Modeling time-varying efficacy of RTS,S/AS01 malaria vaccine across trial settings

Correctly evaluating how vaccine efficacy (VE) changes over time is important for understanding long-term protection and guiding vaccination strategies. Using the 18-month pre-booster data from the multi-center RTS,S/AS01 malaria vaccine trial, we examined how different approaches of modeling time-varying VE affect the interpretation of vaccine protection. Four commonly used linear-in-parameter functions and a four-parameter monotonic function were evaluated and compared with semi-parametric spline benchmarks. Across study sites, the four-parameter function was favored when a sufficient number of clinical malaria episodes were observed. Different VE specifications led to substantially different estimates of vaccine impact under a seasonal delivery strategy. Across settings with a sufficient number of episodes, the four-parameter model recovered an inverse-sigmoid pattern of RTS,S/AS01 protection over 18 months, with rapid early decline followed by stabilization at 20-40%. These findings show that modeling choices influence interpretation of vaccine protection and support the four-parameter approach as a parsimonious, biologically-interpretable parametric form for characterizing waning VE in the RTS,S/AS01 setting, with broader applicability across other vaccine platforms requiring further evaluation.