A few warfarin pharmacogenetic dosing algorithms have been proposed,based on multiethnic or homogeneous populations, to estimate warfarin therapeutic doses. However, it remains to be proven that which algorithm is accurate in predicting warfarin dose in Chinese people. We selected eight warfarin dose predictive pharmacogenetic algorithms and retrospectively assessed the predictive accuracy of each algorithm in a total of 368 eligible outpatients by comparing the actual stable therapeutic dose to the dose predicted by the algorithm. Our results showed that a high level of performance was demonstrated by three algorithms,Gage et al., Anderson et al., and Wu et al., having a similar performance in coefficient of determination (R2) and percentage of patients predicted dose within 20% of actual dose. The Gage et al. algorithm had the lowest mean absolute error (MAE). These results indicated that the algorithm by Gage et al. provided a more accurate prediction than did the others,which suggests that this pharmacogenetic algorithm might be used in clinical practice to guide rational administration of warfarin.