See No Evil, Hear No Evil, Speak No Evil
- Elisa Ballard
- 1 day ago
- 3 min read
How FDA's Vaccine Safety Algorithm May Have Hidden Serious Side Effects — And What Happened to the Scientist Who Tried to Fix
Imagine you're running a giant complaint box for car accidents. Every time someone reports a crash, you throw the note into the box. To spot patterns — like “this one model of car keeps having brake failures” — you use a computer program that compares new complaints against everything else in the box.
That’s basically how the U.S. government’s VAERS system (Vaccine Adverse Event Reporting System) works for vaccines. It collects voluntary reports of possible side effects after shots. Doctors, patients, and manufacturers file them, and safety experts look for “signals” — clusters that suggest a real problem worth investigating.
During the COVID-19 vaccine rollout starting in late 2020, millions of doses were given fast, and VAERS received an unprecedented flood of reports. The FDA used its standard statistical tool called MGPS (Multi-item Gamma Poisson Shrinker) to scan for danger signals.
Here’s the catch: MGPS has a built-in statistical quirk called “masking.” When a ton of similar reports pour in for multiple COVID vaccines at once (Pfizer, Moderna, and Johnson & Johnson), the program’s baseline comparison gets bloated. One vaccine’s problems can get hidden — or “masked” — by the sheer volume from the others. It’s like trying to hear one fire alarm when dozens are going off at the same time.
On March 26, 2021, a highly experienced FDA senior medical officer and data-mining expert, Dr. Ana Szarfman, decided to test a better tool. Working with Dr. William DuMouchel (the statistician who originally invented the MGPS method), she ran an upgraded version called RGPS (Regression-Adjusted Gamma Poisson Shrinker). This newer approach was designed specifically to adjust for masking and cut through the noise.
Her analysis was eye-opening. It found 49 examples of extreme masking. Once the adjustment was applied, more than 20 new statistically significant safety signals popped up for serious conditions that the standard MGPS method had missed. Examples included sudden cardiac death, Bell’s palsy, pulmonary infarction (blood clots in the lungs), acute myocardial infarction, and others tied to the COVID vaccines.
Instead of embracing the improved method or urgently investigating those signals, senior FDA officials reportedly shut it down. Internal documents show Dr. Szarfman was told to stop creating and sharing these reports. By September 2021, the FDA’s Center for Biologics Evaluation and Research Director, Dr. Peter Marks, called the work a “major distraction” that could “feed into anti-vaccination rhetoric.” Her access to run certain analyses was restricted.
Today, April 29, 2026, U.S. Senator Ron Johnson (R-WI) released a new interim report and held a hearing exposing these internal FDA decisions. The report, titled “Unmasked: How Biden Health Officials Purposely Turned a Blind Eye to COVID-19 Vaccine Safety Signals,” argues that top health officials knew about the masking problem, were shown a ready-made fix that revealed hidden risks, and chose to stick with the less-sensitive original algorithm while sidelining the scientist trying to make the system more accurate.
In plain English: The safety watchdog system was using a tool that, under the unique conditions of the COVID rollout, could miss real problems. A senior FDA expert tried to upgrade it — and was reportedly told to stand down. The result, according to the Senate report, is that millions of Americans may never have heard about certain potential risks that the data was actually flagging.
This doesn’t mean the vaccines caused every reported side effect (VAERS is a passive system that needs further study). Still, it does raise serious questions about transparency and whether the public got the fullest possible picture during one of the largest medical rollouts in history.
Link for further reading -
Link to report:
