Review: Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status

John P.A. Ioannidis, Stanford University, [email protected]

[A] Kampf and Fornerod examine age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status using data from England covering the period from April 2021 to May 2023. I have not been able to verify the raw numbers and I advise the authors to examine them again as there are often updates and revisions in these datasets. Also I understand that the study did not have a pre-registered protocol and the analyses are exploratory – if so, this should be stated. Some other aspects of the analytical design are not reported and adherence to STROBE guidelines may improve the manuscript.  It should also be explained why only England data were used and not other parts of the UK and/or other countries, where hopefully some similar data can be retrieved. Also why were the first 4 months of the vaccination campaign excluded from analyses.  

[B] Assuming the raw numbers are correct, the authors observe that the age-adjusted non-COVID-19 mortality rate was higher in COVID-19 vaccinated groups. For doses 1 and 2, the relative increase was highest in the elderly, while for the dose 4 the relative increase was highest in the young (18-39 years old) stratum. The authors then proceed to speculate about various explanations for these observations and many of them center on the possibility of death risk increase related to the COVID-19 vaccines. I find these interpretations very erratic, given that everything that they observe can be readily explained with much anticipated, commonplace selection biases, mostly of the “healthy non-vaccinee bias” type (but also some “healthy vaccine bias”) in these specific data.  I do worry that some people may read the discussion and treatment of the results by Kampf and Fornerod and be misled into thinking that COVID-19 vaccines are responsible for a sizeable part of the excess mortality seen in many countries.

[C] Specifically, all the mortality estimates presented in the paper use only age-adjustment, since only age grouping was available (unclear why gender was also not used, since it must have been readily retrievable). This leaves out all the comorbidities and socioeconomic and other risk factors that largely shape mortality risk in a general population. Even the age-adjustment may be not perfect, since it includes very crude age groups where the upper versus the lower end of each age bracket may differ several fold in mortality risk. Thus even within the same age bin, vaccinated and unvaccinated individuals may have different age distributions (even more so in the upper age stratum) and thus this crude age adjustment does not fully adjust even for age. This is obvious even in the benchmark group, the unvaccinated cohort, where the authors claim that the age-adjusted rates are stable over time, but in fact there are easily 2- and 3-fold differences across different time periods, meaning that the composition and risk factor profile of the unvaccinated group itself changes substantially over time. Besides bias due to simple selection/choice to vaccinate, potential survivorship biases and harvesting effects may all intermingle to create very complex patterns that cannot be dissected simply by the group-level information available to Kampf and Fornerod.  

[D] I wonder whether at least one could obtain information on nursing home status and stratify further accordingly. If so, some further adjustment/stratification for this may modestly improve the situation. Still, large proportions of the risk variance for death risk remain unaccounted and, rationally, one has to assume that any differences in non-COVID-19 deaths may be simply due to such unaccounted differences. Any other complex interpretation is quite implausible compared to this one based on plain Occam’s razor considerations.

[E] For example, it is probably not surprising that people who receive just a single dose are at the highest risk, especially if they are elderly. One may speculate: perhaps many of them who died were so sick that they could not even get a second dose or did not want to get one if they were considered to have terminal disease or if they had a poor tolerance of the first dose given their overall debilitation – impossible to tell exactly what happened without having granular data on individuals and their risk profiles. People with two doses may also have more comorbidities on average than the unvaccinated, especially among the elderly. The elderly who decided to be unvaccinated were probably mostly very healthy. The selection biases can be complex, nevertheless. E.g. based on Figure 2, for 2 doses and 3 doses, it is probable that the first months witnessed mostly a healthy vaccinee bias (perhaps due to eager healthy people who wanted to get quickly vaccinated) but then this shifted to mostly non-healthy vaccine bias. As for the 4th dose, young people who decided to take a 4th dose may have included most of those who had serious comorbidities and strongly wished to get some extra protection.  Again, one needs to have granular, individual level data to validate diverse speculations and make meaningful inferences.  

[F] In all, I see nothing that is not possible to explain with selection biases in the England data. Of course, selection biases of the healthy vaccinee or, conversely, healthy non-vaccinee, types may not be the same in different countries, settings, time periods, and health systems. For example, other countries like Israel, Austria and USA have documented mostly healthy vaccinee bias rather than healthy non-vaccinee bias [1,2,3]. However, attributing these results and patterns not to these commonplace biases but to harmful or beneficial vaccine effects is a dangerous fallacious stretch. Better vaccine pharmacovigilance is definitely needed to understand any potential vaccine harms, but strong mortality extrapolations based on such data cause confusion and probably only show bias against vaccines. Furthermore, the authors do not even restrict themselves to COVID-19 vaccines in the discussion, but also cite selectively data on some other vaccines.

