Impact of AI-quantified fluid dynamics on visual outcomes over 5 years in patients with treatment-naïve nAMD from the FRB! registry

Abstract To investigate the impact of retinal fluid dynamics on visual outcomes in patients with treatment-naïve neovascular age-related macular degeneration (nAMD) treated in the real world over 5 years using approved AI-based fluid monitoring. Real-world data comprising OCT scans and electronic medical records from 148 patients (187 eyes) were extracted from the Fight Retinal Blindness! (FRB! ) Zürich database. OCT scans were analysed using an approved AI algorithm (RetInSight, Vienna, Austria) to quantify fluid volumes by compartements. The impact of fluid persistence and fluctuations on BCVA change was assessed using forward stepwise regression and mixed models. Fluid compartments were further categorized into quartiles (SD-Qs), and the effect of fluid fluctuations on BCVA analysed (SD-Q1 least and SD-Q4 greatest variability of fluctuations). The greatest PED fluctuations in the central 1-mm showed an accentuated BCVA decrease after 2 and 4 years (estimate: -0.07, P = 0.019; estimate: -0.15, P < 0.01). After 4 years, eyes in SD-Q4 compared with SD-Q1 with greater PED fluctuations in the central 1-mm and 6-mm area were affected by a significant mean reduction in BCVA (-5.7 letters (P = 0.013); -6.1 letters (P = 0.015)). Greater intraretinal fluid (IRF) fluctuations (central 1-mm) (SD-Q4 compared with SD-Q1) were associated with a significantly worse mean BCVA by -6.8 letters (P = 0.018) after 5 years. Fluid persistence was not associated with statistically significant BCVA changes. In routine clinical management of nAMD, greater fluctuations of PED and IRF correlate with worse BCVA outcomes over long-term follow-up. A well-suited treatment regimen is required in the real world which can be utilized with AI-based fluid monitoring.