Development of a predictive model for nafamostat mesylate dosing in hemodialysis

Background: Nafamostat mesylate, an anticoagulant with a short half-life, is useful during hemodialysis for patients at high risk of bleeding. Research on the appropriate dosage of nafamostat has been insufficient. This study aimed to investigate the optimal dosage of nafamostat using a dosage prediction model. Methods: We retrospectively analyzed medical records from 12 centers affiliated with Yeolin Medical Foundation over an 8-month period. Candidate predictor variables were evaluated using bootstrapping and stepwise regression to determine feature importance. Predictive models were compared based on performance metrics. Results: In total, 308 sessions from 88 patients without dialyzer clotting were selected. The average nafamostat dose was 21.90 ± 6.82 mg/h. The top four important features were oral anticoagulant use, dry body weight, age, and hemoglobin level. The best-performing model using 12 variables showed a root mean squared error of 4.11 mg/h and an adjusted R² value of 0.49. Multivariable linear regression showed that oral anticoagulant use (coefficient − 14.20, 95% confidence interval [CI] -18.28 to -10.12, P < 0.001) and age (-0.13, 95% CI -0.19 to -0.08, P < 0.001) were associated with a lower nafamostat dose, whereas dry body weight (coefficient 0.15, 95% CI 0.09 to 0.22, P < 0.001) and hemoglobin level (1.13, 95% CI 0.51 to 1.76, P < 0.001) were associated with a higher dose. Conclusion: The nafamostat dosage prediction model can be used to calculate the dose required for individual patients. However, further studies for model improvement and external validation are required. Clinical trial number: Not applicable.