Do Pfizer vaccines "kill" 2 people for every 1 saved? Evaluating Steve Kirsch claims.

Updated: 4 days ago

Summary of Key points:

  • Blatantly misleading headlines have circulated on social media stating that that "FDA experts conclude Pfizer vaccines kill 2 people for every 1 saved" following the FDA meeting on boosters, causing some alarm.

  • These comments were not made by FDA experts, but from a member of the public speaking for 3 minutes during the public comment section of the meeting.

  • Bold claims like this require strong scientific evidence, so I evaluated the evidence the person and his collaborators provided for their claims in the presentation slides and a supporting white paper.

  • In spite of many pages of writing and claims of over a dozen "independent analyses" verifying their results, their evidence falls far short of substantiating these dramatic conclusions, including a claim that vaccines have caused >250k excess deaths in the USA.

  • Their case is built primarily on a VAERs analysis, and one whose results are driven by assumptions of an artificially low background death rate and high underreporting rate with questionable justification. I demonstrate that with more realistic assumptions based on population death rates, the reported VAERs deaths are in line with background rate of deaths in the population, and does not provide evidence of excess deaths caused by vaccination, much less the 250k they claim.

  • They claim to independently validate these results with another dozen or so alternative "methods", but these consist of a series of largely unscientific arguments including anecdotal reports (some unverified), opinions elicited from select clinicians who believe the vaccines are inherently dangerous and/or treatment strategies like Ivermectin are the answer, public polls on whether people "know" more people who have died of COVID-19 or vaccines, and extrapolation of select data with very small sample sizes, often blatantly imputing assumptions that seem directly motivated by their chosen hypothesis.

  • They also cite an analysis of case fatality rate (CFR) that may be interesting for hypothesis generation, but cannot be used to draw any rigorous conclusions given its complete dependence on national case and death rates. Ignoring any of the many potential confounding factors, they conclude the only explanation for increasing CFR after vaccination is widespread fraudulent misclassification of vaccine deaths as COVID-19 deaths that, if true, comprises the greatest and most universal medical conspiracy in history, and leads them to infer estimates of 100k to 200k vaccine caused deaths in the USA.

  • Overall, the level of evidence provided is not even close to what is required to substantiate such bold and dramatic claims.

  • In an appendix, I evaluate even more extreme claims made in a recent paper published by the editor of the journal Toxicology Reports, in which they use VAERs data with even more extreme claims on background death and underreporting rates to suggest that between 225k and 1.4m USA residents have been killed by vaccines within 31 days of vaccination, predicting this is the tip of the iceberg with many more to come.

  • Looking at excess all cause death data for the USA, I show that that these claims are completely implausible, with vaccinations occurring during the lowest excess death periods of the pandemic, and the real spikes in excess deaths occurring during viral surges with high levels of confirmed cases and COVID-attributed deaths.


Following the meeting of the FDA Advisory Committee on Immunization Practices (ACIP) in which they discussed and made recommendations on Pfizer mRNA boosters, a claim circulated on social media that "FDA experts conclude Pfizer vaccines kill 2 people for every 1 saved." This created quite a stir among many circles, with many uncritically retweeting it and others discussing the claim as though it were a talking point of FDA officials, a major reason why boosters were only recommended for those >65yrs old, with the headline strongly implying this intepretation.

This was sparked by the following article published in, an alternative media site in the UK, here screen shot from a retweet by Robert Malone:

I won't comment on the reputation of as an accurate source of news, but obviously their headline was blatantly misleading, one of the worst I've seen during the pandemic.

Blatantly Misleading Headlines

The claim that the Pfizer vaccines were "killing at least 2 people for every 1 life they save" did not come from any FDA experts. They came from a 3 minute presentation during the public comment period of the meeting from Steve Kirsch. Here is a link to the video of his presentation, and here is a pdf of his presentation slides:

Steven Kirsch - VRBPAC meeting 919
Download PDF • 1.91MB

Kirsch is an entrepreneur and philanthropist, and introduced himself as Executive Director of the COVID-19 Early Treatment fund. As stated on his Wikipedia page, he personally donated $1m to start this organization focusing on repurposed drugs for COVID-19. He is an enthusiastic proponent of Ivermectin as an early treatment and alternative to the current vaccines, which he repeatedly claims are deadly dangerous, to the tune of repeated claims that the vaccines have "killed" ~250,000 in the USA.

He starts his presentation by claiming "I have no conflicts." Some might think donating $1m of your personal money to start an early treatment focused fund whose perspective explicitly opposes the current vaccination efforts could be construed as a conflict -- but let's set that aside and simply evaluate the scientific merits of his argument. That is what really matters.

