Pierre Marquet

Background: Ganciclovir (GCV) exposure after standard valganciclovir (VGCV) prophylaxis is highly variable. Guidelines recommend Cockcroft–Gault for renal dose adjustment. We evaluated whether Cockcroft–Gault‐based dosing achieves the target prophylactic exposure (AUC _0‐24h: 40–60 mg.h/L) and whether MDRD or CKD‐EPI alters exposure. Methods: Using a validated population pharmacokinetic model, we simulated 10 000 adult solid organ transplant recipients. VGCV doses were assigned according to SmPC renal function categories based on Cockcroft–Gault, MDRD or CKD‐EPI. Patient reclassification, target attainment and exposure around dosing thresholds (25, 40 and 60 mL/min) were assessed. Results: MDRD and CKD‐EPI reclassified 45.1% and 39.9% of patients, mostly downward, lowering target attainment (20.1% and 22.7% vs. 30.8% with Cockcroft–Gault). Target attainment never exceeded 40% in any category, and underexposure predominated, reaching up to 94.1% in the most impaired renal function categories. Target attainment was consistently lower just below each threshold, and the 25 and 40 mL/min thresholds were set above the renal function ranges where the corresponding regimens achieved target exposure. Conclusions: Renal function category‐based VGCV dosing does not reliably achieve target GCV exposure. Current thresholds are poorly aligned with the exposure‐response relationship and warrant re‐examination. Alternative equations do not reduce the variability, supporting exposure‐guided approaches such as therapeutic drug monitoring and model‐informed precision dosing.

Background: Precision dosing integrates patient-specific factors to optimize pharmacotherapy. However, current evidence often relies on limited single-center studies, restricting generalizability. Methods: A narrative review of literature on real-world evidence, pharmacogenetics, biomarkers, artificial intelligence (AI), and privacy-preserving data strategies relevant to precision dosing was conducted. Results: Single-center data are insufficient for robust precision-dosing evidence because of small subgroups and local practice biases. Multicenter privacy-preserving systems enable richer, more generalizable datasets that link doses, exposures, genotypes, biomarkers, and outcomes. However, multicenter collaborations should remain aligned with the intended population and should not imply indiscriminate pooling of heterogeneous data. Secure pooled environments, federated learning, and synthetic data play distinct method-dependent roles. AI supports exposure estimation, phenotype extraction, and modeling of complex relationships but requires rigorous validation against conventional methods. Causal inference requires epidemiological rigor beyond predictive performance. Conclusions: Advancing precision dosing requires multicenter, purpose-specific, privacy-preserving, methodologically rigorous, and clinically validated data systems guided by clinical pharmacologists, with AI serving as a complementary tool rather than a default solution.

Objectives: The pharmaco‐epidemiological research program in kidney transplantation (PERP‐KT) aims to evaluate, for the principal maintenance immunosuppressive drugs (MISDs): the influence of model‐informed precision dosing on graft and patient survival; long‐term exposure—effects relationships; and the benefit‐harm balance of their combinations and time sequences in patient groups or clinical settings not adequately evaluated in comparative randomized clinical trials. It also aims to develop a hybrid, dynamic, deep learning model capable of predicting the rate of renal graft function decline, thereby providing a platform for individualized prediction of the benefits and risks associated with MISDs. Patients and Methods: After obtaining all regulatory andd ethical approvals, de‐identified extracts of three national databases were linked to create the nationwide PERP‐KT dataset, which is hosted within the highly secure environment of the French Health Data Hub. Results: CRISTAL (the exhaustive registry of the French Agence de la Biomédecine) comprises data from 49 886 kidney donors and the corresponding 47 842 transplant recipients between 2005 and 2020. Following iterative deterministic matching and extensive quality control procedures, CRISTAL data were successfully linked to: the French national Health Data System (SNDS), which records reimbursed healthcare utilization, for 30 782 kidney transplant recipients; and to ISBA, a web‐based platform for Bayesian dose adjustment of MISDs that contains pharmacological data, for 17 700 kidney grafts and 17 576 recipients. Perspectives: More than 20 pharmaco‐epidemiological studies will leverage the database's extensive follow‐up, large population size and richness of clinical, healtcare‐utilization, and pharmacological data.

