Abstract: AIM: To provide reference for clinical application of warfarin PPK/PD model, the prediction accuracy of warfarin PPK/PD model and 6 dose models established by multiple linear regression were compared.  METHODS: Clinical data of inpatients who took warfarin tablets for oral anticoagulant therapy in our hospital were collected, and the predictive values were simulated by PPK/PD model and other 6 models, respectively. SPSS 23.0 software was used for paired t-test of measured value and predicted value. MAE and percentage of prediction deviation were used to evaluate the results, and the prediction deviation box-plot was drawn to compare the total data, different dose groups and different genotypes. RESULTS: A total of 50 patients were included in the study. Among 7 models, only PPK/PD model, Wen et al., and Du Liping et al.'s model had no statistical difference in predicted values and measured values (P>0.05). The prediction accuracy of PPK/PD model was higher among the total data, low and medium doses, and patients with different genotypes.The prediction accuracy of Wen et al. 's model and Li Chuanbao et al.'s model was higher in the high-dose group. CONCLUSION: The PPK/PD model of warfarin has good clinical prediction performance, which is expected to provide reference for accurate administration of warfarin.

Key words: warfarin, the PKK/PD model, dose model, accuracy of prediction