Browse the corpus

us-negative (RV–) diarrhea of any severity, presenting to Matlab Hospital in (A) ISA and (B) GSA areas. Abbreviations: CRT, cluster-randomized controlled trial; GSA, government service area; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RCT, randomized, controlled trial; YR, year. Rotavirus Vaccine Coverage and Timing Rotavirus vaccine was not available in Matlab between February 2000 and March 2007. Between April 2007 and March 2009, 568 infants in ISA villages were randomized to receive PRV and 568 infants were randomized to placebo as part of a multi-site RCT [16]. In the stratified HRV CRT in both ISA and GSA areas, villages were randomized to 2 doses of HRV at 6 and 10 weeks of age or randomized as observed, control-only villages [17]. In the GSA villages, the CRT started in November 2008, and in the ISA villages, the CRT started in April 2009. Follow-ups and vaccinations during the CRT occurred in both the ISA and GSA villages through March 2011. Through a donation of vaccines post-CRT, HRV was provided routinely starting in April 2011. After September 2014, the rotavirus vaccine was unavailable. HRV vaccine coverage levels among children <1 year of age changed during the study period (Figure 2). During the CRT, both the ISA and GSA villages showed similar vaccine coverage levels. Following the CRT, the coverage level among age-eligible children in ISA villages was maintained at between 65–80%, while GSA villages decreased to 42% at the end of the study period.

(See the Editorial Commentary by Steele and Parashar on pages 2071–3.) Globally, an estimated 13 000 deaths due to rotavirus diarrhea occur annually in children <5 years of age, with most of the burden in sub-Saharan Africa and Asia [1]. While diarrhea-associated mortality rates have decreased globally in the last decade, the burden of rotavirus diarrhea remains substantial in low-income settings [2]. In 2006, 2 rotavirus vaccines were introduced worldwide: GlaxoSmithKline’s human rotavirus vaccine (HRV; Rotarix) and Merck’s pentavalent rotavirus vaccine (PRV; RotaTeq). Large, multi-site, randomized, controlled trials (RCTs) of both vaccines in Africa demonstrated moderate vaccine efficacy (VE) against severe rotavirus diarrhea during the first year of life [3, 4]. As of August 2018, 96 countries, of which 46 are Gavi-eligible, have introduced rotavirus vaccines into their regional or national immunization programs [5]. In the World Health Organization (WHO) Africa region, 74% of countries have introduced rotavirus vaccination. Studies in sub-Saharan Africa have shown statistically significant rotavirus vaccine effectiveness and population-level impacts against all-cause and rotavirus diarrhea in children <5 years of age within 2–3 years of the initiation of routine use [6–14].

a region, 74% of countries have introduced rotavirus vaccination. Studies in sub-Saharan Africa have shown statistically significant rotavirus vaccine effectiveness and population-level impacts against all-cause and rotavirus diarrhea in children <5 years of age within 2–3 years of the initiation of routine use [6–14]. Despite the WHO recommendation for rotavirus vaccine use worldwide, only 18% of countries in the WHO southeast Asia region have introduced a rotavirus vaccine [5]. Limited data on vaccine effectiveness and population impacts may have slowed the introduction of rotavirus vaccines in Asia [15]. The only multi-site RCT of PRV in Asia demonstrated moderate vaccine efficacy against severe rotavirus gastroenteritis in the first 2 years of life (Bangladesh VE 42.7%, 95% confidence interval [CI] 10.4–63.9; Vietnam VE 63.9%, 95% CI 7.6–90.9; combined VE 51.0%, 95% CI 12.8–73.3) [16]. In Bangladesh, this RCT included half of the Matlab villages (International Centre for Diarrhoeal Disease Research, Bangladesh [icddr,b] service areas).

n the first 2 years of life (Bangladesh VE 42.7%, 95% confidence interval [CI] 10.4–63.9; Vietnam VE 63.9%, 95% CI 7.6–90.9; combined VE 51.0%, 95% CI 12.8–73.3) [16]. In Bangladesh, this RCT included half of the Matlab villages (International Centre for Diarrhoeal Disease Research, Bangladesh [icddr,b] service areas). To evaluate the effectiveness of HRV on rotavirus diarrhea in Asia, a 2-year cluster-randomized trial (CRT) was conducted in all villages in Matlab, Bangladesh, beginning in 2008 [17]. The overall effectiveness, which assesses the overall reduction in the incidence of acute rotavirus diarrhea, regardless of vaccination status, was 29.0% (95% CI 11.3–43.1) in children <2 years of age. This study provided initial evidence of the potential population impact of routine rotavirus vaccine use in Bangladesh. After the CRT, HRV was provided for routine use among infants in all Matlab villages between March 2011 and September 2014. To evaluate the population-level impact of HRV in Matlab, Bangladesh, during the 3.5 years of routine use following the CRT, we examined trends in the rotavirus-positive (RV+) and rotavirus-negative (RV–) diarrhea incidence rates of any severity presenting to Matlab Hospital between February 2000 and September 2014.

To evaluate the effectiveness of HRV on rotavirus diarrhea in Asia, a 2-year cluster-randomized trial (CRT) was conducted in all villages in Matlab, Bangladesh, beginning in 2008 [17]. The overall effectiveness, which assesses the overall reduction in the incidence of acute rotavirus diarrhea, regardless of vaccination status, was 29.0% (95% CI 11.3–43.1) in children <2 years of age. This study provided initial evidence of the potential population impact of routine rotavirus vaccine use in Bangladesh. After the CRT, HRV was provided for routine use among infants in all Matlab villages between March 2011 and September 2014. To evaluate the population-level impact of HRV in Matlab, Bangladesh, during the 3.5 years of routine use following the CRT, we examined trends in the rotavirus-positive (RV+) and rotavirus-negative (RV–) diarrhea incidence rates of any severity presenting to Matlab Hospital between February 2000 and September 2014. METHODS Study Setting The study utilized diarrheal surveillance data collected among children <2 years of age residing in villages of the Matlab Health and Demographic Surveillance System (HDSS), administered by the icddr,b, and presenting to Matlab Hospital [18]. The HDSS is divided into the icddr,b service area (ISA; 67 villages) and the government service area (GSA; 75 villages). The icddr,b provides ISA villages with child and maternal health intervention programs and the Bangladesh Ministry of Health and Family Welfare provides GSA villages with the government standard of care. The HDSS maintains a census and registration of vital events, including internal and external migration.

villages). The icddr,b provides ISA villages with child and maternal health intervention programs and the Bangladesh Ministry of Health and Family Welfare provides GSA villages with the government standard of care. The HDSS maintains a census and registration of vital events, including internal and external migration. Immunization Records The HDSS also maintains immunization records through a formal record-keeping system. In the ISA villages, community health workers maintain vaccination records, and in the GSA villages, community health workers check vaccination cards or ask mothers if the card is missing. Diarrheal Surveillance Matlab Hospital is the central diarrhea treatment facility for the Matlab HDSS population. This study includes data from children <2 years of age. The incidence rate for presentations to Matlab Hospital of all-cause diarrhea among children from GSA villages has historically been about half of the incidence rate for presentations from ISA villages [17]. Stool specimens are collected from all patients presenting with diarrhea (3 or more loose stools per 24 hours) to Matlab Hospital. The samples are tested for group A rotavirus VP6 antigens using a solid-phase, sandwich-type enzyme immunoassay (Prospect, Oxoid Diagnostics Ltd, Hampshire, United Kingdom).

villages [17]. Stool specimens are collected from all patients presenting with diarrhea (3 or more loose stools per 24 hours) to Matlab Hospital. The samples are tested for group A rotavirus VP6 antigens using a solid-phase, sandwich-type enzyme immunoassay (Prospect, Oxoid Diagnostics Ltd, Hampshire, United Kingdom). Statistical Analysis Interrupted time series, using segmented regression models, were used to estimate the impact of the rotavirus vaccine introduction in Matlab, Bangladesh, among children <2 years of age [19]. The monthly incidence rates of RV+ and RV– diarrhea were examined separately, by age group (0 to <12 months, 12 to <24 months, and combined [0 to <24 months]). The incidence rates were calculated for RV+ and RV– diarrhea with the number of events presenting to Matlab Hospital per month as the numerator and the monthly population at risk, using HDSS census estimates, as the denominator. Due to varied rotavirus vaccine coverage and baseline diarrheal incidences, analyses were conducted separately for the ISA and GSA villages.

Statistical Analysis Interrupted time series, using segmented regression models, were used to estimate the impact of the rotavirus vaccine introduction in Matlab, Bangladesh, among children <2 years of age [19]. The monthly incidence rates of RV+ and RV– diarrhea were examined separately, by age group (0 to <12 months, 12 to <24 months, and combined [0 to <24 months]). The incidence rates were calculated for RV+ and RV– diarrhea with the number of events presenting to Matlab Hospital per month as the numerator and the monthly population at risk, using HDSS census estimates, as the denominator. Due to varied rotavirus vaccine coverage and baseline diarrheal incidences, analyses were conducted separately for the ISA and GSA villages. Among the ISA villages, the pre-vaccine time period was defined as February 2000–February 2007; the RCT period as March 2007–March 2009); the CRT period as April 2009–March 2011; and the HRV introduction period as April 2011–September 2014. Among the GSA villages, the pre-vaccine time period was defined as February 2000–October 2008; the CRT period as November 2008–March 2011; and the HRV introduction period as April 2011–September 2014. During the CRT periods, the villages were stratified by service area and then randomized to control-only (no placebo) or HRV.

the GSA villages, the pre-vaccine time period was defined as February 2000–October 2008; the CRT period as November 2008–March 2011; and the HRV introduction period as April 2011–September 2014. During the CRT periods, the villages were stratified by service area and then randomized to control-only (no placebo) or HRV. We used 2 models to estimate the impact of HRV use on RV+ and RV– diarrhea incidence rates. Model 1 was defined a priori, while Model 2 was defined after examining the count data. Model 1 and Model 2 differ by both the baseline period used as the referent category and the types of villages included (HRV– and/or control-only). In both models, a generalized linear model was fit to the time-series data, assuming a negative, binomial distribution due to over-dispersion of the data [20]. Calendar months were included in each model to account for seasonality, and a sequential, monthly term for every month over the entire time period was included to account for secular trends. The natural log of the monthly population at risk was included in the model as the offset term. The Breusch-Godfrey test identified some autocorrelation; therefore, 95% CIs were estimated using Newey-West heteroskedastic- and autocorrelation-consistent variance estimators, with a lag of 2 [19, 21]. The estimates of the coefficients for each time period were exponentiated to estimate incidence rate ratios (IRRs), compared to the referent category.

d some autocorrelation; therefore, 95% CIs were estimated using Newey-West heteroskedastic- and autocorrelation-consistent variance estimators, with a lag of 2 [19, 21]. The estimates of the coefficients for each time period were exponentiated to estimate incidence rate ratios (IRRs), compared to the referent category. In Model 1, within the ISA and GSA areas separately, the corresponding pre-vaccine time period was used as the referent category. Villages randomized as both HRV and control-only were included in the analysis. To estimate the IRRs and corresponding 95% CIs, the time periods corresponding to the RCT, CRT, and each of the 3.5 years of routine HRV use were modeled with separate indicator variables. This is a conservative model, which directly compares incidence rates in February 2000–February 2007 (ISA villages) and February 2000–October 2008 (GSA villages) to the years of routine HRV use, starting in April 2011 in all ISA and GSA villages. In the secondary analysis (Model 2), only the villages randomized as control-only during the CRT were used. Within the ISA and GSA regions, the pre-vaccine and CRT time periods were combined in the referent category. The time period corresponding to the RCT was excluded. To estimate the IRRs and corresponding 95% CIs, each of the 3.5 years of routine HRV use were modeled with separate indicator variables. This approach directly compared incidence rates in February 2000–March 2011, excluding the RCT time period, to the years of routine HRV use, starting in April 2011 in those ISA and GSA villages randomized as controls.

ding 95% CIs, each of the 3.5 years of routine HRV use were modeled with separate indicator variables. This approach directly compared incidence rates in February 2000–March 2011, excluding the RCT time period, to the years of routine HRV use, starting in April 2011 in those ISA and GSA villages randomized as controls. The monthly vaccine coverage was estimated as the proportion of children 6 to <52 weeks old receiving each HRV dose within regions of Matlab, Bangladesh. Analyses were completed using Stata version 14 (Stata Corporation, College Station, TX). This study was approved by the ethical review committee of icddr,b in Bangladesh and the Fred Hutchinson Cancer Research Center. RESULTS Tables 1 and 2 and Figure 1 show RV+ and RV– counts and average incidence rates over time in the GSA and ISA villages, using the study populations used for Models 1 and 2. Table 1. Trends in Diarrhea Presenting to Matlab Hospital by Time Period and Model in International Centre for Diarrhoeal Disease Research, Bangladesh, Service Area Model 1 Model 2 ISA February 2000– February 2007 (prevaccine) March 2007– March 2009 (RCT) April 2009– March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011– September 2014) February 2000– March 2011 (prevaccine and CRT, exclude RCT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011– September 2014) 0–12 months of age

Model 1 Model 2 ISA February 2000– February 2007 (prevaccine) March 2007– March 2009 (RCT) April 2009– March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011– September 2014) February 2000– March 2011 (prevaccine and CRT, exclude RCT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011– September 2014) 0–12 months of age Population 18 281 5153 4936 2494 2683 2586 1286 9048 12 119 1327 1427 1342 681 4777 RV+, count 738 265 216 64 81 54 24 223 511 41 47 33 15 136 RV–, count 1258 823 380 179 208 166 79 632 865 92 122 97 44 355 RV+ incidence 40 51 44 26 30 21 19 25 42 31 33 25 22 28 RV– incidence 69 160 77 72 78 64 61 70 71 69 85 72 65 74 12–24 months of age Population 18 363 5124 5008 2437 2493 2649 1275 8853 12 243 1280 1325 1408 661 4674 RV+, count 502 200 145 43 41 49 9 142 351 27 20 28 4 79 RV–, count 844 432 185 87 90 89 44 310 558 46 47 53 24 170 RV+ incidence 27 39 29 18 16 19 7 16 29 21 15 20 6 17 RV– incidence 46 84 37 36 36 34 35 35 46 36 35 38 36 36 0–24 months of age Population 36 644 10 276 9945 4930 5176 5235 2561 17 901 24 363 2607 2752 2750 1342 9451 RV+, count 1240 465 361 107 122 103 33 365 862 68 67 61 19 215 RV–, count 2102 1255 565 266 298 255 123 942 1423 138 169 150 68 525 RV+ incidence 34 45 36 22 24 20 13 20 35 26 24 22 14 23 RV– incidence 57 122 57 54 58 49 48 53 58 53 61 55 51 56 Incidence data are per 1000 person-years.

1 24 363 2607 2752 2750 1342 9451 RV+, count 1240 465 361 107 122 103 33 365 862 68 67 61 19 215 RV–, count 2102 1255 565 266 298 255 123 942 1423 138 169 150 68 525 RV+ incidence 34 45 36 22 24 20 13 20 35 26 24 22 14 23 RV– incidence 57 122 57 54 58 49 48 53 58 53 61 55 51 56 Incidence data are per 1000 person-years. Abbreviations: CRT, cluster-randomized controlled trial; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RCT, randomized, controlled trial; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. Table 2. Trends in Diarrhea Presenting to Matlab Hospital by Time Period and Model in Government Service Area Model 1 Model 2 GSA February 2000- October 2008 (prevaccine) November 2008-March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011–September 2014) February 2000– March 2011 (prevaccine and CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) Vaccine Years (April 2011–September 2014) 0–12 months of age Population 22 777 5359 2317 2306 2201 1162 7987 12 163 962 988 934 483 3368 RV+, count 542 144 35 51 37 15 138 285 21 18 14 5 58 RV–, count 671 142 59 80 64 26 229 310 27 32 24 11 94 RV+ incidence 24 27 15 22 17 13 17 23 22 18 15 10 17 RV– incidence 29 26 25 35 29 22 29 25 28 32 26 23 28 12–24 months of age

0–12 months of age Population 22 777 5359 2317 2306 2201 1162 7987 12 163 962 988 934 483 3368 RV+, count 542 144 35 51 37 15 138 285 21 18 14 5 58 RV–, count 671 142 59 80 64 26 229 310 27 32 24 11 94 RV+ incidence 24 27 15 22 17 13 17 23 22 18 15 10 17 RV– incidence 29 26 25 35 29 22 29 25 28 32 26 23 28 12–24 months of age Population 23 168 5641 2224 2312 2305 1111 7951 12 507 939 970 993 479 3381 RV+, count 365 90 46 28 20 8 102 196 25 12 12 1 50 RV–, count 459 94 40 44 24 13 121 249 15 11 11 5 42 RV+ incidence 16 16 21 12 9 7 13 16 27 12 12 2 15 RV– incidence 20 17 18 19 10 12 15 20 16 11 11 10 12 0–24 months of age Population 45 945 10 999 4541 4618 4506 2273 15 938 24 670 1901 1958 1927 963 6748 RV+, count 907 234 81 79 57 23 240 481 46 30 26 6 108 RV–, count 1130 236 99 124 88 39 350 559 42 43 35 16 136 RV+ incidence 20 21 18 17 13 10 15 19 24 15 13 6 16 RV– incidence 25 21 22 27 20 17 22 23 22 22 18 17 20 Incidence data are per 1000 person-years. Abbreviations: CRT, cluster-randomized controlled trial; GSA, government service area; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. Figure 1. Observed counts of rotavirus-positive (RV+) and rotavirus-negative (RV–) diarrhea of any severity, presenting to Matlab Hospital in (A) ISA and (B) GSA areas. Abbreviations: CRT, cluster-randomized controlled trial; GSA, government service area; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RCT, randomized, controlled trial; YR, year.

of age changed during the study period (Figure 2). During the CRT, both the ISA and GSA villages showed similar vaccine coverage levels. Following the CRT, the coverage level among age-eligible children in ISA villages was maintained at between 65–80%, while GSA villages decreased to 42% at the end of the study period. Figure 2. Timing of HRV coverage (dose 1) over time by ISA and GSA villages randomized to HRV or control only in <1-year-olds. Abbreviations: GSA, government service area; HRV, human rotavirus vaccine; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area. ISA, HRV: icddr, b service areas randomized to HRV during the CRT; ISA, Control: icddr, b service areas randomized as control-only villages during the CRT; GSA, HRV: Government service areas randomized to HRV during the CRT; GSA, Control: icddr, b service areas randomized as control-only villages during the CRT; *23 children were vaccinated in GSA Villages in September-October 2008 before the start of the cluster-randomized trial (CRT). This time period is still considered prevaccine due to the small number of children vaccinated. Observed and predicted RV+ diarrhea counts in ISA and GSA villages for both models demonstrated a satisfactory model fit (Supplementary Figures 1–2).

