CIOMS Issues Draft Report on Best Practices for AI Integration in Pharmacovigilance

A key focus of the report is the effective use of vast, diverse datasets, ranging from EHRs and insurance claims to registries and data from connected devices.

The Council for International Organizations of Medical Sciences (CIOMS) has released a draft report proposing a framework to guide the responsible use of artificial intelligence (AI) in pharmacovigilance (PV).

This development reflects a growing global consensus on the importance of advanced technologies in ensuring drug safety, particularly during the COVID-19 pandemic, highlighting the limitations of conventional PV systems due to a dramatic surge in safety reports.

The report is the outcome of a three-year effort by CIOMS' Expert Working Group XIV, established following a concept paper published in February 2022. 

The working group brought together stakeholders from academia, pharmaceutical companies, regulatory agencies, and public health organizations to examine how AI can enhance PV activities. 

The initiative also aligns with increasing interest from international regulators, as noted during a DIA meeting earlier this year.

A key focus of the report is the effective use of vast, diverse datasets—ranging from electronic health records (EHRs) and insurance claims to registries and data from connected devices. 

According to CIOMS, integrating these data sources using AI can accelerate drug development, improve patient care, reduce healthcare costs, and generate actionable insights that would be difficult to uncover through traditional methods.

Guiding Principles for AI Integration

The draft outlines six core principles for implementing AI in PV, starting with a risk-based approach. The level of human oversight and regulatory scrutiny should be proportional to the risks posed by AI applications.

CIOMS stresses that a well-calibrated risk management strategy enables organizations to maximize AI's benefits while ensuring safety and accountability.

Despite the transformative potential of AI, the report emphasizes the irreplaceable role of human judgment. 

It describes two oversight models: "human-in-the-loop," where decisions are made collaboratively by humans and AI, and "human-on-the-loop," where AI makes autonomous decisions later reviewed by human experts. Such oversight mechanisms are vital for ensuring trust and reliability in AI-driven PV systems.

Ensuring Transparency, Performance, & Data Privacy

CIOMS also highlights the importance of transparency in AI processes, calling for clear documentation and communication of how AI systems operate and how decisions are reached. 

Rigorous performance evaluations are necessary to validate the effectiveness and safety of AI models. In parallel, safeguarding patient privacy remains paramount, with the report urging strict adherence to data protection laws and ethical standards.

The report calls for establishing accountability frameworks that assign clear responsibilities to all stakeholders—developers, implementers, regulators, and end-users. 

It also anticipates a shift in workforce dynamics, as AI automation takes over routine tasks traditionally handled by pharmacovigilance professionals. However, it notes that human expertise will continue to be essential for oversight, contextual analysis, and ethical guidance.

An appendix to the draft compares CIOMS' guiding principles with existing standards across regional and international bodies. 

This comparative analysis is intended to promote harmonization of AI governance in PV and encourage adoption of best practices globally.

Founded in 1949 by the World Health Organization (WHO) and UNESCO, CIOMS plays a critical role in shaping global medical research and public health standards.

Stay tuned for more such updates on Digital Health News.