AI, anti-VEGF at ARVO 2019

A few of the highlights from poster and paper presentations at this year’s meeting.

By Michelle Dalton, ELS, contributing editor

While the Association for Research in Vision and Ophthalmology (ARVO) is well-known for presenting cutting-edge research and basic science across all areas of ophthalmology, it’s what happens in the real world clinics after a compound is approved that most affects patients.

Numerous poster and paper presentations at this year’s conference highlighted just how different results from clinical trials and real-world outcomes may be. (All abstracts may be found at https://www.arvo.org/annual-meeting/program/online-planner/ ). Among the highlights:

TELEMEDICINE AND AI

Telemedicine and artificial intelligence (AI) are showing increased potential to alleviate treatment burden for diseases expected to increase exponentially, and they may be able to offset a lull in physician availability.

• AI and DR. Using AI to screen for diabetic retinopathy (DR) with the EyeStar (Visionquest) software system in a clinical setting over the course of two years found high sensitivity (98%) and specificity (77%), with the latter meaning it screened out 77% of patients who did not need a referral to an ophthalmologist (Abstract 5462). Developers and authors E. Simon Barriga and colleagues screened 12,427 patients with diabetes at 13 comprehensive diabetes care clinics in Monterrey, Mexico. The software evaluated images uploaded by a nurse or medical technician and determined a “non-refer” for cases with no DR, mild or moderate non-proliferative DR (NPDR), or “refer to an ophthalmologist” for cases of severe NPDR, proliferative DR or suspect for clinically significant macular edema.
• Telemedicine and ROP. Miles F. Greenwald and colleagues (Abstract 1526) evaluated infants at risk of retinopathy of prematurity (ROP) from 2015-2018 in the Oregon Health & Science University program with Salem (Oregon) Hospital. All patients had been screened with digital fundus imaging (Retcam) taken by a trained pediatric nurse and graded with remote telemedical interpretation by both a trained pediatric ophthalmologist and by the Imaging and Informatics in ROP [i-ROP] deep learning system trained to recognize severe ROP. Infants were transferred to the Portland hospital in cases of treatment-requiring ROP: 81 infants underwent the screening program and deep learning analysis, 79 of whom did not need referral, and two infants were transferred to Portland for treatment of ROP. For the default cutoff i-ROP score of 3, sensitivity was 100% and specificity was 89.7%. In Oregon, hospitals outside the major cities often “do not have a pediatric ophthalmologist on staff, and this system would allow all infants to be screened, per the guidelines,” Greenwald said.

ANTI-VEGF IN THE REAL WORLD

Several clinical trials have suggested treat-and-extend quarterly dosing with anti-vascular endothelial growth factor (VEGF) for the treatment of AMD can reduce treatment burden while maintaining visual gains. Other studies suggest monthly dosing is necessary to maintain vision gains:

• Quarterly dosing and recurrence. A retrospective chart review of 6,809 patients with wet AMD treated at one retina-only practice from 2005-2017 identified 150 eyes (134 patients) intentionally treated at intervals of 11 to 15 weeks, receiving a mean of 9.8 injections over a mean of 29 months. According to Alexander Michael Rusadkevich et al. (Abstract 103), 22% of eyes experienced disease recurrence during quarterly dosing, with a mean 12% cumulative yearly recurrence rate; 38% of eyes that stopped quarterly dosing experienced subsequent exudative disease recurrences after a mean of 21 months. “Recurrence of exudative disease activity is not uncommon,” the group wrote, “even among patients who initially tolerate five or more quarterly injections.” Terminating treatment altogether “appears to heighten the risk of recurrence.”
• Clinical studies vs. real-world data. An analysis of the AAO IRIS Database (Abstract 3865) evaluated 34,972 treatment-naïve eyes starting anti-VEGF therapy for the treatment of diabetic macular edema (DME) between January 2014 and March 2016. While patients had 9.4 ophthalmology encounters and 7.7 retina specialist encounters in the first 12 months, the mean number of injections was substantially lower than in clinical studies, ranging from 3.3 for both ranibizumab or bevacizumab to 3.7 for aflibercept. Within the first year, 4,241 (12.1%) eyes switched agents, predominantly from bevacizumab to aflibercept (n=2,256) or to ranibizumab (n=597). Visual acuity outcomes were also correspondingly lower than reported in the pivotal trials, authors Vanessa Shih et al. said.
• Insufficient treatment. Despite clinical guidelines, most patients do not receive the frequency of injections recommended for the treatment of DME, according to Rajiv Pandey, et al (Abstract 3646). Patients referred from National Diabetic Retinal Screening Programme (RetinaScreen) in Ireland who received anti-VEGF injections from November 2015 to March 2018 were included in their analysis. There were 4,725 referrals for further evaluation and treatment. Of those, 108 patients (28 binocular, 80 monocular) were treated with anti-VEGF injections for center-involving DME but received only 3.5 injections in the first year, Dr. Pandey said. “Our visual and anatomic outcomes are subpar; reasons include personnel shortages and a lack of standard delivery facilities.” He added that bevacizumab is often the first-line treatment, and this initial analysis did not include mixed cohorts (ie, laser). Further, Dr. Pandey said that people with < 20/60 vision may have had different visual outcomes had they been treated with an alternate anti-VEGF as bevacizumab produced poorer results than either aflibercept or ranibizumab over the course of 2 years in Protocol T.
• A look at CRVO treatment. Fewer injections also adversely impacted real-world outcomes in patients with macular edema secondary to central retinal vein occlusion (CRVO) treated at the Royal Free London NHS Foundation Trust (Abstract 4061). Sarah Ah-Moye and colleagues analyzed data from 28 eyes (28 patients) completing 12 months of monotherapy for CRVO, with a mean letter gain from baseline of 9.5 letters, with a median of six injections. In comparison, the Galileo and Crystal clinical trials showed patients gained a mean of 16.9 and 12.3 letters, respectively. “The difference in outcome may partly be explained by reduced injection frequency and different baseline characteristics in real-world practice,” Dr. Ah-Moye said.