If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Cholera, a waterborne disease, is known to be transmitted from person to person.
•
However, the geographical dimensions of this transmission are uncertain.
•
Oral cholera vaccines were used as probes to define these dimensions.
•
Dimensions were 500 m in rural Bangladesh and 150 m in Kolkata, India.
•
Person-to-person transmission of cholera can be sustained over long distances.
Abstract
Background
Cholera is known to be transmitted from person to person, and inactivated oral cholera vaccines (OCVs) have been shown to confer herd protection via interruption of this transmission. However, the geographic dimensions of chains of person-to-person transmission of cholera are uncertain. The ability of OCVs to confer herd protection was used to define these dimensions in two cholera-endemic settings, one in rural Bangladesh and the other in urban India.
Methods
Two large randomized, placebo-controlled trials of inactivated OCVs, one in rural Matlab, Bangladesh and the other in urban Kolkata, India, were reanalyzed. Vaccine herd protection was evaluated by relating the risk of cholera in placebo recipients to vaccine coverage of surrounding residents residing within concentric rings. In Matlab, concentric rings in 100-m increments up to 700 m were evaluated; in Kolkata, 50-m increments up to 350 m were evaluated.
Results
One hundred and eight cholera cases among 24 667 placebo recipients were detected during 1 year of post-vaccination follow-up at Matlab; 128 cholera cases among 34 968 placebo recipients were detected during 3 years of follow-up in Kolkata. Consistent inverse relationships were observed between vaccine coverage of the ring and the risk of cholera in the central placebo recipient for rings with radii up to 500 m in Matlab and up to 150 m in Kolkata.
Conclusions
These results suggest that the dimensions of chains of person-to-person transmission in endemic settings can be quite large and may differ substantially from setting to setting. Using OCVs as ‘probes’ to define these dimensions can inform geographical targeting strategies for the deployment of these vaccines in endemic settings.
Killed oral cholera vaccines (OCVs) are now stockpiled by the World Health Organization (WHO) and are recommended public health tools for the control of cholera outbreaks (
). It is generally accepted that the rational use of such vaccines, either for epidemic or endemic cholera, will require that vaccination be targeted to geographically circumscribed populations at greatest risk of cholera. Knowledge of the geographic dimensions of chains of person-to-person cholera transmission will be important for effective geographic targeting, as OCVs have been demonstrated to confer both direct protection to vaccinees and herd protection to populations, the latter operating via interruption of person-to-person transmission.
The geographic dimensions of chains of person-to-person cholera transmission were estimated in this study. A geographic information system (GIS)-based method that has been employed previously to evaluate whether OCVs confer vaccine herd protection was used (
). It was reasoned that because vaccine herd protection results from interruption of person-to-person cholera transmission, delineation of the dimensions of OCV herd protection should demarcate the dimensions of chains of person-to-person transmission.
Methods
Overview
The vaccine herd protective effects of inactivated OCVs were analyzed in two randomized, placebo-controlled trials, one in Matlab, Bangladesh (
Efficacy and safety of a modified killed whole-cell oral cholera vaccine in India: an interim analysis of a cluster-randomized, double-blind, placebo-controlled trial.
). In this approach, a ‘virtual cluster’ is defined as persons whose residences are within a specified radius of the residence of each person under analysis (termed the ‘focal person’), and the risk of cholera for each focal person under analysis is related to the vaccine coverage in surrounding virtual clusters. An inverse relationship suggests vaccine herd protection, and when the focal person under analysis has received placebo, indirect vaccine protection is measured (
). Conceptually, the maximum radius in which indirect vaccine protection is demonstrated should demarcate the geographical size of a surrounding population that puts an unimmunized individual at risk of becoming infected via person-to-person transmission. In this study, the indirect OCV protection of each focal person (placebo recipient) by vaccination in successive rings of persons in surrounding residences was analyzed, thus identifying the dimensions of chains of person-to-person cholera transmission in the two study settings.
