C. Tang (1), N. Knops (2), R. Faelens (1), D. Kuypers (2), T. Bouillon (1)
(1) Drug Delivery and Disposition, Department of Pharmaceutical and Pharmacological Sciences, Katholieke Universiteit Leuven, BE, (2) Dept. of Pediatric Nephrology and Solid Organ Transplantation, University Hospitals Leuven, BE, (3) Dept. of Nephrology and Renal Transplantation, University Hospitals Leuven, BE.
Objectives: Multiple clinical, demographic and genetic factors affect the pharmacokinetics (PK) of tacrolimus (Tac), necessitating therapeutic drug monitoring (TDM). Due to profound changes in age and size over the course of treatment in the pediatric population (a child can easily double its weight while on maintenance therapy with Tac), a thorough assessment of covariate effects, the projected change of dose with age/size and quantification of the remaining within and between subject variability (WSV, BSV) is warranted, but usually prevented by a small number of patients, brief follow up and lack of diversity of covariates in the analysis datasets. Recently, a distinct age effect on top of weight/BSA on the dose exposure relationship has been postulated in 43 renal transplant patients, using full AUCs as summary metric of exposure [1]. We applied nonlinear mixed effects modeling on individual concentration time courses from this meanwhile extended and growing database.
Methods: The analysis dataset grew from 43 [1] to 79 stable post-transplant patients (renal: 51 (64.5%), liver: 26 (33%), bowel/liver: 2 (2.5%); donor status: 73% deceased, 27% live) included between 1998 and 2017. At time of inclusion (first PK assessment) patients were aged from 1.3-20.7 ys (median=11.3ys), weighing between 9.9-75.5 kg (median= 32.7 kg). CYP3A4-1b, CYP3A5 and CYP3A7 pharmacogenetic information was dichotomized according to the known influence on Tac metabolism (i.e. fast (F) and slow (S) metabolizers (M)). 14% of patients were FM for CYP3A5, 2.5% were FM for both CYP3A4 and CYP3A7. 1-14 (median=4) concentration time course per patient (“occasions”) were available, spaced between 10d to 11.3ys (median 1.0 ys) post transplant. Per occasion, 6 Tac steady state concentrations were collected (0, 1, 2, 4, 6 and 12h), covering the entire interdose interval. The analysis was performed with MONOLIX 2016R1. Compartmental models with first order input into a depot and lag time were tested. Covariate exploration was done treating each concentration time course as a (virtual) individual and repeated with the definitive model including a full covariance matrix for volumes and clearances for both BSV and WSV. Improvement of -2LL, visual inspection and standard errors of the parameters were used for judging goodness of fit and covariate inclusion/deletion.
Results: A two-compartment model with first-order absorption and elimination adequately described the concentration time course of Tac. CYP3A5 fast metabolizers and fast metabolizers for both CYP3A4 and CYP3A7 displayed similar CL values and were lumped into one group (22% increase in CL). Tlag and ka did not differ much between models and were approx. 0.45 h and 0.65 1/min. Volumes and Clearances of the final model are for a weight of 70kg, a hematocrit of 40%, slow metabolizers and, where applicable, an age of 20ys. The allometric coefficients for weight on V1, V2, CL and Q for the final model were (TV+SE): 0.54 (0.16), 0.55 (0.11), 0.45 (0.05) and 0.51 (0.08).
Table: Parameters (TV+SE, random effects (BSV/WSV, SD in log domain)) and significant covariates (p<0.01)
|
|
TV(SE, BSV, WSV) |
|
|
|||
|
Model |
V1 [L] |
V2 [L] |
CL [L/h] |
Q [L/h] |
covariates |
OFV |
|
“virtual ind.” base |
53 (6.3, 1.23,-) |
533 (47, 0.58,-) |
22.1 (0.63, 0.57,-) |
76.3 (4.1, 0.64,-) |
– |
10687 |
|
“virtual ind.” final |
60.2 (1.19, -) |
842 (0.6,-) |
24.7 (0.45,-) |
94.4 (0.53,-) |
WT on V1,V2, CL,Q; Hct+CYP on CL, Age on V1, Q |
10490 |
|
“correct” base |
76.7 (9.4, 0.65, 1.04) |
401 (32, 0.46, 0.42) |
21.6 (1.3, 0.5, 0.31) |
73.4 (5.5, 0.50, 0.52 |
– |
10357 |
|
“correct” final |
86.1 (14, 0.6, 0.98) |
590 (67, 0.4. 0.4) |
24.2 (1.4, 0.38, 0.30) |
94.8 (7.9, 0.41, 0.42) |
WT on V1, V2, CL, Q; Hct+CYP on CL |
10256 |
Conclusions: Size, hematocrit and enzyme activity are the major covariates to be considered in Tac open loop dose adjustment. Unexplained BSV/WSV on CL with the best model are 38% and 30%. The allometric coefficients for clearance were significantly smaller than 0.75 (0.56, 0.45, SE’s 0.05 for both) and we concur with the recommendation for body surface area based dosing by Knops et al [1]. Age was not identified as a covariate in the final analysis, confirming the results of [2,3] and contrary to [1]. Due to mandatory TDM adjusted dosing, this is inconsequential for adjustments of maintenance doses. Care must be taken when deciding on a covariate identification strategy in large unbalanced datasets containing covariates changing per individual and covariates changing per occasion.
References:
[1] Knops N et al. BJCP (2017) 83, 863-74. [2] Zhao W et al. Clin Pharmacol Ther (2009) 86, 609-18. [3] Prytula AA et al. Clin Pharmacokinet (2016) 55, 1129-43
Reference: PAGE 27 (2018) Abstr 8553 [www.page-meeting.org/?abstract=8553]
Poster: Drug/Disease Modelling - Paediatrics