An AI-Based OCT System to Detect Diabetic Macular Edema: A Prospective Validation and Noninferiority Randomized Clinical Trial
Primary result · False-positive DME referral rateResult — go to the resultAbsolute difference −45 % 95% CI −58.2% to −31.9%Authors' conclusion
Compared with standard practice, incorporation of the AI-OCT system as a secondary screening tool was noninferior with respect to false-positive referral rates and was associated with a substantial reduction in potentially unnecessary DME referrals without compromising sensitivity.
Study context
- Population
- Patients with diabetes in Hong Kong: silent-mode validation at a tertiary hospital triage unit (603 patients); RCT of patients with suspected diabetic macular edema (DME) referred from a territory-wide DR screening program (276 patients)RCT
- Participants
- 276 patients
- Comparison
- Referral based on fundus photograph-based screening plus AI-OCT reports vs fundus photograph-based screening reports alone
- Interventionn = 137
- Controln = 139reference
- Primary endpoint
- False-positive DME referral rate (noninferiority margin 20%)
Methodological detailsstepwise evaluation, prospective silent-mode validation, multicenter, noninferiority, noninferiority margin 20%, All participants ultimately underwent specialist evaluation
- Stepwise evaluation
- Prospective silent-mode validation
- Multicenter
- Noninferiority
- Noninferiority margin 20%
- All participants ultimately underwent specialist evaluation
Primary result
False-positive DME referral rate
Absolute difference
−45 %
95% CI −58.2% to −31.9%
P < 0.001 for noninferiority
By group
Interventionvs Control
absolute difference −45 % (95% CI −58.2% to −31.9%); P < 0.001 for noninferiority
As reported in the abstract
Secondary outcomes
DME prevalence (RCT)
| Outcome | Intervention | Control |
|---|---|---|
| Sensitivity for DME referral | 100.0% (95% CI 100.0%–100.0%) | 100.0% (95% CI 100.0%–100.0%) |
| Specificity for DME referral | 86.5% (95% CI 79.3%–92.9%) | 0.0% (95% CI 0.0%–0.0%) |
Silent-mode validation: AI-OCT DME detection
- Ungradable scans
- 86 of 1200 scans (7.2%)
- Gradable scans classified as uncertain
- 49 of 1114 gradable scans (4.4%)
- Sensitivity for DME detection
- 98.8% (95% CI, 94.5%-100.0%)
- Specificity for DME detection
- 90.7% (95% CI, 88.7%-92.4%)
- Prospective silent-mode validation
DME prevalence (RCT)
- DME prevalence was similar in the intervention and control groups
DME among nonreferred participants (intervention)
- DME cases among nonreferred participants in the intervention group
- 0
Safety
- DME among nonreferred participants — Intervention
- 0 events
Original publication
An AI-Based OCT System to Detect Diabetic Macular Edema: A Prospective Validation and Noninferiority Randomized Clinical Trial. JAMA. Published 21 July 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →