Browse the corpus

1 Introduction The convergence of the HIV and tuberculosis (TB) epidemics in Southern Africa has contributed to a significant increase in reported cases of TB and especially of extrapulmonary TB (EPTB).1 Abdominal TB is a frequent localization of EPTB in HIV-positive patients.2 Abdominal symptoms are common at all stages of HIV infection, but particularly at advanced stages of immunosuppression. It can be challenging in a setting with limited diagnostics to differentiate between symptoms caused by HIV itself and symptoms related to opportunistic infection or neoplasm.3 The lack of advanced imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI), and the limited use of cytology, histology, and cultures, compel physicians to make treatment decisions based on history, clinical examination, and basic laboratory tests. Simple non-invasive diagnostic tools are urgently required in resource-limited rural settings. In the rural district hospital setting, ultrasound is often the most commonly available imaging modality.

ures, compel physicians to make treatment decisions based on history, clinical examination, and basic laboratory tests. Simple non-invasive diagnostic tools are urgently required in resource-limited rural settings. In the rural district hospital setting, ultrasound is often the most commonly available imaging modality. World Health Organization guidelines contain criteria for the diagnosis of some forms of EPTB, but do not explicitly include abdominal TB.4 Typical ultrasound findings of abdominal TB have been described in low HIV prevalence settings.5–7 They include retroperitoneal and mesenteric lymphadenopathy with node diameter greater than 1.5 cm, multiple splenic hypoechoic nodules between 0.5 cm and 1 cm, and patterns of ascites. Hepatic hyperechoic nodules, retroperitoneal abscess, and bowel wall thickening are less frequently described.5 Although similar ultrasound findings of lymphadenopathy and splenomegaly are described in African HIV patients, the role of TB co-infection has rarely been explored.8,9 One recent report from Zambia described typical findings of abdominal TB in a high proportion of individuals with low CD4 counts and abdominal symptoms.10 We sought to investigate the role of ultrasound in the diagnosis of abdominal TB in a rural district hospital in an area of high HIV and TB prevalence.

explored.8,9 One recent report from Zambia described typical findings of abdominal TB in a high proportion of individuals with low CD4 counts and abdominal symptoms.10 We sought to investigate the role of ultrasound in the diagnosis of abdominal TB in a rural district hospital in an area of high HIV and TB prevalence. 2 Methods 2.1 Study setting Hlabisa Hospital and its 16 primary healthcare (PHC) clinics serve the Hlabisa sub-district in northern KwaZulu-Natal, South Africa. In an area of approximately 1400 km2 lives a mainly Zulu-speaking population of approximately 228 000. The sub-district is predominantly rural with one peri-urban area. The hospital has 300 beds and is staffed by nine to 12 physicians. Approximately 60 000 outpatient visits and 6500 admissions with an average length of stay of six days occur per year. Care for internal medicine patients is delivered in separate male, female, and TB wards. The diagnostic services available are limited. X-ray and ultrasound can be used as imaging modalities. Basic hematology and biochemistry are available in the hospital laboratory; in addition microscopy of urine and cerebrospinal fluid, TB microscopy, and CD4 counts are done. TB culture and histology are available at the provincial reference hospital, but results rarely influence acute clinical management. Patients can be referred to the provincial hospital for CT and MRI scans, but the distance of approximately 100 km, as well as the limited appointments available, restrict this option to selected cases.

histology are available at the provincial reference hospital, but results rarely influence acute clinical management. Patients can be referred to the provincial hospital for CT and MRI scans, but the distance of approximately 100 km, as well as the limited appointments available, restrict this option to selected cases. The HIV prevalence in the adult resident population (aged 15–49 years) is 21.5%.11 The TB notification rate in Hlabisa sub-district has increased from approximately 700 per 100 000 persons in 2003 to 1700 per 100 000 persons in 2008; the number of reported cases of EPTB has risen over 10-fold from 41 to 522 in the same time-period (personal communication, Department of Health, Umkhanyakude District). The HIV co-infection rate for TB patients is approximately 76%.12 The TB Control Programme adheres to national guidelines: treatment is initiated by nurses for smear-positive pulmonary disease; the diagnosis of smear-negative pulmonary or EPTB is made by a physician on the basis of clinical or radiological findings (rarely with culture or histological confirmation).13 This occurs in hospital, at the central TB clinic, or by physicians serving the decentralized HIV treatment and care program. Patients can be referred to hospital from the clinics for further diagnostic assessment if required.

is of clinical or radiological findings (rarely with culture or histological confirmation).13 This occurs in hospital, at the central TB clinic, or by physicians serving the decentralized HIV treatment and care program. Patients can be referred to hospital from the clinics for further diagnostic assessment if required. 2.2 Study procedure Patients were enrolled into the study from June to August 2008. All participants were adult (≥18 years) HIV-positive patients admitted to the general medical wards. Patients from the tuberculosis ward were not included as they were invariably known TB cases admitted to the ward to receive streptomycin injections or multidrug-resistant TB treatment. Patients were initially seen by admitting hospital staff and were referred to the study if they reported one or more of the following clinical symptoms: abdominal pain, abdominal distension, diarrhea, or weight loss. Patients were subsequently interviewed and examined by the study physician assisted by a nurse (who also translated from isiZulu). A standardized data sheet was used to collect patient information, including clinical symptoms, laboratory results, chest X-ray (CXR) results, CD4 cell count, and history of antiretroviral therapy (ART) use. Patients were included if their HIV diagnosis was known at the time of hospital admission or if the diagnosis was made during the inpatient episode.

o collect patient information, including clinical symptoms, laboratory results, chest X-ray (CXR) results, CD4 cell count, and history of antiretroviral therapy (ART) use. Patients were included if their HIV diagnosis was known at the time of hospital admission or if the diagnosis was made during the inpatient episode. There is no established ultrasound department at the study hospital and all exams were performed by the study physician using a Toshiba Just Vision 400 Model SSA-325A with a 3.5 MHz abdominal probe in a room adjacent to the X-ray facilities. All ultrasound studies were performed by a physician trained in abdominal ultrasound with more than 10 years of experience in the technique. Abdominal sonographic findings were then recorded on the data sheet. Abdominal TB was diagnosed on the basis of typical findings of lymph node enlargement greater than 1.5 cm and/or focal splenic lesions. Ascites was noted as further supportive evidence. If only ascites was present this was not considered diagnostic and the patient was not included in the study group; further investigations for alternative diagnoses (e.g., cirrhosis, congestive heart failure) were undertaken. After TB diagnosis, patients were treated according to the South African national TB guidelines:13 initial treatment for new cases includes rifampin, isoniazid, pyrazinamide and ethambutol for 2 months, followed by rifampin and isoniazid for 4 months.

agnoses (e.g., cirrhosis, congestive heart failure) were undertaken. After TB diagnosis, patients were treated according to the South African national TB guidelines:13 initial treatment for new cases includes rifampin, isoniazid, pyrazinamide and ethambutol for 2 months, followed by rifampin and isoniazid for 4 months. If patients were not already on ART they were offered treatment according to South African national guidelines (EPTB is a stage 4 condition and ART is indicated regardless of CD4 cell count).14 The first-line ART regimen includes stavudine, lamivudine, and either efavirenz or nevirapine (efavirenz preferred in TB co-infection). Follow-up information was obtained at approximately 4 months by the TB coordinator who contacted the patient by phone and asked a series of questions regarding weight gain, resolution of symptoms (abdominal pain, diarrhea), and whether or not ART had been commenced. Ethical approval was obtained from the Hlabisa Hospital Ethics Committee. Verbal consent was obtained from all patients; this was considered adequate as all procedures were considered part of routine clinical care.

If patients were not already on ART they were offered treatment according to South African national guidelines (EPTB is a stage 4 condition and ART is indicated regardless of CD4 cell count).14 The first-line ART regimen includes stavudine, lamivudine, and either efavirenz or nevirapine (efavirenz preferred in TB co-infection). Follow-up information was obtained at approximately 4 months by the TB coordinator who contacted the patient by phone and asked a series of questions regarding weight gain, resolution of symptoms (abdominal pain, diarrhea), and whether or not ART had been commenced. Ethical approval was obtained from the Hlabisa Hospital Ethics Committee. Verbal consent was obtained from all patients; this was considered adequate as all procedures were considered part of routine clinical care. 3 Results One hundred and eighty adult HIV-positive patients with symptoms were screened; amongst these, 30 (16.7%) fulfilled the case definition of abdominal TB based on sonographic features. Patients included in the analysis were between 18 and 49 years of age and 53% were women. The median CD4 count was 78 cells/mm3 (91% with CD4 <200 cells/mm3; 41% with CD4 <50 cells/mm3). Three patients were on ART at the time of the examination (for 1 week, 1 month, and 3 months, respectively). The median body-mass index was 19.9 kg/m2 (range 13.3–34.7) (Table 1). The clinical symptoms of the patients are summarized in Table 2; 86.7% presented with weight loss, 76.7% with abdominal pain, and 60% with diarrhea. Laboratory data from the cases at the time of hospital admission are displayed in Table 3.

ctively). The median body-mass index was 19.9 kg/m2 (range 13.3–34.7) (Table 1). The clinical symptoms of the patients are summarized in Table 2; 86.7% presented with weight loss, 76.7% with abdominal pain, and 60% with diarrhea. Laboratory data from the cases at the time of hospital admission are displayed in Table 3. Abdominal lymph node enlargement was a common finding, present in almost all cases (29/30, 96.7%). Hypoechoic lesions of the spleen were seen in 15/30 (50%) and ascites in 22/30 (73.3%) of the patients. The combinations of findings are shown in Table 4. Examples of abdominal lymphadenopathy and microabscesses in the spleen are shown in Figure 1. A CXR was available for assessment from 23 patients: 18 (78.3%) demonstrated findings compatible with TB, including five (21.7%) that displayed a miliary pattern; five (21.7%) had a normal CXR. Telephonic follow-up information was available for 25 patients a mean period of 17 weeks (range 12–23 weeks) after starting TB treatment. Six patients had died (24%), and the median time between start of TB treatment and death was 2 weeks (range 0–17 weeks). The remaining 19 patients (76%) were still taking TB treatment and reported clinical improvement and weight gain. All but one of the 16 surviving patients not already on ART had started this by the time of follow-up.

died (24%), and the median time between start of TB treatment and death was 2 weeks (range 0–17 weeks). The remaining 19 patients (76%) were still taking TB treatment and reported clinical improvement and weight gain. All but one of the 16 surviving patients not already on ART had started this by the time of follow-up. 4 Discussion Ultrasound is safe, portable, inexpensive, non-invasive, and can investigate most organs affected in HIV-infected patients.15 These features are critical for investigation of patients in resource-limited settings compared to patients in industrialized countries with a preponderance of CT and MRI scanners, and with trained radiologists to guide the interpretation. Sonographic findings suggestive of abdominal TB, especially in HIV-infected patients, have been described in patient populations of different racial backgrounds and have been suggested as part of the work-up in HIV-infected patients with fever of unknown origin.5–7,16 Despite the fact that suggestive findings have been reported in large ultrasound series in African patients, it is unclear whether in the face of other causes of lymphadenopathy the findings are specific enough to be considered diagnostic.8,9 The common dilemma for the physician faced with a heavily immunosuppressed patient in this setting is whether there is evidence of TB and a need for treatment or whether to start ART. This is particularly challenging as almost one in four patients still initiate ART with a CD4 count <50 cells/mm3.17 Commencement of ART without adequate exclusion of TB risks the development of immune reconstitution inflammatory syndrome (IRIS) with consequent morbidity and mortality.18

atment or whether to start ART. This is particularly challenging as almost one in four patients still initiate ART with a CD4 count <50 cells/mm3.17 Commencement of ART without adequate exclusion of TB risks the development of immune reconstitution inflammatory syndrome (IRIS) with consequent morbidity and mortality.18 In our patient series we found that ultrasound is a feasible technique to add weight to a diagnosis of abdominal TB in HIV-infected patients. Most patients have severe immunosuppression and report symptoms of weight loss, abdominal pain, and diarrhea, or a combination thereof. The pre-test probability for TB is often relatively high given the extremely high TB incidence in this community. Laboratory investigations in our patients frequently show anemia, most probably anemia of chronic disease related to both HIV and TB. All patients had a reduced albumin, most probably secondary to malnutrition and possibly malabsorption due to bowel disease. Bilirubin and gamma-glutamyl transferase were elevated in a large proportion of patients, suggesting involvement of the liver either as diffuse infiltration or with obstructive disease of the small bile ducts (no visible biliary dilatation was found in the ultrasound exams).

possibly malabsorption due to bowel disease. Bilirubin and gamma-glutamyl transferase were elevated in a large proportion of patients, suggesting involvement of the liver either as diffuse infiltration or with obstructive disease of the small bile ducts (no visible biliary dilatation was found in the ultrasound exams). In our series, scanning was made easier by the high proportion of underweight patients, as interference by fat and air is minimal. Abdominal TB was diagnosed by ultrasound without further invasive diagnostic steps; treatment was started on the basis of the imaging findings alone. In the majority of patients the approach led to a favorable outcome suggested by weight gain and cessation of symptoms. Of the six patients who died, three died within a week of the ultrasound exam, suggesting they died due to the severity of their disease. It is interesting to note that three patients were on ART at the time of the diagnosis. All of them started within a short period before the examination, suggesting IRIS as a potential contributory factor unmasking underlying abdominal TB.

the ultrasound exam, suggesting they died due to the severity of their disease. It is interesting to note that three patients were on ART at the time of the diagnosis. All of them started within a short period before the examination, suggesting IRIS as a potential contributory factor unmasking underlying abdominal TB. Limitations of our study include the lack of microbiological or histological confirmation of abdominal TB. Fine-needle aspiration biopsies are restricted for suspected drug-resistant cases and for clinical cases that show atypical signs (e.g., solid organ infiltration suggestive of lymphoma). Unfortunately, biopsy needles are scarce and histology and TB culture are not performed on site, but rather are sent to the central histology laboratory in Durban, and results rarely contribute to acute management. One has to consider other causes of abdominal lymphadenopathy. HIV itself can cause abdominal lymphadenopathy throughout the continuum of HIV infection, but this is not usually accompanied by other features.19 Kaposi's sarcoma (KS) is prevalent in the patient population in Hlabisa sub-district, although most patients with disseminated visceral KS have significant skin involvement, which none of our patient group exhibited. Lymphoma is an additional neoplasm that needs to be considered in the differential diagnosis. Infections with non-tuberculous mycobacteria (NTM) cannot be ruled out as a cause, but they are rarely diagnosed due to a lack of mycobacterial blood cultures; the prevalence of NTM disease in this setting is therefore unknown. Nevertheless, in a series of 31 patients, NTM were reported to cause abdominal adenopathy less frequently than Mycobacterium tuberculosis, and splenic microabscesses were not seen with NTM.20

ely diagnosed due to a lack of mycobacterial blood cultures; the prevalence of NTM disease in this setting is therefore unknown. Nevertheless, in a series of 31 patients, NTM were reported to cause abdominal adenopathy less frequently than Mycobacterium tuberculosis, and splenic microabscesses were not seen with NTM.20 We included only patients with symptoms related to the abdomen and did not include patients with isolated fever of unknown origin, so there is a possibility we may have missed additional cases of abdominal TB, particularly those at an earlier, less symptomatic stage. Our exclusion of individuals with ascites alone was based on our experience in this setting that ascites without fibrinous strands and with no other diagnostic signs can usually be attributed to chronic heart failure or liver failure. We may have misclassified some patients by this exclusion, and ascitic fluid culture for Mycobacterium tuberculosis is not routinely practiced here due to the cost and the relatively low yield (less than 20% in most series10). Our results are also hampered by the lack of information on those screened but not diagnosed with abdominal TB. We lost five out of our 30 patients to follow-up, which is most likely due to communication difficulties and patients with a lack of resources to be able to afford either a cell or land-based telephone.

s10). Our results are also hampered by the lack of information on those screened but not diagnosed with abdominal TB. We lost five out of our 30 patients to follow-up, which is most likely due to communication difficulties and patients with a lack of resources to be able to afford either a cell or land-based telephone. Further research is needed to assess the impact of a standardized ultrasound screening as part of work-up before starting ART, especially for those with advanced immunosuppression. It would further be interesting to evaluate, if in a fraction of patients the findings of abdominal TB would render CXR unnecessary. Portable ultrasound machines make it possible to examine patients in the PHC clinic setting, which would reduce the need (and the expenses) of patients traveling to hospital for CXR. As trained sonographers are scarce in the resource-limited setting, the possibility of training a wider range of medical doctors or clinical assistants in rural hospitals in a limited and focused assessment with sonography, similar to the FAST technique (focused abdominal sonography for trauma) widely used in emergency medicine, should be investigated.21 Preliminary results of a training course around focused assessment with sonography for HIV/TB (‘FASH’) have been reported by our group.22 This research into the use of ultrasound obviously does not obviate the need for improved TB diagnostics, particularly point-of-care diagnostics.

medicine, should be investigated.21 Preliminary results of a training course around focused assessment with sonography for HIV/TB (‘FASH’) have been reported by our group.22 This research into the use of ultrasound obviously does not obviate the need for improved TB diagnostics, particularly point-of-care diagnostics. 5 Conclusions Abdominal ultrasound is an effective diagnostic tool to identify lymphadenopathy and focal splenic lesions suggestive of abdominal TB in HIV-infected patients. As a diagnostic gold standard is not available, the sensitivity and specificity of the method are difficult to determine. Nevertheless, our patient series suggests that ultrasound-based diagnosis followed by TB treatment benefits the vast majority of patients and is a feasible approach when resources for further invasive diagnostics are limited. If implemented on a broader scale, ultrasound would assist physicians and other healthcare workers in making more appropriate treatment decisions instead of having to use the ‘treat first, think later’ approach, thus having the potential to save precious time and resources. Ultrasound has been shown to be a useful and financially affordable adjunct to clinical diagnosis in a wide variety of pathologies at the district healthcare level.23 Abdominal ultrasound should be introduced into clinical algorithms for the diagnosis of EPTB. Acknowledgements We would like to thank the medical and nursing staff of Hlabisa Hospital who assisted with patient identification and enrollment. We also thank Lungile Mhlongo who provided follow-up information.

5 Conclusions Abdominal ultrasound is an effective diagnostic tool to identify lymphadenopathy and focal splenic lesions suggestive of abdominal TB in HIV-infected patients. As a diagnostic gold standard is not available, the sensitivity and specificity of the method are difficult to determine. Nevertheless, our patient series suggests that ultrasound-based diagnosis followed by TB treatment benefits the vast majority of patients and is a feasible approach when resources for further invasive diagnostics are limited. If implemented on a broader scale, ultrasound would assist physicians and other healthcare workers in making more appropriate treatment decisions instead of having to use the ‘treat first, think later’ approach, thus having the potential to save precious time and resources. Ultrasound has been shown to be a useful and financially affordable adjunct to clinical diagnosis in a wide variety of pathologies at the district healthcare level.23 Abdominal ultrasound should be introduced into clinical algorithms for the diagnosis of EPTB. Acknowledgements We would like to thank the medical and nursing staff of Hlabisa Hospital who assisted with patient identification and enrollment. We also thank Lungile Mhlongo who provided follow-up information. Funding: TH is supported by the Centre for International Migration and Development (CIM), Gesellschaft für Technische Zusammenarbeit (GTZ), Federal Ministry of Economic Cooperation and Development, Germany. The Africa Centre for Health and Population Studies is supported by a core grant from the Wellcome Trust. Neither funding source had input in the study design, data analysis, writing of the manuscript, or the decision to submit for publication.

t (GTZ), Federal Ministry of Economic Cooperation and Development, Germany. The Africa Centre for Health and Population Studies is supported by a core grant from the Wellcome Trust. Neither funding source had input in the study design, data analysis, writing of the manuscript, or the decision to submit for publication. Conflict of interest: We declare that we have no conflicts of interest ☆ Presented in part at the 4th Southern African AIDS Conference, March 31–April 3, 2009, Durban, South Africa (abstract 278, poster presentation). Figure 1 Sonographic image of (a) enlarged lymph nodes in the periportal area; (b) focal lesions in the spleen of approximately 3 mm in diameter; (c) focal lesions in the spleen of approximately 10 mm in diameter. Table 1 Characteristics of patients with abdominal TB Sex Female 53% Male 47% Age, mean ± SD, years 31 ± 8 CD4 Median (range), cells/mm3 78 (6–408) <200 cells/mm3 91% (95% CI 81–100%) <50 cells/mm3 41% (95% CI 23–59%) BMI, median (range), kg/m2 19.9 (13.3–34.7) SD, standard deviation; CI, confidence interval; BMI, body mass index. Table 2 Symptoms reported by patients with abdominal TB (N = 29a) Symptoms n % 95% CI Weight loss only 3 10% 0–21% Abdominal pain only 3 10% 0–21% Weight loss and abdominal pain 5 16.7% 3–31% Weight loss and diarrhea 3 10% 0–21% Weight loss and abdominal pain and diarrhea 15 50% 31–69% CI, confidence interval. a One patient had only abdominal distension with no further symptoms. Table 3 Laboratory results in patients with abdominal TB

Symptoms n % 95% CI Weight loss only 3 10% 0–21% Abdominal pain only 3 10% 0–21% Weight loss and abdominal pain 5 16.7% 3–31% Weight loss and diarrhea 3 10% 0–21% Weight loss and abdominal pain and diarrhea 15 50% 31–69% CI, confidence interval. a One patient had only abdominal distension with no further symptoms. Table 3 Laboratory results in patients with abdominal TB Laboratory test Valuea 95% CI Hb, g/dl 7.4 ± 2.0 <8 g/dl 65% 45–85% ALT, U/l 47 ± 33 >60 U/l 23% 5–41% GGT, U/l 120 ± 93 >62 U/l 65% 44–86% Bilirubin (total), μmol/l 17 ± 20 >17.0 μmol/l 33% 12–54% Albumin, g/l 16 ± 3.3 <32 g/l 100% 95–100% Total protein, g/l 71 ± 14 >80 g/l 32% 12–52% CI, confidence interval; Hb, hemoglobin; ALT, alanine aminotransferase; GGT, gamma-glutamyl transferase. a Values are mean ± standard deviation. Table 4 Sonographic findings in abdominal TB patients (N = 30) Sonographic findings n % 95% CI Lymph nodes only 4 13% 1–25% Splenic lesions and ascites 1 3% 0–9% Lymph nodes and ascites 11 37% 19–55% Lymph nodes and splenic lesions 4 13% 1–25% Lymph nodes and splenic lesions and ascites 10 33% 16–50% CI, confidence interval.

Introduction Arboviral diseases are among the most important emerging infectious diseases threatening public health in many countries of the world.1 The Democratic Republic of the Congo (DRC) is the second largest African country and shares a long boundary with nine countries, including Congo, Central African Republic (CAR), South Sudan, Uganda, Rwanda, Burundi, Tanzania, Zambia, and Angola. Over the past decade, significant population movements arising from conflicts may have contributed to the introduction of arboviral diseases into new areas. Indeed, 2.4 million people were displaced within DRC because of war and 46 300 refugees have come in from neighbouring and endemic countries.2 Because of its large population of non-human primates and other animal reservoir hosts, DRC is believed to be the origin of several important emerging viruses of humans.3 However, few studies have been conducted on arboviruses within the country, hampering a reliable estimation of the current status and burden of arboviruses in DRC.

large population of non-human primates and other animal reservoir hosts, DRC is believed to be the origin of several important emerging viruses of humans.3 However, few studies have been conducted on arboviruses within the country, hampering a reliable estimation of the current status and burden of arboviruses in DRC. A recent study in the Congo basin that assessed the role of wildlife species as reservoirs for arboviruses (flaviviruses and alphaviruses) by testing sera from various animals, such as buffaloes, elephants, duikers, mandrills, gorillas, monkeys, and chimpanzees, showed the presence of antibodies against chikungunya, o’nyong’nyong, West Nile, dengue, and yellow fever viruses.4 In 1999 and 2000, two outbreaks of febrile illness were reported in humans following heavy rains in Matete and Kingabwa townships of the DRC capital city, Kinshasa. An estimated 50 000 humans cases were reported, and chikungunya virus was identified as the cause of these febrile illness outbreaks.5 A high proportion of tested sera from these cases presenting with febrile illness were found to have immunoglobulin M (IgM) antibodies against chikungunya virus but not against dengue virus, West Nile virus, or bunyaviruses.5 Later on, when the chikungunya viruses were isolated and their partial genomic sequences determined, the isolates from DRC were found to constitute a homogeneous group that was more closely related to CAR and Ugandan isolates than to Tanzanian and South African strains.6 These data called for urgency in conducting and expanding disease surveillance in DRC for emerging viruses that may represent an imminent threat to the population of Africa. However, disease prevention efforts have mostly focused on malaria and little has been done to understand the burden of arboviral disease and to mitigate the risks of possible large-scale outbreaks occurring in DRC.

veillance in DRC for emerging viruses that may represent an imminent threat to the population of Africa. However, disease prevention efforts have mostly focused on malaria and little has been done to understand the burden of arboviral disease and to mitigate the risks of possible large-scale outbreaks occurring in DRC. The present study was conducted to investigate the presence of arboviruses such as yellow fever virus, dengue virus, o’nyong’nyong virus, Rift Valley fever virus, and chikungunya virus, in order to provide information that could serve as an early warning for possible outbreaks and to understand the exposure risk in certain selected areas of Kinshasa. The findings of this study may assist in the development and implementation of strategies for the control of arboviral diseases.

, and chikungunya virus, in order to provide information that could serve as an early warning for possible outbreaks and to understand the exposure risk in certain selected areas of Kinshasa. The findings of this study may assist in the development and implementation of strategies for the control of arboviral diseases. Methods Study area Mosquitoes were collected in selected areas within Kinshasa (Figure 1), the capital city of DRC, located at 4° 19′ 30′′ S and 15° 19′ 20″ E. Kinshasa covers an area of 9965 km2, of which 90% is rural. It borders Brazzaville across the wide Congo River and is surrounded by four provinces, including Kongo-Central in the south and Kwilu, Kwango, and Mai-Ndombe in the north and in the east. The climate of Kinshasa is characterized by two seasons: a dry season from the second half of May to September and a rainy season from October to the first half of May, with a short break in February. Kinshasa experiences an average of 1482 mm of rainfall per year and has an average annual temperature of 25.2 °C and average annual relative humidity of 80.3%. The landscape ranges from the larger plain to some hills on the periphery, and a considerable hydrographic network crosses Kinshasa. The soils are sandy, sandy-argillaceous to argillaceous, with vegetation including steppes, semi deciduous and riverine forest islands, and wooded and grassy savannah. Kinshasa is divided into 24 municipalities.

m the larger plain to some hills on the periphery, and a considerable hydrographic network crosses Kinshasa. The soils are sandy, sandy-argillaceous to argillaceous, with vegetation including steppes, semi deciduous and riverine forest islands, and wooded and grassy savannah. Kinshasa is divided into 24 municipalities. The sampling sites for the present study are known for being the most malaria-endemic areas. These study sites were chosen because they possess favourable mosquito breeding habitats and a history of occurrence of arboviral disease outbreaks. The five study sites selected included Kimwenza, Kingabwa, Ndjili, Kimbanseke, and Ngaba (Figure 1). Kimwenza is a semi-rural area of Mont-Ngafula municipality set on a plateau in western Kinshasa, where a bonobo sanctuary is located and where the last epidemic of chikungunya occurred. Kingabwa is a suburban area of Limete municipality in the north-eastern part of Kinshasa along the Congo River, where rice agriculture is practiced and previous urban chikungunya outbreaks have occurred. Ndjili municipality is located in south-eastern Kinshasa along the Ndjili River, where pig farming, agriculture, and automobile repairs are conducted. Kimbanseke municipality is the most populated suburban area located in the eastern part of Kinshasa and is where vegetable cultivation is practiced and the swampy areas of Mokali and Nsanga are located. Ngaba municipality is an area surrounding the main market in central Kinshasa, characterized by a polluted environment due to the lack of an appropriate drainage sewage system and waste collection.

rn part of Kinshasa and is where vegetable cultivation is practiced and the swampy areas of Mokali and Nsanga are located. Ngaba municipality is an area surrounding the main market in central Kinshasa, characterized by a polluted environment due to the lack of an appropriate drainage sewage system and waste collection. Mosquito collection The present cross-sectional study was conducted between March and May 2014 in order to detect the presence of mosquito-borne viruses in mosquitoes. Adult mosquitoes were collected using unbaited BG-Sentinel traps (Biogents AG, Regensburg, Germany) that were set at 05:00 h and collected 24 h later. Daytime collections were made in order to collect day-biting Aedes mosquitoes. Resting mosquitoes were collected using hand-held battery-powered aspirators. Collections were made over a minimum of 3 days at each site. Collected mosquitoes were killed using alcohol vapour and identified to the genus level with the aid of morphological keys.7, 8, 9 Female mosquitoes were sorted into pools according to their genera and location of collection, and preserved in RNAlater (Ambion, Austin, TX, USA). Each pool contained a maximum of 30 mosquitoes that were afterwards stored at −20 °C before being shipped to Sokoine University of Agriculture, Morogoro, Tanzania for the molecular detection of mosquito-borne viruses.

rding to their genera and location of collection, and preserved in RNAlater (Ambion, Austin, TX, USA). Each pool contained a maximum of 30 mosquitoes that were afterwards stored at −20 °C before being shipped to Sokoine University of Agriculture, Morogoro, Tanzania for the molecular detection of mosquito-borne viruses. Screening for mosquito-borne viruses in mosquitoes RNA extraction After grinding each of the pools containing 30 female mosquitoes preserved in 1.5 ml of RNAlater, RNA was extracted using the QIAamp Viral RNA Mini Kit for nucleic acid extraction (Qiagen, Hilden, Germany); this was done in accordance with the manufacturer’s instructions. Briefly, 560 μl of lysis buffer containing carrier RNA and 140 μl of homogenized mosquito tissue supernatant were added to a 1.5-ml microcentrifuge tube. The contents were pulse-vortexed for 15 s and afterwards incubated at room temperature for 10 min to ensure complete viral particle and cellular lysis. Protein precipitation was conducted by adding 560 μl of absolute ethanol followed by pulse-vortexing for 15 s. Precipitates were settled by centrifugation at 13 000 g for 5 min. The supernatant was carefully withdrawn and passed through a silica-gel column, followed by washing of the column twice with 500 μl of each of the washing buffers AW1 and AW2, respectively. Finally, RNA was eluted by addition of 60 μl of buffer AVE equilibrated at room temperature. The viral RNA extracted was immediately stored at −80 °C until cDNA synthesis and amplification.

a silica-gel column, followed by washing of the column twice with 500 μl of each of the washing buffers AW1 and AW2, respectively. Finally, RNA was eluted by addition of 60 μl of buffer AVE equilibrated at room temperature. The viral RNA extracted was immediately stored at −80 °C until cDNA synthesis and amplification. Arbovirus detection by reverse transcription PCR Virological detection of mosquito-borne viruses in mosquito pools was conducted using reverse transcription PCR (RT-PCR). To convert extracted RNA into cDNA, a total volume of 20 μl containing 4 μl of RNA template added into 16 μl of a master mix containing 4 μl of 5 × VILO reaction mix, 2 μl of 10 × Superscript reverse transcriptase enzyme mix (Invitrogen, Carlsbad, CA, USA), and 10 μl nuclease-free water was used. After mixing by vortexing for 15 s, the tube contents were incubated in a GeneAmp 9700 PCR System (Applied Biosystems, Foster City, CA, USA) at 25 °C for 10 min, 42 °C for 60 min to achieve reverse transcription, followed by enzyme denaturation at 85 °C for 5 min and a hold at 4 °C.

10 μl nuclease-free water was used. After mixing by vortexing for 15 s, the tube contents were incubated in a GeneAmp 9700 PCR System (Applied Biosystems, Foster City, CA, USA) at 25 °C for 10 min, 42 °C for 60 min to achieve reverse transcription, followed by enzyme denaturation at 85 °C for 5 min and a hold at 4 °C. The cDNA was used for PCR amplification using primers targeting Flavivirus, Alphavirus, and Bunyaviridae, as reported previously by Ochieng et al.10 Briefly, a total of 20 μl of master mix was prepared, containing 10 μl of 2 × Dream Taq PCR Master Mix containing DNA polymerase (Thermo Scientific, Carlsbad, CA, USA), 1 μl of each forward and reverse primers, 1 μl of cDNA, and 7 μl of nuclease-free water. The PCR cycling conditions included an initial denaturation step at 94 °C for 15 min, followed by 35 cycles each consisting of a denaturation step at 94 °C for 30 s, an annealing step at 57 °C for 60 s, and an extension step at 72 °C for 30 s, followed by a final extension step at 72 °C for 10 min. Samples that tested strongly positive during screening for Flavivirus, Alphavirus, or Bunyaviridae were further screened for specific viruses using species-specific primers, as reported previously by Ochieng et al.10 For the detection of specific viruses, PCR was performed in a total reaction volume of 25 μl: 12.5 μl of 2 × Dream Taq PCR Master Mix containing DNA polymerase (Thermo Scientific, Carlsbad, USA), 0.5 μl of each forward and reverse primers, 2 μl of cDNA, and 9.5 μl of nuclease-free water.

previously by Ochieng et al.10 For the detection of specific viruses, PCR was performed in a total reaction volume of 25 μl: 12.5 μl of 2 × Dream Taq PCR Master Mix containing DNA polymerase (Thermo Scientific, Carlsbad, USA), 0.5 μl of each forward and reverse primers, 2 μl of cDNA, and 9.5 μl of nuclease-free water. Visualization of RT-PCR products The RT-PCR products were separated by electrophoresis in a 1.5% agarose gel in 0.5 × Tris–borate–ethylenediaminetetraacetic acid (TBE) buffer (SERVA, Heidelberg, Germany) stained with GelRed nucleic acid stain (Phenix Research Products, Candler, NC, USA). Each well was loaded with 5 μl of the PCR product and 1 μl of 6 × blue–orange DNA loading dye (Promega, Madison, USA). Samples were separated along with DNA ladder (Promega, Madison, USA) at 150 V for 30 min. The PCR products were visualized using a gel documentation system (EZ Gel Doc, Bio-Rad, France) and scoring was done based on the size of the PCR products.

oduct and 1 μl of 6 × blue–orange DNA loading dye (Promega, Madison, USA). Samples were separated along with DNA ladder (Promega, Madison, USA) at 150 V for 30 min. The PCR products were visualized using a gel documentation system (EZ Gel Doc, Bio-Rad, France) and scoring was done based on the size of the PCR products. Results Number of mosquitoes collected A total of 2922 mosquitoes were collected from the study sites, comprising 1986 Aedes spp, 631 Culex spp, 283 Anopheles spp, and 22 Mansonia spp (Table 1). The highest number of mosquitoes was collected from Kimwenza (n = 891), followed by Ndjili (n = 697), Kimbanseke (n = 528), Kingabwa (n = 413), and Ngaba (n = 393) (Table 1). Aedes spp were the most commonly collected mosquitoes at these sampling sites, except in Ngaba where Culex spp predominated (Table 1). Female mosquitoes were grouped into 29 pools based on location: Kimwenza (n = 9), Ndjili (n = 7), Kimbanseke (n = 5), Kingabwa (n = 4), and Ngaba (n = 4). These 29 pools of mosquitoes included 20 pools of Aedes spp, six pools of Culex spp, and three pools of Anopheles spp.

Ngaba where Culex spp predominated (Table 1). Female mosquitoes were grouped into 29 pools based on location: Kimwenza (n = 9), Ndjili (n = 7), Kimbanseke (n = 5), Kingabwa (n = 4), and Ngaba (n = 4). These 29 pools of mosquitoes included 20 pools of Aedes spp, six pools of Culex spp, and three pools of Anopheles spp. Mosquito-borne virus detection in mosquitoes by RT-PCR Arboviruses were detected in 12 of the 29 pools (41.4%) and included Alphavirus (eight pools), Flavivirus (nine pools), and Bunyaviridae (five pools) (Table 2). Some pools showed mixed infection with two or three arboviruses. Amplification with specific primers for different members of each genus showed the presence of chikungunya virus, o’nyong’nyong virus (Figure 2), and Rift Valley fever virus in Aedes pools and o’nyong’nyong virus in the Anopheles pool. Positive pools were detected in all of the locations tested within Kinshasa (Table 1). Discussion The abundance of mosquito vectors in an environment can favour the transmission of viruses. Furthermore, multiple feeding by mosquitoes can increase the probability of concurrent infection and viral genetic mixing.11 In the present study, the distribution patterns and abundance of mosquitoes were found to be restricted to certain areas, probably due to ecological and environmental adaptation. These findings are supported by the findings of an earlier study conducted by Karch et al., who found that the distribution of mosquito species in Kinshasa was non-homogeneous.12

n patterns and abundance of mosquitoes were found to be restricted to certain areas, probably due to ecological and environmental adaptation. These findings are supported by the findings of an earlier study conducted by Karch et al., who found that the distribution of mosquito species in Kinshasa was non-homogeneous.12 Although the significant presence of Aedes spp reported in this study could be due to the use of BG-Sentinel traps, which are known to be more effective for catching day mosquitoes, the abundance of Aedes spp in some areas is likely linked to habitat and human activity.13 The presence of vegetable gardens and bamboo, and the inappropriate disposal of containers and used car tyres in Kimbanseke, Ndjili, and Kingabwa, as well as the tree-covered land of Kimwenza, constitute a favourable habitat for Aedes mosquito multiplication. A report from Pakistan demonstrated that the used car tyre trade had contributed to the re-emergence of Aedes aegypti, along with the emergence of diseases such as dengue.14

Kimbanseke, Ndjili, and Kingabwa, as well as the tree-covered land of Kimwenza, constitute a favourable habitat for Aedes mosquito multiplication. A report from Pakistan demonstrated that the used car tyre trade had contributed to the re-emergence of Aedes aegypti, along with the emergence of diseases such as dengue.14 The lack of a sewage drainage system and the uncontrolled population growth in suburban areas of Kinshasa, as well as in its peri-urban areas, has increased the level of environmental pollution. This could explain the abundance of Culex spp in Ngaba, where there is no proper waste collection and disposal or sewage drainage system, increasing the level of pollution and providing a suitable breeding habitat. These observations corroborate the findings of other studies performed in Cameroon15 and Ivory Coast,16 which showed that the strong presence of Culex quinquefasciatus is associated with a polluted environment and that it can be considered as a biological marker of urbanization.

d providing a suitable breeding habitat. These observations corroborate the findings of other studies performed in Cameroon15 and Ivory Coast,16 which showed that the strong presence of Culex quinquefasciatus is associated with a polluted environment and that it can be considered as a biological marker of urbanization. The present study provided an opportunity to report the presence of certain arboviruses in Kinshasa, of which several have been reported previously.6, 7 However, this appears to be the first study to report the presence of Rift valley fever virus in Kinshasa. In general, the presence of most of the arboviruses goes unnoticed in malaria-endemic regions, where febrile illnesses are frequently considered to be malaria because of the similarity in their clinical presentation and lack of appropriate laboratory diagnostic capacity for differential diagnosis.17, 18 The lack of appropriate diagnosis and medical care leads to high mortality and huge economic losses, resulting in worsening poverty in Africa. Arboviral infections can most effectively be controlled through the use of vaccines. However, this is limited by a lack of registered vaccines for the majority of circulating arboviruses in Africa, except for vaccines against yellow fever virus for humans and Rift valley fever virus for livestock.10

poverty in Africa. Arboviral infections can most effectively be controlled through the use of vaccines. However, this is limited by a lack of registered vaccines for the majority of circulating arboviruses in Africa, except for vaccines against yellow fever virus for humans and Rift valley fever virus for livestock.10 The high number of arbovirus-positive pools reported in this study (12 from a total of 29 pools), showing the arboviruses circulating in Kinshasa, is alarming. It is proposed that further work to screen for arboviruses in Kinshasa, especially in the human population, is performed. Surveillance programmes designed to monitor virus activity in vectors provide a system for mapping disease distribution and information needed not only to assess risk, but also to identify vector species for targeted control.19

ork to screen for arboviruses in Kinshasa, especially in the human population, is performed. Surveillance programmes designed to monitor virus activity in vectors provide a system for mapping disease distribution and information needed not only to assess risk, but also to identify vector species for targeted control.19 The presence o’nyong’nyong and Rift valley fever viruses in mosquitoes circulating in Kinshasa may reflect a lack of previous arbovirus surveillance, but could also be a sign of the emergence of these arboviruses. The absence of dengue in screened mosquitoes could also be due to the small sample size of this study. However, it has been shown that A. aegypti originating from Central Africa are less competent in transmitting dengue virus compared to those circulating in East Africa and other locations of the world.20 Field studies from Central Africa conducted in Gabon and Cameroon have shown a lower susceptibility of A. aegypti to become infected with dengue virus and chikungunya virus than Aedes albopictus.21 This variability in vector competence between populations of A. aegypti from different geographical origins is intricately linked to the genetic heterogeneity of the species and could be influenced by environmental factors, human activity, insecticide treatment, and modalities of water supply storage.13 No pool was also positive for Yellow fever virus, while yellow fever outbreaks have recently been reported. This could be ascribed to the absence of outbreaks during the mosquito collection period or to the fact that mosquitoes could be feeding on uninfected people.

treatment, and modalities of water supply storage.13 No pool was also positive for Yellow fever virus, while yellow fever outbreaks have recently been reported. This could be ascribed to the absence of outbreaks during the mosquito collection period or to the fact that mosquitoes could be feeding on uninfected people. The present study findings also suggest that the environment is playing an important role in the occurrence of these arboviruses, as revealed by the frequency of positive pools in the areas surrounding the Ndjili River where pig farms are present and agriculture is practiced, as well as in Kimwenza where a bonobo sanctuary is located. This is supported by findings from Kenya, where the pastoral regions of Garissa and Marigat showed the highest frequency and diversity of arboviruses compared to other regions.10 In conclusion, mosquitoes from five selected areas of Kinshasa were collected and analyzed in this preliminary study, with the objective of detecting the presence of mosquito-borne viruses. Mosquito-borne viruses of public health importance are circulating within mosquitoes in Kinshasa. The health authorities in Kinshasa should conduct surveillance for arboviruses, to establish effective prevention measures. Healthcare providers must take account of arboviral diseases in the differential diagnosis of febrile illnesses. Future studies should sample more areas in the province of Kinshasa and other parts of DRC to detect areas at risk of arboviral disease. Conflict of interest The authors declare that there are no competing interests.

In conclusion, mosquitoes from five selected areas of Kinshasa were collected and analyzed in this preliminary study, with the objective of detecting the presence of mosquito-borne viruses. Mosquito-borne viruses of public health importance are circulating within mosquitoes in Kinshasa. The health authorities in Kinshasa should conduct surveillance for arboviruses, to establish effective prevention measures. Healthcare providers must take account of arboviral diseases in the differential diagnosis of febrile illnesses. Future studies should sample more areas in the province of Kinshasa and other parts of DRC to detect areas at risk of arboviral disease. Conflict of interest The authors declare that there are no competing interests. Author contributions KMM designed the study, participated in the field sample collection, performed laboratory and data analyses, and prepared the manuscript for publication. RW participated to the study design and data analyses, and read and critically revised the manuscript for publication. JPKM participated to the field sample collection, performed laboratory analyses, and read and critically revised the manuscript for publication. JKZ participated in the field sample collection, performed the data analyses, and read and revised the manuscript for publication. FS performed laboratory analyses and read and revised the manuscript for publication. TLB supervised the field sample collections and read and critically revised the manuscript for publication. MNA read and critically revised the manuscript for publication. GM participated in designing field and laboratory studies, supervised the laboratory analyses, performed data analyses, and read and critically revised the manuscript for publication.

lections and read and critically revised the manuscript for publication. MNA read and critically revised the manuscript for publication. GM participated in designing field and laboratory studies, supervised the laboratory analyses, performed data analyses, and read and critically revised the manuscript for publication. Acknowledgements This study was supported by a grant from the Wellcome Trust (grant WT 087546MA) to the Southern Africa Centre for Infectious Diseases Surveillance (SACIDS), Sokoine University of Agriculture (SUA), Morogoro, Tanzania. MKM was supported by a scholarship from the Wellcome Trust (grant WT 087546MA) to SACIDS. The authors are grateful to Mr Carol Mavanga and Dr Tresor Zola for their assistance during the field sample collection. They are also grateful to Ms Mariam R. Makange of the College of Veterinary and Medical Sciences at SUA for her technical assistance during molecular analyses. Finally, the authors are grateful to Mr Seth Irish from CDC Atlanta for his critical revision of the manuscript. Figure 1 Map of Kinshasa showing the mosquito sampling locations. Adult mosquitoes were collected from Kimwenza (Mount Ngafula municipality), Kingabwa (Limete municipality), Area 8 and 9 (Ndjili municipality), Nsanga and Mokali (Kimbanseke municipality), and Ngaba market (Ngaba municipality). The insert shows the location of Kinshasa City in the Democratic Republic of the Congo.

ons. Adult mosquitoes were collected from Kimwenza (Mount Ngafula municipality), Kingabwa (Limete municipality), Area 8 and 9 (Ndjili municipality), Nsanga and Mokali (Kimbanseke municipality), and Ngaba market (Ngaba municipality). The insert shows the location of Kinshasa City in the Democratic Republic of the Congo. Figure 1Figure 2 Detection of chikungunya and o’nyong’nyong viruses in female mosquito pools collected in Kinshasa. Amplification of chikungunya and o’nyong’nyong viruses using reverse transcription PCR (RT-PCR) targeting the 5′ non-translated region. The expected PCR product size for chikungunya was 98 base pairs, while it was 80 base pairs for o’nyong’nyong virus. NC: negative control; LD: ladder. Lane numbers indicate the pool identity. Figure 2Table 1 Different species of mosquitoes and their abundance at each of the sampling sites in Kinshasa, Democratic Republic of the Congo (DRC) Table 1Genus Kimwenza Kingabwa Kimbanseke Ndjili Ngaba Total Aedes 672 304 349 554 107 1986 Culex 118 80 98 77 258 631 Anopheles 97 23 71 64 28 283 Mansonia 4 6 10 2 0 22 Total 891 413 528 697 393 2922 Table 2 Mosquito-borne viruses screened using reverse transcription PCR (RT-PCR) in adult female mosquito pools collected from different locations within Kinshasa, Democratic Republic of the Congoa.

54 107 1986 Culex 118 80 98 77 258 631 Anopheles 97 23 71 64 28 283 Mansonia 4 6 10 2 0 22 Total 891 413 528 697 393 2922 Table 2 Mosquito-borne viruses screened using reverse transcription PCR (RT-PCR) in adult female mosquito pools collected from different locations within Kinshasa, Democratic Republic of the Congoa. Table 2Location Mosquito genus Flavivirus Alphavirus Bunyaviridae CHIKV ONNV YFV RVFV DENV Kimbanseke Culex – – – – Aedes ± ± ± – Aedes – – – – Anopheles – + + – + – – Aedes ± ± – – Kingabwa Aedes – – – – Aedes – – – – Culex ± – – – Aedes – – – – Ndjili Aedes + + + + – – + – Aedes – – – – – Aedes – – – Anopheles – – – – Culex – – – – Aedes – – – – Aedes + + + + – – + – Kimwenza Anopheles – – – – Aedes – – – – Aedes – – – – Culex – – – – Aedes – – – – Aedes – – – – Aedes – – – – Aedes ± ± ± – Aedes ± ± – – Ngaba Aedes ± ± – – Aedes – – – – Culex – – – – Culex – – – – Controlb Dengue + – – – – – – + RNase-free water – – – – – – – – CHIKV, chikungunya virus; ONNV, o’nyong’nyong virus; YFV, yellow fever virus; RVFV, Rift Valley fever virus; DENV, dengue virus. a +, strongly positive reaction; −, negative reaction; ±, weakly positive reaction. b Dengue was used as the positive control, while RNase-free water was used as the negative control.

Background Vector control strategies including universal coverage of long-lasting insecticidal nets (LLINs) during the last decade and a half have significantly contributed to overall reduction in global malaria morbidity by 30%.1 In India, during the last decade, malaria morbidity has been reduced by 45% from an estimated two million reported malaria cases in year 2000 to 1·1 million cases in 2015 due to sustained control efforts.2 Of the remaining malaria cases, more than 80% are contributed by tribal forested areas of ten states including Chhattisgarh.3 Most malaria endemic countries, including India, have expanded their focus from malaria control to elimination, and India has recently launched a framework for a national level malaria elimination programme 2016–2030. This plan envisages scaling up of existing interventions, appropriate vector control measures, capacity building and strengthening drug compliance among malaria positive cases.4 At the time when the malaria control program is focused on targeting malaria vectors through indoor residual spraying (IRS) of insecticide and LLINs, there is a strong need to target malaria parasites and sub clinical carriers to interrupt malaria transmission in endemic areas.5, 6, 7

nce among malaria positive cases.4 At the time when the malaria control program is focused on targeting malaria vectors through indoor residual spraying (IRS) of insecticide and LLINs, there is a strong need to target malaria parasites and sub clinical carriers to interrupt malaria transmission in endemic areas.5, 6, 7 Under the existing health care system, malaria positive cases are captured through active and passive fever surveillance mechanisms. Persons with confirmed blood parasitemia through either microscopy or rapid diagnostic test (RDT) receive anti-malarial treatment. However, a sizeable proportion of the population has been observed to harbour malaria parasites without presentation of any symptom. Such cases are not captured through the routine surveillance system. As carriers, they pose a serious challenge to efforts towards malaria elimination.8, 9 Several studies from malaria endemic countries have reported that irrespective of the intensity of transmission, a large proportion of malarial infections remain asymptomatic and contribute about 20–50% of all malaria transmission.10, 11, 12, 13 At high levels of gametocyte parasitemia, subclinical carriers become the parasite reservoir14, 15 and contribute to persistent transmission of malaria.5, 16 Subclinical infections, also known as asymptomatic or sub-patent infections, have no standard definition, and different studies have used their own definition with subtle differences.9

tocyte parasitemia, subclinical carriers become the parasite reservoir14, 15 and contribute to persistent transmission of malaria.5, 16 Subclinical infections, also known as asymptomatic or sub-patent infections, have no standard definition, and different studies have used their own definition with subtle differences.9 In endemic areas asymptomatic parasitemia may confer partial immunity on repeated exposure and may provide protection against clinical manifestations.17, 18 Malaria parasitemia thus remains asymptomatic and acts as a source for residual transmission. However, the clinical consequences of subclinical malaria are not fully understood as they vary with different ecological, epidemiological and environmental conditions.19 There is a dearth of Indian studies evaluating impact of LLIN use on subclinical malarial infection and malaria transmission in pyrethroid resistant areas. Here, we present the results of a study on the impact of community-wide use of LLIN on the prevalence of subclinical malaria in a cohort of children under age 14 years in the presence of pyrethroid resistance in malaria vectors on the protective effectiveness of LLINs. Methods The study was carried out during the post-monsoon seasons (August − December) of 2014 and 2015 among a cohort of children living in tribal forested villages in southern Chhattisgarh state, India.

There is a dearth of Indian studies evaluating impact of LLIN use on subclinical malarial infection and malaria transmission in pyrethroid resistant areas. Here, we present the results of a study on the impact of community-wide use of LLIN on the prevalence of subclinical malaria in a cohort of children under age 14 years in the presence of pyrethroid resistance in malaria vectors on the protective effectiveness of LLINs. Methods The study was carried out during the post-monsoon seasons (August − December) of 2014 and 2015 among a cohort of children living in tribal forested villages in southern Chhattisgarh state, India. Study area Keshkal (20° 5' 1 N and 81° 35' 12 E), a sub-district of Kondagaon district, has a population of about 90,000 living in 124 villages. A basic health care facility is available through one Community Health Centre (CHC) situated at block headquarter Keshkal, as well as four Primary Health Centres (PHCs), each catering to a population of about 20000–25000. Malaria transmission takes place mainly during the rainy season (June − October). Anopheles culicifacies is the principal vector in the region.20 For the past two decades, synthetic pyrethroid insecticide alpha-cypermethrin (@ 25 mg/m2 and twice a year) has been used in IRS as a major vector control measure by the state health department in the study area. With a consideration on year-round accessibility, 80 clusters were selected for the study. In 2014, 31,000 LLINs were distributed in collaboration with the state health department to cover a population of nearly 74,000 people living in these 80 clusters. In the remaining non-study clusters, the state health department has distributed LLINs covering nearly 16,000 population.

sters were selected for the study. In 2014, 31,000 LLINs were distributed in collaboration with the state health department to cover a population of nearly 74,000 people living in these 80 clusters. In the remaining non-study clusters, the state health department has distributed LLINs covering nearly 16,000 population. Enrolment of children and baseline data collection A cohort of 6582 children aged under 14 years was enrolled (60–80 children/cluster) after reading out the purpose of the study and obtaining written informed consent from the parent/guardian. Peripheral blood smear of all cohort children was prepared for microscopic confirmation of parasitemia to assess parasite prevalence before inclusion. Axial temperature of all study children at the time of survey was recorded with a digital thermometer (Dr. Diaz, Hemodiaz Life sciences Pvt. Ltd, India). Blood smears were stained with Geimsa stain and 100 microscopic fields of thick smear were examined at 1000 x magnification to detect malaria parasites. Parasite density (parasites/μl blood) was counted against 200 White Blood cells (WBCs) considering the average of 8000 WBC/μl. All slide positive malaria cases were treated with anti-malarial drugs according to national drug policy [Plasmodium falciparum (Pf): artesunate + sulfadoxine − pyrimethmine + primaquine; P. vivax(Pv):chloroquine + primaquine]. Follow up slides were prepared seven days after medication to ensure clearance of asexual parasites from the peripheral blood.

ere treated with anti-malarial drugs according to national drug policy [Plasmodium falciparum (Pf): artesunate + sulfadoxine − pyrimethmine + primaquine; P. vivax(Pv):chloroquine + primaquine]. Follow up slides were prepared seven days after medication to ensure clearance of asexual parasites from the peripheral blood. Case definition A slide positive case (presence of asexual parasitemia) with no symptoms of fever (axial temperature ≥ 99·5 °F/37·5 °C) was considered as a case of subclinical malaria. Active case detection For routine fever surveillance, 30 malaria surveillance workers were recruited from different clusters (villages) and trained in malaria surveillance and treatment activities. To assist them in malaria surveillance in houses of cohort children, 124 female Community Health Volunteers i.e. mitanins, were deployed. All cohort children were visited once in a fortnight. A follow up survey of all cohort children was carried out after six months of enrolment. Axial temperature was recorded using a digital thermometer. Self-reported history of fever during the preceding week, previous night and at the time of blood smear collection was recorded. Information on reported use of LLINs during the previous night was also recorded. However, in the follow up survey, ∼10% of children were lost to follow up due mainly to their exiting the study area to pursue higher studies elsewhere.

receding week, previous night and at the time of blood smear collection was recorded. Information on reported use of LLINs during the previous night was also recorded. However, in the follow up survey, ∼10% of children were lost to follow up due mainly to their exiting the study area to pursue higher studies elsewhere. Insecticide resistance assessment In the year 2015, wild caught full-fed An. Culicifacies adult females caught from each cluster (∼100 females per test) were exposed to deltamethrin 0·05% treated papers following WHO test procedures to determine susceptibility status (WHO, 2013). Based on median mortality the clusters were stratified into low resistance (≥84% mortality) and high resistance (<84% mortality) clusters. Thus, 3,249 and 3,333 cohort children were allocated to low resistance and high resistance clusters respectively. Statistical analysis Data were entered in EpiData version 3·1 software. All entries were double checked and further cleaned for error and analysed using IBM SPSS version 20 (IBM Corp, Armonk, NY). Continuous variables were described as mean and standard deviation (M ± SD), and categorical variables were described in percentages. To assess the reduction in prevalence of malaria and subclinical malaria cases in consecutive years, the data for both surveys were combined into one data set and included ‘survey year’ as a fixed effect in the generalised estimated equation (GEE) model together with all the other variables. Reduction in malaria from one survey to the next, adjusted for all the other variables in the model were analyzed.

cutive years, the data for both surveys were combined into one data set and included ‘survey year’ as a fixed effect in the generalised estimated equation (GEE) model together with all the other variables. Reduction in malaria from one survey to the next, adjusted for all the other variables in the model were analyzed. Effect of LLIN use between cluster resistance status was determined in a separate stratified analysis, and then overall cross level interaction between the two was analyzed by adding interaction term in the GEE model to assess impact of insecticide resistance on LLIN effectiveness. Univariate and multivariate analyses were carried out using a generalised estimated equation (GEE) regression model to explore the relation between subclinical malaria and exposure variables. GEE analysis was chosen so that within-cluster correlation can be adjusted and possible interaction of village resistance status can be taken into account during modelling. Response variable presence of subclinical malaria has binary events (Yes/No). Binomial distribution with logit link function was selected. Exchangeable correlation structure was chosen for the model. In the model, explanatory variables that included age group, gender, and last night LLIN use were added and village insecticide status was taken as main effect. Study cluster variable was taken at the subject level. The results of GEE regression analyses were stated as unadjusted and adjusted odds ratio (OR) with 95% confidence interval (CI) and associated P-value.

ed age group, gender, and last night LLIN use were added and village insecticide status was taken as main effect. Study cluster variable was taken at the subject level. The results of GEE regression analyses were stated as unadjusted and adjusted odds ratio (OR) with 95% confidence interval (CI) and associated P-value. Informed consent and ethical clearance The purpose of the study was described in the local language to all study participants, and study related queries such as routine surveillance and blood slide collection procedures etc. were addressed. A written informed consent was collected from the parents or guardians of the enrolled children. This study was undertaken as a part of a WHO-coordinated multi-country project and ethical clearance was obtained from the Institutional Ethics Committee (ECR/NIMR/EC/2010/75). Results Demographic and clinical characteristics Between September and November 2014, 6582 children (Male: Female − 1:1·03) were enrolled from 80 study clusters. The mean age of the study cohort was 6·3 years (SD- 3·4); children of age group 5–9 years constituted 40% of the study population followed by 2–4 years (28·9%). Of all children, 114(1·7%) and 196 (3·3%) reported with a history of fever (i.e., the previous night or at the time of survey) in the baseline and follow up surveys respectively. Detailed demographic and clinical description of the study population is given in Table 1.

he study population followed by 2–4 years (28·9%). Of all children, 114(1·7%) and 196 (3·3%) reported with a history of fever (i.e., the previous night or at the time of survey) in the baseline and follow up surveys respectively. Detailed demographic and clinical description of the study population is given in Table 1. In the baseline survey, peripheral blood samples from a total of 490/6582 (7·4%) children had malarial parasites (Table 2). A total of 398 (81·2%) microscopy positive children presenting without any symptoms at the time of survey had either Pf (331) or Pv (67) parasitemia. One child had a subclinical infection with P. malariae. In all, 91 microscopy positive children (1·4%) had malarial infection (Pf = 87; Pv = 4) associated with clinical symptoms. Among Pf infections, asexual parasitemia was higher (GMD 1680 parasites/μl, range 40–508880) in symptomatic cases as compared to asymptomatic cases (GMD 717 parasites/μl, range 40–46760). Similarly, sexual (gametocyte) parasitemia in symptomatic Pf infections was nearly two times higher than in asymptomatic children (GMD 613 vs 303 parasite/μl). In the follow up survey carried out during 2015, blood samples of 5862 children were collected, of which 82 (1·4%) were positive for malaria parasites. Among the positives, 46 (56·1%) and 19 (23·2%) children had subclinical Pf and Pv infections, respectively.

In the baseline survey, peripheral blood samples from a total of 490/6582 (7·4%) children had malarial parasites (Table 2). A total of 398 (81·2%) microscopy positive children presenting without any symptoms at the time of survey had either Pf (331) or Pv (67) parasitemia. One child had a subclinical infection with P. malariae. In all, 91 microscopy positive children (1·4%) had malarial infection (Pf = 87; Pv = 4) associated with clinical symptoms. Among Pf infections, asexual parasitemia was higher (GMD 1680 parasites/μl, range 40–508880) in symptomatic cases as compared to asymptomatic cases (GMD 717 parasites/μl, range 40–46760). Similarly, sexual (gametocyte) parasitemia in symptomatic Pf infections was nearly two times higher than in asymptomatic children (GMD 613 vs 303 parasite/μl). In the follow up survey carried out during 2015, blood samples of 5862 children were collected, of which 82 (1·4%) were positive for malaria parasites. Among the positives, 46 (56·1%) and 19 (23·2%) children had subclinical Pf and Pv infections, respectively. Comparative analysis of subclinical malaria and total malaria cases in baseline and follow up survey The prevalence of malaria was significantly reduced during the follow-up survey in August 2015 as compared to the baseline survey in November 2014 (SPR = 7·4% vs 1·4%, p < 0·001) after adjusting other factors such as gender and age groups. Similarly, significant reduction was observed in the prevalence of subclinical malaria in children in baseline and follow up surveys (SPR = 6·1% vs. 1%, p < 0·001) keeping other variables (gender and age group) constant. (Table 3).

= 7·4% vs 1·4%, p < 0·001) after adjusting other factors such as gender and age groups. Similarly, significant reduction was observed in the prevalence of subclinical malaria in children in baseline and follow up surveys (SPR = 6·1% vs. 1%, p < 0·001) keeping other variables (gender and age group) constant. (Table 3). Effect of insecticide resistance on protective efficacy of LLIN Mean (range) mortality rate of An. culicifacies to pyrethroid in high and low resistance clusters was 71·8% (53·0–83·6) and 93·8% (84–100) respectively. Malaria positive cases both clinical and subclinical were stratified based on susceptibility test data of An. culicifacies to deltamethrin of year 2015. Clusters with high deltamethrin resistance An. culicifacies showed almost equal numbers of subclinical malaria cases compared to low resistance clusters (1% vs. 1·4%). Overall, no interaction between the LLIN uses and insecticide resistance of study clusters was observed (Adjusted OR,95%CI- 0.58, 0.23-1.4; p = 0.21). (Table 4).

with high deltamethrin resistance An. culicifacies showed almost equal numbers of subclinical malaria cases compared to low resistance clusters (1% vs. 1·4%). Overall, no interaction between the LLIN uses and insecticide resistance of study clusters was observed (Adjusted OR,95%CI- 0.58, 0.23-1.4; p = 0.21). (Table 4). Impact of LLIN uses and other explanatory factors associated with subclinical malaria Table 5 shows results of the GEE univariate and multivariable analysis, which demonstrated that last night use or non-use of LLIN was one of the significant independently associated factors (OR = 0·25, 95% CI = 0·11, 0·58, p = 0·001), with non-use associated with higher prevalence of subclinical malaria. However, malarial infection frequency was nearly similar in both the age groups, 0–4 and 5–14 years (0·9% vs. 1·2%, p = 0·297) and between genders (1·2% vs. 1%, p = 0·811) with no significant association with subclinical malaria identified.

0·001), with non-use associated with higher prevalence of subclinical malaria. However, malarial infection frequency was nearly similar in both the age groups, 0–4 and 5–14 years (0·9% vs. 1·2%, p = 0·297) and between genders (1·2% vs. 1%, p = 0·811) with no significant association with subclinical malaria identified. Discussion Identification and targeting of subclinical malaria cases is one of the key challenges for countries aiming for malaria elimination.9 Children under the age of five years and between 5–14 years old are most vulnerable to malaria infections. Malaria prevalence among these age groups and LLIN usage the previous night are among the few key indicators used to measure outcomes of malaria intervention programs.21 However, asymptomatic malaria cases have also been noted as a suitable indicator to assess the success of programs.22 Systematic reviews have clearly suggested that LLINs, a definitive component of malaria control programs in most settings, if used properly, have significant impact in reducing malaria incidence in endemic communities.23, 24 However, the impact of LLIN use on subclinical malarial infections and residual transmission has largely remained unknown.

arly suggested that LLINs, a definitive component of malaria control programs in most settings, if used properly, have significant impact in reducing malaria incidence in endemic communities.23, 24 However, the impact of LLIN use on subclinical malarial infections and residual transmission has largely remained unknown. The findings of the baseline survey (2014) showed that overall prevalence of subclinical malaria infection was 6·1%, constituting more than three fourths of total slide positive cases (81·3%) among the enrolled children with predominance of P. falciparum infections (67·6%). Studies from different parts of the world have reported that subclinical malaria burden could be 4–5 times of total reported malaria cases,10 even in low transmission areas.25 In India, various sporadic cross-sectional studies have suggested that asymptomatic parasitemia prevalence could range from 2·9% to 8·4% in the transmission season.8, 26, 27, 28 Thus, it can further be speculated that a substantial proportion of subclinical malarial infection may exist, which routinely remains undetected in the existing surveillance system.

nal studies have suggested that asymptomatic parasitemia prevalence could range from 2·9% to 8·4% in the transmission season.8, 26, 27, 28 Thus, it can further be speculated that a substantial proportion of subclinical malarial infection may exist, which routinely remains undetected in the existing surveillance system. An active case surveillance (ACS) system to monitor the fever cases and to ensure regular LLIN use in the selected houses showed that LLIN use among cohort children (under 14 years old) in the study area was 94·8%, which was comparatively higher than reported in other studies.29 This higher LLIN use can be attributed to regular fortnightly surveillance and concurrent health education by field staff during their visits. Another plausible explanation is that regular visits of health workers led to positive response bias among the respondents. However, successive attempts were made to crosscheck the proxy indicators (hanging practices of bed nets) as well as house-to-house health education and awareness campaigns to improve LLIN usage30 since knowledge about malaria transmission and educational level of head of the household is also an important predictive factors for LLIN utilization.29

ere made to crosscheck the proxy indicators (hanging practices of bed nets) as well as house-to-house health education and awareness campaigns to improve LLIN usage30 since knowledge about malaria transmission and educational level of head of the household is also an important predictive factors for LLIN utilization.29 From this study, it can be stated that use of LLINs was equally effective in protecting the cohort from clinical malaria and subclinical malarial infections. A possible hypothesis for the protective effect of LLIN against subclinical infection is that there was interruption of transmission at the household level. It has been observed that if malaria infection were diagnosed from a member of one household there would be increased chance of presence of subclinical infection in other members of the family.31 Secondly, acquired immunity among the children as well as adults11 can also be a plausible reason that these asymptomatic carriers can act as source of infection at household level.

osed from a member of one household there would be increased chance of presence of subclinical infection in other members of the family.31 Secondly, acquired immunity among the children as well as adults11 can also be a plausible reason that these asymptomatic carriers can act as source of infection at household level. Malaria incidence and subclinical malaria prevalence were significantly reduced by 84% in the follow up survey relative to the baseline, which was perhaps attributable to adequate LLIN coverage and high usage rates by the cohort children. Nonetheless, ACS also substantially contributed to the control of malarial infection in the study population. A study from Kenya showed a similar decrease in subclinical malaria due to regular ACS.32 In the Asia Pacific region, different countries have implemented variants of active case surveillance systems to identify and reduce parasitic carriers among reservoirs. Case detection and reactive active case detection were among the approaches taken, although better and more effective methods are still required.33

32 In the Asia Pacific region, different countries have implemented variants of active case surveillance systems to identify and reduce parasitic carriers among reservoirs. Case detection and reactive active case detection were among the approaches taken, although better and more effective methods are still required.33 Parasite density, age and season are essential co-variates in risk of developing fever.34 However, on the contrary, no major difference in age and parasite density among subclinical and total malaria cases was observed in the present study. Asymptomatic parasitemia is usually higher in the dry season as compared to the rainy season35 and the presence of gametocyte density in the dry season probably acts as a source of infection in the next cycle of malaria transmission.36 In the Keshkal sub district, malaria transmission is seasonal and malaria incidence usually peaks in the post monsoon season. It has also been reported that increased breeding potential and subsequently vector density during monsoon and post-monsoon seasons directly affect the malaria transmission in low and meso-endemic regions.29, 34 In view of this, to control for this confounding effect due to seasonal variation, both the surveys were conducted in the same season in the consecutive years of 2014 and 2015. However, there might be a chance of year-to-year variation due to climatic factors.

malaria transmission in low and meso-endemic regions.29, 34 In view of this, to control for this confounding effect due to seasonal variation, both the surveys were conducted in the same season in the consecutive years of 2014 and 2015. However, there might be a chance of year-to-year variation due to climatic factors. The worldwide spread of insecticide resistance among malaria vectors has been well-established.21 It was assumed that pyrethroid resistance can adversely affect efforts of malaria control and elimination, though no concrete evidence is yet available although there have been a few studies37, 38 that suggest an association between insecticide resistance and failure of malaria interventions. However, under laboratory and hut trial conditions it has been shown that LLIN remains effective despite pyrethroid resistance in main malaria vector species.39 In India, occurrence of pyrethroid resistance has been reported in malaria endemic states including Chhattisgarh.20, 40 Susceptibility tests carried out in 2015 have confirmed pyrethroid resistance in An. Culicifacies in most of the study clusters (69) based on WHO cut off ≥98% of mortality rate. Stratified data analysis showed that the occurrence of malaria infection among LLIN users and non-users was similar and resistance status was not behaving as an effect modifier for LLIN usage and malaria cases. We speculate that this could be due to higher LLIN usage among the children and no loss of effectiveness in LLIN efficacy in protecting from malaria and subclinical malaria was observed.

g LLIN users and non-users was similar and resistance status was not behaving as an effect modifier for LLIN usage and malaria cases. We speculate that this could be due to higher LLIN usage among the children and no loss of effectiveness in LLIN efficacy in protecting from malaria and subclinical malaria was observed. Apart from seasonal dynamics of malaria transmission, residing in malaria hotspots at micro-geographical scale in low transmission areas is an independent predictive factor for malaria infections.41 Cluster wise data of both the surveys identified few independent clusters with comparatively higher number of malaria cases, which might be due to their proximity to breeding habitats. Such clusters may serve as a potential reservoir of subclinical cases. Targeting these subsets of the population with appropriate interventions could be cost-effective in clearing the residual parasite reservoir.42, 43 However, this needs to be further corroborated by geospatial assessment, and entomological studies. While this study has focused on the post-monsoon season, further assessment is required during the dry season, when LLIN utilization might reduce due to reduction in vector densities or hot climates that deters usage. Further follow up could delineate the effect in prevalence of subclinical malaria in the low transmission season as well.

s focused on the post-monsoon season, further assessment is required during the dry season, when LLIN utilization might reduce due to reduction in vector densities or hot climates that deters usage. Further follow up could delineate the effect in prevalence of subclinical malaria in the low transmission season as well. This study presents an ideal scenario with high LLIN use, and well-informed household and study findings could only be generalized to other tribal populated area within low transmission settings, if high LLIN coverage is maintained with a robust active surveillance activity in place. Conclusions There was a high proportion of subclinical malaria infections among children in the tribal inhabited malaria endemic area of the study, which may also be the case in areas with similar populations and malaria transmission intensities, and therefore may inform appropriate malaria elimination strategies. These observations indicated that high use of LLINs can significantly reduce the number of subclinical malaria cases, despite vectors having reduced susceptibility to the pyrethroid used in the LLINs. Therefore, high LLIN coverage and usage supported by a strong active case surveillance system would be a useful strategy in low transmission settings where parasite reservoirs exist as subclinical carriers. Conflict of Interest No competing interest declared

Conclusions There was a high proportion of subclinical malaria infections among children in the tribal inhabited malaria endemic area of the study, which may also be the case in areas with similar populations and malaria transmission intensities, and therefore may inform appropriate malaria elimination strategies. These observations indicated that high use of LLINs can significantly reduce the number of subclinical malaria cases, despite vectors having reduced susceptibility to the pyrethroid used in the LLINs. Therefore, high LLIN coverage and usage supported by a strong active case surveillance system would be a useful strategy in low transmission settings where parasite reservoirs exist as subclinical carriers. Conflict of Interest No competing interest declared Role of funding source Funding was provided by the Bill and Melinda Gates Foundation (Grant number OPP 1062754) which has no role in the planning, study design, data collection or write up. This research forms part of a multi-country study coordinated by the Global Malaria Programme of the World Health Organization. Ethics approval and consent to participate The purpose of the study was described in the local language to all study participants, study related queries such as routine surveillance, and blood slide collection procedures etc were addressed. A written informed consent was collected from the parents or guardians of the enrolled children. Ethical clearance was obtained from the Institutional Ethics Committee (ECR/NIMR/EC/2010/75).

study participants, study related queries such as routine surveillance, and blood slide collection procedures etc were addressed. A written informed consent was collected from the parents or guardians of the enrolled children. Ethical clearance was obtained from the Institutional Ethics Committee (ECR/NIMR/EC/2010/75). Acknowledgments We are grateful to the children enrolled in the study and their family members for cooperation and having consented to participate. We would like to thank female community health workers (Mitanin) and Malaria Surveillance Workers (MSW’s)for their excellent efforts in surveillance activity and their altruistic commitment towards community. We would also like to thank technical and laboratory staff of IIR Project Field Site at Kondagaon, NIMR Field Unit, Raipur and NIMR HQ, New Delhi for their support in undertaking field and laboratory activities. ☆ Location of study area: National Institute of Malaria Research (ICMR), IIR-WHO Project, Field Unit- Kondagaon, Chhattisgarh, India. Table 1 Demographic and clinical characteristics of study population (n = 6582).

Acknowledgments We are grateful to the children enrolled in the study and their family members for cooperation and having consented to participate. We would like to thank female community health workers (Mitanin) and Malaria Surveillance Workers (MSW’s)for their excellent efforts in surveillance activity and their altruistic commitment towards community. We would also like to thank technical and laboratory staff of IIR Project Field Site at Kondagaon, NIMR Field Unit, Raipur and NIMR HQ, New Delhi for their support in undertaking field and laboratory activities. ☆ Location of study area: National Institute of Malaria Research (ICMR), IIR-WHO Project, Field Unit- Kondagaon, Chhattisgarh, India. Table 1 Demographic and clinical characteristics of study population (n = 6582). Table 1Variable Category Percent (n) Age (in years, 2014) Mean (SD) 6·3 (3·4) <2 7·3 (480) Age group (in years, 2014) 2–4 28·9 (1899) 5–9 40·0 (2634) 10–14 Years 23·8 (1569) Gender Male 49·3 (3246) Insecticide resistance status (Deltamethrin 0·05%, 2015) Median mortality- 84% Low resistance clusters (Bioassay mortality ≥84%)(40) 50·6 (3333) High resistance clusters (Bioassay mortality <84%) (40) 48·7 (3249) Baseline 2014 (n = 6582) Follow up 2015 (n = 5862) Slide positivity rate Microscopy positive 7·4 (490) 1·4 (82) Clinical malaria (with fever) Yes 1·4 (91) 0·26 (17) Subclinical Malaria (without fever) Yes 6·1 (399) 1 (65) Previous night fever history Yes 1·7 (114) 3·3 (196) Loss to follow up – 10·9 (720/6582) Table 2 Host and parasite characteristics of study cohort during inclusion and follow-up.

opy positive 7·4 (490) 1·4 (82) Clinical malaria (with fever) Yes 1·4 (91) 0·26 (17) Subclinical Malaria (without fever) Yes 6·1 (399) 1 (65) Previous night fever history Yes 1·7 (114) 3·3 (196) Loss to follow up – 10·9 (720/6582) Table 2 Host and parasite characteristics of study cohort during inclusion and follow-up. Table 2Variable Asymptomatic Symptomatic Total malaria positive P. falciparum P. vivax P. falciparum P. vivax Baseline survey (n = 490 positives/6582 slides; SPR = 7·4%) Number (%) 331 (67·6) 67 (13·7) 87 (17·8) 04 (0·8) 490# Age, Mean (years; range) 6·6 (1–12) 6·5 (1–12) 6·6 (1–12) 6·7 (4–11) 6·6 (1–12) Males, n (%) 159 (48·0) 36 (53·7) 41 (47·1) 2 (50) 238 (48·6) Asexual parasite density, geometric mean (per μl, range) 717 415 1680 247 (40–46760) (40–36880) (40–508880) (80–760) Sexual parasite density, geometric mean (per μl, range) 99 303 157 613 (40–1720) (40–49040) (40–1440) (40–31880) Follow up survey (n = 82 positives/5862 slides; SPR = 1·4%) Number (%) 46 (56·1) 19 (23·2) 14 (17·1) 03 (3·7) 82 (100) Age, Mean (years; range) 7 6 7 3 7 (1–12) (1–12) (3–12) (2–4) (1–12) Males, n (%) 21 (45·7) 12 (63·2) 05 (26·3) 01 (33·3) 39 (47·5) Asexual parasite density, geometric mean (μl, range) 892 164 8402 1560 (40–340400) (40–1480) (80–628440) (1560) Sexual parasite density, geometric mean (μl, range) 86 320 Nil 1037 (40–440) (40–8200) (360–5160) # One case of P. Malariae was also identified and is included in the total malaria cases. Table 3 Comparison of subclinical malaria cases in baseline and follow up survey (n = 12444).

Follow up survey (n = 82 positives/5862 slides; SPR = 1·4%) Number (%) 46 (56·1) 19 (23·2) 14 (17·1) 03 (3·7) 82 (100) Age, Mean (years; range) 7 6 7 3 7 (1–12) (1–12) (3–12) (2–4) (1–12) Males, n (%) 21 (45·7) 12 (63·2) 05 (26·3) 01 (33·3) 39 (47·5) Asexual parasite density, geometric mean (μl, range) 892 164 8402 1560 (40–340400) (40–1480) (80–628440) (1560) Sexual parasite density, geometric mean (μl, range) 86 320 Nil 1037 (40–440) (40–8200) (360–5160) # One case of P. Malariae was also identified and is included in the total malaria cases. Table 3 Comparison of subclinical malaria cases in baseline and follow up survey (n = 12444). Table 3Parameters Category Positive (%) Total (N = 12444) Unadjusted Odds ratio (95%CI) P value Adjusted OR (95%CI) P value Survey Year 2015 65(1) 5862 0·196 (0·18–0·26) <0·001 0·194 (0·15–0·26) <0·001 2014 399(6·1) 6582 Gender Male 236(3·7) 6111 1·06 (0·86–1·31) 0·594 1·06 (0·85–1·32) 0·631 Female 228(3·7) 6333 Age groups 5–14 years 325(3·8) 8479 1·04 (0·87–1·3) 0·661 1·20 (0·98–1·54) 0·07 0–4 years 139(3·5) 3965 Subject level- study cluster (n = 80) Comparison of malaria cases in baseline and follow up survey(n = 12444) Survey Year 2015 490(7·4) 5862 0·203 (0·15–0·27) <0·001 0·20 (0·15–0·27) <0·001 2014 82(1·4) 6582 Gender Male 277(4·5) 6111 1·02 (0·86–1·21) 0·810 1·02 (0·85–1·22) 0·87 Female 295(4·7) 6333 Age groups 5–14 years 396(4·7) 8479 1·01 (0·86–1·2) 0·915 1·2 (0·97–1·38) 0·11 0–4 years 176(4·4) 3965 Subject level- study cluster(n = 80) Table 4 Malaria cases stratified according to insecticide resistance status in An. Culicifacies (n = 5862).

