Application of Akaike information criterion to evaluate warfarin dosing algorithm.

Introduction: Several factors responsible for inter-individual differences in response to warfarin have been confirmed; however, unidentified factors appear to remain. The purpose of this study was to examine a simple method to evaluate whether optional variables are appropriate as factors to improve dosing algorithms.

Materials And Methods: All patients were Japanese. Genotyping of selected genes was conducted, and other information was obtained from medical record. Dosing algorithms were constructed by multivariate linear regression analyses and were evaluated by the Akaike Information Criterion (AIC).

Results And Conclusions: Multivariate analysis showed that white blood-cell count (WBC), concomitant use of allopurinol, and CYP4F2 genotype are apparently involved in warfarin dose variation, in addition to well-known factors, such as age and VKORC1 genotype. We evaluated the adequacy of these variables as factors to improve the dosing algorithm using the AIC. Addition of WBC, allopurinol administration and CYP4F2 genotype to the basal algorithm resulted in decreased AIC, suggesting that these factor candidates may contribute to improving the prediction of warfarin maintenance dose. This study is the first to evaluate the warfarin dosing algorithm by AIC. To further improve the dosing algorithm, AIC may be a simple and useful tool to evaluate both the model itself and factors to be incorporated into the algorithm.