Association of acetylcholinesterase inhibitor use with mortality among critically ill patients with sepsis

Acetylcholinesterase inhibitors (AChEIs) may modulate systemic inflammation via the cholinergic anti-inflammatory pathway, but their prognostic significance in sepsis remains unclear. This retrospective cohort study was conducted using real-world intensive care unit (ICU) data and included eligible patients with sepsis. AChEI exposure was defined as any recorded use during the ICU stay, and outcomes were 28-, 90-, and 365-day all-cause mortality. Associations were evaluated using multivariable Cox regression, with inverse probability of treatment weighting (IPTW) and time-dependent Cox analyses performed to address confounding and the timing of AChEI initiation. Subgroup and interaction analyses assessed heterogeneity. Prediction models derived from the multivariable models were implemented as a web-based application and evaluated for discrimination, calibration, overall prediction accuracy, clinical utility, and bootstrap internal validation. A total of 28,838 patients with sepsis were included, of whom 4,123 received AChEIs. AChEI use was significantly associated with lower 28-, 90-, and 365-day mortality (HR 0.25, 95% CI 0.21–0.30, E-value 7.34; HR 0.26, 95% CI 0.23–0.31, E-value 7.05; and HR 0.27, 95% CI 0.23–0.31, E-value 6.99, respectively), with consistent findings in IPTW-weighted and modified Poisson regression analyses. Time-dependent Cox analyses yielded consistent associations at 28, 90, and 365 days (HR 0.28, 95% CI 0.24–0.34; HR 0.29, 95% CI 0.24–0.34; and HR 0.28, 95% CI 0.24–0.33, respectively). Peripherally acting AChEIs ( n = 3,769) were associated with lower mortality estimates than centrally acting AChEIs ( n = 354) at 28, 90, and 365 days (HR 0.28, 95% CI 0.19–0.39; HR 0.30, 95% CI 0.22–0.42; and HR 0.33, 95% CI 0.24–0.45, respectively). Associations were generally consistent across subgroups, with significant interactions for sex, dementia, acute kidney injury, and beta-blocker use. The prediction models showed good discrimination, with C-indices of 0.788, 0.780, and 0.770 for 28-, 90-, and 365-day mortality, respectively, and generally good calibration after bootstrap internal validation. AChEI use showed a significant inverse association with short- and long-term mortality in patients with sepsis, with potential heterogeneity across pharmacological subtypes and certain clinical subgroups. These findings warrant further investigation, while the prediction models require external validation before broader clinical application.