Untargeted metabolomics offers biomedical researchers a powerful means of assessing and comparing human phenotypes via the comparative measurement of the metabolome from different biological samples. Current non-targeted metabolomics methods based on UHPLC-HRMS are often time-consuming and may lack of sufficient robustness to enable successful large-scale cohort analysis (>1000 samples). In that context, and to analyze a cohort of 6000 human plasma samples (PEPR ProPsy project), we are developing an untargeted metabolomics method to reduce analytical times. Our routinely used workflow involves two injections per sample using two complementary methods UHPLC-HRMS platforms, both involving a Q-Orbitrap mass spectrometer: C18 combined with detection in the positive ionization mode over 30 minutes and ZIC-pHILIC with detection in the negative mode over 42 minutes. This proves impractical for a 6000-sample cohort, compromising the robustness and reproducibility of the analyses. The BEH amide column was selected based on an extensive literature search. Mobile phase composition, gradient conditions, and source parameters were thoroughly optimized. Currently, the method is undergoing validation, and robustness tests conducted on 300 plasma injections showing excellent coefficients of variation (0.7% for retention times and 12% for peak areas). By performing 2 injections on the same column, in both positive and negative ionization modes, this method achieves a detection coverage of 90% of our library of 1200-compounds in 38 minutes per sample. Moreover, we will use a dual HPLC system allowing parallel sample analysis, thereby theoretically reducing the time to ~24 minutes per sample, i.e. a third of the time currently required.