Background: Tacrolimus is a first-line immunosuppressant used in kidney transplantation. However, its narrow therapeutic window and variable exposure present challenges. Extended- and prolonged-release formulations, ER-tacrolimus (Advagraf) and LCP-tacrolimus (Envarsus), respectively, improve adherence, and LCP-tacrolimus provides higher bioavailability. The safety and benefits of switching from ER-tacrolimus to LCP-tacrolimus were evaluated. Methods: In this open-label switch study, first-time, adult, kidney transplant recipients with stable graft function, therapeutic ER-tacrolimus levels, and rapid metabolizer status (C/D ratio 0 ; peak blood levels, Cmax; area under the concentration–time curve, AUC). Results: Twelve patients were included. The dose to achieve therapeutic C ₀ (5–8 µg/L) was 30% lower with LCP-tacrolimus than with ER-tacrolimus (6.0 mg versus 8.5 mg, P = 0.004), with similar pill burden and adverse effects. AUC or Cmax were similar [LCP-tac: 14.75 (11.68–16.27); ER-tac: 17.15 (13.53–20.93)]. For CYP3A5 expressors (n = 7; 58%), there was a stronger correlation between C ₀ and AUC for LCP-tacrolimus (0.85, P = 0.016; 0.43, P = 0.34, respectively). No serious adverse events were observed. Conclusions: Conversion to LCP-tacrolimus with a 30% dose reduction achieved therapeutic C ₀ without differences in pill burden or overall exposure. In the CYP3A5 expressors, C ₀ was more strongly correlated with exposure to LCP-tacrolimus, suggesting more reliable trough monitoring. The small sample size, short follow-up period, and selection bias limit interpretation.

Introduction: The sharing of individual patient data is essential for advancing pharmacometrics but is strictly limited by privacy regulations (e.g., GDPR). While synthetic data generation offers a legally compliant alternative, its structural impact on complex nonlinear mixed-effects (NLME) modelling remains largely unexplored. This study aimed to benchmark five generative artificial intelligence algorithms by evaluating the balance between data privacy and the preservation of structural PK properties and clinical dosing guidance. Material & methods: A daptomycin two-compartment PopPK model was used to simulate a reference cohort of 500 patients. Five generative algorithms (Modified AVATAR, Gaussian Copula, Synthpop, TVAE, and CTGAN) produced 100 independent synthetic datasets each. A two-stage evaluation framework was applied: first, a statistical indistinguishability test based on logistic regression (AUC ROC) was used as a macroscopic pre-selection criterion to determine algorithm eligibility for NLME modelling and privacy risk assessment. Privacy risk was independently quantified using the Anonymeter framework (Singling Out and Linkability attacks). Eligible algorithms were further evaluated on PK parameter recovery bias and clinical dosing simulations. Results: Deep learning architectures (TVAE, CTGAN) were excluded at the pre-selection stage due to both biologically implausible covariate generation and high macroscopic detectability (mean AUC ROC = 0.837 and 0.986, respectively). Synthpop, AVATAR, and Gaussian Copula all passed the indistinguishability threshold (AUC ROC = 0.475 ± 0.033, 0.490 ± 0.013, and 0.619 ± 0.031, respectively) and proceeded to NLME evaluation. However, attack-based privacy assessment revealed that Synthpop carried an unacceptable singling-out risk (0.035), disqualifying it from privacy-preserving data sharing. AVATAR and Gaussian Copula demonstrated acceptable privacy profiles (singling-out = 0.004 and 0.001; linkability = 0.010 and 0.003, respectively). At the structural level, Gaussian Copula injected stochastic noise inflating residual error (+157.0%) and VLJ (+25.9%), blunting predicted Cmax and predisposing to empirical dose escalation and risk of toxicity. AVATAR acted aSs a smoothing filter, deflating VLJ (−48.3%) and underestimating CL (−12.9%). Forward clinical simulations confirmed directionally opposed prediction errors: Gaussian Copula consistently underestimated Cmax across standard and renally impaired profiles (−14.5% and -16.0%, respectively), predisposing to empirical dose escalation, whereas AVATAR- and Synthpop-derived models overestimated Cmax and Cmin in the obese infected patient (+14.7% and +8.2%, respectively), compounding the accumulation risk already present in this profile. Conclusion: While no generative algorithm currently offers a perfect solution, AVATAR and Gaussian Copula represent the most viable candidates, being the only methods to satisfy both macroscopic indistinguishability and attack-based privacy criteria. These findings highlight the necessity of a structured, two-stage validation framework and suggest that, when coupled with therapeutic drug monitoring, synthetic datasets could significantly enhance multicentre collaboration while maintaining strict regulatory compliance.

