ScreenTrustCAD

ScreenTrustCAD (NCT04778670; organisation id STGKS001) was a prospective, population-based, paired-reader, non-inferiority study at Capio Sankt Göran Hospital in Stockholm, sponsored by Karolinska University Hospital with Lunit as a collaborator and part funder. Consecutive women aged 40 to 74 without breast implants attending population screening in the hospital's catchment were included. Every examination was read independently by two radiologists and by Lunit INSIGHT MMG. The primary outcome was screen-detected breast cancer within three months of mammography, and the primary analysis tested whether double reading by one radiologist plus AI was non-inferior (margin 0.15 relative reduction in cancer diagnoses) to standard double reading by two radiologists; AI alone and triple reading by two radiologists plus AI were also compared.

From 1 April 2021 to 9 June 2022, 58,344 women were screened and 55,581 were included. Screen-detected cancer followed an initial positive read in 269 women (0.5 percent). One radiologist plus AI detected 261 cancers (0.5 percent) against 250 (0.4 percent) for two radiologists (relative proportion 1.04, 95 percent CI 1.00 to 1.09), meeting non-inferiority. AI alone detected 246 (relative proportion 0.98, 0.93 to 1.04) and two radiologists plus AI detected 269 (1.08, 1.04 to 1.11), both also non-inferior. The authors conclude that replacing one radiologist with AI gave a 4 percent higher, non-inferior detection rate and that the software has potential for controlled implementation with risk management and real-world follow-up.

The registry lists the study as interventional with no drug phase, which is why the phase 2 and 3 pass never attached it. MASAI, the Swedish randomised trial of AI-supported screening, used Transpara and is a separate record.