Quantifying the Impact of Natural Immunity on Rotavirus Vaccine Efficacy Estimates: A Clinical Trial in Dhaka, Bangladesh (PROVIDE) and a Simulation Study

Abstract

Background: The low efficacy of rotavirus vaccines in clinical trials performed in low-resource settings may be partially explained by acquired immunity from natural exposure, especially in settings with high disease incidence.

Methods: In a clinical trial of monovalent rotavirus vaccine in Bangladesh, we compared the original per-protocol efficacy estimate to efficacy derived from a recurrent events survival model in which children were considered naturally exposed and potentially immune after their first rotavirus diarrhea (RVD) episode. We then simulated trial cohorts to estimate the expected impact of prior exposure on efficacy estimates for varying rotavirus incidence rates and vaccine efficacies.

Results: Accounting for natural immunity increased the per-protocol vaccine efficacy estimate against severe RVD from 63.1% (95% confidence interval [CI], 33.0%-79.7%) to 70.2% (95% CI, 44.5%-84.0%) in the postvaccination period, and original year 2 efficacy was underestimated by 14%. The simulations demonstrated that this expected impact increases linearly with RVD incidence, will be greatest for vaccine efficacies near 50%, and can reach 20% in settings with high incidence and low efficacy.

Conclusions: High rotavirus incidence leads to predictably lower vaccine efficacy estimates due to the acquisition of natural immunity in unvaccinated children, and this phenomenon should be considered when comparing efficacy estimates across settings.

Clinical trials registration: NCT01375647.

Figures

Figure 1.

Crude Kaplan–Meier curves for time to first rotavirus diarrhea episode in the per-protocol population (n = 678) of the Performance of Rotavirus and Oral Polio Vaccines in Developing Countries (PROVIDE) study in Dhaka, Bangladesh. Vertical line indicates the beginning of the postvaccination period at 18 weeks. Abbreviation: RV1, monovalent rotavirus vaccine.

Figure 2.

Extended Kaplan–Meier curves [41] for time to first severe rotavirus diarrhea episode in the per-protocol population (n = 678) of the Performance of Rotavirus and Oral Polio Vaccines in Developing Countries (PROVIDE) study in Dhaka, Bangladesh, in which children become previously exposed at the time of their first rotavirus diarrhea episode of any severity. At the end of follow-up, 249 children were vaccinated and rotavirus-naive, 244 children were unvaccinated and rotavirus-naive, and 185 children from both the monovalent rotavirus vaccine (RV1) and no RV1 arms had been previously exposed (potential for natural immunity).

Figure 3.

Expected difference in severe rotavirus vaccine estimates from the expected efficacy in a rotavirus-naive population due to acquisition of natural immunity at varying severe rotavirus diarrhea incidence rates and vaccine efficacies in simulated trials. A, Expected difference at high (90%) and low (50%) vaccine efficacy for a range of severe rotavirus diarrhea incidence rates. B, Expected difference at high (15 cases/100 child-years) and low (5 cases/100 child-years) severe rotavirus incidence rates for a range of severe rotavirus vaccine efficacies. For each, the left graph corresponds to overall impact across year 1 and 2, the center graph corresponds to impact in year 1, and the right graph corresponds to impact in year 2.