Multivariate and Population Pharmacokinetic Analyses for Tacrolimus Area Under The Concentration-Time Curve Prediction in De Novo Renal Transplant Recipients
Musuamba F(1), Mourad M(2), Haufroid V(3), Wallemacq P(4), Verbeeck RK(1)
(1)UCL, Brussels, Belgium, (2) Division of abdominal Transplantation, Cliniques Universitaires St Luc, Belgium, (3)Laboratory of Clinical Chemistry, Cliniques Universitaires St Luc, Belgium.
Objectives: Tacrolimus pharmacokinetics are characterised by a very high variability in the early period after transplantation. The aim of the study was to identify pathophysiological and biochemical determinants of AUC12 (tacrolimus area under the concentration-time curve during one dosing interval) in the early period after renal transplantation and to predict this parameter based on a minimum number of blood samples.
Methods: Data from 63 renal transplant recipients were analyzed. They were immunosuppressed with tacrolimus, mycophenolic acid and methylprednisolone. From day 1 to day 15 post-transplantation, a total of 2184 samples, including daily Cmin levels and a full kinetic profile on day 15, were analyzed for tacrolimus by immunoassay. A stepwise multiple linear regression analysis was performed to predict AUC12 based on a limited number (two) of early (0 to 3 hours) blood concentrations. In addition to timed blood concentrations (C0-C3), the following parameters were checked as covariates in the modelling process: weight, age, sex, weight normalised doses, serum bilirubin, liver enzymes (ASAT, ALAT, GGT, PAL), serum creatinine, and creatinine clearance (CLcr). ABCB1 phenotypes (Rhodamine test) and CYP3A5 genotypes obtained from the graft donors and recipients were also analysed as possible covariates. Linear regression analysis and an agreement test (Bland and Altman, Lancet 1, 307-310, 1981) were used to assess the reliability of the tested models, comparing predicted AUC12 with AUC12 computed by the trapezoidal rule.
Results: The following expression AUC12 = 25 + 8*C3 + 2*C 1.5 + 17 *CYP3A5*1 - 10*Cr - 0.6*CLcr was selected. In addition, data were described by a two compartment model with a weibull-type absorption model and a first order elimination model. A Bayesian estimator was developed after population pharmacokinetic analysis. Here are the predictive performance of both models :
Multivariate model: r2 = 0.78, mean relative prediction error = -9.4 and root mean squared error = 14.7.
Bayesian estimator: r2 = 0.85, mean relative prediction error = 8.9 and root mean squared error = 19.9.
Conclusions: Both models were judged reliable to accurately predict AUC12 from only 2 blood samples and therefore can be used for tacrolimus therapeutic drug monitoring.