Healthy vaccinee effect by study design

The study by Semenzato et al. (1) has repeatedly been presented [(2, 3)] as evidence that the French SARS-CoV-2 vaccination campaign since 2021 was a success, quantified by an approximately 25% reduction in all-cause mortality. In this opinion piece, we highlight an issue that, in our view, makes this claim of success untenable.

To this end, we compare the mean annual values of the mortality rates of both 18–59-year-old (hereafter 18–59 years) study groups with the corresponding annual French national all-cause mortality rate. The latter nationwide value was 0.001715 in 2022, calculated from 5-year group data (4, 5), except for the 18–19-year group, for which the size was assumed to be 40% of that of the 15–19-year group. This estimation—derived by summing all group sizes, each weighted by the group’s respective annual all-cause mortality rate ({0.29,0.43,0.48,0.61,0.87,1.29,1.99,3.20,4.91} per thousand), and dividing this weighted sum by the sum of the group sizes (35,296,713)—closely matches the figure given in an online comment to Semenzato et al. (1) (by X. Azalbert, C. Perronne, M. Zizi, 11 December 2025).

Semenzato et al. (1) reported the following group-mean annual rates, averaged across 39 months: 0.001694 for the unvaccinated (unvacc) and 0.001330 for the vaccinated (vacc). These annual rates were calculated by dividing the sum of all deaths counted in either group during 39 months (1, eTable 2; eFigure 5)—32,662 (unvacc) and 98,429 (vacc)—by the respective group sizes (1, eTable 4) (see also Table 1)—5,932,443 and 22,767,546—and scaling by 3.25 (39months12months). The mortality rate of the entire (18–59 years) unvacc group in (1) being slightly lower than that of the French population may be explained by the members of the group being younger on average (Table 1), a factor that applies to the vacc group to a lesser degree.

Table 1. Sizes of age groups within the overall age group (18–59 years), and percentage fractions therein, at the start of follow-up on 1 November 2021, as in (1).

The study’s (1) overall number of patients analyzed (reimbursed in 2020 by the French National Healthcare Data System, Système National des Données de Santé, SNDS) was approximately 33.9 million (1, eFigure 2), while the total French population in this age group was 35 million in 2020 and 35.3 million in 2022. We conclude that the patient population analyzed in the study closely represents the French population. Approximately 4.05 million 18–59-year patients were excluded from the study because they were vaccinated before 1 May 2021. When combined with two smaller excluded groups—those vaccinated later (about 0.64 million) and those who received non-mRNA-based vaccines (about 0.47 million), respectively (see further below; total vaccX group)—these individuals were on average older or more vulnerable (Table 1). Given the French nationwide all-cause mortality rate (0.001715), the mortality rate of the unvacc group (0.001694) closely matches this nationwide average, while the vacc group exhibits a substantially lower rate (0.001330). Consequently, the vaccX group of size 5,162,545 (1, eFigure 2)—comprising those vaccinated early, those vaccinated during the 6-month lag period (1 May to 31 October 2021), and recepients of non-mRNA vaccines—must have exhibited a mortality rate substantially exceeding the nationwide average.

An all-cause mortality rate, X, of the latter vaccX group can be derived to be approximately twice as high as the nationwide value, namely, X=0.003437, which is not surprising considering that the vaccX group must be older (and likely more vulnerable) on average than both the unvacc and vacc groups (see Table 1). This X value follows from requiring the sum of annual counts (here, using consistent 2022 data) of all-cause deaths within each of the three study groups, the unvacc, vacc, and vaccX groups, to equal the annual count of deaths expected from the nationwide all-cause rate for all three groups combined, thereby assuming that the latter group of all patients actually analyzed in (1) (comprising explicitly the unvacc, vacc and implicitly the vaccX groups: 5,932,443+22,767,546+5,162,545=33,862,534) is closely representative of all individuals aged 18–59 years in France. That is, with the vacc and unvacc rates of 0.001330 and 0.001694, respectively, we foundfor the vaccX group. The mortality rate V of the combined vacc&vaccX group of all vaccinated would beOne may object that an additional 641,910 patients (1, eFigure 2) “left” the unvacc group during the 39 months of follow-up due to being vaccinated after 31 October 2021, meaning their group size should rather be close to 5,290,533=5,932,443−641,910 than 5,932,443. In fact, most of these “leavers” were vaccinated during the first 3 months of follow-up, with 5,400,061 unvacc remaining (1, eFigure 5), meaning 532,382=5,932,443−5,400,061 actually “left” (or died). The study’s unvacc mortality rate would then be estimated by ignoring the first 3 months of follow-up, which yields 32,662−2,9905,400,061⋅12months36months=0.001832 instead of 0.001694; the group size of the vaccX cohort would then be 5,694,927=5,162,545+532,382 instead of 5,162,545, and the correspondingly corrected value of the vaccX rate would beinstead of X=0.003437, which is still by a factor of 1.83 higher than the French nationwide all-cause mortality rate of 0.001715. The correspondingly corrected vacc&vaccX rate would beinstead of V=0.001719.

X=(0.001715⋅33,862,534−0.001694⋅5,932,443−0.001330⋅22,767,546)/5,162,545=0.003437

V=(0.001715⋅33,862,534−0.001694⋅5,932,443)/(22,767,546+5,162,545)=0.001719.

Xcor=(0.001715⋅33,862,534−0.001832⋅5,400,061−0.001330⋅22,767,546)/5,694,927=0.003143

Vcor=(0.001715⋅33,862,534−0.001832⋅5,400,061)/(22,767,546+5,694,927)=0.001693

The corrected mortality rate of 0.001832 in the unvacc group after the 3-month follow-up delay being higher than 0.001694 may reflect disproportionate vaccination among younger individuals during those 3 months, resulting in their removal from the unvacc group, and a corresponding increase in the average mortality rate of the remaining cohort. The study’s reported data (1) lack detailed information about age profile shifts during follow-up in the unvacc group. At least during the summer of 2021, during the 6-month lag period, an increasing accumulation of younger individuals in the vacc group was indeed observable (1, eTable 4).

The study’s (1) vacc rate of 0.001330, which is lower than the unvacc rate ranging between 0.001694 and 0.001832, would yield a hazard ratio (HR) between 0.785 and 0.725. Therefore, the homogeneously positive HR effect of vaccination asserted by Semenzato et al. (1) is fully explainable by the exclusion criteria and constitutes a healthy vaccinee effect of a special kind, primarily driven by the disproportionate exclusion of the oldest (Table 1) and more vulnerable individuals (6, Figure 3a: “… et personnes vulnérables”), (1, author response by M. Zureik et al., 19 December) from the analyzed intervention group (vacc). The complete population of (nearly) all vaccinated French individuals aged 18–59 years comprises both the cohort explicitly analyzed by Semenzato et al. (1) from 1 May 2021 onward (vacc) and the excluded vaccinated individuals (vaccX), those vaccinated between 1 May and 31 October 2021, and those who received a non-mRNA vaccine. Once the all-cause mortality counts of this complete vaccinated population are factored in, the conclusion that SARS-CoV-2 vaccination is associated with, or induces, a reduction in all-cause mortality is no longer tenable.

Correction Note

This article has been corrected with minor changes. These changes do not impact the scientific content of the article.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence, and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Jaroslav Janosek, University of Ostrava, Czechia

Reviewed by: Frank Adusei-Mensah, University of Eastern Finland, Finland

Bernard Black, Northwestern University, United States