Summary

Introduction

Diarrhoea remains the second leading cause of death in children younger than 5 years worldwide, being associated with around 500 000 deaths per year, and ranks sixth in global disability-adjusted life-year burden.^1–3 Public health interventions rely on estimates of pathogen-specific burden for prioritisation. Previous estimates of the infectious causes of diarrhoea have been derived from studies that used varying approaches for pathogen detection,⁴ including different media, stool concentration techniques, and detection methods, from microscopy to antigen detection to PCR. Each method has its own sensitivity, which could substantially affect burden estimates. Furthermore, asymptomatic carriage of some enteropathogens is common in children living in developing countries, for which few studies have attempted to control.^5–7 The quantity of nucleic acid is thought to distinguish clinical disease from asymptomatic shedding.^8–14

We have shown previously that quantitative real-time PCR (qPCR) approaches could detect bacterial, viral, and parasitic pathogens across multiple laboratories with high sensitivity and good reproducibility and quantification.¹⁵ We therefore did this reanalysis of specimens from the Global Enteric Multicenter Study (GEMS), which was a large case-control study of moderate to severe diarrhoea in children younger than 5 years in Africa and Asia,⁶ to inform estimates of the global burden of disease.³ We aimed to reassess estimates of pathogen-specific diarrhoea causes and incidence at the population level, and to provide a case-level understanding of mixed diarrhoeal infections.

Methods

Results

We tested 11 400 specimens from 5700 cases and matched controls in five laboratories that showed good assay reproducibility (table 1). Identical positive-control constructs were provided to all laboratories, with consistent results for assay linearity (appendix). Valid results were obtained from 5304 (93·1%) matched case-control pairs (table 1). When compared with qPCR, the original diagnostic methods showed low and variable sensitivity but high specificity (98% [SD 2%], appendix).¹⁵ Samples positive by the original diagnostic methods had higher quantities with molecular testing methods than did originally negative samples, across all 16 pathogens detected in the original study.

Table 1

Paired case and control samples tested by quantitative real-time PCR, by age stratum and study site

	Basse, The Gambia*	Bamako, Mali†	Manhiça, Mozambique†	Nyanza Province, Kenya*	Kolkata, India†	Mirzapur, Bangladesh*	Karachi (Bin Qasim Town), Pakistan*	Total
Age stratum‡
0–11 months	300 (400)	300 (727)	267 (374)	303 (673)	300 (672)	300 (550)	300 (633)	2070
12–23 months	301 (455)	300 (682)	159 (195)	300 (410)	300 (588)	300 (476)	300 (399)	1960
24–59 months	165 (174)	283 (624)	95 (112)	301 (393)	300 (308)	300 (368)	226 (226)	1670
Valid pairs analysed§	685 (894%)	834 (94·5%)	484 (92·9%)	787 (87·1%)	849 (94·3%)	877 (974%)	788 (954%)	5304 (93·1%)
Reproducibility (%)¶
MS2	6·0%	5·2%	6·0%	6·8%	5·9%	6·2%	5·8%	67%
PhHV	5·2%	4·9%	4·1%	5·5%	5·0%	54%	6·1%	5·5%

Open in a separate window

For each combination of age and site, we aimed to test 300 cases and first available matched controls; for sites with fewer than 300 cases enrolled, all available cases and first available matched controls were tested. PhHV=phocine herpes virus.

^*Extracted and tested on site

^†Extracted at University of Maryland, MD, USA, and tested at University of Virginia, VA, USA.

^‡Numbers in parentheses are the number of cases in the original Global Enteric Multicenter Study.

^§Percentage of case-control pairs that were included in the analysis and did not exhibit extraction blank failures (139 pairs excluded, ranging from 0 to 7·0% by site), amplification failures (220 pairs excluded, ranging from 2·0% to 5·3% by site), or quantitative real-time PCR software flags (77 pairs excluded, ranging from 0·2% to 2·7% by site).

