Nicolas Luyckx (1), Andreas Lindauer (1), Robin Vos (2), Dirk Kuypers (2)
(1) Calvagone SAS, France; (2) KU Leuven, Belgium
Introduction:
Tacrolimus is a critical immunosuppressant drug used in transplantation, but its narrow therapeutic index and highly variable bioavailability make dosing challenging [1]. To address this, KU Leuven and UZ Leuven have collaborated to develop a model-informed precision dosing (MIPD) tool for tacrolimus in renal transplant recipients during the first 14 days post-transplantation [2]. This project has led to the creation of the tdmore framework in R, which allows for individual Bayesian fit of a virtual population, including a novel estimation technique called model-predictive control (MPC) and the exploration of new dosing regimens [3]. The current work aimed at applying the same approach for patients undergoing lung transplantation.
Objectives:
- To develop a population PK model for tacrolimus in lung transplant recipients and identify predictive factors for inter- and intra-individual variability.
- To develop a computer-aided dose-individualization strategy using the MPC approach for this patient population and evaluate its predictive performance retrospectively.
Methods:
Tacrolimus concentration data from 322 lung transplant patients (2015-2020) were analyzed after excluding 38 due to age, re-transplantation, dosing time, or outliers. Data were divided into estimation (209 patients, 2609 concentrations, primarily from twice-daily Prograf capsules or syrup) and validation sets in a 2:1 ratio. Analysis involved developing a structural population PK model with NONMEM (v7.5), automatic covariate modeling, model refinement, and validation through visual checks. The finalized model was implemented in the MPC framework via tdmore (v1.2.0.900) for R, followed by simulations to assess the efficacy of various dose-individualization strategies.
Results:
Model development:
A one-compartment model with time-varying (apparent) clearance provided the best fit to the tacrolimus data. According to the model, apparent clearance increases within about 5 days after surgery from almost 0 to about 20 L/h – which is most likely a reflection of alterations in binding of tacrolimus to blood components and other peri-operative influencial factors (e.g. gastro-intestinal motility, blood flow etc.).
The extended-release formulation showed an about 25% lower bioavailability than the immediate release capsules. Body weight, age, bilirubin and co-medication with triazoles were found to be significant covariates. A ‘reduced’ version of the final model without bilirubin was included in the assessment of dose-adaptation strategies.
Predictive performance:
Models including the time-varying clearance outperformed the base model. The reduced model performed slightly better than the model without any covariates and marginally worse than the full model, indicating that bilirubin measurements are not necessary for dose-adaptation once tacrolimus levels are available.
Dose adaptation algorithm:
An MIPD algorithm was developed, where individual tacrolimus trough measurements are compared to a target or threshold value and future doses are adapted accordingly. To avoid giving a too high dose when the observed levels are below the threshold, the concept of time-to-target (TTT) was introduced, whereby the next doses are adjusted to meet the target concentration not 12 hours after the following dose but some time (TTT) further in the future. This provides more room for the concentrations to stabilize and avoids over-shooting the therapeutic range. As further precautionary measures, dose caps and maximum fold-increase in doses were also explored.
Simulations, based on the retrospective data, identified the following ideal parameters for the MIPD algorithm: ideal time-to-target (TTT) of 36 hours, target threshold of 12.5 ng/mL, maximum allowed dose of 9 mg, maximum fold-increase of 1.5, and a bodyweight-based loading dose of 0.07 mg/kg followed by 3 mg flat (until further adjustment) for patients not treated with triazole antifungals.
Conclusions:
A one-compartment PK model with time-varying apparent clearance and body weight, age, formulation, bilirubin and co-medication with triazole antifungals provided a good fit to the tacrolimus concentration data. The MIPD algorithm potentially increases the percentage of patients with tacrolimus levels within the therapeutic range from day 4 onwards by approx. 48% (current practice) to 57%, while reducing the percentage above the therapeutic range from respectively 29% to 22% and 32% to 23% at day 4 and day 5, the ‘over-shooting’ days.
References:
[1] Kirubakaran, R et al. Clin Pharmacokinet (2020), 59(11): p. 1357-1392.
[2] Faelens, R. et al. CPT:PSP (2022), 11(3): p. 348-361.
[3] Faelens, R., Tdmore: In silico evaluation of precision dosing. (2021)
Reference: PAGE 32 (2024) Abstr 10830 [www.page-meeting.org/?abstract=10830]
Poster: Clinical Applications