) 6111 1·02 (0·86–1·21) 0·810 1·02 (0·85–1·22) 0·87 Female 295(4·7) 6333 Age groups 5–14 years 396(4·7) 8479 1·01 (0·86–1·2) 0·915 1·2 (0·97–1·38) 0·11 0–4 years 176(4·4) 3965 Subject level- study cluster(n = 80) Table 4 Malaria cases stratified according to insecticide resistance status in An. Culicifacies (n = 5862). Table 4Last night LLIN usage vs. malaria in 2015 Insecticide resistance status Variable Category Microscopy Positive 2015 Stratified Odds ratio (OR), 95%CI P value Adjusted OR 95%CI, p value (%) Total Low resistance Last night LLIN use Yes 41 (1·4) 2840 0·30, 0·09–1·00 0·05 0·58 No 05 (4·7) 107 0·23–1·4 Total 46 (1·6) 2947 0·21 High resistance Last night LLIN use Yes 27 (1) 2718 0·21, 0·08–0·55 0·001 No 09 (4·6) 197 Total 36 (1·23) 2915 Last night LLIN usage vs. Subclinical malaria in 2015 Subclinical malaria 2015 Low resistance Last night LLIN use Yes 33 (1·2%) 2840 0·30, 0·08–1·1 0·07 0·56 No 04 (3·7%) 107 0·21–1·4, Total 37 (1·26%) 2947 0·22 High resistance Last night LLIN use Yes 21 (0·8%) 2718 0·21, 0·07–0·66 0·008 No 07 (3·6%) 197 Total 28 (1%) 2915 Table 5 Prevalence of subclinical malaria by use of LLINs, gender, age-group and insecticide resistance status of study clusters (n = 5862).

·1 0·07 0·56 No 04 (3·7%) 107 0·21–1·4, Total 37 (1·26%) 2947 0·22 High resistance Last night LLIN use Yes 21 (0·8%) 2718 0·21, 0·07–0·66 0·008 No 07 (3·6%) 197 Total 28 (1%) 2915 Table 5 Prevalence of subclinical malaria by use of LLINs, gender, age-group and insecticide resistance status of study clusters (n = 5862). Table 5Parameters Category Positive (%) Total (N = 5862) Unadjusted Odds ratio(95%CI) P value Adjusted OR (95%CI) P value Last night LLIN use Yes 54(1) 5558 0·26 (0·11–0·60) 0·002 0·25 (0·11–0·58) 0·001 No 11(3·6) 304 Gender Male 33(1·2) 2865 1·08 (0·6–1·95) 0·799 1·08 (0·6–1·94) 0·811 Female 32(1·1) 2997 Age groups 5–14 years 46(1·2) 3749 1·40 (0·67–2·73) 0·394 1·33 (0·66–2·65) 0·297 0–4 years 19(0·9) 2113 Cluster resistance status High 28(1) 2915 0·76 (0·30–1·98) 0·577 0·70 (0·27–1·81) 0·457 Low 37(1·3) 2947 Subject level- study cluster (n = 80) Overall prevalence of infection by use of LLINs, gender, age-group and insecticide resistance status of study clusters (n = 5862) Last night LLIN use Yes 68(1·2) 5558 0·26 (0·12–0·54) <0·001 0·25 (0·12–0·52) <0·001 No 14(4·6) 304 Gender Male 39(1·4) 2865 0·95 (0·56–1·59) 0·840 0·94 (0·56–1·59) 0·825 Female 43(1·4) 2997 Age groups 5–14 years 64(1·5) 3749 1·32 (0·74–2·38) 0·350 1·29 (0·73–2·29) 0·382 0–4 years 18(1·1) 2113 Clusters resistance status High 36(1·2) 2915 0·80 (0·30–2·07) 0·629 0·72 (0·28–1·88) 0·504 Low 46(1·6) 2947 Subject level- study cluster (n = 80)

Introduction Crimean-Congo hemorrhagic fever (CCHF) is an acute tick-borne viral infection and a major emerging infectious diseases threat. It affects a wide geographical area, centered in Eurasia including Turkey, Russia and Kazakhstan but is under-reported and diagnosis is often delayed. Fever, thrombocytopenia and hemorrhage are the characteristic clinical features, with supportive care forming the mainstay of treatment protocols, although ribavirin is utilized by some centers. Provision of blood product support and access to critical care interventions can improve outcomes, with reported case fatality rates (CFR) being 4–20%.1 The majority of cases of CCHF report a history of tick bite, but healthcare related transmission of CCHF is well reported, and occurs in both high and low-resource settings. Failure to recognize CCHF and as a result implement appropriate infection, prevention and control procedures results in significant nosocomial risk, especially in the context of critical care interventions.2, 3, 4 Retrospective analysis from Turkey has demonstrated that needle stick injuries are the most frequent cause of high risk exposures, followed by ‘splash’ exposures to mucous membranes and highlighted a potential benefit of ribavirin post-exposure prophylaxis.5

ecially in the context of critical care interventions.2, 3, 4 Retrospective analysis from Turkey has demonstrated that needle stick injuries are the most frequent cause of high risk exposures, followed by ‘splash’ exposures to mucous membranes and highlighted a potential benefit of ribavirin post-exposure prophylaxis.5 Although the immune system changes in pregnancy are not completely understood, pregnant woman may be more likely to acquire certain infectious diseases such as toxoplasmosis, and be more severely affected by others such as influenza and varicella.6 Viral hemorrhagic fevers such as Ebola virus disease and Lassa Fever are more severe in pregnancy,7, 8 and frequently result in spontaneous abortion with additional nosocomial risk. Although clinical and epidemiological CCHF data are increasingly reported, few data exist on CCHF in pregnancy.9, 10, 11 The mortality of CCHF disease in pregnant women appears to be higher than in the general population (up to 33%),10 and the severe course of CCHF in pregnant women may also increase risk of nosocomial infection in health care settings.2 In this study we aimed to systematically review the characteristics of CCHF cases in pregnancy, and to report an additional case series of 8 CCHF cases in pregnant women from Russia, Kazakhstan and Turkey.

Although the immune system changes in pregnancy are not completely understood, pregnant woman may be more likely to acquire certain infectious diseases such as toxoplasmosis, and be more severely affected by others such as influenza and varicella.6 Viral hemorrhagic fevers such as Ebola virus disease and Lassa Fever are more severe in pregnancy,7, 8 and frequently result in spontaneous abortion with additional nosocomial risk. Although clinical and epidemiological CCHF data are increasingly reported, few data exist on CCHF in pregnancy.9, 10, 11 The mortality of CCHF disease in pregnant women appears to be higher than in the general population (up to 33%),10 and the severe course of CCHF in pregnant women may also increase risk of nosocomial infection in health care settings.2 In this study we aimed to systematically review the characteristics of CCHF cases in pregnancy, and to report an additional case series of 8 CCHF cases in pregnant women from Russia, Kazakhstan and Turkey. Material and methods We planned and reported this systematic review in accordance with guidelines for performing and reporting systematic reviews and meta-analyses (PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses). We searched PubMED, Science Citation Index (SCI) and Scopus databases for English and foreign language studies published between January 1960 and June 2016. The keyword ‘Crimean Congo H(a)hemorrhagic Fever” was utilised then combined with “pregnancy”; ‘pregnant’ and ‘vertical’. We also searched reference sections; Google scholar and reviews for other studies. Statistical analyses were performed with the use of IBM SPSS Statistics for Windows; Version 24.0. Armonk; NY: IBM Corp.

eyword ‘Crimean Congo H(a)hemorrhagic Fever” was utilised then combined with “pregnancy”; ‘pregnant’ and ‘vertical’. We also searched reference sections; Google scholar and reviews for other studies. Statistical analyses were performed with the use of IBM SPSS Statistics for Windows; Version 24.0. Armonk; NY: IBM Corp. Study Selection Two reviewers (HL & IB) independently screened the titles and abstracts of all studies that were identified through database searches. Inclusion criteria were (1): report of a case of laboratory confirmed Crimean-Congo Hemorrhagic fever OR report of a clinical case with a direct epidemiological link to a laboratory confirmed case of Crimean-Congo hemorrhagic fever and (2); pregnancy. Full-text copies of potentially relevant studies were retrieved and reviewed independently (TF & IB), extracting data from each study meeting inclusion criteria. Standardised data was extracted from each case when available including date, age, gestation, laboratory diagnosis, outcome of mother and foetus, hemorrhagic manifestations and secondary cases.

evant studies were retrieved and reviewed independently (TF & IB), extracting data from each study meeting inclusion criteria. Standardised data was extracted from each case when available including date, age, gestation, laboratory diagnosis, outcome of mother and foetus, hemorrhagic manifestations and secondary cases. Results The initial search results identified the following number of records: Scopus 3275; PubMED 1205; and SCI 1042. After removal of duplicates 3507 records were combined with secondary search terms (Pregnancy: 101, pregnant:114, vertical:109,). An additional search by Google scholar identified 2 further reports. 20 full text articles were then retrieved, with the total number of 34 CCHF cases identified in pregnancy in 15 articles (Figure 1). An additional 8 cases were added from this case series from Russia (5 cases), Kazakhstan (2 cases) and Turkey (1 case), resulting in total of 42 cases of CCHF in pregnancy (Table 1). The first report was from a case in 1979 and the last case occurred in 2016, with cases from Turkey (14 cases), Iran (10 cases), Russia (6 cases), former Yugoslavia (4 cases), Iraq (3 cases), Kazakhstan (2 cases), Mauritania (2 cases) and Bulgaria (1 case). Gestation was reported in 29 cases, with 11 cases occurring in the first 20 weeks of pregnancy and 18 cases occurring in the second 20 weeks of pregnancy.

Turkey (14 cases), Iran (10 cases), Russia (6 cases), former Yugoslavia (4 cases), Iraq (3 cases), Kazakhstan (2 cases), Mauritania (2 cases) and Bulgaria (1 case). Gestation was reported in 29 cases, with 11 cases occurring in the first 20 weeks of pregnancy and 18 cases occurring in the second 20 weeks of pregnancy. Maternal and fetal/neonatal outcome was reported in 41 cases, with 14/41 cases (34%) reporting maternal death, and foetal/neonatal death occurring in 24/41 cases (58.5%). In the first 20 weeks of pregnancy there were 1/11 maternal deaths and 6/11 fetal deaths, and in weeks 20–40 of pregnancy there were 8/18 maternal deaths and 9/18 fetal/neonatal deaths. There was no statistically significant difference between maternal deaths in the 1st 20 weeks of pregnancy compared to weeks 20–40 (p = 0.096–Fisher’s exact test, two-tailed). There was no haemorrhage in 10/41 cases, with maternal and fetal survival occurring in all 10 cases. When maternal haemorrhage was reported (vaginal or other sites), there were 14/31 maternal deaths and 24/31 fetal/neonatal deaths. When compared to the non-haemorrhage group (0/10) differences in outcome were statistically significant (p = 0.009, and p < 0.0001 respectively, Fisher’s exact test, two-tailed). Nosocomial transmission from pregnant CCHF cases was reported in 6/37 cases, resulting in additional 38 cases.

d 24/31 fetal/neonatal deaths. When compared to the non-haemorrhage group (0/10) differences in outcome were statistically significant (p = 0.009, and p < 0.0001 respectively, Fisher’s exact test, two-tailed). Nosocomial transmission from pregnant CCHF cases was reported in 6/37 cases, resulting in additional 38 cases. CCHF case series Case 1 (Stavropol region, Russia, 2002) A 20-year-old pregnant woman (16-weeks gestation) was admitted to the infectious diseases hospital on the 3rd day of disease with fever (38,5 °C), headache, loss of appetite and a petechial rash on her lower limbs. She was an agriculture worker and had contact with animals, but denied any history of tick bite. Initial laboratory tests were: Hb 87 g/l, RBC 3.2 × 1012/L, WBC 3.4 × 109/L, PLT 112 × 109/L, APTT 46 secs. CCHF was suspected and confirmed by PCR on the 4th day of disease, with a positive ELISA result in the second week. Empirical antibiotics and supportive treatment were given (no Ribavirin) and 2 days later her condition improved, and temperature normalized. Laboratory tests improved (WBC 7.5 × 109/L, PLT 144 × 109/L) and she was discharged after a 16-day admission. At 38 weeks of pregnancy she gave birth to a healthy child.

nd week. Empirical antibiotics and supportive treatment were given (no Ribavirin) and 2 days later her condition improved, and temperature normalized. Laboratory tests improved (WBC 7.5 × 109/L, PLT 144 × 109/L) and she was discharged after a 16-day admission. At 38 weeks of pregnancy she gave birth to a healthy child. Case 2 (Stavropol region, Russia, 2003) A 19-year-old pregnant woman (38-weeks gestation) was admitted to the infectious diseases hospital with a one-day history of fever (up to 39 °C), headache, and myalgia. She lived in a rural area and several days previously reported removing ticks from her cow. CCHF was suspected and laboratory tests at admission were: Hb 95 g/L, RBC 3.2 × 1012/L, WBC 3.8 × 109/L, PLT 132 × 109/L and APTT 46 secs. Ribavirin, empiric antimicrobials and supportive treatment were commenced and 2–3 days later her condition improved, and her fever settled. Her laboratory tests improved by day 8 (HB 110 g/l, RBC 3.8 × 1012/L, WBC 4.2 × 109/L, PLT 153 × 109/L, APTT 44 secs) and she had no hemorrhagic manifestations. CCHF was confirmed by ELISA and on day 11 she was transferred to a maternity unit, where 3 days later she gave birth to a healthy baby.

and her fever settled. Her laboratory tests improved by day 8 (HB 110 g/l, RBC 3.8 × 1012/L, WBC 4.2 × 109/L, PLT 153 × 109/L, APTT 44 secs) and she had no hemorrhagic manifestations. CCHF was confirmed by ELISA and on day 11 she was transferred to a maternity unit, where 3 days later she gave birth to a healthy baby. Case 3 (Izobilnensky district, Stavropol region, Russia, 2004) A 17-year-old pregnant woman (30-weeks gestation) was admitted to the gynecology department with a 2-day history of high fever (>39 °C) and back pain. She lived in rural area (with suspected tick contact) and laboratory tests were: HB 102 g/L, RBC 3.5 × 1012/L, WBC 5.6 × 109/L, PLT 153 × 109/L, ESR 20 mm/h, APTT 43 secs, leukocyturia and proteinuria (2 g/l). Acute pyelonephritis was the initial diagnosis and anti-bacterial and supportive treatment was started, but did not result in any significant improvement in her clinical condition.

ests were: HB 102 g/L, RBC 3.5 × 1012/L, WBC 5.6 × 109/L, PLT 153 × 109/L, ESR 20 mm/h, APTT 43 secs, leukocyturia and proteinuria (2 g/l). Acute pyelonephritis was the initial diagnosis and anti-bacterial and supportive treatment was started, but did not result in any significant improvement in her clinical condition. On the 4th day of illness a hemorrhagic rash appeared on her lower limbs, with ecchymosis developing on the skin of her chest and abdomen. She became hypotensive (90/60 mmHg), with increased respiratory rate (24/min), and worsening of her laboratory parameters (Hb72 g/L, RBC 2.1 × 1012/L, WBC 5.6 × 109/L, Plt 32 × 109/L and APTT 48 secs). CCHF was suspected and confirmed by PCR (CCHF IgM negative) and she was transferred to the infectious diseases department. On the 5th day of illness there was onset of premature labour and a stillborn baby was delivered. This was further complicated by post-partum and gastrointestinal tract hemorrhage, development of acute respiratory distress syndrome and the patient died on the 6th day of disease. Ribavirin was not administered to the patient.

5th day of illness there was onset of premature labour and a stillborn baby was delivered. This was further complicated by post-partum and gastrointestinal tract hemorrhage, development of acute respiratory distress syndrome and the patient died on the 6th day of disease. Ribavirin was not administered to the patient. Case 4 (Malgobeksky district, Republic of Ingushetia, Russia, 2005) A 24-year-old pregnant physician (17-weeks gestation) presented with a 4-day history of fever (39 °C), thirst, abdominal pain, vomiting and cough. She had had previous occupational exposure during phlebotomy and processing of blood samples from a 75-year old patient (index patient), who died of massive gastrointestinal hemorrhage.12 Two daughters, who provided care at home also died. The pregnant physician was initially hospitalized in the gynecology department and on the 10th day of illness was transferred to a regional infectious department with suspected viral hepatitis. At this stage she had developed jaundice, ecchymosis at injection sites, a hemorrhagic rash, hepatomegaly, splenomegaly and a right-sided pneumonia. She was managed with antimicrobial therapy and supportive treatment, and was transferred to the National Scientific Research Institute of obstetrics and pediatrics in Moscow (NSRIOP) on the 12th day of disease. She remained febrile, and became critically unwell with respiratory distress and was transferred to the Intensive care unit (ICU) (Hb 60 g/L, PLT 400 × 109/L, WBC 10.7 × 109/L, ESR 40 mm/h, APTT 58 secs).

National Scientific Research Institute of obstetrics and pediatrics in Moscow (NSRIOP) on the 12th day of disease. She remained febrile, and became critically unwell with respiratory distress and was transferred to the Intensive care unit (ICU) (Hb 60 g/L, PLT 400 × 109/L, WBC 10.7 × 109/L, ESR 40 mm/h, APTT 58 secs). She was subsequently transferred to the ICU of National Institute of hematology in a comatose condition with a suspected hematological disease, but following infectious diseases review, CCHF was suspected and she was transferred to the ICU of infectious diseases hospital on the 16 day of disease. Progressive pneumonia, myocarditis, hepatic insufficiency, antenatal fetal death, and a disseminated intravascular coagulation (DIC)-syndrome were diagnosed at this point. A caesarean section was performed on the 18th day of illness and a dead fetus without hemorrhagic manifestations was extracted. The patient remained on a ventilator for several weeks, and after her condition improved she was extubated and then discharged from hospital 2 months later. CCHF RT-PCR from samples taken on day 15 were negative, but the diagnosis confirmed by ELISA with positive CCHF IgM and increasing titers of IgG. The patient was not treated with ribavirin.

ventilator for several weeks, and after her condition improved she was extubated and then discharged from hospital 2 months later. CCHF RT-PCR from samples taken on day 15 were negative, but the diagnosis confirmed by ELISA with positive CCHF IgM and increasing titers of IgG. The patient was not treated with ribavirin. Case 5 (Turkestan city, Southern-Kazakhstan region, Kazakhstan, 2009) A 23-year-old woman was re-admitted to a maternity unit with her 7-day-old newborn 3 days after discharge, with high fever and vaginal bleeding. Post-partum endometritis with secondary post-partum hemorrhage was suspected, and emergency hysterectomy was performed for severe blood loss. Intra-abdominal hemorrhage continued and despite two further laparotomies she died. A few days later her baby also died. CCHF was diagnosed in the woman based on post-mortem results, and it was suspected that she acquired this in the last week of pregnancy. Two clinicians who performed surgical interventions on patient, and the pediatrician who managed the newborn (without direct contact with the mother) also contracted CCHF and died. Diagnosis was confirmed was confirmed by immuno-histochemical analysis of samples at post-mortem. Two more HCWs who cared for woman also developed CCHF, but survived.

d surgical interventions on patient, and the pediatrician who managed the newborn (without direct contact with the mother) also contracted CCHF and died. Diagnosis was confirmed was confirmed by immuno-histochemical analysis of samples at post-mortem. Two more HCWs who cared for woman also developed CCHF, but survived. Case 6 (Turkestan city, Southern-Kazakhstan region, Kazakhstan, 2010) A 21-year-old pregnant woman (34-weeks gestation) was admitted to the maternity hospital with a 4-day history high fever (>39 °C), dizziness, thirst and anorexia. She also complained of left sided abdominal/flank pain, was hypotensive (BP 90/60) and no fetal heart beat was found. She lived in rural area with animal contact, but denied a history of tick bite. The initial diagnosis was acute pyelonephritis and antenatal death, but CCHF was then suspected due to the thrombocytopenia (PLT 23 × 109/L) in her initial blood tests. Hematomas were also observed at injection sites, and she then deteriorated rapidly with loss of consciousness, upper gastrointestinal hemorrhage, and uterine bleeding, dying 7 hours after admission. CCHF viral antigen was subsequently detected by immuno-histochemical staining of post-mortem tissues samples. One health care worker (nurse) who managed the patient wearing only gloves as personal protective equipment, later developed confirmed CCHF.

inal hemorrhage, and uterine bleeding, dying 7 hours after admission. CCHF viral antigen was subsequently detected by immuno-histochemical staining of post-mortem tissues samples. One health care worker (nurse) who managed the patient wearing only gloves as personal protective equipment, later developed confirmed CCHF. Case 7 (Rostov-on-Don, Russia, 2011) A 17-year-old pregnant woman (18-weeks gestation) was admitted to infectious diseases hospital with a 2-day history of fever (up to 39 °C), headache, and myalgia. Her initial laboratory tests were: Hb 90 g/L, RBC 3.4 × 1012/L, WBC 3.2 × 109/L, PLT 162 × 109/L and APTT 44 secs. She had a history of tick bite 7 days previously, CCHF was suspected and confirmed by RT-PCR on day 3 of disease. Ribavirin, antimicrobial and supportive treatment were started and 2–3 days later her condition improved, temperature decreased and laboratory results improved (WBC 8,7 × 109/L, PLT 232 × 109/L and APTT 43 secs). Hemorrhage was not evident during her disease course and she was discharged from hospital after 10 days and at term gave birth to a healthy child.

treatment were started and 2–3 days later her condition improved, temperature decreased and laboratory results improved (WBC 8,7 × 109/L, PLT 232 × 109/L and APTT 43 secs). Hemorrhage was not evident during her disease course and she was discharged from hospital after 10 days and at term gave birth to a healthy child. Case 8 (Tokat, Turkey, 2016) A 20-year-old woman in early pregnancy (estimated 4-weeks gestation) was admitted to Tokat State Hospital 1 day after a tick bite with a history of fever, sore throat, back pain anorexia, nausea and vomiting. Laboratory tests showed: WBC 1.75 × 109/L, Plt 46 × 109/L, APTT 53 secs, PT 17.2, AST 233 IU/L and ALT 98 IU/L. CCHF was suspected and RT-PCR positive from day 2 of illness. On day 3 of illness she developed vaginal bleeding and was referred to Tokat University Hospital with progressive thrombocytopenia (WBC 1.65 × 109/L, Plt 23 × 109/L, APTT 53 sec, PT 17.2, AST 270 IU/L, ALT 127 IU/L). She received 2 units of fresh frozen plasma and 1 unit of platelets and ribavirin was not given. Pelvic ultrasound showed a thickened endometrium with minimal cervical bleeding on examination. Her Beta HCG progressively reduced (2500IU − 143IU) and a complete spontaneous abortion was confirmed. She clinically improved and was discharged 9 days after admission.

a and 1 unit of platelets and ribavirin was not given. Pelvic ultrasound showed a thickened endometrium with minimal cervical bleeding on examination. Her Beta HCG progressively reduced (2500IU − 143IU) and a complete spontaneous abortion was confirmed. She clinically improved and was discharged 9 days after admission. Discussion Crimean-Congo hemorrhagic fever has been designated a priority global diseases threat by the World Health Organization.12 Its pathophysiology remains incompletely understood, particularly in vulnerable groups, despite recent increases in research efforts.13 Other viral hemorrhagic fevers, such as Ebola virus and Lassa fever, are thought to be more severe in pregnancy, resulting in higher maternal mortality rates. Pregnant patients with VHF also present additional infection, prevention and control challenges, due to the interventions required for obstetric complications and potential viral persistence in the fetus/products of conception. Limited data exists describing the course of CCHF in pregnancy, and we aimed to improve knowledge in this area through a systematic review of published cases and multi-national case series.

due to the interventions required for obstetric complications and potential viral persistence in the fetus/products of conception. Limited data exists describing the course of CCHF in pregnancy, and we aimed to improve knowledge in this area through a systematic review of published cases and multi-national case series. The most complete data set previously reported was by Gozel et al. in 2014 who published a case series of 5 pregnant women with CCHF and summary data on 21 other reported cases.10 Our systematic review identified an additional 7 reports (10 cases), that combined with our large additional case series provided a total 42 cases of CCHF in pregnancy for analysis. CCHF appears to be associated with more severe disease in pregnancy, with 14/41 cases (34%) resulting in maternal death. This is higher than overall case fatality rates in Turkey (5%),14 Russia (4%)15 and Kazakhstan (14.8%),16 but may reflect reporting bias, and national sub-group analysis of pregnant cases shows rates to be more comparable (pregnant cases from Turkey 1/14, 9.3%). There is however, a high rate of fetal/neonatal loss occurring in 24/41 cases (58.5%). In the majority of cases this is through spontaneous abortion early in pregnancy or associated with maternal death. Still birth appears to be rare as does neonatal death.

to be more comparable (pregnant cases from Turkey 1/14, 9.3%). There is however, a high rate of fetal/neonatal loss occurring in 24/41 cases (58.5%). In the majority of cases this is through spontaneous abortion early in pregnancy or associated with maternal death. Still birth appears to be rare as does neonatal death. The stage of the pregnancy when CCHF occurs may also influence maternal mortality, with 8/18 mothers dying in the second half of pregnancy (>20 weeks) compared with 1/11 dying in the first half of pregnancy (<20 weeks), although this was not statistically significant (p = 0.096). In our cases series 3/8 cases resulted in maternal death occurring at 34, 38 and 40-weeks gestation. However, as we have highlighted in cases 4 & 8, with appropriate critical care interventions and blood product support, positive outcomes are possible in the context of severe CCHF in pregnancy. This was also demonstrated by Ergonul et al.,9 who reported a case of severe CCHF in late pregnancy that survived, but required 22 units of fresh frozen plasma, 54 units of platelets and repeat surgical intervention for ongoing uterine hemorrhage after caesarian section.

ible in the context of severe CCHF in pregnancy. This was also demonstrated by Ergonul et al.,9 who reported a case of severe CCHF in late pregnancy that survived, but required 22 units of fresh frozen plasma, 54 units of platelets and repeat surgical intervention for ongoing uterine hemorrhage after caesarian section. Due to the risk of nosocomial transmission, surgical interventions in pregnant women with CCHF require careful infection prevention and control planning and risk assessment. Although caesarian section was performed in 2 cases that we have summarized, with no secondary nosocomial transmission, the patients had tested RT-PCR negative in blood prior to the operations. The PCR status of the fetus and amniotic fluid at the time surgery was not known, but in the case reported by Ergonul et al., the neonate subsequently tested CCHF RT-PCR positive and died.9 However, surgery was undertaken for post-partum hemorrhage in case 5 resulting in nosocomial transmission to 2 surgeons, and the neonate also transmitted CCHF to a further 3 healthcare workers. The additional nosocomial risk of pregnancy in other VHFs is not well understood, and during the Ebola outbreak in West Africa there were significant limitations on pregnant women’s access to obstetric care. However, complicated cases were managed successfully,17 and caesarian sections were undertaken by one group in Monrovia with no associated nosocomial transmission (Dr J Brown personal communication).

uring the Ebola outbreak in West Africa there were significant limitations on pregnant women’s access to obstetric care. However, complicated cases were managed successfully,17 and caesarian sections were undertaken by one group in Monrovia with no associated nosocomial transmission (Dr J Brown personal communication). Six of the 42 pregnant CCHF cases we identified did result in subsequent cases of CCHF as a result of nosocomial transmission. Naderi et al.,18 reported a nosocomial outbreak where a pregnant woman was an index case that infected another pregnant woman whilst sharing a room, who then infected another 4 healthcare workers. Nabeth et al.,19 conducted an investigation into a large nosocomial outbreak in Mauritania in 2003, identifying a 30-year old pregnant woman with severe CCHF as the index case. She directly infected a total of 15 healthcare workers, patients and visitors in the ward and emergency room setting with six fatal cases. A more recent report from Russia2 also highlighted the nosocomial risk of critical care interventions, particularly aerosol generating procedures that resulted in 8 healthcare worker infections, from a 23-year old pregnant woman. In all these cases CCHF was not initially suspected. The initial presentation of CCHF is non-specific, and a lack of healthcare worker awareness, that occurs in both endemic settings and in exported cases can result in delayed diagnosis of CCHF. This can delay initiation of supportive treatment for CCHF and also the required infection prevention and control measures. In our series there was a delay in recognition of CCHF in pregnant women in 4/8 cases (cases 3, 4, 5 & 6), all associated with severe/fatal disease and fetal death.

t in delayed diagnosis of CCHF. This can delay initiation of supportive treatment for CCHF and also the required infection prevention and control measures. In our series there was a delay in recognition of CCHF in pregnant women in 4/8 cases (cases 3, 4, 5 & 6), all associated with severe/fatal disease and fetal death. There is currently no specific antiviral therapy for CCHF, and the benefit of ribavirin treatment is controversial with mixed results mainly generated from retrospective studies.20 Its use is contraindicated in pregnancy (FDA Pregnancy Category X) due to significant teratogenecity being demonstrated in all animal species in which adequate studies have been performed.21 However, in ribavirin’s submission for inclusion in the WHO model list of essential medicines, its risk/benefit in pregnancy, in the context of the high mortality in VHF was re-considered.22 This was though predominantly based on its clearer evidence base in the treatment of Lassa fever and Argentinian Hemorrhagic fever23 and early reports of benefit in CCHF.

clusion in the WHO model list of essential medicines, its risk/benefit in pregnancy, in the context of the high mortality in VHF was re-considered.22 This was though predominantly based on its clearer evidence base in the treatment of Lassa fever and Argentinian Hemorrhagic fever23 and early reports of benefit in CCHF. Ribavirin may have greater effect in late stage pregnancy where mortality appears to be higher, risk of teratogenicity lower, and we believe its use may be justified in this scenario. In our combined data when treatment and outcome was known, 5/13 pregnant women died who received ribavirin, compared to 15/23 who died who did not receive ribavirin. The difference is not statistically significant and caution must be applied to any interpretation and conclusions on interventions from this retrospective analysis. Unfortunately, Favipiravir, the only other antiviral that has shown survival benefit in CCHF animal models,24 is also contraindicated in pregnancy due to teratogenicity and embryogenicity25 Pregnancy was also an exclusion criteria for its use in the JIKI Ebola clinical trial.26 Limited data exists on the risk of transmission through breast feeding in CCHF, with only 2 cases reported.27 In both cases RT-PCR was positive in the mother’s blood and negative in breastmilk, and the babies did not go on to develop CCHF. In our units breastfeeding is not recommended during the acute illness phase of CCHF, due to established risks of CCHF transmission to the baby. Samples of breastmilk would then be sent for RT-PCR as the patient recovered.

e in the mother’s blood and negative in breastmilk, and the babies did not go on to develop CCHF. In our units breastfeeding is not recommended during the acute illness phase of CCHF, due to established risks of CCHF transmission to the baby. Samples of breastmilk would then be sent for RT-PCR as the patient recovered. There is currently no CCHF vaccine that is licensed by the European Medicines Agency or US Food and Drug Administration. An inactivated mouse brain vaccine was developed by the former Soviet Union in the 1970s, and is still in use in Bulgaria. However, it has failed to show high neutralizing antibody levels,28 and lacks controlled data demonstrating protective efficacy. A number of promising CCHF vaccine candidates are under development,29, 30 and will soon proceed to phase 1 clinical trials. Pregnant women in endemic areas should be considered a high-risk CCHF group that requires particular consideration with respect to vaccine safety evaluation and during future roll-out strategies.

A number of promising CCHF vaccine candidates are under development,29, 30 and will soon proceed to phase 1 clinical trials. Pregnant women in endemic areas should be considered a high-risk CCHF group that requires particular consideration with respect to vaccine safety evaluation and during future roll-out strategies. Limitations to the systematic review are the incomplete information provided in some case descriptions, particularly on gestation of pregnancy. The total numbers identified also probably represent a significant underestimate due to incomplete reporting of cases from endemic regions and the likelihood of undiagnosed mild CCHF pregnant cases. Due to its non-specific clinical presentation CCHF in pregnancy can be difficult to distinguish from other causes of undifferentiated febrile illness. This is exacerbated by a lack of laboratory capacity in endemic regions combined with a lack of awareness of CCHF by clinicians. Hemorrhagic signs start around the fourth day of illness and it is often only at this stage that CCHF first considered, or as we have reported when there are secondary nosocomial cases.

ness. This is exacerbated by a lack of laboratory capacity in endemic regions combined with a lack of awareness of CCHF by clinicians. Hemorrhagic signs start around the fourth day of illness and it is often only at this stage that CCHF first considered, or as we have reported when there are secondary nosocomial cases. Conclusions The number of reported cases of CCHF in pregnant women is low, but this is an underestimate and the geographical range of CCHF is increasing, including within Europe. Clinicians must maintain a high index of suspicion and undertake risk assessments for CCHF in pregnant women with a fever, who reside in or who have a history of travel to CCHF endemic areas. Early recognition allows appropriate infection prevention and control precautions to be put in place, reducing the demonstrated risk of nosocomial transmission. In accordance with other viral hemorrhagic fevers, the mortality of CCHF in pregnant women appears to be increased, but as we have highlighted supportive care focused on blood product replacement and access to critical care interventions can result in positive outcomes in severe disease. Novel therapeutics are required to improve both maternal and fetal outcomes in CCHF, and as such pregnant woman should be included in future CCHF clinical trials whenever possible. Conflict of interest The authors declare that they have no competing financial interests. The content is solely the responsibility of the authors. Funding source No specific funding. TF is funded by the Wellcome Trust (104480/Z/14/Z) and the UK Ministry of Defence.

Conclusions The number of reported cases of CCHF in pregnant women is low, but this is an underestimate and the geographical range of CCHF is increasing, including within Europe. Clinicians must maintain a high index of suspicion and undertake risk assessments for CCHF in pregnant women with a fever, who reside in or who have a history of travel to CCHF endemic areas. Early recognition allows appropriate infection prevention and control precautions to be put in place, reducing the demonstrated risk of nosocomial transmission. In accordance with other viral hemorrhagic fevers, the mortality of CCHF in pregnant women appears to be increased, but as we have highlighted supportive care focused on blood product replacement and access to critical care interventions can result in positive outcomes in severe disease. Novel therapeutics are required to improve both maternal and fetal outcomes in CCHF, and as such pregnant woman should be included in future CCHF clinical trials whenever possible. Conflict of interest The authors declare that they have no competing financial interests. The content is solely the responsibility of the authors. Funding source No specific funding. TF is funded by the Wellcome Trust (104480/Z/14/Z) and the UK Ministry of Defence. Figure 1 Flowchart of literature search. Figure 1Table 1 Characteristic of pregnant CCHF cases.

Conflict of interest The authors declare that they have no competing financial interests. The content is solely the responsibility of the authors. Funding source No specific funding. TF is funded by the Wellcome Trust (104480/Z/14/Z) and the UK Ministry of Defence. Figure 1 Flowchart of literature search. Figure 1Table 1 Characteristic of pregnant CCHF cases. Table 1Authors Country Year (s) No. cases Diagnosis Age Gestation Maternal outcome Hemorr-hage Fetal/Neo-natal outcome Ribavirin Secondary cases Al-Tikriti SK et al.31 Iraq 1979 3 Viral culture/clinical ND ND Died Yes Died No No ND ND Died Yes Died No No ND ND Survived Yes Died No No Baljosevic. S et al.32 Former Yugoslavia 1986–1995 4 UK 29 32/40 Died Yes Died ND ND 30 24/40 Died Yes Died ND ND 39 38/40 Died Yes Died ND ND 24 11/40 Survived Yes Died ND ND Sharif-Mood B et al.11 Iran 2000–2005 6 PCR/Serology 19–38 ND Survived Yes Died Yes No PCR/Serology ND Survived Yes Died Yes No PCR/Serology ND Survived Yes Died Yes No PCR/Serology ND Survived Yes Survived Yes No PCR/Serology ND Survived Yes Survived Yes No PCR/Serology 16/40 Died Yes Died Yes No Nabeth P et al.19 Mauritania 2003 2 Clinical 30 ND Died Yes Died No Yes (19) Clinical ND ND ND ND ND No ND Ergonul O et al.9 Turkey 2003–2008 3 Serology 40 38/40 Survived Yes Died Yes No PCR 20 19/40 Survived Yes Died No No PCR/Serology 28 28/40 Died Yes Died No No Gozel MG et al.10 Turkey 2007–2011 5 PCR/Serology 35 8/40 Survived Yes Died No No PCR/Serology 30 18/40 Survived No Survived No No PCR 41 20/40 Survived No Survived No No PCR/Serology 19 21/40 Survived No Survived No No PCR/Serology 27 33/40 Survived No Survived No No Oskooei HO et al.33 Iran 2008 1 Serology UK UK Survived No Survived Yes No Dizbay M et al.34 Turkey 2009 1 PCR/Serology 22 36/40 Survived Yes Survived Yes No Naderi HR et al.18 Iran 2009 2 (PCR +ve secondary cases) UK ND Died Yes Died No Yes (1) 31 ND Died Yes Died No Yes (4) Mumdchiev-a H et al.35 Bulgaria 2009 1 PCR/Serology UK 26/40 Survived Yes Survived UK No Aydemir O et al.36 Turkey 2010 1 PCR 29 30/40 Survived No Survived No No Pshenichnya N et al.2 Russia 2011 1 PCR 23 22/40 Died Yes Died No Yes (8) Mardani M et al.37 Iran 2011 1 PCR/Serology 24 16/40 Survived Yes Survived Yes No Duygu F et al.38 Turkey 2011 2 PCR 25 17/40 Survived Yes Survived No No PCR 22 20/40 Survived No Survived No No Ünlüsoy-Aksu A et al.39 Turkey 2014 1 PCR 23 36/40 Survived Yes Survived Yes No Pschenichnaya N et al. Russia 2002 8 PCR/Serology 20 16/40 Survived No Survived No No

logy 24 16/40 Survived Yes Survived Yes No Duygu F et al.38 Turkey 2011 2 PCR 25 17/40 Survived Yes Survived No No PCR 22 20/40 Survived No Survived No No Ünlüsoy-Aksu A et al.39 Turkey 2014 1 PCR 23 36/40 Survived Yes Survived Yes No Pschenichnaya N et al. Russia 2002 8 PCR/Serology 20 16/40 Survived No Survived No No Russia 2003 Serology 19 38/40 Survived No Survived Yes No Russia 2004 PCR 17 30/40 Died Yes Died No No Russia 2005 Serology 24 17/40 Survived Yes Died No No Kazakhstan 2009 Immunohist 23 40/40 Died Yes Died No Yes (5) Kazakhstan 2010 Immunohist 21 34/40 Died Yes Died No Yes (1) Russia 2011 PCR 17 18/40 Survived No Survived Yes No Turkey 2016 PCR 20 04/40 Survived Yes Died No No ND – Not determined. PCR – polymerase chain reaction, Immunohist – Immunohistochemistry.

Introduction Dengue is considered a major cause of morbidity and mortality in tropical and sub-tropical countries.1, 2, 3, 4 It was estimated that in 2015, 79.6 million dengue cases occurred in 196 countries.5 Dengue is caused by any of the four antigenically distinct dengue virus serotypes (DENV1, DENV2, DENV3, and DENV4) and is transmitted to humans by the Aedes mosquitoes, the same vectors as for yellow fever virus, chikungunya virus, and Zika virus.1, 6, 7, 8, 9 Dengue infections range from inapparent or mild forms (with or without warning signs), to severe forms that can eventually lead to a fatal outcome.1, 4, 7, 8, 9, 10 The presence of warning signs and severe forms is often associated with secondary or subsequent infections by a heterotypic serotype, and may be caused by an immune phenomenon called antibody-dependent enhancement (ADE).4, 6, 8, 9, 10 Individuals infected by any of the DENV serotypes develop protective monotypic immunity evidenced by the generation of dengue immunoglobulin M (IgM) and immunoglobulin G (IgG) antibodies. The IgG antibodies may be found in serum at the end of the convalescent period (9–10 days) in primary infections, and may also be detected earlier in the case of secondary infections. IgG is found at higher titers up to 30–40 days after infection, but may remain detectable for decades, which allows the identification of individuals who have been in contact with the virus.4, 9, 10, 11

scent period (9–10 days) in primary infections, and may also be detected earlier in the case of secondary infections. IgG is found at higher titers up to 30–40 days after infection, but may remain detectable for decades, which allows the identification of individuals who have been in contact with the virus.4, 9, 10, 11 In Colombia, there is active circulation of all four serotypes. Epidemic waves have been occurring every 3–4 years, with shorter inter-epidemic periods, and the incidence rate is up to 220 cases per 100 000 inhabitants per year.12, 13, 14, 15 Although there is an existing surveillance system in Colombia to collect and provide information about symptomatic cases, there is limited information on inapparent infection or mild cases of dengue in people who do not seek medical care.12, 15, 16 A dengue vaccine has recently been licensed in some countries of Latin America and Asia, while there are other candidates in the pipeline.17, 18 Since both prevalence and transmission intensity are important factors that affect the efficacy and effectiveness of dengue vaccines,17, 18, 19, 20, 21, 22 it is important to acquire further knowledge about the burden of dengue in potential early dengue vaccine introducer countries such as Colombia.19 In order to better inform dengue vaccine introduction strategies in Colombia, the dengue serological prevalence and the seroconversion rate of individuals between 1 and 65 years of age in Medellin was estimated using a tiered dengue serological survey approach. The details and results of this study are described herein.

n order to better inform dengue vaccine introduction strategies in Colombia, the dengue serological prevalence and the seroconversion rate of individuals between 1 and 65 years of age in Medellin was estimated using a tiered dengue serological survey approach. The details and results of this study are described herein. Methods Study site and population Medellin is the second largest city in Colombia with more than 2.6 million inhabitants. The reported annual incidence of dengue over the last 10 years has ranged between 161 and 745 cases per 100 000 inhabitants. Santa Cruz (Comuna 2) is one of the 16 sub-districts of Medellin and is located in the northeastern part of the city. The population in 2011, distributed across 12 neighborhoods, was 108 706 inhabitants; 57.6% of these inhabitants were female and 80% were aged under 65 years. Approximately 87% of the population in the community is Mestizo or white and 12% are Afro-Colombian. Approximately 96% of households belong to a low socioeconomic stratum and the remaining 4% belong to the ‘low–low’ stratum.23

108 706 inhabitants; 57.6% of these inhabitants were female and 80% were aged under 65 years. Approximately 87% of the population in the community is Mestizo or white and 12% are Afro-Colombian. Approximately 96% of households belong to a low socioeconomic stratum and the remaining 4% belong to the ‘low–low’ stratum.23 Study design A tiered community-based serological survey was conducted in healthy residents of the 12 neighborhoods of Santa Cruz Comuna, from November 2011 to February 2014. To prevent any under-representation, the age and sex distributions of the Santa Cruz population, as well as the reported number of dengue cases in Medellin, were used to calculate sample sizes for age-stratified serum collection; these included the 95% confidence intervals and a margin of error at a fixed significance level within 25% of the true proportion of incidence, giving a relative precision of 75%, considering the gap in evidence for dengue incidence in the study areas. The sample size for precision is equal to24: n={z2[p(1−p)]}(p×0.25)2 where z = 1.96 (i.e., the z-score for the desired 95% confidence interval) and p is the anticipated population (prevalence according to surveillance data). Using the formula above, the sample size for the serological survey was calculated using the adjusted incidence estimates, ranging from 0.116 to 0.312 depending on the age group.14 The age distribution was obtained from census data,23 and the age-specific sample sizes were calculated by adding what was obtained for each age group and assuming a non-response rate or loss to follow-up of about 10%. The total sample size required for individuals between 1 and 65 years of age was 2000 subjects per visit, to be followed up every 6 months during the study period.

age-specific sample sizes were calculated by adding what was obtained for each age group and assuming a non-response rate or loss to follow-up of about 10%. The total sample size required for individuals between 1 and 65 years of age was 2000 subjects per visit, to be followed up every 6 months during the study period. Study outcomes were DENV seroprevalence (i.e., presence of DENV-specific antibodies in serum) and seroconversion (i.e., change from negative DENV antibodies during the first visit to positive DENV antibodies in subsequent visits), both assessed using IgG ELISA. All residents of the district aged between 1 and 65 years were eligible. Individuals from the same household and participants presenting fever at the time of the survey could be enrolled. However, individuals participating in any dengue vaccine trial or with plans to move out of the catchment area during the following 6 months were not eligible.

ct aged between 1 and 65 years were eligible. Individuals from the same household and participants presenting fever at the time of the survey could be enrolled. However, individuals participating in any dengue vaccine trial or with plans to move out of the catchment area during the following 6 months were not eligible. Data collection Participants were located in their households; each household was selected using probabilistic multistage sampling, first randomly selecting the blocks and then selecting the households systematically. If the subjects were considered eligible for enrollment, the study staff described the study, invited them to participate, obtained informed consent, and then proceeded to collect a sample. Arrangements for the second visit, about 6 months after the first visit, were made before leaving the household. The same procedures were followed to take successive blood samples at subsequent visits; laboratory results were provided and explained at these subsequent visits.

nt, and then proceeded to collect a sample. Arrangements for the second visit, about 6 months after the first visit, were made before leaving the household. The same procedures were followed to take successive blood samples at subsequent visits; laboratory results were provided and explained at these subsequent visits. Data were collected on the case record form (CRF) and then double-entered into a Microsoft FoxPro database. CRFs included the subject’s demographic data, medical history (including comorbidities and self-reported previous dengue infection), relevant symptoms (including the presence of fever, headache, muscle pain, and arthralgia, which were then asked about from the second visit onwards), and laboratory data. Financial compensation for lost wages was provided to the participants of the serological survey at each visit (maximum of 6 USD or its equivalent in Colombian Pesos (COP) in a voucher per person). Sample collection and antibody detection Blood samples (7–10 ml) were obtained by trained phlebotomists. The specimens were scanned and logged into a computerized database. Whole blood was centrifuged and serum was separated into cryotubes under sterile conditions; 0.2–0.5 ml serum aliquots were labeled and stored at −80 °C until processing. All samples collected were tested with the Panbio Dengue IgG indirect ELISA.

sts. The specimens were scanned and logged into a computerized database. Whole blood was centrifuged and serum was separated into cryotubes under sterile conditions; 0.2–0.5 ml serum aliquots were labeled and stored at −80 °C until processing. All samples collected were tested with the Panbio Dengue IgG indirect ELISA. Plaque reduction neutralization test (PRNT) The PRNT50 was conducted for samples indicating seroconversion or inconclusive results by IgG indirect ELISA. The PRNT analyses were based on previously described protocols.20 Virus strains used for the PRNT were DENV1 16007, DENV2 16681, DENV3 16562, and DENV4 1036. The reaction was revealed with AEC (3-amino-9-ethylcarbazole) chromogen and the immunofoci were counted using an ELISpot reader (AID Instruments, Germany). The PRNT titer was obtained from the reciprocal of the dilution of serum that reduced the plaque number by at least 50% relative to the virus-only control. Serum dilution values were expressed as the reciprocal of the original serum dilution, including the 1:2 dilution factor introduced during the neutralization step. Positive PRNT samples were defined as having a neutralizing titer of >1:10 to any of the viruses tested.

by at least 50% relative to the virus-only control. Serum dilution values were expressed as the reciprocal of the original serum dilution, including the 1:2 dilution factor introduced during the neutralization step. Positive PRNT samples were defined as having a neutralizing titer of >1:10 to any of the viruses tested. Samples that were first negative but then positive to any DENV serotype in subsequent samples by PRNT were considered as primary infection. Samples with previous neutralizing antibodies and positive to at least two or more DENV serotypes in subsequent samples were defined as secondary infection. Non-reactive cases were determined by the absence of neutralizing antibodies in both samples. Non-recent infections were defined when neutralizing antibodies were present in both samples and no increases or reductions in the PRNT50 titer were observed over time. Monotypic and heterotypic immune responses were defined as positive responses against single or multiple serotypes, respectively. Ethics This study obtained ethical approval from the Ethics Committee of the University of Antioquia, Municipal Health Office (Secretaria de Salud de Medellin) and the Institutional Review Board of the International Vaccine Institute (IVI). Written informed consent was obtained from adult participants and informed consent was obtained from the parents or legal guardians of minors (including the assent of children between 7 and 17 years of age), by a signature or thumbprint prior to blood collection.

view Board of the International Vaccine Institute (IVI). Written informed consent was obtained from adult participants and informed consent was obtained from the parents or legal guardians of minors (including the assent of children between 7 and 17 years of age), by a signature or thumbprint prior to blood collection. Statistical analysis Descriptive methods were used to present the general characteristics of all subjects. Summary statistics are presented as the mean with standard deviation (SD), frequency, or proportion depending on the variable. A mean comparison test was performed to estimate the difference in age as a continuous variable. To determine other statistical differences in the distribution of other key variables, the Kruskal–Wallis equality-of-populations rank test was performed.

deviation (SD), frequency, or proportion depending on the variable. A mean comparison test was performed to estimate the difference in age as a continuous variable. To determine other statistical differences in the distribution of other key variables, the Kruskal–Wallis equality-of-populations rank test was performed. Overall seroprevalence was calculated among subjects participating in at least one survey, and visit-specific seroprevalence rates were calculated using the total number of participants per visit as the denominator. Univariate and multivariate logistic regression was conducted to estimate the odds ratio (OR) with its corresponding 95% confidence interval (95% CI). In addition, multivariate logistic regression using a fixed-effects model adjusted by visits was performed to determine the characteristics related to DENV seroprevalence or seroconversion as the dependent variable. Independent variables in all models included age group, sex, ethnicity, neighborhood, and self-reported variables including previous condition/comorbidities, previous dengue, and yellow fever immunization (YFI) status. For the seroconversion model, presence of fever or other constitutional symptoms and ‘sought care’ during the 6 months prior to the visit were also added. Both models included an interaction term for age and sex, and another for age and previous conditions.

, previous dengue, and yellow fever immunization (YFI) status. For the seroconversion model, presence of fever or other constitutional symptoms and ‘sought care’ during the 6 months prior to the visit were also added. Both models included an interaction term for age and sex, and another for age and previous conditions. Person-months of follow-up time were calculated as the time between the date of the first visit and (1) the visit when seroconversion was identified, and (2) the last date of participation in the study, right-censoring. Age-specific seroconversion rates were calculated in naïve subjects (i.e., subjects with negative IgG indirect ELISA at their first visit) using the proportional follow-up person-months and reported as the age-specific seroconversion incidence rate per 1000 person-months with its corresponding 95% CI. A p-value lower than 0.05 (p < 0.05) was considered as statistically significant. There were no imputations of missing data; therefore, denominators vary by response. The data analysis was conducted using Stata 13 (StataCorp LP, College Station, TX, USA). Maps were created using QGIS 2.14-Essen (http://www.qgis.org/).

Person-months of follow-up time were calculated as the time between the date of the first visit and (1) the visit when seroconversion was identified, and (2) the last date of participation in the study, right-censoring. Age-specific seroconversion rates were calculated in naïve subjects (i.e., subjects with negative IgG indirect ELISA at their first visit) using the proportional follow-up person-months and reported as the age-specific seroconversion incidence rate per 1000 person-months with its corresponding 95% CI. A p-value lower than 0.05 (p < 0.05) was considered as statistically significant. There were no imputations of missing data; therefore, denominators vary by response. The data analysis was conducted using Stata 13 (StataCorp LP, College Station, TX, USA). Maps were created using QGIS 2.14-Essen (http://www.qgis.org/). Results Enrolment strategy and characteristics of enrollees Overall, 3684 individuals participated in at least one survey (Figure 1); 2450 (66.5%) were female, the mean age of all participants was 24.6 years (SD 17.1), and 3596 (97.6%) were Mestizo or white (Table 1). The median follow-up time was 12 months (interquartile range 0–18 months), with 2569 subjects participating in more than two surveys. Of those participants, 39% (n = 1002) were DENV-naïve subjects.

were female, the mean age of all participants was 24.6 years (SD 17.1), and 3596 (97.6%) were Mestizo or white (Table 1). The median follow-up time was 12 months (interquartile range 0–18 months), with 2569 subjects participating in more than two surveys. Of those participants, 39% (n = 1002) were DENV-naïve subjects. Seroprevalence The overall DENV-positive serological prevalence by indirect IgG ELISA was 61% (n = 2246/3684; 95% CI 59.4–62.5%), with 3.2% (n = 73/2246) of seropositive subjects self-reporting a past history of dengue. The seroprevalence rates were as follows: first visit 59.8% (1182/1976), second visit 62% (1247/2012), third visit 62.6% (1259/2012), fourth visit 61.7% (1249/2023), and fifth visit 61.3% (1739/1066). The DENV prevalence status for the new enrollees was not significantly different (Chi-square = 8.1, df = 4, p = 0.087; Kruskal–Wallis test). DENV serological prevalence by neighborhood is presented in Figure 2.

1247/2012), third visit 62.6% (1259/2012), fourth visit 61.7% (1249/2023), and fifth visit 61.3% (1739/1066). The DENV prevalence status for the new enrollees was not significantly different (Chi-square = 8.1, df = 4, p = 0.087; Kruskal–Wallis test). DENV serological prevalence by neighborhood is presented in Figure 2. The mean age of subjects with DENV antibodies was 30.3 years (SD 16.3) and without DENV antibodies was 15.6 years (SD 14.1); the difference between these groups was statistically significant (difference 14.7, 95% CI 13.7–15.7; p < 0.001). After controlling for visits and adjusting for all other covariates, there was a significant increase in the odds of DENV prevalence in subjects over 6 years old, especially in subjects between 41 and 50 years old (OR 47, 95% CI 36.1–61.1), when compared to subjects in the 1–5 years age group. There was also a significant increase in the odds of DENV prevalence in subjects with self-reported previous dengue infection (OR 4.7, 95% CI 3.0–7.4) compared to those who did not report previous dengue infection (Table 2). After including an interaction term between sex and age, it was observed that the odds of DENV prevalence increased when comparing males at 11–15 years to the reference group (OR 4.0, 95% CI 3.0–5.4; p < 0.001), and when comparing males to females at 16–20 years (OR 1.4, 95% CI 1.0–1.8; p = 0.043). However, the odds of seroprevalence decreased when comparing males to females at 11–15 years (OR 0.7, 95% CI 0.6–0.8; p < 0.001). Other interactions for age and sex were not statistically significant (data not shown).

4; p < 0.001), and when comparing males to females at 16–20 years (OR 1.4, 95% CI 1.0–1.8; p = 0.043). However, the odds of seroprevalence decreased when comparing males to females at 11–15 years (OR 0.7, 95% CI 0.6–0.8; p < 0.001). Other interactions for age and sex were not statistically significant (data not shown). Seroconversion There were 122 new infections among the naïve subjects participating in more than one visit (n = 1002) and the mean age of seroconverted subjects was 21.8 years (SD 16.2). The overall seroconversion rate was 8.7% per 1000 person-months (95% CI 7.3–10.4) over 2.5 years. However, infections were occurring at different rates; the highest rate was 22.9% per 1000 person-months, observed during the fifth visit (February 2014) (Figure 3). The highest age-specific seroconversion rate was 17.9% per 1000 person-months (95% CI 10.6–30.3), observed in subjects between 31 and 40 years of age (Table 3). Overall the mean age of DENV prevalent subjects was significantly higher than the mean age of seroconverted subjects (difference 8.4, 95% CI 5.5–11.5; p < 0.001). The seroconversion rates in the various neighborhoods are presented in Figure 4.

I 10.6–30.3), observed in subjects between 31 and 40 years of age (Table 3). Overall the mean age of DENV prevalent subjects was significantly higher than the mean age of seroconverted subjects (difference 8.4, 95% CI 5.5–11.5; p < 0.001). The seroconversion rates in the various neighborhoods are presented in Figure 4. In both the univariate and multivariate analyses, age increased the odds of seroconversion, specifically in subjects between 31 and 40 years old (OR 3.8, 95% CI 1.7–8.6) when compared to subjects between 1 and 5 years old. After adjusting for all other variables, the odds of seroconversion were higher in males (OR 1.31, 95% CI 1.0–1.7), likely due to an interaction with age, and in Afro-Colombians (OR 3.0, 95% CI 1.2–7.4) (Table 4 and Figure 5).

nd 40 years old (OR 3.8, 95% CI 1.7–8.6) when compared to subjects between 1 and 5 years old. After adjusting for all other variables, the odds of seroconversion were higher in males (OR 1.31, 95% CI 1.0–1.7), likely due to an interaction with age, and in Afro-Colombians (OR 3.0, 95% CI 1.2–7.4) (Table 4 and Figure 5). PRNT results A suitable paired sample was not available for 10 of the seroconverted samples; analyses were conducted using only the remaining 112 samples. Overall, 93% (n = 104/112) of samples considered positive for DENV by IgG ELISA were found to be reactive by PRNT. Primary infection with monotypic and heterotypic responses was detected in 2.9% (3/104) and 3.8% (4/104) of samples, respectively. Secondary infection with monotypic and heterotypic responses was found in 3.8% (4/104) and 55.8% (58/104) of samples, respectively. There were 35 samples with neutralizing antibodies in which the PRNT50 did not show significant changes. The circulation of all four DENV serotypes in the study area was confirmed by PRNT identification. Predominant DENV serotypes were DENV1 (n = 36) and DENV4 (n = 26), followed by DENV2 (n = 20) and DENV3 (n = 14).

y. There were 35 samples with neutralizing antibodies in which the PRNT50 did not show significant changes. The circulation of all four DENV serotypes in the study area was confirmed by PRNT identification. Predominant DENV serotypes were DENV1 (n = 36) and DENV4 (n = 26), followed by DENV2 (n = 20) and DENV3 (n = 14). Discussion Dengue is a major public health issue in Colombia. According to the present study findings, there is significant evidence of an important dengue burden in Medellin, with an overall dengue serological prevalence as high as 61% among residents of Santa Cruz aged 1–65 years, and an overall seroconversion rate of 8.7% per 1000 person-months during the study period. Despite this important prevalence, community self-awareness of dengue infection is low − only 3.3% of prevalent subjects self-reported at least one dengue episode. Although similar findings have been reported in other Latin American settings,25, 26 a change in people’s awareness is foreseeable due to the increasing public health strategies in place to diagnose and control the infections transmitted by Aedes mosquitoes, specifically for DENV and Zika virus.

ast one dengue episode. Although similar findings have been reported in other Latin American settings,25, 26 a change in people’s awareness is foreseeable due to the increasing public health strategies in place to diagnose and control the infections transmitted by Aedes mosquitoes, specifically for DENV and Zika virus. The increasing odds of seroprevalence in the older age groups could be attributed to the age effect in terms of time as opportunity of exposure: older people have had more time to be exposed than young people. It could also indicate that in the study population, individuals were exposed to dengue at an older age rather than during the earlier years of life. This could also be attributed to the fact that the study was conducted after one of the biggest outbreaks to have occurred in Colombia (2010), probably increasing the odds of exposure to DENV of susceptible adults before the beginning of this study, increasing the overall seroprevalence and depleting susceptible people.14, 15, 16, 27, 28 This aspect is important at the time of dengue vaccine introduction.18, 19, 28, 29, 30 Although a new dengue vaccine has been licensed for use in individuals between 9 and 45 years of age or between 9 and 60 years of age, the recommendation for introduction according to the license includes endemic settings with around 70% or greater DENV serological prevalence in the group targeted for dengue immunization.17

ugh a new dengue vaccine has been licensed for use in individuals between 9 and 45 years of age or between 9 and 60 years of age, the recommendation for introduction according to the license includes endemic settings with around 70% or greater DENV serological prevalence in the group targeted for dengue immunization.17 Interestingly, in terms of frequencies, the majority of seroconversions were in subjects under 15 years of age; however, rates were higher among adults with lower seroconversion rates at the extremes. These findings are comparable to observations made in Puerto Rico and Nicaragua,25, 31 where dengue incidence seems to increase with age, a pattern that is common in Latin American countries but which contrasts with the pattern observed in Southeast Asia.16, 28, 32, 33 Although it was outside the scope of this study, the assessment of neighborhood-specific strategies for vector control and other environmental characteristics might have been helpful for understanding the relatively homogeneous distribution of DENV prevalence and the more variable behavior for seroconversion. In Latin America, risk assessment by environmental (including entomological evaluation), socioeconomic, and health care access characteristics has been considered a useful tool to identify the dengue burden.15, 25, 34, 35 Moreover, comparing neighborhoods or other small geographic units to identify areas with dengue is key to address vector control and other targeted strategies for disease control.11, 15, 19, 25, 31

mic, and health care access characteristics has been considered a useful tool to identify the dengue burden.15, 25, 34, 35 Moreover, comparing neighborhoods or other small geographic units to identify areas with dengue is key to address vector control and other targeted strategies for disease control.11, 15, 19, 25, 31 In this study there was no significant difference between the sexes or ethnic groups in DENV prevalence, as described previously in other settings.4, 16, 26, 32, 36 However, the seroconversion rate seemed to be increased in Afro-Colombians compared to whites. Although an increase in seroconversion rate was expected in all ethnic groups because a new outbreak was starting at the time of the fourth visit (2013), it will be important to look carefully at the distribution of DENV-susceptible individuals and the seroconversion in the study area, where the majority of the population is Mestizo/white and for which an increased risk of infection and severity have been described.11, 16, 26, 27, 36 In the Colombian context, YFI is recommended but not compulsory for people travelling to endemic areas.37 A relatively small population reported having been immunized against yellow fever (28.8%). After adjusting for all covariates, prior YFI was not associated with seropositivity but was associated with seroconversion. These findings together with the possibility of cross-reactivity should be interpreted with caution and could be compared to what has been reported in other studies performed in Central America and Southeast Asia.4, 16, 33

ariates, prior YFI was not associated with seropositivity but was associated with seroconversion. These findings together with the possibility of cross-reactivity should be interpreted with caution and could be compared to what has been reported in other studies performed in Central America and Southeast Asia.4, 16, 33 Neither fever nor seeking care for fever during the 6 months prior to the visits was associated with seroconversion, although the presence of symptoms was. This is important for the assessment and definition of ‘asymptomatic’ dengue infection.7, 15, 16, 22, 31, 38 In this case, the incidence rate of dengue infections that were either mild (because the subject acknowledged the presence of symptoms) or inapparent is indicated, because only 9.8% of seroconverted subjects reported fever during the previous 6 months. Although similar rates have been observed in Nicaragua and in other endemic countries,16, 27, 31, 33 the rate of ‘subclinical’ dengue infection remains unclear, and in the presence of a low confirmation rate of dengue cases, the accurate identification of dengue burden remains challenging.16, 33, 35

previous 6 months. Although similar rates have been observed in Nicaragua and in other endemic countries,16, 27, 31, 33 the rate of ‘subclinical’ dengue infection remains unclear, and in the presence of a low confirmation rate of dengue cases, the accurate identification of dengue burden remains challenging.16, 33, 35 The four DENV serotypes were identified by PRNT in the study area, which is associated with the observed high seroprevalence and indicates ongoing DENV transmission. The predominance of DENV4 and the low presence of DENV3 in the study area could also explain the presence of mild symptoms in seroconverted subjects.4, 27, 32, 39 This distribution of serotypes could also be related to the important circulation of DENV3 and DENV1 from the years 2000 to 2011,15 and specifically during the 2010’s outbreak.14, 15, 26, 30

low presence of DENV3 in the study area could also explain the presence of mild symptoms in seroconverted subjects.4, 27, 32, 39 This distribution of serotypes could also be related to the important circulation of DENV3 and DENV1 from the years 2000 to 2011,15 and specifically during the 2010’s outbreak.14, 15, 26, 30 Strengths and limitations Despite the known important burden of dengue in Colombia, this study is the first to address serological prevalence and age-specific seroconversion rates in the study population, based on laboratory data. The study period lasted around 2.5 years, allowing incidence rates for new dengue infections to be calculated – information that is needed and has been broadly discussed.15, 35 Knowing the seasonality of dengue, sensitive information about the patterns of dengue distribution were captured for the study area. As in other studies, the level of attrition related to cohort studies constitutes a limitation that has been described previously.25 Part of the large attrition seen at the first and second visits was due to a municipal urbanization relocation plan in one of the neighborhoods (Sinai), which was a slum. Inhabitants of this neighborhood were relocated not only within the study area but also elsewhere, and it was not possible to follow them up. This right-censoring was considered non-informative, and the absence of statistical difference in seroprevalence status in new enrollees eased concerns about the risk of a selection bias. The interpretation of self-reported previous dengue infection and presence of fever or other symptoms may have been affected by the known lack of awareness of the disease in some Latin American contexts,15, 32, 38 as well as the risk of recall bias because of the visit intervals.

rns about the risk of a selection bias. The interpretation of self-reported previous dengue infection and presence of fever or other symptoms may have been affected by the known lack of awareness of the disease in some Latin American contexts,15, 32, 38 as well as the risk of recall bias because of the visit intervals. Although it is important to acknowledge the possible presence of other arboviruses in the study area, this study was conducted prior to the documented introduction of chikungunya and Zika viruses, favoring the reliability of the serological analysis. However, the possibility that some secondary infections were missed by using the IgG ELISA to test for seroconversion and that some of the age effect observed in the seroprevalence may have been related to the intrinsic characteristics of the test is acknowledged. The PRNT was necessary to identify recent and non-recent heterogeneous infections in the presence of a high serological prevalence and the use of an IgG ELISA; however, the presence of heterotypic responses might limit the interpretations about the serotype-specific incident cases.

eristics of the test is acknowledged. The PRNT was necessary to identify recent and non-recent heterogeneous infections in the presence of a high serological prevalence and the use of an IgG ELISA; however, the presence of heterotypic responses might limit the interpretations about the serotype-specific incident cases. Conclusions The role of seroprevalence has been discussed largely in regard to the efficacy and the effectiveness of dengue vaccines.18, 20, 21, 22, 28, 29, 30, 40 From the public health point of view, it is important to have an estimate of age-specific dengue prevalence in endemic countries at the time of dengue vaccine introduction, as suggested by the World Health Organization.11, 29, 30 The information provided in this article has a two-fold impact. First, the age-specific seroprevalence and seroconversion rates will be helpful for policymakers, allowing informed decisions to be made on whether they should consider the introduction of dengue vaccines.11, 12, 19, 22, 30, 41 Specifically, due to the important serological prevalence, the evidence of ongoing transmission, the current re-emergence of other arboviruses, and other demographic characteristics of the study population, the introduction of dengue vaccine, along with a structured vector control program, could be considered when formulating disease control strategies. Secondly, for the developers of new dengue vaccines currently in the pipeline, this study provides evidence of the relative heterogeneity that can be observed in endemic areas such as Colombia and also shows the existing heterotypic response rates, which present a challenge that must be addressed adequately by the new vaccine candidates.17, 18, 19, 21, 22, 28, 29, 30, 33, 40, 41, 42 Overall, this study contributes to a better understanding of dengue in Medellin, Colombia, and in other Latin American settings with similar characteristics (i.e., low socioeconomic urban and peri-urban areas with intermediate to high prevalence and areas endemic for dengue).

18, 19, 21, 22, 28, 29, 30, 33, 40, 41, 42 Overall, this study contributes to a better understanding of dengue in Medellin, Colombia, and in other Latin American settings with similar characteristics (i.e., low socioeconomic urban and peri-urban areas with intermediate to high prevalence and areas endemic for dengue). Appendix A Supplementary data The following is Supplementary data to this article: Acknowledgements We would like to express our gratitude to the population of Santa Cruz Comuna, their community leaders, the staff members of the health care centers, especially Dilia Diaz, all members of Metrosalud E.S.E, and the Secretaria de Salud Municipal de Medellin, especially Rita Almanza. We would like to acknowledge the collaboration of the entire study staff (nurse assistants, laboratory assistants, and couriers) from PECET. We would also like to acknowledge Jose Fernando Florez and Santiago Patino for their collaboration and support as part of the data management team. Funding: This study forms part of the Dengue Vaccine Initiative program, a work supported by the Bill & Melinda Gates foundation (grant number OPP1016669) (URL http://www.gatesfoundation.org/). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of interest: The authors declare no conflict of interest Appendix A Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2017.02.016. Figure 1 Flow chart of enrollment and participation.

Funding: This study forms part of the Dengue Vaccine Initiative program, a work supported by the Bill & Melinda Gates foundation (grant number OPP1016669) (URL http://www.gatesfoundation.org/). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of interest: The authors declare no conflict of interest Appendix A Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2017.02.016. Figure 1 Flow chart of enrollment and participation. Figure 1Figure 2 Overall dengue virus (DENV) seroprevalence by neighborhood. Map of Santa Cruz sub-district, indicating the proportion of DENV serological prevalence by IgG indirect ELISA in each neighborhood. Figure 2Figure 3 Seroconversion rate per visit. Overall seroconversion rate in dengue-naïve subjects at baseline by follow-up time. Includes the number at risk at the beginning of the time period and the number of events (new infections) in parenthesis. The curve shows an increasing probability of seroconversion over time. Figure 3Figure 4 Distribution of seroconversion by neighborhood (rate). Map of Santa Cruz sub-district indicating the proportion of dengue virus seroconversion (negative to positive antibodies) by IgG Indirect ELISA in each of the 11 neighborhoods. Figure 4Figure 5 Overall probability of dengue seroconversion by age group. Seroconversion probabilities adjusted by sex, neighborhood, visits, presence of fever, and all other covariates in the model (n = 1002).

Figure 3Figure 4 Distribution of seroconversion by neighborhood (rate). Map of Santa Cruz sub-district indicating the proportion of dengue virus seroconversion (negative to positive antibodies) by IgG Indirect ELISA in each of the 11 neighborhoods. Figure 4Figure 5 Overall probability of dengue seroconversion by age group. Seroconversion probabilities adjusted by sex, neighborhood, visits, presence of fever, and all other covariates in the model (n = 1002). Figure 5Table 1 General characteristics of study subjects and DENV prevalence at enrolment

Figure 3Figure 4 Distribution of seroconversion by neighborhood (rate). Map of Santa Cruz sub-district indicating the proportion of dengue virus seroconversion (negative to positive antibodies) by IgG Indirect ELISA in each of the 11 neighborhoods. Figure 4Figure 5 Overall probability of dengue seroconversion by age group. Seroconversion probabilities adjusted by sex, neighborhood, visits, presence of fever, and all other covariates in the model (n = 1002). Figure 5Table 1 General characteristics of study subjects and DENV prevalence at enrolment Table 1Characteristics Total (N = 3684)a DENV seropositive (n = 2246)b n (%) n (%) Sex, female 2450 (66.5) 1617 (72.0) Ethnicity White (Hispanic) 109 (3.0) 67 (3.0) Mestizo 3487 (94.6) 2134 (95.0) African descendent 88 (2.4) 45 (2.0) Age, mean ± SD 24.6 ± 17.1 30.3 ± 16.3c Age groups, years 1–5 376 (10.2) 50 (2.2) 6–10 487 (13.2) 145 (6.5) 11–15 552 (15.0) 279 (12.4) 16–20 405 (11) 276 (12.3) 21–30 615 (16.7) 457 (20.3) 31–40 429 (11.6) 341 (15.2) 41–50 414 (11.2) 362 (16.1) 51–60 298 (8.1) 243 (10.8) >60 108 (2.9) 93 (4.1) Neighborhood La Isla 254 (6.9) 168 (7.5) Villa del Socorro 309 (8.4) 157 (7.0) El Playon de los Comuneros 296 (8.0) 178 (7.9) Villa Niza 356 (9.7) 243 (10.8) Pablo VI 275 (7.5) 163 (7.3) Moscu 1 311 (8.4) 189 (8.4) La Frontera 226 (6.1) 134 (6.0) Santa Cruz 399 (10.8) 233 (10.4) La Francia 344 (9.3) 228 (10.1) La Rosa 368 (10.0) 222 (9.9) Andalucía 274 (7.4) 158 (7.0) Sinai 272 (7.4) 173 (7.7) Pre-existing conditionsd 721 (19. 6) 491 (21.9) Self-reported previous dengue infection 84 (2.3) 73 (3.3) Yellow fever vaccinatione No 782 (21.2) 515 (22.9) Yes 1061 (28.8) 573 (25.1) Don’t know 1841 (50.0) 1158 (51.6) DENV, dengue virus; SD, standard deviation.

) Andalucía 274 (7.4) 158 (7.0) Sinai 272 (7.4) 173 (7.7) Pre-existing conditionsd 721 (19. 6) 491 (21.9) Self-reported previous dengue infection 84 (2.3) 73 (3.3) Yellow fever vaccinatione No 782 (21.2) 515 (22.9) Yes 1061 (28.8) 573 (25.1) Don’t know 1841 (50.0) 1158 (51.6) DENV, dengue virus; SD, standard deviation. a Subjects with at least one participation. b DENV IgG indirect ELISA positive at visit 1. c The mean age difference between subjects with and without positive DENV antibodies at enrolment was 14.7 years (95% confidence interval 13.7–15.7 years); p < 0.001. d Self-reported comorbidities or previous conditions such as diabetes, hypertension, hypercholesterolemia, etc. e Self-reported yellow fever immunization status. Table 2 Univariate and multivariate logistic regression analysis for DENV prevalence (IgG Indirect ELISA positive antibodies) at enrollment.

c The mean age difference between subjects with and without positive DENV antibodies at enrolment was 14.7 years (95% confidence interval 13.7–15.7 years); p < 0.001. d Self-reported comorbidities or previous conditions such as diabetes, hypertension, hypercholesterolemia, etc. e Self-reported yellow fever immunization status. Table 2 Univariate and multivariate logistic regression analysis for DENV prevalence (IgG Indirect ELISA positive antibodies) at enrollment. Table 2Characteristics (N = 3684) Univariate analysis Multivariate analysisa OR 95% CI p-Value aOR 95% CI p-Value Sex Female Ref. – – Ref. – – Male 0.54 (0.47–0.62) 0.001 1.01 (0.91–1.13) 0.831 Ethnicity White (Hispanic) Ref – – Ref. – – Mestizo 0.99 (0.67–1.46) 0.935 0.77 (0.58–1.04) 0.084 African descendent 0.66 (0.37–1.16) 0.146 0.77 (0.50–1.18) 0.224 Age, years 1–5 Ref. – – Ref. – – 6–10 2.76 (1.94–3.94) <0.001 2.83 (2.28–3.51) <0.001 11–15 6.66 (4.74–9.37) <0.001 6.07 (4.94–7.46) <0.001 16–20 13.95 (9.70–20.07) <0.001 12.65 (10.1–15.85) <0.001 21–30 18.86 (13.31–26.72) <0.001 20.10 (16.08–25.13) <0.001 31–40 25.26 (17.30–36.90) <0.001 27.14 (21.2–34.76) <0.001 41–50 45.39 (29.94–68.81) <0.001 46.97 (36.09–61.12) <0.001 51–60 28.81 (18.98–43.73) <0.001 34.09 (26.11–44.52) <0.001 >60 40.42 (21.72–5.24) <0.001 32.24 (22.19–46.85) <0.001 Neighborhood La Isla Ref. – – Ref. – – Villa Socorro 0.53 (0.38–0.74) <0.001 0.41 (0.32–0.53) <0.001 El Playon 0.77 (0.54–1.09) 0.146 0.61 (0.47–0.78) <0.001 Villa Niza 1.1 (0.78–1.55) 0.582 0.87 (0.68–1.13) 0.298 Pablo VI 0.75 (0.52–1.06) 0.103 0.71 (0.55–0.91) 0.008 Moscu 1 0.79 (0.56–1.12) 0.188 0.64 (0.49–0.82) <0.001 La Frontera 0.75 (0.51–1.08) 0.121 0.77 (0.59–1.01) 0.057 Santa Cruz 0.72 (0.52–1.00) 0.048 0.53 (0.42–0.68) <0.001 La Francia 1.01 (0.71–1.42) 0.972 0.96 (0.73–1.24) 0.734 La Rosa 0.78 (0.56–1.09) 0.141 0.58 (0.45–0.75) <0.001 Andalucía 0.7 (0.49–0.99) 0.046 0.51 (0.39–0.66) <0.001 Sinai 0.89 (0.63–1.28) 0.542 0.61 (0.47–0.80) <0.001 Pre-existing conditionsb 1.47 (1.24–1.75) <0.001 0.97 (0.85–1.11) 0.678 Previous DENVc 4.36 (2.30–8.24) <0.001 4.70 (2.98–7.41) <0.001 Yellow fever vaccinationd No Ref. – – Ref. – – Yes 0.61 (0.50–0.74) <0.001 1.08 (0.96–1.22) 0.215 Don’t know 0.88 (0.74–1.05) 0.150 1.05 (0.93–1.19) 0.447 DENV, dengue virus; OR, odds ratio; CI, confidence interval; aOR, adjusted odds ratio.

.85–1.11) 0.678 Previous DENVc 4.36 (2.30–8.24) <0.001 4.70 (2.98–7.41) <0.001 Yellow fever vaccinationd No Ref. – – Ref. – – Yes 0.61 (0.50–0.74) <0.001 1.08 (0.96–1.22) 0.215 Don’t know 0.88 (0.74–1.05) 0.150 1.05 (0.93–1.19) 0.447 DENV, dengue virus; OR, odds ratio; CI, confidence interval; aOR, adjusted odds ratio. a Multilevel logistic regression by visit, using a random-effects model, indicating the aOR of DENV prevalence on each covariate. b Self-reported comorbidities or previous conditions such as diabetes, hypertension, hypercholesterolemia, etc. c Self-reported previous dengue infection. d Self-reported yellow fever immunization status. Table 3 Age-specific and sex distribution of seroconversion cases and the corresponding rates per 1000 person-months in naïve subjects (seronegative at visit 1) with more than one visit (n = 1002). Table 3Age group, years Seroconversion Female, n (%) Male, n (%) Total Person-months Rate 95% CI 1–5 5 (6.7) 9 (19.1) 14 2916 4.8 (2.9–8.2) 6–10 8 (10.7) 13 (27.7) 21 3210 6.5 (4.2–10.0) 11–15 12 (16.0) 12 (25.5) 24 3192 7.5 (5.0–11.1) 16–20 7 (9.3) 3 (6.4) 10 1218 8.2 (4.4–15.3) 21–30 14 (18.7) 4 (8.5) 18 1260 14.3 (8.9–22.5) 31–40 12 (16.0) 2 (4.3) 14 780 17.9 (10.6–30.3) 41–50 8 (10.67) 2 (4.3 10 594 16.8 (9.1–31.3) >50 9 (11.9) 2 (4.3) 11 798 13.8 (7.6–24.7) Total 75 (100) 47 (100) 122 14 028 8.7 (7.3–10.4) CI, confidence interval. Table 4 Univariate and multivariate logistic regression for seroconversion in naïve subjects (seronegative at visit 1) with more than one visit (n = 1002).

