Reproducibility of Grading Superficial vs Deep Retinal Hemorrhage in the Study of COmparative Treatments for REtinal Vein Occlusion 2 (SCORE2).

Abstract

Purpose : The intraretinal location of hemorrhage in eyes with central retinal vein occlusion (CRVO) may provide important information regarding disease severity and visual acuity outcomes. The purpose of this post-hoc evaluation is to investigate the reproducibility of grading superficial (nerve fiber layer) vs deep retinal hemorrhage on color fundus photographs (CFP) in participants from the Study of COmparative Treatments for REtinal Vein Occlusion 2 (SCORE2).

Methods : Baseline three-field CFP of 54 study eyes in the SCORE2 cohort were evaluated. A grid was developed to differentiate 9 distinct subfields (Figure 1). Expert graders evaluated for percent involvement of superficial hemorrhage in each of the 9 subfields, and evaluated for percent involvement of deep hemorrhage in 8 subfields, excluding the optic disc. Graders also evaluated eyes for optic disc swelling, disc pallor, cotton wool spots, and retinal folds. Reproducibility was measured using predetermined parameters: exact agreement, agreement within 5%, 10% and 20%. Intergrader reproducibility was calculated using kappa statistics.

Results : All 54 study eyes had gradable images. Average intergrader reproducibility among fields with superficial hemorrhage was 36.6% exact agreement, 56.4% within 5%, 67.5% within 10%, and 85% within 20%. Reproducibility among fields with deep hemorrhage was 66.4% exact agreement, 88% within 5%, 93.1% within 10%, and 98.8% within 20%. Across all fields, the weighted kappa score range for grading percent superficial and deep hemorrhage was 0.43 – 0.71 and 0.49 – 0.69, respectively. Kappa scores for the assessments of disc swelling, disc edema, cotton wool spots, and retinal folds were 0.34, 1.0, 0.8, and 0.64, respectively.

Conclusions : In SCORE2, a novel method for grading area of superficial and deep retinal hemorrhage demonstrated agreement within 20% in ≥85% of the eyes. Intergrader reproducibility for both superficial and deep hemorrhage was moderate to substantial, depending on the field graded. Employing this method in a large scale longitudinal analysis will help define the role of retinal hemorrhage location as a predictive biomarker of therapeutic and visual outcome in CRVO.

This is a 2021 ARVO Annual Meeting abstract.

 

Figure 1. Grid placed on a right eye with CRVO: F1 optic disc; F2-F5 peripapillary region; F6, F8 peri-arcade regions; F7 macula; and F9 nasal retina.