[G] Some other issues also create major problems in these analyses. First, there can be misclassification between COVID-19 and non-COVID-19 deaths. The impact on the types of analyses presented here is unpredictable.

[H] Second, a more appropriate approach would be to use time-to-event analyses with vaccination doses as variables introduced at time t=21 (or whatever is considered appropriate for a time window, even this can be contentious) instead of the crude age-adjustment comparison used here.

[I]Third, the whole discussion about excess deaths and the effort of the authors to raise the possibility that vaccines may have accounted for so many excess deaths is unrealistic. I sympathize with the authors that estimates of excess deaths are speculative and that attributing them mostly, if not entirely, to COVID-19, is probably very wrong. Moreover, it is likely that these estimates are inflated and the true excess deaths are substantially lower [4]. Non-COVID-19 deaths due to disruption of health care (sometimes due to measures taken), increase in drug use and overdose, alcohol, violence, mental health, poor cancer care and more may be key contributors to some of the excess deaths not accounted by COVID-19 deaths. In poor countries and in marginalized populations, starvation and devastation in vulnerable populations may also have contributed. But claiming that vaccines killed large numbers of people is implausible. This does not exclude that some adverse effects from vaccines may have resulted in some deaths. However, the scale of numbers of vaccine-related deaths, based on what we know (and what even the authors cite in the discussion, if the respective papers are read more carefully) is probably many log-scales lower than the excess deaths documented in 2021-2023.

[J] In all, I worry that a reader of this paper may start suspecting that COVID-19 vaccines killed millions of people. It is far more likely that they saved several millions of people. Probably not as many as are sometimes touted – but still a major beneficial contribution. More accurate estimates will require much better, detailed individual-level data and careful approaches to the problem. Even with the best possible data, some uncertainty will remain, given the observational nature of these datasets. Moreover, it is important to have granular, rich data on individual characteristics to understand also the duration of life gained or life lost rather than the simplistic lives saved or killed by adverse events. Eventually, the truth about COVID-19 vaccine effects may be somewhere between the extremes presented by zealot enthusiasts and zealot skeptics. Extreme interpretations will need to be tamed in order to try to understand the risk-benefits in different age groups, time periods, and settings.

References

1. Chalupka A, Riedmann U, Richter L, Chakeri A, El-Khatib Z, Sprenger M, Theiler-Schwetz V, Trummer C, Willeit P, Schennach H, Benka B, Werber D, Høeg TB, Ioannidis JPA, Pilz S. Effectiveness of the First and Second Severe Acute Respiratory Syndrome Coronavirus 2 Vaccine Dose: A Nationwide Cohort Study from Austria on Hybrid Versus Natural Immunity. Open Forum Infectious Diseases, 11:ofae547, 2024. doi: 10.1093/ofid/ofae547.

2. Høeg  TB, Duriseti  R, Prasad  V. Potential “healthy vaccinee bias” in a study of BNT162b2 vaccine against Covid-19. New England Journal of Medicine,  389:284–5, 2023.

3. Xu S, Huang  R, Sy  LS, et al.  COVID-19 vaccination and non-COVID-19 mortality risk—seven integrated health care organizations, United States, December 14, 2020–July 31, 2021. Morbidity and Mortality Weekly Reports, 70:1520–4, 2021.

4. Ioannidis JPA, Zonta F, Levitt M. Flaws and uncertainties in pandemic global excess death calculations. European Journal of Clinical Investigation, 53:e14008, 2023. doi:10.1111/eci.14008.

Competing Interests

None.

Review: Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status

Martin Kulldorff, [email protected]

It is not surprising that we see excess mortality following the pandemic lockdowns. Since lockdowns disrupted regular medical care, we should expect to see worse health outcomes down the road, including excess mortality. For example, cancer detection and treatment were interrupted during the lockdowns, and that is expected to have long-term negative impact on cancer survival. If not, we are wasting medical resources on useless cancer care.