Steve appeared with Robert Malone on Bret Weinstein's YouTube channel back in the summer claiming the spike protein was toxic and the current mRNA vaccines inherently dangerous. This YouTube video (now taken down by YouTube but available as an Apple Podcast) went viral, with millions of views within weeks. Given the visibility and importance of these claims, I wanted to evaluate his arguments based on their scientific merits, and so critiqued a white paper he wrote presenting the purported evidence for his claims in a previous blog post. I now think it equally important to evaluate his arguments leading to the "2 dead for 1 saved" claim given the attention it has received on social media and the alarm it has raised in many who do not know the details. Such bold claims require substantial evidence, and my goal here is to try to evaluate the evidence he presents.

To his credit, he transparently documents his calculations in his presentation slides so we can reproduce his results, and provides further details purportedly substantiating their claims in a supportive "white paper" co-authored by his collaborators Jessica Rose and Mathew Crawford, given here:

Kirsch deaths argument
Download PDF • 1.93MB

While I will focus primarily on the claims he made in his talk at the FDA, in this document I will also comment on some other details in this white paper that are relevant to supporting these claims, as well as some of the numerous "independent analyses" that he asserts validate his conclusions. I deal specifically with his vaccine death claims, both the ">2 dead for 1 saved" claims at the FDA meeting as well as his broader claim that the vaccines have directly caused >250k deaths in the USA.

His Claim

Slide 6 states his key claims:

He mentions 3 different analyses from which he concludes more were killed than saves by vaccines:

  1. An analysis of 6 month follow up data from the Pfizer clinical trial proposing a 5:1 ratio

  2. A VAERs-based analysis proposing a 2:1 ratio

  3. An anecdotal report of 4 deaths at a specific nursing home after boosters were given from which he proposes a 6:1 ratio.

His primary argument is from VAERs, so I will focus most of my attention there, with briefer comments at the end about these other two analyses as well as the many purported "independent methods" he claims validate his conclusions mentioned in the white paper.

His VAERS excess deaths calculations:

On slide 8 of his talk he lays out his VAERs argument.

He generated from VAERs a report of number of deaths after taking the Pfizer vaccine in the USA on 9/3/21, and found a total of 2,927 deaths. Of course, not all of these deaths were "caused" by the vaccines, nor are all events reported to VAERs, and he acknowledges this. Thus, to compute the number of "excess deaths" purportedly caused by vaccines, he subtracts off an estimate of background deaths (500) and multiplies by the assumed underreporting rate (URR) (41x).

From this he gets his estimate of ~100k excess deaths caused by the Pfizer vaccine. Given 217m delivered doses, he thus estimates the death rate is 458 deaths per million doses of vaccine. These are what he considers the number of people "killed" by the vaccine.

Next, to estimate how many lives were "saved" by the vaccine, he assumes that the vaccine prevents 90% of COVID-19 deaths that would have occurred sans vaccination, and with 91,868 COVID-19 deaths in the past 6 months, he estimates 82,681 lives saved. Since an estimated total of 360m doses have been given across all vaccines, this corresponds to ~230 lives saved per million vaccinated. Thus, taking 458/230, he gets an estimate of more than "2 killed by vaccines for every one saved".

Pretty clear presentation. However, his conclusions are completely dependent on the veracity of his assumptions of 500 background deaths and 41x underreporting rate, so it is important to evaluate whether these assumptions are reasonable.

Note that he is assuming all VAERs reports are valid. This not necessarily true given VAERs is an open reporting system for which entries are not fully verified. However, I will not focus on this issue here. as I think the assumed background death and underreporting rates are much more important primary issues, so for sake of argument I will assume all reported events are valid.

VAERs is not intended as a research data set, but as a tool to identify "safety signals", events that seem overrepresented in VAERs relative to background that should be further considered in more rigorous follow up studies involving medical data from health systems. It is a hypothesis generating tool to identify serious adverse events potentially caused by vaccines. The identification of these safety signals relies on estimation of background event and underreporting rates, so the following discussion also has general value beyond the critique of Kirsch's specific claims.

Evaluating the assumptions: Background Death Rate:

Kirsch assumes that only 500 of the deaths reported in VAERs are "background deaths" that would have occurred without vaccines. From whence does he get this assumption of 500?

In his white paper, Kirsch and colleagues state that they consider 500 background deaths "as normal for a year since the propensity to report is the same this year as in previous years." Since there were an average of 500 deaths/year reported in VAERs pre-pandemic, they assume that any more than that in 2021 must represent excess deaths caused by the vaccines. So, they are assuming that reporting practices to VAERs for SARS-CoV-2 vaccines during the pandemic are not any different than for other vaccines for other viruses pre-pandemic. Pages 6-9 of their white paper posted above presents their argument for why they think that is the case.

Given how the pandemic has upended the entire world for more than 18 months now, and given

  • the high level of concern about these specific vaccines using a new technology and being developed so quickly,

  • the increasing prominence of social media in recent years and the countless reports of concern and claims of harm from these vaccines both on social media and in traditional media, and

  • the greater awareness and discussion of the VAERs system in the USA public,

one might reasonably infer that reporting to VAERs right now might not be directly comparable to pre-pandemic levels for other vaccines and viruses. I will not go into the details of their argument here, but I welcome you to check pages 6-9 of their white paper to see why they think reporting has not changed. Suffice it to say, I am not convinced.

