Detection of choroidal neovascular membranes using OCT Angiography: Learnings from the PROCON Study

Abstract

Purpose : The PROCON study evaluated the utility of intravitreal aflibercept injections (IAI) as prophylaxis against conversion to neovascular age-related macular degeneration in high risk eyes. At each quarterly study visit, enrolled subjects were imaged using OCT angiography (OCTA), fluorescein angiography (FA), and spectral domain OCT (SD-OCT). Additionally, we evaluated OCTA images with and without 3D projection artifact removal software. This provided an opportunity to evaluate the effectiveness of OCTA on choroidal neovascular membrane (CNV) detection compared to the gold standard of FA and SD-OCT. Additionally, we were able to evaluate the rates of OCTA detection of non-exudative CNV (neCNV) that are not associated with any fluid or leakage.

Methods : Patients in the PROCON study were randomized in a 1:1 ratio to receive either IAI every 3 months for 24 months or sham injections. Masked readers at the Boston Image Reading Center (BIRC) reviewed all images obtained within the first 12 months of the PROCON study. All OCTA images were evaluated with and without 3D projection artifact removal. Evidence of CNV was independently noted for all imaging modalities.

Results : Of the 128 patients enrolled, 116 patients completed the month 12 visit resulting in 464 study visits. In 22 (4%) of these visits evidence of CNV was identified by either SD-OCT or FA. SD-OCT detected CNV in 21 of the 22 visits (95%). FA detected CNV in 17 of 21 (81%) visits with angiography. OCTA was able to identify CNV in 14 of 22 (63%) visits. Projection artifact removal had no effect in OCTA CNV detection for either 3x3mm or 6x6mm scans. OCTA detected neCNVs in 28 visits with 3x3mm scans, and in 34 visits with 6x6mm scans. Projection artifact removal had resulted in increased neCNV detection with 55 visits demonstrating evidence of neCNV in both the 3x3mm and 6x6mm scans.

Conclusions : In our study, OCTA with and without projection artifact removal did equally well in the detection of CNV in a rigorous reading center setting when exudation on OCT or leakage on FA was already present. However, the introduction of 3D artifact removal significantly enhanced the detection of neCNVs in our study. Future studies evaluating neCNV may benefit from use of 3D projection artifact removal software.

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