Other things may also contribute to the post-pandemic excess mortality. Adverse reactions to the Covid vaccines have been suggested as one possible cause. Modelling studies have come to widely varying conclusions about the total number of lives saved or killed by the vaccines, ranging from 19.8 million saved [1] to 17 million killed [2]. There is enormous uncertainty in such estimates, due to their use of aggregated ecological data, questionable modelling assumptions and unreliable parameter estimates. Trying to make such worldwide estimates may be futile when we don’t even have proper randomized clinical trials (RCTs) to tell us whether the vaccines reduced mortality, as they were only designed to evaluate the efficacy at reducing symptomatic disease. The closest we have are pooled RCT data, with 95% confidence intervals showing that the adenovirus vector vaccine reduced overall mortality by somewhere between 30% and 81%, while the mRNA vaccines caused somewhere between a 37% decrease and a 71% increase in mortality [3]. This was for the mostly young population enrolled in the RCTs. There are no RCT estimates for older adults, who were at the highest risk of Covid mortality.

With the lack of RCT data, and the inherent weakness of aggregated data, we must use individual observational data. There is evidence that the vaccines reduce Covid mortality among older people [4]. To get a complete picture, it is also important to look at the potential effect on non-Covid mortality using individual data. Even if the vaccines saved millions of people from dying from Covid, it is important to know if they caused serious adverse reactions or non-Covid deaths. This is what Kampf and Fornerod have explored using public data from England (Kampf and Fornerod, 2025). Importantly, they have done this for different age groups and different number of vaccine doses.

The most interesting and striking observation is that during the spring and summer of 2021, the age-adjusted non-Covid mortality rate was more than twice as high among those with only one vaccine dose compared to the non-vaccinated, while it was less for those who had two doses compared to the non-vaccinated. What could explain this curious finding?

One possible explanation is that the first dose of the vaccine causes a fatal adverse reaction in some individuals. If they survive the first dose though, it could mean that they are resistant to the fatal adverse reaction. It’s like a situation with a bunch of swimmers and non-swimmers. Those that cannot swim will drown the first time you throw them in the water. If you throw the survivors in the water a second time, they will all survive since they have already proven that they can swim. Those that survive the first dose may also be generally healthier so that they have lower non-Covid mortality than the unvaccinated.

A second possible explanation is that it is not due to the vaccine but to underlying health differences in the different vaccine groups, since this is not a randomized study. If healthy people are more likely to be vaccinated, there is healthy vaccinee bias, which could explain that those with two doses have lower non-Covid mortality than the unvaccinated. On the other hand, if frail patients in for example nursing homes are more likely to be vaccinated, the bias could explain that those with one dose have higher non-Covid mortality than the unvaccinated. What’s curious with these data is that to explain both results, there must be both a health vaccinee bias and a non-healthy vaccinee bias.

Primarily descriptive in nature, the results by Kampf and Fornerod need to be taken very seriously and further investigated. While there is no statistical significance testing or confidence intervals, the results are too consistent over time and age groups to be generated by chance.

To examine these data further, it is necessary to consider different vaccine brands. With different dose recommendations, as well as timing between doses, biases could operate in different directions for the different Covid vaccines. It is also necessary to look at the mortality data by cause of death. If the excess non-Covid mortality is due to one or more of the vaccines, it is likely to be manifest itself for only a few disease outcomes. On the other hand, if the results are due to healthy or unhealthy vaccination bias, one may expect the difference to be manifested across a wide variety of unrelated diseases.

My understanding is that such further analyses cannot be done using the data that is publicly available. In their important study, Kampf and Fornerod have done what is possible to do with the public data. Further analyses require permission from the data holders.

References

1. Watson OJ, Barnsley G, Toor J, Hogan AB, Winskill P, Ghani AC. Global impact of the first year of COVID-19 vaccination: a mathematical modelling study. Lancet Infect Diseases, 22:1293-1302, 2022. doi: 10.1016/S1473-3099(22)00320-6.

2. Rancourt D, Baudin M, Hickey J, Mercier J. COVID-19 vaccine-associated mortality in the Southern Hemisphere. Correlation Report, September 2023.

3. Benn CS, Schaltz-Buchholzer F, Nielsen S, Netea MG, Aaby P. Randomized clinical trials of COVID-19 vaccines: Do adenovirus-vector vaccines have beneficial non-specific effects? iScience, 26:106733, 2023. doi: 10.1016/j.isci.2023.106733.

4. Nordström P, Ballin M, Nordström A. Risk of infection, hospitalisation, and death up to 9 months after a second dose of COVID-19 vaccine: a retrospective, total population cohort study in Sweden. Lancet, 399:814-823, 2022. doi: 10.1016/S0140-6736(22)00089-7.

