Over Five Million Excess COVID Vax Deaths Worldwide: Steve Kirsch

Steve Kirsch's claim is blunt. Averaging across the people who took COVID vaccines, he estimates about one death per thousand vaccinated persons over several years; not per dose, and not confined to the first fortnight after a shot. If roughly 5.5 billion people were vaccinated, that would imply more than five million excess deaths worldwide, or something like a one-to-two per cent lift in global annual mortality. He argues that a signal of that size is easy to miss in crude national totals and easy to dismiss if every source is examined in isolation. The point of his article is that many independent, imperfect sources point in the same direction.

The method he likes best is not an anecdote. It is a matched comparison of Pfizer recipients with Moderna recipients, two vaccinated groups, so the usual healthy-vaccinee selection bias is much reduced. In the Levi–Ladapo analysis he cites, non-COVID all-cause mortality was substantially higher after Pfizer than after Moderna. If one treats Moderna as the safer product and weights the American brand mix, the implied lift in vaccinated mortality is large enough, after allowing for the fact that vaccinated people were healthier to begin with, to land near one death per thousand. The attraction of the design is obvious: it does not require an unvaccinated control group whose health, behaviour and medical contact differ systematically from people who queued for a shot. If the result is real, official silence about replication in other states is part of the case rather than a rebuttal.

He then stacks smaller, messier observations that would be worthless one by one and, in his view, become harder to ignore together. A sole practitioner who knew his patients reported one vaccine-attributed death in nine hundred. A Canadian family doctor with electronic records described a practice death rate that sat flat through 2020 and then doubled for four years, with excess deaths concentrated among the vaccinated and among causes, sudden events, aggressive cancers, neurological collapse, that other clinicians have also linked to the shots. Kirsch treats that not as proof of a 29-per-thousand kill rate, which he regards as an outlier, but as a Poisson problem: if the true rate were one in ten thousand, finding a practice with thirty-two attributed deaths in eleven hundred patients is vanishingly unlikely. The same logic is applied to an FDNY cluster of six active-duty deaths in three months after rollout, to three unexpected deaths in a small rugby side that had had none in a decade, and to a well-networked man who lost fifteen friends after the campaign began, several on the day of vaccination. Each story can be waved away. The claim is that waving all of them away requires believing that many independent clocks jumped at once for unrelated reasons.

VAERS is used the way critics of the system have always used it when the count is high rather than low. Tens of thousands of death reports in 2021–22, minus a modest baseline, multiplied by an under-reporting factor Kirsch puts near twenty-six, produces a national death total well above one per thousand among those vaccinated in the United States. V-safe is turned the other way: if millions enrolled and only a tiny fraction of expected background deaths were flagged, the system is too incomplete to reassure. A wedding guest list split roughly evenly between vaccinated and unvaccinated, followed eight months later, is offered as a miniature trial with seven deaths on one side and none on the other. Family surveys and practitioner tallies from dozens of clinics scatter around and above his central estimate; he concedes that people who saw deaths were more likely to answer, then argues that the raw count of reported family deaths is still compatible with one per thousand once the reach of his readership is taken into account. Ed Dowd's range of one to three per thousand is cited as an independent order-of-magnitude check.

Defending the argument as far as it will go means taking these structural points seriously. First, all-cause mortality is the right endpoint if the concern is net harm rather than a labelled myocarditis line in a trial. Second, comparing product against product, or deaths inside a closed occupational group before and after a simultaneous rollout, is a rational way to dodge healthy-vaccinee bias. Third, passive systems and official narratives have a long record of under-counting rare or delayed harms; insisting that only a pre-registered randomised trial with ten-year follow-up can count is a way of never counting. Fourth, a one-to-two per cent rise in annual deaths is small enough that excess-mortality models can absorb it into ageing, delayed care, Long COVID, or residual pandemic effects unless someone looks specifically at timing after doses and at differences between brands. Fifth, clinicians who watch the same list of patients for twenty-five years are not "anecdotes" in the same sense as a stranger's Facebook post; they are small cohorts with a known denominator.

The argument still has to survive its weakest joints. Many of the practitioner and family numbers are selected, unverified, or volunteered to a writer already known for this thesis. Under-reporting factors for VAERS are disputed by orders of magnitude. A wedding or a rugby club is not a randomised trial. Attribution of a particular cancer or prion disease to a vaccine months or years later is a clinical judgement, not a laboratory isolation of cause. Large observational studies and the original randomised trials did not report all-cause mortality anything like one in a thousand, and critics will say Kirsch is assembling a prosecutor's brief rather than testing a hypothesis that could fail. Those objections do not make the stacked coincidences disappear. They mean the one-per-thousand figure is an estimate from heterogeneous, often hostile-to-official-sources data, not a settled actuarial fact.

What Kirsch is really asking the reader to accept is narrower than the headline. It is that a signal of that size is consistent with several designs that are less contaminated by healthy-vaccinee bias than the usual vaccinated-versus-unvaccinated comparisons; that official surveillance is too blunt and too politically constrained to rule the signal out; and that the polite explanation, that millions of sudden, clustered, post-rollout deaths in young and occupationally fit groups are all something else, has to do more work than it has so far been asked to do. Whether that is enough to move the estimate from "possible" to "probable" is the argument. The essay's force is that if even a fraction of the methods are roughly right, the global death count is not a rounding error, it is a vaccine disaster.

https://kirschsubstack.com/p/how-many-people-did-the-covid-vaccine