OCV trials under analysis
The two randomized, placebo-controlled trials analyzed in this study were conducted in rural Matlab, Bangladesh, an area bisected by the Dhonagoda River, and in urban Kolkata, India, comprising wards 29, 30, and 33 (Figure 1); these trials have been described in detail elsewhere (
Efficacy and safety of a modified killed whole-cell oral cholera vaccine in India: an interim analysis of a cluster-randomized, double-blind, placebo-controlled trial.
). Dosing with inactivated OCV or placebo was conducted in 1985 in Matlab and in 2006 in Kolkata. In Matlab, children aged 2–14 years and non-pregnant female adults (≥15 years) were eligible to participate in the trial. In Kolkata, non-pregnant persons aged ≥1 year were eligible. In Matlab, eligible persons were individually randomized to a three-dose regimen of an oral cholera toxin B subunit-killed whole cell (BS-WC) vaccine, oral killed whole cell (WC)-only vaccine, or oral placebo. In Kolkata, eligible persons were randomized by residential dwelling to a two-dose regimen of an oral killed WC-only vaccine or oral placebo.
In Matlab, surveillance was conducted for all diarrheal patients from the study area who attended either the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) Matlab hospital or two community-operated treatment centers. Diarrhea was defined as the presence of at least three loose or liquid motions in the 24 h before presentation, or one to two or an indeterminate number of loose or liquid stools in the 24 h before presentation with at least two signs of dehydration (poor skin turgor, sunken eyes, dry mucous membranes, weakened radial pulse) on presentation. Stools or rectal swabs were collected from these patients and were tested for Vibrio cholerae O1 using conventional microbiological methods (
In Kolkata, diarrhea surveillance was conducted in nine project health clinics and two governmental hospitals. Diarrhea was defined as having three or more loose stools in the 24 h before presentation, or one to two or an indeterminate number of loose or liquid stools in the 24 h before presentation together with moderate or severe dehydration, according to WHO criteria, on presentation (
). Rectal swabs were collected from all diarrheal patients and tested in the project laboratory for V. cholerae O1 and V. cholerae O139 using conventional methods (
In both studies, cholera was defined as non-bloody diarrhea in which V. cholerae O1 was isolated. V. cholerae O139 was not isolated in the Kolkata trial. In the present analyses, cholera cases occurring during 1 year after the time of vaccination in Matlab and during 3 years after the time of vaccination in Kolkata were considered; these were both intervals in which indirect herd OCV protection was demonstrated in the earlier analyses (
), a statistical criterion was used to define the radius around each placebo recipient to demarcate the virtual cluster of people for determination of vaccine coverage. This criterion was related to the variability of the variances of vaccine coverage across virtual clusters, and was not based on whether vaccine coverage was inversely related to the risk of cholera. This criterion yielded a radius of 500 m for Matlab and 250 m for Kolkata.
In the present analyses, virtual clusters were defined a priori as concentric rings of residential populations in 100-m increments up to 700 m (seven rings centering each placebo recipient) in Matlab, and in 50-m increments up to 350 m (also seven rings centering each placebo recipient) in Kolkata. The difference in the size of increments was related to the dispersed population in Matlab versus the congested urban setting in Kolkata. Vaccine coverage in each ring was computed as the number of recipients of at least two doses of vaccine divided by the number of all eligible residents at the time of the first dose.
Statistical analysis
The earlier analyses found no design effect due to the residence-based allocation in Kolkata; therefore this trial was analyzed as if it had been individually randomized (
Efficacy and safety of a modified killed whole-cell oral cholera vaccine in India: an interim analysis of a cluster-randomized, double-blind, placebo-controlled trial.
