In October, US Food and Drug Administration (FDA) scientists were directed to withdraw two COVID-19 vaccine safety studies that had already been accepted for publication in peer-reviewed journals. In February, top officials refused to sign off on submitting two abstracts on the Shingrix vaccine, used to prevent shingles, to a major drug safety conference. Christina Jewett at the New York Times reported the scope of these decisions.
A spokesman for the Department of Health and Human Services (HHS) confirmed the withdrawals on the record, saying the studies were withdrawn because “the authors drew broad conclusions that were not supported by the underlying data” and that “the F.D.A. acted to protect the integrity of its scientific process and ensure that any work associated with the agency meets its high standards.”
The studies are public. Anyone can read them. The FDA’s own scientists, working with the active surveillance system Congress mandated after the withdrawal of Vioxx, an anti-inflammatory pain reliever, in 2004 after it was tied to increased heart attacks and strokes, produced findings consistent with every major post-market analysis of these vaccines published worldwide since 2023.
The work was buried for reasons that have nothing to do with the underlying data.
What the studies actually found
One of the COVID vaccine studies, involving US adults 65 and older, was withdrawn from the journal Drug Safety after acceptance. It analyzed more than 7 million Medicare beneficiaries who received the 2023-24 vaccine. The investigators evaluated 14 specific health outcomes, ranging from heart attacks and strokes to Guillain-Barré syndrome (GBS), an autoimmune condition that has been linked to certain vaccines.
They identified one statistically meaningful signal: a small elevation in anaphylaxis (a severe allergic reaction) following the Pfizer-BioNTech vaccine. After they adjusted for the possibility that some “anaphylaxis” billing codes did not represent true cases, the signal disappeared. The attributable risk, before that adjustment, was less than one excess case of anaphylaxis per million doses administered. The investigators concluded that no new safety signals had been identified.
The work was buried for reasons that have nothing to do with the underlying data.
The under-65 2023-24 COVID vaccine study, withdrawn from the journal Vaccine, examined 4.2 million recipients aged 6 months to 64 years across 17 health outcomes. The investigators flagged a rare elevation in the risk of febrile seizures (fever-related seizures) on the day of and the day after Moderna vaccination, in children aged 2 to 5 years.
The exact case count was so small the researchers had to mask it for patient privacy. Fewer than 11 cases. A separate small signal for myocarditis (inflammation of the heart muscle) in adolescents lost statistical significance once the analysis accounted for the seasonal pattern of myocarditis in the general population. The investigators concluded that no new safety concerns had been identified.
The Shingrix abstracts blocked from the February conference were routine updates to safety and effectiveness surveillance the FDA has been publishing transparently for years. To understand why their suppression matters, it helps to look at what the agency did with this same vaccine five years ago.
A precedent: How the agency handled a real signal in 2021
In November 2021, an FDA team led by Ravi Goud, MD, MPH, with collaborators at the Centers for Medicare & Medicaid Services (CMS) and the Centers for Disease Control and Prevention (CDC), published an analysis in JAMA Internal Medicine documenting an increased risk of GBS following Shingrix, the recombinant herpes zoster vaccine used to prevent shingles.
The CDC had detected a preliminary statistical signal in its Vaccine Safety Datalink, comparing GBS rates after Shingrix to GBS rates after the older shingles vaccine, Zostavax. Rather than dismiss the signal or sit on it, the agencies launched a full investigation in the Medicare claims database.
The methods were rigorous. The cohort analysis included roughly 850,000 Shingrix vaccinees and 1.8 million Zostavax vaccinees, all aged 65 or older. A self-controlled case series analysis added events from 2.1 million Shingrix recipients, comparing GBS incidence in the 42 days after each dose to incidence in days 43 through 183 in the same patients. To rule out billing artifacts, the team requested actual hospital records and had independent neurologists adjudicate each case against the standardized Brighton Collaboration definition.
The surveillance confirmed a real but exceedingly rare signal. The medical-record-confirmed analysis found a roughly fivefold increase in GBS risk in the 42 days after vaccination, with an absolute attributable risk of 5.17 excess GBS cases per million doses. An extended analysis using a larger pool of cases narrowed the estimate to roughly 3 excess cases per million doses. The case-series design also let the team isolate when the risk occurred. It was concentrated almost entirely after the first dose. After the second dose, the rate ratio dropped to 0.22, with no statistically significant elevation.
The FDA did not bury this finding. The agency added GBS to the Shingrix label as a known side effect. Goud and colleagues laid out the benefit-risk math in plain terms. Using their attributable risk and the trial efficacy of 97%, they estimated that a fully vaccinated population of 1 million people would experience about 6.26 excess cases of GBS, while preventing approximately 7,070 cases of shingles. The authors closed by recommending that clinicians be aware of the risk and that the benefit-risk balance remained clearly in favor of vaccination.
