CLINICAL VERSUS FIXED WARFARIN DOSING AND THE ADDED EFFECT OF A GENOTYPE GUIDED ALGORITHM ON WARFARIN DOSE PREDICTION IN ARABS

Abstract

Warfarin prescribing has been a challenge due to the need for frequent monitoring. Commonly, warfarin is prescribed using a fixed warfarin dosing method (FWD). Genotype guide warfarin dosing (GWD) is a method that uses genetic and clinical factors to guide warfarin dosing, whereas clinical warfarin dosing (CWD) uses clinical factors alone. GWD was compared to CWD or FWD and showed promising results. It is unknown how the non-GWD methods compare to one another. Additionally, the effect of GWD on the Arab population is not well studied. A three-phase project was conducted. The first phase was a prospective clinical study with a historical control comparing CWD to FWD in a group of Arab patients. The Gage et al. algorithm was used to calculate warfarin dose in the CWD arm. No difference was seen between the CWD and FWD in terms of the mean percentage of time in the therapeutic range (62.2 ± 26.2% vs. 58 ± 25.4%, p=0.2). In the second phase, saliva samples were collected from the previous project, and the following genetic variants were genotyped: VKORC-1639G>A, CYP2C9*2, CYP2C9*3, and CYP4F2*3. Those with at least one reduced function allele required a significantly lower median (IQR) weekly dose compared to the normal function allele [24.5 (15.3) mg/week vs. 35 (29.8) mg/ week, p=0.006]. Similar results were seen in variants of VKORC-1639G>A. In the third phase, patients were randomized to a derivation (67%) and a validation cohort. Using univariate and multiple regression analyses, factors affecting warfarin dose were identified, and a regression equation model was developed. The model could explain 56% of warfarin dose variability. Using the validation cohort, the model had a significantly lower median absolute error when compared to the Gage et al. algorithm (9.1 vs 10.8, p=0.02). In conclusion, the results show that FWD and CWD produced similar outcomes, and if pharmacogenetic testing is not available, either method can be utilized. Additionally, variants of VKORC-1639G>A and CYP2C9 but not CYP4F2*3 were associated with warfarin dose in the Arab population. The use of the derived model was more accurate in predicting warfarin dose compared to an internationally validated algorithm.