Rotavirus is the leading cause of severe diarrhea worldwide in children less than 5 years of age, resulting in about 140 million cases of rotavirus-associated gastroenteritis each year.1 Rates of rotavirus disease are similar in developed and developing countries1,2; however, 80%–90% of rotavirus-associated deaths occur in developing countries.3
Vaccination against rotavirus is the single best prevention and control measure implemented to reduce the incidence of rotavirus.4 In 2009, universal rotavirus vaccination was recommended by the World Health Organization.5 Two live, oral rotavirus vaccines are used broadly across the globe (RotaTeq, RV5; Merck & Co., Inc.; Kenilworth, NJ and Rotarix, RV1; GlaxoSmithKline Biologicals, Rixensart, Belgium). As of May 2016, 81 countries have introduced rotavirus vaccines into their national immunization public sector programs while over 100 countries have rotavirus vaccine available through the private market.6 Overall, countries have seen a dramatic health and economic impact as a result of vaccine introduction. In Mexico, an estimated 40%–50% of diarrheal deaths have been prevented in children less than 5 years of age.7,8 Similarly, a recent report demonstrated declines in childhood diarrhea mortality after rotavirus vaccine introduction in Brazil, El Salvador and Nicaragua.9 Comparing pre and post vaccine introduction years (2006 and 2008), the United States saw an 87% reduction in rotavirus hospitalization for 6- to 11-month-old children, and a 96% and 92% reduction for 12- to 23-month-old age groups and 24- to 35-month-old age groups, demonstrating the indirect benefits of vaccination in older children.10
Although there have been reductions in rotavirus-related diarrheal morbidity and mortality, rotavirus vaccines have overall lower estimated efficacy in developing countries when compared with developed countries. The Rotavirus Efficacy and Safety Trial was conducted in 11 developed countries from 2001 to 2004 and the efficacy of RV5 against severe rotavirus gastroenteritis (RVGE) through 1- and 2-year seasons was 98.0% [95% confidence interval (CI): 88.3%–100%] and 88.0% (95% CI: 49.4%–98.7%), respectively.11 Conversely, the estimated efficacies of RV5 against severe RVGE in the trial conducted in Africa (Ghana, Kenya, Mali)12 were lower: 64.2% (95% CI: 40.2%–79.4%) through 1 year and 19.6% (95% CI: −15.7%–44.4%) through 2 years. A similar result was reported in Asian (Bangladesh, Vietnam) developing countries where the efficacies through 1 and 2 years were 51.0% (95% CI: 12.8%–73.3%) and 45.5% (95% CI: 1.2%–70.7%), respectively.13 The disparity in vaccine efficacy between developing and developed countries is similar to results recorded for other live oral vaccines including poliovirus vaccine,14 cholera vaccine15 and typhoid vaccine.16 It is unclear what demographic or clinical characteristics of infants may be responsible for lower efficacy of live oral vaccines. The current study is an exploratory, post hoc analysis of clinical trial data to investigate the effect of baseline infant characteristics on the efficacy of RV5 in Africa (Ghana, Kenya and Mali) and Asia (Bangladesh and Vietnam).
MATERIALS AND METHODS
Study Design and Population
This was a post hoc analysis of data from a previously conducted randomized trial (Clinical Trial Number NCT00362648) of RV5. The design, implementation and results of the study have been previously described in detail.12,13 In brief, this was a double-blind placebo-controlled, multicenter randomized trial conducted to determine the efficacy of 3 doses of RV5 against severe RVGE. The study was conducted from March 2007 to March 2009 in 5 countries eligible for assistance from Gavi: medical facilities in rural Kassena-Nankana district, Ghana; rural Karemo division, Siaya district, Nyanza province, western Kenya; urban Bamako, Mali; rural Matlab, Bangladesh and urban/periurban Nha Trang, Vietnam. Enrolled infants were randomly assigned to receive 3 doses of RV5 or placebo at approximately 4-week intervals (6, 10 and 14 weeks of age) and were followed for approximately 2 years. Stool samples were collected during episodes of gastroenteritis and tested for rotavirus using enzyme immunoassays. Reverse transcription-polymerase chain reaction was used to confirm wild-type rotavirus and to determine rotavirus P and G genotypes. Severity was determined using a 20-point modified Vesikari clinical score for infants with RVGE; a score of ≥11 was considered severe. Infants in Kenya were offered human immunodeficiency virus (HIV) optional counseling and testing for HIV. RV5 efficacy by HIV status has been previously published.17 The study was approved by the Institutional Review Board (Olympia, WA) and institutional review boards and national ethical review committees for each site.
