Author: Tiago Morais Sarmento (Portugal)
Co-authors: Gabriella de Salvo, Sara Vaz-Pereira
Purpose
Fluorescein angiography (FA) has been regarded as the gold standard for proliferative diabetic retinopathy (PDR) assessment, despite the development of other noninvasive imaging modalities, such as MultiColor imaging (MC). This study intended to appraise the accuracy of MC compared to FA in detecting PDR and its features.
Setting/Venue
Department of Ophthalmology, Hospital de Santa Maria, Lisboa, Portugal.
Methods
Fifty-nine eyes from 38 PDR patients were selected. MC images were evaluated by 2 independent masked graders. A qualitative analysis based on the following features was performed: neovascular complexes (NVC), disc neovascularization (NVD), neovascularization elsewhere (NVE), microaneurysm (MA), intraretinal hemorrhage (IRH), vitreous hemorrhage (VH), preretinal hemorrhage (PRH), fibrosis, hard exudates (HE), epiretinal membrane (ERM), diabetic macular edema (DME), ischemia and laser spots (LS). Measures of diagnostic accuracy compared to FA were determined.
Results
The sensitivity for the detection of NVC using MC was 95.1%, with a specificity of 40.0%, a positive predictive value (PPV) of 92.9% and a negative predictive value (NPV) of 50.0%. NVD detection presented higher sensitivity and specificity values (88.9% and 76.9%) while NVE registered higher PPV (88.9%). MC was highly sensitive in detecting IRH, HE, ERM and LS (100%), MA (98.0%) and fibrosis (95.5%). However, the highest specificity value was found in detection of VH (100.0%), DME (100.0%), PRH (98.1%) and LS (89.5%). The area under the receiver-operating characteristic analysis of MC was excellent in NVD (0.83, 95% confidence interval (CI), 0.71-0.95, p<0.001), IRH (0.89, 95% CI 0.74-1.00, p<0.001), VH (0.81, 95% CI 0.60-1.00, p=0.005) and PRH (0.89, 95% CI 0.68-1.00, p=0.004) and outstanding in LS detection (0.95, 95% CI 0.87-1.00, p<0.001). The contrast and quality of the MC probably account for these results, as the green wavelength enables better discrimination. The level of interrater reliability for the MC features appraised was almost perfect with an 0.86 weighted kappa (standard error: 0.02, 95% CI: 0.82–0.88).
Conlusions
MC was more accurate than FA in detecting some PDR features, namely NVD, VH, PRH, IRH and laser spots. These findings validate MC as a useful imaging test not only for the diagnosis but also for the follow-up assessment of PDR, enhancing the noninvasive imaging of this disease. The obtention of more information with less testing and without moving patients to different machines and/or rooms has grown even more pressing with the SARS-CoV-2 pandemic.
Financial Disclosure
Tiago Morais Sarmento has no financial interests to disclose; Gabriella de Salvo reports consultant fees from Allergan, Bayer, Heidelberg Engineering and Novartis; Sara Vaz-Pereira reports consultant fees from Bayer, Novartis, Alimera Sciences and Roche.
Comments
-
