Phesi founder warns AI ‘is scaling, not solving’ clinical trial flaws

Phesi has published an analysis of more than 600,000 clinical trial protocols, which concludes that fewer than one in three are linked to publicly documented patient data and outcomes.

The company said only 29.3% of protocols across all disease areas could be connected to usable patient outcome data. In oncology, the figure was 30.9% across 116,746 protocols.

Phesi said the findings raise concerns about the quality of historical protocols used to guide future clinical trial design, particularly as AI tools become increasingly involved in protocol development and optimisation.

Phesi president and founder Dr Gen Li, warned: “AI can ingest far more historical templates, but without the right logic or judgment it may fail to make that connection,” warned Li.

“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."

Further analysis focused on breast cancer, identified by the company as the world’s most studied disease area over the past five years. Despite the volume of research activity, only 31.2% of 15,977 breast cancer protocols were linked to usable patient outcome data, it revealed.

According to Phesi, the gap between protocol design and real-world patient outcomes risks creating flawed assumptions in future studies. The company said unsuccessful or incomplete trials, alongside studies where patient data is not fully reported, contribute to a disconnect between intended protocol design and observed clinical outcomes.

Phes chief Li said the issue represents a broader structural challenge in drug development. He warned that AI systems may scale underlying weaknesses if the datasets used are not properly contextualised against patient populations and treatment outcomes.

Phesi said its Trial Accelerator platform is designed to address this issue by linking protocol information with contextualised real-world and clinical patient data. The platform draws on data from 375 million patients across 232 countries and territories, using information from clinical trials, observational studies and registries worldwide.

Related Content

Pin It on Pinterest

Share This