doubling) or other hospitals (HR = 1.27; 95% CI, 1.16–1.39 per doubling) than as inpatients (HR in table above) or outpatients/ER/day cases (HR = 0.98; 95% CI, .88–1.10 per doubling; interaction P < .0001).

f Univariable model also adjusts for ever vs never previously admitted. g Effects significantly (P < .0001) stronger if samples taken in primary care (HR = 1.25; 95% CI, 1.16–1.36 per doubling) or other hospitals (HR = 1.27; 95% CI, 1.16–1.39 per doubling) than as inpatients (HR in table above) or outpatients/ER/day cases (HR = 0.98; 95% CI, .88–1.10 per doubling; interaction P < .0001). RESULTS From September 2006 to May 2011, after 14-day deduplication, there were 2745 consecutive toxin-EIA-positive stools in 2222 adults (median age, 78 years; interquartile range [IQR], 67–85 years; 2128 (78%) first ever EIA-positive) and 27 550 consecutive EIA-negative stools in 20 722 adults without a previous positive (median age, 74 years; IQR, 63–83 years). Crude 14-day mortality was similar after first (13%) vs subsequent (13%) EIA-positive cases and first (5%) vs subsequent (7%) EIA-negative controls (Figure 1A). Overall attributable mortality was 7.7% (95% confidence interval [CI], 6.4%–9.0%; P < .0001; Figure 1A). Fourteen-day mortality was lower after EIA-positive/culture-negative cases (8%) than after EIA-positive/culture-positive cases (14%; P < .0001), although still higher than the 5% in EIA-negative/culture-negative controls (P = .002). Figure 1. Fourteen-day mortality after enzyme immunoassay (EIA) tests for Clostridium difficile, overall and by strain. A, Fourteen-day mortality by EIA-negative control vs EIA-positive case and multilocus sequencing type clade if culture positive. B, Fourteen-day mortality by sequence type within clade 1. C, Fourteen-day mortality by age (all tests). Most common ribotypes of isolates from each clade (A) or sequence type (B) shown in brackets. Dashed line in (B) shows overall clade 1 mortality. Clade 4 not shown in (C) due to small numbers (n = 29). Abbreviations: EIA, enzyme immunoassay.

by sequence type within clade 1. C, Fourteen-day mortality by age (all tests). Most common ribotypes of isolates from each clade (A) or sequence type (B) shown in brackets. Dashed line in (B) shows overall clade 1 mortality. Clade 4 not shown in (C) due to small numbers (n = 29). Abbreviations: EIA, enzyme immunoassay. In EIA-positive/culture-positive cases, there were substantial mortality differences between C. difficile clades (P < .0001; Figure 1A). Fourteen-day mortality was highest in clade 5 (25%; all PCR ribotype 078/ST 11 [10]), then clade 2 (20%; 99% PCR ribotype 027/ST 1), clade 4 (14%; 97% A-B+ PCR ribotype 017/ST 37), and clade 1 (12%); lowest mortality occurred in clade 3 (7%; all PCR ribotype 023). The heterogeneous clade 1 had 67 STs, 15 with >20 isolates. Observed mortality varied markedly between common clade 1 STs (median, 11%; range, 4%–16%; Figure 1B), although small numbers limited power to distinguish genuine from chance differences (exact P = .76). Fourteen-day mortality was only 4% in ST 44 (95% CI, .7%–10%; exact P = .01 vs other clade 1, post hoc test). Similar relative differences between clades were observed at all ages (Figure 1C). Over the longer term, mortality was consistently higher in clades 2 and 5 and lower in clades 1 and 3 (Figure 2). In inpatients not dying before 14 days, the median stay post–EIA test was significantly longer in EIA-positive cases (median, 16; IQR, 7–32) than in EIA-negative controls (median, 9; IQR, 3–21; P = .0001) and in clade 2 (median, 19; IQR, 10–34) vs 1 (median, 15; IQR 7–32; P = .005). Figure 2. One-year mortality after first-ever Clostridium difficile enzyme immunoassay–positive test or first negative before positive test by strain. Abbreviation: EIA, enzyme immunoassay.

and 2 cases in the general medicine ward. Only 2 cases appeared to be clustered in space and time. Nine of these 11 cases did not appear to be clustered. There was no consistent pattern (location, seasonality, and timing) that emerged for the 9 nonclustered cases. The source of their infection could not be determined. DISCUSSION The most striking finding from this case series is the prevalence of nonclustered healthcare-associated cases of listeriosis. Eleven of 25 nonmaternal listeriosis cases were healthcare-associated. These infections did not appear to be clustered in time and space. There are rare reports of healthcare-associated transmission of L. monocytogenes via contaminated foods, healthcare workers, and infected patients, but most of these were clustered in time and space. For example, a recent study reported a cluster of 6 L. monocytogenes infections in hospitalized adults during a 10-month period in Brazil [28]. The median age of these patients was 80 years and all had underlying severe comorbidities. Four isolates belonged to a single pulsed-field gel electrophoresis (PFGE) genotype, suggesting a common source. The epidemiological investigation pointed to the hospital kitchen as the possible source of contamination.

negative controls (median, 9; IQR, 3–21; P = .0001) and in clade 2 (median, 19; IQR, 10–34) vs 1 (median, 15; IQR 7–32; P = .005). Figure 2. One-year mortality after first-ever Clostridium difficile enzyme immunoassay–positive test or first negative before positive test by strain. Abbreviation: EIA, enzyme immunoassay. Many potential risk factors were strongly associated with 14-day mortality as expected (Table 1; Supplementary Material). CDI cases, particularly those from clade 2 (PCR ribotype 027/ST 1), were older and generally had more of these risk factors. However, variations in 14-day mortality across C. difficile clades remained after adjustment (P < .0001; Figure 3). Strong evidence of higher mortality after clade 5 (PCR ribotype 078) vs clade 1 CDI (P = .001) and after clade 2 (PCR ribotype 027) vs clade 1 CDI (P = .002) persisted, with a trend toward higher mortality with clade 5 vs clade 2 CDI (P = .09). Further, although clades 3 and 5 are genetically similar in several pathogenicity locus genes [10], mortality differed significantly between clade 5 vs clade 3 CDI (P = .03). Within clade 1, adjusted 14-day mortality risks remained lower in ST 44 (hazard ratio [HR], 0.31 vs other clade 1; 95% CI, .10–.98; interaction P = .05). After adjustment, 14-day mortality decreased year-on-year from 2006 to 2011 in EIA-positive cases (HR per year, 0.88; 95% CI, .80–.96) but not EIA-negative controls (HR, 1.03; 95% CI, .99–1.07; interaction P = .002), with no evidence of differential effects in clade 2 (P = .91). Figure 3. Variation in 14-day mortality risks according to Clostridium difficile clade. Abbreviations: adj, adjusted; CI, confidence interval; cult, culture; EIA, enzyme immunoassay; het, heterogeneity test.

R, 1.03; 95% CI, .99–1.07; interaction P = .002), with no evidence of differential effects in clade 2 (P = .91). Figure 3. Variation in 14-day mortality risks according to Clostridium difficile clade. Abbreviations: adj, adjusted; CI, confidence interval; cult, culture; EIA, enzyme immunoassay; het, heterogeneity test. Variation in biomarkers at CDI diagnosis across clades and associations between excess biomarkers and excess mortality risks broadly followed three patterns. There was strong evidence for higher neutrophils/WBC in EIA-positive cases vs EIA-negative controls and in clades 2, 3, and 5 vs 1 (all P < .01) (Figure 4A and 4B; Supplementary Table 1). In clade 1–5 CDI cases, 31%, 46%, 48%, 21%, and 50%, respectively, had WBC > 15 × 109/L (P < .0001) vs 15% in EIA-negative controls. Excess neutrophils/WBC and excess mortality risks were strongly associated across clades (rho = 0.6). However, clade 3 appeared dissimilar to other clades, with significantly higher neutrophil/WBC vs clade 1, similar to clades 2 (PCR ribotype 027/ST 1) and 5 (PCR ribotype 078/ST 11), despite significantly lower mortality. Variation across clades was similar, but slightly weaker, for CRP (P = .05) and eosinophils (P = .03; Figure 4C and 4D), with more severe (higher) CRP and (lower) eosinophils in clades 3 and 5. Associations between excess biomarker and mortality risks were also weaker (rho = 0.48, −0.35, respectively). At CDI diagnosis, albumin was significantly lower (Figure 4E) and platelets significantly higher (Supplementary Figure 1H) in EIA-positive cases vs EIA-negative controls (P < .0001), but there was no evidence of clade-specific differences (P > .50). In clades 1–5, 8%, 7%, 4%, 5%, and 15%, respectively, had albumin < 25 g/dL (P = .53) vs 5% in EIA-negative controls. However, excess mortality risks tracked reasonably closely with greater albumin reductions vs EIA-negative controls, suggesting that greater patient-level variation may have reduced power. Figure 4. Variation in 7 biomarkers at diagnosis according to Clostridium difficile clade and association with mortality. A, Neutrophils (×109/L). B, White cell count (×109/L). C, C-reactive protein (mg/L). D, Eosinophils (×109/L). E, Albumin (g/dL). F, Sodium (mmol/L). G, Hemoglobin (g/dL).

ion may have reduced power. Figure 4. Variation in 7 biomarkers at diagnosis according to Clostridium difficile clade and association with mortality. A, Neutrophils (×109/L). B, White cell count (×109/L). C, C-reactive protein (mg/L). D, Eosinophils (×109/L). E, Albumin (g/dL). F, Sodium (mmol/L). G, Hemoglobin (g/dL). For each biomarker, left-hand panels show mean (95% confidence interval) values at sample collection for enzyme immunoassay (EIA)–negative controls vs EIA-positive cases; then subdividing EIA-positive cases into culture-negative, not cultured, and culture-positive cases; then subdividing culture-positive cases by clade and comparing sequence type (ST) 44 vs other STs within clade 1; with P values testing for heterogeneity across each group. Means are calculated on BoxCox-transformed values and back-transformed for presentation (see Supplementary Methods). For each clade and EIA-positive/culture-negative cases, the right-hand panels plot the standardized adjusted mean difference vs EIA-negative controls from the left-hand panel (on the BoxCox-transformed scale,±standard error) against the adjusted hazard ratio for mortality vs EIA-negative controls from Table 1. The correlation, ρ, between biomarker and mortality risk excesses was estimated using multivariable random effects meta-analysis (see Supplementary Methods). Diagonal lines show the line of best fit (ie, the best prediction of excess mortality for any given excess in biomarkers compared with EIA-negative controls). If differences in biomarkers across clades completely explained mortality differences (ie, the biomarker was a perfect surrogate for mortality), all the points would lie on the diagonal line. The closer the points are to the diagonal line, the stronger the relationship between biomarker differences and excess mortality risks. Points lying far from the diagonal line indicate a mismatch, either high excess mortality with little difference in biomarkers from EIA-negative controls or vice versa. Abbreviations: CRP, C-reactive protein; cult, culture; EIA, enzyme immunoassay; OUH, Oxford University Hospitals; SE, standard error.

mortality risks. Points lying far from the diagonal line indicate a mismatch, either high excess mortality with little difference in biomarkers from EIA-negative controls or vice versa. Abbreviations: CRP, C-reactive protein; cult, culture; EIA, enzyme immunoassay; OUH, Oxford University Hospitals; SE, standard error. Serum sodium was slightly but significantly lower in EIA-positive cases vs EIA-negative controls (P = .006) and in clade 2 (Figure 4F). Although clades 2 and 5 had highest mortality, if anything, sodium was increased in clade 5 CDI (P = .08 vs clade 2), leading to no overall association between differences in sodium and excess mortality risks across the different clades (rho = 0.02). Hemoglobin was significantly lower in EIA-positive cases vs EIA-negative controls (P < .0001; Figure 4G), but clade-specific variation was restricted to higher hemoglobin in clade 4 (P = .05), with little association with excess mortality (rho = 0.22). Qualitatively, variation across clades in alanine aminotransferase (ALT), creatinine, estimated glomerular filtration rate [23, 24], and serum potassium was similar to hemoglobin (Supplementary Figure 1, I–L). No clear associations were evident for urea or alkaline phosphatase (Supplementary Figure 1N and 1O).

= 0.22). Qualitatively, variation across clades in alanine aminotransferase (ALT), creatinine, estimated glomerular filtration rate [23, 24], and serum potassium was similar to hemoglobin (Supplementary Figure 1, I–L). No clear associations were evident for urea or alkaline phosphatase (Supplementary Figure 1N and 1O). Comparing associations individually for clade 1 STs (Figure 5) supported the partial surrogacy of differences in neutrophils/WBC (rho = 0.48), CRP (rho = 0.43), and eosinophils (rho = −0.45) for excess mortality risk but suggested a stronger relationship with albumin (rho = −0.47). Lack of association for other biomarker changes remained (eg, sodium rho = 0.06; Figure 5D). ST 44 was an outlier, with significantly lower albumin but similar neutrophils/CRP and mortality risk to EIA-negative controls. Figure 5. Impact of Clostridium difficile clade and individual sequence type (ST) on biomarkers compared with mortality. A, Neutrophils (×109/L). B, C-reactive protein (mg/L). C, Albumin (g/dL). D, Sodium (mmol/L). For clades 2–5 (labelled C2, C3, C4, C5) and each clade 1 ST with >20 isolates, the panels plot the standardized adjusted mean difference vs enzyme immunoassay (EIA)–negative controls (on the BoxCox-transformed scale,±standard error) against the hazard ratio for mortality vs EIA-negative controls, adjusted as in Table 1. The correlation, ρ, between biomarker and mortality risk excesses across STs/clades was estimated using multivariable random effects meta-analysis (see Supplementary Methods). Diagonal lines show the line of best fit (ie, the best prediction of excess mortality for any given excess in biomarkers compared with EIA-negative controls), together with a 95% credibility region indicated by the shaded region. If a biomarker was a perfect surrogate for mortality (ie, differences in biomarkers across STs/clades completely explained mortality differences), all the points would lie on the diagonal line. The closer the points are to the diagonal line, the stronger the relationship between biomarker differences and excess mortality risks. Points lying far from the diagonal line indicate a mismatch, either high excess mortality with little difference in biomarkers from EIA-negative controls or vice versa. All clade 1 STs lying outside the 95% credibility region on any of the 4 panels are labelled on each panel; ST 58, which had high mortality in [6], is also labelled.

ar from the diagonal line indicate a mismatch, either high excess mortality with little difference in biomarkers from EIA-negative controls or vice versa. All clade 1 STs lying outside the 95% credibility region on any of the 4 panels are labelled on each panel; ST 58, which had high mortality in [6], is also labelled. Abbreviations: CRP, C-reactive protein; cult, culture; EIA, enzyme immunoassay; HR, hazard ratio; SE, standard error; ST, sequence type. Lastly, we estimated how much of the variation in C. difficile clade-associated mortality risk was related to observed biomarker differences. As expected given large numbers, all biomarkers except ALT independently predicted 14-day mortality in addition to Table 1 factors (Supplementary Table 2). However, association strength varied substantially, with albumin, urea, eosinophils, sodium, and CRP most strongly (and creatinine/estimated glomerular filtration rate most weakly) related to mortality. Adjusting for baseline biomarkers explained 41%, 32%, and 37% of the increased mortality due to clades 1, 2, and 5, respectively (Figure 3). However, even after adjusting for these biomarker differences across C. difficile clades (Figure 4), significant mortality risk variation by clade remained (P = .03), with significantly higher mortality persisting in clade 2 (PCR ribotype 027) vs clade 1 (P = .01) CDIs.

e to clades 1, 2, and 5, respectively (Figure 3). However, even after adjusting for these biomarker differences across C. difficile clades (Figure 4), significant mortality risk variation by clade remained (P = .03), with significantly higher mortality persisting in clade 2 (PCR ribotype 027) vs clade 1 (P = .01) CDIs. DISCUSSION In the largest population-based study of genotype and CDI severity to date, we have exhaustively investigated the relationships between strain types, biomarkers, other risk factors, and mortality. We have demonstrated unequivocally that PCR ribotype 027/NAP1/BI/ST 1 (clade 2) strains have been, and continue to be, associated with greater attributable mortality. This excess risk persists even after adjusting for large differences in severity biomarkers. Further, PCR ribotype 078 (clade 5) CDI has attributable mortality at least as great as PCR ribotype 027/ST 1, in agreement with 1 previous study [13] but in contrast with another [6]. Although PCR ribotype 078/clade 5 strains are currently present at low frequency, prospective surveillance demonstrates their continued expansion [25]; ongoing monitoring therefore remains essential.

ty at least as great as PCR ribotype 027/ST 1, in agreement with 1 previous study [13] but in contrast with another [6]. Although PCR ribotype 078/clade 5 strains are currently present at low frequency, prospective surveillance demonstrates their continued expansion [25]; ongoing monitoring therefore remains essential. Comprehensive simultaneous characterization of the impact of different C. difficile strains on biomarkers and mortality, not previously described to our knowledge, has enabled us to show that strain-type-specific excess mortality risk correlates most closely with strain-type-specific changes in inflammatory biomarkers. Conceptually the framework behind these analyses is similar to that for assessing surrogacy of intermediate for clinical outcomes (eg, blood pressure for cerebrovascular disease) [26]. Some biomarkers, notably renal-related biomarkers (creatinine, eGFR), were prognostic for mortality but did not vary significantly across CDI cases/controls or clades (ie, were acting independently of CDI). Others were prognostic and differed significantly between CDI cases and EIA-negative controls but not across clades. The most prognostic marker, albumin, fell into this category, possibly because of large variability. Biomarkers in the most interesting group, particularly neutrophils/WBC, CRP, and eosinophils, were prognostic and demonstrated evidence of partial surrogacy (ie, greater differences in baseline biomarkers between clades translated into greater differences in 14-day mortality). This has 2 consequences: First, quantitative traits like these biomarkers may provide greater power than time-to-event outcomes to detect effects of polymorphisms in genome-wide association studies. Second, surrogate markers indicate causal mechanisms of bacterial pathogenesis and may identify future therapeutic areas for investigation. Our results implicate inflammatory pathways as the major influence on poor outcome after CDI.

-event outcomes to detect effects of polymorphisms in genome-wide association studies. Second, surrogate markers indicate causal mechanisms of bacterial pathogenesis and may identify future therapeutic areas for investigation. Our results implicate inflammatory pathways as the major influence on poor outcome after CDI. Although we found strong associations between strain-specific biomarkers and mortality overall, we also discovered intriguing exceptions that, as exploratory findings, may indicate important areas for future investigation. Specific genotypes within the large, heterogenous clade 1, notably ST 44, had particularly low 14-day mortality in post hoc analyses. Although ST 44 differs by only 1 nucleotide on MLST from ST 10, respective 14-day mortality was 3% and 11%, the latter typical of clade 1 overall (12%). However, both STs are consistently identified as PCR ribotype 015 [10]. They differ by >1500 single nucleotide polymorphisms across the genome [19] and may also differ in their accessory genomes, suggesting possible areas for future study. In contrast, our data suggest ST 49 (PCR ribotype 014) could be a more severe clade 1 genotype; this is an emergent clone in the United Kingdom [25] and should be monitored closely. Another intriguing finding is the major disconnect between the impact of clade 3 CDI on neutrophils/WBC/CRP and mortality. Similarities between clades 3 and 5 in severity biomarkers might be expected, as the receptor-binding domain of their pathogenicity locus tcdB gene (encoding one of the major known clostridial toxins) is highly genetically similar and their tcdC sequences share the same protein-truncating nucleotide substitution [10]. The latter is phenotypically equivalent to the single nucleotide deletion in the clade 2/PCR ribotype 027 tcdC, which causes a protein-truncating frameshift [10] and possibly leads to hypervirulence through increased toxin expression [27, 28] (although recent studies have questioned this [29]). Clades 2, 3, and 5 are also binary toxin positive [10] (in contrast with clades 1 and 4). However, the substantially lower mortality in clade 3 vs clade 5 highlights the importance of other, as yet undetermined, virulence or host factors to clinical outcomes [30] and suggests that increased toxin production alone in PCR ribotype 078 cannot account for its virulence.

e [10] (in contrast with clades 1 and 4). However, the substantially lower mortality in clade 3 vs clade 5 highlights the importance of other, as yet undetermined, virulence or host factors to clinical outcomes [30] and suggests that increased toxin production alone in PCR ribotype 078 cannot account for its virulence. Overall, we found 30%–40% of differences in mortality risk between strains were due to differences in biomarkers at diagnosis. However, in contrast with a recent much smaller study [31], even after adjusting for biomarker differences (and other factors) significant mortality differences remained across clades; this suggests that further microbial virulence determinants remain to be identified. Of note, the biomarker-adjusted effects of strain (reported in [31]) adjust away any effect of strain on outcome mediated through biomarkers, effects that we show to be substantial (Figure 4).

tality differences remained across clades; this suggests that further microbial virulence determinants remain to be identified. Of note, the biomarker-adjusted effects of strain (reported in [31]) adjust away any effect of strain on outcome mediated through biomarkers, effects that we show to be substantial (Figure 4). Our study has some limitations. The EIA assay used for case ascertainment has suboptimal sensitivity (91.7% in [32]), similar to other toxin EIAs [32, 33]. However, because of widespread concerns about sensitivity, for most of the study (through December 2009), multiple diarrheal samples were submitted from each patient, simultaneously or serially (500–1100 EIA tests performed monthly), reducing the chance of completely missing symptomatic CDI. One consequence is that we almost certainly identified false positives, perhaps explaining some EIA-positive/culture-negative cases [34]. To reduce the impact of false negatives, our controls only included EIA-negative tests >21 days before the first EIA positive result. During the study, there were 9.2 EIA-positive CDIs/10 000 overnight stays in inpatients, compatible with the 3.8–9.5 EIA-positive CDIs/10 000 overnight stays typical in endemic settings [35]. Overall, 14-day mortality attributable to EIA-positive CDI was 7.7%, similar to the 8% in a meta-analysis of 10 975 cases from 27 studies after 2000 [36] and 11% in another large study [37], also suggesting generalizability. By necessity, analyses were limited to available electronic data, which did not include previous/concomitant antibiotics, specific comorbid conditions, or causes of death. Although antibiotics are undoubtedly critical for developing CDI, given the lack of impact of adjusting for other important risk factors on strain–mortality associations, it is plausible that further adjustments would have had little further effect. Although theoretically C. difficile–related deaths should provide a more accurate measure of attributable mortality, practically attributing causes is subjective and usually unaudited. In contrast, all-cause mortality is objective, and differences in early mortality between EIA-positive cases vs EIA-negative diarrhea controls should be directly or indirectly CDI related. Although previous studies have considered 30-day mortality [5], reasonable reinfection rates between 14–30 days [38] influenced our prespecified choice of primary outcome.

ctive, and differences in early mortality between EIA-positive cases vs EIA-negative diarrhea controls should be directly or indirectly CDI related. Although previous studies have considered 30-day mortality [5], reasonable reinfection rates between 14–30 days [38] influenced our prespecified choice of primary outcome. However, strain differences were similar at 30 days, and survival curves were parallel subsequently (Figure 2).

ctive, and differences in early mortality between EIA-positive cases vs EIA-negative diarrhea controls should be directly or indirectly CDI related. Although previous studies have considered 30-day mortality [5], reasonable reinfection rates between 14–30 days [38] influenced our prespecified choice of primary outcome. However, strain differences were similar at 30 days, and survival curves were parallel subsequently (Figure 2). Our study also has important strengths. First is its comprehensive scope, including cases from an entire region over almost 5 years, including 3 hospitals providing acute services and numerous secondary/primary care providers. Second, it included 1893 EIA-positive/culture-positive strain-typed cases, approximately double the largest previous studies (n = 1008 [5]; n = 715 [13]). Study size becomes increasingly important when exploring differences between strains; 700–800 cases are needed to detect an 8% absolute mortality increase (as observed between clade 1 vs clade 2) with 80% power. Inadequate power therefore likely explains why smaller studies failed to identify associations between PCR ribotype 027 and severe outcomes (eg, n = 123 [7]; n = 128 [39]; n = 236 [40]). We were also able to compare strains at the clade/ST level, whereas most previous studies have only compared 027 vs non-027 strains [5], pooling 4 heterogeneous clades. We were unable to confirm previous reports [6] of poorer outcomes with PCR ribotypes 018 (ST 17 [10]) and 056 (ST 34/58 [10]), although longer-term mortality was similar in clade 4 (PCR ribotype 017/ST 37) and clade 2 (PCR ribotype 027/ST 1) as previously reported [11]. Our data confirm that the lack of the large clostridial toxin A (tcdA) in these clade 4 cases does not lead to less severe outcomes. We did not find any evidence of greater year-on-year mortality reductions in PCR ribotype 027/ST 1 (clade 2) compared with other clades [39], suggesting overall improvements in outcome are more likely due to better patient management than strain effects. The other mortality risk factors we identified broadly agree with previous studies [16], mostly reflecting disease severity or subsequent management; however, unlike previous studies, we have adjusted for the potential confounding due to bacterial type.

more likely due to better patient management than strain effects. The other mortality risk factors we identified broadly agree with previous studies [16], mostly reflecting disease severity or subsequent management; however, unlike previous studies, we have adjusted for the potential confounding due to bacterial type. In summary, MLST demonstrates that strain predicts mortality and severity biomarkers at both clade and individual sequence-type level. For patient monitoring, neutrophils/WBC, CRP, and albumin are the key C. difficile–associated biomarkers that are highly prognostic for short-term mortality and also partial surrogates (with the possible exception of clade 3). For surveillance, PCR ribotype 078/ST 11 (clade 5) is associated with severe CDI, and its prevalence provides an important context for hospital mortality data [25]. Lastly, our study demonstrates the power from integrating large electronic databases with molecular sequence-based typing. Using whole-genome sequencing, approximately 85% of an approximately 4.3-Mb reference C. difficile genome can be called using standard mapping [19], providing unparalleled resolution to investigate severity determinants compared with the 7.4-kb MLST sequence used here. Unexpected differences in strains appearing highly similar by MLST and in biomarker vs mortality relationships hint at the advances that pathogen whole-genome association studies will provide in our understanding of bacterial pathogenesis over the next decade.

erminants compared with the 7.4-kb MLST sequence used here. Unexpected differences in strains appearing highly similar by MLST and in biomarker vs mortality relationships hint at the advances that pathogen whole-genome association studies will provide in our understanding of bacterial pathogenesis over the next decade. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. The study was conceived and designed by A. S. W., T. E. A. P., D. W. E., M. H. W., and D. W. C., with analysis performed by A. S. W. D. H. W., J. F., B. S., S. O., L. O. C., K. E. D., A. V., and D. G. contributed to data acquisition. All authors contributed to data interpretation. A. S. W. wrote the first draft, which all authors commented on, and all authors approved the final version. A. S. W. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis and the decision to submit for publication.

data interpretation. A. S. W. wrote the first draft, which all authors commented on, and all authors approved the final version. A. S. W. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis and the decision to submit for publication. We thank all the people of Oxfordshire who contribute to the Infections in Oxfordshire Research Database. Research Database Team: P. Bejon, C. Bunch, D. C. W. Crook, J. Finney, J. Gearing (community), H. Jones, L. O'Connor, T. E. A. Peto (PI), J. Robinson (community), B. Shine, A. S. Walker, D. Waller, and D. Wyllie. Financial support. This work was supported by the National Institute for Health Research (NIHR) under its Oxford Biomedical Research Centre Infection Theme and the UKCRC Modernising Medical Microbiology Consortium, the latter funded under the UKCRC Translational Infection Research Initiative supported by Medical Research Council, Biotechnology and Biological Sciences Research Council and the National Institute for Health Research on behalf of the Department of Health (grant G0800778) and the Wellcome Trust (grant 087646/Z/08/Z). D. W. C. and T. E. A. P. are NIHR Senior Investigators. D. W. E. is an NIHR Doctoral Research Fellow. The views expressed in this publication are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health.

0800778) and the Wellcome Trust (grant 087646/Z/08/Z). D. W. C. and T. E. A. P. are NIHR Senior Investigators. D. W. E. is an NIHR Doctoral Research Fellow. The views expressed in this publication are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health. Potential conflicts of interest. The institution of D. W. C. and T. E. A. P. received per-case funding from Optimer Pharmaceuticals to support fidaxomicin trial patient expenses. D. W. C. and T. E. A. P. also received honoraria from Optimer Pharmaceuticals for participation in additional meetings related to investigative planning for fidaxomicin. M. H. W. has received honoraria for consultancy work, financial support to attend meetings, and research funding from bioMerieux, Optimer, Novacta, Pfizer, Summit, The Medicines Company, and Viropharma. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Fixed-dose combination (FDC) antiretrovirals such as emtricitabine/tenofovir disoproxil fumarate (FTC/TDF) and lamivudine/abacavir (3TC/ABC) allow simplification of regimens to potentially improve outcomes by augmenting adherence [1]. Comparative studies of FTC/TDF to 3TC/ABC-containing regimens tend to favor the FTC/TDF arm in regards to efficacy and/or safety [2, 3]. In a large, prospective, treatment-naive trial, subjects with baseline HIV-1 RNA >100 000 c/mL had a lower rate of virologic failure on FTC/TDF compared to 3TC/ABC-containing regimens [3]. Similarly, the BICOMBO study showed that virologically suppressed subjects on a 3TC-containing regimen had a lower rate of virologic failure when switched to FTC/TDF compared to 3TC/ABC-containing regimens [2]. In fact, the 3TC/ABC arm was not noninferior or comparable to the FTC/TDF arm [2]. The total cholesterol (TC), low-density lipoprotein (LDL) cholesterol, and triglycerides (TG) were significantly lower for subjects on FTC/TDF compared to 3TC/ABC [4]. In another study ROCKET 2, virologically suppressed subjects with dyslipidemia on lopinavir/ritonavir (LPV/r) also showed significant declines in TC, LDL, and TG levels 12 weeks following switch to FTC/TDF-compared to 3TC/ABC-containing regimens [5]. Other studies support similar lipid improvement with FTC/TDF [3, 6]. Finally, some but not all studies have shown an association of 3TC/ABC use with an increased relative risk rate of myocardial infarction (MI) [7–13].

LDL, and TG levels 12 weeks following switch to FTC/TDF-compared to 3TC/ABC-containing regimens [5]. Other studies support similar lipid improvement with FTC/TDF [3, 6]. Finally, some but not all studies have shown an association of 3TC/ABC use with an increased relative risk rate of myocardial infarction (MI) [7–13]. US treatment Guidelines list FTC/TDF as a preferred and 3TC/ABC as an alternative NRTI backbone [1, 14]. In light of this, we undertook a prospective, randomized, open-label trial (SWIFT) to evaluate the virologic efficacy and safety potentials and risks of a nucleos(t)ide backbone switch from 3TC/ABC to FTC/TDF in virologically suppressed subjects receiving a ritonavir-boosted protease inhibitor (PI) based regimen.

14]. In light of this, we undertook a prospective, randomized, open-label trial (SWIFT) to evaluate the virologic efficacy and safety potentials and risks of a nucleos(t)ide backbone switch from 3TC/ABC to FTC/TDF in virologically suppressed subjects receiving a ritonavir-boosted protease inhibitor (PI) based regimen. METHODS The SWIFT study was a 48 week prospective, randomized, open–label, multicenter study to evaluate the safety and efficacy of switching FDCs from 3TC/ABC to FTC/TDF in virologically suppressed, HIV-1 infected patients maintained on their boosted PI. Eligible subjects were ≥18 years old, males and nonpregnant females, receiving 3TC/ABC plus a boosted PI with HIV-1 RNA < 200 copies/mL for at least 3 months prior to study entry and < 200 copies/mL at screening by the COBAS TaqMan version 1.0 assay (TaqMan). Subjects had to have an estimated glomerular filtration rate (eGFR) ≥ 50 mL/minutes by the Cockcroft-Gault (CG) method, AST and ALT ≤ 5 times the upper limit of normal, and, if receiving lipid-lowering agents, the drug and dose had to be stable for ≥3 months. Subjects were excluded if they were receiving antiretroviral agents in addition to 3TC/ABC plus a boosted PI, had known historical resistance to any of the study agents including resistance mutations to FTC/TDF (including K65R, M184V/I, or multiple thymidine analogs) or PIs. Subjects were stratified by LPV/r versus other PIs, and by co-morbidities (diabetes mellitus, hyperlipidemia and cardiovascular disease). Antiviral efficacy was assessed by serial measurements of plasma HIV-1 RNA at baseline, and weeks 4, 12, 24, 36, and 48, and at early study discontinuation, if it occurred. Subjects with HIV-1 RNA >200 copies/mL had the test repeated at the investigator's discretion.

ellitus, hyperlipidemia and cardiovascular disease). Antiviral efficacy was assessed by serial measurements of plasma HIV-1 RNA at baseline, and weeks 4, 12, 24, 36, and 48, and at early study discontinuation, if it occurred. Subjects with HIV-1 RNA >200 copies/mL had the test repeated at the investigator's discretion. The primary objective was to assess non-inferiority of FTC/TDF relative to 3TC/ABC measured by the proportion of subjects who maintained HIV-1 RNA < 200 c/mL through week 48 (intent-to-treat, missing = failure). Secondary objectives included evaluation of safety and tolerability, changes in CD4 cell count, assessment of eGFR using the CG, and the abbreviated modified diet in renal disease (MDRD) methods, and evaluation of change in fasting lipid parameters (TG, TC, LDL, HDL, TC: HDL). In a subset, certain cardiovascular biomarkers (high-sensitivity C-reactive protein [hsCRP], interleukin 6 [IL-6], interleukin 10 [IL-10], tumor necrosis factor α [TNF-α], and fibrinogen) were explored over the 48 weeks. Changes in the risk of coronary heart disease (CHD) outcomes were determined by 10-year Framingham risk scores [15, 16]. STATISTICAL ANALYSIS The treated analysis set, used for safety and outcome summaries, includes subjects who were randomized and received at least 1 dose of study drug. The intent to treat (ITT) analysis set, used for efficacy analysis, excludes those with major protocol violations from the treated analysis set.

The primary objective was to assess non-inferiority of FTC/TDF relative to 3TC/ABC measured by the proportion of subjects who maintained HIV-1 RNA < 200 c/mL through week 48 (intent-to-treat, missing = failure). Secondary objectives included evaluation of safety and tolerability, changes in CD4 cell count, assessment of eGFR using the CG, and the abbreviated modified diet in renal disease (MDRD) methods, and evaluation of change in fasting lipid parameters (TG, TC, LDL, HDL, TC: HDL). In a subset, certain cardiovascular biomarkers (high-sensitivity C-reactive protein [hsCRP], interleukin 6 [IL-6], interleukin 10 [IL-10], tumor necrosis factor α [TNF-α], and fibrinogen) were explored over the 48 weeks. Changes in the risk of coronary heart disease (CHD) outcomes were determined by 10-year Framingham risk scores [15, 16]. STATISTICAL ANALYSIS The treated analysis set, used for safety and outcome summaries, includes subjects who were randomized and received at least 1 dose of study drug. The intent to treat (ITT) analysis set, used for efficacy analysis, excludes those with major protocol violations from the treated analysis set. The primary endpoint was the proportion of subjects with HIV-1 RNA < 200 c/mL through week 48 by time to loss of virologic response (TLOVR) algorithm. TLOVR responders were those who completed the study and maintained HIV-1 RNA < 200 c/mL through week 48 without intervening VF. VF was defined as confirmed on-study HIV-1 RNA ≥ 200 c/mL on 2 successive occasions or the last on-study HIV-1 RNA ≥ 200 c/mL. Subjects were considered failures in the TLOVR analysis if they experienced VF, discontinued study medication before week 48, or changed to a new antiretroviral (ARV) regimen. A 2-sided exact 95% confidence interval (CI) for the difference in treatment group response rate (FTC/TDF minus 3TC/ABC) was constructed using inverted 2 one-sided tests with the standardized statistics. The FTC/TDF group was considered noninferior to the 3TC/ABC group if the lower confidence bound of the responder difference was greater than –12%.

interval (CI) for the difference in treatment group response rate (FTC/TDF minus 3TC/ABC) was constructed using inverted 2 one-sided tests with the standardized statistics. The FTC/TDF group was considered noninferior to the 3TC/ABC group if the lower confidence bound of the responder difference was greater than –12%. Descriptive statistics summarize secondary efficacy endpoints. Confidence intervals (95%) and tests of significance, all 2-sided were also used for measure of interest of secondary efficacy endpoints. Observed values and changes from baseline in the risk of CHD outcomes for the 10-year Framingham risk score were analyzed. The Framingham risk score was calculated based on using both the fasting TC score and also based on the fasting LDL score approaches. Framingham risk scores were summarized using descriptive statistics and differences between treatment groups and were compared using Wilcoxon rank sum test. The HIV-1 RNA threshold for VF was amended in the protocol 1 year into the study from 50 c/mL to 200 c/mL based on data regarding discordance between the COBAS Amplicor and the TaqMan HIV-1 test. The data showed an increased rate of samples with >50 c/mL in the TaqMan assay that were < 50 c/mL in the COBAS Amplicor assay [17]. This protocol change was consistent with the ACTG standard of < 200 c/mL [18]. Subjects meeting criteria for VF had genotypic resistance testing performed on their last available plasma sample if HIV-1 RNA >1000 c/mL.

ate of samples with >50 c/mL in the TaqMan assay that were < 50 c/mL in the COBAS Amplicor assay [17]. This protocol change was consistent with the ACTG standard of < 200 c/mL [18]. Subjects meeting criteria for VF had genotypic resistance testing performed on their last available plasma sample if HIV-1 RNA >1000 c/mL. RESULTS A total of 312 subjects were randomized from 76 North American centers. One subject randomized to FTC/TDF withdrew consent before receiving study treatment and was excluded from the efficacy and safety analysis. Overall, 311 subjects were randomized and treated (155 started FTC/TDF and 156 continued 3TC/ABC). One subject randomized to FTC/TDF was excluded from the ITT analysis set for a major protocol violation (documented prior resistance to study drug). Demographic and baseline disease characteristics are summarized in Table 1. Table 1. Baseline Demographics and Characteristics

started FTC/TDF and 156 continued 3TC/ABC). One subject randomized to FTC/TDF was excluded from the ITT analysis set for a major protocol violation (documented prior resistance to study drug). Demographic and baseline disease characteristics are summarized in Table 1. Table 1. Baseline Demographics and Characteristics Characteristic FTC/TDF + PI/r (N = 155) 3TC/ABC + PI/r (N = 156) Total (N = 311) Age, median (range), years 46 (22, 66) 46 (22, 75) 46 (22, 75) Male sex, No. (%) 129 (83) 134 (86) 263 (85) Race, No. (%) White 96 (62) 106 (68) 202 (65) Black 43 (28) 44 (28) 87 (28) Asian 4 (3) 3 (2) 7 (2) Other 12 (8) 3 (2) 15 (5) Ethnicity Hispanic/Latino 38 (25) 36 (23) 74 (24) Non-Hispanic/Latino 117 (76) 120 (77) 237 (76) HIV-1 RNA c/mL, No. (%) <50 139 (90) 145 (93) 284 (91) 50 to < 200 13 (8) 10 (6) 23 (8) ≥200 3 (2) 1 (1) 4 (1) Time since first ARV therapy, median (IQR), years 4 (2.5, 6.9) 3.7 (2.5, 6.7) 3.8 (2.5, 6.7) CD4 cell count, median (IQR), cells/mm3 532 (354, 725) 532 (382, 728) 532 (363, 725) Comorbidities, No. (%) Hyperlipidemia 81 (52) 96 (62) 177 (57) Hypertension 51 (33) 51 (33) 102 (33) Diabetes 15 (10) 17 (11) 32 (10) Lipid modifying agent, No. (%) 67 (43) 80 (51) 147 (47) PI stratification Lopinavir/ritonavir 48 (31) 53 (34) 101 (32) Non-lopinavir/ritonavir 107 (51) 103 (49) 210 (68) Atazanavir/ritonavir 62 (40) 60 (38) 122 (78) Fosamprenavir/ritonavir 34 (22) 31 (20) 65 (40) Darunavir/ritonavir 9 (6) 11 (7) 20 (13) Other PI 2 (1) 1 (1) 3 (2) eGFR Cockcroft-Gault, mL/min (IQR) 95 (79–110) 96 (77–113) … Abbreviations: ARV, antiretroviral; eGFR, epidermal growth factor receptor; FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; HIV-1, human immunodeficiency virus; IQR, interquartile range; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir.

1 (1) 3 (2) eGFR Cockcroft-Gault, mL/min (IQR) 95 (79–110) 96 (77–113) … Abbreviations: ARV, antiretroviral; eGFR, epidermal growth factor receptor; FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; HIV-1, human immunodeficiency virus; IQR, interquartile range; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir. Efficacy Results At week 48, TLOVR responses were 133 of 155 (86.4%) for the FTC/TDF arm compared to 130 of 156 (83.3%) with continued 3TC/ABC, representing a treatment difference of 3.0% (95% CI, −5.1% to 11.2%), establishing noninferiorty. Additionally, fewer people had virologic failure in the FTC/TDF arm vs 3TC/ABC, 3/155 (1.9%) vs 11/156 (7.1%); P = .034 through week 48 (Figure 1). All 3 subjects who experienced virologic failure in the FTC/TDF arm had low-level viremia (range, 209–452 copies/mL); low adherence was not reported in these subjects with low-level viremia. Two were receiving atazanavir/ritonavir and 1 boosted fosamprenavir. Of the 11 subjects with virologic failure in the 3TC/ABC arm, 3 discontinued study drug early, and 8 subjects experienced viremia (range, 272–6430 copies/mL) at the week 48 visit. Of these 11 subjects, 5 were receiving atazanavir/ritonavir, 4 lopinavir/ritonavir, 1 fosamprenavir/ritonavir, and 1 darunavir/ritonavir. No specific boosted PI regimen was associated with virologic failure. Figure 1. Virologic response and virologic failure by Kaplan-Meier through week 48.

es/mL) at the week 48 visit. Of these 11 subjects, 5 were receiving atazanavir/ritonavir, 4 lopinavir/ritonavir, 1 fosamprenavir/ritonavir, and 1 darunavir/ritonavir. No specific boosted PI regimen was associated with virologic failure. Figure 1. Virologic response and virologic failure by Kaplan-Meier through week 48. Abbreviations: CI, confidence interval; FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; HIV-1, human immunodeficiency virus type 1; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir; TLOVR, time to loss of virologic response. Four virologic failure subjects had HIV-1 RNA values above 1000 copies/mL and had genotypic and phenotypic analyses: 1 subject in the FTC/TDF arm and 3 subjects in the 3TC/ABC arm. No genotypic resistance to study drugs was observed in any subject in either arm through week 48. Note, of the 4 subjects who were suppressed at screening but above the HIV-1 RNA value of 200 copies/mL at baseline, 2 were virologic successes due to post-baseline ongoing virologic suppression, 1 was a virologic failure due to detectable but low-level viremia at week 48 while on FTC/TDF, and 1 subject was excluded from the ITT analysis set due to a major protocol violation. Changes in CD4 count at week 48 were similar between treatment arms with median (IQR) changes of 8 (−49, 80) and 39 (−41, 125) cells/mm3 for the FTC/TDF and the 3TC/ABC arms, respectively (P = .10).

