Can Artificial Intelligence Improve Clinical Trial Enrollment in Polycythemia Vera?

Identifying the right candidates for clinical trials can take time. It is usually a manual process that requires medical resources. New data suggests that artificial intelligence may offer a solution.

While traditionally researchers find patients in manual chart reviews that can take over 30 minutes per patient or during routine office visits. The AI system can scan  4.7 million electronic health records in just one week. 

Can artificial intelligence improve clinical trial efficiency?

Synapsis AI is a medically trained system based on a large language model (LLM). It is designed to automate patient screening. 

In a recent study at the Cleveland Clinic, researchers used a medically trained artificial intelligence model to automate patient screening. This tool helps find eligible participants for a Phase III trial comparing a new treatment, givinostat, with the standard hydroxyurea for high-risk PV.

What is the clinical impact and accuracy of an AI model like Synapsis AI when screening for clinical trials?

The results showed a significant advantage in both speed and volume compared to standard human-led screening. While the medical team was able to identify only nine patients over 12 months using traditional methods, the AI tool achieved a sevenfold increase in candidate identification.

Key findings regarding the system's performance included:

• Accuracy: When research personnel verified the AI's choices, they confirmed a positive predictive value of 100%, meaning every patient identified was truly eligible.
• Faster timelines: By identifying 50 eligible patients before the trial closed, the tool significantly reduced clinical staff workload.
• Accelerated drug development: The time needed to develop new life-saving therapies could be shorter.

While developments are being made, changes in how we implement them are still needed

While the results are promising, there are adjustments to be made in how the new tools are incorporated into daily hospital workflows. While building AI infrastructure is a significant first step, connecting these systems directly to electronic health records and real-time trial databases will be necessary to achieve a measurable impact across different cancer types and institutions.

Another example of how AI can be securely incorporated into healthcare is HealthTree, with AI validators can review and verify records from different facilities a patient has seen and put them all together so they can keep a better track of their health. 

Source: LLM Tool Significantly Reduces Participant Screening Burdens, Improves Enrollment for Phase III Trial in Polycythemia Vera