). Indirect vaccine protection was assessed by relating the risk of cholera in recipients of at least two doses of placebo (focal persons) to vaccine coverage of the defined rings around these recipients. In crude analyses the relationship for each ring between the quintile of vaccine coverage around each focal person (quintiles were ascertained from the distribution of coverages for all rings of the same radius) and the risk of cholera in the focal person, was evaluated, and this relationship was assessed with the Cochran–Armitage test for trend. In multivariable logistic regression models, the relationship between the levels of vaccine coverage of the defined ring (expressed dimensionally and fitted as an independent variable) and the occurrence of cholera for each focal person (the dependent variable) was assessed, controlling for potentially confounding variables. Based on the findings of the earlier work, candidate variables for inclusion as covariates in the models for Matlab were age at the date of first dose, distance to the icddr,b hospital in Matlab, and distance to the Dhonagoda River. Age at date of first dose, sex, owning at least one luxury item, per-capita household expenditure, washing hands with soap and water after defecation, and living in an owned house were considered as candidate covariates in the Kolkata models. To select the covariates for the final models, those candidate variables found to be associated with the risk of cholera at p< 0.10 in the bivariate models were included, and these were retained in the final model if they remained significant at p< 0.20 in a backward elimination process. Multivariable odds ratios (OR) relating each percentage increase in vaccine coverage to the occurrence of cholera were estimated by exponentiation of the coefficient for the vaccine coverage variable in the models; p-values and 95% confidence intervals (CI) for these OR were estimated using the standard errors of these coefficients. All p-values and 95% CI were two-sided.
Ethics
The project in Matlab was approved by the Ethical Review Committees of the icddr,b, Dhaka, Bangladesh and WHO (Geneva). The project in Kolkata was approved by the ethics committees of the National Institute of Cholera and Enteric Diseases (NICED) and the Health Ministry Screening Committee of India, and the International Vaccine Institute Institutional Review Board. Informed consent was obtained from all participants in these studies.
Results
There were 124 035 individuals eligible for vaccination in Matlab and 107 774 individuals eligible for vaccination in Kolkata. Among these populations, 108 cholera cases occurred among 24 667 recipients of at least two doses of placebo during the 1 year of follow-up in Matlab; 128 cholera cases occurred among 34 968 two-dose placebo recipients during the 3 years of follow-up in Kolkata.
The previously published reports that evaluated vaccine coverage from 0 m to 500 m around placebo recipients in Matlab (
Hazard ratio for the relative reduction in the risk of cholera for each percentage increase in vaccine coverage. Adjusted for age, sex, religion, distance from household to nearest river, distance from household to nearest treatment center (kilometers), and experienced dysentery during follow-up.
Odds ratio for the relative reduction in the rate of cholera for each percentage increase in vaccine coverage. Adjusted for age, individuals living in a larger cluster specified in the stratification, wards, monthly per-capita expenditure of the household, individuals living in a household always washing hands with soap and water after defecation, individuals living in their own house, individuals living in a household owning at least one luxury item, and distance from the household to the nearest health clinic.
a Hazard ratio for the relative reduction in the risk of cholera for each percentage increase in vaccine coverage. Adjusted for age, sex, religion, distance from household to nearest river, distance from household to nearest treatment center (kilometers), and experienced dysentery during follow-up.
b Odds ratio for the relative reduction in the rate of cholera for each percentage increase in vaccine coverage. Adjusted for age, individuals living in a larger cluster specified in the stratification, wards, monthly per-capita expenditure of the household, individuals living in a household always washing hands with soap and water after defecation, individuals living in their own house, individuals living in a household owning at least one luxury item, and distance from the household to the nearest health clinic.
Cholera vaccine coverage for each concentric ring in the present analysis is presented in Table 2. Average vaccine coverage in each ring ranged from 41% to 45% in Matlab and from 26% to 30% in Kolkata. In the crude analyses in Matlab, a consistent inverse relationship between the quintile of vaccine coverage within the concentric ring and the risk of cholera in the focal person at the center of the ring was observed for all rings with radii up to 500 m. However, such a relationship was observed only for rings with radii up to 200 m for focal persons in Kolkata (Table 3).
