EMA and FDA set common principles for AI in medicine development
The European Medicines Agency (EMA) and the US FDA have jointly developed ten principles for good Artificial Intelligence (AI) practice in life cycle management for medicinal products. The purpose of these principles is to permit AI used in the development of medicinal products to be expertly managed, including through mitigation of risk. It is anticipated that the principles will be supplemented by EU guidance reflecting relevant existing and future EU regulations.
Although the principles are not binding they reflect what may be anticipated to be the priorities that the EMA will apply in relation to the development and use of AI technologies in medicinal products.
The principles are :
Human-centric by design - development and use of AI technologies should be based on ethical and human-centric values;
Risk-based approach - the development and use of AI technologies should follow a risk-based approach with proportionate validation, risk mitigation, and oversight based on the context of use and determined model risk;
Adherence to standards - AI technologies should adhere to relevant legal, ethical, technical, scientific, cybersecurity, and regulatory standards, including Good Practices (GxP);
Clear context of use - AI technologies should have a well-defined context of use within or in relation to a medicinal product, including definition of its role and an explanation of why it is being used in a particular context;
Multidisciplinary expertise - multidisciplinary expertise covering both the AI technology and its context of use should be integrated throughout the life cycle of the technology;
Data governance and documentation – sources of data, their provenance, processing steps, and analytical decisions should be documented in a detailed, traceable, and verifiable manner, in line with GxP requirements. Appropriate governance, including provisions governing privacy and protection of sensitive data, should be maintained throughout the life cycle of the technology;
Model design and development practices - the development of AI technologies should follow best practices in model and system design and software engineering and leverage data that is fit for use, considering interpretability, explainability, and predictive performance;
Risk-based performance assessment - risk-based performance assessments should be undertaken to evaluate the complete system including human-AI interactions, using fit for use data and metrics appropriate for the intended context of use. These should be supported by validation of predictive performance through appropriately designed testing and evaluation methods;
Life cycle management - risk-based quality management systems should be implemented throughout the life cycle of the AI technology. This should include the support of capture, assessment, and addressing of issues. The AI technology should undergo scheduled monitoring and periodic re-evaluation to ensure adequate performance (e.g., to address data drift);
Clear, essential information - plain language should be used to present clear, accessible, and contextually relevant information to the intended audience, including users and patients, regarding the context of use, performance, limitations, underlying data, updates, and interpretability or explainability of the AI technology.