Purpose The objectives of the study were to establish a dose-response model for warfarin based on the relationship between daily warfarin dose and international normalized ratio (INR) and to evaluate the stability and reliability of the established model using external data. Methods Clinical data were recorded from 676 outpatients with a steady-state warfarin dosage. Demographic characteristics, concomitant medications, daily dosage of warfarin, CYP2C9 and VKORC1 genotypes, and INR were recorded. Data analysis based on the Michaelis-Menten equation to describe the relationship between daily warfarin dose and INR was performed using NONMEM. The reliability and stability of the final model were evaluated using goodness-of-fit plots, resampling techniques with a nonparametric bootstrap, and external data. Results The daily warfarin dose and INR were described by a more pharmacologically expressive model than multivariate linear regression (MLR) model. The population standard value of Km was 3.56mg, and the Hill coefficient was 0.512, with individual variabilities of 53.1% and 55.9%, respectively. CYP2C9 *1/*3, VKORC1 AA, concomitant amiodarone, and nonheart valve replacement reduced the warfarin Km by 30.4%, 74.3%, 34.5%, and 39.4%, respectively. The Km value decreased with age and increased with fat free mass (FFM). INR prediction error (73.0%) of the external datasets was within 20%. Conclusion A dose-response model of warfarin was established based on the relationship between daily warfarin dose and INR. Expected genotype effects on Km and demographic characteristics were confirmed. The model has the potential to be a powerful tool for individualized warfarin therapy for Chinese outpatients.