Estimate a warfarin starting dose in mg per week and per day from age, height, weight, race and interacting drugs, with optional CYP2C9 and VKORC1 genes.

Use this when no CYP2C9 or VKORC1 test result is available. It uses age, body size, race and two drug interactions.

The model uses the decade of age: 60 to 69 counts as 6. Ages 90 and over count as 9.

Self-reported race, in the four groups the model was built with.

These enzyme inducers raise the dose needed. Rifampicin is the same drug as rifampin.

Amiodarone lowers the dose needed.

Optional settings

Calculator Functions

The calculator estimates the steady warfarin dose a person is likely to need for a target INR of 2 to 3, in mg per week and as an average in mg per day. It is written for prescribers and students. Select a function to open it in the calculator.

FunctionUse it whenYou enter
Clinical estimate (no genotype)No CYP2C9 or VKORC1 test result is availableAge, height, weight, race, and whether the person takes an enzyme inducer or amiodarone
Genotype-guided estimateA CYP2C9 or VKORC1 result is availableThe same details plus the VKORC1 and CYP2C9 genotypes, either of which can stay at Not known

Height can be entered in inches or centimeters and weight in pounds or kilograms. Race starts at Mixed or not specified and both drug questions start at No. The result also lists the age group, height and weight the model used. To apply a prescribed percentage change to a dose already being taken, use the Coumadin Adjustment Calculator.

Warfarin Starting Dose Formulas

W = R² [1]
D = W ÷ 7
WEstimated stable warfarin dose (mg per week)
DAverage dose (mg per day)
RSquare root of the weekly dose, which is what the model predicts. It is the sum of the terms in the function below
AAge in decades, rounded down: 1 for ages 10 to 19, 6 for ages 60 to 69, and 9 for 90 and over
HHeight (cm)
MBody weight (kg)

Both functions come from the International Warfarin Pharmacogenetics Consortium (IWPC) model [1]. A starting dose is chosen from the expected maintenance dose [3], which is why an estimate of the steady dose is used when warfarin is started.

Clinical estimate (no genotype)

R = 4.0376 − 0.2546 × A
+ 0.0118 × H + 0.0134 × M
+ race term + drug terms [1]

The race and drug terms are in the Clinical column of the table below. Each one is added only when it applies.

Genotype-guided estimate

R = 5.6044 − 0.2614 × A
+ 0.0087 × H + 0.0128 × M
+ gene terms + race term
+ drug terms [1]

VKORC1 means the −1639G>A variant (rs9923231), and CYP2C9 covers the *2 and *3 variants. A gene left at Not known has its own term, so this function still works with one result.

TermClinicalGenotype-guided
Race: White00
Race: Asian−0.6752−0.1092
Race: Black or African American+0.4060−0.2760
Race: mixed or not specified+0.0443−0.1032
Taking carbamazepine, phenytoin or rifampin+1.2799+1.1816
Taking amiodarone−0.5695−0.5503
VKORC1 G/GNot used0
VKORC1 A/GNot used−0.8677
VKORC1 A/ANot used−1.6974
VKORC1 not knownNot used−0.4854
CYP2C9 *1/*1Not used0
CYP2C9 *1/*2Not used−0.5211
CYP2C9 *1/*3Not used−0.9357
CYP2C9 *2/*2Not used−1.0616
CYP2C9 *2/*3Not used−1.9206
CYP2C9 *3/*3Not used−2.3312
CYP2C9 not knownNot used−0.2188
Terms added to R in each function [1]

Limits of This Estimate

  • The result is a starting point for a prescriber, not a dose to take. Warfarin dosing must be individualized from each person’s INR response, with the INR checked daily after the first dose until it is stable [3]. Nobody should start or change warfarin from this number.
  • The model was built from 5,052 patients with a target INR of 2 to 3. On patients it had not seen, its average miss was 8.5 mg per week with genotypes and 9.9 mg per week without, a little over 1 mg a day [2].
  • It is less predictable for people who need more than 70 mg per week, and it does not account for vitamin K intake, smoking or interacting drugs other than the four listed [2].
  • When genotypes are not known, the U.S. label gives a usual starting dose of 2 to 5 mg once daily, advises lower doses for elderly or debilitated patients, and does not recommend routine loading doses [3]. An estimate far outside that range deserves a second look at the inputs.

Sources