There were 7504 infants randomized to receive RV5 or placebo. Only infants receiving all 3 doses of vaccine (N = 3348) or placebo (N = 3326) were included in the primary analyses. Approximately 10% of randomized infants were excluded, because they violated the protocol, provided samples that could not be evaluated and/or had no follow-up time recorded. The results of the overall RV5 efficacies by country have previously been published.12,13 The original study was powered to estimate efficacy by continent and not by country.
Potential baseline modifying variables of efficacy were collected at the time the first dose was administered. Variables included in the analysis were age at first dose, gender, breastfeeding status and nutrition status. All variables were categorized into binary groups to estimate efficacies within these subgroups (Table 1). These variables were chosen, because they were well measured with limited missing data and could be easily classified into binary groups to estimate subgroup efficacies. The age cutoff was chosen to be 8 weeks to distinguish between groups receiving their first vaccine dose early versus later. Specifically, the 8-week cutoff was chosen, because this was the minimum age where there were relatively large sample sizes in each age-specific stratum to estimate the stratum-specific efficacy of the vaccine while still providing an interesting contrast of ages. Nutrition status was measured using the World Health Organization cutoffs for underweight, stunted and wasting.18
The objective of this research was to estimate vaccine efficacy within subgroups of the trial population to understand potential differences in rotavirus vaccine efficacy by the baseline factors defined (eg, heterogeneity of rotavirus vaccine efficacy). Vaccine efficacy and 95% CIs were estimated using the same method as the original clinical trial but for each stratum of a factor and by continent and country. Sensitivity analyses were done for African continent efficacy estimates by investigating the pairwise combinations of African countries (ie, Ghana and Kenya, Ghana and Mali, Kenya and Mali). This was done to understand important country-level contributions to efficacy estimates by baseline factor. Also, this provided a means to ensure that results were not driven by Mali where there was under ascertainment of the outcome the first year of follow-up,19 which might impact the results presented here. One- and 2-year efficacies were estimated for both RVGE of any severity and severe RVGE as the outcome. Vaccine efficacy was defined as (1 − IRR) × 100%, where the incidence rate ratio (IRR) was equal to the rate of RVGE in the vaccine group divided by the rate in the placebo group. The IRR was estimated using Poisson regression in SAS (version 9.4). To stratify and to investigate heterogeneity by each factor, a model was fit with the vaccination status (vaccine or placebo), the factor of interest and interaction term between vaccination status and factor of interest. The P value of the interaction was analyzed to determine the presence of heterogeneity. Each stratum of the factor analyzed had to have a sample size ≥100 infants for vaccine efficacy to be estimated for that group. An α level of 0.05 was used as the cutoff for statistical significance. Due to the exploratory nature of this analysis, there was no correction for multiple testing.
Because the vaccine and placebo doses were not randomized within strata of factors analyzed, we compared covariate distributions within strata to look for potential imbalance of covariates between the vaccine and placebo group using a χ 2 test and by calculating a standardized proportion difference. The standardized proportion difference was calculated as (p 1 − p 2)/sqrt((p 1(1 − p 1) + p 2(1 − p 2))/2), where p 1 is the proportion (or mean) of the binary covariate in the vaccinated group and p 2 is the proportion in the placebo group. A standardized proportion difference of ≥0.2 was considered a significant imbalance of covariates between vaccinated and placebo groups. Because of small numbers, only crude and not adjusted efficacies were estimated for groups with covariate imbalance; however, these estimates are identified in the results.
The distributions of baseline factors investigated varied by region and are summarized in Table 2. Most infants (~70%–75%) received their first dose before 8 weeks of age in Kenya and Mali, whereas only about half received their first dose at this age in Ghana and Bangladesh and about 10% in Vietnam. About 35% of infants in Kenya and about 65% of infants in Vietnam were exclusively breastfed compared with 99% in Ghana and 90% in Mali and Bangladesh. The prevalence of stunted infants was similar across countries (~5%), except for Kenya where 26% of infants were stunted. About 10% of infants were underweight across all countries except Vietnam (1%) and Ghana (20%). Similarly, 5%–10% of infants were considered wasting across all countries except Ghana (45%). Most infants in Vietnam were neither underweight, stunted or wasting. The length of infants was not measured in Bangladesh; therefore, Z scores could not be calculated.
There was heterogeneity of vaccine efficacy by age at first dose in Africa (Table 3). Lower efficacy was observed in Africa (Ghana, Kenya and Mali) for infants receiving their first dose at <8 weeks of age compared with those vaccinated at ≥8 weeks of age for 2 years of follow-up and RVGE of any severity and severe RVGE as the outcome. The effect in Africa was not statistically significant when Ghana or Mali was removed.