^¶Measured by the coefficient variation of MS2 or PhHV Cqs on the clinical specimens.

Shigella spp or EIEC, adenovirus 40/41, rotavirus, Cryptosporidium spp, and heat-stable enterotoxin-producing E coli (ST-ETEC) had high prevalence and strong quantity-dependent associations with diarrhoea (figure 1). Campylobacter jejuni or C coli, sapovirus, and typical enteropathogenic E coli (tEPEC) were also highly prevalent, but even at the highest quantities were only moderately associated with diarrhoea. Norovirus GII, Helicobacter pylori, astrovirus, Vibrio cholerae, Salmonella spp, Cyclospora cayetanensis, and Entamoeba histolytica also showed associations with diarrhoea but prevalence was low (figure 1). Entero aggregative E coli and Aeromonas spp were associated with diarrhoea only in specific study sites or age strata (enteroaggregative E coli was associated with diarrhoea only in the 12–13 months age group in Mozambique; Aeromonas spp was associated with diarrhoea only in the 0–11 months age group in Pakistan and in the 24–59 months age group in Pakistan, Mozambique, and Bangladesh). Cystoisospora belli was associated only with dysentery. All other enteropathogens tested were not associated with diarrhoea at any quantity or site or in any age group (appendix). The quantitative relations were robust across age groups (appendix).

Open in a separate window

Figure 1

Relation between pathogen quantity and diarrhoea

Pathogens are ordered from top to bottom and left to right by prevalence in cases. Data are numbers of dectections (vertical bars) in cases (dark grey) and controls (light grey) with odds ratios (red lines) and 95% CIs (bands). The model-derived cutoffs used for identification of diarrhoea-associated pathogens in individuals (overlaid in blue) are defined as all detections above the point at which the 95% CI no longer includes 1. EAEC=enteroaggregative E coli. EIEC=enteroinvasive E coli. EPEC=enteropathogenic E coli. LT-ETEC=heat-labile enterotoxin-producing E coli. ST-ETEC=STh-producing enterotoxigenic E coli. STEC=Shiga toxin producing E coli.

Compared with the original microbiological methods, more pathogens were attributed as causes of diarrhoea, with more consistent attribution across sites. When assessed by age strata and study site, the pathogens with the highest attributable fractions for childhood diarrhoea were Shigella spp or EIEC, rotavirus, adenovirus 40/41, ST-ETEC, and Cryptosporidium spp (table 2). Shigella spp or EIEC, rotavirus, and ST-ETEC were consistently high across all age groups and regions. Cryptosporidium spp and adenovirus 40/41 were most evident in children up to age 23 months and the prevalence of H pylori was notable in children aged 24 months and older. C jejuni or C coli were particularly important causes in infants at the Bangladesh, India, Mali, and Mozambique sites. Estimates were robust irrespective of whether raw Cqs or those adjusted to standard curves and external controls to derive a pathogen target copy number were used (appendix).

For most pathogens, the qPCR-derived attributable incidence surpassed the original estimate, including for adenovirus 40/41 (by around five times), Shigella spp or EIEC and C jejuni or C coli (around two times), and ST-ETEC (around 1·5 times; figure 2). With few exceptions, the increase in attribution for these pathogens was consistent across study sites (appendix). For rotavirus and Cryptosporidium spp, the attributable incidence remained similar to the original estimates (obtained with EIA). In infants, six pathogens had the greatest attributable incidence: in descending order, rotavirus (7·9 episodes per 100 child-years, 95% CI 6·2–9·7), adenovirus 40/41 (3·9, 2·4–5·4), Cryptosporidium spp (3·3, 2·2–4·4), C jejuni or C coli (2·4, 1·3–3·4), Shigella spp or EIEC (2·0, 1·4–2·6), and ST-ETEC (2·0, 1·4–2·6). In toddlers, the hierarchy was Shigella or EIEC (7·0, 5·0–9·0), rotavirus (4·1, 3·0–5·1), adenovirus 40/41 (1·8, 0·9–2·6), ST-ETEC (1·7, 1·1–2·3), and Cryptosporidium spp (1·5, 0·8–2·2). In older children, Shigella or EIEC was the dominant pathogen (2·3, 1·2–3·4). By qPCR, 89·3% (95% CI 83·2–96·0) of diarrhoeal episodes at the population level were attributable to a pathogen, as compared with 51·5% (48·0–55·0) by the original workup. Shigella spp, rotavirus, adenovirus 40/41, heat-stable enterotoxin-producing E coli, Cryptosporidium spp, and Campylobacter spp together accounted for 77·8% (74·6–80·9) of all attributable diarrhoea.

