The effect of disease transmission on time-aggregated treatment efficacy estimates: a critical analysis of factors influencing the RTS,S and R21 malaria vaccine phase 3 trials

Abstract

Comparing the efficacy of preventive interventions against infectious diseases, such as vaccines, across different field clinical trials or between subpopulations within the same trial, is common practice. In the case of malaria, WHO has approved two biosimilar subunit vaccines in the past 3 years, both targeting the circumsporozoite protein of the Plasmodium falciparum parasite for mass vaccination. In paediatric phase 3 clinical trials in Africa, the R21 and RTS,S vaccines showed efficacies of 72% (95% CI 69–76) and 55% (51–59) against multiple episodes of clinical malaria in the first year of follow-up, respectively. Notably, R21 exhibited higher efficacy in seasonal transmission areas, whereas RTS,S showed substantial variation in efficacy across the 11 African trial sites, with no clear explanation for this heterogeneity. These efficacy estimates are used to inform public health policies and generate new research hypotheses. However, the fact that efficacy results from clinical trials reflect more than the individual biological protection provided by a treatment, and are also influenced by the intensity and distribution of disease transmission during the follow-up period, is often overlooked. In this Personal View, we review all non-biological factors that can affect efficacy estimates in clinical trials, and particularly focus on one factor that has received little attention despite its importance and ease of identification: the interaction of waning vaccine protection with changes in transmission intensity over time. When efficacy varies over time, typically in the form of waning protection as is the case for R21 and RTS,S efficacy, variations in disease transmission, such as those due to seasonality, outbreak spread, or age-related susceptibility, can cause some periods of the follow-up to have a stronger contribution to the overall estimate than others. This interaction results in real differences in the level of disease prevention achieved, which in turn affects all commonly reported time-aggregated efficacy estimates. Using published results from R21 and RTS,S trials, we show this effect and provide a series of counterfactual predictions, illustrating how vaccine efficacy might differ, by between 10% and 20% in some cases, under alternative vaccination dates. We also discuss how this effect might confound efforts to identify determinants of protective efficacy and offer recommendations to address it in the analysis and reporting of trial results.