Development And Evaluation Of Dose-response Model Of Warfarin In Heart Valvular Replacement Patients

Objectives: In order to establish the dose-response model in heart valvular replacement patients and explore the potential factors that affect the warfarin maintained dose, we compared the predictive abilities of initial and stable response(or dosage)between the dose-response model and multiple linear regression model that was reported in literatures. A more appropriate model for warfarin individualized therapy was suggested,which would provide a reference for warfarin individualized therapy in clinic.Methods:(1) 535 patients who were administered with warfarin for a long time were enrolled in this study. A number of covariates were collected, including demographic characteristics, current medication, combined drugs, operation history, and the results of INR, etc.(2) The genotype of CYP2C9 and VKORC1 were detected by PCR-pyrosequencing.(3) The dose-response model analysis was adopted by NONMEM software. A first-order conditional estimation with interaction(FOCE-I) method within NONMEM was used to estimate the parameters. The covariates, including demographic characteristic, the CYP2C9 and VKORC1 polymorphisms, were evaluated quantitatively. The reliability and stability of the final model were evaluated using resampling techniques with nonparametric bootstraps.(4) Then, 68 patients who were administered with warfarin for the first time were collected. We compared the predictive abilities of initial and stable responses(or dosages)between the dose-response model and multiple linear regression model which were reported in literatures.Results :(1) In 535 heart valve replacement patients, the allele frequencies of CYP2C9(rs1057910)*1 and *3 were 95.2% and 4.8%, respectively, and the allele frequencies of VKORC1(rs9923231) G and A were 8.8% and 91.2%, respectively.(2) Among the investigated patients, the typical value of Km was 1.97 mg, theinter-individual variability of Km was 42.4%, and the intra-individual variability of observed INR was 0.179. The Km value was found to decrease with age and increase with weight. The Km value of CYP2C9*1/*3 patients was about 19.4% lower than that of CYP2C9*1/*1 patients. The Km value of VKORC1 GA or GG patients was about 53.4%higher than that of VKORC1 AA patient. In addition, the Km value decreased about 27.2%when patients were combined with amiodarone.(3) The mean absolute error of the predict response of the final dose-response model were 26.0%(SD:21.9%) and 21.2% for initial warfarin therapy and stable warfarin therapy,respectively. The percentages of the predict response bias that were within 20% were45.6% and 56.4% for initial warfarin therapy and stable warfarin therapy, respectively.Conclusions: The genotype of CYP2C9 and VKORC1, age, weight, and amiodarone were associated with the Km of warfarin in heart valvular replacement patients. The established dose-response model could provide references for individualized use of warfarin in clinic. The dose-response model had a better predictive abilities of initial and stable response(or dosage) than multiple linear regression model that was reported in literatures.