Tacrolimus pharmacokinetics in the early post-liver transplantation period and clinical applicability via Bayesian prediction

Abstract

Purpose: To define and validate a pharmacokinetic (PK) model for tacrolimus (TAC) that includes patient pathophysiology and has clinical applicability in the first 2 weeks post-liver transplantation (PLT).

Methods: Routine monitoring records [dose, trough levels (C(min)), demographics, biochemistry] from 75 patients treated with TAC (Prograf®) PLT were used to develop a population PK model (employing NONMEM®) testing for predictors of oral clearance (CL/F) according to bedside evidence and primarily with aspartate aminotransferase (AST), albumin (ALB), and hematocrit (HCT). Patients were catergorized into subgroups with above and below "normal" thresholds for AST (500 U/L), ALB (2.5 g/dL), and HCT (28 %), respectively. The model was validated with ten patients from the same period and 15 more recent patients. An empirical Bayes method was developed and applied to the prediction of individual profiles serving as a dose adjustment tool.

Results: The number of days PLT (Days PLT) was a key variable during the first 2 weeks, with a dichotomy in the mono-compartmental parameters for 0-3 Days PLT and 4-15 Days PLT. During 0-3 Days PLT, AST levels, indicative of allograft functionality (and TAC metabolism), were crucial predictors of elimination. Three groups were identified with the following clearances: CL/F₀₋₃ = 8.93 L/h for AST ≥ 500 U/L and CL/F₀₋₃ = 11.0 L/h for AST <500 U/L. During 4-15 Day PLT, low values of ALB (<2.5 g/dL) and HCT (<28 %) combined were determinant of a patient subgroup with a tendency to underexposure and complexity in empirical dose adjustment. The CL/F₄₋₁₅ = 25.1 L/h for this subgroup compared to CL/F₄₋₁₅ = 17.1 L/h for the others in that period. The elimination half-life for individual patients varied over tenfold so that a large number of subjects were not at steady state, making the use of a PK model necessary to achieve rapidly and safely the target concentration for TAC in LT. Validation of the model demonstrated that both bias and precision were within acceptable limits.

Conclusion: For TAC therapy, covariate models using mixed effects methods are most useful when combined with patient-specific biochemical assays as well as clinical evidence. In such cases, the observed C(min) and Bayes methods can provide the most likely individual PK parameters, hence the optimal next dose to reach individualized target levels for each patient.