5. Kampf G, Fornerod M, Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status, Journal of the Academy of Public Health, 1, 2025.

Author Rejoinder: Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status

Günter Kampf, Maarten Fornerod

Many thanks for the helpful comments that have helped to improve the quality of the manuscript.

Response to Peer Review by John P.A. Ioannidis

[A] The data were uploaded from the official homepage on May 29, 2024, which is approximately one year after the latest time point described in the table. That is why we think it is rather unlikely that one year later data were updated. That the analyses were exploratory without a pre-registered protocol is now stated in the method section. Data from England were used because the dataset only included data from England. Other sources or data from other parts of the UK were not available at that time. The first 4 months of the UK vaccination were not included by the Office for National Statistics. That is why it was not possible to include in the analyses.

[B] The aspect of a possible healthy vaccine bias has been added to the discussion.

[C] That is correct. Other risk factors are not described, neither in the manuscript nor in the underlying datasets, so that an adjustment to the comorbidities and socioeconomic and other risk factors could not be done by us. This is certainly a relevant limitation of the analyses. At the same time, however, the COVID-19 age-adjusted mortality did not show major deviations according to the vaccination status although the same limitations apply such as lack of adjustment to the comorbidities and socioeconomic and other risk factors. If we assume that the higher non-COVID-19 age-adjusted mortality among the vaccinated is primarily explained by comorbidities and socioeconomic and other risk factors, would it not be plausible to assume that a similar finding is observed for the COVID-19 age-adjusted mortality? But this was not the case. Nevertheless, we included a paragraph in the discussion to explain the major and relevant limitations as suggested by the respected reviewer.

[D] The official data do not provide additional information on the nursing home status. That is why it was not possible to stratify the data accordingly.

[E] These are certainly interesting aspects of relevance. We have added them in the discussion to expand the discussion to other possible explanations for the non-COVID-19 morality rates.

[F] Many thanks for the comment. We have added a short paragraph to the discussion. The last two paragraphs of the discussion about other vaccines have been deleted.

[G] Misclassification is always possible, that is correct. We have added this aspect to the discussion.

[H] Unfortunately, this type of analysis was not possible based on the available data.

[I] We have added this relevant aspect briefly at the end of the discussion including the suggested reference.

[J] We agree. It is not the intention to speculate that COVID-19 vaccines could have killed millions of people. However, in many countries, it seems to be unmentionable in public that the vaccines may have contributed to even a little to the overall excess mortality observed in many countries in recent years. No possible cause should be excluded per se when looking for explanations for excess mortality.

Response to Peer Review by Martin Kulldorff

We agree that there is enormous uncertainty in estimates of post-pandemic excess mortality. The possible explanations for our findings are now included in the discussion, as suggested by both reviewers. We agree that causality remains unknown, as now clearly described at the end of the discussion. We also agree that further stratified analyses are warranted.

Reviewer Closure: Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status

John P.A. Ioannidis, Stanford University, [email protected]

Kampf and Fornerod have made several revisions based on my previous comments and I thank them for these changes. However, they have been unable to address any of the fundamental concerns that I had. They have not performed any new, substantive analyses, since they are limited by the lack of availability of relevant data. Despite a most welcome addition of limitations in the Discussion, their analysis continues to be a very crude probe with extremely limited ability to perform any proper adjustments. Its credibility therefore in making any causal inferences is close to zero.  I continue to disagree strongly with the overall discussion offered by the authors on their findings. Their use and reading of the literature are extremely selective. They try to imply that their data suggest that vaccines may have been responsible for a considerable portion of the observed excess deaths, which is a fundamentally unfounded conclusion.  I dissect here their Discussion, because even though it is egregiously biased, these points are useful to highlight as they represent popular false arguments that are often raised especially in social media and sadly inappropriately undermine vaccines.

1. It is not correct that “rates were expected to be in a similar range in each age group, assuming that there is no association between the vaccination status and the non-COVID-19 mortality rates.” Uncorrected, unadjusted group data are almost certain to show some association, which is almost certainly bias, this is a basic expectation in this type of data.  The next couple of paragraphs of the Discussion are then trying to read signals out of what is probably pure noise and bias.