Four virologic failure subjects had HIV-1 RNA values above 1000 copies/mL and had genotypic and phenotypic analyses: 1 subject in the FTC/TDF arm and 3 subjects in the 3TC/ABC arm. No genotypic resistance to study drugs was observed in any subject in either arm through week 48. Note, of the 4 subjects who were suppressed at screening but above the HIV-1 RNA value of 200 copies/mL at baseline, 2 were virologic successes due to post-baseline ongoing virologic suppression, 1 was a virologic failure due to detectable but low-level viremia at week 48 while on FTC/TDF, and 1 subject was excluded from the ITT analysis set due to a major protocol violation. Changes in CD4 count at week 48 were similar between treatment arms with median (IQR) changes of 8 (−49, 80) and 39 (−41, 125) cells/mm3 for the FTC/TDF and the 3TC/ABC arms, respectively (P = .10). Subjects who switched to FTC/TDF from 3TC/ABC showed reductions from baseline at week 48 in fasting TC (median change of −21 mg/dL vs −3 mg/dL with 3TC/ABC, P < .001), and LDL (−7 mg/dL vs 2 mg/dL with 3TC/ABC; P = .007). There were no differences in lipid lowering agent modification between arms during the study. No differences in HDL (P = .26), TG (P = .074) or HDL/TC ratio (P = .17) were observed (Supplement 1).

g TC (median change of −21 mg/dL vs −3 mg/dL with 3TC/ABC, P < .001), and LDL (−7 mg/dL vs 2 mg/dL with 3TC/ABC; P = .007). There were no differences in lipid lowering agent modification between arms during the study. No differences in HDL (P = .26), TG (P = .074) or HDL/TC ratio (P = .17) were observed (Supplement 1). At baseline, there was no difference in the distribution across National Cholesterol Education Program (NCEP) categories between the 2 treatment groups; NCEP sets cholesterol guidelines in the United States [16]. At week 48, a higher percentage of subjects who switched to FTC/TDF were in the desirable NCEP categories for TC and TG compared to those who remained on 3TC/ABC (TC: 62% vs 45% < 200 mg/dL, P = .005; TG: 60% vs 41% < 150 mg/dL, P = .003) (Figure 2). Figure 2. Fasting total cholesterol and triglycerides by National Cholesterol Education Program classification. Abbreviations: FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir.

At baseline, there was no difference in the distribution across National Cholesterol Education Program (NCEP) categories between the 2 treatment groups; NCEP sets cholesterol guidelines in the United States [16]. At week 48, a higher percentage of subjects who switched to FTC/TDF were in the desirable NCEP categories for TC and TG compared to those who remained on 3TC/ABC (TC: 62% vs 45% < 200 mg/dL, P = .005; TG: 60% vs 41% < 150 mg/dL, P = .003) (Figure 2). Figure 2. Fasting total cholesterol and triglycerides by National Cholesterol Education Program classification. Abbreviations: FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir. Switching to FTC/TDF resulted in improvements in the predicted risk for CHD outcomes as measured by Framingham Risk Scores. Mean (SD) change from baseline in risk by the TC formula was −1.0 (4.32) for the FTC/TDF arm at week 12 (P = .008); this reduction was also maintained through week 48 with a mean (SD) change from baseline of −1.2 (4.39) and P = .006. When the LDL formula was used, mean (SD) change from baseline in Framingham risk was −0.9 (3.07) for the FTC/TDF arm at week 12 (P < .001) and was −0.5 (3.93) at week 48 (P = .21). The mean change for all calculated Framingham Scores in the 3TC/ABC group fluctuated about the baseline level with no statistically significant changes from baseline observed. The difference between groups for the predicted risk of CHD (regardless of method of calculation) only achieved statistical significance at week 24 (P < .05). The FTC/TDF group further demonstrated a shift from higher risk Framingham categories to lower risk categories (Figure 3). Figure 3. Categorical shifts by Framingham 10-year risk scores from baseline to week 48.

sk of CHD (regardless of method of calculation) only achieved statistical significance at week 24 (P < .05). The FTC/TDF group further demonstrated a shift from higher risk Framingham categories to lower risk categories (Figure 3). Figure 3. Categorical shifts by Framingham 10-year risk scores from baseline to week 48. Abbreviations: CHD, coronary heart disease; FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; PI, protease inhibitor; 3TC/ABC, lamivudine/abacavir. Adverse Events The safety and tolerability for both treatment arms in SWIFT were consistent with the known safety profiles of FTC/TDF and 3TC/ABC (Table 2). Similar percentages of subjects in each arm reported any serious adverse event (SAE), any adverse event (AE), or any Grade 3 or 4 treatment-emergent AE. Three subjects died during the study: 1 subject in the FTC/TDF group (suicide) and 2 subjects in the 3TC/ABC group (homicide, lymphoma). None of the deaths or SAEs was considered by the investigator to be related to study (Table 3). There was one pregnancy in the 3TC/ABC arm with a spontaneous abortion, which was considered unrelated to the study drug. Table 2. Summary of Adverse Events (Treated Analysis Set)

s in the 3TC/ABC group (homicide, lymphoma). None of the deaths or SAEs was considered by the investigator to be related to study (Table 3). There was one pregnancy in the 3TC/ABC arm with a spontaneous abortion, which was considered unrelated to the study drug. Table 2. Summary of Adverse Events (Treated Analysis Set) Adverse Event Category, No. (%)a FTC/TDF + PI/r (N = 155) 3TC/ABC + PI/r (N = 156) Total (N = 311) Adverse event 112 (72.3%) 120 (76.9%) 232 (74.6%) Grade 3 or 4 adverse event 13 (8.4%) 16 (10.3%) 29 (9.3%) Adverse event related to study drug 16 (10.3%) 6 (3.8%) 22 (7.1%) Grade 3 or 4 adverse event related to study drug 1 (0.6%) 0 1 (0.3%) Serious adverse event 12 (7.7%) 11 (7.1%) 23 (7.4%) Serious adverse event related to study drug 0 0 0 Adverse event leading to study drug discontinuation 7 (4.5%) 3 (1.9%) 10 (3.2%) Death during study 1 (0.6%) 2 (1.3%) 3 (1.0%) Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; PI,protease inhibitor; 3TC/ABC,lamivudine/abacavir. Table 3. Disposition of Subjects

Adverse Event Category, No. (%)a FTC/TDF + PI/r (N = 155) 3TC/ABC + PI/r (N = 156) Total (N = 311) Adverse event 112 (72.3%) 120 (76.9%) 232 (74.6%) Grade 3 or 4 adverse event 13 (8.4%) 16 (10.3%) 29 (9.3%) Adverse event related to study drug 16 (10.3%) 6 (3.8%) 22 (7.1%) Grade 3 or 4 adverse event related to study drug 1 (0.6%) 0 1 (0.3%) Serious adverse event 12 (7.7%) 11 (7.1%) 23 (7.4%) Serious adverse event related to study drug 0 0 0 Adverse event leading to study drug discontinuation 7 (4.5%) 3 (1.9%) 10 (3.2%) Death during study 1 (0.6%) 2 (1.3%) 3 (1.0%) Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; PI,protease inhibitor; 3TC/ABC,lamivudine/abacavir. Table 3. Disposition of Subjects Subject Dispositiona FTC/TDF + PI/r 3TC/ABC + PI/r Total Subjects randomized 156 156 312 Subjects randomized but not treated 1 0 1 Subjects treated 155 156 311 Completed 48 weeks of studyb 138 (89.0) 139 (89.1) 277 (89.1) Discontinued study drug prematurely 17 (11.0) 17 (10.9) 34 (10.9) Primary reason for premature discontinuation of study Adverse event 7 (4.5) 3 (1.9) 10 (3.2) Death 0 0 0 Pregnancy 0 1 (0.6) 1 (0.3) Lack of efficacy 0 1 (0.6) 1 (0.3) Investigator's discretion 0 3 (1.9) 3 (1.0) Withdrew consent 5 (3.2) 4 (2.6) 9 (2.9) Lost to follow-up 4 (2.6) 5 (3.2) 9 (2.9) Subject noncompliance 0 0 0 Protocol violation 1 (0.6) 0 1 (0.3) Study discontinued by sponsor 0 0 0 Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; PI,protease inhibitor; 3TC/ABC,lamivudine/abacavir. a All percentages are based on the No. of subjects in the treated analysis set.

Subject Dispositiona FTC/TDF + PI/r 3TC/ABC + PI/r Total Subjects randomized 156 156 312 Subjects randomized but not treated 1 0 1 Subjects treated 155 156 311 Completed 48 weeks of studyb 138 (89.0) 139 (89.1) 277 (89.1) Discontinued study drug prematurely 17 (11.0) 17 (10.9) 34 (10.9) Primary reason for premature discontinuation of study Adverse event 7 (4.5) 3 (1.9) 10 (3.2) Death 0 0 0 Pregnancy 0 1 (0.6) 1 (0.3) Lack of efficacy 0 1 (0.6) 1 (0.3) Investigator's discretion 0 3 (1.9) 3 (1.0) Withdrew consent 5 (3.2) 4 (2.6) 9 (2.9) Lost to follow-up 4 (2.6) 5 (3.2) 9 (2.9) Subject noncompliance 0 0 0 Protocol violation 1 (0.6) 0 1 (0.3) Study discontinued by sponsor 0 0 0 Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; PI,protease inhibitor; 3TC/ABC,lamivudine/abacavir. a All percentages are based on the No. of subjects in the treated analysis set. b Subjects completed 48 weeks of the study if the subject completed the protocol-planned duration of the study based on the study completion form.

Subject Dispositiona FTC/TDF + PI/r 3TC/ABC + PI/r Total Subjects randomized 156 156 312 Subjects randomized but not treated 1 0 1 Subjects treated 155 156 311 Completed 48 weeks of studyb 138 (89.0) 139 (89.1) 277 (89.1) Discontinued study drug prematurely 17 (11.0) 17 (10.9) 34 (10.9) Primary reason for premature discontinuation of study Adverse event 7 (4.5) 3 (1.9) 10 (3.2) Death 0 0 0 Pregnancy 0 1 (0.6) 1 (0.3) Lack of efficacy 0 1 (0.6) 1 (0.3) Investigator's discretion 0 3 (1.9) 3 (1.0) Withdrew consent 5 (3.2) 4 (2.6) 9 (2.9) Lost to follow-up 4 (2.6) 5 (3.2) 9 (2.9) Subject noncompliance 0 0 0 Protocol violation 1 (0.6) 0 1 (0.3) Study discontinued by sponsor 0 0 0 Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; PI,protease inhibitor; 3TC/ABC,lamivudine/abacavir. a All percentages are based on the No. of subjects in the treated analysis set. b Subjects completed 48 weeks of the study if the subject completed the protocol-planned duration of the study based on the study completion form. The percentage of subjects who discontinued study drug due to an AE was higher in the FTC/TDF group [4.5% (n = 7/155)], compared to the 3TC/ABC [1.9% (n = 3/156)]. Rash, which was reported in 1.3% (2 subjects) in the FTC/TDF arm, was the only AE reported in more than 1 subject that resulted in study drug discontinuation. A higher percentage of treatment-emergent AEs considered related to study drug by the investigator were reported in the FTC/TDF than in the 3TC/ABC group, 10.3% (n = 16) vs 3.8% (n = 6). Adverse events considered related to the study drug in more than 1 subject included nausea, headache, and dizziness (1.9%, 3 subjects each); diarrhea, flatulence, malaise, and rash (1.3%, 2 subjects each) in the FTC/TDF group; and diarrhea (1.3%, 2 subjects) in the 3TC/ABC group.

group, 10.3% (n = 16) vs 3.8% (n = 6). Adverse events considered related to the study drug in more than 1 subject included nausea, headache, and dizziness (1.9%, 3 subjects each); diarrhea, flatulence, malaise, and rash (1.3%, 2 subjects each) in the FTC/TDF group; and diarrhea (1.3%, 2 subjects) in the 3TC/ABC group. There were no differences observed in renal adverse events between arms (FTC/TDF 4.5% [n = 7]; 3TC/ABC 5.1% [n = 8]). Three subjects in the FTC/TDF arm had renal AEs reported as related to study drug by the investigator: renal impairment (baseline serum creatinine [SCr] of 1.0 mg/dL which subsequently increased to 1.3 mg/dL then decreased to 1.1 mg/dL), abnormal urine odor, and increased SCr with decreased eGFR (grade 1 decrease at discontinuation). Modest decreases from baseline through week 48 in creatinine clearance by the CG method (GFRCG) using ideal body weight occurred within both treatment arms, FTC/TDF (GFRCG−8.3 mL/minutes, P < .001) and 3TC/ABC (GFRCG −4.5 mL/minutes, P = .002). When compared across arms, a statistically significant difference was observed between the groups (P = .012). MDRD GFR estimates gave similar results (Supplement 2).

FRCG) using ideal body weight occurred within both treatment arms, FTC/TDF (GFRCG−8.3 mL/minutes, P < .001) and 3TC/ABC (GFRCG −4.5 mL/minutes, P = .002). When compared across arms, a statistically significant difference was observed between the groups (P = .012). MDRD GFR estimates gave similar results (Supplement 2). Treatment emergent laboratory abnormalities were comparable between the groups. Most laboratory abnormalities were grade 1 or 2, and most common was elevated bilirubin, primarily in subjects on ATV + RTV. There was no grade 2 or higher changes in SCr throughout the study. Grade 1 SCr laboratory changes occurred in 3.2% of subjects on FTC/TDF and 1.9% on 3TC/ABC. No clinically relevant changes in serum phosphorus and in hypophosphatemia were observed. There was no difference in development of proteinuria between the 2 arms when analyzed by change in grade from baseline (Table 4). No patients had confirmed normoglycemic glucosuria in either arm. Table 4. Change From Baseline in Urine Protein by Grade Urine Protein Change in Gradea −2 −1 0 +1 +2 +3 FTC/TDF (n = 148) 2 11 107 25 3 … 3TC/ABC (n = 151) 1 15 114 21 0 … Total (N = 299) 3 26 221 46 3 … Cochran-Mantel-Haenssel statistics (based on table scores). Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; 3TC/ABC,lamivudine/abacavir. a Nonzero correlation value 1.5674, P = .2106; row mean scores diff 1.5674, P = .2106.

Urine Protein Change in Gradea −2 −1 0 +1 +2 +3 FTC/TDF (n = 148) 2 11 107 25 3 … 3TC/ABC (n = 151) 1 15 114 21 0 … Total (N = 299) 3 26 221 46 3 … Cochran-Mantel-Haenssel statistics (based on table scores). Abbreviations: FTC/TDF,emtricitabine/tenofovir disoproxil fumarate; 3TC/ABC,lamivudine/abacavir. a Nonzero correlation value 1.5674, P = .2106; row mean scores diff 1.5674, P = .2106. Given previous reports of increased risks for cardiovascular events, including myocardial infarction, associated with ARV regimens containing ABC, we explored changes in commonly used surrogate cardiovascular biomarkers in a subset of 159 of 312 (51%) patients, 81 randomized to FTC/TDF and 78 to 3TC/ABC. No differences at week 48 compared to baseline were observed between treatment arms for hsCRP, IL-10, IL-6, and TNF-α (Table 5), although there was a trend for differences in fibrinogen (median change, FTC/TDF −10 mg/dL, 3TC/ABC −1 mg/dL, P = .062) at week 48. Table 5. Cardiovascular Biomarkers Change From Baseline at Week 48a Cardiovascular Biomarkers, Median (Q1, Q3) FTC/TDF + PI/r 3TC/ABC + PI/r C-reactive protein (mg/dL) N = 69 −0.013 (−0.123, 0.054) N = 57 0.006 (−0.078, 0.113) P = .19 Fibrinogen (mg/dL) N = 64 −10 (−50, 31) N = 56 −1 (−19, 54) P = .062 IL-10-INF (pg/mL) N = 68 0.0 (0.0, 0.0) N = 56 0.0 (−0.4, 0.0) P = .22 IL-6-INF (pg/mL) N = 68 0.0 (−0.4, 0.2) N = 56 0.0 (−1.2, 0.2) P = .58 TNF-α-INF (pg/mL) N = 68 0.0 (0.0, 0.0) N = 56 0.0 (0.0, 0.0) P = .69 P-values for comparison between treatment groups are from Wilcoxon rank-sum test.

= 56 −1 (−19, 54) P = .062 IL-10-INF (pg/mL) N = 68 0.0 (0.0, 0.0) N = 56 0.0 (−0.4, 0.0) P = .22 IL-6-INF (pg/mL) N = 68 0.0 (−0.4, 0.2) N = 56 0.0 (−1.2, 0.2) P = .58 TNF-α-INF (pg/mL) N = 68 0.0 (0.0, 0.0) N = 56 0.0 (0.0, 0.0) P = .69 P-values for comparison between treatment groups are from Wilcoxon rank-sum test. Abbreviations: 3TC/ABC, lamivudine/abacavir; FTC/TDF, emtricitabine/tenofovir disoproxil fumarate; IL, interleukin; INF, interferon; PI, protease inhibitor; TNF, tumor necrosis factor. a Missing = excluded analysis. DISCUSSION In this large, prospective, randomized trial, the first study specifically designed to evaluate the efficacy and safety of switching from 3TC/ABC to FTC/TDF in HIV-1-infected subjects suppressed on a PI + RTV containing regimen, we have demonstrated that FTC/TDF is noninferior to remaining on 3TC/ABC in maintaining treatment response by TLOVR with fewer VFs, as well as a lower risk of emergent resistance through 48 weeks. Efficacy and virologic failure rates in subjects on FTC/TDF compared to 3TC/ABC arm were comparable to results seen with this FDC seen in the BICOMBO and ASSERT trials [2, 19].

emaining on 3TC/ABC in maintaining treatment response by TLOVR with fewer VFs, as well as a lower risk of emergent resistance through 48 weeks. Efficacy and virologic failure rates in subjects on FTC/TDF compared to 3TC/ABC arm were comparable to results seen with this FDC seen in the BICOMBO and ASSERT trials [2, 19]. We did observe slightly higher rates of discontinuation due to AEs and mild AEs considered study drug-related in the FTC/TDF vs 3TC/ABC arms. This finding is not unexpected as previous studies demonstrate an increase in certain adverse events when stable subjects are switched to a new therapy. Modest declines in eGFR occurred in both arms with the degree of decline significantly greater in FTC/TDF-treated subjects; however, values remained in the normal range. Long-term studies have shown that the use of TDF may be associated with initial declines in GFR within the first few months of starting TDF, which then stabilize [20–22]. Importantly, there were no differences between arms in emergent proteinuria and or normoglycemic glycosuria (Table 4).

remained in the normal range. Long-term studies have shown that the use of TDF may be associated with initial declines in GFR within the first few months of starting TDF, which then stabilize [20–22]. Importantly, there were no differences between arms in emergent proteinuria and or normoglycemic glycosuria (Table 4). Comorbidities (FTC/TDF vs 3TC/ABC) were common in our study population, including diabetes (10% vs11%), hyperlipidemia (52% vs 62%), and hypertension (33% vs 33%). Additionally, 3.2% in both arms had a history of MI. As with previous studies, we demonstrated lipid benefits when switching to FTC/TDF [3–6]. Significant declines in LDL and TC were observed by week 12 in the FTC/TDF arm and significant reductions in TC, LDL, and TG were seen at week 48. By NCEP category criteria [16], higher percentages showed improvements in TC and LDL, as well as improvement (shift from a higher risk to a lower risk category) in the predicted risk for CHD outcomes with FTC/TDF as seen in other comparative studies [2, 5]. In an ad hoc analysis, we found that the predicted Framingham 10-yr Risk Score was more favorable when switching FTC/TDF; particularly, for those with comorbidities, whites, and regimens with a PI other than LPV/r. Such an improvement in Framingham scores is perhaps one of the most novel benefits of switching from an 3TC/ABC to FTC/TDF-containing regimen. It is however, worthwhile to note that the small number of subjects on LPV/r and other confounding factors at baseline may make this a weaker correlation and may limit these results from being generalized. Finally, the changes from baseline in commonly used surrogate cardiovascular biomarkers (hsCRP, IL-10, IL-6, TNF-α, and fibrinogen) between the FTC/TDF and 3TC/ABC arm in a subset of 159 subjects were not significant except a trend toward significance with fibrinogen (P = .062) (Table 5), perhaps with a larger sample size it may have achieved significance.

commonly used surrogate cardiovascular biomarkers (hsCRP, IL-10, IL-6, TNF-α, and fibrinogen) between the FTC/TDF and 3TC/ABC arm in a subset of 159 subjects were not significant except a trend toward significance with fibrinogen (P = .062) (Table 5), perhaps with a larger sample size it may have achieved significance. The SWIFT study showed that high rates of virologic suppression were well maintained through 48 weeks with fewer virologic failures in subjects who switched to FTC/TDF, and also this regimen is well tolerated. Decreases in creatinine clearance did occur with both treatments and were greater in the FTC/TDF arm. In the FTC/TDF arm, improvements in certain lipid parameters and in other measures including in NCEP categories and Framingham predicted risk for CHD outcomes were noted [15, 16]. In summary, switching patients on a boosted PI regimen to FTC/TDF from 3TC/ABC is associated with important metabolic benefits without loss of virologic control. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

rdjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Disclosures. F. B. received travel support to his institution from Metropolis Medical. R. C. received institutional grant support for participation in the study's clinical trials. E. D. J. received travel support through his institution from Gilead. K. H. received institutional grand support from the Minneapolis Medical Research Foundation and also received travel support from Gilead. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

To the Editor—We agree with Dr Halstead that the findings of our trial of early corticosteroid therapy in Vietnamese children and young adults with dengue add an interesting perspective to the long-running debate on the mechanisms underlying disease pathogenesis [1, 2]. The absence of any adverse effects of prednisolone therapy on the virological safety parameters assessed, together with the lack of a reduction in the frequency of shock or other complications, brings into question the concept of an immunopathogenic storm being causally responsible for the microvascular dysfunction. However, we caution against overinterpreting the clinical trial results as representing definitive evidence against a role for T cells or any other immunological mechanisms. Instead, we believe our results emphasize how little we understand about the disease processes responsible for these complications and hope that these results will trigger more expansive efforts in dengue pathogenesis research. Detailed examination of serial measurements of immunological events in study participants will be published elsewhere (manuscript in preparation), but it is clear that the pharmacological “footprint” of prednisolone therapy in this trial was less than anticipated. Perhaps earlier therapy, or higher treatment doses, might have elicited a greater effect on clinical and laboratory end points, but given the dysglycemia encountered with the 2 mg/kg dosage regimen, the latter option would not be feasible for large-scale intervention in the community.

s trial was less than anticipated. Perhaps earlier therapy, or higher treatment doses, might have elicited a greater effect on clinical and laboratory end points, but given the dysglycemia encountered with the 2 mg/kg dosage regimen, the latter option would not be feasible for large-scale intervention in the community. The potential role of dengue nonstructural protein 1 (NS1) in pathogenesis is an important area to consider, but is not without difficulties. Importantly, there are conflicting data on the relationship between concentrations of secreted NS1, plasma viremia, immune status, and clinical disease severity [3–5]. In particular, plasma levels of dengue virus 2–associated NS1 are often low or undetectable in both human studies and mouse models [4, 6], yet dengue virus 2 is well established as a cause of severe disease. Secondly, kinetic studies reveal that although plasma levels of NS1 are often high in early disease, the protein may also persist for several weeks after infection without causing clinical complications [7], suggesting that if NS1 is important in pathogenesis, other factors must also be operating to influence outcome. We agree that complement activation, possibly exacerbated by NS1, is probably important and is a relatively poorly researched area of dengue pathogenesis.

infection without causing clinical complications [7], suggesting that if NS1 is important in pathogenesis, other factors must also be operating to influence outcome. We agree that complement activation, possibly exacerbated by NS1, is probably important and is a relatively poorly researched area of dengue pathogenesis. Sayce and colleagues describe data from a series of experiments that investigate dexamethasone as a dengue therapeutic in primary monocyte-derived macrophages from dengue-naive donors [8]. They observed a significant but transient decrease in viral load on day 1 with dexamethasone treatment, concomitant with a relative reduction in levels of selected inflammatory cytokines, and they comment that the effectiveness of treatment with steroids may depend critically on time of drug administration. However, interpretation of these results in terms of overall disease pathogenesis is difficult, because the true relevance of isolated cell culture systems to human disease processes is doubtful.

es, and they comment that the effectiveness of treatment with steroids may depend critically on time of drug administration. However, interpretation of these results in terms of overall disease pathogenesis is difficult, because the true relevance of isolated cell culture systems to human disease processes is doubtful. The idea of combined antiviral and immunomodulatory therapy is certainly one avenue that may be worth pursuing, particularly if it is possible to initiate the combined therapy very early in the disease evolution. Efforts directed toward improving dengue rapid diagnostics suitable for use early in the febrile phase, together with research to identify risk factors associated with subsequent progression to severe disease, are important if this strategy is to be considered. In the end, if a simple, safe, and effective therapy or combination of therapies does become available, it is crucial that prompt intervention targeted toward high-risk groups is a realistic possibility. Finally, although the mechanisms underlying microvascular dysfunction in dengue are almost certainly multifactorial, it is hoped that the recent momentum in clinical intervention trials will eventually lead to improved treatment options for patients with dengue as well as insights into pathogenesis [2, 7, 9, 10]. Note Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Listeriosis is a relatively uncommon but serious infection caused by Listeria monocytogenes. This organism is ubiquitous in the environment and can survive at temperatures ranging from −7°C to body temperature [1]. The main route of transmission is believed to be through the consumption of contaminated food (processed meats, unpasteurized milk, soft cheeses, and cantaloupes) [2–7] and vertical transmission from mother to child [8, 9]. However, healthcare-associated transmission has also been reported through patient-to-patient transmission, mineral bathing oil, contaminated resuscitation equipment, and the contaminated hands of medical personnel [10–14]. Most of the healthcare-associated infections are clustered and related to food processing [11–13]. Gastroenteritis, bacteremia, and meningitis are the most common manifestations of listeriosis. Because L. monocytogenes has a strong predilection for elderly and immunocompromised persons [15–18], results in poor fetal outcomes [19–21], exhibits poor response to third-generation cephalosporins, and is associated with a high mortality rate, it has become an increasingly important emerging infectious disease [22]. In the United States, L. monocytogenes is the fourth causative microorganism of bacterial meningitis [23]. Among persons aged >65 years, L. monocytogenes is the third leading pathogen [24, 25]. Most listeriosis cases have been reported from industrialized Western countries. Reports from East Asia and developing countries are scarce [26, 27].

. monocytogenes is the fourth causative microorganism of bacterial meningitis [23]. Among persons aged >65 years, L. monocytogenes is the third leading pathogen [24, 25]. Most listeriosis cases have been reported from industrialized Western countries. Reports from East Asia and developing countries are scarce [26, 27]. Our goal was to retrospectively review all culture-proven cases of listeriosis at Peking Union Medical College Hospital (PUMCH) since 1999 and describe the clinical characteristics and outcomes of the infected patients. METHODS PUMCH is an 1800-bed tertiary care hospital in Beijing, China. Founded in 1921 by the Rockefeller Foundation, PUMCH is the national medical technical support center for the diagnosis and treatment of severe and complicated diseases. In 2002, another hospital in Beijing merged with PUMCH and was renamed the Western campus of PUMCH. The latter housed several departments (general medicine, rheumatology, oncology, and breast surgery), and both campuses shared other departments (hematology, gastroenterology). PUMCH provides medical services to patients from surrounding areas (Beijing, and Hebei province) and to patients being referred from various outside institutions throughout China.

s (general medicine, rheumatology, oncology, and breast surgery), and both campuses shared other departments (hematology, gastroenterology). PUMCH provides medical services to patients from surrounding areas (Beijing, and Hebei province) and to patients being referred from various outside institutions throughout China. We retrospectively identified all patients with L. monocytogenes infections based on a list generated from an electronic database in the clinical microbiology laboratory at PUMCH. All positive culture results for L. monocytogenes diagnosed at PUMCH since 1999 are stored in the database. We included all cases from January 1999 to October 2011. Clinical data from the identified cases were abstracted from the medical records. These data included demographic characteristics, comorbidities, known risk factors (immunosuppressive therapy, dietary history, travel, and exposures), the sites from which the organism was isolated, clinical presentation, laboratory data, type of antimicrobial therapy, duration of hospitalization, and outcomes.

cords. These data included demographic characteristics, comorbidities, known risk factors (immunosuppressive therapy, dietary history, travel, and exposures), the sites from which the organism was isolated, clinical presentation, laboratory data, type of antimicrobial therapy, duration of hospitalization, and outcomes. The diagnosis of listeriosis was based on one of the following: isolation of L. monocytogenes from normally sterile clinical specimens (eg, cerebrospinal fluid [CSF], blood, amniotic fluid, uterine swab); isolation of L. monocytogenes from nonsterile specimens (eg, rectal swab, tracheal swab); and histopathology compatible with listeriosis [22]. Cases were categorized as neonatal, maternal, or nonmaternal infections. All maternal cases were in pregnant women who had L. monocytogenes isolated from cultures of normal sterile body sites or vaginal swab [19]. Healthcare-associated cases were defined as onset of listeriosis symptoms >48 hours after admission for medical conditions other than listeriosis. We used descriptive statistics. Where appropriate, we present point estimates with 95% confidence intervals (CIs). This study was reviewed and approved by the Institutional Review Board at PUMCH. RESULTS We identified 38 patients (cases) of listeriosis diagnosed between 1999 and 2011. The demographic characteristics of these cases are summarized in Table 1. There were 5 neonatal, 8 maternal, and 25 nonmaternal infections with L. monocytogenes. Table 1. Characteristics of 38 Cases of Listeriosis

We used descriptive statistics. Where appropriate, we present point estimates with 95% confidence intervals (CIs). This study was reviewed and approved by the Institutional Review Board at PUMCH. RESULTS We identified 38 patients (cases) of listeriosis diagnosed between 1999 and 2011. The demographic characteristics of these cases are summarized in Table 1. There were 5 neonatal, 8 maternal, and 25 nonmaternal infections with L. monocytogenes. Table 1. Characteristics of 38 Cases of Listeriosis Group Neonatal Maternal Nonmaternal Total 5 (13.2) 8 (21.1) 25 (65.8) Male 4 (80) 0 9 (36) Median age (min, max), y NA 30 (26, 33) 47 (18, 79) Median gestation (min, max), wk 37 (27, 39.9) 29 (18.9, 39.9) NA Underlying disease 23 (92) Autoimmune disease 1 (12.5) 10 (40) Neoplasm 10 (40) Diabetes 3 (12) Ulcerative colitis 2 (8) Polycystic kidney and hepatic disease 1 (4) Iatrogenic factors Chronic use of corticosteroids 10 (40)a Chemotherapy 5 (20) Clinical manifestations Fever 4 (80) 6 (75) 24 (96) Gastrointestinal symptoms 5 (62.5) 12 (48) Neurological symptoms 1 (12.5) 16 (68) Laboratory findings Peripheral WBC, mean ± SD, 109/L 13.3 ± 5.1 17.6 ± 6.2 8.3 ± 5.1 CSF WBC median (min, max), cells/µL 1660 (16, 128 300) 200 (36, 2590) CSF neutrophils, %, median (min, max) 65 (62, 97) 40 (10, 96) CSF mononuclear, %, median (min, max) 35 (3, 38) 60 (4, 90) CSF neutrophils >50% 3/3 (100) 7/15 (46.7) CSF protein median (min, max), g/L 1.8 (0.94, 9.16) 1.77 (0.65, 8.45) Mortality 1 (20) 0 9 (36) Data are presented as No. (%) unless otherwise specified.

) CSF neutrophils, %, median (min, max) 65 (62, 97) 40 (10, 96) CSF mononuclear, %, median (min, max) 35 (3, 38) 60 (4, 90) CSF neutrophils >50% 3/3 (100) 7/15 (46.7) CSF protein median (min, max), g/L 1.8 (0.94, 9.16) 1.77 (0.65, 8.45) Mortality 1 (20) 0 9 (36) Data are presented as No. (%) unless otherwise specified. Abbreviations: CSF, cerebrospinal fluid; max, maximum; min, minimum; NA, not applicable; SD, standard deviation; WBC, white blood cell. a On steroid of prednisone equivalent 30–40 mg/d in 4 of 10 cases, >50 mg/d in 6 of 10 cases; of those, 4 patients were on concurrent immunosuppressive therapies. Neonatal Listeriosis Of 26 221 deliveries during this time period, there were 5 cases of neonatal listeriosis identified. Four of 5 cases of neonatal listeriosis were male. All 5 neonatal listeriosis cases were born to symptomatic mothers. All had positive cultures and presented with fetal distress (n = 5), sepsis (n = 4), meningitis (n = 4), Apgar score <5 (n = 3), low birth weight (n = 2), and meconium aspiration (n = 1), suggestive of intrauterine infection. The clinical characteristics and outcomes of these 5 cases are summarized in Table 2. Table 2. Characteristics of 5 Neonatal Cases of Listeriosis

distress (n = 5), sepsis (n = 4), meningitis (n = 4), Apgar score <5 (n = 3), low birth weight (n = 2), and meconium aspiration (n = 1), suggestive of intrauterine infection. The clinical characteristics and outcomes of these 5 cases are summarized in Table 2. Table 2. Characteristics of 5 Neonatal Cases of Listeriosis No. Sex Presentation Maternal Illness Gestation (wk) Culture Sites Initial Antibiotic Switch Antibiotic Intubation Complication Outcome 16 F Fetal distress, Apgar 9, Tmax 37.5°C, low birth weight, WBC 18.3 × 109/L, SpO2 76% on ambient air High fever; positive cultures 37.1 Blood, rectal swab, laryngeal swab Meropenem + PNG No No Sepsis, meningitis, aspiration pneumonia, bilateral IVH Survived 24 M Fetal distress, C-section, SOB, Apgar 5, afebrile, WBC 15 × 109/L, increased ICP, turbid CSF, CSF WBC 1660/µL Diarrhea, fevers; positive cultures 31 Blood, rectal swab Meropenem PNG Yes Sepsis, meningitis, pneumonia, low birth weight, ICH Survived 25 M Fever (38°C), Apgar 1, SOB, cyanosis, rash, hypotension, WBC 16 × 109/L, bloody and turbid CSF, CSF WBC 128 300/µL Headache, fevers, severe abdominal pain. No microbiologic data. 32.7 Blood, laryngeal swab, tracheal tube tip Cefmetazole Meropenem + PNG Yes Sepsis, meningitis, pneumonia NRDS, Bilateral IVH, SAH Survived 36 M Fetal distress, Apgar 9, C-section, meconium aspiration, low fever (37.9°C), WBC 5.39 × 109/L, CSF WBC 0 High fevers; positive cultures 39.9 Blood, laryngeal swab, tracheal tube tip Meropenem Ampicillin/sulbactam + cefepime Yes Sepsis, IVH Survived 28 M Extremely low birth weight (720 g), Apgar 5, WBC 11.4 × 109/L SLE, prednisone 10 mg/d, abdominal pain, no cultures placental pathology: acute chorioamnionitis 27 Rectal swab Ampicillin/sulbactam Meropenem Yes Intrauterine infection, pulmonary hemorrhage (NRDS), neonatal asphyxia, premature birth, extremely low birth weight, sclerema neonatorum Deceased day 2 Abbreviations: C-section, cesarean section; CSF WBC, white blood cell count in cerebrospinal fluid; ICH, intracranial hemorrhage; ICP, intracranial pressure; IVH, intraventricular hemorrhage; NRDS, neonatal respiratory distress syndrome; PNG, penicillin; SAH, subarachnoid hemorrhage; SLE, systemic lupus erythematosus; SpO2, oxygen saturation from pulse oximetry; SOB, shortness of breath; Tmax, maximal temperature; WBC, peripheral white blood cell count.

tracranial pressure; IVH, intraventricular hemorrhage; NRDS, neonatal respiratory distress syndrome; PNG, penicillin; SAH, subarachnoid hemorrhage; SLE, systemic lupus erythematosus; SpO2, oxygen saturation from pulse oximetry; SOB, shortness of breath; Tmax, maximal temperature; WBC, peripheral white blood cell count. Maternal Listeriosis There were 8 maternal cases of listeriosis identified. Six cases were confirmed by culture. Two other cases were suspected based on symptoms and positive cultures in their infants at the time of delivery. The median age was 30 years (range, 26–33 years). The median gestation was 29 weeks (range, 18.9–39.9 weeks). Maternal cases presented with a sudden onset (<1 week from presentation) of symptoms (n = 7), which included high fevers with a maximal temperature >39°C (n = 6), gastrointestinal symptoms (diarrhea, abdominal pain; n = 5), and various obstetrical manifestations (decreased fetal movement in 2 cases, intrauterine fetal death, vaginal bleeding, and acute pyelonephritis) (Table 3). Two maternal cases had L. monocytogenes cultured from blood; all 3 cases whose L. monocytogenes was detected on uterine swabs had histopathologic evidence of either acute chorioamnionitis or intrauterine fetal infection. In one case, L. monocytogenes was cultured from the vaginal swab, placental histopathology demonstrated chorioamnionitis, and the infant had culture-proven listeriosis. The other 2 cases had symptoms consistent with listeriosis, positive listeria cultures in the newborns, and pathologic findings of acute chorioamnionitis (Table 2). None of the mothers had central nervous system (CNS) involvement and all recovered fully after delivery. Table 3. Characteristics of 8 Maternal Cases of Listeriosis

2 cases had symptoms consistent with listeriosis, positive listeria cultures in the newborns, and pathologic findings of acute chorioamnionitis (Table 2). None of the mothers had central nervous system (CNS) involvement and all recovered fully after delivery. Table 3. Characteristics of 8 Maternal Cases of Listeriosis No. Age (y) Gestation (wks) Symptom Duration Presentation Culture Sites Initial Antibiotic Switch Antibiotic Maternal Complications Maternal Outcome Fetal Outcome 19 32 18.9 1 wk Fever (Tmax 39.5°C), chills, headache, dysuria, WBC 12 × 109/L Blood Ceftriaxone→ cefmetazole + clarithromycin Amoxicillin/ clavulanate Pyelonephritis Recovered C-section 5 mo later, healthy baby 34 33 23 1 d Fever (Tmax 39.6°C), diarrhea, WBC 24 × 109/L Blood Ceftriaxone + metronidazole None None Recovered Fetal death; placental pathology:acute chorioamnionitis 6 30 26.7 2 d Fever (Tmax 39.4°C), abdominal pain, vaginal bleeding, WBC 28 × 109/L Uterine swab Cefuroxime + metronidazole No change Late abortion Recovered Fetal death; placental pathology:chorioamnionitis. 23 31 31 3 d Ingestion of roasted lamb and rabbit in a Mongolian village 5 d before, decreased fetal movement 3 d, fever (Tmax 39°C) 1 d, diarrhea, abdominal pain, WBC 19 × 109/L Uterine swab NA NA None Recovered, C-section (severe meconium stained amniotic fluid) Infant listeriosis (case no. 24,Table 2); placental pathology: acute chorioamnionitis 15 28 37.1 1 d Fever (Tmax 39°C), WBC 15 × 109/L Uterine swab Ceftriaxone + metronidazole Ampicillin + metronidazole None Recovered. C-section (meconium stained amniotic fluid) Infant listeriosis (case no. 16,Table 2); placental pathology:chorioamnionitis 35 29 39.9 4 h Fever (Tmax 38.8°C) for 4 hours, decreased fetal movement for 1 d, WBC 9.29 × 109/L Vaginal swab Ceftriaxone + metronidazole No change Intrauterine fetal hypoxia Recovered, C-section (severe meconium stained amniotic fluid) Infant listeriosis (case no. 36,Table 2); Placental pathology:chorioamnionitis 37 26 32.7 2 wk Headache, fever (Tmax 39.8°C) 2 wk, decreased fetal movement 1 wk, severe abdominal pain 1 d NA NA NA Infant listeriosis Recovered, postpartum uterine curettage for retention of fetal membranes Infant listeriosis (case no. 25,Table 2); placental pathology: NA 38 30 27 1 d Sudden onset of lower abdominal pain NA NA NA Premature labor Recovered Late abortion, fetal death (case no.

1 wk, severe abdominal pain 1 d NA NA NA Infant listeriosis Recovered, postpartum uterine curettage for retention of fetal membranes Infant listeriosis (case no. 25,Table 2); placental pathology: NA 38 30 27 1 d Sudden onset of lower abdominal pain NA NA NA Premature labor Recovered Late abortion, fetal death (case no. 28, Table 2); placental pathology: acute chorioamnionitis Abbreviations: C-section, cesarean section; NA, not available; Tmax, maximal temperature; WBC, peripheral white blood cell count. Obstetrical outcomes included 5 cases of listeriosis in the infants postpartum. All 5 cases were the result of listeria infections during the third trimester of gestation, and a single one of these cases was fatal. There were 2 induced/late abortions as a result of listeria infections during the second trimester of gestation, and a normal pregnancy outcome for a single second-trimester infection.

