To evaluate published post-licensure studies of indirect protection from rotavirus vaccine, we performed a literature review of articles indexed in PubMed, Embase, CINAHL, and Scopus databases. We used search terms ((rotavirus AND (vaccin* OR immuni*)) OR ((herd immunity OR herd protection OR indirect immunity OR indirect protection) and (RotaTeq OR Rotarix OR Rotavac OR Rotasiil))) from 1 January 2006 to 31 December 2022. We excluded any reviews, presentations, non-primary research, and publications that did not include estimates specific to internationally licensed human rotavirus. We did not restrict the search by language.

We abstracted information from articles using a standardized collection form in Microsoft Excel. We defined indirect protection as any estimate of the change in rotavirus morbidity among unvaccinated individuals in settings with rotavirus vaccine introduction compared to all individuals in settings without rotavirus vaccine introduction. Any metric that evaluated indirect protection was abstracted, including incidence rate ratios, relative percent change in the proportion of acute gastroenteritis stool specimens positive for rotavirus (henceforth referred to as percent positivity), or relative change in absolute numbers of positive rotavirus tests or medical encounters with an International Classification of Diseases (ICD) code for rotavirus. Incidence rate ratios were directly abstracted if available, back-calculated from VE estimates, or calculated as the ratio of incidence rates. Relative changes in percent positivity or in absolute numbers of rotavirus positive tests or of ICD codes were either directly abstracted or calculated using reported frequencies. We abstracted data on the type of outcome (i.e. hospitalizations, emergency department visits, laboratory detections, outpatient visits, nosocomial infections, mixed medical encounters, community infections, asymptomatic infections). We abstracted estimates stratified by outcome, age, and years since vaccine introduction.

We summarized the frequency and proportion of studies by several factors: World Health Organization (WHO) region, the childhood mortality strata of the country of study, group evaluated for indirect protection (age-eligible unvaccinated, age-ineligible, or both), whether comparisons were made temporally or geographically, the data source, and the methods for ascertaining rotavirus infection. We categorized countries into three child mortality strata: lowest quartile (“low”); 2nd quartile (“medium”); and the 3rd and 4th quartiles (“high”), using the 2014 (median study period year) estimates of <5 year old child mortality from UNICEF [16]. Taiwan was considered to be low mortality based on locally published mortality rates [17]. Data sources were categorized as active hospital surveillance, administrative data with ICD codes, laboratory surveillance, nationally notifiable case database, case-control study, or cluster randomized trial. The methods for ascertaining rotavirus infection included ICD-9 or ICD-10 codes or laboratory tests including enzyme immunoassays (EIA), polymerase chain reactions (PCR), immunochromatography, polyacrylamide gel electrophoresis (PAGE), or unspecified laboratory methods. We summarized incidence rate ratios, relative percent change in percent positivity, and relative percent change in absolute number of rotavirus positive tests or rotavirus-specific ICD codes by age category, outcome, years since introduction, and whether the study collected data annually or only during rotavirus seasons. We sorted the age groups into six strata: infants <2 months, children <5 years, children ≥5 years (maximum age between 5 and 19 years), children and adults ≥5 years (maximum age not specified), adults 18–64 years of age, and older adults ≥65 years of age. Any age groups ranging over multiple age categories were assigned to the category with the most overlap (Supplemental Table 1). For data stratified by single years since vaccine introduction, we evaluated years 0 and 1 since vaccine introduction as single year periods and grouped years 2–4 and 5–12 since vaccine introduction given the sparse data in these later time periods. Given the heterogeneity in the study designs and data quality, we did not use fixed- or random-effects models to pool estimates [18]. Instead, the distributions of incidence rate ratios, relative percent changes in percent positivity, and relative percent change in absolute numbers were summarized as medians and interquartile ranges (IQR). The primary analysis summarized distributions of estimates of indirect protection against hospitalizations due to rotavirus gastroenteritis among unvaccinated children <5 years old, as the majority of estimates were for this outcome and age category. The median and IQR incidence rate ratio for hospitalizations due to rotavirus gastroenteritis among unvaccinated children <5 years old was also stratified by child mortality, given evidence of lower direct protection from rotavirus vaccination in countries with medium and high child mortality [19]. Sensitivity analyses were conducted stratifying estimates by age groups and years since introduction. Private-market rotavirus introduction and seasonal data collection were also considered to be associated with a higher risk of bias and studies with these characteristics were excluded in sensitivity analyses. Data source was also considered a factor that may influence the risk of bias in a study, especially for studies that relied on administrative databases using ICD codes or laboratory surveillance that may not systematically identify cases, thus a sensitivity was also conducted stratifying estimates by data source.

