Wannee Kantasiripitak (1), Karen van Hoeve (2), Sebastian G. Wicha (3), Debby Thomas (1), Ilse Hoffman (2), Marc Ferrante (4,5), Séverine Vermeire (4,5), Erwin Dreesen (1)
(1) Department of Pharmaceutical and Pharmacological Sciences, University of Leuven, Belgium, (2) Department of Paediatric Gastroenterology, Hepatology, and Nutrition, University Hospitals Leuven, Belgium, (3) Department of Clinical Pharmacy, Institute of Pharmacy, University of Hamburg, Germany, (4) Department of Gastroenterology and Hepatology, University Hospitals Leuven, Belgium, (5) Department of Chronic Diseases and Metabolism, University of Leuven, Belgium
Introduction: Higher infliximab (IFX) concentrations during induction are associated with deep remission (DR; combined endoscopic and corticosteroid-free clinical remission) at six months in paediatric patients with inflammatory bowel diseases (IBD) [1]. An IFX trough concentration (TC) of 5.0 mg/L at week (w)12 was defined as the best target for DR [1]. Under the approved weight-based dosing strategy (5 mg/kg), patients with lower bodyweight are expected to have lower IFX TC since the relationship between bodyweight and the pharmacokinetic (PK) parameters of IFX is nonlinear [2]. Model-informed precision dosing (MIPD) has been advocated to appropriately account for PK variability during IFX treatment [3].
Objectives: (i) To develop an exposure-response (ER) model describing the relationship between IFX TC at w12 and probability of attaining DR, (ii) To evaluate the predictive ability of samples taken at different time points using different MIPD approaches, and the predictive ability of IFX clearance (CL) during induction towards DR.
Methods: Data from 31 paediatric IBD patients starting IFX induction therapy were obtained from a prospective study [1].
A logistic regression ER model with Emax function was developed. A bootstrap was performed (N=1000). An exposure target for MIPD was identified.
Eight population PK models of IFX in paediatric IBD patients were identified. The predictive performance of single-model and multi-model approaches (a model selection algorithm [MSA] and a model averaging algorithm [MAA]) using the models jointly was compared for predicting the target [4]. A?priori prediction (based on covariate data) and maximum a?posteriori (MAP) prediction (based on covariate data and TCs at w2 and/or 6) were compared using accuracy (relative bias; rBias) and precision (relative root mean square error; rRMSE) metrics. A weighting scheme based on the maximum likelihood estimate was used in the multi-model approaches of MAP. A ±20% rBias with a 95% confidence interval (CI) including zero was considered clinically acceptable. The rRMSE was to be as low as possible. The classification accuracy at the exposure target was evaluated.
Individual patients’ estimated IFX CL (at w0, 2, 6, and 12; using an average of models’ estimated CL) were compared between those attaining DR and not.
Results: Median age of patients was 14 years (range 4–18). DR was achieved in 18 patients (58%) at six months. A w12 IFX TC of 4.3 mg/L corresponded to a 50% probability of attaining DR (mean 4.5 [95% CI 1.1–7.4] mg/L). An IFX target of 5.0 mg/L was associated with only 54% probability of attaining DR (84% classification accuracy rate). A TC target of 7.5 mg/L, associated with 64% probability of attaining DR (71% classification accuracy rate), was identified as optimal.
A priori predictions with single-model and multi-model approaches were clinically unacceptable (rBias 31–175%, rRMSE 120–272%). Also, a priori prediction had the lowest classification accuracy among all prediction settings (median 51 [IQR 45–58]%; p<0.005). Also, most of the MAP predictions with w2 TC were clinically unacceptable (rBias 20–157%) with high rRMSE (105–282%). MAP prediction with w6 TC resulted in clinically acceptable rBias in five single models (rBias -18– -6%, rRMSE 39–61%) and both multi-model approaches (MSA: rBias -6%, rRMSE 41%; MAA: rBias +10%, rRMSE 43%). Compared to MAP predictions with w2 TC, prediction with w6 TC had a significantly higher classification accuracy (w2 TC: 68 [IQR 61–76]% vs w6 TC: 89 [IQR 82–90]%; p=0.009), and a numerically lower chance of a falsely predicted the w12 TC ≥7.5 mg/L (i.e., risk of loss of response) (w2 TC: 19 [IQR 15–25]% vs w6 TC: 7 [IQR 7–7]%; p=0.050). In general, MAP predictions with TCs at w2 and w6 were worse than the predictions with only w6 TC alone.
At w6, the IFX CL was significantly lower in patients who attained DR in comparison to those who did not (0.21 [IQR 0.16–0.26] L/day vs 0.26 [IQR 0.23–0.31] L/day; p=0.043). The area under the receiver operating characteristics curve (AUROC) of w6 CL was not significantly different from the AUROC of w12 TC in classifying DR (w6 CL: 0.722 [95%CI 0.534–0.910] vs w12 TC: 0.844 [95%CI 0.680–1.000]; p=0.215).
Conclusions: A 7.5 mg/L IFX TC at w12 should be targeted to optimise the probability of attaining DR. The success rate of target attainment could be improved by measuring IFX TC at w6 for early adoption of MIPD, as well as monitoring IFX CL.
References:
[1] Van Hoeve K et al. J Pediatr (2022) 240, 150–157.
[2] Fasanmade AA et al. Clin Ther (2011) 33, 946–964.
[3] Dubinsky MC et al. Inflamm Bowel Dis (2022) doi: 10.1093/ibd/izab285.
[4] Uster DW et al. Clin Pharmacol Ther (2021) 109, 175–183.
Reference: PAGE 30 (2022) Abstr 10141 [www.page-meeting.org/?abstract=10141]
Poster: Drug/Disease Modelling - Paediatrics