Table 2Descriptive statistics for inactivated oral cholera vaccine coverage of populations residing in the different concentric rings surrounding placebo recipients in the Matlab (Bangladesh) and Kolkata (India) trials.
Table 3Risk of cholera among placebo recipients by quintile of vaccine coverage for each concentric ring around the placebo recipients in the Matlab (Bangladesh) and Kolkata (India) trials.
The backward elimination process for selecting covariates for the final model resulted in no covariates retained for Matlab and only age retained for Kolkata. Multivariable logistic regression models showed a significant association between OCV coverage in the ring and the risk of cholera in focal persons for rings with radii up to 500 m in Matlab and up to 150 m for placebo recipients in Kolkata (Table 4). To assess whether the relationships between OCV coverage and cholera risk in focal persons, by distance from the focal person, could be explained simply by variations in population density, population density was added as a covariate in these models. Models that included population density as a covariate again showed significant inverse associations between OCV coverage and the risk of cholera in focal persons for distances up to 500 m in Matlab and 150 m in Kolkata.
Table 4Odds ratios relating the risk of cholera in placebo recipients to inactivated oral cholera vaccine coverage in successive concentric rings of populations in the Matlab (Bangladesh) and Kolkata (India) trials.
Using analyses of the indirect herd protection conferred by OCVs as a probe to define the geographical dimensions of chains of person-to-person transmission of cholera in Matlab, a rural site, and Kolkata, an urban site, an inverse relationship was found between vaccine coverage of the surrounding population and the risk of cholera for placebo recipients for virtual clusters with radii up to 500 m in Matlab and up to 150 m in Kolkata. These results suggest that the dimensions of chains of person-to-person transmission in endemic settings can be quite large and may differ substantially from setting to setting.
Cholera has long been known to occur in geographic clusters. Early studies of space–time clustering of cholera noted clustering of cases within the same household or within small groups of nearby households (
Phage types of Vibrio cholerae O1 biotype El Tor isolated from patients and family which vaccination contacts in Bangladesh: epidemiologic implications.
), leading to the conventional wisdom that the person-to-person spread of cholera is usually limited to household transmission. More recently, two analyses performed in Matlab and Kolkata found an increased risk of clinically significant cholera shortly after the detection of index cases for populations within 450 m of the residence of index cases in Matlab, and within 50 m of residences of index cases in Kolkata (
It is interesting to compare the present study estimates with those of these more recent estimates, particularly since the present study and the earlier studies were done at the same study sites (
). All three analyses identified surrounding populations much larger than the immediate household or clusters of households as being relevant to cholera transmission, and all identified substantially larger zones in the rural Bangladesh site than in the urban India site. However, in contrast to the two earlier analyses in Matlab and Kolkata, which evaluated the population at risk for clinically significant cholera around identified cholera cases, regardless of the type of transmission, the present study analyses specifically addressed the size of surrounding zones in which the use of OCVs had a demonstrable effect on person-to-person transmission.
Whether a simple proxy measure – population density – might serve in lieu of analyses of OCV herd protection to define the geographic dimensions of person-to-person transmission was examined. However, inclusion of population density as an additional covariate in the multivariable models failed to erase the relationships between vaccine coverage and cholera risk in focal persons, by distance from the focal person, indicating that population density cannot be used as a simple proxy measure for estimating the geographical dimensions of chains of person-to-person transmission.
Several potential limitations of this analysis require discussion. First, the impact of vaccine coverage calculated at the time that vaccine doses were given was analyzed, rather than vaccine coverage during post-vaccination follow-up, which would have changed over time due to population births, deaths, and migrations. Such misclassification of coverage over time would most likely have been non-differential with respect to the risk of cholera and would have tended to make the estimates of the radii of persons contributing to person-to-person transmission of cholera conservative. Second, in the analyses the ORs were in the range of 0.974–0.999 for Matlab and 0.949–0.996 for Kolkata, which may seem modest in magnitude. However, it should be appreciated that these ORs reflect relative reductions of risk of cholera per one percentage increase in vaccine coverage. Third, for simplicity, the analyses considered concentric circular geographic rings of populations. It may be that zones of person-to-person transmission in the two settings did not conform geometrically to circles.