Our search criteria yielded 2,293 publications, of which 71 met the inclusion criteria (Figure 1). Among the 71 included studies published between 2009 and 2022, there were 399 estimates of indirect protection from rotavirus vaccines that could be abstracted, stratified by metric, outcome, age group, and years since vaccine introduction. The median number of available estimates per study was 3 (IQR: 2–6.5, range 1–30) (Supplemental Table 1). The majority of studies were from the European Region (42%), the Region of the Americas (23%), and the African region (17%), and were primarily from countries with low childhood mortality (68%) (Table 1). Most studies were among adult and child populations that were age-ineligible at the time of the study for rotavirus vaccination (87%), though 6% of studies were for unvaccinated age-eligible children and 7% of studies included both age-ineligible and age-eligible unvaccinated populations. A variety of data sources and methods were used for ascertaining rotavirus infection. Data came from active hospital surveillance (44%), administrative databases (32%), laboratory surveillance (19%), nationally notifiable case surveillance (3%), case-control study (1%), and cluster randomized trial (1%), while 32% of studies used ICD codes only, 3% of studies used ICD codes and laboratory methods, and 65% used a laboratory method only to determine rotavirus infection.

Thirty-six studies provided 229 incidence rate ratios, 27 studies provided 88 estimates of relative percent change in rotavirus percent positivity, and 14 studies provided 82 estimates of relative percent change in absolute number of rotavirus positive tests or rotavirus-specific ICD codes (Table 2).

While more than half of estimates used hospitalizations due to rotavirus gastroenteritis as the outcome (71% of incidence rate ratios, 68% of relative percent change in percent positivity, and 83% of relative percent change in absolute rotavirus positive tests or rotavirus-specific ICD codes) and children <5 years old as the age group (62% of incidence rate ratios, 72% of relative percent change in percent rotavirus positive, and 61% of relative percent change in absolute rotavirus positive tests or rotavirus-specific ICD codes), there was considerable variability in metrics, outcomes, and age categories used to evaluate indirect protection (Table 1). For example, there were 21 different age groupings in which authors measured indirect protection among unvaccinated children <5 years of age (Supplemental Table 2). There were 52 estimates among adults 18–64 years of age and older adults ≥65 years of age, of which 22 (42%) were from administrative databases with ICD codes, 16 (31%) were from laboratory surveillance, and 14 (27%) were from nationally notifiable case-based surveillance; there were no active surveillance studies with indirect protection estimates among adult populations. Most estimates were from the very early post-introduction period (e.g. 206 [52%] estimates were from 0 or 1 years since vaccine introduction) and there were 35 (9%) estimates from studies that collected data seasonally. There were very few (n=23, 6%) estimates 5 or more years after vaccine introduction.

For hospitalizations due to rotavirus gastroenteritis among unvaccinated children <5 years old, the median incidence rate ratio was 0.60 (IQR: 0.40–0.87, n=110 estimates), the median relative percent change in percent positivity was 25% (IQR: 13–44%, n=49 estimates), and the median relative percent change in absolute number of rotavirus positive tests or rotavirus-specific ICD codes was 42% (IQR: 16–66%, n=40 estimates) (Figure 2). When stratified by under-five mortality strata, the median incidence rate ratio in hospitalized children <5 years old was 0.60 (IQR: 0.40–0.86, n=88 estimates) among 19 low-mortality countries, 0.56 (IQR: 0.34–0.62, n=12 estimates) among three medium mortality countries, and 0.83 (IQR: 0.64–1.11, n=10 estimates) among two high-mortality countries (Figure 3). Compared to children <5 years old, the median incidence rate ratios were closer to the null and there were wider interquartile ranges in studies of indirect protection among unvaccinated children ≥5 years old (median incidence rate ratio 0.74 [IQR: 0.45, 1.07], n=24 estimates), adults (median incidence rate ratio 1.17 [IQR, 0.50, 1.54], n=17 estimates), and older adults (median incidence rate ratio 1.20 [IQR: 0.89, 1.66], n=11 estimates) (Supplemental Table 3). Excluding year 0 data, seasonal estimates, and private market introductions had no substantive impact on the indirect protection estimates against for hospitalization for all ages (Supplemental Table 3) or among children <5 years of age (Supplemental Table 4). There was some variability in the distributions of estimates by data source; the median relative percent change in percent positivity and relative percent change in absolute numbers of rotavirus positive tests or of ICD codes was lower among administrative databases with ICD codes compared with active hospital surveillance among all ages (Supplemental Table 3) and among children <5 years of age (Supplemental Table 4).