Overview
- A randomized phase 3 study of 368 adults with prediabetes compared a fully automated Sweetch app plus connected devices against four accredited, remote human-led Diabetes Prevention Programs over 12 months.
- At one year, 31.7% in the AI arm and 31.9% in human programs met the CDC composite benchmark, satisfying the trial’s prespecified noninferiority margin.
- Engagement favored the AI option, with higher initiation and completion rates than human-led programs (93.4% vs 82.7% initiation; 63.9% vs 50.3% completion).
- Investigators and an accompanying editorial said scalable AI delivery could broaden access if replicated, potentially informing policy on CDC recognition and payer coverage.
- Authors noted limitations and equity concerns, including virtual delivery of human programs during COVID-19, digital access and preference barriers, and disclosed compensation to Sweetch and participating DPPs.