Speaker
Description
Abstract
Background: Physiologically based pharmacokinetic (PBPK) models are increasingly used as surrogates for clinical trials in regulatory submissions, yet features distinguishing regulatory-adopted models from academic research remain poorly characterized. This study aimed to identify factors associated with regulatory adoption of PBPK drug-drug interaction (DDI) models through systematic literature analysis.
Methods: We established a 10-dimension framework encompassing regulatory orientation, methodological rigor, and clinical translatability, informed by US Food and Drug Administration (FDA) credibility assessment principles and regulatory review reports. A reasoning-enabled large language model (LLM; DeepSeek-R1) with human-in-the-loop verification extracted features from 186 cytochrome P450 (CYP) enzyme-mediated PBPK-DDI publications. Statistical analyses included binary comparison, ternary gradient analysis, and Firth penalized logistic regression.
Results: The extraction system achieved 93% accuracy across 17 subdimensions. Among 186 publications, 38 (20.4%) were classified as regulatory-focused based on explicit regulatory submission evidence. Binary comparison revealed significant differences in 10 of 15 features, with in-house data usage (odds ratio [OR]=39.1), direct conclusion orientation (OR=18.5), and industry funding (OR=18.3) showing the strongest associations. Validation rigor and predictive performance did not significantly differentiate the groups. Multivariable regression confirmed industry funding (adjusted OR=10.25), commercial software (adjusted OR=6.74), and development stage clarity (adjusted OR=12.50) as independent associated factors.
Conclusions: Regulatory adoption is primarily associated with factors indicating whether research is embedded within a specific drug development and regulatory decision-making context, rather than with traditional technical quality metrics. These findings provide quantitative support for the fit-for-purpose principle and offer an evidence-based checklist for enhancing regulatory translatability of PBPK models.
