Two e-posters and a paper addressed some key topics in retina: the timing of best-corrected visual acuity (BCVA) gains in ranibizumab-treated patients, factors undermining the reliability of optical coherence tomography angiography (OCT-A), and a clinical trial of the EyeArt artificial intelligence (AI)-based screens for diabetic retinopathy.
BCVA in the HARBOR trial. A post hoc analysis of data from HARBOR found that patients who achieve peak vision two years after starting ranibizumab for neovascular age-related macular degeneration (AMD) ultimately gain more vision than those who peak early on.
The multicenter phase 3 HARBOR trial examined patients treated with monthly or as-needed ranibizumab for wet AMD. An e-poster by Rahul Khurana, MD, and colleagues describes peak and overall BCVA gains in 943 patients. They found that patients who achieved peak vision 18 to 24 months after starting ranibizumab gained up to 20 letters more than those who peaked within the first six months of treatment. The findings held true regardless of dosing regimen.
(You can view this poster via aao.org/mobile. Look for Timing of Peak Vision Gains Is Associated With Greater 24-Month Visual Outcomes: A Post Hoc Analysis of HARBOR, presented by Rahul Khurana, MD, Louis K. Chang, MD, PhD, Bann-Mo Day, PhD, and Ivo Stoilov, MD.)
Reliability of OCT-A readouts. According to an e-poster presented by Sri Meghana Konda, MD, and colleagues, weak signal strength may compromise the reliability of OCT-A. Two certified graders examined 219 Optovue and 187 Zeiss OCT-A scans of eyes with diabetic retinopathy. The authors determined the reliability of algorithm-generated vascular density after accounting for grid decentration, blink, shadow, and other artifacts.
They found that signal strength was considerably lower in unreliable versus reliable scans (p < .001). The diagnostic accuracy of signal strength in determining the reliability of vascular density was good, suggesting that these indices may offer a useful tool for assessing the quality of OCT-A scans and the reliability of quantitative output.
(You can view this poster via aao.org/mobile. Look for Role of Signal Strength in Determining OCT-A Reliability, presented Sri Meghana Konda, MD, Jeong Won Pak, PhD, Kyle McDaniel, Barbara Ann Blodi, MD, Kimberly E. Stepien, MD, and Amitha Domalpally, MBBS.)
EyeArt clinical trial. A multicenter clinical trial of EyeArt, an AI-based screen for diabetic retinopathy (DR), found that the technology performs on par with the 7-field ETDRS stereoscopic reference standard. A paper presented by SriniVas R. Sadda, MD, describes findings from 893 patients with diabetes, who were examined at 15 study centers across the United States.
The EyeArt system detected referable DR with high sensitivity and specificity, they found. It achieved a sensitivity of 95.5%, a specificity of 86%, and a gradability of 87.5% in undilated two-fundus imaging for automated eye-level detection of referable DR. With a dilate-if-ungradable protocol, the system achieved a sensitivity of 95.5%, a specificity of 86.5%, and an improved gradeability of 97.4%.
(Paper Pa073, AI Eye Screening for Diabetic Retinopathy: Results From a Pivotal Multicenter Clinical Trial was authored by Dr. Sadda along with Jennifer Irene Lim, MD, Barbara Ann Blodi, MD, and Carl D. Regillo, MD, FACS.) —Anni Delfaro, PhD
Read more news from AAO 2019 and the Subspecialty Day meetings.