All 5 cases were the result of listeria infections during the third trimester of gestation, and a single one of these cases was fatal. There were 2 induced/late abortions as a result of listeria infections during the second trimester of gestation, and a normal pregnancy outcome for a single second-trimester infection. Nonmaternal Listeriosis Among the 25 nonmaternal cases, the median age was 47 years (range, 18–79 years), and 72% (95% CI, 52.5%–85.7%) were female. Twenty-three (92%; 95% CI, 75.03%–97.78%) infections occurred in patients with significant comorbidities (Table 4). Ten (40%) patients had concurrent neoplasms: 2 cases each of leukemia, multiple myeloma, liver cancer, and rectal cancer, and 1 case each of breast cancer and abdominal malignant metastases from an unknown primary. Ten nonmaternal infections occurred in patients with autoimmune diseases: 6 cases in patients with systemic lupus erythematosus (SLE), 2 cases in patients with dermatomyositis, 1 case in a patient with Still's disease, and 1 in a patient with mixed connective tissue disease. Other comorbidities included diabetes mellitus and polycystic kidney disease with chronic renal failure. Ten (40%) nonmaternal adult listeriosis cases were receiving chronic corticosteroids at the onset of symptoms, and 6 (24%) had received chemotherapy within 2 months before the onset of listeriosis. Table 4. Characteristics of 25 Cases of Nonmaternal Listeriosis

and polycystic kidney disease with chronic renal failure. Ten (40%) nonmaternal adult listeriosis cases were receiving chronic corticosteroids at the onset of symptoms, and 6 (24%) had received chemotherapy within 2 months before the onset of listeriosis. Table 4. Characteristics of 25 Cases of Nonmaternal Listeriosis No. Sex Age (y) Comorbidities Predisposing Factor Healthcare-Associated Duration Presentation Culture Sites Complications Outcome 31 F 24 Acute lymphoblastic leukemia (L2) Chemotherapy, neutropenia Yes 1 d Abdominal pain × 2 wk before admission, sudden fever (Tmax 40.1°C) hospital day 12 Blood, CSF Sepsis (Listeria, E. coli), cerebral hemorrhage, coma Death hospital day 25 3 F 43 Metastatic liver disease; unknown primary Neoplasm No 3 d Intermittent abdominal pain for 1 mo, fever (Tmax 39.3°C), and headache 3 d Blood, CSF Meningitis, coma Death hospital day 6 1 M 53 Multiple myeloma Chemotherapy, chronic use of melphalan, thalidomide No 2 d Fever (Tmax 40.7°C), headache, loss of consciousness Blood, CSF Septic shock, meningitis, ARF, GI perforation Death hospital day 8 4 M 20 None None No 3 wk Sudden onset fever (Tmax 40°C), headache, worsening mental status (delirium, coma), ventricular enlargement, placement of external CSF shunt, intubated CSF, sputum Meningo-encephalitis, pneumonia, MOF, coma, central diabetes insipidus Death hospital day 12 18 F 47 Dermatomyositis Prednisone 40 mg/d No 3 d Fever, dizziness, and dysphagia, sudden cyanosis and coma while in emergency room Blood, CSF Meningitis, HAP (MRSA, Enterobacter) brain stem stroke, brain death Death hospital day 20 2 F 56 SLE and abdominal malignancy of unclear primary Prednisone 30–40 mg/d, CTX 0.4/wk Yes 3 d Admitted with fatigue, edema, and jaundice. Fever (Tmax 38.5°C) started 3 d after admission.

coma while in emergency room Blood, CSF Meningitis, HAP (MRSA, Enterobacter) brain stem stroke, brain death Death hospital day 20 2 F 56 SLE and abdominal malignancy of unclear primary Prednisone 30–40 mg/d, CTX 0.4/wk Yes 3 d Admitted with fatigue, edema, and jaundice. Fever (Tmax 38.5°C) started 3 d after admission. Blood Pneumonia, bacterial sepsis, MOF Death hospital day 25 14 F 23 SLE Prednisone 50–80 mg/d No 1 d Fever (Tmax 39.2°C), epigastric pain for 1 d, epistaxis Blood Acute liver failure hepatic encephalopathy, coma, GI bleed, respiratory failure Death hospital day 7 21 M 71 Rectal cancer, hepatic metastases Chemotherapy No 1 d Fever (Tmax 40°C) after chemotherapy, stool OB(+) Blood Coma, seizure, septic shock Death hospital day 4 27 F 33 SLE with nephropathy Prednisone 60 mg/d, 2 course of MP pulses Yes 2 mo Diarrhea and abdominal pain for 2 mo; sudden onset fever (Tmax 39.2°C) on day 26 after admission Blood Multiple hospital-acquired infections, septic shock Death hospital day 30 5 F 43 Dermatomyositis, DM, HCC Prednisone 80 mg/d, CTX 0.4/wk Yes 4 d Fever (Tmax 39.7°C) started on day 20 after admission, with headache, left hemiplegia Blood Sepsis (meningitis) Recovered 8 F 22 SLE with nephropathy Prednisone 50–60 mg/d, 2 courses of MP pulses + hydroxychloroquine 0.2 bid + CyA/Dapsone/MMF Yes 2 wk Fever (Tmax 40°C) and diarrhea started on day 40 after admission Blood Meningitis Recovered 10 M 53 Still's disease Prednisone 50–60 mg/d or dexamethasone 5 mg/d, methotrexate 15 mg/d Yes 1 d Fever (Tmax 39.6°C) started on day 44 after admission, with headache, vomiting, change in mental status Blood, CSF Meningitis, respiratory failure, MRSA pneumonia Recovered 7 F 18 SLE None No 2 wk Fever (Tmax 39°C) headache and vomiting for 2 wk, and diplopia 1 d Blood, CSF Cryptococcus neoformans also grew from blood cultures Recovered 30 M 74 DM, chronic kidney disease None No 2 d Fever (Tmax 39.2°C), nausea, vomiting Blood, CSF Meningitis, HAP Recovered 17 M 69 None None No 4 d Fever (Tmax 39°C), change in mental status CSF Coma, ARF, pneumonia Recovered 13 F 53 SLE with nephropathy Prednisone 30 mg/d, CTX 0.4/wk No 2 d Fever (Tmax 39.4°C), headache, vomiting, loss of consciousness CSF Meningitis Recovered 12 F 45 Mixed connective tissue disease Prednisone 60 mg/d No 6 d Fever (38.5°C), headache and altered mental status CSF Meningitis, DVT Recovered 20 F 60 Non-Hodgkin lymphoma and lymphoblastic leukemia Chemotherapy and neutropenia Yes 5 d Fever (Tmax 40°C) started 5 d after c

of consciousness CSF Meningitis Recovered 12 F 45 Mixed connective tissue disease Prednisone 60 mg/d No 6 d Fever (38.5°C), headache and altered mental status CSF Meningitis, DVT Recovered 20 F 60 Non-Hodgkin lymphoma and lymphoblastic leukemia Chemotherapy and neutropenia Yes 5 d Fever (Tmax 40°C) started 5 d after c hemotherapy on hospital day 9 Blood Perianal abscess Recovered 26 F 42 Breast cancer with metastases to bone, liver, and lungs Neratinib (HKI-272), neutropenia Yes 3 d Fever (Tmax 39.8°C), oral ulcers, diarrhea, after HKI-272 on hospital day 15 Blood None Recovered 11 F 36 Ulcerative colitis, hepatic cirrhosis, AIH Prednisone 40 mg/d Yes 1 d Fever (Tmax 39.7°C) and hepatitis on hospital day 21 Blood None Recovered 9 F 49 DM None No 5 d Fever (Tmax 40.5°C) abdominal bloating, headaches Blood Urosepsis Recovered 22 M 79 Polycystic kidney disease, CRF None No 4 d Fever (Tmax 39°C), left upper quadrant abdominal pain Blood None Recovered 32 M 59 Ulcerative colitis, rectal cancer with diffuse metastases Chemotherapy Yes 1 d Fever (Tmax 40°C) on hospital day 24 Blood Candidiasis Recovered 33 F 36 Multiple myeloma None Yes 1 d Fever (Tmax 38°C), on hospital day 3 Blood HAP Recovered 29 M 59 DM None No 2 d Fever (Tmax 39.9°C), diarrhea, abdominal pain, mental status change Blood, CSF None Recovered Abbreviations: AIH, autoimmune hepatitis; ARF, acute renal failure; bid, twice daily; CRF, chronic renal failure; CSF, cerebrospinal fluid; CTX, cyclophosphamide; CyA, cyclosporine A; DM, diabetes mellitus; DVT, deep vein thrombosis; E.

x 39.9°C), diarrhea, abdominal pain, mental status change Blood, CSF None Recovered Abbreviations: AIH, autoimmune hepatitis; ARF, acute renal failure; bid, twice daily; CRF, chronic renal failure; CSF, cerebrospinal fluid; CTX, cyclophosphamide; CyA, cyclosporine A; DM, diabetes mellitus; DVT, deep vein thrombosis; E. coli, Escherichia coli; GI, gastrointestinal tract; HAP, hospital-acquired pneumonia; HCC, hepatocellular carcinoma; MMF, mycophenolate mofetil; MOF, multiple organ failure; MP, methylprednisolone; MRSA, methicillin-resistant Staphylococcus aureus; OB, occult blood; SLE, systemic lupus erythematosus; Tmax, maximal temperature. Fever (96%), CNS involvement (64%), and gastrointestinal symptoms (48%) were the most common presentations. Listeria monocytogenes was cultured from blood (n = 13), blood and CSF (n = 8), CSF (n = 3), and CSF and sputum (n = 1).

coli, Escherichia coli; GI, gastrointestinal tract; HAP, hospital-acquired pneumonia; HCC, hepatocellular carcinoma; MMF, mycophenolate mofetil; MOF, multiple organ failure; MP, methylprednisolone; MRSA, methicillin-resistant Staphylococcus aureus; OB, occult blood; SLE, systemic lupus erythematosus; Tmax, maximal temperature. Fever (96%), CNS involvement (64%), and gastrointestinal symptoms (48%) were the most common presentations. Listeria monocytogenes was cultured from blood (n = 13), blood and CSF (n = 8), CSF (n = 3), and CSF and sputum (n = 1). The 2 cases of L. monocytogenes that occurred in otherwise healthy hosts had early CNS involvement, manifested by coma. The first, a 20-year-old patient, experienced sudden onset of diarrhea, fever, and headache and deteriorated rapidly. He was intubated and treated at a local outside hospital first (where no L. monocytogenes was isolated from his cultures), and L. monocytogenes was isolated from sputum and CSF 4 weeks after the onset of gastrointestinal symptoms (Table 4, patient 4). The second, a 69-year-old previously healthy man, developed sudden fever and convulsions (Table 4, patient 17) rapidly progressing to coma complicated by acute renal failure and pneumonia. His condition improved after an extensive hospital stay and he was transferred to an outside institution for further rehabilitation. No long-term follow-up was available.

althy man, developed sudden fever and convulsions (Table 4, patient 17) rapidly progressing to coma complicated by acute renal failure and pneumonia. His condition improved after an extensive hospital stay and he was transferred to an outside institution for further rehabilitation. No long-term follow-up was available. Seventy-two percent of adults were treated empirically with cephalosporins and all were switched to ampicillin after the positive culture results became known. Among the 9 (36%; 95% CI, 20.25%–55.48%) fatal cases, 8 had severe underlying diseases and developed complications after being infected with L. monocytogenes. All died of multiple severe complications within 30 days after the onset of infection. The fatal cases were more likely to have sepsis (n = 9), rapid onset of coma (n = 6), and multiorgan failure (n = 3).

zil [28]. The median age of these patients was 80 years and all had underlying severe comorbidities. Four isolates belonged to a single pulsed-field gel electrophoresis (PFGE) genotype, suggesting a common source. The epidemiological investigation pointed to the hospital kitchen as the possible source of contamination. It is intriguing to speculate whether these healthcare-associated cases were the result of in-hospital acquisition, or whether this was the result of colonization. Until this retrospective case series was conducted, we had absolutely no insight about the frequency of these healthcare-associated cases. The cases were not clustered in time or space so they did not elicit additional surveillance. Although we could not perform PFGE on the specimens from our study, the majority did not cluster in time or space, suggesting that a common source was unlikely. Investigators have recognized for >20 years that L. monocytogenes can be carried in the gastrointestinal tract [29–31]. Listeria monocytogenes can be isolated in the stool of 1%–10% of the population, where it can persist without causing symptoms [32]. Using repeated sampling, Listeria can be detected in the feces of nearly 70% of healthy nonpregnant individuals and 44% of pregnant women [21, 31]. MacGowan et al found that Listeria was isolated from 5.6% (10/177) of renal transplant recipients on 1 or more occasions over the period of a year; moreover, >1 species or serovar of listeria can be isolated from 40% of fecal carriers, and no cases of clinical infection occurred in any fecal carriers [33]. Fecal, cervicovaginal, and oropharyngeal carriage of L. monocytogenes has been reported as a possible predisposing factor for perinatal listeriosis [34, 35]. In one study conducted by Schuchat et al [36], asymptomatic carriage of the illness-associated strain of L. monocytogenes was identified in nearly one-fifth of household contacts of patients with sporadic listeriosis, and no cases of secondary disease were detected within households in this study. Their findings suggest that gastrointestinal carriage of pathogenic strains of L. monocytogenes is not uncommon in contacts of cases, underscoring the critical role that host susceptibility plays in determining whether illness occurs following exposure to this organism. All of our cases of healthcare-associated listeriosis had severe underlying immunosuppression.

carriage of pathogenic strains of L. monocytogenes is not uncommon in contacts of cases, underscoring the critical role that host susceptibility plays in determining whether illness occurs following exposure to this organism. All of our cases of healthcare-associated listeriosis had severe underlying immunosuppression. Besides immunosuppression, many of our patients had underlying diseases involving the gastrointestinal tract, or their therapy could impact the integrity of the intestinal mucosa. So, the role that gastrointestinal colonization of Listeria played in the pathogenesis of these healthcare-associated infections warrants further study. After the discovery of these nonclustered healthcare-associated cases, we have implemented a more aggressive approach: all healthcare-associated cases will be thoroughly investigated for both prehospital and in-hospital exposures. We are also saving all bacterial isolates for DNA fingerprinting. This more aggressive approach may help us better define the source of these infections.

cases, we have implemented a more aggressive approach: all healthcare-associated cases will be thoroughly investigated for both prehospital and in-hospital exposures. We are also saving all bacterial isolates for DNA fingerprinting. This more aggressive approach may help us better define the source of these infections. Among the healthcare-associated listeriosis cases, one patient with diffuse metastatic breast cancer experienced sudden onset of fever, oral ulcers, and diarrhea after 3 days of HKI-272 treatment (Table 4, patient 26). Blood culture yielded L. monocytogenes. The HKI-272 therapy was discontinued and antibiotic treatment was initiated, and the patient fully recovered. HKI-272, also known as neratinib [37], is an oral, irreversible dual EGFR/HER2 inhibitor for breast and non-small-cell lung cancer. Phase 1 and 2 studies reported gastrointestinal adverse events, including diarrhea (89%), nausea (29%–64%), and vomiting (23%–50%). Approximately 30% of patients required discontinuation or dose reduction due to severe diarrhea. Cases of listeriosis were reported among patients undergoing therapy with other biologic agents such as infliximab (antitumor necrosis factor agents) [38–41], etanercept (a tumor necrosis factor antagonist) [42], and trastuzumab (a monoclonal antibody against the HER2 receptor) [43].

e reduction due to severe diarrhea. Cases of listeriosis were reported among patients undergoing therapy with other biologic agents such as infliximab (antitumor necrosis factor agents) [38–41], etanercept (a tumor necrosis factor antagonist) [42], and trastuzumab (a monoclonal antibody against the HER2 receptor) [43]. Forty percent of our cases had underlying rheumatologic diseases. This proportion is higher than what was previously reported in the literature [38]. Although PUMCH does not specifically specialize in the treatment of rheumatic diseases, we do have a large population of such patients. Persons of Asian descent have a higher incidence of SLE, compared with other races [44–46]. Given the paucity of published reports on L. monocytogenes from East Asia, this may explain the higher incidence among patients with rheumatic diseases in our report. This may also have impacted the sex distribution of cases. Traditionally, L. monocytogenes has been reported more often among men than women. The male to female ratio in our study was 1:1.8. This may reflect the increased predisposition of rheumatic diseases among women [47–49]. Comorbidity plays a very important role in the prognosis of listeriosis [18]. Eighty-one percent of 225 patients with listeriosis studied in France had a predisposing immunocompromising condition, whose severity was the major prognostic factor [17]. In our population, 92% of nonmaternal listeriosis cases were immunosuppressed.

ity plays a very important role in the prognosis of listeriosis [18]. Eighty-one percent of 225 patients with listeriosis studied in France had a predisposing immunocompromising condition, whose severity was the major prognostic factor [17]. In our population, 92% of nonmaternal listeriosis cases were immunosuppressed. Our cases of infant listeriosis mirrored the cases reported in the literature, as did their outcomes. We did not observe any late-onset cases of infant listeriosis, as reported by other authors [9, 22, 50–52]. Similarly, the characteristics of our maternal listeriosis were similar to those reported in the literature. This study has several limitations. First, it is a retrospective assessment over a protracted timespan. As such, we were unable to obtain specimens for molecular testing, and we were unable to clarify additional issues relating to certain in-hospital epidemiological exposures. Second, it consists of a relatively small sample size, and our findings may not be necessarily generalizable to other populations or settings. Third, cases of listeriosis in China are not routinely reported to public health authorities. As such, the epidemiology of listeria is not well defined. Our case series reflects a selection bias toward hospitalized (ie, sicker) patients and may not reflect the overall epidemiology of listeria.

populations or settings. Third, cases of listeriosis in China are not routinely reported to public health authorities. As such, the epidemiology of listeria is not well defined. Our case series reflects a selection bias toward hospitalized (ie, sicker) patients and may not reflect the overall epidemiology of listeria. Nonclustered healthcare-associated cases of L. monocytogenes occurred at a large tertiary care hospital in Beijing, China. The source of these infections is unclear. Although rare, in the setting of immunosuppression, Listeria should be considered in the differential diagnosis of healthcare-associated infections—even in the absence of a point-source outbreak. Note Acknowledgments. We thank all healthcare providers who had participated in taking care of our patients. We are grateful to all the medical record staff for their support. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Nontyphoidal Salmonella species are important foodborne pathogens worldwide [1], causing diarrhea, vomiting, nausea, fever, and abdominal pain. Illness has been linked to a wide range of food items including eggs, chicken, beef, pork, salad vegetables, and dairy products, and other risk factors including overseas travel [2–7]. Outbreaks are fairly common [5]. The burden of illness, defined as morbidity and mortality, is substantial. In the United States, nontyphoidal Salmonella species are estimated to cause 1 million foodborne illnesses [8] and are the leading cause of death among foodborne bacterial pathogens [9]. Across the 27 member states of the European Union (EU), there were estimated to be 6.2 million cases of salmonellosis in 2009 [10]. In a population-based study in the United Kingdom (UK) in 2008–2009, there were >38 600 estimated cases and nearly 11 300 patients presenting to a primary care physician [11]. This represented a marked reduction in incidence compared with a similar study conducted more than a decade earlier [12, 13]. The purpose of this article is to discuss the factors associated with a substantial decline in nontyphoidal salmonellosis in the United Kingdom since the mid-1990s.

primary care physician [11]. This represented a marked reduction in incidence compared with a similar study conducted more than a decade earlier [12, 13]. The purpose of this article is to discuss the factors associated with a substantial decline in nontyphoidal salmonellosis in the United Kingdom since the mid-1990s. A BRIEF HISTORY OF NONTYPHOIDAL SALMONELLOSIS IN THE UNITED KINGDOM Remarkable changes in the epidemiology of human nontyphoidal salmonellosis have occurred in the United Kingdom over the last century. Prior to 1942, the dominant foodborne salmonellas causing disease were Salmonella enterica subspecies enterica serovar Typhimurium, Salmonella Enteritidis, Salmonella Thompson, Salmonella Newport, Salmonella Bovismorbificans, and Salmonella Choleraesuis [14]. Salmonella Typhimurium remained the dominant serovar causing human disease for much of the 20th century, although there were fluctuations in other salmonellas in the “top 10” over time. For example, Salmonella Agona emerged as an important serovar in the 1960s following its introduction into pigs and poultry through contaminated fish meal imported from Peru [15]. Salmonella Hadar became the second most commonly isolated cause of human nontyphoidal salmonellosis in the mid-1970s when particular genetic lines of turkeys became infected [15]. Against this background, the incidence of Salmonella Enteritidis increased fairly gradually from around 150 to approximately 900 laboratory-confirmed cases per year between 1961 and 1980 [16]. During this time, phage type (PT) 8 dominated and was responsible for several turkey-associated outbreaks in the late 1960s [16]. By 1975 Salmonella Enteritidis was consistently the second or third most frequently isolated serovar annually [17].

mately 900 laboratory-confirmed cases per year between 1961 and 1980 [16]. During this time, phage type (PT) 8 dominated and was responsible for several turkey-associated outbreaks in the late 1960s [16]. By 1975 Salmonella Enteritidis was consistently the second or third most frequently isolated serovar annually [17]. Between 1981 and 1991, the incidence of nontyphoidal salmonellosis in the United Kingdom rose by >170% [18], driven primarily by an epidemic of Salmonella Enteritidis PT4 [16, 18–20] (Figure 1). In 1981 Salmonella Enteritidis accounted for approximately 10% of human Salmonella illnesses, but by 1993 this proportion had risen to nearly 70% [20]. In the early 1980s, PT4 overtook PT8 to become the predominant phage type in 1983, comprising 46% of isolations that year. By 1988 PT4 had risen to account for 81% of Salmonella Enteritidis strains isolated [16] and had ended the political career of a prominent government minister [21]. The United Kingdom was not alone; analysis of data submitted to the World Health Organization's Salmonella surveillance system showed that Salmonella Enteritidis in the late 1980s was increasing on several continents, with North America, South America, and Europe appearing to bear the brunt [22]. Figure 1. Laboratory reports of human Salmonella cases in the United Kingdom, 1981–2010. Abbreviations: CMO, Chief Medical Officer; PT, phage type.

m showed that Salmonella Enteritidis in the late 1980s was increasing on several continents, with North America, South America, and Europe appearing to bear the brunt [22]. Figure 1. Laboratory reports of human Salmonella cases in the United Kingdom, 1981–2010. Abbreviations: CMO, Chief Medical Officer; PT, phage type. EVIDENCE THAT THE DECLINE IN SALMONELLA IS REAL Compelling evidence that the decline in Salmonella is real is derived from 3 sources. The first comprises 2 population-based prospective cohort studies of infectious intestinal disease (IID) in the community conducted more than a decade apart [11–13]. The primary outcome measures in both studies were estimates of the incidence of IID in the community, presenting to primary healthcare and reported to national surveillance. They were conducted using identical study designs and case definitions and employed similar microbiological methods, the exception being that molecular microbiological techniques were used alongside traditional microbiology in the second study of infectious intestinal disease (IID2). In the first study of infectious intestinal disease (IID1) in 1993–1996, the incidence of nontyphoidal Salmonella in the community in England was 2.2 cases per 1000 person-years (95% confidence interval [CI], 1.1–4.3) but by 2008–2009 this had fallen to 0.7 cases per 1000 person-years (95% CI, .2–3.0). For nontyphoidal Salmonella cases presenting to primary care in England, the incidence rate had fallen from 1.6 cases per 1000 person-years (95% CI, 1.2–2.1) in IID1 to 0.2 cases per 1000 person-years (95% CI, .1–.5) in IID2. The decline in incidence in the community was not statistically significant because in IID2 the study power was insufficient to detect statistically significant changes in organism-specific incidence—to do this would have required >100 000 person-years of follow-up, based on incidence rates in IID1. Nevertheless, the reduction in presentations to primary healthcare was statistically significant.

cause in IID2 the study power was insufficient to detect statistically significant changes in organism-specific incidence—to do this would have required >100 000 person-years of follow-up, based on incidence rates in IID1. Nevertheless, the reduction in presentations to primary healthcare was statistically significant. Second, there has been a substantial fall in laboratory-confirmed Salmonella cases reported to national surveillance (Figure 1). Phage typing of Salmonella Enteritidis was implemented from 1981 as an addition to the centralized, national service already in existence for confirmation and further typing [17], and all clinical diagnostic laboratories have continued to refer all Salmonella isolates to the national reference laboratories since that date. At the beginning of 1992, 2 separate national Salmonella databases were merged to form a single national dataset, which became patient-based rather than isolate-based, thus eliminating potential duplication if people were tested more than once [18]. Laboratory testing methods have remained constant since then and reporting algorithms have not changed [23], suggesting that the reduction in Salmonella is real. When Salmonella Enteritidis PT4 peaked in 1993 in the United Kingdom, >18 000 laboratory-confirmed cases of illness were recorded in national surveillance statistics, yet by 2010 PT4 isolations had fallen to just 459 [24]. Thus, the decline in nontyphoidal salmonellosis witnessed in the United Kingdom in recent years reflects this major contraction in reports of Salmonella Enteritidis PT4.

8 000 laboratory-confirmed cases of illness were recorded in national surveillance statistics, yet by 2010 PT4 isolations had fallen to just 459 [24]. Thus, the decline in nontyphoidal salmonellosis witnessed in the United Kingdom in recent years reflects this major contraction in reports of Salmonella Enteritidis PT4. Finally, outbreaks of salmonellosis have declined. Standardized reporting of outbreaks of gastrointestinal infection was introduced in 1992 in England and Wales and in 1996 in Scotland partly in response to the increase in nontyphoidal salmonellosis. A foodborne outbreak is defined in European legislation as “an incidence, observed under given circumstances, of two of more human cases of the same disease and/or infection, or a situation in which the observed number of human cases exceeds the expected number and where the cases are linked, or are probably linked, to the same source” [25]. Between 1992 and 2008, foodborne Salmonella outbreaks reported to national surveillance fell from nearly 150 per year to just over 20 annually, and the pattern of decline closely mirrors that of laboratory-confirmed cases [25].

ted number and where the cases are linked, or are probably linked, to the same source” [25]. Between 1992 and 2008, foodborne Salmonella outbreaks reported to national surveillance fell from nearly 150 per year to just over 20 annually, and the pattern of decline closely mirrors that of laboratory-confirmed cases [25]. EPIDEMIOLOGY OF SALMONELLA ENTERITIDIS IN THE UNITED KINGDOM Epidemiologic investigations of outbreaks and sporadic cases repeatedly showed that Salmonella Enteritidis PT4 infection in humans was frequently associated with consumption of poultry meat and hens' eggs on both sides of the Atlantic [25–31]. In nearly 2500 foodborne disease outbreaks reported to the UK Health Protection Agency between 1992 and 2008, Salmonella species accounted for 47% of all outbreaks, 46% of cases, 70% of hospital admissions, and 76% of deaths [25]. Salmonella Enteritidis PT4 was the causative organism in 51% of all the Salmonella outbreaks throughout the surveillance period but the percentage of outbreaks caused by Salmonella Enteritidis PT4 declined from the late 1990s onward. At least one food vehicle was identified in 75% of outbreaks reported, and poultry meat was the vehicle most often implicated (19% of outbreaks). Desserts were also implicated commonly (11% of outbreaks), and raw shell eggs were used as an ingredient in 70% of these desserts. Eggs were implicated separately in an additional 6% of outbreaks. Analysis of outbreak data also showed that nearly 50% of foodborne Salmonella outbreaks occurred in the food service/catering sector.

ere also implicated commonly (11% of outbreaks), and raw shell eggs were used as an ingredient in 70% of these desserts. Eggs were implicated separately in an additional 6% of outbreaks. Analysis of outbreak data also showed that nearly 50% of foodborne Salmonella outbreaks occurred in the food service/catering sector. Salmonella Gallinarum and Salmonella Pullorum had been the dominant Salmonella serovars in UK poultry until the early 1970s. These strains both caused clinical disease in the birds and were virtually eradicated by a combination of slaughtering of seropositive hens and vaccination [20]. However, the ecological niche left by these 2 serovars was filled by Salmonella Enteritidis. Complete genome sequencing of a host-promiscuous Salmonella Enteritidis PT4 isolate (P125109) and a chicken-restricted Salmonella Gallinarum isolate (287/91) has indicated that Salmonella Gallinarum 287/91 is a recently evolved descendent of Salmonella Enteritidis [32]. Importantly, Salmonella Enteritidis infects poultry without causing overt disease, which probably facilitated its rapid spread internationally [20]. Another key feature of Salmonella Enteritidis is colonization of the reproductive tissues leading to the production of eggs with Salmonella-positive contents [20, 33] and, in some eggs, the numbers of organisms can be very high [34].

using overt disease, which probably facilitated its rapid spread internationally [20]. Another key feature of Salmonella Enteritidis is colonization of the reproductive tissues leading to the production of eggs with Salmonella-positive contents [20, 33] and, in some eggs, the numbers of organisms can be very high [34]. CONTROLLING SALMONELLOSIS AND OTHER FOODBORNE ILLNESSES In August 1988, as evidence of a link between Salmonella Enteritidis PT4 and raw shell eggs strengthened, the Chief Medical Officer issued advice to consumers to avoid eating raw eggs or uncooked foods in which raw eggs were an ingredient. In December of the same year, he issued further advice to vulnerable people such as the elderly, individuals with chronic illness, infants, and pregnant women. They were counseled only to eat eggs that had been cooked until the yolks and whites were solid [18]. Caterers were encouraged to use pasteurized eggs, especially where foodstuffs were not going to be cooked further (eg, mayonnaise), and it was recommended that eggs be considered short shelf-life products. They should be refrigerated <8°C throughout the production chain and during retail, catering, and domestic storage, and consumed within 3 weeks of the date of lay [18]. In 1989 the government introduced a raft of legislation, including the Zoonoses Order, which required that all Salmonella isolates from live animals or birds, carcasses, or feedstuffs be reported. Movement restrictions were implemented along with compulsory slaughter, compensation, and disinfection procedures. The more draconian procedures were usually reserved for Salmonella Typhimurium and Salmonella Enteritidis [24]. The requirement for compulsory slaughter of poultry flocks was revoked following a recommendation from the Advisory Committee on the Microbiological Safety of Food in 1993 to review the policy in light of the fact that Salmonella Enteritidis in flocks had reduced substantially [18]. In 1989, >600 000 birds from 58 infected flocks were slaughtered. In 1992, <300 000 birds from 38 infected flocks were slaughtered [18]. Alongside legislation was a voluntary, industry-led vaccination scheme that began in broiler-breeder flocks in 1994 and in laying flocks in 1998 [16]. A “Lion Mark,” stamped on eggs, which had been introduced in 1957 but dropped by 1971, was revived in 1998 (http://www.lioneggs.co.uk/page/lionmark).

e slaughtered [18]. Alongside legislation was a voluntary, industry-led vaccination scheme that began in broiler-breeder flocks in 1994 and in laying flocks in 1998 [16]. A “Lion Mark,” stamped on eggs, which had been introduced in 1957 but dropped by 1971, was revived in 1998 (http://www.lioneggs.co.uk/page/lionmark). The Lion Mark can only be used by subscribers to the British Egg Industry Council for eggs that have been produced in accordance with UK and EU law and the Lion Quality Code of Practice. The code of practice requires mandatory vaccination of all pullets destined to lay Lion eggs against Salmonella; independent auditing; full traceability of hens, eggs, and feed; and a “best-before” date stamped on the shell and pack, in addition to on-farm stamping of eggs and packing station hygiene controls.

ractice. The code of practice requires mandatory vaccination of all pullets destined to lay Lion eggs against Salmonella; independent auditing; full traceability of hens, eggs, and feed; and a “best-before” date stamped on the shell and pack, in addition to on-farm stamping of eggs and packing station hygiene controls. When, in 1989, a Junior Health Minister stated in a British television interview that “Most of the egg production in this country, sadly, is now infected with Salmonella,” the sale of eggs collapsed by 60% almost overnight. Moreover, despite government efforts to improve the safety of eggs, sales continued to fall by around 8% per year over the next 10 years, which was a disaster for the industry. The British Egg Industry Service began a major consumer research program in 1997 and, in 1998, the majority of UK producers and packers made a voluntary investment of £8 million to assist the British Egg Industry Council to relaunch British eggs. A total of £4 million was spent on the stringent new Code of Practice described above, and £4 million supported a new promotional campaign to restore consumer confidence and increase consumption. The cost of the vaccination program (including Lion sampling and testing) is estimated to be around £52 million to date (Mark Williams, written personal communication, September 2012). However, between 1998 and 2009, the egg market grew from 9.8 billion to 11 billion eggs per year, and Lion eggs now account for around 85% of the total market. Within the retail sector the market share of Lion eggs share rose from approximately 60% in 1998 to 95% in 2010 (http://www.lioneggs.co.uk/files/lioneggs.co.uk/pdfs/marketing.pdf).

8 and 2009, the egg market grew from 9.8 billion to 11 billion eggs per year, and Lion eggs now account for around 85% of the total market. Within the retail sector the market share of Lion eggs share rose from approximately 60% in 1998 to 95% in 2010 (http://www.lioneggs.co.uk/files/lioneggs.co.uk/pdfs/marketing.pdf). Alas, Salmonella was not the only “food scare” during the 1980s and 1990s. Scandals surrounding, for example, bovine spongiform encephalopathy in the United Kingdom, dioxins in Belgium, and Salmonella EU-wide prompted new legislation providing for a risk-based “farm to fork” approach to food safety policy, which was enacted in 2002 (European General Food Law [Regulation (EC) No. 178/2002]) [24]. EU Zoonoses Regulation (EC) No. 2160/2003 required member states to take effective measures to detect and control Salmonella species of public health significance in specified animal species at all relevant stages of production [24]. Each EU member state was obliged to undertake a standardized baseline survey to determine the prevalence of Salmonella within their industry sectors. EC Regulation (EC) 1168/2006 laid down an annual reduction target for Salmonella Enteritidis and Salmonella Typhimurium for each member state.

stages of production [24]. Each EU member state was obliged to undertake a standardized baseline survey to determine the prevalence of Salmonella within their industry sectors. EC Regulation (EC) 1168/2006 laid down an annual reduction target for Salmonella Enteritidis and Salmonella Typhimurium for each member state. NATIONAL CONTROL PROGRAMS FOR SALMONELLA IN THE POULTRY SECTOR Four National Control Programmes (NCPs) for Salmonella have been implemented in the UK poultry sector between 2007 and 2010. These postdate the rapid decline in Salmonella Enteritidis in the United Kingdom but are designed to achieve and maintain low rates EU-wide. For the most part, the targets set by the EU have already been met or exceeded in the United Kingdom [24]. The NCP for breeding chickens (implemented in 2007): The target for this NCP was that no more than 1% of adult breeding flocks should be infected with 5 specific regulated serovars (Salmonella Enteritidis, Salmonella Typhimurium, Salmonella Hadar, Salmonella Infantis, and Salmonella Virchow) by the end of 2009. Results from UK holdings have been significantly below the EU target of 1% every year for the last 4 years [24].

adult breeding flocks should be infected with 5 specific regulated serovars (Salmonella Enteritidis, Salmonella Typhimurium, Salmonella Hadar, Salmonella Infantis, and Salmonella Virchow) by the end of 2009. Results from UK holdings have been significantly below the EU target of 1% every year for the last 4 years [24]. The NCP for commercial laying chickens (implemented in 2008): An EU-wide baseline survey of commercial laying chicken flock holdings was undertaken in 2004–2005. In a survey of Salmonella species on 454 commercial layer flock holdings in the United Kingdom, 54 (11.7%) were Salmonella positive [35]. Salmonella Enteritidis was the serovar most commonly identified (prevalence = 5.8%) and PTs 4, 6, 7, and 35 comprised 70% of isolates. Salmonella Typhimurium was the second most commonly identified serovar (prevalence = 1.8%). The UK prevalence figures were among the lowest of the major egg-producing countries (7.9% of holdings positive compared with a 20.4% average across the EU) [36]. Across the EU, the incidence rate of salmonellosis in member states varies between 16 and 11 800 per 100 000 population and has been shown to be significantly correlated with the prevalence of Salmonella Enteritidis in laying hens [10], so controlling levels of Salmonella Enteritidis in laying flocks is important for improving public health.

dence rate of salmonellosis in member states varies between 16 and 11 800 per 100 000 population and has been shown to be significantly correlated with the prevalence of Salmonella Enteritidis in laying hens [10], so controlling levels of Salmonella Enteritidis in laying flocks is important for improving public health. The NCP for broilers (implemented in 2009): The target for this NCP was that no more than 1% of flocks should be infected with Salmonella Enteritidis and Salmonella Typhimurium by the end of 2011. In a baseline survey of broiler chickens in 2005–2006 in the United Kingdom, the prevalence of Salmonella Enteritidis and Salmonella Typhimurium was very low (0.2% [37] compared with an EU average of 11.0% [38]) and remains well below the EU target [24]. The NCP for turkeys (implemented in 2010): A baseline survey for Salmonella in turkey breeding and fattening flocks was carried out across the EU in 2006–2007. In the United Kingdom, the prevalence of Salmonella in breeding flock holdings was 20.1% and in fattening flocks the holdings prevalence was 37.7% [39]. The flock prevalence of Salmonella Typhimurium was very low on breeding holdings at 0.7% (EU weighted average = 1.8%) but higher on fattening holdings at 4.6% (EU weighted average = 3.7%) [24]. The target for Salmonella reduction is that only 1% of breeding flocks and 1% of fattening flocks should be positive by the end of 2012. Early indications are that this target will be met.

breeding holdings at 0.7% (EU weighted average = 1.8%) but higher on fattening holdings at 4.6% (EU weighted average = 3.7%) [24]. The target for Salmonella reduction is that only 1% of breeding flocks and 1% of fattening flocks should be positive by the end of 2012. Early indications are that this target will be met. WHAT NEXT? There is no room for complacency. During the 2000s, new Salmonella problems emerged. Notable among these were national outbreaks of Salmonella Enteritidis PT14b linked to raw shell eggs originating in Spain [40, 41]. Unbelievably, perhaps, hospital caterers in the United Kingdom were found serving raw shell eggs again to patients, with consequent outbreaks [42]. The first outbreak of Salmonella Typhimurium PT8 linked to consumption of duck eggs since 1949 occurred in the United Kingdom [43], and Salmonella outbreaks linked to fresh produce were increasingly recognized [44, 45], reflecting a pattern also seen in the United States [46].

patients, with consequent outbreaks [42]. The first outbreak of Salmonella Typhimurium PT8 linked to consumption of duck eggs since 1949 occurred in the United Kingdom [43], and Salmonella outbreaks linked to fresh produce were increasingly recognized [44, 45], reflecting a pattern also seen in the United States [46]. CONCLUSIONS The nature of public health interventions often means that evaluating their impact is complex as they are often implemented in combination and/or simultaneously. It is interesting to reflect on the fact that the various legislative measures in the United Kingdom in the late 1980s and early 1990s appear to have slowed down the increase in Salmonella Enteritidis PT4, whereas the decrease in laboratory-confirmed human cases coincides quite closely with the introduction of vaccination programs in broiler-breeder and laying flocks and prior to much of the EU legislation being implemented. It is probable that no single measure contributed to the decline in Salmonella Enteritidis PT4 and that the combination of measures was successful, but the temporal relationship between vaccination programs and the reduction in human disease is compelling and suggests that these programs have made a major contribution to improving public health. There has also been a reduction in reported human salmonellosis cases across the EU (on average 12% per year between 2005 and 2009). The European Commission and European Food Safety Authority are attributing this, at least in part, to successful control of Salmonella in broiler, laying, and breeding hen flocks and eggs [24].

has also been a reduction in reported human salmonellosis cases across the EU (on average 12% per year between 2005 and 2009). The European Commission and European Food Safety Authority are attributing this, at least in part, to successful control of Salmonella in broiler, laying, and breeding hen flocks and eggs [24]. If success in public health is defined by illnesses averted, then the story of Salmonella Enteritidis PT4 in the United Kingdom, which has come down and stayed down, is good news. However, history teaches us that something else may come along to take its place. Robust surveillance, incorporating state-of-the-art microbiological, epidemiological, and biostatistical methods, and maintaining a prompt and comprehensive response to outbreaks is just as important now as it ever was. Notes Acknowledgments. I thank Mark Williams of the British Egg Industry Council for information on the cost of the vaccination programs. Potential conflicts of interest. Author certifies no potential conflicts of interest. The author has submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

To the Editor—Maternal and neonatal tetanus is a significant cause of mortality, estimated to cause 180 000 deaths annually [1]. Since the mid-1970s, tetanus vaccination of pregnant women has been included in the World Health Organization's (WHO) Expanded Programme on Immunisation (EPI) [2]. Two doses of tetanus toxoid are sufficient to generate an antibody response (immunoglobulin G [IgG] class) capable of protecting neonates from tetanus, 3 doses are recommended for pregnancy, and 5 are recommended for life [3]. Despite these recommendations, WHO has identified a lack of longitudinal data quantifying antitetanus antibody boosting and duration during pregnancy following immunization in the EPI schedule [4]. To address this gap, we determined levels of antitetanus IgG at multiple time points from enrollment to delivery (median, 30 weeks of follow-up) in 376 pregnant women participating in malarial antibody studies at the antenatal clinics of the Shoklo Malaria Research Unit in northwest Thailand (previously published with ethics statement in Fowkes et al [5]). The tetanus vaccination regimen (tetanus toxoid) followed EPI guidelines [3]: dose 1, as early as possible during pregnancy; dose 2, one month after dose 1; dose 3, 6 months after dose 2; dose 4, 1 year after dose 3; and dose 5, one year after dose 4. During the study, 48.9% of women received their first dose, 86.2% received doses 2–4, and 8.2% received the final dose (dose 5).