Fourth, although randomized clinical trials were analyzed, the present analyses of herd protection were observational in design and were subject to the limitations of such a non-randomized design. Fifth, the maximum radius for herd protection may vary not only from site to site, but also from vaccine to vaccine; had the studies been performed with other cholera vaccines, the results may have been different, depending on the capacity of the vaccines studied to interrupt transmission. Nonetheless, the present analyses are of relevance to the deployment of inactivated OCVs, which are currently stockpiled by the WHO for deployment in cholera-affected settings. Sixth, the analyses addressed cholera occurring in endemic settings, in which the population has some level of natural immunity; whether the findings can be generalized to epidemic settings in immunologically naïve populations, where the global OCV stockpile is occasionally deployed, is uncertain and constitutes an important subject for future research. Finally, caution should be exercised in generalizing the different geographical dimensions of person-to-person transmission in Kolkata versus Dhaka to other urban versus rural areas. The Matlab and Kolkata sites differed in multiple ways, not all of which were related to their urban versus rural contexts.
This study has both theoretical and practical implications. The maximum size of virtual clusters for which indirect vaccine protection was detected in the two trials differed from the sizes of virtual clusters that were analyzed in the earlier published analyses of the trials, in which a radius for virtual clusters was determined with an arbitrary statistical criterion (
). This discrepancy suggests that work is required to improve the earlier method for sizing virtual clusters in analyzing vaccine herd protection. On the other hand, it was reassuring that the present study analyses showed a pattern of declining vaccine herd protection with increasing distance of the vaccinated population from the central placebo recipient under analysis, providing evidence of the construct validity of the method. From a practical perspective, the approach used herein may provide guidance to geographically targeted OCV strategies for the control of outbreaks of cholera occurring in endemic settings.
In summary, the results of this study suggest that the dimensions of chains of person-to-person transmission in endemic settings can be quite large and may differ substantially from setting to setting. Using OCVs as ‘probes’ to define these dimensions can inform geographical targeting strategies for the deployment of these vaccines in endemic settings.
Contributors
MA and JC contributed to the study design. MA, DS, SK, SD, FM, SKB, and JC contributed to the implementation and supervision of the study. MA, BM, and DRK analyzed the data and took responsibility for the accuracy of the data analysis. All authors participated in the writing of the manuscript and approved the final version of the manuscript.
Funding
Bill & Melinda Gates Foundation. The icddr,b receives core funding from the governments of Bangladesh, Canada, Sweden, and the UK. The International Vaccine Institute receives core funding from the governments of South Korea and Sweden.
Conflict of interest
None declared.
Acknowledgements
We are grateful to the people of the study areas and the field staff for whom the project was successfully conducted. This study was funded by the Bill & Melinda Gates Foundation in grants to the icddr,b, the International Vaccine Institute (IVI), and the DOVE project of Johns Hopkins Bloomberg School of Public Health (OPP1148763). The study was also supported by core grants to the icddr,b from the governments of Bangladesh, Canada, Sweden, and the UK and to the IVI from the governments of Sweden and South Korea. The sponsors had no role in the study design, in the collection, analysis, and interpretation of the data, in the writing of the manuscript, or in the decision to submit the manuscript for publication.
References
Ali M.
Emch M.
von Seidlein L.
Yunus M.
Sack D.A.
Rao M.
et al.
Herd immunity conferred by killed oral cholera vaccines in Bangladesh: a reanalysis.
Phage types of Vibrio cholerae O1 biotype El Tor isolated from patients and family which vaccination contacts in Bangladesh: epidemiologic implications.
Efficacy and safety of a modified killed whole-cell oral cholera vaccine in India: an interim analysis of a cluster-randomized, double-blind, placebo-controlled trial.
30302-8&id=gr1.jpg){kind=link}