Prediction of Early Heparin Response in Patients With Venous Thromboembolism

Background: Unfractionated heparin (UFH) is widely used for treatment of acute venous thromboembolism (VTE); however, its anticoagulant response is highly variable. Delays in achieving therapeutic anticoagulation have been associated with increased mortality, suggesting a potential role for predictive modeling to guide dosing. Objective: To develop predictive models to estimate probability of therapeutic or supratherapeutic level attainment at the time of first activated partial thromboplastin time (aPTT) and the UFH infusion rate at the time of first therapeutic aPTT. Methods: This retrospective, cohort study included hospitalized patients with a new diagnosis of VTE receiving protocolized UFH (80 units/kg bolus followed by 18 units/kg/h infusion). Patients were excluded for protocol deviations, anticoagulation reversal, hypothermia, or pregnancy. The primary outcome was development of a predictive model for attainment of therapeutic or supratherapeutic aPTT at first follow-up. A secondary model was developed that predicted heparin infusion rate at the time of first therapeutic aPTT. Multivariable logistic and linear regression models were developed with variable selection using least absolute shrinkage and selection operator. Model performance was assessed using discrimination and calibration. Results: Among 398 included patients, 83.2% achieved a therapeutic or supratherapeutic aPTT at initial follow-up. Eleven predictors were included in the primary model (age, sex, race, white blood cell count, creatinine clearance, body mass index, total bilirubin, albumin, platelets, propofol exposure, and coagulopathy) and demonstrated moderate discrimination (area under the curve [AUC] = 0.744) with acceptable calibration for predicting UFH response. Prediction of heparin dose at initial therapeutic aPTT incorporated a subset of these variables along with additional predictors (weight, Glasgow Coma Scale, baseline aPTT, active viral infection, and aspartate aminotransferase) and demonstrated substantial performance ( R 2 = .714). Conclusion and Relevance: Clinical and laboratory factors influence UFH response and dosing requirements. These models may support individualized dosing strategies to improve time to therapeutic anticoagulation, although external validation is needed before clinical application.