Abstract
Survivor bias commonly weakens observational studies, even those published in premier journals. It occurs because patients who live longer are more likely to receive treatment than those who die early. We sought to quantify the effect of survivor bias on the association between valve surgery and mortality in infective endocarditis (IE).
The study cohort included 546 IE patients. We compared the hazard ratios (HR) resulting from two propensity score analysis approaches that adjusted for survivor bias (time-dependent variable and matching on follow-up time) with those achieved using the same models but without that adjustment (time-fixed variable).
In the total cohort, the HR of surgery in the time-dependent model was 1.9 (95% confidence interval [CI] = 1.1–3.2;
P = 0.03) vs. 0.9 (95% CI = 0.5–1.4;
P = 0.53) in the time-fixed model. In the propensity score–matched subset, the HR of surgery was 1.3 (95% CI = 0.5–3.1;
P = 0.56) and 0.8 (95% CI = 0.4–1.7;
P = 0.57) in the subset with and without matching on follow-up time, respectively.
Adjusting for survivor bias changed the conclusion about the association between valve surgery and mortality in IE. Researchers should be aware of this bias when evaluating observational studies of treatment efficacy.