Many oral vaccines, including rotavirus vaccines, provide lower protection from disease to children in low-resource settings than to children in high-resource settings(1). There are four internationally available, World Health Organization (WHO) pre-qualified, live, oral rotavirus vaccines (Rotarix (GlaxoSmithKline Biologicals, Rixensart, Belgium), Rotasiil (Serum Institute of India Pvt. Ltd., Pune, India), RotaTeq (Merck & Co., West Point, PA, USA), and Rotavac (Bharat Biotech International Ltd., Hyderabad, India)) and 2 nationally licensed live, oral rotavirus vaccines ((Rotavin (PolyVac, Hanoi, Vietnam) and Lanzhou Lamb Rotavirus Vaccine (Lanzhou Institute of Biological Products Co. Ltd., Lanzhou, China))(2). Since their first licensure in 2006 and WHO’s expanded implementation recommendation in 2009, rotavirus vaccines have had a tremendous impact on diarrheal disease morbidity and mortality in the >120 countries that have introduced them as part of routine infant vaccination programs(3–7). Randomized controlled clinical trials and post-licensure rotavirus vaccine effectiveness (VE) evaluations have shown a gradient in the level of vaccine-induced protection by national <5-year-old child mortality (U5M) levels, where VE is lower in countries with higher overall U5M and VE is higher in countries with lower overall U5M(8–10). While there are many hypothesized reasons for this disparity including malnutrition, interference from oral polio vaccine, and microbiome differences, data are limited(11). Because diarrhea disease burden is often higher in high child mortality settings than in low mortality settings, even modest protection represents large reductions in illness(12). Despite the success of rotavirus vaccines, rotavirus remains a leading cause of diarrhea hospitalizations among young children globally(13).

Many post-licensure rotavirus VE evaluations have implemented a test-negative case-control design (TND), meaning cases are children who sought healthcare for acute gastroenteritis (AGE) and tested positive for rotavirus and controls are children who sought healthcare for AGE and tested negative for rotavirus(14). The TND helps ensure cases and controls come from the same source population and have similar healthcare-seeking behaviors and it assumes the controls are representative of the general population. TND evaluations are usually less expensive and simpler to implement than VE evaluations using community or hospital non-diarrhea controls. Rotavirus VE results have been similar when using TND and other control groups in the same settings(15–17). TND has been widely used to evaluate other vaccines, including influenza and COVID-19(14, 18).

To improve vaccine performance and further reduce the burden of hospitalizations and deaths due to rotavirus, there is a need to understand the underlying cause of vaccine performance disparities. We propose calculating VE among various subpopulations and against rare but important outcomes such as death, particularly in countries with high U5M where rotavirus VE is lower. With this information, strategies can be developed to optimize the currently available vaccines and prioritize the design of future rotavirus vaccines. Studies conducted in individual countries typically do not have sufficient sample size to explore VE among small subpopulations and rare outcomes(1, 11). Similarly, literature reviews and meta-analyses are limited by the analyses available(8, 19). The Multi-National Subpopulations Study to Evaluate Rotavirus Vaccines (MNSSTER-V) project aims to address this limitation and estimate VE for these subpopulations and outcomes by pooling child-level data from 25 TND rotavirus VE studies. In this report, we summarize the MNSSTER-V project methods used to combine datasets and present complete and partial series rotavirus VE estimates stratified by country-level characteristics.