Irregardless of whether you agree with their argument or not, I wonder why they are computing background deaths like this, by depending completely on a strong controversial assumptions that VAERs reporting has not changed at all during the pandemic. This is completely unnecessary since there are much better ways to estimate background deaths rates from population level data.

As I have presented in a recent blog post, it is possible to look at all cause mortality in the USA in previous years by age groups, vaccination rates by age groups, and USA population by age groups, and pull these together to obtain estimates of background death rates for the relevant vaccinated population in the USA. This is a much stronger and less controversial approach. Here are my own estimates as documented in that blog post.

From these calculations, it is estimated that among the vaccinated population, we would expect ~210k to die in a given month, ~50k in a given week, and ~7k in a given day. This is the relevant background death rate in the population, and comprises the number that we would expect to die within a month, a week, or a day after vaccination (assuming the vaccine is given at a random time).

Since the FDA report was focused on Pfizer, I estimate number of these that are expected to be from Pfizer vaccinated. As of September 23, 2021, a total of 223,782,867 Pfizer doses were given out of 389,726,099 vaccine doses given in the USA with manufacturer identified, which is 57.4%. Taking 57.4% of the total deaths (implicitly assuming Pfizer was uniformly given among all age groups), we would expect the background rate of Pfizer vaccine deaths to be >120k in the month, >28k in the week, and >4k on the day of vaccination.

All of these are supposed to be reported to VAERs -- as health care workers are required by law to report all serious adverse events, including deaths, after vaccination whether thought to be related to vaccination or not, as is clear on the VAERs website:

Of course, we know that in spite of the requirement, not all events are reported to VAERs.

Thus, to assess whether there appears to be a "safety signal" for death evident in the VAERs data base, we must unavoidably consider the underreporting rate (URR) of VAERs, which can be represented as the ratio of true events to reported events.

Evaluate the Assumptions: Underreporting Rate (URR):

In his FDA presentation, Kirsch assumes an URR of 41x, meaning that for every event reported in VAERs, there are 40 more left unreported. He assumes this rate holds across various types of events, including deaths, and claims it is a conservatively low estimate.

The "white paper" linked above documents where he gets this number. His primary derivation comes from a paper looking at anaphylaxis rates in a hospital system in Boston and his secondary derivation is from a paper looking at myopericarditis rates in hospital systems in the Ottawa area.

A JAMA research letter was published March 8, 2021 presenting estimates of anaphylaxis rates for a cohort of Mass General Brigham (MGB) health care workers after their first shot of an mRNA vaccine (Pfizer or Moderna) between 12/16/2020 and 2/12/2021. They found a total of 16 cases of anaphylaxis among 64,900 health care workers, for a rate of 0.0000247, or 247 per million (95% confidence interval 55-438 per million).

Kirsch then ran a VAERs report for all 1st dose cases of anaphylaxis as of March 2021, and found 483 reported cases of anaphylaxis out of 97.5m vaccinated at that time, a rate of 0.000006, or 6 per million.

He then takes 247/6=41, estimating that VAERs has an underreporting rate of 41x (if he acknowledged the uncertainty of the MGB study estimates, the 95% confidence interval for URR would be 9x to 73x).

He reasons that anaphylaxis is a good choice for representing URR since it "is such an obvious association." In fact, he claims this should provide a lower bound for URR for deaths, since "deaths are reported even less frequently than anaphylaxis since deaths are not as proximate to the injection event." He argues that we need to look at a longer time horizon than just the first few days when looking at deaths. Thus, he argues the 41x is a conservative lower bound estimate for URR.

However, he does not mention that in November 2020, a paper was published in the journal Vaccines looking specifically at the question of estimating underreporting rates for VAERs for anaphylaxis (and Guillain Barre syndrome) for 7 different vaccines. They compared VAERs reporting rates to incident rates in the Vaccine Safety Database (VSD) network as a reference. VSD is organized by the CDC consisting of 9 healthcare organizations, shown to be representative of the USA population in many key demographic categories. This study found anaphylaxis had an URR between 1.3x to 8x, depending on the specific vaccine. It is not clear why the underreporting rate would be SO MUCH (5-30x) higher for the SARS-CoV-2 vaccine during the pandemic than it was for these other vaccines several years ago, so this raises some questions about his 41x number.

Kirsch also shows secondary calculation for URR based on a recent paper looking at incidence of myocarditis and pericarditis in June and July 2021 at an academic health care system in Ottawa within 1 month of mRNA vaccination. They found 32 with myocarditis and/or pericarditis out of a purported 32,379 doses of vaccine given in the Ottawa area, suggesting an incidence rate of 32/32,379, for an incidence rate of 1 in 1000 vaccinated, or 100 per 100k. His choice of this preprint was a curious one, given there were 2 other large studies published in peer reviewed journals months ago estimating myocarditis/periocarditis rates that yielded rates 25-35x lower.

