Uncertainty quantification for complex models supporting regulatory decision making

Study type

• Study type:
  Not applicable
• Scope of the study:
  Method development or testing
• Data collection methods:
  Secondary use of data

Study design details

• Study design:
  Literature review
• Main study objective:
  1. To identify and characterize existing UQ methods used in PBPK, QSP, and ML models within the framework of regulatory decision-making in drug development. ; 2. To explore UQ methodologies from broader disciplines, including statistics, toxicology, engineering, and analytical chemistry, that may be adaptable to these complex modeling approaches.
• Data analysis plan:
  Characteristics of articles and proposed methods were described by using descriptive analyses such as line charts and bar charts. The collected categorical variables were summarized by providing absolute and relative frequencies. To enable an overview of the uncertainty quantification methods, model names as they were proposed by the authors were classified into broader classes of models. The main characteristics of these suggested categories are introduced, and examples for the categories were provided. All statistical analyses were performed in R (version 4.5.1).
• Summary results:
  Conclusion and discussion ; This review shows a clear gap in the availability of well-established methods for UQ tailored to complex models in drug development and regulatory evaluation, as illustrated in Figure 8. Nonetheless, a diverse range of UQ techniques originating from disciplines such as mathematics and statistics, engineering, analytical chemistry, and toxicology holds substantial potential for application in this context, as illustrated in Figure 7. Among these, global ; sensitivity analysis, uncertainty propagation, and Bayesian approaches emerged as the most frequently reported and widely applied methods. The findings underscore the critical importance of adapting adequate UQ methodologies to improve the reliability and robustness of PBPK, QSP, and ML models. By identifying promising ; candidate methods from both within and outside drug development, this review lays the groundwork for targeted simulation studies. These next steps will provide a robust evidence base for selecting and refining UQ methods that can meaningfully enhance decision-making in drug development and evaluation.