AI-based fluid volume quantification predicts visual acuity in neovascular AMD better than CST
This randomized phase III clinical trial evaluated AI-based quantification of macular fluid volumes (intraretinal fluid [IRF], subretinal fluid [SRF], pigment epithelial detachment [PED]) versus central subfield thickness (CST) in 290 eyes of 290 participants with active neovascular age-related macular degeneration (nAMD), including both treatment-naïve and previously treated patients in a real-world setting. The primary outcome was the association between best-corrected visual acuity (BCVA) and quantitative macular fluid volumes.
Results showed that IRF volumes within each macular region were significantly greater in treatment-naïve patients. In pretreated patients, larger PED volumes contributed to higher CST values. The 6-mm fluid model (IRF and SRF) explained the largest proportion of BCVA variance, with adjusted R values of 0.140 for IRF and 0.225 for SRF. In contrast, CST explained only half as much BCVA variance in pretreated eyes. However, in the treatment-naïve subgroup, the fluid model fit was poorer compared to the CST model (adjusted R = 0.078 vs. 0.198).
Safety and tolerability data were not reported, and limitations were not specified in the available information. The study did not report follow-up duration, and effect sizes with confidence intervals were not provided for most outcomes, limiting the precision of the findings.
These results suggest that AI-based quantification of IRF and SRF volumes may be a more relevant surrogate for visual function loss or benefit in nAMD than CST, particularly in pretreated patients. However, the weaker performance in treatment-naïve eyes and the lack of reported safety data indicate that further research is needed before clinical adoption.