The Shingrix abstracts blocked in February were routine updates to this same surveillance, conducted by the same career scientists, using the same methodology. According to the Times reporting, one abstract concerned vaccine effectiveness and found it in line with the original clinical trial data. The other concerned safety and found an elevated but low risk of GBS already known to and acted on by the agency. There was no hidden danger in either abstract, only an updated calculation of numbers the FDA itself put on the Shingrix label five years ago.
How the methods hold up to standard objections
A persistent claim from critics of mRNA vaccines, including some who held senior FDA positions during the period of these withdrawals, is that observational vaccine safety studies suffer from “healthy vaccinee bias.” Vaccinated people, the argument goes, are systematically different from unvaccinated people, healthier overall, and more likely to seek preventive care, so comparisons between the two groups can make vaccines look safer than they truly are.
The argument has merit when applied to studies that compare vaccinated to unvaccinated populations. It does not apply to the design used in these studies. A self-controlled case series uses each person as his or her own control, comparing the rate of an outcome in a defined post-vaccination window to the rate of the same outcome in a later control window in the same person. Sex, chronic illness, lifestyle, health-seeking behavior—every personal characteristic that does not change over the study period—is held constant by design.
The investigators went further. They adjusted for the natural seasonal patterns of outcomes like stroke and heart attack, using month-by-month Medicare incidence from the prior year as a baseline, so that winter cardiovascular events following fall vaccination would not be misattributed to the vaccines. They also addressed billing-code noise. A physician might enter a code for “anaphylaxis” or “myocarditis” to justify ordering a test that ultimately rules out the diagnosis, and the code stays in the database. The team used positive predictive value imputation, which uses prior chart-review data to estimate what fraction of recorded codes represent true cases, and filtered out the expected false-positives accordingly.
The anaphylaxis signal in seniors held up after the seasonal adjustment. It disappeared after the misclassification adjustment. The signal was a statistical artifact, identified as such by the same rigorous methods the BEST Initiative was built to apply.
Blocked COVID studies simply verify prior data
A Nordic cohort study of 23.1 million residents of Denmark, Finland, Norway, and Sweden, published in JAMA Cardiology in 2022, characterized the myocarditis risk after the original mRNA COVID vaccines with high precision. The risk was real, concentrated in young males, and highest after the second dose, particularly with Moderna.
The investigators wrote that their data were “compatible with 4 to 7 excess events within 28 days per 100,000 vaccinees after a second dose of BNT162b2 [Pfizer-BioNTech], and 9 to 28 excess events within 28 days per 100,000 vaccinees after a second dose of mRNA-1273 [Moderna].” They concluded that the risk needed to be balanced against the benefits of these vaccines. The paper was not buried. It informed clinical practice.
In the multinational Global Vaccine Data Network analysis published in Vaccine in 2024, Faksová and colleagues pooled data on more than 99 million vaccinated individuals across study sites in eight countries on four continents. Their analysis confirmed pre-established safety signals for myocarditis, pericarditis, GBS, and cerebral venous sinus thrombosis (a rare type of blood clot in the brain). The pattern was internally consistent across the contributing countries.
In a Danish nationwide cohort of more than 1.5 million adults aged 65 or older, or otherwise at high risk, who received the LP.8.1-containing mRNA vaccines used in the 2025-26 season, Hviid et al evaluated 30 prespecified adverse events of special interest in the 28 days after vaccination. Receipt of the vaccine was not associated with a statistically significant increased rate of hospital contacts for any of them.
The suppressed FDA studies are not anomalies. They are corroborations.
Across the 2023-24, bivalent (two-strain), and JN.1-lineage COVID vaccine formulations, post-market surveillance has consistently characterized the same small set of rare, well-defined adverse events. Myocarditis, concentrated in young males and declining substantially with newer formulations and longer dosing intervals (in a systematic review my colleagues and I published in the New England Journal of Medicine in October, the diagnosis occurred at rates of 1.3 to 3.1 per 100,000 doses in male adolescents). Anaphylaxis at a few cases per million doses. Rare GBS signals associated with specific products that have been characterized, labeled, and managed.
That review synthesized 511 studies across the COVID-19, RSV, and influenza vaccines and concluded that “ongoing peer-reviewed evidence supports the safety and effectiveness of immunizations against Covid-19, RSV, and influenza during the 2025-2026 season.”
The suppressed FDA studies are not anomalies. They are corroborations. The agency’s own scientists, looking at the agency’s own data, found exactly what every other major surveillance system in the world has been finding. The decision to bury their work does not change the underlying evidence. It changes only the public’s access to it.
What the BEST Initiative was built to do
The Biologics Effectiveness and Safety (BEST) Initiative is the active surveillance arm of the FDA’s Center for Biologics Evaluation and Research (CBER). It exists because passive systems like the Vaccine Adverse Event Reporting System (VAERS), in which clinicians and patients voluntarily report events they suspect might be related to a vaccine, generate raw counts without denominators or control populations and cannot determine causation.
After the rofecoxib (Vioxx) withdrawal in 2004 made painfully clear how much harm a passive reporting system can miss, Congress passed the FDA Amendments Act of 2007, mandating active post-market surveillance covering at least 100 million Americans. The Sentinel Initiative followed in 2008. BEST, the biologics arm of Sentinel, launched in 2017.
