Right Patient, Right Trial: How AI Can Transform Clinical Trial Recruitment

Finding the best patients for clinical trials is one of the pharmaceutical industry’s most persistent challenges. Recruitment can take months or even years, delaying research and limiting patients’ access to investigational treatments. However, AI may hold the answer to make the process faster, more efficient and potentially more inclusive.

“Anyone who’s considering a clinical trial is making a big decision. AI can help us create more awareness, identify eligible patients, and streamline the process”

Melinda Rottas, Head of Optimisation, Analytics and Recruitment Solutions at Pfizer

Pfizer is experimenting with AI-assisted recruitment to identify eligible patients, optimise the locations of the trials and streamline the design of studies. Rather than relying on manual review of medical records, AI can analyse clinical information against complex eligibility criteria and identify candidates who might otherwise have been overlooked.

The potential benefits are supported by independent research. Researchers from the US National Institutes of Health developed TrialGPT in 2024, an AI system that matches patient information with clinical trial criteria. TrialGPT achieved 87.3% accuracy when assessing individual eligibility criteria and in a small pilot, clinicians reported a reduction in screening time of 42.6% without a significant difference in accuracy.

For pharma companies, greater recruitment efficiency could directly lead to faster trial completion, reduced administrative burdens and shorter drug development timelines. Furthermore, AI could help address inequalities in clinical research. Traditional recruitment often relies on established research centres and referral networks. By analysing where eligible patient populations receive care, AI could help sponsors identify overlooked communities and select more representative trial sites.

More diverse clinical trial populations generate safety evidence across a broader range of demographic and clinical characteristics and will improve understanding of how medicines affect different groups. This is not to say AI is not susceptible to bias; indeed, systems trained on incomplete or unrepresentative healthcare data could reinforce existing inequalities rather than mitigate them. But if implemented responsibly, AI could help clinical research become faster, more accessible and more representative.

For pharmacovigilance professionals, that last benefit may prove particularly valuable: better representation during development could mean better safety knowledge when medicines reach the wider population, yielding drug safety benefits beyond simple efficiency and automation.