He compares this with the 2888 VAERs reports of myopericarditis out of ~200m vaccinated, for a reporting rate of 1.444 per 100k vaccinated and an URR = 100/1.44 =69x. He reasons that "This makes total sense since myopericarditis isn't as serious as anaphylaxis, so the URR would be much higher than for anaphylaxis," so argues that 41x is a conservatively low estimate for URR, implying he thinks that more serious events would tend to have lower URR than less serious events (except for deaths that, although more serious, he apparently thinks should have an even higher URR?).

However, just this week, the authors have retracted this paper, acknowledging they calculated the denominator wrong, as detailed in this news report, which reported the actual number of vaccinations given in the Ottawa area was ~833k, not ~33k. The corrected rate of myocarditis/pericarditis would be 32/833k = 3.8 per 100k vaccinated, which is 1/26k, not 1/1k.

This estimate is nearly identical to two other published reports of myocarditis/pericarditis rates,

  1. A JAMA paper finding 37 cases among >2m vaccinated among 40 hospitals on the USA west coast giving a rate of 2.85 per 100k vaccinated, or ~1/35k

  2. A NEJM paper from a study of ~1m vaccinated persons from the Clalit Health Services covering >1/2 of the population of Israel, suggesting a rate of ~3.9 per 100k vaccinated, or 1/26k.

The latter study is based on a matched case-control study in which the vaccinated cases were matched 1:1 with unvaccinated controls based on key demographic factors, as well as with unvaccinated COVID-19 infected individuals. (BTW, this is the type of rigorous population-based safety study that we need more of). They found the difference in myocarditis/pericarditis rates between vaccinated and unvaccinated was 2.7 per 100k and the risk ratio of vaccinated:unvaccinated to be 3.24x, suggesting that the rates for vaccinated and unvaccinated were 3.9 per 100k and 1.2 per 100k respectively. (Incidentally, the rate of myocarditis/pericarditis was much higher in COVID-19 individuals, 18.3x higher than unvaccinated controls and ~5.6x higher than vaccinated individuals, another benefit of the matched case-control enabling valid comparisons of vaccinated, unvaccinated, and previously infected). So now, all 3 of these studies find rates of myocarditis/pericarditis between 2.85 and 3.9 per 100k. Comparing with the VAERs reporting frequency of 1.44 per 100k reported by Kirsch, this would suggest an URR between 2.0x and 2.7x, not 69x, and far short of the 41x he claims could conservatively be applied to all reported events, including deaths. Recall that Kirsch made the case that "it makes sense" for the URR for anaphylaxis to be even lower than for myocarditis/pericarditis given it is "more serious." Since death is even more serious than anaphylaxis, by his same logic one might conclude the death underreporting rate is less than for myocarditis/pericarditis.

It is difficult to see the plausibility of Kirsch's URR estimate of 41x given results from these 3 studies on myocarditis/pericarditis showing URR from 2.0x-2.7x, and the estimate of 41x from the selected anaphylaxis study contrasts with a study done in November 2020 for other vaccines showing UR of just 1.3x-8x, especially problematic for him based on his claim that reporting rates are similar now as pre-pandemic. To justify the high 41x level, he would need more compelling evidence beyond his current speculative statements on why the URR would be much higher, not lower, for deaths, and why the underreporting in 2021 would be so much higher than previous years.

This discussion highlights how difficult it is to determine the underreporting rate for VAERs, and why it is not considered a useful research data set for estimating population event rates, but only for flagging "safety signals", hypotheses to be tested using other more rigorous studies.

But if one is going to try to make an assessment of whether there are "safety signals" or excess deaths from vaccination based on VAERs data, as Kirsch is trying to do, it is important to consider a range of plausible choices for URR given its uncertainty and importance in the calculation.

Is there Any Real Evidence of Excess Deaths in VAERs?

Based on the discussion above, I computed the expected number of death reports in VAERs after Pfizer vaccination using the background death rates computed in the previous sections,

a range of potentially plausible estimates for URR (2x, 4x, 8x, 16x, 41x), and a given time frame after vaccination in which one might expect a VAERs report to be filed.

I will explain how to interpret this table.

If, in fact, 1/8 of deaths within 1 week of Pfizer vaccination are reported to VAERs (URR=8.0x, time frame=week), we would expect 3,538 deaths occurring within 1 week of vaccination to be reported to VAERs even if none were caused by vaccines -- this is the background rate, and any number over this would be excess deaths potentially caused by vaccines. If you assume 1/16 deaths are reported, you would expect 1,769 deaths occurring within 1 week of vaccination, and 7,666 deaths to be reported to VAERs even if none were caused by vaccines.

