Abstract
Abstract 189
In frail elderly patients, the low warfarin maintenance dose requirements and high risk of thrombosis and bleeding raise specific challenges, especially at treatment initiation. Because of a narrow therapeutic index and a marked interindividual variability in dosage requirements, warfarin induction doses must be tailored to individual and disease-specific factors.
The aim of our multicenter study was to investigate whether VKORC1 and CYP2C9 genotypes helped to predict the warfarin maintenance dose when added to demographic, clinical data and INR values prospectively collected at baseline and during warfarin induction. In a cohort of elderly inpatients, we developed clinical and pharmacogenetic models and evaluated their accuracy in predicting the warfarin maintenance dose comparatively to the accuracy of a dosing algorithm based solely on INR values. The derivation sample consisted in 115 Caucasian inpatients (mean age, 86 years), all initiated using the same warfarin induction protocol designed for the elderly (Siguret, Am J Med 2005, Gouin-Thibault J Am Geriatr Soc 2010): INR was measured at baseline, the day after three 4-mg warfarin intake (Day 3) (INR3) and on Day 6±1: their values allowed to adjust the dose according to the algorithm. The actual daily warfarin maintenance dose was defined as the amount of warfarin required to achieve a stable INR in the 2.0–3.0 range in two consecutive samples at least 48–72 h apart, in the absence of dosage changes within the previous 4 days.
At baseline, the clinical model failed to accurately predict the maintenance dose (R2 <10%). Adding the VKORC1 and CYP2C9 genotypes to the model increased R2 to 31%, indicating that genetic factors were the main determinants of the maintenance dose in our population before warfarin initiation (Table). On Day 3, the predictive information provided by the VKORC1 genotype was completely embedded in the INR3, whereas the CYP2C9 genotype remained (albeit slightly) a significant predictor (Table). After 6±1 days, neither genotype correlated with the warfarin dose (Table). Finally, the maintenance dose was safely predicted by our simple dosing-algorithm solely based on INR3 (R2 0.77) and INR6±1 (R2 0.81), without genetic information: it underestimated the dose by more than 1 mg in fewer than 10% of patients and overestimated the dose by more than 1 mg in fewer than 2%. All clinical models were validated in an independent sample of 55 elderly inpatients, in whom warfarin therapy was initiated using the same dosing algorithm. In 2007, the US FDA added to the warfarin labeling information consideration of the VKORC1 and CYP2C9 genotypes for dose determination. However, our results do not support the routine prospective use of genetically guided dosing in elderly inpatients starting warfarin therapy. Our simple dosing algorithm, which is inexpensive and widely applicable, safely and accurately predicts the warfarin maintenance dose in elderly inpatients at treatment initiation without requiring genetic information.
Clinical and pharmacogenetic regression models for predicting daily warfarin maintenance doses in the derivation sample
| Clinical Models | Pharmacogenetic Models | ||||
|---|---|---|---|---|---|
| Models (M) | Variable | Final model P value | Models (M-G) | Variable | Final model P value |
| M0 | INR0 | 0.0091 | M0-G | Age | 0.0119 |
| Indication (a) | 0.0222 | ||||
| INR0 | 0.0479 | ||||
| CYP2C9, per variant allele (b) | 0.0174 | ||||
| VKORC1 –1639A, per A allele (c) | <0.0001 | ||||
| R2 0.06 | R2 0.31 | ||||
| M3 | Age | 0.0243 | M3-G | Age | 0.0234 |
| Indication (a) | 0.0229 | Indication (a) | 0.0190 | ||
| INR3 | <0.0001 | INR3 | <0.0001 | ||
| CYP2C9, per variant allele (b) | 0.0267 | ||||
| VKORC1 –1639A, per A allele (c) | NS | ||||
| R2 0.52 | R2 0.55 | ||||
| M6 | Age | 0.0313 | M6-G | Age | 0.0313 |
| INR6±1 | <0.0001 | INR6±1 | <0.0001 | ||
| ΔDose6±1 | <0.0001 | ΔDose6±1 | <0.0001 | ||
| INR6±1/ΔDose6±1 | 0.0011 | INR6±1/ΔDose6±1 | 0.0011 | ||
| CYP2C9, per variant allele (b) | NS | ||||
| VKORC1 –1639A, per A allele (c) | NS | ||||
| R2 0.80 | R2 0.80 | ||||
| Clinical Models | Pharmacogenetic Models | ||||
|---|---|---|---|---|---|
| Models (M) | Variable | Final model P value | Models (M-G) | Variable | Final model P value |
| M0 | INR0 | 0.0091 | M0-G | Age | 0.0119 |
| Indication (a) | 0.0222 | ||||
| INR0 | 0.0479 | ||||
| CYP2C9, per variant allele (b) | 0.0174 | ||||
| VKORC1 –1639A, per A allele (c) | <0.0001 | ||||
| R2 0.06 | R2 0.31 | ||||
| M3 | Age | 0.0243 | M3-G | Age | 0.0234 |
| Indication (a) | 0.0229 | Indication (a) | 0.0190 | ||
| INR3 | <0.0001 | INR3 | <0.0001 | ||
| CYP2C9, per variant allele (b) | 0.0267 | ||||
| VKORC1 –1639A, per A allele (c) | NS | ||||
| R2 0.52 | R2 0.55 | ||||
| M6 | Age | 0.0313 | M6-G | Age | 0.0313 |
| INR6±1 | <0.0001 | INR6±1 | <0.0001 | ||
| ΔDose6±1 | <0.0001 | ΔDose6±1 | <0.0001 | ||
| INR6±1/ΔDose6±1 | 0.0011 | INR6±1/ΔDose6±1 | 0.0011 | ||
| CYP2C9, per variant allele (b) | NS | ||||
| VKORC1 –1639A, per A allele (c) | NS | ||||
| R2 0.80 | R2 0.80 | ||||
(a) indication for warfarin: 0 for venous and 1 for arterial thromboembolic disease; (b) CYP2C9 coded 0 (wild-type), 1 (CYP2C9 *2 or CYP2C9 *3 variant allele), or 2 (2 variant alleles); (c)VKORC1 coded 0 (wild-type GG); ΔDose6±1, cumulated dose (mg) between Day 0 and the day on which INR6±1 was measured.
No relevant conflicts of interest to declare.
This icon denotes an abstract that is clinically relevant.
This feature is available to Subscribers Only
Sign In or Create an Account Close Modal