OBJECTIVE: To assess whether the existing three types of pharmacogenetics-based Warfarin dosing algorithms appropriately predict the actual maintenance dose in Han Chinese mechanical heart valve replacement patients (n = 130). METHODS: The patients' CYP2C9 and VKORC1 genetic polymorphisms were detected by PCR-RFLP. The genotype of CYP2C9, VKORC1 and other information were used to calculate predicted doses. Accuracy of the models was assessed using the absolute value of the difference between predicted dose and actual dose, calculated on both an absolute and percentage basis. Actual weekly dose was also regressed on predicted weekly dose, from which we obtained R(2) values. Clinical accuracy of the predictions was assessed by computing the proportion in which the predicted dose was 20% or more below the actual dose (under dosed), within 20% of the actual dose (ideally dosed), or 20% or greater above the actual dose (over dosed). RESULTS: The average absolute error is the smallest for the predictions made by the Wen model (3.74 mg/wk), followed by the Ohno model (4.07 mg/wk) and IWPC model (5.05 mg/wk). R(2) was 40.2% in the Wen model, 38.2% in the Ohno model and 26.7% in the IWPC model. When comparing the percentage of patients for whom the predicted doses were ideal, the Wen model works the best (50.0%) in low-dose group (≤ 21 mg/wk), but the Ohno model works the best (85.29%) in middle-dose group (21 - 49 mg/wk), followed by the Wen model. CONCLUSION: The best accuracy is achieved by the Wen model and the best clinical accuracy is obtained by the Ohno model for predicting the actual maintenance dose in Han Chinese mechanical heart valve replacement patients.
OBJECTIVE: To assess whether the existing three types of pharmacogenetics-based Warfarin dosing algorithms appropriately predict the actual maintenance dose in Han Chinese mechanical heart valve replacement patients (n = 130). METHODS: The patients' CYP2C9 and VKORC1 genetic polymorphisms were detected by PCR-RFLP. The genotype of CYP2C9, VKORC1 and other information were used to calculate predicted doses. Accuracy of the models was assessed using the absolute value of the difference between predicted dose and actual dose, calculated on both an absolute and percentage basis. Actual weekly dose was also regressed on predicted weekly dose, from which we obtained R(2) values. Clinical accuracy of the predictions was assessed by computing the proportion in which the predicted dose was 20% or more below the actual dose (under dosed), within 20% of the actual dose (ideally dosed), or 20% or greater above the actual dose (over dosed). RESULTS: The average absolute error is the smallest for the predictions made by the Wen model (3.74 mg/wk), followed by the Ohno model (4.07 mg/wk) and IWPC model (5.05 mg/wk). R(2) was 40.2% in the Wen model, 38.2% in the Ohno model and 26.7% in the IWPC model. When comparing the percentage of patients for whom the predicted doses were ideal, the Wen model works the best (50.0%) in low-dose group (≤ 21 mg/wk), but the Ohno model works the best (85.29%) in middle-dose group (21 - 49 mg/wk), followed by the Wen model. CONCLUSION: The best accuracy is achieved by the Wen model and the best clinical accuracy is obtained by the Ohno model for predicting the actual maintenance dose in Han Chinese mechanical heart valve replacement patients.