CDC and FDA bypassed PRR signal detection in VAERS during mRNA rollout despite known Bayesian flaw

The investigation documents that FDA medical officer Ana Szarfman and developer William DuMouchel warned in early 2021 that the empirical Bayesian method would miss signals when both Pfizer and Moderna vaccines produced similar elevations in an adverse event. Officials directed Szarfman to cease work on an updated proportional reporting ratio approach. CDC Director Walensky later confirmed PRR analyses were not performed until 2022, after the initial rollout phase. When finally run, the PRR method identified hundreds of signals including myocarditis that the Bayesian tool had missed.

This episode fits a pattern seen in prior vaccine safety systems where high-volume new products saturate passive reporting databases and blunt automated detection. Observational VAERS data alone cannot establish causation, yet the agencies publicly cited the absence of Bayesian alerts as reassurance. Internal records indicate the limitation was known before December 2020 but not disclosed to clinicians or the public during the period when policy decisions relied on rapid safety monitoring.

Future surveillance will require hybrid active-passive systems with pre-specified thresholds and independent audit. Without such changes, similar blind spots remain likely during rapid rollout of novel biologics. Congressional oversight and peer-reviewed re-analysis of the withheld 2021 PRR outputs are the minimum next steps needed to restore transparency.