Genetics of the anticoagulant drug warfarin

In this thesis I use the most widely prescribed anticoagulant drug, warfarin, as a model to investigate the effect of genetic determinants on drug efficacy and safety. Following a literature review of all the genes involved in warfarin pharmacokinetics and pharmacodynamics, 35 candidate genes were selected for investigation. Two independent Swedish cohorts of warfarin-treated patients were analysed. First linkage disequilibrium maps were constructed for each gene. Selected SNPs integrated with putative functional variants were genotyped in 201 patients recruited at the Uppsala University. A panel of 216 haplotype tag SNPs were then derived to analyse an independent cohort of 1496 patients from the prospective Warfarin Genetic Study in Sweden. The two studies were analysed separately for genetic association to warfarin dose requirement (single marker and haplotypic tests).  Common SNPs in the vitamin K epoxide reductase gene (VKORC1) are significantly associated with dose in the Uppsala and WARG studies (p=1.9 x 10^-15 and 6.5 x 10^-100, respectively). Cytochrome P450 2C9 (CYP2C9) has been known to affect dose requirement and was confirmed in both Swedish cohorts (p= 2.3 x 10^-5 and 4.9 x 10^-32). The two genes together explain ~40% of warfarin dose variation. SNPs in microsomal epoxides hydrolase (EPHX1) and orosomucoid 1 (ORM1) genes do not show a broad effect but are associated with dose in both studies. Genes encoding PROC, APOE, CALU, PDIA2 and GGCX showed nominal association with dose in the Uppsala study. Likewise, PROS1, CYP1A1, CYP3A4, PDIA5, PDIA3 and F10 showed nominal association to dose in the WARG study. Most of these minor effects if real are most likely to be population/treatment specific. A model taking in to account genetic factors (VKORC1 and CYP2C9*2/*3) and non genetic factors (age, gender, and drug interaction) together explained more than 50% inter-individual dose variance. We analysed 64 patients from the Uppsala and WARG studies with recorded severe bleeding episodes using the same 216 common SNPs. Case-control analysis found SNPs in PDIA4, P4HB, and NR1I3 to be associated (p≤0.01) with bleeding. Using a recessive model, patients with a gastrointestinal bleeding sub-phenotype in the WARG cohort showed association with common variants in PD1A6 (P=0.0014, odds ratio = 6.98). We sequenced the exons of 11 of the candidate genes in 36 bleeders and 12 non-bleeders (Uppsala study). However, no high prevalence mutation was discovered.