I guidelines [3]: dose 1, as early as possible during pregnancy; dose 2, one month after dose 1; dose 3, 6 months after dose 2; dose 4, 1 year after dose 3; and dose 5, one year after dose 4. During the study, 48.9% of women received their first dose, 86.2% received doses 2–4, and 8.2% received the final dose (dose 5). The boosting and decay of tetanus antibody levels after vaccination was vaccine dose-dependent (Figure 1). In the first 8 days after vaccination, antitetanus IgG increased rapidly at comparable rates in all vaccination groups (P > .85 relative to T1). Interestingly, at 8 days after vaccination, IgG responses peaked and then plateaued in those receiving ≥2 doses. In contrast, IgG responses in those receiving their first vaccination peaked later at 50 days after vaccination (Figure 1, P < .001). After 50 days postvaccination, antitetanus IgG responses declined and calculated IgG half-life was dependent on vaccination dose: 7.12 years (95% confidence interval [CI], 3.02–∞) for dose 1; 10.97 years (6.71–∞) for doses 2–4; and 12.28 years (6.15–∞) for dose 5. These estimates are in concordance with published nondose-specific half-life estimates in nonpregnant American women (10 years, 95% CI, 8–14) [6]. Figure 1. Antitetanus immunoglobulin G (IgG) after vaccination in 376 pregnant women according to vaccination dose. Tetanus IgG levels (optical density) were determined by enzyme-linked immunosorbent assay as previously described [5] with tetanus toxoid coated at 0.4 colony-forming units/mL and sera tested at a 1:500 dilution. Lines represent predicted mean tetanus IgG levels (calculated by mixed linear models with random effect for the intercept, slope, and covariance). The best-fit model had linear splines placed at 8 and 50 days, and each vaccination category had its own slope for antibody level over time since vaccination. Analysis was unadjusted because other potential confounders including gravidity, trimester, chloroquine prophylaxis [7], and Plasmodium species infection [8] did not significantly alter the model outputs. Abbreviations: IgG, immunoglobulin G; OD, optical density.

wn slope for antibody level over time since vaccination. Analysis was unadjusted because other potential confounders including gravidity, trimester, chloroquine prophylaxis [7], and Plasmodium species infection [8] did not significantly alter the model outputs. Abbreviations: IgG, immunoglobulin G; OD, optical density. The close consecutive sampling of antitetanus antibody levels has allowed us to define, in the greatest detail to date, antitetanus IgG kinetics postvaccination, and we provide the first estimates of tetanus IgG half-lives in pregnancy according to vaccination dose in the EPI schedule. These data are important for predicting protection in neonates and are invaluable for understanding the sustainability of protective humoral immunity in high-risk populations such as pregnant women in resource-poor settings. Notes Acknowledgments. We thank the participants and the Karen staff of the Shoklo Malaria Research Unit (SMRU) and Nadia Cross and Gaoqian Feng for technical assistance. We also thank CSL Limited, Australia, for kindly providing tetanus toxoid. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Notes Acknowledgments. We thank the participants and the Karen staff of the Shoklo Malaria Research Unit (SMRU) and Nadia Cross and Gaoqian Feng for technical assistance. We also thank CSL Limited, Australia, for kindly providing tetanus toxoid. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Financial support. This work was supported by the National Health and Medical Research Council of Australia (program grant to J. G. B.; training award to F. J. I. F.; Infrastructure for Research Institutes Support Scheme Grant), Australian Research Council (Future Fellowship to J. G. B.), and Victorian State Government Operational Infrastructure Support grant. SMRU is part of the Mahidol Oxford University Tropical Medicine Research Unit supported by the Wellcome Trust of Great Britain. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Major Article by Sutanto et al, on pages 685–93 and see the Editorial Commentary by Dondorp, on pages 694–6.) The emergence of partial artemisinin resistance in Plasmodium falciparum on the Cambodia-Thailand border and more recently on the Myanmar-Thailand border jeopardizes the renewed global efforts of control and elimination of malaria [1–4]. This poses a danger that malaria could become untreatable, as no good alternatives to artemisinin-based combination therapy (ACT) are currently available [5]. Increasing the dose of the artemisinin component in ACT might partly overcome decreased sensitivity, in case the concentration-effect relationship has merely shifted to higher concentrations in this parasite population. Increasing the dosing frequency could be another strategy to increase artemisinin efficacy. A unique property of the artemisinin drugs is their broad-stage specificity including young ring-stage asexual parasites [6]. There is evidence that artemisinin resistance particularly involves ring-stage parasites [2, 7, 8]. This can affect the dosing frequency of the short-half-life artemisinins, which have therapeutic drug concentrations until approximately 6 hours after the dose [8]. A mathematical model predicted that twice-daily dosing of artesunate (AS) would be more effective than the conventional once-daily dose [8].

tes [2, 7, 8]. This can affect the dosing frequency of the short-half-life artemisinins, which have therapeutic drug concentrations until approximately 6 hours after the dose [8]. A mathematical model predicted that twice-daily dosing of artesunate (AS) would be more effective than the conventional once-daily dose [8]. The current study explored whether increasing or splitting the once-daily AS dose increased efficacy, defined as parasite half-life [9], in patients with uncomplicated malaria in Pailin, western Cambodia, an area of artemisinin resistance, and in Wang Pha, in northwestern Thailand, where ACT has sustained efficacy since 1994 [10], although a recent decline has also been observed here [4]. METHODS Study Design In 2 open-label, randomized clinical trials, we compared different AS dosing regimens in patients with uncomplicated falciparum malaria presenting in Pailin, western Cambodia (2008–2010) and in Wang Pha, northwestern Thailand (2009–2010), areas with low and seasonal transmission. Studies were conducted according to Good Clinical Practices, and monitored independently (Family Health International). A data safety monitoring committee (DSMC) assured safety of the participants. Ethical approval was obtained from the Ministry of Health in Cambodia, the Ethics Committee of the Faculty of Tropical Medicine of Mahidol University in Thailand, the Oxford Tropical Medicine Ethical Committee, and the World Health Organization Research Ethics Review Committee.

DSMC) assured safety of the participants. Ethical approval was obtained from the Ministry of Health in Cambodia, the Ethics Committee of the Faculty of Tropical Medicine of Mahidol University in Thailand, the Oxford Tropical Medicine Ethical Committee, and the World Health Organization Research Ethics Review Committee. Patients Patients with acute uncomplicated falciparum malaria, monoinfection, and asexual stage parasitemia between 10 000 parasites/μL and 175 000 parasites/μL assessed by microscopy were eligible provided that written informed consent was obtained from all patients or their guardian (for children). In Pailin, patients of 6 years and older were studied, whereas in Wang Pha only adult patients (≥18 years) were studied. Pregnant or lactating women were excluded as well as patients with any signs of severe infection [11]. Antimalarial treatment within 48 hours of screening, known hypersensitivity to study drugs, or splenectomy were exclusion criteria.

older were studied, whereas in Wang Pha only adult patients (≥18 years) were studied. Pregnant or lactating women were excluded as well as patients with any signs of severe infection [11]. Antimalarial treatment within 48 hours of screening, known hypersensitivity to study drugs, or splenectomy were exclusion criteria. Randomization and Drug Therapy Patients were randomly allocated to 1 of 4 treatment arms: (1) AS alone in a dose of 6 mg/kg/d for 7 days (AS7); (2) the same total dose, but given as a split twice-daily dose (AS7_split); (3) AS in a dose of 8 mg/kg/d for 3 days, followed by mefloquine in a dose of 15 mg/kg on day 3 and 10 mg/kg on day 4 (8MAS3); (4) the same total dose, but AS given as a split twice-daily dose (8MAS3_split). Arms AS7 and AS7_split were suspended after an association with neutropenia was reported in a separate study [12]. The 7-day regimens using AS 6 mg/kg/d were then replaced by 3 days of AS 6 mg/kg/d either as single or split daily dose, followed by mefloquine 25 mg/kg divided over 2 days (Figure 1). Figure 1. Enrollment, randomization, and follow-up of the patients in the 2 study sites. Please see the Methods section for a description of the study arms. Abbreviations: AS, artesunate; f-up, follow-up.

AS 6 mg/kg/d either as single or split daily dose, followed by mefloquine 25 mg/kg divided over 2 days (Figure 1). Figure 1. Enrollment, randomization, and follow-up of the patients in the 2 study sites. Please see the Methods section for a description of the study arms. Abbreviations: AS, artesunate; f-up, follow-up. Figure 2. Log-linear parasite clearance curves expressed as percentage of admission parasitemia. A, Compares the 4 treatment arms between the 2 study sites in Pailin in western Cambodia and Wang Pha in northwestern Thailand. B, Compares single versus twice-daily dosing between the 2 study sites. Please see the Methods section for a description of the study arms. Randomization was computer generated in blocks of 20. Opaque sealed envelopes were opened sequentially by a study investigator and contained unique study numbers referring to treatment allocations. AS (50-mg tablets; Guilin Pharmaceutical Co, Guangxi, China; repacked by Atlantic Laboratories Corp, Bangkok, Thailand) and mefloquine (250-mg tablets; Atlantic Laboratories Corp, Bangkok, Thailand) doses were calculated according to body weight. Mefloquine was rounded to the nearest quarter tablet and AS to the nearest half tablet in the single daily dose arm and to the nearest quarter tablet in the split-dose arm. Drug administration was directly supervised. The full dose was repeated if vomiting occurred within 30 minutes after drug administration, and half of the dose was given if within 30–60 minutes.

tablet and AS to the nearest half tablet in the single daily dose arm and to the nearest quarter tablet in the split-dose arm. Drug administration was directly supervised. The full dose was repeated if vomiting occurred within 30 minutes after drug administration, and half of the dose was given if within 30–60 minutes. Study Procedures Patients were hospitalized for 7 days and then followed weekly until day 63 after enrollment. A detailed history and physical examination were performed and blood samples were taken for baseline biochemistry and hematology. Parasitemia was assessed at 0, 2, 4, 6, 8, and 12 hours after enrollment and then every 6 hours until 2 consecutive slides were negative. Quality checks were performed on 10% of randomly selected slides. Plasma concentrations of AS and dihydroartemisinin (DHA) were assessed at 0 (predose), 0.25, 0.5, 1, 1.5, 2, 3, 4, 5, 6, and 8 hours after the first drug administration and after the first dose of the last treatment day (day 3 or 7 depending on treatment arm). Recurrent P. falciparum infections were treated with quinine combined with doxycycline or clindamycin (children). Adverse events were recorded on a standard form.

5, 2, 3, 4, 5, 6, and 8 hours after the first drug administration and after the first dose of the last treatment day (day 3 or 7 depending on treatment arm). Recurrent P. falciparum infections were treated with quinine combined with doxycycline or clindamycin (children). Adverse events were recorded on a standard form. Endpoints The primary endpoint was the parasitemia half-life (hereafter stated as “half-life”) defined by the slope of the loge transformed parasite clearance curves [9]. The parasite clearance time (PCT) was defined as time of the first of 2 consecutive negative blood smears. Parasite reduction ratios (PRRs) were calculated as 100 minus the percentage reduction from baseline. PC50, PC90, and PC99 were the times needed to reduce parasitemia with 50%, 90%, and 99%, respectively, of the baseline value. Fever clearance times were defined as the time to first temperature reading <37.5°C (FCTa) and time to defervescence lasting ≥24 hours (FCTb). Efficacy outcomes at day 63, including early treatment failure, late treatment failure, and adequate clinical and parasitological response (ACPR), were classified according to standard definitions [13]. Reinfection was distinguished from recrudescent infections by standard methods using genotyping of the polymorphic genes msp-1, msp-2, and glurp [14].

uding early treatment failure, late treatment failure, and adequate clinical and parasitological response (ACPR), were classified according to standard definitions [13]. Reinfection was distinguished from recrudescent infections by standard methods using genotyping of the polymorphic genes msp-1, msp-2, and glurp [14]. Pharmacokinetics Whole blood samples were collected in prechilled fluoride-oxalate tubes and immediately centrifuged at 4°C. Plasma was stored in liquid nitrogen until analysis at the Mahidol Oxford Research Unit Pharmacology Laboratory in Bangkok. AS and DHA were extracted from plasma samples using solid phase extraction and quantified by high-performance liquid chromatography with tandem mass spectrometry [15]. Individual plasma concentration–time profiles were evaluated with noncompartmental analysis in WinNonlin version 5.0 (Pharsight Corporation). Total exposure up to the last measured concentration was calculated using the linear trapezoidal method for ascending concentrations and the logarithmic trapezoidal method for descending concentrations. The terminal elimination half-life was estimated by log-linear best-fit regression of the observed concentrations in the terminal elimination phase. Complete in vivo conversion of AS into DHA was assumed.

apezoidal method for ascending concentrations and the logarithmic trapezoidal method for descending concentrations. The terminal elimination half-life was estimated by log-linear best-fit regression of the observed concentrations in the terminal elimination phase. Complete in vivo conversion of AS into DHA was assumed. Statistical Analysis A sample size of 40 patients per arm in split versus single daily dosing of AS in artemisinin-resistant malaria was calculated to be sufficient with α = .05 and β = .80 to detect a shortening in half-life of 1.2 hours from a mean of 6.0 (SD, 1.9) hours as observed with once-daily dosing in recent studies in Pailin [2]. Normally distributed continuous data were compared using Student t test or analysis of variance, otherwise Mann-Whitney U or Kruskal-Wallis tests were used. For categorical variables, proportions were examined using χ2 or Fisher exact test. Rates of gametocyte carriage were calculated as total carriage time per person-weeks of follow-up. Efficacy rates were assessed by survival analysis using the Kaplan-Meier method. The rate of increase in reticulocyte counts (up to day 7) and recovery (from day 7 to day 63) was assessed using a random effects model. Analysis was performed using Stata software, version 11.2 (StataCorp).

per person-weeks of follow-up. Efficacy rates were assessed by survival analysis using the Kaplan-Meier method. The rate of increase in reticulocyte counts (up to day 7) and recovery (from day 7 to day 63) was assessed using a random effects model. Analysis was performed using Stata software, version 11.2 (StataCorp). RESULTS A total of 79 patients in Pailin and 80 in Wang Pha were studied between March 2008 and December 2010 (Figure 1). In total, 16 patients (10%), equally divided between sites, were lost to follow-up after discharge from hospital. Severe malaria developed shortly after admission in 1 patient in Pailin and 2 in Wang Pha. One patient in Wang Pha withdrew consent after enrollment. Baseline characteristics are shown in Table 1 and in Supplementary Table 1. Because of the difference in inclusion criteria, in Pailin 22 of 79 (28%) of patients were ≤18 years. Table 1. Baseline Characteristics of Enrolled Patients According to Treatment Arma and Study Site

Wang Pha withdrew consent after enrollment. Baseline characteristics are shown in Table 1 and in Supplementary Table 1. Because of the difference in inclusion criteria, in Pailin 22 of 79 (28%) of patients were ≤18 years. Table 1. Baseline Characteristics of Enrolled Patients According to Treatment Arma and Study Site Pailin, Cambodia Wang Pha, Thailand P Value Characteristic AS7 (n = 25) AS7_split (n = 25) 8MAS3(n = 14) 8MAS3_split (n = 15) AS7 (n = 20) AS7_split(n = 21) 8MAS3(n = 19) 8MAS3_split(n = 20) Pailin vsWang Pha Male sex, No. (%) 23 (92) 22 (88) 11 (79) 10 (67) 19 (95) 19 (90) 18 (95) 18 (90) .08 Age, y .003 Mean (SD) 28 (11) 24 (9) 24 (13) 20 (8) 28 (8) 27 (6) 29 (7) 33 (12) Weight, kg .18 Median 53 51 46 49 52 52 51 52 IQR 50–56 48–54 28–53 25–54 49–55 49–57 48–54 47–56 Temperature, °C .0001 Mean (SD) 38.4 (1.0) 38.6 (1.0) 38.5 (0.9) 38.8 (1.0) 38.1 (1.0) 37.4 (0.6) 37.5 (0.6) 37.8 (0.8) Creatinine, mg/dL .0017 Median 1.00 1.00 0.80 0.90 1.10 1.00 1.00 1.00 IQR 0.80–1.20 0.80–1.10 0.80–1.10 0.80–1.10 0.90–1.25 0.95–1.10 0.90–1.20 0.90–1.20 Alanine aminotransferase, U/L .0001 Median 23 24 21 22 14 11 17 18 IQR 20–33 19–29 16–25 19–26 8–25 8–21 9–38 10–26 Alkaline phosphatase, U/L .0004 Median 74 91 105 105 69 60 70 81 IQR 66–95 83–110 61–159 69–131 64–82 57–78 66–110 69–91 White cell count, ×10³/μL .04 Median 5.9 6.1 6 6.3 5.9 4.5 5.5 4.9 IQR 4.8–7.2 5.1–7.7 5.2–6.7 5.2–8.2 3.9–8.3 3.4–7.0 5.0–7.3 3.8–6.9 Platelet count, ×10³/μL .017 Median 106 118 119 93 96 69 83 85 IQR 85–131 96–140 68–173 75–133 64–128 54–145 61–153 51–120 Parasite density, parasites/μL .18 Geometric mean 54 539 52 237 47 356 58 393 32 507 36 244 58 294 42 541 95% CI 34 151–87 097 33 553–81 324 34 068–65 827 34 299–99 412 18 962–55 727 21 906–59 964 32 295–10 5225 27 240–66 437 Presence of gametocytes, No. (%) 4 (16) 5 (20) 1 (7) 3 (20) 1 (5) 3 (14) 6 (32) 7 (35) .54 Abbreviations: CI, confidence interval; IQR, interquartile range; SD, standard deviation.

6 244 58 294 42 541 95% CI 34 151–87 097 33 553–81 324 34 068–65 827 34 299–99 412 18 962–55 727 21 906–59 964 32 295–10 5225 27 240–66 437 Presence of gametocytes, No. (%) 4 (16) 5 (20) 1 (7) 3 (20) 1 (5) 3 (14) 6 (32) 7 (35) .54 Abbreviations: CI, confidence interval; IQR, interquartile range; SD, standard deviation. a Patients were randomly allocated to 1 of 4 treatment arms: (1) AS alone in a dose of 6 mg/kg/d for 7 days (AS7); (2) the same total dose, but given as a split twice-daily dose (AS7_split); (3) AS in a dose of 8 mg/kg/d for 3 days, followed by mefloquine in a dose of 15 mg/kg on day 3 and 10 mg/kg on day 4 (8MAS3); (4) the same total dose, but AS given as a split twice-daily dose (8MAS3_split). Arms AS7 and AS7_split were suspended after an association with neutropenia was reported in a separate study [12]. The 7-day regimens using AS 6 mg/kg/d were then replaced by 3 days of AS 6 mg/kg/d either as single or split daily dose, followed by mefloquine 25 mg/kg divided over 2 days.

twice-daily dose (8MAS3_split). Arms AS7 and AS7_split were suspended after an association with neutropenia was reported in a separate study [12]. The 7-day regimens using AS 6 mg/kg/d were then replaced by 3 days of AS 6 mg/kg/d either as single or split daily dose, followed by mefloquine 25 mg/kg divided over 2 days. Parasitological Responses Overall, the median half-life was 6.03 (interquartile range [IQR], 4.89–7.28) hours in Pailin versus 3.42 (IQR, 2.20–4.85) hours in Wang Pha (P = .0001). Within sites, there was no difference in half-life between patients receiving 6 or 8 mg/kg per day: median 6.0 (IQR, 4.9–7.5) versus 6.1 (IQR, 4.8–7.1) hours (P = .90) in Pailin and 3.2 (IQR, 2.2–5.1) versus 3.4 (IQR, 2.2–4.8) hours (P = .90) in Wang Pha (Figure 2). In Pailin half-life tended to be shorter with a split daily dose—median 6.5 (IQR, 4.9–7.4) versus 5.4 (IQR, 4.6–6.8) hours—but the difference was not statistically significant (P = .26, Figure 3). In Wang Pha no difference in half-life was observed with a split dose: median 3.4 (IQR, 2.2–5.0) versus 3.3 (IQR, 2.3–4.6) hours (P = .89). Figure 3. A, Parasitemia half-life according to study arm and study site. B, Parasite clearance time according to study arm and study site. Abbreviation: IQR, interquartile range.

In Wang Pha no difference in half-life was observed with a split dose: median 3.4 (IQR, 2.2–5.0) versus 3.3 (IQR, 2.3–4.6) hours (P = .89). Figure 3. A, Parasitemia half-life according to study arm and study site. B, Parasite clearance time according to study arm and study site. Abbreviation: IQR, interquartile range. Other measures of parasite clearance (PRR24, PRR48, PC50, PC90, PC99) confirmed the markedly slower parasite clearance in Pailin compared to Wang Pha (Table 2). The overall median PCT was 78 hours (IQR, 60–96) in Pailin versus 48 hours (IQR, 36–72) in Wang Pha (P = .0001) and correlated with admission parasitemia in both sites (Pailin: Kendall τ = 0.25, P = .001 and Wang Pha: Kendall τ = 0.40, P ≤ .001). Corresponding to this, at 48 hours 66 of 78 (85%) patients remained parasite positive (PPR) in Pailin versus 32 of 77 (42%) in Wang Pha (P < .001). At 72 hours the PPR was 41 of 78 (53%) in Pailin compared to 17 of 77 (22%) in Wang Pha (P < .001). Table 2. Clinical and Parasitological Responses in the Study Patientsa, According to Location

, at 48 hours 66 of 78 (85%) patients remained parasite positive (PPR) in Pailin versus 32 of 77 (42%) in Wang Pha (P < .001). At 72 hours the PPR was 41 of 78 (53%) in Pailin compared to 17 of 77 (22%) in Wang Pha (P < .001). Table 2. Clinical and Parasitological Responses in the Study Patientsa, According to Location Pailin, Cambodia P Value Wang Pha, Thailand P Value Variable AS7 AS7_split 8MAS3 8MAS3_split Singlevs Split AS7 AS7_split 8MAS3 8MAS3_split Singlevs Split Pailin vs Wang Pha Patients, No. 25 25 14 14 20 20 19 18 Parasite clearance time, h .15 .76 .0001 Median 78 66 84 78 48 48 48 51 IQR 66–90 54–96 72–96 54–84 36–78 36–66 36–72 42–66 Time to 50% clearance of parasite density (PC50), h .16 .99 .003 Median 10 9 9 8 8 6 5 7 IQR 8–15 5–11 3–16 4–14 5–9 3–10 2–11 3–9 Time to 90% clearance of parasite density (PC90), h .19 .69 .0001 Median 25 18 26 22 15 13 16 14 IQR 19–30 15–28 17–32 16–29 12–21 9–19 6–22 10–19 Time to 99% clearance of parasite density (PC99), h .17 .82 .0001 Median 49 37 44 45 27 24 27 25 IQR 41–55 32–51 37–56 37–48 20–37 18–32 12–44 22–36 Parasite reduction ratio (PRR) After 24 h (PRR24) .47 .64 .0001 Median 0.85 0.9 0.91 0.89 0.99 0.99 0.99 0.99 IQR 0.77–0.94 0.81–0.97 0.76–0.98 0.83–0.98 0.96–1.00 0.96–1.00 0.94–1.00 0.94–1.00 After 48 h (PRR48) .24 .89 .0001 Median 0.98 1 0.99 0.99 1 1 1 1 IQR 0.97–1.00 0.98–1.00 0.93–1.00 0.97–1.00 1.00–1.00 1.00–1.00 1.00–1.00 1.00–1.00 Patient parasitemic, No. (%) At 24 h 24/25 (96) 25/25 (100) 14/14 (100) 14/14 (100) 1 18/20 (90) 16/20 (80) 18/19 (95) 18/18 (100) .71 .03 At 48 h 20/25 (80) 21/25 (84) 13/14 (93) 12/14 (86) 1 09/20 (45) 07/20 (35) 07/19 (37) 9/18 (50) 1 <.001 At 72 h 14/25 (56) 10/25 (40) 09/14 (64) 08/14 (57) .26 05/20 (25) 04/20 (20) 04/19 (21) 04/18 (22) .83 <.001 Parasitemia half life, h .26 .89 .0001 Median 6.5 5.2 5.9 6.2 3.8 3.1 3.2 3.6 IQR 4.9–7.4 4.9–6.7 5.0–7.1 4.6–6.8 2.4–4.8 2.0–5.1 1.9–4.8 2.3–4.3 Fever clearance (temp <37.5°C) Time to first occurrence, h .7 .5 .0005 Median 18 12 15 15 8 0 4 4 IQR 0–24 6–24 0–24 6–30 0–18 0–12 0–18 0–12 Time to first 24-h period, h .23 .52 .0005 Median 36 30 33 30 18 24 24 30 IQR 24–48 18–48 30–54 30–54 8–36 0–36 18–36 18–36 Recrudescence, No. (%) 4 (16) 0 0 0 2 (10) 1 (5) 2 (10) 0 .68c Reinfection, No. (%) 1 (4) 2 (8) 0 1 (7) 0 0 0 0 Recurrent infection, No.

2 15 15 8 0 4 4 IQR 0–24 6–24 0–24 6–30 0–18 0–12 0–18 0–12 Time to first 24-h period, h .23 .52 .0005 Median 36 30 33 30 18 24 24 30 IQR 24–48 18–48 30–54 30–54 8–36 0–36 18–36 18–36 Recrudescence, No. (%) 4 (16) 0 0 0 2 (10) 1 (5) 2 (10) 0 .68c Reinfection, No. (%) 1 (4) 2 (8) 0 1 (7) 0 0 0 0 Recurrent infection, No. (%) 5 (20) 2 (8) 0 1 (7) 5 (25) 3 (15) 2 (10) 0 .55c Duration of gametocyte carriageb, days .34 0.16 .72 Median 5.5 2 4 6 11 2 9.5 6 IQR 3.5–11 2–4 4–4 1–22 11–11 1–3 2–10 1–10 Abbreviation: IQR, interquartile range. a Patients were randomly allocated to 1 of 4 treatment arms: (1) AS alone in a dose of 6 mg/kg/d for 7 days (AS7); (2) the same total dose, but given as a split twice-daily dose (AS7_split); (3) AS in a dose of 8 mg/kg/d for 3 days, followed by mefloquine in a dose of 15 mg/kg on day 3 and 10 mg/kg on day 4 (8MAS3); (4) the same total dose, but AS given as a split twice-daily dose (8MAS3_split). Arms AS7 and AS7_split were suspended after an association with neutropenia was reported in a separate study [12]. The 7-day regimens using AS 6 mg/kg/d were then replaced by 3 days of AS 6 mg/kg/d either as single or split daily dose, followed by mefloquine 25 mg/kg divided over 2 days. b Duration of gametocyte carriage is reported only for patients who had gametocytemia at any time. c P value from log-rank test. The parasitemia half-life is directly proportional to the clearance rate constant describing the slope of the log-linear parasite clearance curve, a measure used in earlier publications (half-life = 0.693/parasite clearance rate).

b Duration of gametocyte carriage is reported only for patients who had gametocytemia at any time. c P value from log-rank test. The parasitemia half-life is directly proportional to the clearance rate constant describing the slope of the log-linear parasite clearance curve, a measure used in earlier publications (half-life = 0.693/parasite clearance rate). The mean duration of patent gametocytemia per person-weeks of follow-up was 0.018 (95% confidence interval [CI], 0.014–0.022) person-weeks in Pailin and 0.029 (95% CI, 0.024–0.035) in Wang Pha (P = .0005). Thirteen of 79 (16%) patients in Pailin and 17 of 80 (21%) in Wang Pha had gametocytemia at some point during the study (P = .54). In gametocytemic patients, the median duration of gametocyte carriage was 4 (IQR, 2–7) days in Pailin and 3 (IQR, 2–10) days in Wang Pha (P = .72). Gametocyte carriage in patients presenting with gametocytemia was overall shorter in patients treated with a split dose of AS (median, 2.5 [IQR, 1–7] days, n = 18) compared to a single daily dose (median, 8 [IQR, 3.5–10.5] days, n = 12; P = .045).

(IQR, 2–7) days in Pailin and 3 (IQR, 2–10) days in Wang Pha (P = .72). Gametocyte carriage in patients presenting with gametocytemia was overall shorter in patients treated with a split dose of AS (median, 2.5 [IQR, 1–7] days, n = 18) compared to a single daily dose (median, 8 [IQR, 3.5–10.5] days, n = 12; P = .045). Clinical Responses FCT was shorter in patients in Wang Pha compared to Pailin, whereas there was no difference in FCT between single and split daily-dose arms in either site (Table 2). ETF defined as parasitemia and fever ≥72 hours [13] occurred in 6 of 79 patients (8%) in Pailin, compared to 4 of 80 (5%) in Wang Pha (P = .53). Late parasitologic failure, defined as polymerase chain reaction (PCR)–confirmed recrudescence of P. falciparum infection ≥7 days after the start of treatment [13], occurred in 4 of 79 (5%) patients in Pailin after a median of 28 days (range, 28–42 days) and in 5 of 80 patients (6%) in Wang Pha, after a median of 28 days (range, 21–35 days). At day 63, the PCR unadjusted and adjusted Kaplan-Meier estimates for ACPR were 87.9% (95% CI, 78.0%–93.5%) and 93.3% (95% CI, 84.6–97.2%) in Pailin compared to 84.0% (95% CI, 73.6%–90.6%) and 90.4% (95% CI, 81.1–95.4%) in Wang Pha (P = .44 and P = .51, respectively). There was no difference in ACPR after implementation of the changes in the study arms receiving 6 mg/kg AS.

estimates for ACPR were 87.9% (95% CI, 78.0%–93.5%) and 93.3% (95% CI, 84.6–97.2%) in Pailin compared to 84.0% (95% CI, 73.6%–90.6%) and 90.4% (95% CI, 81.1–95.4%) in Wang Pha (P = .44 and P = .51, respectively). There was no difference in ACPR after implementation of the changes in the study arms receiving 6 mg/kg AS. Nadir in hematocrit level was reached at day 7 in both sites, with a mean daily decrease of 0.45% per day (95% CI, .26%–.63%) in Pailin and 0.50% (95% CI, .40%–.60%) in Wang Pha. This initial drop in hematocrit level was not correlated significantly with reticulocyte counts. Reticulocyte counts on admission and days 6–7 were available from 36 patients in Pailin and 33 in Wang Pha. Compared to admission, there was a median fractional reduction after 5 days of 66.7% (IQR, 50%–82%) in Pailin and 66.7% (IQR, 0%–80%) in Wang Pha. In both sites this was not different between patients receiving 6 mg/kg/d versus 8 mg/kg/d AS (P = .54 and P = .87, respectively). Adjusted for baseline hematocrit level, recovery of reticulocyte counts after day 6 was faster in Pailin (4% per day [95% CI, 2%–5%]) compared to Wang Pha (1% [95% CI, 2%–5%]) and counts were maximal at day 14 in both sites.

t between patients receiving 6 mg/kg/d versus 8 mg/kg/d AS (P = .54 and P = .87, respectively). Adjusted for baseline hematocrit level, recovery of reticulocyte counts after day 6 was faster in Pailin (4% per day [95% CI, 2%–5%]) compared to Wang Pha (1% [95% CI, 2%–5%]) and counts were maximal at day 14 in both sites. The number of patients reporting minor adverse events were more frequent in Pailin (Supplementary Table 2). In Pailin, a 29-year-old man treated with a total of 24 mg/kg of AS (8MAS3) developed transient afebrile neutropenia with a nadir of 0.78 × 103/μL on day 14, which normalized on day 17. In total, 2 patients in Wang Pha and 1 patient in Pailin developed severe malaria shortly after enrollment and received treatment with intravenous AS and quinine. Of these, a 30-year-old woman in Pailin unfortunately died on the fourth day of hospitalization. DSMC evaluation of the case concluded that severe malaria was the likely direct cause of death.

d 1 patient in Pailin developed severe malaria shortly after enrollment and received treatment with intravenous AS and quinine. Of these, a 30-year-old woman in Pailin unfortunately died on the fourth day of hospitalization. DSMC evaluation of the case concluded that severe malaria was the likely direct cause of death. Pharmacokinetics The pharmacokinetic profiles of AS and DHA were comparable between sites (Table 3 and Figure 4). However, a relative small but significant higher maximum concentration was observed for both AS and DHA in Wang Pha. With increased dosing, there was a proportional increase in the maximum concentration (Cmax) and exposure (AUC0-∞) to both drugs (Figure 4). Compared to admission, day 7 elimination clearance values were higher and drug exposure lower for both AS (P = .0003) and DHA (P = .0001). There was no correlation between half-life and the individual exposure or maximum concentration of AS or DHA or both drugs combined. Pharmacokinetic parameters were otherwise similar to that previously reported in the same populations [2]. Table 3. Pharmacokinetic Parameter Estimates From the Noncompartmental Analysis of the First Dose According to Study Site

e and the individual exposure or maximum concentration of AS or DHA or both drugs combined. Pharmacokinetic parameters were otherwise similar to that previously reported in the same populations [2]. Table 3. Pharmacokinetic Parameter Estimates From the Noncompartmental Analysis of the First Dose According to Study Site Wang Pha (n = 79),Median (Range) Pailin (n = 40),Median (Range) P Valuea Body weight (kg) 51.0 (39.0–69.0) 52.0 (24.0–64.9) .2636 Total artesunate dose (mg/kg)a 4.25 (2.88–8.19) 5.56 (2.86–8.33) .2777 Artesunate Cmax (ng/mL) 533 (71.1–3330) 341 (44.4–3610) .0153 Cmax/dose ([ng/mL]/mg) 1.92 (0.203–8.83) 1.35 (0.222–15.6) .0474 Tmax (h) 0.500 (0.180–5.03) 0.983 (0.217–4.98) .2296 CL/F (L/h) 645 (290–2950) 659 (136–2250) .5699 V/F (L) 255 (65.8–3720) 358 (40.1–2430) .0434 T1/2 (h) 0.279 (0.0900–2.21) 0.392 (0.125–2.73) .0135 AUC0-∞ (h × ng/mL) 386 (102–1230) 375 (77.7–1470) .3006 AUC0-∞/dose ([h × ng/mL]/mg) 1.55 (0.339–3.45) 1.51 (0.444–7.35) .4925 Dihydroartemisinin Cmax (ng/mL) 2430 (583–8160) 1960 (427–4050) .0130 Cmax/dose ([ng/mL]/mg) 14.2 (3.54–31.7) 10.3 (2.60–27.3) .0299 Tmax (h) 1.00 (0.500–5.03) 1.48 (0.483–4.98) .8077 CL/F (L/h) 37.1 (15.0–82.3) 41.8 (15.8–91.3) .2442 V/F (L) 45.9 (18.0–185) 55.7 (13.1–166) .0802 T1/2 (h) 0.842 (0.426–2.77) 0.836 (0.442–2.08) .6896 AUC0-∞ (h × ng/mL) 4540 (1440–19700) 4680 (1260–12900) .1630 AUC0-∞/dose ([h × ng/mL]/mg) 26.9 (12.2–66.7) 23.9 (7.88–63.5) .2330 Abbreviations: AUC0-∞, predicted area under the plasma concentration time curve after the first dose from zero time to infinity; CL, elimination clearance; Cmax, maximum observed plasma concentration after oral administration; F oral bioavailability; Tmax, observed time to reach Cmax; T1/2, terminal elimination half-life; V, apparent volume of distribution.

nder the plasma concentration time curve after the first dose from zero time to infinity; CL, elimination clearance; Cmax, maximum observed plasma concentration after oral administration; F oral bioavailability; Tmax, observed time to reach Cmax; T1/2, terminal elimination half-life; V, apparent volume of distribution. a P values are given using the Mann-Whitney test. b The median dose of artesunate is lower in Pailin compared to Wang Pha because of discontinuation of the 8 mg/kg/d arm in Pailin (see Figure 1). Figure 4. Total artesunate (A) and dihydroartemisinin (B) exposure and maximum artesunate (C) and dihydroartemisinin (D) concentrations in Pailin (filled squares) and Wang Pha (open circles) after the first dose in the different treatment arms (3 mg/kg and 4 mg/kg are twice-daily dose groups and 6 mg/kg and 8 mg/kg are once-daily dose groups). All exposure measurements are dose (mg) normalized. Abbreviations: AUC0-∞, predicted area under the plasma concentration time curve after the first dose from zero time to infinity; Cmax, maximum observed plasma concentration after oral administration.

ily dose groups and 6 mg/kg and 8 mg/kg are once-daily dose groups). All exposure measurements are dose (mg) normalized. Abbreviations: AUC0-∞, predicted area under the plasma concentration time curve after the first dose from zero time to infinity; Cmax, maximum observed plasma concentration after oral administration. DISCUSSION This study evaluated the use of an increased AS dose and an increased dosing frequency (ie, a split dose) in uncomplicated falciparum malaria with varying degrees of resistance to artemisinins. The common dose of AS is 4 mg/kg/d as a once-daily dose for 3 days as part of an ACT. This study showed that increasing the AS dose up to 8 mg/kg/d did not affect parasite clearance parameters, strongly suggesting a reduction in the maximum obtainable effect [6, 16]. There was a slight trend that splitting the AS dose in a twice-daily administration did accelerate parasite clearance in artemisinin-resistant malaria in Pailin, but this did not reach statistical significance. A mathematical model predicted a larger effect on half-life with increasing dosing frequency than observed in the current study [8]. This could be related to reduced sensitivity of not only ring-stage parasites (as assumed in the model), but also to some extent in more mature stages. Gametocyte clearance in patients presenting with gametocytemia was slightly faster with split dosing of AS, but the significance of this finding will need further study.

his could be related to reduced sensitivity of not only ring-stage parasites (as assumed in the model), but also to some extent in more mature stages. Gametocyte clearance in patients presenting with gametocytemia was slightly faster with split dosing of AS, but the significance of this finding will need further study. Further increase of the total dose of AS is limited by toxicity. A study from western Cambodia has recently shown that AS in a dose of 6 mg/kg/d for 7 days resulted in transient neutropenia with a nadir at 2 weeks after treatment. Severe neutropenia (<1.0 × 103/μL) was observed in 19% of patients [12]. The findings prompted the current study to amend the dosing schedules in arm 1 and 2. In the current study, one patient treated with 8 mg/kg/d for 3 days had grade 3 transient neutropenia below 1.0 × 103/μL, with no signs of concomitant infection [17]. Although transient, this dose-related neutropenia restricts increasing the currently recommended AS dose [12]. The study also showed transient reticulocytopenia that was possibly partly attributable to AS therapy, which could delay recovery of anemia. Reticulocytopenia was not dose dependent. This drug class–related effect has also been described in patients with severe malaria treated with artemisinin, artemether, and artesunate [18, 19], as well as in animal models [20].

nia that was possibly partly attributable to AS therapy, which could delay recovery of anemia. Reticulocytopenia was not dose dependent. This drug class–related effect has also been described in patients with severe malaria treated with artemisinin, artemether, and artesunate [18, 19], as well as in animal models [20]. Compared to our study in 2007–2008, there was an increase in the PPR at 72 hours (from 8% to 22%) in Wang Pha, northwestern Thailand [2]. The emergence of artemisinin resistance in northwestern Thailand was recently confirmed in a longitudinal study showing a progressive reduction in parasite clearance rates from 2001 to 2010 [4]. In the current study, the distribution of peripheral blood parasite half-life in Wang Pha did not have a unimodal shape, suggesting a separate population with prolonged parasite clearance. The emergence of artemisinin resistance on the Myanmar-Thailand border is evidently very worrying. In the current study, an unexpected 3 of 159 (1.9%) patients with uncomplicated malaria developed severe malaria after start of antimalarial treatment, an unusual sequence in susceptible infections. Whether AS-resistant falciparum malaria is indeed associated with a larger proportion of patients developing severe disease under ACT requires further study.

3 of 159 (1.9%) patients with uncomplicated malaria developed severe malaria after start of antimalarial treatment, an unusual sequence in susceptible infections. Whether AS-resistant falciparum malaria is indeed associated with a larger proportion of patients developing severe disease under ACT requires further study. Recrudescence rates, which strongly depend on parasite sensitivity to the partner drug, were not different between treatment arms and sites. Despite marked slower parasite clearance in Pailin, the AS-mefloquine combination was borderline efficacious in both sites with PCR-adjusted cure rates at day 63 of 93.3% in Pailin and 90.4% in Wang Pha. In the current study with a relatively small sample size, there was no clear relationship between gametocyte carriage and asexual stage resistance to AS. A larger study is under way to address this important determinant of transmissibility [21, 22].

djusted cure rates at day 63 of 93.3% in Pailin and 90.4% in Wang Pha. In the current study with a relatively small sample size, there was no clear relationship between gametocyte carriage and asexual stage resistance to AS. A larger study is under way to address this important determinant of transmissibility [21, 22]. In agreement with previous studies, considerable interindividual variation was observed in the pharmacokinetic profiles for AS and the active metabolite DHA, but there were no major differences between sites. There was a higher maximum concentration (Cmax) of AS and DHA in Wang Pha compared to Pailin, but individual parasite clearance times were not correlated with Cmax. Total drug exposure was not different between sites. Differences in half-life between western Cambodia and northwestern Thailand can thus not be explained by differences in pharmacokinetics. There was a proportional increase in exposure with increased dosing of AS, which supports dose-independent absorption and clearance of AS and DHA. Doubling the frequency of dosing proportionately increased the duration of plasma parasiticidal drug concentrations, but this did not significantly accelerate parasite half-life. The marked decrease in total exposure for AS and DHA on day 7 compared to admission is likely to be a disease-related effect resulting from an increase in relative bioavailability during the acute phase of the disease. This has previously been reported in pregnant women on the Thai-Myanmar border [23].

asite half-life. The marked decrease in total exposure for AS and DHA on day 7 compared to admission is likely to be a disease-related effect resulting from an increase in relative bioavailability during the acute phase of the disease. This has previously been reported in pregnant women on the Thai-Myanmar border [23]. In conclusion, this study suggests that slow parasite clearance in artemisinin-resistant falciparum malaria cannot be mitigated by increasing the daily dose or dose frequency of artesunate. Although ACTs were still effective in 2009–2010, this now heavily depends on efficacy of the partner drug. New classes of drugs are urgently needed because of the threat of untreatable falciparum malaria in the region. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://www.oxfordjournals.org/our_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

r_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We dedicate this paper to Niklas Lindegardh, who established and led the pharmacology department at Mahidol Oxford Research Unit. We thank the staff of the Pailin Referral Hospital and the Shoklo Malaria Research Unit for their excellent work and dedication. We also thank the Village Malaria Workers Network in Pailin; Institute of Pasteur, Cambodia; Family Health International; The National Center for Parasitology, Entomology and Malaria Control, Cambodia; and the World Health Organization for their extended support and collaboration. We are grateful to the DSMC members Ric Price (Chair), Hien Tran Tinh, and Karen I. Barnes. We thank Song Nagh (Family Health International) and Oliver Koch for help in executing the study in Cambodia. Disclaimer. The views expressed in this article are those of the authors and do not necessarily represent the decisions, policy, or views of the World Health Organization. P. R. is a staff member of the WHO. Financial support. This study was funded by the Bill & Melinda Gates Foundation (grant 48821) and was part of the Wellcome Trust–Mahidol University, Oxford Tropical Medicine Research Programme (Major Overseas Programme–Thailand Unit Core Grant).

