Background: Adverse effects resulting from Capecitabine-based chemotherapy pose a significant concern in clinical practice. While the pharmacokinetics (PK) approach has been commonly used to investigate the toxicity of Capecitabine, its predictive ability remains limited. This study aims to assess whether pre-chemotherapy endogenous plasma metabolome can offer improved predictive values for Capecitabine chemotherapy-related adverse events (CRAEs). Methods: Plasma samples were collected from colorectal cancer patients at different time points: 0 hours (before), and 1, 2.5, 4 hours after oral Capecitabine administration, to assess individual variations in exposure levels of Capecitabine and its metabolites. Additionally, the endogenous metabolome profile was analyzed using UHPLC-MS/MS and UHPLC/Q-TOF-MS. Results: Capecitabine and its metabolites can predict two CRAEs, with 5-FU, 5’-DFCR, and FUH2 exposures being associated with diarrhea and thrombocytopenia, respectively. In contrast, identified plasma engougenous biomarker metabolites can predict all seven observed CRAEs. These CRAE-related endogenous plasma metabolites are involved in various physiological functions, including cell proliferation, maintenance, and inflammation. Pre-chemotherapy endogenous plasma metabolites established superior predictive performance for CRAEs (AUROC values ranging from 0.834 to 0.984) compared to conventional drug exposure (AUROC values ranging from 0.737 to 0.773). Additionally, the endogenous plasma metabolome demonstrated a strong correlation with Capecitabine and its metabolite exposures. Conclusions: Our study demonstrates that pre-chemotherapy endogenous plasma metabolome serve as superior biomarkers for predicting Capecitabine-related adverse effects, surpassing the value of exposures to Capecitabine and its metabolites. This finding holds significant potential in guiding personalized medicine approaches.