What is the Gage warfarin algorithm?

The Gage algorithm (Gage et al, Clinical Pharmacology & Therapeutics, 2008) is a validated pharmacogenetic model that predicts a patient's warfarin maintenance dose. It combines body surface area, age, CYP2C9 and VKORC1 genotype, and interacting factors such as amiodarone, smoking status and indication to estimate the weekly dose.

Do I need genetic testing to use it?

No. Genotype fields are optional. If CYP2C9 or VKORC1 results are unavailable, leave them as wild type or unknown and the estimate uses clinical factors alone. Genetic data, when present, refines the prediction because both genes strongly influence warfarin sensitivity.

Why is body surface area used instead of weight alone?

The Gage model uses Mosteller body surface area, calculated as the square root of (height in cm times weight in kg, divided by 3600). Surface area correlates better with warfarin volume of distribution than weight alone, so the algorithm uses it as the size term.

Is this a loading dose?

No. The output is an estimated maintenance dose, not an initial loading regimen. Loading typically follows a separate protocol with frequent INR checks. Always titrate the final dose to the patient's measured INR rather than relying on any prediction.

Can I use this for a patient already on warfarin?

It is intended for dose initiation. For a patient already stabilised, the best predictor of future dose is their current dose plus INR trend. Use this estimator mainly when starting therapy or as a sanity check against a markedly different prescribed dose.

Warfarin Dose Estimator (Gage Algorithm)

Warfarin starting-dose estimator

Warfarin has a narrow therapeutic window and huge between-patient variability — the required maintenance dose can differ more than tenfold. This tool estimates a sensible starting dose using the Gage 2008 pharmacogenetic algorithm, which combines patient size, age, two key genotypes and common interacting factors. It is designed for anticoagulation-clinic pharmacists and prescribers who want a defensible initial estimate before INR titration.

How it works

The algorithm predicts the natural log of the weekly dose, then exponentiates it. The size term uses Mosteller body surface area:

BSA (m²) = sqrt( height_cm × weight_kg / 3600 )

Each factor shifts the predicted log-dose up or down. Two genes dominate:

CYP2C9 metabolises warfarin. The *2 and especially *3 variants slow clearance, so carriers need lower doses — each *3 allele subtracts from the estimate.
VKORC1 (the −1639 G>A promoter variant) sets the drug target’s expression. The A allele lowers the dose requirement, so AG and AA genotypes pull the dose down.

Clinical terms then adjust further: amiodarone and a DVT/PE indication reduce the dose, while smoking and a high target INR raise it. The result is divided by seven to give an equivalent daily dose.

Notes and example

For a 65-year-old, 170 cm, 75 kg patient with wild-type genotype and no interacting drugs, the model estimates roughly 35 mg/week (about 5 mg/day). Adding VKORC1 AA and CYP2C9 *3/*3 would cut that estimate sharply, reflecting a slow metaboliser with a sensitive target.

Always round to available tablet strengths, confirm against local anticoagulation policy, and titrate to the measured INR. This is a decision-support estimate, not a prescription, and it should never replace clinical judgement or INR monitoring.