Tacrolimus dose individualization in "de novo" patients after 10 years of experience in liver transplantation: pharmacokinetic considerations and patient pathophysiology.

AIM: To determine how changes in tacrolimus (TAC) immunosuppression clinical practice, in the first 15 days post liver transplantation (LT) and across a decade, impact a clinical covariate - pharmacokinetic (PK) model, developed in data from 1998, thus testing its utility in dose individualization across time. Patient cohorts from 1998 (Reference: R-1998) and 2007 (EVALUATION: E-2007) were compared. METHODS: Analysis of monitoring observations (Cmin and Cmin/dose) and the biochemical variables aspartate aminotransferase (AST), hematocrit (HCT), albumin (ALB) and serum creatinine (SCr) was done for 0 - 3 and 4 - 15 days post transplantation (PT). The population PK model developed for R-1998 [1] was re-evaluated for the two cohorts. RESULTS: Significant differences in R-1998 vs. E-2007 existed in Cmin and Cmin/dose and in covariates AST (as hepatic function marker) and SCr (as toxicity marker). E-2007 had lower levels of Cmin and Cmin/dose (1/CL), lower AST with faster recovery and lower variability in Cmin/dose for similar dose. AST was a covariate on CL/F in the 0 - 3 day PT period. In 4 - 15 days PT for E-2007, low levels of HCT and ALB as CL/F predictors confirmed a subgroup with higher CL/F (23.8 l/h vs. 19.3 l/h). The R-1998 model's original structure was confirmed. CONCLUSIONS: Ten years of use of TAC shows gain in therapeutic targeting efficiency, due to improvement in LT methods, knowledge of the drug and consideration of PK steady state. The remaining uncertainty with TAC monitoring in LT can be resolved with application of PK principles combined with patients' diosyncrasies in the model developed for TAC dose individualization in R-1998. The applicability of the model as nucleus in Bayes individualization remains intact across a decade.

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

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.

Pathophysiological idiosyncrasies and pharmacokinetic realities may interfere with tacrolimus dose titration in liver transplantation.

PURPOSE: To explore the main factors that make it difficult to empirically monitor tacrolimus (TAC) in the early period post-liver transplantation (LTx), with a specific focus on those aspects related to patient idiosyncrasy and clinical status as well as to the pharmacokinetic (PK) assumptions on which drug individualization in clinical practice is based. METHODS: Retrospective monitoring data from 75 de novo liver transplant patients treated with twice daily with TAC and followed for up to 15 days were analyzed. An extensive battery of laboratory measurements were available. Dose adjustment was performed empirically using trough levels (C(min)). The population was separated into two major background groups according to low or high values of aspartate aminotransferase (AST) (Group 1 and 2, respectively) based on AST measurements made during the first 4 days post-LTx. Each of these two major groups was then further subdivided into two subgroups based on elevated (Groups 1A, 2A) or reduced (Groups 1B, 2B) combined albumin (cut-off 2.5 g/dl) and hematocrit (cut-off 28%). RESULTS: The C(min)/Dose ratio [inversely proportional to systemic clearance (CL)] had a variability [coefficient of variation (CV) >80%) that was incongruently higher for the ratio than for C(min) and Dose separately. This was attributed to most patients not being at steady state or physiologically stable in the early post-LTx period. Group 1 patients were more predictable than Group 2 patients, who were responsible for the variability in the ratio. C(min) was lower in the reduced ALB and HCT patient groups when AST conditions were similar (1A vs. 1B and 2A vs. 2B), likely due to increased TAC metabolic clearance (reduced C(min)/Dose). This situation existed for two periods: 0-15 days post-LTx and 4-15 days post-LTx observations. Group 2A patients were the main source of the paradoxical variability in C(min)/Dose (higher ratio of 2.7; CV = 100%), suggesting a lower clearance and difficulty in the recovery of stability. In contrast, Group 2B patients had the lower ratio (1.4; 47%) but required the highest number of dose adjustments as the variability was hard to identify clinically. Group 1A patients were the most predictable empirically. When observations from 15 new patients who entered the clinic in 2007 and 2008 were used for the analysis, the same sub-groups existed in the same proportions in both years. CONCLUSION: The difficulty in empirical dose adjustment of TAC is associated to the inevitable non-fulfillment of PK assumptions early post-LTx and also to the inherent complexity of the clinical condition, leading to increased uncertainty for the clinician regarding dose selection. Identifying these sub-categories provides a rational means of classifying patients akin to a phenotype. The complexity of the kinetics in LTx and TAC treatment does not invalidate C(min) as a biomarker, but a Bayes algorithm including a full PK structure and these covariates would be optimal.