[Cyclosporin pharmacokinetic modeling in renal transplant patients]

To characterize the pharmacokinetic behavior of oral cyclosporin (CsA) in renal transplant patient, based on through blood concentration (C0) value, and to develop and to evaluate a Bayesian method for the individualized adjustment of CsA daily dose (DD).

Sixty-seven renal allograft recipients (42 men and 25 women) who had been treated with CsA (Sandimmun Neoral) associated with mycophenolate mofetil (2g daily) and prednisone (0,5-1 mg/kg daily) were randomly divided into two groups. Group A (N=48) was used to characterize CsA pharmacokinetic behavior and Group B (N=19) to evaluate Bayesian predictive performance for the model developed. We evaluated different structural models using non linear mixed effects modeling implemented in the NONMEN computer program in order to quantify the relationship between DD and C0. Accuracy and precision were evaluated by the mean standardized prediction error and its standard deviation.

The Michaelis-Menten model was found to be optimum for quantifying the relationship between DD and C0. This model includes time-dependent parameters such as the Michaelis-Menten constant (Km) and daily maximum dose (Dmax) as well as first order autoregressive terms DD and C0 included in the structural model in an additive way. In the final model, the Dmax parameter is affected by plasmatic urea values and shows a half-life stabilization time of 90.90 days (95% CI: 52.60 to 250 days). Plasmatic urea values of 50 mg/dL are related to an initial Dmax value of 3 mg/kg daily (95% CI: 1.81 to 4.19 mg/kg daily) which decreases exponentially throughout the post-transplant period until it reaches a constant value of 2.16 mg/kg daily (95% CI: 1.41 to 2.91 mg/kg daily) In the same way, the Km parameter presents a central tendency value of 93.60 ng/mL (95% CI: 28.60 to 158.60 ng/mL) and the half-life necessary for its stabilization is 12.70 days (95% CI: 9.80 to 17.90 days). The residual variability of the model is 8.2%. The mean value of standardized prediction errors for populations and its standard deviation, as well as its confidence intervals of 95%, confirm the appropriate accuracy and precision of both a priori and a posteriori predictions with this model. Also, it reached between 70 and 100% a posteriori sequential predictions with prediction errors below 10%.

The characterization of the pharmacokinetic behavior of CsA requires us to consider parameters such as Dmax and Km as non lineal functions of time, while the first order autoregressive terms DD and C0 must also be incorporated into the model.