Nafamostat mesylate versus low-molecular-weight heparin during continuous renal replacement therapy in critically Ill adults: a single-center retrospective cohort study

Background Nafamostat mesylate (NM) is frequently used for continuous renal replacement therapy (CRRT) anticoagulation when clinicians wish to avoid sustained systemic anticoagulation, whereas low-molecular-weight heparin (LMWH) is commonly selected when systemic anticoagulation is considered acceptable. However, direct evidence comparing NM with LMWH in adult intensive care unit (ICU) patients undergoing CRRT remains limited. This study compared filter lifespan, circuit-related failure and safety outcomes between NM and LMWH in a real-world ICU CRRT cohort. Methods This single-center retrospective cohort study included adult ICU patients receiving CRRT with NM or LMWH between January 2022 and January 2023. The primary outcome was filter lifespan. In the primary time-to-event analysis, circuit-related filter failure was defined as clotting-related failure or access/device-related failure, while planned and clinical-event-related terminations were censored. Kaplan–Meier analysis, Cox regression with patient-level cluster-robust standard errors, shared-frailty Cox regression, and propensity score–matched analyses were used to compare filter survival, account for repeated circuits within patients, adjust for CRRT prescription factors, and assess robustness. Results A total of 116 patients contributing 344 filters were included: 31 patients with 132 filters in the NM group and 85 patients with 212 filters in the LMWH group. NM-treated patients had greater illness severity and more pronounced coagulation abnormalities. Unadjusted Kaplan–Meier analysis showed shorter filter survival with NM than with LMWH (log-rank P = 0.007). However, anticoagulant strategy was not significantly associated with circuit-related filter failure in the primary Cox model (HR 1.41; 95% CI 0.88–2.25; P = 0.15), shared-frailty Cox model (HR 1.57; 95% CI 0.85–2.90; P = 0.15), extended Cox model (HR 1.16; 95% CI 0.70–1.90; P = 0.56), or propensity score–matched analyses. Bleeding events were infrequent, precluding a precise comparison of bleeding risk between groups. Conclusions Anticoagulant strategy was not significantly associated with circuit-related filter failure after adjustment for measured covariates. Because of substantial baseline differences, residual confounding, and limited statistical precision, these findings should not be interpreted as evidence of comparative efficacy, equivalence, noninferiority, or comparable safety. Larger prospective studies with standardized treatment protocols are needed.