The final analysis included 6,626 children 3–59 months old with AGE who tested positive for rotavirus (cases) and 19,459 children 3–59 months old who tested negative for rotavirus (controls) from 25 datasets collected in 24 countries, 16 of which used Rotarix, 5 used RotaTeq, and 1 each used Rotasiil, Rotavin, and both Rotarix and RotaTeq (Table 1, Supplementary Table 1). The number of children 3–59 months old from each country is visualized in Figure 1. The largest proportion of children were 6–11 months old (n=11,410; 44%), followed by 12–23 months old (n=8,830; 34%). Rotavirus positivity was similar between age groups, ranging from 24% in 3–5 and 6–11 month-old children to 28% in 12–23 and 24–59 month old children. Thirteen countries in the WHO African Region contributed data from 8,562 children; 5 countries in the European Region contributed data from 4,370 children; 4 countries in the Americas Regions contributed data from 7,090 children; and 2 countries from other regions (Eastern Mediterranean and Western Pacific Regions) together contributed data from 6,063 children. Rotavirus positivity ranged from 14% in the European region to 31% in the combined Eastern Mediterranean and Western Pacific regions. Three countries in the combined low and medium U5M quartiles contributed data from 2,818 children. There were 4 countries in the high U5M strata (n=6,955 children), 10 countries in the higher U5M strata (n=9,060 children), and 7 countries in the highest U5M strata (n=7,252 children). Rotavirus positivity increased with U5M strata, from 12% in the low and medium U5M strata to 30% in the highest U5M strata. There were 7 countries contributing data from 6,939 children in Gavi’s initial self-financing group; 4 countries contributing data from 4,440 children in the preparatory transition group; 3 countries contributing data from 1,050 children in the accelerated transition group; and 6 countries contributing data from 9,204 children in the fully self-financing group. Among countries whose rotavirus vaccine introduction was supported by Gavi, rotavirus positivity increased from 20% in the currently fully self-financing group to 31% in the initial self-financing group.

Overall, 18,709 children 3–59 months old received a complete rotavirus vaccine series, of which 24% tested positive for rotavirus (Table 1). There were 4,412 children 3–59 months old who received a partial rotavirus vaccine series, of which 25% tested positive for rotavirus, and 2,781 children who did not receive any rotavirus vaccine doses, of which 31% tested positive for rotavirus. Adjusted ≥1 rotavirus vaccine dose VE among children 3–59 months old was 50% (95%CI: 45, 55) (Figure 2, Supplementary Table 2). Adjusted complete series VE of any vaccine brand among children 6–59 months old was 53% (95%CI: 48, 58) and partial series VE was 42% (95%CI: 32, 50). For both Rotarix and RotaTeq, complete series VE point estimate was higher than partial series VE (Supplementary Table 2).

The adjusted complete series VE point estimate declined modestly by age group from 56% (95%CI: 48, 62) among children 6–11 months old to 40% (95%CI: 3, 63) among children 24–59 months old (Figure 2, Supplementary Table 2). The low and medium U5M strata had significantly higher adjusted complete series VE (74% (95%CI: 64, 81)) compared to all groups within the high quantiles. Among the high quantile groups, VE point estimates were 50% (95%CI: 38, 60), 52% (95%CI: 42, 60), and 46% (95%CI: 31, 57) in the high, higher, and highest mortality strata, respectively. Partial series VE point estimates were somewhat lower than complete series, decreasing from 66% (95%CI: 38, 81) in the low and medium strata to 37% (95%CI: 13, 55), 37% (95%CI: 20, 51) and 35% (95%CI: 12, 51) in the high, higher, and highest mortality strata, respectively. Complete series rotavirus VE point estimates also declined as the current level of Gavi support increased, from 65% (95%CI: 57, 71) among fully self-financing countries to 44% (95%CI: 29, 56) among initial self-financing countries.

In sensitivity analysis, after excluding 8 datasets that did not include symptom onset date, the adjusted VE was 51% (95%CI: 45, 58) both when doses given >14 days before arrival to the surveillance hospital and when doses given >14 days before symptom onset were considered to be administered. Second, complete series VE among children 8–59 months old was 50% (95%CI: 44, 56) with any vaccine brand, 48% (95%CI: 40, 55) with Rotarix, and 60% (95%CI: 46, 70) with RotaTeq. Finally, 7 strata necessitated a weighted analysis because one country represented 50% of observations (high U5M strata; highest U5M strata; Americas WHO region; initial self-financing, accelerated transition, and fully self-financing Gavi groups; and RotaTeq complete series). The VE point estimate did not change by more than 3 percentage points in any strata when large datasets were down weighted.