Disclaimer. The views expressed in this article are those of the authors and do not necessarily represent the decisions, policy, or views of the World Health Organization. P. R. is a staff member of the WHO. Financial support. This study was funded by the Bill & Melinda Gates Foundation (grant 48821) and was part of the Wellcome Trust–Mahidol University, Oxford Tropical Medicine Research Programme (Major Overseas Programme–Thailand Unit Core Grant). Potential conflicts of interest. All authors: No potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Cytomegalovirus (CMV) is a common herpesvirus that infects the majority of the population worldwide. After primary infection, usually an undifferentiated febrile illness, the virus remains latent, under the control of the immune system, and is asymptomatic for the lifetime of the host [1]. Evidence of past CMV infection is the presence of immunoglobulin G (IgG) antibodies for CMV in the peripheral blood (seropositivity). If CMV escapes immunological control and reactivates from latency [1], it can cause severe disease and excess mortality, as has been observed among severely unwell [2] and immunocompromised individuals [3, 4], CMV infection in transplantation has been associated with direct organ damage as well as indirect consequences, such as transplant-associated vasculopathy, which can manifest as coronary artery stenosis and graft rejection, and an increased risk of opportunistic infection [5].

and immunocompromised individuals [3, 4], CMV infection in transplantation has been associated with direct organ damage as well as indirect consequences, such as transplant-associated vasculopathy, which can manifest as coronary artery stenosis and graft rejection, and an increased risk of opportunistic infection [5]. However, experimental and epidemiological data also point to the direction of CMV as a potential factor in the development of cardiovascular disease among immunocompetent individuals [1]. Some studies have reported associations between CMV seropositivity or level of CMV IgG antibody and cardiovascular disease [6–8], cardiovascular or all-cause mortality [9–16], or even cancer [17]. CMV infection has been hypothesized to increase mortality risk through its association with cardiovascular disease and frequent silent CMV reactivations that are driving chronic inflammation [13]. An alternative explanation could be that the reported associations are subject to residual confounding, given the social patterning of CMV infection [18]. CMV is only one of many pathogens that have been in the past examined for their possible associations with cardiovascular diseases. The list of candidates includes other viruses like Coxsackieviruses as well as bacteria; however, CMV remains of major interest because of possible dose effects described above as well as biologic plausibility [19, 20], which raise the possibility that CMV might be causal for increased morbidity and mortality.

iseases. The list of candidates includes other viruses like Coxsackieviruses as well as bacteria; however, CMV remains of major interest because of possible dose effects described above as well as biologic plausibility [19, 20], which raise the possibility that CMV might be causal for increased morbidity and mortality. In the current study, we examine whether seropositivity for CMV and the level of CMV IgG antibody are associated with all-cause and cause-specific mortality among participants in a population-based cohort study and whether markers of inflammation and socioeconomic status confound any observed associations. METHODS Ethics Statement All volunteers gave written informed consent, and the study was approved by the Norfolk Research Ethics Committee. Cohort Characteristics We studied this association in the European Prospective Investigation of Cancer (EPIC)–Norfolk study, a well-described UK population-based cohort [21], which recruited 25 639 men and women between 1993 and 1997, who were aged 40–79 years at baseline. From the participants with available sera, we randomly selected 13 090 for measurement of CMV antibodies.

uropean Prospective Investigation of Cancer (EPIC)–Norfolk study, a well-described UK population-based cohort [21], which recruited 25 639 men and women between 1993 and 1997, who were aged 40–79 years at baseline. From the participants with available sera, we randomly selected 13 090 for measurement of CMV antibodies. Questionnaires and Biochemical and Hematological Analyses Anthropometric measurements were taken, and a health and lifestyle questionnaire that included questions on housing, occupational social class, educational level, use of tobacco, and physical activity was completed at baseline. A validated 4-point ordered categorical physical activity index was used [21]. The participants were asked to report if they had ever been diagnosed with a “heart attack” (myocardial infarction) or stroke or diabetes. Individual data were linked with the East Anglia Cancer Registry database and participants were coded as having had a history of cancer if they had been diagnosed with any cancer except for nonmelanoma skin cancers at study entry. Serum high-sensitivity C-reactive protein (CRP) was measured using a Dade Behring Dimension ARx automated system (Deerfield, Illinois) and ferritin with a 2-step time-resolved fluoroimmunoassay (Wallac Oy, Turku, Finland). All other biochemical and hematological indices were measured using standard assays [21].

Questionnaires and Biochemical and Hematological Analyses Anthropometric measurements were taken, and a health and lifestyle questionnaire that included questions on housing, occupational social class, educational level, use of tobacco, and physical activity was completed at baseline. A validated 4-point ordered categorical physical activity index was used [21]. The participants were asked to report if they had ever been diagnosed with a “heart attack” (myocardial infarction) or stroke or diabetes. Individual data were linked with the East Anglia Cancer Registry database and participants were coded as having had a history of cancer if they had been diagnosed with any cancer except for nonmelanoma skin cancers at study entry. Serum high-sensitivity C-reactive protein (CRP) was measured using a Dade Behring Dimension ARx automated system (Deerfield, Illinois) and ferritin with a 2-step time-resolved fluoroimmunoassay (Wallac Oy, Turku, Finland). All other biochemical and hematological indices were measured using standard assays [21]. Mortality Ascertainment All participants were flagged for death certification at the Office of National Statistics (United Kingdom) with vital status ascertained on the entire cohort, and mortality data for all EPIC-Norfolk participants were available up to 31 March 2011. Coding of death certificates was executed by trained nosologists according to the International Classification of Diseases (ICD), Ninth or Tenth Revision. All deaths with ICD-9 codes were recoded into comparable ICD-10 codes. Only underlying causes of death were used for these analyses. A “cardiovascular death” was defined as a death where the underlying cause of death belonged to ICD-10 codes I10–I79 inclusive and a “cancer death” to codes C00–C97. The cardiovascular codes were selected to reflect postulated CMV pathogenetic mechanisms and included causes of death such as hypertensive heart disease, atherosclerosis including coronary and cerebrovascular disease, heart failure, aortic aneurysms, and thrombosis [22].

inclusive and a “cancer death” to codes C00–C97. The cardiovascular codes were selected to reflect postulated CMV pathogenetic mechanisms and included causes of death such as hypertensive heart disease, atherosclerosis including coronary and cerebrovascular disease, heart failure, aortic aneurysms, and thrombosis [22]. CMV Measurements CMV-specific IgG measurements were performed on sera from 13 090 participants, by the Cambridge University Hospitals Virology Laboratory, using an indirect chemiluminescence immunoassay (Liaison, Diasorin, Saluggia, Italy). The amount of isoluminol-antibody conjugate is measured by a photomultiplier as relative light units (RLUs). The machine, using an internal algorithm, converts RLUs to antibody levels. The coefficient of variation for the assay is <8%, specificity 99.65%, sensitivity 99.88%, and repeatability >98%. The assay compares favorably to other CMV IgG assays for confirmation of past CMV infections [23]. Samples were shipped to the laboratory in randomly ordered boxes, with no identifiers other than a barcode. A sample was defined as being negative, equivocal, or positive for CMV IgG antibody using the clinical antibody cutoffs of the assay (<0.4 IU/mL, 0.4–0.6 IU/mL, and >0.6 IU/mL, respectively). Ninety-one participants had equivocal test results and were excluded. Because of low variable missingness (<5% for all variables), no other exclusions were applied and models were performed on the maximum number possible. Results from 7113 women and 5886 men were available for analysis.

, and >0.6 IU/mL, respectively). Ninety-one participants had equivocal test results and were excluded. Because of low variable missingness (<5% for all variables), no other exclusions were applied and models were performed on the maximum number possible. Results from 7113 women and 5886 men were available for analysis. Statistical Analyses RLUs were standardized by calculating the difference from the mean and dividing by the standard deviation within the day of measurement, to account for any differences in assay performance from day to day. Standardized RLUs among participants with positive tests only were grouped into thirds of the distribution. We summarized baseline characteristics within the cohort using means and standard deviations for continuous variables with an approximately symmetric distribution, medians and interquartile ranges for continuous variables with a skewed distribution (Townsend deprivation index, alcohol consumption, fibrinogen, ferritin, plasma glucose, cholesterol, low-density lipoprotein, high-density lipoprotein, triglycerides), and percentages for binary variables. Among participants with positive tests, a P value for linear trend in the baseline characteristic across thirds of the distribution was calculated using linear regression (continuous variables) or logistic regression (binary variables), adjusted for age and sex. Continuous variables with skewed distributions were log transformed. Similar methods were used to test the difference between participants with positive and negative tests.

the distribution was calculated using linear regression (continuous variables) or logistic regression (binary variables), adjusted for age and sex. Continuous variables with skewed distributions were log transformed. Similar methods were used to test the difference between participants with positive and negative tests. We used Cox proportional hazards regression models to estimate age- and sex-adjusted mortality rates (using the age/sex distribution of our dataset as the standard population) and hazard ratios of death, comparing seropositive vs seronegative individuals, and also comparing thirds of RLU values, with seronegative individuals as the reference group. We also estimated the effect of each exposure on cardiovascular death, cancer-related death, and noncancer, noncardiovascular-related death, separately. The assumption of proportional hazards was tested by plotting and inspecting the relevant Kaplan-Meier survival curves. Within the Cox model we tested for interactions between the exposure of interest (either seropositive vs seronegative or thirds of RLU values) and sex, age at time of entry to the study (continuous), socioeconomic status (manual vs nonmanual employment), physical activity, body mass index (BMI; continuous), and inflammatory markers (ferritin, fibrinogen, and CRP, all continuous). We fit multivariable models adjusting for various potential confounders. We did not assume any mediation effects.

the study (continuous), socioeconomic status (manual vs nonmanual employment), physical activity, body mass index (BMI; continuous), and inflammatory markers (ferritin, fibrinogen, and CRP, all continuous). We fit multivariable models adjusting for various potential confounders. We did not assume any mediation effects. Covariates were chosen as possible confounders (age, sex, Townsend deprivation index, smoking, educational level, physical activity, social class, BMI, waist-to-hip ratio [WHR], total cholesterol, CRP) based on a priori hypotheses only. We additionally performed all of the above analyses (1) including the 91 participants with the equivocal results initially in the seronegative and subsequently in the seropositive group (lower antibody category), and (2) excluding all seronegative individuals and repeating all the analyses within the seropositive group only, using the lower antibody group as the baseline. All statistical analyses were performed using Stata/SE 12.0 (StataCorp, College Station, Texas). RESULTS A total of 59% of the participants were seropositive for CMV, with seropositivity being slightly more common in women (60%) compared to men (57%, χ2 P < .001) and at older ages. Higher CMV IgG antibody levels among seropositive participants were associated with older age and female sex (Table 1). Table 1. Baseline Demographic, Lifestyle, and Comorbidity Characteristics of Participants of the European Prospective Investigation of Cancer–Norfolk Cohort Cytomegalovirus (CMV) Study (N = 12 999), Grouped by CMV Antibody Status

s among seropositive participants were associated with older age and female sex (Table 1). Table 1. Baseline Demographic, Lifestyle, and Comorbidity Characteristics of Participants of the European Prospective Investigation of Cancer–Norfolk Cohort Cytomegalovirus (CMV) Study (N = 12 999), Grouped by CMV Antibody Status Characteristic Negative for CMV Antibodies (n = 5366) Positive for CMV IgG Antibodies P Valueb Low Antibody Group (n = 2545) Middle Antibody Group (n = 2544) High Antibody Group (n = 2544) P Valuea Age at study entry, y, mean (SD) 56.9 (9) 59.0 (9) 59.6 (9) 61.1 (9) <.001 <.001 Female sex 53% 50% 56% 62% <.001 <.001 Education, A levelc and above 58% 52% 52% 46% .26 <.001 Social class, nonmanual employment 63% 58% 60% 58% .97 <.001 Townsend deprivation index, median (IQR) −2.7 (−3.7 to −1.3) −2.6 (−3.6 to −1.0) −2.6 (−3.7 to −1.0) −2.5 (−3.7 to −0.9) .61 <.001 Ever smokers 50% 56% 55% 56% .11 <.001 Alcohol consumption, units/wk, median (IQR) 4.0 (1–10) 4.0 (1–10) 3.5 (1.9) 2.5 (1–8) .08 .14 Physical activity, moderately or very active 44% 41% 41% 37% .73 .25 With prevalent diabetes mellitus 2.5% 3.4% 3.3% 3.9% .58 .04 With prevalent myocardial infarction 2.6% 3.4% 3.7% 4.2% .10 .03 With prevalent stroke 1.2% 1.3% 1.6% 1.8% .42 .87 With prevalent cancer, excluding nonmelanoma skin cancer 4.7% 4.6% 5.6% 5.8% .35 .95 Abbreviations: CMV, cytomegalovirus; IgG, immunoglobulin G; IQR, interquartile range; SD, standard deviation. a Age- and sex-adjusted P value for trend among thirds of antibody within participants with positive IgG antibodies.

Characteristic Negative for CMV Antibodies (n = 5366) Positive for CMV IgG Antibodies P Valueb Low Antibody Group (n = 2545) Middle Antibody Group (n = 2544) High Antibody Group (n = 2544) P Valuea Age at study entry, y, mean (SD) 56.9 (9) 59.0 (9) 59.6 (9) 61.1 (9) <.001 <.001 Female sex 53% 50% 56% 62% <.001 <.001 Education, A levelc and above 58% 52% 52% 46% .26 <.001 Social class, nonmanual employment 63% 58% 60% 58% .97 <.001 Townsend deprivation index, median (IQR) −2.7 (−3.7 to −1.3) −2.6 (−3.6 to −1.0) −2.6 (−3.7 to −1.0) −2.5 (−3.7 to −0.9) .61 <.001 Ever smokers 50% 56% 55% 56% .11 <.001 Alcohol consumption, units/wk, median (IQR) 4.0 (1–10) 4.0 (1–10) 3.5 (1.9) 2.5 (1–8) .08 .14 Physical activity, moderately or very active 44% 41% 41% 37% .73 .25 With prevalent diabetes mellitus 2.5% 3.4% 3.3% 3.9% .58 .04 With prevalent myocardial infarction 2.6% 3.4% 3.7% 4.2% .10 .03 With prevalent stroke 1.2% 1.3% 1.6% 1.8% .42 .87 With prevalent cancer, excluding nonmelanoma skin cancer 4.7% 4.6% 5.6% 5.8% .35 .95 Abbreviations: CMV, cytomegalovirus; IgG, immunoglobulin G; IQR, interquartile range; SD, standard deviation. a Age- and sex-adjusted P value for trend among thirds of antibody within participants with positive IgG antibodies. b Age- and sex-adjusted P value comparing participants with positive IgG antibodies to participants with negative IgG antibodies. c “A levels” corresponds to 12 years of school education in the United Kingdom.

a Age- and sex-adjusted P value for trend among thirds of antibody within participants with positive IgG antibodies. b Age- and sex-adjusted P value comparing participants with positive IgG antibodies to participants with negative IgG antibodies. c “A levels” corresponds to 12 years of school education in the United Kingdom. After age and sex adjustment, positive IgG antibody levels for CMV were associated with markers of lower social status, lifetime exposure to smoking, higher BMI and WHR, higher total cholesterol, and a history of diabetes mellitus and myocardial infarction (Tables 1 and 2). Among seropositive participants and after age and sex adjustment, there was an association between higher IgG antibody levels and higher total cholesterol, lower high-density lipoprotein, and higher triglycerides. Table 2. Baseline Anthropometric and Biochemical Characteristics of Participants of the European Prospective Investigation of Cancer–Norfolk Cohort Cytomegalovirus (CMV) Study (N = 12 999), Grouped by CMV Antibody Status

body levels and higher total cholesterol, lower high-density lipoprotein, and higher triglycerides. Table 2. Baseline Anthropometric and Biochemical Characteristics of Participants of the European Prospective Investigation of Cancer–Norfolk Cohort Cytomegalovirus (CMV) Study (N = 12 999), Grouped by CMV Antibody Status Negative for CMV Antibodies (n = 5366) Positive for CMV IgG Antibodies P Valueb Low Antibody Group (n = 2545) Middle Antibody Group (n = 2544) High Antibody Group (n = 2544) P Valuea CMV IgG antibody, IU/mL, median (IQR) 0.2 (0.2–0.2) 3.0 (1.9–4.1) 6.9 (5.7–8.4) 14.1 (10.7–20.8) <.001 <.001 Body mass index, kg/m2, mean (SD) 26.1 (3.7) 26.4 (3.7) 26.3 (3.9) 26.5 (4.1) .45 <.001 Waist-to-hip ratio, mean (SD) 0.85 (0.1) 0.86 (0.1) 0.86 (0.1) 0.86 (0.1) .29 .02 C-reactive protein, mg/L, mean (SD) 2.9 (5.9) 3.0 (5.6) 3.3 (6.5) 3.5 (7.6) .05 .07 Ferritin, pmol/L, mean (SD) 206.5 (172) 206.4 (170) 204.6 (172) 194.1 (161) .74 .14 Fibrinogen, µmol/L, median (IQR) 8.3 (7.1–9.8) 8.5 (7.1–9.9) 8.5 (7.1–9.9) 8.8 (7.4–9.9) .18 .61 Glucose, mmol/L, median (IQR) 4.0 (3.5–4.7) 4.0 (3.5–4.7) 4.1 (3.5–4.8) 4.1 (3.6–4.8) .26 .77 Cholesterol, mmol/L, median (IQR) 6.0 (5.3–6.8) 6.1 (5.4–6.9) 6.1 (5.4–6.9) 6.1 (5.4–6.9) .07 .15 LDL, mmol/L, median (IQR) 3.8 (3.2–4.6) 3.9 (3.3–4.6) 3.8 (3.2–4.6) 3.9 (3.3–4.6) .13 .63 HDL, mmol/L, median (IQR) 1.4 (1.1–1.7) 1.4 (1.1–1.6) 1.3 (1.1–1.6) 1.4 (1.1–1.7) .001 <.001 Triglycerides, mmol/L, median (IQR) 1.5 (1.0–2.2) 1.5 (1.1–2.2) 1.6 (1.1–2.3) 1.6 (1.1–2.3) .64 <.001 Systolic blood pressure, mm Hg, mean (SD) 134.2 (17) 135.1 (18) 135.9 (18) 137.2 (19) .05 .59 Diastolic blood pressure, mm Hg, mean (SD) 82.1 (11) 82.3 (11) 82.5 (11) 82.9 (11) .09 .89 To obtain cholesterol, LDL, and HDL in mg/dL, divide mmol/L by 0.0259. To obtain triglycerides in mg/dL, divide mmol/L by 0.0113. To obtain glucose in mg/dL, divide mmol/L by 0.0555. To obtain fibrinogen in mg/dL, divide µmol/L by 0.0294. To obtain ferritin in ng/mL, divide pmol/L by 2.247.

82.5 (11) 82.9 (11) .09 .89 To obtain cholesterol, LDL, and HDL in mg/dL, divide mmol/L by 0.0259. To obtain triglycerides in mg/dL, divide mmol/L by 0.0113. To obtain glucose in mg/dL, divide mmol/L by 0.0555. To obtain fibrinogen in mg/dL, divide µmol/L by 0.0294. To obtain ferritin in ng/mL, divide pmol/L by 2.247. Abbreviations: CMV, cytomegalovirus; HDL, high-density lipoprotein; IgG, immunoglobulin G; IQR, interquartile range; LDL, low-density lipoprotein; SD, standard deviation. a Age- and sex-adjusted P value for trend among thirds of antibody within participants with positive IgG antibodies. b Age- and sex-adjusted P value comparing participants with positive IgG antibodies to participants with negative IgG antibodies.

Abbreviations: CMV, cytomegalovirus; HDL, high-density lipoprotein; IgG, immunoglobulin G; IQR, interquartile range; LDL, low-density lipoprotein; SD, standard deviation. a Age- and sex-adjusted P value for trend among thirds of antibody within participants with positive IgG antibodies. b Age- and sex-adjusted P value comparing participants with positive IgG antibodies to participants with negative IgG antibodies. Mortality After a mean of 14.3 years (SD, 3.3) of follow-up, 2514 deaths occurred. Age- and sex-adjusted mortality rates were 12.4 (95% confidence interval [CI], 11.3–13.2) per 1000 person-years at risk among seronegative participants and 14.2 (95% CI, 13.5–15.0) among seropositive participants. Among seropositive participants, rates increased across thirds of IgG antibody levels (score test of trend for rates P < .0001). A total of 851 deaths were attributed to cardiovascular diseases and 955 to cancer, and 708 had an underlying cause other than cancer or cardiovascular disease. Of the noncardiovascular, noncancer-related deaths, 12% (n = 76) had respiratory diseases, 16% (n = 100) had gastrointestinal diseases, and 21% (n = 130) had central and peripheral nervous system diseases coded as the underlying cause. The rest of the deaths were attributed to various causes including infection, kidney, blood, joint, endocrine, and rheumatologic diseases as well as alcohol abuse or intoxication and accidents.

gastrointestinal diseases, and 21% (n = 130) had central and peripheral nervous system diseases coded as the underlying cause. The rest of the deaths were attributed to various causes including infection, kidney, blood, joint, endocrine, and rheumatologic diseases as well as alcohol abuse or intoxication and accidents. Associations Between CMV and All-Cause Mortality CMV seropositivity was associated with all-cause mortality, after age and sex adjustment as well as after adjustment for socioeconomic factors and BMI, WHR, total cholesterol, and CRP (Table 3). Hazard ratios for mortality were higher for the group of participants with the highest antibody levels for all of the models examined (Table 3). Hazard ratios for mortality were similar after excluding participants who died within 2 years from study entry or had prevalent myocardial infarction, cerebrovascular accident, or cancer at study entry (Table 3). No interactions with sex, age, socioeconomic status, physical activity, BMI, and inflammatory markers were identified. A statistically significant association was noted between death not attributed to cancer and cardiovascular disease and CMV seropositivity (Table 4). Hazard ratios for mortality were identical to those in Tables 3 and 4 when the 91 participants with the equivocal results were included in the seronegative or in the seropositive lower antibody category group. Limiting the analyses within the seropositive participants only and using the low antibody group as the reference category confirmed that participants with higher antibody groups had higher hazard ratios for all-cause and cause-specific mortality (Supplementary Tables 1 and 2). Table 3. Hazard Ratios for All-Cause Mortality

ing the analyses within the seropositive participants only and using the low antibody group as the reference category confirmed that participants with higher antibody groups had higher hazard ratios for all-cause and cause-specific mortality (Supplementary Tables 1 and 2). Table 3. Hazard Ratios for All-Cause Mortality Model Model Adjustments No. All With Positive CMV IgG Antibodies Low Antibody Group Middle Antibody Group High Antibody Group Model 1 Age, sex 12 999 1.16 (1.07–1.26) 1.09 (.98–1.22) 1.14 (1.02–1.27) 1.26 (1.13–1.39) Model 2 Age, sex, Townsend, smoking, educational level, physical activity, social class 12 612 1.14 (1.04–1.24) 1.07 (.95–1.20) 1.12 (.99–1.25) 1.23 (1.10–1.37) Model 3 Age, sex, Townsend, smoking, educational level, physical activity, social class, BMI, WHR, total cholesterol 12 425 1.15 (1.05–1.25) 1.08 (.96–1.21) 1.13 (1.01–1.27) 1.24 (1.11–1.38) Model 4 Age, sex, Townsend, smoking, educational level, physical activity, social class, BMI, WHR, total cholesterol, CRP 11 840 1.13 (1.03–1.24) 1.06 (.94–1.19) 1.11 (.99–1.24) 1.23 (1.09–1.37) Model 4 sensitivity analysis Excluding deaths during first 2 y of follow-up 11 707 1.12 (1.02–1.23) 1.06 (.94–1.19) 1.08 (.96–1.22) 1.23 (1.09–1.38) Participants without baseline cancer, MI, or CVA 10 684 1.11 (1.00–1.23) 1.06 (.93–1.21) 1.08 (.95–1.23) 1.19 (1.06–1.36) Data are presented as hazard ratio (95% confidence interval). Reference group is participants with negative IgG antibody for CMV, the European Prospective Investigation of Cancer–Norfolk cohort CMV study (N = 12 999). All models are adjusted for age at recruitment to the study, sex, Townsend deprivation index [32], smoking, educational level, physical activity, social class, body mass index, waist-to-hip ratio, total cholesterol, C-reactive protein (see table below).

ive Investigation of Cancer–Norfolk cohort CMV study (N = 12 999). All models are adjusted for age at recruitment to the study, sex, Townsend deprivation index [32], smoking, educational level, physical activity, social class, body mass index, waist-to-hip ratio, total cholesterol, C-reactive protein (see table below). Abbreviations: BMI, body mass index; CMV, cytomegalovirus; CRP, C-reactive protein; CVA, cerebrovascular accident; IgG, immunoglobulin G; MI, myocardial infarction; Townsend, Townsend deprivation index [32]; WHR, waist-to-hip ratio. Table 4. Hazard Ratios for Mortality Grouped by Different Attributable Cause and Seropositivity and Immunoglobulin G Antibody Levels for Cytomegalovirus

Abbreviations: BMI, body mass index; CMV, cytomegalovirus; CRP, C-reactive protein; CVA, cerebrovascular accident; IgG, immunoglobulin G; MI, myocardial infarction; Townsend, Townsend deprivation index [32]; WHR, waist-to-hip ratio. Table 4. Hazard Ratios for Mortality Grouped by Different Attributable Cause and Seropositivity and Immunoglobulin G Antibody Levels for Cytomegalovirus Model Outcome No. of Deaths All With Positive CMV IgG Antibodies Low Antibody Group Middle Antibody Group High Antibody Group Death attributed to cardiovascular diseases 851 1.06 (.91–1.24) 1.02 (.83–1.25) 0.91 (.75–1.13) 1.24 (1.03–1.49) Death attributed to cancer 955 1.13 (.98–1.31) 1.09 (.90–1.31) 1.18 (.99–1.42) 1.13 (.94–1.36) Death attributed to causes other than cardiovascular diseases or cancer 708 1.23 (1.04–1.47) 1.09 (.87–1.37) 1.25 (1.01–1.55) 1.35 (1.09–1.67) Data are presented as hazard ratio (95% confidence interval). Reference group is participants with negative IgG antibody, the European Prospective Investigation of Cancer–Norfolk cohort CMV study (N = 12 999). All models are adjusted for age at recruitment to the study, sex, Townsend deprivation index [32], smoking, educational level, physical activity, social class, body mass index, waist-to-hip ratio, total cholesterol, C-reactive protein. Abbreviations: CMV, cytomegalovirus; IgG, immunoglobulin G. DISCUSSION This is the largest ever population-based study examining all-cause and specific-cause mortality in association with CMV IgG antibody levels.

Model Outcome No. of Deaths All With Positive CMV IgG Antibodies Low Antibody Group Middle Antibody Group High Antibody Group Death attributed to cardiovascular diseases 851 1.06 (.91–1.24) 1.02 (.83–1.25) 0.91 (.75–1.13) 1.24 (1.03–1.49) Death attributed to cancer 955 1.13 (.98–1.31) 1.09 (.90–1.31) 1.18 (.99–1.42) 1.13 (.94–1.36) Death attributed to causes other than cardiovascular diseases or cancer 708 1.23 (1.04–1.47) 1.09 (.87–1.37) 1.25 (1.01–1.55) 1.35 (1.09–1.67) Data are presented as hazard ratio (95% confidence interval). Reference group is participants with negative IgG antibody, the European Prospective Investigation of Cancer–Norfolk cohort CMV study (N = 12 999). All models are adjusted for age at recruitment to the study, sex, Townsend deprivation index [32], smoking, educational level, physical activity, social class, body mass index, waist-to-hip ratio, total cholesterol, C-reactive protein. Abbreviations: CMV, cytomegalovirus; IgG, immunoglobulin G. DISCUSSION This is the largest ever population-based study examining all-cause and specific-cause mortality in association with CMV IgG antibody levels. Our findings support an independent association between seropositivity for CMV and 14-year mortality among adult participants, as shown by the previous population study [16] as well as previous small studies limited to older women [9], patients with high-level stenosis of the coronary arteries and high CRP [10], or patients undergoing coronary angiography [11]. Our findings also support an association between the level of CMV IgG antibodies and mortality.

the previous population study [16] as well as previous small studies limited to older women [9], patients with high-level stenosis of the coronary arteries and high CRP [10], or patients undergoing coronary angiography [11]. Our findings also support an association between the level of CMV IgG antibodies and mortality. Higher CMV IgG antibody levels as well as CMV DNA in the urine but not in the blood are more frequently found among older individuals [24], suggesting asymptomatic CMV reactivation in older age. CMV-specific antibody levels correlate well with the numbers of circulating CMV-specific memory B cells and are not due to a generally higher antibody production by the individuals [25]. Previous studies have attempted to examine the association between CMV IgG antibody levels and mortality. These studies have been limited to highly selected groups including Latinos over the age of 60 [13], community-dwelling older women [14], and community-dwelling older adults with stable cardiovascular disease [15]. All 3 studies found evidence of an association between higher antibody levels and mortality but had a low discriminatory power to examine the effect of seropositivity vs seronegativity because of prevalent seropositivity, and 2of them grouped participants negative for CMV IgG antibody into the lower antibody level groups, making interpretations about any possible effect of CMV difficult. Our study shows that high CMV antibody levels are associated with all-cause mortality.

ositivity vs seronegativity because of prevalent seropositivity, and 2of them grouped participants negative for CMV IgG antibody into the lower antibody level groups, making interpretations about any possible effect of CMV difficult. Our study shows that high CMV antibody levels are associated with all-cause mortality. CMV seropositivity is thought to contribute to oncogenesis [26, 27], but any excess of cancer-related deaths is not apparent in a population study of this size. Seropositivity for CMV and CMV latency has been associated with telomere shortening [28], making it possible that CMV could be contributing to mortality through its effects on immunity. The relations that we observe appear to be independent of CRP level, suggesting that we are not observing a general association with other intercurrent infection or inflammation. Measurement error might have affected our results. The assay used is not affected by other common viral antibodies and has good performance characteristics [29]. Error might have also been introduced by inaccuracies in death certificate information, although it is unlikely that such error would have been differential by CMV antibody status.

t have affected our results. The assay used is not affected by other common viral antibodies and has good performance characteristics [29]. Error might have also been introduced by inaccuracies in death certificate information, although it is unlikely that such error would have been differential by CMV antibody status. Residual confounding might have also affected our results. In this study we measured CMV IgG antibody titers at a single time point and did not directly measure CMV reactivation. Although it is postulated that higher CMV IgG antibody levels represent more frequent or intense subclinical CMV reactivation from latency, this has not been conclusively proven. CMV IgG antibody levels correlate well with CMV-specific B-cell numbers [25], but significant intraindividual variation exists [25], and antibody levels do not always reflect the presence of CMV DNA [30]. Other events that might affect CMV IgG levels acutely or chronically have not been studied. It is possible, therefore, that CMV IgG antibody levels are elevated through an unknown mechanism and that the underlying cause is also associated with the risk of death. In this context, CMV infection might even be an innocent bystander or a measure of a failing immune system [31]. The true cause of increased mortality is less likely to be contemporary social deprivation, as previously hypothesized, as after adjusting for multiple variables associated with social status in our analysis, the observed hazard ratios did not appreciably change.

bystander or a measure of a failing immune system [31]. The true cause of increased mortality is less likely to be contemporary social deprivation, as previously hypothesized, as after adjusting for multiple variables associated with social status in our analysis, the observed hazard ratios did not appreciably change. CONCLUSIONS This is the first population cohort study to show an association between seropositivity as well as CMV antibody level and mortality. We show that the association with all-cause mortality is not explained by various measures of social deprivation or inflammation and is not limited to cardiovascular disease. Studies attempting to correlate the levels of IgG CMV antibody with longitudinal measurements of viral load or other direct measurements of viral reactivation will be necessary to further explore the strong observed associations. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

rdjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Financial support. This work was supported by the Medical Research Council (Special Populations Fellowship to E. G.-K. and support to EPIC-Norfolk); Cancer Research UK; the Stroke Association; the British Heart Foundation; the Department of Health; the Commissions of the European Union's Europe Against Cancer Programme; the Food Standards Agency; the Department of Environment, Food and Rural Affairs; and the World Health Organization. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

The nonnucleoside reverse-transcriptase inhibitor nevirapine is widely used as a first-line treatment of human immunodeficiency virus (HIV) [1] infection in developing countries because of its low cost. Nevirapine is usually given in combination with 2 nucleoside reverse-transcriptase inhibitors (stavudine or zidovudine and lamivudine). Though effective [2], nevirapine causes hypersensitivity in 6%–10% of patients [3, 4], which can manifest clinically in various ways from nevirapine-induced rash (without any systemic manifestations) to severe blistering skin reactions such as Stevens-Johnson syndrome and toxic epidermal necrolysis [5] (1–2 per 1000 exposed individuals [6]). Extracutaneous involvement typically manifests as hepatotoxicity [7]. Reactions most commonly manifest within the first 6 weeks of starting therapy. Immunogenetic factors, including a number of human leukocyte antigen (HLA) alleles, have been previously identified as risk factors for hypersensitivity reactions to the antiretroviral abacavir [8] and many other classes of drugs, including the antiepileptic drug carbamazepine [9, 10] and the antigout drug allopurinol [11]. Research has also focused on immunogenetic risk factors for nevirapine hypersensitivity, identifying the HLA-DRB1*01:01 (whites [12–14]), HLA-C*04 (Thai [15], Chinese [16], and blacks [14]), HLA-C*08 (Japanese [17]) and HLA-B*35:05 (Thai [14, 18]) as risk alleles (Table 1). Table 1. Previously Reported Human Leukocyte Antigen Allele Associations With Nevirapine Hypersensitivity

nevirapine hypersensitivity, identifying the HLA-DRB1*01:01 (whites [12–14]), HLA-C*04 (Thai [15], Chinese [16], and blacks [14]), HLA-C*08 (Japanese [17]) and HLA-B*35:05 (Thai [14, 18]) as risk alleles (Table 1). Table 1. Previously Reported Human Leukocyte Antigen Allele Associations With Nevirapine Hypersensitivity Phenotype [Reference] Population Cases/Controls (No.) Risk Protective HLA Allele OR (95% CI) P Value HLA Allele OR (95% CI) P Value 1) HSRs (12 DILI) [16] Han Chinese 32/71 HLA-C*04 3.61 (1.13–11.49) .03 HLA-DRB1*15 0.34 (.12–.99) .049 2) Rash [15] Thai 39/60 HLA-C*04 3.18 (1.33–8.63) .009 HLA-C*03 0.27 (.09–.82) .01 3) Isolated rash [13] White 6/15 HLA-DRB1*01 70.0 (3.65–1342.66) .004a 4) Rash/systemic with associated hepatitis [12] White (Australia) 15/64 HLA-DRB1*01 4.77 (1.55–14.73) .01 5) All HSRs Sardinian 13/36 HLA-C*08/B*14 14.57 (2.42–87.73) .004a 3 systemic with rash and/or liver toxicity 5 extensive skin rash 5 isolated hepatotoxicity [40] 6) Rash [18] Thai 143/181 HLA-B*35:05 18.96 (4.87–73.44) 4.9 × 10−8a HLA-C*07:02 0.40 (.20–.78) .0067 7) All HSRs Japanese 41/41 HLA-C*08 6.19 (1.18–32.40) .03 8 rash; 3 rash + fever; 1 DILI [17] 8) Cutaneous HSRs [14] Mixed 175/587 HLA-C*04 2.51 (1.73–3.62) 6.7 × 10−7a Black 27/77 5.17 (2.01–13.30) 9.5 × 10−4a Asian 71/233 2.55 (1.41–4.60) .0028a Asian 71/227 HLA-B*35/C*04 18.34 (5.10–65.99) 2.4 × 10−7a Thai 52/173 13.49 (3.56–52.20) 3.4 × 10−5a 9) Isolated hepatotoxicity [14] White 57/277 HLA-DRB1*01 3.02 (1.66–5.49) 5.7 × 10−4a 10) Rash, 4 SJS, 3 with hepatitis [41] Indian 40/40 HLA-B*35 3.38 (1.54–7.41) .003 HLA-B*8 0.29 (.12–.71) .008 Where odds ratios and 95% CIs are not reported, a 2 × 2 χ2 test was performed on the data available.

.56–52.20) 3.4 × 10−5a 9) Isolated hepatotoxicity [14] White 57/277 HLA-DRB1*01 3.02 (1.66–5.49) 5.7 × 10−4a 10) Rash, 4 SJS, 3 with hepatitis [41] Indian 40/40 HLA-B*35 3.38 (1.54–7.41) .003 HLA-B*8 0.29 (.12–.71) .008 Where odds ratios and 95% CIs are not reported, a 2 × 2 χ2 test was performed on the data available. Abbreviations: CI, confidence interval; DILI, drug-induced liver injury; HLA, human leukocyte antigen; HSR, hypersensitivity reaction; OR, odds ratio; SJS, Stevens-Johnson syndrome. a Denotes reported associations which withstood correction for multiple testing (Pcorrected < 0.05). To date, little is known regarding genetic risk factors for nevirapine-induced hypersensitivity in sub-Saharan African HIV-infected populations. Using a cohort of carefully phenotyped Malawian patients, we have undertaken high-resolution sequence-based genotyping to determine whether alleles in 5 loci in the class I and II major histocompatibility complex (MHC) regions on chromosome 6 (HLA- DRB1, DQB1, A, B, or C) are predisposing factors for nevirapine hypersensitivity.

hort of carefully phenotyped Malawian patients, we have undertaken high-resolution sequence-based genotyping to determine whether alleles in 5 loci in the class I and II major histocompatibility complex (MHC) regions on chromosome 6 (HLA- DRB1, DQB1, A, B, or C) are predisposing factors for nevirapine hypersensitivity. PATIENTS AND METHODS Patients Between March 2007 and December 2008, we prospectively recruited 1117 antiretroviral-naive adult patients from the outpatient clinic at the Queen Elizabeth Central Hospital, Blantyre, Malawi. At time of recruitment, this clinic had approximately 10 000 patients registered as having started on antiretroviral therapy since 2004. Patients were self-reported black African; were older than 16 years; and gave informed consent approved by the research ethics committees at the College of Medicine Research and Ethics Committee, Malawi, and Liverpool School of Tropical Medicine. Patients presenting with jaundice at baseline were excluded. Patients commenced antiretroviral therapy as recommended by the World Health Organization eligibility criteria at the time of recruitment. All patients were diagnosed as clinical stage 3/4 or had a CD4+ count <250 cells/µL; commenced preparations, which contained a fixed dose of nevirapine, lamivudine, and stavudine; and were followed up for 26 weeks. Clinical and laboratory parameters including CD4+ count and liver function tests were monitored at 0, 6, 14, and 18 weeks.

diagnosed as clinical stage 3/4 or had a CD4+ count <250 cells/µL; commenced preparations, which contained a fixed dose of nevirapine, lamivudine, and stavudine; and were followed up for 26 weeks. Clinical and laboratory parameters including CD4+ count and liver function tests were monitored at 0, 6, 14, and 18 weeks. The study was a nested case-control study; however, because of the low incidence of hypersensitivity in the prospectively recruited cohort, an additional 177 patients attending the same outpatient clinic who developed nevirapine hypersensitivity were also recruited, either prospectively (n = 149) or identified from patient records retrospectively (n = 28). Of 177 patients, 65 were excluded owing to insufficient DNA quality and quantity. Careful clinical assessment of all patients was undertaken to identify and characterize the hypersensitivity reactions, using the Naranjo causality assessment tool [19, 20]. Phenotypes were retrospectively reviewed independently by a dermatologist using both clinical data and photographs. These were defined as: Nevirapine-induced rash (NIR): widespread maculopapular rash without systemic manifestations and getting worse on treatment continuation. Hypersensitivity syndrome (HSS; also known as drug reaction with eosinophilia and systemic symptoms or drug-induced hypersensitivity syndrome): widespread rash and systemic manifestations such as fever, cough, or abnormal liver function tests.

Careful clinical assessment of all patients was undertaken to identify and characterize the hypersensitivity reactions, using the Naranjo causality assessment tool [19, 20]. Phenotypes were retrospectively reviewed independently by a dermatologist using both clinical data and photographs. These were defined as: Nevirapine-induced rash (NIR): widespread maculopapular rash without systemic manifestations and getting worse on treatment continuation. Hypersensitivity syndrome (HSS; also known as drug reaction with eosinophilia and systemic symptoms or drug-induced hypersensitivity syndrome): widespread rash and systemic manifestations such as fever, cough, or abnormal liver function tests. Stevens-Johnson syndrome (SJS): extensive rash with the involvement of at least 2 mucous membranes or blistering eruptions affecting <10% of body surface area. Toxic epidermal necrolysis (TEN) [5]; blistering rash affecting >30% of body surface area and mucous membrane involvement as per SJS [21]. Blistering between 10% and 30% of body surface area was termed overlap syndrome. Drug-induced liver injury (DILI) [7]: jaundice and abnormal alanine aminotransferase level.

Stevens-Johnson syndrome (SJS): extensive rash with the involvement of at least 2 mucous membranes or blistering eruptions affecting <10% of body surface area. Toxic epidermal necrolysis (TEN) [5]; blistering rash affecting >30% of body surface area and mucous membrane involvement as per SJS [21]. Blistering between 10% and 30% of body surface area was termed overlap syndrome. Drug-induced liver injury (DILI) [7]: jaundice and abnormal alanine aminotransferase level. Patients meeting criteria for drug-induced reactions had nevirapine withdrawn in accordance with international guidelines. It is important to note that as part of the Malawian treatment guidelines, liver function tests are not routine, and therefore, abnormal tests without clinical jaundice did not fulfill criteria for treatment cessation and were not included as cases. Furthermore, some patients who developed transient nonsevere rash without systemic symptoms underwent close observation, were treated continuously with rash resolution, and again were not classified as cases. Control patients (n = 155) were identified from the prospective cohort and followed up for at least 6 months while taking nevirapine without developing any signs of hypersensitivity. Cases and controls were matched by age and sex, and were also from the same region of Malawi.

Patients meeting criteria for drug-induced reactions had nevirapine withdrawn in accordance with international guidelines. It is important to note that as part of the Malawian treatment guidelines, liver function tests are not routine, and therefore, abnormal tests without clinical jaundice did not fulfill criteria for treatment cessation and were not included as cases. Furthermore, some patients who developed transient nonsevere rash without systemic symptoms underwent close observation, were treated continuously with rash resolution, and again were not classified as cases. Control patients (n = 155) were identified from the prospective cohort and followed up for at least 6 months while taking nevirapine without developing any signs of hypersensitivity. Cases and controls were matched by age and sex, and were also from the same region of Malawi. DNA Extraction and High-Resolution Sequence-Based HLA Typing DNA was extracted from whole blood using a salt precipitation protocol. High-resolution, sequencing-based HLA typing of 5 loci (HLA-A, B, C, DRB1, and DQB) was undertaken by Histogenetics (Ossining, New York). Sequencing data files were analyzed using Histogenetics’ proprietary analysis software (Histomatcher and HistoMagic) for HLA genotype calling. Allele assignments are based on IMGT/HLA Database release version 2.21.0, dated April 2008 (http://www.ebi.ac.uk/imgt/hla/).

as undertaken by Histogenetics (Ossining, New York). Sequencing data files were analyzed using Histogenetics’ proprietary analysis software (Histomatcher and HistoMagic) for HLA genotype calling. Allele assignments are based on IMGT/HLA Database release version 2.21.0, dated April 2008 (http://www.ebi.ac.uk/imgt/hla/). Statistical Analysis Sample size calculations were performed assuming that a 10% background frequency of an HLA allele would provide 80% power (α = .05) to detect an odds ratio (OR) of 3.0 (and 90% power to detect an OR of 3.4). We included all patients with hypersensitivity in the analysis. Subgroup analyses were performed for all phenotypes, (DILI, SJS/TEN, HSS, NIR) where we compared the frequency of HLA alleles in patients with nevirapine-induced adverse reaction with the frequency in tolerant individuals.

