Can AI improve clinical trial recruitment? Yes, and here's proof.

Quick summary:

Retroactive use of AI data analysis and patient matching tool led to significant improvements in the patient recruitment process for 3 oncology studies.

• Efficiency: Significantly reduces the time required for patient identification.

• Increased Eligibility: Identifies more potentially eligible patients, ensuring a more representative sample.

• Reduced Resource Expenditure: Saves valuable human resources, making them available for other crucial tasks.

• Faster Access to Treatment: Shortens the time between eligibility and actual screening, benefiting patients.

• Improved Site Selection: Helps sponsors choose the right sites for their trials.

• Reduced Biases: Minimises biases in patient selection, enhancing the generalisability of patients selected

An image generated by Google's Deep Mind.

Delays in recruiting people for clinical trials can have huge impact on the costs, if there aren’t enough patients, it simply can’t complete. In some cases, each day a trial runs over, there are an estimated $8million dollars lost. Delays and difficulties in identifying eligible patients, and potential biases in patient selection continue to be major concerns for sponsors, CROs, sites, and patients. An AI tool focused on physical and digital data indexing helps overcome these challenges.

Methods:

The tool was retrospectively applied to two completed oncology studies (one in breast cancer and another in lung cancer) and one study that failed to enroll patients with lung cancer at the Comprehensive Blood and Cancer Center (yes, we know retroactive data isn't quite proof in action, but when people are reluctant to adopt technology, how else can it be proven?). The study compared the results achieved through standard prescreening practices with those using AI. Key outcome variables included the number of potentially eligible patients identified and the time between eligibility and identification.

Results:

• In trials that successfully enrolled patients, AI increased the number of potentially eligible patients by 24% to 50% compared to standard practices.

• The AI did not miss any patients identified as eligible by standard practices.

• For a non-enrolling trial, neither approach identified suitable patients.

• The time between eligibility and identification was significantly reduced with the AI, taking only minutes compared to several days for standard practices.

Conclusion:

The study highlights the potential of AI in clinical trial recruitment. By integrating artificial intelligence, agencies can accelerate the identification of eligible patients, reduce resource expenditure, and improve the quality of their services. AI-driven tools represent a promising future for clinical trial recruitment and offer a win-win for sponsors, sites, and patients alike.

https://assets-global.website-files.com/5d89d3efdbb3a843ec64b147/651f33309e2ae9f5c5434014_Mendel%20Article%20-%202020%20Improving%20Clinical%20Trial%20Participation.pdf