Lessons Learned from the Bayesian Design and Analysis for the BNT162b2 COVID-19 Vaccine Phase 3 Trial

TL;DR

This paper reviews the Bayesian design and analysis of the BNT162b2 COVID-19 vaccine trial, highlighting common misinterpretations and proposing clearer reporting practices to improve clinical and regulatory understanding.

Contribution

It identifies reporting issues in Bayesian trial results and offers four recommendations to enhance clarity and interpretation in clinical research.

Findings

Bayesian credible intervals are often mislabeled as confidence intervals.

Proper presentation of Bayesian results facilitates better clinical decision-making.

Recommendations improve transparency and understanding of Bayesian analyses in trials.

Abstract

The phase III BNT162b2 mRNA COVID-19 vaccine trial is based on a Bayesian design and analysis, and the main evidence of vaccine efficacy is presented in Bayesian statistics. Confusion and mistakes are produced in the presentation of the Bayesian results. Some key statistics, such as Bayesian credible intervals, are mislabeled and stated as confidence intervals. Posterior probabilities of the vaccine efficacy are not reported as the main results. We illustrate the main differences in the reporting of Bayesian analysis results for a clinical trial and provide four recommendations. We argue that statistical evidence from a Bayesian trial, when presented properly, is easier to interpret and directly addresses the main clinical questions, thereby better supporting regulatory decision making. We also recommend using abbreviation "BI" to represent Bayesian credible intervals as a differentiation to "CI" which stands for confidence interval.