Abstract 19509: Identifying Genotype-Phenotype Relations in Electronic Medical Record Systems: Application to Warfarin Pharmacogenomics.
Introduction: We have shown that the associations between genetic variation and common diseases can be replicated in electronic medical record (EMR) systems. However, the extension of this approach to pharmacogenetics has not been evaluated. We tested whether the association between steady state warfarin dose and variants in VKORC1 and CYP2C9 could be replicated in BioVU, the Vanderbilt DNA Databank that links DNA samples from >86,000 patients to de-identified EMRs.
Methods: Stable warfarin dose was defined as the median weekly dose taken from the first window in which a patient had a consistently therapeutic INR between 2 and 3 for at least 3 weeks (with no out-of-range INRs). We used natural language processing algorithms to identify records with stable warfarin dose and these were verified by human review. Eight VKORC1 and CYP2C9 SNPs were genotyped in 1,007 European Americans and four passed quality control criteria: minor allele frequencies >1%, genotypes in Hardy Weinberg Equilibrium, and call rates >95%.
Results: Log-transformed median steady-state warfarin dosage was strongly associated with 3 VKORC1 SNPs all in linkage disequilibrium (p<10−54, beta = −0.32 to −0.35) and CYP2C9 rs4917639, which tags both *2 and *3 variants (p = 8.5x10−31, beta = −0.32). The association was seen equally in subjects followed in specialized “coumadin clinics” and those managed by providers. The VKORC1 SNPs explained ∼22% of the dose variability and CYP2C9 rs4917639 explained ∼12%. Extraction of warfarin dosing achieved a positive predictive value of 98%. Identification of subjects and their validation took approximately two months.
Conclusion: This is an initial demonstration that DNA repositories linked to drug outcome data in an EMR can replicate previously known pharmacogenetic associations. Deploying advanced informatics in EMR systems can rapidly generate very large datasets for replication and discovery in pharmacogenetics.
- © 2010 by American Heart Association, Inc.