OBJECTIVESTo establish an algorithm to predict the warfarin maintenance dose in Chinese Han population and validate the accuracy of this algorithm.

METHODSA total of 488 Chinese Han patients, hospitalized in Fuwai hospital and had a stable dose of warfarin and a target international normalized ratio (INR) of 1.5 to 3.0, were recruited. Indications for warfarin use included prosthetic heart valve, atrial fibrillation and pulmonary embolism. These patients were divided into derivation group (n = 323) and validation group (n = 165) according to the enrollment time. A warfarin maintenance dose algorithm was established based on genetic information, demographic characteristics and concomitant medications by multiple linear regression analysis in derivation group. In the validation group, we evaluated the accuracy of our algorithm by comparing the predicted dose with the actual dose.

RESULTSOur algorithm included VKORC1-1639G > A, CYP2C9*3 and CYP4F2 genotype, age, Body hight, body weight, amiodarone and digoxin use (R(2) = 0.652, P < 0.001) .In the validation group, the average predicted dose by our algorithm had no statistical difference with the actual dose [(3.51 ± 1.03) mg vs. (3.53 ± 1.41) mg, P = 0.779]. Our algorithm identified 100 out of 165 (60.6%) patients in the validation group, whose predicted dose of warfarin was within 20% of the actual dose, and predicted warfarin dose was underestimated in 17.6% (29/165) patients and overestimated in 21.8% (36/165) patients.

CONCLUSIONOur algorithm based on VKORC1, CYP2C9 and CYP4F2 polymorphisms can help to predict the warfarin maintenance dose in Chinese Han Population.