From case-control study to longitudinal patient’s follow-up using NMR-based metabolomics: back to the “real-life” of clinics. The case of neovascular Age-related Macular Degeneration (nAMD).

Title: From case-control study to longitudinal patient’s follow-up using NMR-based metabolomics: back to the “real-life” of clinics. The case of neovascular Age-related Macular Degeneration (nAMD).

Introduction: The success of clinical metabolomics in identifying metabolites as potential biomarkers of disease in controlled studies is well documented. However, the translation of these findings into real life and clinical settings is not without difficulties and raises many questions and challenges. This is particularly true when metabolomics is applied to personalized medicine. This approach represents the most important paradigm shift in medical care, requires the development of innovative tools and is usually associated with patient follow-up. Thanks to the dynamic changes that link metabolites to the evolution of patient health, metabolomics is particularly adapted to this approach. That’s why we decided to apply it to neovascular age-related macular degeneration (nAMD). This pathology is characterized by alternating unpredictable outbreaks of choroidal neovascularization (CNV) and periods of stabilization. This led to a very complicated and empirical management, that could obviously be improved by a more personalized approach that requires precise detection of the transition from the stabilized to the CNV form.
Objectives: In a previous case-control study, we identified, through NMR-based metabolomics, some biomarkers that are associated with the CVN phase compared with controls and stabilized patients. Based on these results, we designed an experimental study to follow nAMD patients for 2 years (32 patients, 10-20 expected time points), using NMR-based blood metabolomics with the aim of correlating the levels of certain metabolites with the evolution of patient status. Indeed, establishing a predictive model that could indicate how and when the patients will negatively evolve could help clinicians to adapt treatment.
Results: The shift from a case-control study to a patient follow-up study first highlighted the difficulty of applying our longitudinal experimental design in a clinical setting. Indeed, the reality of patient follow-up generated several major concerns. First, the impossibility of having patients examined with regular and identical timing for all. The spacing between two visits obviously depended on their clinical condition and the choice of the clinician. Secondly, the clinical data on which clinicians base their assessment of the patient's condition are very empirical and not very well adapted. This led to a very incomplete data matrix and the need to re-analyze the clinical data to extract relevant and useful values. Moreover, the number of clinical “events” in this matrix is very low which means that during the 2 years of follow-up the status of the patients are poorly changing. At the analytical level, the NMR measurements of the samples in different batches emphasized some problems in the final comparison of the data. Finally, this approach requires the use of specific and complex statistical methods. In our case, some models including all the data allow the identification of some metabolites correlated with the clinical data. However, the individual analysis of the data and the metabolomics trajectories demonstrate the great inter-individual variability and highlight the difficulty of establishing predictive global models.
Conclusion: This project emphasizes the difficulty of moving from a case-control study to a longitudinal clinical real case. In this sense, it is very informative on the next challenges for the application of metabolomics in clinical practices and personalized medicine.