A predictive assay for chemotherapeutic efficacy of stage IV colorectal cancer using serum metabolic features profiling

Background: Currently, there is no unbiased colorectal cancer (CRC) prognostic and predictive model based on serum molecular biomarkers to evaluate potential treatment outcomes and risk of CRC relapse for stage IV CRC patients. In addition, criteria to identify likely CRC patient populations at high risk and might benefit from additional chemotherapeutics have not yet been investigated, and it is an unmet clinical need. This study aims to develop a potential predictive risk discrimination model using serum metabolomic features generated from high-resolution mass spectrometry. Methods Using global serum metabolic pathway analysis and machine learning approaches, we have constructed a risk discrimination model to predict stage IV CRC patients' treatment efficacy and survival outcomes. This risk assessment model is further tested and validated in CRC patient cohorts via progressive free survival and overall survival with variable subset classifications such as the first-line treatment types, age, location of the primary tumor, and metastasis status. Results This study established an effective predictive model that can accurately discriminate stage IV CRC patients' progression-free survival (PFS) length regardless of the treatment types, age, and primary and metastatic tumor locations. Conclusions We have demonstrated a serum metabolomic pathway-based discriminating model to predict treatment outcomes of stage IV CRC patients under standard chemotherapeutics.