AI tools are scaling flaws in clinical trials instead of solving them, analysis shows. 

Analytics and data company Phesi revealed the findings from a global analysis of more than 600,000 clinical trial protocols.  

Results show fewer than one in three (29.3%) protocols are linked to publicly documented patient data and outcomes, with only 30.9% of 116,746 protocols in oncology linked to usable patient data. 

Phesi also conducted analysis into breast cancer, the world’s most studied disease over the past five years. Despite the scale and maturity of research in breast cancer, just 31.2% of 15,977 protocols are linked to trials with usable patient data.  

The findings underline that, even in the most data-rich disease area, high volumes of research do not automatically create the reliable, outcome-driven evidence base needed for future trial design or AI models. 

“This is a systemic issue in drug development,” said Dr Gen Li, President and Founder of Phesi. 

“It is standard practice to design protocols based on existing, similar protocols, but when those templates are disconnected from the target patient population, they contribute to amendments and recruitment challenges.  

“In the past, protocol writers could be selective about the protocols they used and apply human judgment to the connection between design and the target patient population. AI can ingest far more historical templates, but without the right logic or judgment it may fail to make that connection.  

“A clinical protocol is effectively a business plan for an investment of tens or hundreds of millions of dollars, so AI must be guided by the right data foundations. In essence, flaws are being scaled, not solved.” 

“There is huge opportunity for AI to optimise clinical development,” Li later added, “but only when the platforms being used as the basis for AI can identify protocols with patients and outcomes reported.”