Figure 2. Timing of HRV coverage (dose 1) over time by ISA and GSA villages randomized to HRV or control only in <1-year-olds. Abbreviations: GSA, government service area; HRV, human rotavirus vaccine; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area. ISA, HRV: icddr, b service areas randomized to HRV during the CRT; ISA, Control: icddr, b service areas randomized as control-only villages during the CRT; GSA, HRV: Government service areas randomized to HRV during the CRT; GSA, Control: icddr, b service areas randomized as control-only villages during the CRT; *23 children were vaccinated in GSA Villages in September-October 2008 before the start of the cluster-randomized trial (CRT). This time period is still considered prevaccine due to the small number of children vaccinated. Observed and predicted RV+ diarrhea counts in ISA and GSA villages for both models demonstrated a satisfactory model fit (Supplementary Figures 1–2). Diarrhea Incidence Trends: International Centre for Diarrhoeal Disease Research, Bangladesh, Service Area Villages Using Model 1, with the pre-vaccine time period as the referent category, RV+ diarrhea rates increased during the RCT period and the CRT period in both age groups in ISA villages (Table 3; Figure 3). During periods of routine HRV use, there was a downward trend that was not statistically significant in RV+ diarrhea incidences after each additional year of vaccine use. During the entire 3.5 years of routine use, there was no meaningful decrease in RV+ diarrhea rates in 0- to <12-month-old children (IRR 0.72, 95% CI 0.39–1.33) or 12- to <24-month-old children (IRR 0.91, 95% CI 0.46–1.83). Using Model 2, combining the pre-vaccine time period and the CRT time period in the reference category and using control-only villages, there was a downward trend in the RV+ diarrhea incidence rates after each additional year of routine HRV use in both age groups (Table 4; Figure 3). During 3.5 years of routine HRV use, there was a statistically significant, 41% decrease in RV+ diarrhea rates in 0- to <12-month-old children (IRR 0.59, 95% CI 0.43–0.80), a 35% decrease in 12- to <24-month-old children (IRR 0.65, 95% CI 0.42–1.02), and a statistically significant, 39% decrease in children 0 to <24 months of age (IRR 0.61, 95% CI 0.45–0.82).

e was a statistically significant, 41% decrease in RV+ diarrhea rates in 0- to <12-month-old children (IRR 0.59, 95% CI 0.43–0.80), a 35% decrease in 12- to <24-month-old children (IRR 0.65, 95% CI 0.42–1.02), and a statistically significant, 39% decrease in children 0 to <24 months of age (IRR 0.61, 95% CI 0.45–0.82). Table 3. Diarrhea Trends in International Centre for Diarrhoeal Disease Research, Bangladesh, Service Area Region (Model 1) ISA Feb 2000- Feb 2007 (prevaccine) March 2007– March 2009 (RCT) April 2009– March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) April 2011– September 2014 (vaccine years) IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI 0–12 months of age RV+ REF 1.32 0.91 1.92 1.16 0.72 1.88 0.67 0.37 1.21 0.85 0.41 1.75 0.57 0.27 1.20 0.63 0.26 1.49 0.72 0.39 1.33 RV– REF 2.93 2.03 4.24 1.51 1.14 2.00 1.49 1.01 2.19 1.71 1.18 2.49 1.46 0.97 2.19 1.25 0.79 1.96 1.59 1.09 2.31 12–24 months of age RV+ REF 1.84 1.19 2.84 1.45 0.86 2.46 0.86 0.41 1.80 0.94 0.40 2.17 1.08 0.47 2.45 0.62 0.25 1.56 0.91 0.46 1.83 RV– REF 1.95 1.42 2.69 0.86 0.60 1.22 0.85 0.54 1.34 0.91 0.52 1.58 0.83 0.45 1.54 0.67 0.35 1.30 0.88 0.56 1.38 0–24 months of age

RV+ REF 1.32 0.91 1.92 1.16 0.72 1.88 0.67 0.37 1.21 0.85 0.41 1.75 0.57 0.27 1.20 0.63 0.26 1.49 0.72 0.39 1.33 RV– REF 2.93 2.03 4.24 1.51 1.14 2.00 1.49 1.01 2.19 1.71 1.18 2.49 1.46 0.97 2.19 1.25 0.79 1.96 1.59 1.09 2.31 12–24 months of age RV+ REF 1.84 1.19 2.84 1.45 0.86 2.46 0.86 0.41 1.80 0.94 0.40 2.17 1.08 0.47 2.45 0.62 0.25 1.56 0.91 0.46 1.83 RV– REF 1.95 1.42 2.69 0.86 0.60 1.22 0.85 0.54 1.34 0.91 0.52 1.58 0.83 0.45 1.54 0.67 0.35 1.30 0.88 0.56 1.38 0–24 months of age RV+ REF 1.50 1.03 2.19 1.26 0.80 2.00 0.72 0.40 1.31 0.90 0.44 1.84 0.74 0.36 1.52 0.66 0.29 1.51 0.79 0.43 1.43 RV– REF 2.55 1.91 3.41 1.24 0.98 1.55 1.23 0.89 1.70 1.40 0.99 1.97 1.19 0.82 1.73 1.01 0.68 1.49 1.31 0.95 1.79 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; IRR, incidence rate ratio; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RCT, randomized, controlled trial; REF, referent; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. Figure 3. Observed incidences and IRRs of RV+ and RV– diarrhea of any severity presenting to Matlab Hospital in ISA villages using Models 1 and 2 in (A) 0– to <12-month-old children and (B) 12 to <24-month-old children. Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; IRR, incidence rate ratio; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RV–, rotavirus negative; RV+, rotavirus positive; RCT, randomized, controlled trial; YR, year.

) 12 to <24-month-old children. Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; IRR, incidence rate ratio; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RV–, rotavirus negative; RV+, rotavirus positive; RCT, randomized, controlled trial; YR, year. Table 4. Diarrhea Trends, International Centre for Diarrhoeal Disease Research, Bangladesh, Service Area Region (Model 2) ISA Feb 2000-March 2011 (prevaccine and CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) April 2011– September 2014 (vaccine years) IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI 0–12 months of age RV+ REF 0.61 0.43 0.86 0.68 0.46 1.00 0.46 0.29 0.72 0.51 0.29 0.90 0.59 0.43 0.80 RV– REF 1.07 0.74 1.52 1.33 1.05 1.69 1.14 0.87 1.49 0.89 0.62 1.28 1.15 0.91 1.47 12–24 months of age RV+ REF 0.72 0.44 1.17 0.58 0.29 1.12 0.74 0.41 1.33 0.33 0.11 0.99 0.65 0.42 1.02 RV– REF 1.00 0.72 1.39 1.05 0.62 1.79 1.10 0.66 1.84 0.86 0.56 1.32 1.03 0.74 1.43 0–24 months of age RV+ REF 0.64 0.46 0.89 0.66 0.44 0.98 0.55 0.35 0.85 0.48 0.25 0.91 0.61 0.45 0.82 RV– REF 1.05 0.80 1.38 1.27 0.98 1.64 1.12 0.85 1.46 0.89 0.70 1.13 1.12 0.91 1.37 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; IRR, incidence rate ratio; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RV–, rotavirus negative; RV+, rotavirus positive; YR, year.

0 1.38 1.27 0.98 1.64 1.12 0.85 1.46 0.89 0.70 1.13 1.12 0.91 1.37 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; IRR, incidence rate ratio; ISA, International Centre for Diarrhoeal Disease Research, Bangladesh, service area; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. In Model 1, RV– diarrhea rates increased during the RCT period and the CRT period in both age groups. During periods of routine HRV use, there was an increased risk of RV– diarrhea in 0- to <12-month-old children (IRR 1.59, 95% CI 1.09–2.31) and no meaningful change in 12- to <24-month-old children. In Model 2, there were no statistically significant changes in RV– diarrhea rates during periods of HRV routine use.

ups. During periods of routine HRV use, there was an increased risk of RV– diarrhea in 0- to <12-month-old children (IRR 1.59, 95% CI 1.09–2.31) and no meaningful change in 12- to <24-month-old children. In Model 2, there were no statistically significant changes in RV– diarrhea rates during periods of HRV routine use. Diarrhea Incidence Trends: Government Service Area Villages Using Model 1, with the pre-vaccine time period as the referent category, the incidence of RV+ diarrhea increased during the CRT period in 0- to <12-month-old children, but did not meaningfully change in 12- to <24-month-old children (Table 5; Figure 4). During periods of routine HRV use, there was an upward trend in the RV+ diarrhea incidence after each additional year of vaccine use in 0- to <12-month-old children, but no clear trends in 12- to <24-month-old children. During 3.5 years of routine use, there was no statistically significant change in the incidences of RV+ diarrhea in 0- to <12-month-old children (IRR 1.25, 95% CI 0.78–2.01) or in 12- to <24-month-old children (IRR 1.00, 95% CI 0.52–1.92). Using Model 2, there was a downward trend in the RV+ diarrhea incidence after each additional year of routine HRV use in both age groups (Table 6; Figure 4). However, during 3.5 years of routine HRV use, there was no meaningful change in the RV+ diarrhea rate in either age group. In Models 1 and 2, there were no statistically significant changes in RV– diarrhea rates during periods of HRV routine use. Table 5. Diarrhea Trends, Government Service Area Region (Model 1)

Diarrhea Incidence Trends: Government Service Area Villages Using Model 1, with the pre-vaccine time period as the referent category, the incidence of RV+ diarrhea increased during the CRT period in 0- to <12-month-old children, but did not meaningfully change in 12- to <24-month-old children (Table 5; Figure 4). During periods of routine HRV use, there was an upward trend in the RV+ diarrhea incidence after each additional year of vaccine use in 0- to <12-month-old children, but no clear trends in 12- to <24-month-old children. During 3.5 years of routine use, there was no statistically significant change in the incidences of RV+ diarrhea in 0- to <12-month-old children (IRR 1.25, 95% CI 0.78–2.01) or in 12- to <24-month-old children (IRR 1.00, 95% CI 0.52–1.92). Using Model 2, there was a downward trend in the RV+ diarrhea incidence after each additional year of routine HRV use in both age groups (Table 6; Figure 4). However, during 3.5 years of routine HRV use, there was no meaningful change in the RV+ diarrhea rate in either age group. In Models 1 and 2, there were no statistically significant changes in RV– diarrhea rates during periods of HRV routine use. Table 5. Diarrhea Trends, Government Service Area Region (Model 1) GSA Feb 2000– Oct 2008 (prevaccine) Nov 2008– March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) April 2011– September 2014 (vaccine years) IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI 0–12 months of age

Table 5. Diarrhea Trends, Government Service Area Region (Model 1) GSA Feb 2000– Oct 2008 (prevaccine) Nov 2008– March 2011 (CRT) April 2011– March 2012 (YR1) April 2012– March 2013 (YR2) April 2013– March 2014 (YR3) April 2014– September 2014 (YR3.5) April 2011– September 2014 (vaccine years) IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI 0–12 months of age RV+ REF 1.46 1.02 2.09 0.99 0.63 1.54 1.58 0.90 2.77 1.27 0.75 2.16 1.55 0.66 3.61 1.25 0.78 2.01 RV– REF 1.01 0.75 1.37 0.96 0.62 1.47 1.33 0.87 2.04 1.13 0.70 1.82 0.79 0.43 1.43 1.10 0.73 1.65 RV+ REF 0.98 0.63 1.53 1.31 0.73 2.33 0.82 0.38 1.77 0.60 0.28 1.25 0.89 0.36 2.22 1.00 0.52 1.92 RV– REF 0.97 0.72 1.32 1.06 0.70 1.62 1.16 0.76 1.75 0.65 0.37 1.13 0.67 0.40 1.14 0.98 0.64 1.50 0–24 months of age RV+ REF 1.26 0.89 1.77 1.18 0.76 1.85 1.24 0.72 2.12 0.96 0.57 1.63 1.34 0.61 2.95 1.16 0.73 1.85 RV– REF 0.98 0.77 1.26 0.99 0.71 1.38 1.25 0.88 1.78 0.92 0.61 1.40 0.75 0.47 1.19 1.04 0.74 1.47 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; GSA, government service area; IRR, incidence rate ratio; RV–, rotavirus negative; RV+, rotavirus positive; YR, year.

.61 2.95 1.16 0.73 1.85 RV– REF 0.98 0.77 1.26 0.99 0.71 1.38 1.25 0.88 1.78 0.92 0.61 1.40 0.75 0.47 1.19 1.04 0.74 1.47 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; GSA, government service area; IRR, incidence rate ratio; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. Figure 4. Observed incidence and IRRs of RV+ and RV– diarrhea of any severity presenting to Matlab Hospital in GSA villages using Models 1 and 2 in (A) 0 to <12-month-old children and (B) 12 to <24-month-old children. Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; GSA, government service area; IRR, incidence rate ratio; RV–, rotavirus negative; RV+, rotavirus positive; RCT, randomized, controlled trial; YR, year. Table 6. Diarrhea Trends, Government Service Area Region (Model 2) GSA Feb 2000–March 2011 (prevaccine and CRT) April 2011–March 2012 (YR1) April 2012–March 2013 (YR2) April 2013–March 2014 (YR3) April 2014–September 2014 (YR3.5) April 2011–September 2014 (vaccine years) IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI IRR 95% CI 0–12 months of age RV+ REF 0.91 0.58 1.43 0.79 0.39 1.61 0.62 0.34 1.14 0.65 0.29 1.46 0.78 0.48 1.27 RV– REF 1.38 0.93 2.06 1.66 1.00 2.75 1.36 0.85 2.17 1.06 0.60 1.85 1.43 0.97 2.12 12–24 months of age RV+ REF 1.38 0.91 2.07 0.63 0.34 1.16 0.61 0.36 1.03 0.23 0.03 1.53 0.87 0.53 1.43 RV– REF 0.98 0.56 1.72 0.72 0.43 1.22 0.72 0.40 1.30 0.61 0.33 1.10 0.79 0.51 1.23 0–24 months of age

RV+ REF 0.91 0.58 1.43 0.79 0.39 1.61 0.62 0.34 1.14 0.65 0.29 1.46 0.78 0.48 1.27 RV– REF 1.38 0.93 2.06 1.66 1.00 2.75 1.36 0.85 2.17 1.06 0.60 1.85 1.43 0.97 2.12 12–24 months of age RV+ REF 1.38 0.91 2.07 0.63 0.34 1.16 0.61 0.36 1.03 0.23 0.03 1.53 0.87 0.53 1.43 RV– REF 0.98 0.56 1.72 0.72 0.43 1.22 0.72 0.40 1.30 0.61 0.33 1.10 0.79 0.51 1.23 0–24 months of age RV+ REF 1.12 0.80 1.55 0.72 0.48 1.08 0.62 0.41 0.94 0.50 0.24 1.06 0.82 0.57 1.19 RV– REF 1.20 0.87 1.67 1.24 0.84 1.84 1.06 0.71 1.58 0.86 0.56 1.31 1.15 0.86 1.53 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; GSA, government service area; IRR, incidence rate ratio; RV–, rotavirus negative; RV+, rotavirus positive; YR, year.

.41 0.94 0.50 0.24 1.06 0.82 0.57 1.19 RV– REF 1.20 0.87 1.67 1.24 0.84 1.84 1.06 0.71 1.58 0.86 0.56 1.31 1.15 0.86 1.53 Abbreviations: CI, confidence interval; CRT, cluster-randomized controlled trial; GSA, government service area; IRR, incidence rate ratio; RV–, rotavirus negative; RV+, rotavirus positive; YR, year. DISCUSSION Our study demonstrates a decreasing trend in RV+ diarrhea incidences among children <2 years of age from ISA villages presenting to Matlab Hospital during 3.5 years of routine HRV use. Using a conservative model to estimate pre-vaccination rotavirus diarrhea trends (Model 1), the results were not statistically significant. However, by restricting the analysis to control-only villages, we gained an additional 2 years of recent, pre-vaccine time to model baseline trends (Model 2), and found a statistically significant, 39% reduction in RV+ diarrhea rates in children 0 to <24 months of age. No significant impact of HRV on the RV+ diarrhea incidence among children from GSA villages was observed using either model. Differences in the population-level impacts between ISA and GSA villages are likely due to lower HRV coverage and lower reported diarrhea incidences in GSA areas, compared to ISA villages.

f age. No significant impact of HRV on the RV+ diarrhea incidence among children from GSA villages was observed using either model. Differences in the population-level impacts between ISA and GSA villages are likely due to lower HRV coverage and lower reported diarrhea incidences in GSA areas, compared to ISA villages. Our study also examined changes in the rate of RV– diarrhea as a control outcome, with the assumption that HRV introduction should have no significant impact on RV– diarrhea [22]. In Model 1, using only the pre-vaccine period in the referent category, we observed an increasing trend in both RV+ and RV– diarrhea rates in children 0 to <24 months of age in ISA villages during the RCT and CRT time periods. While other interventions or unmeasured biases may have influenced the all-cause gastroenteritis incidence, we believe this increase was due to changes in health-care–seeking behaviors due to the RCT. During the RCT, field staff visited the homes of infants enrolled in the study to remind parents to bring their child to the hospital for episodes of diarrhea [16]. A change in community health-care–seeking behavior is the most likely explanation, as there was no significant change in all-cause diarrhea in the corresponding time period in the GSA villages, where no RCT took place (Figure 1), and no specific pathogen was identified as a cause of the increase in all-cause diarrhea. The most conservative model to estimate the HRV impact (Model 1) modelled the RCT and CRT time periods separately and directly compared the pre-vaccine time period to the years of routine HRV use in both ISA and GSA villages. However, if increased health-care–seeking behaviors were sustained, results from Model 1 would underestimate the population-level impact of HRV.

pact (Model 1) modelled the RCT and CRT time periods separately and directly compared the pre-vaccine time period to the years of routine HRV use in both ISA and GSA villages. However, if increased health-care–seeking behaviors were sustained, results from Model 1 would underestimate the population-level impact of HRV. In the secondary analysis (Model 2), both to increase power and to include relevant health-care–seeking behaviors to estimate the baseline incidence, we restricted the analysis to those villages randomized as control-only during the CRT period, and assessed the impact of routine HRV use on diarrhea over time. The referent category combined the pre-vaccine time period and the CRT time period. These models showed a significant impact of routine HRV use on RV+ diarrhea rates in 0- to <24-month-old children in ISA villages, but not in GSA villages. RV– diarrhea rates did not significantly change over time using this model. Notably, both models showed a decreasing trend in RV+ diarrhea in ISA villages during sustained HRV coverage. This analysis demonstrates the importance of using the appropriate baseline incidences and underlying trends in time-series analyses.

RV– diarrhea rates did not significantly change over time using this model. Notably, both models showed a decreasing trend in RV+ diarrhea in ISA villages during sustained HRV coverage. This analysis demonstrates the importance of using the appropriate baseline incidences and underlying trends in time-series analyses. Despite the potential differences in health-care–seeking behavior over time, our results are similar to the RCT and CRT conducted in Matlab, Bangladesh, with the greatest impact of rotavirus vaccine on children 0 to <12 months of age. To our knowledge, no other population-level impact analyses have been reported in Asia with rotavirus diarrhea as the outcome, though a study in the Philippines saw a 60% (95% CI 55–64%) reduction in all-cause diarrhea hospitalizations within 4 years after rotavirus vaccine introduction [23]. Similar time-series analyses conducted 2–3 years after rotavirus introduction found a 49% (95% CI 32–63%) decrease in rotavirus diarrhea in <5-year-old children in Ghana [12], a 54% (95% CI 33–69%) decrease in rotavirus diarrhea in <1-year-old children in Malawi [11], a 33% (95% CI 25–41%) reduction in rotavirus diarrhea in <5-year-old children in Botswana [14], and a 38% reduction in rotavirus positivity among children <5 years old in Zambia [10]. Long-term impacts were also observed in Ghana [24] and Zambia [25]. Importantly, in these studies, >90% vaccine coverage for 1 or 2 doses of rotavirus vaccine were reported within 1 year of vaccine introduction. In our study, the maximum, 2-dose HRV coverage of 68% was attained in the ISA villages during the second year of routine use.

were also observed in Ghana [24] and Zambia [25]. Importantly, in these studies, >90% vaccine coverage for 1 or 2 doses of rotavirus vaccine were reported within 1 year of vaccine introduction. In our study, the maximum, 2-dose HRV coverage of 68% was attained in the ISA villages during the second year of routine use. Our study has limitations. As in any time-series analysis, our study may have been confounded by other interventions or other unmeasured factors associated with RV+ diarrhea and the timing of the vaccine introduction. However, our confidence in the impact of HRV is increased, because no meaningful changes in RV– diarrhea incidences were observed. Second, while the Matlab HDSS database shows lower vaccine coverage in GSA areas, coverage may be underestimated or inaccurate due to the lack of recording on health cards in this region and potential reliance on maternal reports. Though measured with the same potential biases, during the study period, the average coverage for 3 doses of Diphtheria-Pertussis-Tetanus (DTP3) was 97% in ISA villages and 91% in GSA villages [26]. Third, with the available data, we were unable to assess the impact of the rotavirus vaccine on severe rotavirus diarrhea, as indicated by a Vesikari score ≥11, which is the outcome used in rotavirus vaccine clinical trials.

es of Diphtheria-Pertussis-Tetanus (DTP3) was 97% in ISA villages and 91% in GSA villages [26]. Third, with the available data, we were unable to assess the impact of the rotavirus vaccine on severe rotavirus diarrhea, as indicated by a Vesikari score ≥11, which is the outcome used in rotavirus vaccine clinical trials. This study provides initial evidence of the population-level impact of rotavirus vaccines in children <2 years of age in regions of high vaccine coverage in Matlab, Bangladesh. Pecenka et al [27] estimated that, with a Gavi subsidy in Bangladesh, the averted cost/disability adjusted life year (DALY) ratio ranged between $58/DALY and $142/DALY, indicating a highly cost-effective vaccine, given 94% coverage of DTP3 in Bangladesh [27, 28] In our study, during the pre-vaccine period, rotavirus was detected in 34.5% of diarrhea cases in children <5 years of age presenting to Matlab Hospital. Other regions of Bangladesh show an average of 64% of diarrhea instances being due to rotavirus in children <5 years of age [29]. With sustained vaccine coverage and a considerable nationwide burden of rotavirus diarrhea, larger impacts of HRV on rotavirus gastroenteritis are likely to be observed long-term in Bangladesh. This may provide additional evidence to influence other countries in the region to introduce the rotavirus vaccine.

age [29]. With sustained vaccine coverage and a considerable nationwide burden of rotavirus diarrhea, larger impacts of HRV on rotavirus gastroenteritis are likely to be observed long-term in Bangladesh. This may provide additional evidence to influence other countries in the region to introduce the rotavirus vaccine. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. ciz133_suppl_Supplementary_Figure_1AB Click here for additional data file. ciz133_suppl_Supplementary_Figure_2AB Click here for additional data file. Notes Acknowledgments.The authors thank PATH for its commitment to their research efforts and for the negotiation of vaccines for Matlab, Bangladesh, following the cluster-randomized study. They thank the governments of Bangladesh, Canada, Sweden, and the United Kingdom for providing core/unrestricted support. They thank the families of the Matlab field area who participated in this study, and the field staff of the International Centre for Diarrhoeal Disease Research, Bangladesh, without whom they could not have completed this research. They thank PATH and GlaxoSmithKline for the donation of vaccines to the study and the study population, both during the cluster-randomized trial and for 3 years following the study.

d staff of the International Centre for Diarrhoeal Disease Research, Bangladesh, without whom they could not have completed this research. They thank PATH and GlaxoSmithKline for the donation of vaccines to the study and the study population, both during the cluster-randomized trial and for 3 years following the study. Disclaimer.The findings and conclusions contained within are those of the authors and do not necessarily reflect the positions or policies of the Bill & Melinda Gates Foundation or the National Institutes of Health. The funders of the study had no role in the study design, data collection, data analysis, data interpretation, writing of the report, or decision to submit the article for publication. Financial support.This work was supported by the Bill & Melinda Gates Foundation (grant number OPP1097672) and the National Institute of Allergy and Infectious Diseases (NIAID) of the National Institutes of Health (NIH) (grant number MERIT R37 AI032042). 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.