Table 3Age group, years Seroconversion Female, n (%) Male, n (%) Total Person-months Rate 95% CI 1–5 5 (6.7) 9 (19.1) 14 2916 4.8 (2.9–8.2) 6–10 8 (10.7) 13 (27.7) 21 3210 6.5 (4.2–10.0) 11–15 12 (16.0) 12 (25.5) 24 3192 7.5 (5.0–11.1) 16–20 7 (9.3) 3 (6.4) 10 1218 8.2 (4.4–15.3) 21–30 14 (18.7) 4 (8.5) 18 1260 14.3 (8.9–22.5) 31–40 12 (16.0) 2 (4.3) 14 780 17.9 (10.6–30.3) 41–50 8 (10.67) 2 (4.3 10 594 16.8 (9.1–31.3) >50 9 (11.9) 2 (4.3) 11 798 13.8 (7.6–24.7) Total 75 (100) 47 (100) 122 14 028 8.7 (7.3–10.4) CI, confidence interval. Table 4 Univariate and multivariate logistic regression for seroconversion in naïve subjects (seronegative at visit 1) with more than one visit (n = 1002). Table 4Seroconverted subjects (n = 122/1002) Univariate analysis Multivariate analysisa Characteristics n (%) OR 95% CI p-Value aOR (95% CI) p-Value Sex Male 47 (38.52) 0.84 (0.57–1.24) 0.379 1.31 (1.02–1.68) 0.036 Ethnicity White (Hispanic) 3 (2.46) Ref. – – Ref – – Mestizo 113 (92.62) 1.04 (0.31–3.52) 0.950 1.06 (0.52–2.18) 0.866 African descendent 6 (4.92) 1.92 (0.43–8.58) 0.395 3.01 (1.22–7.42) 0.017 Age groups, years (mean 21.8, SD 16.2) 1–5 14 (11.48) Ref. – – Ref. – – 6–10 21 (17.21) 1.3 (0.65–2.66) 0.445 1.43 (0.95–2.14) 0.087 11–15 24 (19.67) 1.66 (0.84–3.31) 0.148 1.67 (1.12–2.48) 0.011 16–20 10 (8.20) 1.76 (0.75–4.13) 0.195 1.60 (0.94–2.72) 0.085 21–30 18 (14.75) 3.09 (1.46–6.5) 0.003 4.17 (2.66–6.54) <0.001 31–40 14 (11.48) 4.68 (2.08–10.57) <0.001 4.73 (2.9–7.72) <0.001 41–50 10 (8.20) 4.57 (1.86–11.22) 0.001 5.68 (3.33–9.7) <0.001 51–60 10 (8.20) 4.29 (1.75–10.47) 0.001 5.36 (3.11–9.24) <0.001 >60 1 (0.82) 1.37 (0.16–11.49) 0.771 1.69 (0.55–5.15) 0.357 Neighborhood La Isla 11 (9.02) Ref. – – Ref. – – Villa Socorro 12 (9.84) 0.48 (0.20–1.16) 0.105 0.45 (0.27–0.76) 0.003 El Playon 6 (4.92) 0.34 (0.12–0.98) 0.046 0.32 (0.17–0.6) <0.001 Villa Niza 7 (5.74) 0.46 (0.17–1.28) 0.138 0.39 (0.21–0.73) 0.003 Pablo VI 6 (4.92) 0.32 (0.11–0.93) 0.036 0.35 (0.19–0.65) 0.001 Moscu 1 16 (13.11) 1.0 (0.43–2.34) 0.995 1.02 (0.61–1.7) 0.95 La Frontera 14 (11.48) 1.05 (0.44–2.53) 0.908 1.45 (0.87–2.42) 0.152 Santa Cruz 13 (10.66) 0.56 (0.23–1.35) 0.195 0.47 (0.28–0.81) 0.007 La Francia 6 (4.92) 0.42 (0.15–1.22) 0.112 0.46 (0.24–0.88) 0.019 La Rosa 8 (6.56) 0.45 (0.17–1.19) 0.108 0.53 (0.29–0.94) 0.031 Andalucía 5 (4.10) 0.31 (0.10–0.95) 0.041 0.34 (0.18–0.65) 0.001 Sinai 18 (14.75) 1.31 (0.56–3.03) 0.53 1.24 (0.74–2.08) 0.411 Pre-existing conditionb Yes 24 (19.67) 1.33 (0.82–2.15) 0.249 1.16 (0.86–1.55) 0.328 Previous DENVc Yes 4 (3.28) 4.94 (1.37–17.75) 0.014 4.12 (1.88–9.0) <0.001 Yellow fever vaccinationd No 27 (22.13) Ref. – – Ref.

.31 (0.10–0.95) 0.041 0.34 (0.18–0.65) 0.001 Sinai 18 (14.75) 1.31 (0.56–3.03) 0.53 1.24 (0.74–2.08) 0.411 Pre-existing conditionb Yes 24 (19.67) 1.33 (0.82–2.15) 0.249 1.16 (0.86–1.55) 0.328 Previous DENVc Yes 4 (3.28) 4.94 (1.37–17.75) 0.014 4.12 (1.88–9.0) <0.001 Yellow fever vaccinationd No 27 (22.13) Ref. – – Ref. – – Yes 53 (43.44) 1.08 (0.66–1.77) 0.749 1.41 (1.06–1.86) 0.017 Don’t know 42 (34.43) 1.82 (1.08–3.06) 0.023 1.78 (1.32–2.42) <0.001 Fevere 12 (9.84) 0.97 (0.51–1.83) 0.924 1.18 (0.83–1.67) 0.367 Sought caref 6 (4.92) 1.42 (0.58–3.47) 0.446 1.56 (0.94–2.6) 0.084 Symptomsg 111 (90.98) 1.44 (0.75–2.76) 0.271 1.85 (1.16–2.94) 0.009 OR, odds ratio; CI, confidence interval; aOR, adjusted odds ratio. a Multivariate logistic regression adjusted by visit, using a fixed-effects model, indicating the aOR of seroconversion on each covariate. b Self-reported comorbidities or previous conditions such as diabetes, hypertension, hypercholesterolemia, etc. c Self-reported previous dengue infection. d Self-reported yellow fever immunization status. e Self-reported presence of fever during the 6 months prior to the visit. f Self-reported medical consultation due to the febrile episode. g Self reported constitutional symptoms/ dengue-like symptoms (headache, myalgia, arthralgia, etc.)

Introduction Antimicrobial resistance (AMR) is now a global problem, and resistance in Enterobacteriaceae, specifically Escherichia coli and Klebsiella pneumoniae, is a critical threat to human health (World Health Organization, 2014, WHO, 2018). Infections caused by third-generation cephalosporin-resistant (3GCR) Enterobacteriaceae are associated with increased mortality, length of stay, and costs compared with drug-sensitive strains (Stewardson et al., 2016). Carbapenems are less reliable as last-resort antibiotics because of increasing resistance (Gelband et al., 2015). AMR already imposes a heavy economic burden on health systems (Stewardson et al., 2016). Projecting future prevalence of AMR may help prioritize research projects and interventions. The aims of this study were to estimate global trends in AMR in E. coli and K. pneumoniae and to project future AMR prevalence to 2030. Methods Data on population and gross national income per capita (GNIPC) from the World Bank and on AMR from ResistanceMap, a global repository of AMR data from quality-assured and accredited hospitals and laboratory networks, were used (CDDEP, 2018). Countries for which samples were obtained from a single hospital were excluded from this study. Annual AMR data that had fewer than 30 isolates were also excluded (Agresti and Caffo, 2000).

p, a global repository of AMR data from quality-assured and accredited hospitals and laboratory networks, were used (CDDEP, 2018). Countries for which samples were obtained from a single hospital were excluded from this study. Annual AMR data that had fewer than 30 isolates were also excluded (Agresti and Caffo, 2000). With few exceptions, low- and middle-income countries are less likely to monitor AMR; therefore, high-income countries are overrepresented in the ResistanceMap database. Not taking this into account may lead to selection bias and an underestimation of the prevalence of AMR because of the strong negative association between GNIPC and the prevalence of AMR (Alvarez-Uria et al., 2016). To overcome this problem, inverse probability of inclusion (IPI) weighting, a method analogous to the use of inverse probability weights, was used to account for non-responders in surveys (Dugoff et al., 2014). IPI weights were calculated based on the inverse probability of being included in the study, using a logistic regression model that included data from countries in the world for which GNIPC data were available (countries with no GNIPC data comprised 1.3% of the world population) (The World Bank, 2018). In this logistic regression model, the availability of national AMR data (thus being included in the study) was the dependent variable, and orthogonal cubic spline transformations of 2014 GNIPC and 2014 country populations were the independent covariates (Dugoff et al., 2014). IPI weights gave more ‘weight’ to countries that were less likely to have AMR data in the ResistanceMap database, based on their GNIPC and population. This method helps generalize the results of the study to the world population. IPI weights were multiplied by population weights, which gives more weight to countries with larger populations, and the results were used as probability or sample weights in the final mixed model with random intercept and slopes (Dugoff et al., 2014). The mixed models were used to project AMR up to 2030.

d population. IPI weights were multiplied by population weights, which gives more weight to countries with larger populations, and the results were used as probability or sample weights in the final mixed model with random intercept and slopes (Dugoff et al., 2014). The mixed models were used to project AMR up to 2030. Results The study included 45 countries with AMR data for E. coli and 43 countries with AMR data for K. pneumoniae. In countries with E. coli AMR data, the median number of AMR point estimates was 14 (interquartile range 1–15), and 31 were high-income countries. In countries with K. pneumoniae AMR data, the median number of AMR point estimates was 10 (interquartile range 2–14), and 28 were high-income countries. No country had AMR data beyond 2015.

h E. coli AMR data, the median number of AMR point estimates was 14 (interquartile range 1–15), and 31 were high-income countries. In countries with K. pneumoniae AMR data, the median number of AMR point estimates was 10 (interquartile range 2–14), and 28 were high-income countries. No country had AMR data beyond 2015. Forecast estimates of global AMR are presented in Figure 1. The estimated prevalence of AMR in 2015 was 64.5% (95% confidence interval (CI) 42–87%) for 3GCR E. coli, 5.8% (95% CI 1.8–9.7%) for carbapenem-resistant (CR) E. coli, 66.9% (95% CI 47.1–86.8%) for 3GCR K. pneumoniae, and 23.4% (95% CI 7.4–39.4%) for CR K. pneumoniae. The projected annual variation (slope) of AMR was 0.83% (95% CI 0.73–0.93%) for 3GCR E. coli, 0.4% (95% CI 0.12–0.68%) for CR E. coli, −0.58% (95% CI −1.46% to 0.3%) for 3GCR K. pneumoniae, and 1.96% (95% CI 0.59–3.33%) for CR K. pneumoniae. The projected AMR prevalence in 2030 was 77% (95% CI 55–99.1%) for 3GCR E. coli, 11.8% (95% CI 3.7–19.9%) for CR E. coli, 58.2% (95% CI 50.2–66.1%) for 3GCR K. pneumoniae, and 52.8% (95% CI 16.3–89.3%) for CR K. pneumoniae. Projections for individual countries with at least four AMR point estimates using simple linear regression are presented in the Supplementary material.Figure 1 Forecast estimates with 95% confidence intervals of global resistance of Escherichia coli (A) and Klebsiella pneumoniae (B) to third-generation (3G) cephalosporins and carbapenems based on population weighted mixed models with random slopes and intercepts. Figure 1

Forecast estimates of global AMR are presented in Figure 1. The estimated prevalence of AMR in 2015 was 64.5% (95% confidence interval (CI) 42–87%) for 3GCR E. coli, 5.8% (95% CI 1.8–9.7%) for carbapenem-resistant (CR) E. coli, 66.9% (95% CI 47.1–86.8%) for 3GCR K. pneumoniae, and 23.4% (95% CI 7.4–39.4%) for CR K. pneumoniae. The projected annual variation (slope) of AMR was 0.83% (95% CI 0.73–0.93%) for 3GCR E. coli, 0.4% (95% CI 0.12–0.68%) for CR E. coli, −0.58% (95% CI −1.46% to 0.3%) for 3GCR K. pneumoniae, and 1.96% (95% CI 0.59–3.33%) for CR K. pneumoniae. The projected AMR prevalence in 2030 was 77% (95% CI 55–99.1%) for 3GCR E. coli, 11.8% (95% CI 3.7–19.9%) for CR E. coli, 58.2% (95% CI 50.2–66.1%) for 3GCR K. pneumoniae, and 52.8% (95% CI 16.3–89.3%) for CR K. pneumoniae. Projections for individual countries with at least four AMR point estimates using simple linear regression are presented in the Supplementary material.Figure 1 Forecast estimates with 95% confidence intervals of global resistance of Escherichia coli (A) and Klebsiella pneumoniae (B) to third-generation (3G) cephalosporins and carbapenems based on population weighted mixed models with random slopes and intercepts. Figure 1 Discussion The projections of AMR in this study signal a potentially serious shortage of effective antimicrobials for common causes of infection by 2030. Under current trends, over three-fourths of E. coli globally will be 3GCR, and over half of K. pneumoniae invasive isolates will be CR. The consequences of the high prevalence of AMR could be devastating for health systems (World Health Organization, 2014, de Kraker et al., 2011).

or common causes of infection by 2030. Under current trends, over three-fourths of E. coli globally will be 3GCR, and over half of K. pneumoniae invasive isolates will be CR. The consequences of the high prevalence of AMR could be devastating for health systems (World Health Organization, 2014, de Kraker et al., 2011). The models showed that the annual variation in the prevalence of 3GCR K. pneumoniae was not significantly different from zero, with narrowing of the confidence interval over time. This can be explained by the fact that countries with initial low prevalence of 3GCR K. pneumoniae showed a rising trend over time, while the trends were stable or mildly decreasing in countries with initial high prevalence, such as India and South Africa (Supplementary material, Figure S3). CR K. pneumoniae had the highest annual increase of AMR, which could reach 53% by 2030, but the confidence intervals were wide, indicating uncertainty of the projections. The projected increase in the prevalence of CR E. coli was more modest. However, empirical treatment of infections will need to cover 3GCR, leading to an increased use of carbapenems, and this, in turn, may accelerate the pace of CR.

30, but the confidence intervals were wide, indicating uncertainty of the projections. The projected increase in the prevalence of CR E. coli was more modest. However, empirical treatment of infections will need to cover 3GCR, leading to an increased use of carbapenems, and this, in turn, may accelerate the pace of CR. Enterobacteriaceae are part of the human gut microbiota, and the spread of AMR is facilitated by conditions that are more common in resource-poor settings, such as suboptimal sewage systems and a lack of access to clean water (Holmes et al., 2016). Previous studies have shown that resistance in Enterobacteriaceae can emerge anywhere and spread around the globe (Nordmann et al., 2011). Isolated interventions in high-income countries alone, without intervention efforts in low- and middle-income countries, may be ineffective in a globalized world (Nordmann et al., 2011).

ous studies have shown that resistance in Enterobacteriaceae can emerge anywhere and spread around the globe (Nordmann et al., 2011). Isolated interventions in high-income countries alone, without intervention efforts in low- and middle-income countries, may be ineffective in a globalized world (Nordmann et al., 2011). This study has important limitations. The total population of all countries included in the study was approximately a third of the world population and was biased towards high-income countries. While IPI models were used to attempt to correct for underrepresentation of low- and middle-income countries, more surveillance data are urgently needed to improve current and future estimates of AMR. The projections for future levels of AMR were based on linear models, which assumed no changes in the growth rate of resistance. They also did not account for saturation or stabilization of AMR levels, as was observed with 3GCR K. pneumoniae. In addition, it was not possible to distinguish the case mix of community- and hospital-acquired infections among the countries included in the study, and the high prevalence of AMR in some countries could be influenced by a higher proportion of hospital-acquired infections (Dat et al., 2017, Thaden et al., 2017).

iae. In addition, it was not possible to distinguish the case mix of community- and hospital-acquired infections among the countries included in the study, and the high prevalence of AMR in some countries could be influenced by a higher proportion of hospital-acquired infections (Dat et al., 2017, Thaden et al., 2017). These results suggest that if current trends were to continue, third-generation cephalosporins and carbapenems could become ineffective against E. coli and K. pneumoniae in most parts of the world in the not-too-distant future. Empirical antimicrobial therapy for sepsis or for urinary tract or abdominal infections might shift to non-beta-lactam antibiotics, which, in turn, may lead to an increase in AMR in other antibiotic groups. These results underscore the need to improve the judicious use of antimicrobials and support recent World Health Organization recommendations to prioritize the research, discovery, and development of new and effective antibiotic treatments for beta-lactam-resistant Enterobacteriaceae (WHO, 2018). Funding The funders had no role in the study design, in the collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the article for publication. Ethics approval This study used data available in the public domain and thus did not require ethics approval. Conflict of interest There are no conflicts of interest to disclose. Appendix A Supplementary data The following are Supplementary data to this article:

Funding The funders had no role in the study design, in the collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the article for publication. Ethics approval This study used data available in the public domain and thus did not require ethics approval. Conflict of interest There are no conflicts of interest to disclose. Appendix A Supplementary data The following are Supplementary data to this article: Acknowledgments The research leading to these results received support from the Bill & Melinda Gates Foundation for the ResistanceMap project (Investment ID OPP1112355) and from the Innovative Medicines Initiative Joint Undertaking under grant agreement number 115618 (Driving re-investment in R&D and responsible antibiotic use, DRIVE-AB, www.drive-ab.eu) (for SG, RL), resources of which are composed of financial contributions from the European Union Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations (EFPIA) companies in kind contribution. Appendix A Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.ijid.2018.01.011.

Background Environmental enteropathy is a widespread subclinical disturbance found in low-income countries in individuals exposed over time to poor sanitation and hygiene. It has been identified as a possible cause of stunting and malnutrition, oral vaccine failure and impaired development in children from low-income countries. Although much about its cause remains unknown, it has been proposed to result from increased T-cell activation arising from the repeated exposure to insanitary conditions in many poor regions of the world (Korpe and Petri, 2012, Veitch et al., 2001). Several pathways have been proposed, all leading to specific morphological and functional derangements involving intestinal mucosal disruption and increased susceptibility to infection (Naylor et al., 2015, Prendergast and Kelly, 2012).

ons in many poor regions of the world (Korpe and Petri, 2012, Veitch et al., 2001). Several pathways have been proposed, all leading to specific morphological and functional derangements involving intestinal mucosal disruption and increased susceptibility to infection (Naylor et al., 2015, Prendergast and Kelly, 2012). Matrix metalloproteinases (MMPs) are a group of zinc dependent enzymes which degrade structural proteins. Specifically it has been documented that MMP-1 and −8 degrade many types of collagen as well as proteoglycans, while MMP-2 and −9 degrade gelatin, type IV collagen, fibronectin and elastins (Lee, 2014, Ravi et al., 2007). They have been implicated in the pathogenesis of inflammatory bowel disease due to their role in tissue resorption and re-modelling of the extracellular matrix (O’Sullivan et al., 2015, Ravi et al., 2007). It has been suggested that the degradation of both extracellular and intracellular structural molecules by MMPs can contribute to impairment of the intestinal epithelial barrier seen in inflammatory bowel disease. MMPs are activated by growth factors, hormones, various cytokines, and cellular transformation (de Bruyn et al., 2016).

uggested that the degradation of both extracellular and intracellular structural molecules by MMPs can contribute to impairment of the intestinal epithelial barrier seen in inflammatory bowel disease. MMPs are activated by growth factors, hormones, various cytokines, and cellular transformation (de Bruyn et al., 2016). Nontuberculous mycobacteria are common environmental pathogens that are frequently being found to be responsible for disease in humans (Falkinham, 2002, Mirsaeidi et al., 2013). Given the ubiquity of NTM, their constant passage through the gut (Chongwe et al., 2017) and their ability to induce various cytokines through T-cell activation, notwithstanding that the exact mechanism for enteropathy is still unknown, we hypothesised that these NTMs could play a role in the pathogenesis. We set out to investigate the role of Mycobacterium avium in inducing inflammatory cytokine secretion as well as the secretion of MMPs in human gut biopsies in a population known to have widespread environmental enteropathy (Kelly et al., 2016, Kelly et al., 2004).

that these NTMs could play a role in the pathogenesis. We set out to investigate the role of Mycobacterium avium in inducing inflammatory cytokine secretion as well as the secretion of MMPs in human gut biopsies in a population known to have widespread environmental enteropathy (Kelly et al., 2016, Kelly et al., 2004). Methods Patients and recruitment We recruited participants from among consenting adults 18 years and older reporting for routine endoscopy at the University Teaching Hospital, Lusaka between July and December 2014. Selection of participants was based on a simple random sampling from a list of booked patients on each clinic day. A structured questionnaire was used to extract baseline data and to collect information on possible risk factors for tuberculosis or nontuberculous mycobacterial infection. This included collection of information on HIV status as well as having a previous Bacillus Calmette–Guerin (BCG), vaccination (verified by a scar on the arm) and self-reported history of TB.

ct baseline data and to collect information on possible risk factors for tuberculosis or nontuberculous mycobacterial infection. This included collection of information on HIV status as well as having a previous Bacillus Calmette–Guerin (BCG), vaccination (verified by a scar on the arm) and self-reported history of TB. Biopsy collection and processing Diagnostic endoscopy procedures were performed using Pentax EG2990i endoscopes (Pentax, Tokyo, Japan). During endoscopy three biopsies were taken from the second part of the duodenum, in those patients in whom the endoscopy was found to be normal, and immediately placed into a 2 ml cryovial filled with 1 ml culture media composed of five volumes Dulbecco’s modified Eagle’s medium, five volumes National Cancer Tissue Culture-135 medium and one volume of newborn calf serum, all from Sigma Aldrich, (Dorset, UK). The biopsies were cultured with M. avium lysate in tissue culture medium within two hours in an environment with 95% O2/5% CO2 at 37 °C for 24 h as previously described (Dhaliwal et al., 2003, Dhaliwal et al., 2009). For each participant, the first biopsy was placed on a centre well culture dish (Sigma Aldrich, Dorset, UK) with no stimulant added and this served as a negative control, the second biopsy was stimulated with 10 μl of M. avium lysate while the third biopsy was stimulated with 10 μl Staphylococcus enterotoxin B (SEB) antigen. An additional positive control to confirm the validity of the experimental model using Salmonella typhimurium lipopolysaccharide (LPS) was also used (results not shown).

econd biopsy was stimulated with 10 μl of M. avium lysate while the third biopsy was stimulated with 10 μl Staphylococcus enterotoxin B (SEB) antigen. An additional positive control to confirm the validity of the experimental model using Salmonella typhimurium lipopolysaccharide (LPS) was also used (results not shown). Assays for MMPs and cytokines After 24 h of incubation, the supernatant was collected and stored at −80 °C. To quantify cytokines, the human cytometric bead array Th1/Th2/Th17A kit (Becton, Dickinson and Company, San Jose, California, USA) was used to detect interleukin 2 (IL-2), interleukin 4 (IL-4), interleukin 6 (IL-6), interleukin 10 (IL-10), interleukin 17A (IL-17A), Tumour Necrosis Factor alpha (TNF-α) and Interferon Gamma (IFN-γ). This was performed on a BD FACSVerse flow cytometer, and analysed using BD FCAP array software. Measurement of IL-1β and IL-12 were performed by ELISA as they are not included in the CBA kit. MMP-1, -2, -8, and -9 were also quantified using ELISA assays. Information was also collected on history of tuberculosis (TB) and having a BCG scar, while HIV testing was done on all participants. Cell viability (results not shown) was assessed using the MTT-based in vitro toxicology assay kit (Sigma, Dorset, UK) which confirmed viability of the tissue during incubation. The results were exported into Stata 14 software (College Station, Texas, USA) for analysis. A subset of 13 samples chosen at random were used to quantify levels of MMP-1, -2, -8 and -9 secretion using ELISA methods (R&D Systems, Abingdon, UK).

gma, Dorset, UK) which confirmed viability of the tissue during incubation. The results were exported into Stata 14 software (College Station, Texas, USA) for analysis. A subset of 13 samples chosen at random were used to quantify levels of MMP-1, -2, -8 and -9 secretion using ELISA methods (R&D Systems, Abingdon, UK). Analysis MMP or cytokine quantities were expressed in nanograms per ml for MMPs and picograms per ml for other cytokines and this data was entered in Excel and analysed using Stata version 14 (Stata Corp., College Station, Texas). The Shapiro–Wilk test for normality showed that these data were non-normally distributed and so the Wilcoxon signed rank test was used to compare cytokines or MMP concentrations in supernatants from M. avium stimulated samples and unstimulated control samples from the same individuals. Spearman’s correlation was used to check for association between and among the different cytokines and MMPs. Secretion of cytokines or MMPs was stratified by HIV status, having a previous Bacillus Calmette–Guerin (BCG) vaccination (verified by a scar on the arm) and self-reported history of TB. The difference in cytokine secretion between each M. avium-stimulated sample and their negative control was estimated using panel random effects multivariable linear regression analysis with maximum likelihood estimation (xtreg in Stata), using an investigator-led backward regression. Akaike and Bayesian information criteria were used for model selection, and a p-value of less than 0.05 was considered significant. Positive controls were included in all experiments but used only to confirm the viability of each experiment and were not included in pairwise comparisons. The study was approved by the University of Zambia Biomedical Research Ethics Committee (Reference # 015-07-12).

value of less than 0.05 was considered significant. Positive controls were included in all experiments but used only to confirm the viability of each experiment and were not included in pairwise comparisons. The study was approved by the University of Zambia Biomedical Research Ethics Committee (Reference # 015-07-12). Results Overall (n = 48) we recruited 21 men and 27 women whose median age was 35 (IQR 27.5, 50.5) years; women were slightly older than men (Table 1). Six (13%) of the participants were HIV positive. Most of the participants (74%) had evidence of previous BCG vaccination, while 15% had a previous history of TB.Table 1 Characteristics of participants recruited for the in-vitro study to determine the secretion of cytokines and MMPs in duodenal samples from 48 individuals. Table 1Variable Overall [n (%)] Female [n (%)] Male [n (%)] Age* 35.0 (27.5, 50.5)* 35.5 (26, 52.5)* 35.0 (29.5, 41.0)* Sex [n (%)] 48 27 (54) 21 (46) BCG scar Present 23 10 (44) 13 (56) History of TB Present 7 3 (43) 4 (57) HIV status Seropositive 6 4 (67) 2 (33) Presenting symptoms Abdominal pain 39 (81) 22 (56) 17 (44) Vomiting 8 (17) 5 (62) 3 (38) Diarrhoea 7 (15) 3 (43) 4 (57) Night sweats 4 (8.3) 1 (25) 3 (75) Cough 3 (7.4) 2 (67) 1 (33) Note: *Age is expressed in median (Interquartile Range, IQR).

History of TB Present 7 3 (43) 4 (57) HIV status Seropositive 6 4 (67) 2 (33) Presenting symptoms Abdominal pain 39 (81) 22 (56) 17 (44) Vomiting 8 (17) 5 (62) 3 (38) Diarrhoea 7 (15) 3 (43) 4 (57) Night sweats 4 (8.3) 1 (25) 3 (75) Cough 3 (7.4) 2 (67) 1 (33) Note: *Age is expressed in median (Interquartile Range, IQR). The most common presenting symptom was abdominal pain, which was the indication for endoscopy in 39 (82%) patients. Other symptoms present at the time of endoscopy were vomiting (17%), diarrhoea (15%) and cough (7.4%) as shown in Table 1. The majority of tissue, therefore, came from patients with functional dyspepsia.

Presenting symptoms Abdominal pain 39 (81) 22 (56) 17 (44) Vomiting 8 (17) 5 (62) 3 (38) Diarrhoea 7 (15) 3 (43) 4 (57) Night sweats 4 (8.3) 1 (25) 3 (75) Cough 3 (7.4) 2 (67) 1 (33) Note: *Age is expressed in median (Interquartile Range, IQR). The most common presenting symptom was abdominal pain, which was the indication for endoscopy in 39 (82%) patients. Other symptoms present at the time of endoscopy were vomiting (17%), diarrhoea (15%) and cough (7.4%) as shown in Table 1. The majority of tissue, therefore, came from patients with functional dyspepsia. M. avium induces IL-1β and IL-6 secretion in duodenal biopsies Interleukin 1β and IL-6 were significantly higher (p = 0.002 and 0.04 respectively) in supernatants from duodenal biopsies from the same individuals cultured with M. avium than from controls (Table 2). The secretion of IL-6 in M. avium stimulated samples was higher in patients with a history of TB than those without a history of TB (Figure 1). Intestinal tissue from HIV negative participants secreted more IL-6 but, not in any other cytokines. Induction of IL-2, IL-6, IL-10 (not shown), IL-17A, IFN-g and was higher in participants with BCG scars, but only IL-6 were induced more in participants with no past history of tuberculosis (Figure 1). Stratifying the IL-1, IL4 and IL-12 by HIV status, having a BCG scar or history of TB did not show any differences in our patients. Similarly, the secretion of the cytokines did not show any differences when stratified by age group and sex.Figure 1 Secretion of some gut inflammatory cytokines among endoscopy patients in Lusaka, stratified by HIV, BCG and history of TB.

having a BCG scar or history of TB did not show any differences in our patients. Similarly, the secretion of the cytokines did not show any differences when stratified by age group and sex.Figure 1 Secretion of some gut inflammatory cytokines among endoscopy patients in Lusaka, stratified by HIV, BCG and history of TB. Notes: p values comparing Mycobacterium avium versus negative controls (NC) calculated using the Wilcoxon signed rank test as sets of biopsies were collected from the same individuals. Figure 1Table 2 Secretion of cytokines by duodenal explants into supernatant during stimulation with Mycobacterium avium lysate.

having a BCG scar or history of TB did not show any differences in our patients. Similarly, the secretion of the cytokines did not show any differences when stratified by age group and sex.Figure 1 Secretion of some gut inflammatory cytokines among endoscopy patients in Lusaka, stratified by HIV, BCG and history of TB. Notes: p values comparing Mycobacterium avium versus negative controls (NC) calculated using the Wilcoxon signed rank test as sets of biopsies were collected from the same individuals. Figure 1Table 2 Secretion of cytokines by duodenal explants into supernatant during stimulation with Mycobacterium avium lysate. Table 2Cytokine* Number Amount secreted after stimulation with M. avium (pg/ml)a Amount secreted by the negative control (pg/ml)a P valueb Interleukin 1β 13 23.6 (10.0, 83.0) 0.14 (0.00, 1.36) 0.002§ Interleukin 2 30 2.18 (0.34, 4.32) 0.26 (0.00, 4.66) 0.25 Interleukin 4 30 0.18 (0.00, 0.70) 0.18 (0.00, 1.84) 0.31 Interleukin 6 30 77.1 (0.94, 408.4) 16.2 (0.00, 104.4) 0.04 Interleukin 10 30 0.08 (0.00, 0.49) 0.01 (0.00, 0.92) 0.36 Interleukin 12 13 7.84 (5.10, 15.8) 9.48 (4.29, 14.6) 0.97 Interleukin 17A 30 12.0 (2.68, 28.3) 1.15 (0.00, 19.6) 0.25 TNF – α 30 0.31 (0.00, 2.34) 0.24 (0.00, 2.71) 0.91 Interferon γ 30 1.27 (0.28, 3.16) 0.23 (0.00, 3.39) 0.57 Notes: aAll the figures are medians with interquartile range, in pg/ml. bp-value calculated using the Wilcoxon signed-rank test. *Except for IL-1Β and 12, for which ELISA was used, the other cytokines were quantified using flow cytometry via the cytometric bead array. §: Boldface figure signifies statistical significance.

es: aAll the figures are medians with interquartile range, in pg/ml. bp-value calculated using the Wilcoxon signed-rank test. *Except for IL-1Β and 12, for which ELISA was used, the other cytokines were quantified using flow cytometry via the cytometric bead array. §: Boldface figure signifies statistical significance. In multivariable regression analysis of data from duodenal samples (Table 3), the results showed that M. avium-stimulated duodenal samples expressed 156 pg of IL-6 more than unstimulated samples (95% CI 7.6, 304; p = 0.04) after adjusting for the effect of history of TB, previous BCG vaccination, age and sex. We did not detect any differences in secretion of IL-17A, IL-10, IL-4, IL-2, IFN-γ and TNF-α in the gut between stimulated samples and unstimulated samples.Table 3 Panel linear regression analyses showing the effect of Mycobacterium avium stimulation on secretion of each cytokine. Table 3Cytokine Unadjusted Adjusted Mean cytokine secretion (SD) [pg/ml] Mean difference (95%CI) p value* Mean difference (95%CI) p value* Intra-cluster correlation coefficient (Rho) IL-17A Not stimulated 19.3 (30.2) Ref 0.14 Ref 0.07 – Stimulated with M. avium 20.5 (24.5) 1.72 (−8.56, 12.0) 11.8 (−0.97, 24.7)1 IL-10 Not stimulated 0.60 (1.01) Ref 0.64 Ref 0.52 0.33 Stimulated with M. avium 0.57 (1.24) 0.09 (−0.27, 0.44) 0.08 (−0.17, 0.35)2 IL-6 Not stimulated 202.0 (365.5) Ref 0.36 Ref 0.04§ 0.42 Stimulated with M. avium 271.7 (440.1) 74.6 (−84.0, 233.5) 155.8 (7.64, 304.0)3

Table 3Cytokine Unadjusted Adjusted Mean cytokine secretion (SD) [pg/ml] Mean difference (95%CI) p value* Mean difference (95%CI) p value* Intra-cluster correlation coefficient (Rho) IL-17A Not stimulated 19.3 (30.2) Ref 0.14 Ref 0.07 – Stimulated with M. avium 20.5 (24.5) 1.72 (−8.56, 12.0) 11.8 (−0.97, 24.7)1 IL-10 Not stimulated 0.60 (1.01) Ref 0.64 Ref 0.52 0.33 Stimulated with M. avium 0.57 (1.24) 0.09 (−0.27, 0.44) 0.08 (−0.17, 0.35)2 IL-6 Not stimulated 202.0 (365.5) Ref 0.36 Ref 0.04§ 0.42 Stimulated with M. avium 271.7 (440.1) 74.6 (−84.0, 233.5) 155.8 (7.64, 304.0)3 IL-4 Not stimulated 1.31 (2.43) Ref 0.46 Ref 0.60 0.02 Stimulated with M. avium 0.93 (1.70) −0.26 (−0.93, 0.42) −0.06 (−0.16, 0.28)4 IL-2 Not stimulated 3.46 (7.98) Ref 0.82 Ref 0.36 0.75 Stimulated with M. avium 6.46 (23.6) 0.28 (−2.16, 2.71) 3.84 (−4.47, 12.1)5 IFNG Not stimulated 3.52 (8.33) Ref 0.74 Ref 0.45 0.64 Stimulated with M. avium 3.68 (9.02) −0.41 (−2.80, 1.99) 1.41 (−2.23, 5006)6

IL-4 Not stimulated 1.31 (2.43) Ref 0.46 Ref 0.60 0.02 Stimulated with M. avium 0.93 (1.70) −0.26 (−0.93, 0.42) −0.06 (−0.16, 0.28)4 IL-2 Not stimulated 3.46 (7.98) Ref 0.82 Ref 0.36 0.75 Stimulated with M. avium 6.46 (23.6) 0.28 (−2.16, 2.71) 3.84 (−4.47, 12.1)5 IFNG Not stimulated 3.52 (8.33) Ref 0.74 Ref 0.45 0.64 Stimulated with M. avium 3.68 (9.02) −0.41 (−2.80, 1.99) 1.41 (−2.23, 5006)6 TNF Not stimulated 4.03 (10.9) Ref 0.40 Ref 0.97 0.74 Stimulated with M. avium 2.53 (5.61) −1.32 (−4.43, 1.79) −0.01 (−0.39, 0.41)7 Notes: *p value was calculated using panel random effects linear regression with maximum likelihood estimation. §: Boldface figure signifies statistical significance. ICC: Intracluster Correlation Coefficient; each row represents a separate multiple linear regression model. 1: adjusted for BCG vaccination, age and history of TB; 2: adjusted for sex, age, history of TB, HIV and previous BCG vaccination; 3: adjusted for sex, age, history of TB and previous BCG vaccination; 4: adjusted for sex, age, history of TB, HIV and previous BCG vaccination; 5: adjusted for sex, age, history of TB, previous BCG vaccination; 6: adjusted for sex, age and BCG vaccination; 7: adjusted for sex and BCG vaccination.

ious BCG vaccination; 3: adjusted for sex, age, history of TB and previous BCG vaccination; 4: adjusted for sex, age, history of TB, HIV and previous BCG vaccination; 5: adjusted for sex, age, history of TB, previous BCG vaccination; 6: adjusted for sex, age and BCG vaccination; 7: adjusted for sex and BCG vaccination. M. avium induces the secretion of Th1 and Th2 response but not Th17 response in whole blood M. avium lysate stimulation of whole blood samples showed that M. avium induced the secretion of much higher levels of cytokines than those produced by duodenal tissue. Unlike in the gut, M. avium significantly increased whole blood secretion of the regulatory cytokine IL-10, (p < 0.001), and the pro-inflammatory cytokines IL-2, IL-6, TNF-α and IFN-γ were all significantly higher in stimulated samples compared to unstimulated samples (Figure 2). IL-17A was not induced by M. avium stimulation. Multiple linear regression (Table 5) showed that M. avium induced the secretion of Th1 and Th2 cytokines but not Th17 cytokines in whole blood in this experiment.Figure 2 Secretion of cytokines in whole blood after stimulation with Mycobacterium avium complex. Notes: p values comparing Mycobacterium avium versus negative controls (NC) calculated using the Wilcoxon matched pairs signed rank test. Figure 2

M. avium induces the secretion of Th1 and Th2 response but not Th17 response in whole blood M. avium lysate stimulation of whole blood samples showed that M. avium induced the secretion of much higher levels of cytokines than those produced by duodenal tissue. Unlike in the gut, M. avium significantly increased whole blood secretion of the regulatory cytokine IL-10, (p < 0.001), and the pro-inflammatory cytokines IL-2, IL-6, TNF-α and IFN-γ were all significantly higher in stimulated samples compared to unstimulated samples (Figure 2). IL-17A was not induced by M. avium stimulation. Multiple linear regression (Table 5) showed that M. avium induced the secretion of Th1 and Th2 cytokines but not Th17 cytokines in whole blood in this experiment.Figure 2 Secretion of cytokines in whole blood after stimulation with Mycobacterium avium complex. Notes: p values comparing Mycobacterium avium versus negative controls (NC) calculated using the Wilcoxon matched pairs signed rank test. Figure 2 MMPs Having observed these changes in cytokine secretion in blood, and IL-6 in intestinal tissue, we went on to test the hypothesis that MMPs might be induced alongside them in response to M. avium lysate. From among the 48 participants, we took a non-sex differentiated random sub-sample of 14 participants and this was based on what could be handled by the ELISA kit. It turned out that there were five males and eight females in this sub-sample. The median age for this group was 32.5 years (IQR 27, 36), with males having a median age of 31.5 (IQR 27, 36) years and females with 34 (IQR 25, 40) years. They were therefore similar to the whole group.

t could be handled by the ELISA kit. It turned out that there were five males and eight females in this sub-sample. The median age for this group was 32.5 years (IQR 27, 36), with males having a median age of 31.5 (IQR 27, 36) years and females with 34 (IQR 25, 40) years. They were therefore similar to the whole group. The secretion of MMP-1 in supernatants from duodenal tissue stimulated with M. avium was increased compared to the negative controls, but secretion of MMP-2, -8 and -9 in M. avium stimulated tissues was not significantly increased (Table 4).Table 4 Panel linear regression analysis showing the effect of M. avium stimulation on cytokine secretion in whole blood. Table 4Cytokine Unadjusted Adjusted Mean cytokine secretion (SD) [pg/ml] Mean difference (95%CI) p value§ Mean difference (95%CI) P value Intra-cluster correlation coefficient (rho) IL-17A Not stimulated 17.2 (14.4) Ref 0.63 Ref 0.91 0.002 Stimulated with M. avium 19.4 (24.1) 2.19 (−6.76, 11.2) 0.52 (−8.60, 9.64)1 IL-10 Not stimulated 9.59 (17.5) Ref <0.01§ Ref <0.01 0.00 Stimulated with M. avium 434.2 (641.0) 422.3 (203.1, 641.6) 432.3 (213.3, 651.4)2 IL-6 Not stimulated 2028.6 (2666.8) Ref <0.01 Ref <0.01 0.19 Stimulated with M. avium 7511.2 (2513.5) 5447.2 (4370.8, 6523.6) 5423.2 (4315.8, 6530.7)3 IL-4 Not stimulated 0.43 (1.22) Ref 0.05 Ref 0.04 0.00 Stimulated with M. avium 1.58 (3.59) 1.17 (−0.01, 2.34) 1.27 (0.07, 2.46)4 IL-2 Not stimulated 11.1 (51.5) Ref 0.03 Ref 0.02 0.00 Stimulated with M. avium 271.9 (669.9) 260.7 (31.45, 490.0) 277 (54.21, 500.0)5

IL-6 Not stimulated 2028.6 (2666.8) Ref <0.01 Ref <0.01 0.19 Stimulated with M. avium 7511.2 (2513.5) 5447.2 (4370.8, 6523.6) 5423.2 (4315.8, 6530.7)3 IL-4 Not stimulated 0.43 (1.22) Ref 0.05 Ref 0.04 0.00 Stimulated with M. avium 1.58 (3.59) 1.17 (−0.01, 2.34) 1.27 (0.07, 2.46)4 IL-2 Not stimulated 11.1 (51.5) Ref 0.03 Ref 0.02 0.00 Stimulated with M. avium 271.9 (669.9) 260.7 (31.45, 490.0) 277 (54.21, 500.0)5 IFNG Not stimulated 5.26 (17.9) Ref <0.01 Ref <0.01 0.00 Stimulated with M. avium 241.86 (424.3) 236.6 (91.65, 381.5) 216.5 (82.64, 350.3)6 TNF Not stimulated 197.0 (594.6) Ref <0.01 Ref <0.01 0.42 Stimulated with M. avium 1366.9 (1673.8) 1098.2 (664.0, 1532.4) 1083.2 (634.2, 1532.3)7 Notes: §p value was calculated using panel random effects linear regression with the maximum likelihood estimation. §: Boldface figure signifies statistical significance. Each row represents a separate multiple linear regression model. 1: adjusted for history of TB and sex; 2: adjusted for sex, age, history of TB, HIV and previous BCG vaccination; 3: history of TB; 4: adjusted for sex, history of TB and previous BCG vaccination; 5: adjusted for sex and history of TB; 6: adjusted for sex, and History of TB; 7: adjusted for sex, age, history of TB, HIV and previous BCG vaccination. Table 5 Secretion of MMPs in duodenal tissue in response to M. avium stimulation.

TNF Not stimulated 197.0 (594.6) Ref <0.01 Ref <0.01 0.42 Stimulated with M. avium 1366.9 (1673.8) 1098.2 (664.0, 1532.4) 1083.2 (634.2, 1532.3)7 Notes: §p value was calculated using panel random effects linear regression with the maximum likelihood estimation. §: Boldface figure signifies statistical significance. Each row represents a separate multiple linear regression model. 1: adjusted for history of TB and sex; 2: adjusted for sex, age, history of TB, HIV and previous BCG vaccination; 3: history of TB; 4: adjusted for sex, history of TB and previous BCG vaccination; 5: adjusted for sex and history of TB; 6: adjusted for sex, and History of TB; 7: adjusted for sex, age, history of TB, HIV and previous BCG vaccination. Table 5 Secretion of MMPs in duodenal tissue in response to M. avium stimulation. Table 5Cytokine Stimulated with M. avium (ng/ml)a Negative control (ng/ml)a p valueb MMP-1 1.23 (0.02, 3.95) 0.04 (0.01, 0.97) 0.002 MMP-8 0.23 (0.00, 0.49) 0.00 (0.00, 0.23) 0.06 MMP 2 0.36 (0.07, 0.85) 0.23 (0.02, 0.26) 0.12 MMP-9 0.98 (0.00, 2.25) 0.56 (0.00, 1.38) 0.50 Notes: n = 13. aAll the figures are medians with interquartile range, in ng/ml. bp-value calculated using the Wilcoxon signed-rank test.

1.23 (0.02, 3.95) 0.04 (0.01, 0.97) 0.002 MMP-8 0.23 (0.00, 0.49) 0.00 (0.00, 0.23) 0.06 MMP 2 0.36 (0.07, 0.85) 0.23 (0.02, 0.26) 0.12 MMP-9 0.98 (0.00, 2.25) 0.56 (0.00, 1.38) 0.50 Notes: n = 13. aAll the figures are medians with interquartile range, in ng/ml. bp-value calculated using the Wilcoxon signed-rank test. We analysed the changes in secretion of all the MMPs according to HIV status, having a BCG scar, gender and age group. The induction of MMP-1 was greatest in those who were HIV negative, in those with BCG vaccination scars, in younger participants and in women (Figure 3). However stratified analysis for MMP-2, -8 and -9 did not show any differences.Figure 3 Stratified analysis of MMP-1 against HIV, BCG vaccination status, sex and age. Figure 3 The secretion of MMP-1 in whole blood stimulated with M. avium was significantly higher than in unstimulated blood, p = 0.02. However, the level of MMP-1 expressed in blood was lower than that expressed by duodenal samples. When the subgroups HIV, previous BCG vaccination and history of TB were considered, the patterns of secretion of MMP-1 in the blood was similar to that expressed by duodenal samples.

r than in unstimulated blood, p = 0.02. However, the level of MMP-1 expressed in blood was lower than that expressed by duodenal samples. When the subgroups HIV, previous BCG vaccination and history of TB were considered, the patterns of secretion of MMP-1 in the blood was similar to that expressed by duodenal samples. Correlation analysis Using Spearman’s correlation analysis, duodenal MMP-1 secretion was positively correlated with that of IL-6 (ρ = 0.70, p = 0.01), while duodenal IL-17A secretion was correlated with that of IL-10, IL-6, IL-4, IL-2 interferon gamma and TNF (Table 6). Interleukin 10 secretion was correlated with the secretion of IL-4 and that of IL-2 (Table 6).Table 6 Spearman’s correlation Matrix showing the association among Cytokines and Matrix Metalloproteinases in intestinal tissue. Table 6 IL-17A IL-17A 1 IL-12 IL-12 – 1 IL-10 IL-10 0.57 – 1 IL-6 IL-6 0.52 – 0.33 1 IL-4 IL-4 0.58 – 0.56 – 1 IL-2 IL-2 0.58 – 0.54 0.47 0.48 1 IL-1β IL-1β – – – – – – 1 INF-γ INF-γ 0.66 – 0.54 0.42 0.58 0.80 – 1 TNF-α TNF- α 0.51 – 0.50 – 0.74 0.39 – 0.56 1 MMP-1 MMP-1 – – – 0.70 – – – – – 1 MMP-2 MMP-2 – – – – – – – – – 1 MMP-8 MMP-8 – 0.61 – – – – 0.59 – – 0.60 – 1 MMP-9 MMP-9 – – – 0.73 – – – – – 0.90 – – 1 Note: the matrix is showing only correlations that were statistically significant.

NF-γ 0.66 – 0.54 0.42 0.58 0.80 – 1 TNF-α TNF- α 0.51 – 0.50 – 0.74 0.39 – 0.56 1 MMP-1 MMP-1 – – – 0.70 – – – – – 1 MMP-2 MMP-2 – – – – – – – – – 1 MMP-8 MMP-8 – 0.61 – – – – 0.59 – – 0.60 – 1 MMP-9 MMP-9 – – – 0.73 – – – – – 0.90 – – 1 Note: the matrix is showing only correlations that were statistically significant. Discussion Our results have shown that a lysate of Mycobacterium avium bacteria induced the secretion of IL-1β and MMP-1 in duodenal biopsies of healthy patients undergoing endoscopy. MMPs have long been implicated in the pathogenesis of inflammatory bowel disease, as well as other diseases that are characterized by destruction of the extracellular tissue such as rheumatoid arthritis, periodontal disease (Naito and Yoshikawa, 2005, Quiding-Järbrink et al., 2001) and tuberculosis. Although we did not directly investigate a link between enteropathy and MMPs in this study, this finding suggests a mechanism by which M. avium, through induction of MMP-1 (Biancheri et al., 2013, Giuffrida et al., 2014, Steck et al., 2012), could at least be partially responsible for some of the morphological changes seen in enteropathy. MMP-1 has previously been shown to drive the immunopathological process in Mycobacterium tuberculosis infection in the lung (Elkington et al., 2011). Lamina propria mononuclear cells in the duodenum have been shown to express multiple MMPs after cytokine stimulation (Ciccocioppo et al., 2005).

nges seen in enteropathy. MMP-1 has previously been shown to drive the immunopathological process in Mycobacterium tuberculosis infection in the lung (Elkington et al., 2011). Lamina propria mononuclear cells in the duodenum have been shown to express multiple MMPs after cytokine stimulation (Ciccocioppo et al., 2005). Clearly, the use of tissue explants in vitro does not fully represent the real impact of mycobacterial pathogen-associated molecular patterns (PAMPs) on the mucosal immune response in vivo. There are several limitations to this experimental system – the tissue is traumatised during biopsy, the range of mycobacterial molecules used for stimulation is limited, the time course of the experiment is limited and fixed, and we were unable to perform a full range of dose ranging experiments as the amount of tissue available was limited. Nevertheless, the tissue is human intestinal tissue and the mucosa being stimulated has the full complement of epithelial, stromal and immune cells. Each of the tissues from the same individual were subjected to different stimuli, namely nothing (as a negative control), SEB and S. typhimurium LPS (as a positive controls) and M. avium as the main experimental stimuli. This implies that any differences in the secretion of cytokines between the negative control and M. avium stimulated tissues would be solely due to the M. avium. Furthermore, the results of the cytotoxicity assay done on the intestinal samples were within normal range. It is therefore reasonable to assume that the secretion of the different cytokines seen in this study was due to the stimulation by M. avium.

l and M. avium stimulated tissues would be solely due to the M. avium. Furthermore, the results of the cytotoxicity assay done on the intestinal samples were within normal range. It is therefore reasonable to assume that the secretion of the different cytokines seen in this study was due to the stimulation by M. avium. There is already existing evidence suggesting that environmental enteropathy is virtually ubiquitous in communities of low socio-economic status in Africa (Kelly et al., 2004, Lindenbaum et al., 1972, Prendergast and Kelly, 2012), so we postulated that non-tuberculous mycobacteria, which are also virtually ubiquitous, could contribute to intestinal mucosal re-modelling. M. avium is just one of the NTMs which could be implicated, and further work is required to ascertain if the effects we report here might be observed with other species.

, 2012), so we postulated that non-tuberculous mycobacteria, which are also virtually ubiquitous, could contribute to intestinal mucosal re-modelling. M. avium is just one of the NTMs which could be implicated, and further work is required to ascertain if the effects we report here might be observed with other species. We were unable to demonstrate the effect of M. avium on the secretion of MMPs 2, 8 or 9 in duodenal tissue, though it is possible that MMP-8 might have been significant in a larger study. It has been shown that MMP-1, -3 and -9 are involved in the pathogenesis of gluten sensitive enteropathy through tissue remodelling (Mohamed et al., 2006). In the lungs, it is well known that MMP-1 and MMP-9 are involved in the pathogenesis of TB. For MMP-9, this occurs through its effect on recruitment of macrophages and role in the formation of granulomas (Taylor et al., 2006) in addition to its effect on extracellular matrix. Mycobacterium avium paratuberculosis has been shown to enhance the expression MMP-2, -9, -13, -14, and Tissue Inhibitor of Metalloproteinase -1 (TIMP-1), in addition to IL1β and TNF-α in a murine model (Roderfeld et al., 2012).

ation of granulomas (Taylor et al., 2006) in addition to its effect on extracellular matrix. Mycobacterium avium paratuberculosis has been shown to enhance the expression MMP-2, -9, -13, -14, and Tissue Inhibitor of Metalloproteinase -1 (TIMP-1), in addition to IL1β and TNF-α in a murine model (Roderfeld et al., 2012). We were also unable to demonstrate the production of IL-10 or IL-4 in duodenal samples, although we showed significant amounts of both cytokines in blood (Table 5). It is generally accepted that gut microbiota provide continuous antigenic stimulation that leads to activation of T-cells leading to intestinal injury (Sartor and Mazmanian, 2012), Gut microbiota have been implicated in the induction of regulatory B cells in the spleen and mesenteric lymph nodes through the production of IL-1β and IL-6. This inflammatory response in turn leads to the production of the anti-inflammatory IL-10 (Rosser et al., 2014).

o intestinal injury (Sartor and Mazmanian, 2012), Gut microbiota have been implicated in the induction of regulatory B cells in the spleen and mesenteric lymph nodes through the production of IL-1β and IL-6. This inflammatory response in turn leads to the production of the anti-inflammatory IL-10 (Rosser et al., 2014). We found evidence that M. avium induced the secretion of IL-1β in intestinal tissue. IL-1 is a pro-inflammatory cytokine that is activated via a variety of microbial and non-microbial mechanisms including by Mycobacterium avium (Elkington et al., 2011, Weber et al., 2010). IL-1β is known to be a potent stimulator of the extracellular tissue to produce MMPs including MMP-1, leading to tissue re-modelling (Dinarello et al., 2012, Garlanda et al., 2013). It has been postulated that the tissue destruction seen in environmental enteropathy could be as a result of constant activation of T-cells caused by the presence of intestinal pathogens in the lumen (Korpe and Petri, 2012, Prendergast and Kelly, 2012).

ssue re-modelling (Dinarello et al., 2012, Garlanda et al., 2013). It has been postulated that the tissue destruction seen in environmental enteropathy could be as a result of constant activation of T-cells caused by the presence of intestinal pathogens in the lumen (Korpe and Petri, 2012, Prendergast and Kelly, 2012). This study showed that M. avium-stimulated duodenal tissue secreted IL-6 more than unstimulated controls. IL-6 is a potent pleotropic inflammatory cytokine known to be involved in epithelial proliferation and wound repair (Kuhn et al., 2014), and has been shown to induce the secretion of MMP-1, -2 and -9 (Kothari et al., 2014, Sengupta and MacDonald, 2007). It has also been found to be involved in the pathogenesis of inflammatory bowel disease (Wang et al., 2003). In the present study, there was a strong correlation (ρ = 0.70) between IL-6 and MMP-1, suggesting that this could be one mechanism through which NTMs could lead to tissue re-modelling in enteropathy.

ld, 2007). It has also been found to be involved in the pathogenesis of inflammatory bowel disease (Wang et al., 2003). In the present study, there was a strong correlation (ρ = 0.70) between IL-6 and MMP-1, suggesting that this could be one mechanism through which NTMs could lead to tissue re-modelling in enteropathy. In conclusion, we found that M. avium induced the secretion of MMP-1 in duodenal tissue and in peripheral blood. M. avium also induced the secretion of a restricted set of cytokines in duodenal tissue, namely IL-1β and IL-6 as well as eliciting a Th1 and Th2 response in the blood. We speculate that the induction of these cytokines by M. avium suggests a possible pathway through which NTMs, and M. avium in particular, could remodel the mucosa and lead to environmental enteropathy. The induction of MMP-1 and cytokines by M. avium in small intestinal tissue from a tropical population suggests that environmental mycobacteria could contribute to the epithelial disruption seen in environmental enteropathy in exposed populations. Further work will be required to demonstrate if MMP-mediated mucosal re-modelling actually operates in vivo. Competing interests No competing interests declared.

In conclusion, we found that M. avium induced the secretion of MMP-1 in duodenal tissue and in peripheral blood. M. avium also induced the secretion of a restricted set of cytokines in duodenal tissue, namely IL-1β and IL-6 as well as eliciting a Th1 and Th2 response in the blood. We speculate that the induction of these cytokines by M. avium suggests a possible pathway through which NTMs, and M. avium in particular, could remodel the mucosa and lead to environmental enteropathy. The induction of MMP-1 and cytokines by M. avium in small intestinal tissue from a tropical population suggests that environmental mycobacteria could contribute to the epithelial disruption seen in environmental enteropathy in exposed populations. Further work will be required to demonstrate if MMP-mediated mucosal re-modelling actually operates in vivo. Competing interests No competing interests declared. Funding Financial support for this study was provided by the Research Support Centre at the University of Zambia, School of Medicine (UNZA-SOM), through the Southern African Consortium for Research Excellence (SACORE), which is part of the African Institutions Initiative Grant (Grant code 087537) of the Wellcome Trust (Company Number 2711000), a charity (No. 210183) registered in England. Neither Wellcome Trust nor SACORE had a role in the design, conduct and interpretation of the study.

n Consortium for Research Excellence (SACORE), which is part of the African Institutions Initiative Grant (Grant code 087537) of the Wellcome Trust (Company Number 2711000), a charity (No. 210183) registered in England. Neither Wellcome Trust nor SACORE had a role in the design, conduct and interpretation of the study. Authors’ contributions GC, CM and PK took part in the planning of the study, data collection, analysis and writing of the manuscript. FD took part in the planning of the study, analysis and writing of the manuscript. ES, VK, JBN took part in data collection, analysis and writing of the manuscript. All authors read and approved the final manuscript. Availability of data and materials The datasets analyzed during the current study are available from the corresponding author on reasonable request. Consent statement The study was approved by the University of Zambia Biomedical Research Ethics Committee. Written informed consent was obtained from all patients who took part in the study. Acknowledgements The authors would like to express their gratitude to the following individuals for the invaluable support rendered during the conduct of the study and report writing: Ms Rose Soko, Mr Themba Banda, Mr Patrick Kaonga, Ms Ellen Besa, Ms Kanekwa Zyambo.

Introduction HIV is a significant public health problem worldwide. Globally, it is estimated that 2.1 million children under 15 years of age are infected by HIV, with about 160 000 new infections and 120 000 related deaths annually (World Health Organization, 2017a), the majority of which occur in Sub-Saharan Africa. Most of these children acquire the infection by vertical transmission from their HIV-infected mothers during pregnancy, delivery, or through breastfeeding (World Health Organization, 2017b). HIV-infected children become progressively immunosuppressed, a situation that increases their risk of acquiring common and life-threatening childhood infections (Humphreys et al., 2010). The HIV pandemic has hit Mozambique particularly hard. This is currently one of the most affected countries in the world, with a nationally reported HIV prevalence in adults of 13% (Instituto Nacional de Saúde (INS) and Instituto Nacional de Estatística (INE), 2017), although community cross-sectional studies conducted in recent years in the district of Manhiça have repeatedly estimated the HIV prevalence in this specific district to be as high as 40% (Gonzalez et al., 2015, Gonzalez et al., 2012). In this setting, vertical transmission rates in the first month of life for children born to HIV-infected mothers have been estimated at around 9% (Moraleda et al., 2014).

Manhiça have repeatedly estimated the HIV prevalence in this specific district to be as high as 40% (Gonzalez et al., 2015, Gonzalez et al., 2012). In this setting, vertical transmission rates in the first month of life for children born to HIV-infected mothers have been estimated at around 9% (Moraleda et al., 2014). Mozambique has implemented a series of diagnostic and preventive services to reduce the disproportionate burden imposed by HIV. Pregnant women attending antenatal clinics are routinely screened for HIV according to the national guidelines, and antiretroviral treatment (ART) is offered free of charge regardless of CD4 count, whenever the test is positive. Appropriate prophylaxis against vertical transmission or post-delivery ART treatment of the newborn is instituted for those children born to HIV-infected mothers, while awaiting definitive molecular confirmation of the infection (República de Moçambique et al., 2016). However, the prevention of mother-to-child transmission (PMTCT) remains suboptimal, and many incident infections still occur, leading to a relatively high HIV prevalence among children under the age of 5 years.

Education 1748 (75·7) Primary 222 (12·7) 125 (56·3) 1·1 (0·8–1·5) 0·637 Middle/JHS 637 (36·4) 347 (54·5) Reference Secondary 429 (24·5) 232 (54·1) 1·0 (0·8–1·3) 0·899 Tertiary 190 (10·9) 110 (57·9) 1·1 (0·8–1·6) 0·405 No education 270 (15·4) 141 (52·2) 0·9 (0·7–1·2) 0·534 Occupation 1722 (74·6) Unemployed 390 (22·6) 208 (53·3) 0·9 (0·7–1·1) 0·423 Unskilled 951 (55·2) 530 (55·7) Reference Skilled 381 (22·1) 198 (52·0) 0·9 (0·7–1·1) 0·213 Monthly income (GH¢) 1622 (70·2) None 371 (22·9) 213 (57·4) Reference <301 807 (49·7) 438 (54·3) 0·9 (0·7–1·1) 0·315 301–1000 407 (25·1) 218 (53·6) 0·8 (0·6–1·1) 0·281 >1000 37 (2·3) 15 (40·5) 0·5 (0·3–1·0) 0·052 Religion 1771 (76·7) Christian 1361 (76·9) 739 (54·3) Reference Islam 302 (17·0) 161 (53·3) 1·0 (0·7–1·2) 0·755 Other 26 (1·5) 14 (53·9) 1·0 (0·4–2·1) 0·963 Not religious 82 (4·6) 49 (59·7) 1·2 (0·8–2·0 0·366 Ethnicity 1760 (76·4) Akan 570 (32·3) 309 (54·2) Reference Ewe 259 (14·7) 143 (55·2) 1·0 (0·8–1·4) 0·788 Ga/Adangbe 544 (30·8) 310 (57·0) 1·1 (0·9–1·4) 0·352 Other 392 (22·2) 196 (50·0) 0·8 (0·6–1·1) 0·199 Marital status 1758 (76·1) Single 766 (43·6) 431 (56·3) Reference Married 742 (42·2) 395 (53·2) 0·9 (0·7–1·1) 0·237 Divorced 167 (9·5) 99 (59·3) 1·1 (0·8–1·6) 0·476 Widowed 83 (4·7) 35 (42·2) 0·6 (0·3–0·9) 0·015 Smear positivity 2208 (95·6) Scanty 1–9 173 (7·8) 96 (55·5) 1·1 (0·8–1·5) 0·714 1+ 474 (21·5) 237 (50·0) 0·9 (0·7–1·1) 0·151 2+ 546 (24·7) 294 (53·9) 1·0 (0·8–1·2) 0·957 3+ 1015 (46·0) 548 (54·0) Reference Previous TB treatment 1737 (75·2) Yes 291 (16·8) 153 (52·6) 0·9 (0·7–1·2) 0·535 No 1446 (83·2) 789 (54·6) Reference

rs, while awaiting definitive molecular confirmation of the infection (República de Moçambique et al., 2016). However, the prevention of mother-to-child transmission (PMTCT) remains suboptimal, and many incident infections still occur, leading to a relatively high HIV prevalence among children under the age of 5 years. Similarly, diarrheal diseases still represent a major cause of morbidity and mortality in childhood, in spite of decreasing trends in recent years. Diarrheal diseases are estimated to account for around 578 000 annual child deaths, i.e. approximately 9% of all under-five deaths globally (Liu et al., 2015). The geographical distribution of these diarrhea deaths remains disproportionately shifted to resource-constrained settings, with Sub-Saharan Africa and Southeast Asia contributing to approximately 78% of deaths due to diarrhea worldwide (Humphreys et al., 2010, Boschi-Pinto et al., 2008). In these settings, challenging socio-economic and hygiene conditions, fragile health systems, and limitations in the deployment of already proven or existing preventive strategies such as oral rehydration solution (ORS) or enteric vaccines, all contribute to maintain the significant public health impact imposed by diarrhea (Liu et al., 2015, Bern et al., 1992, Kosek et al., 2003). Furthermore, the HIV/AIDS pandemic has significantly worsened the profile of diarrheal disease in these regions (Thea et al., 1993, van Eijk et al., 2010).

or enteric vaccines, all contribute to maintain the significant public health impact imposed by diarrhea (Liu et al., 2015, Bern et al., 1992, Kosek et al., 2003). Furthermore, the HIV/AIDS pandemic has significantly worsened the profile of diarrheal disease in these regions (Thea et al., 1993, van Eijk et al., 2010). The etiological agents causing diarrhea vary greatly depending on the region of the world, but a recent multicenter study carefully characterized those of major relevance using a case–control approach enrolling not only moderate-to-severe diarrhea (MSD) or less-severe diarrhea (LSD) cases, but also matched healthy community controls (Kotloff et al., 2013, Liu et al., 2016). Diarrhea cases among HIV-infected children have been reported to be more severe, longer-lasting, cause longer hospital admissions, and be related to more co-morbidities and higher case-fatality rates (Groome and Madhi, 2012a, Chhagan and Kauchali, 2006, Keusch et al., 1992, Johnson et al., 2000). Related data regarding the specific etiology of diarrhea in HIV-infected children are still scarce for most of Sub-Saharan Africa. Previous mortality studies among young children with diarrhea in various settings in Africa have identified immunosuppression by HIV infection as an important factor for diarrhea and deaths (van Eijk et al., 2010).

e specific etiology of diarrhea in HIV-infected children are still scarce for most of Sub-Saharan Africa. Previous mortality studies among young children with diarrhea in various settings in Africa have identified immunosuppression by HIV infection as an important factor for diarrhea and deaths (van Eijk et al., 2010). In Manhiça District, in southern Mozambique, where HIV prevalence remains exceedingly high (Gonzalez et al., 2015, Gonzalez et al., 2012), diarrhea is the third leading cause of hospital admission among children 0–14 years and the fourth leading cause of death among children 12–59 months of age (Mandomando et al., 2007, Sacarlal et al., 2009, Nhampossa et al., 2015). The aim of this study was to evaluate the role of HIV infection in modifying the frequency and etiology of MSD, and to determine its impact on mortality among children less than 5 years of age with diarrhea in Manhiça District, rural Mozambique. Materials and methods Study area and population This study was conducted by the Centro de Investigação em Saúde de Manhiça (CISM) in Manhiça District, a rural area in southern Mozambique, under the framework of The Global Enteric Multicenter Study (GEMS) – “Diarrheal Disease in Infants and Young Children in Developing Countries” – coordinated by the Center of Vaccine Development (CVD) at the University of Maryland, Baltimore, Maryland, USA.

CISM) in Manhiça District, a rural area in southern Mozambique, under the framework of The Global Enteric Multicenter Study (GEMS) – “Diarrheal Disease in Infants and Young Children in Developing Countries” – coordinated by the Center of Vaccine Development (CVD) at the University of Maryland, Baltimore, Maryland, USA. Manhiça District has an estimated population of 183 000 inhabitants. During the study period, more than half of the inhabitants were living in an area under continuous demographic surveillance as part of CISM Demographic Surveillance System (DSS). The area has been described in detail elsewhere (Sacoor et al., 2013). The CISM DSS is linked to a morbidity surveillance system at Manhiça District Hospital, the referral health facility for Manhiça District, and is also ongoing at five additional smaller health posts within the district (Guinovart et al., 2008, Bassat et al., 2011).

n described in detail elsewhere (Sacoor et al., 2013). The CISM DSS is linked to a morbidity surveillance system at Manhiça District Hospital, the referral health facility for Manhiça District, and is also ongoing at five additional smaller health posts within the district (Guinovart et al., 2008, Bassat et al., 2011). Manhiça District Hospital is a 110-bed referral hospital and has a 16-bed specific malnutrition ward. It admits an average of 3500 children annually. Standardized forms are routinely completed for all outpatients and inpatients, and include demographic, clinical, and laboratory data. Weight is measured for all outpatients, but height is only measured for admitted patients. Malaria is proactively screened in all patients with a fever or a referred history of fever in the preceding 24 h. Upon admission, a single blood culture is routinely performed for all children under the age of 2 years and for older children with a documented temperature ≥39 °C with severe malnutrition or other signs of severe disease, according to clinical judgment (Sigaúque et al., 2009).

d history of fever in the preceding 24 h. Upon admission, a single blood culture is routinely performed for all children under the age of 2 years and for older children with a documented temperature ≥39 °C with severe malnutrition or other signs of severe disease, according to clinical judgment (Sigaúque et al., 2009). Data collection and clinical management The GEMS case–control study was conducted between December 2007 and November 2012, and has been described extensively elsewhere (Kotloff et al., 2013). In brief, the study targeted three age strata: infants aged 0–11 months, toddlers aged 12–23 months, and children aged 24–59 months. All children in these age strata from the DSS population who sought care at the health centers within the DSS area were screened for diarrhea, defined as ≥3 loose stools within the preceding 24 h. Study clinicians assessed each child with diarrhea for eligibility. To be included, the episode had to be new (onset after ≥7 diarrhea-free days), acute (onset within the previous 7 days), and fulfill at least one of the following criteria for MSD: sunken eyes, loss of skin turgor (abdominal skin pinch with slow (≤2 s) or very slow (>2 s) recoil), intravenous hydration administered or prescribed, dysentery (visible blood in loose stools), or admission to hospital with diarrhea or dysentery. Additionally, a group of LSD cases, not fulfilling any of the aforementioned criteria for MSD, was also recruited (Kotloff et al., 2013, Kotloff et al., 2012). For each MSD case, at least one to three healthy children from the community, with no history of diarrhea in the previous 7 days, were recruited as controls. A single control matched to each LSD case was also recruited. Controls were randomly selected from the neighborhood in which the case resided using the DSS database within 14 days of presentation of the index case. Controls were also matched by age and sex. Written informed consent was sought from the child’s representative before clinical and epidemiological data were obtained and anthropometric measurements were taken. A stool sample was collected from case and control subjects within 12 h of registration for microbiological identification. Additionally, a follow-up home visit was made approximately 60 days after enrollment to assess vital status (Kotloff et al., 2012).

l data were obtained and anthropometric measurements were taken. A stool sample was collected from case and control subjects within 12 h of registration for microbiological identification. Additionally, a follow-up home visit was made approximately 60 days after enrollment to assess vital status (Kotloff et al., 2012). Three consecutive attempts were made to re-contact children who were absent at their 60-day visit, in accordance with the study protocol, but if this was unsuccessful, they were considered lost to follow-up. The laboratory procedures undertaken for the extensive microbiological investigations conducted to characterize each diarrheal episode have been described previously (Kotloff et al., 2013, Liu et al., 2016, Panchalingam et al., 2012).

study protocol, but if this was unsuccessful, they were considered lost to follow-up. The laboratory procedures undertaken for the extensive microbiological investigations conducted to characterize each diarrheal episode have been described previously (Kotloff et al., 2013, Liu et al., 2016, Panchalingam et al., 2012). HIV counseling and testing Between May 2010 and November 2012, the parents or caretakers of case and control subjects were offered HIV counseling and testing for their children. After providing specific additional informed consent, a blood sample was obtained by finger prick for serological testing on site, according to Mozambican national guidelines. The Determine HIV-1/2 Rapid Test (Abbott Laboratories, Abbott Park, IL, USA) was used. A positive Determine HIV-1/2 result was confirmed using the Uni-Gold Rapid Test (Trinity Biotech Co., Wicklow, Ireland). DNA PCR was required to confirm the HIV infection in children under 18 months of age with a positive HIV serology. HIV infection was diagnosed in a child older than 18 months of age with a positive serological test, or in a child less than 18 months old with a positive DNA PCR. Children testing positive were managed according to the national HIV/AIDS treatment guidelines. For children ≥18 months of age, HIV results became available within approximately 60 min after enrollment in the diarrhea study, whereas for children <18 months of age, the results were only available within the month after enrolment, as a result of delays in PCR confirmation. Children testing positive and confirmed to be HIV-infected were referred for HIV clinical care according to national guidelines.

in after enrollment in the diarrhea study, whereas for children <18 months of age, the results were only available within the month after enrolment, as a result of delays in PCR confirmation. Children testing positive and confirmed to be HIV-infected were referred for HIV clinical care according to national guidelines. Data management and analysis Overall data were recorded using specific questionnaires, which were checked on site and scanned and sent to a data collection centre in the USA. Queries were resolved on site, and the original source was also stored there. HIV data were collected using a specific form and double-entered into OpenClinica version 2.5 at CISM; discrepancies in the data were resolved by referring to the original form.

ite and scanned and sent to a data collection centre in the USA. Queries were resolved on site, and the original source was also stored there. HIV data were collected using a specific form and double-entered into OpenClinica version 2.5 at CISM; discrepancies in the data were resolved by referring to the original form. Analyses were conducted using Stata/SE software version 13.1 and the package coxphf from R version 3.1.1 (StataCorp, 2013, Ploner and Heinze, 2015, R Core Team, 2014). Appropriate weights based on sampling weights were constructed for the estimation of disease burden, such as the prevalence of HIV in the population and the mortality rate among cases. In parallel, propensity scores estimated by logistic regression were used to adjust these weights for potential non-response bias in HIV status (Rosenbaum and Rubin, 1983, Little and Vartivarian, 1986, Little and Vartivarian, 2003). A crude model of conditional logistic regression with Firth’s penalized likelihood was estimated to assess the association between HIV and MSD, with HIV status as the independent variable. Firth’s penalized likelihood approach is a method of addressing issues of separability, small sample sizes, and bias of the parameter estimates (Firth, 1993, Heinze and Schemper, 2002).

Firth’s penalized likelihood was estimated to assess the association between HIV and MSD, with HIV status as the independent variable. Firth’s penalized likelihood approach is a method of addressing issues of separability, small sample sizes, and bias of the parameter estimates (Firth, 1993, Heinze and Schemper, 2002). Conditional logistic regression models with Firth’s penalized likelihood were estimated to assess the impact of HIV on the association of each pathogen with MSD or LSD. The modification of the pathogen-specific association with MSD or LSD was assessed by the odds ratio (OR) corresponding to the interaction coefficient, e.g. the ratio and 95% confidence intervals (95% CI) of the OR among HIV-infected children over the OR among the uninfected. Differences between HIV-infected and HIV-uninfected children in clinical presentations and microbiological etiologies at recruitment were evaluated within the MSD and LSD cases, and statistical tests for independent groups were performed (e.g., Chi-square or Fisher’s exact test for categorical variables and t-test for continuous variables). Mortality analyses were performed within the MSD and LSD cases. A study death was defined as any death occurring in a case or a control within the 60 days up to the follow-up home visit. The time to death (or censoring) was calculated as the difference between the date of death (or 60-day visit), minus the enrolment date, plus the number of days with diarrhea on the day of enrolment.

study death was defined as any death occurring in a case or a control within the 60 days up to the follow-up home visit. The time to death (or censoring) was calculated as the difference between the date of death (or 60-day visit), minus the enrolment date, plus the number of days with diarrhea on the day of enrolment. Kaplan–Meier survival curves were estimated, and the difference between the HIV-infected and HIV-uninfected children was assessed by weighted Cox regression estimation. Mortality rates and appropriate centiles of survival time were also described. Cox regression with Firth’s penalized likelihood was used to estimate the associations with time-to-death adjusted for age. Similarly, the modification of the pathogen-specific association with time-to-death (interaction) was assessed by the hazard ratio (HR) corresponding to the interaction coefficient, e.g. the ratio and 95% CI of the HR among HIV-infected children over the HR among the HIV-uninfected.

with time-to-death adjusted for age. Similarly, the modification of the pathogen-specific association with time-to-death (interaction) was assessed by the hazard ratio (HR) corresponding to the interaction coefficient, e.g. the ratio and 95% CI of the HR among HIV-infected children over the HR among the HIV-uninfected. Ethical approval This study was a site-specific sub-analysis conducted under the framework of the Global Enteric Multicenter Study, a large multicenter study conducted in six other developing countries investigating the burden, etiology, and sequelae of diarrheal disease in infants and young children. The overall protocol and informed consent, including for the HIV sub-study, were both approved by the National Bioethics Committee for Health of Mozambique (CNBS −IRB00002657), the Ethics Committee of Hospital Clínic of the University of Barcelona, and the Institutional Review Board at the University of Maryland. Parents/caregivers of participants provided written informed consent, countersigned by an impartial witness in the case of illiterate parents/caretakers.

ambique (CNBS −IRB00002657), the Ethics Committee of Hospital Clínic of the University of Barcelona, and the Institutional Review Board at the University of Maryland. Parents/caregivers of participants provided written informed consent, countersigned by an impartial witness in the case of illiterate parents/caretakers. Results HIV testing and counseling A total of 2014 children were enrolled during the study period, including 293 MSD cases with their matched 859 controls, and 431 LSD cases with their matched 431 controls. Seventy three percent (214/293) of MSD cases, 49% (418/859) of MSD controls, 81% (349/431) of LSD cases, and 50% (214/431) of LSD controls provided consent for HIV testing, including 324 matched case–control pairs for MSD and 164 matched case–control pairs for LSD (Figure 1).Figure 1 Study profile: HIV counseling and testing among moderate-to-severe diarrhea (MSD) and less-severe diarrhea (LSD) cases and their matched controls. Figure 1 Overall, HIV prevalence was higher among MSD cases (23%, 49/214; 95% CI 17.72–29.06) than among MSD matched controls (4%, 17/418; 95% CI 2.54–6.46). HIV weighted prevalence increased to 25% in the MSD group (95% CI 17.98–32.60) and decreased to 3% in the control group (95% CI 1.48–7.97). In the 214 MSD cases tested for HIV, males (accounting for 60% of the participants enrolled, 128/214) showed a non-significantly lower HIV positivity rate (26/128, 20%) than females (23/86, 27%) (p = 0.2723).

eased to 25% in the MSD group (95% CI 17.98–32.60) and decreased to 3% in the control group (95% CI 1.48–7.97). In the 214 MSD cases tested for HIV, males (accounting for 60% of the participants enrolled, 128/214) showed a non-significantly lower HIV positivity rate (26/128, 20%) than females (23/86, 27%) (p = 0.2723). HIV prevalence in the LSD group was similar among cases and controls: 7% (26/349; 95% CI 5.11–10.74) in LSD cases and 7% (14/214; 95% CI 3.90–10.78) in their matched LSD controls. Furthermore, the weighted prevalence rates adjusted for non-response were also very similar: 6% (95% CI 4.28–9.47) for LSD cases and 6% (95% CI 3.00–13.25) for their matched LSD controls. HIV-infected children had a higher risk of presenting with MSD compared to uninfected children (OR 5.7, 95% CI 3.05–11.32; p < 0.0001). This association was not observed in the LSD group, where the HIV infection did not appear to increase the odds of suffering an episode of LSD compared to their matched controls (OR 0.7, 95% CI 0.30–1.60; p = 0.403).

er risk of presenting with MSD compared to uninfected children (OR 5.7, 95% CI 3.05–11.32; p < 0.0001). This association was not observed in the LSD group, where the HIV infection did not appear to increase the odds of suffering an episode of LSD compared to their matched controls (OR 0.7, 95% CI 0.30–1.60; p = 0.403). Pathogen profile After adjusting for HIV, Shigella spp (OR 96.8; p < 0.0001), rotavirus (OR 4.3; p < 0.0001), and Cryptosporidium (OR 2.7; p = 0.0040) were all significant risk factors for MSD episodes, whereas Giardia (OR 0.59; p = 0.060) and Clostridium difficile (positivity for glutamate dehydrogenase antigen (GDH Ag)) (OR 0.24; p = 0.0029) were pathogens protective for MSD episodes. Regarding LSD cases, enterotoxigenic Escherichia coli producing heat-stable toxin (ETEC of any sequence type (ST); OR 3.2; p = 0.0050), enteroaggregative Escherichia coli (EAEC, aatA only; OR 10.1; p = 0.0373), Shigella spp (OR 14.4; p = 0.0080), and rotavirus (OR 2.8; p = 0.0018) were all risk factors for LSD, whereas Giardia (OR 0.36; p = 0.0007), EAEC (aaiC only; OR 0.49; p = 0.046), and C. difficile (GDH Ag-positive) (OR 0.27; p = 0.0023) were the pathogens found to protect against LSD.