Open in a separate window

Figure 2

Attributable incidence of pathogen-specific moderate to severe diarrhoea per 100 child-years, by age stratum, across study sites, in this and the original GEMS

For each age stratum, any pathogens significantly associated with diarrhoea by one or both diagnostic approaches are shown. GEMS=Global Enteric Multicenter Study. EAEC=enteroaggregative E coli. EIEC=enteroinvasive E coli. tEPEC=typical enteropathogenic E coli. ST-ETEC=STh-producing enterotoxigenic E coli. qPCR=quantitative realtime PCR. *Indicates the microbiology in the original GEMS did not test for H pylori or C cayetanensis.

To further understand the increase in burden attributable to Shigella spp or EIEC, we examined the qPCR and culture results by site. Attribution was higher at all study sites with qPCR than with the previous diagnostic methods, but to a lesser degree in Bangladesh than in the others, which was due mainly to high quantities of Shigella spp and high sensitivity of the culture at that site (appendix).

Among analysed diarrhoeal episodes, 4077 (76·9%) of 5304 were described as being watery, and gross blood was reported as being seen in 1227 (23·1%). As expected, Shigella spp or EIEC was the dominant pathogen associated with dysentery (attributable fraction 63·8%, 95% CI 61·6–68·0), but was also associated with the second-highest burden of watery diarrhoea (12·9%, 11·0–16·4, figure 3), and, overall, 527·7 (40·3%) of 1310·3 attributable cases of diarrhoea attributable to Shigella spp or EIEC were non-dysenteric. Generally, quantities of Shigella spp or EIEC in dysentery cases were higher than those in cases of watery diarrhoea (mean Cq 22·0 [4·6] vs 26·2 [5·8], p<0·0001; appendix). Confirmation of Shigella spp or EIEC detection in 450 case-control pairs showed similar attributable fractions with different PCR targets (appendix). Additionally, qPCR for gene regions specific to S flexneri and S sonnei showed these species account for more than 70% of all Shigella spp or EIEC attributable fractions. Adenovirus 40/41 and C jejuni or C coli were also associated with a substantial burden of dysentery (attributable fractions 10·4%, 95% CI 4·5–16·7 and 7·8%, 5·5–12·9, respectively; these estimates control for the presence of other pathogens, such as Shigella spp).

Open in a separate window

Figure 3

Relative attribution of watery diarrhoea and dysentery to each pathogen

Data are overall adjusted attributable fractions (vertical bars) with 95% CIs. Pathogens are ordered by the proportion of total attributable cases that were watery diarrhoea (dotted line). All pathogens significantly associated with either dysentery or watery diarrhoea are shown. ST-ETEC=STh-producing enterotoxigenic E coli. tEPEC=typical enteropathogenic E coli. EIEC=enteroinvasive E coli.

