Covid-19 ‘vaccinations’ have not protected against death from Covid-19 or other causes
Our analysis of the Semenzato study has been published
On 4 December 2025, a sigh of relief swept through the press: a large French study had allegedly shown that people who had been vaccinated against Covid-19 were less likely to die from Covid-19, as well as from other illnesses [1]. The data came from the French national health service and covered more than 22 million vaccinated people and almost 6 million unvaccinated people. According to the authors, the vaccinated had a 74 percent reduced risk of dying from Covid-19 and a 25 percent reduced risk of dying from all possible causes (‘all-cause mortality’). The authors had statistically controlled for all possible confounding factors. Everything seemed to be going swimmingly. The mainstream press cheered. Politicians breathed a sigh of relief. Lauterbach, former German Minister for Health, triumphed.
I was skeptical, took a closer look at the publication, and saw what I had suspected: several groups had been excluded from the analysis – those who were vaccinated early on, people in poorer health, and so on. This is permissible, provided it is defined in advance in an analysis protocol. I wrote to the senior author, Prof. Mahmoud Zureik, and asked for the analysis protocol, which, strangely enough, was not mentioned or linked to anywhere. I also asked the obvious question: whether this selection had been defined in advance. I received no reply.
Study and analysis protocols, as required by good scientific practice, should be formulated in advance and, ideally, also registered somewhere, e.g., on websites or in study registries. This makes it possible to verify whether the authors have adhered to their own guidelines, definitions, and procedures. I have been doing this as standard for many years with all major studies. For significant studies, I publish the protocols or post them on relevant websites and provide links to the protocols in my publications. It is very strange that reviewers at a reputable journal such as the JAMA group failed to spot this and raise the issue.
If one fails to do this and, as in this case, is working with an existing dataset, the following can happen: You tweak the dataset until you find what you expect to or want to be correct. Of course, it cannot be proven whether the authors actually did this. But conversely, the authors would have to demonstrate that they did not do so. However, they have not done so, and they have evaded questions on this point.
We therefore took a closer look at this publication and wrote a critique, initially as a ‘Letter to the Editor’ of the journal. The editor did not find the letter ‘sufficiently interesting for the readership’. We therefore tried submitting it to various other outlets. It has now been published in Frontiers of Epidemiology and is freely available [2].
My colleague Michael Günther in particular sat down to pull an all-nighter and compare the mortality data for the entire group – and especially for the unvaccinated – with that of the general French population. Lo and behold: they are largely identical. This means that the mortality data for the unvaccinated do not differ in any way from that of the general population. To put it another way: the vaccination cannot have had any effect on mortality whatsoever. This, too, can be demonstrated.
The vaccinated that had been included in the study, on the other hand, had a significantly lower mortality rate than the general population. So what happened? Due to the selection of the vaccinated – and the exclusion of the sicker individuals amongst them, namely those who were vaccinated earlier – the group of vaccinated people is, on the whole, slightly younger and healthier. It is therefore no longer comparable to the unvaccinated. Nor can this be addressed through statistical adjustment of relatively coarse variables such as different disease categories.
My colleague demonstrates this by calculating the mortality rate of those excluded – that is, those vaccinated earlier who were not included in the study – and showing that it is roughly twice as high as the national average.
What we see in the study, therefore, is a classic selection effect: the so-called ‘healthy-vaccinee’ effect. Because the vaccinated group consists primarily of healthier individuals – since those in poorer health were excluded by the selection criterion based on timing – mortality is naturally lower in every respect. When this is then compared with the unvaccinated group, which is very similar to the general population, the impression arises that the vaccination is responsible for this. But it is not. The selection of the study population is responsible. And that was selected by the authors. As long as they do not disclose the nature of this selection, it can be assumed that it was carried out deliberately in order to achieve a desired outcome. Many have suspected that this study result was brought about by selection. This now proves it.
Sources
- Semenzato L, Le Vu S, Botton J, Bertrand M, Jabagi M-J, Drouin J, et al. COVID-19 mRNA vaccination and 4-year all-cause mortality among adults aged 18 to 59 years in France. JAMA Network Open. 2025;8(12):e2546822-e. doi: https://doi.org/10.1001/jamanetworkopen.2025.46822.
- Günther M, Walach H, Rockenfeller R, Reitzner M, Meyer C. Healthy vaccinee effect by study design. Frontiers in Epidemiology. 2026; Volume 6 – 2026. doi: https://doi.org/10.3389/fepid.2026.1837295.