Ella Signe Kassandra Widigson 1,2, Katharina Braun 1,2, Zrinka Duvnjak 1,2, Marian Klose 1,2, Camilla Frimor 3, Kunal Anilkumar, Wilhelm Huisinga 2,4, Casper Steenholdt, Mark Andrew Ainsworth 3,5, Charlotte Kloft 1,2
1 Freie Universität Berlin, Institute of Pharmacy, Department of Clinical Pharmacy and Biochemistry (Berlin, Germany), 2 Graduate Research Training Program PharMetrX (Berlin/Potsdam, Germany), 3 Odense University Hospital, Department of Medical Gastroenterology (Odense, Denmark), 4 University of Potsdam, Institute of Mathematics (Potsdam, Germany), 5 University of Southern Denmark, Department of Clinical Research- Research Unit of Medical Gastroenterology (Odense, Denmark)
Background
Monoclonal antibodies (mAbs) such as infliximab, ustekinumab (UST), and vedolizumab (VDZ) represent effective treatment options for patients with inflammatory bowel diseases (IBD) [1]. While biologic therapy optimisation using MIPD has been shown to result in higher clinical remission rates and cost-effectiveness of infliximab treatment, dosing optimisation strategies for UST and VDZ remain largely empirical, typically relying on symptom-based dose escalation [2,3]. Evidence supporting model-informed Therapeutic Drug Monitoring for UST and VDZ is still limited [4].
Current MIPD approaches predominantly rely on maximum a posteriori (MAP)-based Bayesian estimation, which provides point estimates of individual pharmacokinetic parameters [5]. Approximating the posterior distribution via the Laplace approximation enables additionally the quantification of individual parameter uncertainty, which can be propagated into stochastic simulations of concentration-time profiles to derive the probability of target attainment (PTA) for evaluated dosing regimens [6].
The prospective randomised controlled trial (RCT) MOVE-IT [7] was designed to compare conventional dosing with an MIPD-based dosing strategy, additionally incorporating parameter uncertainty to estimate PTA.
Objectives
To develop a MIPD framework for UST and VDZ incorporating Laplace-approximated posterior uncertainty from MAP Bayesian estimation, for implementation within the interventional arm of the prospective RCT MOVE-IT.
Methods
A literature search was conducted to identify drug- and administration route-specific minimum serum concentration (Cmin) thresholds associated with endoscopic remission during maintenance treatment (ER), as well as published NLME pharmacokinetic models for UST and VDZ. NLME model selection criteria included structural plausibility, inclusion of clinically relevant covariates, support for both intravenous (IV) and subcutaneous (SC) administration routes, and similarity with the MOVE-IT study population. The selected NLME models were implemented in NONMEM® (version 7.5.1) and the R package mrgsolve (version 1.5.2)[8].
MAP Bayesian estimation provided empirical Bayes estimates (EBEs) of the individual random effects (η) [9]. The associated variance–covariance matrix of these effects (NONMEM®: ETC matrix), obtained from the Laplace approximation of the individual posterior distribution, was used to characterise parameter uncertainty [6,10]. This uncertainty was propagated into simulations to calculate PTA, defined as the proportion of simulations achieving Cmin above the drug- and route-specific efficacy threshold at steady-state. The dosing recommendation for the RCT was the longest dosing interval achieving PTA ≥ 80 %.
Nine virtual patient scenarios (three per drug/route: UST SC, VDZ IV, VDZ SC) with predefined exposure categories (under-, adequate, overexposure relative to established Cmin–ER thresholds) were used to verify correct algorithm behaviour. Each scenario incorporated full patient characteristics and three iteratively added serum concentrations.
Results
Published NLME models for UST SC, Adedokun et al. (2022) [11], and VDZ IV, Rosario et al. (2015) [12], extended to SC application by Rosario et al. (2019) [13], were implemented as described. The identified Cmin–ER thresholds were 3.5 mg/L (UST SC), 10.4 mg/L (VDZ IV), and 34.3 mg/L (VDZ SC) [14].
To realise the incorporation of parameter uncertainty into dosing decisions, Monte Carlo simulations (n = 1000) were performed by sampling η-vectors from the multivariate normal distribution with mean equal to the MAP estimates and covariance equal to the ETC matrix. For each sampled parameter set, concentration-time profiles were simulated using mrgsolve for clinically relevant alternative dosing intervals (UST SC and VDZ IV: every 4–12 weeks; VDZ SC: every 1–3 weeks) until steady-state. Application of the predefined PTA-based decision rule enabled identification of the recommended dosing interval for each patient within the RCT.
Across all 27 tests, the algorithm recommended the expected dosing interval adjustment: shortening for underdosed, continuation for adequate, and extension for overdosed patients. For all drug/route combinations, PTA at the recommended interval exceeded 80%. Recommendations remained consistent across occasions, varying by at most one dosing interval step as further observations were iteratively included.
The framework was operationalised within the interventional arm of the prospective MOVE-IT workflow, in which enrolled patients undergo 4-weekly drug concentration measurements over 48 weeks. Dose recommendations are initiated after the second sample to enable iterative MIPD-guided dosing-interval optimisation, according to the described algorithm.
Conclusions
We established an uncertainty-informed MIPD framework for VDZ and UST and demonstrated its practical integration into a prospective RCT. By incorporating parameter uncertainty into Bayesian forecasting, the approach enables probabilistic, target-oriented dose individualisation for mAbs in IBD. Clinical evaluation within the ongoing RCT MOVE-IT will determine its impact on therapeutic outcomes.
References:
[1] K.H. Katsanos et al. Clin. Immunol. 206: 9–14 (2019).
[2] A.R. Bourgonje et al. Trends Pharmacol. Sci. 46: 9–19 (2025).
[3] K. Papamichael et al. Lancet Gastroenterol. Hepatol. 7: 171–185 (2022).
[4] X. Roblin et al. Expert Rev. Gastroenterol. Hepatol. 18: 575–586 (2024).
[5] F. Kluwe et al. Clin. Pharmacol. Ther. 109: 29–36 (2021).
[6] D. Kang et al. Korean J. Physiol. Pharmacol. 16: 97 (2012).
[7] C. Frimor et al. BMJ Open Gastroenterol. 12: (2025).
[8] A. Elmokadem et al. CPT Pharmacomet. Syst. Pharmacol. 8: 883–893 (2019).
[9] L.B. Sheiner, S.L. Beal. J. Pharm. Sci. 71: 1344–1348 (1982).
[10] R. Wolfinger. Biometrika 80: 791–795 (1993).
[11] O.J. Adedokun et al. Clin. Ther. 44: 1336–1355 (2022).
[12] M. Rosario et al. Aliment. Pharmacol. Ther. 42: 188–202 (2015).
[13] M. Rosario et al. J. Crohns Colitis 13: S357 (2019).
[14] C. Frimor et al. J. Crohns Colitis 20: jjaf231.1318 (2026).
Reference: PAGE 34 (2026) Abstr 12045 [www.page-meeting.org/?abstract=12045]
Poster: Clinical Applications