Pharmacokinetic and pharmacodynamic re-evaluation of a genetic-guided warfarin trial

Abstract

Purpose

A previous trial failed to demonstrate the superiority of a demographic-genetic algorithm in predicting warfarin (W) dose over a standard clinical approach. The purpose of the present study is to re-analyse the results in subgroups of patients with differing baseline sensitivity to W, integrated with additional pharmacokinetic data.

Methods

The original trial allocated 180 treatment-naïve patients with non-valvular atrial fibrillation to a control arm (CTL, n = 92) or a genetic-guided arm (GEN, n = 88). Before starting anticoagulation treatment, all patients were genotyped for CYP2C9, VKORC1 and CYP4F2 variants and classified into four quartiles (Q1, Q2, Q3, Q4) according to the algorithm-predicted W maintenance dose. International normalised ratios (INR) and plasma concentrations of S-warfarin [S-W]s and R-warfarin [R-W]s were measured at baseline and on days 5, 7, 9, 12, 15 and 19 of therapy.

Results

In the lowest dose quartile (Q1), the number of INRs > 3 and mean INR values on days 5 and 7 were significantly higher in CTL than in GEN. In Q3 and Q4, the mean INR values reached therapeutic level (> 2) 2 days later in CTL than in GEN. During follow-up, the mean time courses of INRs and [S-W]s in GEN were remarkably stable in all dose quartiles. Thus, mean changes from starting to final doses were significantly smaller in GEN than in CTL. Plasma concentrations of R-W (a partially active enantiomer) steadily increased from day 5 to day 19 in all Qs in both CTL and GEN, except in the Q1 CTL group, due to the marked dose reduction required.

Conclusions

This analysis showed that the demographic-genetic algorithm used to predict the W dose can identify patients with differing degrees of sensitivity to W and to ‘normalise’ their average anticoagulant responses. The progressive rise in [R-W]s throughout the 19-day follow-up indicates that the (partial) contribution of R-W to the W anticoagulant effect changes continually during the early phase of treatment.

Appendix

Loading dose algorithm

The personalised loading dose (LD) was calculated with the standard formula for the one-compartment pharmacokinetic model:

$$ \mathrm{LD}\ \left(\mathrm{mg}\right)=\mathrm{volume}\ \mathrm{of}\ \mathrm{distribution}\ \left(\mathrm{L}\right)\times \mathrm{target}\ \mathrm{steady}\hbox{-} \mathrm{state}\ \mathrm{plasma}\ \mathrm{concentration}\ \left(\mathrm{mg}/\mathrm{L}\right). $$

The volume of distribution was estimated according to body weight (0.14 L/Kg) [22].

The ideal steady-state plasma concentration (Css) target, which should theoretically produce an INR in the centre of the therapeutic range (2.5), was calculated on the basis of the VKORC1 genotype according to a PK-PD model described elsewhere [23], as follows:

• (a) 0.25 mg/mL for VKORC1 AA;

• (b) 0.34 mg/mL for VKORC1 GA;

• (c) 0.52 mg/mL for VKORC1 GG.

Maintenance dose algorithm

The estimated weekly maintenance dose (MD) was [11]:

$$ \mathrm{MD}\ \left(\mathrm{mg}/\mathrm{week}\right)={\left[7.39764\hbox{--} \left(0.02734\times \mathrm{age}\right)+\left(1.06287\times \mathrm{BSA}\right)\hbox{--} \left(1.04468\ \mathrm{for}\ VKORC1\ AG\right)\hbox{--} \left(2.12117\ \mathrm{for}\ VKORC1\ AA\right)\hbox{--} \left(0.78983\ \mathrm{for}\ CYP 2C{9}^{\ast }{1}^{\ast } 2\right)\hbox{--} \left(1.17138\ \mathrm{for}\ CYP 2{C9}^{\ast }{1}^{\ast } 3\right)\hbox{--} \left(1.81292\ \mathrm{for}\ CYP 2{C9}^{\ast }{2}^{\ast } 2\ {\mathrm{or}}^{\ast }{2}^{\ast } 3\ {\mathrm{or}}^{\ast }{3}^{\ast } 3\right)\hbox{--} \left(0.46723\ \mathrm{for}\ CYP 4{F2}^{\ast }{1}^{\ast } 3\right)\hbox{--} \left(0.71528\ \mathrm{for}\ CYP 4{F2}^{\ast }{1}^{\ast } 1\right)\right]}^2 $$

where age is entered as years; BSA (body surface area) is calculated as weight(kg)^0.425 × height(cm)^0.725/139.2. VKORC1 AG, VKORC1 AA, CYP2C9*1*2, CYP2C9 *1*3, CYP2C9 *2*2 or *2*3 or *3*3, CYP4F2 *1*3 and CYP4F2 *1*1 are coded 0 if absent and 1 if present.