Warfarin: two genes, an algorithm, and still a blood test
Warfarin is the drug that taught medicine to expect genetic variation in dosing, and it is also the clearest example of why a genotype is an input rather than an answer.
Also sold as: Marevan, Coumadin.
What warfarin is
Warfarin is an anticoagulant, taken daily to prevent clots in atrial fibrillation, after some heart-valve surgery and after venous thrombosis. It works by blocking vitamin K recycling, which the liver needs to make several clotting factors. Its dose is not fixed: it is titrated against a blood test called the INR until the blood clots at the intended rate, and it stays under monitoring for as long as the drug is taken.
Why your genes come into it
Two genes carry most of the inherited variation, and they act at different points. CYP2C9 clears warfarin out of the body, so reduced function means the drug lingers and a lower daily dose produces the same effect. VKORC1 codes for the target that warfarin blocks, and common variation near it changes how sensitive that target is, so some people respond strongly to a small dose. The published dosing algorithms combine both genes with age, weight, other medications and the reason for treatment. That combination is the point. Neither gene on its own predicts a dose, and neither replaces the INR.
What published guidance says
The rows below restate CPIC's published guidance for each possible CYP2C9 result. Note what they do not do: none of them names a milligram figure, because the guideline routes through an algorithm rather than a lookup. Warfarin dosing is set by a prescriber against INR measurements, and a genotype changes the starting estimate rather than the destination.
CYP2C9
| If your CYP2C9 result is | What that means | Published CPIC guidance |
|---|---|---|
| Normal metabolizer | Normal warfarin clearance. | CPIC: warfarin dosing follows the genotype-guided algorithm together with VKORC1 and clinical factors; no CYP2C9-driven adjustment beyond it. |
| Intermediate metabolizer | Slower clearance generally predicts a lower stable dose and a higher early-bleeding risk. | CPIC: the genotype-guided warfarin algorithm (CYP2C9 + VKORC1 + clinical factors) applies; reduced function typically lowers the predicted dose. |
| Poor metabolizer | Substantially slower clearance; a notably lower dose is often required and bleeding risk is higher. | CPIC: the genotype-guided warfarin algorithm applies; poor metabolism predicts a markedly lower dose and closer INR monitoring. |
Can your DNA file answer this?
Both warfarin genes are readable from a plain consumer file, which is unusual. The two common CYP2C9 reduced-function alleles, *2 and *3, are single positions that consumer arrays commonly carry, and VKORC1's key position is a single common variant too. There is an asymmetry worth stating: the table above is CYP2C9 only, because CPIC's pinned release has no standalone diplotype-to-phenotype table for VKORC1 that Aimosti can restate, so the report reads VKORC1 as a genotype with a warfarin-sensitivity interpretation rather than as its own drug-guidance table. Rarer CYP2C9 alleles, several of which matter more in African and Asian populations, are not on consumer arrays and will read as reference.
A genotyping chip reads only a few hundred thousand pre-selected positions, about 0.02% of your genome, chosen for common variation. It does not sequence the rest, so it cannot find the rare or novel pathogenic variants the clinical and carrier modules look for: a “no finding” from chip data means “this chip never looked”, far more so than with whole-genome sequencing. Chip data suits common-variant traits, pharmacogenomic tag SNPs, and haplogroup ancestry, not clinical or carrier screening.
Common questions
Can I work out my warfarin dose from my genotype?
No. The genotype-guided algorithms produce a starting estimate that still has to be checked and corrected against INR measurements over the first weeks. A dose derived from a page like this and acted on without monitoring would be dangerous in both directions.
I am already stable on warfarin. Is this useful?
Less so, and that is worth saying plainly. Genotype-guided dosing earns its keep at initiation, when nobody yet knows your stable dose and the early bleeding risk is highest. Once you have been stable for months, your own INR history is a better guide than any prediction, because it reflects everything the algorithm approximates.
I take apixaban or rivaroxaban instead. Does this apply to me?
No. The direct oral anticoagulants are not dosed on CYP2C9 or VKORC1 and are not covered by this guidance. If you were switched off warfarin, this page is history rather than instruction.
Why does VKORC1 not have its own table here?
Because CPIC's pinned release does not give it a standalone diplotype-to-phenotype table that can be restated the way CYP2C9's can. VKORC1 is genuinely important for warfarin and it is read in the report, but its contribution enters through the dosing algorithm rather than as its own set of recommendations. Showing an empty table would look more complete and be less true.