Deep learning-assisted analysis of biomarker changes after increase of dosing from aflibercept 2 mg to 8 mg in therapy-resistant neovascular age-related macular degeneration.

Abstract

Purpose: Age-related macular degeneration (AMD) remains the leading cause of blindness in developed countries. There are many different intravitreal anti-vascular endothelial growth factor (VEGF) drugs available for the treatment of neovascular AMD (nAMD). Unfortunately, not all patients respond equally well to the drugs, and some show recurrences during treatment. Since 01/2024, aflibercept 8 mg represents an additional treatment option and contains a four times higher dosage than the already known aflibercept 2 mg.

Methods: To evaluate the real-world efficacy of aflibercept 8 mg in refractory nAMD patients, focusing on changes in key optical coherence tomography biomarkers over a follow-up period of the first four aflibercept 8 mg injections using a deep learning-based semantic segmentation algorithm. Inclusion criteria were: switch to aflibercept 8 mg after insufficient response to aflibercept 2 mg, marked by persistent retinal fluid or inability to extend treatment beyond 6 weeks; completion of at least 3 months (90 days) follow-up under treat-and-extend treatment regime; and no confounding conditions like intraocular infection, uveitis or other retinal diseases.

Results: 23 eyes of 21 patients with therapy-resistant nAMD were switched to aflibercept 8 mg. All patients had previously received aflibercept 2 mg, with an average of 30.7 previous anti-VEGF injections. Significant reductions in intraretinal fluid and fibrovascular pigment epithelial detachment at timepoint V3 were observed. The decrease in subretinal fluid and central retinal thickness at V3 was not significant. Treatment intervals extended significantly by 24%, from a baseline average of 34 days to 42 days. Best-corrected visual acuity remained stable throughout the study period.

Conclusions: Aflibercept 8 mg demonstrated significant efficacy and durability in reducing nAMD biomarkers and extending intervals in a real-world setting. The use of deep learning for biomarker quantification highlighted its potential for enhancing treatment monitoring and decision-making. Future studies with a larger patient cohort and prospective study setting should explore long-term outcomes and integration of artificial intelligence-driven analysis.