The original GEMS microbiology interrogated for ST-ETEC producing the STh subtype (STh-ETEC) and heat-labile enterotoxin-producing E coli (LT-ETEC), but not ST-ETEC producing the STp subtype, which might have led to underestimation of the burden for ST-ETEC. qPCR showed that most of the burden was attributable to the STh subtype and was mainly evident at specific study sites (appendix). Inclusion of STp subtype in the definition of ST-ETEC increased the overall burden estimates for this pathogen by 15%. There was essentially no burden attributable to LT-ETEC, and the presence of heat-labile enterotoxin did not modify the strength of the association between ST-ETEC producing the STh subtype and case or control status (with heat-labile enterotoxin, OR 0·92, 95% CI 0·74–1·13). Therefore, to be consistent with the original study, we categorised ST-ETEC producing the STh subtype with or without heat-labile enterotoxin. In 249 (43·9%) of 505 samples in which ST-ETEC producing STh subtype was detected in diarrhoea-associated quantities, heat-labile enterotoxin was detected in similarly high quantities, which suggests that these infections might be due to both toxins and the remainder were due to ST-ETEC producing STh subtype only.

To examine the causes of diarrhoea in individual children, we identified model-derived quantitative cutoffs for diarrhoea association (figure 1, appendix). Diarrhoea-associated quantities varied greatly by pathogen, ranging from 7·3 × 10³ copies per g of stool for V cholerae to 2·7 × 10⁹ copies per g for tEPEC. For the subset of 15 pathogens for which cutoff values could be identified, at least one was detected at a diarrhoea-associated quantity in 4317 (81·4%) of 5304 cases (one in 2254 [42·5%] and more than one in 2063 [38·9%]; figure 4). Combining all study sites and age strata, Shigella spp or EIEC was the most frequent cause, followed by adenovirus 40/41, rotavirus, Cryptosporidium spp, and ST-ETEC (figure 4). In 1517 (79·6%) of 1905 cases in which Shigella spp or EIEC and 984 (84·2%) of 1168 in which rotavirus was present in diarrhoea-associated quantities, these were the primary pathogens (ie, the only diarrhoea-associated pathogen or at a quantity that delivered the highest diarrhoea association). The predominance of Cryptosporidium spp and ST-ETEC was slightly less (313 [58·0%] of 540 and 332 [65·6%] of 506 cases, respectively). The opposite was seen for adenovirus 40/41, which in most cases (1070 [71·4%] of 1499), even at diarrhoea-associated quantities, occurred with pathogens more strongly associated with diarrhoea (appendix). In general, we saw no preferential co-occurrence between particular diarrhoea-associated pathogens, although Shigella spp or EIEC was frequently identified as a diarrhoea-associated co-infection when H pylori was detected as the primary diarrhoea-associated pathogen (appendix). In many samples, quantities of C jejuni or C coli, tEPEC, H pylori, Cryptosporidium spp, ST-ETEC, and norovirus GII were low and the pathogens were not associated with diarrhoea.

Open in a separate window

Figure 4

Detection of co-infections in diarrhoeal cases

(A) Numbers of pathogens at diarrhoea-associated quantities and any quantity in individual cases of diarrhoea. (B) Distribution of pathogens, alone and in co-infections, by quantity and association with diarrhoea. The quantification cycle cutoff used to identify diarrhoea-associated detections is shown in parentheses after each pathogen name. Cq=quantification cycle. EIEC=enteroinvasive E coli. ST-ETEC=STh-producing enterotoxigenic Escherichia coli. tEPEC=typical enterpathogenic E coli.

Discussion

In this reanalysis of the causes of diarrhoea in GEMS, we have shown with a quantitative molecular diagnostic approach that the estimated burdens of Shigella spp, adenovirus 40/41, ST-ETEC, and C jejuni or C coli were underestimated with the original diagnostic methods. By contrast, estimates for rotavirus and Cryptosporidium spp, for which detection was based on EIA diagnostics, changed little. These six pathogen groups were responsible for 77·8% of attributable diarrhoea in this study and, therefore, targeted interventions could have larger public health benefits than previously projected.⁴

