Matthieu Schoumacher ^1,2 Vincent Lambert ^3,4 Pierre Blaise ^3,4 Bénédicte Locht ^3,4 Michéle Thys ^3,4 Edouard Duchateau ^3,4 Etienne Cavalier ^1,2,5 Jean-Marie Rakic ^3,4 Agnès Noël ⁶ Pascal De Tullio ^1,2*

• ¹ Center for Interdisciplinary Research on Medicines, Faculty of Medicine, University of Liège, Liege, Liege, Belgium
• ² Clinical Metabolomics group (CliMe), Liège, Liège, Belgium
• ³ University Hospital of Liège, Liège, Liège, Belgium
• ⁴ Department of Ophtalmology, CHU Liège, Liège, Liège, Belgium
• ⁵ Department of Medical Chemistry, CHU LIège, Liège, Liège, Belgium
• ⁶ Laboratory of Biology of Tumor and Development, GIGA-Cancer, GIGA Institute, University of Liège, Liege, Liege, Belgium

This study explores the application of NMR-based metabolomics to investigate neovascular age-related macular degeneration (nAMD), with a focus on addressing challenges related to patient management, disease progression evaluation, and treatment response assessment. A twoyear follow-up was conducted with 29 nAMD patients undergoing treatment, totaling 231 time points. At each time point, blood samples, clinical data, and optical coherence tomography (OCT) images were collected to assess the evolution of choroidal neovascularization. This longitudinal approach enabled the correlation of changes in the metabolome with ocular alterations during disease progression. Despite a rigorous analytical workflow that allowed for the precise quantification of over 60 metabolites, and the use of statistical tools designed to handle complex datasets, consistent results were difficult to obtain. The final dataset failed to meet the criteria necessary for drawing conclusive conclusions across the entire cohort. Our study's "real-life clinical practice" design, which reflects the heterogeneity and individualized nature of nAMD, posed challenges in achieving global, consistent results. While a personalized, patient-by-patient analysis was conducted, extracting a general model proved difficult. Nevertheless, we demonstrated that individual correlation models can be generated for each patient. This study underscores the challenges of creating a universal metabolomic model to monitor and predict the progression of nAMD. These difficulties may stem from the experimental design, which, although aligned with clinical practice, led to incomplete and heterogeneous datasets with limited variability across visits in both metabolites and ophthalmological data. Our findings also highlight the complexities of translating discoveries from controlled studies into clinical settings, where factors like visit frequency, treatment variation, and disease progression cannot be controlled. Ultimately, the study emphasizes the critical role of experimental design, particularly in longitudinal studies. With a more personalized approach and improved study design, we believe it would be possible to develop a robust, clinically relevant tool for the management and follow-up of nAMD patients.