C, aatA only; OR 10.1; p = 0.0373), Shigella spp (OR 14.4; p = 0.0080), and rotavirus (OR 2.8; p = 0.0018) were all risk factors for LSD, whereas Giardia (OR 0.36; p = 0.0007), EAEC (aaiC only; OR 0.49; p = 0.046), and C. difficile (GDH Ag-positive) (OR 0.27; p = 0.0023) were the pathogens found to protect against LSD. For all study cases in which a pathogen was detected, differences according to HIV status were also assessed for every given pathogen. For MSD cases, only Cryptosporidium seemed to border statistical significance, being more prevalent among HIV-positive cases than among HIV-negative ones (31% vs. 18%; p = 0.0608), while for LSD cases, only ETEC (any ST) was more frequent among HIV-infected cases (38% vs. 13%; p = 0.0016), although Cryptosporidium was also borderline significantly associated with HIV infection (27% vs. 13%, p = 0.0759). An assessment of whether HIV-positive status modified the relationship between pathogens and being a case of MSD or LSD was additionally performed. For MSD, no pathogen significantly increased the risk of being an MSD case among HIV-infected children.

For all study cases in which a pathogen was detected, differences according to HIV status were also assessed for every given pathogen. For MSD cases, only Cryptosporidium seemed to border statistical significance, being more prevalent among HIV-positive cases than among HIV-negative ones (31% vs. 18%; p = 0.0608), while for LSD cases, only ETEC (any ST) was more frequent among HIV-infected cases (38% vs. 13%; p = 0.0016), although Cryptosporidium was also borderline significantly associated with HIV infection (27% vs. 13%, p = 0.0759). An assessment of whether HIV-positive status modified the relationship between pathogens and being a case of MSD or LSD was additionally performed. For MSD, no pathogen significantly increased the risk of being an MSD case among HIV-infected children. Clinical characteristics The analysis of MSD cases showed that HIV-infected children were more likely to be malnourished (defined according to Z-scores; 59% vs. 35%; p = 0.0020), wasted (33% vs. 19%; p = 0.0447), underweight (49% vs. 21%; p = 0.0001), or stunted (46% vs. 23%; p = 0.0020). Furthermore, they were more likely to be dehydrated with sunken eyes (57% vs. 39%%; p = 0.0227) or wrinkled skin (43% vs. 27%; p = 0.0381) and more likely to be lethargic or show decreased consciousness (59% vs. 35%; p = 0.0026) or fast breathing (37% vs. 21%; p = 0.0208). In addition, HIV-infected children were more likely to be admitted to hospital than those not infected (80% vs. 64%; p = 0.0435) (Table 1). Of those who were admitted, the mean hospital stay was significantly longer for the HIV-infected: 7.0 days (range 1–24 days) for HIV-infected children vs. 4.3 days (range 0–17 days) for those who were not infected (p = 0.0001).Table 1 Clinical characteristics of children less than 5 years of age with moderate-to-severe diarrhea (MSD) at enrollment, according to HIV status.

significantly longer for the HIV-infected: 7.0 days (range 1–24 days) for HIV-infected children vs. 4.3 days (range 0–17 days) for those who were not infected (p = 0.0001).Table 1 Clinical characteristics of children less than 5 years of age with moderate-to-severe diarrhea (MSD) at enrollment, according to HIV status. Table 1Variable HIV-status p-Value HIV-uninfected (n = 165) HIV-infected (n = 49) Weight-for-age Z-score −1.19 (1.35) −2.14 (1.80) 0.0001 Length/height-for-age Z-score −1.21 (1.23) −2.08 (1.44) 0.0001 Underweight 34/165 (21%) 24/49 (49%) 0.0001 Malnutrition (Z-scores) 57/165 (35%) 29/49 (59%) 0.0020 Sunken eyes 64/165 (39%) 28/49 (57%) 0.0227 Loss of skin turgor 45/165 (27%) 21/49 (43%) 0.0381 Intravenous rehydration 68/165 (41%) 25/49 (51%) 0.2239 Dysentery 34/165 (21%) 3/49 (6%) 0.0186 Hospitalized 106/165 (64%) 39/49 (80%) 0.0435 Vomiting 69/165 (42%) 27/49 (55%) 0.1007 Fever 67/165 (41%) 18/49 (37%) 0.6268 Lethargy or loss of consciousness 58/165 (35%) 29/49 (59%) 0.0026 Convulsions 10/165 (6%) 4/49 (8%) 0.5303 Irritable or restless 21/165 (13%) 5/49 (10%) 0.6669 Belly pain 38/165 (23%) 13/49 (27%) 0.6136 Dry mouth 50/165 (30%) 22/49 (45%) 0.0576 Fast breathing 34/165 (21%) 18/49 (37%) 0.0208

165 (41%) 18/49 (37%) 0.6268 Lethargy or loss of consciousness 58/165 (35%) 29/49 (59%) 0.0026 Convulsions 10/165 (6%) 4/49 (8%) 0.5303 Irritable or restless 21/165 (13%) 5/49 (10%) 0.6669 Belly pain 38/165 (23%) 13/49 (27%) 0.6136 Dry mouth 50/165 (30%) 22/49 (45%) 0.0576 Fast breathing 34/165 (21%) 18/49 (37%) 0.0208 On discharge, a diagnosis of dysentery was less frequent among HIV-infected children with MSD than among uninfected ones (4% vs. 20%; p = 0.0082). A diagnosis of malnutrition at discharge was more frequent among HIV-infected children with MSD than among HIV-uninfected cases (16% vs. 6%; p = 0.0365), and a diagnosis of invasive bacterial infection was borderline more frequent in the HIV-infected children (4% vs. 0%; p = 0.0516) (Table 2).Table 2 Diagnosis at the time of discharge for moderate-to-severe diarrhea (MSD) cases admitted to the hospital, according to HIV status. Table 2Variable HIV-status p-Value Negative (n = 165) Positive (n = 49) Diarrhea 123/165 (75%) 43/49 (88%) 0.0516 Dysentery 33/165 (20%) 2/49 (4%) 0.0082 Pneumonia/lower respiratory tract infection 28/165 (17%) 9/49 (18%) 0.8203 Malaria 22/165 (13%) 7/49 (14%) 0.8642 Malnutrition 10/165 (6%) 8/49 (16%) 0.0365 Other diagnosis 42/165 (25%) 19/49 (39%) 0.0697 Multiple diagnoses 77/165 (47%) 30/49 (61%) 0.0735

(75%) 43/49 (88%) 0.0516 Dysentery 33/165 (20%) 2/49 (4%) 0.0082 Pneumonia/lower respiratory tract infection 28/165 (17%) 9/49 (18%) 0.8203 Malaria 22/165 (13%) 7/49 (14%) 0.8642 Malnutrition 10/165 (6%) 8/49 (16%) 0.0365 Other diagnosis 42/165 (25%) 19/49 (39%) 0.0697 Multiple diagnoses 77/165 (47%) 30/49 (61%) 0.0735 Table 3 summarizes the clinical characteristics of LSD cases at enrollment, according to HIV status. HIV-infected children were more likely to be malnourished (62% vs. 28%, p = 0.003) or stunted (58% vs. 23%; p = 0.0001) than uninfected ones. The stool characteristics of the HIV-infected children were different to those of the HIV-uninfected (p = 0.0187): they less frequently presented with simple watery diarrhea (58% vs. 78%) and more commonly showed pathological products in their stool (mucus, 42% vs. 22%).Table 3 Clinical characteristics of children less than 5 years of age with less severe diarrhea (LSD) at enrollment, according to HIV status.

infected (p = 0.0187): they less frequently presented with simple watery diarrhea (58% vs. 78%) and more commonly showed pathological products in their stool (mucus, 42% vs. 22%).Table 3 Clinical characteristics of children less than 5 years of age with less severe diarrhea (LSD) at enrollment, according to HIV status. Table 3Variable HIV p-Value Negative (n = 323) Positive (n = 26) Weight-for-age Z-score −0.73 (1.21) [323] −1.12 (1.33) [25] 0.1282 Length/height-for-age Z-score −1.17 (1.19) [323] −1.63 (1.51) [26] 0.0688 Underweight 44/323 (14%) 7/25 (28%) 0.0719 Stunted 74/323 (23%) 15/26 (58%) 0.0001 Malnutrition (Z-scores) 89/323 (28%) 16/26 (62%) 0.0003 Vomiting 89/323 (28%) 10/26 (38%) 0.2352 Fever 91/323 (28%) 9/26 (35%) 0.4846 Lethargy 80/323 (25%) 6/26 (23%) 0.8474 Lethargy or loss of consciousness 72/323 (22%) 5/26 (19%) 0.7173 Belly pain 61/323 (19%) 3/26 (12%) 0.4397 Dry mouth 3/323 (1%) 1/26 (4%) 0.2673 Fast breathing 24/323 (7%) 1/26 (4%) 1.0000 Stool characteristics at enrollment Simple watery 252/323 (78%) 15/26 (58%) 0.0187 Sticky/mucoid 71/323 (22%) 11/26 (42%) –

0.8474 Lethargy or loss of consciousness 72/323 (22%) 5/26 (19%) 0.7173 Belly pain 61/323 (19%) 3/26 (12%) 0.4397 Dry mouth 3/323 (1%) 1/26 (4%) 0.2673 Fast breathing 24/323 (7%) 1/26 (4%) 1.0000 Stool characteristics at enrollment Simple watery 252/323 (78%) 15/26 (58%) 0.0187 Sticky/mucoid 71/323 (22%) 11/26 (42%) – Mortality analysis Of all MSD cases with a known HIV status, 89% (191/214) successfully completed their 60-day follow-up home visit. Seventeen deaths occurred among the study participants: 25% (10/40) among the HIV-infected children and 5% (7/151) among the uninfected. Three deaths occurred in the LSD group: 4% (1/25) of the HIV-infected children died, compared to 0.6% (2/303) of the HIV-uninfected. Figure 2 shows the Kaplan–Meier survival curve for children with MSD according to HIV status.Figure 2 Survival at the 60-day follow-up home visit according to HIV status among moderate-to-severe diarrhea (MSD) cases. Figure 2 Mortality rates were higher among HIV-infected cases with MSD than among the uninfected: 33.84 (95% CI 15.28–81.04) vs. 5.49 (95% CI 1.92–21.41) deaths/1000 person-weeks at risk (p = 0.0039). The difference in infant mortality rates was not statistically significant (11.50, 95% CI 3.73–50.20 vs. 40.33, 95% CI 15.03–135.82 deaths/1000 person-weeks at risk; p = 0.08), while in the other age groups a larger difference was observed between HIV-infected and HIV-uninfected children, although it could not be evaluated due to the low (or null) number of deaths among the HIV-uninfected children.

, 95% CI 3.73–50.20 vs. 40.33, 95% CI 15.03–135.82 deaths/1000 person-weeks at risk; p = 0.08), while in the other age groups a larger difference was observed between HIV-infected and HIV-uninfected children, although it could not be evaluated due to the low (or null) number of deaths among the HIV-uninfected children. According to the survival time centiles, 5% of the HIV-infected children had already died at 3 days after the onset of the MSD episode. This was not observed in the HIV-uninfected until 38 days after the onset of the episode. Similarly, mortality rates were higher among HIV-infected children with LSD than among the uninfected (5.09 vs. 0.58 deaths/1000 person-weeks at risk, respectively), although the difference did not reach statistical significance (p = 0.0748). Finally, it was aimed to investigate the specific effect of the different pathogens (n = 29) analyzed as part of the study in terms of the likelihood of affecting survival among MSD and LSD cases. None of the pathogens seemed to have an independent effect in terms of modifying the risk of death, neither for MSD nor for LSD cases.

was aimed to investigate the specific effect of the different pathogens (n = 29) analyzed as part of the study in terms of the likelihood of affecting survival among MSD and LSD cases. None of the pathogens seemed to have an independent effect in terms of modifying the risk of death, neither for MSD nor for LSD cases. Discussion This article describes a novel approach that was used to explore the role of HIV in diarrheal disease, based on a very robust case–control study design, which adds an important level of confidence to the results. The findings confirm that HIV infection plays a significant role in the epidemiology of MSD and its associated mortality in this particular rural area of southern Mozambique, therefore explaining the high mortality rates reported previously (Kotloff et al., 2013, Sow et al., 2016). The strong association between HIV infection, MSD, and mortality in children is in accordance with the findings of previous studies (Irena et al., 2011).

lity in this particular rural area of southern Mozambique, therefore explaining the high mortality rates reported previously (Kotloff et al., 2013, Sow et al., 2016). The strong association between HIV infection, MSD, and mortality in children is in accordance with the findings of previous studies (Irena et al., 2011). The overall high prevalence of HIV infection (25%) among children with MSD seeking care at health centers is in line with data from other studies recruiting sick children within the health system (34% of the children in a pneumonia study being HIV-positive (Bassat et al., 2011)), and with the prevalence rate observed among pregnant women during antenatal consultations in Manhiça District (29%) (Gonzalez et al., 2012), in spite of the many advances in the implementation of mother-to-child vertical transmission programs conducted in the whole country, including in Manhiça District. It therefore appears crucial to continue pushing for the improvement and scale-up of such programs (including monitoring adherence, as up to 25% of children born to HIV-positive mothers are infected in the first year of life), along with other strategies designed to decrease the consequences of HIV infection, such as the scale-up of ART among infected children. This would likely reduce not only the frequency of diarrheal episodes but also the complications that are often observed among HIV-infected children. These findings also corroborate the results of previous studies, in which HIV infection was shown to increase the risk of diarrhea, worsening the symptoms and prolonging its duration (Merchant and Lala, 2012).

requency of diarrheal episodes but also the complications that are often observed among HIV-infected children. These findings also corroborate the results of previous studies, in which HIV infection was shown to increase the risk of diarrhea, worsening the symptoms and prolonging its duration (Merchant and Lala, 2012). One of the relevant findings of this study was that the likelihood of developing an MSD episode was strongly associated with being HIV-positive (OR 5.68; p < 0.0001), while the likelihood of developing an LSD episode was not. In this respect, HIV infection may condition the host’s response to pathogens causing diarrhea, commonly triggering more severe episodes. The fact that differences in relation to HIV as the cause of the more ordinary LSD episodes were not identified may also reflect diarrheal infections affecting those hosts with more recently acquired HIV infections, and therefore less immunosuppression, or otherwise occurring among slow progressors for AIDS. In any case, the World Health Organization now recommends systematic HIV counseling and testing as part of the Integrated Management of Childhood Illness (IMCI) (World Health Organization, 2013), for all children seeking care in health centers, irrespective of the motive of consultation, as a better approach to readily recognizing HIV-infected children and thus triggering appropriate management, and thereby improving the prognosis and reducing the incidence of opportunistic diseases, complications, and deaths.

for all children seeking care in health centers, irrespective of the motive of consultation, as a better approach to readily recognizing HIV-infected children and thus triggering appropriate management, and thereby improving the prognosis and reducing the incidence of opportunistic diseases, complications, and deaths. In this study, Cryptosporidium was among the non-bacterial pathogens most frequently detected, and this organism was associated with an increased risk of MSD among HIV-infected children. This finding is in concordance with the results reported in previous studies (Adamu et al., 2013, Mathur et al., 2013, Paboriboune et al., 2014, Ramakrishnan et al., 2007, Tumwine et al., 2005). The study was conducted in an area where malnutrition is highly endemic (Instituto Nacional de estatistica (INE), 2008, Nhampossa et al., 2013), and such a pre-existing nutritional impairment, even in the absence of HIV infection, may have negatively affected the immune systems of the children, possibly leading to an increase in the incidence of Cryptosporidium infections, thereby explaining the diagnosis of these parasites even among HIV-negative children. However, Cryptosporidium and other parasites have more traditionally been described among chronic and persistent diarrhea cases in HIV-infected patients (Mathur et al., 2013, Paboriboune et al., 2014). The recruitment of subjects into this study was restricted to acute cases – no stool samples were collected from persistent or chronic diarrhea cases.

er parasites have more traditionally been described among chronic and persistent diarrhea cases in HIV-infected patients (Mathur et al., 2013, Paboriboune et al., 2014). The recruitment of subjects into this study was restricted to acute cases – no stool samples were collected from persistent or chronic diarrhea cases. This study also found Shigella and rotavirus to be pathogens with a strong and significant association with MSD according to HIV status. Other studies conducted in Africa have reported similar evidence, with the incidence of rotavirus being higher among HIV-infected children than among those not infected (Groome and Madhi, 2012b). Another study also conducted in the same African region, documented a relatively higher prevalence of Shigella among HIV-infected individuals when compared to uninfected ones (Rowe et al., 2010). In the present study, the number of Shigella isolated was very small, probably explaining the scarcity of negative outcome events. Conventional logistic regression is not a good option for this type of data, and for this reason it was analyzed using penalized estimation. Despite this, the ORs remained very large (OR 96.8) and fairly implausible. However, it is believed that they are interesting results that suggest a very strong association, as also reported in other studies. The associations of pathogens and LSD, in the context of HIV status, were also similar to those of MSD, although Cryptosporidium was not found to be a significant risk factor, while ETEC (any ST) and EAEC (aatA only) were positively associated.

This study also found Shigella and rotavirus to be pathogens with a strong and significant association with MSD according to HIV status. Other studies conducted in Africa have reported similar evidence, with the incidence of rotavirus being higher among HIV-infected children than among those not infected (Groome and Madhi, 2012b). Another study also conducted in the same African region, documented a relatively higher prevalence of Shigella among HIV-infected individuals when compared to uninfected ones (Rowe et al., 2010). In the present study, the number of Shigella isolated was very small, probably explaining the scarcity of negative outcome events. Conventional logistic regression is not a good option for this type of data, and for this reason it was analyzed using penalized estimation. Despite this, the ORs remained very large (OR 96.8) and fairly implausible. However, it is believed that they are interesting results that suggest a very strong association, as also reported in other studies. The associations of pathogens and LSD, in the context of HIV status, were also similar to those of MSD, although Cryptosporidium was not found to be a significant risk factor, while ETEC (any ST) and EAEC (aatA only) were positively associated. Certain pathogens in this case–control study were found to be negatively associated with the risk of MSD or LSD, even after adjusting for HIV. Indeed, Giardia lamblia and C. difficile infection (determined by positivity for GDH antigen) were significantly associated with protection against MSD and LSD, and EAEC (aaiC only) was found to be protective against LSD episodes only. For Giardia, these results are in concordance with the general findings of the GEMS (Kotloff et al., 2012) and other similarly designed studies (Tellevik et al., 2015), whereas Giardia was identified significantly more frequently in controls than in patients with MSD. Although some authors have suggested that this pathogen may be a significant risk factor for diarrhea among HIV-infected individuals (Moolasart, 1999), this remains to be substantiated. Pavlinac et al. described Giardia infections as occurring more frequently among HIV-uninfected individuals than in infected individuals (Pavlinac et al., 2014).

suggested that this pathogen may be a significant risk factor for diarrhea among HIV-infected individuals (Moolasart, 1999), this remains to be substantiated. Pavlinac et al. described Giardia infections as occurring more frequently among HIV-uninfected individuals than in infected individuals (Pavlinac et al., 2014). Regarding C. difficile infection, the results are more difficult to interpret. The prolonged use of antibiotics has been described as a clear risk factor for this infection (Bignardi, 1998). C. difficile is a bacterium that is naturally present in the intestinal flora of about 3% of adults and 66% of children. Such bacteria do not generally cause problems among healthy people; however, some antibiotics used to treat other health conditions may significantly alter the balance of ‘good bacteria’ in the intestinal flora. When this happens, C. difficile can multiply and cause symptoms such as diarrhea. In contrast to what would be expected in Manhiça District, an area of high HIV endemicity where antibiotics such as co-trimoxazole are used in a massive and prolonged way among HIV-infected individuals for the prevention of opportunistic infections, C. difficile infections were negatively associated with the incidence of MSD episodes when adjusted by HIV status. A possible explanation for this is that in such a population, the use of antibiotics favors the proliferation of C. difficile infections among healthy individuals (controls) rather than the proliferation of C. difficile-associated disease (cases). Other studies have reported a high prevalence of C. difficile infections and associated disease among HIV-infected individuals (Haines et al., 2013), or an increased incidence of C. difficile in the general population (Rupnik et al., 2009). Further research will be needed to clarify the role of C. difficile infections as a risk or protective factor for MSD among HIV-infected individuals.

ns and associated disease among HIV-infected individuals (Haines et al., 2013), or an increased incidence of C. difficile in the general population (Rupnik et al., 2009). Further research will be needed to clarify the role of C. difficile infections as a risk or protective factor for MSD among HIV-infected individuals. The study also revealed an interesting finding in relation to the decreased association of dysentery (both on admission but also upon discharge from hospital) with HIV infection, in the context of MSD. This may also be related to the high intake of prophylactic antibiotics among HIV-infected individuals, which may have contributed to the reduction in dysentery episodes.

lation to the decreased association of dysentery (both on admission but also upon discharge from hospital) with HIV infection, in the context of MSD. This may also be related to the high intake of prophylactic antibiotics among HIV-infected individuals, which may have contributed to the reduction in dysentery episodes. Study limitations Several limitations need to be taken into consideration when interpreting the results of this study. While interpreting the effect of HIV on diarrheal diseases, it would have been useful to have more detailed information on the immunosuppression status of all confirmed HIV cases. Unfortunately, the study did not collect data on (1) the use and duration of ART, (2) the use of co-trimoxazole as prophylaxis for opportunistic infections, or (3) CD4 counts or HIV viral load, which could have allowed a more detailed evaluation of the degree of immunosuppression and how this or any given treatment could have influenced the incidence and prevalence of enteric pathogens in HIV-infected children. Moreover, the fact that HIV testing was not mandatory and required an additional patient consent, led to a considerable proportion of participants not having HIV results − MSD and LSD cases and their respective controls − hindering the interpretation of the study results.

of enteric pathogens in HIV-infected children. Moreover, the fact that HIV testing was not mandatory and required an additional patient consent, led to a considerable proportion of participants not having HIV results − MSD and LSD cases and their respective controls − hindering the interpretation of the study results. Conclusions The prevalence of HIV remains high among children with MSD in Manhiça District despite the efforts made in the implementation of programs for the prevention of vertical transmission and the roll-out of ART for those children diagnosed to be HIV-infected. HIV infection was found to increase the risk of having MSD caused by rotavirus, Shigella, and Cryptosporidium. These findings emphasize the need for routine screening for HIV infection among children presenting with diarrhea, as the prognosis is worse and the mortality risk is higher in HIV-infected children with this symptomatology. These findings may additionally support the need to strengthen all efforts to prevent vertical transmission in the antenatal clinics, a critical step to reduce HIV transmission and thus all the undesirable consequences.

ognosis is worse and the mortality risk is higher in HIV-infected children with this symptomatology. These findings may additionally support the need to strengthen all efforts to prevent vertical transmission in the antenatal clinics, a critical step to reduce HIV transmission and thus all the undesirable consequences. Acknowledgements We would like to thank the participants in this study and their parents for allowing the collection of samples and data. The authors would also like to thank Sónia Machevo, Abel Nhama, Kizito Gondo, and all medical officers and management staff at the different peripheral health posts and Manhiça District Hospital. The authors also thank the CISM laboratory technicians for their support in sample processing, the field workers and other staff from CISM and Manhiça District Hospital for collaborating on the data collection and processing, and João Campos Mucasse and Juvêncio Joaquim for their contribution to the data management. Funding source The study on which this work was based (Project GEMS) was supported by the Bill & Melinda Gates Foundation (Project OPP 38874). Core funding for CISM is provided by the Spanish Agency for International Cooperation and Development (AECID). ISGlobal is a member of the CERCA Programme, Generalitat de Catalunya. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Introduction Tuberculosis (TB) is a global health emergency; in 2016 an estimated 10.4 million people got sick, while 1.7 million died of TB (WHO, 2017). In 1993, the World Health Organization (WHO) declared TB a global health emergency and called for more efforts and resources to fight TB. Due largely to the inefficacy of the bacillus Calmette–Guérin (BCG) vaccine against pulmonary TB in adults, the current TB control strategy relies on case detection and treatment under the directly observed therapy short course (DOTs) strategy. The conventional indicators used to assess national control programs under this strategy focus on the proportion of cases that are cured at the end of treatment or whose sputum microscopy becomes negative after the first 2 months of treatment. Such indicators ignore equally important aspects of TB control, which include the duration of infectivity, the frequency of reactivation, and the risk of progression among the infected contacts, as well as the proportion of TB due to recent transmission.

icroscopy becomes negative after the first 2 months of treatment. Such indicators ignore equally important aspects of TB control, which include the duration of infectivity, the frequency of reactivation, and the risk of progression among the infected contacts, as well as the proportion of TB due to recent transmission. Understanding transmission dynamics will contribute to knowledge on factors that enhance the spread of the disease, which is useful for developing preventive interventions. Molecular epidemiological studies have been very useful in a number of countries, identifying populations at risk and areas of high transmission, as well as providing much understanding on the prevalence of different Mycobacterium tuberculosis complex (MTBC) strains with varied virulence and drug resistance rates (Anderson et al., 2014, Malm et al., 2017, Seto et al., 2017, Varghese et al., 2013, Walker et al., 2014, Yang et al., 2016). These studies have shown that the dynamics of TB transmission vary greatly geographically. Even though Africa harbors a large proportion of the global TB cases, with a current incidence of 254 per 100 000 population (WHO, 2017), population-based molecular epidemiological studies needed to understand transmission patterns are rare. The few studies conducted have not been population-based and have lacked an in-depth analysis of the transmission dynamics of MTBC strains belonging to different lineages (Asante-Poku et al., 2016, Glynn et al., 2010, Mulenga et al., 2010).

ar epidemiological studies needed to understand transmission patterns are rare. The few studies conducted have not been population-based and have lacked an in-depth analysis of the transmission dynamics of MTBC strains belonging to different lineages (Asante-Poku et al., 2016, Glynn et al., 2010, Mulenga et al., 2010). The molecular typing tools – spacer oligonucleotide typing (spoligotyping) and mycobacterial interspersed repetitive unit variable number tandem repeat (MIRU-VNTR) typing – have been used successfully for strain differentiation in TB transmission studies due to their combined high discriminatory power and reproducibility; furthermore, in combination with epidemiological data, they have been used for the detection of recent TB transmission and outbreaks (Anderson et al., 2014, Barnes and Cave, 2003, Maguire et al., 2002, Surie et al., 2017, Varghese et al., 2013). Currently, the high cost and expertise needed for whole genome sequencing and analysis have precluded its use in population-based studies, and considering capacity building in a low-resource setting like Ghana, spoligotyping and MIRU-VNTR typing remain good alternatives.

002, Surie et al., 2017, Varghese et al., 2013). Currently, the high cost and expertise needed for whole genome sequencing and analysis have precluded its use in population-based studies, and considering capacity building in a low-resource setting like Ghana, spoligotyping and MIRU-VNTR typing remain good alternatives. TB in humans is caused mainly by Mycobacterium tuberculosis sensu stricto (MTBss) and Mycobacterium africanum (MAF), which are further divided into seven lineages: MTBss lineages 1–4 and 7 (L1–L4 and L7); MAF lineages 5 and 6 (L5 and L6) (Blouin et al., 2012, de Jong et al., 2010). While MTBss is distributed globally, MAF is restricted to West Africa, where it is responsible for up to 50% of TB cases (Gagneux and Small, 2007). Nevertheless, reports mainly from the Gambia where L6 is prevalent, suggest MAF is attenuated compared to MTBss, hence could be outcompeted by MTBss (de Jong et al., 2010, de Jong et al., 2008, Kallenius et al., 1999). However, an 8-year study recently conducted in Ghana found the prevalence of MAF to be fairly constant at approximately 20%, indicating that MAF and MTBss may be transmitted equally (Yeboah-Manu et al., 2016). The objective of this study was to determine the transmission dynamics of TB caused by MTBss and MAF in Ghana.

However, an 8-year study recently conducted in Ghana found the prevalence of MAF to be fairly constant at approximately 20%, indicating that MAF and MTBss may be transmitted equally (Yeboah-Manu et al., 2016). The objective of this study was to determine the transmission dynamics of TB caused by MTBss and MAF in Ghana. Methods Study design and population This study was a population-based prospective study in which sputum samples were collected from consecutive clinically diagnosed pulmonary TB patients reporting to 12 selected health facilities within an urban setting (Accra Metropolitan Assembly (AMA)) and the rural setting of East Mamprusi District (MamE) (Supplementary material, Figure S1). The study was conducted from July 2012 to December 2015. A pulmonary TB case was defined as an individual with a case of TB that was confirmed both clinically and bacteriologically. Detailed demographic and epidemiological data were obtained from consented participants. Mycobacterial isolation, species identification, and drug susceptibility testing The sputum samples were decontaminated and cultured on Lowenstein–Jensen medium to obtain mycobacterial isolates. These isolates were confirmed as MTBC by detecting the MTBC-specific insertion sequence IS6110 using PCR (Yeboah-Manu et al., 2001). In vitro drug susceptibility to isoniazid and rifampicin were determined using either the microplate Alamar Blue cell viability assay, as described elsewhere (Otchere et al., 2016), and/or the GenoType MTBDRplus assay (Hain Lifescience), following the manufacturer’s protocol (Barnard et al., 2008).

-Manu et al., 2001). In vitro drug susceptibility to isoniazid and rifampicin were determined using either the microplate Alamar Blue cell viability assay, as described elsewhere (Otchere et al., 2016), and/or the GenoType MTBDRplus assay (Hain Lifescience), following the manufacturer’s protocol (Barnard et al., 2008). Lineage and strain classification Lineage and strain classification of the MTBC was achieved in a stepwise manner using large sequence polymorphism typing identifying regions of difference 4, 9, 12, 702, and 711 (de Jong et al., 2010, Gagneux and Small, 2007), single nucleotide polymorphism typing, spoligotyping (Kamerbeek et al., 1997), and MIRU-VNTR typing (Supply et al., 2006). For MIRU-VNTR typing, a customized set of 8 MIRU loci was first used, as described by Asante-Poku et al. (2014), and clustered cases were resolved by analyzing the remaining 7 loci of the standard MIRU-15 loci set (Supply et al., 2006). All assays were well controlled with PCR amplifications and pre-PCR procedures conducted in physically separated compartments to avoid laboratory cross-contamination. The presence of more than one allelic repeat number (multiple allele) for any given locus is suggestive of laboratory cross-contamination, multiple strain infection, or microevolution of a single strain. To prevent bias resulting from cross-contamination and multiple strain infection, isolates with multiple alleles at more than one MIRU locus (described as ‘untypeable’) were excluded from further analysis. Isolates with only one multiple allele at any given locus were, however, included due to the possibility of microevolution.

t bias resulting from cross-contamination and multiple strain infection, isolates with multiple alleles at more than one MIRU locus (described as ‘untypeable’) were excluded from further analysis. Isolates with only one multiple allele at any given locus were, however, included due to the possibility of microevolution. The spoligotyping patterns and assigned shared type numbers obtained were defined according to the SITVITWEB database (http://www.pasteur-guadeloupe.fr: 8081/SITVIT_ONLINE/), while sub-lineages were assigned based on the MIRU-VNTRplus database (http://www.miru-vntrplus.org) (Allix-Beguec et al., 2008). Strains with no lineage nomenclature data were further identified using the TB lineage database (Shabbeer et al., 2012) or otherwise regarded as orphan strains. A strain was defined as an MTBC isolate with a unique molecular signature, and thus a unique spoligotype pattern and/or a unique MIRU-VNTR allelic pattern for the number of investigated MIRU loci.

data were further identified using the TB lineage database (Shabbeer et al., 2012) or otherwise regarded as orphan strains. A strain was defined as an MTBC isolate with a unique molecular signature, and thus a unique spoligotype pattern and/or a unique MIRU-VNTR allelic pattern for the number of investigated MIRU loci. Clustering analysis and risk factor assessment Clustering analysis was performed using the categorical parameter and the unweighted pair group method with arithmetic mean (UPGMA) coefficient from a constructed phylogenetic tree using the online MIRU-VNTR tool. Clustering analysis was based on the assumption that strains with the same DNA fingerprint may be epidemiologically linked and associated with recent TB transmission (Hall, 1996). A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci (VNTR 2163, 3690, and 4156). The size of a cluster was also defined using the total number of isolates in the cluster classified into categories of small (2 isolates), medium (3–5 isolates), large (6–20 isolates), and very large (>20 isolates). The recent transmission rate was estimated using the n − 1 formula (Glynn et al., 1999): (nc−c)n, where nc is the total number of clustered cases, c is the number of clusters, and n is the total number of cases in the sample.

Clustering analysis and risk factor assessment Clustering analysis was performed using the categorical parameter and the unweighted pair group method with arithmetic mean (UPGMA) coefficient from a constructed phylogenetic tree using the online MIRU-VNTR tool. Clustering analysis was based on the assumption that strains with the same DNA fingerprint may be epidemiologically linked and associated with recent TB transmission (Hall, 1996). A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci (VNTR 2163, 3690, and 4156). The size of a cluster was also defined using the total number of isolates in the cluster classified into categories of small (2 isolates), medium (3–5 isolates), large (6–20 isolates), and very large (>20 isolates). The recent transmission rate was estimated using the n − 1 formula (Glynn et al., 1999): (nc−c)n, where nc is the total number of clustered cases, c is the number of clusters, and n is the total number of cases in the sample. Only one strain per participant was included in the analysis, and follow-up cases were excluded. The clustering analysis was stratified first by location and then by MTBC lineage. The spatial distribution and clustering among all of the observed Spoligo/MIRU strain types were studied by constructing a minimum spanning tree (MST) with Bionumerics software (Applied Maths, Sint-Marteen-Latem, Belgium).

were excluded. The clustering analysis was stratified first by location and then by MTBC lineage. The spatial distribution and clustering among all of the observed Spoligo/MIRU strain types were studied by constructing a minimum spanning tree (MST) with Bionumerics software (Applied Maths, Sint-Marteen-Latem, Belgium). Data management and analysis Both molecular and epidemiological data were analyzed. Epidemiological data retrieved from all participants with positive MTBC cultures were included in the analysis while excluding data from those with no growth, contaminated cultures, and isolated non-tuberculous mycobacterial species. All statistical analyses were conducted using the Stata statistical package version 14.2 (Stata Corp., College Station, TX, USA). The association of specific lineages and/or sub-lineages of the MTBC with time and/or geographical locations were explored using the Chi-square test and a logistic regression model. For the determination of independent predictive factors for recent TB transmission, a multivariate analysis (forward stepwise approach with a probability entry of 0.1) was conducted using a logistic regression model while estimating the odds ratios (OR). p-Values of <0.05 were considered significant. The study is reported according to the Strengthening the Reporting of Molecular Epidemiology for Infectious Diseases (STROME-ID) guidelines (Field et al., 2016).

Data management and analysis Both molecular and epidemiological data were analyzed. Epidemiological data retrieved from all participants with positive MTBC cultures were included in the analysis while excluding data from those with no growth, contaminated cultures, and isolated non-tuberculous mycobacterial species. All statistical analyses were conducted using the Stata statistical package version 14.2 (Stata Corp., College Station, TX, USA). The association of specific lineages and/or sub-lineages of the MTBC with time and/or geographical locations were explored using the Chi-square test and a logistic regression model. For the determination of independent predictive factors for recent TB transmission, a multivariate analysis (forward stepwise approach with a probability entry of 0.1) was conducted using a logistic regression model while estimating the odds ratios (OR). p-Values of <0.05 were considered significant. The study is reported according to the Strengthening the Reporting of Molecular Epidemiology for Infectious Diseases (STROME-ID) guidelines (Field et al., 2016). Results Characteristics of study participants A total 3303 sputum smear-positive pulmonary TB cases were recruited, 382 (11.6%) from the rural setting and 2921 (88.4%) from the urban setting; 2604 (78.8%) MTBC isolates were obtained from these cases (Supplementary material, Table S1). After excluding 13 Mycobacterium bovis and isolates that were untypeable (described in the Methods section), 2309 of 2604 isolates (88.7%) were included for clustering analysis. The participants comprised 1631 (71%) males and 663 (29%) females (there was no record of sex for 15 participants) with a median age of 39 years (range 3–91 years) and 33 years (range 4–90 years), respectively (Figure 1; Supplementary material, Table S1). The male-to-female ratio observed was comparable to the national average of approximately 2:1.Figure 1 Pipeline for recruited participants and culture-positive TB cases included in the clustering analysis.

ars (range 3–91 years) and 33 years (range 4–90 years), respectively (Figure 1; Supplementary material, Table S1). The male-to-female ratio observed was comparable to the national average of approximately 2:1.Figure 1 Pipeline for recruited participants and culture-positive TB cases included in the clustering analysis. *Category described as untypeable for MIRU-VNTR includes isolates with ≥2 MIRU loci unamplified (n = 164, 71.3%) and isolates with a double allele at ≥2 MIRU loci (n = 66, 28.7%). These isolates were described as suspected mixed infection or laboratory contamination and hence were excluded from further analysis. #Frequency was expressed as the total number of Mycobacterium tuberculosis complex (MTBC) isolates obtained. Figure 1 Of the 2309 participants with MTBC genotyping results, 201 (8.7%) were from the rural setting and 2108 (91.3%) from the urban setting. Among this study cohort, 7.4% (184/2482) of participants were previously treated cases including relapse, which is similar to the national value of 7.0% (WHO, 2015). Seventy-one percent (1561/2208) presented with a sputum smear microscopy bacterial burden result of at least 2+ and 33% (544/1665) admitted having contact with at least one TB patient. In a multivariate logistic regression analysis, it was found that male patients were less likely to be infected with a L5 strain (adjusted OR 0.7, 95% confidence interval (CI) 0.5–0.9) and individuals living in villages were more likely to be infected with a L6 strain (OR 6.6, 95% CI 1.2–36.1) (Supplementary material, Table S2).

variate logistic regression analysis, it was found that male patients were less likely to be infected with a L5 strain (adjusted OR 0.7, 95% confidence interval (CI) 0.5–0.9) and individuals living in villages were more likely to be infected with a L6 strain (OR 6.6, 95% CI 1.2–36.1) (Supplementary material, Table S2). Population structure and recent transmission rate estimation Among the 2309 MTBC isolates analyzed for clustering, 1870 (81.0%) were MTBss and 439 (19.0%) were MAF. Six of the seven human-adapted MTBC lineages were found, with L4, L5, and L6 being most frequent: 1741 (75.4%), 289 (12.5%), and 150 (6.5%) isolates, respectively (Table 1). The relative proportions of the most frequent MTBC lineages remained constant over the entire 3.5-year study period (ptrend: L4 p = 0.72, L5 p = 0.84, L6 p = 0.25; Figure 2).Figure 2 Temporal distribution of 2309 Mycobacterium tuberculosis complex (MTBC) isolates stratified by lineage. Lineages are color-coded with the universally accepted color codes for the main MTBC lineages. Figure 2Table 1 Geographical distribution and population structure of MTBC in Ghana by spoligotyping. Table 1 Rural, n (%) Urban, n (%) Combined, n (%)a MTBC isolates 204 (8.8) 2118 (91.2) 2322 Species distribution M. tuberculosis 172 (9.2) 1698 (90.8) 1870 (80.5) M. africanum 29 (6.6) 410 (93.4) 439 (18.9) Animal 3 (23.1) 10 (76.9) 13 (0.6)

Figure 2Table 1 Geographical distribution and population structure of MTBC in Ghana by spoligotyping. Table 1 Rural, n (%) Urban, n (%) Combined, n (%)a MTBC isolates 204 (8.8) 2118 (91.2) 2322 Species distribution M. tuberculosis 172 (9.2) 1698 (90.8) 1870 (80.5) M. africanum 29 (6.6) 410 (93.4) 439 (18.9) Animal 3 (23.1) 10 (76.9) 13 (0.6) Human adapted MTBC lineage distribution Lineage_1 4 (10.5) 34 (89.5) 38 (1.6) Lineage_2 14 (21.5) 51 (78.5) 65 (2.8) Lineage_3 1 (3.8) 25 (96.2) 26 (1.1) Lineage_4 153 (8.8) 1588 (91.2) 1741 (75.4) Lineage_5 15 (5.2) 274 (94.8) 289 (12.5) Lineage_6 14 (9.3) 136 (90.7) 150 (6.5) Lineage_4 sub-lineage distribution Cameroon 77 (7.4) 969 (92.6) 1046 (60.1) Ghana 50 (13.3) 326 (86.7) 376 (21.6) Haarlem 12 (7.7) 144 (92.3) 156 (9.0) LAM 7 (14.0) 43 (86.0) 50 (2.9) Uganda 1 (2.5) 39 (97.5) 40 (2.3) Other (S, U, X, NEW-1) 5 (9.8) 46 (90.2) 51 (2.9) Not determined 1 (4.5) 21 (95.5) 22 (1.3) MTBC, Mycobacterium tuberculosis complex. a Proportions stated here are column-wise distributions with respect to the categories of species, lineages or sub-lineages.

Lineage_4 sub-lineage distribution Cameroon 77 (7.4) 969 (92.6) 1046 (60.1) Ghana 50 (13.3) 326 (86.7) 376 (21.6) Haarlem 12 (7.7) 144 (92.3) 156 (9.0) LAM 7 (14.0) 43 (86.0) 50 (2.9) Uganda 1 (2.5) 39 (97.5) 40 (2.3) Other (S, U, X, NEW-1) 5 (9.8) 46 (90.2) 51 (2.9) Not determined 1 (4.5) 21 (95.5) 22 (1.3) MTBC, Mycobacterium tuberculosis complex. a Proportions stated here are column-wise distributions with respect to the categories of species, lineages or sub-lineages. Of the 2309 isolates included for clustering analysis, 1227 (53.1%) isolates clustered in 276 different clusters with a mean cluster size of 4 (range 2–35) and 1082 (46.9%) unique isolates were identified, giving a total of at least 1358 different MTBC strains circulating within the study population (Table 2a). Using the n − 1 method, the overall clustering rate (reflecting the recent transmission rate) was estimated to be 41.2%. Lineages 2, 4, and 5 contributed high clustering rates of 53.8%, 44.9%, and 31.8%, respectively (Table 2a). The Cameroon, Ghana, and Haarlem sub-lineages of L4 were the most abundant sub-lineages and, compared to the LAM sub-lineage, contributed significantly to the observed high L4 clustering rate (p < 0.05) (Figure 3). There was no significant difference in the clustering rate between the Cameroon and Ghana sub-lineages (p = 0.57) (Figure 3). While no significant difference in the recent transmission rates was seen between members of MAF (L5 and L6, p = 0.118), it was found that L4 was transmitted significantly more (p < 0.001), with seven of its clusters having very large cluster sizes (>20 isolates per cluster) made up of the Ghana sub-lineage (four very large clusters) and Cameroon sub-lineage (three very large clusters) (Figure 3; Supplementary material, Figure S2). Notwithstanding the lower transmissibility of L5 and L6 compared to L4, four large clusters were also observed for each of these lineages. The urban and rural settings had estimated recent transmission rates of 41.7% and 9.0%, respectively.Figure 3 Cluster distribution and size stratified by lineage (panel A and C) and sub-lineage (panel B and D). *p < 0.001, #p = 0.118, ¤p = 0.565.

to L4, four large clusters were also observed for each of these lineages. The urban and rural settings had estimated recent transmission rates of 41.7% and 9.0%, respectively.Figure 3 Cluster distribution and size stratified by lineage (panel A and C) and sub-lineage (panel B and D). *p < 0.001, #p = 0.118, ¤p = 0.565. Figure 3Table 2a Clustering analysis stratified by lineages and major sub-lineage populations of MTBC. Table 2aLineage Isolates (n) Clustered cases (c) Clustered strains (nc) Single cases (s) Total strain types (s + c) Clustering ratea (%) Lineage 1 38 3 7 31 34 10.5 Lineage 2 65 8 43 22 30 53.8 Lineage 3 26 2 4 22 24 7.7 Lineage 4 1741 201 982 759 960 44.9 Cameroonb 1046 123 614 432 555 46.9 Ghanab 376 36 206 170 206 45.2 Haarlemb 156 23 91 65 88 43.6 LAMb 50 6 25 25 31 38.0 Ugandab 40 5 16 24 29 27.5 Lineage 5 289 51 143 146 197 31.8 Lineage 6 150 11 48 102 113 24.7 Summaryc 2309 276 1227 1082 1358 41.2 MTBC, Mycobacterium tuberculosis complex. a The clustering rate was used to estimate the recent transmission rate. b Major lineage 4 sub-population. c The summary was calculated using only the items in cells corresponding to the six main lineages.

Table 2aLineage Isolates (n) Clustered cases (c) Clustered strains (nc) Single cases (s) Total strain types (s + c) Clustering ratea (%) Lineage 1 38 3 7 31 34 10.5 Lineage 2 65 8 43 22 30 53.8 Lineage 3 26 2 4 22 24 7.7 Lineage 4 1741 201 982 759 960 44.9 Cameroonb 1046 123 614 432 555 46.9 Ghanab 376 36 206 170 206 45.2 Haarlemb 156 23 91 65 88 43.6 LAMb 50 6 25 25 31 38.0 Ugandab 40 5 16 24 29 27.5 Lineage 5 289 51 143 146 197 31.8 Lineage 6 150 11 48 102 113 24.7 Summaryc 2309 276 1227 1082 1358 41.2 MTBC, Mycobacterium tuberculosis complex. a The clustering rate was used to estimate the recent transmission rate. b Major lineage 4 sub-population. c The summary was calculated using only the items in cells corresponding to the six main lineages. Exploring the diversity and clustering within the MTBC lineages Very large molecular clusters (clusters with >20 isolates; defined in the Methods section) were observed for L4, in addition to one strikingly large cluster belonging to the Beijing family of lineage 2 (Figure 4; Supplementary material, Figure S3). Generally, only a few multidrug-resistant MTBC strains were observed across all the major lineages (Supplementary material, Figures S4–S6). There was no single large cluster with all isolates being multidrug-resistant (Supplementary material, Figure S4). The spatial distributions of the isolates constituting each cluster stratified by study setting are shown in the Supplementary material, Figures S7–S9.Figure 4 Minimum spanning tree (MST) representation of the clustering of 2322 Mycobacterium tuberculosis complex (MTBC) isolates from Accra Metropolitan Assembly and East Mamprusi District built with Bionumerics software. The color code reflects the main MTBC lineages 1 to 6 with the size depicting the number of clustered isolates with an identical strain type.

entation of the clustering of 2322 Mycobacterium tuberculosis complex (MTBC) isolates from Accra Metropolitan Assembly and East Mamprusi District built with Bionumerics software. The color code reflects the main MTBC lineages 1 to 6 with the size depicting the number of clustered isolates with an identical strain type. Figure 4 Molecular epidemiology and factors associated with clustering: logistic regression modeling Risk factors associated with recent TB transmission were sought. A total of 675 individuals belonging to either large (6–20 isolates) or very large (>20 isolates) molecular clusters were identified, with a combined median cluster size of 14 (range 6–35). The majority of the individuals belonging to very large clusters were male, with a male-to-female ratio of approximately 3:1, significantly higher than the 2:1 ratio observed in the general TB patient population (p = 0.022). Three large clusters – cluster ID MSC4193, MSC5003.X, and MSC4107, with cluster sizes of 9, 7, and 7 respectively – involved only male subjects (Table 3).Table 2b Clustering analysis stratified by study setting and lineages/major sub-lineage populations of MTBC.

bserved in the general TB patient population (p = 0.022). Three large clusters – cluster ID MSC4193, MSC5003.X, and MSC4107, with cluster sizes of 9, 7, and 7 respectively – involved only male subjects (Table 3).Table 2b Clustering analysis stratified by study setting and lineages/major sub-lineage populations of MTBC. Table 2bLineage Isolates (n) Clustered cases (c) Clustered strains (nc) Single cases (s) Total strain types (s + c) Clustering ratea (%) Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Lineage 1 34 4 3 0 7 0 27 4 30 4 11.8 0 Lineage 2 51 14 5 1 33 4 18 10 23 11 54.9 21.4 Lineage 3 25 1 2 0 4 0 21 1 23 1 8 0 Lineage 4 1588 153 183 10 907 25 681 128 864 138 45.6 9.8 Cameroonb 969 77 112 5 575 10 394 67 506 72 47.8 6.5 Ghanab 326 50 32 4 182 12 144 38 176 42 46 16 Haarlemb 144 12 20 1 81 3 63 9 83 10 42.4 16.7 LAMb 43 7 6 0 25 0 18 7 24 7 44.2 0 Ugandab 39 1 5 0 16 0 23 1 28 1 28.2 0 Lineage 5 274 15 49 0 137 0 137 15 186 15 32.1 0 Lineage 6 136 14 10 0 43 0 93 14 103 14 24.3 0 Summaryc 2108 201 252 11 1131 29 977 172 1229 183 41.7 9 MTBC, Mycobacterium tuberculosis complex. a The clustering rate was used to estimate the recent transmission rate. b Major lineage 4 sub-population. c The summary was calculated using only the items in cells corresponding to the six main lineages. Table 3 Characteristics of large molecular clusters resulting from combined 15-MIRU and spoligotyping cluster analysis.

Table 2bLineage Isolates (n) Clustered cases (c) Clustered strains (nc) Single cases (s) Total strain types (s + c) Clustering ratea (%) Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Lineage 1 34 4 3 0 7 0 27 4 30 4 11.8 0 Lineage 2 51 14 5 1 33 4 18 10 23 11 54.9 21.4 Lineage 3 25 1 2 0 4 0 21 1 23 1 8 0 Lineage 4 1588 153 183 10 907 25 681 128 864 138 45.6 9.8 Cameroonb 969 77 112 5 575 10 394 67 506 72 47.8 6.5 Ghanab 326 50 32 4 182 12 144 38 176 42 46 16 Haarlemb 144 12 20 1 81 3 63 9 83 10 42.4 16.7 LAMb 43 7 6 0 25 0 18 7 24 7 44.2 0 Ugandab 39 1 5 0 16 0 23 1 28 1 28.2 0 Lineage 5 274 15 49 0 137 0 137 15 186 15 32.1 0 Lineage 6 136 14 10 0 43 0 93 14 103 14 24.3 0 Summaryc 2108 201 252 11 1131 29 977 172 1229 183 41.7 9 MTBC, Mycobacterium tuberculosis complex. a The clustering rate was used to estimate the recent transmission rate. b Major lineage 4 sub-population. c The summary was calculated using only the items in cells corresponding to the six main lineages. Table 3 Characteristics of large molecular clusters resulting from combined 15-MIRU and spoligotyping cluster analysis. Table 3Number Cluster codea Number of cases in cluster Sex, male: female Median age (IQR) Diagnosis lapseb (months) Same residential districtc Known risk factor (number)d Lineage (sub-lineage) Drug resistancee 1 MSC4063.X 35 31:4 34 (26–44) 40 7/5/5/4/4/5 Smoking (6) Other (8) L4 (Cameroon) 3 2 MSC4060.X 34 24:10 34 (25–45) 41 6/4/3/3/3/3 Smoking (6) Other (5) L4 (Cameroon) 4 3 MSC4045.X 30 26:4 40 (29–48) 39 7/3/3/3/3 Smoking (5) HIV (4) Other (3) L4 (Cameroon) 2 4 MSC2001 27 22:5 35 (27–48) 37 8/5 Smoking (8) HIV (4) Other (2) L2 (Beijing) 1 5 MSC4031 26 19:7 41 (33–52) 36 6/4/3 Smoking (6) HIV (3) Other (1) L4 (Ghana) 11 6 MSC4110 26 21:5 38 (28–51) 39 6/3 Smoking (5) HIV (1) Other (4) L4 (Ghana) ND 7 MSC4095 24 16:8 35 (24–45) 39 7/6 Smoking (5) Other (2) L4 (Ghana) 9 8 MSC4027 21 16:5 27 (25–45) 40 6/3 Smoking (3) L4 (Ghana) 3 9 MSC4063.3 19 18:1 28 (21–45) 41 5/5 Smoking (7) L4 (Cameroon) ND 10 MSC4063.18 18 10:7 35 (24–41) 36 6/3 Smoking (4) HIV (1) L4 (Cameroon) ND 11 MSC4013 15 13:2 42 (32–55) 32 4/3 Smoking (3) HIV (2) Other (2) L4 (Haarlem) 2 12 MSC4136 15 13:2 36 (28–44) 34 6 Smoking (2) HIV (1) Other (3) L4 (Haarlem) ND 13 MSC4040 14 8:6 31 (27–45) 33 3/3 HIV (1) Other (4) L4 (Cameroon) 1 14 MSC4069.X 14 11:3 27 (23–38) 34 6/3 Smoking (2) HIV (2) Other (1) L4 (Cameroon) ND 15 MSC4073 14 9:5 40 (29–47) 24 5/4 Smoking (3) L4 (Cameroon) 3 16 MSC5002.X 14 7:7 40 (38–53) 28 5 HIV (2) Smoking (1) L5 (West African I) 2 17 MSC4063.2 13 8:4 37 (27–44) 38 4 Smoking (2) Other (3) L4 (Cameroon) ND 18 MSC4068.X 13 9:4 35 (30–44) 27 5/3 Smoking (6) HIV (2) Other (2) L4 (Cameroon) ND 19 MSC4024 12 6:5 28 (26–42) 37 3 Smoking (2) Other (1) L4 (X3) 4 20 MSC4060.18 12 7:5 35 (32–40) 36 3/3 Smoking (2) HIV (2) Other (3) L4 (Cameroon) 1 21 MSC4063.17 12 7:5 26 (24–51) 39 ND Smoking (4) HIV (1) Other (3) L4 (Cameroon) 2 22 MSC4138 11 7:4 41 (30–48) 28 4 Smoking (4) L4 (LAM) ND 23 MSC4069.3 10 5:5 32 (24–39) 31 3 Smoking (1) HIV (1) L4 (Cameroon) 2 24 MSC4104 10 7:2 35 (25–54) 34 5 Smoking (3) Other (1) L4 (Ghana) 6 25 MSC6006 10 4:6 41 (35–47) 33 5/3 Smoking (1) Other (2) L6 (West African II) ND 26 MSC4045.3 9 7:2 43 (32–50) 33 3 Smoking (1) Other (1) L4 (Cameroon) 1 27

oking (4) L4 (LAM) ND 23 MSC4069.3 10 5:5 32 (24–39) 31 3 Smoking (1) HIV (1) L4 (Cameroon) 2 24 MSC4104 10 7:2 35 (25–54) 34 5 Smoking (3) Other (1) L4 (Ghana) 6 25 MSC6006 10 4:6 41 (35–47) 33 5/3 Smoking (1) Other (2) L6 (West African II) ND 26 MSC4045.3 9 7:2 43 (32–50) 33 3 Smoking (1) Other (1) L4 (Cameroon) 1 27 MSC4060.21 9 6:3 32 (26–43) 22 ND HIV (1) Other (2) L4 (Cameroon) ND 28 MSC4060.3 9 5:4 32 (25–53) 34 3 Smoking (1) Other (2) L4 (Cameroon) ND 29 MSC4193 9 9:0 36 (30–41) 28 ND Smoking (6) HIV (1) Other (2) L4 (Cameroon) ND 30 MSC4068.3 8 6:2 45 (34–54) 27 2 Smoking (2) HIV (1) Other (1) L4 (Cameroon) ND 31 MSC4022 7 6:1 50 (46–62) 36 4 Smoking (1) Other (1) L4 (Haarlem) ND 32 MSC4060.4 7 3:4 34 (30–49) 22 5 Smoking (1) L4 (Cameroon) ND 33 MSC4080.13 7 4:3 24 (17–50) 20 3 Smoking (1) L4 (Cameroon) 1 34 MSC4082 7 6:1 35 (28–40) 33 3 Smoking (1) L4 (Ghana) ND 35 MSC4107 7 7:0 38 (29–53) 28 3/3 Smoking (2) Other (1) L4 (Ghana) 1 36 MSC5003.2 7 4:3 35 (26–57) 33 ND HIV (1) L5 (West African I) 1 37 MSC5003.X 7 7:0 43 (26–66) 34 3 Smoking (2) Other (1) L5 (West African I) ND 38 MSC6004 7 5:2 44 (36–50) 31 ND HIV (1) Other (3) L6 (West African II) 3 MIRU, mycobacterial interspersed repetitive unit; L2, lineage 2; L4, lineage 4; L5, lineage 5; L6, lineage 6; ND, none determined; IQR, interquartile range. a Cluster codes in bold font involved evidence of household transmission. b Time lapse (in months) between first diagnosed case and last diagnosed case.

MSC4060.21 9 6:3 32 (26–43) 22 ND HIV (1) Other (2) L4 (Cameroon) ND 28 MSC4060.3 9 5:4 32 (25–53) 34 3 Smoking (1) Other (2) L4 (Cameroon) ND 29 MSC4193 9 9:0 36 (30–41) 28 ND Smoking (6) HIV (1) Other (2) L4 (Cameroon) ND 30 MSC4068.3 8 6:2 45 (34–54) 27 2 Smoking (2) HIV (1) Other (1) L4 (Cameroon) ND 31 MSC4022 7 6:1 50 (46–62) 36 4 Smoking (1) Other (1) L4 (Haarlem) ND 32 MSC4060.4 7 3:4 34 (30–49) 22 5 Smoking (1) L4 (Cameroon) ND 33 MSC4080.13 7 4:3 24 (17–50) 20 3 Smoking (1) L4 (Cameroon) 1 34 MSC4082 7 6:1 35 (28–40) 33 3 Smoking (1) L4 (Ghana) ND 35 MSC4107 7 7:0 38 (29–53) 28 3/3 Smoking (2) Other (1) L4 (Ghana) 1 36 MSC5003.2 7 4:3 35 (26–57) 33 ND HIV (1) L5 (West African I) 1 37 MSC5003.X 7 7:0 43 (26–66) 34 3 Smoking (2) Other (1) L5 (West African I) ND 38 MSC6004 7 5:2 44 (36–50) 31 ND HIV (1) Other (3) L6 (West African II) 3 MIRU, mycobacterial interspersed repetitive unit; L2, lineage 2; L4, lineage 4; L5, lineage 5; L6, lineage 6; ND, none determined; IQR, interquartile range. a Cluster codes in bold font involved evidence of household transmission. b Time lapse (in months) between first diagnosed case and last diagnosed case. c Number of participants with the same district of residence. Only >2 individuals in the same residential district are indicated. ‘/’ is used to separate individuals from different districts. d ‘Other’ in this category refers to alcohol or substance abuse. e Number of participants carrying strains with drug resistance to either isoniazid or rifampicin.

c Number of participants with the same district of residence. Only >2 individuals in the same residential district are indicated. ‘/’ is used to separate individuals from different districts. d ‘Other’ in this category refers to alcohol or substance abuse. e Number of participants carrying strains with drug resistance to either isoniazid or rifampicin. Epidemiological investigations revealed both localized and dispersed recent transmission among the clustered cases, with suggested evidence of household transmission in at least six large clusters (MSC4063.X, MSC2001, MSC4095, MSC4063.18, MSC4069.X, and MSC4104). Specifically, the same L4 strain (part of cluster MSC4069.X) was found among three individuals belonging to the same household, with the oldest person (age 49 years) reporting having contact with his son who had TB 4 months prior to his episode (suggestive of household transmission). The majority of the large clusters involved TB strains circulating over almost the entire study period (Supplementary material, Figure S10). Apart from three Ghana sub-lineage clusters (MSC4104, MSC4031, and MSC4095) and one L6 cluster (MSC6004), with respectively 60% (6/10), 42% (11/26), 38% (9/24), and 43% (3/7) of isolates showing resistance to rifampicin and/or isoniazid (Table 3), such high levels of drug resistance were not observed in the other large and very large clusters. Only 2% of the isolates belonging to large and very large clusters were multidrug-resistant TB strains and this was significantly lower than that for small (2 isolates) and medium (3–5 isolates) (4%) clusters (p = 0.031).

such high levels of drug resistance were not observed in the other large and very large clusters. Only 2% of the isolates belonging to large and very large clusters were multidrug-resistant TB strains and this was significantly lower than that for small (2 isolates) and medium (3–5 isolates) (4%) clusters (p = 0.031). For the determination of possible factors associated with recent TB transmission, a general logistic regression model including all MTBC lineages was first performed, using the event of belonging to a clustered case as the outcome variable and participant variables as possible predictors (Table 4). In a separate logistic regression model, risk factors associated with recent TB transmission were tested stratified independently by L4 and L5 (Table 5), excluding L6 due to the limited sample size. In the multivariable analysis for the general logistic regression model, it was found that harboring either an isoniazid- or rifampicin-resistant MTBC strain (adjusted OR 0.7, 95% CI 0.5–0.9) was associated with a lower odds of belonging to a clustered case (Table 4). All other factors such as education status, occupation, income level, ethnicity, religion, and HIV status had no association with recent TB transmission.Table 4 Logistic regression analysis of risk factors associated with TB clustering (recent TB transmission).

h a lower odds of belonging to a clustered case (Table 4). All other factors such as education status, occupation, income level, ethnicity, religion, and HIV status had no association with recent TB transmission.Table 4 Logistic regression analysis of risk factors associated with TB clustering (recent TB transmission). Table 4Variable MTBC (N = 2309) Univariate Multivariatea Total TB cases, n (%) Clustered casesb, n (%) OR (95% CI) p-Value Adjusted OR (95% CI) p-Value Year diagnosed 2309 (100) 1229 (53·2) 2012 244 (10·6) 147 (60·3) 1·4 (1·0–1·8) 0·043 1·3 (0·9–1·7) 0·113 2013 776 (33·6) 410 (52·8) Reference 2014 707 (30·6) 365 (51·6) 1·0 (0·8–1·2) 0·642 0·9 (0·7–1·1) 0·203 2015 582 (25·2) 307(52·8) 1·0 (0·8–1·2) 0·975 1·0 (0·8–1·2) 0·703 Sex 2294 (99·4) Male 1631 (71·1) 863 (52·9) 1·0 (0·8–1·2) 0·685 Female 663 (28·9) 357 (53·8) Reference Age (years)c 2224 (96·3) <15 37 (1·7) 25 (67·6) 1·6 (0·8–3·3) 0·183 1·6 (0·8–3·2) 0·221 15–29 639 (28·7) 360 (56·3) Reference 30–39 570 (25·6) 307 (53·9) 0·9 (0·7–1·1) 0·387 0·9 (0·7–1·2) 0·688 40–59 778 (35·0) 398 (51·2) 0·8 (0·7–1·0) 0·052 0·9 (0·7–1·1) 0·241 >59 200 (9·0) 97 (48·5) 0·7 (0·5–1·0) 0·053 0·9 (0·6–1·1) 0·211 Nationality 1781 (77·1) Ghanaian 1714 (96·2) 932 (54·4) Reference Other 67 (3·8) 38 (56·7) 1·1 (0·7–1·8) 0·706 Locality 2309 (100) 1229 (53·2) Rural 201 (8·7) 74 (36·8) Reference Urban 2108 (91·3) 1155 (54·8) 2·1 (1·5–2·8) <0·001 Residence classification 1642 (71·1) Village 69 (4·2) 27 (39·1) 0·5 (0·3–0·8) 0·007 Town 182 (11·1) 96 (52·7) 0·9 (0·6–1·2) 0·415

Nationality 1781 (77·1) Ghanaian 1714 (96·2) 932 (54·4) Reference Other 67 (3·8) 38 (56·7) 1·1 (0·7–1·8) 0·706 Locality 2309 (100) 1229 (53·2) Rural 201 (8·7) 74 (36·8) Reference Urban 2108 (91·3) 1155 (54·8) 2·1 (1·5–2·8) <0·001 Residence classification 1642 (71·1) Village 69 (4·2) 27 (39·1) 0·5 (0·3–0·8) 0·007 Town 182 (11·1) 96 (52·7) 0·9 (0·6–1·2) 0·415 City residential area 52 (3·2) 27 (51·9) 0·8 (0·5–1·5) 0·564 City suburb 1136 (69·2) 636 (56·0) Reference City slum 203 (12·4) 112 (55·2) 1·0 (0·7–1·3) 0·83 Residential district 1538 (66·6) Ablekuma 545 (35·4) 298 (54·7) Reference Ashiedu Keteke 170 (11·1) 100 (58·8) 1·2 (0·8–1·7) 0·343 Ayawaso 220 (14·3) 124 (56·4) 1·1 (0·8 to 1·5) 0·672 Kpeshie 224(14·6) 121 (54·0) 1·0 (0·7–1·3) 0·867 Mamprusi East 70 (4·6) 22 (31·4) 0·4 (0·2–0·6) <0·001 Okaikoi 176 (11·4) 98 (55·7) 1·0 (0·7 to 1·5) 0·816 Osu Klottey 133 (8·6) 78 (58·7) 1·2 (0·8–1·7) 0·409 Household type 1624 (70·3) Self-contained 412 (25·4) 221 (53·6) 1·0 (0·8–1·2) 0·797 Compound house 1212 (74·6) 659 (54·4) Reference Education 1748 (75·7) Primary 222 (12·7) 125 (56·3) 1·1 (0·8–1·5) 0·637 Middle/JHS 637 (36·4) 347 (54·5) Reference Secondary 429 (24·5) 232 (54·1) 1·0 (0·8–1·3) 0·899 Tertiary 190 (10·9) 110 (57·9) 1·1 (0·8–1·6) 0·405 No education 270 (15·4) 141 (52·2) 0·9 (0·7–1·2) 0·534 Occupation 1722 (74·6) Unemployed 390 (22·6) 208 (53·3) 0·9 (0·7–1·1) 0·423 Unskilled 951 (55·2) 530 (55·7) Reference Skilled 381 (22·1) 198 (52·0) 0·9 (0·7–1·1) 0·213

Marital status 1758 (76·1) Single 766 (43·6) 431 (56·3) Reference Married 742 (42·2) 395 (53·2) 0·9 (0·7–1·1) 0·237 Divorced 167 (9·5) 99 (59·3) 1·1 (0·8–1·6) 0·476 Widowed 83 (4·7) 35 (42·2) 0·6 (0·3–0·9) 0·015 Smear positivity 2208 (95·6) Scanty 1–9 173 (7·8) 96 (55·5) 1·1 (0·8–1·5) 0·714 1+ 474 (21·5) 237 (50·0) 0·9 (0·7–1·1) 0·151 2+ 546 (24·7) 294 (53·9) 1·0 (0·8–1·2) 0·957 3+ 1015 (46·0) 548 (54·0) Reference Previous TB treatment 1737 (75·2) Yes 291 (16·8) 153 (52·6) 0·9 (0·7–1·2) 0·535 No 1446 (83·2) 789 (54·6) Reference Risk of TB contact Close friend/household 1665 (72·1) No contact 1121 (67·3) 594 (53·0) Reference 1 contact 212 (12·7) 118 (55·7) 1·1 (0·8–1·5) 0·475 2–5 contacts 309 (18·6) 179 (57·9) 1·2 (0·9–1·6) 0·123 6–10 contacts 23 (1·4) 15 (65·2) 1·7 (0·7–4·0) 0·249 Imprisonment 1660 (71·9) Yes 97 (5·8) 56 (57·7) 1·1 (0·8–1·7) 0·513 No 1563 (94·2) 849 (54·3) Reference Health/laboratory worker 1661 (71·9) Yes 47 (2·8) 25 (53·2) 0·9 (0·5–1·7) 0·85 No 1614 (97·2) 881 (54·6) Reference Immunosuppressive condition 1695 (73·4) Any 893 (52·7) 488 (54·6) 1·0 (0·9–1·2) 0·747 None 802 (47·3) 432 (53·9) Reference Diabetes mellitus 534 (23·1) Yes 104 (19·5) 54 (51·9) 1·0 (0·7–1·5) 0·957 No 430 (80·5) 222 (51·6) Reference HIV status 1166 (50·5) Positive 144 (12·3) 82 (56·9) 1·1 (0·8–1·6) 0·481 Negative 1022 (87·7) 550 (53·8) Reference Smoking 1518 (65·7) Yes 434 (28·6) 237 (54·6) 1·0 (0·8–1·2) 0·949 No 1084 (71·4) 590 (54·4) Reference Substance abuse (excluding alcohol) 1401 (60·7) Yes 140 (10·0) 84 (60·0) 1·3 (0·9–1·8) 0·172 No 1261 (90·0) 680 (53·9) Reference

HIV status 1166 (50·5) Positive 144 (12·3) 82 (56·9) 1·1 (0·8–1·6) 0·481 Negative 1022 (87·7) 550 (53·8) Reference Smoking 1518 (65·7) Yes 434 (28·6) 237 (54·6) 1·0 (0·8–1·2) 0·949 No 1084 (71·4) 590 (54·4) Reference Substance abuse (excluding alcohol) 1401 (60·7) Yes 140 (10·0) 84 (60·0) 1·3 (0·9–1·8) 0·172 No 1261 (90·0) 680 (53·9) Reference Substance abuse (including alcohol) 1474 (63·8) Yes 460 (31·2) 250 (54·3) 1·0 (0·8–1·3) 0·858 No 1014 (68·8) 546 (53·8) Reference Lineage 2309 (100) Lineage 1 38 (1·7) 7 (18·4) 0·2 (0·08–0·4) <0·001 0·13 (0·05–0·36) <0·001 Lineage 2 65 (2·8) 43 (66·2) 1·5 (0·9–2·5) 0·126 1·5 (0·9–2·5) 0·155 Lineage 3 26 (1·1) 4 (15·4) 0·1 (0·05–0·4) <0·001 0·15 (0·05–0·45) 0·001 Lineage 4 1741 (75·4) 984 (56·5) Reference Lineage 5 289 (12·5) 143 (49·5) 0·8 (0·6–1·0) 0·026 0·7 (0·6–0·9) 0·032 Lineage 6 150 (6·5) 48 (32·0) 0·4 (0·3–0·5) <0·001 0·3 (0·2–0·5) <0·001 Lineage 4 sub-lineage Cameroon 1046 (60·1) 616 (58·9) Reference Ghana 376 (21·6) 206 (54·8) 0·8 (0·7–1·1) 0·167 Haarlem 156 (9·0) 91(58·3) 1·0 (0·7–1·4) 0·895 LAM 50 (2·9) 25 (50·0) 0·7 (0·4–1·2) 0·215 Uganda 40 (2·3) 16 (40·0) 0·5 (0·2–0·9) 0·02 Other 51 (2·9) 26 (51·0) 0·7 (0·4–1·3) 0·265 Not determined 22 (1·3) 4 (18·2) 0·2 (0·1–0·5) 0·001 Drug resistance 2300 (99·6) Any 313 (13·6) 138 (44·1) 0·6 (0·5–0·8) <0·001 0·7 (0·5–0·9) 0·002 None 1987 (86·4) 1090 (54·9) Reference Isoniazid mono-resistant 2300 (99·6) Yes 295 (12·8) 129 (43·7) 0·6 (0·5–0·8) <0·001 No 2005 (87·2) 1099 (54·8) Reference Multidrug resistant (MDR) 2300 (99·6) Yes 81 (3·5) 35 (43·2) 0·7 (0·4–1·0) 0·063 No 2219 (96·5) 1193 (53·8) Reference

Drug resistance 2300 (99·6) Any 313 (13·6) 138 (44·1) 0·6 (0·5–0·8) <0·001 0·7 (0·5–0·9) 0·002 None 1987 (86·4) 1090 (54·9) Reference Isoniazid mono-resistant 2300 (99·6) Yes 295 (12·8) 129 (43·7) 0·6 (0·5–0·8) <0·001 No 2005 (87·2) 1099 (54·8) Reference Multidrug resistant (MDR) 2300 (99·6) Yes 81 (3·5) 35 (43·2) 0·7 (0·4–1·0) 0·063 No 2219 (96·5) 1193 (53·8) Reference Cluster size (n) 1227 (53·1) Small (2) 290 (23·6) Medium (3–5) 262 (21·4) Large (6–20) 452 (36·8) Very large (>20) 223 (18·2) MTBC, Mycobacterium tuberculosis complex; TB, tuberculosis; OR, odds ratio; CI, confidence interval; JHS, junior high school; GH¢, Ghanaian cedi. a For the multivariate model, only variables with p < 0.1 and with at least 90% of available data were included. However ‘locality’ was excluded due to the small sample size from the rural setting. Residence classification, marital status, isoniazid mono-resistance, and MDR were excluded due to collinearity with other variables in the model. b A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci. c A significant decreasing trend in the probability of belonging to a clustered case was found with increasing age category (p = 0.004). Table 5 Risk factors associated with TB clustering: logistic regression analysis stratified by lineage.a

b A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci. c A significant decreasing trend in the probability of belonging to a clustered case was found with increasing age category (p = 0.004). Table 5 Risk factors associated with TB clustering: logistic regression analysis stratified by lineage.a Table 5Variables Lineage 4 (n = 1741) Univariate Multivariateb Lineage 5 (n = 289) Univariate TB cases, n (%) Clustered casesc, n (%) OR (95% CI) Adjusted OR (95% CI) p-Value TB cases, n (%) Clustered casesc, n (%) OR (95% CI) p-Value Year diagnosed 1741 (100) 289 (100) 2012 183 (10·5) 120 (65·6) 1·5 (1·1–2·1)* 1·4 (1·0–2·1) 0·062 26 (9·0) 14 (53·8) 1·2 (0·5–2·9) 0·659 2013 568 (32·6) 318 (56·0) Reference 98 (33·9) 48 (49·0) Reference 2014 548 (31·5) 300 (54·7) 1·0 (0·8–1·2) 1·0 (0·7–1·3) 0·847 92 (31·8) 43 (46·7) 0·9 (0·5–1·6) 0·757 2015 442 (25·4) 244 (55·2) 1·0 (0·8–1·2) 1·0 (0·7–1·3) 0·955 73 (25·3) 38 (52·1) 1·2 (0·6–2·1) 0·691 Age (years) 1672 283 <15 27 (1·6) 20 (74·1) 2·1 (0·9–5·0) 5 (1·8) 3 (60·0) 15–29 497 (29·7) 289 (58·2) Reference 78 (27·6) 42 (53·8) 30–39 432 (25·8) 252 (58·3) 1·0 (0·8–1·3) 68 (24·0) 31 (45·6) 40–59 580 (34·7) 315 (54·3) 0·9 (0·7–1·1) 94 (33·2) 48 (51·1) >59 136 (8·1) 71 (52·2) 0·8 (0·5–1·2) 38 (13·4) 16 (42·1)

Table 5Variables Lineage 4 (n = 1741) Univariate Multivariateb Lineage 5 (n = 289) Univariate TB cases, n (%) Clustered casesc, n (%) OR (95% CI) Adjusted OR (95% CI) p-Value TB cases, n (%) Clustered casesc, n (%) OR (95% CI) p-Value Year diagnosed 1741 (100) 289 (100) 2012 183 (10·5) 120 (65·6) 1·5 (1·1–2·1)* 1·4 (1·0–2·1) 0·062 26 (9·0) 14 (53·8) 1·2 (0·5–2·9) 0·659 2013 568 (32·6) 318 (56·0) Reference 98 (33·9) 48 (49·0) Reference 2014 548 (31·5) 300 (54·7) 1·0 (0·8–1·2) 1·0 (0·7–1·3) 0·847 92 (31·8) 43 (46·7) 0·9 (0·5–1·6) 0·757 2015 442 (25·4) 244 (55·2) 1·0 (0·8–1·2) 1·0 (0·7–1·3) 0·955 73 (25·3) 38 (52·1) 1·2 (0·6–2·1) 0·691 Age (years) 1672 283 <15 27 (1·6) 20 (74·1) 2·1 (0·9–5·0) 5 (1·8) 3 (60·0) 15–29 497 (29·7) 289 (58·2) Reference 78 (27·6) 42 (53·8) 30–39 432 (25·8) 252 (58·3) 1·0 (0·8–1·3) 68 (24·0) 31 (45·6) 40–59 580 (34·7) 315 (54·3) 0·9 (0·7–1·1) 94 (33·2) 48 (51·1) >59 136 (8·1) 71 (52·2) 0·8 (0·5–1·2) 38 (13·4) 16 (42·1) Locality 1741 (100) 289 (100) Rural 153 (8·8) 59 (38·6) Reference 15 (5·2) 4 (26·7) Reference Urban 1588 (91·2) 923 (58·1) 2·2 (1·6–3·1)** 274 (94·8) 139 (50·7) 2·8 (0·9–9·1) 0·081

Age (years) 1672 283 <15 27 (1·6) 20 (74·1) 2·1 (0·9–5·0) 5 (1·8) 3 (60·0) 15–29 497 (29·7) 289 (58·2) Reference 78 (27·6) 42 (53·8) 30–39 432 (25·8) 252 (58·3) 1·0 (0·8–1·3) 68 (24·0) 31 (45·6) 40–59 580 (34·7) 315 (54·3) 0·9 (0·7–1·1) 94 (33·2) 48 (51·1) >59 136 (8·1) 71 (52·2) 0·8 (0·5–1·2) 38 (13·4) 16 (42·1) Locality 1741 (100) 289 (100) Rural 153 (8·8) 59 (38·6) Reference 15 (5·2) 4 (26·7) Reference Urban 1588 (91·2) 923 (58·1) 2·2 (1·6–3·1)** 274 (94·8) 139 (50·7) 2·8 (0·9–9·1) 0·081 Residential district 1165 189 Ablekuma 412 (35·4) 237 (57·5) Reference 77 (40·7) 39 (50·7) Reference Ashiedu Keteke 132 (11·3) 81 (61·4) 1·2 (0·8–1·8) 13 (6·9) 5 (38·5) 0·6 (0·2–2·0) 0·419 Ayawaso 178 (15·3) 111 (62·4) 1·2 (0·8–1·8) 21 (11·1) 7 (33·3) 0·5 (0·2–1·4) 0·163 Kpeshie 166 (14·2) 88 (53·0) 0·8 (0·6–1·2) 37 (19·6) 25 (67·6) 2·0 (0·9–4·6) 0·091 Mamprusi East 56 (4·8) 19 (33·9) 0·4 (0·2–0·7)* 4 (2·1) 1 (25·0) 0·32 (0·03–3·26) 0·339 Okaikoi 134 (11·5) 80 (59·7) 1·1 (0·7–1·6) 24 (12·7) 12 (50·0) 1·0 (0·4–2·4) 0·956 Osu Klottey 87 (7·5) 58 (66·7) 1·5 (0·9–2·4) 13 (6·9) 5 (38·5) 0·6 (0·2–2·0) 0·419 Monthly income (GH¢) 1222 None 275 (22·5) 167 (60·7) Reference <301 605 (49·5) 351 (58·0) 0·9 (0·7–1·2) 301–1000 314 (25·7) 184 (58·6) 0·9 (0·7–1·3) >1000 28 (2·3) 11 (39·3) 0·4 (0·2–0·9)* Marital status 1322 Single 591 (44·7) 355 (60·1) Reference Married 549 (41·5) 312 (56·8) 0·9 (0·7–1·1) 0·9 (0·7–1·2) 0·589 Divorced 124 (9·4) 78 (62·9) 1·1 (0·8–1·7) 1·1 (0·7–1·7) 0·543 Widowed 58 (4·4) 24 (41·4) 0·5 (0·3–0·8)* 0·5 (0·3–0·8) 0·011

Monthly income (GH¢) 1222 None 275 (22·5) 167 (60·7) Reference <301 605 (49·5) 351 (58·0) 0·9 (0·7–1·2) 301–1000 314 (25·7) 184 (58·6) 0·9 (0·7–1·3) >1000 28 (2·3) 11 (39·3) 0·4 (0·2–0·9)* Marital status 1322 Single 591 (44·7) 355 (60·1) Reference Married 549 (41·5) 312 (56·8) 0·9 (0·7–1·1) 0·9 (0·7–1·2) 0·589 Divorced 124 (9·4) 78 (62·9) 1·1 (0·8–1·7) 1·1 (0·7–1·7) 0·543 Widowed 58 (4·4) 24 (41·4) 0·5 (0·3–0·8)* 0·5 (0·3–0·8) 0·011 Lineage 4 sub-lineage Cameroon 1046 (60·1) 614 (58·7) Ghana 376 (21·6) 206 (54·8) 0·9 (0·7–1·1) 0·9 (0·7–1·2) 0·403 Haarlem 156 (9·0) 91 (58·3) 1·0 (0·7–1·4) 1·0 (0·7–1·5) 0·87 LAM 50 (2·9) 25 (50·0) 0·7 (0·4–1·2) 0·7 (0·4–1·4) 0·354 Uganda 40 (2·3) 16 (40·0) 0·5 (0·2–0·9)* 0·4 (0·2–0·8) 0·013 Other 51 (2·9) 26 (51·0) 0·7 (0·4–1·3) 0·8 (0·4–1·6) 0·558 Not determined 22 (1·3) 4 (18·2) 0·2 (0·1–0·5)* 0·10 (0·03–0·35) <0·001 Drug resistance 1736 Any 241 (13·9) 114 (47·3) 0·7 (0·5–0·9)* 0·7 (0·5–1·0) 0·059 None 1495 (86·1) 867 (58·0) Reference TB, tuberculosis; OR, odds ratio; CI, confidence interval; GH¢, Ghanaian cedi. a Only variables with p < 0.1 from the general logistic regression model in Table 4 were included in this analysis. *p < 0.05; **p < 0.001. b For the multivariate model, only variables with p < 0.1 and with at least 90% of available data were included. c A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci.

b For the multivariate model, only variables with p < 0.1 and with at least 90% of available data were included. c A cluster was defined as two or more isolates (same strain) that share an indistinguishable spoligotype and 15-locus MIRU-VNTR allelic pattern, but allowing for one missing allelic data at any one of the difficult-to-amplify MIRU loci. Finally, using adjusted predictions, it was found that the probability of belonging to a clustered case decreased with age and increased with the number of TB contacts (Figure 5). In a separate logistic regression analysis, including age as a continuous variable with belonging to a clustered case as the outcome variable, it was found that each year increase in age was significantly associated with an approximately 1% (95% CI 0.13–2.00%) decrease in the odds of a TB patient being part of a recent transmission event (p = 0.007).Figure 5 Adjusted predictions of the probability of belonging to a clustered case with 95% confidence interval: (A) at the year of diagnosis, (B) while ageing, (C) considering the number of close TB contact(s), and (D) considering the number of circulating Mycobacterium tuberculosis complex (MTBC) lineages. Figure 5 Discussion The aims of this study were to conduct a population-based prospective molecular epidemiological study to analyze the transmission dynamics of MTBC strains circulating in Ghana and to identify risk factors associated with recent TB transmission.

