Effectiveness of the Oral Human Attenuated Rotavirus Vaccine: A Systematic Review and Meta-analysis—2006–2016

Abstract Background Gastroenteritis caused by rotavirus accounts for considerable morbidity in young children. We aimed to assess the vaccine effectiveness (VE) of the oral rotavirus vaccine Rotarix, as measured by laboratory-confirmed rotavirus infection after referral to hospital and/or emergency departments in children aged <5 years with gastroenteritis. Methods We performed a systematic search for peer-reviewed studies conducted in real-life settings published between 2006 and 2016 and a meta-analysis to calculate the overall Rotarix VE, which was further discriminated through stratified analyses. Results The overall VE estimate was 69% (95% confidence interval [CI], 62% to 75%); stratified analyses revealed a non-negligible impact of factors such as study design and socioeconomic status. Depending on the control group, VE ranged from 63% (95% CI, 52% to 72%) to 81% (95% CI, 69% to 88%) for unmatched and matched rotavirus test–negative controls. VE varied with socioeconomic status: 81% (95% CI, 74% to 86%) in high-income countries, 54% (95% CI, 39% to 65%) in upper-middle-income countries, and 63% (95% CI, 50% to 72%) in lower-middle-income countries. Age, rotavirus strain, and disease severity were also shown to impact VE, but to a lesser extent. Conclusions This meta-analysis of real-world studies showed that Rotarix is effective in helping to prevent hospitalizations and/or emergency department visits due to rotavirus infection.

Rotavirus (RV) is the major cause of severe gastroenteritis (GE) diseases, which amount to a considerable burden of disease in children younger than age 5 years [1]. Although a global decline in mortality was observed in the last decades, RV diseases still accounted for an estimated 215 000 deaths in this age group in 2013 [2].
Vaccination is the best preventive approach against RV diseases [1]. The oral live-attenuated human RV vaccine Rotarix (GSK, Belgium) was introduced in routine immunization programs as of 2006. In 2009, the World Health Organization recommended the global implementation of RV vaccination in infants [3]. By June 2017, 85 countries had introduced RV vaccination in their national immunization program (NIP), with an additional 7 countries including it in subnational programs and other countries making the vaccines available for private market use [4,5].
Rotarix is a 2-dose-shedule oral live-attenuated human RV vaccine, recommended for active immunization against GE due to RV infection (RVGE) in infants aged 6-24 weeks [6]. The 2 doses of Rotarix should be administered at least 4 weeks apart, and the vaccination course must be completed by 24 weeks of age [6]. Rotarix was shown to have a favorable benefit/risk profile in infants and was efficacious against severe GE-or RV-associated hospitalization in several large clinical trials conducted worldwide [7][8][9][10][11][12][13][14], with a vaccine efficacy of 85%-96% demonstrated against these end points [13,14]. Favorable data from clinical trials were further supported by postlicensure studies conducted over a period of more than 10 years since the introduction of the vaccine in routine immunization programs [5]. Rotarix was shown to provide broad protection against severe RVGE caused by nonvaccine RV strains; that is, efficacy or effectiveness has been demonstrated against 9 different strains [6].
This study evaluated the vaccine effectiveness (VE) of Rotarix, as measured by laboratory-confirmed RV infection after referral to hospitals and/or the emergency department (ED), in children with GE diseases in real-world settings. We conducted a systematic literature review and a meta-analysis of the VE of 2-or 1-dose Rotarix vaccination data published between January 1, 2006, and July 7, 2016. Figure 1 represents a "Focus on the Patient Section, " which elaborates on the clinical relevance and impact of the study, to be shared with patients by health care professionals.