Community-acquired respiratory viruses cause lower respiratory tract disease (LRTD) after hematopoietic cell transplantation (HCT), which is associated with high morbidity and mortality [1, 2]. The definition of LRTD associated with respiratory viruses differs in various studies of outcomes and risk factors for mortality following LRTD [3–8]. The strictest definition of LRTD requires LRTD symptoms, abnormal chest radiography, and viral detection in lower respiratory samples, whereas a less stringent definition often consists of LRTD symptoms and a positive nasopharyngeal sample. This spectrum of definitions for LRTD likely includes a broad range of disease stages, resulting in difficulty interpreting and comparing published study results. Parainfluenza virus (PIV) is a respiratory virus that frequently infects HCT recipients [9–11]. In previous studies, the mortality among patients with LRTD ranged from 13% to 63% [3–6, 11, 12]. Although factors such as copathogens, mechanical ventilation, underlying disease, and steroid use have been reported as risk factors for mortality [4–6], it is not known whether these factors independently predict poor outcomes in HCT recipients.

mortality among patients with LRTD ranged from 13% to 63% [3–6, 11, 12]. Although factors such as copathogens, mechanical ventilation, underlying disease, and steroid use have been reported as risk factors for mortality [4–6], it is not known whether these factors independently predict poor outcomes in HCT recipients. We hypothesized that the differences in definitions of LRTD used in prior studies are a major determinant of the observed variability in outcome. The purpose of this study was to classify PIV LRTD according to the site of virologic detection and clinical manifestations, and to compare clinical outcomes among the groups. Moreover, we examined various risk factors for overall mortality and death due to respiratory failure. METHODS Study Design This retrospective cohort study includes patients who first received transplant between December 1990 and December 2011 at the Fred Hutchinson Cancer Research Center (FHCRC) and had documented PIV infection after HCT that was virologically diagnosed at the University of Washington Virology Laboratories (a subset of patients were included in a previously published report [4]). Only an individual's first episode of PIV infection was analyzed. Patients' demographic data and transplant information closest to the PIV infection were retrieved from the FHCRC database, and other data related to the clinical course of PIV infections were collected by medical record review. The study was approved by the Institutional Review Board at FHCRC.

V infection was analyzed. Patients' demographic data and transplant information closest to the PIV infection were retrieved from the FHCRC database, and other data related to the clinical course of PIV infections were collected by medical record review. The study was approved by the Institutional Review Board at FHCRC. Definitions PIV detection was performed by conventional culture, direct fluorescent antibody tests, and/or reverse transcription polymerase chain reaction (RT-PCR) assay in respiratory samples. PIV upper respiratory tract infection (URTI) was defined as PIV detection in a nasopharyngeal or sputum sample, with URTI symptoms but no new pulmonary infiltrates. LRTD was divided into 3 groups: possible, probable, and proven. Possible LRTD was defined as PIV detection in a nasopharyngeal or sputum sample with new pulmonary infiltrates (but without confirmation of PIV in the lower respiratory tract) with or without LRTD signs or symptoms (eg, cough, wheezing, rales, tachypnea, shortness of breath, dyspnea, or hypoxia). Probable LRTD was defined as PIV detection in a bronchoalveolar lavage (BAL) or lung biopsy sample with LRTD symptoms, with or without pulmonary function decline, and without new pulmonary infiltrates. The definition of proven LRTD was PIV detection in a BAL or biopsy sample with new pulmonary infiltrates with or without LRTD symptoms.

defined as PIV detection in a bronchoalveolar lavage (BAL) or lung biopsy sample with LRTD symptoms, with or without pulmonary function decline, and without new pulmonary infiltrates. The definition of proven LRTD was PIV detection in a BAL or biopsy sample with new pulmonary infiltrates with or without LRTD symptoms. Viral load was determined by quantitative RT-PCR using stored frozen repository samples [13]. Peak steroid dose was recorded from the period within 2 weeks before PIV infection in patients with URTI. In PIV LRTD cases, peak steroid doses were recorded from within 2 weeks before and after LRTD diagnosis, respectively, and exact steroid dose at 1 month after diagnosis was also collected. Death caused by respiratory failure was defined as any death caused exclusively or predominantly by respiratory failure [14].

URTI. In PIV LRTD cases, peak steroid doses were recorded from within 2 weeks before and after LRTD diagnosis, respectively, and exact steroid dose at 1 month after diagnosis was also collected. Death caused by respiratory failure was defined as any death caused exclusively or predominantly by respiratory failure [14]. Statistical Analysis Patients' demographic characteristics were summarized and compared among disease categories using χ2 or Fisher exact test for categorical variables, and Wilcoxon rank-sum test for continuous variables (as appropriate). Wilcoxon rank-sum or Kruskal-Wallis test was utilized for comparisons of continuous variables. The probability of overall survival was estimated using the Kaplan-Meier method. The probability of mortality caused by respiratory failure and incidence of mechanical ventilation were estimated by cumulative incidence curves, treating death due to other causes as a competing risk event. The log-rank test was used to compare univariable hazards of time-to-event outcomes between disease categories. Cox proportional hazards models were used to evaluate unadjusted and adjusted hazard ratios (HRs) for mortality or respiratory mortality, and associated 95% confidence intervals (CIs) were reported. Variables with P ≤ .1 in the univariable models were candidates for multivariable models. Two-sided P values <.05 were considered statistically significant.

e used to evaluate unadjusted and adjusted hazard ratios (HRs) for mortality or respiratory mortality, and associated 95% confidence intervals (CIs) were reported. Variables with P ≤ .1 in the univariable models were candidates for multivariable models. Two-sided P values <.05 were considered statistically significant. RESULTS Patient Characteristics A total of 544 patients had PIV infection following HCT; 345 (63%) and 199 (37%) had URTI and LRTD disease, respectively. LRTD classification of patients included 78 (39%) possible, 19 (10%) probable, and 102 (51%) proven cases of LRTD. Characteristics of each PIV infection group are shown in Table 1. The median time to PIV infection and PIV LRTD after HCT was 71.5 days (range, 0–3140 days) and 78 days (range, 3–3140), respectively. Table 1. Characteristics of All Patients With Parainfluenza Virus Infection Characteristic Total (N = 544) URTI (n = 345) LRTD P Value Possible Probable Proven (n = 78) (n = 19) (n = 102) Sex .11 Male 321 (59) 201 (58) 40 (51) 15 (79) 65 (64) Female 223 (41) 144 (42) 38 (49) 4 (21) 37 (36) Age at transplant, y .14 ≤20 105 (19) 71 (20) 19 (24) 3 (16) 12 (12) 21–60 365 (67) 227 (66) 45 (58) 15 (79) 78 (77) >60 74 (14) 47 (14) 14 (18) 1 (5) 12 (12) Transplant yeara <.001 1990–2000 269 (49) 175 (51) 25 (32) 15 (79) 54 (53) 2001–2011 275 (51) 170 (49) 53 (68) 4 (21) 48 (47) Transplant No.

(49) 4 (21) 37 (36) Age at transplant, y .14 ≤20 105 (19) 71 (20) 19 (24) 3 (16) 12 (12) 21–60 365 (67) 227 (66) 45 (58) 15 (79) 78 (77) >60 74 (14) 47 (14) 14 (18) 1 (5) 12 (12) Transplant yeara <.001 1990–2000 269 (49) 175 (51) 25 (32) 15 (79) 54 (53) 2001–2011 275 (51) 170 (49) 53 (68) 4 (21) 48 (47) Transplant No. .55 First 488 (89) 311 (90) 69 (88) 19 (100) 89 (87) Second 53 (10) 31 (9) 9 (12) 0 (0) 13 (13) Third 3 (1) 3 (1) 0 (0) 0 (0) 0 (0) Cell source <.001 Bone marrow 251 (46) 168 (49) 23 (30) 16 (84) 44 (43) Peripheral blood stem cell 269 (49) 167 (48) 46 (59) 3 (16) 53 (52) Cord blood 24 (4) 10 (3) 9 (11) 0 (0) 5 (5) Donor type <.001 Autologous 104 (19) 71 (21) 16 (21) 0 (0) 17 (17) Related 202 (37) 135 (39) 23 (29) 8 (42) 36 (35) Unrelated 238 (44) 139 (40) 39 (50) 11 (58) 49 (48) Conditioning regimenb .83 MAC 447 (82) 281 (82) 63 (78) 16 (84) 87 (85) RIC 97 (18) 64 (18) 15 (22) 3 (16) 15 (15) Days between transplant and PIV infection .036 ≤30 119 (22) 66 (19) 22 (28) 3 (16) 28 (28) 31–365 349 (64) 238 (69) 40 (51) 14 (74) 57 (56) >365 76 (14) 41 (12) 16 (21) 2 (10) 17 (17) PIV typec PIV-1 52 (10) 26 (8) 16 (21) 1 (5) 9 (9) PIV-2 30 (5) 24 (7) 3 (4) 0 (0) 3 (3) PIV-3 434 (80) 275 (80) 53 (68) 17 (90) 89 (87) PIV-4 22 (4) 16 (5) 5 (6) 1 (5) 0 (0) Unclassified 6 (1) 4 (1) 1 (1) 0 (0) 1 (1) Quantitative viral load, median (range) 5.0 × 106 (1.0 × 102–1.1 × 109) 4.8 × 105 (1.0 × 102–3.3 × 108) 7.4 × 106 (2.7 × 103–1.1 × 109) .51 Copathogend <.001 No 457 (84) 313 (91) 61 (78) 16 (84) 67 (66) Yes 87 (16) 32 (9) 17 (22) 3 (16) 35 (34) Oxygen at diagnosis <.001 No 472 (87) 342 (99) 63 (81) 13 (68) 54 (53) Yes 72 (13) 3 (1) 15 (19) 6 (32) 48 (47) White blood cell count <.001 >1000 cells/µL 447 (85) 299 (90) 59 (76) 19 (100) 70 (69) ≤1000 cells/µL 82 (15) 32 (10) 19 (24) 0 (0) 31 (31) Lymphocyte count .026 >300 cells/µL 332 (64) 224 (68) 44 (58) 11 (58) 53 (53) ≤300 cells/µL 191 (36) 104 (32) 32 (42) 8 (42) 47 (47) Neutrophil count <.001 >1000 × 106 cells/L 407 (77) 270 (82) 56 (73) 18 (95) 63 (63) ≤1000 × 106 cells/L 120 (23) 61 (18) 21 (27) 1 (5) 37 (37) Monocyte count <.001 >100 × 106 cells/L 381 (73) 266 (81) 53 (70) 11 (58) 51 (51) ≤100 × 106 cells/L 141 (27) 61 (19) 23 (30) 8 (42) 49 (49) Steroid dose before diagnosisc No 242 (47) 160 (50) 40 (53) 5 (26) 37 (38) <1 mg/kg 133 (26) 78 (24) 25 (34) 4 (21) 26 (26) 1–2 mg/kg 124 (24) 77 (24) 9 (12) 7 (37) 31 (32) >2 mg/kg 13 (3) 5 (2) 1 (1) 3 (16) 4 (4) Ribavirin usee <.001 No 483 (89) 336 (97) 73 (94) 12 (63) 62 (61) Yes

(27) 61 (19) 23 (30) 8 (42) 49 (49) Steroid dose before diagnosisc No 242 (47) 160 (50) 40 (53) 5 (26) 37 (38) <1 mg/kg 133 (26) 78 (24) 25 (34) 4 (21) 26 (26) 1–2 mg/kg 124 (24) 77 (24) 9 (12) 7 (37) 31 (32) >2 mg/kg 13 (3) 5 (2) 1 (1) 3 (16) 4 (4) Ribavirin usee <.001 No 483 (89) 336 (97) 73 (94) 12 (63) 62 (61) Yes 61 (11) 9 (3) 5 (6) 7 (37) 40 (39) IVIG use <.001 No 373 (69) 251 (73) 58 (74) 11 (58) 53 (52) Low-dosef 138 (26) 89 (26) 17 (22) 3 (16) 29 (29) High-dose 31 (6) 4 (1) 3 (4) 5 (26) 19 (19) All values are indicated as No. (%). Additional baseline parameters (disease risk at transplant, graft-vs-host disease prophylaxis, recipient cytomegalovirus serostatus, percentage of forced expiratory volume in 1 second/forced vital capacity before PIV infection, and percentage of predicted total lung capacity before PIV infection) were examined and did not show statistical differences between groups. Abbreviations: IVIG, intravenous immunoglobulin; LRTD, lower respiratory tract disease; MAC, myeloablative conditioning; PIV, parainfluenza virus; RIC, reduced-intensity conditioning; URTI, upper respiratory tract infection. a Five patients with multiple transplants had their reference transplant after 2011. b The MAC and RIC regimens were defined as previously described [14]. c Exact P value could not be calculated. d A copathogen was defined as a significant pathogen detected in concurrent nasopharyngeal, bronchoalveolar lavage, or lung biopsy samples, or in a blood sample obtained within 2 days of diagnosis of PIV infection.

b The MAC and RIC regimens were defined as previously described [14]. c Exact P value could not be calculated. d A copathogen was defined as a significant pathogen detected in concurrent nasopharyngeal, bronchoalveolar lavage, or lung biopsy samples, or in a blood sample obtained within 2 days of diagnosis of PIV infection. e Ribavirin was administered as follows: aerosolized in 45 patients, systemic in 10, both in 1, and unknown in 5. f To maintain levels of >400 mg/dL, as needed. Mortality in Each Disease Category The probabilities of overall survival and mortality from respiratory failure at 90 days following PIV diagnosis in patients with URTI or LRTD are shown in Figure 1A and 1B (overall survival: 91% in URTI, 62% in LRTD; mortality from respiratory failure: 2.5% in URTI, 28% in LRTD) (P < .001 for both comparisons). Next, we analyzed the probabilities of overall survival and death caused by respiratory failure among LRTD cases, comparing the 3 disease categories: possible, probable, and proven (Figure 1C and 1D). The probabilities of 90-day survival after PIV LRTD were 87%, 58%, and 45% in possible, probable and proven cases, respectively (P < .001; Figure 1C). The presence of LRTD symptoms in possible or proven cases resulted in worse survival than when no LRTD symptoms were present (Supplementary Figure 1A and 1B). Mortality rates in each group are shown in Table 2. Table 2. Mortality Rates in Each Group

ssible, probable and proven cases, respectively (P < .001; Figure 1C). The presence of LRTD symptoms in possible or proven cases resulted in worse survival than when no LRTD symptoms were present (Supplementary Figure 1A and 1B). Mortality rates in each group are shown in Table 2. Table 2. Mortality Rates in Each Group Category URTI (n = 345) LRTD Possible Probable Proven (n = 22) (n = 56) (n = 19) (n = 14) (n = 88) Positive nasopharyngeal test Yes Yes Yes a a a Positive BAL/biopsy NT/No NT/No NT/No Yes Yes Yes Positive radiography No Yes Yes No Yes Yes Lower respiratory tract symptoms b No Yes Yes No Yes Overall survival by day 90 (%) 91 100 82 58 57 43 Respiratory death by day 90 (%) 3  0  9 11 29 47 Abbreviations: BAL, bronchoalveolar lavage; LRTD, lower respiratory tract disease; NT, not tested; URTI, upper respiratory tract infection. a Any result (yes or no) or nonnasopharyngeal test. b Any result (yes or no).

Category URTI (n = 345) LRTD Possible Probable Proven (n = 22) (n = 56) (n = 19) (n = 14) (n = 88) Positive nasopharyngeal test Yes Yes Yes a a a Positive BAL/biopsy NT/No NT/No NT/No Yes Yes Yes Positive radiography No Yes Yes No Yes Yes Lower respiratory tract symptoms b No Yes Yes No Yes Overall survival by day 90 (%) 91 100 82 58 57 43 Respiratory death by day 90 (%) 3  0  9 11 29 47 Abbreviations: BAL, bronchoalveolar lavage; LRTD, lower respiratory tract disease; NT, not tested; URTI, upper respiratory tract infection. a Any result (yes or no) or nonnasopharyngeal test. b Any result (yes or no). Figure 1. Probability of overall survival and death caused by respiratory failure. A, Kaplan-Meier estimate of overall survival according to classification of parainfluenza virus (PIV) infection in hematopoietic cell transplant (HCT) recipients (P < .001). B, Cumulative incidence of death caused by respiratory failure according to classification of PIV infection in HCT recipients (P < .001). C, Kaplan-Meier estimate of overall survival according to lower respiratory tract disease (LRTD) classification (P < .001). D, Cumulative incidence of death caused by respiratory failure according to LRTD classification (P < .001). Abbreviations: LRTD, lower respiratory tract disease; PIV, parainfluenza virus; URTI, upper respiratory tract infection.

all survival according to lower respiratory tract disease (LRTD) classification (P < .001). D, Cumulative incidence of death caused by respiratory failure according to LRTD classification (P < .001). Abbreviations: LRTD, lower respiratory tract disease; PIV, parainfluenza virus; URTI, upper respiratory tract infection. Univariable analysis of risk factors for overall mortality demonstrated that probable and proven LRTD was significantly associated with higher mortality compared with URTI (hazard ratio [HR], 5.87; 95% confidence interval [CI], 2.7–12.8; P < .001 in probable cases, and HR, 9.23; 95% CI, 5.9–14.3; P < .001 in proven cases), whereas possible LRTD was not (HR, 1.49; 95% CI, .7–3.0; P = .27). Similar results were obtained in the univariable analysis of risk factors for mortality from respiratory failure (HR, 2.56; 95% CI, .9–7.6; P = .09 in possible cases; HR, 4.95; 95% CI, 1.1–22.9; P = .041 in probable cases; HR, 24.4; 95% CI, 11.9–50.0; P < .001 in proven cases). Because only 10 patients with possible LRTD and 8 with probable LRTD died by day 90 after diagnosis, it was not feasible to evaluate the impact of each disease category on mortality in multivariable analyses.

