Comparative diagnostic accuracy of photographic methods for detecting diabetic retinopathy: A systematic review and meta-analysis

Diabetic retinopathy (DR) remains a leading cause of vision loss globally. Timely detection through photographic screening is key to prevention, yet considerable variability exists across imaging protocols. This systematic review and meta-analysis evaluates the diagnostic accuracy of various photographic strategies for DR detection to inform evidence-based screening practices. We systematically searched six databases and trial registries up to June 2023. Eligible studies included adults (≥18 years) with type 1 or 2 diabetes undergoing DR screening using fundus photography, compared against a reference standard (7-field ETDRS or dilated ophthalmoscopy). We performed bivariate meta-analyses for direct comparisons (e.g., mydriatic vs. non-mydriatic), and indirect armbased meta-analyses. Eighty studies (over 50,000 participants) were included. Direct comparisons including 8 studies (16 arms, 2967 participants), found that the sensitivity for detecting any DR was comparable between mydriatic and non-mydriatic imaging (90% vs. 89%), though specificity was lower without dilation (85% vs. 90%). Sensitivity improved with the number of fields (from 0.82 with 1-field to 0.98 with ≥4-fields), particularly in nonmydriatic settings. Results also showed an observed heterogeneity in accuracy of screening across settings showing the need for continuous auditing of screening performance. Photographic screening for DR was effective across multiple configurations. Two-field mydriatic protocols offer high diagnostic accuracy, while trained graders and portable devices can enhance scalability. Findings support the design of structured screening programs adapted to local resources and clinical priorities.