Systematic Literature Review
We performed a systematic search of the PubMed and Cochrane databases for peer-reviewed articles published from January 1, 2006, to July 7, 2016, using prespecified terms related to RV vaccines, as detailed in the Supplementary Data. We included papers in any language reporting data from postlicensure or original studies assessing the VE of Rotarix (full inclusion and exclusion criteria are presented in the Supplementary Data).
Relevant references were selected by a 3-step selection procedure. First, titles and abstracts identified through the search were screened based on their relevance to the objectives, with a random sample of 30% of titles and abstracts being screened in duplicate. Second, a full-text review of articles selected during the first step was performed, with the first 10% of the articles being appraised by 2 reviewers. Third, further scrutiny of the articles during the data extraction phase was applied. For example, when 2 included articles described results of the same study, we only included 1 of the articles in the meta-analysis to avoid double inclusion of data (ie, the article published most recently or with the most relevant data). In addition, the reference list of meta-analyses or systematic reviews was checked for relevant articles that could have been missed. The quality of the selected articles was assessed using the Coordination of Cancer Clinical Practice Guidelines (CoCanCPG) [15].
We extracted and summarized the following data as a minimum: study design, setting and period, study objectives, study country and its socioeconomic status (SES; according to the World Bank list of economies classification [16]), type of control group used (matched/unmatched hospital, test-negative, community/neighborhood controls), clinical setting (hospitalizations or ED visits), RV strain type (homotypic, fully/partly heterotypic), disease severity (mild, moderate, severe, and very severe, according to the Vesikari score list [17]), reports on vaccine introduction in the NIP, and vaccination coverage, when available.

Meta-analyses
A meta-analysis was performed to assess the overall VE of Rotarix, as measured by laboratory-confirmed RV infection after referral to hospitals and/or EDs in children with GE under 5 years of age, as reported by observational studies identified by the systematic literature search (see the Supplementary Data for a full list of inclusion criteria).
We estimated the overall VE in children receiving 2 doses of Rotarix (main analysis). The secondary objectives of the meta-analysis were to assess the VE according to the number of Rotarix doses provided (1 or 2), type of controls used in the studies, the SES of the country, RV strain type, age (<1 or ≥1 year), and disease severity. Stratification analyses were carried out on 1 level (age, RV strain, SES, and disease severity) for 1 Rotarix dose and the complete schedule, whereas 2-level stratified analyses were conducted only for 2 doses of Rotarix (by age and SES and by strain and SES). A meta-analysis was performed only when at least 4 VE estimates could be included. To investigate the effect of specific study parameters on the VE, sensitivity analyses were conducted by excluding studies reporting on ED referrals only or primary health care centers, high-income countries and unmatched control groups, or unadjusted data.

Statistical Methods
In performing the analyses, several considerations were predefined for each included study. If VE estimates were presented for multiple control groups, only 1 estimate was selected as follows: hospital controls were preferred above community controls; in case of multiple hospital controls, matched controls were used; and when nonmatched control groups were studied, test-negative controls were preferred. When both crude and covariate-adjusted VE were provided in studies, adjusted estimates were used for the meta-analysis.
VE was defined as the percent reduction in the odds of referral to hospitals and/or EDs due to RVGE disease among vaccinated children compared with unvaccinated children. Meta-analyses were performed on odds ratios (ORs) and 95% confidence intervals (CIs); log(OR) and standard error of log(OR) were computed. If the ORs and 95% CIs were not available in the included studies, these were calculated from the estimates included in the articles using the formula OR = 1 -(VE/100) [18].
The random-effect model (using the DerSimonian-Laird approach) [19] was used for the main model, but the fixed-effect model (using the inverse variance method) [20] was also employed to calculate pooled ORs. The level of study heterogeneity was assessed by computing the Higgins I 2 test, along with visual assessment of the funnel plots [21]. I 2 values of <25%, 25%-50%, 50%-75%, and >75% were considered very low, low, medium, and high heterogeneity, respectively [22].
Publication bias was investigated for the overall 1-and 2-dose VE analyses, by visual assessment of funnel plots, and by Egger's weighted regression test, with a 2-sided P value of <.10 considered significant [23].
All analyses were performed without any adjustment for multiplicity using STATA v13.1.

Focus on the Patient
Rotavirus is the main cause of acute diarrhea among young children worldwide. As the disease can cause hospitalization and even death, vaccines against rotavirus have been developed for the pediatric population. Rotarix, an oral rotavirus vaccine, has been introduced in national immunization programs in more than 80 countries worldwide resulting in significant reduction in childhood morbidity and mortality due to acute diarrhea.
We reviewed published data on Rotarix, from its introduction in national immunization programs in 2006 and over a 10-year time period, in order to assess the e ect of the vaccine in real-life settings. We showed that the two-dose rotavirus vaccination reduced hospitalization and/or emergency department referrals due to rotavirus gastroenteritis by an average of 69%. The reduction was 81% in high-income countries, 54% in upper-middle-income countries and 63% in lower-middle-income countries.
Our analysis of published data provides further evidence of the e ective protection o ered by Rotarix against rotavirus gastroenteritis in real-world settings.
What is the context?

What is new?
What is the impact? Meta-analysis of Rotarix Effectiveness • OFID • 3

Systematic Literature Review
After removal of duplicates, the search strategy yielded 2890 unique records. Following the screening of titles and abstracts, we retained 261 articles for full-text review, from which 32 studies  were identified as relevant for the assessed outcomes. Figure 2 gives a schematic overview of the selection procedure used. Study characteristics are summarized in Table 1.

Meta-analysis
The meta-analysis included 29 studies ( Figure 2). For the 3 studies excluded [53][54][55], RV disease was only confirmed based on International Classification of Diseases codes and/or electronic medical records and was not based on laboratory results ( Table 1).
All 29 studies included in the meta-analysis were case-control studies and fulfilled the quality criteria of the CoCanCPG checklist.

Meta-analysis of the Effectiveness of Rotarix Vaccine
The main analysis included 27 studies evaluating the overall VE of 2 doses of Rotarix. Two [32,45] of the 29 studies identified by the systematic search were excluded from the main analysis, as only stratified data were presented without reporting overall results. The reported or calculated ORs per study ranged from 0.06 to 0.84. The overall VE estimate was 69% (95% CI, 62% to 75%), with medium heterogeneity (I 2 = 67%; 95% CI, 50% to 78%) observed between studies (Table 2, Figure 3). Among the 12 studies included in the secondary analysis for 1 dose of Rotarix, the pooled VE was 46% (95% CI, 34% to 57%) with an I 2 of 34% (95% CI, 0% to 67%), showing low between-study heterogeneity (Supplementary Figure 1).
Stratified analyses on 2 levels performed for the 2-dose VE confirmed a higher VE in children aged <1 year than in those ≥1 year of age, in both upper-middle-and lower-middle-income countries (Supplementary Table 1).
When excluding data from studies conducted only in ED settings or primary health care centers (not in hospital settings), sensitivity analyses yielded a 2-dose VE of 69% (95% CI, 61% to 76%), similar to that obtained from the main analysis. When omitting data from high-income countries, VE estimates were found to be in the same ranges as those obtained for the main analysis (ie, 60%; 95% CI, 51% to 67%; and 42%; 95% CI, 32% to 50%; for 2 doses and 1 dose, respectively). Higher VE estimates for both 2 doses and 1 dose of Rotarix were observed when excluding unmatched controls or studies where the matching process was not clear ( Table 2).

