The FDA Didn't Bury the Data. It Buried Its Own Voice.

In October, FDA scientists were told to withdraw two Covid vaccine safety studies that had already cleared peer review at medical journals. In February, top officials at the agency declined to sign off on submitting safety abstracts about Shingrix, the shingles vaccine, to a major drug-safety conference. The studies, built on millions of patient records and millions of dollars in public funding, found the same thing essentially every other large analysis has found: serious side effects from these vaccines are very rare. An HHS spokesperson confirmed the withdrawals on the record, saying the authors had drawn "broad conclusions" not supported by the data¹.

The immediate reaction in much of the press has been to treat this as an attack on vaccine science. I think that framing misses the actual scandal, and in a way that lets the agency off easy. The findings themselves are recoverable. The data sit in federal surveillance systems, the conclusions match a wall of independent literature, and the authors can, in principle, publish elsewhere. What is not recoverable is the institutional decision rule the FDA has now set on the public record: that reassuring findings can be pulled back after journals have accepted them, when leadership doesn't like the conclusions. That rule, not the manuscripts, is the durable damage.

Start with what actually happened. The Food and Drug Administration is the federal agency that licenses vaccines and is supposed to communicate their risks and benefits. Inside HHS (the Department of Health and Human Services, FDA's parent), leadership has shifted sharply toward vaccine skepticism over the past year. According to reporting in the New York Times² and the Washington Post³, the two Covid studies were already accepted for publication when senior officials ordered them withdrawn. Former FDA Commissioner Robert Califf told reporters he could not recall a comparable post-acceptance withdrawal. Harvard's Aaron Kesselheim and Public Citizen's Robert Steinbrook described it as a departure from past FDA practice. Former CDC vaccine official Fiona Havers, speaking to ABC News⁴, described a pattern of HHS leadership blocking studies whose results clash with their views.

The agency's defenders make a real point that I want to engage seriously before saying why I think it fails. The argument goes like this: the underlying safety data are not gone. VAERS⁵, the public adverse-event reporting system, was actually expanded in May 2025 to include more reports. The FDA's Sentinel Initiative⁶, which uses insurance claims from roughly 138 million people to monitor drug and vaccine safety, has produced hundreds of peer-reviewed papers and continues to publish through 2026. The FDA's own Covid-19 vaccine safety surveillance page⁷ still tells the public that serious adverse events are vanishingly rare. Internal disagreements over the scope of conclusions in agency-authored manuscripts are, in fact, routine. On this view, calling a clearance dispute "the real scandal" inflates a normal editorial fight into an institutional indictment, and that inflation is itself politicizing.

I take that argument seriously. I also think it dissolves on contact with three specific facts.

First, the timing is wrong for a routine clearance dispute. Pre-publication scope arguments between agency authors and reviewers happen all the time, and the standard remedy is revision. Authors rewrite the abstract, soften a sentence, add a limitations paragraph, and the paper goes out. What does not happen, in the recollection of a former commissioner who ran the agency twice, is that papers already through journal peer review get yanked. If HHS's objection were genuinely methodological, the fix was a revise-and-resubmit, not a withdrawal. The choice of withdrawal over revision is the part that reveals the operative criterion.

Second, the agency stated its rule on the record. The HHS spokesperson did not say the studies were being revised, or that there was a technical disagreement being worked through. He said the authors had drawn conclusions the data did not support, full stop, and the papers were being pulled. Peer reviewers at the journals that accepted the manuscripts had already tested exactly that question. So the agency's stated rationale is: peer review found the conclusions defensible, and we overrode peer review anyway. That is a decision rule, articulated by the agency itself. Whether it is applied once or a hundred times, the public now knows it exists.

Third, the directional pattern is real, and the selection-bias rebuttal doesn't dispatch it. Over the past year, HHS leadership has canceled roughly $500 million in mRNA vaccine development⁸, replaced all 17 sitting members of ACIP⁹, the CDC's outside vaccine advisory panel, voted to recommend removing thimerosal from flu vaccines against the longstanding evidence base, and edited a CDC webpage to imply the autism-vaccine question had been ignored. Now add manuscript withdrawal. The defense is that this list is cherry-picked from news coverage, while routine pro-vaccine activity (ongoing BEST surveillance, formulation approvals) goes unmentioned. That's true at the level of agency throughput. But the relevant universe is not "all FDA actions." It's discretionary, leadership-level communication choices. At that level, I cannot find a single 2025-2026 example running the other direction. A pattern with no counter-examples at the relevant decision tier is a pattern, not a perception artifact.

The deeper point is about how regulator credibility actually works. A regulator's authority is not the sum of its outputs. It is the public's estimate of its decision rule. When the FDA tells you a drug is safe, what you are buying is not the specific finding but a probability that the agency would have told you if it weren't. That probability is exactly what gets discounted when the agency demonstrates, on the record, that it will withhold reassuring conclusions when leadership disapproves. Every future FDA reassurance now carries an implicit asterisk: this is what they're willing to say given current leadership. The reassurances skeptics themselves may someday want, on a future vaccine, a future drug, a future food additive, are now worth less.

This is why I think "the data are still accessible" misses the mechanism. Of course they are. But the FDA's job is not to host datasets. Anyone can host a dataset. The agency's distinctive function is to author authoritative interpretations the public can rely on without becoming epidemiologists. The Medicare-cohort studies that were pulled were precisely that kind of authoritative interpretation, designed to answer the rare-serious-adverse-event question that drives most actual vaccine hesitancy. A motivated parent searching "Covid vaccine myocarditis" does not land on Sentinel row counts. They land on whatever the agency is willing to say in a journal. Removing the agency's voice from those journals while pointing at raw data downloads is not transparency. It is abdication dressed up as openness.

The strongest counter I have not yet addressed is that the durable credibility harm I'm describing is speculative. No one has measured a trust decline traceable specifically to these withdrawals, as opposed to broader HHS posture. Fair. I would only note that the absence of a measurement is not evidence of absence, and that the standard credibility literature on regulators (FDA, central banks, food-safety authorities) treats perceived independence as the load-bearing variable. We will get the measurement eventually. By the time we have it, the rule will have been operating for years.

What to watch next. Two indicators will tell us whether this is a one-off or a settled practice. The first is whether the original authors publish the studies independently, with or without FDA affiliation, and whether HHS attempts to block that. If the authors are quietly told they cannot publish in personal capacity either, the suppression reading hardens into fact. The second is the next time a Sentinel or BEST analysis returns a reassuring result on a vaccine HHS leadership dislikes: does it appear in a journal, or does it stall in clearance? One more withdrawal, and "clearance dispute" stops being a defensible framing for anyone.