2. The study from Norway did raise some genuine concern early on. Based on this and other similar observations, vaccination in moribund nursing home individuals was probably reduced. However, even if you take these results literally, the total number of expected deaths at a global level from that reason are likely to be in the range of 15,000 [1]. This is a miniscule number compared to the supposedly many millions of deaths that are unexplained according to Kampf and Fornerod. Moreover, the life-time lost would be a few days or weeks for these people, even if we were to accept that their demise was accelerated by vaccination. The same applies to the alluded observations from Germany.

3. Non-specific effects on death remain a speculative issue. The evidence for or against such non-specific effects for vaccines in general, and even more for COVID-19 vaccines, is even more thin and bias-prone that the evidence provided by Kampf and Fornerod in their study.

4. The Japanese paper cited by Kampf and Fornerod has been retracted. The retraction notice runs: “The Editors-in-Chief have retracted this article. Upon post-publication review, it has been determined that the correlation between mortality rates and vaccination status cannot be proven with the data presented in this article. As this invalidates the conclusions of the article, the decision has been made to retract. The authors disagree with this retraction.” I agree that the paper made unbelievable strong statements given the type of data that it had.

5. The study on rat cells has unknown relevance to humans, let alone to human deaths. The cited article on vaccine-associated myocarditis mortality was also withdrawn by the previous journal where it had been published, and the retraction note runs: “This Article-in-Press has been withdrawn at the request of the Editors-in-Chief. Members of the scientific community raised concerns about this Article-in-Press following its posting online. The concerns encompassed. • Inappropriate citation of references. • Inappropriate design of methodology. • Errors, misrepresentation, and lack of factual support for the conclusions. • Failure to recognise and cite disconfirming evidence. The concerns were shared with the authors, who prepared a response and submitted a revised manuscript for consideration by the journal. In consideration of the extent of the concerns raised and the responses from the authors, the journal sent the revised manuscript to two independent peer-reviewers. The peer-reviewers concluded that the revised manuscript did not sufficiently address the concerns raised by the community and that it was not suitable for publication in the journal. The authors disagree with this withdrawal and dispute the grounds for it.”  Myocarditis is indeed a recognized adverse event for mRNA vaccines, especially in young males and some deaths may have been caused by it. However, the best estimates would suggest roughly a few hundred deaths at a global level (or far fewer) over the entire pandemic: the risk of myocarditis seems to be less than 150 per million in children and adolescents and much lower in higher ages and among cases 1-4% may die [2-4] – in fact, these rates may also be inflated. At any rate, such cardiac deaths would account for an extremely miniscule portion of any observed excess deaths during the pandemic.

6. There is absolutely nothing striking about the age pattern; for this type of data, this (or more extreme even) patterns may easily occur without the vaccines having anything to do with it. Totally or partly different patterns have been seen in similar groupings/analyses in other countries, now cited by Kampf and Fornerod in their revised paper.

7. The meandering extrapolation used to suggest “an overall negative benefit-risk” is completely unfounded. It is very likely indeed that effectiveness against Omicron was less versus effectiveness for previous variants, this has been widely documented [5-7]. But the logical jump from reduced effectiveness to net harm is totally unjustified.

8. There is absolutely no way that one can account for 12.3 million deaths from vaccine adverse events, each of which seems likely to account for a few hundred or a few thousands of deaths globally at most, based on the incidence data we have [1].  Non-COVID-19 reasons like deaths from health care disruption, poor health care and limited access, overdose, deaths of despair, and many other problems that accompanied the pandemic and the often highly detrimental pandemic response can readily account for millions of deaths. Blaming these deaths to vaccines only shows bias and, worse, takes our efforts away from fixing major real problems that have plagued public health even before the pandemic and have decimated communities during the pandemic. The roots of these problems may continue unaddressed, while weird vaccine blame games are being played. 

9. I am a strong supporter of the need for randomized trials, and it is a pity that we did not get more randomized evidence. However, extrapolating from 21 and 17 deaths with such short follow-up to millions of deaths from harms is among the most improbable extrapolations I have seen ever made in the medical literature. 

10. A single case adds negligible weight to the claims made here. As above, it is indeed possible that some deaths occurred due to myocarditis, but the scale of numbers is 1000-fold or more lower than the phenomenon that Kampf and Fornerod try to explain.