HR, 4.95; 95% CI, 1.1–22.9; P = .041 in probable cases; HR, 24.4; 95% CI, 11.9–50.0; P < .001 in proven cases). Because only 10 patients with possible LRTD and 8 with probable LRTD died by day 90 after diagnosis, it was not feasible to evaluate the impact of each disease category on mortality in multivariable analyses. Hypoxemia and Hospitalization in Each LRTD Category The probabilities of requiring mechanical ventilation by 28 days after PIV LRTD were 10%, 18%, and 41% in possible, probable, and proven cases, respectively (P < .001; Figure 2). Oxygen-free days in each group were shown in Table 3. Days alive without hospitalization by 28 days after PIV LRTD were longer in order of possible, probable, and proven cases (mean: 19 [SD, 10] days in possible cases, 15 [SD, 11] days in probable cases, 8.0 [SD, 9] days in proven cases, respectively, P < .001). Table 3. Oxygen-Free Days According to Parainfluenza Virus Lower Respiratory Tract Disease Category Outcome Possible Probable Proven P Value Any oxygen-free days By day 14 after PIV LRTD 12 (4) 10 (6) 6 (6) <.0001 By day 28 after PIV LRTD 24 (8) 20 (11) 12 (11) <.0001 >2 L/min oxygen-free days By day 14 after PIV LRTD 13 (3) 11 (4) 9 (4) <.0001 By day 28 after PIV LRTD 26 (4) 22 (9) 18 (9) <.0001 All values are presented as mean (standard deviation). Abbreviations: LRTD, lower respiratory tract disease; PIV, parainfluenza virus.

Outcome Possible Probable Proven P Value Any oxygen-free days By day 14 after PIV LRTD 12 (4) 10 (6) 6 (6) <.0001 By day 28 after PIV LRTD 24 (8) 20 (11) 12 (11) <.0001 >2 L/min oxygen-free days By day 14 after PIV LRTD 13 (3) 11 (4) 9 (4) <.0001 By day 28 after PIV LRTD 26 (4) 22 (9) 18 (9) <.0001 All values are presented as mean (standard deviation). Abbreviations: LRTD, lower respiratory tract disease; PIV, parainfluenza virus. Figure 2. Probability of mechanical ventilation after parainfluenza virus lower respiratory tract disease (LRTD). Cumulative incidence of requirements of mechanical ventilation according to LRTD classification (P < .001).

Outcome Possible Probable Proven P Value Any oxygen-free days By day 14 after PIV LRTD 12 (4) 10 (6) 6 (6) <.0001 By day 28 after PIV LRTD 24 (8) 20 (11) 12 (11) <.0001 >2 L/min oxygen-free days By day 14 after PIV LRTD 13 (3) 11 (4) 9 (4) <.0001 By day 28 after PIV LRTD 26 (4) 22 (9) 18 (9) <.0001 All values are presented as mean (standard deviation). Abbreviations: LRTD, lower respiratory tract disease; PIV, parainfluenza virus. Figure 2. Probability of mechanical ventilation after parainfluenza virus lower respiratory tract disease (LRTD). Cumulative incidence of requirements of mechanical ventilation according to LRTD classification (P < .001). Risk Factors for Mortality From All Causes or Respiratory Failure in Proven/Probable LRTD Cases We focused on proven and probable LRTD cases to analyze the risk factors for mortality. A univariable analysis of risk factors for overall mortality showed that days between transplant and PIV infection, oxygen use, cell counts, and steroid dose >2 mg/kg/day before or after diagnosis were significant risk factors (Table 4). The same risk factors as well as PIV type were also significant for mortality from respiratory failure (Table 4). Among white blood cell populations, a monocyte count <100 cells/µL was identified as the most important factor for mortality in the multivariable analysis (Table 5). These results were similar to those obtained from patients with only proven LRTD, except that low monocyte counts become a significant risk factor for overall mortality in a multivariable analysis (adjusted HR [aHR], 2.01; 95% CI, 1.1–3.8; P = .029). Lower monocyte counts had a statistically significant effect on overall survival and risk of pulmonary death when analyzed in combination with oxygen requirements at diagnosis (Figure 3A and 3B). Table 4. Univariable Analysis of Risk Factors for Mortality From All Causes or Respiratory Failure by Day 90 After Parainfluenza Infection Among Proven/Probable Cases (n = 121)

t effect on overall survival and risk of pulmonary death when analyzed in combination with oxygen requirements at diagnosis (Figure 3A and 3B). Table 4. Univariable Analysis of Risk Factors for Mortality From All Causes or Respiratory Failure by Day 90 After Parainfluenza Infection Among Proven/Probable Cases (n = 121) Variables Overall Mortality Mortality From Respiratory Failure HR (95% CI) P Value HR (95% CI) P Value Disease category Probable 1.00 1.00 Proven without symptoms 1.07 (.4–3.1) .91 2.82 (.5–15.4) .23 Proven with symptoms 1.61 (.8–3.4) .21 5.19 (1.3–21.5) .023 Presence of LRTD symptoms No 1.00 1.00 Yes 1.39 (.6–3.2) .44 1.54 (.6–4.3) .41 Transplant year 1990–2000 1.00 1.00 2001–2011 0.62 (.4–1.0) .07 0.70 (.4–1.3) .23 Conditioning regimen MAC 1.00 1.00 RIC 0.50 (.2–1.2) .11 0.58 (.2–1.5) .26 Days between transplant and PIV infection ≤365 1.00 1.00 >365 2.95 (1.2–7.4) .020 5.46 (1.3–22.5) .019 PIV type PIV-1, -2, -4 1.00 1.00 PIV-3 0.63 (.3–1.2) .18 0.44 (.2–.9) .021 Quantitative viral load (log10)a 1.01 (.9–1.1) .84 1.07 (.9–1.2) .32 Diagnostic methods Conventional methods 1.00 1.00 PCR alone 0.10 (.3–1.1) .10 0.59 (.3–1.3) .18 Copathogen No 1.00 1.00 Yes 1.56 (1.0–2.6) .08 1.76 (1.0–3.2) .06 Oxygen at diagnosis No 1.00 1.00 Yes 2.49 (1.4–4.5) .002 4.25 (1.9–9.5) <.001 Oxygen after diagnosisb ≤2 L 1.00 1.00 >2 L 2.09 (1.8–2.4) <.001 2.62 (.9–7.4) .07 Mechanical ventilation 2.59 (2.2–3.1) <.001 17.93 (8.7–37.0) <.001 White blood cell count >1000 cells/µL 1.00 1.00 ≤1000 cells/µL 2.27 (1.4–3.8) .002 2.80 (1.6–5.1) <.001 Lymphocyte count >300 cells/µL 1.00 1.00 ≤300 cells/µL 1.73 (1.1–2.9) .032 1.82 (1.0–3.3) .045 Neutrophil count >1000 cells/µL 1.00 1.00 ≤1000 cells/µL 2.35 (1.4–3.9) <.001 2.71 (1.5–4.9) <.001 Monocyte count >100 cells/µL 1.00 1.00 ≤100 cells/µL 2.30 (1.4–3.8) .001 2.96 (1.6–5.4) <.001 Steroid dose before diagnosis No 1.00 1.00 <1 mg/kg 1.05 (.6–2.0) .88 1.06 (.5–2.1) .88 1–2 mg/kg 0.70 (.4–1.3) .28 .62 (.3–1.3) .22 >2 mg/kg 3.23 (1.5–7.2) .004 1.41 (.4–4.8) .58 Steroid dose after diagnosisb No 1.00 1.00 <1 mg/kg 0.64 (.3–1.4) .24 0.65 (.3–1.5) .31 1–2 mg/kg 1.09 (.6–2.1) .80 1.04 (.5–2.3) .92 >2 mg/kg 4.39 (1.9–9.9) <.001 4.12 (1.8–9.7) .001 Ribavirin useb No 1.00 1.00 Yes 0.68 (.4–1.1) .15 0.95 (.5–1.7) .88 IVIG use No/low-dose 1.00 1.00 High-dose 1.17 (.7–2.1) .60 1.29 (.7–2.5) .44 All variables in Table 1 were used for the univariable analysis. Only variables with P < .1 in any analysis are shown in this table.

g 4.39 (1.9–9.9) <.001 4.12 (1.8–9.7) .001 Ribavirin useb No 1.00 1.00 Yes 0.68 (.4–1.1) .15 0.95 (.5–1.7) .88 IVIG use No/low-dose 1.00 1.00 High-dose 1.17 (.7–2.1) .60 1.29 (.7–2.5) .44 All variables in Table 1 were used for the univariable analysis. Only variables with P < .1 in any analysis are shown in this table. The following parameters were also shown regardless of P values: presence of symptoms, quantitative viral load (log10), ribavirin use, and IVIG use. Abbreviations: CI, confidence interval; HR, hazard ratio; IVIG, intravenous immunoglobulin; LRTD, lower respiratory tract disease; MAC, myeloablative conditioning; PCR, polymerase chain reaction; PIV, parainfluenza virus; RIC, reduced-intensity conditioning. a Viral titer was analyzed as a continuous variable. b These variables are analyzed as time dependent. Table 5. Multivariable Analysis of Risk Factors for Mortality From All Causes or Respiratory Failure by Day 90 After Diagnosis in Proven/Probable Cases (n = 121)

Abbreviations: CI, confidence interval; HR, hazard ratio; IVIG, intravenous immunoglobulin; LRTD, lower respiratory tract disease; MAC, myeloablative conditioning; PCR, polymerase chain reaction; PIV, parainfluenza virus; RIC, reduced-intensity conditioning. a Viral titer was analyzed as a continuous variable. b These variables are analyzed as time dependent. Table 5. Multivariable Analysis of Risk Factors for Mortality From All Causes or Respiratory Failure by Day 90 After Diagnosis in Proven/Probable Cases (n = 121) Final Model Steroid Dose Ribavirin IVIG HR (95% CI) P Value HR (95% CI) P Value HR (95% CI) P Value HR (95% CI) P Value Overall mortality Oxygen at diagnosis (yes vs no) 2.44 (1.3–4.4) .003 2.19 (1.2–4.1) .013 2.28 (1.3–4.2) .007 2.63 (1.5–4.8) .001 Days between transplant and PIV infection (≤365 vs >365) 2.26 (.9–5.8) .09 2.46 (1.0–6.3) .06 2.34 (.9–6.1) .08 Transplant year (2001–2011 vs 1990–2000) 0.72 (.4–1.2) .22 0.68 (.4–1.2) .16 0.77 (.4–1.3) .34 Monocyte count (<100 vs ≥100 cells/µL) 1.63 (.9–2.9) .09 2.33 (1.4–4.0) .002 1.97 (1.2–3.3) .010 Neutrophil count (<1000 vs ≥ 1000 cells/µL) 1.58 (.9–2.8) .12 Steroid dose after diagnosis (>2 vs ≤2 mg/kg)a 3.80 (2.0–7.4) <.001 Ribavirin use (yes vs no)a 0.51 (.3–.9) .021 IVIG use (high-dose vs no/low-dose) 0.99 (.5–1.8) .97 Mortality from respiratory failure Oxygen at diagnosis (yes vs no) 3.96 (1.7–9.1) .001 3.59 (1.6–8.2) .002 4.06 (1.8–9.3) <.001 4.26 (1.9–9.7) <.001 Days between transplant and PIV infection (≤365 vs >365) 4.14 (1.0–17.4) .052 4.25 (1.0–17.9) .049 4.13 (1.0–17.5) .054 PIV type (PIV-3 vs PIV-1, -2, -4) 0.54 (.3–1.1) .10 0.53 (.3–1.1) .09 0.54 (.3–1.2) .11 Monocyte counts (<100 vs ≥100 cells/µL) 2.07 (1.0–4.2) .041 2.48 (1.3–4.7) .006 2.34 (1.3–4.4) .008 Neutrophil counts (<1000 vs ≥1000 cells/µL) 1.36 (.7–2.7) .38 Steroid dose after diagnosis (>2 vs ≤2 mg/kg)a 3.27 (1.5–6.9) .008 Ribavirin use (yes vs no)a 0.87 (.5–1.6) .66 IVIG use (high-dose vs no/low-dose) 1.09 (.5–2.2) .81 Abbreviations: CI, confidence interval; HR, hazard ratio; IVIG, intravenous immunoglobulin; PIV, parainfluenza virus.

vs ≥1000 cells/µL) 1.36 (.7–2.7) .38 Steroid dose after diagnosis (>2 vs ≤2 mg/kg)a 3.27 (1.5–6.9) .008 Ribavirin use (yes vs no)a 0.87 (.5–1.6) .66 IVIG use (high-dose vs no/low-dose) 1.09 (.5–2.2) .81 Abbreviations: CI, confidence interval; HR, hazard ratio; IVIG, intravenous immunoglobulin; PIV, parainfluenza virus. a These variables are analyzed as time dependent. Figure 3. Probability of overall survival and death caused by respiratory failure according to monocyte count and oxygen requirement in proven/probable lower respiratory tract disease (LRTD) cases. A, Kaplan-Meier estimate of overall survival by monocyte count and oxygen requirement in proven/probable cases (P < .001). B, Cumulative incidence of death caused by respiratory failure according to LRTD classification in proven/probable cases (P < .001). A statistically significant difference in mortality was seen between probable and proven LRTD with symptoms (Table 4). The presence of clinical LRTD symptoms tended to be associated with worse outcome, although the effect did not reach statistical significance (Table 4).

Figure 3. Probability of overall survival and death caused by respiratory failure according to monocyte count and oxygen requirement in proven/probable lower respiratory tract disease (LRTD) cases. A, Kaplan-Meier estimate of overall survival by monocyte count and oxygen requirement in proven/probable cases (P < .001). B, Cumulative incidence of death caused by respiratory failure according to LRTD classification in proven/probable cases (P < .001). A statistically significant difference in mortality was seen between probable and proven LRTD with symptoms (Table 4). The presence of clinical LRTD symptoms tended to be associated with worse outcome, although the effect did not reach statistical significance (Table 4). Association Between Steroid Dose or Antiviral Treatment and Outcomes in Proven/Probable LRTD Cases In the univariable analysis, steroid doses >2 mg/kg/day both before and after diagnosis of PIV LRTD were significantly associated with increased mortality, whereas steroid doses <2 mg/kg/day did not have any dose-dependent effect on mortality (Table 4). Because only a small number of the patients receiving steroid doses >2 mg/kg/day died by day 90 after diagnosis, the effect of steroid dose was evaluated while adjusting only for oxygen requirement at diagnosis, which is the most important risk factor for mortality. In this adjusted model, steroid dose >2 mg/kg/day after diagnosis remained significantly associated with mortality (Table 5).

g/kg/day died by day 90 after diagnosis, the effect of steroid dose was evaluated while adjusting only for oxygen requirement at diagnosis, which is the most important risk factor for mortality. In this adjusted model, steroid dose >2 mg/kg/day after diagnosis remained significantly associated with mortality (Table 5). The use of ribavirin was significantly associated with reduced overall mortality, but not with mortality from respiratory failure in multivariable analyses (Table 5). Among the patients with proven LRTD, however, ribavirin did not affect the overall or respiratory failure–related mortality (aHR, 0.64; 95% CI, .4–1.2; P = .13 in overall mortality, and aHR, 0.96; 95% CI, .5–1.8; P = .91 in respiratory mortality). High-dose intravenous immunoglobulin (IVIG) also had no effect on mortality after PIV LRTD (Table 5). Association Between Diagnostic Method or BAL Viral Load and Outcomes in Proven/Probable LRTD Cases Among 121 proven/probable LRTD cases, 31 (26%) patients were diagnosed using PCR alone. Although patients diagnosed with PCR alone had lower viral load (P < .001; Supplementary Figure 2A), there was no statistically significant survival advantage (Table 4).

d or BAL Viral Load and Outcomes in Proven/Probable LRTD Cases Among 121 proven/probable LRTD cases, 31 (26%) patients were diagnosed using PCR alone. Although patients diagnosed with PCR alone had lower viral load (P < .001; Supplementary Figure 2A), there was no statistically significant survival advantage (Table 4). In proven/probable LRTD cases, 48 stored BAL samples were available and tested by quantitative RT-PCR. There was no statistically significant difference between median viral load in subjects with proven and probable LRTD (Table 1, Supplementary Figure 2B). An association between viral load and mortality after PIV LRTD was not observed even after adjusting for oxygen use at diagnosis (HR, 1.03; 95% CI, .93–1.13; P = .60 for overall mortality, and HR, 1.09; 95% CI, .96–1.23; P = .19 for mortality from respiratory failure; Table 4 and data not shown). DISCUSSION This study demonstrates that the clinical outcome of patients with PIV detection in the lung (which we termed “proven” or “probable” LRTD) is significantly worse than that of patients with PIV detection only in upper respiratory tract samples (termed “possible” LRTD) (Figure 1C and 1D). Outcomes in patients with possible LRTD were similar to that of URTI. Our study also defined important clinical risk factors associated with poor outcomes of PIV LRTD, including oxygen requirement at diagnosis, low monocyte count, and steroid use >2 mg/kg/day.

tract samples (termed “possible” LRTD) (Figure 1C and 1D). Outcomes in patients with possible LRTD were similar to that of URTI. Our study also defined important clinical risk factors associated with poor outcomes of PIV LRTD, including oxygen requirement at diagnosis, low monocyte count, and steroid use >2 mg/kg/day. Numerous studies of respiratory viruses in immunocompromised hosts have examined LRTD either as an endpoint or a risk factor for mortality. However, previously published results for both the incidence and risk factors for progression and outcome vary widely [3–6, 11, 12]. Data presented here suggest that the variance in definitions of LRTD is one important reason for the observed differences. To address this issue, we conducted the present study to define the optimized diagnostic criteria that correlate with outcome. Based on the site of virologic detection and clinical manifestations, we classified LRTD into 3 groups (possible, probable, and proven). The recently published Fourth European Conference on Infections in Leukaemia (ECIL-4) guidelines also proposed a classification based on levels of certainty (possible, probable, confirmed) [15]. However, the ECIL-4 classification is for all respiratory viruses, applies to both UTRI and LRTD, and uses symptoms, exposure, and any viral detection (independent of site) as classifiers. In contrast, our classification is strictly among LRTD cases that were all virologically confirmed [15]. In our paper, we specifically focused on the “possible” category, which includes patients with nasopharyngeal tests demonstrating PIV infection and abnormal chest radiography, but without confirmation of virus in the lower respiratory tract. Numerous publications as well as society guidelines have generally accepted that these patients have LRTD [5–8, 11, 15–19]. Our data demonstrate that the mortality in patients with possible LRTD is significantly better than that in patients with probable and proven disease (Figure 1C and 1D) and similar to that of patients with URTI (Table 2). The effect was persistent throughout the study period (data not shown). Interestingly, possible LRTD patients without symptoms tended to have even better outcome than URTI (Supplementary Figure 1). Nevertheless, many studies include this category as LRTD [6–8, 16, 17]. Therefore, we suggest that this group we defined as “possible” LRTD should be separated from the patient group with viral detection in the lung (proven and probable LRTD cases).

tended to have even better outcome than URTI (Supplementary Figure 1). Nevertheless, many studies include this category as LRTD [6–8, 16, 17]. Therefore, we suggest that this group we defined as “possible” LRTD should be separated from the patient group with viral detection in the lung (proven and probable LRTD cases). We developed our definitions a priori based on clinical practices across centers and the spectrum of definitions used in the literature. The “possible” and “proven” categories are by far the most common categories documented, both in our series and elsewhere. Bronchoscopy based on PIV detection in the upper respiratory tract and the presence of LRTD symptoms or significant signs of obstruction by spirometry (“probable” category) was performed during the 1990s at our center, but it has become very uncommon in recent years. Nevertheless, these patients are informative as their outcome appears to be similar to that seen in proven cases.

