Time-aligned association between ICU statin exposure and 28-day mortality in sepsis: a MIMIC-IV cohort study

Purpose Observational associations between statin exposure and mortality in sepsis may be distorted by immortal time bias and covariate timing. We reassessed this association using time-updated exposure and complementary landmark analyses in MIMIC-IV. Methods Adults with sepsis in MIMIC-IV version 3.1 were followed from ICU admission for 28-day all-cause mortality. Statin exposure changed from unexposed to exposed at the first ICU statin prescription in a counting-process Cox model adjusted for admission-known covariates. Twenty-four-hour and 48-hour landmark analyses were performed. A sensitivity analysis across five multiply imputed datasets additionally included first-day severity, laboratory variables, and treatments re-extracted within 0–24 h. Results Among 10,156 eligible patients, 2,535 received a statin during ICU stay. Time-updated statin exposure was associated with lower 28-day mortality after adjustment for admission-known covariates (HR 0.82 (95% CI 0.74-0.90; P < 0.001)). The 24-hour and 48-hour landmark estimates were similar (HR 0.84 (95% CI 0.76-0.93; P < 0.001) and HR 0.84 (95% CI 0.76−0.93; P < 0.001), respectively). The association attenuated after adjustment for first-day severity, laboratory variables, and corrected 0−24-hour treatments (multiply imputed HR 0.95 (95% CI 0.87−1.05; P = 0.310)). Conclusions Time-updated ICU statin exposure was associated with lower 28-day mortality in models using admission-known covariates and in the 24-hour and 48-hour landmark analyses. Attenuation after adding first-day clinical status indicates sensitivity to covariate specification and possible residual confounding; a causal survival benefit cannot be established.