BEST scans electronic health records on tens of millions of people in near-real time, applying rigorous epidemiologic designs that control for the biases that skeptics have long argued contaminate observational vaccine research. The system is built to publish what it finds, whether the finding is reassuring or whether it identifies a small signal that warrants regulatory action.
The previously published BEST surveillance studies of COVID vaccines illustrate what this transparency has looked like in practice. Wong and colleagues openly published four early-warning statistical signals in elderly Medicare patients receiving the original mRNA vaccines in 2023. Shoaibi and colleagues followed up later that year with refined analyses that resolved most of those signals as not consistently elevated.
Each of these papers had authors in common with the studies just suppressed. They followed the same methodology. They were published openly—because that is what the system was built to do.
When the system stops producing public outputs, the loss is not simply a few papers. It is the credibility of the surveillance enterprise itself, which is the only thing that allows post-market evidence to function as a legitimate public good rather than an exercise in agency self-protection.
The other side of the same misleading operation
Last August, HHS Secretary Robert F. Kennedy Jr. terminated $500 million in BARDA contracts for mRNA vaccine research and pointed to a 181-page bibliography as his scientific justification. I read that document carefully and wrote about it in STAT shortly afterward.
The lead compiler was a dentist, not an immunologist, virologist, or vaccine expert. The document originated as a research collection assembled for a book whose foreword was written by Sen Ron Johnson (R-Wis). It was not a systematic review. It was a curated reading list, organized around a conclusion the compilers had already reached.
The substance of the bibliography did not support the conclusion the secretary drew from it. Most of the cited papers were in vitro laboratory experiments documenting effects of the SARS-CoV-2 spike protein produced during natural infection, not during vaccination. Many used experimental routes of administration with no relevance to human vaccination, including direct injection into the brain and infusion into cerebrospinal fluid. Several of the cited papers explicitly concluded that the benefits of vaccination outweighed the risks of the rare neurologic events they described.
The Commonwealth Fund estimate that COVID-19 vaccines prevented approximately 3.2 million American deaths through 2022 was not cited. The Global Vaccine Data Network analysis of 99 million vaccinated individuals was not cited.
The bibliography manufactured the appearance of an evidence base for ending mRNA vaccine research. The actual literature, including the literature the bibliography itself purported to summarize, did not support that decision.
One half of the operation amplifies the science that fits a predetermined conclusion. The other half buries the science that does not.
What has happened at the FDA and CDC in the months since is the other side of the same operation. The Kennedy bibliography elevated misrepresented evidence to manufacture concern about mRNA vaccines. The FDA withdrawals and the CDC suppression last month of the VISION Network’s COVID vaccine effectiveness paper—showing modest protection for 2025-26 vaccines—have suppressed rigorous evidence that would provide reassurance about those same vaccines.
One half of the operation amplifies the science that fits a predetermined conclusion. The other half buries the science that does not. Both push in the same direction. The public ends up with less reliable information either way.
What we lose when safety findings don’t reach the public
The career scientists at the FDA and CDC and the contractors who run BEST built something genuinely remarkable over the past 15 years. Active surveillance covering tens of millions of Americans. Methods sensitive enough to detect adverse events occurring at rates of a few per million doses. Findings made public routinely, including findings of rare risks that warranted regulatory action and label changes.
The infrastructure exists because Congress decided after Vioxx that the public deserved it. It works only if the findings reach the public.
The studies the FDA tried to bury would have told the American public that the system Congress built is doing its job. The 2023-24 COVID vaccines, evaluated across more than 7 million Medicare beneficiaries and 4.2 million commercial enrollees, showed no new safety signals. The 2025-26 COVID vaccines, evaluated across the CDC’s VISION Network of more than 250 emergency departments and urgent care centers and 179 hospitals, reduced emergency department visits by half and hospitalizations by 55% in immunocompetent adults.
None of this would have surprised anyone fluent in the post-market vaccine literature. All of it would have been useful to the patients and clinicians making real decisions about real vaccines this fall.
The deal Congress struck after Vioxx was straightforward. The federal government would conduct active surveillance on the vaccines it licenses and recommends, and tell the public what it finds. When that deal holds, the system works. When it doesn’t, what looks like editorial discretion from the inside reads from the outside as the cover-up vaccine skeptics have spent two decades alleging.
The career scientists who actually did the rigorous work do not get to defend it, because their papers are not out. The vaccinated and the vaccine-confident lose access to evidence that supports their decisions. The skeptics get to point at the silence and claim vindication.
The studies suppressed here were not buried because they showed the vaccines were dangerous. They were buried because they showed the system was working.
Dr. Scott is a Clinical Associate Professor of Infectious Diseases at Stanford University School of Medicine and a co-author of “Updated Evidence for Covid-19, RSV, and Influenza Vaccines for 2025-2026“ in the New England Journal of Medicine.
The opinions voiced in CIDRAP Op-Ed pieces are the authors’ own and do not necessarily represent the official position of CIDRAP.