The increases in estimated burden were a function of improved sensitivity with molecular diagnostics and the higher resolution provided by pathogen quantification. Previous studies of the causes of diarrhoea have generally used non-quantitative diagnostics, which yield dichotomous results at detection limits that might not be clinically relevant. Such results become potentially problematic for the study of causes of diarrhoea in children in resource-limited settings because the rate of enteropathogen carriage shortly after birth is high.^7,21

We identified a high burden of diarrhoea attributable to Shigella spp or EIEC, which we suspect is due to the fastidious nature of this pathogen in culture, particularly when quantities are low and in areas where antibiotics are frequently used. Underdetection of Shigella spp by culture compared with PCR has been seen in various settings worldwide.^8,22–25 The ipaH target has been widely used for detection of Shigella spp, and its presence has only been described in Shigella spp and EIEC, both of which genera are highly similar at the DNA level.²⁶ Nevertheless, EIEC has not been a major diarrhoea-associated pathogen in similar settings.⁷ The detection of additional gene regions clarified that Shigella spp explained most of the burden estimate. Shigella spp were particularly associated with dysentery, which was a case definition inclusion criterion in GEMS and was common at the Bangladesh site, but were also notably associated with watery diarrhoea across all study sites, which raises questions about the treatment guidelines for infection with Shigella spp, which emphasise the presence of blood in stools as an indication.²⁷ The higher quantities of Shigella spp in samples from children who had dysentery than in those who had watery diarrhoea without a diagnosis of dysentery supports previous findings²⁶ and suggests that watery diarrhoea might indicate a less advanced stage of this disease. Development of a vaccine against Shigella spp is underway, and these findings and the high detection rate in samples in this study suggest that it should be prioritised. Vaccination of children early in life could be beneficial.

The importance of rotavirus and Cryptosporidium spp was reaffirmed, with burden estimates being similar to those in the original study, whereas the importance of ST-ETEC and Campylobacter spp was increased. The association of adenovirus 40/41 with diarrhoea has been noted previously,²⁸ including by use of EIA in the original GEMS study,⁶ but molecular diagnostics revealed a substantially higher prevalence and retained diarrhoea association. Our array card for qPCR did detect panadenovirus of any serotype, but this was not associated with diarrhoea after controlling for the presence of adenovirus 40/41. The association of H pylori with diarrhoea was unexpected. This pathogen has been reported in children with diarrhoea, most often those with chronic diarrhoea or hypochlorhydria,^29,30 but other studies have identified no risk or suggested protection.^31,32 We found frequent co-infection with H pylori and Shigella spp, which has been reported previously³³ and suggests that infection with H pylori could be a risk factor for infection with Shigella spp. Further study is needed to understand the role of H pylori in diarrhoea.

By identifying model-derived quantitative cutoff values for detection of diarrhoea-associated pathogens, we were able to describe the causes of diarrhoea in individual children, which is crucial for clinical diagnosis or case definitions. Quantitative cutoffs have previously been described for tEPEC, norovirus GII, and rotavirus,^9–14 mostly in comparisons of quantities in cases versus controls, but those studies involved fewer samples and did not adjust for other pathogens or provide the resolution afforded by the present analytical methods. We were able to identify cutoff values for 15 pathogens. This approach worked well for most pathogens but less so for others. For example, Campylobacter spp had weak but significant associations with diarrhoea that yielded poorly discriminatory quantitative cutoff values despite a substantial burden in the overall population due to the high prevalence of this group of pathogens. The cutoffs associated with diarrhoea in this study varied by pathogen. The ipaH Cq for detection of Shigella spp that maximally separated cases and controls was previously estimated from a subset of GEMS sites of approximately 10⁷ copies per g of stool.⁸ We found a slightly lower value of 2 × 10⁶ copies per g. The quantitative relations we identified were reasonably generalisable across age groups for the pathogens with the greatest attributions and, therefore, these cutoff values should be useful for future studies when control stools are not available.

