Comparison of nonmydriatic handheld retinal imaging with early treatment diabetic retinopathy study (ETDRS) 7-standard field photography for diabetic retinopathy (DR) and diabetic macular edema (DME)

Purpose : To compare nonmydriatic handheld retinal imaging with ETDRS standard 7-field 300 fundus photographs (ETDRS photos) for assessment of DR and DME severity.
Methods : Following a standard imaging protocol, nonmydriatic retinal images were taken using handheld retinal cameras [Aurora(AU), Smartscope(SS), RV700(RV)] and compared to dilated ETDRS photos. Images were evaluated at a centralized reading center by 4 independent graders (2 certified graders, 1 ophthalmologist, 1 retina specialist) using the International DR/DME classification. A senior retina specialist adjudicated all differences. Kappa statistics [simple (K),weighted (KW)] assessed agreement for DR/DME. Sensitivity and specificity for any DR, referable DR [(refDR) moderate nonproliferative DR (NPDR) or worse, any DME or ungradable images] and vision threatening DR [(vtDR) severe NPDR or worse, clinically significant DME (CSME) or ungradable images] were calculated.
Results : Images from 177 eyes of 92 patients with diabetes were evaluated. Severity by ETDRS photos, DR: no DR 40.1% eyes, mild NPDR 19.2%, moderate 14.7%, severe 10.2%, and proliferative DR 15.8%; DME: No DME 72.9%, DME 6.8%, ciDME 17.0%, 3.4% ungradable. Ungradable rate for DR/DME for AU: 13.0%/15.8%; SS: 15.3%/18.1% and RV: 35.6%/36.7%. Agreement of clinical DR grading between handheld retinal and ETDRS photos are shown in table 1. Among the devices, AU (exact 58.8%; w/in 1-step 83.1%) had the highest agreement with ETDRS photos. These are higher than SS (56.5%; 80.2%) and RV (47.5%; 62.7%). Ungradable images were associated with a higher rate of refDR on corresponding ETDRS photos (AU 4.8x, SS 2.9x, RV 2.1x, p<0.0001). Table 1 shows the sensitivity/specificity for any DR, refDR and vtDR. AU, SS and RV had 71-97% specificity and 82-96% sensitivity for any DR, refDR and vtDR.
Conclusions : Despite a standardized protocol of image capture and evaluation, the ungradable rate of these devices varies from 13%-36%. In this cohort, there was a 2.1 to 4.8-fold increased risk of refDR among ungradable images. Thus, although handheld nonmydriatic retinal devices are able to achieve substantial agreement with DR in some cases, additional methods may be needed to reduce ungradable rates and appropriately triage eyes that require specialized care.