New EMA Reflection Paper on the use of Artificial Intelligence

Recommendation

22-24 May 2024
Copenhagen, Denmark

IT / OT Infrastructure Qualification and Operation in a GMP Environment

The topic of artificial intelligence (AI) has been gaining more and more importance recently. The success of ChatGPT and the corresponding reporting have made the topic accessible to a broad public.

But also in the life science industry, the topic has been playing an important role for years in the entire life cycle of medicinal products. The use of AI in the diagnosis of skin cancer is one of the better-known examples. In addition, there are now also more and more applications of AI and ML (machine learning) in the production of drugs, e.g. in the visual inspection for particles in parenterals.

After the FDA had already published a first discussion paper on the use of AI at the beginning of 2023, the EMA (European Medicines Agency) followed on 19 July 2023 with a "Reflection Paper". The comment period runs until 31. December 2023. This document describes the EMA's current experience in this environment, where scientific knowledge is rapidly evolving. It is important to read the document in the context of the legal requirements and the overarching EU principles on AI, data protection and medicines regulation.

General considerations

In general, it is mentioned that AI and ML, if used correctly, can effectively support the aquisition, transformation, analysis and interpretation of data within the medicinal products lifecycle. In doing so, a risk-based approach should enable developers to proactively define risks to be managed throughout the lifecycle of AI and MS tools. Recommendations on risk management will be reflected in future regulatory guidance. It is important to bear in mind that the degree of risk depends not only on the AI technology, but also on the context of use and the degree of influence that the AI technology exerts.

A key principle is that it is the responsibility of the marketing authorisation applicant or MAH to ensure that any algorithms, models, datasets, etc. used are fit for purpose and meet ethical, technical, scientific and regulatory standards. 

Content of the document

The main part of the document addresses the following topics:

• AI in the lifecycle of medicinal products
• Regulatory interactions
• Technical aspects
• Governance
• Data protection
• Integrity aspects
• Ethical aspects and trustworthy AI

"AI in the lifecycle of medicinal products":

• Drug discovery
• Non-clinical development
• Clinical trials
• Precision medicine
• Product information
• Manufacturing
• Post-authorisation phase

and "Technical aspects" are further subdivided :

• Data acquisition and augmentation
• Training, validation, and test data
• Model development
• Performance assessment
• Interpretability and explainability
• Model deployment

Final remark

The use of AI in the life cycle of medicines should always be in accordance with existing legal requirements, taking into account ethics and its underlying principles, and with due respect for fundamental rights. A human-centred approach should be taken in the development and use of AI and ML.

Sources:
EMA Reflection paper on the use of Artificial intelligence in the lifecycle of medicines

Draft reflection paper