Finally, using adjusted predictions, it was found that the probability of belonging to a clustered case decreased with age and increased with the number of TB contacts (Figure 5). In a separate logistic regression analysis, including age as a continuous variable with belonging to a clustered case as the outcome variable, it was found that each year increase in age was significantly associated with an approximately 1% (95% CI 0.13–2.00%) decrease in the odds of a TB patient being part of a recent transmission event (p = 0.007).Figure 5 Adjusted predictions of the probability of belonging to a clustered case with 95% confidence interval: (A) at the year of diagnosis, (B) while ageing, (C) considering the number of close TB contact(s), and (D) considering the number of circulating Mycobacterium tuberculosis complex (MTBC) lineages. Figure 5 Discussion The aims of this study were to conduct a population-based prospective molecular epidemiological study to analyze the transmission dynamics of MTBC strains circulating in Ghana and to identify risk factors associated with recent TB transmission. A high MTBC isolate recovery rate of 78.8% was obtained, higher than that reported in similar studies (Hamblion et al., 2016, Mears et al., 2015) and this strengthens the power of the sample size to make assessments of the TB transmission rate in Ghana. This study identified a high TB clustering (recent TB transmission) rate of 41.2%, which is quite alarming, with the urban and rural areas having estimated rates of 41.7% and 9.0%, respectively (Table 2b). These findings call for intensifying community outreach programs to encourage early case reporting and infection control. Moreover, the analysis predicted the probability of clustering to generally increase with the increase in the number of TB contacts (Figure 5). This means that a susceptible individual is likely to have TB and be involved in a recently transmitted event as the number of TB contacts increases.

ing and infection control. Moreover, the analysis predicted the probability of clustering to generally increase with the increase in the number of TB contacts (Figure 5). This means that a susceptible individual is likely to have TB and be involved in a recently transmitted event as the number of TB contacts increases. Within the study population, no association of recent TB transmission was found with education status, occupation, income level, ethnicity, religion, or HIV status. However, it was observed that individuals below the age of 30 years were associated with recent TB transmission, and this is similar to observations made elsewhere (Hamblion et al., 2016, Vluggen et al., 2017). Also in this study, it was observed that each year increase in age was associated with an approximately 1% (95% CI 0.13–2.00; p = 0.007) decrease in the odds of a TB patient being part of a recent transmission event, implying that compared to younger individuals, older individuals are more likely to get active TB disease by reactivation of latent TB infection rather than through a recent transmission event (Hamblion et al., 2016). This finding puts age as a risk factor for recent TB transmission in Ghana. However, this finding was largely driven by L4 and L5, since separate analysis was not valid for L6 due to the small sample size. Furthermore, it was found that the male-to-female ratio among very large clusters was significantly higher than that observed in the general TB patient population (p = 0.022). This finding, together with the observation that some large clusters involved only male subjects, also indicates that males have a higher risk of recent TB transmission compared to females, suggesting that males may engage in certain social activities that predispose them to belonging to a recent transmission event.

). This finding, together with the observation that some large clusters involved only male subjects, also indicates that males have a higher risk of recent TB transmission compared to females, suggesting that males may engage in certain social activities that predispose them to belonging to a recent transmission event. A lower rate of multidrug-resistant TB was seen among large clustered cases compared to the general population (2% vs. 4%, p = 0.031), indicating a low multidrug-resistant TB transmissibility within the study population. This finding further suggests that the majority of drug-resistant TB cases in Ghana acquired the drug resistance during treatment, which indicates poor patient compliance (Danso et al., 2015). Moreover, it was also found that compared to drug (isoniazid and/or rifampicin)-sensitive MTBC strains, it was unlikely to find MTBC strains with isoniazid and/or rifampicin resistance involved in a recent transmission event (adjusted OR 0.7, 95% CI 0.5–0.9).

hich indicates poor patient compliance (Danso et al., 2015). Moreover, it was also found that compared to drug (isoniazid and/or rifampicin)-sensitive MTBC strains, it was unlikely to find MTBC strains with isoniazid and/or rifampicin resistance involved in a recent transmission event (adjusted OR 0.7, 95% CI 0.5–0.9). Within the study setting, a reduced transmission of MAF (L5: 31.8%, L6: 24.7%) compared to MTBss L4 (44.9%) was observed. The high recent transmission rate observed for L4 was driven by both the Cameroon and Ghana sub-lineages, with no difference in their transmissibility, hence identifying these sub-lineages as very important pathogens. The high recent transmission of the Ghana sub-lineage coupled with recently reported association with drug resistance (Otchere et al., 2016) is of public health importance and hence calls for the national tuberculosis control program to support peripheral diagnostic laboratories with facilities to accurately detect and help control the spread of the Ghana sub-lineage.

age coupled with recently reported association with drug resistance (Otchere et al., 2016) is of public health importance and hence calls for the national tuberculosis control program to support peripheral diagnostic laboratories with facilities to accurately detect and help control the spread of the Ghana sub-lineage. The higher recent transmission rate for L4 compared to L5 and L6 may not necessarily imply the outcompeting of L5 and L6 by L4, as their relative proportions remained constant over the entire study period (Figure 2) and also based on previous reports (Yeboah-Manu et al., 2016). Despite the low transmissibility of MAF, the observed stable relative proportion over the entire study period may be because the pathogen has adapted to infecting specific host populations (possibly due to unidentified host genetic or environmental factors peculiar to some West African inhabitants), hence enabling the maintenance of a stable prevalence over time. Using adjusted predictions for the probability of clustering, it was found that MAF L5 may still have the propensity to transmit equally to lineage 4 (Figure 5), not forgetting the confounding effect of a higher diversity in spoligotype pattern of L5 compared to L4 and hence reduced clustering of the former (Asante-Poku et al., 2016). Compared to L4, a significant association of L6 with individuals living in villages was found (OR 6.6, p < 0.05; Supplementary material, Table S2). The low recent TB transmission in the villages coupled with an association of L6 could be the reason why low frequencies of L6 strains were observed within the study setting.

mpared to L4, a significant association of L6 with individuals living in villages was found (OR 6.6, p < 0.05; Supplementary material, Table S2). The low recent TB transmission in the villages coupled with an association of L6 could be the reason why low frequencies of L6 strains were observed within the study setting. This report could be limited by the possibility of an underestimation of the recent transmission rate resulting from the misclassification of strains as unique if they were actually clustered outside of the restricted geographic sampling site and sampling period. However, measures were taken to address the underestimation of recent TB transmission by recruiting up to 90% of the diagnosed TB cases spanning a 3.5-year period. In addition, the possibility of overestimating recent TB transmission rates is also possible considering that the basis of the clustering analysis was done using combined 15-locus MIRU-VNTR typing and spoligotyping, whereas whole genome sequencing could have offered a better resolution of strains. Overall, the findings indicate high recent TB transmission, suggesting the occurrence of unsuspected outbreaks. The intensification of community education is recommended to improve early case reporting and infection control. Funding This research was funded by a Wellcome Trust Intermediate Fellowship Grant (097134/Z/11/Z) to Dorothy Yeboah-Manu. The funding source had no role in the study design, collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the paper for publication.

Overall, the findings indicate high recent TB transmission, suggesting the occurrence of unsuspected outbreaks. The intensification of community education is recommended to improve early case reporting and infection control. Funding This research was funded by a Wellcome Trust Intermediate Fellowship Grant (097134/Z/11/Z) to Dorothy Yeboah-Manu. The funding source had no role in the study design, collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the paper for publication. Ethical approval The Scientific and Technical Committee and then the Institutional Review Board at NMIMR, University of Ghana (FWA00001824) reviewed and approved the study. Conflict of interest We declare that we have no competing interest. Appendix A Supplementary data The following is Supplementary data to this article: Acknowledgements The authors are grateful for the administrative support of Frank Bonsu, National Tuberculosis Control Program, Ghana, and to the laboratory heads and nurses at various facilities who recruited cases and all study participants. We thank the national service personnel for providing great help in completing the questionnaires and making sputum samples available for laboratory investigations. We thank Vida Yirenkyiwaa Adjei and Portia Abena Morgan of Noguchi Memorial Institute for Medical Research for their assistance with some of the laboratory procedures. Prince Asare was supported by a West African Centre for Cell Biology of Infectious Pathogens (WACCBIP)–World Bank ACE PhD Studentship.

Acknowledgements The authors are grateful for the administrative support of Frank Bonsu, National Tuberculosis Control Program, Ghana, and to the laboratory heads and nurses at various facilities who recruited cases and all study participants. We thank the national service personnel for providing great help in completing the questionnaires and making sputum samples available for laboratory investigations. We thank Vida Yirenkyiwaa Adjei and Portia Abena Morgan of Noguchi Memorial Institute for Medical Research for their assistance with some of the laboratory procedures. Prince Asare was supported by a West African Centre for Cell Biology of Infectious Pathogens (WACCBIP)–World Bank ACE PhD Studentship. Appendix A Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.ijid.2018.05.014.

Introduction Hand, foot and mouth disease (HFMD) is a common, mild viral infection that mostly affects children under 5 years of age. The causative agent is typically enterovirus A (EV-A), which includes the serotypes Coxsackievirus A (CV-A) 2–8, 10, 12, 14, and 16 and enterovirus A71, A76, and A89–92; serotypes of enterovirus B are sometimes detected (Wang et al., 2018, Yang et al., 2011, Park et al., 2011). Since 1997 attention has focused on the Asia-Pacific region due to the occurrence of several large outbreaks, with millions of reported cases, including a small proportion of cases progressing to severe disease and even death (Chan et al., 2000, Ho et al., 1999, Zhang et al., 2010, Khanh et al., 2012). Many countries in the Asia-Pacific region now have established surveillance systems for HFMD, including China, Singapore, Taiwan, Malaysia, and Vietnam. The incidence of EV-A71-associated HFMD has increased dramatically since the Sarawak outbreak in 1997, although it varies according to country and year (Zhao et al., 2015).

countries in the Asia-Pacific region now have established surveillance systems for HFMD, including China, Singapore, Taiwan, Malaysia, and Vietnam. The incidence of EV-A71-associated HFMD has increased dramatically since the Sarawak outbreak in 1997, although it varies according to country and year (Zhao et al., 2015). Currently, there is no specific antiviral treatment for HFMD, but as the causative agent of the large outbreaks with severe illness has been EV-A71, attention has focused particularly on agents active against this virus. Several antivirals with activity against enteroviruses (including EV-A71) have been developed, but few have undergone extensive clinical evaluation (Tian et al., 2012). Vaccine development has been more successful with the completion of three phase 3 trials in 2013–14 and registration of two inactivated EV-A71 vaccines in China in 2016 (Zhu et al., 2013, Zhu et al., 2014). These vaccines have an efficacy of over 90% against EV-A71-associated HFMD, but weak protection against HFMD caused by other circulating EV-A serotypes. Protection against severe disease and fatality also remains inconclusive.

ion of two inactivated EV-A71 vaccines in China in 2016 (Zhu et al., 2013, Zhu et al., 2014). These vaccines have an efficacy of over 90% against EV-A71-associated HFMD, but weak protection against HFMD caused by other circulating EV-A serotypes. Protection against severe disease and fatality also remains inconclusive. In Vietnam, EV-A71 was first detected in 2003. The first outbreak of HFMD was reported in 2005, with CV-A16 and EV-A71 being the leading causes, although EV-A71 was responsible for cases with neurological complications (Tu et al., 2007a). Since then, increasing numbers of severe cases of HFMD have occurred. In 2008, HFMD became a notifiable disease, and on average 10 000 cases were recorded annually between 2008 and 2010. In 2011–2012, there was a dramatic increase in hospitalized and fatal cases, with a total of over 200 000 children admitted and more than 200 deaths (Khanh et al., 2012). HFMD currently occurs throughout the country, but the highest burden of disease is recorded in the southern provinces in terms of absolute numbers and proportions of severe and fatal cases (Khanh et al., 2012).

nd fatal cases, with a total of over 200 000 children admitted and more than 200 deaths (Khanh et al., 2012). HFMD currently occurs throughout the country, but the highest burden of disease is recorded in the southern provinces in terms of absolute numbers and proportions of severe and fatal cases (Khanh et al., 2012). A better understanding of the underlying aetiology of HFMD is needed for optimal therapeutic selection and vaccine development, but data on the current situation in Vietnam are limited. Most previously published work has focused on hospitalized patients with EV-A71-associated disease, and the epidemiology and aetiology of milder disease is poorly characterized. However, in recent years, clinical and national surveillance data from the region have shown the emergence of other EV-A serotypes (including CV-A6 and CV-A10) as the cause of hospitalized cases of HFMD, including cases of severe disease (Xing et al., 2014, Yang et al., 2014). Until now, most studies have focused on severe disease and inpatients. In the present study >1500 patients were enrolled, covering a wider spectrum of severity and including outpatients. This article presents a detailed characterization of HFMD in southern Vietnam. The aim was to describe the aetiological agents associated with HFMD occurring between July 2013 and July 2015, to compare the associated epidemiological and clinical characteristics of HFMD in patients infected by different pathogens, and to identify potential risk factors associated with progression to more severe illness.

m was to describe the aetiological agents associated with HFMD occurring between July 2013 and July 2015, to compare the associated epidemiological and clinical characteristics of HFMD in patients infected by different pathogens, and to identify potential risk factors associated with progression to more severe illness. Materials and methods Study setting A prospective observational study was conducted in the outpatient clinics, clinical wards, and paediatric intensive care units (PICU) of three tertiary referral hospitals in southern Vietnam: Children’s Hospital 1, Children’s Hospital 2, and the Hospital for Tropical Diseases, all located in Ho Chi Minh City. For outpatient enrolment, a dedicated study room was assigned at the outpatient clinic of each hospital and patients were enrolled daily between 7.00 am and 4.00 pm. Clinically diagnosed HFMD patients seen by clinicians in other rooms were sent to the study room for screening. For inpatient enrolment, new patients were invited to participate in the study by study doctors on the clinical wards and in the PICU. Patient enrolment All patients presenting to the three study hospitals between July 2013 and July 2015 with a clinical diagnosis of HFMD according to the guidelines of the Vietnamese Ministry of Health (Vietnam Ministry of Health, 2012) were eligible to participate in the study.

Materials and methods Study setting A prospective observational study was conducted in the outpatient clinics, clinical wards, and paediatric intensive care units (PICU) of three tertiary referral hospitals in southern Vietnam: Children’s Hospital 1, Children’s Hospital 2, and the Hospital for Tropical Diseases, all located in Ho Chi Minh City. For outpatient enrolment, a dedicated study room was assigned at the outpatient clinic of each hospital and patients were enrolled daily between 7.00 am and 4.00 pm. Clinically diagnosed HFMD patients seen by clinicians in other rooms were sent to the study room for screening. For inpatient enrolment, new patients were invited to participate in the study by study doctors on the clinical wards and in the PICU. Patient enrolment All patients presenting to the three study hospitals between July 2013 and July 2015 with a clinical diagnosis of HFMD according to the guidelines of the Vietnamese Ministry of Health (Vietnam Ministry of Health, 2012) were eligible to participate in the study. HFMD case classification Patients were classified as having a severe or non-severe illness using a grading system issued by the Vietnamese Ministry of Health based on Taiwanese guidelines using clinical signs and symptoms to guide hospital management (Huang et al., 1999). Briefly grade 1 includes patients with mouth ulcers or vesicles/papules on the hands, feet or buttocks, with or without mild fever (<39 °C). Grade 2 includes features of central nervous system (CNS) involvement, such as myoclonus, fever >39 °C, or ataxia. Grade 2 is divided into subgroups according to the reporting of myoclonus: 2A is myoclonus reported by the parents/care-givers only and grade 2B is when myoclonus is observed by a physician. Grade 2B is divided into two further subgroups: 2B1 is considered when myoclonus is observed by medical staff or there is a history of myoclonus and lethargy or pulse higher than 130 bpm; 2B2 is considered when ataxia, nystagmus, limb weakness, cranial nerve palsies, persistent high fever, or pulse higher than 150 bpm is recorded. Grade 3 includes autonomic dysfunction with sweating, hypertension, tachycardia, and tachypnea. In grade 4 disease, additional cardiopulmonary compromise with pulmonary oedema or shock syndrome are observed (Khanh et al., 2012). Patients with grade 2A and above need to be admitted to the hospital and patients with grade 2B and above are monitored in the ICU or emergency room of the infectious diseases ward (Supplementary material, Appendix 1). In this study severe HFMD was defined as grade 2B, 3, and 4.

drome are observed (Khanh et al., 2012). Patients with grade 2A and above need to be admitted to the hospital and patients with grade 2B and above are monitored in the ICU or emergency room of the infectious diseases ward (Supplementary material, Appendix 1). In this study severe HFMD was defined as grade 2B, 3, and 4. Clinical data and samples Demographic and clinical data were recorded for all patients on enrolment to the study and daily until discharge or day 7 for inpatients (whichever was sooner), using standard paper case record forms (CRFs). For outpatients, similar data were collected by telephone on days 2, 4, and 6. Inpatients who were discharged before day 7 and all outpatients returned for a follow-up visit between days 7 and 14 after enrolment. CRF completion was double-checked before entering data into an electronic data entry system. Rectal and throat swabs were collected at enrolment, in addition to the baseline haematology and biochemistry data obtained as part of routine care. An additional blood sample was collected at the follow-up visit. The enrolment process is described in detail in the Supplementary material (Appendix 2A).

Clinical data and samples Demographic and clinical data were recorded for all patients on enrolment to the study and daily until discharge or day 7 for inpatients (whichever was sooner), using standard paper case record forms (CRFs). For outpatients, similar data were collected by telephone on days 2, 4, and 6. Inpatients who were discharged before day 7 and all outpatients returned for a follow-up visit between days 7 and 14 after enrolment. CRF completion was double-checked before entering data into an electronic data entry system. Rectal and throat swabs were collected at enrolment, in addition to the baseline haematology and biochemistry data obtained as part of routine care. An additional blood sample was collected at the follow-up visit. The enrolment process is described in detail in the Supplementary material (Appendix 2A). Enterovirus detection and serotype determination Collected swabs were tested with a multiplex real-time RT-PCR using previously published generic EV and EV-A71 primers and probes (Thanh et al., 2015). For swab samples, rectal swabs were tested if throat swabs were negative (Thanh et al., 2015). Samples that were EV-positive but EV-A71-negative were then subjected to a genotyping step using a previously described VP1 sequencing assay (Perera et al., 2010, Kroneman et al., 2011), targeting the most variable and antigenic region of the viral genome. The VP1 sequences obtained were first assembled and then manually edited using Contig Express, a component of Vector NTI Suite 7 (Informax Inc., NY, USA). Finally, the sequences were analysed using an online EV typing tool available at http://www.rivm.nl/mpf/enterovirus/typingtool/to determine specific enterovirus serotypes (Kroneman et al., 2011).

e first assembled and then manually edited using Contig Express, a component of Vector NTI Suite 7 (Informax Inc., NY, USA). Finally, the sequences were analysed using an online EV typing tool available at http://www.rivm.nl/mpf/enterovirus/typingtool/to determine specific enterovirus serotypes (Kroneman et al., 2011). Statistical analysis Data were analysed with R 3.2.5 (R Development Core Team, 2010). Missing data were tested for ‘missing at random’. As no variable was ‘missing not at random’ and the missing data rate was less than 20%, cases with missing data were removed (list-wise deletion). Continuous variables were expressed using the median and interquartile range (IQR) and categorical variables as the frequency. Fisher’s exact test with p-value by Monte-Carlo simulation, the t-test, the Kruskal–Wallis test, and the independent test were used to test categorical, continuous (normal distribution), continuous (skewed distribution), and categorical variables with ordinal variables, respectively. Multiple comparisons were adjusted using Bonferroni–Holm correction. To determine independent factors associated with disease progression, a primary analysis was first performed by comparing all variables of non-progressing and progressing patients. Patients who progressed to a more severe HFMD grade during hospitalization were divided into two groups for analysis: group 1 comprised those progressing to severe illness (i.e., from grade 1/2A to grade 2B1/2B2/3/4) and group 2 comprised those progressing to cardiopulmonary complications (i.e., from grade 1/2A/2B1/2B2 to grade 3/4). All factors with p-values <0.05 were tested in a generalized linear regression model. Variables were backward-selected and all variables with a p-value of >0.05 were removed.

1/2A to grade 2B1/2B2/3/4) and group 2 comprised those progressing to cardiopulmonary complications (i.e., from grade 1/2A/2B1/2B2 to grade 3/4). All factors with p-values <0.05 were tested in a generalized linear regression model. Variables were backward-selected and all variables with a p-value of >0.05 were removed. Results Patients and characteristics Between July 2013 and July 2015, a total of 1547 patients were enrolled in the study. Baseline characteristics are given in Table 1. Patients were enrolled relatively early after the onset of symptoms (median 1 day, IQR 1–2 days). The median age of the children was 21.2 months (IQR 12.4–25.7 months). A total of 1423 (92%) patients had mild HFMD (grade 1 or 2A) on admission/enrolment and 1355 (87.6%) patients were classified as having mild disease at discharge, with 119 patients progressing to a higher grade during hospitalization (Table 1). Three patients were discharged with limb weakness; no fatalities were recorded in the study group (Supplementary material, Appendix 2B).Table 1 Patient characteristics and comparison between inpatients and outpatients.a

ease at discharge, with 119 patients progressing to a higher grade during hospitalization (Table 1). Three patients were discharged with limb weakness; no fatalities were recorded in the study group (Supplementary material, Appendix 2B).Table 1 Patient characteristics and comparison between inpatients and outpatients.a Table 1 Total (n = 1547) Outpatients (n = 590) Inpatients (n = 957) p-Valueb Demographics Sex (male) 933 (61.3) 344 (58.3) 589 (61.5) 0.2 Age (months) 21.2 (12.4–25.7) 20.2 (13.2–30.3) 16.8 (12.1–24) *** Ho Chi Minh City origin 893 (57.7) 357 (60.5) 536 (56) 0.08 Illness day on admission (days) 1.4 (1–2) 1.2 (1–2) 1.5 (1–2) *** Duration of hospital stay (days) 4.1 (2–5) NA 4.1 (2–5) NA Clinical signs and symptoms Fever 1311 (73.2) 377 (63.9) 755 (78.9) *** Rash 1310 (85.6) 541 (91.7) 769 (82.3) *** Type of rash *** Mostly vesicular 63 (4.9) 46 (8.5) 17 (2.2) * Mostly macular 601 (46.2) 215 (39.7) 386 (50.9) Both 636 (48.9) 281 (51.8) 355 (46.8) Mouth lesion 1345 (88.2) 506 (85.8) 839 (89.7) * Lips 64 (5) 61 (10.3) 3 (0.4) *** Tongue 512 (39.9) 188 (31.9) 324 (46.8) *** Palate 959 (74.6) 434 (73.6) 525 (75.8) 0.5 Buccal 256 (20) 201 (34.1) 55 (7.9) *** Headache 22 (1.9) 19 (5.5) 3 (0.4) *** Cough 347 (22.5) 207 (35.1) 140 (14.7) *** Runny nose 326 (21.2) 214 (36.3) 112 (11.8) *** Conjunctivitis 17 (1.1) 12 (2) 5 (0.5) * Vomiting 331 (21.4) 145 (24.6) 186 (19.5) * Diarrhoea 86 (5.6) 33 (5.6) 53 (5.6) 1 Drowsiness 35 (2.3) 9 (1.5) 26 (2.8) 0.1 Irritability 213 (13.8) 145 (24.6) 68 (7.2) *** Myoclonus 261 (16.9) 64 (10.8) 197 (20.6) *** Sweating 13 (0.84) 6 (1) 7 (0.7) 0.5 Lethargy 12 (0.8) 2 (0.3) 10 (1.1) 0.1 Ataxia 4 (0.4) 0 4 (0.4) 1 Tremor 30 (3.1) 0 30 (3.2) 1 Nystagmus 3 (0.3) 0 3 (0.3) 1 Limb weakness 5 (0.5) 0 5 (0.5) 1 Temperature (°C) 37.5 (35.6–38.2) 37 (36.9–37.5) 38 (37.4–38.5) *** Pulse (bpm) 120 (110–130) 110 (100–110) 124 (120–130) *** Systolic blood pressure (mmHg) 90 (85–90) 90 (80–90) 90 (90–97) *** Diastolic blood pressure (mmHg) 55 (50–60) 50 (50–60) 60 (55–60) *** Body mass index (kg/m2) 17.5 (15.6–18.6) 17.7 (16–18.7) 17.4 (15.5–18.5) 0.2 Blood results White blood cell count (109/l) 12.5 (9.7–15.9) 11 (8.8–13.6) 13.6 (10.7–17.1) *** Neutrophil (%) 48.7 (38.4–59.3) 46.2 (37.1–57.5) 50.3 (39.5–60.6) *** Lymphocyte (%) 37.6 (27.8–47.7) 39.2 (29.3–48.2) 36.5 (26.9–46.5) ** Platelet count (109/ml) 300 (248–359) 306 (256.5–354) 296 (241–364) 0.4 Blood glucose (mg/dl) 99 (83–114) 112 (102–126) 88 (76–101) *** C-reactive protein (mg/l) 7.6 (2.4–22) 3

*** Neutrophil (%) 48.7 (38.4–59.3) 46.2 (37.1–57.5) 50.3 (39.5–60.6) *** Lymphocyte (%) 37.6 (27.8–47.7) 39.2 (29.3–48.2) 36.5 (26.9–46.5) ** Platelet count (109/ml) 300 (248–359) 306 (256.5–354) 296 (241–364) 0.4 Blood glucose (mg/dl) 99 (83–114) 112 (102–126) 88 (76–101) *** C-reactive protein (mg/l) 7.6 (2.4–22) 3 (1–7.5) 13 (6–28.1) *** NA, not applicable. a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables; denominators may vary slightly. b p-Values: *p < 0.05, **p < 0.01, ***p < 0.001. Enterovirus serotype detection and monthly distribution Overall, 1327 patients (85.8%) were positive for enterovirus by multiplex real-time PCR. Serotyping and specific PCR analyses identified a total of 20 EV serotypes in HFMD patients in southern Vietnam during the study period (Figure 1A, B). The four dominant serotypes were EV-A71 (n = 378, 24.4%), CV-A6 (n = 337, 21.8%), CV-A10 (n = 122, 7.9%), and CV-A16 (n = 167, 10.8%), with EV-A71 being the most common (Figure 1A). The predominant sub-genogroup of EV-A71 in this cohort was B5 (unpublished data).Figure 1 A) pie chart showing the proportions of four dominant genotypes of EV-A detected among 1547 patients with clinical HFMD enrolled between July 2013 and July 2015; B) pie chart showing the proportions of other genotypes of EV detected among 138 patients with clinical HFMD enrolled between July 2013 and July 2015. Figure 1

Enterovirus serotype detection and monthly distribution Overall, 1327 patients (85.8%) were positive for enterovirus by multiplex real-time PCR. Serotyping and specific PCR analyses identified a total of 20 EV serotypes in HFMD patients in southern Vietnam during the study period (Figure 1A, B). The four dominant serotypes were EV-A71 (n = 378, 24.4%), CV-A6 (n = 337, 21.8%), CV-A10 (n = 122, 7.9%), and CV-A16 (n = 167, 10.8%), with EV-A71 being the most common (Figure 1A). The predominant sub-genogroup of EV-A71 in this cohort was B5 (unpublished data).Figure 1 A) pie chart showing the proportions of four dominant genotypes of EV-A detected among 1547 patients with clinical HFMD enrolled between July 2013 and July 2015; B) pie chart showing the proportions of other genotypes of EV detected among 138 patients with clinical HFMD enrolled between July 2013 and July 2015. Figure 1 Other EV serotypes were detected in 138 patients, with CV-A4 being identified in 21.7%, followed by CV-A12 in 19.6%, CV-A2 in 15.2%, CV-A8 in 7.2%, CV-A5 in 5.1%, other enterovirus A in 1.4%, and enterovirus B in 23.9% (Figure 1B). Temporally, the four most common enterovirus serotypes (EV-A71, CV-A6, CV-A10, and CV-A16) replaced each other over the entire study period (Figure 2A). This phenomenon was seen in both the inpatient and outpatient groups (Figure 2B, C).Figure 2 Monthly distribution of enterovirus genotypes over the study period (July 2013 and July15); A) all 1547 out- and inpatients; B) 590 outpatients and C) 947 inpatients. Figure 2

Temporally, the four most common enterovirus serotypes (EV-A71, CV-A6, CV-A10, and CV-A16) replaced each other over the entire study period (Figure 2A). This phenomenon was seen in both the inpatient and outpatient groups (Figure 2B, C).Figure 2 Monthly distribution of enterovirus genotypes over the study period (July 2013 and July15); A) all 1547 out- and inpatients; B) 590 outpatients and C) 947 inpatients. Figure 2 Demographics, clinical characteristics, and laboratory data by pathogens When stratifying the data by pathogens, there was considerable homogeneity in terms of demographics, clinical characteristics, and laboratory data between patients infected with the different enterovirus A serotypes. Variables for which there were statistically significant differences between patient groups are shown in Table 2.Table 2 Clinical characteristics and laboratory results by pathogen.a

erms of demographics, clinical characteristics, and laboratory data between patients infected with the different enterovirus A serotypes. Variables for which there were statistically significant differences between patient groups are shown in Table 2.Table 2 Clinical characteristics and laboratory results by pathogen.a Table 2Characteristic CV-A10 (n = 122) CV-A16 (n = 167) CV-A6 (n = 337) EV-A71 (n = 378) p-Value Age (months) 14.3 (10.8–19.7)b 19.1 (14.2–27) 16.5 (11.9–22.4) 21.9 (14.2–32.1) <0.001 Length of hospital stay (days) 3 (2–5) 3 (2–4) 3 (2–5) 4 (3–6)b <0.001 Myoclonus 15 (12%) 18 (11%) 49 (15%) 90 (24%)b 0.002 Irritability 13 (11%) 18 (11%) 49 (15%) 89 (24%)b <0.001 Tremor 0 (0) 0 (0) 1 (1%) 25 (12%)b <0.001 Erythema 78 (64%)b 155 (93%) 330 (98%) 345 (95%) <0.001 Mouth ulcers 117 (96%)b 156 (93%) 294 (87%) 311 (85%) <0.001 White blood cell count (109/l) 14.65 (12.18–18.1)b 12.2 (9.2–14.5) 12.5 (9.9–16.6) 11.7 (9.5–14.7) <0.001 Platelet count (109/l) 294 (250–364) 298 (251–342) 308 (249.7–359.2) 309 (261–376.75) 0.026 Glucose (mg/dl) 94.95 (82–111) 99.5 (81.1–112.2) 101 (85.36–117.5) 102 (86–118) 0.01 C-reactive protein (mg/l) 17.35 (6–34.23)b 7 (1.4–20) 10 (4–26) 5 (1.2–10) <0.001 CV, Coxsackievirus; EV, enterovirus. a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables; denominators may vary slightly. p-Values are adjusted for multiple comparison. b Factors that are significantly different when compared with other groups.

Table 2Characteristic CV-A10 (n = 122) CV-A16 (n = 167) CV-A6 (n = 337) EV-A71 (n = 378) p-Value Age (months) 14.3 (10.8–19.7)b 19.1 (14.2–27) 16.5 (11.9–22.4) 21.9 (14.2–32.1) <0.001 Length of hospital stay (days) 3 (2–5) 3 (2–4) 3 (2–5) 4 (3–6)b <0.001 Myoclonus 15 (12%) 18 (11%) 49 (15%) 90 (24%)b 0.002 Irritability 13 (11%) 18 (11%) 49 (15%) 89 (24%)b <0.001 Tremor 0 (0) 0 (0) 1 (1%) 25 (12%)b <0.001 Erythema 78 (64%)b 155 (93%) 330 (98%) 345 (95%) <0.001 Mouth ulcers 117 (96%)b 156 (93%) 294 (87%) 311 (85%) <0.001 White blood cell count (109/l) 14.65 (12.18–18.1)b 12.2 (9.2–14.5) 12.5 (9.9–16.6) 11.7 (9.5–14.7) <0.001 Platelet count (109/l) 294 (250–364) 298 (251–342) 308 (249.7–359.2) 309 (261–376.75) 0.026 Glucose (mg/dl) 94.95 (82–111) 99.5 (81.1–112.2) 101 (85.36–117.5) 102 (86–118) 0.01 C-reactive protein (mg/l) 17.35 (6–34.23)b 7 (1.4–20) 10 (4–26) 5 (1.2–10) <0.001 CV, Coxsackievirus; EV, enterovirus. a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables; denominators may vary slightly. p-Values are adjusted for multiple comparison. b Factors that are significantly different when compared with other groups. Patients with EV-A71-associated HFMD were significantly older than the others (21.9 months, IQR 14.2–32.1 months), while patients with CV-A10-associated HFMD were the youngest (14.3 months, IQR 10.8–19.7 months) (Table 2).

a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables; denominators may vary slightly. p-Values are adjusted for multiple comparison. b Factors that are significantly different when compared with other groups. Patients with EV-A71-associated HFMD were significantly older than the others (21.9 months, IQR 14.2–32.1 months), while patients with CV-A10-associated HFMD were the youngest (14.3 months, IQR 10.8–19.7 months) (Table 2). EV-A71 was detected at a higher frequency among patients with severe disease (grade 2B or more) compared to the other viruses (p < 0.01) (Figure 3). Likewise, other clinical presentations reflecting clinically severe HFMD (including duration of hospital stay, myoclonal jerks, tremors, and irritability) were more often observed among EV-A71-infected patients (Table 2). Compared with the other groups, CV-A10-infected patients had fewer skin lesions but more often had mouth ulcers (78/122 (64%) and 117/122 (96%), respectively (Table 2)). CV-A10-infected patients also had higher values for both white blood cell count and C-reactive protein (CRP) than the other groups (Table 2).Figure 3 Distribution of pathogen detected by severity. Figure 3

EV-A71 was detected at a higher frequency among patients with severe disease (grade 2B or more) compared to the other viruses (p < 0.01) (Figure 3). Likewise, other clinical presentations reflecting clinically severe HFMD (including duration of hospital stay, myoclonal jerks, tremors, and irritability) were more often observed among EV-A71-infected patients (Table 2). Compared with the other groups, CV-A10-infected patients had fewer skin lesions but more often had mouth ulcers (78/122 (64%) and 117/122 (96%), respectively (Table 2)). CV-A10-infected patients also had higher values for both white blood cell count and C-reactive protein (CRP) than the other groups (Table 2).Figure 3 Distribution of pathogen detected by severity. Figure 3 Comparison between inpatients and outpatients There were considerable differences in terms of demographics, clinical presentations, and laboratory test results between inpatients and outpatients (Table 1), while the aetiological patterns and monthly distribution of the detected enterovirus serotypes were similar in the two groups (Figure 2B, C). Notably, outpatients were older, more likely to have a rash, and had higher blood glucose levels than inpatients (Table 1). Meanwhile, inpatients presented to the hospital at a later time of illness, had a higher temperature, and had a higher white blood cell count and CRP than outpatients (Table 1).

ps (Figure 2B, C). Notably, outpatients were older, more likely to have a rash, and had higher blood glucose levels than inpatients (Table 1). Meanwhile, inpatients presented to the hospital at a later time of illness, had a higher temperature, and had a higher white blood cell count and CRP than outpatients (Table 1). Factors associated with progression to severe disease One hundred and nineteen inpatients (8.8%) progressed to a more severe grade of HFMD, of whom 98 (6.3%) progressed from non-severe grades to severe grades. A further 21 patients progressed within the mild grade category, i.e. from grade 1 to grade 2A (Table 3). Patients progressing from mild to severe grades were less likely to have a rash and mouth lesions on admission than those who did not. The types of rash also differed between the two groups. Vesicular and mixed rashes were seen more frequently in the non-progressing group, while progressing patients more often presented with a macular rash (Table 4). Myoclonus, tremor, and limb weakness at presentation were also less likely to be seen in the non-progressing group. Patients who progressed to the most severe grades (grade 3 and 4) with cardiopulmonary complications and/or autonomic dysfunction had a higher incidence of drowsiness and lethargy at presentation. Patients in this group also presented with a higher heart rate and systolic blood pressure, in addition to higher blood glucose and lower CRP than non-progressing patients or patients with milder disease (Table 4).Table 3 Number of progressing cases.a

n had a higher incidence of drowsiness and lethargy at presentation. Patients in this group also presented with a higher heart rate and systolic blood pressure, in addition to higher blood glucose and lower CRP than non-progressing patients or patients with milder disease (Table 4).Table 3 Number of progressing cases.a Table 3 aGrey squares indicate patients with the same grade recorded at admission and discharge (non-progression). bPatients who progressed from low grades to a higher grade (more severe) during their hospital stay. Table 4 Comparison between progressing and non-progressing cases.a

n had a higher incidence of drowsiness and lethargy at presentation. Patients in this group also presented with a higher heart rate and systolic blood pressure, in addition to higher blood glucose and lower CRP than non-progressing patients or patients with milder disease (Table 4).Table 3 Number of progressing cases.a Table 3 aGrey squares indicate patients with the same grade recorded at admission and discharge (non-progression). bPatients who progressed from low grades to a higher grade (more severe) during their hospital stay. Table 4 Comparison between progressing and non-progressing cases.a Table 4 Did not progress from non-severe to severe (n = 859) Progressed from non-severe to severe (n = 98) p-Value1 Did not progress to cardiopulmonary complication (n = 907) Progressed to autonomic dysfunction and/or cardiopulmonary complication (n = 50) p-Value2 Demographics Sex (male) 517 (61.7) 72 (60.5) 0.8 558 (61.5) 31 (62) 1 Age (months) 16.6 (12.1–23.5) 18.7 (14.9–26.9) 0.2 16.7 (12.1–23.7) 18.7 (11.2–26.9) 0.7 Ho Chi Minh City origin 474 (55.2) 62 (63.3) 0.1 505 (55.7) 31 (62) 0.6 Illness day on admission (days) 1 (1–2) 2 (1–2) 0.002 1 (1–2) 2 (1–2) 0.05 Clinical signs and symptoms Rash 698 (83.4) 71 (73.2) 0.02 737 (83.3) 32 (65.3) 0.001 Type of rash <0.001 <0.001 Mostly vesicular 13 (1.9) 4 (5.6) 15 (2.1) 2 (6.1) Mostly macular 339 (49.3) 47 (66.2) 360 (49.7) 26 (78.8) Both 335 (48.8) 20 (28.2) 350 (48.3) 5 (15.2) Mouth lesion 767 (91.5) 72 (74.2) <0.001 808 (91.2) 31 (63.3) <0.001 Lips 3 (0.5) 0 1 3 (0.4) 0 1 Tongue 305 (46.8) 19 (46.3) 1 314 (46.2) 10 (74.1) 0.06 Palate 499 (76.5) 26 (63.4) 0.06 518 (76.5) 7 (43.8) 0.005 Buccal 55 (8.4) 0 0.07 55 (8.1) 0 0.6 Headache 3 (0.4) 0 1 3 (0.4) 0 1 Cough 127 (14.9) 13 (13.5) 0.9 133 (14.8) 7 (13.5) 1 Runny nose 104 (12.2) 8 (8.2) 0.3 107 (11.9) 5 (10.2) 1 Conjunctivitis 4 (0.5) 1 (1) 0.4 4 (0.5) 1 (2) 0.2 Vomiting 167 (19.5) 19 (19.4) 1 175 (19.3) 11 (22) 0.6 Diarrhoea 50 (5.8) 3 (3.1) 0.3 51 (5.6) 2 (4.1) 1 Drowsiness 20 (2.4) 6 (6.6) 0.03 21 (2.4) 5 (11.6) 0.005 Irritability 61 (7.2) 7 (7.2) 1 66 (7.3) 2 (4.1) 0.6 Myoclonus 167 (19.5) 30 (30.6) 0.01 185 (20.4) 12 (24) 0.6 Sweating 6 (0.7) 1 (1) 0.5 7 (0.8) 0 1 Lethargy 7 (0.8) 3 (3.1) 0.08 7 (0.8) 3 (6) 0.01 Ataxia 4 (0.5) 0 1 4 (0.5) 0 1 Tremor 22 (2.6) 8 (8.8) 0.005 29 (3.2) 1 (2.3) 1 Nystagmus 3 (0.4) 0 1 3 (0.3) 0 1 Limb weakness 2 (0.2) 3 (3.3) 0.008 4 (0.5) 1 (2.4) 0.2 Temperature (°C) 38 (37.3–38.5) 38 (37.5–38.7) 0.4 38 (37.4–38.5) 37.8 (37.5–38.2) 0.3 Pulse (bpm) 122 (120–130) 130 (122–142) <0.001 124 (120–130) 130 (120–142) 0.001 Systolic blood pressure (mmHg) 90 (90–95) 98 (90–104.5) <0.001 90 (90–95) 98 (90–105) <0.001 Diastolic blood pressure (mmHg) 60 (55–60) 60 (55–60) 0.5 60 (55–60) 60 (54–60) 0.3 Body mass index (kg/m2) 17 (15.5–18.6) 16

37.8 (37.5–38.2) 0.3 Pulse (bpm) 122 (120–130) 130 (122–142) <0.001 124 (120–130) 130 (120–142) 0.001 Systolic blood pressure (mmHg) 90 (90–95) 98 (90–104.5) <0.001 90 (90–95) 98 (90–105) <0.001 Diastolic blood pressure (mmHg) 60 (55–60) 60 (55–60) 0.5 60 (55–60) 60 (54–60) 0.3 Body mass index (kg/m2) 17 (15.5–18.6) 16 (15.5–18.2) 0.2 16.8 (15.5–18.6) 16.2 (14.9–17.9) 0.1 Blood results White blood cell count (109/l) 13.6 (10.6–17.2) 13.1 (11.5–16.5) 1 13.6 (10.6–17.2) 13.2 (11.7–17.1) 0.7 Neutrophil (%) 50 (39.5–60.4) 45.6 (40.6–63.1) 0.054 50.2 (39.5–60.5) 54.5 (40.1–63.3) 0.3 Lymphocyte (%) 37 (27–46.6) 33 (25–45.8) 0.06 36.7 (27.1–46.4) 35.2 (23.9–48.2) 0.3 Platelet count (109/ml) 294 (237.5–361.5) 320 (277–412) 0.051 295 (240–364) 322 (299.8–383.5) 0.1 Blood glucose (mg/dl) 87 (75–100) 97 (84.3–108.5) <0.001 88 (75–101) 94 (85–106.7) 0.008 C-reactive protein (mg/l) 14 (6–29) 6 (2.1–21) <0.001 13.2 (6–28.2) 6 (3–22.5) 0.004 a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables.

83.5) 0.1 Blood glucose (mg/dl) 87 (75–100) 97 (84.3–108.5) <0.001 88 (75–101) 94 (85–106.7) 0.008 C-reactive protein (mg/l) 14 (6–29) 6 (2.1–21) <0.001 13.2 (6–28.2) 6 (3–22.5) 0.004 a Note: data are presented as the number (%) for categorical variables and as the median (interquartile range) for continuous variables. p-Value1: p-values for cases that progressed from a non-severe grade to a severe grade. p-Value2: p-values for cases that progressed from a grade without autonomic dysfunction/cardiopulmonary involvement to a grade with autonomic dysfunction/cardiopulmonary involvement. The progressed from non-severe to severe group included patients admitted to the hospital with grade 1 or 2A who then progressed to grade 2B or higher. The progressed to autonomic dysfunction and/or cardiopulmonary complication group included patients admitted to the hospital with grade 1/2A who then progressed to grade 3 or grade 4.

from non-severe to severe group included patients admitted to the hospital with grade 1 or 2A who then progressed to grade 2B or higher. The progressed to autonomic dysfunction and/or cardiopulmonary complication group included patients admitted to the hospital with grade 1/2A who then progressed to grade 3 or grade 4. Multivariate analysis showed that admission features of vesicular rash, increased blood pressure, high heart rate, and elevated blood glucose were associated with an increased risk of progression to severe grades (Table 5). For every 10 bpm increase in heart rate, the risk of progression increased by 0.54 (95% confidence interval 1.2–2). Similarly, for every 10 mmHg increase in systolic blood pressure, the risk of progression increased by 1.3 (95% confidence interval 1.4–3.8). However, when analysing only cases who progressed to very severe disease (grade 3 and 4, with autonomic dysfunction and/or cardiopulmonary complications), the presence of a skin rash was associated with protection from progression. When rash was present, a macular rash was significantly associated with an increased risk of progression (Table 6).Table 5 Multivariate analysis of factors associated with disease progression from non-severe (grade 1–2A) to severe (all severe cases, grade 2B or higher).a Table 5Factors OR (95% CI) p-Value Type of rash (vesicular) 8.7 (1.4–43) 0.01 Blood glucose 2.9 (1.04–8.3) 0.04 Pulse 1.54 (1.2–2) <0.001 Systolic blood pressure 2.3 (1.4–3.8) 0.001 OR, odds ratio; CI, confidence interval.

Multivariate analysis showed that admission features of vesicular rash, increased blood pressure, high heart rate, and elevated blood glucose were associated with an increased risk of progression to severe grades (Table 5). For every 10 bpm increase in heart rate, the risk of progression increased by 0.54 (95% confidence interval 1.2–2). Similarly, for every 10 mmHg increase in systolic blood pressure, the risk of progression increased by 1.3 (95% confidence interval 1.4–3.8). However, when analysing only cases who progressed to very severe disease (grade 3 and 4, with autonomic dysfunction and/or cardiopulmonary complications), the presence of a skin rash was associated with protection from progression. When rash was present, a macular rash was significantly associated with an increased risk of progression (Table 6).Table 5 Multivariate analysis of factors associated with disease progression from non-severe (grade 1–2A) to severe (all severe cases, grade 2B or higher).a Table 5Factors OR (95% CI) p-Value Type of rash (vesicular) 8.7 (1.4–43) 0.01 Blood glucose 2.9 (1.04–8.3) 0.04 Pulse 1.54 (1.2–2) <0.001 Systolic blood pressure 2.3 (1.4–3.8) 0.001 OR, odds ratio; CI, confidence interval. a Note: in the model, laboratory values of blood glucose were calculated by ln2. For pulse and systolic blood pressure, changes were evaluated for every 10 bpm and 10 mmHg, respectively. Table 6 Multivariate analysis of factors associated with disease progression from non-severe (grade 1–2A) to severe with autonomic dysfunction and/or cardiopulmonary complications (grade 3–4).

a Note: in the model, laboratory values of blood glucose were calculated by ln2. For pulse and systolic blood pressure, changes were evaluated for every 10 bpm and 10 mmHg, respectively. Table 6 Multivariate analysis of factors associated with disease progression from non-severe (grade 1–2A) to severe with autonomic dysfunction and/or cardiopulmonary complications (grade 3–4). Table 6Factors OR (95% CI) p-Value Type of rash (macular) 3.2 (1.09–11.7) 0.04 Having skin lesions 0.03 (0.0007–1.15) 0.03 OR, odds ratio; CI, confidence interval.

Table 6 Multivariate analysis of factors associated with disease progression from non-severe (grade 1–2A) to severe with autonomic dysfunction and/or cardiopulmonary complications (grade 3–4). Table 6Factors OR (95% CI) p-Value Type of rash (macular) 3.2 (1.09–11.7) 0.04 Having skin lesions 0.03 (0.0007–1.15) 0.03 OR, odds ratio; CI, confidence interval. Discussion This article reports a novel longitudinal study and in-depth analysis of the aetiology and associated clinical phenotypes of both inpatients and outpatients with HFMD admitted to three tertiary referral hospitals in southern Vietnam, a region where the majority of HFMD cases in Vietnam have been reported to date. Between July 2013 and July 2015, a total of 1547 HFMD patients were recruited. Between 2011 and 2012 Vietnam experienced an explosive outbreak of HFMD in terms of both hospitalized cases and fatal cases, with over 200 000 cases and 200 deaths (Khanh et al., 2012). The number and proportion of severe cases recorded in the present study was less than anticipated based on the previous 2 years of observations. Whilst the underlying mechanism remains unknown, previous research by the present authors showed that genetically EV-A71, the main cause of severe HFMD, had undergone a switch from sub-genogroup C4 to sub-genogroup B5 by the end of 2012 (Yi et al., 2017, Geoghegan et al., 2015), and since then B5 has continued to circulate in southern Vietnam. Intriguingly, the cyclical 2–3 year pattern of HFMD and its pathogens reported from Japan and Malaysia has not been seen in Vietnam (NikNadia et al., 2016, Iwai et al., 2009).

sub-genogroup B5 by the end of 2012 (Yi et al., 2017, Geoghegan et al., 2015), and since then B5 has continued to circulate in southern Vietnam. Intriguingly, the cyclical 2–3 year pattern of HFMD and its pathogens reported from Japan and Malaysia has not been seen in Vietnam (NikNadia et al., 2016, Iwai et al., 2009). This study is one of the few to record the characteristics of mild cases of HFMD, finding a significant proportion of EV-A71 among these non-severe cases (168/590, 28.5%). The proportion of EV-A71-infected patients in mild and moderately severe cases (grade 1 and 2A) was remarkable in this study, with a detection rate of around 20%. All of these patients recovered completely without any complications. On the other hand, as EV-A71 was detected more frequently among severe patients, a higher rate of neurological symptoms among EV-A71-positive patients was also found. This is the first study to report CV-A6-associated HFMD, previously not detected in Vietnam, including in the 2005 HFMD outbreak (Tu et al., 2007b); indeed this was the second most frequently detected pathogen after EV-A71. From October 2013 to October 2014, CV-A6 emerged, replacing CV-A16 and together with EV-A71 becoming the two most common pathogens in southern Vietnam, similar to reports from mainland China. Likewise other countries worldwide have also described the simultaneous emergence of CV-A6 (Yang et al., 2014, Lu et al., 2012, Kobayashi et al., 2013, Hayman et al., 2014, Hongyan et al., 2014, Ramirez-Fort et al., 2014).

coming the two most common pathogens in southern Vietnam, similar to reports from mainland China. Likewise other countries worldwide have also described the simultaneous emergence of CV-A6 (Yang et al., 2014, Lu et al., 2012, Kobayashi et al., 2013, Hayman et al., 2014, Hongyan et al., 2014, Ramirez-Fort et al., 2014). The four main serotypes (EV-A71, CV-A6, CV-A16, and CV-A10) continuously varied in dominance and detection rates and replaced each other during the study period, while accounting for 1005/1327 (75.7%) of all EV-positive HFMD cases. This suggests complex patterns of aetiology and serotype replacement among HFMD patients, and emphasizes the need for active surveillance of circulating pathogens and to develop bivalent/multivalent vaccines to control HFMD. To support this, it is also important to assess the inter-relatedness among HFMD-causing pathogens as well as between other enteroviruses, particularly the level of cross-immunity in natural infection and post-vaccination.

rveillance of circulating pathogens and to develop bivalent/multivalent vaccines to control HFMD. To support this, it is also important to assess the inter-relatedness among HFMD-causing pathogens as well as between other enteroviruses, particularly the level of cross-immunity in natural infection and post-vaccination. Several differences in clinical features of the cases were detected. Inpatients presented significantly less often with rash than outpatients. This may represent either a health-seeking or diagnostic bias: either parents are prompted to go to hospital earlier when their children have a rash, or doctors only diagnose HFMD with subtle or no rash as HFMD when (characteristic) more severe symptoms of neurological involvement develop or on laboratory confirmation. Clinical phenotypes also appeared to differ according to the underlying viral aetiology. Patients testing positive for CV-A10 had a typical clinical presentation, with low frequency of skin rash but high frequency of mouth ulcers. Furthermore, they had higher white blood cell counts and CRP levels, which may suggest more severe illness or bacterial co-infection. As fever for more than 48 h is one of the admission criteria for HFMD according to the Vietnamese Ministry of Health guidelines, this finding suggest that patients with fever for more than 48 h only may not need to be admitted. Other signs of neurological involvement and/or laboratory test results such as CRP should also be considered for admission.

one of the admission criteria for HFMD according to the Vietnamese Ministry of Health guidelines, this finding suggest that patients with fever for more than 48 h only may not need to be admitted. Other signs of neurological involvement and/or laboratory test results such as CRP should also be considered for admission. The lower median age of CV-A6- and CV-A10-infected patients suggests the absence of background (maternal) immunity and recent introductions into Vietnam, which merits further analysis into the evolutionary process of the pathogens and the level of cross-immunity between common HFMD-causing enteroviruses.

one of the admission criteria for HFMD according to the Vietnamese Ministry of Health guidelines, this finding suggest that patients with fever for more than 48 h only may not need to be admitted. Other signs of neurological involvement and/or laboratory test results such as CRP should also be considered for admission. The lower median age of CV-A6- and CV-A10-infected patients suggests the absence of background (maternal) immunity and recent introductions into Vietnam, which merits further analysis into the evolutionary process of the pathogens and the level of cross-immunity between common HFMD-causing enteroviruses. Similar to previous reports from Taiwan and China, several clinical features on admission that were indicative of progression to more severe stages of the disease were documented. A vesicular rash was shown to be associated with progression to more severe illness in the first regression model. However, due to the small number of events, the odds ratio confidence interval of this sign was relatively wide, making this finding inconclusive. The 50 patients who progressed to very severe disease (grade 3 and 4) less frequently presented with rash and mouth lesions and if presenting with rash, showed a macular rash as the predominant type. This could be explained by different types of rash caused by different pathogens. EV-A71, detected in 70% of very severe cases (grade 3 and 4), was more likely to be associated with a macular rash (72% of 378 EV-A71 cases in this study; data not shown), while other non-EV-A71 enteroviruses, dominant among patients with milder disease, were more often associated with a vesicular rash (67% of 1169 non-EV-A71 cases; data not shown).

severe cases (grade 3 and 4), was more likely to be associated with a macular rash (72% of 378 EV-A71 cases in this study; data not shown), while other non-EV-A71 enteroviruses, dominant among patients with milder disease, were more often associated with a vesicular rash (67% of 1169 non-EV-A71 cases; data not shown). The study approach, including inpatients and outpatients attending all three referral hospitals for HFMD treatment in the southern part of Vietnam, was unique and comprehensive and allowed the epidemiological patterns of HFMD to be captured. However, this study is subject to the following limitations. First, in terms of the clinical aspects, building and validating a prognostic model for severe HFMD was unachievable due to the unexpectedly low numbers of severe cases admitted during the study period and the small numbers of events (grade progression during admission), and only factors potentially associated with progression of severity could be identified, without validation. These factors will need to be validated in another HFMD cohort. Second, the case distribution could have been biased by the study setting and study staff capacity, especially for outpatients, as parents of mild cases may prefer to attend private clinics or lower level hospitals, and even if they had attended the study hospitals, the study staff may not have been able to approach all of them (i.e., patients came to the hospitals outside of working hours, when the study room was not operational). Therefore, the case distribution among enrolled cases may not reflect the distribution seen in the community. Third, although it was sought to minimize confounding factors by setting a quorum of patients enrolled every week, this quorum could not always be reached, since the number of patients who came to the three study hospitals varied with time. Therefore, the time-series findings should be interpreted with consideration. Lastly, the list-wise deletion method might have produced biased parameters, and other methods to impute missing data should be considered to minimize possible bias.

e number of patients who came to the three study hospitals varied with time. Therefore, the time-series findings should be interpreted with consideration. Lastly, the list-wise deletion method might have produced biased parameters, and other methods to impute missing data should be considered to minimize possible bias. In conclusion, this study represents the most comprehensive descriptive HFMD study from Vietnam. The analysis of 1547 inpatients and outpatients revealed an overall dominance of EV-A71, especially among those with more a severe illness, but also among outpatients. Furthermore, it revealed detailed serotype emergence and replacement, including the emergence of CV-A6 and CV-A10 (in accordance with the global epidemiology), specific serotype-associated clinical features, and clinical and laboratory data associated with progression to more severe illness. These findings are essential for planning public health measures, particularly vaccine development and implementation. The clinical information could also be used as a reference for changing/modifying admission criteria to ensure that high-risk patients are closely monitored in the hospital and appropriate patients are hospitalized without placing a burden on the healthcare system through unnecessary admission. As it was not possible to build an effective model for progression, a further study with a larger sample size and more severe cases should be conducted to identify reliable risks factors for progression. Appendix A Supplementary data The following are Supplementary data to this article:

In conclusion, this study represents the most comprehensive descriptive HFMD study from Vietnam. The analysis of 1547 inpatients and outpatients revealed an overall dominance of EV-A71, especially among those with more a severe illness, but also among outpatients. Furthermore, it revealed detailed serotype emergence and replacement, including the emergence of CV-A6 and CV-A10 (in accordance with the global epidemiology), specific serotype-associated clinical features, and clinical and laboratory data associated with progression to more severe illness. These findings are essential for planning public health measures, particularly vaccine development and implementation. The clinical information could also be used as a reference for changing/modifying admission criteria to ensure that high-risk patients are closely monitored in the hospital and appropriate patients are hospitalized without placing a burden on the healthcare system through unnecessary admission. As it was not possible to build an effective model for progression, a further study with a larger sample size and more severe cases should be conducted to identify reliable risks factors for progression. Appendix A Supplementary data The following are Supplementary data to this article: Acknowledgements We would like to thank Ms Le Kim Thanh for her logistical support. We are indebted to the patients and their parents for their participation in this study, and all of the nursing and medical staff of the PICU, infectious diseases wards, and outpatient clinics at Children’s Hospital 1, Children’s Hospital 2, and the Hospital for Tropical Diseases who provided care for the patients and helped to collect the clinical data.

their parents for their participation in this study, and all of the nursing and medical staff of the PICU, infectious diseases wards, and outpatient clinics at Children’s Hospital 1, Children’s Hospital 2, and the Hospital for Tropical Diseases who provided care for the patients and helped to collect the clinical data. Funding source This work was supported by the Wellcome Trust, UK (101104/Z/13/Z, 106680/B/14/Z, and 204904/Z/16/Z). The funding body did not have any influence on the study design, study conduct, preparation of the manuscript, or decision to publish. Ethical approval The study was approved by the Oxford Tropical Research Ethics Committee (OxTREC) and by the institutional review boards of all enrolling hospitals. Families/guardians of children all gave informed consent to their child participating in the study. Conflict of interest All authors declare that they have no conflicts of interest. Appendix A Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.ijid.2018.12.004.

Background Pneumonia is the major cause of post-neonatal mortality in children under five years of age, contributing annually to over a million deaths, of which two thirds occur in low and middle income countries (LMIC) (Rudan et al., 2008). The World Health Organisation (WHO) uses clinical syndromal definitions according to severity. The WHO currently recommends antibiotic treatment for children aged 2–59 months with suspected lower respiratory tract infection to cover the possibility of bacterial infections (World Health Organization, 2014). Treatment allocation is made according to the severity of illness which is based on clinical criteria made by observation. Until 2014, classification was made into four categories: no pneumonia, mild (essentially fast breathing alone), severe (with chest indrawing with or without fast breathing) and very severe a definition requiring additional danger signs. The first two categories are felt appropriate for primary health care and home management with oral antibiotics: the third requires secondary centre referral, monitoring and parenteral antibiotic use.

e (with chest indrawing with or without fast breathing) and very severe a definition requiring additional danger signs. The first two categories are felt appropriate for primary health care and home management with oral antibiotics: the third requires secondary centre referral, monitoring and parenteral antibiotic use. The broad recommendation for children with ‘fast breathing pneumonia without danger signs’ is based on the assumption that a proportion of children in the most resource limited settings will not have the means to re-consult should the picture change. However, evidence for the guidance is weak and infections are often viral and self-limiting. This has generated substantial debate among experts (Hazir et al., 2011, Awasthi et al., 2008). There is equipoise regarding utility of antibiotics in fast breathing pneumonia and WHO has repeatedly identified a need for research for providing high quality evidence regarding appropriate management of community-acquired pneumonia (CAP). In 2014 a Cochrane review investigated the existing evidence comparing antibiotic to no antibiotic treatment for fast breathing pneumonia. The study found a lack of research in this area and concluded that “we do not currently have evidence to support or challenge the continued use of antibiotics for the treatment of non-severe (reclassified fast breathing) pneumonia, as suggested by WHO guidelines” (Lassi et al., 2014).

ment for fast breathing pneumonia. The study found a lack of research in this area and concluded that “we do not currently have evidence to support or challenge the continued use of antibiotics for the treatment of non-severe (reclassified fast breathing) pneumonia, as suggested by WHO guidelines” (Lassi et al., 2014). Moreover, increasing global coverage of effective vaccines (Pneumococcal and Haemophilus influenza type b) against the two major bacterial causes of childhood pneumonia in GAVI-eligible countries, including Pakistan. The epidemiology is changing and, though non-vaccine serotypes may become more prevalent, data to date suggest that these infections are likely to become less important contributors to pneumonia morbidity (Levine et al., 2006, Cowgill et al., 2006) and that the proportion of viral cases likely to increase. The changing epidemiology of the disease, therefore, requires a re-evaluation of practice related to use of antibiotics. This review presents the scientific rationale of performing non-inferiority studies in children with fast breathing pneumonia, comparing amoxicillin (control) to a placebo intervention. There is such a trial underway in Pakistan, the results of which should provide evidence to support or refute current WHO guidance (Jehan et al., 2016).

ts the scientific rationale of performing non-inferiority studies in children with fast breathing pneumonia, comparing amoxicillin (control) to a placebo intervention. There is such a trial underway in Pakistan, the results of which should provide evidence to support or refute current WHO guidance (Jehan et al., 2016). Main text Scientific rationale Withholding WHO recommended antibiotic treatment has a sound scientific rationale essentially because WHO-defined “fast breathing pneumonia” is a misclassification in the majority of cases (Izadnegahdar et al., 2013). Tabish et al in a study of 1848 children with fast breathing in Pakistan found that only 14% had radiological evidence of pneumonia, while the rest had either normal chest X-rays (82%) or bronchiolitis (4%) (Hazir et al., 2006). Previous studies have shown a high rate of resolution without treatment and there is evidence that amoxicillin has only partial efficacy in resolving this sign. In some settings, up to 65% of non-severe pneumonia is viral in aetiology with a bacterial viral co-infection in about 30% (Grant et al., 2009, Ruuskanen et al., 2011). Spontaneous remissions are frequent that may render antibiotics partly or completely ineffective. Current management guidelines prioritise sensitivity over specificity, resulting in widespread use of antibiotics when they are not needed (Izadnegahdar et al., 2013, Qazi and Were, 2015, English and Scott, 2008, Maitland, 2014).

eous remissions are frequent that may render antibiotics partly or completely ineffective. Current management guidelines prioritise sensitivity over specificity, resulting in widespread use of antibiotics when they are not needed (Izadnegahdar et al., 2013, Qazi and Were, 2015, English and Scott, 2008, Maitland, 2014). A fundamental principle of medical practice is to “do no harm.” By prescribing antibiotics to children that do not need them, there are potential risks and negative consequences at both the individual and population levels. Risks to children include an increased exposure to adverse events associated with antibiotics, which may be both unpleasant and dangerous. Moreover, early life exposure to antibiotics has shown to increase the risk of allergic disease in childhood (Kuo et al., 2013). There is also a potential long-term deleterious effect on the native gut microbiota which may be altered immune processing resulting in long-term risk of subsequent infections (Kristinsson, 1997, Uzuner et al., 2007, Woolfson et al., 1997, Murni et al., 2014, Rizal et al., 2010, Jonathan and Stoltenberg, 2012, Mauri and D’Agostino, 2017). At the population level, indiscriminate/injudicious use of antibiotics has increased risk of antimicrobial resistance (Kristinsson, 1997, Uzuner et al., 2007, Woolfson et al., 1997), resulting in the rise of antibiotic-resistant strains of bacteria and the need to use more expensive alternatives with greater risk of adverse events (Murni et al., 2014). Good antibiotic stewardship is increasingly important for amoxicillin to remain a long-term solution for treating childhood pneumonia worldwide (Rizal et al., 2010, Jonathan and Stoltenberg, 2012).

ains of bacteria and the need to use more expensive alternatives with greater risk of adverse events (Murni et al., 2014). Good antibiotic stewardship is increasingly important for amoxicillin to remain a long-term solution for treating childhood pneumonia worldwide (Rizal et al., 2010, Jonathan and Stoltenberg, 2012). Feasibility of non-inferiority placebo controlled design Testing a placebo intervention against an active control requires a non-inferiority trial, which works on the basis of the margin of failure being within a margin deemed a priori to be acceptable. Employing a non-inferiority trial is much more complex in the design, implementation and analysis (Mauri and D’Agostino, 2017). It is impossible to establish non-inferiority of no antibiotics to existing treatment without undertaking a robustly performed and adequately powered randomized controlled trial with low attrition and per-protocol analysis (Lewis et al., 2002). The most important aspect of such placebo-controlled trials is patient safety and it is fundamental to follow patients in the first 72–96 h after recruitment to guarantee safety. If this is made in a site with HDSS this might be facilitated though is not compulsory. Moreover, these trials must be designed in such a way that continued surveillance and easy re-access to health facilities is feasible and that rescue treatment introduction possible in case of deterioration and failure of expected resolution. (Lewis et al., 2002). Such trials should be blinded and randomized to reduce potential bias and enhance the quality and generalizability of study results, considered the “most important design techniques for avoiding bias in clinical trials” (International Conference on Harmonisation EEWG, 1999). In addition, these trials should be scrutinized for protocol deviations or violations and failures by external oversight by both a data safety monitoring board and trial steering committee. Furthermore, the participant exposure to placebo should be made for short duration and it is necessary to ensure careful and regular monitoring to detect early treatment failure signs through a robust safety net. Such active surveillance could result in relatively better standard of care in comparison to cases outside a trial setting, so the risk of harm may be further reduced (Lewis et al., 2002).

n and it is necessary to ensure careful and regular monitoring to detect early treatment failure signs through a robust safety net. Such active surveillance could result in relatively better standard of care in comparison to cases outside a trial setting, so the risk of harm may be further reduced (Lewis et al., 2002). Ethical issues Ethical analysis permits the use of placebo where the obligation is to determine the efficacy or safety of an intervention (in this case absence of treatment) provided there are sound methodological reasons and justification for using placebo and patients who receive placebo or no treatment will not be subject to any risk of serious harm (Millum and Grady, 2013), or subjects may benefit from being in the placebo group (van der Graaf, 2015). Placebo-controlled trials are justified when there is genuine equipoise and participants are not exposed to harm. There must be close clinical supervision, and a position of genuine informed consent (Lewis et al., 2002, van der Graaf, 2015). In these situations, international ethical standards in research allow for placebo to be used even if a known intervention exists. The Declaration of Helsinki discusses the use of placebo and notes that it may be used “where for compelling and scientifically sound methodological reasons the use of any intervention less effective than the best proven one, the use of placebo, or no intervention is necessary to determine the efficacy or safety of an intervention” (World Medical Association, 2013).

placebo and notes that it may be used “where for compelling and scientifically sound methodological reasons the use of any intervention less effective than the best proven one, the use of placebo, or no intervention is necessary to determine the efficacy or safety of an intervention” (World Medical Association, 2013). Most children require treatment with oral antibiotic solutions, which cost more and require refrigeration. This places a financial burden on families who bear these expenses out of their pocket and it also puts a strain on already under-resourced programmes in low-income settings. Dispersible Amoxicillin tablets are available through UNICEF, at a lower cost to the consumer, but availability is still non-uniform. Conclusion There are sound biological and societal reasons for revisiting the management of fast breathing pneumonia in children. Equipoise in treatment, low risk of harm and the potential benefits of rationing antibiotic use are strong justifications for a non-inferiority trial. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Availability of data and material Not applicable. Competing interests The authors declare that they have no competing interests.

Conclusion There are sound biological and societal reasons for revisiting the management of fast breathing pneumonia in children. Equipoise in treatment, low risk of harm and the potential benefits of rationing antibiotic use are strong justifications for a non-inferiority trial. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Availability of data and material Not applicable. Competing interests The authors declare that they have no competing interests. Funding This manuscript is based on RETAPP trial which is jointly funded by the MRC-Wellcome-DFID through the Joint Global Health Trials, [grant number MR/L004283/1] and the Bill & Melinda Gates Foundation [grant number OPP1158281]. Fyezah Jehan and Imran Nisar received funding from the National Institute of Health’s Fogarty International Centre [Grant Number 1 D43 TW007585-01]. Authors’ contributions FJ, AZ, IN conceived the idea. FJ, IN and SK were responsible for drafting the manuscript. FJ, SK, NB and GA revised the manuscript critically for important intellectual content. All authors reviewed and approved the final version. Acknowledgement None.

Introduction Hepatitis B virus (HBV) infection remains a serious health problem worldwide. Globally, HBV infection is one of the most dangerous infectious diseases, causing significant mortality (Mokaya et al., 2018). Infection by HBV results in inflammation of the liver. The infection can progress from acute hepatitis to chronic infection. The World Health Organization (WHO) estimated that in the year 2015, 257 million people worldwide were living with chronic HBV (CHB) infection, and 68% of those infected were living in the African and Western Pacific regions (WHO, 2017). This estimation showed that the second largest number of chronically infected HBV patients are living in the African continent, with 6.1% of the adult population infected (WHO, 2017). In the Central African Republic (CAR), a previous study reported a high prevalence of HBV in rural areas, with overall hepatitis B core antibody (HBc antibody) prevalence of 27.1% and hepatitis B surface antigen (HBsAg) prevalence of 10.6% (Komas et al., 2013). This finding highlights the possible impact of HBV on the economy of CAR, because these regions are the key farming regions.

high prevalence of HBV in rural areas, with overall hepatitis B core antibody (HBc antibody) prevalence of 27.1% and hepatitis B surface antigen (HBsAg) prevalence of 10.6% (Komas et al., 2013). This finding highlights the possible impact of HBV on the economy of CAR, because these regions are the key farming regions. The United Nations Sustainable Development Goals set out the challenge of the elimination of viral hepatitis as a public health threat by the year 2030 (WHO, 2016). The most effective measure to reduce the global incidence of hepatitis B remains vaccination. A mathematical model for the estimation of the global incidence of HBV showed that without vaccination, 64.8 million of the surviving birth cohort for the year 2000 would be infected by HBV and 1.4 million would die because of HBV complications (Goldstein et al., 2005). HBV is categorized into eight genotypes, A–H, based on divergence of more than 8% in the entire nucleotide sequence of the viral DNA; two additional genotypes I and J have also been proposed (Lin and Kao, 2011, Kramvis, 2014). There are also subgenotypes with 4–8% genomic variability. The HBV genotypes show a distinct geographical distribution worldwide, and genotypes A, D, and E are the most frequently found in Africa. In Western and Central Africa, genotype E is the dominant strain (Kramvis, 2014). A previous study in the CAR reported a prevalence of 94% genotype E, 4.5% genotype D, and 1.5% subgenotype A1 (Bekondi et al., 2007). In 2013, 100% of 19 isolated HBV were identified as genotype E (Komas et al., 2013).

ly found in Africa. In Western and Central Africa, genotype E is the dominant strain (Kramvis, 2014). A previous study in the CAR reported a prevalence of 94% genotype E, 4.5% genotype D, and 1.5% subgenotype A1 (Bekondi et al., 2007). In 2013, 100% of 19 isolated HBV were identified as genotype E (Komas et al., 2013). Although the genome of HBV is made up of DNA, the virus transitions to an unstable RNA state during replication, where its proof-reading deficient reverse transcriptase does not correct errors. This results in the accumulation of mutations. Mutations in the HBV genome of each genotype may affect prevention strategies, the diagnosis techniques, the response to treatment, and the course of the disease (Mokaya et al., 2018). Mutations in HBV that are associated with vaccine escape and detection escape have been documented in many countries worldwide; however, these remain poorly documented in some African countries (Mokaya et al., 2018). In the CAR, there is a lack of robust information in regards to HBV vaccine escape mutations and the accumulation of polymerase mutations. Previous studies in the country have reported the presence of recombinant genotype E/D (Bekondi et al., 2007) and a suspicion of vaccine escape mutants (Komas et al., 2013), without further investigation into their impact on the prevention and diagnosis measures in place.

s and the accumulation of polymerase mutations. Previous studies in the country have reported the presence of recombinant genotype E/D (Bekondi et al., 2007) and a suspicion of vaccine escape mutants (Komas et al., 2013), without further investigation into their impact on the prevention and diagnosis measures in place. Materials and methods Study site and sample collection This study was conducted at Institut Pasteur de Bangui in the CAR, a landlocked country in Central Africa. Serum samples were collected both retrospectively and prospectively between the years 2017 and 2019 in the Viral Hepatitis Laboratory, Institut Pasteur de Bangui. All patients included in this study had no previous history of any vaccination or treatment for HBV, as seen in their records at the institute. For the archived samples, the medical information of the patients was used to collect only confirmed HBsAg-positive samples. Fifty-three samples were collected from the archive (years 2017–2018). Older samples had been discarded from the institute to allow a new archiving process. The remaining 75 samples were collected prospectively from January to February 2019. All of the patients who were positive for HBsAg on screening and who were also confirmed to be HBsAg-positive had samples collected immediately and stored at −20 °C. The total number of samples collected was 118. All of the patients included in this study came to the institute for HBV screening and were enrolled after being consented.

nts who were positive for HBsAg on screening and who were also confirmed to be HBsAg-positive had samples collected immediately and stored at −20 °C. The total number of samples collected was 118. All of the patients included in this study came to the institute for HBV screening and were enrolled after being consented. HBV serological assays An ELISA (Abbott-Murex kit, United Kingdom) for the detection of HBsAg marker was performed for all samples, in accordance with the manufacturer’s instructions. HBsAg-positive samples were confirmed by Murex HBsAg confirmatory test version 3 and tested for hepatitis B e antigen (HBeAg) using the Creative Diagnostic Human HBeAg ELISA Kit (catalog number DEIA003). Confirmed HBsAg-positive samples were screened for hepatitis delta virus (HDV) using a DiaSorin kit (DiaSorin Italy). All results were interpreted in accordance with the manufacturers’ instructions. Viral DNA extraction Viral hepatitis DNA was extracted from 200 μl of serum using the QIAamp DNA Mini Kit (catalog number 51306; Qiagen, Germany) according to the manufacturer’s instructions.

