Systematic review: Can non-mydriatic cameras accurately detect diabetic retinopathy?

AIMS: Diabetes puts many individuals at risk for developing diabetic retinopathy (DR). Non-mydriatic (NM) retinal photography without pharmacological pupil dilation can be used to detect DR, however, its value in a general clinical setting has not been established. The objective of this review is to evaluate the validity of NM retinal photos, as compared to the reference standard of seven standard stereoscopic 30 degree field photographs (7SF), for their use of detecting DR in community-dwelling adults. METHODS: English articles were identified through MEDLINE (1992-2016), EMBASE (1992-2016), Center for Disease Control and Prevention, Food and Drug Administration, Health Canada, and Google Scholar (1992-2016). Only studies that examined the validity of NM retinal photos, without pharmacological pupil dilation, as compared to 7SF, used in a community-dwelling adult population were included. Risk of bias was assessed using the QUADAS-2. RESULTS: Of the 368 articles identified in the search, 6 articles met eligibility requirements. Sensitivities ranged from 64 to 97.9% and specificities ranged from 65.6 to 98%. Studies that used multiple field images to detect DR in a younger population generally produced higher quality photos and more valid results. CONCLUSIONS: The strongest validity indicators were reported in studies capturing multiple field images.