Cohort designs may underestimate rare vaccine adverse events due to healthy vaccinee bias, as evidenced by a spurious protective effect of COVID-19 vaccination on stroke risk in a nationwide cohort that contrasts with self-controlled case series findings of a small increase in post-vaccination stroke risk.
Key Findings
Results
Stroke incidence was lower in vaccinated persons than in matched unvaccinated persons during the 42 days following COVID-19 vaccination, suggesting a spurious protective effect.
8 strokes occurred in the vaccinated cohort and 19 in the unvaccinated cohort during follow-up
Incidence rates were 0.2 (95% CI, 0.1–0.3) per 100,000 person-weeks in the vaccinated group and 0.4 (95% CI, 0.2–0.6) per 100,000 person-weeks in the unvaccinated group
The adjusted hazard ratio comparing vaccinated with unvaccinated individuals was 0.42 (95% CI, 0.18–0.96)
The study period was 1 December 2020 to 11 April 2023 in Qatar
Methods
The nationwide retrospective cohort included 1,111,939 vaccinated persons matched individually to an equal number of unvaccinated persons.
Individual matching was used to create comparable vaccinated and unvaccinated groups
The cohort was nationwide in Qatar, providing a population-level assessment
The risk window assessed was 42 days following vaccination
This is described as a retrospective cohort design
Results
Cohort study findings of a protective effect of vaccination on stroke risk diverged from self-controlled case series (SCCS) results from the same and other populations, which indicated a small increase in post-vaccination stroke risk.
SCCS analyses from the same population and other populations indicated a small increase in post-vaccination stroke risk
The cohort design found an adjusted hazard ratio of 0.42 (95% CI, 0.18–0.96), suggesting apparent protection
The divergence between cohort and SCCS results was interpreted as potentially reflecting healthy vaccinee bias in between-person comparisons
The SCCS design allows only indirect estimation of absolute risk differences, whereas cohort designs allow direct estimation
Discussion
Healthy vaccinee bias is proposed as the likely explanation for the discrepancy between cohort and SCCS findings.
Healthy vaccinee bias arises in between-person comparisons when vaccinated individuals are systematically healthier than unvaccinated individuals
The authors argue this bias may lead cohort designs to underestimate rare vaccine adverse events
This finding challenges the exclusive use of cohort designs for estimating absolute risk differences following vaccination
The authors suggest that SCCS results, which use within-person comparisons, may be less susceptible to this form of bias
Conclusions
The authors conclude that cohort designs should not be used exclusively for estimating absolute risk differences for rare vaccine adverse events.
The study evaluated whether absolute measures should be estimated exclusively using cohort designs by comparing results to prior SCCS estimates
The cohort finding of a spurious protective effect contrasts with SCCS results indicating increased stroke risk
The authors describe this divergence as 'challenging the exclusive use of cohort designs for estimating absolute risk differences'
The SCCS design is described as 'widely used for short-term vaccine safety assessment' but limited to indirect estimation of absolute risk differences
What This Means
This research compared two different study methods for detecting rare side effects of COVID-19 vaccines, using stroke as an example. The researchers conducted a large cohort study in Qatar, following over one million vaccinated people and an equal number of unvaccinated people for 42 days after vaccination. Surprisingly, the cohort study found that vaccinated people appeared to have a lower risk of stroke than unvaccinated people — a result that seems biologically implausible as a true vaccine benefit, and which contradicts findings from a different type of study design (called a self-controlled case series, or SCCS) that found a small increase in stroke risk after vaccination.
The researchers believe the cohort study's misleading result is due to a phenomenon called 'healthy vaccinee bias.' This occurs because people who choose to get vaccinated tend to be healthier overall than those who do not — meaning any comparison between vaccinated and unvaccinated groups is not truly apples-to-apples. The SCCS method avoids this problem by comparing each person's own health outcomes at different time points (before and after vaccination), rather than comparing vaccinated people to different unvaccinated people.
This research suggests that cohort studies — a commonly used approach in medical research — may miss or undercount rare side effects of vaccines because of this bias. The findings imply that relying solely on cohort studies to measure absolute risks of rare vaccine side effects could be misleading, and that self-controlled case series or other within-person comparison methods may provide more accurate estimates in these situations. The study highlights the importance of using multiple study designs when assessing vaccine safety.
Chemaitelly H, Akhtar N, Al Jerdi S, Kamran S, Joseph S, Morgan D, et al.. (2026). Cohort studies may underestimate rare vaccine adverse events: evidence from stroke incidence following COVID-19 vaccination.. Expert review of vaccines. https://doi.org/10.1080/14760584.2026.2720953