HBV serological assays An ELISA (Abbott-Murex kit, United Kingdom) for the detection of HBsAg marker was performed for all samples, in accordance with the manufacturer’s instructions. HBsAg-positive samples were confirmed by Murex HBsAg confirmatory test version 3 and tested for hepatitis B e antigen (HBeAg) using the Creative Diagnostic Human HBeAg ELISA Kit (catalog number DEIA003). Confirmed HBsAg-positive samples were screened for hepatitis delta virus (HDV) using a DiaSorin kit (DiaSorin Italy). All results were interpreted in accordance with the manufacturers’ instructions. Viral DNA extraction Viral hepatitis DNA was extracted from 200 μl of serum using the QIAamp DNA Mini Kit (catalog number 51306; Qiagen, Germany) according to the manufacturer’s instructions. Amplification of the surface and polymerase genes The primer pair Pol forward and reverse was used, as described previously (Sayan et al., 2010), to amplify a 742-bp DNA fragment spanning the surface and polymerase gene regions (Table 1). In brief, the PCR mix was carried in a total volume of 50 μl with a final concentration of 0.1 pmol of each primer, 2.5 U of Taq DNA polymerase, 250 μM of each dNTP, 10 mM of Tris–HCl (pH 9.0), 30 mM of KCl, 1.5 mM of MgCl2, stabilizer and tracking dye, and 0.8–2 ng/μl of DNA template. A ProFlex PCR Thermal Cycler (Applied Biosystems) was used for thermal cycling, as follows: 95 °C for 5 min, and then 45 cycles consisting of 95 °C for 45 s, 56 °C for 45 s, and 72 °C for 45 s. A final elongation was set at 72 °C for 10 min.Table 1 List of primers used in the study for PCR amplification and sequencing.

A ProFlex PCR Thermal Cycler (Applied Biosystems) was used for thermal cycling, as follows: 95 °C for 5 min, and then 45 cycles consisting of 95 °C for 45 s, 56 °C for 45 s, and 72 °C for 45 s. A final elongation was set at 72 °C for 10 min.Table 1 List of primers used in the study for PCR amplification and sequencing. Table 1Primer Sequence Used Region Product size (bp) Reference Pol F 5′ TCGTGGTGGACTTCTCTCAATT 3′ PCR and sequencing Polymerase 740 Sayan et al. (2010) Pol R 5′ CGTTGACAGACTTTCCAATCAAT 3′ PCR and sequencing Polymerase 740 Sayan et al. (2010) WA-L 5′ ACTGTTCAAGCCTCCAAGCTGTGC 3′ PCR Whole genome 3200 Zhang et al. (2007) WA-R 5′ AGCAAAAAGTTGCATGGTGCTGGT 3′ PCR Whole genome 3200 Zhang et al. (2007) A3-L 5′ CTGCTGGTGGCTCCAGTT 3′ Sequencing Polymerase 1059 Zhang et al. (2007) A3-R 5′ GCCTTGTAAGTTGGCGAGAA 3′ Sequencing Polymerase 1059 Zhang et al. (2007) A4-L 5′ GTATTGGGGGCCAAGTCTGT 3′ Sequencing Polymerase 1072 Zhang et al. (2007) A4-R 5′ AAAAAGTTGCATGGTGCTG 3′ Sequencing Polymerase 1072 Zhang et al. (2007)

(2007) A3-L 5′ CTGCTGGTGGCTCCAGTT 3′ Sequencing Polymerase 1059 Zhang et al. (2007) A3-R 5′ GCCTTGTAAGTTGGCGAGAA 3′ Sequencing Polymerase 1059 Zhang et al. (2007) A4-L 5′ GTATTGGGGGCCAAGTCTGT 3′ Sequencing Polymerase 1072 Zhang et al. (2007) A4-R 5′ AAAAAGTTGCATGGTGCTG 3′ Sequencing Polymerase 1072 Zhang et al. (2007) Amplification of the HBV whole genome In an additional investigation, the complete genome (3.2 kb) of one sample, for which the region spanning the surface and polymerase gene was successfully sequenced in the present study, was amplified with the primers WA-L and WA-R, as described previously (Zhang et al., 2007). In brief, the PCR mix was prepared as described for the surface and polymerase gene PCR, using the primer pair WA-L and WA-R (Table 1). The ProFlex PCR Thermal Cycler (Applied Biosystems) was used for thermal cycling, as follows: 95 °C for 5 min, and then 30 cycles consisting of 95 °C for 30 s, 58 °C for 1 min, and 72 °C for 3 min 30 s. A final elongation was set at 72 °C for 10 min. Purification and sequencing of PCR products All PCR products were resolved on 1% agarose gels stained with GelRed and viewed using a UV transilluminator. Sanger sequencing was performed by Macrogen (Netherlands). The fragment spanning the surface and polymerase gene regions was sequenced using the same PCR primer pair, and the PCR product of the whole genome was sequenced with a couple of overlapping primers A3-L/A3-R and A4-L/A4-R, as described previously (Zhang et al., 2007), to generate the sequence of the entire reverse transcriptase of HBV (Table 1).

rface and polymerase gene regions was sequenced using the same PCR primer pair, and the PCR product of the whole genome was sequenced with a couple of overlapping primers A3-L/A3-R and A4-L/A4-R, as described previously (Zhang et al., 2007), to generate the sequence of the entire reverse transcriptase of HBV (Table 1). Sequence cleaning up and assembly All of the sequences were analyzed and assembled using CLC Genomic Workbench 8.0.3 (https://www.qiagenbioinformatics.com/blog/discovery/publications-citing-clc-genomics-workbench/) and then subjected to NCBI nucleotide BLAST for quality check. Sequence genotyping/subgenotyping and serotyping A phylogenetic analysis was done using MEGA version 10.0.5. (Kumar et al., 2018). The analysis was done using the neighbor-joining statistical method, the Kimura-2 parameter model, and the bootstrap method of 1000 replicates. Genotypes and subgenotypes were confirmed with geno2pheno HBV (https://hbv.geno2pheno.org/index.php) and the genotyping tool of NCBI (https://www.ncbi.nlm.nih.gov/projects/genotyping/formpage.cgi). Serotyping was done based on amino acids at positions s122, s127, s140, s159, and s160 (Swenson et al., 1991, Bell and Kramvis, 2015) by aligning the surface antigen amino acid sequence of 51 isolates against the reference sequences (gnl|hbvcds|AB014370 genotype A and gnl|hbvcds|AB091255 genotype E) in BioEdit.

Sequence genotyping/subgenotyping and serotyping A phylogenetic analysis was done using MEGA version 10.0.5. (Kumar et al., 2018). The analysis was done using the neighbor-joining statistical method, the Kimura-2 parameter model, and the bootstrap method of 1000 replicates. Genotypes and subgenotypes were confirmed with geno2pheno HBV (https://hbv.geno2pheno.org/index.php) and the genotyping tool of NCBI (https://www.ncbi.nlm.nih.gov/projects/genotyping/formpage.cgi). Serotyping was done based on amino acids at positions s122, s127, s140, s159, and s160 (Swenson et al., 1991, Bell and Kramvis, 2015) by aligning the surface antigen amino acid sequence of 51 isolates against the reference sequences (gnl|hbvcds|AB014370 genotype A and gnl|hbvcds|AB091255 genotype E) in BioEdit. Analysis of mutations in HBsAg and polymerase The overlapping surface (S) and polymerase gene sequences obtained were translated to the protein sequences and aligned with the reference sequence AB014370 for genotype A and AB091255 for genotype E in BioEdit, for the analysis of mutations. Subsequently, sequences were submitted to geno2pheno HBV for mutations analysis confirmation. Amino acid exchanges in each sequence were recorded and searched in the scientific literature. Nucleotide sequence accession numbers The sequences of the regions spanning the surface and polymerase gene reported in this study have been submitted to GenBank and assigned the accession numbers MN047437 and MN420406 to MN420455.

Analysis of mutations in HBsAg and polymerase The overlapping surface (S) and polymerase gene sequences obtained were translated to the protein sequences and aligned with the reference sequence AB014370 for genotype A and AB091255 for genotype E in BioEdit, for the analysis of mutations. Subsequently, sequences were submitted to geno2pheno HBV for mutations analysis confirmation. Amino acid exchanges in each sequence were recorded and searched in the scientific literature. Nucleotide sequence accession numbers The sequences of the regions spanning the surface and polymerase gene reported in this study have been submitted to GenBank and assigned the accession numbers MN047437 and MN420406 to MN420455. Accession numbers of HBV sequences retrieved from the HBV database are mentioned in each analysis and those used for phylogenetic analysis are shown in the phylograms. Ethical approval The Scientific Committee of the Faculty of Medicine, University of Bangui, CAR, approved the study protocol (Ref. No. 21/UB/FACSS/CSCVPER/19).

Accession numbers of HBV sequences retrieved from the HBV database are mentioned in each analysis and those used for phylogenetic analysis are shown in the phylograms. Ethical approval The Scientific Committee of the Faculty of Medicine, University of Bangui, CAR, approved the study protocol (Ref. No. 21/UB/FACSS/CSCVPER/19). Results General patient data All 118 samples included in the study were confirmed to be HBsAg-positive by ELISA. Patients were between 14 and 64 years old, with approximately 68 of the 118 patients aged between 25 and 60 years (Table 2). By year, 64% of samples were collected in the year 2019, 28% in 2018, and 8% in 2017. There were more male patients (n = 84, 71%) than female patients (n = 34, 29%); the male to female sex ratio was 2.47. The serological tests showed that 23/118 (19%) patients were co-infected with HDV and 16/118 (14%) patients were HBeAg-positive.Table 2 Age and sex distribution of the HBsAg-positive study population and distribution of patients positive for HBeAg and HDV; Institut Pasteur de Bangui 2017–2019. Table 2Ages groups (years) Male Female Total HBsAg frequency (%) HBeAg-positive HDV-positive HBsAg frequency (%) HBeAg-positive HDV-positive HBsAg frequency >18 7 (5.93) 2 1 0 (0.0) 0 0 7 18–24 4 (3.39) 1 1 4 (3.39) 1 1 8 25–40 49 (41.52) 7 8 19 (16.1) 2 3 68 41–60 22 (18.64) 1 4 11 (9.32) 1 5 33 <60 2 (1.69) 0 0 0 (0.0) 0 0 2 Total 84 10 14 34 4 9 118 HBsAg, hepatitis B surface antigen; HBeAg, hepatitis B e antigen; HDV, hepatitis delta virus.

HDV-positive HBsAg frequency >18 7 (5.93) 2 1 0 (0.0) 0 0 7 18–24 4 (3.39) 1 1 4 (3.39) 1 1 8 25–40 49 (41.52) 7 8 19 (16.1) 2 3 68 41–60 22 (18.64) 1 4 11 (9.32) 1 5 33 <60 2 (1.69) 0 0 0 (0.0) 0 0 2 Total 84 10 14 34 4 9 118 HBsAg, hepatitis B surface antigen; HBeAg, hepatitis B e antigen; HDV, hepatitis delta virus. PCR and sequencing All 118 HBsAg-positive samples were subjected to viral DNA extraction. Subsequently, surface and polymerase genes were amplified by PCR and sequenced. Only 57/118 (48.30%) samples were PCR-positive and 51/57 PCR-positive samples were successfully sequenced; six of the PCR-positive samples gave sequences that were not exploitable. The 51 patients whose samples were successfully sequenced were aged between 17 and 64 years; 69.05% were male and 30.95% were female, giving a male to female sex ratio of 2.4. Of these samples, 27.45% were positive for HBeAg and 13.72% were positive for co-infection with HDV. None of these patients had a history of treatment for HBV or history of HBV vaccination. Basic data of these 51 patients are given in Table 3.Table 3 Basic data of the HBV patients whose samples were successfully sequenced (n = 51). Table 3Characteristics Values Sex (male:female; sex ratio) 29:13; 2.4 Age (years), mean (range) 37 (17–64) HBeAg-positive, n (%) 14 (27.45) HDV, n (%) 7 (13.72) Vaccination No record Treatment No record HBV, hepatitis B virus; HBeAg, hepatitis B e antigen; HDV, hepatitis delta virus.

The 51 patients whose samples were successfully sequenced were aged between 17 and 64 years; 69.05% were male and 30.95% were female, giving a male to female sex ratio of 2.4. Of these samples, 27.45% were positive for HBeAg and 13.72% were positive for co-infection with HDV. None of these patients had a history of treatment for HBV or history of HBV vaccination. Basic data of these 51 patients are given in Table 3.Table 3 Basic data of the HBV patients whose samples were successfully sequenced (n = 51). Table 3Characteristics Values Sex (male:female; sex ratio) 29:13; 2.4 Age (years), mean (range) 37 (17–64) HBeAg-positive, n (%) 14 (27.45) HDV, n (%) 7 (13.72) Vaccination No record Treatment No record HBV, hepatitis B virus; HBeAg, hepatitis B e antigen; HDV, hepatitis delta virus. HBV genotypes/subgenotypes and serotypes All 51 contig sequences of the isolated HBV showed significant similarity to the HBV sequences after being subjected to Nucleotide BLAST search on the NCBI BLAST webpage. The phylogram (Figure 1) of the 51 HBV isolates as well as sequences of the eight HBV genotypes A–H retrieved from the HBV database (http://hbvdb.ibcp.fr/HBVdb/) revealed that 49/51 (96.08%) belonged to genotype E and 2/51 (3.92%) belonged to genotype A, more specifically subgenotype A1, after being subjected to geno2pheno HBV and the genotyping tool of NCBI.Figure 1 Phylogenetic tree of the partial polymerase (742 bp) of the HBV isolates. The tree was constructed in MEGA using the neighbor-joining statistical method, the Kimura-2 parameter model, and the bootstrap method of 1000 replicates. The black squares represent the reference sequences and the black circles represent the isolated HBV sequences discussed in the present study.

) of the HBV isolates. The tree was constructed in MEGA using the neighbor-joining statistical method, the Kimura-2 parameter model, and the bootstrap method of 1000 replicates. The black squares represent the reference sequences and the black circles represent the isolated HBV sequences discussed in the present study. Figure 1 Using amino acid residues at positions s122, s127, s140, s159, and s160, the protein sequences of the 51 isolated HBV alongside the reference sequences of the HBV surface gene (gnl|hbvcds|AB014370 genotype A and gnl|hbvcds|AB091255 genotype E) showed that all 49 isolates previously genotyped as E were serotyped ayw4 and the two genotype A subgenotype A1 isolates were serotyped ayw1. Mutations identified in surface antigen Alignment of the surface protein of the 51 isolated HBV alongside reference genotype A and E (Figure 2) and subsequent sequence submission to geno2pheno HBV on the web for analysis confirmation, led to the detection of 43 amino acid exchanges (Supplementary material). Six of these (sL127I, sA128V, sG130S, sM133T, sF134I, and S140T) were located in the ‘a’ determinant region, among which three immune escape mutants (IEMs) (sY100C, sA128V, and sM133T) were identified.Figure 2 Mutation detected in the ‘a’ determinant. Genotype A isolates alongside a reference genotype A (GenBank reference AB014370) and genotype E isolates alongside a reference genotype E (GenBank reference AB091255). The box highlights the ‘a’ determinant region. Figure 2

Mutations identified in surface antigen Alignment of the surface protein of the 51 isolated HBV alongside reference genotype A and E (Figure 2) and subsequent sequence submission to geno2pheno HBV on the web for analysis confirmation, led to the detection of 43 amino acid exchanges (Supplementary material). Six of these (sL127I, sA128V, sG130S, sM133T, sF134I, and S140T) were located in the ‘a’ determinant region, among which three immune escape mutants (IEMs) (sY100C, sA128V, and sM133T) were identified.Figure 2 Mutation detected in the ‘a’ determinant. Genotype A isolates alongside a reference genotype A (GenBank reference AB014370) and genotype E isolates alongside a reference genotype E (GenBank reference AB091255). The box highlights the ‘a’ determinant region. Figure 2 Mutations identified in the polymerase gene Alignment of the protein sequences of the partial polymerase gene of the 51 isolated HBV alongside reference genotype A and E, followed by sequence submission to geno2pheno HBV on the web for analysis confirmation, led to the detection of 43 amino acid exchanges (Supplementary material). These included the previously reported mutations rtL91I, rtM129L, rtW153R, rtS213T, rtV214A, rtN238T, rtN248H, and rtI269L (Yamani et al., 2017, Choi et al., 2018, Shaha et al., 2018, Ababneh et al., 2019). Of the 49 genotype E identified, 29 had amino acid exchanges in the polymerase gene. The two genotype A studied harbored 13 and 14 amino acid exchanges, respectively.

1I, rtM129L, rtW153R, rtS213T, rtV214A, rtN238T, rtN248H, and rtI269L (Yamani et al., 2017, Choi et al., 2018, Shaha et al., 2018, Ababneh et al., 2019). Of the 49 genotype E identified, 29 had amino acid exchanges in the polymerase gene. The two genotype A studied harbored 13 and 14 amino acid exchanges, respectively. Discussion HBV IEMs This study is novel in investigating HBV escape mutations in the CAR. The study did not cover the entire CAR or Bangui populations, but was focused on individuals who were diagnosed at Institut Pasteur de Bangui from 2017 to 2019. This study detected three IEMs, sY100C, sA128V, and sM133T, similar to previous studies (Mello et al., 2011, Kwei et al., 2013, Cremer et al., 2018). Although data on IEMs in CAR are limited, the hypothesis of escape mutation among circulating HBV strains in the CAR was referred to previously in a study performed in 2013 (Komas et al., 2013). In that previous study, some individuals screened negative for HBsAg but positive for antibodies against HBc, with viral DNA detected in five of these individuals. The present study confirms this hypothesis. Therefore, there is the need for further investigations on HBV IEMs in the CAR.

ed in 2013 (Komas et al., 2013). In that previous study, some individuals screened negative for HBsAg but positive for antibodies against HBc, with viral DNA detected in five of these individuals. The present study confirms this hypothesis. Therefore, there is the need for further investigations on HBV IEMs in the CAR. sG145R is the most identified of the HBV IEMs and has been well-characterized, but was not identified in this study. The three IEMs identified in the present study (sY100C, sA128V, and sM133T) were confirmed by geno2pheno HBV. Mutation sY100C was previously assessed by Mello et al. (Mello et al., 2011), who showed a reduction in HBsAg detection level by commercial ELISA kit, but without statistical significance, leading the authors to conclude that mutation sY100C alone did not play a role in reducing the HBsAg affinity of this commercial ELISA assay. As suggested by Mello et al. (Mello et al., 2011), other mutations like K122R and F134I detected in the ‘a’ determinant support the idea of IEMs associated with strains harboring the sY100C as discussed in the present study. The mutation sA128 V was confirmed by geno2pheno HBV, and this IEM was associated with vaccine escape in a recent study conducted by Cremer et al. (Cremer et al., 2018), who linked it with HBsAg test failure. Many studies (Protzer-Knolle et al., 1998, Beckebaum et al., 2003, Cheung et al., 2010) have demonstrated that M133T itself is frequently associated with occult HBV infection and is also often associated with some mutations in the ‘a’ determinant such as G130N, F134L, D144A, D144G, G145A, G145K, and G145R or failed hepatitis B immune globulin (HBIG) prophylaxis. It has also been reported that the naturally occurring mutation sM133T can create a novel N-linked glycosylation site in the viral envelope proteins (Ito et al., 2010). The findings of the present study indicate the need for a large scale study that also includes HBsAg-negative patients in order to investigate occult HBV infections in the CAR population, as has been suggested previously (Komas et al., 2013).

-linked glycosylation site in the viral envelope proteins (Ito et al., 2010). The findings of the present study indicate the need for a large scale study that also includes HBsAg-negative patients in order to investigate occult HBV infections in the CAR population, as has been suggested previously (Komas et al., 2013). In addition to the three IEMs identified, several surface protein mutations were identified (sP56Q, sT57I, sN59S, sP62L, sI110L, sS140T, sE164G, sS207N, and sL216*), which have also been reported in previous studies (Mello et al., 2011, Lin et al., 2013, Munshi et al., 2017, Qin and Liao, 2018). The most frequently occurring surface protein mutation was sN59S, harbored by 46/51 isolates. This was previously reported to affect T helper/cytotoxic T lymphocyte (TH/CTL) inter-epitopes (Lin et al., 2013, Qin and Liao, 2018). Mutations sP56Q and sP62L, harbored by two HBeAg-negative patients, have previously been reported to be associated with an increased risk of hepatocellular carcinoma (HCC) (Lin et al., 2013, Munshi et al., 2017, Qin and Liao, 2018); the serological analysis identified 102/118 patients in this study to be HBeAg-negative. Indeed, the loss of HBeAg has been reported to be a marker associated with the end of active viral replication and improvement in HBV disease (Lazarevic, 2014). Currently, the increasing mutation in the HBV genome is making serologically negative HBeAg results uncertain in some African countries (Olinger et al., 2006, Belyhun et al., 2018, Mokaya et al., 2018). HBeAg, hepatitis B core antigen (HBcAg), and the core-related protein p22 are all proteins produced by the pre-core/core HBV gene and they share the same 149 amino acid sequence; therefore, quantification of the HBV core-related antigen (HBcrAg) has been shown to be more accurate (Hadziyannis and Laras, 2018) than HBeAg serology alone. Co-infection with HDV was found in 23 of the 118 patients, a prevalence similar to that reported previously in CAR (Komas et al., 2018).

149 amino acid sequence; therefore, quantification of the HBV core-related antigen (HBcrAg) has been shown to be more accurate (Hadziyannis and Laras, 2018) than HBeAg serology alone. Co-infection with HDV was found in 23 of the 118 patients, a prevalence similar to that reported previously in CAR (Komas et al., 2018). HBV polymerase mutants This study identified several known polymerase mutations (rtL91I, rtM129L, rtW153R, rtS213T, rtV214A, rtN238T, rtN248H, and rtI269L), recorded in many previous studies (Locarnini, 2008, Abdelnabi et al., 2014, Gomes-Gouvêa et al., 2015, Zhang and Ding, 2015, Kim et al., 2017, Yamani et al., 2017, Choi et al., 2018, Shaha et al., 2018, Ababneh et al., 2019). The most frequently identified polymerase mutation was rtL91I, harbored by six isolates. The study did not detect any drug resistance mutations among the polymerase mutants recorded. Since the records of the HBsAg-positive patients at Institut Pasteur de Bangui indicated that they had not received any nucleoside/nucleotide analog (NA) treatment, the probability of detecting HBV drug resistance was expected to be very low: as discussed before in California, Brazil, and elsewhere, HBV drug resistance is very rare in treatment-naïve patients (Nguyen et al., 2009, Abdelnabi et al., 2014, Gomes-Gouvêa et al., 2015, Rugieri Pacheco et al., 2017).

log (NA) treatment, the probability of detecting HBV drug resistance was expected to be very low: as discussed before in California, Brazil, and elsewhere, HBV drug resistance is very rare in treatment-naïve patients (Nguyen et al., 2009, Abdelnabi et al., 2014, Gomes-Gouvêa et al., 2015, Rugieri Pacheco et al., 2017). It is important to note that there is currently no therapeutic support of chronically infected HBV patients in the CAR. Most of the patients used traditional medicine to overcome this issue. HBV patients received treatment in the case of HIV/HBV co-infection, and a very limited number of patients who had sufficient finance to support their treatment also received HBV treatment. This is one of the challenges for investigating HBV drug resistance properly, unless HIV/HBV co-infected patients are included. The inclusion of HIV/HBV co-infected patients is thus recommended for further investigations into HBV drug resistance in the CAR.

inance to support their treatment also received HBV treatment. This is one of the challenges for investigating HBV drug resistance properly, unless HIV/HBV co-infected patients are included. The inclusion of HIV/HBV co-infected patients is thus recommended for further investigations into HBV drug resistance in the CAR. HBV genotypes and serotypes Although the present study identified two HBV genotypes (A and E), HBV genotype E was the most identified genotype and this remains the dominant strain circulating in the CAR. This is in agreement with previous studies showing genotype E to be dominant in Western and Central Africa (Bekondi et al., 2007, Komas et al., 2013, Kramvis, 2014). In fact, HBV genotype E has low diversity and emerged more recently, within the last 200 years (Kramvis, 2014). The two subgenotype A1 isolates (Pol33 and Pol40) harbored 14 and 13 amino acid exchanges, respectively, in the polymerase gene, among which nine (rtN122H, rtN124H, rtM129L, rtS137T, rtW153R, rtV163I, rtS256C, rtT259S, and rtI269L) were not present in genotype E (of the 49 isolated genotype E, 22 did not have any amino acid exchanges in the polymerase gene), showing that genotype A circulating in CAR is likely to undergo mutations that will affect the diagnosis, as well as the prevention and control measures for HBV in the CAR population.

tI269L) were not present in genotype E (of the 49 isolated genotype E, 22 did not have any amino acid exchanges in the polymerase gene), showing that genotype A circulating in CAR is likely to undergo mutations that will affect the diagnosis, as well as the prevention and control measures for HBV in the CAR population. In addition to genotype classification, this study serotyped the isolated strains. Two serotypes were identified: ayw4 accounted for all genotype E and ayw1 accounted for all subgenotype A1. This result is not surprising, as the HBV serotype is strongly associated with the genotype (Kramvis et al., 2008). However, this information is still useful, since data on isolated HBV from CAR are limited. Conclusions Genotype E, presenting less genetic variability, remains the dominant strain circulating in the CAR. This study identified potential IEMs and several polymerase mutants among the limited population of patients attending Institut Pasteur de Bangui, confirming the hypothesis of circulating HBV IEM strains in the CAR population. The results of this study raise awareness of the need for further studies to be conducted on a large scale, including patients under treatment, in order to better understand IEMs of HBV for improved disease prevention and control strategies in the CAR.

he hypothesis of circulating HBV IEM strains in the CAR population. The results of this study raise awareness of the need for further studies to be conducted on a large scale, including patients under treatment, in order to better understand IEMs of HBV for improved disease prevention and control strategies in the CAR. Funding and support This research was supported by the Pan African University of the African Union Commission for Science and Technology. Partial funding was also received from the Japan International Corporation Agency (JICA) through the AFRICA-ai-JAPAN PROJECT and from the BecA-ILRI Hub program and ILRI through the Africa Biosciences Challenge Fund (ABCF) program. The ABCF program is funded by the Australian Department for Foreign Affairs and Trade (DFAT) through the BecA-CSIRO partnership and by the Syngenta Foundation for Sustainable Agriculture (SFSA), the Bill & Melinda Gates Foundation (BMGF), the UK Department for International Development (DFID), and the Swedish International Development Cooperation Agency (Sida).

n Department for Foreign Affairs and Trade (DFAT) through the BecA-CSIRO partnership and by the Syngenta Foundation for Sustainable Agriculture (SFSA), the Bill & Melinda Gates Foundation (BMGF), the UK Department for International Development (DFID), and the Swedish International Development Cooperation Agency (Sida). Author contributions Conceptualization: Giscard Wilfried Koyaweda and revised by Juliette Rose Ongus, Rosaline Macharia, Narcisse Patrice Komas, and Roger Pelle. Data collection: Giscard Wilfried Koyaweda, guided by Narcisse Patrice Komas. Investigation: Giscard Wilfried Koyaweda and Eunice Machuka. Supervision: Rosaline Macharia, Juliette Rose Ongus, Narcisse Patrice Komas, and Roger Pelle. Data analyses: Giscard Wilfried Koyaweda, Rosaline Macharia, John Juma, Eunice Machuka, and Roger Pelle. Writing – original draft: Giscard Wilfried Koyaweda and Juliette Rose Ongus. Writing – review and editing: Giscard Wilfried Koyaweda, Juliette Rose Ongus, Rosaline Macharia, Eunice Machuka, Narcisse Patrice Komas, and Roger Pelle. Conflict of interest The authors declare that they have no conflict of interest. Appendix A Supplementary data The following is Supplementary data to this article: Acknowledgements This work forms part of the Master’s Thesis of Giscard Wilfried Koyaweda and was performed in part at the Viral Hepatitis Laboratory at Institut Pasteur de Bangui and the BecA ILRI Laboratory. We thank the team of the Serology Laboratory of Institut Pasteur de Bangui for sample screening for HBV and HDV markers.

ements This work forms part of the Master’s Thesis of Giscard Wilfried Koyaweda and was performed in part at the Viral Hepatitis Laboratory at Institut Pasteur de Bangui and the BecA ILRI Laboratory. We thank the team of the Serology Laboratory of Institut Pasteur de Bangui for sample screening for HBV and HDV markers. Appendix A Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.ijid.2019.10.039.

when compared to their non-infected counterparts.5 An increase in the rates of THA and TKA revision, with infection playing a role in up to 15% of cases, is occurring.6,7 While the overall infection rates among primary arthroplasty are less than 2%, rates have increased significantly for revision arthroplasty surgery.7 Factors that are associated with PJI include rheumatoid arthritis (RA), underlying malignancy, use of corticosteroids, and increased body mass index (BMI); the highest rates of infection occur in patients with these risk factors undergoing THA.4 Staphylococcus aureus and coagulase-negative staphylococci (CoNS) are the most common causes of PJI. Currently, Gram-negative bacteria are responsible for a substantial proportion of PJI, ranging from 5% to 23% of cases, especially among the elderly. Both Gram-negative and Gram-positive bacteria have been associated with device-related biofilms, which protect the organisms from many antimicrobial agents and the host immune system.8 However, the clinical outcomes of PJI caused by Gram-negative bacteria are reportedly less favorable than those of infection caused by Gram-positive bacteria.9–11 The emergence of resistance to antibiotics among Gram-negative bacteria that cause PJIs is also a major concern. The emergence of resistance to fluoroquinolones is linked to failure of open debridement and loss of the prosthesis.12

reportedly less favorable than those of infection caused by Gram-positive bacteria.9–11 The emergence of resistance to antibiotics among Gram-negative bacteria that cause PJIs is also a major concern. The emergence of resistance to fluoroquinolones is linked to failure of open debridement and loss of the prosthesis.12 In the past two decades, Klebsiella pneumoniae has emerged as a multi- and extremely-drug resistant Gram-negative pathogen.13 Strains of K. pneumoniae have acquired plasmids with myriad mechanisms of antibiotic resistance, such as qnr against fluoroquinolones, 16S rRNA methylases against aminoglycosides, and, against cephalosporins and carbapenems, extended-spectrum beta-lactamases (ESBLs), New Delhi metallo-beta-lactamase (NDM), and K. pneumoniae carbapenemase (KPC).14–17 Because of the paucity of antibiotic options to treat them, infections caused by carbapenem-resistant K. pneumoniae (CRKP) pose a significant threat to our health care system. Vulnerable patients in acute and long-term care facilities experience bloodstream, respiratory, and urinary tract infections that often lead to unwanted outcomes.18–20 CRKP-related PJIs may be particularly complicated by the development of biofilms. Although CRKP biofilms have not been documented in PJIs, they have been associated with endoscopes.21 Thus, the combination of plasmid-acquired and biofilm-associated microbial resistance may explain the severe outcomes described here.

20 CRKP-related PJIs may be particularly complicated by the development of biofilms. Although CRKP biofilms have not been documented in PJIs, they have been associated with endoscopes.21 Thus, the combination of plasmid-acquired and biofilm-associated microbial resistance may explain the severe outcomes described here. In this report, we recount our experience with three cases of CRKP-related PJI. This single institution case series illustrates the unique management challenges faced by clinicians and the adverse clinical outcomes experienced by patients in an era of potentially ‘untreatable infections’.22

1. Introduction Primary total hip arthroplasty (THA) and total knee arthroplasty (TKA) are among the most common operations in orthopedic surgery, with nearly 800 000 THAs/TKAs performed annually in the USA.1 Demand for THA and TKA is projected to significantly increase in the next two decades.2–4 One of the most devastating complications of THA and TKA is infection of the prosthesis. Patients with prosthetic joint infection (PJI) demonstrate a greater morbidity, prolonged hospital stay, and incur additional costs of care when compared to their non-infected counterparts.5 An increase in the rates of THA and TKA revision, with infection playing a role in up to 15% of cases, is occurring.6,7 While the overall infection rates among primary arthroplasty are less than 2%, rates have increased significantly for revision arthroplasty surgery.7

20 CRKP-related PJIs may be particularly complicated by the development of biofilms. Although CRKP biofilms have not been documented in PJIs, they have been associated with endoscopes.21 Thus, the combination of plasmid-acquired and biofilm-associated microbial resistance may explain the severe outcomes described here. In this report, we recount our experience with three cases of CRKP-related PJI. This single institution case series illustrates the unique management challenges faced by clinicians and the adverse clinical outcomes experienced by patients in an era of potentially ‘untreatable infections’.22 2. Materials and methods We conducted a retrospective study of all patients at a tertiary care institution (Cleveland Clinic Foundation, Cleveland, Ohio, USA) with CRKP isolated from cultures of clinical samples between January 2007 and December 2010. CRKP was defined as K. pneumoniae isolates having a minimum inhibitory concentration (MIC) ≥2 μg/ml against ertapenem, meropenem, or imipenem and a positive modified Hodge test (Clinical and Laboratory Standards Institute (CLSI) 2009). CRKP-related PJI were diagnosed if CRKP was recovered from intraoperative prosthetic joint and tissue specimens, synovial fluid culture, and/or from a sinus tract communicating with the prosthesis. Demographic data, type and number of procedures, involved organisms, hospitalization cost and length of stay, antibiotic treatments, and outcomes were ascertained for cases of CRKP-related PJI. Antimicrobial susceptibility testing was performed on CRKP isolates, including the following antibiotics: ciprofloxacin, amikacin, gentamicin, ceftazidime, piperacillin–tazobactam, doxycycline, tigecycline, and colistin.

length of stay, antibiotic treatments, and outcomes were ascertained for cases of CRKP-related PJI. Antimicrobial susceptibility testing was performed on CRKP isolates, including the following antibiotics: ciprofloxacin, amikacin, gentamicin, ceftazidime, piperacillin–tazobactam, doxycycline, tigecycline, and colistin. The mechanism of carbapenem resistance was ascertained by PCR amplification of blaKPC, blaNDM, blaVIM, and blaIMP.14–17,19 Genetic similarity among CRKP strains was investigated by repetitive sequence-based PCR (rep-PCR) using the DiversiLab strain typing system (Bacterial BarCodes; bioMérieux) (as validated in Rice et al.14). Isolates with >95% similarity were considered of the same clonal type.16 Multilocus sequence typing (MLST) was performed on all CRKP strains as described by Diancourt et al.23 DNA sequences of seven housekeeping genes (rpoB, gapA, mdh, pgi, phoE, infB, and tonB) were compared with the MLST database (http://www.pasteur.fr/recherche/genopole/PF8/mlst/).

e considered of the same clonal type.16 Multilocus sequence typing (MLST) was performed on all CRKP strains as described by Diancourt et al.23 DNA sequences of seven housekeeping genes (rpoB, gapA, mdh, pgi, phoE, infB, and tonB) were compared with the MLST database (http://www.pasteur.fr/recherche/genopole/PF8/mlst/). 3. Results Between the years 2007 and 2010, 221 patients were identified as having cultures of clinical samples with CRKP. Twenty-three (10.4%) patients with CRKP possessed a bone and joint-related infection. Three (1.3%) of these cases involved an infected orthopedic joint prosthesis. All cases occurred in patients with TKA, and in all cases CRKP were recovered as part of a polymicrobial or ‘complex’ infection. The initial pathogen was methicillin-susceptible S. aureus (MSSA) in two cases, whereas the other case was a polymicrobial infection with vancomycin-resistant enterococci (VRE), vancomycin-susceptible enterococci (VSE), and Proteus mirabilis. Demographic data, comorbidities, number of procedures, organisms involved, hospitalization cost (on one patient only) and length of stay, antibiotic treatments, and outcomes are summarized in Table 1.

infection with vancomycin-resistant enterococci (VRE), vancomycin-susceptible enterococci (VSE), and Proteus mirabilis. Demographic data, comorbidities, number of procedures, organisms involved, hospitalization cost (on one patient only) and length of stay, antibiotic treatments, and outcomes are summarized in Table 1. A total of 10 CRKP isolates were saved from the three patients. Results of antimicrobial susceptibility testing of CRKP isolates from case 3 are presented in Table 2. Using validated PCR primers and controls, all CRKP isolates were determined to harbor blaKPC. Genetic typing with rep-PCR demonstrated a high percentage of similarity between isolates belonging to two of the patients (cases 1 and 2). Case 3 was infected with CRKP with a different rep-PCR pattern. However, the six CRKP isolates obtained from this patient were similar to each other (Figure 1). MLST revealed that all strains belonged to sequence type (ST) 258. Of note, ST258 and the rep-PCR strain types identified in these three cases were similar to the predominant CRKP strains in our institution (data not shown).

owever, the six CRKP isolates obtained from this patient were similar to each other (Figure 1). MLST revealed that all strains belonged to sequence type (ST) 258. Of note, ST258 and the rep-PCR strain types identified in these three cases were similar to the predominant CRKP strains in our institution (data not shown). 4. Case studies 4.1. Case 1 A 58-year-old male suffering from osteoarthritis and diabetes mellitus presented to the Cleveland Clinic with left knee pain and swelling, fever, and hypotension. Clinical evaluation indicated that the infection originated from an infected left TKA implanted 5 years earlier. Blood and synovial fluid cultures obtained upon admission grew methicillin-susceptible S. aureus (MSSA) (Table 1). Antibiotic treatment with intravenous (IV) oxacillin was started and a two-stage left knee revision arthroplasty was planned, with initial explantation of the prosthesis and placement of an antibiotic spacer. Two weeks after explantation and spacer placement, the patient had a wound infection due to Alcaligenes faecalis. The wound was debrided and antibiotics were changed to piperacillin–tazobactam resulting in a good initial response.

anned, with initial explantation of the prosthesis and placement of an antibiotic spacer. Two weeks after explantation and spacer placement, the patient had a wound infection due to Alcaligenes faecalis. The wound was debrided and antibiotics were changed to piperacillin–tazobactam resulting in a good initial response. Two months later, the patient presented with wound dehiscence and cement exposure. Tissue cultures from the knee capsule grew CRKP (see Figure 1, case 1, isolate 1) and VRE. After left medial and lateral gastrocnemius muscle flaps, split thickness skin grafting, and exchange of the antibiotic spacer, he was started on IV daptomycin and oral doxycycline. Within 2 months of this last procedure he underwent another spacer exchange and radical knee debridement: he required patellectomy and quadriceps plasty with rotation of the muscle into the open wound area. Purulence was present in the femoral and the tibial canals and in the posterior recesses of the joint. Intraoperative cultures were positive for CRKP, Acinetobacter baumannii, and Candida parapsilosis. He was discharged from the hospital on IV tigecycline and oral fluconazole only to be readmitted a week later with wound drainage, fever, and hypotension. Synovial fluid and blood cultures were still positive for CRKP. Despite left above-the-knee amputation, maximum medical support, and combined antibiotic therapy with IV colistin, amikacin, and tigecycline, the patient died on postoperative day 3.

o be readmitted a week later with wound drainage, fever, and hypotension. Synovial fluid and blood cultures were still positive for CRKP. Despite left above-the-knee amputation, maximum medical support, and combined antibiotic therapy with IV colistin, amikacin, and tigecycline, the patient died on postoperative day 3. His C-reactive protein (CRP) levels did not change significantly (26.7 mg/dl at the time of diagnosis, 24.6 mg/dl 24 h prior to death); his white blood cell count (WBC) decreased abnormally from 16.2 × 109/l to 1.9 × 109/l at the time of death.

o be readmitted a week later with wound drainage, fever, and hypotension. Synovial fluid and blood cultures were still positive for CRKP. Despite left above-the-knee amputation, maximum medical support, and combined antibiotic therapy with IV colistin, amikacin, and tigecycline, the patient died on postoperative day 3. His C-reactive protein (CRP) levels did not change significantly (26.7 mg/dl at the time of diagnosis, 24.6 mg/dl 24 h prior to death); his white blood cell count (WBC) decreased abnormally from 16.2 × 109/l to 1.9 × 109/l at the time of death. 4.2. Case 2 A 72-year-old male underwent left TKA for osteoarthritis, complicated 3 years later by a late PJI with development of a fistulous tract. The patient was treated with a two-stage revision with interval placement of an antibiotic spacer. Peri-articular soft tissue cultures obtained intraoperatively were positive for Proteus mirabilis, VRE, and VSE. Treatment with oral ciprofloxacin, linezolid, and rifampin were initiated. He was discharged to a skilled nursing facility. Eight weeks after the placement of the antibiotic spacer, he developed wound dehiscence and required re-revision and spacer exchange; intraoperative cultures were sterile. He was started on IV daptomycin and ciprofloxacin. Seven days later, the patient developed bacteremia with methicillin-resistant S. aureus (MRSA) and underwent evacuation of a hematoma and removal of the tissue expander. At this time, intraoperative cultures grew carbapenem-resistant A. baumannii and CRKP (see Figure 1, case 2, isolate 1). He was treated with IV vancomycin and tigecycline for 3 months, followed by oral doxycycline for 2 months.

S. aureus (MRSA) and underwent evacuation of a hematoma and removal of the tissue expander. At this time, intraoperative cultures grew carbapenem-resistant A. baumannii and CRKP (see Figure 1, case 2, isolate 1). He was treated with IV vancomycin and tigecycline for 3 months, followed by oral doxycycline for 2 months. Twelve months after the second revision surgery, the patient underwent re-implantation of the knee prosthesis. Intraoperative cultures were positive for MSSA, and he was treated with IV oxacillin. Three weeks postoperatively, he developed purulent wound drainage and wound cultures grew carbapenem-resistant A. baumannii, while synovial fluid from the knee grew CRKP (see Figure 1, case 2, isolate 2). He underwent synovectomy and polyethylene removal, and all prosthetic joint components were exchanged. After a second washout 1 week later, cultures remained positive for CRKP (see Figure 1, case 2, isolate 3). He completed an 8-week course of IV oxacillin and tigecycline and was placed on a long-term suppressive regimen with oral doxycycline. Unfortunately, the patient died 4 months later. His CRP levels decreased from 2.4 mg/dl at the time of diagnosis to 0.4 at time of discharge. Likewise, his WBC decreased from 12.5 × 109/l to 8.2 × 109/l at the time of discharge.

Twelve months after the second revision surgery, the patient underwent re-implantation of the knee prosthesis. Intraoperative cultures were positive for MSSA, and he was treated with IV oxacillin. Three weeks postoperatively, he developed purulent wound drainage and wound cultures grew carbapenem-resistant A. baumannii, while synovial fluid from the knee grew CRKP (see Figure 1, case 2, isolate 2). He underwent synovectomy and polyethylene removal, and all prosthetic joint components were exchanged. After a second washout 1 week later, cultures remained positive for CRKP (see Figure 1, case 2, isolate 3). He completed an 8-week course of IV oxacillin and tigecycline and was placed on a long-term suppressive regimen with oral doxycycline. Unfortunately, the patient died 4 months later. His CRP levels decreased from 2.4 mg/dl at the time of diagnosis to 0.4 at time of discharge. Likewise, his WBC decreased from 12.5 × 109/l to 8.2 × 109/l at the time of discharge. 4.3. Case 3 A 70-year-old female underwent right knee arthroplasty revision 1 month after a primary TKA, because of recurrent dislocation of the prosthesis. She had a history of rheumatoid arthritis treated with methotrexate and hydroxychloroquine. An area of purulence within the subcutaneous tissue was noted during the surgical revision, although it did not track to the prosthesis. All hardware was removed, and an antibiotic-impregnated cement was placed. Intraoperative tissue cultures were positive for Corynebacterium sp and VSE. She received IV vancomycin for 6 weeks followed by TKA re-implantation. The postoperative period was complicated by multiple episodes of infection at the surgical site, which required wound debridement and wound therapy with negative-pressure. She was discharged to a skilled nursing facility.

orynebacterium sp and VSE. She received IV vancomycin for 6 weeks followed by TKA re-implantation. The postoperative period was complicated by multiple episodes of infection at the surgical site, which required wound debridement and wound therapy with negative-pressure. She was discharged to a skilled nursing facility. Five months after the primary TKA, CRKP (see Figure 1, case 3, isolate 1) was isolated from the surgical wound and she underwent hardware explantation. Peri-articular tissue cultures also grew CRKP (see Figure 1, case 3, isolate 2). Initially, she was treated with IV tigecycline, but it was stopped due to nausea. Therapy with IV colistin was initiated, but was later discontinued due to circumoral paresthesias. The patient was again treated with IV tigecycline, but she developed drug-induced cholestasis and acute kidney injury. Therefore, antibiotics were held. Two years after the primary TKA, she underwent a spacer exchange; intraoperative cultures remained positive for CRKP (see Figure 1, case 3, isolate 3), which was now resistant to amikacin. Despite treatment with IV tigecycline (which was tolerated well), a new lateral sinus tract developed requiring excision and spacer exchange. Tissue cultures, however, were negative.

acer exchange; intraoperative cultures remained positive for CRKP (see Figure 1, case 3, isolate 3), which was now resistant to amikacin. Despite treatment with IV tigecycline (which was tolerated well), a new lateral sinus tract developed requiring excision and spacer exchange. Tissue cultures, however, were negative. Six months after the spacer was exchanged, a second revision was needed because the spacer fractured. She underwent a muscle flap from the left thigh to the right knee, which was complicated by wound failure requiring debridement and an infection in the wound from the donor area caused by Streptococcus pneumoniae and MRSA. She completed a course of IV tigecycline and vancomycin, followed by long-term oral ciprofloxacin and clindamycin.

uscle flap from the left thigh to the right knee, which was complicated by wound failure requiring debridement and an infection in the wound from the donor area caused by Streptococcus pneumoniae and MRSA. She completed a course of IV tigecycline and vancomycin, followed by long-term oral ciprofloxacin and clindamycin. Six months later, the spacer was removed and a distal femoral TKA was placed. Cultures were negative, although she had been empirically re-started on IV tigecycline before the surgery. On postoperative day 7, she developed partial dehiscence of the surgical wound and cultures grew CRKP, while cultures of the synovial fluid remained negative. Despite aggressive surgical debridement, the patient became septic requiring amputation above the right knee, followed by right hip disarticulation. Therapy with IV colistin was initiated, but changed to IV tigecycline and amikacin after development of acute kidney injury. Additionally, she required five subsequent wound debridements. Culture of tissue obtained from these procedures grew CRKP, eventually resistant to amikacin and colistin. Amikacin was replaced with ciprofloxacin, but CRKP (see Figure 1, case 3, isolate 6) persisted. As there were no signs of systemic infection, all antibiotics were discontinued. At 8-month follow-up, her surgical site appeared well-healed with no signs of infection. Her CRP levels decreased significantly from 27.1 mg/dl at the time of diagnosis to 0.6 mg/dl 24 h prior to discharge; her WBC improved from 27.4 × 109/l to 8.1 × 109/l at the time of discharge.

Six months later, the spacer was removed and a distal femoral TKA was placed. Cultures were negative, although she had been empirically re-started on IV tigecycline before the surgery. On postoperative day 7, she developed partial dehiscence of the surgical wound and cultures grew CRKP, while cultures of the synovial fluid remained negative. Despite aggressive surgical debridement, the patient became septic requiring amputation above the right knee, followed by right hip disarticulation. Therapy with IV colistin was initiated, but changed to IV tigecycline and amikacin after development of acute kidney injury. Additionally, she required five subsequent wound debridements. Culture of tissue obtained from these procedures grew CRKP, eventually resistant to amikacin and colistin. Amikacin was replaced with ciprofloxacin, but CRKP (see Figure 1, case 3, isolate 6) persisted. As there were no signs of systemic infection, all antibiotics were discontinued. At 8-month follow-up, her surgical site appeared well-healed with no signs of infection. Her CRP levels decreased significantly from 27.1 mg/dl at the time of diagnosis to 0.6 mg/dl 24 h prior to discharge; her WBC improved from 27.4 × 109/l to 8.1 × 109/l at the time of discharge. 5. Discussion This unique case series serves to highlight the opportunistic, deleterious, and chronic nature of CRKP as a cause of PJI. In our series, CRKP PJIs exacted a tremendous cost in terms of morbidity, disability, health care expenses, and lost lives. Poor clinical outcomes occurred with CRKP PJI despite the implementation of intensive medical and surgical treatment regimens. Patients with multiple comorbidities may be mostly affected, and the MDR profile of the causative organisms may dramatically impact effective antibiotic therapy. A notable aspect of these cases was the persistence of CRKP, which also influenced length of stay and the need for recovery in post-acute care facilities. The sums of these factors lead us to conclude that CRKP PJIs are potentially incurable infections. In addition we found that: (1) CRKP was very difficult to eradicate (persistent) in PJI; (2) multiple surgeries and antibiotic courses need to be undertaken and patients require a prolonged length of stay; (3) resistance to ‘last line agents’ (colistin and amikacin) emerges on therapy; (4) the same strain of CRKP may be responsible for relapse of infection.

very difficult to eradicate (persistent) in PJI; (2) multiple surgeries and antibiotic courses need to be undertaken and patients require a prolonged length of stay; (3) resistance to ‘last line agents’ (colistin and amikacin) emerges on therapy; (4) the same strain of CRKP may be responsible for relapse of infection. When compared to 19 previously reported Gram-negative PJIs with non-resistant strains,24 it is clear that CRKP-related infections are more severe. Specifically, the median duration of hospital stay was longer (101 days vs. 31 days), median WBC was higher (14.92 × 109/l vs. 8.1 × 109/l), and mortality was higher (67% vs. 5%) for patients with CRKP PJIs. Another case series of 31 patients with Gram-negative PJIs, reported that irrigation and debridement alone was successful in eradicating 70% of infections.9 This was in stark contrast to our CRKP patients, who underwent 10 or more procedures. Interestingly, all of the previously reported Gram-negative PJIs were monomicrobial, while CRKP arose at least 2 months after prosthetic infection with a different primary organism. Thus, early and adequate treatment of primary PJIs may help prevent CRKP adverse outcomes.

patients, who underwent 10 or more procedures. Interestingly, all of the previously reported Gram-negative PJIs were monomicrobial, while CRKP arose at least 2 months after prosthetic infection with a different primary organism. Thus, early and adequate treatment of primary PJIs may help prevent CRKP adverse outcomes. The steady increase in the rates of CRKP infections in the USA is worrisome. In 2007, up to 8% of all K. pneumoniae isolates reported to the US Centers for Disease Control and Prevention (CDC) were carbapenem-resistant (compared to less than 1% in 2000) due to the widespread dissemination of the K. pneumoniae carbapenemase (KPC) gene, blaKPC.25 Isolates of blaKPC- harboring CRKP responsible for PJI in this report belonged to ST258, the predominant lineage of KPC-harboring organisms in the USA and in other parts of the world.26,27 The factors underlying the success of this particular lineage or clonal group remain unclear, although its MDR profile likely confers a selective advantage in the setting of broad-spectrum antibiotic therapy.

d to ST258, the predominant lineage of KPC-harboring organisms in the USA and in other parts of the world.26,27 The factors underlying the success of this particular lineage or clonal group remain unclear, although its MDR profile likely confers a selective advantage in the setting of broad-spectrum antibiotic therapy. These cases of PJI illustrate the potential for ST258 KPC-producing CRKP to cause persistent infection, as documented with the use of molecular typing techniques (Figure 1). The long duration of CRKP infection and colonization may create additional opportunities for its dissemination in the health care system. This may be of particular importance in long-term care facilities, which have emerged as ‘silent reservoirs’ of MDR organisms, and where it is more difficult to implement antimicrobial stewardship and infection control programs aimed at controlling CRKP.28

opportunities for its dissemination in the health care system. This may be of particular importance in long-term care facilities, which have emerged as ‘silent reservoirs’ of MDR organisms, and where it is more difficult to implement antimicrobial stewardship and infection control programs aimed at controlling CRKP.28 The successful management of PJIs depends on the combination of surgical and antibiotic therapy, through different approaches, including one-stage irrigation and debridement (with possible polyethylene exchange), two-stage revision procedure, and a hybrid modality with partial component exchange and retention of the prosthesis.29,30 The risk of failure of these different approaches is considerable and depends on the type of surgery performed. A case series of 53 first-time PJI, secondary to Gram-negative organisms, reported a 2-year survival rate of 27% (95% confidence interval (CI) 16–34%) for debridement and retention of the prosthesis, 69% (95% CI 59%–84%) for resection arthroplasty, and 87% (95% CI 80–99%) for two-stage exchange.10 Ineffective antibiotic therapy may affect these outcomes, as demonstrated by treatment failures when prostheses are retained and organisms become resistant to fluoroquinolones.12 Of note, CRKP are often resistant to fluoroquinolones (as well as beta-lactams), leaving aminoglycosides, tigecycline, and polymyxins (chiefly in the form of colistin) as the only active antibiotics against this type of organism. Use of these agents may be limited by side effects and toxicity (gastrointestinal in the case of tigecycline, renal with aminoglycosides and polymyxins).31 Unfortunately, the doses of polymyxins that are commonly used to treat CRKP may lead to a relatively high rate of nephrotoxicity.32 Our series demonstrated that resistance to amikacin and colistin can also occur during the course of therapy. We note that in the case of amikacin resistance, we suspect that either horizontal gene transfer has occurred (acquisition of an aminoglycoside modifying enzyme on a mobile plasmid) or there has been up-regulation of an efflux pump. The mechanism of colistin resistance in K. pneumoniae is not fully known, but likely does not involve a plasmid-mediated process.

tance, we suspect that either horizontal gene transfer has occurred (acquisition of an aminoglycoside modifying enzyme on a mobile plasmid) or there has been up-regulation of an efflux pump. The mechanism of colistin resistance in K. pneumoniae is not fully known, but likely does not involve a plasmid-mediated process. In device-associated infections, it is possible that bacteria are enclosed in slime-forming biofilms.33 This complicates both treatment strategies to eradicate biofilms and diagnostics, as bacteria tend to conglomerate, lowering the culture yield. Sonication of the prosthesis could aid in higher culture results.34 Whether CRKP biofilm formation plays a role in the resilient nature of these infections is unknown. Further complicating the treatment options for CRKP-related PJI, pharmacologic studies suggest that the synovial fluid and bone distribution of the remaining active antibiotics against CRKP is limited. Aminoglycosides appear to be less active in synovial fluid and bone, perhaps because of the acidic environment of the synovial fluid.35 Reports also indicate that colistin distribution to bone is only between 15% and 25%.36 However, when used in combination with rifampin and imipenem,37 or with tigecycline,38 colistin has been reported to successfully treat MDR Pseudomonas aeruginosa-related osteomyelitis.

he acidic environment of the synovial fluid.35 Reports also indicate that colistin distribution to bone is only between 15% and 25%.36 However, when used in combination with rifampin and imipenem,37 or with tigecycline,38 colistin has been reported to successfully treat MDR Pseudomonas aeruginosa-related osteomyelitis. Evidence exists suggesting that the outcomes of serious infections (e.g., bacteremias) caused by KPC-producing CRKP are improved with the use of combination therapy, either carbapenems in conjunction with tigecycline or with colistin,39 or tigecycline combined with colistin.40 Interestingly, tigecycline alone also seems to have poor distribution into the synovial fluid and bone. As shown in a report by Ji et al., the synovial and/or bone concentrations of tigecycline ranged from 31% to 41% of those found in the serum,41 which is typically below the MIC of most Gram-negative bacteria. One of the few antibiotics with consistent evidence of high bone penetration is fosfomycin (not available for intravenous use in the USA). Although the bone concentrations of fosfomycin may reach up to 43% of the serum concentration, these concentrations are above the MIC of most bacteria.42 Fosfomycin retains excellent activity against contemporary KPC-producing CRKP isolates, and synergistic bactericidal activity against CRKP has been demonstrated between fosfomycin and carbapenems.43,44 Therefore, fosfomycin, administered as part of combination therapy, has the potential to become a preferred antibiotic for CRKP bone-related infections. However, the use of fosfomycin in CRKP PJI has yet to be validated by clinical experience.

against CRKP has been demonstrated between fosfomycin and carbapenems.43,44 Therefore, fosfomycin, administered as part of combination therapy, has the potential to become a preferred antibiotic for CRKP bone-related infections. However, the use of fosfomycin in CRKP PJI has yet to be validated by clinical experience. In conclusion, the devastating effects of CRKP PJI and their almost intractable nature underscore the crisis precipitated by the emergence of multidrug-resistant Gram-negative organisms. That such infections can complicate TKA, an increasingly common procedure aimed at improving function in older adults with disability, should serve as a warning to health care professionals and the public. Efforts to prevent and control CRKP applied throughout the continuum of health care are justified to avoid this type of infection. The future availability of new drugs containing beta-lactamase inhibitors such as avibactam (formerly designated as NXL-104) may offer a reprieve against KPC-producing organisms, but would not overcome other carbapenemase types now circulating worldwide (e.g., NDM, VIM, or IMP metallo-beta-lactamases).45,46 In the meantime, the combined use of currently available antibiotics as part of the management of these uniquely challenging infections needs to be investigated further.

producing organisms, but would not overcome other carbapenemase types now circulating worldwide (e.g., NDM, VIM, or IMP metallo-beta-lactamases).45,46 In the meantime, the combined use of currently available antibiotics as part of the management of these uniquely challenging infections needs to be investigated further. This work was supported in part by the Veterans Affairs Merit Review Program (RAB), the National Institutes of Health (Grant AI072219-05 and AI063517-07 to RAB), and the Geriatric Research Education and Clinical Center VISN 10 (RAB). FP is supported as the Louis Stokes Cleveland Department of Veterans Affairs Medical Center and Case Western Reserve University School of Medicine Scholar in Infectious Diseases and the Geriatric Research Education and Clinical Center VISN 10. ☆ This manuscript was presented at the 49th Annual Meeting of the Infectious Diseases Society of America, Boston, Massachusetts, USA, October 20–23, 2011. Ethical approval: This work was approved by the Cleveland Clinic Institutional Review Board and thereby complies with the policy of the International Journal of Infectious Diseases on ethical consent. Conflict of interest: The authors declare no conflict of interest. Figure 1 Dendrogram illustrating the results of molecular typing with rep-PCR of carbapenem-resistant Klebsiella pneumoniae isolated from three cases of prosthetic joint infection. Isolates obtained from cases 1 and 2 share >97% similarity, indicating that they belong to the same strain type. Isolates from case 3 belong to a different rep-PCR strain type, but are similar to each other.

h rep-PCR of carbapenem-resistant Klebsiella pneumoniae isolated from three cases of prosthetic joint infection. Isolates obtained from cases 1 and 2 share >97% similarity, indicating that they belong to the same strain type. Isolates from case 3 belong to a different rep-PCR strain type, but are similar to each other. Table 1 Characteristics and clinical outcomes of cases of prosthetic joint infection caused by carbapenem-resistant Klebsiella pneumoniae

h rep-PCR of carbapenem-resistant Klebsiella pneumoniae isolated from three cases of prosthetic joint infection. Isolates obtained from cases 1 and 2 share >97% similarity, indicating that they belong to the same strain type. Isolates from case 3 belong to a different rep-PCR strain type, but are similar to each other. Table 1 Characteristics and clinical outcomes of cases of prosthetic joint infection caused by carbapenem-resistant Klebsiella pneumoniae Variable Case 1 Case 2 Case 3 Age (years), sex 58, male 72, male 70, female Comorbidities Osteoarthritis, diabetes Osteoarthritis, coronary artery disease, congestive heart failure RA on immunosuppression with methotrexate and hydroxychloroquine Onset of first PJI (months from index surgery) 60 36 1 Primary organism PJI MSSA VSE, VRE, Proteus mirabilis Corynebacterium sp and VSE Onset of CRKP PJI (months from first PJI) 2 2 5 Number of procedures (n) 10 12 57 Antibiotics Oxacillin; piperacillin–tazobactam; daptomycin and oral doxycycline; tigecycline and fluconazole; colistin, amikacin, and tigecycline Ciprofloxacin, linezolid, and rifampin; daptomycin and ciprofloxacin; vancomycin and tigecycline → doxycycline; oxacillin, oxacillin and tigecycline → doxycycline Vancomycin; tigecycline; colistin; tigecycline; tigecycline; tigecycline and vancomycin → oral ciprofloxacin and clindamycin; tigecycline; colistin; tigecycline, and amikacin; ciprofloxacin WBC ×109/l (median (IQR)) 9.07 (0.63, 12.49) 8.45 (7.73, 9.75) 8.92 (7.40, 11.68) Hospital LOS (days) 51 101 225 Hospitalization costs ($) N/A N/A 850 000 Functional status Above-the-knee amputation Full Disarticulated Outcomes Died Died Alive with major disability RA, rheumatoid arthritis; PJI, prosthetic joint infection; MSSA, methicillin-susceptible Staphylococcus aureus; VSE, vancomycin-susceptible Enterococcus sp; VRE, vancomycin-resistant Enterococcus sp; CRKP, carbapenem-resistant Klebsiella pneumoniae; WBC, white blood cell count; IQR, interquartile range; LOS, length of stay; N/A, not available.

PJI, prosthetic joint infection; MSSA, methicillin-susceptible Staphylococcus aureus; VSE, vancomycin-susceptible Enterococcus sp; VRE, vancomycin-resistant Enterococcus sp; CRKP, carbapenem-resistant Klebsiella pneumoniae; WBC, white blood cell count; IQR, interquartile range; LOS, length of stay; N/A, not available. Table 2 Antimicrobial susceptibility testing of carbapenem-resistant Klebsiella pneumoniae isolates from prosthetic joint infections Antibiotics Case 3, isolate 1 MIC (in μg/ml) and interpretation Case 3, isolate 6 MIC (in μg/ml) and interpretation Amikacin 4 S >64 R Colistin <2 S >8 R Gentamicin >16 R >16 R Tigecycline 1 S 1 S Tobramycin >16 R >16 R Imipenem >16 R >16 R MIC, minimum inhibitory concentration; S, susceptible; R, resistant.

1. Introduction Adjuvants are crucial for the generation of an optimal immune response to purified protein vaccines. Recent advances in our understanding of innate immunity have led to the identification of immune pathways and adjuvant formulations more suitable for clinical advancement. One area of particular interest is the discovery of agonists that target the toll-like receptors (TLRs). Signaling from the TLRs expressed on monocytes and monocyte-derived dendritic cells (moDCs), through recognition of various pathogen-associated molecular patterns, induces these cells to secrete distinct cytokines, which in turn influence T-cell differentiation.1 Recent research has demonstrated that microbial stimulation promotes monocyte differentiation into DC-SIGN/CD209+ moDCs in vivo and these moDCs exhibit a greater capacity than lymphoid resident dendritic cells (DCs) to stimulate T-cell proliferation once they acquire the antigens together with TLR4 ligands.2 Cervarix, a prophylactic vaccine against human papillomavirus (HPV) types 16 and 18, recently received approval from the US Food and Drug Administration (FDA).3 In this vaccine, viral antigens are formulated with monophosphoryl lipid A, a TLR4-targeted adjuvant, which confers protective immunity against HPV and promotes immune response broadening. Other adjuvants targeting TLRs are in development for new therapeutic vaccine candidates for cancers and some chronic infectious diseases.3,4

ccine, viral antigens are formulated with monophosphoryl lipid A, a TLR4-targeted adjuvant, which confers protective immunity against HPV and promotes immune response broadening. Other adjuvants targeting TLRs are in development for new therapeutic vaccine candidates for cancers and some chronic infectious diseases.3,4 The idea of utilizing immunotherapy for chronic hepatitis B virus (HBV) infection is supported by findings that bone marrow transplantation of anti-HBV immunity to the recipient could cure chronic HBV infection.5,6 A therapeutic vaccine, which represents one of the immunotherapy strategies, has been developed in different forms.7–10 However, the clinical response to these vaccines has been poor, probably because of immune tolerance to HBV viral antigens.11,12 Patients who recover from acute HBV infections usually have vigorous antibody responses, with antibodies against hepatitis B surface antigen (anti-HBs) easily detectable, and polyclonal T-cell responses against multiple HBV antigens (HBV-Ag).13,14 Therefore, it is important for an effective therapeutic vaccine to induce multiple HBV antigen-specific responses by activating both antigen-specific CD4+ and CD8+ T-cells in the immune tolerant state.

ace antigen (anti-HBs) easily detectable, and polyclonal T-cell responses against multiple HBV antigens (HBV-Ag).13,14 Therefore, it is important for an effective therapeutic vaccine to induce multiple HBV antigen-specific responses by activating both antigen-specific CD4+ and CD8+ T-cells in the immune tolerant state. Previously, we reported that human monocytes differentiated into moDCs when they phagocytosed dead cells containing ssRNA, the TLR7/8 agonist, and induced strong CD8+ T-cell responses to the cell-associated antigens.15 Using chemically synthesized TLR7/8 agonists we demonstrated that CL075 and CL097 stimulated newly recruited monocyte-derived cells into potent antigen-presenting cells (APCs) that enhance hepatitis B surface antigen (HBsAg) immunogenicity in both humans and mice.16 TLR7/8 agonists conjugated to HIV Gag protein have been shown to enhance the magnitude and quality of Th1 and CD8+ T-cell responses in non-human primates.17,18 TLR7/8 agonists appear to be good candidate adjuvants for prophylactic vaccines to induce Th1 responses in normal animals.16–20 However, it is unknown whether TLR7/8 agonist-conjugated vaccines could break the established antigen-specific tolerance and induce antigen-specific immune responses.

non-human primates.17,18 TLR7/8 agonists appear to be good candidate adjuvants for prophylactic vaccines to induce Th1 responses in normal animals.16–20 However, it is unknown whether TLR7/8 agonist-conjugated vaccines could break the established antigen-specific tolerance and induce antigen-specific immune responses. 2. Materials and methods 2.1. Mice and reagents C57BL/6 male wild-type mice and two independently generated HBV transgenic (HBV-Tg) mouse colonies (males, 7–8 weeks) with C57BL/6 background were used. C57BL/6-HBV-1.3 genome-eq transgenic mice were generated in the Transgenic Laboratory, Infectious Disease Center, Guangzhou.21 HBsAg-transgenic C57BL/ 6J-TgN (AlblHBV) 44Bri/J mice, which were originally generated in the laboratory of Dr Chisari, were purchased from Peking University, China. Both colonies constitutively express HBsAg in liver cells and secrete HBsAg in serum, as reported previously.22 All procedures involving mice were approved by the Institutional Animal Care and Use Committee of the Cancer Institute, Chinese Academy of Medical Sciences.

sari, were purchased from Peking University, China. Both colonies constitutively express HBsAg in liver cells and secrete HBsAg in serum, as reported previously.22 All procedures involving mice were approved by the Institutional Animal Care and Use Committee of the Cancer Institute, Chinese Academy of Medical Sciences. Recombinant HBsAg (yeast) was from Dalian Hissen Bio-pharm Inc.; recombinant influenza A H1N1 virus M1 protein (Escherichia coli) was from QuantoBio Biotechnology. Endotoxin levels of these proteins were less than 1.0 EU/μg protein. Recombinant hepatitis B core antigen (HBcAg) (E. coli) was from Jianan Biotechnology, China (endotoxin level 0.14 EU/μg protein). ELISA kits for quantifying HBsAg and anti-HBs were purchased from Wantai Biological Pharmacy, China. The TLR7/8 agonist, CL097, was purchased from InvivoGen (CA, USA); aluminum hydroxide gel was from Sigma-Aldrich (MO, USA). All chemicals for ultra performance liquid chromatography (UPLC) were analytical grade. Detailed information on the materials and methods are provided in the Supplementary Material. 2.2. Preparation of vaccines and immunization protocol To prepare the TLR7/8 agonist-conjugated vaccine, HBsAg and HBcAg (5 μg of each) were diluted in 50 μl normal saline and mixed with the same volume of aluminum hydroxide (500 μg) at room temperature; 5 μg CL097 was then added. For immunization, each mouse received 5 μg HBV-Ag every 2 weeks.

es and immunization protocol To prepare the TLR7/8 agonist-conjugated vaccine, HBsAg and HBcAg (5 μg of each) were diluted in 50 μl normal saline and mixed with the same volume of aluminum hydroxide (500 μg) at room temperature; 5 μg CL097 was then added. For immunization, each mouse received 5 μg HBV-Ag every 2 weeks. 2.3. Confirmation of CL097 conjugation with alum-absorbed HBV-Ag by UPLC-Q/TOF MS UPLC quadrupole time-of-flight mass spectrometry (UPLC-Q/TOF MS) was performed using an Acquity UPLC BEH C18 column on the Acquity UPLC system, equipped with SYNAPT G2 HDMS (Waters Corp., UK) with an electrospray ionization source (ESI) in positive ion mode. The column was maintained at 40 °C and eluted at a flow rate of 0.45 ml/min, using a mobile phase of 0.1% formic acid in water (18.2 mΩ) and acetonitrile (HPLC grade; J.T. Baker, Phillipsburg, NJ, USA). Leucine-enkephalin was used as the lock mass in all analyses (m/z 556.2771) at 0.5 μg/ml, with a flow rate of 5 ml/min. Data were collected in centroid mode from m/z 100 to m/z 1500. A series of standard working solutions was prepared and 5 μl of each was injected into the UPLC system for analysis after centrifugation at 6500 g for 5 min. CL097-conjugated HBV-Ag solution was divided into two parts after the same centrifugation. The pellet was resuspended in 20% sodium citrate solution and incubated at 37 °C for 2 h followed by the same centrifugation. All the supernatant was analyzed by UPLC-Q/TOF MS.

r analysis after centrifugation at 6500 g for 5 min. CL097-conjugated HBV-Ag solution was divided into two parts after the same centrifugation. The pellet was resuspended in 20% sodium citrate solution and incubated at 37 °C for 2 h followed by the same centrifugation. All the supernatant was analyzed by UPLC-Q/TOF MS. 2.4. Flow cytometry analysis (FACS; fluorescence activated cell sorting) Fluorescence-conjugated antibodies to mouse CD3, CD4, CD8, interferon gamma (IFN-γ), CD25, and FoxP3 were purchased from eBioscience (CA, USA). Flow cytometry staining for cell surface markers or intracellular cytokines was done using standard laboratory protocols. Data were acquired in an LSR-II (Becton-Dickinson, San Diego, CA) and analyzed using Flowjo software (Tree Star Inc, Asland, OR). 2.5. Assays for HBsAg-specific antibody-secreting cells (ASCs) The presence of HBsAg-specific ASCs in immunized mouse splenocytes was determined by ELISPOT assay, as described in the literature.23 Further details of the materials and methods are provided in the Supplementary Material. 2.6. Assays for HBsAg-specific T-cell cytokine production Splenocytes were isolated from immunized normal or HBV-Tg mice and cultured in the presence of 5 mg HBsAg for 72 h, followed by the addition of 1 × Brefeldin A Solution (Becton-Dickinson, San Diego, CA) for another 6 h. Cells were collected and stained using standard FACS staining protocols.

T-cell cytokine production Splenocytes were isolated from immunized normal or HBV-Tg mice and cultured in the presence of 5 mg HBsAg for 72 h, followed by the addition of 1 × Brefeldin A Solution (Becton-Dickinson, San Diego, CA) for another 6 h. Cells were collected and stained using standard FACS staining protocols. 2.7. Cell sorting and transfer Single splenocytes of HBV-1.3 genome-eq transgenic mice that were untreated or immunized with CL097-conjugated HBV-Ag, were labeled with fluorescein isothiocyanate (FITC)-conjugated CD4 and PE-Cy5-conjugated CD25 antibodies. The cells were sorted into two parts: CD4+CD25+ (Treg cells) and CD4+CD25+-depleted populations (non-Treg) using a FACSAria (BD, USA). Because Treg is only about 10% of peripheral CD4+ T-cells,24 each Naїve C57BL/6J mouse received 2 ×105 Treg or 2 ×106 non-Treg cells from sham-treated or immunized mice (in 200 μl phosphate buffered saline) via the tail vein.25 On the following day, all recipient mice were challenged intramuscularly with 5μg recombinant HBsAg or 5 μg recombinant influenza A H1N1 M1 proteins. Serum samples were collected on days 3, 6, 9, and 12. Anti-HBs serum levels were quantified using a commercialized ELISA kit; anti-M1 levels were determined using a direct ELISA assay developed in our laboratory. 2.8. Statistical analysis The statistical analysis was conducted using SPSS software (11.0). Statistical tests were performed using all-pairs Tukey–Kramer analysis and/or the two-tailed Student's t-test.

2.7. Cell sorting and transfer Single splenocytes of HBV-1.3 genome-eq transgenic mice that were untreated or immunized with CL097-conjugated HBV-Ag, were labeled with fluorescein isothiocyanate (FITC)-conjugated CD4 and PE-Cy5-conjugated CD25 antibodies. The cells were sorted into two parts: CD4+CD25+ (Treg cells) and CD4+CD25+-depleted populations (non-Treg) using a FACSAria (BD, USA). Because Treg is only about 10% of peripheral CD4+ T-cells,24 each Naїve C57BL/6J mouse received 2 ×105 Treg or 2 ×106 non-Treg cells from sham-treated or immunized mice (in 200 μl phosphate buffered saline) via the tail vein.25 On the following day, all recipient mice were challenged intramuscularly with 5μg recombinant HBsAg or 5 μg recombinant influenza A H1N1 M1 proteins. Serum samples were collected on days 3, 6, 9, and 12. Anti-HBs serum levels were quantified using a commercialized ELISA kit; anti-M1 levels were determined using a direct ELISA assay developed in our laboratory. 2.8. Statistical analysis The statistical analysis was conducted using SPSS software (11.0). Statistical tests were performed using all-pairs Tukey–Kramer analysis and/or the two-tailed Student's t-test. 3. Results 3.1. Binding of TLR7/8 agonist with particulate antigens is necessary to induce optimal humoral and cellular immune responses against HBsAg Under physiological conditions (pH 7.2–7.4), alum hydroxide adjuvants have a positive surface charge, while CL097, HBsAg, and HBcAg have a negative surface charge. Theoretically, CL097 binds to alum via the surface charge using the bond present in the N atom (Figure 1A). Using UPLC-Q/TOF MS, we confirmed the conjugation of CL097 to alum (Figure 1B). The conjugation was stable at normal pH but was partially dissociated when sodium citrate solution was added (Table 1). Conjugation of HBV-Ag, the HBsAg and HBcAg proteins, with alum was confirmed using the bicinchonnic acid (BCA) assay (Table 1).

PLC-Q/TOF MS, we confirmed the conjugation of CL097 to alum (Figure 1B). The conjugation was stable at normal pH but was partially dissociated when sodium citrate solution was added (Table 1). Conjugation of HBV-Ag, the HBsAg and HBcAg proteins, with alum was confirmed using the bicinchonnic acid (BCA) assay (Table 1). In comparison to the immunized mice with alum absorbed HBV-Ag (G1 in Figure 1C), the serum level of anti-HBs in the immunized mice using CL097-conjugated HBV-Ag through alum was significantly increased (G2, in Figure 1C). The antibodies against HBcAg in the two groups of mice showed a similar pattern (data not shown). However, no similar effect on antibody production was observed when CL097 was injected in free form, without being absorbed with HBV-Ag through alum (G3 in Figure 1C), or mixed with HBV-Ag without alum (G4 in Figure 1C). The percentage of HBsAg-specific, HBcAg-specific IFN-γ-producing T-cells was also significantly increased in the mice immunized with CL097-conjugated HBV-Ag (with CL097, in Figure 1D; Supplementary Material, Figure S1) compared to the mice immunized with HBV-Ag alone (without CL097, in Figure 1D; Supplementary Material, Figure S1).

e of HBsAg-specific, HBcAg-specific IFN-γ-producing T-cells was also significantly increased in the mice immunized with CL097-conjugated HBV-Ag (with CL097, in Figure 1D; Supplementary Material, Figure S1) compared to the mice immunized with HBV-Ag alone (without CL097, in Figure 1D; Supplementary Material, Figure S1). 3.2. Immunization with CL097-conjugated HBV-Ag results in the generation of anti-HBs antibodies and HBsAg-specific T-cells in HBV-Tg mice The effects of TLR7/8 agonist-conjugated HBV-Ag were then examined in HBV-Tg mice. To confirm the results, we tested it in two independently prepared HBV-Tg mouse colonies. In the AlblHBV 44Bri/J mice, a total of 8/11 (72.7%) generated serum anti-HBs antibodies following immunization with four doses of CL097-conjugated HBV-Ag, averaging 636 ± 258 mIU/ml (Figure 2A). In the C57BL/6-HBV-1.3 genome-eq mice, a total of 10/13 (76.9%) of the mice had serum detectable anti-HBs antibodies, with an average of 425 ± 118 mIU/ml (Figure 2B). The effect of the serum HBsAg concentration on anti-HBs production was analyzed in the C57BL/6-HBV-1.3 genome-eq mice. Generally, no correlation was found between serum anti-HBs levels and the corresponding pre-immunized serum HBsAg concentration in each mouse (R2 = 0.077, p = 0.357) (Figure 2C). However, anti-HBs was detectable in all mice with less than 3 μg/ml serum HBsAg (Figure 2C).

analyzed in the C57BL/6-HBV-1.3 genome-eq mice. Generally, no correlation was found between serum anti-HBs levels and the corresponding pre-immunized serum HBsAg concentration in each mouse (R2 = 0.077, p = 0.357) (Figure 2C). However, anti-HBs was detectable in all mice with less than 3 μg/ml serum HBsAg (Figure 2C). Anti-HBs persisted for more than 20 weeks after antibodies were initially induced. Following a single dose of vaccine booster, the concentration of anti-HBs increased dramatically (Figure 2D). Mice treated with alum-conjugated CL097 alone (sham-immunization) had no detectable serum anti-HBs (Figure 2, A and B). Serum alanine aminotransferase (ALT) levels were elevated in some vaccine-immunized and some sham-immunized HBV-Tg mice. On weeks 10 and 20, serum ALT returned to the same levels as untreated HBV-Tg mice (Figure 2E). Representative mice that had anti-HBs generated were euthanized at week 22 after the vaccine booster. Splenocytes from the mice immunized with CL097-conjugated HBV-Ag contained more HBsAg-specific B-cells than those from sham-immunized mice (Figure 3, A and B). After the splenocytes were stimulated with HBsAg in vitro, IFN-γ-producing CD4+ and CD8+ T-cells were also detected in immunized HBV-Tg mice with a more obvious increase in the number of IFN-γ-producing HBsAg-specific CD8+ T-cells compared to the sham-immunized mice (Figure 3C). These results indicate that inclusion of TLR7/8 agonist CL097 in the alum-absorbed vaccine results in induction of a robust immune response against tolerant HBsAg antigens in HBV-Tg mice.

more obvious increase in the number of IFN-γ-producing HBsAg-specific CD8+ T-cells compared to the sham-immunized mice (Figure 3C). These results indicate that inclusion of TLR7/8 agonist CL097 in the alum-absorbed vaccine results in induction of a robust immune response against tolerant HBsAg antigens in HBV-Tg mice. 3.3. Immunization with CL097-conjugated HBV-Ag breaks antigen-specific immune tolerance in HBV-Tg mice CD4+CD25+ Treg cells play an important role in maintaining immune tolerance. The percentage of CD4+CD25+FoxP3+ Treg cells decreased significantly in the vaccine-immunized (2.6%) versus sham-immunized mice (6.72%) (Figure 4A). We then sorted splenocytes from the immunized mice into two populations – CD4+CD25+ and CD4+CD25+-depleted cells – and transferred each of them intravenously into naïve C57BL/6J mice, followed by challenge with HBsAg or non-relevant influenza A virus M1protein (depicted in Figure 4B). The mice that received CD4+CD25+-depleted splenocytes from immunized mice generated anti-HBs immediately after HBsAg challenge, while those receiving cells from sham-immunized mice did not. Antibodies were detected as early as on day 3 and reached a plateau by day 9 (Figure 4C). The mice that received CD4+CD25+ cells from the vaccine-immunized mice generated anti-HBs antibodies by day 12 (Figure 4C). Mice receiving splenocytes from vaccine- or sham-immunized HBV-Tg exhibited similar responses to non-relevant M1 proteins (Figure 4D). These results suggest that immune tolerance to HBsAg was broken and specific immune responses against HBsAg were generated in HBV-Tg mice by immunization with CL097-conjugated HBV-Ag.

