Enter Note Done
Go to previous page in this tab
Event
  • Scientific Session
  • Bookmark Icon
  • Advancing Next-Generation Risk Assessment Through Bayesian Uncertainty Characterization
  • Schedule
    Notes
  • Stars at Night Ballroom B2‚ Convention Center
    Set Timezone


Chair: Mary Jacketti, Applied Research Associates Inc.
Co-Chair: Chao Ji, CDC/ATSDR

Primary Endorser: Computational Toxicology Specialty Section


Endorser(s): Biological Modeling Specialty Section


Next generation risk assessment (NGRA) is defined as an exposure-led, hypothesis-driven risk assessment approach that integrates human-relevant in silico, in chemico, and in vitro methods to assess the possible hazards associated with chemicals and components. While NGRA offers a promising framework for advancing more human-relevant and mechanistically informed risk assessment, important scientific and practical challenges remain. These challenges include incomplete biological coverage across relevant cell types, pathways, endpoints, exposure scenarios, life stages, and susceptible populations; variability in assay reliability and reproducibility; uncertainties in in vitro to in vivo extrapolation (IVIVE) and toxicokinetic modeling; and a lack of standardized methods for integrating evidence generated from diverse new approach methodologies (NAMs). Given these challenges, NGRA applications must both transparently document qualitative uncertainty and quantitatively characterize uncertainty across the assessment. Qualitative characterization should identify key assumptions, data gaps, biological and model-domain limitations, confidence in assay relevance, and the role of expert judgment in evidence integration. Quantitative characterization should evaluate how uncertainty propagates through exposure estimates, assay-derived points of departure (PoDs), concentration-response modeling, toxicokinetic parameters, IVIVE assumptions, benchmark dose estimates, and margins-of-safety. However, to date, there are no standardized approaches for characterizing and propagating uncertainty across all NGRA inputs and decision points. As a result, most current NGRA applications rely on qualitative descriptions of uncertainty, without consistent quantitative methods for evaluating how uncertainties from individual assays, in silico models, and heterogeneous data integration influence the overall risk predictions. Bayesian inference can be used to address these challenges by explicitly representing uncertainty from multiple NGRA inputs and propagating that uncertainty through the assessment. Prior information from existing toxicological data, chemical similarity, high-throughput assays, mechanistic models, and expert judgment can be formally combined with new NAM-based evidence. The resulting posterior distributions provide quantitative estimates of uncertainty in parameters, predictions, and points of departure. Bayesian hierarchical models can also support NGRA by distinguishing measurement and model uncertainty from biological variability, allowing population-level differences in exposure, toxicokinetics, biological response, and susceptibility to be represented directly in the assessment.

This session will focus on how Bayesian approaches can be translated from technical methods into practical NGRA decision-making frameworks. The presentations will provide complementary perspectives on four key areas: quantifying human population variability and probabilistic toxicity values; applying Bayesian and model-integration approaches within NGRA workflows; combining multiple NAM-derived PoDs into unified uncertainty distributions; and employing practical, accessible Bayesian methods to support toxicological research. Together, these presentations will address the scientific basis, applied utility, and implementation considerations for Bayesian approaches in NGRA, including how uncertainty estimates may be interpreted, communicated, and incorporated into risk assessment decision-making.

Following the presentations, a moderated discussion period will invite audience participation and facilitate dialogue on practical application, regulatory relevance, and future development of Bayesian approaches in NGRA. Attendees will leave with a clearer understanding of how Bayesian methods can support transparent uncertainty characterization, improve confidence in NAM-based decisions, and help move NGRA toward more consistent use in regulatory and applied safety assessment contexts.

Session and Activity Type
Presentations