Objective: The Veterans Affairs Diabetes Trial (VADT) is a 20-medical center, prospective, randomized study of 1792 Type 2 diabetic individuals primarily aimed at determining whether intensive glycemic control prevents macrovascular events. We report a comparison of fundus photographs and ophthalmologic examination at baseline, permitting an evaluation of multiple settings similar to common clinical practice.
Research design and methods: A 340-patient subset had both local dilated fundus examinations and centrally read seven-field stereo fundus photographs completed within 60 days of each other (median 28 days). Local examiners were unaware of the stereo photographs.
Results: Overall, agreement within one step was 76% and exact agreement between ophthalmoscopy and central gradings of fundus photographs on a five-step retinopathy severity scale was 43% (weighted kappa 0.42, CI 0.35-0.48). In about 90% of disagreements the severity level was higher by photographic grading. The sensitivity for ophthalmoscopy compared to grading of fundus photographs for the detection of any retinopathy was 51% and specificity was 91%. For proliferative diabetic retinopathy (PDR), sensitivity was 61% and specificity 98%. Only one eye was high-risk PDR, and it was detected by both methods. For clinically significant macular edema, these measures were 24% and 98%, respectively. The disagreements were of possible clinical importance in three cases (<1%).
Conclusion: Most disagreements occurred in eyes rated near the milder end of a category and/or resulted from small differences between the ophthalmoscopic and photographic definitions used in classifying severity. There were reasonably few disagreements of possible clinical significance.