. Mice receiving splenocytes from vaccine- or sham-immunized HBV-Tg exhibited similar responses to non-relevant M1 proteins (Figure 4D). These results suggest that immune tolerance to HBsAg was broken and specific immune responses against HBsAg were generated in HBV-Tg mice by immunization with CL097-conjugated HBV-Ag. 4. Discussion Most chronic HBsAg carriers are infected in early life,26 and viral persistence has been associated with a defect in the development of specific immunity against HBV viral proteins.5,6,27 It is important for therapeutic vaccines to overcome immune tolerance and induce both specific antibodies against HBsAg and T-cell immunity against HBV structural proteins in the presence of chronic HBV infection.11,12,27 Our results demonstrated that binding of TLR7/8 agonist to alum-absorbed HBV-Ag was able to break the immune tolerance and induce humoral and cellular immune responses against HBsAg in HBV-Tg mice.

bodies against HBsAg and T-cell immunity against HBV structural proteins in the presence of chronic HBV infection.11,12,27 Our results demonstrated that binding of TLR7/8 agonist to alum-absorbed HBV-Ag was able to break the immune tolerance and induce humoral and cellular immune responses against HBsAg in HBV-Tg mice. Many TLR agonists can be potent adjuvant candidates for improving the magnitude and quality of memory T-cell responses.4,17,18 However, systemic activation of DCs by TLR ligands inhibits their antigen presentation function.28 Monocytes express a variety of TLRs, patrol various tissues for signs of infection and inflammation, and are approximately 20 times more abundant than DCs in blood and bone marrow. Therefore, monocytes are a promising target for the design of prophylactic and therapeutic vaccines against some chronic diseases.2,29 TLR7/8 are mainly expressed in intracellular compartments and sense the ligands when the cells take up the ligands.30 In this study, we demonstrated that binding CL097 with HBV-Ag through alum resulted in better immune responses than did the free form of CL097. We reported previously that the immunogens were mainly taken up by monocytes and monocyte-derived cells after immunization.16 Therefore, TLR7/8 conjugation with immunogens appeared important for inducing monocyte differentiation into potent moDCs that could prime naïve T-cells.

esponses than did the free form of CL097. We reported previously that the immunogens were mainly taken up by monocytes and monocyte-derived cells after immunization.16 Therefore, TLR7/8 conjugation with immunogens appeared important for inducing monocyte differentiation into potent moDCs that could prime naïve T-cells. TLR expression patterns are different in mice and humans. Distinct DC subsets also have different TLR expression patterns.1 TLR7 and TLR8 are intracellularly expressed and recognize ssRNA and synthetic antiviral imidazoquinoline components.1,30 In humans, TLR7 is mainly expressed in plasmacytoid DCs (pDCs), which produce high levels of type I IFNs in response to microbial stimulation. TLR8 is expressed in various tissues, with the highest levels in myeloid DCs and monocytes.1,31 In mice, TLR8 is nonfunctional and TLR7/8 agonists work through TLR7, which is expressed in murine monocytes and moDCs. The effect in humans might be different from what we observed in this murine system. However, stimulation with TLR7/8 agonists in both humans and mice has been shown to generate Th1-polarizing moDCs.16,20,32

, TLR8 is nonfunctional and TLR7/8 agonists work through TLR7, which is expressed in murine monocytes and moDCs. The effect in humans might be different from what we observed in this murine system. However, stimulation with TLR7/8 agonists in both humans and mice has been shown to generate Th1-polarizing moDCs.16,20,32 An elevation of serum ALT levels was observed in our vaccine-immunized mice, which might be harmful for patients with chronic HBV infection. Clinically, patients exhibit increased serum ALT levels during and after seroconversion from the status of hepatitis e antigen (HBeAg)-positive to anti-HBe-positive as a result of the immunological elimination of HBV-expressing liver cells. Consequently, this translates into reduced viral loads and clinical remission.33 In this study, HBsAg-specific B-cells and IFN-γ-producing CD4+ and CD8+ T-cells were only detected in vaccine-immunized but not in sham-treated HBV-Tg mice. The elevated serum ALT levels in vaccine-immunized mice were probably due to the effect of immunological elimination on HBsAg-expressing liver cells. However, we could not exclude the possibility of non-specific damage caused by the TLR7/8 agonists on liver cells, since the sham-treated mice also exhibited elevated ALT. In conclusion, HBV-Ag conjugated with TLR7/8 agonists reversed the HBsAg-specific non-responses in HBV-Tg mice. TLR7/8 agonists appear to be good candidate adjuvants for therapeutic vaccines to induce Th1 responses in the presence of a tolerant state, in addition to their reported effects when used in prophylactic vaccines.16–20

ion, HBV-Ag conjugated with TLR7/8 agonists reversed the HBsAg-specific non-responses in HBV-Tg mice. TLR7/8 agonists appear to be good candidate adjuvants for therapeutic vaccines to induce Th1 responses in the presence of a tolerant state, in addition to their reported effects when used in prophylactic vaccines.16–20 Supplementary Material Supplemental Material We are grateful to Dr Eskild Petersen at the Department of Infectious Diseases, Aarhus University Hospital, Denmark, for his valuable comments on the manuscript. Financial support: Supported by NFSC (81172888, 81161120495) and the PhD Program Foundation of the Ministry of Education of China (20111106110016). The grant sponsors had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. Conflict of interest: None to declare. Appendix A. Supplementary data: Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2014.07.015. The corresponding author had full access to all study data and final responsibility for the decision to submit for publication.

Conflict of interest: None to declare. Appendix A. Supplementary data: Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2014.07.015. The corresponding author had full access to all study data and final responsibility for the decision to submit for publication. Figure 1 Effectof TLR7/8 agonist on specific immune responses against HBsAg in wild-type C57BL/6 mice. (A) Structure of CL097 and its ESI mass spectrum (m/z 243→197) in positive ion mode performed in UPLC-Q/TOF MS. (B) Relative peak intensity of CL097 bound with alum. (C) Serum levels of anti-HBs measured 2 weeks after the second immunization (five mice per group, each was repeated three times). G1: immunization with 5 μg HBV-Ag absorbed to alum; G2: immunization with 5 μg HBV-Ag and 5 μg CL097 absorbed to alum; G3: immunization as for G1 in one side of the mice, 5 μg CL097 solution was injected in the other side; G4: immunization with 5 μg HBV-Ag and 5 μg CL097 without absorbing to alum. No anti-HBs was detectable in the naïve mice or the mice immunized with aluminum hydroxide alone (Alum) or CL097 alone (CL097); *p < 0.05, **p < 0.01 determined by all-pairs Tukey–Kramer analysis. (D) Frequency of HBsAg-specific, HBcAg-specific CD4+ and CD8+ IFN-γ-producing T-cells in the group pooled splenocytes based on FACS staining (Supplementary Material, Figure S1) (five mice per group, each was repeated three times); *p < 0.05, **p < 0.01 determined by two-tailed Student's t-test.

s Tukey–Kramer analysis. (D) Frequency of HBsAg-specific, HBcAg-specific CD4+ and CD8+ IFN-γ-producing T-cells in the group pooled splenocytes based on FACS staining (Supplementary Material, Figure S1) (five mice per group, each was repeated three times); *p < 0.05, **p < 0.01 determined by two-tailed Student's t-test. Figure 2 Humoral immune responses to HBsAg in HBV-Tg mice immunized with CL097-conjugated HBV-Ag. (A,B) Serum anti-HBs levels measured at 2weeks after four-dose immunization with CL097-conjugated HBV-Ag (HBV-Ag) in Alb1HBV 44Bri/J mice (A, n = 11) and in HBV-1.3 genome-eq mice (B, n = 13). Alum absorbed CL097 was used as the sham-immunization (sham). Each dot represents one mouse. (C) Spearman's correlation analysis between serum anti-HBs levels and the pre-immunized serum concentration of HBsAgin individual HBV-1.3 genome-eq mice. (D) Serum concentration of anti-HBs measured 20 weeks after the fourth immunization (Pre-Bst) and 2 weeks after one dose of booster (Af-Bst) using CL097-conjugated HBV-Ag (n = 5); **p < 0.01 determined by two-tailed Student's t-test. (E) Serum ALT levels measured at different time points before (pre-immu) and after the immunization. Each dot represents one mouse.

20 weeks after the fourth immunization (Pre-Bst) and 2 weeks after one dose of booster (Af-Bst) using CL097-conjugated HBV-Ag (n = 5); **p < 0.01 determined by two-tailed Student's t-test. (E) Serum ALT levels measured at different time points before (pre-immu) and after the immunization. Each dot represents one mouse. Figure 3 HBsAg-specific B-cells and T-cells in immunized HBV-Tg mice. (A) The numbers of HBsAg-specific antibody secreting cells (ASC) were determined by ELISPOT assay 2 weeks after boosting with CL097-conjugated HBV-Ag (CL097). Samples from the mice immunized with alum absorbed CL097 alone (sham) were also assayed. Wells not coated with HBsAg were used as spontaneous release controls (Med); *p < 0.05 determined by two-tailed Student's t-test. (B) Images showing representative ASC spots in each sample well. Each spot (black arrows) represents one HBsAg-specific B-cell. (C) Splenocytes isolated from mice immunized with CL097-conjugated HBV-Ag (HBV-Ag) or the mice immunized with alum absorbed CL097 alone (Sham) were re-stimulated with HBsAg. HBsAg-specific IFN-γ-producing T-cells were determined by intracellular cytokine staining (representative of five independent experiments).

-cell. (C) Splenocytes isolated from mice immunized with CL097-conjugated HBV-Ag (HBV-Ag) or the mice immunized with alum absorbed CL097 alone (Sham) were re-stimulated with HBsAg. HBsAg-specific IFN-γ-producing T-cells were determined by intracellular cytokine staining (representative of five independent experiments). Figure 4 Frequency and function of Treg cells in the mice immunized with CL097-conjugated HBV-Ag. (A) Frequency of CD3+CD4+CD25+FoxP3+Treg cells in splenocytes of the sham-immunized (sham) or the mice immunized with CL097-conjugated HBV-Ag (CL097) (representative of three independent experiments). (B) Sorting strategy of CD4+CD25+ Treg (G1) and CD4+CD25+-depleted (G2) cell populations. Each of the sorted cell populations, 2 × 105 cells/mouse for CD4+CD25+ and 2 × 106 cells/mouse for CD4+CD25+-depleted cells, were injected intravenously into Naїve C57BL/6 mice (n = 3). Dot plot indicating the splenocytes from mice immunized with CL097-conjugated HBV-Ag. On the day of cell transfer, each mouse received 5 μg HBsAgor 5 μg influenza A virus M1 protein intramuscularly. (C) Anti-HBs in mice that received the G1 or G2 cells, as indicated in Figure 4B, from sham-immunized mice was undetectable (dashed line). Serum anti-HBs levels in mice that received cells from CL097-conjugated HBV-Ag immunized mice are shown as solid lines. (D) Serum levels of antibodies against influenza A virus M1 proteins in the same mice on day 12.

the G1 or G2 cells, as indicated in Figure 4B, from sham-immunized mice was undetectable (dashed line). Serum anti-HBs levels in mice that received cells from CL097-conjugated HBV-Ag immunized mice are shown as solid lines. (D) Serum levels of antibodies against influenza A virus M1 proteins in the same mice on day 12. Table 1 Binding of CL097 and HBV antigens with different amounts of aluminum hydroxide Aluminum hydroxide (μg) Added CL 097 (5 μg) Added HBV antigens (10 μg) Normal saline 20% sodium citrate Normal solution 20% sodium citrate Bound Free Bound Free Bound Free Bound Free 0 0 4.98 0 4.98 0 10.8 0 10 200 4.23 0.75 2.56 2.42 8.8 0 0 8.8 500 4.95 0.03 2.05 2.93 10 0 0 10 750 4.73 0.25 2.04 2.94 10.1 0 0 10 1000 4.93 0.05 2.78 2.2 10.2 0 0 10.2 HBV, hepatitis B virus.

1. Introduction Virus-related cancers are a leading cause of death in Caribbean countries,1,2 and have been identified as an important public health problem by the French West Indies cancer registries. The most common of these viruses associated with cancer are the human papillomavirus (HPV) types designated as ‘high risk’ (HR); these are implicated in 99% of cervical cancer, in 40–80% of anogenital cancers other than cervical, and in approximately 25% of head and neck cancers.3–9 The most prevalent HR HPV types in invasive cervical cancers worldwide are types 16 and 18, which are the primary targets of current HPV vaccination programs.10–12 Although the prevalence of HPV infection and incidence rates of cervical cancer are high in the Caribbean, limited data are available on the prevalence of HR HPV cervical infections among healthy Caribbean women.1,2,10,13–17 In addition, very few studies have been performed to describe the HR HPV type distribution in this population. Investigators of the African Caribbean Cancer Consortium have reported a high prevalence rate of HPV 45 rather than HPV 16 or 18 in Tobago, Jamaica, and Barbados.14–16 No data are available for the French West Indies. The recent US Food and Drug Administration (FDA) approval of a second-generation HPV vaccine that targets five HR HPV types (i.e., 31, 33, 45, 52, and 58) in addition to 16 and 18 warrants the evaluation of HR HPV cervical infections in the whole Caribbean.18 This new prophylactic vaccine may be more appropriate in Caribbean vaccination programs for the prevention of cervical cancer.10,19

eneration HPV vaccine that targets five HR HPV types (i.e., 31, 33, 45, 52, and 58) in addition to 16 and 18 warrants the evaluation of HR HPV cervical infections in the whole Caribbean.18 This new prophylactic vaccine may be more appropriate in Caribbean vaccination programs for the prevention of cervical cancer.10,19 Guadeloupe, an overseas department of France, is the largest island of the French West Indies (405 000 inhabitants, mostly Black Caribbean persons of African and European descent). A retrospective study was conducted on the island of Guadeloupe in order to assess the prevalence of cervical HR HPV infections and the type distribution of HR HPV among healthy women in this Caribbean country. 2. Patients and methods 2.1. Data and sample collection The details of consecutive non-pregnant women who attended cervical cancer screening and had HPV genotyping performed at the largest pathology laboratory on the island from January 1, 2013 to December 31, 2014 were recorded. All cases with available HPV genotyping results of liquid-based cytology samples were included in the study.

f consecutive non-pregnant women who attended cervical cancer screening and had HPV genotyping performed at the largest pathology laboratory on the island from January 1, 2013 to December 31, 2014 were recorded. All cases with available HPV genotyping results of liquid-based cytology samples were included in the study. Pathology charts were reviewed retrospectively using the computerized database of the laboratory and DIAMIC software. The following data were recorded: resident status to confirm permanent residence on the island of Guadeloupe during the study period, date of birth, date of sampling, and cytology results according to the Bethesda System 2001.20 Samples were classified as atypical squamous cells of undetermined significance (ASC-US), low-grade squamous intraepithelial lesion (LSIL), or high-grade squamous intraepithelial lesion (HSIL). 2.2. HPV testing and genotyping The genotyping analysis of liquid-based cytology specimens collected in PreservCyt (Hologic Corp.) was performed independently from the histopathological examination, using a semi-genotyping kit (Cobas 4800 HPV test, Roche Diagnostics). This qualitative multiplex assay provides specific genotyping of HPV 16 and 18 while concurrently detecting the other HR HPV types (i.e., 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) in a pooled analysis using amplification of the target DNA by PCR and nucleic acid hybridization.

00 HPV test, Roche Diagnostics). This qualitative multiplex assay provides specific genotyping of HPV 16 and 18 while concurrently detecting the other HR HPV types (i.e., 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) in a pooled analysis using amplification of the target DNA by PCR and nucleic acid hybridization. 2.3. Statistical analysis Results were recorded as numbers and frequencies. The prevalence rates of HPV infection were calculated with 95% confidence intervals (CIs). Age was described using the median and interquartile range (Q1–Q3). Stratification according to age was performed to analyze the HR HPV type distribution by cytology grade within each age group: 18–24, 25–29, 30–39, 40–49, 50–64, and ≥65 years. Cross-tabulations were analyzed by Chi-square test or Fisher’s exact test when appropriate. Comparisons of prevalence rates with other countries were based on the Chi-square test. For all analyses, two-sided p-values of less than 0.05 were considered statistically significant. Data management and the statistical analysis were performed using SPSS v.17.0 software (SPSS Inc., Chicago, IL, USA). 2.4. Ethical considerations This retrospective study was approved by the Ethics Committee for Non-interventional Research of Rouen University Hospital (registered number E2015–37).

For all analyses, two-sided p-values of less than 0.05 were considered statistically significant. Data management and the statistical analysis were performed using SPSS v.17.0 software (SPSS Inc., Chicago, IL, USA). 2.4. Ethical considerations This retrospective study was approved by the Ethics Committee for Non-interventional Research of Rouen University Hospital (registered number E2015–37). 3. Results A total of 618 consecutive HR HPV genotyping results from 618 women were collected. The median age of the study population was 42 years (Q1–Q3 32–52 years). The overall prevalence rate of HR HPV cervical infection was 36.1% (95% CI 32.3–40.0%). The distribution of HR HPV types was as follows: HPV 16 or 18 irrespective of other HPV types, 7.3% (95% CI 5.4–9.6%); other HR HPV types excluding HPV 16 and 18, 28.8% (95% CI 25.3–32.5%). Cytology results were available for 592 of the 618 cases and were classified as normal for 447 women, ASC-US for 106 women, LSIL for 31 women, and HSIL for eight women. Interestingly, the overall prevalence rate of HR HPV cervical infection increased significantly (p < 0.001) with the severity of cytology grade from normal to LSIL, as did the prevalence rate of HR HPV other than 16 or 18 infection. These varied from 25.1% (95% CI 21.1–29.3%) and 19.7% (95% CI 16.1–23.7%) for normal cytology to 65.1% (95% CI 55.2–74.1%) and 53.8% (95% CI43.8–53.5%) in ASC-US and 77.4% (95% CI 58.9–90.4%) and 67.7% (95% CI 48.6–83.3%) in LSIL (Table 1). The HSIL group was not considered for the analysis because of its small size (n = 8).

varied from 25.1% (95% CI 21.1–29.3%) and 19.7% (95% CI 16.1–23.7%) for normal cytology to 65.1% (95% CI 55.2–74.1%) and 53.8% (95% CI43.8–53.5%) in ASC-US and 77.4% (95% CI 58.9–90.4%) and 67.7% (95% CI 48.6–83.3%) in LSIL (Table 1). The HSIL group was not considered for the analysis because of its small size (n = 8). Table 2 shows the HR HPV type distribution by country (i.e., Guadeloupe, France, developed regions, and Tobago) for normal cytology; data were obtained from Heard et al.,21 the World Health Organization/Institut Català d’Oncologia (WHO/ICO) 2010 report,10 and a secondary analysis of data published previously from the Tobago study performed by Ragin et al.14 The overall prevalence rate of cervical HR HPV infection, as well as the prevalence rate of HR HPV other than 16 or 18 infection, was significantly higher in Guadeloupe than in France and in developed regions (p < 0.001). Conversely, the crude prevalence of HPV 16 and/or 18 cervical infection was not significantly different between Guadeloupe and other countries. Interestingly, data from the Caribbean island of Tobago were of the same order of magnitude as those recorded in Guadeloupe.

han in France and in developed regions (p < 0.001). Conversely, the crude prevalence of HPV 16 and/or 18 cervical infection was not significantly different between Guadeloupe and other countries. Interestingly, data from the Caribbean island of Tobago were of the same order of magnitude as those recorded in Guadeloupe. Figure 1 presents the crude prevalence of overall HR HPV, HPV 16 and/or 18, and HR HPV other than 16 or 18 cervical infections by grade of cytology for each age group. Regardless of the cytology grade, the overall rate of HR HPV cervical infection decreased with age for women older than 29 years, except for the age group ≥65 years. The largest drop was observed in ASC-US for women aged 50–64 years (i.e., overall HR HPV of 25% (95% CI 8.7–49.1), HPV 16 and/or 18 of 10% (95% CI 1.2–31.7), HR HPV other than 16 or 18 of 15% (95% CI 3.2–37.9) versus 74% (p < 0.001), 12% and 62% (p < 0.001), respectively, for younger women). Interestingly, whatever the age group, the cervical infection rate of HR HPV other than 16 or 18 was always higher than the corresponding rate of HPV 16 and/or 18 infection for each cytology grade from normal to LSIL. This was particularly obvious for the latter cytology grade, for which the rate of HPV 16 and/or 18 cervical infection was 0% in all age groups except for women aged 30–39 years (i.e., 27.3%).

or 18 was always higher than the corresponding rate of HPV 16 and/or 18 infection for each cytology grade from normal to LSIL. This was particularly obvious for the latter cytology grade, for which the rate of HPV 16 and/or 18 cervical infection was 0% in all age groups except for women aged 30–39 years (i.e., 27.3%). 4. Discussion This study is the first report of HR HPV cervical infection among healthy women from the general population of the French West Indies. Moreover this is the first study describing the HR HPV distribution type among age groups and cervical cytology grades in healthy Caribbean women. A high overall HR HPV cervical infection crude prevalence of 36.1% (95% CI 32.3–40.0) was demonstrated. This high prevalence rate is in accordance with those of all previous Caribbean studies, in which the crude prevalence of HR HPV cervical infection has varied between 35.4% in Tobago2,14 and 70.8% in Jamaica.15 As in previous worldwide studies and in the largest French study, it was found that the crude prevalence of HR HPV cervical infection was highest in young women, decreasing in women of older age,10,21,22 but increasing slightly in women ≥65 years of age compared to women aged 50–64 years.21

A high overall HR HPV cervical infection crude prevalence of 36.1% (95% CI 32.3–40.0) was demonstrated. This high prevalence rate is in accordance with those of all previous Caribbean studies, in which the crude prevalence of HR HPV cervical infection has varied between 35.4% in Tobago2,14 and 70.8% in Jamaica.15 As in previous worldwide studies and in the largest French study, it was found that the crude prevalence of HR HPV cervical infection was highest in young women, decreasing in women of older age,10,21,22 but increasing slightly in women ≥65 years of age compared to women aged 50–64 years.21 The main result of this study is the very high prevalence rate of HR HPV infection with genotypes other than 16 and 18 in Guadeloupe, irrespective of age and the cytology grade from normal to LSIL. The higher prevalence rate of infections caused by HR HPV types other than 16 or 18 compared to the rate for HPV 16 and/or 18 has been highlighted in previous studies from developed regions,10,22 as well as in the large French series published by Heard et al.,21 but this was not of the same order of magnitude as in the present study. Indeed, in the present study, the crude prevalence of HR HPV cervical infections with genotypes other than 16 or 18 was found to vary from 19.7% (95% CI 16.1–23.7%) for normal cytology to 53.8% (95% CI 43.8–53.5%) in ASC-US and 67.7% (95% CI 48.6–83.3%) in LSIL. These results are in accordance with those of a previous study in Tobago, which demonstrated a crude prevalence rate of 18.1% (95% CI 13.1–24.1%) for infection with HR HPV types other than 16 or 18 versus 2.5% for infection with HPV 16 and/or 18 in healthy women with normal cytology (Table 2).14

CI 48.6–83.3%) in LSIL. These results are in accordance with those of a previous study in Tobago, which demonstrated a crude prevalence rate of 18.1% (95% CI 13.1–24.1%) for infection with HR HPV types other than 16 or 18 versus 2.5% for infection with HPV 16 and/or 18 in healthy women with normal cytology (Table 2).14 The recent worldwide study by Serrano et al., including 10 575 invasive cervical cancer cases from 48 countries except the Caribbean, was performed in order to estimate the relative contribution of the nine HPV types targeted in the second-generation HPV vaccine; this study confirmed the variation in relative contribution of HPV 16/18 among geographical regions.19 Thus the recently FDA-approved nine-valent HPV vaccine, which covers five additional HR HPV types (i.e., 31, 33, 45, 52, and 58) to those covered by current HPV vaccines, could help to increase protection against HPV infection-related cancers, particularly in regions with high prevalence rates of HR HPV cervical infection with genotypes other than 16 or 18, such as the Caribbean. This is supported by the recent report on 131 invasive cervical cancers from Martinique (French West Indies), which showed a slightly heterogeneous distribution of HR HPV types: HPV 16 (47%), HPV 35 (14%), HPV 18 (10%), HPV 33 (8%), HPV 45 (6%), and HPV 51 (6%).23

h genotypes other than 16 or 18, such as the Caribbean. This is supported by the recent report on 131 invasive cervical cancers from Martinique (French West Indies), which showed a slightly heterogeneous distribution of HR HPV types: HPV 16 (47%), HPV 35 (14%), HPV 18 (10%), HPV 33 (8%), HPV 45 (6%), and HPV 51 (6%).23 As with all retrospective studies, this study has various possible biases. To limit the selection bias due to the single-center design, all consecutive HPV genotyping results registered in the largest pathology laboratory of Guadeloupe, which handles more than 10 000 Pap smears per year from women living in all parts of the island, were included exhaustively. The absence of major selection bias in this series is supported by the similarities in clinical features (i.e., median age 42 years) and in the crude prevalence of overall HR HPV, HR HPV other than 16 or 18, and HPV 16 or 18 infection in this series compared to those found in previous Caribbean studies.14–16

austively. The absence of major selection bias in this series is supported by the similarities in clinical features (i.e., median age 42 years) and in the crude prevalence of overall HR HPV, HR HPV other than 16 or 18, and HPV 16 or 18 infection in this series compared to those found in previous Caribbean studies.14–16 In conclusion, the present study showed a high prevalence rate of HR HPV cervical infection of 36.1% in healthy women from Guadeloupe, attributable to a high rate of infection with HR HPV genotypes other than 16 or 18 (i.e., 28.8%). These findings suggest that larger Caribbean studies should be performed in order to assess the potential impact of the new HPV vaccine on the Caribbean HPV ecology and related cervical lesions. This second-generation HPV vaccine might potentially be helpful in the development of an appropriate Caribbean vaccination program for the prevention of cervical cancer, which is a current public health problem in this region. ☆ Nadège Cordel and Camille Ragin are members of the African Caribbean Cancer Consortium.. Funding: None. Ethical approval: E2015–37. Conflict of interest: None. Figure 1 Crude prevalence of overall HR HPV, HPV 16 and/or 18, and HR HPV other than 16 or 18 cervical infections by grade of cytology for each age group. Table 1 Prevalence of HR HPV cervical infections by cytology grade Cytology grade

☆ Nadège Cordel and Camille Ragin are members of the African Caribbean Cancer Consortium.. Funding: None. Ethical approval: E2015–37. Conflict of interest: None. Figure 1 Crude prevalence of overall HR HPV, HPV 16 and/or 18, and HR HPV other than 16 or 18 cervical infections by grade of cytology for each age group. Table 1 Prevalence of HR HPV cervical infections by cytology grade Cytology grade Normal (n = 447) ASC-US (n = 106) LSIL (n = 31) p-Value n % 95% CI n % 95% CI n % 95% CI Overall HR HPV 112 25.1 21.1–29.3 69 65.1 55.2–74.1 24 77.4 58.9–90.4 <0.001 HR HPV 16 and/or 18 irrespective of other HR HPV types 24 5.4 3.5–7.9 12 11.3 6.0–18.9 3 9.7 2.0–25.8 0.060 Other HR HPV excluding 16 and 18 88 19.7 16.1–23.7 57 53.8 43.8–63.5 21 67.7 48.6–83.3 <0.001 HR, high-risk; HPV, human papillomavirus; ASC-US, atypical squamous cell of undetermined significance; LSIL, low-grade squamous intraepithelial lesion; CI, confidence interval. Table 2 HR HPV type distribution by country for normal cytology Guadeloupe (French West Indies) France (Heard et al., 2013) Developed regions (WHO/ICO 2010) Tobago (Ragin et al., 2007) % 95% CI % 95% CI p-Valuea % 95% CI p-Valuea % 95% CI p-Valuea HR HPVb 25.1 21.1–29.3 13.7 11.7–15.6 <0.001 10.3 10.2–10.4 <0.001 20.6 15.3–26.8 0.192 HPV 16 and/or 18c 5.4 3.5–7.9 3.9 2.8–5.1 0.089 3.6 3.5–3.7 0.017 2.5 0.8–5.6 0.089 HR HPV other than 16 and/or 18d 19.7 16.1–23.7 9.7 8.0–11.4 <0.001 6.0 5.9–6.1 <0.001 18.1 13.1–24.1 0.616 HR, high-risk; HPV, human papillomavirus; WHO/ICO, World Health Organization/Institut Català d’Oncologia; CI, confidence interval. a Comparison to Guadeloupe.

% 95% CI % 95% CI p-Valuea % 95% CI p-Valuea % 95% CI p-Valuea HR HPVb 25.1 21.1–29.3 13.7 11.7–15.6 <0.001 10.3 10.2–10.4 <0.001 20.6 15.3–26.8 0.192 HPV 16 and/or 18c 5.4 3.5–7.9 3.9 2.8–5.1 0.089 3.6 3.5–3.7 0.017 2.5 0.8–5.6 0.089 HR HPV other than 16 and/or 18d 19.7 16.1–23.7 9.7 8.0–11.4 <0.001 6.0 5.9–6.1 <0.001 18.1 13.1–24.1 0.616 HR, high-risk; HPV, human papillomavirus; WHO/ICO, World Health Organization/Institut Català d’Oncologia; CI, confidence interval. a Comparison to Guadeloupe. b At least one HR HPV type among 16, 18, and other HR types (i.e., 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68). c HPV 16 and/or 18 irrespective of other HR HPV types. d Other HR HPV types, excluding HPV 16 and 18.

1 Introduction In April 2009, an outbreak of respiratory disease in young people was reported in Mexico City,1 and the etiologic agent was identified as influenza A virus H1N1pdm09.2 This outbreak and its consequences compelled the Ministry of Health of Mexico and the US National Institute of Allergy and Infectious Diseases (NIAID) to collaborate in establishing a Mexican Emerging Infectious Diseases Clinical Research Network (La Red). Acute respiratory infections are estimated to cause 3.9 million deaths annually,3 many of which are influenza-like illnesses (ILI). The etiology of ILI in Mexican individuals seeking medical care is not well-defined. It has been reported in other populations that multiple respiratory viruses such as parainfluenza virus, rhinovirus, adenovirus, metapneumovirus, respiratory syncytial virus (RSV), and coronavirus can cause ILI,4 but the virologic etiology varies greatly among geographic locales, age groups, seasons, and years.5, 6, 7, 8, 9, 10, 11, 12, 13 Some studies have suggested that specific symptoms could be used for the clinical prediction of the etiology of ILI,14, 15, 16, 17, 18 though most of these studies only compared influenza to non-influenza etiology.

ogy varies greatly among geographic locales, age groups, seasons, and years.5, 6, 7, 8, 9, 10, 11, 12, 13 Some studies have suggested that specific symptoms could be used for the clinical prediction of the etiology of ILI,14, 15, 16, 17, 18 though most of these studies only compared influenza to non-influenza etiology. Although the outcomes of subjects infected with various influenza subtypes have been well documented,19, 20, 21 there is relatively little description of outcomes from other viral etiologies of ILI. Shlomai et al. compared outcomes of those hospitalized with an ILI, and found mortality of 4.6% in influenza patients vs. 7% in influenza-negative ILI patients.18 As further demonstration of the burden that non-influenza etiological agents contribute to ILIs, Widmer et al. reported that RSV hospitalization rates exceeded those of influenza (15.01 and 11.81 per 10 000, respectively).22 It is important to know the characteristics and seasonal behavior of the different types of virus causing ILI. Although current clinical diagnostic tests of all viruses are high in cost, improved diagnosis would allow a more judicious use of empirical wide-spectrum antibiotic treatment, which in many cases could be suspended, and thus reduce the cost of hospitalization. Recognizing the need to investigate both influenza and non-influenza ILIs in the Mexican population, La Red implemented a study with the objective of describing the etiology, symptoms, and outcomes of subjects presenting with ILI in Mexico City.

It is important to know the characteristics and seasonal behavior of the different types of virus causing ILI. Although current clinical diagnostic tests of all viruses are high in cost, improved diagnosis would allow a more judicious use of empirical wide-spectrum antibiotic treatment, which in many cases could be suspended, and thus reduce the cost of hospitalization. Recognizing the need to investigate both influenza and non-influenza ILIs in the Mexican population, La Red implemented a study with the objective of describing the etiology, symptoms, and outcomes of subjects presenting with ILI in Mexico City. 2 Methods 2.1 Study design and sites Beginning in April 2010, this hospital-based observational cohort study was conducted at five hospitals in Mexico City. These hospitals are located in the south of Mexico City, and include a general hospital, a tertiary care hospital that serves mainly those with respiratory problems, a tertiary care hospital that serves the metabolic/surgical population, and two pediatric centers. These centers were chosen given their capacity to conduct clinical research.

ocated in the south of Mexico City, and include a general hospital, a tertiary care hospital that serves mainly those with respiratory problems, a tertiary care hospital that serves the metabolic/surgical population, and two pediatric centers. These centers were chosen given their capacity to conduct clinical research. 2.2 Case definition and study population The study population included subjects aged >3 months who presented with an ILI. ILI was defined by the presence of at least one respiratory symptom (e.g., shortness of breath, postnasal drip, and cough) and one of the two following criteria: (1) fever (≥38 °C by any method) on examination, or participant-reported fever (≥38 °C), or feverishness in the past 24 h; (2) one or more non-respiratory symptoms (e.g., malaise, headache, myalgia, and chest pain). The subjects included were those who sought medical attention at our centers and agreed to participate in the study.

38 °C by any method) on examination, or participant-reported fever (≥38 °C), or feverishness in the past 24 h; (2) one or more non-respiratory symptoms (e.g., malaise, headache, myalgia, and chest pain). The subjects included were those who sought medical attention at our centers and agreed to participate in the study. 2.3 Study procedures Demographic data were collected at enrollment. A nasopharyngeal swab (Copan, Brescia, Italy), or a nasal aspirate in children, was obtained for PCR detection of respiratory pathogens and sample storage. The samples were placed in transport medium, maintained at 4 °C, and sent within 72 h to a central facility (Molecular Biology Laboratory, Infectious Diseases Department, INNSZ, Mexico City), where they were stored at −70 °C. Clinical laboratory tests were carried out at enrollment. If available, the results of tests taken for standard clinical care (chest X-ray, bacteriological cultures, arterial blood gases, and liver tests) were also abstracted from the medical records. Subjects were seen again on day 28, and follow-up information was also obtained by phone on days 14 and 60. At each visit, clinical information (symptoms, chronic medical conditions, previous treatment, impact on daily function, hospitalizations, and death) was assessed. 2.4 Virology All samples were tested by real-time reverse transcriptase PCR for influenza A following the Centers for Disease Control and Prevention (CDC) protocol (CDC 2009), as described previously.23

2.3 Study procedures Demographic data were collected at enrollment. A nasopharyngeal swab (Copan, Brescia, Italy), or a nasal aspirate in children, was obtained for PCR detection of respiratory pathogens and sample storage. The samples were placed in transport medium, maintained at 4 °C, and sent within 72 h to a central facility (Molecular Biology Laboratory, Infectious Diseases Department, INNSZ, Mexico City), where they were stored at −70 °C. Clinical laboratory tests were carried out at enrollment. If available, the results of tests taken for standard clinical care (chest X-ray, bacteriological cultures, arterial blood gases, and liver tests) were also abstracted from the medical records. Subjects were seen again on day 28, and follow-up information was also obtained by phone on days 14 and 60. At each visit, clinical information (symptoms, chronic medical conditions, previous treatment, impact on daily function, hospitalizations, and death) was assessed. 2.4 Virology All samples were tested by real-time reverse transcriptase PCR for influenza A following the Centers for Disease Control and Prevention (CDC) protocol (CDC 2009), as described previously.23 For multi-pathogen detection, the samples were tested with the RespiFinder19 kit (previously RespiFinder Plus, from PathoFinder BV, Maastricht, The Netherlands), as described previously.24 This multiplex PCR test can detect and differentiate 15 viruses (coronavirus NL63, OC43, and 229E, human metapneumovirus, influenza A, influenza A H5N1, influenza B, parainfluenza virus types 1 to 4, RSV A and B, rhinovirus, and adenovirus), as well as four bacteria (Bordetella pertussis, Chlamydophila pneumoniae, Legionella pneumophila, and Mycoplasma pneumoniae). The analytical sensitivity (reported by the manufacturer) varies between 5 and 50 copies per reaction for most targets.25

ainfluenza virus types 1 to 4, RSV A and B, rhinovirus, and adenovirus), as well as four bacteria (Bordetella pertussis, Chlamydophila pneumoniae, Legionella pneumophila, and Mycoplasma pneumoniae). The analytical sensitivity (reported by the manufacturer) varies between 5 and 50 copies per reaction for most targets.25 For the samples with discordant real-time reverse transcriptase PCR results and multiplex PCR for influenza, the samples were further tested with a third method, based on a final-point nested PCR protocol.26 For these samples, the result of this test was considered the final result. 2.5 Statistical analysis Data were analyzed with the Statistical Analysis System (SAS Institute, Cary, NC, USA). Median (range) and percentage were used to summarize quantitative and qualitative variables, respectively. To discard the role of chance, the Mann–Whitney test and the Chi-square or Fisher's exact test were used as appropriate. Univariate logistic regression analyses were performed to evaluate the effect of each baseline covariate on the risk of death. Odds ratios (OR) and the corresponding 95% confidence intervals (CI) were reported based on the fitted logistic regression models. All p-values were two-sided, with no correction for multiple comparisons; p < 0.05 was considered statistically significant.

uate the effect of each baseline covariate on the risk of death. Odds ratios (OR) and the corresponding 95% confidence intervals (CI) were reported based on the fitted logistic regression models. All p-values were two-sided, with no correction for multiple comparisons; p < 0.05 was considered statistically significant. 3 Results 3.1 Patients From April 11, 2010 through April 10, 2011, 1065 subjects were enrolled in this study. A total of 86 withdrew or were withdrawn from the study (Figure 1 ). Characteristics of the study participants are listed in Table 1 . Just under a quarter of the study subjects were children and 58.8% were female. Four hundred thirty-six (40.9%) subjects were hospitalized. Asthma and cardiovascular disease were the most common chronic medical conditions, though neither condition accounted for more than 17% of subjects.Figure 1 Enrollment and follow-up. Table 1 Baseline characteristics

3 Results 3.1 Patients From April 11, 2010 through April 10, 2011, 1065 subjects were enrolled in this study. A total of 86 withdrew or were withdrawn from the study (Figure 1 ). Characteristics of the study participants are listed in Table 1 . Just under a quarter of the study subjects were children and 58.8% were female. Four hundred thirty-six (40.9%) subjects were hospitalized. Asthma and cardiovascular disease were the most common chronic medical conditions, though neither condition accounted for more than 17% of subjects.Figure 1 Enrollment and follow-up. Table 1 Baseline characteristics Demographics Total, N = 1065 Children, n = 262 Adults, n = 803 Age, years Median 33 Range 0–96 ≤18 262 (24.6%) 19–59 614 (57.6%) ≥60 189 (17.8%) Sex Male 439 (41.2%) 132 (50.4%) 307 (38.2%) Female 626 (58.8%) 130 (49.6%) 496 (61.8%) Pregnant 4 4 Ethnicity Mixed 1026 (96.4%) 252 (96.2%) 775 (96.5%) White 33 (3.1%) 10 (3.8%) 23 (2.9%) Other 5 (0.5%) 5 (0.6%) Medical history Influenza vaccination 504 (47.3%) 121 (46.2%) 383 (47.7%) Asthma 151 (14.2%) 16 (6.1%) 135 (16.8%) CVD 147 (13.8%) 17 (6.5%) 130 (16.2%) Diabetes 99 (9.3%) 99 (12.3%) Immunodeficiency 48 (4.5%) 8 (3.1%) 40 (5.0%) Renal disorder 42 (3.9%) 4 (1.5%) 38 (4.8%) COPD 36 (3.4%) 36 (4.5%) BMI categorya Obese 207 (19.9%) 23 (9.2%) 184 (23.3%) Overweight 313 (30.2%) 42 (16.1%) 273 (34.6%) Normal 417 (40.2%) 121 (48.6%) 296 (37.5%) Underweight 101 (9.7%) 65 (26.1%) 36 (4.6%) Median days since symptoms onset 1 Inpatient at baseline 436 (40.9%) 113 (43.1%) 323 (40.2%) Outpatient at baseline 629 (59.1%) 149 (56.9%) 480 (59.8) BMI, body mass index; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; SD, standard deviation.

7.5%) Underweight 101 (9.7%) 65 (26.1%) 36 (4.6%) Median days since symptoms onset 1 Inpatient at baseline 436 (40.9%) 113 (43.1%) 323 (40.2%) Outpatient at baseline 629 (59.1%) 149 (56.9%) 480 (59.8) BMI, body mass index; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; SD, standard deviation. a BMI categories: underweight = BMI < 18.5 kg/m2 (or <1 SD for children); normal weight = 18.5–24.9 kg/m2; overweight = 25–29.9 kg/m2 (>1SD for children); obese = BMI of ≥30 kg/m2 (>2 SD for children). 3.2 Etiology A total of 821 etiologic agents were identified among the 678 (64% of total) subjects from whom a pathogen was isolated. Rhinovirus was the most frequently isolated pathogen, infecting 15.3% of enrolled subjects and accounting for 25.3% of all isolates. Rhinovirus was followed by influenza, which was detected in 14.3% of subjects, and 24% of isolates. Adenovirus, coronavirus, metapneumovirus, and RSV all had an isolate frequency of 9% or greater. No pathogen was detected for 35.5% of subjects (Table 2 ).Table 2 Frequency distribution of viral isolates

of all isolates. Rhinovirus was followed by influenza, which was detected in 14.3% of subjects, and 24% of isolates. Adenovirus, coronavirus, metapneumovirus, and RSV all had an isolate frequency of 9% or greater. No pathogen was detected for 35.5% of subjects (Table 2 ).Table 2 Frequency distribution of viral isolates Pathogen Age <18a, n (%) Age 18–59a, n (%) Age ≥60a, n (%) Totala, n (%) Totalb, n (%) Adenovirus 17 (6.6) 15 (2.5) 3 (1.6) 35 (3.3) 74 (9.0) Bordetella pertussis 2 (0.8) 3 (0.5) 0 (0) 5 (0.5) 9 (1.1) Coronavirus 12 (4.7) 50 (8.2) 15 (8.2) 77 (7.3) 118 (14.4) 229E 0 11 2 13 22 NL63 2 8 1 11 18 OC43 10 31 12 53 78 Influenza 43 (16.7) 89 (14.6) 18 (9.8) 150 (14.3) 197 (24.0) A 3 5 0 8 11 A H1N1 2009 0 5 1 6 7 A H3N2 16 46 11 73 91 B 24 33 6 63 88 Metapneumovirus 11 (4.3) 22 (3.6) 9 (4.9) 42 (4.0) 78 (9.5) Mycoplasma pneumoniae 1 (0.4) 4 (0.7) 1 (0.5) 6 (0.6) 10 (1.2) Parainfluenza virus 6 (2.3) 12 (1.9) 2 (1.1) 20 (1.9) 42 (5.1) 1 0 0 0 0 3 2 0 0 0 0 3 3 6 7 1 14 24 4 0 5 1 6 12 Rhinovirus 29 (11.2) 100 (16.4) 32 (17.4) 161 (15.3) 208 (25.3) RSV 34 (13.2) 15 (2.5) 8 (4.3) 57 (5.4) 85 (10.3) A 33 15 6 54 72 B 1 0 2 3 13 Mixed infections 51 (19.8) 60 (9.9) 14 (7.6) 125 (11.9) Negative cases 52 (20.2) 238 (39.1) 83 (45.1) 373 (35.5) Total 258 (100) 609 (100) 184 (100) 1051 (100)c 821 (100) RSV, respiratory syncytial virus. a Frequency by subject. Infections with more than one virus are counted as a single category (mixed infection). b Frequency by isolates. Excludes negative cases. c Fourteen subjects did not have a baseline sample for virologic testing.

Pathogen Age <18a, n (%) Age 18–59a, n (%) Age ≥60a, n (%) Totala, n (%) Totalb, n (%) Adenovirus 17 (6.6) 15 (2.5) 3 (1.6) 35 (3.3) 74 (9.0) Bordetella pertussis 2 (0.8) 3 (0.5) 0 (0) 5 (0.5) 9 (1.1) Coronavirus 12 (4.7) 50 (8.2) 15 (8.2) 77 (7.3) 118 (14.4) 229E 0 11 2 13 22 NL63 2 8 1 11 18 OC43 10 31 12 53 78 Influenza 43 (16.7) 89 (14.6) 18 (9.8) 150 (14.3) 197 (24.0) A 3 5 0 8 11 A H1N1 2009 0 5 1 6 7 A H3N2 16 46 11 73 91 B 24 33 6 63 88 Metapneumovirus 11 (4.3) 22 (3.6) 9 (4.9) 42 (4.0) 78 (9.5) Mycoplasma pneumoniae 1 (0.4) 4 (0.7) 1 (0.5) 6 (0.6) 10 (1.2) Parainfluenza virus 6 (2.3) 12 (1.9) 2 (1.1) 20 (1.9) 42 (5.1) 1 0 0 0 0 3 2 0 0 0 0 3 3 6 7 1 14 24 4 0 5 1 6 12 Rhinovirus 29 (11.2) 100 (16.4) 32 (17.4) 161 (15.3) 208 (25.3) RSV 34 (13.2) 15 (2.5) 8 (4.3) 57 (5.4) 85 (10.3) A 33 15 6 54 72 B 1 0 2 3 13 Mixed infections 51 (19.8) 60 (9.9) 14 (7.6) 125 (11.9) Negative cases 52 (20.2) 238 (39.1) 83 (45.1) 373 (35.5) Total 258 (100) 609 (100) 184 (100) 1051 (100)c 821 (100) RSV, respiratory syncytial virus. a Frequency by subject. Infections with more than one virus are counted as a single category (mixed infection). b Frequency by isolates. Excludes negative cases. c Fourteen subjects did not have a baseline sample for virologic testing. Among the 125 subjects (11.9%) with mixed virus infections, 62 unique combinations were recorded. The most common combinations were adenovirus/rhinovirus in nine subjects (7.2% of combinations), adenovirus/metapneumovirus in eight subjects (6.4%), and influenza B/metapneumovirus in eight subjects (6.4%). Influenza was the most common virus involved in co-infections (when considering all subtypes), present in 45/125 (36%) combinations. Most combinations were limited to two viruses.

jects (7.2% of combinations), adenovirus/metapneumovirus in eight subjects (6.4%), and influenza B/metapneumovirus in eight subjects (6.4%). Influenza was the most common virus involved in co-infections (when considering all subtypes), present in 45/125 (36%) combinations. Most combinations were limited to two viruses. 3.3 Seasonal distribution of viruses The four most common pathogens (rhinovirus, influenza, coronavirus, and RSV) all had seasonal distributions with more isolates observed during cooler months (December–March). The seasonal variation of rhinovirus and RSV, however, was less pronounced than that of coronavirus and influenza (Figure 2 ). During April – November, rhinovirus accounted for 20–60% of all ILI. However, between December and March – a period during which the total number of ILI cases increased – the relative contribution of rhinovirus to ILI was only 10–20%.Figure 2 Monthly distribution of viral isolates. Total monthly enrollment (line), and the four most prevalent viruses (A). Monthly isolates by virus subtype (bars) and percent positive are demonstrated for the five most common viruses: influenza (B), coronavirus (C), respiratory syncytial virus (D), metapneumovirus (E), and rhinovirus (F). 3.4 Vaccination Previous influenza vaccination was reported for 46.2% of the children and 47.7% of the adults. We found no difference in the proportion of vaccinated and unvaccinated subjects in the persons with virus or without virus detection, neither for individual agents nor as a group.

3.3 Seasonal distribution of viruses The four most common pathogens (rhinovirus, influenza, coronavirus, and RSV) all had seasonal distributions with more isolates observed during cooler months (December–March). The seasonal variation of rhinovirus and RSV, however, was less pronounced than that of coronavirus and influenza (Figure 2 ). During April – November, rhinovirus accounted for 20–60% of all ILI. However, between December and March – a period during which the total number of ILI cases increased – the relative contribution of rhinovirus to ILI was only 10–20%.Figure 2 Monthly distribution of viral isolates. Total monthly enrollment (line), and the four most prevalent viruses (A). Monthly isolates by virus subtype (bars) and percent positive are demonstrated for the five most common viruses: influenza (B), coronavirus (C), respiratory syncytial virus (D), metapneumovirus (E), and rhinovirus (F). 3.4 Vaccination Previous influenza vaccination was reported for 46.2% of the children and 47.7% of the adults. We found no difference in the proportion of vaccinated and unvaccinated subjects in the persons with virus or without virus detection, neither for individual agents nor as a group. 3.5 Symptoms Evaluation across the four most common pathogens showed that children were more likely to have fever, productive cough, and nausea, whereas adults were more likely to have fatigue, headache, and sore throat (Table 3 ). Respiratory symptoms, such as dyspnea, were more common among those with rhinovirus or RSV, and gastrointestinal symptoms, such as nausea or diarrhea, were more common among those with influenza.Table 3 Demographics and symptoms by age and pathogen

were more likely to have fatigue, headache, and sore throat (Table 3 ). Respiratory symptoms, such as dyspnea, were more common among those with rhinovirus or RSV, and gastrointestinal symptoms, such as nausea or diarrhea, were more common among those with influenza.Table 3 Demographics and symptoms by age and pathogen Children Adults Coronavirus Influenza Rhinovirus RSV Coronavirus Influenza Rhinovirus RSV Number enrolled 12 43 29 34 65 107 132 23 Any antiviral 0% 9% 7% 6% 28% 33% 27% 35% Chronic medical condition 42% 28% 38% 26% 42% 44% 63% 74% Asthma 8% 7% 10% 3% 11% 21% 19% 26% CVD 0% 0% 10% 6% 15% 10% 20% 17% COPD 0% 0% 0% 0% 5% 6% 2% 13% DM 0% 0% 0% 0% 8% 6% 19% 13% Antibiotics 50% 44% 52% 71% 40% 48% 66% 65% Inhaled steroids 8% 9% 21% 24% 18% 20% 14% 35% Systemic steroids 25% 9% 17% 21% 22% 24% 28% 43% Pregnant 0% 0% 0% 0% 0% 1% 0% 0% Non-smoker 0% 0% 0% 0% 72% 51% 60% 65% Current/former smoker 0% 0% 0% 0% 28% 49% 40% 35% Seasonal flu vaccine 50% 47% 45% 38% 60% 39% 39% 52% Outpatient 67% 81% 38% 35% 68% 67% 52% 57% Hospitalized 33% 19% 62% 65% 32% 33% 48% 43% Symptoms Rales/crepitations 33% 16% 55% 79% 25% 29% 39% 26% Wheezing 25% 16% 21% 29% 12% 22% 25% 35% Fever 67% 95% 69% 100% 45% 79% 65% 70% Productive cough 75% 65% 83% 88% 52% 59% 58% 57% Diarrhea 0% 9% 0% 12% 6% 11% 4% 4% Dyspnea 0% 16% 31% 38% 22% 33% 39% 43% Fatigue 0% 58% 34% 32% 71% 64% 41% 57% Headache 0% 51% 21% 12% 74% 75% 55% 39% Nausea 17% 42% 14% 24% 15% 21% 9% 22% Red eyes 0% 49% 17% 9% 37% 27% 13% 17% Sore throat 25% 60% 14% 15% 65% 61% 47% 52% COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; DM, diabetes mellitus; RSV, respiratory syncytial virus.

58% 34% 32% 71% 64% 41% 57% Headache 0% 51% 21% 12% 74% 75% 55% 39% Nausea 17% 42% 14% 24% 15% 21% 9% 22% Red eyes 0% 49% 17% 9% 37% 27% 13% 17% Sore throat 25% 60% 14% 15% 65% 61% 47% 52% COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; DM, diabetes mellitus; RSV, respiratory syncytial virus. 3.6 Outcomes Hospitalization was required for 585 subjects (54.9%) (Table 4 ). A lower percentage of subjects infected with coronavirus and influenza were hospitalized (32% and 29%) than those infected with rhinovirus and RSV (51% and 56%, respectively) (Table 3). In the pediatric population, RSV infections had the highest rate of hospitalization (76.5%), as well as the highest absolute number of hospitalizations (n = 26, 24.4% of all hospitalized pediatric subjects). Mixed infections (n = 25, 20.7%), rhinovirus (n = 17, 14.0%), and influenza (n = 11, 9.1%) also contributed significantly to hospitalization in the pediatric cohort. Rhinovirus was also the most common pathogen isolated from adults and the elderly who were hospitalized, followed by mixed infections, influenza, and metapneumovirus.Table 4 Hospitalizations and deaths by pathogen

), and influenza (n = 11, 9.1%) also contributed significantly to hospitalization in the pediatric cohort. Rhinovirus was also the most common pathogen isolated from adults and the elderly who were hospitalized, followed by mixed infections, influenza, and metapneumovirus.Table 4 Hospitalizations and deaths by pathogen Pathogen Hospitalized % (n) Death % (n) Age <18 Age 18–59 Age ≥60 Age <18 Age 18–59 Age ≥60 Adenovirus 41.8% (7) 66.7% (10) 100% (3) 6.7% (1) Bordetella pertussis 100% (2) 33.3% (1) Coronavirus 33.3% (4) 24% (12) 73.3% (11) 2% (1) 13.3% (2) 229E 45.5% (5) 100% (2) NL63 25% (2) 100% (1) OC43 40% (4) 16.1% (5) 66.7% (8) 3.2% (1) 16.7% (2) Influenza 25.6% (11) 43.8% (39) 72.2% (13) 11.1% (2) A 60% (3) A H1N1 2009 80% (4) A H3N2 25% (4) 50% (23) 72.7% (8) 9.1% (1) B 29.2% (7) 27.3% (9) 83.3% (5) 16.7% (1) Metapneumovirus 63.6% (7) 50% (11) 88.9% (8) 22.2% (2) Mycoplasma pneumoniae 100% (4) 100% (1) 100% (1) Parainfluenza virus 66.7% (4) 41.7% (5) 100% (2) 16.7% (1) 1 2 3 66.7% (4) 42.9% (3) 100% (1) 16.7% (1) 4 40% (2) 100% (1) Rhinovirus 58.6% (17) 59% (59) 81.3% (26) 7% (7) 12.5% (4) Respiratory syncytial virus 76.5% (26) 53.3% (8) 100% (8) 5.9% (2) 25% (2) A 75.8% (25) 53.3% (8) 100% (6) 6.1% (2) 33.3% (2) B 100% (1) 100% (2) Mixed infections 49% (25) 56.7% (34) 100% (14) 10% (6) 7.1% (1) Negative cases 34.6% (18) 51.7% (123) 88% (73) 4.6% (11) 27.7% (23) Total 46.2% (121) 49.8% (306) 84.1% (159) 1.5% (4) 4.2% (26) 19% (36)

76.5% (26) 53.3% (8) 100% (8) 5.9% (2) 25% (2) A 75.8% (25) 53.3% (8) 100% (6) 6.1% (2) 33.3% (2) B 100% (1) 100% (2) Mixed infections 49% (25) 56.7% (34) 100% (14) 10% (6) 7.1% (1) Negative cases 34.6% (18) 51.7% (123) 88% (73) 4.6% (11) 27.7% (23) Total 46.2% (121) 49.8% (306) 84.1% (159) 1.5% (4) 4.2% (26) 19% (36) Of the 1065 patients enrolled, 66 died (6.2%). Of the patients who died, 54% had no pathogen identified. Rhinovirus was the most common etiologic agent identified in subjects who died (37.9% of deaths with a pathogen identified, and 17.5% of all deaths). Only four pediatric subjects died: two with RSV, one with parainfluenza, and one with Mycoplasma. Subjects ≥60 years of age infected with coronavirus, influenza, metapneumovirus, rhinovirus, or RSV all had mortality rates exceeding 10%. The highest mortality rate (27.7%, n = 23) was seen in the cohort of patients ≥60 years old with no identified pathogen, followed by rhinovirus. Mixed viral infections did not have an appreciably larger rate of hospitalization or death.

fluenza, metapneumovirus, rhinovirus, or RSV all had mortality rates exceeding 10%. The highest mortality rate (27.7%, n = 23) was seen in the cohort of patients ≥60 years old with no identified pathogen, followed by rhinovirus. Mixed viral infections did not have an appreciably larger rate of hospitalization or death. Subjects who died were more likely to have elevated creatinine, lactate dehydrogenase, and C-reactive protein, lower lymphocyte counts, and higher neutrophil counts (Table 5 ). Subjects who died were more likely to have underlying medical conditions (except asthma). Subjects on systemic steroids prior to enrollment were also more likely to die. Headaches and sore throats were seen more commonly in non-fatal cases, whereas rales (representing lower respiratory tract disease) were more likely in fatal cases. A productive cough did not seem to differentiate fatal from non-fatal cases.Table 5 Outcomes

mic steroids prior to enrollment were also more likely to die. Headaches and sore throats were seen more commonly in non-fatal cases, whereas rales (representing lower respiratory tract disease) were more likely in fatal cases. A productive cough did not seem to differentiate fatal from non-fatal cases.Table 5 Outcomes Variable at baseline Patients who survived Patients who died Odds ratio p-Value Median (range), n = 999 Median (range), n = 66 Estimate (95% CI) Creatinine phosphokinase (U/l) 95 (9–7160) 90.5 (10–1840) 1.04 (1.00, 1.08) 0.066 Creatinine (mg/dl) 0.6 (0–14) 1.1 (0.1–9.7) 1.37 (1.19, 1.58) <0.001 C-reactive protein (mg/l) 1.4 (0–36) 14.25 (1.6–34.1) 1.12 (1.09, 1.16) <0.001 Hematocrit (%) 41.7 (14.3–76.1) 34.9 (13.2–57.5) 0.73 (0.66, 0.8) <0.001 Hemoglobin (g/dl) 14.1 (4.1–25.4) 11.7 (4.7–19) 0.73 (0.66, 0.8) <0.001 Lactate dehydrogenase (IU/l) 224 (26–4003) 338 (31–5175) 1.19 (1.08, 1.32) 0.001 Lymph (%) 18 (0–92) 7 (1–35) 0.92 (0.90, 0.95) <0.001 Lymph absolute (×109/l) 1.4 (0–35.1) 0.7 (0.02–6.8) 0.92 (0.88, 0.95) <0.001 Neutrophil (%) 70 (0–97) 87 (16–98) 1.06 (1.04, 1.08) <0.001 Neutrophil absolute (×109/l) 5.7 (0–26.9) 8.6 (0.27–27.2) 1.01 (1.00, 1.01) 0.001a Platelets (×109/l) 231 (10–744) 140 (3–551) 0.91 (0.88, 0.94) <0.001 White blood count (×109/l) 8.5 (0.1–54) 10.4 (0.3–29.2) 1.03 (0.99, 1.07) 0.179 Patients who survived Patients who died Odds ratio p-Value Percent Percent Estimate (95% CI) Any antiviral medications 21.2 39.4 2.41 (1.44, 4.05) <0.001a Inhaled steroids 18.4 18.2 1.00 (0.52, 1.91) 1.00 Systemic steroids 23.7 36.4 1.88 (1.11, 3.1) 0.02 History of smoking 30.3 50.0 2.30 (1.39, 3.79) <0.001a Medical history Asthma 15.0 1.5 0.09 (0.01, 0.63) 0.02a Cardiovascular disorder 12.7 30.3 2.99 (1.71, 5.21) <0.001a COPD 3.0 9.1 3.23 (1.29, 8.06) 0.01a Diabetes 8.7 18.2 2.33 (1.20, 4.52) 0.01 Immunodeficiency 3.7 16.7 5.20 (2.52, 10.75) <0.001 Renal disorder 3.5 10.6 3.27 (1.39, 7.67) 0.01a Signs and symptoms Productive cough 62.9 51.5 0.63 (0.38, 1.03) 0.07 Diarrhea 7.5 3.0 0.39 (0.09, 1.60) 0.19 Dyspnea 32.5 39.4 1.35 (0.81, 2.25) 0.25 Fatigue 51.8 42.4 0.69 (0.42, 1.14) 0.14 Fever 70.9 75.8 1.28 (0.72, 2.29) 0.40 Headache 53.2 21.2 0.24 (0.13, 0.43) <0.001 Nausea 18.9 1.5 0.07 (0.01, 0.48) 0.01a Rales/crepitations 36.7 83.3 8.61 (4.45, 16.66) <0.001 Red eyes 21.6 4.5 0.17 (0.05, 0.56) <0.001a Sore throat 46.4 21.2 0.31 (0.17, 0.57) <0.001 Wheezing 23.2 9.1 0.33 (0.14, 0.78) 0.01a CI, confidence interval; COPD, chronic obstructive pulmonary disease.

24 (0.13, 0.43) <0.001 Nausea 18.9 1.5 0.07 (0.01, 0.48) 0.01a Rales/crepitations 36.7 83.3 8.61 (4.45, 16.66) <0.001 Red eyes 21.6 4.5 0.17 (0.05, 0.56) <0.001a Sore throat 46.4 21.2 0.31 (0.17, 0.57) <0.001 Wheezing 23.2 9.1 0.33 (0.14, 0.78) 0.01a CI, confidence interval; COPD, chronic obstructive pulmonary disease. a When analysis for deaths attributed to ILI, the p-value was no longer <0.05. 4 Discussion This study presents the most thorough description of the causes of ILI in Mexico City to date. Using a multiplex PCR test, one or more pathogens were identified in 65% of enrolled subjects. The most common pathogen identified in the study was rhinovirus (25% of all isolates), and influenza was the second most common pathogen (24% of all isolates). This is consistent with other series of adult ILI in various geographic regions, where rhinovirus has been reported to account for 25–50% of adult ILI isolates, and influenza for 14–23%.10, 13, 27, 28 Coronavirus – most commonly OC43, but also NL63 and 2229E – was also a significant contributor to the burden of ILI (14% of all isolates). Recent studies have reported coronavirus to be the third to fifth most common pathogen.10, 13

ed to account for 25–50% of adult ILI isolates, and influenza for 14–23%.10, 13, 27, 28 Coronavirus – most commonly OC43, but also NL63 and 2229E – was also a significant contributor to the burden of ILI (14% of all isolates). Recent studies have reported coronavirus to be the third to fifth most common pathogen.10, 13 Among children in this study, the most common pathogen was influenza, followed by RSV and rhinovirus. Similar significant contributions of RSV and influenza have been noted in prior studies.29, 30, 31 Rhinovirus, however, was rarely tested in previous studies. In one other prospective cohort study of children with lower respiratory tract infections, rhinovirus was reported as second only to RSV in causing hospitalizations.32 The frequency at which it was isolated in our hospitalized children (14%, 17/121) suggests that rhinovirus is a previously under-appreciated pathogen for this age group. Rhinovirus was a cause of significant morbidity in all age groups of our study population, such that 60–80% of individuals infected with rhinovirus required hospitalization. Rhinovirus was also the most common isolate from subjects who died (isolated from 11/66 subjects). Given the study design, a clear attribution to the cause of hospitalization or death is not possible.

age groups of our study population, such that 60–80% of individuals infected with rhinovirus required hospitalization. Rhinovirus was also the most common isolate from subjects who died (isolated from 11/66 subjects). Given the study design, a clear attribution to the cause of hospitalization or death is not possible. Coronavirus species made up 14% of all isolates in our population. Coronavirus can present with a similar syndrome to influenza.33 Twenty-two percent of all subjects with isolated coronavirus were hospitalized, but in the population aged ≥60 years this increased to 73%. While our study design precludes the determination of exact hospitalization rates as a result of coronavirus infections, the hospitalization rates were similar to those for influenza.

nty-two percent of all subjects with isolated coronavirus were hospitalized, but in the population aged ≥60 years this increased to 73%. While our study design precludes the determination of exact hospitalization rates as a result of coronavirus infections, the hospitalization rates were similar to those for influenza. No pathogen was identified in 35% of subjects. This was more common in adults (no pathogen detected in 41% (321/793) of all adult subjects aged ≥18 years) than in children (no pathogen detected in 20% of subjects). It is unclear if these subjects had bacterial infections, pathogens not tested for in our assay, or non-infectious etiologies of their ILI. Unfortunately the lack of paired sputum limited our ability to determine etiology. In a cohort in New Caledonia, paired sputum samples were required to determine the etiologic agent in 23% of subjects with lower respiratory tract infections.32 In a population of hospitalized children in Mexico with lower respiratory tract infections, viral etiologies were determined in only 47%.34 This study found a virologic agent in 80% of children. This improved detection is likely in part due to the extended panel of viruses that were tested with the multiplex PCR.

ns.32 In a population of hospitalized children in Mexico with lower respiratory tract infections, viral etiologies were determined in only 47%.34 This study found a virologic agent in 80% of children. This improved detection is likely in part due to the extended panel of viruses that were tested with the multiplex PCR. More than one virus was isolated in 11.9% of subjects. It is not possible to determine the contribution of each virus to the clinical syndrome, nor is it possible to determine if some viruses that were detected were not causing disease. While PCR significantly improves the sensitivity of detecting respiratory viruses, the detection of viral nucleic acids may not always represent causation. In prior case–control studies, virus nucleic acids could be isolated from 20–27% of asymptomatic subjects.35, 36 This might be supported by the data showing that among those with mixed infections, the outcomes were not significantly worse than those for subjects with single infections. While all subjects enrolled were symptomatic, and this would increase the likelihood that isolated pathogens were causative, cohort studies such as ours are not able to directly determine causation. With 62 unique combinations in 125 different subjects, the assessment of outcomes lacks statistical power to determine differences between the various combinations of mixed infections.

crease the likelihood that isolated pathogens were causative, cohort studies such as ours are not able to directly determine causation. With 62 unique combinations in 125 different subjects, the assessment of outcomes lacks statistical power to determine differences between the various combinations of mixed infections. The determination of non-influenza respiratory viruses in ILI has several implications. First, increasing knowledge that acute respiratory illness and its associated morbidity are often caused by respiratory viruses may stimulate the development of more affordable rapid diagnostic tests as well as therapeutics for these viruses. With the consideration of increasing antibiotic resistance, the diagnosis of respiratory viruses could decrease the quantities of empiric broad-spectrum antibiotics used. The aim of this study was not the epidemiological surveillance of respiratory diseases. We included only patients who by the severity of their illnesses sought medical attention. The attending physicians determined the need for hospitalization or outpatient treatment. This limited our population to a particular group of patients with clinical manifestations of the disease. With this strategy we seek to find the more serious cases. From this we believe we can more easily determine the factors associated with severe disease. It is a limitation of this study, however, that we lose a part of the population with mild or asymptomatic disease.

tients with clinical manifestations of the disease. With this strategy we seek to find the more serious cases. From this we believe we can more easily determine the factors associated with severe disease. It is a limitation of this study, however, that we lose a part of the population with mild or asymptomatic disease. In conclusion, viruses that were traditionally believed to produce minor disease have the capacity, in certain subjects, to cause severe cases, as suggested by the high proportion of adult patients with rhinovirus infection who required in-hospital treatment. We believe that these findings mark the need for specific studies that will help us to better understand the behavior of these viruses, and the factors associated with a more serious presentation. Acknowledgements La Red is funded by the Mexico Ministry of Health and the U.S. National Institute of Allergy and Infectious Diseases. This project has been funded in part by funding provided by: CONACYT through FONSEC SSA/IMSS/ISSSTE No. 71260 and No. 127088. Intramural Research Programs of the National Institute of Allergy and Infectious Diseases, National Institutes of Health. National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH) through a contract with Westat, Inc., Contract Number: HHSN2722009000031, Task Order Number: HHSN27200002. And through the National Cancer Institute, National Institutes of Health, 318 under Contract No. HHSN261200800001E.

ational Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH) through a contract with Westat, Inc., Contract Number: HHSN2722009000031, Task Order Number: HHSN27200002. And through the National Cancer Institute, National Institutes of Health, 318 under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, or Westat, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government. Ethical approval: The Ethics committee of each institution reviewed the study, and the study was conducted following the principles of the ICH GCP, Declaration of Helsinki, and the Mexican General Health Law. All subjects provided informed consent. All data was de-identified during data collection. The project was registered on clinicaltrials.gov (NCT01418287). Conflict of interest: The authors have no competing interests to declare.

1 Introduction Since the Middle East respiratory syndrome coronavirus (MERS-CoV) was first described in September 2012,1 there have been a total of 206 cases of MERS-CoV infection with 86 deaths (41.7% mortality rate) reported to the WHO.2 All cases have had links to the Middle East and the majority of cases (156 with 63 deaths (40% mortality) have been reported from KSA as of March 15, 2014. We previously reported family3 and healthcare associated4 case clusters of MERS-CoV infections where human-to-human transmission occurred between index cases and their contacts. Whilst camels have been implicated as the reservoir of MERS-CoV,5, 6, 7 the exact source(s) and mode of transmission for most patients remain unknown. Serology consistent with a common MERS-CoV like virus in camels has been demonstrated by several studies5, 6, 7 and recently evidence has emerged of a MERS-CoV infection in a camel and in humans in contact with these camels.8, 30

CoV,5, 6, 7 the exact source(s) and mode of transmission for most patients remain unknown. Serology consistent with a common MERS-CoV like virus in camels has been demonstrated by several studies5, 6, 7 and recently evidence has emerged of a MERS-CoV infection in a camel and in humans in contact with these camels.8, 30 During a 3 month period, May 31, 2013 to August 31, 2013, there were 12 confirmed MERS-CoV cases reported from the Hafr Al-Batin district in the north east region of the Kingdom of Saudi Arabia (KSA). Hafr Al-Batin has the biggest camel market in the entire Kingdom with 500,000 camels being reared there. Hafr Al-Batin annually hosts drovers of more than 100 camel herds, comprising around 10,000 camels, from various regions of KSA, Kuwait, Qatar and the United Arab Emirates. This annual festival is known locally as “Mazayin al-Ibl, meaning “The Best of the Herds,” and attracts more than 160,000 people9 from November- December to March each year. This festival was the first to be established in the region and subsequently other camel festivals were started in neighboring countries: Qatar, Kuwait and United Arab Emirates. Since evidence is accumulating that camels are a zoonotic reservoir for MERS-CoV, we conducted a detailed epidemiological, clinical and genomic study to ascertain common exposure and transmission patterns of all cases of MERS reported from Hafr Al-Batin and relate it to other available genomic sequences from KSA and globally.

vidence is accumulating that camels are a zoonotic reservoir for MERS-CoV, we conducted a detailed epidemiological, clinical and genomic study to ascertain common exposure and transmission patterns of all cases of MERS reported from Hafr Al-Batin and relate it to other available genomic sequences from KSA and globally. 2 Methods 2.1 Selection of MERS-CoV cases MERS-CoV cases reported from the Hafr Al-Batin region were selected for study. Epidemiological, clinical and laboratory details were collected. Clinical information included demographic data, clinical symptoms and signs, co-morbidities, contact with animals and travel history

vidence is accumulating that camels are a zoonotic reservoir for MERS-CoV, we conducted a detailed epidemiological, clinical and genomic study to ascertain common exposure and transmission patterns of all cases of MERS reported from Hafr Al-Batin and relate it to other available genomic sequences from KSA and globally. 2 Methods 2.1 Selection of MERS-CoV cases MERS-CoV cases reported from the Hafr Al-Batin region were selected for study. Epidemiological, clinical and laboratory details were collected. Clinical information included demographic data, clinical symptoms and signs, co-morbidities, contact with animals and travel history 2.2 MERS-CoV testing and genomic analyses All suspected cases meeting the basic MERS-CoV infection criteria were confirmed in Saudi Ministry of Health regional laboratories by reverse transcription, real-time-PCR as previously described.4 MERS-CoV genomic sequences were available from a subset of the Hafr Al-Batin MERS cases.10 These viral sequences were used to test possible transmission routes for the virus and establish the plausibility of epidemiologically suspected virus transmissions using a previously described statistical test of transmission.4 Briefly, the expected number of sequence changes between two sequences was calculated as the product of the time interval between sampling, the evolutionary rate of the virus, and the maximum length of sequence shared by the two virus genomes. If the number of differences between two sequences accumulating in a given time is assumed to follow a Poisson distribution, with λ equal to the expected number of mutations, the probability of finding this number of differences between the two sequences by chance can be calculated from the cumulative density function of the Poisson distribution. A transmission pair was rejected if the number of observed mutations exceeded the 95% upper cumulative probability value. To reduce the chance of type 1 statistical errors due to multiple testing, a Bonferroni correction was applied to the significance cutoff, resulting in an adjusted significance level of 3.85 × 10–3. The rate of evolution of MERS-CoV has been estimated at 1.12 × 10–3 substitutions per site per year (95% credible interval [95% CI], 8.76 × 10–4; 1.37 × 10–3).10 To account for uncertainty in the evolutionary rate of the virus, the transmission tests were repeated at the 95% upper and lower credible intervals and checked for consistency in the results.

has been estimated at 1.12 × 10–3 substitutions per site per year (95% credible interval [95% CI], 8.76 × 10–4; 1.37 × 10–3).10 To account for uncertainty in the evolutionary rate of the virus, the transmission tests were repeated at the 95% upper and lower credible intervals and checked for consistency in the results. In addition, a plausibility test was added. A reproductive time for MERS-CoV has been estimated at 7-12 days,11 which represents the time from symptom onset in a primary case to symptom onset in a secondary case. This estimate is largely derived from hospital-based infections which may be dominated by patients with renal failure and other co-morbidities,4, 12 as well as close contact with infected cases resulting in an underestimate of the generation time. For testing the global transmission of the virus, we included an asymptomatic period when a patient might still be infectious, estimating that a case remains infectious for 14 days, and assuming that identification would occur within 7 days of infection. Thus, any two cases might be plausibly directly linked if they meet the statistical sequence test and the two sample dates differ by 21 days or less. This calculation was used to assess the likelihood that virus transmission occurred directly between two test cases and was applied to all cases infected with the Hafr-Al-Batin_1 MERS-CoV variant. 2.3 Statistical analysis A chi-squared test was used to assess the associations of comorbidities and clinical presentations, with p values <0.05 considered significant.

In addition, a plausibility test was added. A reproductive time for MERS-CoV has been estimated at 7-12 days,11 which represents the time from symptom onset in a primary case to symptom onset in a secondary case. This estimate is largely derived from hospital-based infections which may be dominated by patients with renal failure and other co-morbidities,4, 12 as well as close contact with infected cases resulting in an underestimate of the generation time. For testing the global transmission of the virus, we included an asymptomatic period when a patient might still be infectious, estimating that a case remains infectious for 14 days, and assuming that identification would occur within 7 days of infection. Thus, any two cases might be plausibly directly linked if they meet the statistical sequence test and the two sample dates differ by 21 days or less. This calculation was used to assess the likelihood that virus transmission occurred directly between two test cases and was applied to all cases infected with the Hafr-Al-Batin_1 MERS-CoV variant. 2.3 Statistical analysis A chi-squared test was used to assess the associations of comorbidities and clinical presentations, with p values <0.05 considered significant. 3 Results 3.1 MERS-CoV cases and clusters Between May 31, 2013 and August 31, 2013, there were 12 confirmed MERS-CoV cases in the Hafr Al-Batin area. Two index cases and two clusters were noted. In Cluster 1, the index case Patient 1 (as defined by first date of diagnosis) was a 21 year-old non-Saudi shepherd with onset of symptoms on May 31, 2013 and a secondary case, Patient 2, a healthcare worker contact had who had an asymptomatic infection (Figure 1 A). Cluster 2 involved a 38 year-old Saudi male, Patient 3, who owned and directly cared for camels, and had onset of symptoms on August 8, 2013. Patient 3 was closely associated with five additional MERS cases (Patients 4-8) and Patient 8, one of the secondary contacts of Patient 3, was associated with an additional four MERS cases (Patients 9-12, Figure 1A).Figure 1 (A) The epidemiologically defined transmission pathway of the two MERS-CoV clusters. Markers for each MERS case were placed by sample date (x-axis) and colored dark grey for fatal outcome or green for non-fatal outcome. Patient numbers are indicated within each marker. The age and gender of each case are indicated, and epidemiologically-possible contacts of Patients 1, 3 and 8 are marked by arrows. (B) Determination of genetically plausible transmission for the Hafr Al-Batin Cluster 2. Time of sample collection is indicated on the x-axis. Colored markers indicate the MERS case with the following code. Blue: MERS-CoV sequence was available for that case, grey: no sequence was available, orange: no sequence was available but the ancestral reconstructed sequence for the clade was used. Green arrows indicate statistically plausible transmission between the pair (with no direction implied), red arrows indicate that transmission between the pair is not supported statistically. Statistical tests of transmission were performed as previously described.4

estral reconstructed sequence for the clade was used. Green arrows indicate statistically plausible transmission between the pair (with no direction implied), red arrows indicate that transmission between the pair is not supported statistically. Statistical tests of transmission were performed as previously described.4 3.2 Comorbidity and Clinical Presentations Of the 12 cases, five (41.7%) had contacts with camels. The first index (June 2013) case (Patient 1) was a shepherd. In the two clusters, five patients died including the 2 index cases and 3 of their close contacts (four out of the five mortalities had comorbidities). Of those with comorbidities, all had diabetes mellitus, three had hypertension, and one was also obese and was a smoker. Comorbidities were present in four (80%) of the five symptomatic cases and in one (14%) of the seven asymptomatic cases (p = 0.07) (Table 1 ).Table 1 shows symptomatic and asymptomatic cases and comorbid conditions. Cases Age Gender Comorbidity Outcome Animal Contact Pt1* Index case Cluster 1 21 M none died yes Pt2 56 F none alive no Pt3* Index case Cluster 2 38 M DM died yes Pt4* 79 F DM, HTN died no Pt5 26 M none alive yes Pt6 16 M none alive yes Pt7 7 F none alive no Pt8* 47 M Obesity, DM, HTN, smoking, HD died yes Pt9 46 F DM, HTN alive no Pt10 3 F none alive no Pt11 18 M none alive no Pt12* 74 F DM, HTN died no * Symptomatic case; DM=Diabetes Mellitus,HTN= hypertension, HD= Hemodialysis; M = male; F= female.

no Pt5 26 M none alive yes Pt6 16 M none alive yes Pt7 7 F none alive no Pt8* 47 M Obesity, DM, HTN, smoking, HD died yes Pt9 46 F DM, HTN alive no Pt10 3 F none alive no Pt11 18 M none alive no Pt12* 74 F DM, HTN died no * Symptomatic case; DM=Diabetes Mellitus,HTN= hypertension, HD= Hemodialysis; M = male; F= female. All symptomatic cases had fever, cough, shortness of breath and four (80%) complained of sore throat. Two (40%) had headache, one (20%) complained of hemoptysis and one (20%) had nausea (Table 2 ).Table 2 Clinical Presentations among symptomatic cases. Symptom No. of cases % fever 5 100 sore throat 4 80 cough 5 100 shortness of breath 5 100 hemoptysis 1 20 nausea 1 20 headache 2 40 3.3 Contact Investigation The contact investigation was carried out for both family contacts and healthcare worker contacts. Among the family contacts, 7 out of 36 (19.4%) tested positive and 1 of 51 (2%) healthcare worker contacts tested positive for MERS-CoV (p = 0.0078).

