Improving Efficacy Of Anti-Vegf Injections In Neovascular Age-Related Macular Degeneration: Lessons Learnt From Ai-Based Fluid Monitoring For A Comprehensive Patient Management In The Elderly Population

Age-related macular degeneration is a frequent diagnosis in a busy surgical clinic and a common comorbidity of patients undergoing cataract surgery. However, treatment outcomes for neovascular age-related macular degeneration (nAMD) lag behind those from clinical trials in the real world. Precise monitoring and measureable endpoints could lead to improved anti-VEGF injection management in the real world. The purpose of this investigation is to predict anti-VEGF treatment requirements, visual acuity, and morphologic outcomes in treatment-naive nAMD using artificial intelligence (AI)-based fluid quantification in a cohort from the Fight Retinal Blindness! (FRB!) Registry.

AI-quantified fluid analysis from the Fight Retinal Blindness! (FRB!) Registry in Zürich, Switzerland.

OCT data of patients with treatment-naive nAMD from FRB! were analyzed. Volumes of intraretinal fluid (IRF), subretinal fluid (SRF), and pigment epithelial detachment (PED) were quantified using an approved deep learning algorithm (Vienna Fluid Monitor, RetInSight, Austria). Fluid volumes, number of anti-VEGF treatments, and the impact of fluid volumes after initial therapy were examined over 4 years. Shift of macular fluid volumes was assessed over time in different retinal compartments and investigated during treatment breaks. Quantitative features from the OCT were used to create a predictive machine learning model for future treatment requirements and morphologic outcomes including long-term development of macular atrophy and fibrosis.

209 eyes (mean age: 78.3 years) were evaluated. Eyes with high volumes (top 25%) of IRF after initial therapy differed by -7.4 letters (p=0.007) compared to eyes with low IRF volumes while receiving significantly more injections (+4.1; p=0.006) after 4 years. Patients with high SRF or PED volumes had comparable visual outcomes but received significantly more injections (SRF +8.5; p<0.001, PED +6.0; p<0.001). Fluid was located predominantly in the subretinal compartment in treatment-naïve patients, whereas after treatment breaks fluid increased more for IRF than SRF. In a predictive machine learning model, first-year injection frequency was predicted with an accuracy of 0.77 AUC, atrophy with an AUC of 0.70, and fibrosis with an AUC of 0.74.

AMD is a frequent comorbidity affecting patients that are scheduled for cataract surgery. Adequate monitoring and objective criteria for repeated anti-VEGF injections are crucial to manage these patients. Eyes with high fluid volumes are at risk for long-term visual impairment, which cannot be compensated for in IRF by more frequent injections. Higher treatment frequency for high SRF and PED volumes adds no superior outcome. Localization and quantification of fluid compartments can predict long-term outcomes. Efficient management of the elderly population with cataract and AMD requires a comprehensive infrastructure.