tract and the presence of LRTD symptoms or significant signs of obstruction by spirometry (“probable” category) was performed during the 1990s at our center, but it has become very uncommon in recent years. Nevertheless, these patients are informative as their outcome appears to be similar to that seen in proven cases. Although our study focused on patients infected with PIV, we hypothesize that the outcome differences seen among the “possible” disease category may also exist for other respiratory viruses. Indeed, most studies and guidelines of non-PIV respiratory viruses include “possible” cases in their definitions of LRTD [7, 15–19]. Additional studies are needed to extend these results to other respiratory virus infections. More than a decade ago, the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) established the standard definitions for invasive fungal disease [20, 21]. These guidelines are widely accepted [22, 23], and resulted in improved reproducibility of study results between centers. These guidelines have greatly aided consensus endpoint definitions for clinical trials in invasive fungal disease. As new drugs for some respiratory viruses including PIV (eg, DAS181) are advancing through clinical development [24, 25], careful consideration of LRTD definitions that correlate with outcome is as important as clinical trial endpoints, entry criteria, and stratification variables.

al trials in invasive fungal disease. As new drugs for some respiratory viruses including PIV (eg, DAS181) are advancing through clinical development [24, 25], careful consideration of LRTD definitions that correlate with outcome is as important as clinical trial endpoints, entry criteria, and stratification variables. Another novel observation of this study is that the monocyte count at diagnosis of LRTD appears to be critical to survival after PIV LRTD. Previous studies of respiratory viral infections in immunocompromised hosts suggest that lymphocytes or neutrophils are important cell components for mortality [5, 26]. In our study, however, lymphocyte counts appeared to be less important than monocyte or neutrophil counts, and monocytes seemed to comprise an important cell component (Tables 4 and 5). Indeed, patients who were both monocytopenic and required oxygen at diagnosis of LRTD had a particularly high mortality (Figure 3). Monocytes have been known to be important to protect from viral infections [27–31]. Neutrophils may also have an important role at the time of pneumonia, and these 2 cells interact with each other [32–34]. These data from basic research support our finding that monocytes and neutrophils are important for outcomes after LRTD.

tes have been known to be important to protect from viral infections [27–31]. Neutrophils may also have an important role at the time of pneumonia, and these 2 cells interact with each other [32–34]. These data from basic research support our finding that monocytes and neutrophils are important for outcomes after LRTD. This is the first outcome study of PIV that used multivariable modeling for evaluating treatment strategies. Previous studies suggested that steroid use was associated with high mortality [5, 6, 12]. Our results showed that steroid use of ≤2 mg/kg/day after PIV LRTD was not associated with mortality. Based on our results, high-dose steroids for acute lung injury are not recommended, and a rapid taper of steroids for graft-vs-host disease may not be required when lower doses are used.

with high mortality [5, 6, 12]. Our results showed that steroid use of ≤2 mg/kg/day after PIV LRTD was not associated with mortality. Based on our results, high-dose steroids for acute lung injury are not recommended, and a rapid taper of steroids for graft-vs-host disease may not be required when lower doses are used. Because no antiviral drug is currently approved for the treatment of PIV disease, ribavirin and IVIG are often used off-label in PIV LRTD cases. Previous reports showed conflicting results ranging from no apparent efficacy on mortality after PIV LRTD [3–6, 12, 35], to moderate efficacy [36–39]. Most studies had a small sample size, and no statistical adjustments were made for disease severity. Thus, the effect of ribavirin is still controversial [40]. Our results show conflicting results for ribavirin treatment. Although the overall mortality model suggested a survival benefit among patients with proven and probable LRTD (Table 5), no significant effect was seen for death due to respiratory failure and for patients with proven LRTD. Some important variables such as high-dose steroids or the effect of combination therapy with IVIG could not be adequately assessed due to the small number of the patients with proven/probable LRTD who received these treatment combinations.

as seen for death due to respiratory failure and for patients with proven LRTD. Some important variables such as high-dose steroids or the effect of combination therapy with IVIG could not be adequately assessed due to the small number of the patients with proven/probable LRTD who received these treatment combinations. Recently, PCR testing has become more widely utilized as a diagnostic method and the frequency of diagnoses of viral infections has been increasing [13]. However, the association of quantitative RNA viral load on mortality after PIV LRTD has not been well evaluated. In our study, BAL samples that were also positive by conventional methods generally had a higher viral load (Supplementary Figure 2A), consistent with previous reports in nasal wash samples [13]. The results also show that the “proven” and “probable” disease categories do not differ with regard to viral load and that the presence of clinical LRTD symptoms was not associated with higher viral load (Supplementary Figure 2B). These findings provide support for combining both groups in outcome analyses, as proposed in this study. Our results did not identify RNA viral load in the BAL as a significant factor for mortality, although the number of subjects with available viral load was relatively small.

her viral load (Supplementary Figure 2B). These findings provide support for combining both groups in outcome analyses, as proposed in this study. Our results did not identify RNA viral load in the BAL as a significant factor for mortality, although the number of subjects with available viral load was relatively small. This study has several limitations. Although this study is the largest cohort of confirmed PIV infections in HCT recipients (which includes our previously reported cases [4] as well as subsequent cases through 2011), the sample size was still not sufficient to perform multivariable analyses to evaluate the impact of LRTD disease categories or the effect of combination of several drugs on mortality. Because of the retrospective nature of the study, BAL samples were available in only half of the patients for viral load quantification, which limited the multivariable modeling. The use of bronchoscopy for the workup of LRTD is largely based on protocols at our center, with almost all patients undergoing this procedure when radiographic signs or LRTD symptoms occur. Nevertheless, the final decision to perform bronchoscopy rests with the attending physician and some patients may not have undergone this procedure. Therefore, some probable or proven cases may have been missed, but this is unlikely to affect the major results of this study.

n radiographic signs or LRTD symptoms occur. Nevertheless, the final decision to perform bronchoscopy rests with the attending physician and some patients may not have undergone this procedure. Therefore, some probable or proven cases may have been missed, but this is unlikely to affect the major results of this study. In conclusion, our data demonstrate that patients with PIV detection in the lungs (proven/probable LRTD) had worse outcomes compared with those with PIV detection in nasopharyngeal samples alone (possible LRTD). The outcome of possible LRTD was comparable to that of URTI. This LRTD definition could be useful for future outcome studies independent of virus types, and studies are needed to validate the results for other viruses and immunocompromised host settings. Consensus definitions in accordance with outcomes should be developed for respiratory viral disease. 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 Zachary Stednick for database services, Terry Stevens-Ayers and Elsa Garnace for laboratory assistance, and Louise E. Kimball, Farah T. Sahoo, and Sonia Goyal for data collection.

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 Zachary Stednick for database services, Terry Stevens-Ayers and Elsa Garnace for laboratory assistance, and Louise E. Kimball, Farah T. Sahoo, and Sonia Goyal for data collection. Financial support. This work was supported by the National Institutes of Health (grant numbers CA18029, CA15704, HL081595, HL93294, K23HL091059, and L40AI071572) and Ansun Biopharma, Inc. S. S. is a recipient of a fellowship from the Joel Meyers Memorial Fund. A. P. C. also received support from the Seattle Children's Center for Clinical and Translational Research and Clinical and Translational Science Award (ULI RR025014). Potential conflicts of interest. M. B. has received research support from Ansun Biopharma, Inc. 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.

Human metapneumovirus (HMPV) is a newly recognized respiratory pathogen first identified in 2001 [1] and is structurally similar to respiratory syncytial virus (RSV) [2]. Serological epidemiology shows a 90% exposure by adulthood [3]. HMPV infection occurs in approximately 5% of hematopoietic cell transplantation (HCT) recipients [4, 5], and can cause severe or sometimes fatal lower respiratory tract disease (LRD) [6–9]. The mortality rate after HMPV LRD in patients with hematological malignancy or HCT recipients ranges from 10% to 40% [9–12]. To date, most studies in HCT recipients are very small and have focused on the LRD characteristics and outcome [7, 8, 10, 12]. Minimal data exist on the timing of, proportion of, and risk factors for progression from upper respiratory tract infection (URI) to LRD. These data are critical for the design of clinical trials that aim to prevent progression from URI to LRD. The purpose of this study was to examine the seasonality of HMPV infections in HCT recipients and to identify patient, transplantation, and viral risk factors for progression to LRD.

tract infection (URI) to LRD. These data are critical for the design of clinical trials that aim to prevent progression from URI to LRD. The purpose of this study was to examine the seasonality of HMPV infections in HCT recipients and to identify patient, transplantation, and viral risk factors for progression to LRD. METHODS Study Design This retrospective cohort study includes patients who underwent transplantation between March 2004 and April 2014 at the Fred Hutchinson Cancer Research Center (FHCRC) and had documented HMPV infection during pretransplant conditioning or after transplantation. Only an individual's first episode of HMPV infection was analyzed. Patients' demographic data and transplantation information were retrieved from the FHCRC database, and other data related to the clinical course of HMPV infections were collected by medical record reviews. The data about HMPV infection in the community were based on the University of Washington Molecular Virology Laboratory data, which include data from Seattle Children's Hospital and from samples that were sent to the laboratory from regional providers (http://depts.washington.edu/rspvirus/respiratory.htm). The study was approved by the FHCRC Institutional Review Board.

based on the University of Washington Molecular Virology Laboratory data, which include data from Seattle Children's Hospital and from samples that were sent to the laboratory from regional providers (http://depts.washington.edu/rspvirus/respiratory.htm). The study was approved by the FHCRC Institutional Review Board. Laboratory Testing Nasopharyngeal samples were collected when HCT recipients had URI symptoms. We also collected a bronchoalveolar lavage (BAL) sample when patients had lower respiratory tract symptoms and a radiographic abnormality. Follow-up nasopharyngeal samples were obtained in a subset of patients. We initially applied direct fluorescent antibody testing and individual reverse transcription polymerase chain reaction (RT-PCR) assay to detect HMPV and started multiplex PCR testing in 2008 [13, 14]. Samples were considered positive if the PCR amplification plot crossed the threshold at <40 cycles. Viral load was determined by quantitative RT-PCR using nasopharyngeal samples at diagnosis of URI [15].

cription polymerase chain reaction (RT-PCR) assay to detect HMPV and started multiplex PCR testing in 2008 [13, 14]. Samples were considered positive if the PCR amplification plot crossed the threshold at <40 cycles. Viral load was determined by quantitative RT-PCR using nasopharyngeal samples at diagnosis of URI [15]. Definitions URI was defined as HMPV detection in a nasopharyngeal sample with URI symptoms, such as rhinorrhea, stuffy nose, or sore throat. LRD was defined as detection of HMPV in a BAL or lung biopsy sample with new pulmonary infiltrates detected on radiographic studies; in analogy to a previous article on parainfluenza virus LRD, this manifestation was termed “proven LRD” [16]. “Possible LRD” was defined as viral detection in a nasopharyngeal sample and new pulmonary infiltrates without bronchoscopic examination, and these cases were included as URI in this study. Patients diagnosed with LRD ≥2 days after URI diagnosis were considered to have progression to LRD. Duration of viral shedding was defined as days between the first and the last date of HMPV detection. A copathogen was defined as a significant pathogen when detected in a concurrent respiratory sample or in a blood sample obtained within 2 days of diagnosis of HMPV infection. Coagulase-negative staphylococcal bacteremia and cytomegalovirus viremia or antigenemia were not considered significant copathogens. Peak steroid dose before diagnosis was recorded from the period within 2 weeks [16].

current respiratory sample or in a blood sample obtained within 2 days of diagnosis of HMPV infection. Coagulase-negative staphylococcal bacteremia and cytomegalovirus viremia or antigenemia were not considered significant copathogens. Peak steroid dose before diagnosis was recorded from the period within 2 weeks [16]. Statistical Analysis Patients' demographic characteristics were compared among disease categories using χ2 or Fisher exact test for categorical variables and Wilcoxon rank-sum test for continuous variables (as appropriate). The probability of progression to LRD among patients who presented with URI was estimated by cumulative incidence curves, treating death as a competing risk. Cox proportional hazards models were used to evaluate unadjusted and adjusted hazard ratios for progression to LRD. Logistic regression models were used to evaluate cross-sectional association between each risk factor and occurrence of LRD among all patients (including patients who presented with LRD). Risk factors for duration of shedding were assessed using linear regression on the ranks of days of shedding by day 100. Two-sided P values <.05 were considered statistically significant. All statistical analyses were performed using SAS version 9.3 for Windows (SAS Institute, Cary, North Carolina).

resented with LRD). Risk factors for duration of shedding were assessed using linear regression on the ranks of days of shedding by day 100. Two-sided P values <.05 were considered statistically significant. All statistical analyses were performed using SAS version 9.3 for Windows (SAS Institute, Cary, North Carolina). RESULTS Patient Characteristics and Seasonality HMPV was detected in 118 HCT recipients; 88 (75%) and 30 (25%) had URI alone and LRD, respectively. Characteristics of each HMPV infection group are shown in Table 1. The median time to HMPV detection after HCT was 278.5 days (range, −5 to 2773 days), and the median viral load in nasal wash samples was 1.1 × 106 copies/mL (range, 3.3 × 102–1.7 × 109 copies/mL). The median day of engraftment was 16 days (range, 0–39 days), and only 1 patient who did not have engraftment prior to death had LRD, which was documented 63 days after HCT. Earlier transplantation year, HMPV infection sooner after HCT, copathogens, and low monocyte counts were observed more frequently in LRD cases than URI-only cases. There were 21 “possible LRD” cases, of which 8 (38%) and 13 (62%) received their transplantation before and after 2009, respectively. Other respiratory viral copathogens such as rhinovirus (n = 13 [11%]), coronaviruses (n = 11 [9%]), or parainfluenza virus (n = 5 [4%]) were occasionally detected in the nasopharyngeal or BAL samples. Nonviral copathogens in BAL included Pseudomonas aeruginosa (n = 1 [1%]) and Aspergillus fumigatus (n = 2 [2%]). Four patients had bacteremia due to Stenotrophomonas maltophilia, Acinetobacter baumannii, Enterococcus faecium, or Streptococcus pneumoniae at diagnosis. Table 1. Characteristics of All Patients With Human Metapneumovirus Infection (N = 118)

d Pseudomonas aeruginosa (n = 1 [1%]) and Aspergillus fumigatus (n = 2 [2%]). Four patients had bacteremia due to Stenotrophomonas maltophilia, Acinetobacter baumannii, Enterococcus faecium, or Streptococcus pneumoniae at diagnosis. Table 1. Characteristics of All Patients With Human Metapneumovirus Infection (N = 118) Characteristic Total (N = 118) URI Alone (n = 88) LRD (n = 30) P Value Sex .53 Male 63 (53) 45 (51) 18 (60) Female 55 (47) 43 (49) 12 (40) Age at HCT, y .62 ≤20 19 (16) 16 (18) 3 (10) 21–60 81 (69) 59 (67) 22 (73) >60 18 (15) 13 (15) 5 (17) Disease risk at HCT .12 Standard 78 (66) 62 (70) 16 (53) High 40 (34) 26 (30) 14 (47) Transplantation year .011 2004–2009 58 (49) 37 (42) 21 (70) 2010–2014 60 (51) 51 (58) 9 (30) Transplantation number .19 First 94 (80) 73 (83) 21 (70) Second 22 (18) 14 (16) 8 (27) Third 2 (2) 1 (1) 1 (3) Stem cell source .27 Bone marrow 22 (18) 14 (16) 8 (27) Peripheral blood stem cell 81 (69) 61 (69) 20 (67) Cord blood 15 (13) 13 (15) 2 (6) Donor type 1.00 Autologous 25 (21) 19 (22) 6 (20) Related 30 (26) 22 (25) 8 (27) Unrelated 63 (53) 47 (53) 16 (53) Conditioning regimen .23 MA including high-dose TBI (≥12 Gy) 22 (19) 18 (20) 4 (13) MA ± low-dose TBI (≤2 Gy) 53 (45) 42 (48) 11 (37) Reduced intensity 43 (36) 28 (32) 15 (50) GVHD prophylaxis .64 CNI + MTX 38 (41) 30 (43) 8 (33) CNI + MMF 49 (53) 35 (51) 14 (59) Others 6 (6) 4 (6) 2 (8) Days between HCT and infection .027 ≤30 18 (15) 9 (10) 9 (30) 31–365 52 (44) 39 (44) 13 (43) >365 48 (41) 40 (46) 8 (27) Quantitative viral load at diagnosis, median (range) 1.1 × 106 (3.3 × 102–1.7 × 109) 2.6 × 106 (3.3 × 102–1.7 × 109) 4.2 × 105 (5.4 × 102–1.3 × 108) .15 Copathogen .038 None 82 (69) 66 (75) 16 (53) Any pathogen 36 (31) 22 (25) 14 (47) %FEV1/FVC pre HMPV LRD .58 ≥70 76 (78) 57 (80) 19 (73) <70 21 (22) 14 (20) 7 (27) %TLC pre HMPV LRD 1.00 ≥80 77 (85) 55 (85) 22 (85) <80 14 (15) 10 (15) 4 (15) White blood cell count at diagnosis .15 >1000 × 106/L 96 (90) 73 (92) 23 (82) ≤1000 × 106/L 11 (10) 6 (8) 5 (18) Neutrophil count at diagnosis .14 >1000 × 106/L 89 (84) 68 (87) 21 (75) ≤1000 × 106/L 17 (16) 10 (13) 7 (25) Lymphocyte count at diagnosis .06 >300 × 106/L 83 (78) 65 (83) 18 (64) ≤300 × 106/L 23 (22) 13 (17) 10 (36) Monocyte count at diagnosis .025 >300 × 106/L 65 (61) 53 (68) 12 (43) ≤300 × 106/L 41 (39) 25 (32) 16 (57) Steroid dose before diagnosis .10 None 59 (50) 44 (50) 15 (52) <1 mg/kg 51 (44) 41 (47) 10 (34) ≥1 mg/kg 7 (6) 3 (3) 4 (14) Intravenous immunoglobulin .015 No 115 (97) 88 (

64) ≤300 × 106/L 23 (22) 13 (17) 10 (36) Monocyte count at diagnosis .025 >300 × 106/L 65 (61) 53 (68) 12 (43) ≤300 × 106/L 41 (39) 25 (32) 16 (57) Steroid dose before diagnosis .10 None 59 (50) 44 (50) 15 (52) <1 mg/kg 51 (44) 41 (47) 10 (34) ≥1 mg/kg 7 (6) 3 (3) 4 (14) Intravenous immunoglobulin .015 No 115 (97) 88 ( 100) 27 (90) Yes 3 (3) 0 (0) 3 (10) Ribavirin <.001 No 101 (86) 85 (97) 16 (53) Yes 17 (14) 3 (3) 14 (47) Data are presented as No. (%) unless otherwise specified. All variables in Table 1 were used for the univariate analyses. Abbreviations: CNI, calcineurin inhibitor; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; GVHD, graft-vs-host disease; HCT, hematopoietic cell transplantation; HMPV, human metapneumovirus; LRD, lower respiratory tract disease; MA, myeloablative; MMF, mycophenolate mofetil; MTX, methotrexate; TBI, total body irradiation; %TLC, percentage of predicted total lung capacity; URI, upper respiratory tract infection.