Even if you assume Kirsch's estimate of 1/41 reported to VAERS (which I think is far too high), then you would expect 2,991 deaths within 1 month of vaccination to be reported to VAERs, even if none were caused by vaccines. Given his argument that "deaths are not as proximate (sic) to the injection event" on page 2 of his white paper, it would be hard for him to argue 1 month is too long of a time frame to be relevant.

The number of background deaths expected to be reported to VAERs but not caused by vaccines appears to be far larger than 500, even if you agree with Kirsch's URR estimate of 41x (which I don't).

Considering the 2,927 deaths reported in VAERs after Pfizer vaccination, if the URR were 41x, we would expect this number of VAERs-reported deaths within 31 days of Pfizer vaccination even if none were caused by vaccine. If URR=16x, we'd expect this number to be reported within 12 days, for URR=8x, 6 days, URR=4x, 3 days, and if URR=2x, we'd expect more than this number to be reported within 2 days of Pfizer vaccination.

It is difficult when considering the background death rates of the population to conclude there are any excess vaccine-caused deaths based on VAERs reports, much less 100,000, irrespective of which of the URR is used. This is the conclusion that researchers at the FDA and CDC have repeatedly reached in their own analysis, and which this group claims is fraudulently mistaken.

Note that this does not mean that none of these deaths could have been caused by the vaccines. For example, we know there are individual cases of VITT after AstraZeneca/Johnson and Johnson vaccines that have directly led to deaths, so there is always potential for individual cases of minority harm when giving any medical vaccine or treatment to billions of people. It is important to detect and characterize these risks as was done for VITT. However, the number of reported counts of deaths in VAERs are not sufficient to conclude excess deaths caused by vaccines, and do not support the dramatic claims made by Kirsch and colleagues.

The crux of Kirsch's argument that VAERs counts support an inference that vaccines are killing more than they save does not hold up. His claim that >100k people in the USA have been killed by Pfizer vaccines, and ~250k killed by vaccines in general, seem preposterous.

Other Claims in the FDA report

While the VAERs analysis comprised his main point, Kirsch also had two other claims in his FDA report of evidence that the vaccines were killing more than they were saving:

Claims of 5 deaths per life saved from Pfizer Clinical Trial

Kirsch argues based on updated results from the Pfizer Phase 3 trial that the Pfizer vaccine has killed an estimated 5 for every one it saves. Here is the slide from his presentation:

Here is the medRXiv paper from July 2021 and the version published in NEJM on September 15, with other details of the study provided in the FDA report from December 2020.

This trial was double-blinded, meaning neither the participants nor clinicians nor researchers knew who was randomized to vaccine or who was randomized to placebo. At some point, the blinding was broken and placebo patients were allowed to cross over to the vaccine arm and receive the vaccine, as is typical in many studies, before which the final safety comparisons of placebo vs. vaccine were done. After that point, no legitimate causal comparison of vaccine and placebo groups is possible.

The post-blinding analysis found the following results for deaths:

  • Vaccinate group: 15 deaths among 20,998 (rate of 71.4 per 100k, 95% confidence interval 35.3-107.6 per 100k)

  • Placebo control group: 14 deaths among 21,096 (rate of 66.4 per 100k, 95% confidence 31.6-101 per 100k)

One might think this shows no excess deaths from vaccines, given the rates for the vaccinated and placebo groups were nearly identical, with almost completely overlapping confidence intervals. And unlike VAERs, a double-blind placebo controlled randomized trial provides a true valid control group with matching demographics that enables an unbiased estimate of the causal effect of vaccination on death. But Kirsch doesn't interpret the data that way. Kirsch adds in 3 more reported deaths after unblinding for the vaccinated group, but not for the placebo group, to obtain 18 vaccine caused deaths that he compares to the 14 placebo deaths in the blinded period. This is an invalid scientific comparison, and drives his conclusion of "5 vaccine caused deaths per 1 saved" that he makes. If he wanted to add in the 3 deaths to the vaccine side in an intent to treat analysis, he would have to list the 2 placebo randomized patients who died after crossover as well, or alternatively if we wanted to count all deaths after unblinding he would need to adjust for the fact that most placebo patients crossed over to account for the greater number of vaccinated than unvaccinated patients after crossover, and also adjust for potential confounders since the comparison is no longer randomized. But he doesn't do any of this, but illegitimately chooses to just add the 3 post unblinding deaths to the vaccine group, which produces his desired result.

It is hard to believe that he considers a study in which the last valid blinded vaccine vs. placebo comparison found nearly equal deaths in the vaccinated and placebo groups in a phase 3 study to be evidence that vaccines "killed 5 for every 1 saved" and worth mentioning in a presentation to the FDA. This argument he makes would not fly in any scientific venue.

Claims of 6 deaths for 1 saved in anecdotal report from nursing home. On slide 10 of his FDA report, Kirsch infers from reports of 4 deaths at a specific nursing home, Sunnycrest Nursing Home in Whitby Ontario, that the vaccines are killing 6 for every one they save.

