On Day 2 of the 52nd Annual Scientific Congress of The Royal Australian and New Zealand College of Ophthalmologists (RANZCO Brisbane 2022), yet another Rapid Fire session heated up the show. This time it was all about the retina — highlighting the latest in the treatment and management of age-related macular degeneration (AMD).
Promising Phase 3 Trial Results for nAMD Treatment
Results from the Archway phase 3 trial of the port delivery system (PDS) with ranibizumab for neovascular age-related macular degeneration (nAMD) showed that the PDS was non-inferior to monthly ranibizumab.
The PDS is an investigational drug delivery system designed for continuous intravitreal ranibizumab release through a surgically implanted, refillable ocular implant. The phase 3 Archway trial evaluated the safety and efficacy of PDS for the treatment of nAMD.
Patients were randomized 3:2 to treatment with the PDS with ranibizumab 100 mg/ml with fixed 24-week (Q24W) refill-exchanges or intravitreal ranibizumab 0.5 mg injections every four weeks.
Results showed that the PDS Q24W was non-inferior to monthly ranibizumab at W44/48. Over 90% of PDS Q24W patients did not receive supplemental ranibizumab treatment.
The implant insertion procedure has seven key steps: peritomy, implant preparation, scleral dissection, pars plana laser ablation, pars plana incision, implant insertion, and conjunctival and Tenon’s closure.
According to Assoc. Prof. Andrew Chang, medical director of Sydney Retina Clinic, PDS Q24W was non-inferior to monthly ranibizumab, with vision outcomes for the average of W44/48 consistent with the primary analysis.
The PDS was generally well tolerated, with a favorable benefit-risk profile. “Adherence to a specific methodology for the implant insertion procedure is key for maximizing optimal outcomes with PDS procedures,” he noted.
Standardized surgical training and ongoing procedural evaluations are important to enhance surgeon proficiency, he added.
Exploring the Long-term Effects of AMD Treatments
Meanwhile, Dr. Zhichao Wu from the Centre for Eye Research Australia gave his presentation on an observational extension study to the laser intervention in the early stages of age-related macular degeneration (LEAD) clinical trial.
The purpose of the study was to examine the long-term effect of subthreshold nanosecond laser (SNL) treatment with the Retinal Rejuvenation Therapy (2RT®) device on disease progression in the early stages of AMD or age-related macular degeneration.
The 36-month LEAD trial included 292 participants with bilateral large drusen, who were randomized to receive either SNL or sham treatment in one eye at six-monthly intervals up to 30-months. The difference in progression to late AMD between treatment arms for these participants was observed.
Overall, the rate of progression over a 60-month period was not significantly different between the SNL and sham group (p = 0.098); a similar finding to the 36-month LEAD study.
“In conclusion, this 24-month observational extension study confirmed findings from the LEAD trial that overall, SNL treatment did not significantly reduce the rate of progression to late AMD,” explained Dr. Wu.
Automating Diagnoses of AMD with the Help of AI
On the other hand, Dr. David Squirrell, from the University of Auckland, New Zealand, showed how the diagnosis of advanced age-related macular degeneration and high-risk intermediate age-related macular degeneration can be automated, based on the Age-Related Eye Disease Study (AREDS) scoring system.
The aim was to develop a deep learning algorithm to detect advanced AMD, and high-risk intermediate AMD based on images extracted from the AREDS dataset.
The AREDS image datasets comprised a total of 118,254 images. An artificial intelligence (AI) system was trained to remove very poor quality images, as well as those with dense cataracts.
From the remaining images, 8627 were labeled as advanced AMD. A total of 58,249 images had drusen labels (0-4) and 59,165 images had pigmentary labels. Color balancing, contrast enhancement and image normalization techniques were applied to homogenize the dataset. The data was then split 70/15/15 into training, validation and test sets.
Finally, an ensemble of AI algorithms was trained to detect advanced AMD and high-risk intermediate AMD models. A “waterfall” of AIs was then trained to automatically reproduce the scoring system in the AREDS Research Group’s report 18.
The results showed that compared to the ground truth provided by the AREDS dataset, the final patient-level accuracy of detecting advanced AMD is 92.75%, with a sensitivity of 94.86%, a specificity of 92.24%, and a negative predictive value of 98.66%.
Compared to the ground truth provided by the AREDS dataset, the final accuracy of detecting high-risk intermediate dry macular degeneration was 97.19%, with a sensitivity of 94.7%, a specificity of 97.31%, and negative predictive value of 99.74%.
“In conclusion, our model was able to accurately identify both advanced AMD and intermediate AMD in this large publicly available dataset. Further validatory trials are required to test its generalisability to other datasets,” concluded Dr. Squirrell.
Editor’s Note: A versión of this article was first published in Issue 2 of CAKE & PIE POST (52nd RANZCO Brisbane 2022 Edition). The 52nd Annual Scientific Congress of The Royal Australian and New Zealand College of Ophthalmologists (RANZCO Brisbane 2022) is being held virtually from February 26 to March 1. Reporting for this story took place during the event.