HD, graft-vs-host disease; HCT, hematopoietic cell transplantation; HMPV, human metapneumovirus; LRD, lower respiratory tract disease; MA, myeloablative; MMF, mycophenolate mofetil; MTX, methotrexate; TBI, total body irradiation; %TLC, percentage of predicted total lung capacity; URI, upper respiratory tract infection. The number of HMPV infections by month between July 2008 and June 2014 is shown in Figure 1A. Most of the infections (93%) were identified between December and May of each calendar year. The incidence of HMPV infections in HCT recipients was similar to that in the community based on University of Washington reference laboratory data (Figure 1B). The numbers of HMPV cases by year are shown in Figure 1C and 1D. When HMPV infections were frequently detected, LRD occurred more frequently (Figure 1A and 1C). Figure 1. Monthly distribution of human metapneumovirus (HMPV) infection. A, Number of cases with HMPV infection by month, July 2008–June 2014, at our transplantation center. B, Number of cases with HMPV infection by month, July 2008–June 2014, diagnosed by the University of Washington (UW) reference laboratory, which tests samples from both Seattle and regional hospitals and healthcare providers. C, Number of cases with HMPV infection by year, January 2009–June 2014, at our center. *Number of cases in 2014 was obtained between January and June. D, Number of cases with HMPV infection by year, January 2009–June 2014, diagnosed by the UW reference laboratory. *Number of cases in 2014 was obtained between January and June. Abbreviations: HCT, hematopoietic cell transplantation; LRD, lower respiratory tract disease; URI, upper respiratory tract infection.

nd June. D, Number of cases with HMPV infection by year, January 2009–June 2014, diagnosed by the UW reference laboratory. *Number of cases in 2014 was obtained between January and June. Abbreviations: HCT, hematopoietic cell transplantation; LRD, lower respiratory tract disease; URI, upper respiratory tract infection. Progression From URI to LRD Among 30 patients with HMPV LRD, 17 (57%) had detected HMPV URI before diagnosis of LRD and progressed to LRD in a median of 7 days (range, 2–63 days) after URI. In total, among 105 patients with URI, the probability of progression to LRD within 40 days was 16% (Figure 2). Approximately 75% (14/17) of the patients with URI who progressed to LRD did so within 2 weeks after URI. Figure 2. Probability of progression to lower respiratory tract disease (LRD) after human metapneumovirus (HMPV) upper respiratory tract infection (URI) diagnosis.

gression to LRD within 40 days was 16% (Figure 2). Approximately 75% (14/17) of the patients with URI who progressed to LRD did so within 2 weeks after URI. Figure 2. Probability of progression to lower respiratory tract disease (LRD) after human metapneumovirus (HMPV) upper respiratory tract infection (URI) diagnosis. Risk Factors for Progression to LRD In the univariate Cox model, the factors shown in Table 1 were used to identify risk factors for progression to LRD among patients who presented with URI. Early HMPV infection after HCT, low lymphocyte count, and steroid use of ≥1 mg/kg before diagnosis were significantly associated with disease progression (Table 2), of which steroid dose remained significant in one bivariate model (Table 3). Probabilities of progression to LRD by these factors are shown in Figure 3. Approximately 60% of HCT recipients with low lymphocyte count or steroid use of ≥1 mg/kg before virologic diagnosis subsequently progressed to LRD. Table 2. Risk Factors for Progression From Upper Respiratory Tract Infection to Lower Respiratory Tract Disease (n = 105)

ion to LRD by these factors are shown in Figure 3. Approximately 60% of HCT recipients with low lymphocyte count or steroid use of ≥1 mg/kg before virologic diagnosis subsequently progressed to LRD. Table 2. Risk Factors for Progression From Upper Respiratory Tract Infection to Lower Respiratory Tract Disease (n = 105) Risk Factor Univariate Analysis HR 95% CI P Value Transplantation yeara 0.90 .73–1.10 .31 Days between HCT and infection ≤30 1.00 >30 3.54 1.31–9.60 .013 Copathogen None 1.00 Any pathogen 2.42 .93–6.27 .07 White blood cell count at diagnosis >1000 × 106/L 1.00 ≤1000 × 106/L 2.71 .88–8.33 .08 Lymphocyte count at diagnosis >300 × 106/L 1.00 ≤300 × 106/L 3.43 1.32–8.90 .011 Monocyte count at diagnosis >300 × 106/L 1.00 ≤300 × 106/L 2.31 .89–6.00 .08 Steroid dose before diagnosis None 1.00 <1 mg/kg 1.27 .43–3.77 .67 ≥1 mg/kg 5.74 1.62–20.40 .007 Viral load Low 1.00 High 0.47 .17–1.28 .14 All variables in Table 1 were used for the univariable analysis. Only variables with P < .1 are shown in this table. Transplantation year and viral load were shown regardless of P values. Abbreviations: CI, confidence interval; HCT, hematopoietic cell transplantation; HR, hazard ratio. a This variable was analyzed as continuous. Table 3. Bivariate Analysis of Risk Factors for Progression From Upper Respiratory Tract Infection to Lower Respiratory Tract Disease (n = 105)

Risk Factor Univariate Analysis HR 95% CI P Value Transplantation yeara 0.90 .73–1.10 .31 Days between HCT and infection ≤30 1.00 >30 3.54 1.31–9.60 .013 Copathogen None 1.00 Any pathogen 2.42 .93–6.27 .07 White blood cell count at diagnosis >1000 × 106/L 1.00 ≤1000 × 106/L 2.71 .88–8.33 .08 Lymphocyte count at diagnosis >300 × 106/L 1.00 ≤300 × 106/L 3.43 1.32–8.90 .011 Monocyte count at diagnosis >300 × 106/L 1.00 ≤300 × 106/L 2.31 .89–6.00 .08 Steroid dose before diagnosis None 1.00 <1 mg/kg 1.27 .43–3.77 .67 ≥1 mg/kg 5.74 1.62–20.40 .007 Viral load Low 1.00 High 0.47 .17–1.28 .14 All variables in Table 1 were used for the univariable analysis. Only variables with P < .1 are shown in this table. Transplantation year and viral load were shown regardless of P values. Abbreviations: CI, confidence interval; HCT, hematopoietic cell transplantation; HR, hazard ratio. a This variable was analyzed as continuous. Table 3. Bivariate Analysis of Risk Factors for Progression From Upper Respiratory Tract Infection to Lower Respiratory Tract Disease (n = 105) Variables Model 1 Model 2 Model 3 HR 95% CI P Value HR 95% CI P Value HR 95% CI P Value Days between HCT and infection >30 1 1 ≤30 1.89 .60–5.99 .28 2.58 .87–7.64 .09 Lymphocyte count at diagnosis >300/µL 1 1 ≤300/µL 2.60 .86–7.85 .09 2.60 .90–7.49 .08 Steroid dose before diagnosis <1 mg/kg 1 1 ≥1 mg/kg 3.44 1.01–11.70 .048 2.72 .78–9.45 .12 Abbreviations: CI, confidence interval; HCT, hematopoietic cell transplantation; HR, hazard ratio.

1 1 ≤30 1.89 .60–5.99 .28 2.58 .87–7.64 .09 Lymphocyte count at diagnosis >300/µL 1 1 ≤300/µL 2.60 .86–7.85 .09 2.60 .90–7.49 .08 Steroid dose before diagnosis <1 mg/kg 1 1 ≥1 mg/kg 3.44 1.01–11.70 .048 2.72 .78–9.45 .12 Abbreviations: CI, confidence interval; HCT, hematopoietic cell transplantation; HR, hazard ratio. Figure 3. Incidence of progression to lower respiratory tract disease (LRD) after diagnosis of human metapneumovirus (HMPV) upper respiratory tract infection (URI). A, Cumulative incidence of progression to LRD by days between hematopoietic cell transplantation (HCT) and diagnosis of URI (global P = .01, log-rank test). B, Cumulative incidence of progression to LRD by lymphocyte count (global P = .0007). C, Cumulative incidence of progression to LRD by steroid dose before diagnosis of URI (global P = .006).

progression to LRD by days between hematopoietic cell transplantation (HCT) and diagnosis of URI (global P = .01, log-rank test). B, Cumulative incidence of progression to LRD by lymphocyte count (global P = .0007). C, Cumulative incidence of progression to LRD by steroid dose before diagnosis of URI (global P = .006). In univariate logistic regression models among all patients, factors associated with the development of LRD were early onset of HMPV infection after HCT (odds ratio [OR], 3.70; 95% confidence interval [CI], 1.33–11.1; P = .010), low lymphocyte count (OR, 2.78; 95% CI, 1.05–7.37; P = .040), earlier transplantation year (OR, 1.22; 95% CI, 1.02–1.46; P = .028), presence of copathogens (OR, 2.63; 95% CI, 1.11–6.23; P = .029), and low monocyte count (OR, 3.00; 95% CI, 1.23–7.30; P = .016); use of steroids (≥1 mg/kg) approached statistical significance (OR, 4.53; 95% CI, .95–21.62; P = .06). In a multivariable logistic model that included transplantation year, low lymphocyte count, and high steroid use before diagnosis, earlier transplantation year (adjusted OR [aOR], 1.33; 95% CI, 1.09–1.63; P = .005) and low lymphocyte count (aOR, 3.38; 95% CI, 1.07–10.67; P = .038) remained statistically significant. When early onset after transplantation was included in the model, lymphocytopenia was no longer significant, whereas early onset remained significant (aOR, 3.30; 95% CI, 1.01–10.83; P = .048).

CI, 1.09–1.63; P = .005) and low lymphocyte count (aOR, 3.38; 95% CI, 1.07–10.67; P = .038) remained statistically significant. When early onset after transplantation was included in the model, lymphocytopenia was no longer significant, whereas early onset remained significant (aOR, 3.30; 95% CI, 1.01–10.83; P = .048). Viral Shedding and Viral Load in Nasopharyngeal Samples Thirty-nine patients had continuous HMPV detection for up to 182 days. Only 2 patients who died before day 100 had prolonged shedding: 1 patient who died at day 20 had 20 days of shedding, and another who died at day 31 had 17 days of shedding. In a bivariate analysis of risk factors for duration of shedding that included the same 3 variables as in Table 3, steroid use of ≥1 mg/kg was associated with longer shedding (P = .030). Among 101 patients with detected viruses in nasal wash samples, we compared disease status by viral load. There was no evidence of a correlation between disease status and viral load in the nasal wash samples at presentation (data not shown). DISCUSSION This study demonstrated that low lymphocyte count and steroid use of ≥1 mg/kg at the time of URI diagnosis are associated with an increased risk of progression to LRD with progression rates approaching 60% in patients in the highest-risk categories. The viral load in nasopharyngeal samples at the time that HMPV URI was diagnosed did not appear to predict the risk of progression to LRD.

≥1 mg/kg at the time of URI diagnosis are associated with an increased risk of progression to LRD with progression rates approaching 60% in patients in the highest-risk categories. The viral load in nasopharyngeal samples at the time that HMPV URI was diagnosed did not appear to predict the risk of progression to LRD. HMPV is a relatively newly discovered respiratory pathogen and the effect of HMPV infections on HCT recipients is poorly understood. Numerous studies of respiratory disease due to RSV, a genetically similar respiratory virus, in transplantation recipients have been reported with progression rates to LRD and mortality after LRD ranging from 20% to 50% and 15% to 50%, respectively [17–21]. Due to the high incidence of RSV-related LRD and subsequent high mortality from RSV infection, RSV infections after HCT are considered a very important complication. Mortality following HMPV is comparably as high as that seen with RSV, at rates of approximately 40% [10]. HMPV is difficult to diagnose with culture methods, but several multiplex PCR platforms are now available that permit rapid, sensitive, and specific detection of HMPV.

are considered a very important complication. Mortality following HMPV is comparably as high as that seen with RSV, at rates of approximately 40% [10]. HMPV is difficult to diagnose with culture methods, but several multiplex PCR platforms are now available that permit rapid, sensitive, and specific detection of HMPV. The progression rate to LRD may be affected by the underlying conditions of the HCT recipient. We identified steroid use of ≥1 mg/kg before diagnosis, low lymphocyte count, and onset of HMPV infection during the first month after HCT as important risk factors for progression. As shown in Figure 3, 60% of the patients with steroid use of ≥1 mg/kg or with low lymphocyte count had progression. These 2 characteristics are generally associated with high mortality after LRD by respiratory virus [10, 16, 20]. The logistic regression model provided additional insights. Lymphopenia was important but lost significance when the timing after transplantation was included. This suggests that other host defense mechanisms in the early time period after transplantation, such as monocytes, are also important. With only 30 events of LRD, we were unable to fit a model including all factors that are potentially important.

ut lost significance when the timing after transplantation was included. This suggests that other host defense mechanisms in the early time period after transplantation, such as monocytes, are also important. With only 30 events of LRD, we were unable to fit a model including all factors that are potentially important. Disease progression following infection with RSV, a virus that is genetically similar to HMPV, has been related to patient characteristics including neutropenia, lymphocytopenia, unrelated donor status, conditioning including total body irradiation, fungal coinfections, and smoking history [17]. In our study, however, none of these factors except lymphocytopenia was associated with progression. Although we did not collect data on smoking history, we evaluated pulmonary function prior to HMPV infection, and no association between pulmonary dysfunction and progression of HMPV infections was detected.

story [17]. In our study, however, none of these factors except lymphocytopenia was associated with progression. Although we did not collect data on smoking history, we evaluated pulmonary function prior to HMPV infection, and no association between pulmonary dysfunction and progression of HMPV infections was detected. To improve patient outcome of HMPV infection, strategies need to be developed to prevent progression to LRD. Our study identified high-risk clinical situations during which early intervention strategies could be studied. To date, there is no proven effective treatment for HMPV infection [22, 23]. Several small reports support the use of ribavirin with or without intravenous immunoglobulin (IVIG) for HMPV infection [24–27]. A larger study, however, showed no protective effect of ribavirin for HMPV LRD [10]. All studies were nonrandomized. In our study, a maintenance dose of IVIG was not associated with a lower progression rate (data not shown). We were unable to analyze the effect of ribavirin or high-dose IVIG because the number of patients who received it before progression was low. Based on our finding, reduction of steroid doses if patients receive high-dose steroid may be a possible strategy. However, whether such an approach would be effective cannot be determined from this study.

e effect of ribavirin or high-dose IVIG because the number of patients who received it before progression was low. Based on our finding, reduction of steroid doses if patients receive high-dose steroid may be a possible strategy. However, whether such an approach would be effective cannot be determined from this study. Other important findings in this study are that steroid use of ≥1 mg/kg before diagnosis was associated with longer viral shedding and that viral load in a nasopharyngeal sample was not correlated with progression to LRD. As for the effect of viral load, we may have failed to show the relationship due to issues related to host immune function, which we did not quantify in this study. We generally perform nasopharyngeal testing when patients have URI symptoms. Although URI symptoms are based on host immune response, the level of the immune response in HCT recipients may not be correlated to viral load. Another possible reason is that progression to pneumonia may not be related to viral load in the upper respiratory tract. Previous studies of parainfluenza virus or pathogens detected in patients with idiopathic pneumonia syndrome after HCT failed to show an association of viral load in a BAL sample with the level of lung injury or outcome, which is consistent with the finding in the current study [16, 28]. Although we did not demonstrate a correlation between viral load and progression to LRD, it is possible that larger studies with standardized BAL viral load measurements that control for dilution may improve our understanding of the role of viral load in disease progression.

h the finding in the current study [16, 28]. Although we did not demonstrate a correlation between viral load and progression to LRD, it is possible that larger studies with standardized BAL viral load measurements that control for dilution may improve our understanding of the role of viral load in disease progression. This study has several limitations. Although this cohort represents the largest number of HCT recipients infected with HMPV who received a quantitative virological workup of both the upper and lower respiratory tract samples, the sample size was too small to definitively assess all the patient characteristics we studied and the role of individual copathogens. We also did not have enough events for multivariable analyses. Nevertheless, we were able to perform several different bivariate Cox models. Another limitation is the long observation period, during which various transplantation procedures have changed. Low lymphocyte count or steroid dose before diagnosis are important factors for progression to LRD. Reduced intensity conditioning regimens that are related to mild lymphocytopenia are frequently used. Our center has recently adopted lower dose of steroids for the initial treatment of acute graft-vs-host disease [29], which may have affected the incidence of LRD in recent years. Although we included possible factors affecting the progression rate in the analyses, we may have missed some important factors that changed during the study period. The main methods to detect HMPV also changed because of the long observation period. However, there was no apparent bias in the way samples were collected for testing. We performed bronchoscopy when patients had lower respiratory tract symptoms and radiographic abnormalities. Nevertheless, the final decision was left to the attending physician. Therefore, an aggressive workup was declined in some “possible LRD” cases, especially in cases late after HCT, which may have resulted in underreporting of progression to LRD in some cases. The virological analyses were also limited because of inconsistent surveillance of viral shedding and incomplete ascertainment of viral load by dilution factor of the nasal wash samples, although a previous study suggested that the dilution effect is rather limited [13].

erreporting of progression to LRD in some cases. The virological analyses were also limited because of inconsistent surveillance of viral shedding and incomplete ascertainment of viral load by dilution factor of the nasal wash samples, although a previous study suggested that the dilution effect is rather limited [13]. In conclusion, our findings show that HMPV infections in HCT recipients occurred primarily in the winter and spring. The progression rate from URI to LRD approached 60% in patients who acquired HMPV infection and had low lymphocyte count or use of ≥1 mg/kg of steroids prior to URI. These data provide the necessary foundation to design clinical trials to study the efficacy of new antiviral compounds, which are urgently needed. Further studies are needed to define the role of viral load in the pathogenesis of progressive disease. Notes Acknowledgments. We thank Terry Stevens-Ayers for laboratory assistance, Angela P. Campbell for collection of laboratory data, and Brad Edmison and Lisa Chung for assistance with creating figures. Financial support. This work was supported by the National Institutes of Health (grant number K24HL093294). S. S. is a recipient of the Joel Meyers Endowment Scholarship. Potential conflicts of interest. M. B. has served as a consultant to Humabs Biomed and has received research funding from Ansun Biopharma. 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.

s of LRTD was estimated using the Kaplan-Meier method. The log-rank test was used to compare mortality curves among subgroups. Two-sided P values <.05 were considered statistically significant. All statistical analyses were performed using SAS software version 9.4 for Windows (SAS Institute, Inc, Cary, North Carolina). RESULTS Patient and Viral Characteristics We identified a total of 35 patients (37 episodes) with HCoV detected by RT-PCR from BAL samples. Table 1 shows characteristics of HCT recipients and patients with HM. Two patients developed HCoV LRTD twice. Two HCT recipients had a history of lung transplantation: 1 underwent lung transplantation for bronchiolitis obliterans related to previous HCT and the other received lung transplantation for cystic fibrosis before HCT. Only 1 pediatric patient (8-year-old male) was identified in this cohort. The median time to HCoV LRTD after HCT in 28 recipients was 302 days (range, 8–7045 days): 20 (71%) and 12 (43%) patients developed HCoV LRTD >100 days and >365 days following transplant, respectively. All but 1 of the 20 patients with HCoV LRTD >100 days following transplant received either immunosuppressive therapy or chemotherapy to control their underlying disorders (eg, relapse of hematologic malignancy, graft-vs-host disease) prior to diagnosis of LRTD. Twenty-three recipients were transplanted after 1 May 2006 when respiratory viral PCR panel testing became routine. The median time to HCoV LRTD after HCT in these 23 patients was 340 days (range, 8–3618 days), which was similar to that of entire cohort. At the time of BAL, acute respiratory symptoms and new pulmonary infiltrates were present in the majority of episodes (Table 2). Among 23 available frozen BAL samples, 11 (48%) were identified as OC43, 5 (22%) as NL63, 4 (17%) as 229E, and 3 (13%) as HKU1. The majority of episodes occurred in the winter and spring regardless of strain type (Figure 1A).