More than 80% of samples revealed pathogens at diarrhoea-associated quantities, which is a much greater diagnostic yield than has been found with routine microbiological methods.^4,34,35 Among those containing a diarrhoea-associated pathogen, roughly half had more than one. This extent of mixed infections, even after filtering out probable asymptomatic infections, was substantial. Some pathogens, such as Shigella spp and rotavirus, were most frequently detected as sole pathogens in diarrhoea-associated quantities, meaning that misattribution of diarrhoea to these pathogens would be uncommon. Many other pathogens, such as Cryptosporidium spp, ST-ETEC, norovirus GII, adenovirus 40/41, and Campylobacter spp, were not detected in diarrhoea-associated quantities or present in quantities associated with diarrhoea but with other pathogens that showed stronger associations (figure 4). This finding underscores the difficulty in attributing diarrhoea to such pathogens without quantification or considering other pathogens. It will also challenge vaccine trials, because efficacy would be underestimated if a non-quantitative PCR result were used that might over-attribute diarrhoea to the pathogen of interest. For such trials, more stringent quantitative cutoffs could be applied, for example using the quantity of the highly diarrhoea-associated ST-ETEC to improve the specificity of the case definition for ETEC-associated diarrhoea (appendix). Other criteria, such as clinical severity scores or inflammatory markers, could be assessed to see whether they increase the specificity of detection.

Achieving optimum clinical management in the era of molecular diagnostic panels for enteropathogens is a work in progress. Commonly available commercial panels are qualitative PCR-based tests that do not distinguish between low-level enteropathogen detection of unclear relevance and infections that are more clearly clinically important.³⁶ For example, detection of the Clostridium difficile toxin gene by PCR in the setting of a negative toxin EIA is not clinically relevant.³⁷ We believe that the quantitative cutoff concept we have assessed will be helpful, although it places additional requirements on commercial assay developers, and the exact cutoffs used might depend on the clinical setting. Management guidelines for acute diarrhoea recommend supportive care with rehydration and zinc supplementation without antibiotics unless signs are indicative of dysentery or cholera is suspected,²⁷ but could be adapted as pathogen-specific diagnosis improves and hopefully becomes more widely available. For a high-mortality entity, such as childhood diarrhoea, an overly stringent cutoff could lead to underdiagnosis of treatable causes. Thus further assessment of quantitative cutoff values for clinical care is needed, along with studies that assess therapeutic response, for instance whether treatment of diarrhoea associated with Shigella spp in children in low-resource settings that is negative to culture but positive on PCR improves outcomes. The necessity of treating multiple pathogens during mixed infection is also unknown. Finally, the existing quantitative PCR assays for enteropathogens do not assess antimicrobial resistance and, therefore, for now these emerging molecular technologies should ideally be used in conjunction with conventional culture methods.

This study had some limitations. The quantitative approach, while incrementally useful, will inherently function less well for pathogens that are shed with high frequency, in high quantities, and for an extended duration in the absence of diarrhoea. Longitudinal studies might further refine understanding for such pathogens. Theoretically, the sum of the attributable fractions for individual pathogens can be greater than 100%, although this possibility is the same for estimates derived from molecular and non-molecular diagnostic approaches, and is implicit in any clinical entity with co-infections. Although GEMS was carefully designed to study the causes of diarrhoea and the study sites were broadly representative of countries with moderate to high mortality in children younger than 5 years, some heterogeneity in the hierarchy of pathogens was noted between sites. Generalisability, therefore, should be audited by local epidemiology. Finally, vaccine development often relies on speciation and subtyping of infections,³⁸ and our molecular assays did not provide such information.

This reanalysis of the causes of diarrhoea in young children by use of molecular diagnostic methods substantially increased attribution at the population level and clinical diagnostic yields. We found a high burden of diarrhoea associated with six main pathogens: Shigella spp, rotavirus, adenovirus 40/41, ST-ETEC, Cryptosporidium spp, and Campylobacter spp. Mixed infection with diarrhoea-associated pathogens was common and has implications for vaccine efficacy trials, responses to treatments, and the clinical diagnosis of pathogen-specific diarrhoea.