11. Still a disservice, I think, to make the conclusion “that the COVID-19 vaccines may have contributed to the excess mortality”. Given that more than 13 billion doses of COVID-19 vaccines were given, some deaths due to adverse events are likely to have happened, but they are likely to be overall far fewer than the lives saved. Vaccines in the balance saved more lives, especially among the elderly and the most vulnerable. Of course, risk-benefits and cost-effectiveness need to be carefully studied for different age and risk profile strata and one size does not fit all. However, I am afraid that analyses that use such crude data to make such claims as Kampf and Fornerod did, do not help to advance our knowledge on this front.

References

1. Ioannidis JPA, Pezzullo AM, Cristiano A, Boccia S. Global estimates of lives and life-years saved by COVID-19 vaccination during 2020-2024. medRxiv 2024; doi: https://doi.org/10.1101/2024.11.03.24316673. JAMA Health forum (accepted).

2. Patone M, Mei XW, Handunnetthi L, Dixon S, Zaccardi F, Shankar-Hari M, Watkinson P, Khunti K, Harnden A, Coupland CAC, Channon KM, Mills NL, Sheikh A, Hippisley-Cox J. Risk of Myocarditis After Sequential Doses of COVID-19 Vaccine and SARS-CoV-2 Infection by Age and Sex. Circulation. 146:743–754, 2022.

3. Pillay J, Gaudet L, Wingert A, Bialy L, Mackie AS, Paterson DI, Hartling L. Incidence, risk factors, natural history, and hypothesised mechanisms of myocarditis and pericarditis following covid-19 vaccination: living evidence syntheses and review. BMJ. 378:e069445, 2022.

4. Cho JY, Kim KH, Lee N, Cho SH, Kim SY, Kim EK, Park JH, Choi EY, Choi JO, Park H, Kim HY, Yoon HJ, Ahn Y, Jeong MH, Cho JG. COVID-19 vaccination-related myocarditis: a Korean nationwide study. European Heart Journal, 44:2234–2243, 2023.

5. Chalupka A, Richter L, Chakeri A, El-Khatib Z, Theiler-Schwetz V, Trummer C, Krause R, Willeit P, Benka B, Ioannidis JPA, Pilz S. Effectiveness of a fourth SARS-CoV-2 vaccine dose in previously infected individuals from Austria. European Journal of Clinical Investigation, 54:e14136, 2024. doi: 10.1111/eci.14136.

6. Chalupka A, Riedmann U, Richter L, Chakeri A, El-Khatib Z, Sprenger M, Theiler-Schwetz V, Trummer C, Willeit P, Schennach H, Benka B, Werber D, Høeg TB, Ioannidis JPA, Pilz S. Effectiveness of the First and Second Severe Acute Respiratory Syndrome Coronavirus 2 Vaccine Dose: A Nationwide Cohort Study From Austria on Hybrid Versus Natural Immunity. Open Forum Infectious Diseases, 11:ofae547, 2024. doi: 10.1093/ofid/ofae547.

7. Ioannou GN, Berry K, Rajeevan N, Li Y, Yan L, Huang Y, Lin HM, Bui D, Hynes DM, Rowneki M, Bohnert A, Boyko EJ, Iwashyna TJ, Maciejewski ML, Smith VA, Berkowitz TSZ, O'Hare AM, Viglianti EM, Aslan M, Bajema KL. Effectiveness of the 2023-to-2024 XBB.1.5 COVID-19 Vaccines Over Long-Term Follow-up: A Target Trial Emulation. Annals of Internal Medicine, 2025;178:348-359. [Epub 4 February 2025]. doi:10.7326/ANNALS-24-01015

Closure: Age-adjusted non-COVID-19 mortality rates according to the COVID-19 vaccination status

Martin Kulldorff, [email protected]

Descriptive epidemiology is useful for raising issues that need thorough investigation. Access to and adjustment for potential confounding variables are needed before conclusions can be made from these UK data. This article will hopefully open the access to the additional variables needed to conduct a thorough observational study. Until then, the authors have wisely concluded that “excess deaths are subject to many uncertainties”, that “causal explanations require great caution”, that “a causal relationship … has not been firmly established” and that the question “requires further research”.

The unfortunate failure to conduct randomized trials to evaluate whether the Covid vaccines reduce mortality has led to observational and modelling studies with widely divergent claims concerning the number of lives saved or caused by the vaccines.  This requires open and honest scientific discussions about the strengths and weaknesses of both data and methods. With traditional publishing, this manuscript might have been rejected by some journals before eventually being published without the reader having access to any of the peer-reviews. The way forward is not to hide descriptive epidemiological studies but to openly and honestly discuss them.