In Ghana, the effect of age at first dose, gender and underweight status were marginally significant (P ~ 0.06). Infants receiving their first dose at <8 weeks of age compared with those vaccinated at ≥8 weeks of age for 2 years of follow-up and RVGE of any severity as the outcome. Male infants had lower vaccine efficacy when compared with female infants for 2 years of follow-up and RVGE of any severity and severe RVGE as the outcome. Infants who were underweight (WFA Z score < −2) had a lower vaccine efficacy when compared with infants who were not underweight for 2 years of follow-up and severe RVGE as the outcome.
No heterogeneity of vaccine efficacy was observed for breastfeeding status, being stunted, or wasting status. No statistically significant differences were detected in Asia (Bangladesh and Vietnam) at P value <0.05; however, the effect of gender was marginally significant (P < 0.1), where males had significantly higher vaccine efficacy when compared with females during 1 year of follow-up and for severe RVGE (Table 4). All subgroup efficacy results are presented in Tables 3 and 4.
In general, there was a balance of covariates between vaccinated and placebo groups within strata of factors analyzed. However, there were some imbalances that are summarized in Table, Supplemental Digital Content 1, http://links.lww.com/INF/C553. Crude vaccine efficacies for strata with covariate imbalances are identified in Tables 3 and 4 (34 of 242 subgroup estimates reported) and should be interpreted with caution. In addition, unmeasured confounders could potentially be imbalanced between all efficacies reported.
In this post hoc, exploratory analysis, statistically significant heterogeneity of vaccine efficacy was observed by age of first dose in African countries (Ghana, Kenya and Mali). Marginally statistically significant differences were observed by age at first dose, gender and underweight status in Ghana and by gender in Asian countries. No statistically significant heterogeneity of vaccine efficacy was observed for breastfeeding, stunted or wasting status in any of the countries analyzed.
In African countries, lower vaccine efficacy estimates were observed for infants receiving their first dose at younger ages compared with those receiving first dose at slightly later ages; however, the age of infants at their subsequent visits was not evaluated. One reason why a younger age at first dose of a live oral vaccine may result in lower efficacy is that, unlike in developed countries, there may be high levels of circulating antirotavirus antibodies in adults living in developing countries.20 Circulating maternal antirotavirus immunoglobulin G antibodies can cross the placenta,21 which can potentially interfere with vaccine immune responses.
In this study, there were marginally significant differences in efficacy by gender that differed based on region. In Ghana, efficacy was lower in males as compared with females, whereas in Asia (Bangladesh and Vietnam), this effect was reversed. The variation in the vaccine efficacy by region could possibly be explained by differences in baseline rates of severe RVGE by gender and region. For example, Morris et al22 reported higher rotavirus-associated mortality in Indian females as compared with males. Although there could potentially be differences in susceptibility to severe rotavirus infection, many other studies have not reported differences by gender,2,23 and it is impossible to know if the differences are because of case capture or random variation in the studies that have reported differences.
Lower efficacy subgroup estimates of marginal statistical significance were observed for infants in Ghana who were underweight (Z score < −2) as compared with those who were not (Z score ≥ −2). The result reported here is similar to a recent cohort study in Bangladesh (N = 258) that found children with who were underweight had significantly lower antibody titers following oral polio vaccination compared with those who were not underweight.24
There are some limitations of this analysis. First, the clinical trial study was not designed to estimate efficacy by subgroup. Therefore, the power to detect differences, particularly within a country, is low because of small sample size. As a result, some factors may modify efficacy, but the difference between efficacy estimates among factors was not large enough to detect. Second, heterogeneity of vaccine efficacy was only analyzed on a multiplicative scale, meaning we only captured variables that modify vaccine efficacy on a relative scale. We expect that the factors that modify vaccine efficacy on the additive (absolute) scale would differ from the multiplicative (efficacy) scale and would be important to identify when considering the population effect of the vaccine. Third, a large number of tests (N = 122) were conducted. Therefore, we have likely identified some “false positives” in our results (ie, factors that do not truly modify efficacy). This was an exploratory analysis with low power to detect differences in efficacy by factor. It is possible that not all factors identified are modifiers of vaccine efficacy and some factors not identified could be modifiers of vaccine efficacy.25 Given this limitation, it is important that these estimates are validated in future studies. Fourth, there was imbalance of covariates between some vaccinated and placebo subgroups, which may result in residual confounding. Because of small numbers, these efficacy estimates could not be adjusted, and they must be interpreted in that context. Also, only the timing of the first dose was evaluated; thus, vaccine efficacy based on the age of the infants at subsequent doses is not known. Similarly, there are other factors not measured or assessed in this analysis that could contribute to lower vaccine efficacy. Finally, the settings (ie, urban vs. rural) varied by country, so we cannot distinguish potential country-level effects from the population-density effects.