Assuming a 3% infection fatality rate for the elderly (clearly should be MUCH higher for a nursing home) and that 30% get COVID, with a booster "lasting 6 months", he estimates boosting 136 residents would save 0.68 lives. Given 4 reported deaths (all of which he assumes are caused by vaccines), he estimates 4/0.68 = 6 deaths per life saved based on this report.

The reference he provides documenting this event is a tweet, which in the followup he claims "the data is from someone who works there. Obviously the source has to be confidential or the source would be fired." Yes, obviously. What strong evidence the vaccines are killing people -- please stop the vaccinations everywhere in the world on the basis of this story.

Regardless of whether this anecdotal report actually happened or not, which we really don't know, the fact that he thinks this is evidence worth mentioning in an FDA meeting says a lot about his sense of what qualifies as scientific evidence.

So, his dramatic claim that "more people are killed by vaccines than are saved" is solely driven by his VAERs analysis whose conclusions fall apart when carefully considering background death rates in the population. Kirsch would argue that he has "over 7 methods" independently validating his claims, as detailed in his white paper, as he has done in claiming vaccines have caused >250k deaths in the USA. I will briefly consider these arguments.

>250k vaccine caused deaths in USA?

Even more extreme than his claim of ~100k Pfizer caused vaccine deaths in the FDA report, Kirsch keeps repeating claims that the vaccines have caused >250k excess deaths, which would amount to >1/850 of the 213m who have received at least 1 dose of vaccine. He is claiming that the vaccine has caused the death of 1 in every 850 who have been vaccinated in the USA!

His argument is laid out in the white paper above and overviewed here.

Similar to his VAERs argument for Pfizer cases above, his primary method is to take VAERs deaths (7149) and subtract their "estimate" of background death (he uses 1000 instead of 500 to be "conservative") and then multiplies by the assumed URR 41x, to get (7149-1000)*41=252,109.

Of course, on the basis of the excess death calculations described above, 7149 is well within the range of expected background deaths after vaccination for a large range of proposed URR if we consider all vaccinations, not just Pfizer, as we can see:

If the URR=41x as Kirsch claims, we'd expect this number of deaths in 42 days post vaccination, if URR=16x, within 16 days, if URR=8x, within 8 days, if URR=4x. within 4 days, and if URR=2x, we'd expect this number of reported deaths within 2 days of vaccination even if none were, in fact, caused by vaccine.

He claims that he has two other "methods" corroborating his estimates of ~250k deaths.

  1. He cites a paper by Scott Mclachlin that evaluated 250 early vaccine death records in VAERs, admittedly including many of the most vulnerable members of society such as nursing home patients, and found that "vaccines could be ruled out" for just 14% of them. Interpreting this to mean that we can confidently include that 86% MUST have been actually caused by vaccines since they can't be ruled out (do you see how his statistical logic works?), he projects 7149 x 86% x 41% URR = 252, 073. He calls this one independent validation of his VAERs results.

  2. He claims that Dr. Peter Schirmacher, a German pathologist, stated to him that 30-40% of autopsies he performed within 2 weeks of vaccination were "clearly caused" by vaccine. No link to a paper, not detail on how many autopsies he did or how old they were, and no details on how he determined they were "clearly caused." But, hey, Steve says he is "one of the world's top pathologists", so it must be true even though we have not seen any other such reports or any primary document. Further, Kirsch assumes that the 30-40% number must be low because he believed "in making a potentially career ending revelation" that he was being "extremely conservative and only estimating what he was 100% certain of proving." So it must be higher. He then highlights a report from a doctor in Norway who was quoted as saying that 13 of 23 early deaths reported January 2021 (in nursing homes) might have been related to vaccines, saying “The reports might indicate that common side effects from mRNA vaccines, such as fever and nausea, may have led to deaths in some frail patients.” Kirsch interprets this say saying 100% of reported deaths were found to be caused by the vaccine. Thus, he assumes from these two reports that 60% seems like a "relatively conservative" number to assume for % of VAERs deaths caused by vaccines, since it is "between 30% and 100%, a little closer to 30%", from which he estimates 7149 x 60% x 41% URR = 175,865. This is scientific logic?

Therefore he claims he has 3 independent methods that confirm his estimates of 175k to 250k vaccine caused deaths.

To put this claim into perspective, the TOTAL SUM OF ALL COVID-19 DEATHS in the USA from January 1 to July 7, 2021 is less than 252,109, while most vaccination in the USA occurred between February and April.

Thinking about the magnitude of this claim, one might reasonably ask, " How could we have had such an enormous number of vaccine deaths and have it go unnoticed?" Where were the refrigerator trucks parked outside of hospitals during the vaccination push? The overfull morgues with vaccine deaths? The reports from ICU nurses overwhelmed from treating all of those dying from vaccine-induced complication? Their answer is conspiracy -- a massive international conspiracy unlike any the world has ever known to suppress the hordes of vaccine-induced deaths. That HCW are so scared for their jobs that none will speak up about the overwhelming number of vaccine-caused deaths (across all age groups), and the conspiracy is tight -- the information does not get out.

