Internal and external evaluation of a K-PD model for warfarin using prediction corrected visual predictive check (PC-VPC)
AK Hamberg (1), M Wadelius (1), M Pirmohamed (2), EN Jonsson (3)(4)
(1) Department of Medical Sciences, Clinical Pharmacology, Uppsala University Hospital, Uppsala, Sweden (2) Department of Pharmacology and Therapeutics, University of Liverpool, UK (3) Department of Pharmaceutical Biosciences, Division of Pharmacokinetics and Drug Therapy, Uppsala University, Sweden (4) Exprimo NV, Mechelen, Belgium
Objectives: To evaluate the performance of a warfarin K-PD model using PC-VPCs in both an internal [1] and [2] external dataset.
Background: Warfarin therapy is challenging due to its narrow therapeutic range and pronounced variability in individual dose requirements. Variability in both PD and PK contribute to the more than 20-fold difference in maintenance dose. We have developed a population model for the relationship between warfarin dose and anticoagulant response (INR) using data from a Swedish study with patients starting on warfarin (n=1426) [1]. The model includes age and genetic variations in CYP2C9 and VKORC1 as covariates, and together these factors explain more than a 10-fold difference in dose. Visual predictive checks (VPCs) are rapidly becoming an important diagnostic tool for model evaluation [3]. A prediction corrected VPC (PC-VPC) is an adaptation of the standard VPC, which is more suited for data collected in studies with adaptive design [4], as for warfarin where the dose is individually adjusted based on anticoagulant response.
Methods: The final model and parameter estimates obtained from the analyses of the Swedish study [1] were used in the internal and external model evaluation. PC-VPCs were constructed with median, 5th and 95th percentiles for the observed data. Model predictions were based on 100 simulated datasets and presented as non-parametric 95% confidence intervals for the median, 5th and 95th percentiles. The procedure was repeated on an external dataset derived from a British study of patients starting warfarin therapy.
Results: The PC-VPCs based on internal data did not indicate any major differences between observations and model predictions. Preliminary results suggest that the model was not able to describe the external dataset appropriately.
Conclusions: Preliminary results suggest a difference in the dose-response relationship for warfarin between Swedish and British patients. Reasons for this may for example be differences in INR methods, demographics, compliance patterns, and/or vitamin K intake between the two countries.
Acknowledgement:The authors would like to acknowledge Andrea Jorgensen and Steven Lane at The University of Liverpool for facilitating access to the British data and Prof. Anders Rane and Jonatan Lindh for access to data from the Swedish study. Anna-Karin Hamberg is financially supported by a personal grant from the Ränk family via the Swedish Heart and Lung foundation.
References:
[1] A.K Hamberg et al. A pharmacometric model describing the relationship between warfarin dose and INR response taking variations in CYP2C9, VKORC1 and age into account. Accepted for publication in Clin Pharmacol Ther
[2] A.L Jorgensen et al, Genetic and environmental factors determining clinical outcomes and cost of warfarin therapy: a prospective study. Pharmacogenetics and Genomics 2009; 19: 800-812
[3] M.O Karlsson, N Holford. A Tutorial on Visual Predictive Checks. PAGE 17 (2008) Abstract 1434 [www.page-meeting.org/?abstract=1434]
[4] M Bergstrand, A.C Hooker, J.E Wallin, M.O Karlsson. Prediction Corrected Visual Predictive Checks. ACoP (2009) Abstract F7. [http://www.go-acop.org/sites/all/assets/webform/Poster_ACoP_VPC_091002_two_page.pdf]
