Objective: To compare diabetic retinopathy (DR) severity as evaluated by digital and film images in a long-term multicenter study, as the obsolescence of film forced the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study (DCCT/EDIC) to transition to digital after 25 years.
Methods: At 20 clinics from 2007 through 2009, 310 participants with type 1 diabetes with a broad range of DR were imaged, per the Early Treatment Diabetic Retinopathy Study (ETDRS) protocol, with both film and digital cameras. Severity of DR was assessed centrally from film and tonally standardized digital cameras. For retinopathy outcomes with greater than 10% prevalence, we had 85% or greater power to detect an agreement κ of 0.7 or lower from our target of 0.9.
Results: Comparing DR severity, digital vs film yielded a weighted κ of 0.74 for eye level and 0.73 for patient level ("substantial"). Overall, digital grading did not systematically underestimate or overestimate severity (McNemar bias test, P = .14). For major DR outcomes (≥3-step progression on the ETDRS scale and disease presence at ascending thresholds), digital vs film κ values ranged from 0.69 to 0.96 ("substantial" to "nearly perfect"). Agreement was 86% to 99%; sensitivity, 75% to 98%; and specificity, 72% to 99%. Major conclusions were similar with digital vs film gradings (odds reductions with intensive diabetes therapy for proliferative DR at EDIC years 14 to 16: 65.5% digital vs 64.3% film).
Conclusion: Digital and film evaluations of DR were comparable for ETDRS severity levels, DCCT/EDIC design outcomes, and major study conclusions, indicating that switching media should not adversely affect ongoing studies.