Respiratory viruses can cause lower respiratory tract disease (LRTD) in immunocompromised hosts, which is associated with significant morbidity and mortality [1–4]. With the development and widespread use of new molecular diagnostic techniques, the clinical impact of previously underdiagnosed respiratory viruses in this population remains uncertain [5]. This is particularly true of human coronavirus (HCoV). In addition to severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS) coronaviruses, 4 other strains of HCoV (229E, OC43, NL63, and HKU1) are now acknowledged to be human pathogens [6, 7]. Previous studies have demonstrated that HCoV is now the second most common virus identified from the upper respiratory tract in hematopoietic cell transplant (HCT) recipients [8]. Cases of fatal pneumonia related to HCoV without copathogens have also been reported mainly in HCT populations [9–12]. Two previous studies describe the possible role of HCoV in LRTD; however, these studies included only limited numbers of HCT recipients and patients with hematologic malignancy (HM), and outcome analyses could not be done [13, 14]. The purpose of this study was to describe the clinical characteristics and outcomes of HCT recipients and patients with HM with HCoV detected in the lower respiratory tract based on testing of bronchoalveolar lavage (BAL) fluid. Mortality rates were compared among HCT recipients with LRTD caused by HCoV, respiratory syncytial virus (RSV), influenza virus, or parainfluenza virus (PIV).

ics and outcomes of HCT recipients and patients with HM with HCoV detected in the lower respiratory tract based on testing of bronchoalveolar lavage (BAL) fluid. Mortality rates were compared among HCT recipients with LRTD caused by HCoV, respiratory syncytial virus (RSV), influenza virus, or parainfluenza virus (PIV). METHODS Study Design We identified all HCT recipients and patients with HM with HCoV detected in clinical BAL samples from patients at the Fred Hutchinson Cancer Research Center, University of Washington, or Seattle Children’s Hospital from May 2006 through February 2016. We identified 3 additional HCT recipients with HCoV detected in BAL samples from a previously reported cohort [15]. Demographic and clinical data were collected from the above-mentioned institutions’ databases and medical chart review. We also compared HCT recipients with HCoV LRTD to previously reported cohorts of HCT recipients with LRTD caused by RSV, influenza virus, and PIV [16–18]. The study was approved by the Institutional Review Board at the Fred Hutchinson Cancer Research Center.

e-mentioned institutions’ databases and medical chart review. We also compared HCT recipients with HCoV LRTD to previously reported cohorts of HCT recipients with LRTD caused by RSV, influenza virus, and PIV [16–18]. The study was approved by the Institutional Review Board at the Fred Hutchinson Cancer Research Center. Laboratory Testing and Definitions Reverse-transcription polymerase chain reaction (RT-PCR) was performed for HCoV on BAL samples, serum specimens, lung biopsy, and autopsy samples according to a previously published protocol [19]. Viral load of HCoV was determined by quantitative RT-PCR using BAL samples. HCoV was identified from BAL specimens using the consensus HCoV assay, which is part of a multiplex PCR used to detect 12 respiratory viruses. Strain-specific PCR was performed using saved BAL samples as described previously [19]. We performed RT-PCR to detect HCoV RNA in frozen serum samples that were drawn between 23 days before and 23 days after the BAL. When adequate lung tissue was available, curls were cut from fresh frozen tissue or formalin-fixed, paraffin-embedded (FFPE) tissue blocks for RT-PCR. FFPE samples and frozen samples were extracted using RNAeasy FFPE kit and RNAeasy mini kit, respectively (Qiagen, Hilden, Germany). All samples underwent fragment size analysis for quality with RT-PCR targeting amplicons from housekeeping genes with sizes ranging from 100 to 600 base pairs (bp).

ssue blocks for RT-PCR. FFPE samples and frozen samples were extracted using RNAeasy FFPE kit and RNAeasy mini kit, respectively (Qiagen, Hilden, Germany). All samples underwent fragment size analysis for quality with RT-PCR targeting amplicons from housekeeping genes with sizes ranging from 100 to 600 base pairs (bp). HCoV LRTD was defined as HCoV detection in a BAL sample from a patient with signs of LRTD (eg cough, dyspnea) or new pulmonary infiltrates. All BAL specimens underwent broad diagnostic tests including conventional cultures for bacteria, fungi, mycobacteria, and viruses, shell vial culture for cytomegalovirus, immunofluorescent antibody staining for Pneumocystis jirovecii and Legionella, fungal PCR, Aspergillus galactomannan enzyme-linked immunosorbent assay, and cytopathologic examination. HCoV was considered the sole respiratory pathogen if all above-mentioned microbiological test results on BAL specimens were negative. Pulmonary bacterial coinfection was defined as bacterial load of >103 colony-forming units per milliliter of BAL specimen with compatible radiological findings and clinical course. Any virus or fungus detected in BAL samples was considered a respiratory copathogen. Highest steroid doses in the 2 weeks prior to HCoV LRTD and cell counts most immediately prior to HCoV LRTD were recorded. Oxygen-free days and ventilator-free days are defined as days alive and free from oxygen support and mechanical ventilation, respectively [16]. Respiratory death was defined as any death occurring as a consequence of respiratory failure.

or to HCoV LRTD and cell counts most immediately prior to HCoV LRTD were recorded. Oxygen-free days and ventilator-free days are defined as days alive and free from oxygen support and mechanical ventilation, respectively [16]. Respiratory death was defined as any death occurring as a consequence of respiratory failure. Morphologic re-review of available BAL samples, lung biopsies, and autopsy lung tissues was performed on hematoxylin and eosin–stained sections by a board-certified pathologist with expertise in transplant pathology.

or to HCoV LRTD and cell counts most immediately prior to HCoV LRTD were recorded. Oxygen-free days and ventilator-free days are defined as days alive and free from oxygen support and mechanical ventilation, respectively [16]. Respiratory death was defined as any death occurring as a consequence of respiratory failure. Morphologic re-review of available BAL samples, lung biopsies, and autopsy lung tissues was performed on hematoxylin and eosin–stained sections by a board-certified pathologist with expertise in transplant pathology. Statistical Analysis Patients’ outcomes were compared using χ2 or Fisher exact test for categorical variables and Wilcoxon rank-sum test for continuous variables, as appropriate. Summary of the various patient cohorts according to analysis type is shown in Supplementary Figure 1. Only the first episode of HCoV LRTD per subject was used for outcome analyses. We also excluded 2 HCT recipients with a history of lung transplantation for outcome analyses except for evaluation of risk factors for mortality following HCoV LRTD. Univariable Cox proportional hazards models were used to evaluate risk factors for overall mortality by day 90 after the diagnosis of HCoV LRTD. Variables with a P value ≤.2 in the univariable models were candidates for multivariable models. A multivariable Cox regression model adjusted for respiratory viruses (HCoV, RSV, influenza, and PIV), cell source, neutrophil counts, lymphocyte counts, monocyte counts, presence of copathogens, steroid dosage, and oxygen use at diagnosis was performed. Patients with any respiratory viral copathogens were excluded for this analysis. The probability of overall mortality in HCT recipients by day 90 following the diagnosis of LRTD was estimated using the Kaplan-Meier method. The log-rank test was used to compare mortality curves among subgroups. Two-sided P values <.05 were considered statistically significant. All statistical analyses were performed using SAS software version 9.4 for Windows (SAS Institute, Inc, Cary, North Carolina).

ory symptoms and new pulmonary infiltrates were present in the majority of episodes (Table 2). Among 23 available frozen BAL samples, 11 (48%) were identified as OC43, 5 (22%) as NL63, 4 (17%) as 229E, and 3 (13%) as HKU1. The majority of episodes occurred in the winter and spring regardless of strain type (Figure 1A). Table 1. Characteristics of All Patients With Human Coronavirus Lower Respiratory Tract Disease Characteristic Total (N = 35) Hematopoietic Cell Transplant Recipients (n = 28) Patients With Hematologic Malignancy (n = 7) Female sex 10 (29) 10 (36) 0 Age, y, median (range) 53 (8–68) 53.5 (24–67) 52 (8–68) Underlying pulmonary disordera 10 (29) 10 (36) 0 Immunosuppressive therapy or chemotherapy 34 (97) 27 (96) 7 (100) Transplant year 1996–2006 6 (21) 2007–2015 22 (79) Transplant number ≥2 10 (36) Cell source Cord 1 (11) Bone marrow 6 (21) PBSC 21 (75) Donor type Autologous 4 (14) Related 10 (36) Unrelated 14 (50) Days between transplant and HCoV LRTD, median (range) 302 (8–7045) Data are presented as No. (11) unless otherwise indicated. Abbreviations: HCoV, human coronavirus; LRTD, lower respiratory tract disease; PBSC, peripheral blood stem cell. a Bronchiolitis obliterans (n = 4), lung transplantation for bronchiolitis obliterans (n = 1), lung transplantation for cystic fibrosis (n = 1), radiation pneumonia (n = 2), asthma (n = 1), prolonged acute respiratory distress syndrome (n = 1), diffuse alveolar hemorrhage (n = 1), cystic fibrosis (n = 1). Table 2. Presentation of Human Coronavirus Lower Respiratory Tract Disease Episodes

a Bronchiolitis obliterans (n = 4), lung transplantation for bronchiolitis obliterans (n = 1), lung transplantation for cystic fibrosis (n = 1), radiation pneumonia (n = 2), asthma (n = 1), prolonged acute respiratory distress syndrome (n = 1), diffuse alveolar hemorrhage (n = 1), cystic fibrosis (n = 1). Table 2. Presentation of Human Coronavirus Lower Respiratory Tract Disease Episodes Characteristic Totala (n = 37) Hematopoietic Cell Transplant Recipients (n = 30) Patients With Hematologic Malignancy (n = 7) Respiratory symptomsb 34 (92) 27 (90) 7 (100) Abnormal lung examinationc 25 (68) 21 (70) 4 (57) Abnormal findings on chest imagingd 34 (92) 27 (90) 7 (100) HCoV strain OC43 11 (30) 10 (33) 1 (14) NL63 5 (14) 4 (13) 1 (14) 229E 4 (11) 4 (13) 0 HKU1 3 (11) 3 (10) 0 Unknown 14 (38) 9 (30) 5 (71) Respiratory copathogen 21 (57) 18 (60) 3 (43) None 16 (43) 12 (40) 4 (57) Viruses 5 (13) 3 (10) 2 (29) Bacteria 4 (11) 4 (13) Fungi 4 (11) 4 (13) Multiple 8 (22) 7 (23) 1 (14) Quantitative viral load, log10 copies/mL, median (range) 5.4 (2.4–9.0) 5.3 (2.4–7.8) 6.1 (3.4–7.4) WBC count ≤1000 × 106 cells/L 11 (30) 7 (23) 4 (57) Lymphocyte count ≤300 × 106 cells/L 19 (51) 15 (50) 4 (57) Neutrophil count ≤500 × 106 cells/L 14 (38) 9 (30) 5 (71) Monocyte count ≤300 × 106 cells/L 24 (65) 19 (63) 5 (71) Steroid dosee None 14 (38) 7 (23) 7 (100) ≤1 mg/kg 13 (35) 13 (43) 0 >1 mg/kg 10 (27) 10 (33) 0 Oxygen requirement at diagnosis 23 (62) 20 (67) 3 (43) Data are presented as No. (11) unless otherwise indicated. Abbreviations: HCoV, human coronavirus; WBC, white blood cell.

Characteristic Totala (n = 37) Hematopoietic Cell Transplant Recipients (n = 30) Patients With Hematologic Malignancy (n = 7) Respiratory symptomsb 34 (92) 27 (90) 7 (100) Abnormal lung examinationc 25 (68) 21 (70) 4 (57) Abnormal findings on chest imagingd 34 (92) 27 (90) 7 (100) HCoV strain OC43 11 (30) 10 (33) 1 (14) NL63 5 (14) 4 (13) 1 (14) 229E 4 (11) 4 (13) 0 HKU1 3 (11) 3 (10) 0 Unknown 14 (38) 9 (30) 5 (71) Respiratory copathogen 21 (57) 18 (60) 3 (43) None 16 (43) 12 (40) 4 (57) Viruses 5 (13) 3 (10) 2 (29) Bacteria 4 (11) 4 (13) Fungi 4 (11) 4 (13) Multiple 8 (22) 7 (23) 1 (14) Quantitative viral load, log10 copies/mL, median (range) 5.4 (2.4–9.0) 5.3 (2.4–7.8) 6.1 (3.4–7.4) WBC count ≤1000 × 106 cells/L 11 (30) 7 (23) 4 (57) Lymphocyte count ≤300 × 106 cells/L 19 (51) 15 (50) 4 (57) Neutrophil count ≤500 × 106 cells/L 14 (38) 9 (30) 5 (71) Monocyte count ≤300 × 106 cells/L 24 (65) 19 (63) 5 (71) Steroid dosee None 14 (38) 7 (23) 7 (100) ≤1 mg/kg 13 (35) 13 (43) 0 >1 mg/kg 10 (27) 10 (33) 0 Oxygen requirement at diagnosis 23 (62) 20 (67) 3 (43) Data are presented as No. (11) unless otherwise indicated. Abbreviations: HCoV, human coronavirus; WBC, white blood cell. aTwo patients had separated HCoV lower respiratory tract disease (LRTD) episodes. The first patient developed LRTD 361 days and 415 days following hematopoietic cell transplant (HCT), respectively. The second patient developed LRTD 425 days before and 11 days after HCT, respectively. bCough or dyspnea. cCrackles, wheeze, rhonchi, or decreased breath sound. dAny new abnormal lung findings except for single nodule.

aTwo patients had separated HCoV lower respiratory tract disease (LRTD) episodes. The first patient developed LRTD 361 days and 415 days following hematopoietic cell transplant (HCT), respectively. The second patient developed LRTD 425 days before and 11 days after HCT, respectively. bCough or dyspnea. cCrackles, wheeze, rhonchi, or decreased breath sound. dAny new abnormal lung findings except for single nodule. eMaximum daily dose within 2 weeks prior to diagnosis.

aTwo patients had separated HCoV lower respiratory tract disease (LRTD) episodes. The first patient developed LRTD 361 days and 415 days following hematopoietic cell transplant (HCT), respectively. The second patient developed LRTD 425 days before and 11 days after HCT, respectively. bCough or dyspnea. cCrackles, wheeze, rhonchi, or decreased breath sound. dAny new abnormal lung findings except for single nodule. eMaximum daily dose within 2 weeks prior to diagnosis. Figure 1. A, Human coronavirus strain. Seasonal distribution of human coronavirus lower respiratory tract disease (LRTD). B, Respiratory copathogens in human coronavirus LRTD. Each color indicates category of copathogen as follows: white (viruses), gray (fungi), and dark gray (bacteria). Respiratory viral copathogens were detected in 12 patients, fungal copathogens were detected in 10 patients, and bacterial copathogens were detected in 8 patients. The number of patients (n = 30) with other respiratory copathogens does not equal the sum of detections for each respiratory copathogen (n = 36) owing to codetections of multiple copathogens in some subjects.Abbreviations: ADV, adenovirus; A. fumigutus, Aspergillus fumigatus; B. cepacia, Burkholderia cepacia; CMV, cytomegalovirus; C. neoformans, Cryptococcus neoformans; E. faecium, Enterococcus faecium; H. influenzae, Haemophilus influenzae; HMPV, human metapnuemovirus; NT, nontypeable strain due to unavailable sample; P. aeruginosa, Pseudomonas aeruginosa; PIV, parainfluenza virus; PJP, Pneumocystis jirovecii; RSV, respiratory syncytial virus; RV, rhinovirus; S. aureus, Staphylococcus aureus; VGS, Viridans group streptococci.

nzae, Haemophilus influenzae; HMPV, human metapnuemovirus; NT, nontypeable strain due to unavailable sample; P. aeruginosa, Pseudomonas aeruginosa; PIV, parainfluenza virus; PJP, Pneumocystis jirovecii; RSV, respiratory syncytial virus; RV, rhinovirus; S. aureus, Staphylococcus aureus; VGS, Viridans group streptococci. Other respiratory pathogens were detected in BAL samples in 21 episodes (57%), including viruses (12 episodes), fungi (10 episodes), and bacteria (8 episodes) (Figure 1B). Two or more other respiratory copathogens were detected in approximately half of these episodes (10/21). Two patients had respiratory copathogen as well as concomitant bacteremia/fungemia; 1 patient was found to have Staphylococcus aureus in blood and BAL, and the other had both Clostridium non-perfringens and Candida glabrata in the blood only.

ry copathogens were detected in approximately half of these episodes (10/21). Two patients had respiratory copathogen as well as concomitant bacteremia/fungemia; 1 patient was found to have Staphylococcus aureus in blood and BAL, and the other had both Clostridium non-perfringens and Candida glabrata in the blood only. Respiratory copathogens were found in 82% (9/11) of episodes with OC43 and only 42% (5/12) of episodes with other strains (P = .089). The median viral loads of HCoV in BAL samples did not differ among strains (Figure 2). No HCoV RNA was detected in serum samples prior to and following HCoV LRTD available from 21 episodes. Five lung biopsy samples and 4 lung autopsy samples were tested for RT-PCR among 6 patients (3 FFPE samples and 6 fresh frozen samples), all obtained within 67 days after LRTD diagnosis. Quality control fragment size analysis by RT-PCR of the RNA from these samples shows that all FFPE specimens could be reliably amplified to 100 bp while all frozen specimens were reliably amplified to 600 bp. Only 1 sample (lung tissue) was positive for HCoV by RT-PCR, which had been obtained on the same day as BAL. Figure 2. Viral load of human coronavirus in bronchoalveolar lavage samples. The bars indicate median values and first and third quartiles.

Respiratory copathogens were found in 82% (9/11) of episodes with OC43 and only 42% (5/12) of episodes with other strains (P = .089). The median viral loads of HCoV in BAL samples did not differ among strains (Figure 2). No HCoV RNA was detected in serum samples prior to and following HCoV LRTD available from 21 episodes. Five lung biopsy samples and 4 lung autopsy samples were tested for RT-PCR among 6 patients (3 FFPE samples and 6 fresh frozen samples), all obtained within 67 days after LRTD diagnosis. Quality control fragment size analysis by RT-PCR of the RNA from these samples shows that all FFPE specimens could be reliably amplified to 100 bp while all frozen specimens were reliably amplified to 600 bp. Only 1 sample (lung tissue) was positive for HCoV by RT-PCR, which had been obtained on the same day as BAL. Figure 2. Viral load of human coronavirus in bronchoalveolar lavage samples. The bars indicate median values and first and third quartiles. Outcomes Patients’ outcomes were summarized after excluding patients with second episodes of HCoV LRTD or a history of lung transplantation (Tables 3 and 4). Outcomes by day 28 and 90 after LRTD diagnosis were compared between HCT recipients and patients with HM. HCT recipients were more likely to have fewer oxygen and ventilator-free days than patients with HM. Outcomes by day 28 and 90 after HCoV diagnosis were also compared between patients with and without respiratory copathogens, with no statistical differences found. Table 3. Outcome of Patients With Human Coronavirus Lower Respiratory Tract Disease

Outcomes Patients’ outcomes were summarized after excluding patients with second episodes of HCoV LRTD or a history of lung transplantation (Tables 3 and 4). Outcomes by day 28 and 90 after LRTD diagnosis were compared between HCT recipients and patients with HM. HCT recipients were more likely to have fewer oxygen and ventilator-free days than patients with HM. Outcomes by day 28 and 90 after HCoV diagnosis were also compared between patients with and without respiratory copathogens, with no statistical differences found. Table 3. Outcome of Patients With Human Coronavirus Lower Respiratory Tract Disease Outcome Total (N = 33) Hematopoietic Cell Transplant Recipients (n = 26) Patients With Hematologic Malignancy (n = 7) P Value Outcome by day 28 after diagnosis Mechanical ventilation requirement, No. (11) 7 (21) 7 (27) 0 .30 Oxygen-free days 17.0 (11.8) 15.0 (12.2) 24.7 (5.6) .04 Ventilator-free days 22.1 (9.7) 20.5 (10.4) 28.0 (0.0) .03 Days alive without hospitalization 11.7 (10.9) 10.5 (10.7) 16.1 (11.5) .22 Outcome by day 90 after diagnosis Any death, No. (11) 18 (55) 16 (62) 2 (29) .20 Respiratory death, No. (11) 10 (30) 9 (35) 1 (14) .40 Data are presented as mean (standard deviation) unless otherwise indicated. Table 4. Outcome of Human Coronavirus Lower Respiratory Tract Disease With and Without Respiratory Copathogens

Outcome Total (N = 33) Hematopoietic Cell Transplant Recipients (n = 26) Patients With Hematologic Malignancy (n = 7) P Value Outcome by day 28 after diagnosis Mechanical ventilation requirement, No. (11) 7 (21) 7 (27) 0 .30 Oxygen-free days 17.0 (11.8) 15.0 (12.2) 24.7 (5.6) .04 Ventilator-free days 22.1 (9.7) 20.5 (10.4) 28.0 (0.0) .03 Days alive without hospitalization 11.7 (10.9) 10.5 (10.7) 16.1 (11.5) .22 Outcome by day 90 after diagnosis Any death, No. (11) 18 (55) 16 (62) 2 (29) .20 Respiratory death, No. (11) 10 (30) 9 (35) 1 (14) .40 Data are presented as mean (standard deviation) unless otherwise indicated. Table 4. Outcome of Human Coronavirus Lower Respiratory Tract Disease With and Without Respiratory Copathogens Outcome Total (N = 33) HCoV as Sole Respiratory Pathogen (n = 14) HCoV Coinfected With Other Respiratory Pathogens (n = 19) P Value Outcome by day 28 after diagnosis Mechanical ventilation requirement, No. (11) 7 (21) 2 (14) 5 (26) .67 Oxygen-free days 17.0 (11.8) 19.0 (11.3) 15.6 (12.2) .36 Ventilator-free days 22.1 (9.7) 24.4 (8.1) 20.4 (10.6) .16 Days alive without hospitalization 11.7 (10.9) 13.4 (11.0) 10.4 (11.0) .43 Outcome by day 90 after diagnosis Any death, No. (11) 18 (55) 7 (50) 11 (58) .65 Respiratory death, No. (11) 10 (30) 2 (14) 8 (42) .13 Data are presented as mean (standard deviation) unless otherwise indicated. Abbreviation: HCoV, human coronavirus.

