FDA Knew of Better Method to Detect COVID Vaccine Safety Signals — But Refused to Use It

Key Takeaways

• FDA stuck with outdated EB tool despite evidence of superiority.
• Improved method uncovered 25 COVID‑vaccine safety signals missed previously.
• Senior officials ordered “cease and desist” to prevent anti‑vaccine backlash.
• Emails show attempts to limit FOIA disclosures of safety data.
• Peer‑reviewed study validated new tool but FDA did not adopt it.

Pulse Analysis

The Vaccine Adverse Event Reporting System (VAERS) is a cornerstone of U.S. post‑marketing surveillance, relying on statistical algorithms to flag unexpected safety patterns. Historically, the FDA has employed an empirical Bayesian (EB) data‑mining approach known as the Multi‑item Gamma Poisson Shrinker (MGPS). While useful for broad signal detection, MGPS can mask rare events when multiple vaccines are reported simultaneously, a limitation that became starkly apparent during the rapid rollout of COVID‑19 immunizations.

In 2020, FDA medical officer Ana Szarfman and Oracle statistician William DuMouchel introduced an upgraded algorithm—Regression‑Adjusted Gamma Poisson Shrinker (RGPS)—designed to correct for masking effects. Their peer‑reviewed analysis demonstrated that RGPS uncovered roughly 25 statistically significant adverse‑event signals, such as sudden cardiac death and Bell’s palsy, that the older MGPS missed. Despite these findings, senior FDA officials instructed Szarfman to halt dissemination of the new analyses, citing concerns that the results could “sic anti‑vaccination rhetoric.” Internal emails reveal a deliberate choice to retain the legacy tool, even as the agency faced mounting FOIA requests for transparent safety data.

The episode underscores a broader tension between rapid public‑health communication and rigorous pharmacovigilance. By sidelining a more sensitive detection method, the FDA may have postponed critical safety alerts, eroding confidence in vaccine oversight. The incident also highlights the need for independent review mechanisms and clearer statutory guidance on data‑mining standards. As new vaccine platforms emerge, adopting state‑of‑the‑art analytics will be essential to safeguard public health and maintain trust in regulatory institutions.