When I first covered this story, the CDC had delayed publication of internal data suggesting that updated COVID boosters reduced emergency department visits and hospitalisations by around 50 to 55 percent. At the time, I raised a series of methodological questions about what that number actually meant. The data has now been released, and I want to revisit what we can and cannot conclude.
Let me be clear: I believe the report should have been published. Withholding data — particularly data that its authors considered favourable — undermines public trust regardless of which direction the findings point. Transparency is not optional in public health. If the data supports the intervention, publish it. If it raises concerns, publish it. The principle is the same.
But having now seen what was shared, I find myself more concerned than before — not because of what the report contains, but because of what it omits.
The Missing Variable That Changes Everything
The headline finding compares those who remained up to date with seasonal boosters against those who did not. On the surface, a 50 to 55 percent relative risk reduction sounds compelling.
But here is the critical gap: what was the vaccination status of those who were not up to date?
This is not a minor detail. The comparator group — the people classified as “not up to date” — could include individuals who received one or two primary doses years ago and stopped. It could include people who continued through six, seven, or eight boosters before discontinuing. It could, in theory, include people who were never vaccinated at all, though the framing strongly implies otherwise.
We are not told. And without that stratification, the finding is very difficult to interpret.
If someone who had eight prior doses and then stopped is being compared to someone who had nine, we are looking at a very narrow immunological question. If someone who had two doses four years ago is being grouped with someone who had eight doses and recently stopped after an adverse reaction, we are collapsing fundamentally different clinical histories into a single category.
The fact that this information was not provided — or not analysed — suggests something I find troubling. It looks less like an objective attempt to understand post-pandemic immune dynamics and more like an exercise designed to promote continued boosting. Why Cohort Definition Is Not Academic
A recent study published in JAMA Network Open looked at the method most commonly used to measure how well COVID vaccines work in the real world — a method known as the test-negative design. The researchers compared results from this method against the gold standard: randomised clinical trials.
What they found was telling. The method works well when the groups being compared are properly balanced and potential biases are accounted for. But when people are not randomly assigned — when they self-select into one group or another — the results can be misleading unless you understand and adjust for the reasons behind that self-selection.
That is exactly the problem here. The CDC’s booster comparison is not a randomised trial. The people who continued boosting and the people who stopped were not assigned to those groups. They chose. And the reasons for that choice are neither reported nor adjusted for.
The Question That Still Has No Answer
I keep returning to the same clinical observation. In real-world practice, patients rarely alter long-established behaviour without a reason.
If someone has taken eight or nine previous doses without incident and then declines the next update, something changed. Was it inconvenience? Shifting risk perception? Prior infection conferring a sense of natural protection? Or was it because they experienced an adverse response that made them reluctant to continue?
If the discontinuation group contains a disproportionate number of individuals who had already become unwell — who developed inflammatory symptoms, post-vaccination fatigue, or suspected immune-mediated reactions — then higher subsequent hospitalisation rates in that group may not reflect the absence of the latest booster at all.
They may reflect the fact that this was already a less healthy cohort at baseline.
Without stratification by prior dose number and reason for discontinuation, we cannot distinguish protection from selection.
Relative Risk Without Context
A 50 percent relative risk reduction sounds dramatic. But relative figures without baseline risk distort perception.
If the underlying hospitalisation risk in a healthy adult cohort is already very low, halving it produces only a modest absolute difference. A drop from 2 in 1,000 to 1 in 1,000 is statistically notable, but it is very different from a drop from 20 in 100 to 10 in 100.
Without the denominator — and without clear stratification by age, comorbidity, prior infection status, and cumulative dose number — the public is left with a number that sounds far more conclusive than it may be.
We need the full data, not the headline.
The Broader Signal Does Not Fit
What makes me particularly cautious is that broader post-pandemic population trends do not support the idea that the population as a whole is doing better.
Across older adult cohorts, we continue to see concerning shifts in inflammatory and cardiovascular diagnoses. I have previously highlighted a 237 percent increase in acute myocarditis in 60-to-74-year-olds — a group in which the vast majority are vaccinated. That signal is not consistent with a narrative of broad protection.
It suggests we are dealing with a far more complex interaction between prior immune priming, repeated antigen exposure, ongoing viral circulation, and host susceptibility — an interaction that a simple boosted-versus-unboosted comparison cannot capture.
A Plausible but Incomplete Mechanism
There is a plausible biological explanation for why continued boosting could reduce COVID-specific hospital admissions in a previously primed population.
If an updated booster increases short-term circulating antibody titres and modifies immune recall, those individuals would plausibly experience less severe symptomatic COVID in the months immediately following the dose. That part makes mechanistic sense.
But the report also appeared to reference non-COVID hospitalisation differences. If that signal is real, the interpretation becomes much more complicated. Immune protection against unrelated causes of hospitalisation is not a straightforward claim. At that point, one has to consider behavioural differences, healthcare-seeking patterns, and pre-existing illness in the discontinuation group as alternative explanations.
Cohort definition is not an academic detail. It is central to the truth of the finding.
What Objective Science Would Look Like
If the CDC genuinely wanted to advance understanding rather than promote an intervention, the path would be straightforward.
Publish the full vaccination history of both cohorts — not just “up to date” versus “not up to date,” but the actual number of prior doses in each group. Stratify by reason for discontinuation. Report absolute risk alongside relative risk. Provide age-stratified and comorbidity-stratified tables. And include the never-vaccinated as a reference group, rather than quietly dropping them from the analysis.
The fact that none of this appears to have been done — or if done, not shared — raises a question that goes beyond methodology. It raises the question of intent. Where This Leaves Us
I welcome the release of this data. Delayed transparency is still better than none. But publication alone does not equal rigour. And what has been published does not answer the question that matters most.
Until we know the full vaccination history of the comparator group — until we understand why previously compliant individuals stopped, and whether their subsequent outcomes reflect the absence of a booster or the presence of prior harm — the interpretation remains incomplete.
The most important question is still unanswered: what is the biology and behaviour driving the divergence between these two highly vaccinated cohorts?
That is where the real science begins. And it is precisely the science that this report chose not to do.
Vejon COVID-19 Review is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
