The Influence of Genotype on Warfarin Dose Predictions

Background & Aims: Warfarin is the most commonly prescribed oral anticoagulant worldwide, however, a narrow therapeutic range poses a barrier to safe and effective therapy. Common methods to predict warfarin dose requirements are biased at the extremes. When evaluated by simulation, Bayesian dose forecasting using a theory-based warfarin PKPD model achieves unbiased and precise dose predictions across the full range of clinical doses (1). Despite the association between genotype and warfarin dose requirements, current evidence is not sufficiently robust to support genotype-guided warfarin therapy. The theory-based PKPD model for warfarin accounts for the influence of genotype on warfarin PKPD (2).

Aim 1: evaluate the performance of the model against an external, clinically derived dataset (3).

Aim 2: evaluate the influence of genotype knowledge on the performance of warfarin dose predictions.

Methods: NONMEM 7.4.1 was used to simulate 1000 virtual patients by sampling sex, age, weight, CYP2C9, VKORC1 and CYP4F2 covariates before warfarin doses were individualised using a genotype-known or genotype-missing model. The model predicted maintenance dose was used to individualise doses on days 1-3, before INR measurements on days 3, 7, 10, 14, 21, 28, 35, 42, 49, and 56 were used to individualise daily doses to achieve an INR of 2.5.

The performance of genotype-known and genotype-missing dose individualisation were compared using measures of bias (mean prediction error, ME), imprecision (root mean square error, RMSE) and time within the therapeutic range (INR 2.0-3.0) during days 4-14 (TTR4-14) and days 15-28 (TTR15-28).

An external evaluation of the model was performed by using the model to predict the maintenance dose for 138 patients (data provided by Dr Dan Wright (New Zealand) and Dr Alison Thompson (Scotland)). The model predicted maintenance dose was compared with the clinically observed target dose.

Results: Measures of predictive performance were similar for the genotype guided (ME: -0.016 mg/day, 95% CI: -0.186, 0.155 mg/day; RMSE: 0.45 mg/day) and genotype missing simulations (ME: 0.0004 mg/day; 95% CI: -0.169, 0.17 mg/day; RMSE: 0.512 mg/day) over the simulated dose range of 0.77-27 mg/day.

Genotype-guided dosing (TTR4-14: 29%; TTR15-28: 69%) was not  clinically different from genotype-missing dosing (TTR4-14: 30%; TTR15-28: 69%).

External evaluation of warfarin was unbiased (ME: 0.12, 2.5th percentile: -0.91 mg/day, 97.5th percentile: 1.45 mg/day) and precise (RMSE: 0.58 mg/day) over the actual dose range of 0.75-11 mg/day.

Conclusion: Unbiased and precise warfarin dose predictions were achieved using the theory-based PKPD model in simulation, and external evaluation. The addition of genotype knowledge does not improve Bayesian dose forecasting using the theory-based PKPD model for warfarin. The minimal benefit for TTR4-14 when genotype information is used is consistent with the magnitude of effects observed in clinical trials but without the bias attributable to different dosing methods.


  1. Ma G, Holford N. Performance of a Theory-Based Mechanistic Model for Predicting the Target Dose of Warfarin.  Population Approach Group of Australia and New Zealand Conference 2018; Melbourne, Australia. 2018.
  2. Xue L, Holford N, Ding XL, Shen ZY, Huang CR, Zhang H, et al. Theory-based pharmacokinetics and pharmacodynamics of S- and R-warfarin and effects on international normalized ratio: influence of body size, composition and genotype in cardiac surgery patients. British Journal of Clinical Pharmacology. 2017;83(4):823-35.
  3. Saffian SM, Duffull SB, Roberts RL, Tait RC, Black L, Lund KA, et al. Influence of Genotype on Warfarin Maintenance Dose Predictions Produced Using a Bayesian Dose Individualization Tool. Therapeutic Drug Monitoring. 2016;38(6):677-83.

Guangda Ma

  • The University of Auckland