Mathew Crawford has presented some analyses of cases fatality rate (CFR = # deaths/# confirmed cases) that he claims is evidence that these hordes of vaccine deaths are being hidden by fraudulently misclassifying them as COVID-19 deaths. He performed two "excess CFR analyses" one in Europe and one using a mishmash of 23 countries from around the world that appeared to have a surge in COVID-19 deaths around the time the vaccination program started. In these analyses, he demonstrates an increase in CFR after vaccination, inferring that it must be caused by misclassified vaccine-induced deaths. I find his analysis an interesting idea for hypothesis generation and signal detection, but it cannot be used to draw any rigorous conclusions given its complete dependence on national case and death rates over time, ignoring all potential confounding factors, including vaccination status, age, previous infection status, time effects, what variant was hitting the country at that time, and how much population immunity the country had from previous infection. It is important to know the details of how this unusual set of 23 countries were selected, many of which still have very low population vaccination rates, since there might be some selection bias if these tend to follow the hypothesis better than the others left out.

Ignoring any potential confounding factors (Surge of aggressive Delta variant in countries with low levels of immune protection?), he concludes the only plausible explanation for this increase in CFR is widespread fraudulent misclassification of vaccine deaths as COVID-19 deaths, and that leads him to infer 72m to 180m or more vaccine-caused deaths in the USA. Considering a total of ~250m COVID-19 deaths between January 1 and July 1, 2021, this conclusion requires an extremely high proportion of all attributed COVID-19 deaths to in fact be deaths caused by vaccine. I find this conclusion to be overreaching and implausible, and such dramatic claims would require rigorous independent corroboration of some sort. And if driven by the fraudulent misclassification of vaccine deaths as COVID-19 deaths would comprise the worst medical conspiracy in history.

Kirsch states these two excess CFR analyses are "the most credible" among the 7 "methods" that he considered as a sanity check to assess whether their estimates "are reasonable." He frequently cites these analyses as independent corroboration for his initial results.

I will briefly summarize the rest of the 7 methods he mentions:

  1. His third "method" links to a "small island study" done by a member of their collaborative team Marc Girardot that estimated 171k vaccine deaths, with no details.

  2. His fourth "method" was that Norway reported 23 deaths in its first 40k vaccinations, figuring this rate of 1 in 1700 suggests 150k deaths in the USA (he doesn't mention anything about deaths being within 6 days of vaccination, and being in people old and frail at end of life, or in nursing homes).

  3. His 5th "method" is an analysis done estimating 174k vaccine caused deaths based on a poll asking people how many people they know who have died of COVID-19 or vaccines.

  4. His 6th "method" is asking several of his "doctor friends who are clued in that vaccines can cause death", and he found several of them consistently stated about 1 death in 1000 vaccinations and suggesting ~200k estimates.

  5. His 7th "method" is that 4 British Airways pilots died in 1 month after the vaccines rolled out. In spite of the fact that the vaccination status of each pilot was unknown, he assumes that 1 was "just bad luck" and the other 3 were excess deaths caused by vaccines, so with 3000 vaccinated pilots this 1/1000 death rate suggests ~200k deaths.


In conclusion, based primarily on a VAERs analysis of deaths after vaccination, Kirsch claims that he has evidence that the vaccines have killed >250k Americans, more than they have saved. I argue that this VAERs analysis makes questionable assumptions on background death rate and underreporting rate, and when population-based background death rates are used, the counts in VAERs fall within the range of expected number of VAERs death reports within days or weeks of vaccination even if none were in fact caused by vaccines. And this conclusion holds for a wide range of assumed underreporting rates, including the 41x value Kirsch assumes (that I think is too high).

Although he claims to independently validate these results with another dozen or so alternative "methods", most consist of a serious of largely unscientific arguments including anecdotal reports (some unverified), opinions elicited from select clinicians within their circle who believe vaccines are inherently dangerous, public polls on whether people "know" more people who have died of COVID-19 or vaccines, and extrapolations of select data sets with small sample sizes, often blatantly imputing assumptions that seem directly motivated by their chosen hypotheses. He also cites an excess CFR analyses that generates interesting hypotheses but but does not adjust for any relevant contextual factors so not sufficient do draw rigorous conclusions, certainly not of dramatic claim of ~100k-200k vaccine-induced deaths fraudulently misclassified as COVID-19 deaths.

Bold and dramatic claims like these require high levels of evidence to substantiate, and the presented analyses do not even come close to what is required to support them.

