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

Advocate A's framing rests on a hidden premise that deserves scrutiny: that an unpublished FDA manuscript creates a permanent "information vacuum" so corrosive it constitutes the "real scandal." That premise is the weakest link, and it collapses under three observations.

First, the underlying safety data on COVID and Shingrix vaccines is not vacated by a withdrawn manuscript. VAERS public datasets were actually expanded on May 8, 2025, to include secondary reports, and CDC WONDER continues to provide open access to de-identified records. FDA's Sentinel/BEST system covers roughly 138.7 million members and has produced 264 published scientific papers since founding, including ongoing 2026 peer-reviewed analyses of RSVpreF, COVID, and biologic safety in journals like JAMA Network Open and BMJ Open. Independent academic literature on Shingrix and mRNA COVID safety is voluminous (e.g., the Lancet 2022 VAERS/v-safe analysis of 298 million doses). The claim that suppression "removes evidence from the public record entirely" is empirically false — the evidence base is the data, not the agency-branded write-up.

Second, A conflates "withdrawal of an FDA-authored manuscript" with "suppression of findings." Per the same NYT/CNBC reporting A cites, HHS objected that authors drew "broad conclusions...not supported by the underlying data." Whether that rationale is correct or pretextual, internal disagreement over scope of conclusions is a routine feature of agency clearance — FDA scientists do not have an unconditional right to publish under the agency's name. Authors retain the option to revise, publish independently, or release data through FOIA. None of the reporting indicates the underlying datasets were destroyed or sequestered.

Third, A's "pattern" argument elides directionality problems. The ACIP reshuffle, mRNA funding cuts, and thimerosal vote are policy decisions one can criticize on the merits, but they are not the same act as halting a manuscript. Aggregating heterogeneous decisions into a single "capture" narrative is precisely the kind of motivated pattern-matching that A accuses HHS of. Kesselheim's quote in the NYT calling this "censorship" is one expert's normative judgment — a steel-manned opposing view (e.g., that agency clearance reasonably includes editorial control) is absent from A's argument.

My affirmative evidence: (1) Public access to vaccine safety data has been expanded, not contracted, in 2025–2026 (CDC May 8, 2025 announcement). (2) FDA's Sentinel/BEST infrastructure continues to publish peer-reviewed safety findings into 2026, including on COVID and adjuvanted vaccines. (3) The reassuring conclusion (rare serious adverse events) is already the documented consensus across CDC, FDA, and independent literature; one delayed manuscript does not reverse that record.

Normatively, I concede transparency norms favor publication. But calling delay "the real scandal" overstates harm and risks the very politicization A decries.

Advocate A's Round 2 contains a quiet but decisive concession: A withdraws the claim that suppression "removes evidence from the public record entirely" and reframes the harm as removal of the FDA's "institutional voice." That reframing matters because it shifts the debate from a falsifiable empirical claim (information vacuum) to a softer reputational claim (agency branding). The headline framing — "permanent information vacuum" — was the load-bearing premise; once it falls, what remains is a normative complaint about agency communication strategy, not a scandal of the magnitude the headline asserts.

A's three remaining moves do not rescue the original framing.

First, on "institutional voice": A argues lay readers Google headlines, not VAERS rows. True, but FDA's institutional voice on COVID and Shingrix safety is already extensive and ongoing. FDA's own COVID-19 Vaccine Safety Surveillance page publicly states serious adverse events occur in fewer than 1 in 200,000 vaccinated individuals, citing BEST/Sentinel active surveillance. CBER has monitored roughly 15 prespecified adverse events using Medicare claims through BEST. The reassuring agency-voiced conclusion is already on fda.gov and in CBER-authored peer-reviewed publications. One additional Medicare-cohort manuscript would marginally add to a documented body of agency communication, not break a silence.

Second, on "post-acceptance withdrawal is unprecedented": A cites Califf's recollection and Kesselheim/Steinbrook's normative judgments. These are expert opinions, not base-rate evidence. Neither A nor the underlying NYT reporting provides systematic data on how often FDA clearance disputes occur post-journal-acceptance, because such disputes are typically handled quietly through revision. Absence of recalled precedent is not proof of absence; it is proof that comparable cases, if they existed, were not newsworthy. The authors also retain the option to publish independently with disclaimers — a routine path for federal scientists testifying or writing in personal capacity (see Jake Scott's HSGAC testimony as one example of the personal-capacity channel).

Third, on the directional pattern: A's challenge — name a counter-example — confuses sample selection with directional bias. The universe of "HHS vaccine actions in 2025–2026" was itself selected by news coverage filtering for controversial moves; routine pro-vaccine agency activity (continued BEST surveillance publications, ongoing VAERS expansion announced May 2025, FDA approvals of updated COVID formulations) is invisible in that frame precisely because it is uncontroversial. The pattern A perceives is partly a function of which decisions reach the front page.

The normative core: A frames regulator credibility as "asymmetric" — easily lost. I accept that. But credibility is also lost when agencies are accused of "capture" every time internal review delays a manuscript. Calling delay "the real scandal" — bigger than the underlying findings — inflates a clearance dispute into an institutional indictment. That inflation is itself corrosive.

Steel-manning A: The strongest version of A's case is that regulator credibility is governed by perceived decision rules, not marginal information output, and HHS has now established on the record — through Nixon's confirmed rationale and post-acceptance withdrawal of peer-reviewed studies — a rule that reassuring findings can be withheld when leadership disapproves. Combined with a directional pattern of discretionary leadership actions (ACIP, mRNA, thimerosal, autism-page edit) running uniformly one way, that rule durably discounts every future FDA reassurance, which is harm of a different and deeper kind than any single manuscript's content. On this framing, 'scandal' properly tracks the institutional decision rule, and the rule has been revealed.

Why my position is still stronger: A's argument requires that a decision rule has actually been established, not merely alleged. But the evidentiary record supports a thinner reading — a contested clearance dispute over scope of conclusions, justified (rightly or wrongly) on methodological grounds, in an environment where FDA's institutional voice on COVID and Shingrix safety remains publicly documented through fda.gov surveillance pages, CBER publications, and expanded VAERS access (May 2025). A's 'directional pattern' rests on news-salience selection: discretionary leadership actions that confirm the narrative are visible; routine pro-vaccine clearances, approvals of updated formulations, and ongoing BEST/Sentinel publications are not coded as 'discretionary' precisely because they are uncontroversial. That is definitional, not empirical. Further, A's claim that post-acceptance withdrawal is unprecedented relies on Califf's recollection and two normative expert quotes — not base-rate data. Without systematic comparison across administrations, 'unprecedented' is an inference, not a measurement.

Most importantly, A's framing — calling delay 'the real scandal, bigger than the findings' — is itself a politicization risk. The reassuring findings are the public-health-relevant fact; elevating an internal clearance dispute above them inverts the informational hierarchy and feeds the same distrust dynamic critics claim to oppose. Transparency norms favor publication, and I have conceded the withdrawal is criticizable. But 'criticizable agency communication choice' and 'the real scandal' are different magnitudes. A has not bridged that gap with evidence proportional to the headline.

The durable harm A posits — discounted future reassurances — is plausible but speculative; it depends on public perception dynamics that have not been measured here. The reassuring evidence base, by contrast, is concretely documented and accessible.