The Department of Health and Human Services has confirmed that studies examining the safety of COVID-19 and shingles vaccines were withdrawn from publication or conference submission, an unusual dispute over federally conducted vaccine surveillance whose scientific details remain largely out of public view.
An HHS spokesperson said the authors had drawn broad conclusions that were not supported by the underlying data, according to ABC News and the Guardian, which published their accounts May 5. But neither the withdrawn COVID-19 manuscripts nor the Shingrix abstracts described in those reports has been publicly released. That leaves outside scientists unable to examine the protocols, statistical work, peer-review comments or the precise language that HHS considered too broad.
What HHS confirmed — and what remains unclear
The HHS statement confirms that studies were withdrawn and provides the agency’s rationale: protecting the integrity of the scientific process because the conclusions were not supported by the data. It does not identify the disputed analyses, say who made each publication decision, or specify whether the concern was about study design, statistical interpretation, presentation of results or final wording.
News accounts describe the action in somewhat different terms. ABC News reported that the Food and Drug Administration withdrew publication of research tracking COVID-19 and shingles vaccine safety. The Guardian reported that the FDA blocked publication or submission of several studies. Those accounts may reflect different stages of a publication process, but the available public statements do not resolve the internal mechanics for each paper or abstract.
The reported work matters because it involved post-market safety surveillance, the kind of research used to detect uncommon medical events after vaccines reach broad populations. Such analyses can inform regulators, clinicians and the public even when they find no new risk signal. Withholding or withdrawing completed work can also prevent other researchers from evaluating its strengths, limitations and interpretation.
Reported findings came from large health-record analyses
The Guardian, citing reporting that first appeared in the New York Times, said one withdrawn COVID-19 analysis examined records from about 7.5 million Medicare beneficiaries and assessed 14 health outcomes. The account said anaphylaxis, at roughly one case per million Pfizer-BioNTech vaccine recipients, was the only outcome identified as concerning in that analysis. Because the paper itself is unavailable, those figures and that characterization cannot be independently verified from the manuscript.
A second reported study included 4.2 million people ages 6 months to 64 years and found rare febrile-seizure and myocarditis cases, the Guardian reported. The news report did not provide enough information to establish how the researchers defined cases, selected comparison periods, handled prior infection or accounted for seasonal shifts in illness and health-care use. Details of the Shingrix research are thinner still: two abstracts reportedly were not submitted to a drug-safety conference after FDA officials did not sign off in time.
These appear to have been observational analyses of health records, not randomized clinical trials. Surveillance studies can measure whether diagnoses arise more often during a defined period after vaccination than expected under a study’s comparison method. They can identify associations and help estimate risks, but they do not independently prove that a vaccine caused an individual event.
Some observational designs compare people with themselves at different times, reducing bias from stable personal characteristics such as genetics or longstanding medical conditions. Their findings can still be affected by diagnostic coding, missed cases, timing choices, changes in disease circulation and testing many outcomes at once. Those are reasons scientists need access to methods and data summaries before judging whether a conclusion is justified.
Public FDA information already recognizes some rare risks
The publication dispute should not be read as evidence that COVID-19 vaccines have no known safety risks. In a June 2025 safety communication, the FDA required updated warnings about myocarditis and pericarditis in the labeling for the Moderna and Pfizer-BioNTech mRNA vaccines. The agency said information about those conditions had appeared in mRNA vaccine labeling since 2021.
For the 2023-24 formulation, FDA estimated an unadjusted incidence of myocarditis or pericarditis within one to seven days after vaccination at about eight cases per million doses among people ages 6 months through 64 years. The estimate was about 27 cases per million doses among males ages 12 through 24, the group with the highest observed rate. Those estimates cannot be directly compared with the reported unpublished studies because the populations, vaccine formulations, time periods, outcome definitions and analytic approaches may differ.
The FDA’s communication also described required post-approval follow-up research, illustrating why continued monitoring does not end with authorization or updated labeling. In a separate May 2026 statement on the 2026-27 COVID-19 vaccine formulation, the agency said it would continue to monitor vaccine safety and effectiveness.
Transparency questions remain
HHS has not publicly described which conclusions it regarded as unsupported. That distinction is consequential: a concern about a causal claim, for example, is different from a concern that a result was imprecise, insufficiently adjusted or overstated in a news release or abstract. The available accounts do not establish which issue, if any, applied to the withdrawn work.
Releasing the manuscripts, protocols or a detailed agency explanation would not settle every scientific disagreement. It would allow researchers to assess the choice of comparison groups, outcome definitions, risk windows and sensitivity analyses, and to determine whether the reported findings align with other vaccine-safety evidence. Until then, it is not possible to tell whether the disagreement concerns statistical methods, interpretation, wording or some combination of all three.