Emmanuel Niyigena1, Yannick Hoffert1, Laurent Peyrin-Biroulet2, Waqqas Afif3, Alessandro Pedicelli3, Xavier Roblin4, Jurij Hanžel5, Konstantinos Papamichael6, Taku Kobayashi7, Zhigang Wang1, Bram Verstockt8, Séverine Vermeire8, Niels Vande Casteele9, Robert Battat10, Erwin Dreesen1
1Department of Pharmaceutical and Pharmacological Sciences, KU Leuven, 2INFINY Institute, Nancy University Hospital, 3Division of Gastroenterology, McGill University Health Centre, 4Gastroenterology, University Hospital of Saint Etienne, 5Department of Gastroenterology, University Medical Centre Ljubljana, 6Center for Inflammatory Bowel Diseases, Division of Gastroenterology, Beth-Israel Deaconess Medical Center, Harvard Medical School, 7Center for Advanced IBD Research and Treatment, Kitasato University Kitasato Institute Hospital, 8Translational Research in Gastrointestinal Disorders, Department of Chronic Diseases and Metabolism, KU Leuven, 9Department of Medicine, University of California, La Jolla, 10Center for Clinical Excellence and Translational Research in Inflammatory Bowel Diseases, University of Montreal Hospital Centre (CHUM)
Introduction: Acute severe ulcerative colitis (ASUC) is a medical emergency that may require urgent colectomy. Infliximab is effective for patients with steroid-refractory ASUC, but increased clearance—due to target-mediated disposition and protein loss in stool—is linked to more severe symptoms and higher colectomy rates.[1] Consequently, patients with ASUC may benefit from personalized infliximab dosing to account for disease activity. Yet, the optimal dosing strategy remains unknown, illustrated by the variety of dosing protocols across medical centres. Escalated dosing regimens (higher doses and shorter intervals) are frequently used to compensate for increased drug clearance, but they have not led to better clinical outcomes compared to standard dosing.[2] Objectives: Our objectives were to (i) pool real-world clinical data from patients with steroid-refractory ASUC, (ii) develop and evaluate a data-driven dose–exposure–response model, (iii) develop and evaluate a dosing algorithm to guide infliximab rescue therapy and (iv) integrate the algorithm into an interactive software tool for personalized infliximab dosing. Methods: A multicentre, retrospective study was performed using pooled data from hospitalized patients (=18 years) with corticosteroid-refractory ASUC diagnosed according to Truelove and Witts criteria.[3] The infliximab exposure–response relationship was characterized using population pharmacokinetic (popPK) modelling in NONMEM v7.5. A 1- or 2-compartment model structure, various residual error models, and parameter correlations were tested. Data below quantification limits were handled with the M3 method. Baseline covariates (body weight, BMI, CRP, serum albumin) were evaluated through stepwise covariate modelling (forward p <5%; backward p <1%). Model selection was based on objective function value, diagnostic plots, and visual predictive checks. Parameter uncertainty was evaluated via nonparametric bootstrap (n=2000). A parametric time-to-event model was developed to describe time to colectomy over 90 days post-infliximab infusion, testing multiple hazard functions. Predictors included baseline covariates and infliximab exposure metrics (empirical Bayes estimated clearance and area under the concentration–time curve; AUC). An optimal cutoff for colectomy risk stratification was identified from ROC analysis using the Youden J statistic. PopPK and time-to-event models informed precision dosing to minimize colectomy risk. A Shiny-based risk and dosing tool was developed in R (v1.7.4). Results: Data from 74 ASUC patients across 8 centres yielded 157 infliximab concentrations. Eleven patients (15%) required colectomy within 90 days. A 1-compartment model with first-order elimination best described infliximab popPK. Higher clearance (CL) and volume of distribution (V) were associated with higher body weight (allometric scaling); higher CRP was associated with higher CL. For a typical 63-kg patient with CRP 30 mg/L, CL was 0.49 L/day (14%), and V was 13.0 L (20%). Interindividual variability was 67.5% (CL) and 100.4% (V); residual variability followed a proportional error model (16%). A Weibull hazard function best described the hazard risk of colectomy, which decreased as the logarithm of week 2-to-4 AUC-to-infliximab CL ratio (AUCw2-4/CL) increased. The Log(AUCw2-4/CL) ratio demonstrated discriminatory ability corresponding to an AUROC of 0.79 (95% CI, 0.53–1.00). A Log(AUCw2-4/CL) threshold of 5.81 discriminated between high-risk (<5.81) and low-risk (=5.81) patients (sensitivity 83%, specificity 85%). High-risk patients had Log(AUCw2-4/CL) ratios 1.34–5.77 with 31% of colectomy events, while low-risk patients had ratios 5.81–8.85 with 4% colectomy events. Classification accuracy was 84% (95% CI 73–92%, p=0.01). An interactive tool was developed to support individualized infliximab dosing using drug concentrations, Bayesian forecasting, and AUCw2-4/CL-based risk prediction. Conclusion: This is the first dose–exposure–response study of infliximab in patients with ASUC. Our work presents a model-based algorithm to inform personalized infliximab dosing in patients with ASUC. We demonstrated that infliximab exposure and infliximab clearance best predict time-to-colectomy.
[1] Seow CH et al. Gut. (2010) 59, 49–54 [2] Choy MC et al. Lancet Gastroenterol Hepatol. (2024) 9, 981–96 [3] Truelove SC, Witts LJ. Br Med J. (1955) 2, 1041–8
Reference: PAGE 33 (2025) Abstr 11597 [www.page-meeting.org/?abstract=11597]
Poster: Drug/Disease Modelling - Other Topics