Symptom No. of cases % fever 5 100 sore throat 4 80 cough 5 100 shortness of breath 5 100 hemoptysis 1 20 nausea 1 20 headache 2 40 3.3 Contact Investigation The contact investigation was carried out for both family contacts and healthcare worker contacts. Among the family contacts, 7 out of 36 (19.4%) tested positive and 1 of 51 (2%) healthcare worker contacts tested positive for MERS-CoV (p = 0.0078). 3.4 Genetic tracing of MERS-CoV transmission and possible linkages For the Hafr Al-Batin Cluster 1 we obtained a full MERS-CoV genome from the index Patient 1 (Hafr-Al-Batin_1_2013), however no useful sequence could be obtained from the only positive contact in that cluster (Patient 2, a health care worker). For Cluster 2, MERS-CoV genomic sequences were obtained from Patient 4 (Hafr-Al-Batin_5_2013), Patient 5, (Hafr-Al-Batin_4_2013), Patient 8 (Hafr-Al-Batin_6_2013) and Patient 12 (Hafr-Al-Batin_2_2013). We were unable to obtain sequence from Patient 3, however an ancestral sequence for the entire family clade was reconstructed and used as a surrogate for the Patient 3 sequence. Possible transmission routes are indicated in Figure 1B. Each patient is indicated by a filled circle placed by sample date, blue-filled circles indicate patients with sequence, grey-filled circles indicate cases with no available sequence and the orange-filled circle represents the reconstructed ancestral sequence.

e. Possible transmission routes are indicated in Figure 1B. Each patient is indicated by a filled circle placed by sample date, blue-filled circles indicate patients with sequence, grey-filled circles indicate cases with no available sequence and the orange-filled circle represents the reconstructed ancestral sequence. Individual cases in the second Hafr Al-Batin transmission cluster were plotted by date of the sequenced sample (Figure 2 ). Blue-filled circles indicate cases with sequence data, the orange-filled circle for Patient 3 indicates that an ancestor sequence for the clade was reconstructed and used as a surrogate, grey-filled circles indicate cases with no available sequence. Some transmissions are statistically allowed including Patient 3 to Patient 8 and Patient 4, 5 or 8 to patient 12 (green arrows). However, transmission from Patient 3 to Patient 4 or Patient 5 is not likely to have occurred (dashed red arrows); other sources of the infection should therefore be considered for Patient 4 and Patient 5.Figure 2 Statistically possible transmission pairs between MERS cases infected with the Hafr-Al-Batin_1 clade of MERS-CoV. All MERS cases known to be infected with the Hafr-Al-Batin_1 clade of MERS-CoV were plotted by sample date (x-axis) and color-coded by Hafr-Al-Batin_1 subclade as previously described 10. Upper panel: All statistically supported transmission pairs are marked by arcs. Lower panel: Only potential transmission pairs whose collection dates differ by 21 days or less are marked by arcs. See Methods section for additional details.

(x-axis) and color-coded by Hafr-Al-Batin_1 subclade as previously described 10. Upper panel: All statistically supported transmission pairs are marked by arcs. Lower panel: Only potential transmission pairs whose collection dates differ by 21 days or less are marked by arcs. See Methods section for additional details. To examine MERS-CoV sources on a broader scale, the possible transmissions amongst all patients known to have been infected with the Haf Al-Batin variant were tested (Figure 2). The Hafr-Al-Batin_1 clade was first observed in Patient 1 (Hafr-Al-Batin_1) in May 201313 and since then has been identified in 19 MERS patients10 and 1 camel8 in Riyadh, Hafr Al-Batin, Madinah, and Qatar.

ansmissions amongst all patients known to have been infected with the Haf Al-Batin variant were tested (Figure 2). The Hafr-Al-Batin_1 clade was first observed in Patient 1 (Hafr-Al-Batin_1) in May 201313 and since then has been identified in 19 MERS patients10 and 1 camel8 in Riyadh, Hafr Al-Batin, Madinah, and Qatar. All MERS cases were depicted by sample date and color-coded by virus clade as previously described 10 (Figure 2, upper panel). All statistically supported transmissions are marked by arcs connecting the relevant patients. Several important patterns appear. The viruses Hafr-Al-Batin_1, Riyadh_8, Riyadh_12 are linked to a large number of possible pairs. Hafr-Al-Batin_1_2013 is an early virus in this clade and may be representative of the camel to human zoonosis that gave rise to this clade, showing linkage to 10 of the 19 cases infected with the Hafr-Al-Batin_1 clade virus. The 4 linkages from Riyadh_8, 3 linkages from Riyadh_11 and 4 linkages from Riyadh_12 viruses may provide important clues. These are viruses from patients in Riyadh with no direct links to the Hafr Al-Batin region and no contact with animals, including camels. Within a period of one month, closely-related viruses were observed in Riyadh, in the Hafr Al-Batin region as well as in Madinah, with statistical support for direct transmission events. Any MERS-CoV transmission model must account for this rapid virus movement and MERS-CoV infection with no apparent live animal contact. Furthermore, sequences from three viruses in the recently described camel/human transmission cluster in Qatar also fall within the Hafr Al-Batin cluster.8 The transmission testing supports the conclusion that the Qatar human and camel viruses are directly related to the Saudi Hafr Al-Batin cluster.

live animal contact. Furthermore, sequences from three viruses in the recently described camel/human transmission cluster in Qatar also fall within the Hafr Al-Batin cluster.8 The transmission testing supports the conclusion that the Qatar human and camel viruses are directly related to the Saudi Hafr Al-Batin cluster. Adding the 21-day plausibility filter, (see Methods) the pattern is reduced in complexity; however important features remain (Figure 2, lower panel). The linkages between the Riyadh_12 patient and the patients in the Hafr Al-Batin family cluster (Hafr-Al-Batin_4, 5 and 6) remain. In searching for alternative sources of the infection of Patient 4 and 5, Riyadh_12 or the source of the infection of Riyadh_12 should be considered, including exposure to health care facilities, or health care workers, consumption or exposure to uncooked animal products, exposure to camels or other animals directly. A second network of transmissions passes the plausibility test with transmissions between the Riyadh_8, Riyadh_11 and Riyadh 17 cases and beyond to Madinah_1 and Madinah_3.

health care facilities, or health care workers, consumption or exposure to uncooked animal products, exposure to camels or other animals directly. A second network of transmissions passes the plausibility test with transmissions between the Riyadh_8, Riyadh_11 and Riyadh 17 cases and beyond to Madinah_1 and Madinah_3. 4 Discussion In this report we describe the possible transmission dynamics of MERS-CoV in community case clusters from the Hafr Al-Batin region. A cluster was defined by WHO as the occurrence of > 2 patients with onset of symptoms within an incubation period of 14 days. The transmission occurs in the same setting such as a classroom, workplace, household, extended family, or hospital.14 Since the emergence of MERS-CoV, a number of clusters involving more than two people14 have been reported from France,15, 16 Italy, Jordan,17 KSA,3, 4, 18 Tunisia,19 UAE, UK20 and Qatar. The known 14 primary cases in these clusters were adult men.21 Of the involved individuals, 26% occurred in healthcare setting.21 The largest healthcare associated MERS-CoV cluster was reported from Al-Hasa, Saudi Arabia.4 In an earlier family cluster from Saudi Arabia, an adult male index case resided in an extended household of 10 other adults and 18 children. Secondary cases were identified in two sons, and a grandson.3 The findings from previous clusters showed that immunocompetent contacts exhibit mild symptoms.20, 22 In the UK cluster of MERS-CoV, two cases of MERS-CoV infection were confirmed and one of the two cases had severe illness. None of the 59 healthcare workers contacts had infection.20 In the Hafr Al-Batin cases reported here, 19.4% among the family contacts tested positive and 2% of the healthcare worker contacts were positive for MERS-CoV (p = 0.0078).

cases of MERS-CoV infection were confirmed and one of the two cases had severe illness. None of the 59 healthcare workers contacts had infection.20 In the Hafr Al-Batin cases reported here, 19.4% among the family contacts tested positive and 2% of the healthcare worker contacts were positive for MERS-CoV (p = 0.0078). One of the differences between primary cases and secondary cases in MERS-CoV clusters is that primary cases frequently have no contact with a known MERS cases and are therefore thought to acquire infection through contact with non-human sources of the virus.23 The Hafr Al-Batin MERS Cluster 2 showed the spectrum of illness of MERS-CoV from asymptomatic to a fulminant disease as observed previously.3 In the current study, animal contact was reported in 41.7% of all cases. The presence of animal contact among asymptomatic family contacts further complicate the issue of having secondary cases as a result of direct contact or the result of exposure to the same source or host of MERS-CoV that lead to the index infection. In fact, for family cluster 2, our genetic data indicate that while Patient 8 is likely to have acquired the infection from the index Patient 3, at least two of the infected contacts (Patients 4 and 5) could not have been directly infected from Patient 3 and alternate source should be considered.

o the index infection. In fact, for family cluster 2, our genetic data indicate that while Patient 8 is likely to have acquired the infection from the index Patient 3, at least two of the infected contacts (Patients 4 and 5) could not have been directly infected from Patient 3 and alternate source should be considered. Although, there have been a number of transmissions observed within healthcare setting and interfamilial, the number of secondary transmissions seem be limited and is consistent with the R0 estimates for MERS being less than 1.11, 32 This finding is similar to previous observations from known clusters4, 24, 25 and that secondary attack rates among family members of patients in other clusters appear to be low.3, 4, 17, 25, 26 In a recent large screening study, family contacts had a higher positivity rate (3.6%) than HCW contacts (1.12%).27 Systematic implementation of infection prevention and control measures in reported clusters involving healthcare settinghas appeared to limit onward transmission to HCW and hospitalized patients.4, 15, 20, 25, 28, 29

large screening study, family contacts had a higher positivity rate (3.6%) than HCW contacts (1.12%).27 Systematic implementation of infection prevention and control measures in reported clusters involving healthcare settinghas appeared to limit onward transmission to HCW and hospitalized patients.4, 15, 20, 25, 28, 29 A large annual camel fair takes place in Hafr Al-Batin each November-December to March with movement of a large number of camels both to and from Hafr Al-Batin.9 It seems that the timing of the Hafr Al-Batin sequence divergence is consistent with the annual fair and animal movement. It remains unclear if primary cases had acquired MERS-CoV from direct animal contact or as a result of contact or consumption of animal products, unpasteurized camel milk and products (ice cream). Evidence that camels are a source of human MERS-CoV infections is accumulating. Two sets of patients with camel contacts have now been analyzed and the sequence data support direct transmission between camel and human, although the sequences do not allow direction of transmission.8, 30 Serological evidence that the camel infections predate the human infection has been provided.30 Furthermore, a full MERS-CoV genome has been obtained from a camel in Egypt;31 this combined with the high prevalence of MERS-CoV seropositivity indicates that camels may be frequently infected with the virus. More detailed case control studies with detailed case histories, epidemiological information and genomic analysis are being conducted to delineate the missing pieces in the transmission dynamics of MERS-CoV outbreak.

As of June 24, 2015, there have been 1354 laboratory-confirmed cases of Middle East respiratory syndrome coronavirus (MERS-CoV) infection. The first reported case was in September 2012. Since then, sporadic cases, clusters of cases, and outbreaks have occurred in 24 countries, with more than 1000 cases reported in Saudi Arabia. In May–June 2015, a large outbreak occurred in the Republic of Korea. Both MERS-CoV and influenza viruses are respiratory viruses, and outbreaks of infection can have a significant public health impact. MERS-CoV is a zoonotic infection with limited human-to-human transmissibility, and sporadic cases have sometimes led to clusters of infection. Influenza A(H5N1) is also a zoonotic infection, with a similar degree of human-to-human transmissibility in households as MERS-CoV.1, 2 On the other hand, human influenza viruses spread easily between humans, and large epidemics occur at least once per year in most countries. Zoonotic and human-to-human transmission of these pathogens could be affected by common factors such as meteorological factors (e.g., daily temperature and humidity), travel restrictions, and enhanced surveillance efforts. The practice of testing for either virus could also impact on detection of the other. It is therefore of interest to study the patterns of laboratory-confirmed influenza cases in the Middle East and the global MERS-CoV cases.

e.g., daily temperature and humidity), travel restrictions, and enhanced surveillance efforts. The practice of testing for either virus could also impact on detection of the other. It is therefore of interest to study the patterns of laboratory-confirmed influenza cases in the Middle East and the global MERS-CoV cases. The patterns of MERS-CoV cases occurring from May 1, 2012 to May 31, 2015 were studied; 82.1% of these cases were reported in Saudi Arabia or the United Arab Emirates (UAE). Data were obtained from the EMPRES-i Global Animal Disease Information System (http://empres-i.fao.org; accessed June 25, 2015). For influenza, as surveillance data were not publicly available in Saudi Arabia or the UAE, influenza data from the neighbouring countries – Bahrain, Iraq, Jordan, Oman, Qatar and Egypt – for the same period, were used. These data were downloaded from the World Health Organization FluNet (http://www.who.int/influenza/gisrs_laboratory/flunet/en/; accessed June 25, 2015).

y available in Saudi Arabia or the UAE, influenza data from the neighbouring countries – Bahrain, Iraq, Jordan, Oman, Qatar and Egypt – for the same period, were used. These data were downloaded from the World Health Organization FluNet (http://www.who.int/influenza/gisrs_laboratory/flunet/en/; accessed June 25, 2015). The patterns of weekly confirmations of influenza A(H1N1)pdm09, A(H3N2), and A(H5N1) were studied, and also the weekly total numbers of specimens tested for influenza virus. The weekly total numbers of influenza specimens tested in the six countries were compared with the weekly numbers of confirmed MERS-CoV cases (Figure 1A). The weekly influenza A(H1N1)pdm09, A(H3N2), and A(H5N1) confirmations were also compared with the weekly global numbers of confirmed MERS-CoV cases (Figure 1B, C).Figure 1 Weekly MERS-CoV and influenza confirmations. Weekly MERS-CoV confirmations as shaded regions versus weekly specimens processed for influenza in six countries as coloured curves (panel A); weekly influenza A(H1N1)pdm09 and influenza A(H3N2) confirmations (panel B) and influenza A(H5N1) laboratory-confirmed human cases in Egypt (panel C). There were a number of peaks in the weekly total influenza specimens processed that coincided with the major MERS-CoV waves from March to July 2014: these took place in Iraq in early 2013, Egypt in early 2014 and from late 2014 to early 2015, and Qatar in 2013/14 and 2014/15 influenza seasons (Figure 1A).3 Minor oscillating MERS-CoV waves were also noted in October 2014, March 2015, and May 2015.

that coincided with the major MERS-CoV waves from March to July 2014: these took place in Iraq in early 2013, Egypt in early 2014 and from late 2014 to early 2015, and Qatar in 2013/14 and 2014/15 influenza seasons (Figure 1A).3 Minor oscillating MERS-CoV waves were also noted in October 2014, March 2015, and May 2015. The influenza A(H1N1)pdm09 wave in early 2014 in Egypt closely led the major MERS-CoV wave in May 2014, while the influenza A(H3N2) wave in late 2014 in Egypt closely led the MERS-CoV wave in early 2015 (Figure 1B). The second MERS-CoV wave in late 2014 and the third MERS-CoV wave in early 2015 appeared to be split by the influenza A epidemic. The 2014/15 influenza A wave in Egypt appeared to fit in the time interval between the occurrences of the two MERS-CoV waves. The influenza A(H5N1) wave in Egypt co-occurred with a MERS-CoV wave in early 2015 (Figure 1C). These patterns showed that the major MERS-CoV waves in 2014 and 2015 closely followed the human influenza A epidemics in the Middle East region, and most of the influenza waves did not co-occur with the MERS-CoV waves. This implies that the factors affecting seasonality of MERS-CoV infections may differ from the factors affecting the seasonality of human influenza virus infections. Further studies on the social or environmental factors that may increase the risk of human infection with MERS-CoV would be worthwhile.

he MERS-CoV waves. This implies that the factors affecting seasonality of MERS-CoV infections may differ from the factors affecting the seasonality of human influenza virus infections. Further studies on the social or environmental factors that may increase the risk of human infection with MERS-CoV would be worthwhile. Acknowledgements This work received financial support from the Harvard Center for Communicable Disease Dynamics from the National Institute of General Medical Sciences (Grant No. U54 GM088558), the Research Council of Hong Kong Food and Health Bureau (Project No. 13121382), and the Research Grants Council of the Hong Kong Special Administrative Region, China (Project Nos. T11-705/14N and 25100114). The funding bodies had no role in the study design, data collection and analysis, preparation of the manuscript, or the decision to publish. Conflict of interest: BJC has received research funding from MedImmune Inc. and Sanofi Pasteur, and consults for Crucell NV. The authors report no other potential conflicts of interest.

1 Introduction Middle East respiratory syndrome (MERS), caused by a novel coronavirus (MERS-CoV), was first identified in 2012 in patients with severe respiratory disease in Jordan and Saudi Arabia.1 Since its discovery, approximately 1600 cases have been reported, amounting to about 40 cases per month. While this number is low, the worrisome features of the disease are its propensity to cause severe disease in patients with underlying conditions, including diabetes, renal disease, lung disease, or an immunocompromised state, and its apparent ability to readily spread within hospital settings.2 In addition, MERS-CoV has been identified in camel populations throughout the Arabian Peninsula and Africa,3, 4, 5 and epidemiological evidence suggests that it is periodically introduced into human populations.6 Further, coronaviruses have a well-described propensity to mutate and recombine.7 Consistent with this propensity, the genomic sequence of MERS-CoV has changed since it first entered human populations in 2012, but these changes have not enhanced the ability to effect human-to-human transmission.8 This lack of increased transmissibility is encouraging, but, on the other hand, the continued introduction into human populations from infected camels coupled with coronavirus mutability means that measures to prevent infection are important to develop anticipatorily.

the ability to effect human-to-human transmission.8 This lack of increased transmissibility is encouraging, but, on the other hand, the continued introduction into human populations from infected camels coupled with coronavirus mutability means that measures to prevent infection are important to develop anticipatorily. Following the demonstration of the key role of hospitals in secondary spread,9, 10 efforts were made to introduce careful infection control measures into affected hospitals. These appear to have been effective in reducing virus transmission and greatly decreasing the number of MERS cases. However, these measures do not affect the acquisition of primary cases of MERS, which likely occur either directly or indirectly from camels. These primary cases are the source for subsequent hospital outbreaks, so preventing transmission from camels or within the community might be the best way to provide subsequent secondary cases and hospital spread.

ct the acquisition of primary cases of MERS, which likely occur either directly or indirectly from camels. These primary cases are the source for subsequent hospital outbreaks, so preventing transmission from camels or within the community might be the best way to provide subsequent secondary cases and hospital spread. In addition to the appropriate infection control measures, virus transmission would be most effectively prevented by a combination of rapid and efficient diagnosis, treatment with antiviral therapy to decrease virus loads, and prophylactic treatment with an intervention that prevents infection or at least disease manifestations. Most often, the latter approach involves passive or active immunization, which will be discussed in this review. Efforts to prevent MERS by immunization are based in part on the extensive information gained from studies of coronavirus vaccines used to prevent infections in domesticated and companion animals. Additionally, a key piece of information required for the rational design of vaccines is knowledge of a protective immune response. Immune responses to some non-human coronaviruses have been characterized and these responses are also described below.

nes used to prevent infections in domesticated and companion animals. Additionally, a key piece of information required for the rational design of vaccines is knowledge of a protective immune response. Immune responses to some non-human coronaviruses have been characterized and these responses are also described below. 2 Protective immune response in animals experimentally infected, or patients naturally or experimentally infected with coronaviruses In general, protective immune responses to coronaviruses involve a combination of virus-specific antibody and T-cell responses.11 The neutralizing antibody response is primarily directed against the surface (S) protein, responsible for binding to the host cell receptor. The N terminal S1 fragment of the S protein binds to the host cell receptor, elicits neutralizing antibody, and perhaps not surprisingly, is also the part of the virus that is most variable between isolates.12 This variability explains why neutralizing antibodies are generally virus strain-specific and do not provide cross-reactive protection against even closely related coronaviruses.13 On the other hand, coronavirus-specific CD8 and CD4 T-cells recognize epitopes from across the genome, some of which are in conserved proteins, which do not readily undergo mutation.

antibodies are generally virus strain-specific and do not provide cross-reactive protection against even closely related coronaviruses.13 On the other hand, coronavirus-specific CD8 and CD4 T-cells recognize epitopes from across the genome, some of which are in conserved proteins, which do not readily undergo mutation. Prior to the onset of severe acute respiratory syndrome (SARS) and MERS, many studies on protective immune responses used mice infected with the murine coronavirus, mouse hepatitis virus (MHV). These studies showed that virus clearance from infected mice required the development of an effective T-cell response. Both CD4 and CD8 T-cells were required for optimal kinetics of clearance.14 The studies also showed that the T-cell response could be immunopathological.14, 15, 16 Thus when irradiated mice or mice lacking T- and B-cells were infected with a strain of MHV that causes demyelination, the mice developed minimal clinical disease and showed no evidence of demyelination. However, within a few days of receiving virus-specific T-cells, severe myelin destruction occurred, along with hind limb paralysis. Neutralizing antibodies were also important in immune protection, serving at least two roles. First, in the absence of neutralizing antibody, MHV was cleared to very low levels by T-cells, but later recrudesced, resulting in lethal disease.17 Second, virus-specific antibodies were most important for protecting mice against further challenge. Of note, immune protection was long-lived in immunocompetent mice that survived experimental infection with MHV, possibly because the infection was systemic, involving the central nervous system, or in some cases, the liver.

d, virus-specific antibodies were most important for protecting mice against further challenge. Of note, immune protection was long-lived in immunocompetent mice that survived experimental infection with MHV, possibly because the infection was systemic, involving the central nervous system, or in some cases, the liver. In marked contrast, coronaviruses that are primarily mucosal induce short-lived protection. This is most evident in studies of patients or human volunteers infected with respiratory coronaviruses such as HCoV-229E or HCoV-OC43.18, 19 These viruses generally cause mild upper respiratory tract disease and only rarely cause severe disease. In human volunteer studies, the presence of pre-existing anti-HCoV-OC43 or HCoV-229E antibodies did not provide protection against experimental challenge with the same virus, in terms of clinical disease or virus titers. Similarly, experimental challenge provided only partial protection against subsequent re-challenge and this protection waned over several months. In these studies, systemic antibodies were generally measured, so less is known about the levels of IgA, which are likely most important for protection against viruses that remain confined to the upper respiratory tract.

y partial protection against subsequent re-challenge and this protection waned over several months. In these studies, systemic antibodies were generally measured, so less is known about the levels of IgA, which are likely most important for protection against viruses that remain confined to the upper respiratory tract. From these data, one might predict that infection with MERS-CoV or SARS-CoV would result in a long-lived protective response, since SARS-CoV and MERS-CoV cause severe respiratory illness based in the lungs, and SARS-CoV (and perhaps MERS-CoV) causes a systemic infection.20 However, this may not be the case. While only a few SARS survivors have been followed longitudinally, anti-SARS-CoV antibody titers were not detectable after 6 years.21 Longitudinal studies of T-cell responses in these patients are even fewer in number, but T-cell responses were detected at low levels in some survivors.21, 22, 23, 24 While these data suggest that coronavirus-specific T-cells are more likely to persist than B-cells, it is still possible that there are sufficient numbers of residual memory T- and B-cells to protect patients from infection or severe disease on rechallenge.

s were detected at low levels in some survivors.21, 22, 23, 24 While these data suggest that coronavirus-specific T-cells are more likely to persist than B-cells, it is still possible that there are sufficient numbers of residual memory T- and B-cells to protect patients from infection or severe disease on rechallenge. 3 Previous studies of coronavirus-vaccinated domesticated and companion animals Prior to the outbreak of SARS, coronaviruses were considered most important as causes of infections of domesticated and companion animals. Vaccines to prevent several of these diseases were developed over the years, but none were very successful in preventing disease. Infectious bronchitis virus (IBV) is an economically important infection of young chickens, causing bronchitis as well as renal disease (reviewed by Cavanagh25). Live attenuated vaccines were developed, which were efficacious in providing short-term protection to challenge with homologous but not heterologous IBV strains. Levels of circulating IBV did not diminish substantially because many strains of IBV co-circulate in chicken populations. Recombination between the vaccine and circulating strains resulted in the emergence of novel strains of IBV.

ding short-term protection to challenge with homologous but not heterologous IBV strains. Levels of circulating IBV did not diminish substantially because many strains of IBV co-circulate in chicken populations. Recombination between the vaccine and circulating strains resulted in the emergence of novel strains of IBV. Live attenuated vaccines were also developed for a swine coronavirus, transmissible gastroenteritis virus (TGEV), which causes fatal diarrhea with associated high mortality in very young pigs.26 These vaccines were administered to pregnant sows but did not protect piglets to a great extent; the use of virulent virus in sows was more successful in protecting baby animals from lethal disease. Remarkably, however, a deletion variant of TGEV, porcine respiratory coronavirus (PRC), appeared in swine populations in North America and Eurasia.27 PRC caused only a mild respiratory disease, but induced an immune response that was cross-reactive and protective against TGEV, resulting in the disappearance of TGEV from most locales.

ever, a deletion variant of TGEV, porcine respiratory coronavirus (PRC), appeared in swine populations in North America and Eurasia.27 PRC caused only a mild respiratory disease, but induced an immune response that was cross-reactive and protective against TGEV, resulting in the disappearance of TGEV from most locales. Finally, feline infectious peritonitis virus (FIPV) causes a lethal granulomatous disease in domestic cats and other felines, with wet (pyogranulomatous, effusive) and dry (classic granulomatous) forms.28 FIP is uncommon and most often occurs in animals chronically infected with feline coronavirus (FCV), which mutates during the course of persistence. A vaccinia virus-based vaccine expressing the FIPV surface (S) glycoprotein was developed, and was shown to induce high levels of anti-FIPV neutralizing antibody.29 However, this anti-S antibody was not protective against challenge with virulent FIPV. Rather, it induced an antibody-dependent accelerated and enhanced disease after challenge. Of note, antibody-dependent enhancement has never been observed in naturally infected felines, but the possibility that it might develop has been a concern as vaccines for SARS-CoV and MERS-CoV are developed.30

ith virulent FIPV. Rather, it induced an antibody-dependent accelerated and enhanced disease after challenge. Of note, antibody-dependent enhancement has never been observed in naturally infected felines, but the possibility that it might develop has been a concern as vaccines for SARS-CoV and MERS-CoV are developed.30 4 Development of anti-SARS-CoV and MERS-CoV vaccines Vaccines useful for preventing SARS or MERS have been developed, based on information learned from the studies described above (Table 1 ). Because both SARS and MERS tend to spread extensively within hospital settings, initial efforts were directed at developed reagents that could be used for passive immunization; more recent efforts have focused on methods useful for active immunization. In this section, vaccines targeting SARS-CoV are described first, since many of the approaches used in developing MERS vaccines were initially investigated in the context of SARS.Table 1 Middle East respiratory syndrome coronavirus vaccines

on; more recent efforts have focused on methods useful for active immunization. In this section, vaccines targeting SARS-CoV are described first, since many of the approaches used in developing MERS vaccines were initially investigated in the context of SARS.Table 1 Middle East respiratory syndrome coronavirus vaccines Table 1Vaccine Target Use Advantages Problems Anti-MERS-CoV monoclonal antibodies Surface (S) glycoprotein Passive immunization; prophylaxis or treatment at early times p.i. High titer preparations; can be produced in large amounts Short half-life; needs to be re-administered for continued efficacy Human polyclonal anti-MERS-CoV antibodies Virus structural proteins Passive immunization; treatment at early times p.i. Polyclonal antibody so antibody escape unlikely; human antibody Short half-life; needs to be re-administered for continued efficacy; few MERS survivors available as donors Inactivated virion vaccines Virus structural proteins; anti-S neutralizing antibodies most important Active immunization High titer antibody to S protein Response may not be long term; on challenge may induce immunopathological disease; may be ineffective in aged populations Live attenuated vaccines (e.g., viruses deleted in envelope (E) protein; viruses with reduced fidelity (mutated in nsp14) Mostly virus structural proteins Active immunization Generally safe; induce antibody and T-cell responses; long-term immunity May not be safe in immunocompromised patients; may regain virulence by reversion or recombination with circulating CoV Viral vector (attenuated) vaccines: poxvirus, AAV adenovirus, parainfluenza virus, rabies virus, measles virus, VSV S protein Active immunization Safe; non-replicating; induce antibody and T-cell responses Long-term immunity, but not as long as live attenuated vaccines Replicon particles (e.g., VEEV or VSV-based) S protein or any viral protein Active immunization Safe; non-replicating; induce antibody and T-cell responses; useful for mucosal immunity Production is complex Subunit vaccines (e.g.

ibody and T-cell responses Long-term immunity, but not as long as live attenuated vaccines Replicon particles (e.g., VEEV or VSV-based) S protein or any viral protein Active immunization Safe; non-replicating; induce antibody and T-cell responses; useful for mucosal immunity Production is complex Subunit vaccines (e.g. RBD of S protein) Generally S protein Active immunization Safe; non-replicating; induce high antibody titers; may also induce T-cell responses Duration of response not known DNA vaccines Generally S protein Active immunization Safe; induce high antibody titers and T-cell responses Immunogenicity variable; may induce anti-DNA immune response MERS-CoV, Middle East respiratory syndrome coronavirus; p.i., post infection; AAV, adeno-associated virus; VSV, vesicular stomatitis virus; VEEV, Venezuelan equine encephalitis virus; RBD, receptor binding domain.

uce high antibody titers and T-cell responses Immunogenicity variable; may induce anti-DNA immune response MERS-CoV, Middle East respiratory syndrome coronavirus; p.i., post infection; AAV, adeno-associated virus; VSV, vesicular stomatitis virus; VEEV, Venezuelan equine encephalitis virus; RBD, receptor binding domain. 4.1 Passive immunization 4.1.1 SARS Monoclonal antibodies (mAb) with neutralizing activity against SARS-CoV have been isolated from non-immune human volunteers.31, 32 The advantage of this approach is that protective antibodies can be isolated, cloned, and propagated without the need to obtain patient specimens. Other approaches have included identifying and cloning memory B-cells obtained from SARS survivors and amplifying those that produce the most potently neutralizing antibodies.33 In all of these vaccines, neutralizing antibodies have been directed against the S protein. Stockpiled anti-SARS-CoV antibodies would be especially useful in the healthcare or family setting to provide prophylaxis or treatment if administered very soon after exposure.

roduce the most potently neutralizing antibodies.33 In all of these vaccines, neutralizing antibodies have been directed against the S protein. Stockpiled anti-SARS-CoV antibodies would be especially useful in the healthcare or family setting to provide prophylaxis or treatment if administered very soon after exposure. Convalescent sera from SARS survivors have also been used to treat patients.34, 35 Efficacy was not demonstrated, but this may well have reflected the administration of sera after disease had already developed; the clinical presentation of SARS is non-specific, making it difficult to identify infected patients at an early time during the disease course. Convalescent sera would be most useful in an outbreak setting in which a large fraction of patients with respiratory disease might be expected to have SARS and therefore benefit from treatment.

ntation of SARS is non-specific, making it difficult to identify infected patients at an early time during the disease course. Convalescent sera would be most useful in an outbreak setting in which a large fraction of patients with respiratory disease might be expected to have SARS and therefore benefit from treatment. 4.1.2 MERS Similar strategies have been used to isolate and amplify antibodies with MERS-CoV neutralizing activity. Initial reports described the isolation of neutralizing antibodies from naïve human antibody populations using phage display and yeast display.36, 37, 38 In another approach, a mAb with high avidity for the MERS-CoV S proteins was isolated from B-cells harvested from a MERS patient after cloning into a mammalian expression system.39 This antibody was shown to efficiently accelerate the kinetics of virus clearance and diminish pathological changes in mice infected with MERS-CoV. Mice are not naturally infectable by MERS-CoV because the virus cannot use the mouse MERS-CoV receptor (dipeptidyl peptidase, DPP4) to enter cells. In this instance, mice were sensitized to MERS-CoV by prior transduction with an adenovirus engineered to express human DPP4 (hDPP4).40 In another approach, fully human antibodies with MERS-CoV neutralizing activity were developed using mice that expressed human antibody heavy and κ light chains. In this study, efficacy was examined in mice that had been engineered to express hDPP4 in lieu of mDPP4 (‘knock-in’, KI mice).41 The humanized anti-S mAbs accelerated virus clearance and reduced pathological changes in mouse lungs.

tivity were developed using mice that expressed human antibody heavy and κ light chains. In this study, efficacy was examined in mice that had been engineered to express hDPP4 in lieu of mDPP4 (‘knock-in’, KI mice).41 The humanized anti-S mAbs accelerated virus clearance and reduced pathological changes in mouse lungs. The use of convalescent sera from MERS survivors has been proposed based on studies of SARS patients.35 However, the limited availability of convalescent sera may make its use infeasible. Camels are considered the primary reservoir for human MERS and appear to be periodically reinfected by the virus. Consequently, MERS-CoV antibody titers are elevated. The administration of sera from previously infected camels to MERS-CoV challenged hDPP4-transduced mice was shown to accelerate MERS-CoV clearance and reduce pathological changes in the lungs.42

ir for human MERS and appear to be periodically reinfected by the virus. Consequently, MERS-CoV antibody titers are elevated. The administration of sera from previously infected camels to MERS-CoV challenged hDPP4-transduced mice was shown to accelerate MERS-CoV clearance and reduce pathological changes in the lungs.42 4.2 Active immunization 4.2.1 SARS Most vaccines have been directed at developing anti-S neutralizing antibody responses. Vaccines have included inactivated whole virus vaccines, live attenuated virus DNA vaccines, viral vector vaccines, subunit vaccines, and DNA vaccines. DNA vaccines were shown to induce anti-S antibodies in mice and were later shown to induce virus-specific neutralizing antibody and T-cell responses in a phase I human trial.43, 44 Inactivated SARS-CoV vaccines were developed and tested in experimentally infected animals as well as in phase I human trials.45 These vaccines induced strong anti-S antibody responses if administered with adjuvants such as β-propiolactone or formalin,46 but subsequent studies suggested that they also induced eosinophilia and other signs of immunopathological disease upon challenge.47, 48 Human phase I trials testing this reagent were halted based on these putative immunopathological changes. Live attenuated vaccines offer the best opportunity for developing both antibody and T-cell responses without eosinophilic infiltration or other manifestations of immunopathological disease in the lungs.

nge.47, 48 Human phase I trials testing this reagent were halted based on these putative immunopathological changes. Live attenuated vaccines offer the best opportunity for developing both antibody and T-cell responses without eosinophilic infiltration or other manifestations of immunopathological disease in the lungs. Engineering of live attenuated vaccines has been facilitated by the development of reverse genetics systems for SARS-CoV, as well as other coronaviruses, including MERS-CoV.49, 50 Using one of these methodologies, viruses deleted in the small envelope (E) protein were developed. These viruses were shown to be attenuated and to induce protective humoral and cell-based immune responses in hamsters and mice after SARS-CoV challenge.51, 52, 53 Further investigations showed that this vaccine was not genetically stable, with partial duplication of the transmembrane (M) protein detected upon repeated passage. Remarkably, this genetic change resulted in re-acquisition of a PDZ binding motif (PBM) important for protein–protein interactions.54 If the E protein was only partially deleted so that the PBM was retained, the virus was genetically stable, attenuated, and immunogenic. In another approach, virus mutated at the catalytic site of a protein critical for genome fidelity during replication (nsp14) resulted in an attenuated virus that did not revert upon repeated passage in cells and mice, was safe even in highly immunocompromised mice, and induced a strong anti-S antibody response.55 Further efforts to maximize the biosafety of these live attenuated vaccines include the introduction of additional mutations into non-essential proteins or into non-coding regions of the genomic RNA, which minimize the likelihood of a virulent virus arising after recombination with circulating coronavirus strains.56, 57

urther efforts to maximize the biosafety of these live attenuated vaccines include the introduction of additional mutations into non-essential proteins or into non-coding regions of the genomic RNA, which minimize the likelihood of a virulent virus arising after recombination with circulating coronavirus strains.56, 57 Attenuated poxvirus,58 adenovirus,59, 60 adeno-associated virus,61 parainfluenza virus,62 rabies virus,63 measles virus,64, 65 and vesicular stomatitis virus66 vectors expressing either full-length SARS-CoV S protein or the S1 extracellular domain have been engineered. These vaccines also induced high levels of SARS-CoV neutralizing antibody titers. Venezuelan equine encephalitis virus replicons (VRPs) expressing viral proteins have been shown to induce potent T-cell and antibody responses and to act as self-adjuvants. A major advantage of VRPs is that since they are non-replicating, they are not infectious and will not recombine with circulating CoV to generate new variants. VRPs expressing the S or nucleocapsid (N) protein, like other vaccines, were found to be less effective in senescent mice and VRP-N was reported to induce immunopathological disease.67 Similar immunopathology was observed after SARS-CoV challenge of mice previously immunized with the N protein.68

generate new variants. VRPs expressing the S or nucleocapsid (N) protein, like other vaccines, were found to be less effective in senescent mice and VRP-N was reported to induce immunopathological disease.67 Similar immunopathology was observed after SARS-CoV challenge of mice previously immunized with the N protein.68 Potent neutralizing antibody responses have also been induced in mice immunized with constructs expressing the receptor binding domain, the part of the SARS-CoV protein that actually binds to the receptor (angiotensin-converting enzyme 2, ACE2) on target cells.69 These subunit vaccines exhibit high safety profiles and have minimal side effects in addition to being immunogenic. 4.2.2 MERS Many of the approaches described above have been used to develop MERS vaccines. Recombinant adenoviruses,70 poxviruses,71 and measles virus72 expressing full-length S protein or the extracellular S1 domain have been engineered and tested in experimentally infected animals. All were able to induce an anti-S protein antibody response. Although not examined, it is likely that some or all of them also induced CD8 and CD4 T-cell responses. Live attenuated MERS-CoV vaccines have not yet been described, although it is likely, based on SARS-CoV data,55 that virus mutated in the catalytic site of the exonuclease of nsp14 would be an excellent vaccine candidate.

ugh not examined, it is likely that some or all of them also induced CD8 and CD4 T-cell responses. Live attenuated MERS-CoV vaccines have not yet been described, although it is likely, based on SARS-CoV data,55 that virus mutated in the catalytic site of the exonuclease of nsp14 would be an excellent vaccine candidate. DNA vaccines that induce MERS-CoV-specific antibody and T-cell responses have been described and shown to be efficacious in non-human primates.73 Vaccines expressing the MERS-CoV receptor binding domain (RBD) induced potent neutralizing antibodies in mice and neutralizing antibodies and T-cell responses in non-human primates.74, 75 Vaccination resulted in accelerated virus clearance and diminished pathological changes, but did not prevent infection. While vaccine development usually targets human populations, MERS-CoV infects a much greater number and higher percentage of camels than humans in the Arabian Peninsula and in Africa. Thus camel vaccination is an approach to decrease the amount of circulating virus and also diminish the amount of virus secreted by infected animals. A recent study showed that this approach is feasible. Camels were immunized with an orthopoxvirus vector (modified vaccinia virus Ankara, MVA) expressing the S protein.76 After challenge with MERS-CoV at 3 weeks after boosting, clinical signs (rhinitis) and infectious virus titers in the upper respiratory tract, the main site of replication in the camel, were diminished in immunized compared to control animals.

rus vector (modified vaccinia virus Ankara, MVA) expressing the S protein.76 After challenge with MERS-CoV at 3 weeks after boosting, clinical signs (rhinitis) and infectious virus titers in the upper respiratory tract, the main site of replication in the camel, were diminished in immunized compared to control animals. 5 Conclusions and future directions While several promising MERS vaccine candidates are under development, several issues need to be resolved. First, an important consideration is whether humans or, alternatively, camels should be vaccinated. Humans but not camels develop severe respiratory disease, but only a relatively small total number of infected individuals have been identified and many of these have had co-morbidities, which would impair vaccine responsiveness. In the absence of a greater disease burden in human populations, it seems unlikely that human vaccination would ever be economically viable. On the other hand, a high percentage of camels are infected with MERS-CoV and vaccination reduces virus load, although without inducing sterilizing immunity. The longevity of the camel immune response is not known but may be short, since camels appear to be readily re-infected with MERS-CoV. Further, the large size of camels plus the number of camels potentially requiring immunization would cause logistical problems.

rus load, although without inducing sterilizing immunity. The longevity of the camel immune response is not known but may be short, since camels appear to be readily re-infected with MERS-CoV. Further, the large size of camels plus the number of camels potentially requiring immunization would cause logistical problems. Second, most vaccines induce anti-S neutralizing antibody responses. The receptor binding part of the S protein, the target for most neutralizing antibodies, is the most variable part of the S protein so that antibodies are highly strain-specific. While the S protein of MERS-CoV has not shown evidence of mutations that result in antibody evasion, this is still a possibility because coronaviruses are prone to mutation and recombination. Thus, targeting the S protein may provide protection against MERS-CoV but may not be useful against either a closely related strain or one that evolves in response to immune or other pressure in humans. This possibility was highlighted in a recent study that showed that two bat strains of SARS-like CoV were closely related to human SARS-CoV and used the same ACE2 receptor to enter cells.77, 78 However, anti-SARS-CoV S antibodies could not neutralize one of these strains.78

sponse to immune or other pressure in humans. This possibility was highlighted in a recent study that showed that two bat strains of SARS-like CoV were closely related to human SARS-CoV and used the same ACE2 receptor to enter cells.77, 78 However, anti-SARS-CoV S antibodies could not neutralize one of these strains.78 Third, T-cell responses in MERS and SARS survivors have not been investigated widely, but these tend to target more conserved regions of the viral genome and will provide protection against strains that differ in the RBD. Analysis of the T-cell response is facilitated by the identification of CD8 or CD4 T-cell epitopes. T-cell epitopes have been identified in some inbred strains of mice,40 but are more difficult to identify in human populations because of inter-individual human leukocyte antigen (HLA) diversity. The extent to which vaccines should be formulated to induce T-cell responses as well as neutralizing antibody responses is not yet resolved. Fourth, prior to use in humans, vaccines need to be carefully evaluated in experimentally infected animals. No laboratory animal infected with MERS-CoV develops disease with the same pathogenesis as occurs in patients with severe respiratory disease. Marmosets develop severe disease in some laboratory settings but not all.79, 80 Even if a lethal mouse-adapted MERS-CoV is identified, disease in the mouse and untoward effects of vaccines may not mirror the human infection sufficiently. Better animal models for MERS would facilitate more useful and accurate in vivo vaccine evaluation.

severe disease in some laboratory settings but not all.79, 80 Even if a lethal mouse-adapted MERS-CoV is identified, disease in the mouse and untoward effects of vaccines may not mirror the human infection sufficiently. Better animal models for MERS would facilitate more useful and accurate in vivo vaccine evaluation. In conclusion, it was learned from the Ebola pandemic that preparedness for epidemic spread of a virus that has never exhibited such spread in the past is critical. So far, there has not been an upsurge in MERS cases during the Hajj or Umrah pilgrimages. Nevertheless, consideration of how to develop tools for passive and active immunization is critical. Acknowledgements Supported in part by grants from the National Institutes of Health, USA (RO1 AI091322 and PO1 AI060699). Conflict of interest: The authors report no conflicts of interest, financial or otherwise.

rn after the initiation of MDA are of particular interest, as they provide an evaluation sample on the interruption of transmission. The current study monitored the impact of 12 rounds of MDA and the deployment of a universal LLIN coverage intervention on the prevalence of CFA in standard one school children in Rufiji. Materials and methods Study site and population All study sites are located in Rufiji District in southeastern Tanzania (7°57′ S and 38°43′ E), which had a population of 217 274 people in 2012 (National Census 2012, National Bureau of Statistics, Tanzania). The district was purposively selected for this study due to its history of high LF transmission, with a baseline prevalence of W. bancrofti CFA reported at 49% before the start of control activities (MoHSW 2012, unpublished). The study sampled participants from the community (heads of households) and primary schools (standard one children) in the villages of Nyamisati, Mchukwi, Nyanjati, Bungu, and Nyambili in Rufiji. The villages were purposively selected to represent the diverse geographical features of the district. Nyamisati village lies in the coastal belt of the Indian Ocean, while Mchukwi, Nyanjati, Nyambili, and Bungu are inland villages. Each village has only one primary school and all standard one pupils in each school were recruited for the study.

ely selected to represent the diverse geographical features of the district. Nyamisati village lies in the coastal belt of the Indian Ocean, while Mchukwi, Nyanjati, Nyambili, and Bungu are inland villages. Each village has only one primary school and all standard one pupils in each school were recruited for the study. A list of heads of households was obtained from the village officials in the enrolled villages. A simple random sampling technique was applied to obtain participants from each village. The selected participants were identified and recruited into the study. The enrolled heads of households were interviewed to gather information on MDA coverage and ITN use. The NLFEP in Rufiji District was implemented in 2002. By 2014, the programme had administered 12 rounds of annual MDA, with an interruption in 2005 due to logistical issues. The current study was conducted from April to May in 2012 and was repeated at the same time in 2015, coinciding with the 9th and 12th rounds of MDA, respectively.

Introduction Lymphatic filariasis (LF) is a disease of major public health significance, affecting people in the tropical areas of Africa, India, and South and Central America (Michael et al., 1996). It is caused by the filarial nematode Wuchereria bancrofti and is transmitted by a variety of genera of mosquitoes, including Culex, Anopheles, Aedes, and Mansonia (Bockarie Moses et al., 2009). In Sub-Saharan Africa, the most important species of mosquito vector are Anopheles and Culex (Bockarie Moses et al., 2009). Members of the Anopheles gambiae complex Anopheles funestus group and Culex quinquefasciatus have been identified as important filarial vectors in Tanzania (Rwegoshora et al., 2005, Simonsen et al., 2010). Globally, it is estimated that 120 million people are infected with the parasite and approximately one billion are at risk of infection (WHO, 2005). Manifestations of the disease include chronic lymphoedema, elephantiasis involving the limbs and sometimes the genital area, chyluria, and attacks of acute adenolymphangitis (Gasarasi et al., 2000). LF has been ranked as the leading cause of permanent disability in the world (Zeldenryk et al., 2011).

ction (WHO, 2005). Manifestations of the disease include chronic lymphoedema, elephantiasis involving the limbs and sometimes the genital area, chyluria, and attacks of acute adenolymphangitis (Gasarasi et al., 2000). LF has been ranked as the leading cause of permanent disability in the world (Zeldenryk et al., 2011). The World Health Organization (WHO) launched a global programme for the elimination of LF (GPELF) in 2000, with the goal of eliminating the disease globally by 2020 (WHO, 2002). The main strategy proposed to achieve LF elimination is the provision of repeated mass drug administration (MDA) with a combination of albendazole and either diethylcarbamazine or ivermectin to people living in endemic areas who are at risk of the disease (Gyapong et al., 2005, Tisch et al., 2005). These drug combinations are mainly microfilaricidal, aimed at reducing the transmission of the parasite. It is assumed that once the community has been treated long enough, levels of microfilariae will be reduced below the required threshold to sustain transmission (Ottesen, 2000). The period required to achieve this goal has been estimated at 4 to 6 years, which corresponds to the reproductive lifespan of the adult W. bancrofti worm (Ottesen, 2012).

ommunity has been treated long enough, levels of microfilariae will be reduced below the required threshold to sustain transmission (Ottesen, 2000). The period required to achieve this goal has been estimated at 4 to 6 years, which corresponds to the reproductive lifespan of the adult W. bancrofti worm (Ottesen, 2012). It has been suggested that the interruption of transmission of this infection depends on the proportion of the population receiving antifilarial drugs every year. Furthermore, it was estimated that four to six annual rounds of MDA, with a minimum effective coverage of 60–70% of the target population, would be sufficient to interrupt transmission (Michael et al., 2004). However 5 years after the launch of the GPELF, it was concluded that more than four to six annual rounds of MDA may be required to interrupt transmission in many endemic areas (Swaminathan et al., 2012).

e coverage of 60–70% of the target population, would be sufficient to interrupt transmission (Michael et al., 2004). However 5 years after the launch of the GPELF, it was concluded that more than four to six annual rounds of MDA may be required to interrupt transmission in many endemic areas (Swaminathan et al., 2012). Tanzania initiated the implementation of LF control with the launch of the Tanzanian National Lymphatic Filariasis Elimination Programme (NLFEP) in 1997 (Malecela Mwele et al., 2009). The control strategy adopted by NLFEP was to apply annual MDA with a combination of ivermectin (150–200 mg/kg) and albendazole (400 mg) to all individuals aged 5 years and above living in selected endemic areas. In the year 2000, the Tanzanian NLFEP launched its first MDA campaign in Mafia District, in which 45 000 people were treated (Malecela Mwele et al., 2009). The programme has expanded to cover more than 13 million people treated at least once, and the goal is to expand the programme to the entire at-risk population of around 39 million people (Kisoka et al., 2014).

its first MDA campaign in Mafia District, in which 45 000 people were treated (Malecela Mwele et al., 2009). The programme has expanded to cover more than 13 million people treated at least once, and the goal is to expand the programme to the entire at-risk population of around 39 million people (Kisoka et al., 2014). Rufiji District in southeastern Tanzania started implementing MDA in 2002, with a baseline W. bancrofti circulating filarial antigen (CFA) prevalence ranging from 49% to 64% among the community members aged 5 years and above (Ministry of Health and Social Welfare (MoHSW) 2012, unpublished). From 2012, the MDA programme coincided with the universal distribution of long-lasting insecticide-treated nets (LLINs) by the malaria control programme in Tanzania, which aimed to cover 80% of the general population (West et al., 2012). The use of insecticide-treated nets (ITNs) for malaria control has shown a significant effect in lowering filarial rates (Odermatt et al., 2008, Ashton et al., 2011, Njenga et al., 2011).

d nets (LLINs) by the malaria control programme in Tanzania, which aimed to cover 80% of the general population (West et al., 2012). The use of insecticide-treated nets (ITNs) for malaria control has shown a significant effect in lowering filarial rates (Odermatt et al., 2008, Ashton et al., 2011, Njenga et al., 2011). Monitoring the effect of MDA on the transmission of lymphatic filariasis is crucial for measuring progress towards the elimination goal and also forms the basis for establishing treatment endpoints. Studies have indicated that the monitoring of circulating filarial antigen in young children who are born during the intervention period is a good indicator for assessing the impact of MDA in a community (Simonsen et al., 2010). In endemic regions, children are most susceptible to acquiring the infection because of their lack of immunity and high exposure to infective larvae; infections established in childhood may act as the reservoir for future disease later in life (Lammie et al., 1994).

impact of MDA in a community (Simonsen et al., 2010). In endemic regions, children are most susceptible to acquiring the infection because of their lack of immunity and high exposure to infective larvae; infections established in childhood may act as the reservoir for future disease later in life (Lammie et al., 1994). The use of new sensitive and highly specific diagnostic tools for the detection of CFA released by adult W. bancrofti parasites has shown that many children acquire the infection earlier than was previously thought and that often a considerable proportion of young children are CFA-positive. As reduced transmission will lead to reduced acquisition of infection, it has been suggested that young children be screened to assess the effectiveness of transmission interventions during the LF elimination programme. Children born after the initiation of MDA are of particular interest, as they provide an evaluation sample on the interruption of transmission. The current study monitored the impact of 12 rounds of MDA and the deployment of a universal LLIN coverage intervention on the prevalence of CFA in standard one school children in Rufiji.

iji District was implemented in 2002. By 2014, the programme had administered 12 rounds of annual MDA, with an interruption in 2005 due to logistical issues. The current study was conducted from April to May in 2012 and was repeated at the same time in 2015, coinciding with the 9th and 12th rounds of MDA, respectively. Detection of CFA and MDA participation On the survey dates, all standard one children were invited to participate in the blood screening using the class registers containing the names and sexes of the pupils. They were examined for CFA using immunochromatographic test cards (ICT) (BinaxNOW Filariasis; Inverness Medical Innovations Inc.); the manufacturer’s instructions were followed. Prior to the field survey, two cards from the lot were tested in the laboratory using a positive control obtained from the National Institute for Medical Research, Tanzania. The results for both cards were positive. In brief, a finger prick was conducted using a sterile disposable lancet after cleaning the finger with an alcohol swab. One hundred microlitres of finger-prick blood was collected using a sterile disposable capillary tube and applied to the test card. The results were read after exactly 10 min as positive, negative, or undetermined. Just after the blood test, the children were interviewed on their age and participation in the previous MDA rounds. To guide the pupils, the drugs used in the MDA were displayed to them and they were then asked if they had swallowed the drugs or not.

ere read after exactly 10 min as positive, negative, or undetermined. Just after the blood test, the children were interviewed on their age and participation in the previous MDA rounds. To guide the pupils, the drugs used in the MDA were displayed to them and they were then asked if they had swallowed the drugs or not. Assessment of community participation in MDA and utilization of ITNs The treatment coverage was obtained using two different methods. The first method involved obtaining the official programme coverage for all administrative wards from the Rufiji District neglected tropical disease (NTD) coordination office. The second method involved interviews with the heads of households within the study villages to determine their participation in the MDA campaign and utilization of ITNs. Participants were asked whether they had participated in previous MDAs and whether household members slept under ITNs. The investigators asked if they could examine the respondents’ bedrooms for the presence of ITNs in each household. Moreover, demographic information for the respondents and their reasons for not taking drugs during MDA rounds were also recorded.

icipated in previous MDAs and whether household members slept under ITNs. The investigators asked if they could examine the respondents’ bedrooms for the presence of ITNs in each household. Moreover, demographic information for the respondents and their reasons for not taking drugs during MDA rounds were also recorded. Data analysis Data were entered into Excel and subsequently analyzed using IBM SPSS Statistics for Windows, version 20.0 (IBM Corp., Armonk, NY, USA). For the data analysis, respondents were categorized into three age groups: 15–29 years, 30–49 years, and ≥50 years. Educational status was categorized into four groups: those with no formal education, primary education, secondary education, and post-secondary education. The duration of stay in the village was categorized into four groups: since birth, less than 1 year, 1–5 years, and more than 5 years. The different categories were compared with the Chi-square test. p-Values of ≤0.05 were considered statistically significant.

ucation, secondary education, and post-secondary education. The duration of stay in the village was categorized into four groups: since birth, less than 1 year, 1–5 years, and more than 5 years. The different categories were compared with the Chi-square test. p-Values of ≤0.05 were considered statistically significant. Ethics This study was reviewed and approved by the Muhimbili University of Health and Allied Sciences Research Ethics Review Board, as well as the WHO Ethics Review Committee. Permission to conduct the study was obtained from the Executive Director of Rufiji District. The parents/guardians and community were informed of the purpose of the survey through the village meetings. Children were recruited at their respective schools and given consent forms (in Kiswahili) to give to their parents/guardians for consenting. Parents/guardians or school pupils who refused to participate in the surveys were excluded. Verbal consent to participate in the community component of the study was sought from the heads of households prior to the interviews.

ols and given consent forms (in Kiswahili) to give to their parents/guardians for consenting. Parents/guardians or school pupils who refused to participate in the surveys were excluded. Verbal consent to participate in the community component of the study was sought from the heads of households prior to the interviews. Results CFA prevalence and treatment coverage in pupils In 2012, 413 children aged between 6 and 9 years were screened for CFA in five selected primary schools. Of those screened, 236 (57.1%) were female and 177 (42.9%) were male. A total of 59 (14.3%) screened pupils were positive for W. bancrofti CFA (Table 1). The prevalence of CFA increased significantly with increasing age of the pupils (Chi-square test of trends, p = 0.016; Table 1). Just a third (33.7%) of the children screened in 2012 reported having swallowed the drugs in the 2011 MDA round. The prevalence of CFA was not significantly different among those who participated (12.2%) and those who did not participate (15.3%) in the 2011 MDA (Chi-square test, p = 0.4).Table 1 CFA status among standard one school children aged 6 to 9 years in five different primary schools in Rufiji District.

he 2011 MDA round. The prevalence of CFA was not significantly different among those who participated (12.2%) and those who did not participate (15.3%) in the 2011 MDA (Chi-square test, p = 0.4).Table 1 CFA status among standard one school children aged 6 to 9 years in five different primary schools in Rufiji District. Table 1Name of school 2012 Survey 2015 Survey No. tested Sex ratioa Mean age (years) CFA positive (%) No. tested Sex ratioa Mean age (years) CFA positive (%) Nyanjati 76 0.6 6.9 7 (9.2) 173 0.7 6.87 0.0 Nyamisati 81 0.7 7.6 15 (18.5) 132 1.8 7.86 0.0 Mchukwi 79 0.9 7.5 12 (15.1) 130 1.2 7.68 0.0 Bungu 97 0.7 7.8 16 (16.5) 122 1.1 7.66 0.0 Nyambili 80 0.9 7.4 9 (11.3) 102 1.6 7.73 0.0 Total 413 0.8 7.4 59 (14.3) 659 1.1 7.51 0.0 CFA, circulating filarial antigen. a Ratio of males to females. In 2015, a total of 659 standard one children aged between 6 and 9 years (mean 7.5 years) were tested for W. bancrofti CFA in the five selected schools. Of the tested pupils, 346 (52.5%) were male and 313 (47.5%) were female. None of the pupils examined tested positive for W. bancrofti CFA (Table 1). With respect to participation in MDA, 54.3% (n = 358) of the tested pupils reported that they had always swallowed the drugs when given out in each MDA round. The same proportion of pupils (54.3%) reported having swallowed the drugs during the last MDA that took place in October 2014 (12th round). Statistical analysis revealed that participation in the MDA was not significantly different with regard to sex (Chi-square test, p = 0.012) or age (Chi-square test, p = 0.15) of the pupils.

proportion of pupils (54.3%) reported having swallowed the drugs during the last MDA that took place in October 2014 (12th round). Statistical analysis revealed that participation in the MDA was not significantly different with regard to sex (Chi-square test, p = 0.012) or age (Chi-square test, p = 0.15) of the pupils. Reported and surveyed community treatment coverage MDA treatment coverage was assessed using two methods: the official NLFEP-reported coverage from the district NTD office and interviews conducted in the study villages (Table 2, Table 3). The NLFEP reported annual treatment coverage for Rufiji District ranging from 54.3% to 94.0% during the years 2002–2014. MDA was not administered in 2005 (Table 2).Table 2 Official NLFEP treatment coverage for Rufiji District where the five study schools are located. Table 2Year Target population No. treated % treated 2014 226 939 154 319 68.7 2013 217 274 168 930 77.7 2012 203 835 124 747 61.2 2011 170 606 150 133 88 2010 166 682 113 677 68.2 2009 162 848 130 279 80 2008 159 103 115 350 72.5 2007 155 443 116 583 75 2006 151 868 85 046 56 2005 – – – 2004 144 962 136 267 94 2003 141 628 123 216 87 2002 138 370 75 135 54.3 NLFEP, Tanzanian National Lymphatic Filariasis Elimination Programme. Source: Tanzania Neglected Tropical Disease Control Programme. Table 3 Drug uptake and ownership and use of insecticide-treated nets in relation to population characteristics among interviewed adults in Rufiji District, southeastern Tanzania.

Table 2Year Target population No. treated % treated 2014 226 939 154 319 68.7 2013 217 274 168 930 77.7 2012 203 835 124 747 61.2 2011 170 606 150 133 88 2010 166 682 113 677 68.2 2009 162 848 130 279 80 2008 159 103 115 350 72.5 2007 155 443 116 583 75 2006 151 868 85 046 56 2005 – – – 2004 144 962 136 267 94 2003 141 628 123 216 87 2002 138 370 75 135 54.3 NLFEP, Tanzanian National Lymphatic Filariasis Elimination Programme. Source: Tanzania Neglected Tropical Disease Control Programme. Table 3 Drug uptake and ownership and use of insecticide-treated nets in relation to population characteristics among interviewed adults in Rufiji District, southeastern Tanzania. Table 32012 survey 2015 survey Population characteristics No. of people (% per category) Drug uptake (%) p-Value Population characteristics No. of people (% per category) Drug uptake (%) p-Value Gender (n = 246) Gender (n = 868) Female 159 (64.6) 72 (45.2) 0.007  Female 471 (54.2) 277 (58.2) 0.254 Male 87 (35.4) 55 (63.2)  Male 397 (45.8) 218 (54.9 Age group (years) (n = 246) Age groups (years) (n = 868) 15–29 94 (38.2) 43 (45.7) 0.001  15–29 345 (39.7) 182 (52.7) 0.000 30–49 115 (46.7) 65 (56.5)  30–49 357 (41.1) 228 (63.7) ≥50 37 (15.1) 19 (51.3)  ≥50 166 (19.1) 85 (73.2) Education (n = 246) Education (n = 868) Not gone to school 40 (25.8) 16 (40) 0.001  No education 224 (25.8) 119 (53.6) 0.001 Primary education 88 (35.8) 45 (51.1)  Primary education 487 (56.2) 286 (58.7) Secondary education 86 (35.0) 44 (51.1)  Secondary education 139 (16.0) 80 (57.6) Post-secondary education 32 (13.0) 20 (62.5)  Post-secondary education 18 (2.1) 10 (55.6) Stay in the village (n = 246) Stay in the village (n = 868) Since birth 116 (47.2) 64 (55.2) 0.7  Since birth 413 (47.6) 206 (49.9) 0.004 Less than 1 year 70 (28.5) 33 (47.1)  Less than 1 year 48 (5.5) 23 (47.9) 1–5 years 39 (15.9) 20 (51.3)  1–5 years 103 (11.9) 69 (66.9) More than 5 years 21 (8.5) 10 (41.7)  More than 5 years 279 (32.2) 172 (61.3) No answer 0 0  No answer 25 (2.8) 25 (100) Participated in the last MDA (n = 246) Participated in the last MDA (n = 868) Swallowed drugs 127 (51.6) 0  Swallowed drugs 495 (57.0) 0.000 Did not swallow drugs 119 (48.4)  Did not swallow drugs 341 (39.3) Don’t remember 32 (3.6) Possess LLINs (n = 246) Possess LLINs (n = 868) Own LLINs 156 (63.4) 79 (50.6) 0.684  Own LLINs 803 (92.5) 440 (54.7) 0.000 Do not possess LLINs 90 (36.6) 48 (53.5)  Do not possess LLINs 65 (7.5) 55 (84.6) Used LLINs last night (n = 246) Used LLINs last night (n = 764) Used 146 (59.2) 74 (50.7) 0.721  Used 576 (75.4) 321 (5.8) 0.004 Not used 100 (40.7) 53 (53)  Not used 188 (24.6) 153 (81.4) MDA, mass drug administration; LLINs, long-lasting insecticide-treated nets.

.6) 48 (53.5)  Do not possess LLINs 65 (7.5) 55 (84.6) Used LLINs last night (n = 246) Used LLINs last night (n = 764) Used 146 (59.2) 74 (50.7) 0.721  Used 576 (75.4) 321 (5.8) 0.004 Not used 100 (40.7) 53 (53)  Not used 188 (24.6) 153 (81.4) MDA, mass drug administration; LLINs, long-lasting insecticide-treated nets. Surveyed treatment coverage was assessed by interviews conducted in 2012 and 2015. Demographic information of the respondents in both surveys is presented in Table 3. In the 2012 survey, in which 246 heads of households were interviewed, it was found that the drug uptake in the 2011 MDA campaign was 51.6%. Moreover, participation in more than four previous rounds of MDA was reported to be 33.0%. Analysis by gender, age group, and education level revealed that individuals above 30 years of age were significantly more compliant with the MDA (Table 3).

d, it was found that the drug uptake in the 2011 MDA campaign was 51.6%. Moreover, participation in more than four previous rounds of MDA was reported to be 33.0%. Analysis by gender, age group, and education level revealed that individuals above 30 years of age were significantly more compliant with the MDA (Table 3). In the 2015 survey, a total of 868 heads of households were interviewed, of whom 54.2% were female and 45.8% were male. Of these respondents, more than half (57.0%) reported having swallowed the drugs in the 2014 MDA campaign. Only 22.1% reported having swallowed the drugs more than four times in the previous MDA rounds. Analysis by age group revealed that individuals above 30 years of age were significantly more compliant with the MDA (Chi-square test, p = 0.000; Table 3). Being absent during drug distribution was cited as the major reason for not participating in MDA rounds, as presented in Figure 1.Figure 1 Reasons given for not taking drugs in the last round of mass drug administration (MDA); results are expressed as the percentage of the respondents who did not take drugs in the last MDA in 2015 (n = 370). Figure 1

In the 2015 survey, a total of 868 heads of households were interviewed, of whom 54.2% were female and 45.8% were male. Of these respondents, more than half (57.0%) reported having swallowed the drugs in the 2014 MDA campaign. Only 22.1% reported having swallowed the drugs more than four times in the previous MDA rounds. Analysis by age group revealed that individuals above 30 years of age were significantly more compliant with the MDA (Chi-square test, p = 0.000; Table 3). Being absent during drug distribution was cited as the major reason for not participating in MDA rounds, as presented in Figure 1.Figure 1 Reasons given for not taking drugs in the last round of mass drug administration (MDA); results are expressed as the percentage of the respondents who did not take drugs in the last MDA in 2015 (n = 370). Figure 1 Community utilization of LLINs Heads of households were interviewed on their possession and utilization of LLINs, and if allowed, the interviewer verified the availability of the LLINs. In the 2012 survey, 63.4% of the 246 heads of households interviewed possessed LLINs and 59.2% reported having slept under LLINs the previous night (Table 3). In contrast, in the 2015 survey of 868 heads of households, possession was reported to be 92.5% and sleeping under LLINs the previous night was reported to be 75.4% (Table 3).

rvey, 63.4% of the 246 heads of households interviewed possessed LLINs and 59.2% reported having slept under LLINs the previous night (Table 3). In contrast, in the 2015 survey of 868 heads of households, possession was reported to be 92.5% and sleeping under LLINs the previous night was reported to be 75.4% (Table 3). Discussion Since the launch of the GPELF in 2000, countries in endemic areas have initiated LF transmission control activities based on MDA, and progress towards the elimination goal has been reported elsewhere (Ottesen et al., 2008). Moreover, studies have indicated that ITNs lower the prevalence and transmission of LF (Odermatt et al., 2008, Ashton et al., 2011, Njenga et al., 2011). As a reduction in transmission will lead to a simultaneous decrease in acquisition of infection, it has been suggested that young children be screened to assess the effectiveness of transmission control interventions during LF elimination programmes (Witt and Ottesen, 2001). The current study monitored the effect of MDA on LF infection in school children living in the endemic district of Rufiji in southeastern Tanzania. Compliance with MDA and the utilization of ITNs were assessed in the study areas.

nsmission control interventions during LF elimination programmes (Witt and Ottesen, 2001). The current study monitored the effect of MDA on LF infection in school children living in the endemic district of Rufiji in southeastern Tanzania. Compliance with MDA and the utilization of ITNs were assessed in the study areas. Before the onset of LF control activities in Rufiji District, the baseline W. bancrofti CFA in adults was reported to range from 49% to 64% (MoHSW 2012, unpublished). In the current study, surveys conducted in 2012 and 2015, which coincided with the 9th and 12th rounds of MDA, reported W. bancrofti CFA prevalence of 14.3% and 0.0%, respectively. The findings of the 2015 survey indicate a remarkable reduction in W. bancrofti CFA prevalence in young children after 12 rounds of MDA. Based on other studies that have reported a reduction in CFA as a result of MDA (Simonsen et al., 2011, Simonsen et al., 2014), the findings suggest a decrease in acquisition of new infections by young school children as a result of MDA.

reduction in W. bancrofti CFA prevalence in young children after 12 rounds of MDA. Based on other studies that have reported a reduction in CFA as a result of MDA (Simonsen et al., 2011, Simonsen et al., 2014), the findings suggest a decrease in acquisition of new infections by young school children as a result of MDA. Treatment coverage and community compliance are important factors for successful LF elimination through the MDA strategy. It has been shown that in areas with high pre-MDA levels of infection, maintaining high drug intake during MDA is crucial in order to reach the elimination goal within a reasonable time frame (Michael et al., 2004, Michael et al., 2006). In the current study, the overall surveyed drug uptake for 2011 (surveyed in 2012) was 51.6% and for 2014 (surveyed in 2015) was 57.4%. The official reported treatment coverage for the respective years was 88.0% and 68.7%. This suggests that reported MDA coverage was relatively higher than surveyed coverage, as has also been reported elsewhere (Simonsen et al., 2013). The surveyed percentages of drug uptake in the current study were lower than the minimum of 65% required to attain the interruption of LF transmission, and this observation has been documented in other places implementing the programme (World Health Organization, 2011). Lower than optimal drug uptake is a challenge to the success of MDA campaigns in some areas of Tanzania (Kisoka et al., 2014, Simonsen et al., 2014). In the current study, being absent during drug distribution was cited as the major reason for not taking the drugs. This was also noted in studies conducted previously in other parts of Tanzania (Kisoka et al., 2014, Simonsen et al., 2014).

mpaigns in some areas of Tanzania (Kisoka et al., 2014, Simonsen et al., 2014). In the current study, being absent during drug distribution was cited as the major reason for not taking the drugs. This was also noted in studies conducted previously in other parts of Tanzania (Kisoka et al., 2014, Simonsen et al., 2014). Despite the lower than optimal treatment coverage observed in this study, the survey for CFA in school pupils in 2015 indicated a marked reduction in transmission when compared to 2012 infection levels. These contrasting findings may be due to an increase in ownership and utilization of ITNs, found to be 92.5% and 75.4%, respectively, in the 2015 survey. Of particular relevance, it has been shown that the use ITNs for malaria control has had an effect in reducing filarial rates (Odermatt et al., 2008, Ashton et al., 2011, Kelly-Hope et al., 2013). Moreover, a study conducted in the Coast Region of Kenya reported a sustained reduction in LF transmission despite missed rounds of MDA, due to the impact of the ITN intervention (Njenga et al., 2011). It is thus suggested that the lower drug uptake in this study was complemented by high ITN coverage, which sustained the CFA reductions observed. Although this study recorded a marked reduction in acquisition of new infection by young school pupils in the study areas, a formal transmission assessment survey (TAS) is recommended to establish whether the transmission of LF has been interrupted in Rufiji District.

erage, which sustained the CFA reductions observed. Although this study recorded a marked reduction in acquisition of new infection by young school pupils in the study areas, a formal transmission assessment survey (TAS) is recommended to establish whether the transmission of LF has been interrupted in Rufiji District. In conclusion, the findings of this study suggest that 12 rounds of MDA complemented with vector control through the use of ITNs resulted in a marked reduction in W. bancrofti CFA in young school children. It is recommended that a formal TAS is conducted to establish whether the transmission of LF has been interrupted in Rufiji District. Funding The study received financial support from TDR, Special Programme for Research and Training in Tropical Diseases, co-sponsored by UNICEF, UNDP, the World Bank, and the WHO (project number HQTDR1409931). The sponsor did not take part in the writing or submission of the manuscript. Conflict of interest The authors declare that they have no competing interests. Acknowledgements The authors are grateful to the parents in the study areas for allowing their children to participate in the study. We thank the pupils and their teachers in the study schools for their participation. The late Mzee Maembe, Justine Mkeni, and Shaban are thanked for their skilled assistance during the field work. The authors acknowledge the district authorities for granting permission to conduct the study in Rufiji.

1. Introduction According to the World Health Organization (WHO), Georgia has a high incidence of tuberculosis (TB) cases (116/100.000 habitants), with a high Multi-Drug Resistant TB (MDR-TB) burden (51% of all declared cases in 2015).1 In countries with high incidences, therapeutical surgery is still a good option to cope with TB complications and sequelae, as well as to reduce the bacilli burden2; and might be essential as an adjuvant to the appropriate chemotherapy in the MDR-TB cases.3 In this manuscript, we present a study that aimed to retrospectively compare clinical data and characteristics of removed TB lesions of a cohort of patients undergoing therapeutical surgery for their TB, in the National Center for Tuberculosis and Lung Diseases (NCTLD, Tblisi, Georgia). 2. Materials and methods 2.1. Design The project was reviewed and approved by the Ethics Committee IRB00007705 NCTLD Georgia #1, IORG0006411 (Tbilisi, Georgia) and the Ethics Committee of Germans Trias i Pujol Hospital (Badalona, Catalonia, Spain) to ensure compliance with all current national and European laws on clinical studies. The study was registered at the clinicaltrials.gov database with the identifier NCT02715271. The results presented here were extracted from a retrospective cohort including all patients operated per clinical routine at the NCTLD.

ain) to ensure compliance with all current national and European laws on clinical studies. The study was registered at the clinicaltrials.gov database with the identifier NCT02715271. The results presented here were extracted from a retrospective cohort including all patients operated per clinical routine at the NCTLD. 2.2. Data Data were recorded retrospectively and anonymously from the 2014 and 2015 surgical notebooks in a Spreadsheet created ad-hoc, including different categories on both the patients and the lesions’ characteristics. WHO definitions were used whenever possible.4 Table 1 includes all data categories recorded, definitions and possible answers. Graph Pad Prism 6 software (La Jolla, CA, USA) was used to draw the figures. Statistical analysis was done using the independent 2 samples t-test to compare continuous variables. The associations with other categorical variables were tested with the Chi-square test or Fisher exact test. All tests were two-tailed, and p-values less than 0.05 were considered statistically significant.

he figures. Statistical analysis was done using the independent 2 samples t-test to compare continuous variables. The associations with other categorical variables were tested with the Chi-square test or Fisher exact test. All tests were two-tailed, and p-values less than 0.05 were considered statistically significant. 3. Results 3.1. Demographic and epidemiological features A total of 137 patients underwent therapeutical surgery for their TB at the NCTLD during the years 2014 and 2015, 96 males (69.8%) and 41 females (30.8%). Men were older (median age = 43.5 years old) than women (28 years old). 25% of men had comorbidities: Hepatitis C Virus Infection (n = 15), Diabetes mellitus (n = 5), HCV and Diabetes (n = 3), and HCV plus HIV infection (n = 1). A total of 49 patients were smokers, 48 on a daily or mostly daily basis; representing a 48.9% and 2.4% of total male and female populations, respectively. The majority of women (92.6%) declared never drinking alcohol. The alcohol intake in men increased with age. While men declaring themselves abstemious (13.5%) had a median age of 26 years old, those declaring drinking on less than monthly (33.3%), monthy (52%) or weekly basis (8.3%) had a higher median age (45.6 years old). No statistically significant differences were encountered between DS-TB and MDR/XDR-TB groups.

ith age. While men declaring themselves abstemious (13.5%) had a median age of 26 years old, those declaring drinking on less than monthly (33.3%), monthy (52%) or weekly basis (8.3%) had a higher median age (45.6 years old). No statistically significant differences were encountered between DS-TB and MDR/XDR-TB groups. 3.2. TB data The 92.7% of total cases recorded were pulmonary TB exclusively (127 patients), while in 7.3% (10 patients) pleura involvement (as empyema) was found (being more frequent in males, not statistically significant). Pleura involvement was present in the 80% of MDR/XDR-TB cases. Approximately the half of the patients included were DS-TB (56.9%), and the other half, MDR or XDR (40.1%). Mono resistance was rare. Table 2 shows the number of cases for each category, as well as the number and percentage according to gender. The percentage of men and women were very similar regardless of the drug-sensitivity of the strains involved. The chest X-ray assay (previous to surgery) showed the presence of a single lesion (only cavities and tuberculomas considered) for most of the patients (76%), while 19.2% showed 2 lesions, and 4.8% up to three. Most patients undergoing surgery had been considered bacteriology cured according to WHO’s definitions (96.3%). No statistically significant differences according to gender or drug susceptibility were found in the non microbiologically cured group of patients.

The chest X-ray assay (previous to surgery) showed the presence of a single lesion (only cavities and tuberculomas considered) for most of the patients (76%), while 19.2% showed 2 lesions, and 4.8% up to three. Most patients undergoing surgery had been considered bacteriology cured according to WHO’s definitions (96.3%). No statistically significant differences according to gender or drug susceptibility were found in the non microbiologically cured group of patients. Figure 1 shows the results of time to culture’s negativization, in totals and stratified by gender and by the drug-sensitivity of the strains. As women included in this cohort were younger (28 vs 43.5 years old), we stratified the results by dividing the male cases in ≤43 and >43 years old, to see if the gender itself could be an influence free from the age input. Differences were encountered between females and males of ≤43 years old, being statistically significant for DS-TB cases (p ≤ 0.0001). A high percentage of patients negativized the culture in ≤2 months (51.8%), especially if they were women or younger (median age of 33). The culture was always negative in a non despricable 31.3% of the cases; and up to 40.6% of male cases when stratified by gender. Patients with culture negative at ≤2 months were younger than those in whom it had always been negative (median age of 33 vs 43 years old, respectively; p < 0.0001). When stratified by gender, this difference was only seen in males (median age: 37 vs 47 years old; p = 0.022); but the overall percentage of negative cultures in females was significantly lower. Statistically significant differences in terms of negativization of culture were encountered between DS-TB and MDR/XDR-TB, both in total (p ≤ 0.0001) and when stratifying the results by age and gender (males ≤43, p = 0.0009; males >43, p = 0.0007; and females, p = 0.0169).