0% power to detect an OR of 3.4). We included all patients with hypersensitivity in the analysis. Subgroup analyses were performed for all phenotypes, (DILI, SJS/TEN, HSS, NIR) where we compared the frequency of HLA alleles in patients with nevirapine-induced adverse reaction with the frequency in tolerant individuals. Nongenetic factors identified a priori as being of importance, such as CD4+ count, were first tested univariately for association with hypersensitivity reaction (all cases) using the Student t test. The distribution of CD4+ count was skewed, and a square-root transformation was used to achieve normality. CD4+ count for 20 cases was missing, and these observations were substituted by the mean-transformed CD4+ count for all cases where CD4+ count was observed. Differences in frequencies of alleles in individual HLA locus between tolerant patients and each of the hypersensitivity phenotypes were determined from 2 × N contingency tables using a χ2 test within the CLUMP software package (http://www.smd.qmul.ac.uk/statgen/dcurtis/lc/clump.html). To determine association with specific alleles within hypersensitivity-linked HLA loci, 2 logistic regression models were fitted. The first included covariates representing the nongenetic factors identified from univariate analysis (P < .05). The second included a covariate to represent HLA alleles assuming a dominant mode of inheritance. Rare alleles were grouped into a single allele category and, because this represented the largest category, it was assumed to be the baseline allele category for the purpose of regression modeling. To assess for significance of the genetic locus, a likelihood-ratio test was undertaken comparing the models and the P value was recorded. Analyses were undertaken in R version 2.13.0. To account for multiple comparisons (5 phenotypes and 5 loci), we used the false-discovery rate method within the “p.adjust” function of R. The HLA multiple locus haplotypes were generated using PyPop 0.7.0 software [22].

n comparing the models and the P value was recorded. Analyses were undertaken in R version 2.13.0. To account for multiple comparisons (5 phenotypes and 5 loci), we used the false-discovery rate method within the “p.adjust” function of R. The HLA multiple locus haplotypes were generated using PyPop 0.7.0 software [22]. A random-effects OR meta-analysis of pooled data from our study and previously published data was undertaken using StatsDirect version 2.6.8 (StatsDirect Ltd, Atrincham, UK). RESULTS From the prospective cohort (n = 1117), 57 patients developed hypersensitivity (5.1%), of whom 31 were successfully HLA-typed. Of the 149 supplementary hypersensitive patients, 86 were HLA-typed, giving a total of 117 HLA-typed hypersensitive patients (15 DILI, 33 SJS/TEN, 20 HSS, and 46 NIR, plus 3 individuals with the DILI and SJS/TEN phenotype). One control sample failed HLA typing, leaving 154 HLA-typed drug-tolerant controls. The overall HLA-allele call rates were 182 of 271 (67%) for DRB1*; 241 of 271 (89%) for DQB1*; and 296 of 271 (99%) for A, B, and C. A summary of the HLA allele frequencies for each of the phenotypes and controls is provided in Supplementary Table 1. Median CD4+ cell count at the start of antiretroviral therapy was 235 cells/µL (interquartile range [IQR], 128–424 cells/µL) in cases and 166 cells/µL (IQR, 83–250 cells/µL) in controls. This represented a statistically significant difference; thus, CD4+ cell count was adjusted for in the analyses of association with genetic loci.

cell count at the start of antiretroviral therapy was 235 cells/µL (interquartile range [IQR], 128–424 cells/µL) in cases and 166 cells/µL (IQR, 83–250 cells/µL) in controls. This represented a statistically significant difference; thus, CD4+ cell count was adjusted for in the analyses of association with genetic loci. We undertook χ2 analyses in CLUMP focusing on the association of each locus with the different phenotypic manifestations (Table 2). After correction for multiple comparisons, we identified HLA-DQB1 as the only significant (Pcorrected < .05) HLA locus for nevirapine-induced hypersensitivity, when all phenotypes were combined, and with SJS/TEN specifically. The locus-specific analysis provided an indication that the HLA-DQB1 region protected against nevirapine hypersensitivity. Given the high degree of linkage disequilibrium in the MHC, and the multiple alleles present within each locus, we then undertook an analysis of the individual HLA alleles (Table 3). Consistent with the locus-specific data, a number of HLA-DQB1 alleles were found to protect against nevirapine hypersensitivity when compared to the “rare allele” group. These included 6 different DQB1* alleles (02:01G, 03:02:01, 05:01:01, 06:02, 06:03:01, and 06:09) associated with a decreased risk of all hypersensitivity reactions with ORs ranging from 0.17 (95% CI, .05–.6) to 0.41 (95% CI, .18–.96). DQB1*05:0101 was protective for SJS (OR = 0.11 [95% CI, .02–.56]) and HSS (OR = 0.17 [95% CI, .04–.80]); and 06:02 for DILI (OR = 0.09 [95% CI, .01–.52]) and HSS (OR = 0.16 [95% CI, .04–.75]). One DRB1* allele (15:03) protected against DILI (OR = 0.08 [95% CI, .01–.59]). Table 2. Logistic Regression Analysis for 5 Human Leukocyte Antigen Loci in All Patients With Nevirapine-Induced Hypersensitivity

.17 [95% CI, .04–.80]); and 06:02 for DILI (OR = 0.09 [95% CI, .01–.52]) and HSS (OR = 0.16 [95% CI, .04–.75]). One DRB1* allele (15:03) protected against DILI (OR = 0.08 [95% CI, .01–.59]). Table 2. Logistic Regression Analysis for 5 Human Leukocyte Antigen Loci in All Patients With Nevirapine-Induced Hypersensitivity Locus P Value All HSR SJS/TEN DILI NIR HSS HLA-A 894 .015 .429 .672 .135 HLA-B 111 .160 .068 .478 .066 HLA-C 036 .014 .179 .489 .564 HLA-DQB1 002* .003* .006 .148 .030 HLA-DRB1 137 .123 .018 .455 .477 P values are derived from a likelihood ratio test comparing a logistic regression model both with and without a covariate representing the alleles observed at the genetic locus. Statistically significant findings (P < .05) are indicated in bold while associations withstanding correcting for multiple comparisons (false discovery rate P < .05) are indicated by an asterisk (*). Abbreviations: DILI, drug-induced liver injury; HLA, human leukocyte antigen; HSR, hypersensitivity reaction; HSS, hypersensitivity syndrome; NIR, nevirapine-induced rash; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis. Table 3. Association of Specific Human Leukocyte Antigen Alleles With Nevirapine-Induced Hypersensitivity Phenotypes Compared to the Baseline “Rare Allele” Group

Abbreviations: DILI, drug-induced liver injury; HLA, human leukocyte antigen; HSR, hypersensitivity reaction; HSS, hypersensitivity syndrome; NIR, nevirapine-induced rash; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis. Table 3. Association of Specific Human Leukocyte Antigen Alleles With Nevirapine-Induced Hypersensitivity Phenotypes Compared to the Baseline “Rare Allele” Group Loci Allele Odds Ratio (95% CI) All HSR (n = 116) SJS/TEN (n = 36) DILI (n = 18) HSS HLA-A* 01:01 n/a 02:01 1.63 (.45–5.93) 02:05 1.19 (.18–7.97) 03:01 3.11 (.62–15.59) 23:01 .52 (.12–2.30) 29:02:01 1.81 (.42–7.71) 30:01 1.04 (.28–3.84) 30:02 .59 (.16–2.22) 34:02 3.59 (.89–14.57) 36:01 2.23 (.51–9.75) 66:01 n/a 68:01 n/a 68:02 2.80 (.81–9.72) 74:01 .36 (.06–2.14) HLA-C* 02:10 .56 (.21–1.49) 1.38 (.22–8.72) 03:02 1.03 (.24–4.4) 3.51 (.25–49.79) 03:03 1.42 (.31–6.43) 1.59 (.08–31.76) 03:04:02 1.43 (.46–4.41) 4.85 (.8–29.54) 04:01 2.64 (1.13–6.18) 17.52 (3.31–92.8) 06:02 .67 (.27–1.67) 1.25 (.18–8.57) 07:01 2.04 (.81–5.14) 4.00 (.73–21.84) 07:02 1.08 (.33–3.55) 3.06 (.35–26.72) 07:04 1.40 (.36–5.42) 6.68 (.61–72.57) 08:02 .99 (.37–2.68) 2.40 (.39–14.8) 12:03 .82 (.19–3.6) 1.69 (.15–19.67) 16:01:01 .90 (.29–2.78) 3.78 (.52–27.35) 17:01 1.43 (.56–3.63) 5.95 (.94–37.55) 18:01 .92 (.34–2.49) 2.02 (.29–14.08) Allele All HSR (n = 106) SJS/TEN (n = 35) DILI (n = 14) HSS (n = 17) HLA-DQB1* 02:01G .41 (.18–.96) .98 (.30–3.24) .08 (.01–.61) .26 (.06–1.19) 03:01G 1.12 (.45–2.75) 2.26 (.60–8.54) 1.13 (.23–5.69) .83 (.19–3.57) 03:02:01 .32 (.08–.94) .62 (.09–4.29) .40 (.03–5.25) n/a 04:02 .39 (.13–1.16) .67 (.12–3.59) .14 (.01–1.81) .60 (.10–3.48) 05:01:01 .27 (.12–.63) .11 (.02–.56) .29 (.06–1.45) .17 (.04–.80) 06:02 .30 (.13–.72) .87 (.22–3.45) .09 (.01–.52) .16 (.04–.75) 06:03:01 .17 (.05–.60) .27 (.04–1.80) .26 (.02–3.40) n/a 06:04:01 .22 (.05–1.34) n/a n/a .40 (.06–2.95) 06:09 .16 (.05–.56) .30 (.05–1.88) n/a .32 (.05–2.04) HLA-DRB1* 01:02:01 .16 (.02–1.61) 03:01:01 n/a 03:02:01 .35 (.03–3.66) 07:01:01 .10 (.01–1.06) 09:01:02 .25 (.02–2.62) 11:01 .17 (.02–1.28) 11:02:01 n/a 12:01 n/a 13:01:01 .14 (.01–1.39) 13:02:01 n/a 15:03 .08 (.01–.59) Only loci/phenotype associations determined as significant in Table 2 are included. Results in bold are those allele/phenotypes associations where the 95% confidence interval for the odds ratio excludes 1.

25 (.02–2.62) 11:01 .17 (.02–1.28) 11:02:01 n/a 12:01 n/a 13:01:01 .14 (.01–1.39) 13:02:01 n/a 15:03 .08 (.01–.59) Only loci/phenotype associations determined as significant in Table 2 are included. Results in bold are those allele/phenotypes associations where the 95% confidence interval for the odds ratio excludes 1. Abbreviations: CI, confidence interval; DILI, drug-induced liver injury; HLA, human leukocyte antigen; HSR, hypersensitivity reaction; HSS, hypersensitivity syndrome; n/a, odds ratios were not possible to calculate due to the low frequency of the allele in that phenotype; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis. Our analysis showed that HLA-C*04:01 predisposed to nevirapine hypersensitivity. Individuals who carry HLA-C*04:01 were at higher risk of developing hypersensitivity reactions in general (OR = 2.64 [95% CI, 1.13–2.64]), and, specifically SJS/TEN (OR = 17.52 [95% CI, 3.31–92.80]) when exposed to nevirapine than were carriers of the “rare alleles,” the most common group of HLA alleles in C locus. This association was not observed with any other phenotype.

developing hypersensitivity reactions in general (OR = 2.64 [95% CI, 1.13–2.64]), and, specifically SJS/TEN (OR = 17.52 [95% CI, 3.31–92.80]) when exposed to nevirapine than were carriers of the “rare alleles,” the most common group of HLA alleles in C locus. This association was not observed with any other phenotype. Multilocus haplotypes for class I and II HLA loci were generated to determine the structure of haplotypes across multiple loci in our cohort from Malawi (Table 4). The data suggest high linkage disequilibrium between the HLA-B and C loci in both the nevirapine hypersensitive and tolerant patients (D′ = 0.946 and 0.924, respectively). Significant linkage disequilibrium was observed in the hypersensitive and tolerant groups between the DQB1 and DRB1 loci (D′ = 0.890 and 0.898 respectively). Haplotype frequencies were calculated for 5-loci haplotypes and for combinations of HLA-B, C, DRB1, and DQB1 loci haplotypes containing the HLA-C*04:01 allele (Table 5 and Supplementary Table 2). The frequency of the HLA B53:01:01/C*04:01 haplotype was significantly higher in the hypersensitive cohort (0.121) than the tolerant group (0.039). There was no difference in DRB1/DQB1 haplotype frequencies in haplotypes containing the DQB1*05:01:01 allele (Table 5). Table 4. Linkage Disequilibrium Analysis of Class I and II Human Leukocyte Antigen Loci in Nevirapine-Hypersensitive and -Tolerant Malawian Cohorts

tive cohort (0.121) than the tolerant group (0.039). There was no difference in DRB1/DQB1 haplotype frequencies in haplotypes containing the DQB1*05:01:01 allele (Table 5). Table 4. Linkage Disequilibrium Analysis of Class I and II Human Leukocyte Antigen Loci in Nevirapine-Hypersensitive and -Tolerant Malawian Cohorts Locus Pair D′ All HSR Tolerant 1 A|B 0.747 0.746 2 A|C 0.659 0.666 3 A|DRB1 0.642 0.621 4 A|DQB1 0.599 0.523 5 B|C 0.946 0.924 6 B|DRB1 0.760 0.743 7 B|DQB1 0.650 0.592 8 C|DRB1 0.632 0.635 9 C|DQB1 0.581 0.538 10 DRB1|DQB1 0.890 0.898 Data represent the D′ value for each pairwise analysis as determined by PyPop 0.7.0 software. Abbreviation: HSR, hypersensitivity reaction. Table 5. Frequency of Human Leukocyte Antigen Haplotype Frequencies in the Nevirapine-Hypersensitive and -Tolerant Cohorts

Locus Pair D′ All HSR Tolerant 1 A|B 0.747 0.746 2 A|C 0.659 0.666 3 A|DRB1 0.642 0.621 4 A|DQB1 0.599 0.523 5 B|C 0.946 0.924 6 B|DRB1 0.760 0.743 7 B|DQB1 0.650 0.592 8 C|DRB1 0.632 0.635 9 C|DQB1 0.581 0.538 10 DRB1|DQB1 0.890 0.898 Data represent the D′ value for each pairwise analysis as determined by PyPop 0.7.0 software. Abbreviation: HSR, hypersensitivity reaction. Table 5. Frequency of Human Leukocyte Antigen Haplotype Frequencies in the Nevirapine-Hypersensitive and -Tolerant Cohorts Hypersensitive (n = 116) Tolerant (n = 153) Frequency Counts Frequency Counts B|C Haplotype 53:01:01|04:01 0.121 26 0.039 12 44:03|04:01 0.069 16 0.059 18 35:01|04:01 0.026 6 0.023 7 15:10|04:01 0.013 3 0.006 2 58:02|04:01 0.004 1 15:03|04:01 0.004 1 37:01:01|04:01 0.004 1 81:01|04:01 0.004 1 42:01|04:01 0.003 1 57:01:01|04:01 0.006 2 58:01|04:01 0.003 1 Cases (n = 93) Controls (n = 89) DRB1|DQB1 Haplotype 01:02:01|05:01:01 0.038 7 0.073 13 12:01|05:01:01 0.038 7 0.062 11 13:01:01|05:01:01 0.032 6 0.028 5 01:01:01|05:01:01 0.011 2 0.006 1 10:01|05:01:01 0.011 2 0.011 2 13:02:01|05:01:01 0.005 1 14:01|05:01:01 0.006 1 15:03|05:01:01 0.006 1 Only B|C haplotypes containing the C*04:01 and DRB1\DQB1 haplotypes containing DQB1*05:01:01 are listed. Frequency data for class I and class II haplotypes are listed in Supplementary Table 2.

05:01:01 0.011 2 0.006 1 10:01|05:01:01 0.011 2 0.011 2 13:02:01|05:01:01 0.005 1 14:01|05:01:01 0.006 1 15:03|05:01:01 0.006 1 Only B|C haplotypes containing the C*04:01 and DRB1\DQB1 haplotypes containing DQB1*05:01:01 are listed. Frequency data for class I and class II haplotypes are listed in Supplementary Table 2. Subsequent analysis was undertaken incorporating all HLA-typed individuals to determine the absolute risk of SJS/TEN and predictive values of HLA-C*04:01, the HLA-B*53:01:01/C*04:01 haplotype, and DQB1*05:0101 carriage (Table 6). The OR for overall risk of SJS/TEN associated with HLA-C*04:01 was 5.17 (95% CI, 2.39–11.18; P < .0001). Positive predictive values (PPVs) and negative predictive values (NPVs), based on a SJS/TEN prevalence of 1.07%, were 2.6% and 99.2%, respectively. The OR for overall risk of SJS/TEN associated with carriage of the HLA-B*53:01:01/C*04:01 haplotype (5.17 [95% CI, 1.83–14.28]) was comparable to HLA-C*04:01 alone, although the NPV was lower (91.6%). For DQB*05:0101 the OR was 0.17 (95% CI, .05–.60), and the PPV and NPV were 0.3% and 98.6%, respectively. Other HLA allele/hypersensitivity phenotype associations noted in Table 3 demonstrated PPVs between 0.08% and 3.1% and NPVs between 22.3% and 36.2% (data not shown). Table 6. Predictive Value for Stevens-Johnson Syndrome/Toxic Epidermal Necrolysis Risk of HLA-C*04:01, B*53:01:01|*04:01, and DQB1*05:01:01 Carriage

vely. Other HLA allele/hypersensitivity phenotype associations noted in Table 3 demonstrated PPVs between 0.08% and 3.1% and NPVs between 22.3% and 36.2% (data not shown). Table 6. Predictive Value for Stevens-Johnson Syndrome/Toxic Epidermal Necrolysis Risk of HLA-C*04:01, B*53:01:01|*04:01, and DQB1*05:01:01 Carriage SJS/TEN Tolerant Total HLA-C*04:01 Positive 23 39 62 PPV = 2.6% Negative 13 114 127 NPV = 99.2% All 36 153 Sensitivity = 63.9% Specificity = 74.4% OR = 5.17 (95% CI, 2.39–11.18), P < .0001 HLA-B*53:01:01|HLA-C*04:01 Positive 11 12 33 PPV = 4.2% Negative 25 141 176 NPV = 91.6% All 36 153 Sensitivity = 31.4% Specificity = 92.2% OR = 5.17 (95% CI, 1.83–14.28), P = .0002 HLA-DQB1*05:01:01 Positive 3 47 50 PPV = 0.3% Negative 32 88 120 NPV = 98.6% All 35 135 Sensitivity = 0.9% Specificity = 65.1% OR = 0.17 (95% CI, .05–.60), P = .0024 Positive and negative predictive values as well as sensitivity and specificity are shown. Prevalence of SJS/TEN is assumed at 1.07% based on an incidence of 12 of 1117 observed in the prospective study. Odds ratio with 95% confidence and P value are determined using a 2 × 2 χ2 test. Abbreviations: CI, confidence interval; HLA, human leukocyte antigen; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis.

SJS/TEN Tolerant Total HLA-C*04:01 Positive 23 39 62 PPV = 2.6% Negative 13 114 127 NPV = 99.2% All 36 153 Sensitivity = 63.9% Specificity = 74.4% OR = 5.17 (95% CI, 2.39–11.18), P < .0001 HLA-B*53:01:01|HLA-C*04:01 Positive 11 12 33 PPV = 4.2% Negative 25 141 176 NPV = 91.6% All 36 153 Sensitivity = 31.4% Specificity = 92.2% OR = 5.17 (95% CI, 1.83–14.28), P = .0002 HLA-DQB1*05:01:01 Positive 3 47 50 PPV = 0.3% Negative 32 88 120 NPV = 98.6% All 35 135 Sensitivity = 0.9% Specificity = 65.1% OR = 0.17 (95% CI, .05–.60), P = .0024 Positive and negative predictive values as well as sensitivity and specificity are shown. Prevalence of SJS/TEN is assumed at 1.07% based on an incidence of 12 of 1117 observed in the prospective study. Odds ratio with 95% confidence and P value are determined using a 2 × 2 χ2 test. Abbreviations: CI, confidence interval; HLA, human leukocyte antigen; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis. DISCUSSION Nevirapine-induced hypersensitivity reactions have shown an association with a number of HLA alleles (Table 1), which vary according to ethnicity and the phenotype of the reaction [23]. The main finding of the present study is that HLA-C*04:01 predisposes to nevirapine-cutaneous reactions with the greatest risk observed with SJS/TEN (OR = 5.17 [95% CI, 2.39–11.18]; Table 3), the severest form of hypersensitivity in terms of mortality [24–26]. The risk associated with HLA-B*53:01:01/C*04:01 haplotype carriage was comparable. Its sensitivity as a biomarker for SJS/TEN was 31.4%, compared to 63.9% for HLA-C*04:01 alone (Table 6), suggesting that the association is driven by carriage of 1 allele at a single HLA locus. This is supported by the haplotype analysis of the HLA loci in this particular Malawian population (Table 4). Although HLA-C*04 (along with B*35) has been associated with the development of AIDS in whites [27], no association between HLA-C*04 and HIV has been reported in African populations or any other ethnic group. This is the first report of an association between nevirapine-induced SJS/TEN and HLA-C*04:01, but is consistent with previous studies in black African (OR = 5.17) [14], Thai (OR = 3.79) [15], and Chinese (OR = 3.23) [16] populations that have reported an association with HLA-C*04 nevirapine-cutaneous reactions. A meta-analysis of our data, related to HLA-C*04:01 carriage and cutaneous nevirapine hypersensitivity reactions (n = 102), but excluding patients who had DILI only (n = 15), with the only eligible previous study in a black American population (OR = 5.17 [95% CI, 1.81–14.78]) [14] gave a combined OR of 3.34 (95% CI, 1.60–4.98) (Supplementary Figure 1).

ed to HLA-C*04:01 carriage and cutaneous nevirapine hypersensitivity reactions (n = 102), but excluding patients who had DILI only (n = 15), with the only eligible previous study in a black American population (OR = 5.17 [95% CI, 1.81–14.78]) [14] gave a combined OR of 3.34 (95% CI, 1.60–4.98) (Supplementary Figure 1). Although it would be useful to replicate in other populations, the data available to date strongly suggest that HLA-C*04:01 predisposes to nevirapine-induced cutaneous reactions of different severities (including SJS/TEN) in several ethnic groups. The predictive value of HLA-C*04:01 as a biomarker of nevirapine-induced SJS/TEN is limited (Table 4). The incidence of nevirapine-induced SJS/TEN in our prospective study was 12 of 1117 patients (1.07%). This gives a PPV of 2.6%, which is of no diagnostic value. Although the NPV is 99.2%, it does not reach 100%, which has been recommended [23]. It is important to note that in our population, HLA-C*04:01 was associated with SJS/TEN, which is almost always drug-related, severe, and perhaps more easily recognizable than other drug-induced adverse phenotypes. This contrasts with abacavir hypersensitivity, which varies in severity and can be more difficult to differentiate from other causes. This is reflected in the fact that the NPV of HLA-B*57:01 was 95.5% for clinically diagnosed abacavir-induced hypersensitivity and 100% for immunologically confirmed abacavir hypersensitivity [28].

sts with abacavir hypersensitivity, which varies in severity and can be more difficult to differentiate from other causes. This is reflected in the fact that the NPV of HLA-B*57:01 was 95.5% for clinically diagnosed abacavir-induced hypersensitivity and 100% for immunologically confirmed abacavir hypersensitivity [28]. We did not replicate a previous associations between HLA-DRB1*01:01 and nevirapine-induced hypersensitivity [12–14] in whites. This is possibly due to ethnic differences in the frequency of HLA alleles. HLA-DRB1*01:01 was observed at a low frequency in our Malawian cohort (0.008) with 1 tolerant and 2 hypersensitive carriers out of 182 individuals genotyped.

ions between HLA-DRB1*01:01 and nevirapine-induced hypersensitivity [12–14] in whites. This is possibly due to ethnic differences in the frequency of HLA alleles. HLA-DRB1*01:01 was observed at a low frequency in our Malawian cohort (0.008) with 1 tolerant and 2 hypersensitive carriers out of 182 individuals genotyped. The observation of an apparent protective effect of 6 different HLA-DQB1 (Table 3) across a number of different phenotypes is interesting. A number studies have identified protective HLA alleles (Table 1), but there is no common pattern. These may not represent the actual protective alleles given the high degree of linkage disequilibrium across the MHC. Further work in larger populations is needed to elucidate the interaction between risk and protective HLA alleles in predisposing to different forms of nevirapine hypersensitivity. Of note here is that the pathogenesis of nevirapine hypersensitivity is immune-mediated, as shown by a positive lymphocyte transformation test in a patient with DILI [29]. However the mechanism by which this occurs is unknown. Three possible hypotheses have been suggested, including the hapten hypothesis [30], pharmacological interaction hypothesis [31], and altered peptide binding profile [32]. It is possible that based on the HLA profile of an individual, the interaction between nevirapine (and its antigen) and the HLA molecules leads to either a protective or a predisposing effect. Such an allele-competing effect has been postulated for the HLA-associated disease narcolepsy [33].

ptide binding profile [32]. It is possible that based on the HLA profile of an individual, the interaction between nevirapine (and its antigen) and the HLA molecules leads to either a protective or a predisposing effect. Such an allele-competing effect has been postulated for the HLA-associated disease narcolepsy [33]. Studies have also evaluated the role of CYP2B6, which metabolizes nevirapine, in predisposing to hypersensitivity. CYPB6 shows wide interindividual variability in expression and activity in human livers [34]. It contains a functional exonic variant (c.516G>T), which causes loss of enzymatic function [35]; is associated with higher plasma concentrations in black and white populations [36, 37]; and has been associated with nevirapine-induced cutaneous adverse reactions [14] and neuropsychological toxicity [36], though not hepatotoxicity [38]. The combination of c.516G>T and HLA-Cw*04 alleles showed a stronger association in black, white, and Asian populations than c.516 G>T alone [14]. In our cohort, however, the CYP2B6 c.516G>T polymorphism was not a significant risk factor for any of the hypersensitivity phenotypes (data not shown); therefore, a combination analysis has not been undertaken.

Cw*04 alleles showed a stronger association in black, white, and Asian populations than c.516 G>T alone [14]. In our cohort, however, the CYP2B6 c.516G>T polymorphism was not a significant risk factor for any of the hypersensitivity phenotypes (data not shown); therefore, a combination analysis has not been undertaken. Our study has several strengths: (1) we investigated a Malawian HIV cohort originating from a highly homogeneous population from a small geographic area; thus, the effect of ethnicity admixture is likely to be minimal; (2) sex- and age-matching of our tolerant controls also minimized the effect of these nongenetic factors; (3) when compared to previous studies, our sample size was larger; (4) we used strict phenotypic characterization with independent adjudication by a dermatologist; (5) the majority of patients were recruited prospectively where detailed phenotypic data could be gathered, although we did include 28 retrospectively identified patients; (6) close monitoring of patients with nonsevere rash, treated through and excluded as cases, further strengthened the phenotype by omitting potential false positives; and (7) we used sequence-based HLA typing to at least 4 digits, which is particularly important in this population because of the presence of some rare alleles. However, there are also some limitations. First, despite a large sample size, when subdividing the groups into phenotypes, the numbers within categories fell, limiting our power to detect true associations. This is a recognized drawback of studying nevirapine hypersensitivity, where the phenotypic manifestations not only vary, but have different allelic associations (Table 1). Second, we could not genotype all patients, particularly for the class II HLA alleles, which would have strengthened haplotype analysis, nevertheless, our study was larger than previous studies despite the missing data. Third, given the homogeneity of the Malawian population, it is possible that although the HLA association identified here is relevant, it may not be applicable to other ethnicities including other African populations.

haplotype analysis, nevertheless, our study was larger than previous studies despite the missing data. Third, given the homogeneity of the Malawian population, it is possible that although the HLA association identified here is relevant, it may not be applicable to other ethnicities including other African populations. In conclusion, we have identified an association between the HLA-C*04:01 allele nevirapine-induced hypersensitivity phenotypes, including the first report of an association between HLA-C*04:01 and the most severe phenotype, SJS/TEN. Our study appears to replicate previous observations [14] of an association between HLA-C*04:01 and risk of nevirapine cutaneous adverse drug reactions in a black population. Further work is required to replicate the association identified here, and to evaluate in more detail the effects of risk and competing HLA alleles. Additionally, functional in vitro or in silico models are needed to clarify the mechanisms of the immune-mediated response to nevirapine and its metabolites [39]. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

rdjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. The authors thank Dr Gavin Wong, University Hospital of South Manchester (Wythenshawe, UK) for independently reviewing the hypersensitivity phenotypes; and the patients and staff of the antiretroviral therapy clinic of Queen Elizabeth Central Hospital (Blantyre, Malawi), in particular S. Kaunda, Clinical Officer. Financial support. M. C. was funded by a 3-year Wellcome Trust training fellowship (WT078857MA) administered through the University of Liverpool. The Malawi-Liverpool-Wellcome Trust Clinical Research Programme is funded through a Core Programme Grant award from the Wellcome Trust. Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Hui and Hayden on pages 1104–6.) The emergence of human infections with avian influenza A(H7N9) virus further widens the spectrum of novel influenza A viruses that currently pose a threat to public health [1]. Although H7N9 virus has not been shown to transmit efficiently between humans, there are indications that the recently emerged H7N9 viruses are better adapted to replication in mammalian cells than other avian influenza A viruses and represent a plausible pandemic threat [2, 3]. H7N9 viruses isolated from human cases have amino acid sequences in the hemagglutinin (HA) protein that are associated with improved binding to α2–6-linked sialidases that are abundant on human respiratory epithelial cells, and in the polymerase and other proteins that are associated with increased virulence and transmissibility in mammals [2–4]. In ferret experiments, H7N9 virus replicates well in the upper respiratory tract following intranasal inoculation, causes relatively mild illness, and is efficiently transmitted by direct contact, but less so by respiratory droplets [2, 3, 5]. Intratracheal inoculation of ferrets results in severe pneumonia and high mortality [6]. In a ferret model, therefore, H7N9 virus possesses a constellation of features that are intermediate between highly pathogenic H5N1 viruses and fully adapted but less virulent human influenza A viruses such as influenza A subtypes H3N2 and pandemic H1N1/2009 (pH1N1).

virus replication, and experimental H7N9 virus infection of ferrets has provided little evidence of systemic replication [2, 34, 35]. H7N9 viral RNA has been detected in the serum, urine, and feces of H7N9 patients but it is not known if this represents viral replication occurring outside of the respiratory tract [35]. Hospitalized H7N9 patients had a case fatality risk that was intermediate between pH1N1 and H5N1 patients, and a more protracted clinical course than either H5N1 or pH1N1 patients, with the longest median time to death and the longest hospitalization. Whether this reflects the natural history of severe H7N9 virus infection, patient characteristics, or differences in the clinical management of patients with severe H7N9 compared with H5N1 patients, including increased frequency of rescue modalities such as extracorporeal membrane oxygenation, is unknown.

results in severe pneumonia and high mortality [6]. In a ferret model, therefore, H7N9 virus possesses a constellation of features that are intermediate between highly pathogenic H5N1 viruses and fully adapted but less virulent human influenza A viruses such as influenza A subtypes H3N2 and pandemic H1N1/2009 (pH1N1). Despite meeting the criteria for a low pathogenic phenotype in birds, H7N9 virus has caused severe and fatal disease in humans [7]. However, the demographic profile of patients with H7N9 virus infection is unusual, with a high median age and an excess of males [8]. Although this might be due to age and sex differences in exposures to infected poultry or settings contaminated by infected poultry, the pattern differs markedly from H5N1 cases, and would also be consistent with age-dependent biological cofactors contributing to pathogenesis and disease severity [8]. An assessment of the clinical severity of human infections with H7N9 virus has concluded that many mild cases may have occurred and the overall symptomatic case fatality risk is estimated to be <3% [7]. Understanding the determinants of the severity of disease due to H7N9 virus infection is important both for the identification and clinical management of high-risk cases and for the purposes of public health risk assessment and contingency planning.

the overall symptomatic case fatality risk is estimated to be <3% [7]. Understanding the determinants of the severity of disease due to H7N9 virus infection is important both for the identification and clinical management of high-risk cases and for the purposes of public health risk assessment and contingency planning. To assess whether the H7N9 virus genotype translates into a distinct clinical phenotype in humans, and to provide insights into the pathogenesis of H7N9 virus infection, we compared the risk factors, clinical presentation, and progression of patients hospitalized with H7N9, H5N1, and pH1N1 virus infections.

the overall symptomatic case fatality risk is estimated to be <3% [7]. Understanding the determinants of the severity of disease due to H7N9 virus infection is important both for the identification and clinical management of high-risk cases and for the purposes of public health risk assessment and contingency planning. To assess whether the H7N9 virus genotype translates into a distinct clinical phenotype in humans, and to provide insights into the pathogenesis of H7N9 virus infection, we compared the risk factors, clinical presentation, and progression of patients hospitalized with H7N9, H5N1, and pH1N1 virus infections. METHODS Subject Ascertainment All subjects with influenza virus infection reported in this manuscript were hospitalized patients. The patients with laboratory-confirmed H7N9 infection were all hospitalized in China between 25 February and 4 May 2013. The Chinese H5N1 cases represent all hospitalized cases of laboratory-confirmed H5N1 virus infection detected between 30 November 2003 and 8 February 2012. The Vietnamese H5N1 cases represent all hospitalized cases of laboratory-confirmed H5N1 virus infection detected between 25 December 2003 and 14 March 2009 [9]. A comparison of the Chinese and Vietnamese H5N1 cases showed similar demographic characteristics, underlying medical conditions, and behavioral risk factors (Supplementary Data). Patients with pH1N1 virus infection in China were ascertained through hospitals designated for the treatment of severe cases. The case definitions and time periods for ascertaining patients hospitalized with influenza A H5N1, H7N9, and pH1N1 virus infections are available in the Supplementary Data.

Supplementary Data). Patients with pH1N1 virus infection in China were ascertained through hospitals designated for the treatment of severe cases. The case definitions and time periods for ascertaining patients hospitalized with influenza A H5N1, H7N9, and pH1N1 virus infections are available in the Supplementary Data. Clinical and laboratory data were abstracted retrospectively from original medical records for cases of H7N9, H5N1, and pH1N1 virus infections. Laboratory values were presented as medians with interquartile ranges and were dichotomized into normal or abnormal based on normal ranges for children and adults (Supplementary Table 1). Because the only subjects aged <29 days were 5 subjects with pH1N1 virus infection, and normal laboratory values are different in neonates compared with other age groups, we excluded all subjects aged <29 days from the assessment of laboratory results. We excluded pH1N1 cases from the analysis of signs and symptoms on admission as the ascertainment process for these cases required the presence of 1 or more symptoms, many of which were severe. Ethics Statement The Chinese National Health and Family Planning Commission determined that the collection of data from H5N1, H7N9, and pH1N1 cases was part of public health investigations of emerging influenza outbreaks and was exempt from institutional review board assessment. The Science and Ethics Committee of the Ministry of Science and Technology of Vietnam approved the collection of clinical data from Vietnamese subjects with H5N1 virus infection.

1N1 cases was part of public health investigations of emerging influenza outbreaks and was exempt from institutional review board assessment. The Science and Ethics Committee of the Ministry of Science and Technology of Vietnam approved the collection of clinical data from Vietnamese subjects with H5N1 virus infection. Risk Factors for Hospitalization and Death To assess the importance of putative risk factors for hospitalization with each influenza A subtype, we estimated the relative risk of being hospitalized in subjects with and without risk factors. Data on the prevalence of each risk factor in the general Chinese population were used as denominators for the risk estimates and to weight (adjust) the overall relative risk estimates by age and sex. Data on age- and sex-specific population prevalence were available for coronary heart disease, chronic renal disease, diabetes, hypertension, smoking, and obesity; age-specific but not sex-specific population prevalence was available for asthma and chronic obstructive pulmonary disease (COPD) [10–13]. The definitions for these conditions are shown in the Supplementary Data. The age- and sex-stratified population prevalence of chronic heart disease (CHD; excluding isolated hypertension) was estimated from a study that recorded a prior history of hospitalization with coronary artery disease (A history of hospitalization for myocardial infarction or a surgical history of coronary balloon angioplasty, or coronary stent implantation or coronary artery bypass.) [10]. We assumed that the age distribution of coronary artery disease is a valid proxy for the age distribution of CHD. Where surveys assessed disease prevalence only in older adults, we assumed that prevalence was zero in those younger than the lower age limit of the survey. Because we were not able to source relevant baseline data for Vietnam, we have assumed that the age distribution of chronic diseases is similar in the Chinese and Vietnamese populations.

disease prevalence only in older adults, we assumed that prevalence was zero in those younger than the lower age limit of the survey. Because we were not able to source relevant baseline data for Vietnam, we have assumed that the age distribution of chronic diseases is similar in the Chinese and Vietnamese populations. Statistical Methods We compared the characteristic of patients infected by different subtypes using Fisher exact test or χ2 test for comparing proportions and Wilcoxon signed-rank test for comparing medians of continuous variables. To evaluate the association between risk factors and the risk of hospitalization, Poisson regression was used to estimate the incidence rate ratios associated with each risk factor, adjusted for age and sex. The association between risk factors and the risk of death among hospitalized cases was assessed using multivariable logistic regression to estimate the odds ratios associated with each risk factor, adjusted for age and sex. In both analyses a spline function was used for age to allow for the possibly nonlinear effect of age on risk. We used the Kaplan-Meier method to estimate survival curves for death and the hospitalized fatality risk. We used the same approach to estimate the time to invasive mechanical ventilation. The censoring time of each recovered or nonventilated patient was set to 90 days. The 95% confidence intervals (CIs) for the cumulative proportion of subjects requiring invasive ventilation and with a fatal outcome were estimated using bootstrapping with 1000 resamples.

mate the time to invasive mechanical ventilation. The censoring time of each recovered or nonventilated patient was set to 90 days. The 95% confidence intervals (CIs) for the cumulative proportion of subjects requiring invasive ventilation and with a fatal outcome were estimated using bootstrapping with 1000 resamples. We used maximum likelihood to estimate the distribution of the number of days of hospitalization, and compared alternative parametric distributions including γ, Weibull, and log-normal distributions, selecting the best parametric distribution on the basis of the Akaike information criterion. RESULTS As of 6 August 2013, 133 laboratory-confirmed influenza A(H7N9) cases have been officially recorded in mainland China. Among these, 123 requiring hospitalization for medical reasons were included in this study [7]. Ten laboratory-confirmed mild cases were excluded [14]. Data were included for 119 patients hospitalized with H5N1 (Vietnam = 76; China = 43), and 3486 patients hospitalized with pH1N1.

officially recorded in mainland China. Among these, 123 requiring hospitalization for medical reasons were included in this study [7]. Ten laboratory-confirmed mild cases were excluded [14]. Data were included for 119 patients hospitalized with H5N1 (Vietnam = 76; China = 43), and 3486 patients hospitalized with pH1N1. The median age of subjects hospitalized with H7N9 was 63 years, compared to 26 years for H5N1 patients and 25 years for pH1N1 patients (P < .001). A higher proportion of H7N9 subjects were male compared with H5N1 (P = .019) or pH1N1 subjects (P = .001). Subjects hospitalized with H7N9 had the highest prevalence of chronic medical conditions traditionally associated with an increased risk of severe seasonal influenza disease (Table 1). CHD and diabetes were the commonest medical risk factors reported among H7N9 patients, and the prevalence of smoking and hypertension was higher in subjects with H7N9 compared with the other patient groups. Pregnancy was more common in subjects hospitalized with pH1N1. Table 1. Characteristics of Subjects Hospitalized With Influenza A Virus Subtypes H7N9, H5N1, and pH1N1

ors reported among H7N9 patients, and the prevalence of smoking and hypertension was higher in subjects with H7N9 compared with the other patient groups. Pregnancy was more common in subjects hospitalized with pH1N1. Table 1. Characteristics of Subjects Hospitalized With Influenza A Virus Subtypes H7N9, H5N1, and pH1N1 Characteristic H7N9a H5N1 P Value pH1N1 P Value Age, y, median (range) 63 (4–91) 26 (1–75) <.001 25 (0–100) <.001 Interval from onset, admission days (IQR) 4 (3–6) 5 (3–6) .155 4 (3–6) .244 Male sex 87/123 (71%) 67/119 (56%) .019 1937/3486 (56%) .001 Any coexisting chronic medical conditions 42/105 (40%) 11/104 (11%) <.001 748/3485 (21%) <.001 Chronic heart disease 12/105 (11%) 1/102 (1%) .001 147/3457 (4%) .003 Chronic lung disease 10/105 (10%) 6/100 (6%) .344 305/3397 (9%) .849 Chronic renal disease 1/105 (1%) 1/102 (1%) .984 91/3450 (3%) .221 Chronic liver disease 5/105 (5%) 1/101 (1%) .092 27/3478 (1%) .002 Chronic neurological disease 3/105 (3%) 0/39 (0%) .166 55/3472 (2%) .356 Diabetes 18/105 (17%) 1/100 (1%) <.001 185/3470 (5%) <.001 Asthma 0/105 (0%) 0/0   NA 102/3442 (3%) .013 Immune compromise 2/105 (2%) 1/100 (1%) .586 86/3433 (3%) .685 Hypertension 51/105 (49%) 2/41 (5%) <.001 366/3479 (11%) <.001 Malignancy 6/105 (6%) 1/41 (2%) .375 92/3468 (3%) .096 Pregnancy 2/105 (2%) 5/106 (5%) .246 400/3436 (12%) <.001 Smoking history 26/105 (25%) 10/88 (11%) .015 541/3431 (16%) .02 Obesity (BMI ≥30) 3/45 (7%) 0/10 (0%) .265 175/2018 (9%) .623 Any coexisting chronic medical conditions are any of the following: asthma, diabetes, chronic respiratory disease, chronic heart disease, chronic renal disease, chronic hepatic (liver) disease, chronic neurological disease, immune compromise (see Supplementary Data for definitions).