DISCUSSION
This meta-analysis of the VE of the Rotarix vaccine includes peer-reviewed data publicly available on its use in real-world settings over a period of 10 years. Although several metaanalyses assessing Rotarix effectiveness against various end points have already been published [5,[56][57][58], we included worldwide data and performed subgroup analyses, thus providing an exhaustive view on the VE of Rotarix against RVGE-related hospitalizations or ED visits. Our systematic review and meta-analysis showed that programmatic use of Rotarix prevents hospital admission or ED visits due to RVGE in children under 5 years of age.
We analyzed data from 29 studies assessing postlicensure VE, conducted in various geographical regions, and covering the entire range of SES. Most of the studies included in the review were undertaken in countries where Rotarix is implemented in the NIP. Studies from Taiwan and Portugal, where the vaccine is only available on the private market, were also included.
The meta-analysis showed that a 2-dose schedule of Rotarix provided considerable prevention of RV disease-related hospitalizations or referrals to ED, with a pooled VE estimate of 69%, but increasing to up to 81% in high-income countries. The estimated VE for 1 dose was 46%, indicating that partial reduction of hospitalization is also provided by 1 dose of Rotarix. However, our results highlight the importance of a full vaccination schedule.
Overall VE values obtained from stratified analyses were the highest in high-income countries. This is consistent with results from clinical trials, showing lower vaccine efficacy in African countries [11] than in industrialized countries [13]. Lower VE in low-income countries was also noted for other live oral vaccines and has been attributed to differences between countries of different SES in breastfeeding practices, micronutrient malnutrition, gut flora, RV epidemiology, and underlying medical conditions [59].
Because case-control study design is a widely used method to assess VE, only this type of study was included in our meta-analysis. In addition, to ensure high specificity of the outcome and avoid any potential misclassification, only studies with laboratory-confirmed RVGE diagnosis were considered when conducting the meta-analysis. The type of controls used was previously shown to impact VE to some extent, as observed in a review performed to assess VE of RV vaccines in Latin America [56]. In our study, we considered all types of controls, in line with other meta-analyses performed for RV vaccines [5,56,57] and selected the control groups according to a prespecified selection method. Only peer-reviewed studies that were checked for quality before inclusion in the meta-analysis were used; therefore, an adequate homogeneity between cases and controls in terms of exposure to the disease was assumed. Of the selected studies, when available, we preferentially included in the meta-analysis those reporting estimates based on hospital controls over community controls and studies using matched  controls to limit bias due to any potential confounders. To assess the choice of controls on the overall VE estimate, we also performed subgroup analyses by type of control. Similar to previous meta-analyses [56], we evidenced a relationship between VE and the type of controls used, as studies using hospital matched controls yielded higher VE of the 2-dose schedule of Rotarix than those with unmatched or neighborhood/ community controls, with values varying from 81% to 63%. Furthermore, RV test-negative control types are highly specific for non-RV diseases and thereby increase the robustness of the estimate. The overall VE of 2 Rotarix doses obtained using matched RV test-negative controls was 81%, but only 4 studies could be included in the analysis. Vaccine efficacy as assessed in clinical trials was previously shown to increase with the severity of the disease [13], a finding partially confirmed in real-world settings [57]. In our analysis, Rotarix VE was similar against disease of high and very high severity regardless of the number of doses but could not be estimated against mild and moderately severe disease due to the lack of data. VE against RVGE-related hospitalizations or ED visits varied between 64% and 89% with the type of RV strain, in line with efficacy results observed in clinical trials that demonstrated A rigorous quality control procedure has been applied for both the systematic review and the meta-analysis. Additional analyses investigating the publication bias have been performed. The study's main strength was the inclusion of robust case-control studies using laboratory confirmation of RV status. To minimize the risk of bias, adjusted results were preferentially included in the analysis. Nevertheless, although the case-control design allowed for stratified meta-analyses performed by type of control, this type of study also presents a certain risk of bias, as it relies on retrospectively assessed chart-based data and thorough documentation of vaccination history that might not have been correctly reported.
The study also has several limitations. Although covering various geographical settings, the data used in the meta-analysis originated from only 21 countries, with a small amount of data from low-income countries. Similarly, a relatively small number of studies reported strain-specific data. For certain subgroups, only a limited number of studies could be included in the analyses, or no meta-analyses could be performed due to insufficient data. The funnel plots suggest publication bias for some small studies with high OR (or low VE). However, as these studies had very low weights, their impact on the overall estimates is low.
Importantly, as previously discussed in relation with postlicensure studies [60], values derived from case-control studies can only provide information on the direct effectiveness of a vaccine and do not fully account for indirect protection afforded by herd effect, following the implementation of a vaccination program [60]. Given that a disease reduction of up to 75% was estimated in age groups that were not vaccine-eligible in countries with national RV immunization programs [61][62][63], VE values from our study are likely to underestimate the true global impact of Rotarix vaccination.

CONCLUSIONS
This meta-analysis provides strong evidence that vaccination with 2 doses of Rotarix has a substantial preventive effect against hospitalizations and ED visits due to RVGE, further confirming