EMA details lifecycle approach to AI/ML drug development in new reflection paper
Regulatory News | 20 July 2023 |
In a draft reflection paper published this week, European regulators said they view the use of artificial intelligence/machine learning (AI/ML) to develop and market drugs in a total product lifecycle and risk-based context. The reflection paper provides insights on when AI/ML technology may be used to develop products and how such technologies can be used in the postmarket setting.
The reflection paper was jointly developed by the European Medicines Agency’s (EMA) Big Data Steering Group (BDSG) and the Heads of Medicines Agencies (HMA).
“With this paper, we are opening a dialogue with developers, academics, and other regulators, to discuss ways forward, ensuring that the full potential of [AI/ML] innovations can be realized for the benefit of patients’ and animal health,” said Peter Arlett, EMA’s head of data analytics and methods, co-chair of the BDSG.
The purpose of its paper is to consider the use of AI/ML throughout the lifecycle of drugs including during product development, authorization, and post-authorization, EMA said. Given rapid developments in the AI/ML field, the agency’s aim is to reflect on the scientific principles that are relevant for regulatory evaluation when such emerging technologies are used to support developing new products.
“It is crucial to identify aspects of AI/ML that would fall within the remit of EMA or the National Competent Authorities of the Member States as the level of scrutiny into data during assessment will depend on this remit,” the paper states. “This reflection paper focuses only on the use of AI in the medicinal product lifecycle and any references to qualification of novel methodologies for medicines development, interaction etc. are to be understood within this scope.”
“However, medical devices with AI/ML technology can be used within the context of clinical trials to generate evidence in support of a marketing authorization application and/or can be combined with the use of a medicinal product,” the paper notes. “In such cases EMA will be involved in the assessment on whether the characteristics of the device is adequate to generate evidence, supporting a EU marketing authorization. Similarly, if a device is used to provide recommendations in the Summary of Product Characteristics, e.g. on posology or monitoring, the EMA will assess all relevant aspects of the proposed combined use.”
EMA noted that AI/ML tools can potentially support the acquisition, transformation, analysis and interpretation of data across the medical product’s lifecycle. For example, EMA said that AI/ML modelling could be used to replace, reduce and refine the use of animal models during the preclinical development. In clinical trials, AI/ML systems could potentially support the selection of patients based on certain disease characteristics or other clinical parameters. Regulators also said AI/ML tools have the potential to support data recording and analyses which can be used by regulators to make marketing decisions.
“At the marketing-authorization stage, AI applications include tools to draft, compile, translate, or review data to be included in the product information of a medicine,” said EMA. “In the post-authorization phase, such tools can effectively support, for example, pharmacovigilance activities including adverse event report management and signal detection.”
“This range of applications brings with it challenges such as the understanding of the algorithms, notably their design and possible biases, as well as the risks of technical failures and the wider impact these would have on AI uptake in medicine development and health,” the agency added.
To account for the challenges, the reflection paper emphasizes that a human-centric approach should guide all development and deployment of AI/ML technologies. It also states use of such technology in the medicinal product lifecycle should always occur in compliance with the existing legal requirements, consider ethics and ensure due respect of fundamental rights.
If AI/ML technology is used in the context of medical product development, evaluation or monitoring, and is expected to have an impact on the benefit-risk assessment of the product, product developers should ask for early input from the agency.
“Advice on risk management will be further reflected in future regulatory guidance, as the impact of system malfunction or degradation of model performance can range from minimal to critical or even life-threatening,” the reflection paper notes. “The degree of risk may depend not only on the AI technology, but also on the context of use and the degree of influence the AI technology exerts.”
“In addition, the degree of risk may vary throughout the lifecycle of the AI-system,” the paper added. “Marketing authorization applicants or marketing authorization holders (MAHs) planning to deploy AI/ML technology are expected to consider and systematically manage relevant risks from early development to decommissioning.”
