Bhattacharya accused of blocking CDC study — but flawed methods escape scrutiny
A row over a “blocked” CDC study has centred on censorship rather than the study’s flawed methods or the journal it was headed for.
Jay Bhattacharya has faced a wave of criticism after reports he stopped publication of a CDC study said to show the “benefits” of COVID vaccination while overseeing the agency.
The paper had been slated for the CDC’s in-house journal, Morbidity and Mortality Weekly Report (MMWR).
The reaction from public health figures and legacy media was immediate and largely one-sided, focusing on suppression rather than the evidence itself.
Now the study has been released.
Published exclusively on Inside Medicine, it reports a 50–55% reduction in COVID-related urgent care visits and hospitalisations — a figure already being used to reinforce claims about vaccine effectiveness.
But the central question has been largely sidestepped — is the study itself reliable?
Critics have not been deterred by the weak study design. If anything, they have doubled down and defended it.
What actually happened
On April 22, 2026, The Washington Post reported that a CDC study had been “blocked from being published.” Journalist Lena H. Sun framed the episode as a dispute over suppression rather than scientific standards, triggering immediate backlash.
Paediatrician Paul Offit said the decision reflected an effort to “suppress studies showing vaccines are safe and effective.”
Former CDC adviser Fiona Havers described it as “an escalation of this administration’s undermining of CDC science,” warning about interference in scientific work.
Those reactions assumed the study was valid — but now that it is public, its methods raise serious doubts.
The problem with the test-negative design
The study relies on what is known as a test-negative design — a method widely used during the pandemic to generate rapid estimates of vaccine effectiveness.
At a basic level, it works like this.
Patients who present to the hospital with respiratory symptoms are tested.
Those who test positive for COVID are counted as ‘cases,’ and those who test negative for COVID are treated as ‘controls.’ The controls are likely to have other respiratory illnesses (although flu and RSV are excluded from both groups).
The basic idea is that since the COVID vaccine does not protect against other respiratory viruses, the proportion of vaccinated among the controls will approximate their proportion in the general population.
It seems straightforward, but in practice it is unlikely to hold up.
This is because if vaccination against COVID affects susceptibility to those illnesses, or if vaccinated individuals tend to test more (or less) frequently, the comparison becomes distorted.
That problem is compounded by prior infection, differences in health-seeking behaviours, and fundamental differences in the health status of vaccinated and unvaccinated people.
Put simply, a test-negative design cannot tell you what parents and doctors actually care about — the overall health impact on people, nor does it have the power to detect serious outcomes such as death or long-term harm.
ACIP warning was ignored
These concerns are not new.
At a September 2025 meeting of the Advisory Committee on Immunisation Practices, MIT professor Retsef Levi outlined the weaknesses of the test-negative approach in detail.
He described the method as “prone to many biases” and said it relies on an assumption that cannot be verified in practice — that the tested groups reflect the broader vaccinated and unvaccinated populations.
Levi pointed to existing knowledge that the design “cannot be used for studying the mortality effects of vaccines” and is “problematic” for analysing hospitalisation outcomes.
At best, he said, it provides an indication of effectiveness under highly controlled conditions “that are often far away from reality,” and highlighted alternative approaches for more reliable assessment of vaccine efficacy.
Finally, Levi criticised the CDC’s working definition of ‘COVID hospitalisation,’ which relies on a positive PCR test and overlooks the fact that many of these cases are incidental — with patients admitted for other serious conditions, not COVID.
However, Levi’s warnings were ignored.
Three months later, the CDC published a paediatric study using the same weak method (see story).
It compared COVID-positive children with a mixed group of other respiratory illnesses, did not adequately account for prior infection, and relied on small sample sizes that made the results highly sensitive to misclassification.
The decision not to publish
The CDC study at the centre of the current dispute was reviewed by Bhattacharya in his role as acting director.
According to a spokesperson from the Department of Health and Human Services, the manuscript was “not accepted for publication” due to “concerns about the methodological approach.”
Declining to publish a paper because the method is weak and cannot support the conclusion isn’t censorship, it’s basic editorial judgment.
Inside Medicine author Jeremy Faust defended the study’s design, arguing that the same methods have appeared in top-tier journals, including The New England Journal of Medicine.
But just because a method is repeatedly used, it does not make it reliable. Junk science is still junk science, and there is plenty of it polluting the medical literature.
The debate has been cast as censorship versus transparency. But the question is simpler — does the method actually support the conclusions being drawn?
In this case, the answer is no.
Defending the study while brushing aside limitations of this magnitude shifts the focus away from the actual evidence and onto process.
“This is how it’s always been done” is not a valid scientific argument.
MMWR needs an overhaul
Estimates of vaccine effectiveness are only as strong as the methods used to generate them.
When a study design cannot measure overall health outcomes, cannot account for key sources of bias, and cannot assess serious endpoints, its conclusions need to be treated with caution — as Bhattacharya appears to have done.
MMWR is often presented as a leading scientific journal, but it operates without external peer review and has increasingly functioned as a policy-aligned outlet for the CDC.
During the pandemic, it played a central role in advancing mask mandates, school closures, and other interventions, often on the basis of studies using weak methods.
Bhattacharya has indicated that reforms are being considered, including moves toward external peer review.
But this is not about one paper. It reflects a system that repeatedly turns weak studies into strong claims.
MMWR needs a major overhaul.
Until then, the cycle will persist, and confidence in the evidence will continue to erode.
"When the debate is lost, slander becomes the tool of the loser."
— Socrates
[arguing that the same methods have appeared in top-tier journals, including The New England Journal of Medicine.] Well, that's a pretty low bar, since Richard Horton, ex editor-in-chief of The Lancet said '....much of the scientific literature, perhaps half, may simply be untrue. Afflicted by studies with small sample sizes, tiny effects, invalid exploratory analyses,...', and John Ioannidis’s famous 2005 paper "Why Most Published Research Findings Are False"