The FDA will move to recall a product if it has been determined to be relatively unsafe for use by the general public. In 2024, the FDA recalled just over 2,000 products:
[click to enlarge]
About 12% of the recalls that FDA has historically issued have been for drugs, and another 25% of the recalls are for biologics such as immunologic products and vaccines. This report will dissect a recall made back in 2000 for the ingredient, phenylpropanolamine (PPA) found in various cough/cold and OTC diet products.
PPA had been suspected of increasing the susceptibility to hemorrhagic strokes, where there is a bleed in the brain instead of the other type of stroke, the ischemic stroke, where there is a blockage/stoppage of blood flow in the brain. In either case, you can refer to the event as a cerebrovascular accident (medical term for “stroke”).
FDA got tipped off that PPA might increase risk of stroke when a team at Yale did a retrospective case-control study on people with hemorrhagic stroke, along with matched controls. Such a study looks backwards in time to find out the share of all “cases” who had been exposed to a substance vs. the share of non-cases exposed.
While FDA prefers prospective studies (either cohorts or clinical trials) to establish causality, the slide presentation given by the Yale team to the FDA committee members argued against that:
So this FDA recall would have to rely on a case-control study, and on some other factors, to establish enough suspected causality to reach actionable levels for a recall. After the presentation by the Yale group, 90% of the FDA committee voted to remove the “General Recognized as Safe” (GRAS) status that PPA had enjoyed for decades:
The FDA acted on the Committee vote by issuing a Public Health Advisory for PPA a few weeks later:
But what level of evidence was the FDA sitting on when it admitted that it had collected an actionable level of evidence for issuing a recall?
Disproportionality: Proportional Reporting Ratio
The share of all spontaneous adverse event reports (AERs) collected specifically for cerebrovascular accidents (CVA) — out of all the AERs involving PPA — was 17.5 times higher than the share of all AERs for CVA involving all other products:
But with regard to reports involving COVID shots in 2021, both the April PRR for Pulmonary Embolism and the May PRR for myocarditis were over 25.0! But the FDA requires more than a data-mining detection of such safety signals, so let us find out what else is required to reach an actionable level and pull a product from the market.
Case Reviews
Case reviews show a tight temporal association, with half of the CVA events occuring within 4 hours of a dose, and the other half occuring after 4 hours had gone by. 82% of the CVA events were within 3 days of starting to take the product, though some of the people would have been taking more than one dose per day for those 3 days.
Population Exposure
With 6 billion doses sold in 1999, the level of exposure to this drug was pretty high. This means that even small increases in risk can magnify through the population, resulting in a sizeable number of CVA events attributed to PPA.
The Yale group had found an adjusted odds ratio (aOR) of 3.13 for women upon first dose, and 16.58 for women taking diet products containing PPA:
But to find the number of CVA events attributable to PPA, you have to incorporate the background incidence of CVA:
With 10,000 yearly strokes happening in the background, you can apply the OR to discover the number of strokes attributable to PPA (the “excess strokes” from it):
A yearly estimate spanning from 210 to 510 strokes attributable to PPA was enough to reach an actionable level and to go ahead and pull-the-plug on this product. But with COVID shots, even just the spontaneous reports of CVA tower far over this number, with 8000 reports coming in for those dosed from Dec 2020 to Dec 2021:
March 2021 alone had 1,300 reports of strokes after COVID shots, and the minimum multiplication factor for serious AERs in spontaneous reporting systems like VAERS is 5x — meaning that over 6,000 actual strokes had occurred after COVID shots during that month. There were 18 stroke reports for each million two-dose series of Moderna.
For J&J’s single-dose COVID shot, there were over 21 stroke reports per million doses.
There were 6 criteria for establishing causality in the FDA recall of PPA, which were stated as being largely met:
The evidence suggests that at least as much evidence exists for COVID shots as what existed for PPA.