While the clinical trial results suggested serious adverse events and deaths had similar rates in placebo as vaccinated groups, it was not powered to detect rare events of incidence <1/10k or so. Thus, post approval monitoring of vaccine safety is crucial to identify any serious adverse events that might be caused by vaccines, even if very rare. The passive and active monitoring systems designed by the FDA and CDC should be actively used in these goals, and countries with centralized records should perform population studies comparing serious adverse event rates between vaccinated individuals and unvaccinated individuals matched on key demographic and clinical factors to adjust for systematic differences between vaccinated and unvaccinated cohorts, just as Israeli researchers have done in this NEJM paper. While international efforts have already uncovered various rare but potentially dangerous risks, anaphylaxis and myocarditis/pericarditis for mRNA vaccines and VITT or GBS for viral vector vaccines, one could argue more should be done to rigorously check population-level data to identify and characterize any other risks not yet discovered or acknowledged. The companies should be held accountable. I strongly agree.

But over the top, dramatic unsubstantiated claims like "vaccines are killing more than they are saving" does not help.

Following is an Excel spreadsheet containing the data I used to compute the background death analysis explained in the previous blog post. I encourage others to download, evaluate, and adapt/update with their own assumptions if they would like.

Estimating background death rate relevant for Pfizer vaccinated Sept21
Download XLSX • 19KB

Appendix: Evaluating a "peer reviewed" paper providing purported support for these claims

After writing this evaluation, Kirsch pointed out to me a paper published in Toxicology Reports included an analysis purporting to estimate causal vaccine deaths in the US claiming between 227,792 and 1,381,429 have died "from" the vaccine in the USA within 31 days of inoculation.

Here is the paper. Note that the senior author of the paper is the lead editor of the journal, and the editor who handled the peer review of this paper. This is unusual.

Like Kirsch's VAERs analysis above, their conclusions are completely driven by the assumptions they make on the background death rates and underreporting rate. However, their assumptions as detailed in his appendix are even more extreme than Kirsch's.

Implicitly assuming reporting rates to VAERs are constant as a function of the number of days since inoculation, the authors posit that the background rate of reported VAERs deaths not caused by vaccine can be estimated from the typical daily VAERs death reports >31 days after inoculation, and then assume that any reported deaths in the first 31 days above this number must have been caused by vaccine.

They subsequently double down on this assumption by using it to justify very high underreporting rates between ~125x and ~500x from which they extrapolate these numbers up to his enormous projected numbers. If events are more likely to be reported closer to vaccination, e.g. a death occurring the day after vaccination much more likely to be reported to VAERs than one occurring 2 months later, then this methodology will produce extremely biased estimates of purported vaccine-caused deaths.

It is especially strange they make this assumption given that elsewhere in the paper they acknowledge that deaths right after inoculation are far more likely to be reported than those much later in making a different point (see page 1675:)

Furthermore, they emphasize that these 225k to 1.4m deaths they concluded were caused by vaccines within 31 days of inoculation are only the tip of the iceberg, claiming many more deaths will happen months and years after inoculation. In fact, most of the paper focuses on these deaths they speculate will happen much later, laying out their concerns, but not providing any evidence that these things are actually occurring.

To my eyes, the paper read more like a commentary, filled with oft-repeated talking points from people who think that COVID-19 is not dangerous but that the SARS-CoV-2 vaccines are deadly so, than a research paper.

Also, like Kirsch, they never assess the plausibility of their assertions. If 225k-1.4m deaths were caused by vaccines within 31 days of inoculation, were is the evidence of this and how would it go unnoticed? Here are the excess deaths in the USA from ourworldindata

The majority of USA vaccinated residents were vaccinated between March 1 and July 31, a time during which excess death counts lower than any time since the beginning of the pandemic. How could this be if their claims were true? Tracking the excess deaths over time, we see they have spiked precisely at the times of high levels of confirmed COVID-19 cases and COVID-attributed deaths, including the April 2020 spikes and summer 2020 pre-vaccination spikes, and the winter 2020 spike that started pre-vaccination but was ongoing when the public vaccination commenced. Could the winter 2020 spike be caused by misattributed vaccine deaths as they might claim? Well, this spike in excess deaths started increasing precipitously in November 2020 and reached its peak in early January, when vaccinations had just started, and then sharply declined in January and February as vaccinations of seniors and other vulnerable members of society were ramping up. And then this lull of the lowest weekly excess death numbers of the pandemic kicked in from March through July, the time period during which most vaccinated USA residents received their inoculations, before increasing again in late July and early August as the Delta surge kicked in (also a time with very low daily inoculation rates).

Where would all these hundreds of thousands or millions of purported vaccine-induced deaths be hiding?

To further illuminate this point consider that @hmatejx plotted excess deaths vs. vaccinations and covid cases over time for 100 countries in the world. Here is the USA.

It is clear that that the excess deaths have tracked closely with confirmed case counts, so in spite of imperfections in SARS-CoV-2 testing and reporting, and in COVID-19 death attribution, the order of magnitude seems right.

It is also clear that it is not possible that the vaccines have caused a substantial number of deaths, and highlights the cartoonish nature of Kirsch and Kostoff, et al.'s projections, and their clear implausibility.

7,600 views55 comments