Concordance of OCT and FA-based MNV classifications in neovascular AMD and correlation with fluid volumes

Abstract

Purpose : To examine the concordance of the classical fluorescein angiography-based macular MNV classification of classic, predominantly classic, minimally classic, occult and retinal angiomatous proliferation (RAP) compared with the optical coherence tomography (OCT)-based classification of type 1, type 2 and type 3 MNV. Furthermore, we examined the influence of OCT-based MNV types on the distribution of fluid accumulation in different retinal compartments.

Methods : In a post-hoc analysis baseline FA and OCT images of 625 eyes of 625 patients from two multicenter randomized clinical trials were independently graded at the Vienna Reading Center by trained masked graders. In OCT mixed types MNVs were categorically separated in to predominantly type 1, predominantly type 2 or type 3 based on predefined definitions. For this analysis, FA MNV types were also separated into predominantly classic, predominantly occult types or RAP. The quantification of retinal fluid compartments was done using a commercially available AI tool (RetInSight Fluid Monitor Version 2). Statistical analysis was performed using Kruskal-Wallis and Wilcoxon rank sum tests.

Results : The concordance of FA and OCT MNV types was 77% (type 1/occult), 47% (type 2/ classic) and 68% (type 3/RAP). The volume of fluid in the different compartments such as IRF, SRF and PED differed significantly between the three OCT-based MNV types (p<0.0001). In a pairwise group comparison fluid volumes were statistically different except for SRF volume between type 1 and type 2 MNV (p=0.29) and PED volume between type 1 and type 3 MNVs (p=0.79).

Conclusions : In this analysis FA and ICG-based MNV classifications had a fairly good concordance in type 1 and type 3 MNVs, but type 2 MNVs were only identified as classic MNVs in less than 50% of cases. On the other hand, OCT-based MNV types had a significant effect on the distribution of fluid. Thus, AI-based retinal fluid segmentation allows the differentiation between MNV types and an AI-assisted OCT-based MNV classification can be used in the future in clinical trials and real-world practice.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.