International group proposes best practices for AI in pharmacovigilance

The Council for International Organizations of Medical Sciences (CIOMS) has issued a draft report detailing a set of best practices to guide the integration of artificial intelligence in pharmacovigilance (PV) activities.

The report was partly prompted by COVID-19, which “emphasized the need for advanced methods in PV, as it has led to a significant rise in safety reports.”

The report is the result of three years of work. In February 2022, the group released a concept paper suggesting the formation of an Expert Working Group XIV to investigate the use of AI in PV. The group consisted of academics, pharmaceutical companies, regulatory authorities, and national and international organizations.

During a DIA meeting last February, a representative from the US Food and Drug Administration (FDA) mentioned that there is increasing global interest among regulators in the potential use of AI for pharmacovigilance. (RELATED: Official: FDA modernizing pharmacovigilance oversight with AI tools, Regulatory Focus 6 February 2024).

The report highlights that a crucial factor for the success of AI in PV is the ability to link and analyze large volumes of diverse data from various sources, including electronic health records (EHRs), claims databases, registries, the Internet of Things (IoT), and connected devices. Effectively utilizing health data can facilitate the faster development of new treatments, enhance patient outcomes, and lower healthcare costs. Additionally, this capability has the potential to reveal valuable and actionable insights that may not have been discovered otherwise.

The report outlines six guiding principles for establishing an AI system in pharmaceutical governance for activities related to PV. These principles include:

A risk-based approach that prioritizes risks associated with the implementation of AI;
Human oversight is needed to ensure that human judgment complements AI decisions;
Guaranteeing that AI systems are effective and reliable through performance evaluation;
Promoting clear communication about the AI system's processes and outcomes;
Safeguarding sensitive information and maintaining patient confidentiality; and
Establishing frameworks that hold stakeholders responsible for the AI system's functions.

The report notes that “a sound risk-based approach, in which the human oversight in the development and deployment of AI is commensurate with these risks, enables organisations to make the most of AI capabilities while ensuring that neither patient safety nor PV stakeholders are adversely affected.”

An AI system should also include human oversight to optimize the performance of AI in PV and increase trustworthiness and accountability. “The extent and nature of human oversight for an AI solution should be risk-based, incorporating quality assurance principles. The human oversight might be “human-in-the-loop" meaning that the decision is the end results of a human-machine interaction, while in “human-on-the-loop", the machine autonomously makes a decision or otherwise returns a result that is checked by a human,” said the report.

The report also notes that as AI capabilities become more integrated, there will likely be a decreased reliance on teams of PV professionals. This is due to automation and AI taking over some tasks traditionally performed by these experts.

The report includes an appendix comparing the CIOMS Working Group XIV guiding principles for artificial intelligence across regional and country government institutions and international organizations.

CIOMS is an international non-profit organization established in 1949 by the World Health Organization and UNESCO. It aims to advance public health by providing guidance on medical product development.

The deadline for submitting comments is 6 June. Comments should be submitted using this form.

CIOMS draft report

International group proposes best practices for AI in pharmacovigilance

Related topics

Recommended content

Welcome to the new RAPS Digital Experience