We found that rotavirus VE was significantly lower in countries with high overall U5M than in countries with low and medium overall U5M. Literature reviews found rotavirus VE to be between 85–95% in low U5M countries, 60–80% in medium U5M countries, and 35–65% in high U5M countries (8, 9). Our results showing 74% VE in low and medium U5M countries and 46–52% VE in the three high U5M strata concur with the published literature, though our analysis was able to estimate VE with more precision. We found VE varied somewhat among the high, higher, and highest U5M strata, suggesting some heterogeneity in vaccine performance within the high mortality groups. Using the Gavi classification of countries, following a different set of criteria based on country income, we found a similar, significant gradient in vaccine performance. The MNSSTER-V dataset’s size and the diversity of contributing countries allows for comprehensive consideration of potential drivers of these disparities in rotavirus vaccine performance.

Partial series effectiveness was only 42%, compared to 50% for the complete series. In previous literature reviews and metanalyses, information about partial dose rotavirus VE was very limited, and, when available, estimates were imprecise(26). Because of the substantial sample size in this dataset, we quantified partial dose series VE with relative precision. Although partial series protection was demonstrated, we found a benefit to complete series vaccination over partial series vaccination with both Rotarix and RotaTeq and at all U5M levels. This information is important for decisionmakers and vaccinators to promote complete series rotavirus vaccination with any rotavirus vaccine brand to optimize protection and further reduce the burden of rotavirus hospitalizations and deaths.

This pooled dataset of 26,085 children from 25 studies across 24 countries holds promise as the largest global dataset to date to address questions regarding heterogeneity in rotavirus VE estimates in different geographic and socioeconomic settings. Combining multiple datasets allows estimating VE against outcomes and exposures that are not available from individual datasets or metanalyses. Additionally, this analysis confirms disparities in rotavirus vaccine performance; however, we did not further address the potential causes for differences in VE, such as levels of malnutrition in the population and waning immunity. These reasons require more detailed exploration and analyses than space allows and future analyses of this dataset are planned.

We used logistic regression models for this analysis, rather than fixed or random effects models(27–29). Although there were some differences in quality and cleanliness across the included datasets, the enrollment protocols and designs were essentially identical across studies and any implementation differences were assumed to be comparable to differences between sites in a multi-site evaluation within a single country. The outcome (rotavirus diarrhea) was measured using the same type of test in all but one dataset. For the exposure, we used strict, predefined criteria to interpret and include vaccine card data across sites. Our standardized case definition in this analysis should minimize the impact of enrollment implementation differences between evaluations(20). Sensitivity analyses demonstrated that large datasets did not overwhelmingly influence the VE point estimates or precision.

These findings should be considered alongside a few limitations. First, we assumed all children within a country were vaccinated with the vaccine brand provided through the national immunization program. Although rare, other vaccine brands may have been available on the private market and children who moved between countries may have been vaccinated with another brand. Vaccine brand was not captured during data collection in any country except Guatemala. Similarly, national U5M does not account for in-country heterogeneity and we assumed all children’s situations would be reflected in national U5M. Second, only a small number of datasets were available for certain strata and those VE results are not presented. WHO regions in Asia (Eastern Mediterranean, Southeast Asia, and Western Pacific) are notably underrepresented in this analysis, which reflects delayed rotavirus vaccine implementation in many Asian countries. If more datasets become available, particularly from Asian countries or using alternate rotavirus vaccine brands, additional analyses will be informative. Finally, not all variables in the protocol were available from all the datasets, either because the information was not collected or because the original data management team determined the quality was insufficient. For this analysis, we limited variables in the descriptive analysis and in the logistic regression models to those available from all datasets so as not to exclude individual countries. Future analyses using this dataset may choose to maximize the possible confounding variables under consideration in exchange for inclusiveness.

The MNSSTER-V project used a combined dataset to assess rotavirus VE for a range of country-level stratifications. Our findings demonstrated reduced partial series VE compared to complete series VE and support continued efforts to increase complete series rotavirus vaccine coverage. Finally, these results agree with literature reviews and metanalyses showing differential rotavirus vaccine performance between high, medium, and low U5M countries. This dataset offers many possibilities to better understand actionable causes of these differences to maximize rotavirus vaccines’ impact on diarrhea morbidity and mortality.