Effectiveness and Safety of Screening for Diabetic Retinopathy with Topcon CRW6 Non-Mydriatic Digital Images Centered on the Optic Disc and on the Macula as Compared with the Seven Standard Stereoscopic Photographic Fields

Abstract

Abstract: : Purpose: Prospective, masked, clinic-based cross-sectional study to compare the assessment of severity grading of DR of a limited portion of the fundus (two 45 degree non-mydriatic camera fields centered on the optic disc and on the macula) with the 7 stereoscopic standard photographic fields and with a dilated ophthalmologic examination of 98 patients. Methods: Unweighted k analysis for agreement of the grading of DR. These analysis were used to determine safe screening guidelines to identify the patients needing referral to an ophthalmologist Results: Substantial agreement (k=0.758 ) is demonstrated between the clinical DR level of disease assessed with two NMC camera fields centered on the disc and on the macula and the seven standard photographic fields and fair agreement ( k =0.581) is shown between the dilated ophthalmologic examination and the seven standard photographic fields. Threshold for referral to an ophthalmologic specialist chosen at very mild DR (ETDRS severity level 20), mild diabetic DR (ETDRS severity level 35) and moderate DR (ETDRS severity level 43) for NMC screening with this limited portion of the fundus respectively correctly identifies 100%, 49.9 and 48.6% of eyes with severe non proliferative diabetic retinopathy and or proliferative retinopathy . A low threshold (very mild DR) NMC screening leads to the referral to an ophthalmologist of a proportion of patients approximating the rate of incidence DR in that population, with a 6.0% rate of false positives and the need to refer 17 % of the diabetic patients to an ophthalmologist because of insufficient image quality in at least one eye for reliable grading. Conclusions: Such a screening protocol associated with the choice of a low DR severity threshold for ophthalmologic referral identifies safely most diabetic patients with vision-threatening retinopathy. Its good sensitivity and negative predictive value make it a useful tool for mass screening of DR, reduces significantly the total number of patients needing referral to an ophthalmologist and helps improve the implementation of the current DR screening recommendations.

Keywords: diabetic retinopathy • clinical (human) or epidemiologic studies: sys • imaging/image analysis: clinical