Adaptive Sequential Multiple Hypotheses Testing for Concomitant Vaccine Safety Surveillance

When multiple vaccines are simultaneously recommended, a statistical method to monitor adverse events accounting for multiple combinations of vaccine exposure rather than each vaccine individually is preferred. We introduce an adaptive multiple hypotheses test method for rapidly detecting increased risks of adverse events from one or more combinations of simultaneous vaccine exposure, for example, exposed to influenza vaccination only; exposed to COVID and influenza vaccinations; exposed to COVID, influenza, and respiratory syncytial virus (RSV) vaccinations. We do this by extending the binomial maximized sequential probability ratio test (MaxSPRT) method to a multinomial probability model. With an exact analytical alpha spending approach, the computationally feasible limit of multiple exposures is likely limited to two vaccines. For more complex situations with three or more vaccines and multiple adverse event endpoints, we demonstrate a valid Monte Carlo approach. Illustrative examples and simulation studies were performed with the R Sequential package.