Yannick Hoffert 1, Barbara Deleenheer 2,3,4, Caroline Boelhouwer 2,5, Nele Van den Eede 6, Emmanuel Niyigena 1,7, Laurens Ceulemans 4,8, Laure Elens 7, Diethard Monbaliu 5,9, Jacques Pirenne 5,10, Lucas Wauters 3,4,11, Tim Vanuytsel 3,4,11, Erwin Dreesen 1
1 Katholieke Universiteit Leuven (Leuven, Belgium), 2 Hospital Pharmacy, University Hospitals (UZ) Leuven (Leuven, Belgium), 3 Leuven Intestinal Failure and Transplantation (LIFT), UZ Leuven (Leuven, Belgium), 4 Department of Chronic Diseases and Metabolism, KU Leuven (Leuven, Belgium), 5 Department of Microbiology, Immunology and Transplantation, KU Leuven (Leuven, Belgium), 6 Laboratory Medicine, UZ Leuven (Leuven, Belgium), 7 Louvain Drug Research Institute (LDRI), UCLouvain (Brussels, Belgium), 8 Department of Thoracic Surgery, UZ Leuven (Leuven, Belgium), 9 Transplantoux Foundation (Leuven, Belgium), 10 Abdominal Transplant Surgery, UZ Leuven (Leuven, Belgium), 11 Department of Gastroenterology and Hepatology, UZ Leuven (Leuven, Belgium)
Objectives
Tacrolimus is the cornerstone immunosuppressant after intestinal transplantation, a rare but life saving option for irreversible complicated intestinal failure. Its dosing is challenged by a narrow therapeutic index and highly variable pharmacokinetics, especially early after surgery when inflammation, physiological instability and impaired enteral function cause unpredictable exposure putting the patient at risk for rejection of the graft.¹ No population pharmacokinetics (popPK) and exposure–response models are available for intestinal transplant recipients, limiting quantitative support for dosing.²
The objectives were:
i) to develop an integrated popPK and exposure–response framework of tacrolimus disposition and risk of rejection;
ii) to quantify predictors of tacrolimus exposure and biopsy-proven rejection;
iii) to integrate the framework into a clinical decision tool to compare tacrolimus initiation strategies.
Methods
We performed a 15 year (2007–2022) single center retrospective study of routine-care data from intestinal transplant recipients at University Hospital Leuven (Belgium). Data extracted from electronic health records included intravenous and oral tacrolimus dosing histories, concentrations, laboratory markers, and clinical outcomes (biopsy proven rejection and death) from transplantation to one year.
A popPK model was built using nonlinear mixed effects modeling. A parametric multistate time to event model linked tacrolimus exposure to three clinical states:
i) ongoing tacrolimus therapy without complication (S1);
ii) biopsy proven rejection (S2);
iii) death (S3).
Transitions were parameterized using Weibull (S1→S2) and exponential functions (S1→S3, S2→S3). To mitigate immortal time and time‑varying confounding, tacrolimus exposure was computed from individual Bayesian posteriors and evaluated as time‑dependent predictors (area under the curve, AUC0–t) along with longitudinal clinical covariates (hematocrit, C-reactive protein, serum albumin, age, bodyweight).
Parameter uncertainty was assessed by relative standard error (RSE) and Sampling Importance Resampling (SIR). Goodness-of-fit plots and visual predictive checks (VPCs) were used to assess model calibration. The AUC under the receiver operating characteristic curve (AUROC) was used to assess discrimination, including the Youden J index for threshold identification.
All modeling and simulation tasks were performed in NONMEM (v.7.5.1) supported by high performance cluster computing (VSC).
Results
Twenty recipients (median age 41 years; 11 male) contributed data. Fourteen experienced rejection (median time-to-first rejection 23 days, range 13–264), and four died (time-to death 33, 116, 171, and 254 days) within one year.
Tacrolimus pharmacokinetics was best described by a one compartment model with first order absorption and postoperative time dependent clearance (hockey-stick function). Clearance increased linearly from 1.59 L/h (RSE 21%) at day 0 to stabilize by day 84 (RSE 3%) at 8.72 L/h, with a corresponding half life shortening from 16.2 h to 2.9 h. Oral bioavailability was low (26.9%, RSE 28%) and variable (50.7% coefficient of variation). Besides postoperative time, no covariates were retained on pharmacokinetic parameters.
In the multistate exposure–response model, baseline transition parameters were estimated for
i) S1→S2 (n=14) λ₁₂=1/121 t^-1 (RSE 37%), γ₁₂=0.737 (RSE 20%);
ii) S2→S3 (n=2) λ₂₃=1/1960 t^-1 (RSE 70%);
iii) S1→S3 (n=2) λ₁₃=1/1250 t^-1 (RSE 71%).
Tacrolimus exposure during the first 13 days (AUC₀-₁₃) was the sole significant predictor of rejection (slope -5.99∙10-4, RSE 23%) with higher AUC₀-₁₃ associated with lower hazard of transitioning to S2 (Equation 1).
h₁₂,i(t) = λ₁₂∙γ₁₂∙(λ₁₂∙t)^(γ₁₂-1)∙(1+(-5.99∙10^-4∙(AUC₀-₁₃,i-x))); Equation 1
With λ₁₂ depicting the baseline hazard, γ₁₂ the shape parameter of the Weibull function, and x the median of the predictor in our cohort.
Both popPK and exposure–response models were well-calibrated, but discrimination for rejection was modest-to-uncertain (AUROC 0.77, 95% confidence interval 0.48–1.00), indicating limited risk stratification performance. The optimal AUC₀-₁₃ threshold was 4,884.6 ng∙h/mL (~Cavg 15.7 ng/mL; sensitivity 0.79, specificity 0.83).
The linked popPK–multistate model was implemented in a Shiny open-source interactive tool that (i) performs Bayesian forecasting from patient concentrations, (ii) simulates candidate intravenous‑to‑oral initiation regimens, and (iii) reports predicted exposure metrics and rejection probability to support individualized tacrolimus dosing (https://gitlab.kuleuven.be/pharmacometrics).
Conclusions
This is the first popPK and multistate exposure–response framework for tacrolimus in intestinal transplantation. Tacrolimus clearance showed marked postoperative time dependency, underscoring the need for dynamic early dosing. Early exposure predicted biopsy-proven rejection within one year. These findings provide a quantitative basis for model informed tacrolimus initiation. Updating of the exposure–response model with additional data will be required before external validation and clinical implementation. The open source tool translates the models into a simulation workflow to facilitate external evaluation of individualized tacrolimus initiation in the clinic.
References:
¹Venkataramanan R, Swaminathan A, Prasad T, et al. Clinical Pharmacokinetics of Tacrolimus. Clin Pharmacokinet. 1995;29(6):404-430. doi:10.2165/00003088-199529060-00003
²Hoffert Y, Dia N, Vanuytsel T, et al. Model-Informed Precision Dosing of Tacrolimus: A Systematic Review of Population Pharmacokinetic Models and a Benchmark Study of Software Tools. Clin Pharmacokinet. 2024;63(10):1407-1421. doi:10.1007/s40262-024-01414-y
Reference: PAGE 34 (2026) Abstr 12121 [www.page-meeting.org/?abstract=12121]
Poster: Clinical Applications