LAUSR

Evaluation of COVID-19 vaccines has proceeded at unprecedented speed and required the quick development of statistical methods to emerging challenges in design and analysis. Amajor issue emerged shortly after the trials started enrolling. What to do with the placebo group after efficacy was confirmed? Although continued follow-up of the original arms was ideal to assess durability and performance against emerging strains, maintenance of the placebo arm with available efficacious vaccines was both ethically and practically untenable. A remarkable result was that the placebo controlled vaccine efficacy profile could be recovered long after the placebo group had been vaccinated. This result allows for the assessment of the durability of vaccine effect and can inform when booster vaccinations might become necessary. However, individual initiated unblinding, vaccination, and associated differential behavior can jeopardize the unbiased estimation of vaccine efficacy. In this thoughtful and elegant work Tsiatis and Davidian (TD) developmethods to fix the potential bias, motivated by and applied to the Moderna phase III clinical trial. The results are important and timely as multiple trials do and will grapple with this issue. In this comment, I provide some detail about how time varying vaccine efficacy is recovered to facilitate understanding and the potential for bias. I discuss possible extensions and provide some details about practical implementation. 2 THE BACKGROUND