Mycophenolic acid (MPA) is a widely used immunosuppressant whose use is often limited by gastrointestinal toxicity. Gut bacterial hydrolysis of liver‐derived MPA glucuronides increases local exposure to MPA, potentially impairing epithelial barrier function and cellular metabolism. To explore the effects of MPA on gut barrier integrity and metabolic pathways in gut epithelial cells, Caco‐2 cells were exposed to MPA (10 or 100 μM), and barrier function was assessed by transepithelial electrical resistance (TEER) and lucifer yellow (LY) permeability in both differentiated and early‐stage monolayers, while intracellular metabolic changes were investigated using targeted LC–MS/MS metabolomics. In differentiated monolayers, MPA did not significantly alter LY transport or TEER measurements. In contrast, MPA exposure during the early stages of monolayer formation reduced TEER values, 3 days after MPA withdrawal (Day 6; p < 0.01). The effects of 100‐μM MPA were still noticeable at Day 10, as confirmed by LY permeability ( p < 0.05). Metabolomic profiling clearly separated exposed from control cells (PCA, PC1 + PC2 = 92% variance). At 10 and 100 μM, 9 and 8 metabolites were significantly altered, with 6 common to both doses. Pathway enrichment revealed perturbations mainly in nucleotide synthesis, consistent with altered metabolic activity.

In lifelong immunosuppressive therapy, missed doses of prolonged‐release tacrolimus are almost inevitable and may reduce drug exposure, compromising transplant outcomes. Using pharmacokinetic models, we simulated delayed and missed doses in virtual kidney and liver transplant recipients. Our simulations showed that a missed dose significantly lowers exposure, with recovery taking 2–4 days. For delays shorter than 12 h, the full dose should be taken immediately; for delays between 12 and 24 h, half the dose should be administered; for a missed dose, 150% should be given at the next intake. These strategies may help maintain therapeutic exposure and preserve transplant outcomes.

Background.: Many tools have been developed to estimate the risk of graft loss for kidney transplant patients, but only a few have been thoroughly validated. Methods.: We previously developed the adjustable graft failure score (AdGFS), an easy-to-use tool combining 2–7 variables routinely collected at 1 y posttransplant in all of the transplantation centers. During model development, we selected the variables using random survival forest, a machine learning method, to come up with a simple decision tree based on: donor age; pretransplant non-donor-specific anti-HLA antibodies (presence/absence); kinetic profile of serum creatinine using its values at months 1, 3, 6, and 12 posttransplant; serum creatinine and proteinuria (g/L or g/24 h) at 1 y posttransplantation; occurrence of de novo donor-specific anti-HLA antibodies (yes/no); occurrence of a first episode of acute rejection. The aim of the present study was to validate AdGFS on 2 large, independent, longitudinal cohorts of kidney transplant patients from Leuven, Belgium (n = 930 patients) and Lyon, France (n = 307 patients). Results.: The 4 risk groups obtained actually showed very large and significant differences in graft survival at 10 y posttransplantation: 89%, 81%, 59%, and 23% in the low, intermediate, high, and very high-risk groups, respectively. At 2, 5, and 10 y posttransplantation, the area under the curve (95% confidence interval) of the receiver operating characteristic curve of AdGFS was 0.86 (0.76-0.96), 0.76 (0.70-0.82), and 0.73 (0.69-0.77), respectively. Conclusions.: This external validation study confirmed that AdGFS is clinically applicable for the long-term risk graduation of graft loss at the first transplantation anniversary. It suggests that it may be used to tailor the overall management of the patient to the individual risk and as a surrogate outcome in clinical trials.

Aims: Model‐informed precision dosing of tacrolimus is crucial to minimize toxicity and prevent graft‐ vs.‐ host disease in children undergoing haematopoietic stem cell transplantation (HSCT). Therefore, we developed a population pharmacokinetic model of tacrolimus oral suspension in paediatric HSCT patients that could be used for routine care. Methods: Sixteen children receiving continuous intravenous (IV) tacrolimus who were switched to an oral suspension prepared at the hospital pharmacy were included. A total of 339 whole‐blood samples were collected and quantified by immunoassay: 260 during IV administration and 79 from 12‐h pharmacokinetic profiles (at 0, 20 min, 1, 3 and 12 h) after oral administration. Data analysis was performed using a nonlinear mixed‐effects approach with Monolix considering demographic, clinical and pharmacological factors as potential covariates influencing tacrolimus pharmacokinetics. Results: A 1‐compartment model with linear elimination adequately described tacrolimus pharmacokinetics. The typical estimated values (% relative standard error) for clearance, volume of distribution and absolute bioavailability were 1.5 L/h (9.5%), 42.3 L (17.5%) and 40.0% (14.8%), respectively. A mean lag time of 0.3 h best characterized the delay in the absorption phase. Body weight was significantly associated with clearance, volume of distribution and bioavailability. Conclusion: A population pharmacokinetic model was developed to describe the pharmacokinetics of tacrolimus during continuous IV infusion or oral suspension administration in paediatric HSCT patients, which may be used for individualized dose adjustment. The mean bioavailability of the oral suspension was 40%, suggesting a dose conversion factor of 2.5 from IV to oral suspension therapy in our clinical setting.