II-32 Guangda Ma

Evaluating the influence of genotype on warfarin dose predictions made using a theory-based PKPD model.

Guangda Ma (1), Nick Holford (1), Jacqui Hannam (1), Jeff Harrison (1)

(1) The University of Auckland

Objectives: 

Warfarin continues to be the mainstay of oral anticoagulation therapy, however, a narrow therapeutic range poses a barrier to safe and effective therapy. Published evidence suggests that current methods to predict warfarin maintenance dose requirements are biased at the extremes [2]. In contrast, Bayesian dose forecasting using a theory-based warfarin PKPD model can achieve unbiased and precise dose predictions across a full range of clinical doses [3].

Despite the association between genotype and warfarin dose requirements, current evidence is not sufficiently robust to support clinical benefits of genotype-guided warfarin therapy [4-6]. Bayesian dose forecasting may overcome the need for prior genotype data [7]. A theory-based PKPD model for warfarin that accounts for the influence of genotype on warfarin PKPD can be used to evaluate this[8].

Objective 1: External Evaluation
Evaluate the predictive performance of the theory-based model against an external, clinically derived dataset and evaluate whether genotype knowledge influences model predictive performance.

Objective 2: Simulation-Estimation
Use simulation and estimation techniques to evaluate whether genotype knowledge influences the predictive performance and potential clinical utility of Bayesian forecasting using the theory-based model.

Methods:

Estimation and simulation was performed using NONMEM 7.4.1.

External Evaluation
The external evaluation dataset consisted of 138 individuals which has been previously used to evaluate an empirical warfarin dosing model [9]. The model was used to predict the maintenance dose needed to achieve the observed stable INR given the full dosing and INR history for each individual. The model predicted maintenance dose was compared with that clinically observed.

Simulation-Estimation

Bayesian dose individualisation using the model was evaluated using a simulation-estimation procedure. A virtual study population (n=1000) was created by sampling sex, weight, CYP4F2, CYP2C9, and VKORC1 genotypes as covariates. Following initial dose individualisation based upon each individuals’ covariates and population parameters (days 1-3), INR measurements on days 3, 7, 10, 14, 21, 28, 35, 42, 49, and 56 were then used successively to individualise daily warfarin doses.

The predictive performance of the model was evaluated using measures of bias (mean prediction error, ME) and imprecision (root mean square error, RMSE). Clinical utility was evaluated using the percentage of time within the therapeutic range (INR 2.0-3.0) during days 4-14 (TTR4-14), and 15-28 (TTR15-28).

Influence of Genotype
To evaluate the influence of genotype, the simulation-estimation and external evaluation investigations were conducted using a genotype-guided model and compared to a genotype-missing model.

Results:

External Evaluation

External evaluation of the genotype-guided and genotype-missing models was unbiased and precise over the actual dose range of 0.75-11 mg/day. Improvements in predictive performance following the addition of genotype knowledge were not apparent:

Genotype

ME

2.5%ile

97.5%ile

RMSE

Yes

0.115

-0.91

-0.53

0.58

Missing

0.08

-0.93

-0.56

0.54

Simulation-Estimation
Based on covariates alone the model predictions were initially biased and imprecise, however, unbiased and precise dose predictions across the simulated range of doses (0.77-27 mg/day) were achieved as time progressed and more INR measurements and dose adjustments were made:

INR and Doses Genotype ME 2.5%ile 97.5%ile RMSE
Day 3 (one INR & dose adjustment)   Yes -0.76 -7.01 1.8 2.42
  Missing -0.77 -7.55 2.09 2.62
Day 21 (five INRs & dose adjustments) Yes -0.15 -2.45 1.23 0.99
  Missing -0.16 -2.74 1.25 1.04
Day 56 (ten INRs & dose adjustments) Yes -0.02 -1.06 0.79 0.45
  Missing 0.0004 -1.11 0.80 0.51

Measures of predictive performance were similar for the genotype guided and genotype missing simulations across the entire simulated dose range.

 The time within the therapeutic range with genotype-guided dosing (TTR4-14: 29%; TTR15-28: 69%) was similar to genotype-missing dosing (TTR4-14: 30%; TTR15-28: 69%).

Conclusions:

Unbiased and precise warfarin dose predictions were achieved using the theory-based PKPD model based on external evaluation. Genotypes as covariates did not improve the predictions. The simulated use of genotype information is consistent with the small effects observed in clinical trials but without the bias associated with empirical dosing algorithms.

References:

  1. Sheiner, L.B., Computer-aided long-term anticoagulation therapy. Comput Biomed Res, 1969. 2(6): p. 507-18.
  2. Saffian, S.M., S.B. Duffull, and D.F.B. Wright, Warfarin Dosing Algorithms Underpredict Dose Requirements in Patients Requiring ≥7 mg Daily: A Systematic Review and Meta-analysis. Clinical Pharmacology & Therapeutics, 2017. 102(2): p. 297-304.
  3. Holford, N., G. Ma, and Y. Tsuji, Using biomarkers to predict the target dose of warfarin and linezolid. PAGE, 2018. 27 (2018) Abstr 8562 [www.page-meeting.org/?abstract=8562].
  4. Gage, B.F., et al., Effect of genotype-guided warfarin dosing on clinical events and anticoagulation control among patients undergoing hip or knee arthroplasty: The gift randomized clinical trial. JAMA, 2017. 318(12): p. 1115-1124.
  5. Stergiopoulos, K. and D.L. Brown, Genotype-guided vs clinical dosing of warfarin and its analogues: meta-analysis of randomized clinical trials. JAMA Intern Med, 2014. 174(8): p. 1330-8.
  6. Kimmel, S.E., et al., A pharmacogenetic versus a clinical algorithm for warfarin dosing. N Engl J Med, 2013. 369(24): p. 2283-93.
  7. Wright, D.F. and S.B. Duffull, A Bayesian dose-individualization method for warfarin. Clin Pharmacokinet, 2013. 52(1): p. 59-68.
  8. Xue, L., 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. Br J Clin Pharmacol, 2017. 83(4): p. 823-835.
  9. Saffian, S.M., et al., Influence of Genotype on Warfarin Maintenance Dose Predictions Produced Using a Bayesian Dose Individualization Tool. Ther Drug Monit, 2016. 38(6): p. 677-683.

Reference: PAGE 28 (2019) Abstr 9028 [www.page-meeting.org/?abstract=9028]

Poster: Clinical Applications