%) 0/10 (0%) .265 175/2018 (9%) .623 Any coexisting chronic medical conditions are any of the following: asthma, diabetes, chronic respiratory disease, chronic heart disease, chronic renal disease, chronic hepatic (liver) disease, chronic neurological disease, immune compromise (see Supplementary Data for definitions). Abbreviations: BMI, body mass index; IQR, interquartile range; pH1N1, 2009 pandemic H1N1 virus. a Reference group. Compared with subjects without CHD, the presence of CHD was associated with an increased risk of hospitalization with H7N9 (relative risk [RR], 9.68; 95% CI, 5.24–17.9; Table 2). CHD was also a risk factor for hospitalization with pH1N1 (RR, 16.51; 95% CI, 13.68–19.91). Hypertension was not associated with an increased risk of hospitalization in any group, whereas a history of smoking was associated with a reduced risk of hospitalization. Chronic renal disease was associated with a reduced risk of hospitalization in H7N9 and pH1N1 patients. Once patients were hospitalized, the odds of death were not significantly increased in subjects with any of the risk factors examined (Table 3). Table 2. Age- and Sex-Adjusted Risk Factors for Hospitalization

zation. Chronic renal disease was associated with a reduced risk of hospitalization in H7N9 and pH1N1 patients. Once patients were hospitalized, the odds of death were not significantly increased in subjects with any of the risk factors examined (Table 3). Table 2. Age- and Sex-Adjusted Risk Factors for Hospitalization Risk Factora Source of Baseline Prevalence Data H7N9 H5N1 pH1N1 RR (95% CI)b RR (95% CI)b RR (95% CI)b Asthmac [12, 15] NC NC 1.76 (1.43–2.15) COPDc (assume zero prevalence aged <40 y) [11] 0.73 (.35–1.52) 4.25 (1.34–13.48) 1.76 (1.43–2.18) Diabetes (assume zero prevalence aged <20 y) [10] 1.11 (.67–1.87) 0.23 (.03–1.67) 1.11 (.94–1.30) Chronic heart disease (assume zero prevalence aged <20 y) [10] 9.68 (5.24–17.9) NC 16.51 (13.68–19.91) Chronic renal disease (assume zero prevalence aged <18 y) [13] 0.07 (.01–.54) NC 0.47 (.37–.58) Hypertension (assume zero prevalence aged <20 y) [10] 1.28 (.85–1.91) 0.45 (.10–1.99) 0.63 (.55–.71) Smokingd [10] 0.38 (.24–.60) 0.41 (.20–.88) 0.74 (.66–.84) Obesity (BMI ≥30)c [10] 1.16 (.36–3.74) NC 2.42 (2.03–2.88) Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; NC, not calculable due to insufficient data; RR, relative risk. a See the Supplementary Data for definitions. b Adjusted for cubic spline for age (continuous) and sex where data were available. c Sex-specific data not available. d Restricted to subjects aged ≥20 years only. Table 3. Age- and Sex-Adjusted Risk Factors for Death Among Hospitalized Patients

Risk Factora Source of Baseline Prevalence Data H7N9 H5N1 pH1N1 RR (95% CI)b RR (95% CI)b RR (95% CI)b Asthmac [12, 15] NC NC 1.76 (1.43–2.15) COPDc (assume zero prevalence aged <40 y) [11] 0.73 (.35–1.52) 4.25 (1.34–13.48) 1.76 (1.43–2.18) Diabetes (assume zero prevalence aged <20 y) [10] 1.11 (.67–1.87) 0.23 (.03–1.67) 1.11 (.94–1.30) Chronic heart disease (assume zero prevalence aged <20 y) [10] 9.68 (5.24–17.9) NC 16.51 (13.68–19.91) Chronic renal disease (assume zero prevalence aged <18 y) [13] 0.07 (.01–.54) NC 0.47 (.37–.58) Hypertension (assume zero prevalence aged <20 y) [10] 1.28 (.85–1.91) 0.45 (.10–1.99) 0.63 (.55–.71) Smokingd [10] 0.38 (.24–.60) 0.41 (.20–.88) 0.74 (.66–.84) Obesity (BMI ≥30)c [10] 1.16 (.36–3.74) NC 2.42 (2.03–2.88) Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; NC, not calculable due to insufficient data; RR, relative risk. a See the Supplementary Data for definitions. b Adjusted for cubic spline for age (continuous) and sex where data were available. c Sex-specific data not available. d Restricted to subjects aged ≥20 years only. Table 3. Age- and Sex-Adjusted Risk Factors for Death Among Hospitalized Patients Risk Factora H7N9 H5N1 pH1N1 Deathb, OR (95% CI) Deathb, OR (95% CI) Deathb,OR (95% CI) Asthma NC NC 0.24 (.06–1.01) COPD 2.55 (.38–17.20) 0.92 (.12–6.83) 0.98 (.51–1.89) Diabetes 3.68 (.97–14.03) NC 0.85 (.51–1.44) Chronic heart disease 0.96 (.18–5.17) NC 1.22 (.72–2.08) Chronic renal disease NC NC 1.56 (.86–2.80) Hypertension 1.06 (.36–3.13) 0.24 (.01–6.92) 0.87 (.58–1.29) Smoking 0.66 (.20–2.17) 1.23 (.25–5.99) 1.12 (.79–1.60) Obesity (BMI ≥30) NC NC 0.96 (.59–1.56) Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; NC, not calculable due to insufficient data; OR, odds ratio.

rtension 1.06 (.36–3.13) 0.24 (.01–6.92) 0.87 (.58–1.29) Smoking 0.66 (.20–2.17) 1.23 (.25–5.99) 1.12 (.79–1.60) Obesity (BMI ≥30) NC NC 0.96 (.59–1.56) Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; NC, not calculable due to insufficient data; OR, odds ratio. a See Supplementary Data for definitions. b Adjusted for cubic spline for age (continuous) and sex. Signs and symptoms at hospital admission were compared for H7N9 and H5N1 cases. Subjects with H7N9 virus infection were more likely to report a fever, a productive cough, and hemoptysis than those with H5N1 virus infection (Table 4). Gastrointestinal symptoms were most common in H5N1 cases. Table 4. Signs and Symptoms on Admissiona Sign or Symptom H7N9 H5N1 P Value Fever (temp ≥37.8) 99/105 (94%) 75/102 (74%) <.001 Any cough 96/105 (91%) 89/106 (84%) .097 Productive cough 59/104 (57%) 35/94 (37%) .006 Dry cough 17/105 (16%) 45/94 (48%) <.001 Yellow sputum 33/105 (31%) 10/61 (16%) .029 Hemoptysis 25/105 (24%) 5/61 (8%) .008 Myalgia 21/105 (20%) 12/50 (24%) .572 Fatigue 38/105 (36%) 9/37 (24%) .179 Shortness of breath 62/105 (59%) 54/93 (58%) .889 Gastrointestinal symptoms 15/105 (14%) 17/53 (32%) .01 Diarrhea 10/105 (10%) 6/50 (12%) .64 Vomiting 4/105 (4%) 10/54 (19%) .003 Nausea 6/105 (6%) 7/50 (14%) .093 Central nervous system symptoms 4/105 (4%) 8/113 (7%) .285 a Or earliest available time point after admission.

rtness of breath 62/105 (59%) 54/93 (58%) .889 Gastrointestinal symptoms 15/105 (14%) 17/53 (32%) .01 Diarrhea 10/105 (10%) 6/50 (12%) .64 Vomiting 4/105 (4%) 10/54 (19%) .003 Nausea 6/105 (6%) 7/50 (14%) .093 Central nervous system symptoms 4/105 (4%) 8/113 (7%) .285 a Or earliest available time point after admission. The values of hematological, liver, and renal function tests, and markers of inflammation on admission are shown in Table 5. H7N9 and H5N1 patients showed similar patterns of elevated alanine aminotransferase, creatinine kinase, C-reactive protein, and lactate dehydrogenase, which were all significantly higher than in pH1N1 patients. Leukopenia and thrombocytopenia were equally common in patients with H7N9 and H5N1 virus infections, and more common than in those with pH1N1 virus infection. Lymphopenia was more common in patients with H7N9 compared with H5N1 (88% vs 55%; P < .001), and neutropenia was more common in H5N1 patients. Neutrophilia was equally common in H5N1 and pH1N1 patients, and least common in H7N9 patients. Table 5. Laboratory Results on Admissiona

han in those with pH1N1 virus infection. Lymphopenia was more common in patients with H7N9 compared with H5N1 (88% vs 55%; P < .001), and neutropenia was more common in H5N1 patients. Neutrophilia was equally common in H5N1 and pH1N1 patients, and least common in H7N9 patients. Table 5. Laboratory Results on Admissiona Result H7N9b H5N1 P Value pH1N1 P Value White cell count 4.5 (2.9–6.2) 3.9 (2.5–7.1) .805 6 (4.2–8.8) <.001 Lymphocyte count 0.5 (0.3–0.7) 0.9 (0.6–1.4) <.001 1 (0.6–1.5) <.001 Neutrophil count 3.3 (2.2–5.4) 3 (1.5–5.4) .203 4.3 (2.6–6.9) .004 Platelet count 114 (82–147.5) 126 (86–196) .203 173 (132–229.8) .004 AST 53 (38–96.5) 100 (47–233) .076 40 (26.4–68.5) <.001 ALT 35.5 (24–64.5) 48.5 (29.5–99.5) <.001 24 (15.6–44) <.001 Serum creatinine 70.7 (58.3–85) 83 (54–100) .028 62 (45.4–81) <.001 CK 195 (96–562) 552 (126.5–939.8) .255 120 (62–304) <.001 CRP 65 (25–113) 51 (14.2–118.3) .191 25.4 (7.9–75.5) <.001 LDH 498 (388–661) 1025 (334.8–1832.5) .525 307 (217–491) <.001 Leukopenia 48/105 (46%) 54/107 (50%) .489 736/3305 (22%) <.001 Lymphopenia 88/99 (89%) 54/98 (55%) <.001 1601/2891 (55%) <.001 Neutropenia 13/103 (13%) 24/97 (25%) .027 221/2891 (8%) .086 Neutrophilia 5/103 (5%) 15/97 (15%) .011 477/2891 (16%) <.001 Thrombocytopenia 80/104 (77%) 69/105 (66%) .073 1106/3066 (36%) <.001 Elevated AST 54/103 (52%) 41/54 (76%) .004 1165/3197 (36%) .001 Elevated ALT 34/100 (34%) 25/52 (48%) .093 668/3167 (21%) .003 Elevated serum creatinine 11/103 (11%) 9/62 (15%) .469 201/3054 (7%) .129 Elevated CK 48/98 (49%) 13/20 (65%) .188 1018/2951 (34%) .004 Elevated CRP 83/92 (90%) 9/12 (75%) .162 1193/1708 (70%) <.001 Elevated LDH 89/98 (91%) 17/21 (81%) .218 1617/2922 (55%) <.001 Data are presented as median (IQR) or No. (%).

(21%) .003 Elevated serum creatinine 11/103 (11%) 9/62 (15%) .469 201/3054 (7%) .129 Elevated CK 48/98 (49%) 13/20 (65%) .188 1018/2951 (34%) .004 Elevated CRP 83/92 (90%) 9/12 (75%) .162 1193/1708 (70%) <.001 Elevated LDH 89/98 (91%) 17/21 (81%) .218 1617/2922 (55%) <.001 Data are presented as median (IQR) or No. (%). Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; CK, creatine kinase; CRP, C-reactive protein; IQR, interquartile range; LDH, lactate dehydrogenase; pH1N1, 2009 pandemic H1N1 virus. a Or earliest available time point after admission. b Reference group.

(21%) .003 Elevated serum creatinine 11/103 (11%) 9/62 (15%) .469 201/3054 (7%) .129 Elevated CK 48/98 (49%) 13/20 (65%) .188 1018/2951 (34%) .004 Elevated CRP 83/92 (90%) 9/12 (75%) .162 1193/1708 (70%) <.001 Elevated LDH 89/98 (91%) 17/21 (81%) .218 1617/2922 (55%) <.001 Data are presented as median (IQR) or No. (%). Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; CK, creatine kinase; CRP, C-reactive protein; IQR, interquartile range; LDH, lactate dehydrogenase; pH1N1, 2009 pandemic H1N1 virus. a Or earliest available time point after admission. b Reference group. The risk of invasive ventilation and death among hospitalized cases by influenza A virus subtype are shown in Figure 1. The cumulative proportion of hospitalized subjects requiring invasive ventilation differs between subtypes, reaching 62% (95% CI, 53%–71%) for H7N9, 54% (95% CI, 45%–63%) for H5N1, and 17% (95% CI, 15%–18%) for pH1N1. Among those ventilated, the interval from onset to invasive ventilation was a median of 7 days for both H7N9 and H5N1 cases (P = .651), and 6 days for pH1N1 cases. The hospitalized case fatality risk was highest for H5N1 (55%; 95% CI, 47%–64%) and death occurred earlier, with a median time from onset to death of 11 days for H5N1, compared with 15 days for pH1N1 patients (P = .154) and 18 days for H7N9 (P = .002). H7N9 patients were hospitalized for a longer duration than either H5N1 (P < .001) or pH1N1 patients (P < .001) (Figure 2). Figure 1. Case fatality risk and invasive ventilation risk in hospitalized patients. A and B, Case fatality risk by influenza A virus subtype and day of hospitalization (A) and day of illness onset (B). C and D, Invasive ventilation risk by influenza A virus subtype and day of hospitalization (C) and day of illness onset (D). Abbreviation: pH1N1, 2009 pandemic H1N1 virus.

risk in hospitalized patients. A and B, Case fatality risk by influenza A virus subtype and day of hospitalization (A) and day of illness onset (B). C and D, Invasive ventilation risk by influenza A virus subtype and day of hospitalization (C) and day of illness onset (D). Abbreviation: pH1N1, 2009 pandemic H1N1 virus. Figure 2. Distribution of the number of days of hospitalization for patients with H7N9, H5N1, and pH1N1.

risk in hospitalized patients. A and B, Case fatality risk by influenza A virus subtype and day of hospitalization (A) and day of illness onset (B). C and D, Invasive ventilation risk by influenza A virus subtype and day of hospitalization (C) and day of illness onset (D). Abbreviation: pH1N1, 2009 pandemic H1N1 virus. Figure 2. Distribution of the number of days of hospitalization for patients with H7N9, H5N1, and pH1N1. DISCUSSION One of the most striking differences in this and other comparative analysis is the high median age of H7N9 patients [16]. This age distribution is unlikely to be due to differences in humoral immunity as the prevalence of neutralizing antibodies to H7N9 virus is probably low in all ages [17–20]. It might arise either because elderly people are more often exposed to the animal or environmental reservoir of H7N9 viruses, or because elderly people have a greater propensity to become infected or severely ill following exposure. After adjusting for the age- and sex-specific prevalence of chronic illnesses in the general Chinese population, we found that CHD was associated with an increased risk of hospitalization with H7N9 virus infection (RR, 9.68; 95% CI, 5.24–17.9). The age distribution of H7N9 patients may therefore be partially explained by an increased propensity in persons with CHD (who are mostly older) to develop severe disease following infection with H7N9 virus. The overrepresentation of males among H7N9 patients may also be partially explained by this association, because in China coronary heart disease is commoner in males than females (male prevalence, 0.74%; female prevalence, 0.51%) [10]. In agreement with our results, an age- and sex-matched case control study of 25 H7N9 cases has reported that the presence of a preexisting chronic medical condition (excluding hypertension) was associated with H7N9 disease (odds ratio, 5.1; 95% CI, 1.5–16.9) [21]. Although only 11% of H7N9 patients reported a history of CHD, unrecognized CHD may have been present in some individuals, and other unmeasured age-related factors, such as impaired innate and cell-mediated immunity, might also contribute to the observed age distribution of hospitalized H7N9 cases [17, 22, 23]. H7N9 viruses isolated from humans exhibit a mixed receptor specificity, binding both α2–6- and α2–3-linked sialidases [3, 4, 20]. H7N9 virus can infect cells of both the upper and lower respiratory tract of humans and ferrets, and disease in ferrets is more severe following intratracheal inoculation [5, 6, 20].

2, 23]. H7N9 viruses isolated from humans exhibit a mixed receptor specificity, binding both α2–6- and α2–3-linked sialidases [3, 4, 20]. H7N9 virus can infect cells of both the upper and lower respiratory tract of humans and ferrets, and disease in ferrets is more severe following intratracheal inoculation [5, 6, 20]. This raises the possibility that susceptibility of humans to severe H7N9 disease may be a consequence of an impaired ability to control virus replication in the lower respiratory tract. A history of chronic renal disease was associated with a reduced risk of hospitalization with H7N9 virus infection, but the number of patients with this condition was small, so this finding should be interpreted with caution. A history of smoking was associated with a reduced risk of hospitalization with H7N9, H5N1, and pH1N1 virus infections. This is an unexpected finding that might be biased by inconsistent definitions and methods of ascertaining smoking history, which were not standardized in the clinical datasets.

erpreted with caution. A history of smoking was associated with a reduced risk of hospitalization with H7N9, H5N1, and pH1N1 virus infections. This is an unexpected finding that might be biased by inconsistent definitions and methods of ascertaining smoking history, which were not standardized in the clinical datasets. The clinical presentation and laboratory indices at hospital admission are similar for H7N9 and H5N1 patients, except that a productive cough, hemoptysis, lymphopenia, and neutropenia were more common in H7N9 patients. Neutropenia, thrombocytopenia, and elevated liver enzymes are common in H5N1 patients and have been associated with more severe outcomes [9, 24–29]. A low absolute lymphocyte count has been associated with poor outcomes in patients hospitalized with pH1N1, H5N1, and severe acute respiratory syndrome [9, 30–32]. The hematological and serum chemistry abnormalities suggest that subjects hospitalized with H7N9 have a severe systemic illness. It remains to be determined if this is a consequence of severe pneumonia and poor tissue oxygenation or is the result of an excessive inflammatory response (as is seen with H5N1 virus infection) [33]. High levels of chemokines and cytokines have been identified in patients with H7N9 virus infection [20]. Extrapulmonary virus replication is an alternative explanation for the severity of hospitalized H7N9 cases, but H7N9 virus does not posses the polybasic amino acid motif at the HA cleavage site normally associated with extrapulmonary virus replication, and experimental H7N9 virus infection of ferrets has provided little evidence of systemic replication [2, 34, 35]. H7N9 viral RNA has been detected in the serum, urine, and feces of H7N9 patients but it is not known if this represents viral replication occurring outside of the respiratory tract [35].

pitalization. Whether this reflects the natural history of severe H7N9 virus infection, patient characteristics, or differences in the clinical management of patients with severe H7N9 compared with H5N1 patients, including increased frequency of rescue modalities such as extracorporeal membrane oxygenation, is unknown. The comparisons we have made are limited by a lack of standardization of the methods of case ascertainment and inclusion, and of the recording of clinical and other data. As such, the patients and data included in this study may be subject to unmeasured selection and information biases and differences in practices over time and between locations. However, we have tried to minimize these potential biases by restricting our analysis only to hospitalized subjects and to variables where data were available for a reasonable proportion of all cases. Although the H5N1 patients from China and Vietnam had very similar demographic characteristics, underlying medical conditions, and behavioral risk factors, there were some differences in clinical presentation (Supplementary Data), and we cannot exclude that the clinical phenotype of H5N1 virus infections may be heterogeneous. We used univariate analysis that adjusted for age and sex to explore possible risk factors for hospitalization with H7N9; interactions effects were not assessed and the estimated odds ratios and RRs might be confounded by other unmeasured confounders; as such, these risk factors should not be considered to be causal without further validation.

is that adjusted for age and sex to explore possible risk factors for hospitalization with H7N9; interactions effects were not assessed and the estimated odds ratios and RRs might be confounded by other unmeasured confounders; as such, these risk factors should not be considered to be causal without further validation. In conclusion, this comparative analysis shows that patients hospitalized with H7N9 virus infection share some risk factors with those hospitalized with pH1N1 infection but have a clinical profile more closely resembling that of H5N1 patients. The identification in H7N9 patients of known risk factors for severe seasonal influenza and the more protracted clinical course compared with H5N1 patients suggests that host factors may be an important contributor to the severity of H7N9 virus infection. This is consistent with the observation that there have probably been a large number of undetected mild H7N9 virus infections, and to date the patients with detected mild infection have been predominantly young (mean age, 13 years) [7, 14]. H7N9 virus has recently reemerged in China. People with chronic medical conditions that are traditionally associated with a higher risk of severe complications following seasonal influenza virus infection should be targeted for preventive interventions and for early treatment with antiviral drugs should they develop a respiratory illness.

tly reemerged in China. People with chronic medical conditions that are traditionally associated with a higher risk of severe complications following seasonal influenza virus infection should be targeted for preventive interventions and for early treatment with antiviral drugs should they develop a respiratory illness. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

ordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We thank staff members of the Bureau of Disease Control and Prevention and Health Emergency Response Office of the National Health and Family Planning Commission and provincial and local departments of health for providing assistance with administration and data collection; staff members at county-, prefecture-, and provincial-level CDCs at the provinces where human H7N9, H5N1, and pandemic H1N1 cases occurred for providing assistance with field investigation, administration, and data collection. We thank Dr Jian Hua Hu of Zhejiang University for her help in data collection. We thank staff members of the National Hospital for Tropical Diseases, the National Pediatric Hospital, and the Hospital for Tropical Diseases, Ho Chi Minh City, Vietnam for assistance with enrolling patients with H5N1 infection and collection of data. Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the China Centers for Disease Control and Prevention or the US Centers for Disease Control and Prevention. The funding bodies had no role in study design, data collection and analysis, preparation of the manuscript, or the decision to publish.

ose of the authors and do not necessarily represent the official position of the China Centers for Disease Control and Prevention or the US Centers for Disease Control and Prevention. The funding bodies had no role in study design, data collection and analysis, preparation of the manuscript, or the decision to publish. Financial support. This study was funded by the US National Institutes of Health (Comprehensive International Program for Research on AIDS, grant number U19 AI51915); the Ministry of Science and Technology, China (grant number 2012 ZX10004-201); the National Program for Prevention and Control of human infections by avian-origin H7N9 influenza A virus (grant number KJYJ-2013-01); the National Natural Science Foundation of China (grant numbers 81070005/H0104, 81030032/H19 and 81271840); the National Major S & T Research Projects for the Control and Prevention of Major Infectious Diseases in China (grant numbers 2012ZX10004-210, 2012ZX10004–206); the Technology Group Project for Infectious Disease Control of Zhejiang Province (grant number 2009R50041); and the Fundamental Research Funds for the Central Universities. P. W. H. is supported by the Wellcome Trust (grant numbers 089276/Z/09/Z and 089276/B/09/Z). Potential conflicts of interest. B. J. C. has received research funding from MedImmune Inc, and consults for Crucell NV. D. K. M. I. has received research funding from Hoffmann-La Roche. G. M. L. has received speakers’ honoraria from HSBC and CLSA. All other authors report no potential conflicts.

Financial support. This study was funded by the US National Institutes of Health (Comprehensive International Program for Research on AIDS, grant number U19 AI51915); the Ministry of Science and Technology, China (grant number 2012 ZX10004-201); the National Program for Prevention and Control of human infections by avian-origin H7N9 influenza A virus (grant number KJYJ-2013-01); the National Natural Science Foundation of China (grant numbers 81070005/H0104, 81030032/H19 and 81271840); the National Major S & T Research Projects for the Control and Prevention of Major Infectious Diseases in China (grant numbers 2012ZX10004-210, 2012ZX10004–206); the Technology Group Project for Infectious Disease Control of Zhejiang Province (grant number 2009R50041); and the Fundamental Research Funds for the Central Universities. P. W. H. is supported by the Wellcome Trust (grant numbers 089276/Z/09/Z and 089276/B/09/Z). Potential conflicts of interest. B. J. C. has received research funding from MedImmune Inc, and consults for Crucell NV. D. K. M. I. has received research funding from Hoffmann-La Roche. G. M. L. has received speakers’ honoraria from HSBC and CLSA. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Combination antiretroviral therapy (cART) has led to declining morbidity and mortality in resource-poor settings [1, 2], and scale-up at the end of 2012 had reached 9.7 million human immunodeficiency virus (HIV)–infected individuals worldwide [3]. Optimal utilization of first-line cART and switch to second-line therapy in resource-poor settings is a priority. World Health Organization (WHO) guidelines recommend routine viral load monitoring (VLM), and switch to second-line therapy is recommended after 2 viral load measurements >1000 copies/mL following adherence counseling [4]. However, the WHO document recognizes that the evidence base for VLM itself is weak. Given that such a monitoring strategy is likely to be a huge burden for most resource-limited settings, it is important to increase the evidence base. Furthermore, there is a limited body of data on how viremia evolves on therapy in absence of VLM, and the impact on emergence of drug resistance; such information is needed to inform treatment guidelines.

gy is likely to be a huge burden for most resource-limited settings, it is important to increase the evidence base. Furthermore, there is a limited body of data on how viremia evolves on therapy in absence of VLM, and the impact on emergence of drug resistance; such information is needed to inform treatment guidelines. The Development of Anti Retroviral Therapy in Africa (DART) study compared clinical monitoring only with clinical and laboratory monitoring (CD4 and routine blood tests including biochemistry and full blood count), with switch to second-line therapy on clinical and immunologic criteria. This study demonstrated good clinical outcomes in both arms over the 5-year follow-up period [5]. In a substudy of DART, a comparison of zidovudine (ZDV)/lamivudine (3TC)/nevirapine (NVP) vs ZDV/3TC/abacavir (ABC) (NORA Study) showed triple nucleosides (ZDV/3TC/ABC) to be associated with higher rates of virologic and immunologic failure than the nonnucleoside reverse transcriptase inhibitor (NNRTI)–based regimen (ZDV/3TC/NVP) at 48 weeks [6, 7]. Here we report virologic outcomes at 96 weeks, demonstrating substantial resuppression following earlier viremia despite not switching.

) to be associated with higher rates of virologic and immunologic failure than the nonnucleoside reverse transcriptase inhibitor (NNRTI)–based regimen (ZDV/3TC/NVP) at 48 weeks [6, 7]. Here we report virologic outcomes at 96 weeks, demonstrating substantial resuppression following earlier viremia despite not switching. METHODS The NORA Study enrolled 600 previously untreated and asymptomatic Ugandan participants with CD4 counts of <200 cells/µL, randomly assigned to coformulated ZDV/ 3TC and either ABC and NVP placebo (ABC arm), or ABC placebo and NVP (NVP arm). Each drug was taken twice daily. After 24 weeks, participants continued to receive the study drugs open label and were followed as part of DART for a minimum of 4 years. In a separate randomized substudy, participants with a CD4 count ≥300 cells/µL at 48 or 72 weeks after ART initiation were eligible to be randomized to continuous therapy or structured treatment interruption (STI) with repeated 12-week periods on or off therapy [8]. Viral loads were retrospectively measured using Roche Amplicor 1.5.

randomized substudy, participants with a CD4 count ≥300 cells/µL at 48 or 72 weeks after ART initiation were eligible to be randomized to continuous therapy or structured treatment interruption (STI) with repeated 12-week periods on or off therapy [8]. Viral loads were retrospectively measured using Roche Amplicor 1.5. Analyses were based on participants who were alive, in follow-up, and still on first-line therapy at week 96, and who were not randomized to the STI arm in the STI substudy. Although the latter exclusion was essential because of the effect of STIs on viral load (and also possibly development of drug resistance), it introduces a different bias as eligibility for the STI substudy was related to earlier viral load values via the CD4 count inclusion criterion. The effect of this is the selective exclusion of participants with a good early virologic response and therefore, in crude analyses, underestimation of the rate of viral suppression at week 96. To account for this bias, inverse probability weights (separate for the 2 NORA arms) were used to up-weight participants who were randomized to continuous therapy. Week 96 samples with a viral load >1000 copies/mL underwent resistance testing by standard population sequencing of pol [6]. The frequencies of resistance-associated mutations [9] were calculated both for all participants (intention-to-treat) and for participants who had made no major substitutions (defined in the Results section) to their initial regimen (on-treatment). Participants with baseline resistance were excluded from analyses of resistance.

encies of resistance-associated mutations [9] were calculated both for all participants (intention-to-treat) and for participants who had made no major substitutions (defined in the Results section) to their initial regimen (on-treatment). Participants with baseline resistance were excluded from analyses of resistance. Ethics approval both for DART and the NORA substudy was obtained both in Uganda (Uganda Research Unit on AIDS Science and Ethics Committee) and the United Kingdom (Imperial College). RESULTS Of the 600 participants randomized in NORA (300 ABC arm, 300 NVP arm), 32 died before week 96 (13 ABC, 19 NVP), 21 were lost to follow-up (10 ABC, 11 NVP), and 107 were randomized to structured treatment interruption (37 ABC, 70 NVP). Seven participants (4 ABC, 3 NVP) switched to a second-line regimen based on lopinavir/ritonavir after week 48 and are excluded from all analyses; all achieved virologic suppression by 96 weeks. The number left for evaluation at week 96 was 236 and 197 in the ABC and NVP arms, respectively (Supplementary Figure 1). Twenty-five (11%) participants made a substitution in the ABC arm (from ABC to NVP or tenofovir [TDF]) and 28 (14%) in the NVP arm (from NVP to ABC or TDF).

hieved virologic suppression by 96 weeks. The number left for evaluation at week 96 was 236 and 197 in the ABC and NVP arms, respectively (Supplementary Figure 1). Twenty-five (11%) participants made a substitution in the ABC arm (from ABC to NVP or tenofovir [TDF]) and 28 (14%) in the NVP arm (from NVP to ABC or TDF). Consistent with previously reported week 48 data [6], the distribution of viral load at week 96 differed between the 2 arms (P < .001; Table 1), with a greater proportion of participants in the NVP arm achieving viral load suppression <1000 copies/mL. The viral load in the majority of participants with suppression was <200 copies/mL in both arms (91% and 95% of those <1000 copies/mL in the ABC and NVP arms, respectively). Table 1 shows the association between viral load at week 48 and week 96 for individual participants. Participants with viral load <1000 copies/mL at week 48 were likely to remain <1000 copies/mL at 96 weeks, although more so in the NVP arm (96% [149/156]) than in the ABC arm (82% [148/180]) (P = .003). Table 1. Distribution of Week 96 HIV RNA Viral Load by Week 48 HIV RNA and Treatment Arm

for individual participants. Participants with viral load <1000 copies/mL at week 48 were likely to remain <1000 copies/mL at 96 weeks, although more so in the NVP arm (96% [149/156]) than in the ABC arm (82% [148/180]) (P = .003). Table 1. Distribution of Week 96 HIV RNA Viral Load by Week 48 HIV RNA and Treatment Arm Viral Load ABC NVP Week 96 VL, Copies/mL Week 96 VL, Copies/mL <1000 1000–9999 10 000–99 999 ≥100 000 Total <1000 1000–9999 10 000–99 999 ≥100 000 Total Week 48 VL, copies/mL <1000 148 16 12 4 180 149 2 5 0 156 1000–9999 9 7 5 1 22 0 3 1 0 4 10 000–99 999 1 1 6 6 14 4 1 6 1 12 ≥100 000 2 0 4 4 10 3 0 2 3 8 Total 160 24 27 15 226 156 6 14 4 180 Adjusted total, % (95% CI) 71.1 (65.4–76.1) 11.6 (7.9–16.7) 11.5 (8.1–16.1) 5.8 (3.8–8.9) 88.7 (84.6–91.9) 3.6 (1.6–8.0) 5.9 (3.8–9.0) 1.7 (0.8–3.9) Adjusted percentages were calculated using inverse probability weights to account for missing values due to structured treatment interruption randomization (37 ABC, 70 NVP) or missing sample/failed assay (9 ABC, 11 NVP). Participants who died, were lost to follow-up, or started second-line treatment before week 96 were excluded (27 ABC, 33 NVP), as were 7 (1 ABC, 6 NVP) participants with no week 48 VL. Abbreviations: ABC, abacavir; CI, confidence interval; NVP, nevirapine; VL, viral load.

Viral Load ABC NVP Week 96 VL, Copies/mL Week 96 VL, Copies/mL <1000 1000–9999 10 000–99 999 ≥100 000 Total <1000 1000–9999 10 000–99 999 ≥100 000 Total Week 48 VL, copies/mL <1000 148 16 12 4 180 149 2 5 0 156 1000–9999 9 7 5 1 22 0 3 1 0 4 10 000–99 999 1 1 6 6 14 4 1 6 1 12 ≥100 000 2 0 4 4 10 3 0 2 3 8 Total 160 24 27 15 226 156 6 14 4 180 Adjusted total, % (95% CI) 71.1 (65.4–76.1) 11.6 (7.9–16.7) 11.5 (8.1–16.1) 5.8 (3.8–8.9) 88.7 (84.6–91.9) 3.6 (1.6–8.0) 5.9 (3.8–9.0) 1.7 (0.8–3.9) Adjusted percentages were calculated using inverse probability weights to account for missing values due to structured treatment interruption randomization (37 ABC, 70 NVP) or missing sample/failed assay (9 ABC, 11 NVP). Participants who died, were lost to follow-up, or started second-line treatment before week 96 were excluded (27 ABC, 33 NVP), as were 7 (1 ABC, 6 NVP) participants with no week 48 VL. Abbreviations: ABC, abacavir; CI, confidence interval; NVP, nevirapine; VL, viral load. Nineteen of 70 (27%) of individuals (12/46 ABC vs 7/24 NVP; P = .82) with viral load >1000 copies/mL at week 48 experienced resuppression by week 96, indicating issues with adherence. Sixty-seven of these 70 patients had drug resistance data at week 48; 10 of 57 (18%) individuals with at least 1 major mutation at week 48 had experienced resuppression by 96 weeks. Resistance patterns present in these 10 individuals were M184V (n = 3), M184V + D67N, M184V + T215Y, M184V + Y181C, M184V + D67N + K70R (n = 3), and Y188C. Among the remaining 10 individuals who had no resistance mutations at week 48, 7 (70%) were resuppressed, suggesting an improvement in adherence after week 48. Two of 3 individuals with no resistance result available at week 48 experienced resuppression by week 96.

M184V + Y181C, M184V + D67N + K70R (n = 3), and Y188C. Among the remaining 10 individuals who had no resistance mutations at week 48, 7 (70%) were resuppressed, suggesting an improvement in adherence after week 48. Two of 3 individuals with no resistance result available at week 48 experienced resuppression by week 96. Of 91 participants with viral load ≥1000 copies/mL) at week 96, 87 (96%) had a genotype available. Five (4 ABC, 1 NVP) participants with baseline resistance were excluded, leaving 82 (59 ABC, 23 NVP) patients for analysis. The frequencies of mutations for both the intention-to-treat and on-treatment populations are given in Supplementary Table 2. The following description focuses on the on-treatment population for simplicity. A high proportion of failures in the NVP arm had major NNRTI resistance at week 96 (95%). Thirteen (68%) had only 1 NNRTI mutation, and 5 (26%) participants had 2 NNRTI mutations. The M184V mutation, conferring resistance to 3TC, was highly prevalent (90% ABC, 89% NVP). The proportion of participants with ≥3 thymidine analogue mutations (TAMs) at week 96 was similar between the ABC group (49%) and the NVP group (42%) (P = .79). In the former group, ABC-specific mutations L74V and K65R were each seen in 1 individual. The pan–nucleoside resistance mutation Q151M was not observed in any individual.

ion of participants with ≥3 thymidine analogue mutations (TAMs) at week 96 was similar between the ABC group (49%) and the NVP group (42%) (P = .79). In the former group, ABC-specific mutations L74V and K65R were each seen in 1 individual. The pan–nucleoside resistance mutation Q151M was not observed in any individual. DISCUSSION We present 2-year virologic data from the DART-NORA study, highlighting the very good suppression rates achieved using ZDV/3TC and NVP. Viral failure as defined by WHO [4] was almost 3-fold higher with triple nucleoside reverse transcriptase inhibitors (NRTIs) containing ZDV/3TC/ABC compared to that seen in ZDV/3TC/NVP–treated individuals, and supports the recommendation that this combination not be used for first-line therapy in adults when alternative drugs are available. There was a high prevalence of extensive NRTI cross-resistance following viral failure at week 96, with almost half of patients in each treatment arm having ≥3 TAMs, consistent with other studies in resource-limited settings [10, 11]. Nonetheless, in vivo residual activity of approximately 1 log in viral load was observed in both NORA treatment groups overall [12]. The residual activity in the NVP group was lower than the triple NRTI group, consistent with NNRTI mutations conferring high-level resistance [13].

in resource-limited settings [10, 11]. Nonetheless, in vivo residual activity of approximately 1 log in viral load was observed in both NORA treatment groups overall [12]. The residual activity in the NVP group was lower than the triple NRTI group, consistent with NNRTI mutations conferring high-level resistance [13]. We noted that most individuals treated with NVP who developed NNRTI resistance had a single mutation only, consistent with previous reports examining viral failure with both efavirenz- and NVP-containing regimens [14–17]. This questions the assumption that prolonged viral failure necessarily leads to accumulation of NNRTI mutations.

most individuals treated with NVP who developed NNRTI resistance had a single mutation only, consistent with previous reports examining viral failure with both efavirenz- and NVP-containing regimens [14–17]. This questions the assumption that prolonged viral failure necessarily leads to accumulation of NNRTI mutations. Most importantly, we found that one-quarter of individuals with viral failure (>1000 copies/mL) at week 48 experienced resuppression at 96 weeks even though real-time viral load testing was not undertaken. This was most likely due to an improvement in adherence. It is notable that resuppression occurred in the presence of major resistance mutation(s) at week 48 and no change in therapy, suggesting that strong antiviral activity is possible despite reduced viral susceptibility, although the role of adherence cannot be ignored. Drug substitutions due to poor tolerability/side effects did not account for the observed changes in viral load. In South Africa, where real-time viral monitoring has taken place, substantial rates of resuppression without modification of ART have also been reported, even in patients with NNRTI resistance [18]. Where VLM is introduced more widely, our data support WHO recommendations that suspected viral failure should be addressed by adherence counseling as well as repeat measurement before consideration of treatment switch. Such counseling might identify specific issues with the regimen and culminate in a treatment substitution to achieve a better fit for the patient and therefore better adherence.

at suspected viral failure should be addressed by adherence counseling as well as repeat measurement before consideration of treatment switch. Such counseling might identify specific issues with the regimen and culminate in a treatment substitution to achieve a better fit for the patient and therefore better adherence. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We thank all the participants and staff from all the centers participating in the NORA and DART trial.

Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author. Supplementary Data Notes Acknowledgments. We thank all the participants and staff from all the centers participating in the NORA and DART trial. MRC/UVRI Uganda Research Unit on AIDS, Entebbe: H. Grosskurth, P. Munderi, G. Kabuye, D. Nsibambi, R. Kasirye, E. Zalwango, M. Nakazibwe, B. Kikaire, G. Nassuna, R. Massa, K. Fadhiru, M. Namyalo, A. Zalwango, L. Generous, P. Khauka, N. Rutikarayo, W. Nakahima, A. Mugisha, J. Todd, J. Levin, S. Muyingo, A. Ruberantwari, P. Kaleebu, D. Yirrell, N. Ndembi, F. Lyagoba, P. Hughes, M. Aber, A. Medina Lara, S. Foster, J. Amurwon, B. Nyanzi Wakholi. Joint Clinical Research Centre, Kampala, Uganda: P. Mugyenyi, C. Kityo, F. Ssali, D. Tumukunde, T. Otim, J. Kabanda, H. Musana, J. Akao, H. Kyomugisha, A. Byamukama, J. Sabiiti, J. Komugyena, P. Wavamunno, S. Mukiibi, A. Drasiku, R. Byaruhanga, O. Labeja, P. Katundu, S. Tugume, P. Awio, A. Namazzi, G. T. Bakeinyaga, H. Katabira, D. Abaine, J. Tukamushaba, W. Anywar, W. Ojiambo, E. Angweng, S. Murungi, W. Haguma, S. Atwiine, J. Kigozi. University of Zimbabwe, Harare: A. Latif, J. Hakim, V. Robertson, A. Reid, E. Chidziva, R. Bulaya-Tembo, G. Musoro, F. Taziwa, C. Chimbetete, L. Chakonza, A. Mawora, C. Muvirimi, G. Tinago, P. Svovanapasis, M. Simango, O. Chirema, J. Machingura, S. Mutsai, M. Phiri, T. Bafana, M. Chirara, L. Muchabaiwa, M. Muzambi. Infectious Diseases Institute (formerly the Academic Alliance), Makerere University, Mulago, Uganda: E. Katabira, A. Ronald, A. Kambungu, F. Lutwama, A. Nanfuka, J. Walusimbi, E. Nabankema, R. Nalumenya, T. Namuli, R. Kulume, I. Namata, L. Nyachwo, A. Florence, A. Kusiima, E. Lubwama, R. Nairuba, F. Oketta, E. Buluma, R. Waita, H. Ojiambo, F. Sadik, J. Wanyama, P. Nabongo. The AIDS Support Organisation, Uganda: R. Ochai, D. Muhweezi. Imperial College, London, UK: C. Gilks, K. Boocock, C. Puddephatt, D. Winogron, J. Bohannon. MRC Clinical Trials Unit, London, UK: J. Darbyshire, D. M. Gibb, A. Burke, D. Bray, A. Babiker, A. S. Walker, H. Wilkes, M. Rauchenberger, S. Sheehan, L. Peto, K. Taylor, M. Spyer, A. Ferrier, B. Naidoo, D. Dunn, R. Goodall. Independent DART Trial Monitors: R. Nanfuka, C. Mufuka-Kapuya. DART Virology Group: P. Kaleebu (Co-Chair), D. Pillay (Co-Chair), P. Awio, M. Chirara, D. Dunn, D. M. Gibb, C. Gilks, R. Goodall, A. Kapaata, M. Katuramur, F. Lyagoba, R.

S. Sheehan, L. Peto, K. Taylor, M. Spyer, A. Ferrier, B. Naidoo, D. Dunn, R. Goodall. Independent DART Trial Monitors: R. Nanfuka, C. Mufuka-Kapuya. DART Virology Group: P. Kaleebu (Co-Chair), D. Pillay (Co-Chair), P. Awio, M. Chirara, D. Dunn, D. M. Gibb, C. Gilks, R. Goodall, A. Kapaata, M. Katuramur, F. Lyagoba, R. Magala, B. Magambo, K. Mataruka, A. McCormick, L. Mugarura, T. Musunga, M. Nabankkema, J. Nkalubo, P. Nkurunziza, C. Parry, V. Robertson, M. Spyer, D. Yirrell. DART Health Economics Group: A. Medina Lara (Chair), S. Foster, J. Amurwon, B. Nyanzi Wakholi, J. Kigozi, L. Muchabaiwa, M. Muzambi. Trial Steering Committee: I. Weller (Chair), A. Babiker (Trial Statistician), S. Bahendeka, M. Bassett, A. Chogo Wapakhabulo, J. Darbyshire, B. Gazzard, C. Gilks, H. Grosskurth, J. Hakim, A. Latif, C. Mapuchere, O. Mugurungi, P. Mugyenyi; Observers: C. Burke, S. Jones, C. Newland, S. Rahim, J. Rooney, M. Smith, W. Snowden, J.-M. Steens. Data and Safety Monitoring Committee: A. Breckenridge (Chair), A. McLaren (Chair—deceased), C. Hill, J. Matenga, A. Pozniak, D. Serwadda. Endpoint Review Committee: T. Peto (Chair), A. Palfreeman, M. Borok, E. Katabira.