HMA/EMA plan to hold a workshop 20-21 November 2023 to discuss the draft reflection paper and stakeholders can comment on the document until 31 December.
AI/ML draft reflection paper
The reflection paper was jointly developed by the European Medicines Agency’s (EMA) Big Data Steering Group (BDSG) and the Heads of Medicines Agencies (HMA).
“With this paper, we are opening a dialogue with developers, academics, and other regulators, to discuss ways forward, ensuring that the full potential of [AI/ML] innovations can be realized for the benefit of patients’ and animal health,” said Peter Arlett, EMA’s head of data analytics and methods, co-chair of the BDSG.
The purpose of its paper is to consider the use of AI/ML throughout the lifecycle of drugs including during product development, authorization, and post-authorization, EMA said. Given rapid developments in the AI/ML field, the agency’s aim is to reflect on the scientific principles that are relevant for regulatory evaluation when such emerging technologies are used to support developing new products.
“It is crucial to identify aspects of AI/ML that would fall within the remit of EMA or the National Competent Authorities of the Member States as the level of scrutiny into data during assessment will depend on this remit,” the paper states. “This reflection paper focuses only on the use of AI in the medicinal product lifecycle and any references to qualification of novel methodologies for medicines development, interaction etc. are to be understood within this scope.”
“However, medical devices with AI/ML technology can be used within the context of clinical trials to generate evidence in support of a marketing authorization application and/or can be combined with the use of a medicinal product,” the paper notes. “In such cases EMA will be involved in the assessment on whether the characteristics of the device is adequate to generate evidence, supporting a EU marketing authorization. Similarly, if a device is used to provide recommendations in the Summary of Product Characteristics, e.g. on posology or monitoring, the EMA will assess all relevant aspects of the proposed combined use.”
EMA noted that AI/ML tools can potentially support the acquisition, transformation, analysis and interpretation of data across the medical product’s lifecycle. For example, EMA said that AI/ML modelling could be used to replace, reduce and refine the use of animal models during the preclinical development. In clinical trials, AI/ML systems could potentially support the selection of patients based on certain disease characteristics or other clinical parameters. Regulators also said AI/ML tools have the potential to support data recording and analyses which can be used by regulators to make marketing decisions.
“At the marketing-authorization stage, AI applications include tools to draft, compile, translate, or review data to be included in the product information of a medicine,” said EMA. “In the post-authorization phase, such tools can effectively support, for example, pharmacovigilance activities including adverse event report management and signal detection.”
“This range of applications brings with it challenges such as the understanding of the algorithms, notably their design and possible biases, as well as the risks of technical failures and the wider impact these would have on AI uptake in medicine development and health,” the agency added.
To account for the challenges, the reflection paper emphasizes that a human-centric approach should guide all development and deployment of AI/ML technologies. It also states use of such technology in the medicinal product lifecycle should always occur in compliance with the existing legal requirements, consider ethics and ensure due respect of fundamental rights.
If AI/ML technology is used in the context of medical product development, evaluation or monitoring, and is expected to have an impact on the benefit-risk assessment of the product, product developers should ask for early input from the agency.
“Advice on risk management will be further reflected in future regulatory guidance, as the impact of system malfunction or degradation of model performance can range from minimal to critical or even life-threatening,” the reflection paper notes. “The degree of risk may depend not only on the AI technology, but also on the context of use and the degree of influence the AI technology exerts.”
“In addition, the degree of risk may vary throughout the lifecycle of the AI-system,” the paper added. “Marketing authorization applicants or marketing authorization holders (MAHs) planning to deploy AI/ML technology are expected to consider and systematically manage relevant risks from early development to decommissioning.”
HMA/EMA plan to hold a workshop 20-21 November 2023 to discuss the draft reflection paper and stakeholders can comment on the document until 31 December.
AI/ML draft reflection paper
© 2025 Regulatory Affairs Professionals Society.