Bayesian network analysis of risk classification strategies in the regulation of cellular products

Area of Science:

• Regulatory Science
• Biotechnology
• Data Science

Background:

• Cellular therapies require robust regulatory frameworks, but global consensus on risk classification is lacking.
• Existing regulatory approaches face challenges in effectively managing the risks associated with novel cell-based products.
• A need exists for data-driven methods to refine risk assessment in cell therapy regulation.

Purpose of the Study:

• To compare and evaluate risk classification strategies for cellular products from FDA, MHLW, and WHO using Bayesian network analysis.
• To assess the appropriateness of key risk factors and their impact on clinical safety within different regulatory frameworks.
• To propose improvements for the regulatory system of cellular products, focusing on risk-based classification and leveraging machine learning.

Main Methods:

• Bayesian network analysis was applied to compare risk classification strategies from major regulatory bodies.
• Real-world data was utilized to validate the developed Bayesian network models.
• A substudy specifically examined risk factors for chimeric antigen receptor (CAR) T cell therapy.

Main Results:

• The study identified areas for refining risk classification approaches for cellular products.
• Key risk factors, including CAR targets, tumor types, and costimulatory domains, were found crucial for CAR T cell therapy safety assessment.
• The analysis demonstrated the utility of Bayesian networks in regulatory decision-making for cell product risk classification.

Conclusions:

• There is a current gap in real-world data-driven regulatory frameworks and risk-based review methods for cellular products.
• Implementing machine learning, such as Bayesian networks, can enhance the assessment of cellular product safety.
• This research provides a foundation for improving global regulatory harmonization and safety oversight for advanced cell therapies.