Observers: C. Burke, S. Jones, C. Newland, S. Rahim, J. Rooney, M. Smith, W. Snowden, J.-M. Steens. Data and Safety Monitoring Committee: A. Breckenridge (Chair), A. McLaren (Chair—deceased), C. Hill, J. Matenga, A. Pozniak, D. Serwadda. Endpoint Review Committee: T. Peto (Chair), A. Palfreeman, M. Borok, E. Katabira. Author contributions. The NORA substudy was conducted by P. M. and C. K., and C. F. G. was part of the UK coordinating team. C. F. G., D. P., P. K., and R. L. G. were involved in the design and coordination of the virology substudy. L. M. carried out HIV RNA assays, and F. L. conducted the genotyping. R. L. G. conducted the analyses with M. R. All authors contributed to interpretation of the data. R. K. G. and R. L. G. wrote the first draft of the paper. All authors revised the manuscript critically and approved the final version. R. L. G. had full access to all the data in the study and takes responsibility for the integrity of the data, the accuracy of the data analysis, and the decision to submit for publication. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author contributions. The NORA substudy was conducted by P. M. and C. K., and C. F. G. was part of the UK coordinating team. C. F. G., D. P., P. K., and R. L. G. were involved in the design and coordination of the virology substudy. L. M. carried out HIV RNA assays, and F. L. conducted the genotyping. R. L. G. conducted the analyses with M. R. All authors contributed to interpretation of the data. R. K. G. and R. L. G. wrote the first draft of the paper. All authors revised the manuscript critically and approved the final version. R. L. G. had full access to all the data in the study and takes responsibility for the integrity of the data, the accuracy of the data analysis, and the decision to submit for publication. Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Financial support. DART was funded by the UK Medical Research Council, the UK Department for International Development, and the Rockefeller Foundation. First-line drugs for NORA were provided by GlaxoSmithKline and Boehringer Ingelheim. Additional support for viral load and resistance assays in NORA was provided by GlaxoSmithKline. This work was partly supported by the European Community's Seventh Framework Programme (FP7/2007–2013) under the project “Collaborative HIV and Anti-HIV Drug Resistance Network (CHAIN)” (grant agreement number 223 131). R. K. G. is funded by a Wellcome Trust Fellowship (WT093722MA). Potential conflicts of interest. All authors: No reported conflicts.

Financial support. DART was funded by the UK Medical Research Council, the UK Department for International Development, and the Rockefeller Foundation. First-line drugs for NORA were provided by GlaxoSmithKline and Boehringer Ingelheim. Additional support for viral load and resistance assays in NORA was provided by GlaxoSmithKline. This work was partly supported by the European Community's Seventh Framework Programme (FP7/2007–2013) under the project “Collaborative HIV and Anti-HIV Drug Resistance Network (CHAIN)” (grant agreement number 223 131). R. K. G. is funded by a Wellcome Trust Fellowship (WT093722MA). Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Cryptococcal meningitis (CM) is a leading cause of death in human immunodeficiency virus (HIV)–infected individuals in low-resource settings [1]. The causative organism, Cryptococcus neoformans, is a facultative intracellular pathogen that has developed numerous strategies allowing it to survive and replicate inside macrophages [2, 3]. Environmental exposure to Cryptococcus is universal [4]. In the context of impaired adaptive immune responses, the ability of Cryptococcus to evade macrophage killing leads to dissemination, disease, and ultimately death [5]. Although the primary immune defect leading to development of cryptococcal meningitis is impairment of CD4+ T-cell (CD4) responses, usually secondary to HIV infection [6], the effectiveness of macrophage recognition, processing, and killing of Cryptococcus is likely to play an important role in the evolution of infection [2, 3, 7]. Vitamin D is required for effective macrophage responses to a number of intracellular pathogens including Mycobacterium tuberculosis complex (MTB), where it plays a critical role in macrophage activation following Toll-like receptor (TLR) signaling, tumor necrosis factor alpha (TNF-α) release, interferon gamma (IFN-γ)–mediated cathelicidin function, phagolysosome maturation and autophagy, and intracellular killing of mycobacteria [8–13]. Macrophages from HIV-infected patients have particularly impaired antituberculous activity in the absence of adequate vitamin D levels [8, 14], consistent with the markedly increased susceptibility to tuberculosis during HIV infection [15].

ome maturation and autophagy, and intracellular killing of mycobacteria [8–13]. Macrophages from HIV-infected patients have particularly impaired antituberculous activity in the absence of adequate vitamin D levels [8, 14], consistent with the markedly increased susceptibility to tuberculosis during HIV infection [15]. Similar to tuberculosis, CM is caused by an inhaled pathogen that evades effective intracellular killing by alveolar macrophages, often establishes a latent infection in the lung, and disseminates and causes disease when effective T-cell–mediated immune responses are depleted in HIV infection [5]. Data show that HIV-infected patients who have had pulmonary tuberculosis are at increased risk of developing CM [16], raising the possibility of a shared immune defect over and above CD4+ T-cell depletion. We hypothesized that vitamin D deficiency may impair immune responses to Cryptococcus, leading to similar increases in susceptibility to disease and impairments of microbiological clearance to those seen in MTB infection.

g CM [16], raising the possibility of a shared immune defect over and above CD4+ T-cell depletion. We hypothesized that vitamin D deficiency may impair immune responses to Cryptococcus, leading to similar increases in susceptibility to disease and impairments of microbiological clearance to those seen in MTB infection. To test this hypothesis, we performed a study in Cape Town, South Africa, consisting of 3 parts: (1) 25-hydroxyvitamin D (25[OH]D) levels were measured in patients presenting with CM and control patients with comparable CD4 counts drawn from the same population who did not have CM to determine whether vitamin D deficiency was associated with the development of CM; (2) 25(OH)D levels in the study population were analyzed for evidence of seasonality corresponding to sunshine hours, and Western Cape CM notifications from the South African National Institute for Communicable Diseases (NICD) covering the study period were analyzed for evidence of reciprocal seasonality; and (3) associations between 25(OH)D levels and disease severity, immune responses, and microbiological clearance rates were examined in patients with CM.

ifications from the South African National Institute for Communicable Diseases (NICD) covering the study period were analyzed for evidence of reciprocal seasonality; and (3) associations between 25(OH)D levels and disease severity, immune responses, and microbiological clearance rates were examined in patients with CM. METHODS Participants and Procedures Participants were recruited at GF Jooste Hospital, Cape Town, South Africa, between July 2005 and May 2010. One hundred fifty participants were HIV-infected adults (aged ≥21 years) with a first episode of CM (cases), diagnosed by cerebrospinal fluid (CSF) India ink or cryptococcal antigen testing (titers ≥1:1024; Meridian Cryptococcal Latex Agglutination System, Meridian Bioscience, Cincinnati, Ohio), who were enrolled sequentially in 2 clinical trials examining different amphotericin B–based induction regimens [17, 18]. The studies were approved by the Research Ethics Committee of the University of Cape Town, and patients gave informed consent for blood and CSF samples to be used for research purposes. The component trials had the same inclusion and exclusion criteria, and have been described elsewhere [17, 18]. On study enrollment, history and clinical examination findings were recorded. Blood samples taken prior to antifungal therapy were used for plasma vitamin D quantification. Lumbar punctures (LPs) with quantitative CSF cultures were performed on days 1, 3, 7, and 14. Cryptococcal clearance (early fungicidal activity [EFA]) was calculated as the rate of decrease in log colony-forming units (CFU) per milliliter of CSF per day derived from the slope of the linear regression of log CFU per milliliter against time for each patient [19]. The CSF cell count and protein and glucose levels were determined. CSF interferon gamma (IFN-γ), tumor necrosis factor alpha (TNF-α), and interleukin 6 (IL-6) concentrations were measured in all patients using the Luminex multianalyte platform (Luminex) and Bio-Rad cytokine kits (Bio-Rad) [20]. CSF soluble CD14 (sCD14) and neopterin concentrations were measured for a subset of 90 sequential patients using Bio-Rad kits and manual enzyme-linked immunosorbent assay (ELItest Neopterin, BRAHMS Aktiengesellschaft, Hennigsdorf, Germany), respectively. Baseline CD4 cell counts were recorded for all patients. Patients were followed for 1 year and mortality outcomes recorded.

measured for a subset of 90 sequential patients using Bio-Rad kits and manual enzyme-linked immunosorbent assay (ELItest Neopterin, BRAHMS Aktiengesellschaft, Hennigsdorf, Germany), respectively. Baseline CD4 cell counts were recorded for all patients. Patients were followed for 1 year and mortality outcomes recorded. Recruited concurrently were 150 hospital-based control patients, who were HIV-infected adults (aged ≥21 years) with a nadir CD4 count ≤100 cells/µL and no current evidence of or prior history of cryptococcal disease, attending the hospital for management of either newly diagnosed HIV infection or an opportunistic infection other than CM. These patients were drawn from the same population as the cases during the same “risk period,” and would have been included as a case in the study had they developed CM. Basic demographic data, medical history, and current CD4 count were recorded, and a blood sample was taken for plasma vitamin D quantification. Among cases and controls, all patients currently taking antituberculosis medication with a clinical diagnosis of tuberculosis (both sputum acid-fast bacillus smear positive and smear negative) were defined as having active tuberculosis. Written informed consent was obtained from each control participant, and the study was approved by the Research Ethics Committee of the University of Cape Town.

cation with a clinical diagnosis of tuberculosis (both sputum acid-fast bacillus smear positive and smear negative) were defined as having active tuberculosis. Written informed consent was obtained from each control participant, and the study was approved by the Research Ethics Committee of the University of Cape Town. Vitamin D Levels Plasma 25(OH)D concentrations were measured in stored baseline blood samples at St George's University of London using Immunodiagnostics Systems’ 25(OH)D kit (REF IS2700) on the iSYS multidiscipline autoanalyzer. Vitamin D status was defined according to standard criteria as normal (>75 nmol/L), insufficient (≤75 nmol/L), deficient (≤50 nmol/L), or severely deficient (≤25 nmol/L) [13, 21]. Cryptococcal Meningitis Notifications All incident laboratory-confirmed cases of cryptococcal disease from the Western Cape were reported to the NICD during the study period with date of specimen collection; surveillance audits were conducted to ensure complete reporting. A case of incident cryptococcosis was defined as the first episode of laboratory-confirmed disease in a patient (encapsulated yeasts observed by microscopic examination of an India ink–stained fluid, or a positive cryptococcal antigen test or culture of Cryptococcus species from any body site) diagnosed at a clinical laboratory in the Western Cape Province.

efined as the first episode of laboratory-confirmed disease in a patient (encapsulated yeasts observed by microscopic examination of an India ink–stained fluid, or a positive cryptococcal antigen test or culture of Cryptococcus species from any body site) diagnosed at a clinical laboratory in the Western Cape Province. Statistical Analysis Data were analyzed using Stata version 12.0 (StataCorp, College Station, Texas), R version 3.0.2 (R foundation for Statistical Computing), and GraphPad Prism version 6 (Graphpad Software Inc, San Diego, California). Variables were compared across groups using unpaired t tests, 1-way analysis of variance, Kruskal-Wallis, χ2, or Fisher exact tests as appropriate. The 25(OH)D results were log transformed, geometric means and 95% confidence intervals (CIs) presented, and log-transformed results used in regression analyses. For the case-control analysis, crude and adjusted odds ratios (ORs) exploring the association between vitamin D deficiency and CM, and potential confounders in this relationship, were obtained using logistic regression analysis. Evidence for seasonality in 25(OH) D levels and cryptococcal case notifications was examined using Poisson regression models, which modeled monthly data using a general trend plus a sinusoidal wave for seasonal effect (cosinor regression modeling [22]). Assessment of seasonality was made by comparing the Akaike information criterion of models including or jointly omitting the sine and cosine terms using a likelihood ratio test. Among the CM cases, associations between 25(OH)D levels and disease severity at presentation, baseline CSF immune responses, rate of clearance of infection, and mortality were examined using linear and Cox regression modeling. Statistical significance was defined as P ≤ .05.

cosine terms using a likelihood ratio test. Among the CM cases, associations between 25(OH)D levels and disease severity at presentation, baseline CSF immune responses, rate of clearance of infection, and mortality were examined using linear and Cox regression modeling. Statistical significance was defined as P ≤ .05. RESULTS Demographic and clinical characteristics of patients are summarized in Table 1. Patients with CM had a median CD4 count of 32 cells/µL, and severe disease at presentation, with high CSF fungal burdens (median, 5.3 [interquartile range, 4.3–5.8 log10 CFU/mL]) and a high proportion of altered mental status (13%). All CM patients were antiretroviral therapy (ART) naive. The control patients were similar to cases in terms of age and CD4 count, although a larger proportion was female, and 30% were already taking ART. Sixty-three (42%) control patients had a current diagnosis of tuberculosis, compared with 53 (35%) of the CM cases (P = .24). Thirty-four (23%) controls had advanced HIV infection alone, and the remaining 53 (35%) were being investigated or treated for opportunistic infections or complications of HIV infection (including 13 with gastroenteritis, 10 with pneumonia or bacterial sepsis, 6 with anemia, and 24 with other conditions including Pneumocystis pneumonia, Kaposi sarcoma, and candidiasis). All patients were black Africans. Table 1. Patient Characteristics and 25-Hydroxyvitamin D Levels

s or complications of HIV infection (including 13 with gastroenteritis, 10 with pneumonia or bacterial sepsis, 6 with anemia, and 24 with other conditions including Pneumocystis pneumonia, Kaposi sarcoma, and candidiasis). All patients were black Africans. Table 1. Patient Characteristics and 25-Hydroxyvitamin D Levels Characteristic CM Cases (n = 150) Controls (n = 150) Adjusted ORa P Value Age, y 32 (28–38) 32 (27–37) .337 Male sex, % (No.) 41% (62) 17% (26) <.001 CD4 count, cells/µL 32 (13–58) 40 (19–79) .13 Active tuberculosis, % (No.) 35% (53) 42% (63) .236 On ART, % (No.) 0% (0) 30% (45) <.001 Duration of ART, d … 55 (21–99) … Vitamin D, nmol/Lb 38 (35–41) 36 (33–39) .367 Vitamin D ≤75 nmol/L, % (No.) 93% (139) 95% (142) .338 Vitamin D ≤50 nmol/L, % (No.) 75% (112) 72% (108) .669 Vitamin D ≤25 nmol/L, % (No.) 18% (27) 26% (38) .116 Fungal burden, log1 0 CFU/mL 5.3 (4.3–5.8) … … Altered mental status, % (No.) 13% (19) … … EFA, log10 CFU/mL/d −0.52 (−0.39 to −0.69) … … Mortalityc, % (No.) 28% (41) … … Vitamin D >50 nmol/L, % (No.) 25% (38) 28% (41) 1 (base) .796 Vitamin D ≤50 nmol/L, % (No.) 75% (112) 72% (108) 0.93 (95% CI, .54–1.61) Data presented are median (interquartile range) or percentage (No.). Significance testing was performed using Kruskal-Wallis, χ2, or Student t test as appropriate. Abbreviations: ART, antiretroviral therapy; CFU, colony-forming units; CI, confidence interval; CM, cryptococcal meningitis; EFA, early fungicidal activity; OR, odds ratio; vitamin D, 25-hydroxyvitamin D.

Characteristic CM Cases (n = 150) Controls (n = 150) Adjusted ORa P Value Age, y 32 (28–38) 32 (27–37) .337 Male sex, % (No.) 41% (62) 17% (26) <.001 CD4 count, cells/µL 32 (13–58) 40 (19–79) .13 Active tuberculosis, % (No.) 35% (53) 42% (63) .236 On ART, % (No.) 0% (0) 30% (45) <.001 Duration of ART, d … 55 (21–99) … Vitamin D, nmol/Lb 38 (35–41) 36 (33–39) .367 Vitamin D ≤75 nmol/L, % (No.) 93% (139) 95% (142) .338 Vitamin D ≤50 nmol/L, % (No.) 75% (112) 72% (108) .669 Vitamin D ≤25 nmol/L, % (No.) 18% (27) 26% (38) .116 Fungal burden, log1 0 CFU/mL 5.3 (4.3–5.8) … … Altered mental status, % (No.) 13% (19) … … EFA, log10 CFU/mL/d −0.52 (−0.39 to −0.69) … … Mortalityc, % (No.) 28% (41) … … Vitamin D >50 nmol/L, % (No.) 25% (38) 28% (41) 1 (base) .796 Vitamin D ≤50 nmol/L, % (No.) 75% (112) 72% (108) 0.93 (95% CI, .54–1.61) Data presented are median (interquartile range) or percentage (No.). Significance testing was performed using Kruskal-Wallis, χ2, or Student t test as appropriate. Abbreviations: ART, antiretroviral therapy; CFU, colony-forming units; CI, confidence interval; CM, cryptococcal meningitis; EFA, early fungicidal activity; OR, odds ratio; vitamin D, 25-hydroxyvitamin D. a Variables that were associated with both case status and vitamin D deficiency with a P value ≤0.1 were considered to be potential confounders in the relationship between vitamin D deficiency and development of CM. The only variable meeting these criteria was season, which was adjusted for in the analysis reported here. Levels of 25-hydroxyvitamin D varied by season, with the highest levels in the first quarter of the year (mean, 48 nmol/L [95% CI, 43–52 nmol/L]), lower levels in the second quarter (mean, 33 nmol/L [95% CI, 29–38 nmol/L]), the lowest levels in the third quarter of the year (mean, 32 nmol/L [95% CI, 28–35 nmol/L]), and increasing levels in the fourth quarter (mean, 38 nmol/L [95% CI, 34–42 nmol/L]), analysis of variance P = .005. Further adjustment for sex, CD4 count, and ART status did not alter the findings (adjusted OR, 0.82 [95% CI, .44–1.51]; P = .523).

the third quarter of the year (mean, 32 nmol/L [95% CI, 28–35 nmol/L]), and increasing levels in the fourth quarter (mean, 38 nmol/L [95% CI, 34–42 nmol/L]), analysis of variance P = .005. Further adjustment for sex, CD4 count, and ART status did not alter the findings (adjusted OR, 0.82 [95% CI, .44–1.51]; P = .523). b Log-normal distribution; geometric mean and 95% CIs are presented. c Mortality at 10 weeks.

the third quarter of the year (mean, 32 nmol/L [95% CI, 28–35 nmol/L]), and increasing levels in the fourth quarter (mean, 38 nmol/L [95% CI, 34–42 nmol/L]), analysis of variance P = .005. Further adjustment for sex, CD4 count, and ART status did not alter the findings (adjusted OR, 0.82 [95% CI, .44–1.51]; P = .523). b Log-normal distribution; geometric mean and 95% CIs are presented. c Mortality at 10 weeks. Vitamin D Deficiency Is Common, and Clear Seasonal Variations Are Observed The mean 25(OH)D concentration of the total study population (cases and controls combined) was 38 nmol/L (Figure 1). Only 18 (6%) had adequate 25(OH)D levels (>75 nmol/L). Two hundred twenty (74%) had vitamin D deficiency (≤50 nmol/L), and 65 (22%) were severely vitamin D deficient (≤25 nmol/L). Levels of 25(OH)D varied by season, with the highest levels in the first quarter of the year (mean, 48 nmol/L [95% CI, 43–52 nmol/L]), corresponding to the southern hemisphere summer and highest number of sunshine hours, and the lowest levels in the third quarter of the year (mean, 32 nmol/L [95% CI, 28–35 nmol/L]), corresponding to the winter months and lowest number of sunshine hours. Cosinor regression modeling confirmed the presence of significant seasonality in vitamin D levels (P < .001). The 25(OH)D levels did not differ by sex and were not associated with age, CD4 count, or ART status. Figure 1. Plasma 25-hydroxyvitamin D (25[OH]D) levels by cryptococcal meningitis status, tuberculosis status, and season. A, Plasma 25(OH)D levels of the whole study population (cases and controls combined), with dashed lines at 75 nmol/L (vitamin D insufficiency), 50 nmol/L (vitamin D deficiency), and 25 nmol/L (severe vitamin D deficiency). B and C, Plasma 25(OH)D levels according to cryptococcal meningitis case status (B) and tuberculosis status (C), with lines at the geometric mean and error bars showing 95% confidence intervals. Levels of 25(OH)D were significantly lower in individuals with tuberculosis than in those without tuberculosis (*34 nmol/L vs 39 nmol/L; P = .029). D, Average number of sunshine hours per month in Cape Town (source: National Oceanic and Atmospheric Administration, available at: www.noaa.gov). E, Levels of 25(OH)D by month (averaged over the 5-year study period) with cosinor regression line. F, Monthly cryptococcal notification rates (averaged over the period 2005–2011) with best-fit regression line. Abbreviations: CM, cryptococcal meningitis; TB, tuberculosis.

ic Administration, available at: www.noaa.gov). E, Levels of 25(OH)D by month (averaged over the 5-year study period) with cosinor regression line. F, Monthly cryptococcal notification rates (averaged over the period 2005–2011) with best-fit regression line. Abbreviations: CM, cryptococcal meningitis; TB, tuberculosis. No Seasonal Trends Are Evident in Cryptococcal Meningitis Notification Rates in the Western Cape Region To examine associations between vitamin D status and the risk of developing CM, the Western Cape region CM notification rates for the 7-year period 2005–2011 were analyzed for seasonal trends. Despite the seasonal variation in 25(OH)D levels seen in this patient population, cosinor regression modeling did not demonstrate any seasonal trend in CM notification rates (P > .7), with an average of 39 cases per month during the period and very little monthly variation (Figure 1).

were analyzed for seasonal trends. Despite the seasonal variation in 25(OH)D levels seen in this patient population, cosinor regression modeling did not demonstrate any seasonal trend in CM notification rates (P > .7), with an average of 39 cases per month during the period and very little monthly variation (Figure 1). Vitamin D Deficiency Is Not Associated With Cryptococcal Meningitis, but Is Associated With Active Tuberculosis Levels of 25(OH)D levels did not differ between CM cases and control patients (mean, 38 nmol/L vs 36 nmol/L; P = .367; Table 1). Vitamin D deficiency was not associated with CM (OR, 1.12 [95% CI, .7–1.9]; P = .669), and this remained the case in a multivariable logistic regression model adjusted for season (adjusted OR [aOR], 0.93 [95% CI, .6–1.6]; P = .796). A sensitivity analysis restricted to ART-naive patients yielded the same findings (aOR, 0.82 [95% CI, .44–1.51]; P = .523), as did the equivalent analysis looking at severe vitamin D deficiency (aOR, 0.62 [95% CI, .32–1.31]; P = .223). Conversely, 25(OH)D levels were lower in patients with active tuberculosis compared with those without (34 nmol/L vs 39 nmol/L; P = .029), and this difference remained significant after adjusting for CM case status and CD4 count (P = .04). In both CM cases and controls, vitamin D deficiency was associated with increased odds of active tuberculosis, with some evidence for an increasing trend with worsening deficiency (OR, 1.47 [95% CI, .5–4.7] for vitamin D insufficiency; OR, 1.51 [95% CI, .5–4.5] for vitamin D deficiency; and OR, 2.52 [95% CI, .8–7.9] for severe vitamin D deficiency, all compared to a baseline of normal vitamin D status; P for trend = .069).

sis, with some evidence for an increasing trend with worsening deficiency (OR, 1.47 [95% CI, .5–4.7] for vitamin D insufficiency; OR, 1.51 [95% CI, .5–4.5] for vitamin D deficiency; and OR, 2.52 [95% CI, .8–7.9] for severe vitamin D deficiency, all compared to a baseline of normal vitamin D status; P for trend = .069). Vitamin D Status Is Not Associated With Disease Severity, Host Immune Response, or Microbiological Clearance in Patients With HIV-Associated Cryptococcal Meningitis Among the 150 CM cases studied, there were no associations between 25(OH)D level and either fungal burden at disease presentation, the host immune response at the site of infection, or the rate of clearance of infection (Figure 2 and Table 2). Mean fungal burden was very similar in those with and without vitamin D deficiency (5.1 log10 CFU/mL vs 5.0 log10 CFU/mL; P = .687), as were CSF lymphocyte counts (15 × 106/L vs 19 × 106/L; P = .897), CSF TNF-α levels (0.84 log10 pg/mL vs 0.81 log10 pg/mL; P = .697), CSF IL-6 levels (2.44 log10 pg/mL vs 2.28 log10 pg/mL; P = .540), and CSF IFN-γ levels (1.62 log10 pg/mL vs 1.61 log10 pg/mL; P = .988). Regression modeling confirmed the absence of significant associations between 25(OH)D levels and fungal burden, CSF lymphocytes, CSF TNF-α levels, CSF IL-6 levels, and CSF IFN-γ levels (Table 2). Given evidence that in the context of tuberculosis infection the activation of macrophages by IFN-γ is vitamin D dependent [11], we examined the ratio of IFN-γ to the macrophage activation markers sCD14 and neopterin. The IFN-γ:sCD14 ratios (0.26 vs 0.25; P = .788) and IFN-γ:neopterin ratios (0.82 vs 0.83; P = .914) were similar in patients with and without vitamin D deficiency, providing no evidence for differential macrophage activation in CM patients according to vitamin D status. Table 2. Associations Between Vitamin D Status and Fungal Burden, Immune Responses, and Rate of Clearance of Infection in Patients With Cryptococcal Meningitis

ar in patients with and without vitamin D deficiency, providing no evidence for differential macrophage activation in CM patients according to vitamin D status. Table 2. Associations Between Vitamin D Status and Fungal Burden, Immune Responses, and Rate of Clearance of Infection in Patients With Cryptococcal Meningitis Variable Vitamin D>50 nmol/L Vitamin D≤50 nmol/L P Value β Coefficienta P Value Baseline fungal burden, log10 CFU/mL 5.0 (4.6–5.3) 5.1 (4.8–5.3) .687 0.07 (−.32 to .47) .702 CSF lymphocytes, ×106/Lb 19 (1–67) 15 (1–88) .896 −39 (−93 to 14) .148 CSF TNF-α, log10 pg/mL 0.81 (.70–.92) 0.84 (.76–0.92) .697 −0.09 (−.21 to .03) .148 CSF IFN-γ, log10 pg/mL 1.61 (1.41–1.81) 1.62 (1.49–1.74) .988 −0.09 (−.30 to .11) .374 CSF IL-6, log10 pg/mL 2.28 (1.84–2.72) 2.43 (2.19–2.69) .540 −0.26 (−.68 to .17) .231 CSF sCD14, log10 pg/mL 6.02 (5.91–6.11) 6.02 (5.97–6.09) .834 0.03 (−.07 to .12) .596 CSF neopterin, log10 pg/mL 1.85 (1.75–1.95) 1.90 (1.82–1.95) .522 −0.05 (−.16 to .06) .366 CSF IFN-γ:sCD14 ratio 0.25 (.21–.29) 0.26 (.23–.28) .788 −0.02 (−.05 to .02) .309 CSF IFN-γ:neopterin ratio 0.82 (.68–.97) 0.82 (.74–.90) .914 −0.04 (−.17 to .08) .497 Early fungicidal activity, log10 CFU/mL/d −0.56 (−0.46 to −0.66) −0.56 (−.51 to −0.60) .847 −0.02 (−.09 to .06) .701 Data are presented as means and 95% confidence intervals for the vitamin D–deficient and vitamin D–nondeficient groups. Abbreviations: CFU, colony-forming units; CSF, cerebrospinal fluid; IFN, interferon; IL, interleukin; sCD14, soluble CD14; TNF, tumor necrosis factor; vitamin D, 25-hydroxyvitamin D.

Variable Vitamin D>50 nmol/L Vitamin D≤50 nmol/L P Value β Coefficienta P Value Baseline fungal burden, log10 CFU/mL 5.0 (4.6–5.3) 5.1 (4.8–5.3) .687 0.07 (−.32 to .47) .702 CSF lymphocytes, ×106/Lb 19 (1–67) 15 (1–88) .896 −39 (−93 to 14) .148 CSF TNF-α, log10 pg/mL 0.81 (.70–.92) 0.84 (.76–0.92) .697 −0.09 (−.21 to .03) .148 CSF IFN-γ, log10 pg/mL 1.61 (1.41–1.81) 1.62 (1.49–1.74) .988 −0.09 (−.30 to .11) .374 CSF IL-6, log10 pg/mL 2.28 (1.84–2.72) 2.43 (2.19–2.69) .540 −0.26 (−.68 to .17) .231 CSF sCD14, log10 pg/mL 6.02 (5.91–6.11) 6.02 (5.97–6.09) .834 0.03 (−.07 to .12) .596 CSF neopterin, log10 pg/mL 1.85 (1.75–1.95) 1.90 (1.82–1.95) .522 −0.05 (−.16 to .06) .366 CSF IFN-γ:sCD14 ratio 0.25 (.21–.29) 0.26 (.23–.28) .788 −0.02 (−.05 to .02) .309 CSF IFN-γ:neopterin ratio 0.82 (.68–.97) 0.82 (.74–.90) .914 −0.04 (−.17 to .08) .497 Early fungicidal activity, log10 CFU/mL/d −0.56 (−0.46 to −0.66) −0.56 (−.51 to −0.60) .847 −0.02 (−.09 to .06) .701 Data are presented as means and 95% confidence intervals for the vitamin D–deficient and vitamin D–nondeficient groups. Abbreviations: CFU, colony-forming units; CSF, cerebrospinal fluid; IFN, interferon; IL, interleukin; sCD14, soluble CD14; TNF, tumor necrosis factor; vitamin D, 25-hydroxyvitamin D. a The β coefficients are from linear regression analyses where the clinical and immunological parameters were considered individually as dependent variables, and 25-hydroxyvitamin D levels (log transformed) were considered as the explanatory variable. The coefficients shown thus represent the average increase in the dependent variable for each single unit increase (log10 nmol/L) in 25-hydroxyvitamin D concentration.

eters were considered individually as dependent variables, and 25-hydroxyvitamin D levels (log transformed) were considered as the explanatory variable. The coefficients shown thus represent the average increase in the dependent variable for each single unit increase (log10 nmol/L) in 25-hydroxyvitamin D concentration. b Heavily positively skewed; median values with interquartile ranges are shown. Figure 2. Fungal burden, cerebrospinal fluid (CSF) immune responses, and rate of clearance of infection in cryptococcal meningitis patients with and without vitamin D deficiency. The baseline CSF fungal burden (QCC), rate of clearance of infection (EFA), baseline CSF lymphocyte count, CSF TNF-α concentration, and CSF IFN-γ concentration are shown according to whether patients were vitamin D deficient (plasma 25-hydroxyvitamin D ≤50 nmol/L). Lines indicate the mean in the vitamin D–deficient patients and in those without vitamin D deficiency. No significant differences were present between the vitamin D–deficient and –sufficient groups in any of the variables shown. Abbreviations: CFU, colony-forming units; CSF, cerebrospinal fluid; EFA, early fungicidal activity; IFN, interferon; QCC, quantitative cryptococcal culture; TNF, tumor necrosis factor.

ficiency. No significant differences were present between the vitamin D–deficient and –sufficient groups in any of the variables shown. Abbreviations: CFU, colony-forming units; CSF, cerebrospinal fluid; EFA, early fungicidal activity; IFN, interferon; QCC, quantitative cryptococcal culture; TNF, tumor necrosis factor. In keeping with the absence of any observed impact of vitamin D status on the immune response to cryptococcal disease, rates of clearance of Cryptococcus from the CSF were not associated with 25(OH)D levels (β coefficient −0.015 [95% CI, −.09–.06]; P = .701). The mean rate of clearance was −0.56 in those with vitamin D deficiency vs −0.56 in those without (P = .847). Mortality at 10 weeks was 30% (n = 33) in patients with vitamin D deficiency vs 22% (n = 8) in those without (P = .367). After adjustment for CD4 count and the other key predictors of mortality, baseline fungal burden and abnormal mental status [23], the hazard of death was 1.35 (95% CI, .7–2.6; P = .375) in vitamin D–deficient patients compared with those non–vitamin D–deficient patients.

ency vs 22% (n = 8) in those without (P = .367). After adjustment for CD4 count and the other key predictors of mortality, baseline fungal burden and abnormal mental status [23], the hazard of death was 1.35 (95% CI, .7–2.6; P = .375) in vitamin D–deficient patients compared with those non–vitamin D–deficient patients. DISCUSSION Vitamin D deficiency was prevalent in this population of HIV-infected black African patients in Cape Town, consistent with previous findings in HIV-infected and uninfected populations in this setting, and, in keeping with previous reports, showed a marked seasonal variation closely related to sunshine exposure [14]. Also consistent with recent studies from Cape Town [14] was the observed association of vitamin D deficiency with active tuberculosis. We did not find any evidence for an association between vitamin D status and either susceptibility to CM or the immune response to CM and microbiological clearance in patients who had developed CM. Levels of 25(OH)D levels did not differ between the cohort of patients with CM and the control patients with comparable CD4 counts but no history of cryptococcal disease. This remained the case in sensitivity analysis adjusting for ART status, the only important factor differing between the CM cases and controls. Further evidence for an absence of association between vitamin D status and susceptibility to CM was the lack of seasonal trend in CM notifications, despite the clear seasonal variation in 25(OH)D levels in this population [14]. Consistent with these observations were our findings that vitamin D deficiency was not associated with fungal burden at CM presentation, did not influence the CSF immune response, and had no bearing on the rate at which infection was cleared from the CSF. As in prior studies [14], mean 25(OH)D levels in the studied population were low. Nevertheless, variation within a range of relatively low levels was associated with important differences in susceptibility to tuberculosis in this and other studies [14], arguing against the possibility that the lack of association seen with cryptococcal disease was due to low population vitamin D status.

on were low. Nevertheless, variation within a range of relatively low levels was associated with important differences in susceptibility to tuberculosis in this and other studies [14], arguing against the possibility that the lack of association seen with cryptococcal disease was due to low population vitamin D status. Very few prior studies have examined vitamin D in the context of other fungal infections, and the reported results do not show a consistent association with vitamin D status, which may be related to the diverse host defense mechanisms involved [24, 25]. Our findings suggest that immune control and clearance of Cryptococcus is not via vitamin D–dependent pathways. Given the immunomodulatory effects of vitamin D on both innate and adaptive immunity [8–13, 26], plus reports demonstrating impaired immune responses and increased susceptibility to HIV and HIV-associated opportunistic infections such as tuberculosis, respiratory tract infections, and candidiasis [8, 9, 14, 26–28], the lack of any observed association with CM is perhaps surprising. The bulk of the data concerning the role of vitamin D in immunity to infectious diseases come from studies of tuberculosis. Convincing evidence shows that vitamin D deficiency is a risk factor for the development of tuberculosis [14, 27, 28], and data from controlled trials suggest that vitamin D replacement may improve outcomes in patients with tuberculosis [29]. Macrophages from vitamin D–deficient HIV-infected patients demonstrate impaired intracellular signalling and TNF-α expression in response to TLR2/4 signaling by MTB [8], and these responses are restored by vitamin D supplementation in vitro. Activation of MTB-infected macrophages by T-cell–derived IFN-γ is dependent on vitamin D [11], and can be restored in macrophages from vitamin D–deficient patients by vitamin D supplementation. Importantly, for restriction of MTB growth in macrophages, vitamin D promotes phagolysosome fusion and maturation [9, 11], the generation of reactive oxygen and nitrogen species [30, 31], production of antimicrobial cathelicidins [9, 11, 32], and induction of autophagy [9, 11]. These mechanisms overcome the immune evasion mechanisms employed by MTB of blocking phagosome maturation, and inhibiting phagosome-lysosome fusion [32–34].

[9, 11], the generation of reactive oxygen and nitrogen species [30, 31], production of antimicrobial cathelicidins [9, 11, 32], and induction of autophagy [9, 11]. These mechanisms overcome the immune evasion mechanisms employed by MTB of blocking phagosome maturation, and inhibiting phagosome-lysosome fusion [32–34]. In contrast to MTB, Cryptococcus does not need to prevent phagosome maturation or phagosome-lysosome fusion for intracellular survival, and is able to thrive in the acidic phagolysosome, protected by a thick polysaccharide capsule and virulence factors such as the ability to produce melanin using laccase, which protects against the oxidative burst [2, 3]. It is thus probable that vitamin D–dependent promotion of phagolysosome fusion and maturation has little effect on anticryptococcal immunity. Similarly, the promotion of cathelicidin production and autophagy, neither of which have a proven role in the innate response to cryptococcal infection [2], is unlikely to have significant anticryptococcal activity.

pendent promotion of phagolysosome fusion and maturation has little effect on anticryptococcal immunity. Similarly, the promotion of cathelicidin production and autophagy, neither of which have a proven role in the innate response to cryptococcal infection [2], is unlikely to have significant anticryptococcal activity. Activation of Cryptococcus-infected macrophages by T-cell–derived IFN-γ is likely to be critical for effective control of cryptococcal infection [35–37]. IFN-γ levels in the CSF are strongly associated with fungal burden and the rate of fungal clearance in patients with HIV-associated CM [20, 23], and exogenous IFN-γ has been shown to significantly increase the rate of clearance of cryptococci from the CSF [18]. Although we can only infer indirectly from our results, we found no evidence to suggest that IFN-γ–induced macrophage activation was vitamin D dependent, unlike in IFN-γ–induced activation of MTB-infected macrophages [11]. Levels of the macrophage activation markers sCD14 and neopterin, and the IFN-γ:sCD14 and IFN-γ:neopterin ratios did not differ according to vitamin D status.

ound no evidence to suggest that IFN-γ–induced macrophage activation was vitamin D dependent, unlike in IFN-γ–induced activation of MTB-infected macrophages [11]. Levels of the macrophage activation markers sCD14 and neopterin, and the IFN-γ:sCD14 and IFN-γ:neopterin ratios did not differ according to vitamin D status. Interestingly, there are limited data to suggest that the protective effects of vitamin D in the host response to MTB are due to anti-inflammatory properties, with inhibition of Th1-type immune responses [38, 39], faster resolution of inflammation [10], and limitation of the tissue damage associated with active MTB infection [26, 40]. Again in contrast to tuberculosis, tissue damage resulting from excessive inflammation is not a prominent feature of HIV-associated CM [41]. Rather, a lack of Th1-type inflammatory responses and high organism burdens are associated with poor outcomes in HIV-associated CM [18, 23, 37, 42], underlining the differing immune responses required for effective control of the opportunistic intracellular pathogens MTB and Cryptococcus. In summary, we found no evidence that vitamin D deficiency predisposes to the development of CM, or leads to impaired immune responses or microbiological clearance in HIV-infected patients with CM. These data suggest that, in contrast to tuberculosis, vitamin D–dependent pathways are not of key importance in the host immune response to cryptococcal infection.

amin D deficiency predisposes to the development of CM, or leads to impaired immune responses or microbiological clearance in HIV-infected patients with CM. These data suggest that, in contrast to tuberculosis, vitamin D–dependent pathways are not of key importance in the host immune response to cryptococcal infection. Notes Acknowledgments. The authors thank G. Ntombomzi Williams and Nomqondiso Sidibana for assistance with patient recruitment and for providing clinical care to the patients in Cape Town. We acknowledge the support of the clinical and administrative staff of the Department of Health (Provincial Government of the Western Cape), and the GERMS-SA surveillance network for reporting cases of cryptococcal disease to the NICD. Financial support. This work was supported by the Wellcome Trust (training fellowship to J. N. J., WT081794 and G.M., WT098316) and the British Infection Society (fellowship to T. B.). Potential conflicts of interest. N. P. G. has received grants from Pfizer South Africa and personal fees from Pfizer South Africa, MSD South Africa, and Fujifilm Pharmaceuticals. J. R. P. has received research grants and advisory board/consulting fees from Merck, Pfizer, Astellas, F2G, Viamet, and Scynexis. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

(See the Editorial Commentary by Pappas on pages 1615–17.) Despite recognition of the burden and advancements in treatment, acute mortality from human immunodeficiency virus (HIV)-associated cryptococcal meningitis remains high with 17%–50% mortality within 2 weeks of diagnosis among individuals in sub-Saharan Africa [1–9]. One complication of cryptococcal meningitis is elevated intracranial pressure (ICP), defined as a cerebral spinal fluid (CSF) opening pressure >250 mmH2O, and prior literature suggests there is higher mortality among cryptococcal patients with raised ICP [10–12].

diagnosis among individuals in sub-Saharan Africa [1–9]. One complication of cryptococcal meningitis is elevated intracranial pressure (ICP), defined as a cerebral spinal fluid (CSF) opening pressure >250 mmH2O, and prior literature suggests there is higher mortality among cryptococcal patients with raised ICP [10–12]. Raised ICP is common at the time of diagnosis and frequently leads to changes in mental status, headache, loss of vision and hearing, or death. Aggressive management of ICP is therefore suggested in treatment guidelines for cryptococcal meningitis, including daily therapeutic lumbar punctures (LPs) until pressures and symptoms have normalized [13, 14]. With these recommendations, elevated ICP typically resolves over the first 2 weeks of antifungal therapy. Prior studies have not found an association between baseline opening pressure and 2-week mortality attributed, in part, to aggressive control of ICP with therapeutic LPs [4, 15, 16]. A recent comparison to historical data also suggested that following a strict schedule of therapeutic LPs may have led to lower 30-day mortality in a hospital in Tanzania [17]. We aimed to add to the current body of literature and estimate the direct effect of therapeutic LPs on acute mortality in a prospective cohort of HIV-infected individuals with cryptococcal meningitis in Uganda and South Africa.

le of therapeutic LPs may have led to lower 30-day mortality in a hospital in Tanzania [17]. We aimed to add to the current body of literature and estimate the direct effect of therapeutic LPs on acute mortality in a prospective cohort of HIV-infected individuals with cryptococcal meningitis in Uganda and South Africa. METHODS Study Population Data from the Cryptococcal Optimal ART Timing (COAT) trial, conducted from November 2010 to April 2012, and an observational cohort of patients with cryptococcal meningitis, from April 2012 through December 2012, were used in this analysis. Ethical approval was granted from the Uganda National Council of Science and Technology, South African Medicines Control Council, and the Institutional Review Boards at the University of Minnesota, Makerere University, University of Cape Town, and Mbarara University of Science and Technology.