Comparative performance of pharmacogenetics-based warfarin dosing algorithms derived from Caucasian, Asian, and mixed races in Thai population

Abstract

© 2017 John Wiley & Sons Ltd Aim: This study was conducted to compare predictive accuracy of the available pharmacogenetics (PGx)-guided warfarin dosing algorithms derived from Caucasian, Asian, and mixed population to identify a suitable algorithm for Thai population. Methods: Ten warfarin dosing algorithms derived from different population including Caucasian, East Asian, South-East Asian, and mixed races were selected and tested with clinical and genetic data of Thai patients. Comparative performances of these algorithms were tested using mean dose error (MDE) between actual warfarin maintenance dose (AWMD) and predicted dose generated by each dosing algorithm, and percentage of ideal dose prediction (IDP). Sensitivity analysis for predictive accuracy was also conducted by stratifying patients into low (AWMD ≤21 mg/wk), intermediate (AWMD >21 to <49 mg/wk), and high maintenance dose (AWMD ≥49 mg/wk) groups. Results: Data of 165 patients were included for the analyses. Mean actual warfarin dose of the study population was 25.03 ± 10.53 mg/wk. Large variability of MDE, ranging from −12.11 to 11.24 mg/wk, among algorithms was observed. International Warfarin Pharmacogenetics Consortium, Gage et al, and Ohno et al algorithms had comparable performances to Sangviroon et al algorithm, as observed by MDE of <1 mg/wk with percentage of IDP ≥40%. Further sensitivity analyses among patients requiring low and intermediate maintenance doses confirmed such findings with IDP percentage ranging from 37.8% to 59.2%. Among high-dose group, only Ohno et al and Sarapakdi et al algorithms had acceptable performance. Conclusions: Warfarin PGx-guided dosing algorithms derived from large, mixed population performed comparably to Sangviroon et al algorithm. Certain algorithms should be avoided due to significant dose prediction error.