Background: Warfarin is a highly effective anticoagulant however its effectiveness relies on maintaining INR in therapeutic range. Finding the correct dose is difficult due to large inter-individual variability. Two genes, CYP2C9 and VKORC1, have been associated with this variability, leading to genotype-guided dosing tables in warfarin labeling. Nonetheless, it remains unclear how genotypic information should be used in practice. Navigating the literature to determine how genotype will influence warfarin response in a particular patient is difficult, due to significant variation in patient ethnicity, outcomes investigated, study design, and methodological rigor. Our systematic review was conducted to enable fair and accurate interpretation of which variants affect which outcomes, in which patients, and to what extent.
Methodology/principal findings: A comprehensive search strategy was applied and 117 studies included. Primary outcomes were stable dose, time to stable dose and bleeding events. Methodological quality was assessed using criteria of Jorgensen and Williamson and data synthesized in meta-analyses using advanced methods. Pooled effect estimates were significant in most ethnic groups for CYP2C9*3 and stable dose (mutant types requiring between 1.1(0.7-1.5) and 2.3 (1.6-3.0)mg/day). Effect estimates were also significant for VKORC1 and stable dose for most ethnicities, although direction differed between asians and non-asians (mutant types requiring between 0.8(0.4-1.3) and 1.5(1.1-1.8)mg/day more in asians and between 1.5(0.7-2.2) and 3.1(2.7-3.6)mg/day less in non-asians). Several studies were excluded due to inadequate data reporting. Assessing study quality highlighted significant variability in methodological rigor. Notably, there was significant evidence of selective reporting, of outcomes and analysis approaches.
Conclusions/significance: Genetic associations with warfarin response vary between ethnicities. In order to achieve unbiased estimates in different populations, a high level of methodological rigor must be maintained and studies should report sufficient data to enable inclusion in meta-analyses. We propose minimum reporting requirements, suggest methodological guidelines and provide recommendations for reducing the risk of selective reporting.