Personalized Dosing System Guided by Cross-Scale Metabolic Imaging Biomarkers in Neovascular Age-Related Macular Degeneration.

PURPOSE: Long-term anti-vascular endothelial growth factor (VEGF) response in neovascular age-related macular degeneration (nvAMD) varies widely, with no validated predictive biomarkers. We aimed to establish a metabolic imaging biomarker-based framework for personalized efficacy prediction and dosing.

METHODS

In this single-center retrospective study, patients with clinically confirmed nvAMD and age-matched controls (cataract surgery with normal fundus or non-neovascular AMD) were enrolled. Aqueous humor was analyzed by ultra-high-performance liquid chromatography tandem mass spectrometry. Patients with nvAMD received anti-VEGF therapy (three doses then as needed, followed for ≥12 months) and were classified by structural/functional outcomes as responders or nonresponder. Comparative metabolomics identified phenotype-associated metabolites. An integrated prediction model was built using orthogonal partial least squares discriminant analysis, receiver operating characteristic analysis, and multivariate logistic regression.

RESULTS

The aqueous humor of patients with nvAMD showed lipid remodeling and suppressed purine metabolism. Eight key metabolites were identified as phenotype-associated biomarkers (indole-3-acetic acid, N-acetyl-L-alanine, xanthine, arachidonoyl lysophosphatidic acid, taurine, indole-3-lactic acid, 4-acetoxyphenol, and Lys-Ile). Among them, indole-3-acetic acid and N-acetyl-L-alanine showed the highest predictive performance for long-term treatment response (area under the curve = 0.802 each). Integrating metabolic markers with baseline imaging and clinical parameters improved predictive accuracy (area under the curve = 0.86). Longitudinal analysis showed partial normalization of metabolite levels in responders after anti-VEGF therapy.

CONCLUSIONS

nvAMD involves responder-specific metabolic reprogramming. Integrating aqueous metabolites with imaging biomarkers enables accurate pretreatment prediction of anti-VEGF efficacy, supporting personalized therapy.

TRANSLATIONAL RELEVANCE: This work links aqueous metabolic profiling to clinical decision-making by identifying biomarkers predictive of long-term anti-vascular endothelial growth factor response. Their integration with routine imaging yields a pretreatment stratification model (area under the curve = 0.86), which may inform individualized monitoring and earlier regimen adjustment in predicted nonresponders.