Bayesian forecasting and prediction of tacrolimus concentrations in pediatric liver and adult renal transplant recipients

Aim: To test the predictive capacity of two recently derived population pharmacokinetic models and the usefulness of Bayesian forecasting to predict tacrolimus blood concentrations in pediatric liver and adult kidney transplant recipients. Materials and Methods: New databases were added to the Abbottbase PKS (Bayesian dosage prediction) program to incorporate the population pharmacokinetic models developed for tacrolimus. Two independent populations of transplant recipients were used to predict tacrolimus trough concentrations. Pharmacokinetic, demographic, and covariate data were collected from patient records. Different time weighting factors were tested (1, 1.005, 1.01) and the influence of excluding data collected in the first 5 day, post-transplant examined. Concentrations were predicted until the 10(th) tacrolimus measurement. Actual tacrolimus concentrations were compared with those predicted by the PKS pro.-ram and bias and precision determined. Results: Tacrolimus concentrations predicted by the PKS pro-ram were, on average, C C unbiased for the pediatric liver population, but were over-predicted (9%) for the adult renal population. In both populations predictions were not precise (imprecision ranged from 39 to 50%). Conclusions: Due to the imprecision seen in this study, these models could not be used in clinical practice in the immediate post-transplant period. Poor precision may be due to reliance on routine drug monitoring data alone, difficulties with expression of covariates in continuous modeling relationships in the PKS program, lack of accurate quantitative measures of liver function, or large, random intraindividual variability in the bioavailability of tacrolimus.