Quantitative Comparison of the Predictive Accuracy of Warfarin Pharmacogenetic Dosing Algorithms Derived From Population Data of Different Ethnicities in the Chinese Population

This study systematically evaluated the predictive performance of 10 international warfarin dosing algorithms (originating from the United States, China, Singapore, Thailand, India, United Kingdom, Japan, and South Korea) in 87 Chinese patients, aiming to identify optimal algorithms for warfarin dose optimization. Clinical and genetic data were analyzed using mean dose error (MDE) and ideal dose prediction (IDP) rate metrics, with sensitivity analysis stratifying patients into low-dose (≤14 mg/week, n = 21), medium-dose (14-21 mg/week, n = 43), and high-dose (≥21 mg/week, n = 23) groups based on actual weekly maintenance dose (mean: 18.9 ± 8.8 mg/week). Results revealed significant variation in MDEs (-6.6 to 11.3 mg/week) across algorithms. The Chinese-developed Huang algorithm and Thai-developed Sangviroon algorithm demonstrated superior overall accuracy, both achieving MDEs <1 mg/week and IDPs >40%. In medium-dose patients, their performance was particularly robust (Huang IDP: 65.1%; Sangviroon IDP: 74.4%). However, both algorithms showed limitations at dose extremes: they overestimated doses in 90.48% of low-dose patients and underestimated doses in 60.9%-65.2% of high-dose patients. This evidence indicates that region-specific algorithms (Huang and Sangviroon) outperform internationally recommended models (e.g., IWPC/Gage endorsed by CPIC) for warfarin dosing in Chinese populations. Locally derived algorithms may thus offer greater clinical utility despite current international guidelines.