Predicting Tolerance to Extended Interval Dosing in Diabetic Macular Edema and Retinal Vein Occlusion via Subvisual Feature Assessment of Ultra-widefield Angiography: Preliminary Findings in the PERMEATE study

Abstract

Purpose : Limited durability of anti-VEGF response in DME and RVO remains a significant clinical challenge. The purpose of this study is to evaluate the association of higher order quantitative imaging metrics from baseline ultra-widefield angiography (UWFA) images with durability of therapeutic response.

Methods : The PERMEATE study is an IRB-approved prospective open-label consecutive case series for treatment-naïve eyes with DME or RVO treated with intravitreal aflibercept (IAI) to evaluate UWFA features and dynamics. Eyes were initially treated with q4 weeks and then switched to q8 dosing at month 6. Extended interval intolerance was defined as increased macular edema and visual acuity following switch to q8 dosing. For this analysis, leakage foci and vascular tree segmentation is performed with a semi-automated platform. Subsequently, a novel quantitative assessment of morphological and anatomical characteristics of leakage patterns as well as curvature-based vascular tortuosity is performed. These baseline biomarkers are then used with a machine learning classifier (Linear Discriminant Analysis) to predict tolerance to extended interval dosing.

Results : Twenty-nine eyes of 29 patients were included in this study. 12 eyes (41%) exhibited rebound following the switch to q8 dosing. 79 features were extracted from the baseline UWFA scans. Median variance of distance of leakage nodes to vessels (p < 0.001) and median tortuosity of the vessel network (p 0.07) are identified as top 2 distinguishing features. Eyes that exhibited rebound demonstrated increased vascular tortuosity and more variance in the co-localization of leakage patterns to vessels. The cross-validated area under the ROC curve, across 5 runs, was found to be 0.933 ± 0.01%. Traditional clinical parameters (e.g., visual acuity, central subfield thickness) were not associated with treatment durability.

Conclusions : In this study, machine-based quantitative assessment of UWFA images successfully discriminates between eyes that tolerated extended interval dosing and those that did not. This analysis provides an exciting new opportunity for potential imaging biomarker identification. Additional research will focus on further characterization of subvisual features and independent validation of these findings.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.