Prediction of blood cyclosporine concentrations in haematological patients with multidrug resistance by one-, two- and three-compartment models using Bayesian and non-linear least squares methods.

The blood cyclosporine (CsA) concentration-time profile in each of 24 adult haematological patients with multidrug resistance taking the first course of CsA treatment was fitted by one-, two- and three-compartment models to obtain relevant pharmacokinetic parameters. The pharmacokinetic parameters obtained were implemented into the PKS program (Abbottbase Pharmacokinetic System) as the population pharmacokinetic parameters used to predict blood CsA concentrations in adult haematological patients with multidrug resistance. The predictions of blood CsA concentrations by one-, two- and three-compartment models using the Bayesian method (BM) and the non-linear least squares method (NLLSM) were evaluated employing 11 patients who took the second course of CsA treatment. While the Akaike's information criterion (AIC) favoured the two-compartment model to describe CsA concentration-time profiles in patients taking the first and second courses of CsA treatment, the predictive performance analyses showed that both two- and three-compartment models were better than the one-compartment model for prediction, but the three-compartment model was slightly superior to the two-compartment model. The results also show that the predictions using BM were slightly better than those using NLLSM. Several factors affecting BM predictions and the possible difference among AIC, BM and predictive performance analyses were also addressed.