Forecasting of Blood Tacrolimus Concentrations Based on the Bayesian Method in Adult Patients Receiving Living-Donor Liver Transplantation

Objective: To evaluate Bayesian prediction of blood tacrolimus concentrations in adult patients receiving living-donor liver transplantation (LDLT) using previously obtained population pharmacokinetic parameters.

Patients and methods: Data were retrospectively collected from 47 adult patients receiving LDLT who were not included in the estimation of population pharmacokinetic parameters. Blood tacrolimus concentrations were predicted without or with the empirical Bayesian method using sparse samples obtained in the previous week. Predictive performance of the concentrations was evaluated by the mean prediction error (ME), mean absolute prediction error (MAE) and root mean square error (RMSE) as well as the percentage of successful predictions (percentage of absolute prediction error less than 3 μg/L, %PRED3).

Results: Concentrations predicted by the population mean pharmacokinetic parameter values coincided well with observed concentrations during the period of tacrolimus infusion immediately after the operation. For concentrations during subsequent oral therapy with tacrolimus, predictability by the population mean pharmacokinetic parameter values alone was not satisfactory. Bayesian forecasting using one or two blood concentrations obtained in the previous week significantly decreased (p < 0.05) MAE and RMSE compared with predictions based on the population mean pharmacokinetic parameters on postoperative days 21 and 28, but not on day 14. During postoperative days 15–21, %PRED3 was increased to 68.6% or 71.2% with the Bayesian method using one or two blood concentrations, respectively, from 44.9% with the population mean pharmacokinetic parameter values.

Conclusion: The present study demonstrated the applicability of the Bayesian method with use of one or two samples for prediction of blood tacrolimus concentrations in adult patients receiving LDLT.