CYP2C9
CYP2C9 is the enzyme that clears warfarin — the classic example of pharmacogenomics. It also metabolizes phenytoin, several NSAIDs, and some diabetes drugs. People who carry reduced-function alleles break warfarin down slowly, need much lower doses, and face a higher bleeding risk while their dose is found. Together with the drug-target gene VKORC1, CYP2C9 explains a large share of why the right warfarin dose varies so widely from person to person.
The warfarin enzyme.
CYP2C9 is a hepatic cytochrome P450 enzyme that inactivates a set of narrow-therapeutic-window drugs — the kind where a little too much or too little matters. Its headline substrate is S-warfarin, the more potent enantiomer of the anticoagulant warfarin (Coumadin). It also metabolizes phenytoin (an anti-seizure drug), NSAIDs such as celecoxib, ibuprofen, and flurbiprofen, several sulfonylureas used for type 2 diabetes, losartan, and the multiple-sclerosis drug siponimod.
Because these drugs are inactivated by CYP2C9, a reduced-function genotype means the parent drug is cleared slowly and accumulates. For warfarin that translates into a lower dose requirement and a higher early bleeding risk; for phenytoin, a risk of dose-dependent neurotoxicity; for NSAIDs, greater exposure and gastrointestinal risk.
Warfarin is special because two genes act together: CYP2C9 controls how fast you clear the drug, while VKORC1 — the gene encoding warfarin's target enzyme — controls how sensitive you are to it. CPIC dosing algorithms combine CYP2C9 star alleles, the VKORC1 -1639 variant, and sometimes CYP4F2, alongside age and weight, to estimate a starting dose.
Star alleles, not single SNPs.
CYP2C9 is described using star-allele haplotype nomenclature. The most clinically relevant alleles:
Adding the two alleles' activity gives a phenotype:
- Normal metabolizer (NM) — two normal-function alleles (*1/*1).
- Intermediate metabolizer (IM) — one reduced- or no-function allele (e.g. *1/*2, *1/*3).
- Poor metabolizer (PM) — two reduced- or no-function alleles (e.g. *2/*3, *3/*3); the lowest warfarin dose requirement and highest early bleeding risk.
The textbook pharmacogenomics gene.
CYP2C9 has formal dosing guidelines from the Clinical Pharmacogenetics Implementation Consortium (CPIC) and the Dutch Pharmacogenetics Working Group (DPWG), and the FDA warfarin label includes a pharmacogenomic dosing table. Examples:
- Warfarin: CPIC's guideline combines CYP2C9 genotype with VKORC1 (and CYP4F2) to recommend a starting dose. Poor metabolizers and VKORC1-sensitive patients need markedly lower doses; getting the initial dose closer to the maintenance dose reduces time out of therapeutic range and early bleeding events.
- Phenytoin: CPIC recommends lower starting doses for intermediate and poor metabolizers (combined with HLA-B*15:02 screening for skin-reaction risk).
- NSAIDs (celecoxib, ibuprofen, flurbiprofen, meloxicam): CPIC suggests reduced doses or alternatives for poor metabolizers to limit exposure and gastrointestinal risk.
- Siponimod: the label contraindicates the drug in CYP2C9 *3/*3 poor metabolizers.
"Warfarin is the drug that put pharmacogenomics in the clinic. CYP2C9 tells you how fast you clear it; VKORC1 tells you how sensitive you are — and you need both to predict a dose."
Primary sources: CPIC — CYP2C9 gene page; Johnson et al., Clin Pharmacol Ther 2017 (CPIC warfarin guideline); PharmGKB CYP2C9; PharmVar CYP2C9.
What consumer arrays do — and don't — tell you about CYP2C9.
CYP2C9 is relatively array-friendly: the two main European alleles are tagged by single, commonly genotyped SNPs, and warfarin's partner variant in VKORC1 is often genotyped too. What you can usually look up:
- rs1799853 (*2 marker) — usually genotyped. A T allele indicates *2.
- rs1057910 (*3 marker) — usually genotyped. A C allele indicates the stronger-effect *3.
- rs9923231 (VKORC1 -1639G>A) — often genotyped; the main driver of warfarin sensitivity, complementary to CYP2C9.
The limits matter, and they are not symmetric across ancestries: consumer chips frequently miss the *5, *6, *8, and *11 alleles that are important in people of African ancestry, so a "normal" raw-data readout can be falsely reassuring. And a raw file is never a substitute for INR monitoring. For anyone actually on or starting warfarin, dosing is a clinical decision made with lab monitoring and a validated algorithm — your raw data is context, not a prescription.