Hyeseon Jeon 1, Chaebin Yoon 1, Min-Gul Kim 2,3, Soyoung Lee 1,4, Jung-woo Chae 1,4,5, Hwi-yeol Yun 1,5, Andrew Hooker 6, Jae Hyun Kim 7
1 College of Pharmacy, Chungnam National University (, Republic of Korea), 2 Department of Pharmacology, Jeonbuk National University Medical School (Jeonju, Republic of Korea), 3 Centre for Clinical Pharmacology and Biomedical Research Institute, Jeonbuk National University Hospital (Jeonju, Republic of Korea), 4 Institute of Drug Research and Development, Chungnam National University (, Republic of Korea), 5 Department of Bio-AI Convergence, Chungnam National University (, Republic of Korea), 6 Department of Pharmacy, Uppsala University (Uppsala, Sweden), 7 School of Pharmacy and Institute of New Drug Development, Jeonbuk National University (Jeonju, Republic of Korea)
[Background] PopPK modeling using therapeutic drug monitoring (TDM) data from routine practice is often challenged by sparse sampling. In such settings, literature-informed priors can improve model stability [1], but inappropriate prior specification may lead to excessive prior dependency and reduce the extent to which the model reflects observed patient data. We aimed to develop a robust infliximab popPK model for pediatric inflammatory bowel disease (IBD) patients by leveraging literature-informed prior information, and to compare prior specification strategies based on individual published models versus meta-analytically combined priors.
[Methods] Infliximab TDM data from 31 patients were analyzed (mean age 15.45 ± 3.61 years; 25 males/6 females; Crohn’s disease [CD] n=26, ulcerative colitis [UC] n=5). Published PopPK models were screened based on similarity in age distribution, disease composition, and overlap of key covariates with our dataset; models with insufficient information were excluded. Candidate priors were derived from selected studies [2-4]. PopPK modeling was performed in NONMEM, with PsN and R used for model diagnostics and comparisons. Two prior approaches were evaluated: (1) priors based on parameter estimates from individual publications and (2) priors combined via meta-analysis. All four candidate models were estimated and compared comprehensively using prediction performance, estimation stability, physiological plausibility, visual predictive checks (VPC), objective function value (OFV), shrinkage, and precision metrics (RSE). To quantify the contribution of prior information relative to the observed data, we calculated the Np value following Shoji et al. [5]. This represents the relative information content of the prior for each parameter, expressed as how many times as informative the prior is as the observed data. It was derived from parameter variances obtained using the NONMEM $DESIGN subroutine with and without prior information [6]. Np≫1 was interpreted as prior-dominant estimation, whereas Np ≪ 1 indicated data-dominant estimation.
[Results] A robust PopPK model of infliximab for pediatric IBD patients was developed using sparse TDM data. Among the four candidate models, the meta-analysis based model was selected as the final model based on an integrated assessment of predictive diagnostics (including visual predictive checks, objective function value, shrinkage and relative standard errors). The parameter estimates were physiologically reliable, including CL 0.355 L/day, V1 3.04 L, and V2 1.89 L, with acceptable precision for major structural parameters. The final model showed smaller Np values than the Xiong-based model despite estimating one additional parameter, indicating lower prior dependency while maintaining adequate model performance. Although the Bauman-based model showed the smallest Np values, these should be interpreted cautiously because multiple parameters were fixed and VPC performance was inadequate.
[Conclusions] Meta-analytic prior integration provided a better balance between prior information and observed data than priors derived from individual studies, reducing prior dependency while preserving model stability, physiological plausibility, and predictive performance in sparse real-world pediatric infliximab TDM data.
*Soyoung Lee, Jung-woo Chae, Hwi-yeol Yun, Andrew Hooker and Jae Hyun Kim contributed equally to this work as co-correspondances.
References:
[1] Chan Kwong, Anna H-XP, et al. “Prior information for population pharmacokinetic and pharmacokinetic/pharmacodynamic analysis: overview and guidance with a focus on the NONMEM PRIOR subroutine.” Journal of Pharmacokinetics and Pharmacodynamics 47.5 (2020): 431-446.
[2] Fasanmade, Adedigbo A., et al. “Pharmacokinetic properties of infliximab in children and adults with Crohn’s disease: a retrospective analysis of data from 2 phase III clinical trials.” Clinical therapeutics 33.7 (2011): 946-964.
[3] Bauman, Laura E., et al. “Improved population pharmacokinetic model for predicting optimized infliximab exposure in pediatric inflammatory bowel disease.” Inflammatory Bowel Diseases 26.3 (2020): 429-439.
[4] Xiong, Ye, et al. “Real‐world infliximab pharmacokinetic study informs an electronic health record‐embedded dashboard to guide precision dosing in children with Crohn’s disease.” Clinical Pharmacology & Therapeutics 109.6 (2021): 1639-1647.
[5] Shoji, Satoshi., et al. “Evaluation of information of prior relative to current data in analysis with prior.” [Poster presentation] 2014 PAGE, Spain. https://www.page-meeting.org/wp-content/uploads/pdf_assets/9748-PAGE_2014_Poster_v1.pdf
[6] Bauer, Robert J., Andrew C. Hooker, and France Mentre. “Tutorial for $ DESIGN in NONMEM: Clinical trial evaluation and optimization.” CPT: pharmacometrics & systems pharmacology 10.12 (2021): 1452-1465.
Reference: PAGE 34 (2026) Abstr 11869 [www.page-meeting.org/?abstract=11869]
Poster: Drug/Disease Modelling - Other Topics