In the multicenter PolyDeep advance 3 randomized clinical trial, PolyDeep-assisted colonoscopy did not improve adenoma detection rate in colorectal cancer screening: ADR was 60.8% with AI vs 57.7% with conventional colonoscopy (aRR 1.05, 95% CI 0.94-1.17) among 827 analyzed patients. The study also found no significant gains in serrated lesion, polyp, advanced adenoma, or advanced serrated lesion detection.

Why It Matters To Your Practice

• CADe tools may not automatically improve key detection outcomes in real-world screening and surveillance colonoscopy.

• This trial focused on patients aged 40-79 undergoing colonoscopy after positive FIT or for surveillance, making the findings relevant to everyday colorectal cancer screening workflows.

• For clinicians evaluating AI adoption, the result suggests performance claims should be judged on randomized outcome data, not just technical promise.

Clinical Implications

• ADR, the primary endpoint, was not significantly different with PolyDeep vs conventional colonoscopy.

• Secondary endpoints were also neutral: SLDR 8.8% vs 7.2%, PDR 69.6% vs 68.8%, AADR 23.6% vs 24.4%, and ASLDR 4.9% vs 4.8%.

• Subgroup analyses by indication, endoscopist ADR, and bowel cleansing did not show a meaningful benefit.

Insights

• The trial randomized 857 patients and analyzed 827, with comparable baseline characteristics between groups.

• Mean age was 61.7 years, 490 participants were male, and 552 colonoscopies followed a positive FIT.

• In a setting with already high detection performance, incremental benefit from CADe may be harder to demonstrate.

The Bottom Line

• In this RCT, PolyDeep did not improve ADR or other detection metrics in a colorectal cancer screening program.

• For now, clinicians should view CADe as a tool that may support practice in some settings, but not as a proven way to raise ADR across screening colonoscopy.