Supplementary Material

Acknowledgments

Footnotes

Contributor Information

Jie Liu, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

James A Platts-Mills, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Jane Juma, Center for Global Health Research, Kenya Medical Research Institute, Nairobi, Kenya.

Furqan Kabir, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Joseph Nkeze, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Catherine Okoi, Medical Research Council Unit, Banjul, The Gambia.

Darwin J Operario, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Jashim Uddin, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR, B), Dhaka, Bangladesh.

Shahnawaz Ahmed, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR, B), Dhaka, Bangladesh.

Prof Pedro L Alonso, Centro de Investigação em Saúde da Manhiça, Maputo, Mozambique; Barcelona Centre for International Health Research (CRESIB, Hospital Clinic-Universitat de Barcelona), Barcelona, Spain.

Prof Martin Antonio, Medical Research Council Unit, Banjul, The Gambia.

Stephen M Becker, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Prof William C Blackwelder, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Prof Robert F Breiman, Global Disease Detection Division, Kenya Office of the US Centers for Disease Control and Prevention, Nairobi, Kenya.

Abu S G Faruque, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR, B), Dhaka, Bangladesh.

Barry Fields, Global Disease Detection Division, Kenya Office of the US Centers for Disease Control and Prevention, Nairobi, Kenya.

Jean Gratz, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Rashidul Haque, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR, B), Dhaka, Bangladesh.

Anowar Hossain, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR, B), Dhaka, Bangladesh.

M Jahangir Hossain, Medical Research Council Unit, Banjul, The Gambia.

Sheikh Jarju, Medical Research Council Unit, Banjul, The Gambia.

Farah Qamar, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Najeeha Talat Iqbal, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Brenda Kwambana, Medical Research Council Unit, Banjul, The Gambia.

Inacio Mandomando, Centro de Investigação em Saúde da Manhiça, Maputo, Mozambique.

Timothy L McMurry, Public Health Sciences, University of Virginia, Charlottesville, VA, USA.

Caroline Ochieng, Center for Global Health Research, Kenya Medical Research Institute, Nairobi, Kenya.

John B Ochieng, Center for Global Health Research, Kenya Medical Research Institute, Nairobi, Kenya.

Melvin Ochieng, Center for Global Health Research, Kenya Medical Research Institute, Nairobi, Kenya.

Clayton Onyango, Global Disease Detection Division, Kenya Office of the US Centers for Disease Control and Prevention, Nairobi, Kenya.

Sandra Panchalingam, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA.

Adil Kalam, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Fatima Aziz, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Shahida Qureshi, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Thandavarayan Ramamurthy, National Institute of Cholera and Enteric Diseases, Kolkata, India.

James H Roberts, Public Health Sciences, University of Virginia, Charlottesville, VA, USA.

Debasish Saha, Medical Research Council Unit, Banjul, The Gambia.

Prof Samba O Sow, Centre pour le Développement des Vaccins, Bamako, Mali.

Suzanne E Stroup, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Dipika Sur, National Institute of Cholera and Enteric Diseases, Kolkata, India.

Boubou Tamboura, Centre pour le Développement des Vaccins, Bamako, Mali.

Mami Taniuchi, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

Sharon M Tennant, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Deanna Toema, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Yukun Wu, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Prof Anita Zaidi, Department of Paediatrics and Child Health, Aga Khan University, Karachi, Pakistan.

Prof James P Nataro, Department of Pediatrics, University of Virginia, Charlottesville, VA, USA.

Prof Karen L Kotloff, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Pediatrics, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Prof Myron M Levine, Center for Vaccine Development and Institute of Global Health, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Pediatrics, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA.

Prof Eric R Houpt, Division of Infectious Diseases and International Health, University of Virginia, Charlottesville, VA, USA.

References