Outcome Total (N = 33) HCoV as Sole Respiratory Pathogen (n = 14) HCoV Coinfected With Other Respiratory Pathogens (n = 19) P Value Outcome by day 28 after diagnosis Mechanical ventilation requirement, No. (11) 7 (21) 2 (14) 5 (26) .67 Oxygen-free days 17.0 (11.8) 19.0 (11.3) 15.6 (12.2) .36 Ventilator-free days 22.1 (9.7) 24.4 (8.1) 20.4 (10.6) .16 Days alive without hospitalization 11.7 (10.9) 13.4 (11.0) 10.4 (11.0) .43 Outcome by day 90 after diagnosis Any death, No. (11) 18 (55) 7 (50) 11 (58) .65 Respiratory death, No. (11) 10 (30) 2 (14) 8 (42) .13 Data are presented as mean (standard deviation) unless otherwise indicated. Abbreviation: HCoV, human coronavirus. Pathology Results Twenty-eight patients had samples available for pathologic review, including 25 BAL samples, 5 lung biopsy samples, and 4 autopsy lung specimens; 6 of 7 patients without any other respiratory copathogens had either nonspecific findings of multinucleated giant cells or nuclear enlargement (Supplementary Figure 2A and 2B). Lung tissue from 1 patient was positive for HCoV by RT-PCR; the morphologic features noted in the lung biopsy were inflamed tissue with lymphocytes, neutrophils, and cytologic atypia (Supplementary Figure 3A and 3B).

pecific findings of multinucleated giant cells or nuclear enlargement (Supplementary Figure 2A and 2B). Lung tissue from 1 patient was positive for HCoV by RT-PCR; the morphologic features noted in the lung biopsy were inflamed tissue with lymphocytes, neutrophils, and cytologic atypia (Supplementary Figure 3A and 3B). Comparison of Mortality With Other Respiratory Viruses and Risk Factors for Mortality There was a total of 286 HCT recipients with a single respiratory virus identified in BAL samples for whom comparable clinical data were available (HCoV [n = 18], RSV [n = 113], influenza virus [n = 36], and PIV [n = 119]); demographics of these are shown in Supplementary Table 1. Overall mortality rates by day 90 following viral LRTD caused by HCoV, RSV, influenza virus, and PIV among HCT recipients without respiratory viral copathogens and without any copathogens were not different (P = .78 and P = .47, respectively) (Figure 3A and 3B). Furthermore, no difference was seen when the cohort was stratified by those with and without oxygen requirement at the time of LRTD diagnosis (P = . 78 for both) (Figure 3C and 3D). Univariable Cox regression models were used to evaluate risk factors for overall mortality in HCT recipients with LRTD caused by HCoV, RSV, influenza virus, or PIV without respiratory viral copathogens (Table 5). In multivariable models, cell source (bone marrow), respiratory bacterial or fungal copathogens, low neutrophil counts, low monocyte counts, steroid use, and oxygen requirement at diagnosis were associated with overall mortality (Table 6). Mortality due to HCoV LRTD was not significantly different from RSV LRTD (adjusted hazard ratio, 1.34 [95% confidence interval, .66–2.71], P = .41). Similarly, risk factors for overall mortality by day 90 in HCoV LRTD patients alone were evaluated using univariable and multivariable Cox regression models in HCT recipients; no risk factors significantly associated with mortality were found (Supplementary Tables 2 and 3).

95% confidence interval, .66–2.71], P = .41). Similarly, risk factors for overall mortality by day 90 in HCoV LRTD patients alone were evaluated using univariable and multivariable Cox regression models in HCT recipients; no risk factors significantly associated with mortality were found (Supplementary Tables 2 and 3). Table 5. Univariable Cox Regression Analysis for Overall Mortality by Day 90 After Diagnosis of Lower Respiratory Tract Disease (n = 286) Covariates Category Hazard Ratio (95% CI) P Value Cell source Peripheral blood stem cell 1 Bone marrow 1.69 (1.22–2.36) <.01 Cord 0.51 (.16–1.62) .25 Transplant year 1993–2006 1 2007–2015 0.87 (.60–1.25) .45 Respiratory copathogen None 1 Nonrespiratory virusa ± bacteria/fungi 1.61 (.93–2.80) .09 Bacteria/fungi 1.54 (1.09–2.19) .02 Days between transplant and diagnosis ≤30 1 31–365 0.94 (.65–1.34) .71 >365 0.55 (.32–.94) .03 White blood cell count, 106 cells/L ≤1.0 1.69 (1.21–2.37) <.01 >1.0 1 Neutrophil count, 106 cells/L <0.5 1.76 (1.26–2.47) <.01 ≥0.5 1 Lymphocyte count, 106 cells/L <0.3 1.63 (1.17–2.29) <.01 ≥0.3 1 Monocyte count, 106 cells/L <0.3 2.38 (1.53–3.70) <.01 ≥0.3 1 Steroid use within 2 weeks before diagnosis No 1 <1 mg/kg 0.96 (.62–1.47) .85 1–2 mg/kg 1.48 (1.00–2.20) .05 >2 mg/kg 2.23 (1.16–4.27) .02 Oxygen use at diagnosis No 1 Any 2.51 (1.72–3.66) <.01 Respiratory virus Respiratory syncytial virus 1 Parainfluenza virus 1.16 (.81–1.68) .42 Influenza virus 1.08 (.63–1.84) .78 Human coronavirus 1.32 (.69–2.53) .40 Abbreviation: CI, confidence interval.

.85 1–2 mg/kg 1.48 (1.00–2.20) .05 >2 mg/kg 2.23 (1.16–4.27) .02 Oxygen use at diagnosis No 1 Any 2.51 (1.72–3.66) <.01 Respiratory virus Respiratory syncytial virus 1 Parainfluenza virus 1.16 (.81–1.68) .42 Influenza virus 1.08 (.63–1.84) .78 Human coronavirus 1.32 (.69–2.53) .40 Abbreviation: CI, confidence interval. a Cytomegalovirus, herpes simplex virus, human herpesvirus 6, and Epstein-Barr virus. Table 6. Multivariable Cox Regression Analysis for Overall Mortality by Day 90 After Diagnosis of Lower Respiratory Tract Disease (n = 286) Covariates Categories Adjusted HR (95% CI) P Value Cell source Peripheral blood stem cell 1 Bone marrow 1.64 (1.13–2.40) .01 Cord 0.74 (.23–2.41) .62 Respiratory copathogen None 1 Nonrespiratory virusa ± bacteria/fungi 1.80 (1.00–3.26) .05 Bacteria/fungi 1.66 (1.12–2.45) .01 Neutrophil count, 106 cells/L <0.5 1.61 (1.00–2.58) .05 ≥0.5 1 Lymphocytes count, 106 cells/L <0.3 0.95 (.62–1.45) .81 ≥0.3 1 Monocyte count, 106 cells/L <0.3 1.87 (1.12–3.13) .02 ≥0.3 1 Steroid use within 2 wk before diagnosis No 1 <1 mg/kg 1.27 (.77–2.08) .35 1–2 mg/kg 1.38 (.87–2.20) .18 >2 mg/kg 2.40 (1.15–5.03) .02 Oxygen use at diagnosis No 1 Any 3.00 (1.98–4.53) <.01 Respiratory virus Respiratory syncytial virus 1 Parainfluenza virus 1.13 (.77–1.67) .52 Influenza virus 0.88 (.47–1.66) .70 Human coronavirus 1.34 (.66–2.71) .41 Abbreviations: CI, confidence interval; HR, hazard ratio. a Cytomegalovirus, herpes simplex virus, human herpesvirus 6, and Epstein-Barr virus.

Covariates Categories Adjusted HR (95% CI) P Value Cell source Peripheral blood stem cell 1 Bone marrow 1.64 (1.13–2.40) .01 Cord 0.74 (.23–2.41) .62 Respiratory copathogen None 1 Nonrespiratory virusa ± bacteria/fungi 1.80 (1.00–3.26) .05 Bacteria/fungi 1.66 (1.12–2.45) .01 Neutrophil count, 106 cells/L <0.5 1.61 (1.00–2.58) .05 ≥0.5 1 Lymphocytes count, 106 cells/L <0.3 0.95 (.62–1.45) .81 ≥0.3 1 Monocyte count, 106 cells/L <0.3 1.87 (1.12–3.13) .02 ≥0.3 1 Steroid use within 2 wk before diagnosis No 1 <1 mg/kg 1.27 (.77–2.08) .35 1–2 mg/kg 1.38 (.87–2.20) .18 >2 mg/kg 2.40 (1.15–5.03) .02 Oxygen use at diagnosis No 1 Any 3.00 (1.98–4.53) <.01 Respiratory virus Respiratory syncytial virus 1 Parainfluenza virus 1.13 (.77–1.67) .52 Influenza virus 0.88 (.47–1.66) .70 Human coronavirus 1.34 (.66–2.71) .41 Abbreviations: CI, confidence interval; HR, hazard ratio. a Cytomegalovirus, herpes simplex virus, human herpesvirus 6, and Epstein-Barr virus. Figure 3. Kaplan-Meier overall survival curve by day 90 after diagnosis of lower respiratory tract disease without respiratory viral copathogens according to respiratory virus classification in hematopoietic cell transplant recipients. A, Kaplan-Meier overall survival curve in overall cohort (n = 286) (log-rank test, P = .78). B, Kaplan-Meier overall survival curve in patients without other copathogens (n = 173) (log-rank test, P = .47). C, Kaplan-Meier overall survival curve in patients with oxygen requirement at diagnosis (n = 178) (log-rank test, P = .78). D, Kaplan-Meier overall survival curve in patients without oxygen requirement at diagnosis (n = 108) (log-rank test, P = .78).

atients without other copathogens (n = 173) (log-rank test, P = .47). C, Kaplan-Meier overall survival curve in patients with oxygen requirement at diagnosis (n = 178) (log-rank test, P = .78). D, Kaplan-Meier overall survival curve in patients without oxygen requirement at diagnosis (n = 108) (log-rank test, P = .78). DISCUSSION In this study, we demonstrated that the presence of HCoV in BAL samples in immunocompromised hosts was significantly associated with high rates of respiratory support and mortality. HCT recipients appeared to be more affected than patients with HM. Although respiratory copathogens were frequently detected, the clinical outcomes of these patients were similar to those without copathogens. The mortality rate of HCT recipients by day 90 after developing HCoV LRTD was similar to rates seen with established respiratory pathogens including RSV, influenza virus, and PIV (Figure 3 and Table 6) [16–18]. All 4 HCoV strains were identified in BAL samples regardless of the presence of copathogens, and at least 2 HCoV strains were present nearly half of the year.

ter developing HCoV LRTD was similar to rates seen with established respiratory pathogens including RSV, influenza virus, and PIV (Figure 3 and Table 6) [16–18]. All 4 HCoV strains were identified in BAL samples regardless of the presence of copathogens, and at least 2 HCoV strains were present nearly half of the year. SARS and MERS are recognized as highly human-pathogenic coronaviruses, causing acute, severe, frequently fatal LRTD [20–22]. Although 4 other strains of HCoV (229E, OC43, NL63, and HKU1) are also human pathogens, the clinical impact of HCoV LRTD remains unclear, especially in immunocompromised patients [4, 14, 23]. A previous prospective study with weekly nasal surveillance sampling during the first 100 days after HCT demonstrated prolonged HCoV shedding in the upper respiratory tract (>3 weeks) in half of subjects including asymptomatic patients [8]. In addition, respiratory copathogens were identified in more than half the episodes in this study (57%). Given the prolonged asymptomatic shedding and frequent detection of respiratory copathogens, attributing poor clinical outcomes to HCoV in the lower respiratory tract may be difficult. In the current study, follow-up BAL procedures were not performed to assess prolonged shedding in the lower respiratory tract; however, a prior study that included only 4 patients with cancer demonstrated that only 1 of 10 cases had HCoV detected in follow-up BAL specimens, arguing against asymptomatic prolonged shedding in the lower respiratory tract [14]. More data are needed to define shedding duration in the immunocompromised population. In our study, clinical outcomes including intensity of respiratory support, days alive without hospitalization, and mortality were not significantly different between patients with and without other copathogens, suggesting that HCoV in the lower respiratory tract can contribute to severity of LRTD regardless of copathogens. Lung tissue from 1 patient was positive for HCoV by RT-PCR. This patient subsequently developed prolonged oxygen requirement, which also supports the potential pathogenicity of HCoV.

hout other copathogens, suggesting that HCoV in the lower respiratory tract can contribute to severity of LRTD regardless of copathogens. Lung tissue from 1 patient was positive for HCoV by RT-PCR. This patient subsequently developed prolonged oxygen requirement, which also supports the potential pathogenicity of HCoV. To further demonstrate the clinical significance of HCoV LRTD in HCT recipients, we compared mortality rates in HCT recipients with HCoV LRTD to other respiratory viruses using multivariable Cox regression analysis and found mortality rates in HCoV LRTD were comparable to those seen with RSV, influenza virus, and PIV. Given the common perception of HCoV as a relatively benign pathogen based on limited data [5, 8], our data are somewhat surprising. The adverse impact of oxygen requirement at the time of diagnosis on subsequent clinical outcome has been suggested [16, 24]. Once substantial acute lung injury occurs, clinical outcome can potentially be affected by inflammation rather than virus itself. Therefore, mortality rates by day 90 following LRTD caused by HCoV, RSV, influenza virus, and PIV were also compared according to oxygen requirement at the time of LRTD diagnosis; no statistically significant difference was found. These data combined suggest that HCoV LRTD is significantly associated with poor clinical outcome in this immunocompromised population.

LRTD caused by HCoV, RSV, influenza virus, and PIV were also compared according to oxygen requirement at the time of LRTD diagnosis; no statistically significant difference was found. These data combined suggest that HCoV LRTD is significantly associated with poor clinical outcome in this immunocompromised population. Previous studies, mainly in immunocompetent hosts, have not demonstrated a distinct association between particular HCoV strains in upper respiratory tract samples and disease severity in LRTD [25–29]. Although there are case reports with each strain identified in lower respiratory tract as a sole pathogen in HCT recipients, systematic data are limited [9–12]. This is the first study to describe that all 4 HCoV strains can be detected in BAL samples with and without any other respiratory copathogens in a large immunocompromised population. However, the few instances of each strain limited our ability to examine if specific HCoV strains are associated with increased disease severity in LRTD. Further studies with larger sample sizes will help to characterize the role of particular HCoV strains.

spiratory copathogens in a large immunocompromised population. However, the few instances of each strain limited our ability to examine if specific HCoV strains are associated with increased disease severity in LRTD. Further studies with larger sample sizes will help to characterize the role of particular HCoV strains. Our study showed relatively late presentation of HCoV LRTD following transplantation. The median time to HCoV LRTD following HCT was 302 days, which was longer than median days to LRTD caused by RSV (52.5 days), PIV (78 days), and influenza (95 days) [16, 17, 30]. Because this may in part be due to the fact our cohort included some patients who were transplanted prior to the introduction of routine HCoV PCR testing, we separately analyzed patients who underwent transplantation after introduction of a respiratory viral PCR panel in 2006 and determined the median time of diagnosis remained similar (340 days). The majority of patients who developed HCoV LRTD >100 days following transplant had received either chemotherapy or immunosuppressive therapy as predisposing factors. This does not explain why there is a relatively lower incidence of HCoV LRTD early after transplant. Differences in infection control practices early after transplant and factors that affect progression to LRTD early vs late after transplant may play a role. Further studies are needed to determine why HCoV often causes LRTD late after transplant.

ere is a relatively lower incidence of HCoV LRTD early after transplant. Differences in infection control practices early after transplant and factors that affect progression to LRTD early vs late after transplant may play a role. Further studies are needed to determine why HCoV often causes LRTD late after transplant. This study evaluated the largest cohort of HCoV LRTD confirmed with BAL in patients with HCT and HM by HCoV strain-specific and quantitative PCR in BAL samples. In addition, RT-PCR was performed on serum specimens, lung biopsy samples, and autopsy samples. The main limitation of this study was the relatively small sample size, which prevented us from detecting small differences and performing multivariable analyses to evaluate risk factors for mortality in patients with HCoV LRTD. Another limitation is the fact that BAL samples were available in only two-thirds of the patients for strain identification, which limited our ability to compare clinical and virological differences among each HCoV strain. Among a total of 9 lung biopsy and autopsy samples, only 1 lung biopsy sample, which was taken on the same day as the BAL, was positive for HCoV by RT-PCR. The lower rate of detection may be due to the fact that the negative samples were obtained from 19 days to 67 days after the diagnosis of LRTD, suggesting that the timing of collecting samples may have been too late to identify HCoV in lung specimens. Furthermore, not all samples were optimally preserved for RT-PCR, and thus the sensitivity for HCoV may have been suboptimal.

he negative samples were obtained from 19 days to 67 days after the diagnosis of LRTD, suggesting that the timing of collecting samples may have been too late to identify HCoV in lung specimens. Furthermore, not all samples were optimally preserved for RT-PCR, and thus the sensitivity for HCoV may have been suboptimal. We demonstrated high rates of respiratory support including oxygen use and mechanical ventilation requirement as well as a high mortality in immunocompromised patients with HCoV identified in the lower respiratory tract. Mortality rates associated with HCoV LRTD in transplant recipients were similar to those seen with other respiratory viral pathogens including RSV, influenza virus, and PIV. Thus, we conclude that HCoV appears to be a significant respiratory pathogen in the populations studied. This is an important observation because HCoVs are highly prevalent in immunocompromised hosts. The appreciation of HCoV as an important lower respiratory tract pathogen could impact clinical management including risk stratification in future studies and provide a rationale to develop antiviral therapies. Further studies are needed to clarify if particular HCoV strains and viral load are correlated with clinical outcome and to identify risk factors for progression to LRTD.

act pathogen could impact clinical management including risk stratification in future studies and provide a rationale to develop antiviral therapies. Further studies are needed to clarify if particular HCoV strains and viral load are correlated with clinical outcome and to identify risk factors for progression to LRTD. Supplementary Data Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material Supplementary_Table_and_Figures Click here for additional data file. Notes Acknowledgments. We thank Zachary Stednick for database services; Terry Stevens-Ayers, Reigran Sampoleo, Isabel Palileo, Kristen Shimp, Petrina Mulhauser, and Catherine Spurgeon for laboratory assistance. Financial support. This work was supported by the National Institutes of Health (grant numbers K24HL093294 to M. B., K23 AI114844 to A. W., CA18029 to W. L., clinical database, CA15704 to H. X.); the Fred Hutchinson Cancer Research Center Vaccine and Infectious Disease Division (biorepository); T32HD00723332and Pediatric Infectious Diseases Society Fellowship Award funded by Horizon Pharma to C. O.

grant numbers K24HL093294 to M. B., K23 AI114844 to A. W., CA18029 to W. L., clinical database, CA15704 to H. X.); the Fred Hutchinson Cancer Research Center Vaccine and Infectious Disease Division (biorepository); T32HD00723332and Pediatric Infectious Diseases Society Fellowship Award funded by Horizon Pharma to C. O. Potential conflicts of interest. M. B. has received research support from and served as a consultant to Gilead Sciences, Merck, Ansun Bioscience, and Aviragen Therapeutics, and has served as a consultant for Humabs Biomed. J. A. E. has received research support from GlaxoSmithKline, Gilead, Pfizer, and Chimerix, and has served as a consultant for Pfizer and GlaxoSmith Kline (data safety monitoring board). C. Y. has received a research grant from Gilead Sciences for unrelated research. 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.