Abstarct:
Immortal time bias (ITB) is common in cohort studies and biases the association estimates between the treated and untreated individuals. We used data from an Italian study on COVID-19 vaccine effectiveness, with a large cohort, long follow-up, and adjustment for confounding factors, but affected by ITB. We verified the real impact of the vaccination campaign, comparing the risk of all-cause death between the vaccinated and the unvaccinated population.
We aligned all subjects on a single index date and considered the “all-cause deaths” outcome to compare the survival distributions of the unvaccinated group versus various vaccination statuses. The all-cause-death hazard ratios (HRs) in univariate analysis for vaccinated people with 1, 2, and 3/4 doses versus unvaccinated people were 0.88, 1.23, and 1.21, respectively. The multivariate values were 2.40, 1.98, and 0.99. Possible explanations of this trend of the HRs as vaccinations increase could be: a harvesting effect; a calendar-time bias, accounting for seasonality and pandemic waves; a case-counting window bias; a healthy-vaccinee bias; or some combination of these factors. With 2 and even with 3/4 doses, the calculated Restricted Mean Survival Time and Restricted Mean Time Lost showed a small but significant downside for the vaccinated populations.
The study results should lead to rethink political choices about pandemic management, supporting greater caution in the future.