In this exploratory analysis, we observed a suggestion of heterogeneity of vaccine efficacy estimates across some baseline infant characteristics, which varied based on geographic region. These results are meant to be hypothesis generating; additional studies designed specifically to evaluate infant demographics and clinical characteristics are needed.
The authors thank Jon Stek for his administrative support.
1. Dennehy PH. Rotavirus
vaccines: an overview. Clin Microbiol Rev. 2008;21:198–208.
2. Shaw AR. The rotavirus
saga revisited. Annu Rev Med. 2013;64:165–174.
vaccines. WHO position paper - January 2013. Releve epidemiologique hebdomadaire / Section d’hygiene du Secretariat de la Societe des Nations = Weekly epidemiological record / Health Section of the Secretariat of the League of Nations. 2013;88:49–64.
5. Meeting of the immunization Strategic Advisory Group of Experts, April 2009–conclusions and recommendations. Releve epidemiologique hebdomadaire / Section d’hygiene du Secretariat de la Societe des Nations = Weekly epidemiological record / Health Section of the Secretariat of the League of Nations. 2009;84:220–236.
7. Richardson V, Hernandez-Pichardo J, Quintanar-Solares M, et al. Effect of rotavirus
vaccination on death from childhood diarrhea in Mexico. N Engl J Med. 2010;362:299–305.
8. Richardson V, Parashar U, Patel M. Childhood diarrhea deaths after rotavirus
vaccination in Mexico. N Engl J Med. 2011;365:772–773.
9. Paternina-Caicedo A, Parashar UD, Alvis-Guzmán N, et al. Effect of rotavirus vaccine
on childhood diarrhea mortality in five Latin American countries. Vaccine
10. Payne DC, Staat MA, Edwards KM, et al; New Vaccine
Surveillance Network (NVSN). Direct and indirect effects of rotavirus
vaccination upon childhood hospitalizations in 3 US Counties, 2006-2009. Clin Infect Dis. 2011;53:245–253.
11. Vesikari T, Matson DO, Dennehy P, et al; Rotavirus
Efficacy and Safety Trial (REST) Study Team. Safety and efficacy of a pentavalent human-bovine (WC3) reassortant rotavirus vaccine
. N Engl J Med. 2006;354:23–33.
12. Armah GE, Sow SO, Breiman RF, et al. Efficacy of pentavalent rotavirus vaccine
against severe rotavirus gastroenteritis
in infants in developing countries
in sub-Saharan Africa: a randomised, double-blind, placebo-controlled trial. Lancet. 2010;376:606–614.
13. Zaman K, Dang DA, Victor JC, et al. Efficacy of pentavalent rotavirus vaccine
against severe rotavirus gastroenteritis
in infants in developing countries
in Asia: a randomised, double-blind, placebo-controlled trial. Lancet. 2010;376:615–623.
14. John TJ. Antibody response of infants in tropics to five doses of oral polio vaccine
. Br Med J. 1976;1:812.
15. Suharyono, Simanjuntak C, Witham N, et al. Safety and immunogenicity of single-dose live oral cholera vaccine
CVD 103-HgR in 5-9-year-old Indonesian children. Lancet. 1992;340:689–694.
16. Simanjuntak CH, Paleologo FP, Punjabi NH, et al. Oral immunisation against typhoid fever in Indonesia with Ty21a vaccine
. Lancet. 1991;338:1055–1059.
17. Feikin DR, Laserson KF, Ojwando J, et al. Efficacy of pentavalent rotavirus vaccine
in a high HIV prevalence population in Kenya. Vaccine
. 2012;30(suppl 1):A52–A60.
19. Sow SO, Tapia M, Haidara FC, et al. Efficacy of the oral pentavalent rotavirus vaccine
in Mali. Vaccine
. 2012;30(suppl 1):A71–A78.
20. Chan J, Nirwati H, Triasih R, et al. Maternal antibodies to rotavirus
: could they interfere with live rotavirus
vaccines in developing countries
21. Patel M, Shane AL, Parashar UD, et al. Oral rotavirus
vaccines: how well will they work where they are needed most? J Infect Dis. 2009;200 Suppl 1:S39–S48.
22. Morris SK, Awasthi S, Khera A, et al; Million Death Study Collaborators. Rotavirus
mortality in India: estimates based on a nationally representative survey of diarrhoeal deaths. Bull World Health Organ. 2012;90:720–727.
23. Parashar UD, Nelson EA, Kang G. Diagnosis, management, and prevention of rotavirus gastroenteritis
in children. BMJ. 2013;347:f7204.
24. Haque R, Snider C, Liu Y, et al. Oral polio vaccine
response in breast fed infants with malnutrition and diarrhea. Vaccine
25. Peduzzi P, Henderson W, Hartigan P, et al. Analysis of randomized controlled trials. Epidemiol Rev. 2002;24:26–38.