First-line treatment efficacy and prognostic model in RAS-mutant metastatic colorectal cancer: a real-world study.

[OBJECTIVE] This study aimed to construct and validate a dedicated prognostic model for patients with RAS-mutant metastatic colorectal cancer (mCRC) using real-world data from two centers and explore the differences in the efficacy of first-line standard chemotherapy regimens in this population to provide an evidence-based foundation for individualized prognostic assessment and the selection of first-line treatment strategies.

[METHODS] Clinical, pathological, and follow-up data from 275 patients with RAS-mutant mCRC treated in two hospitals from January 2016 to December 2023 were retrospectively collected. Prognosis-related candidate variables were screened by univariate Cox regression and LASSO regression, and a prognostic model was constructed using a stepwise multivariate Cox proportional hazards model. For variables violating the proportional hazards assumption, a time-dependent Cox model was further used for correction. Model performance was evaluated by the concordance index (C-index), time-dependent receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). Leave-one-out cross-validation (LOOCV) was used to test its stability, and the tercile method was adopted for stratified validation to assess the stratification efficiency. In addition, 129 patients receiving first-line standard chemotherapy were selected from the total cohort to form an efficacy analysis cohort. The Kaplan-Meier method and the Cox proportional hazards regression model combined with multivariate adjustment were used to explore the differences in efficacy of different chemotherapy regimens with or without bevacizumab.

[RESULTS] The final prognostic model included four independent prognostic factors: lesion resection status, number of metastatic organs, CA19-9 level, and serum ALB concentration. The C-index of the model was 0.730 (95% CI: 0.689-0.771), and the pooled C-index verified by LOOCV was 0.705 (95% CI: 0.662-0.747). Time-dependent ROC analysis at 1, 2, and 3 years revealed that the AUC values of the model were 0.806, 0.781, and 0.772, respectively. The calibration curve had a good fit, and DCA confirmed that the model had clinical net benefit over a wide threshold range. There were significant differences in survival among patients in the high-, medium-, and low-risk groups (P < 0.001), indicating good stratification efficiency of the model. Exploratory efficacy analysis revealed that compared with single two-drug chemotherapy, two-drug chemotherapy combined with bevacizumab significantly prolonged the median progression-free survival (mPFS) of patients (9.61 vs. 6.74 months, P = 0.025), and this benefit remained stable after multivariate adjustment (HR = 0.437; 95% CI: 0.241-0.791; P = 0.006), but there was no significant difference in overall survival (OS) between the two groups. No statistically significant differences were observed in the objective response rate (ORR), PFS, or OS between patients receiving oxaliplatin-based or irinotecan-based two-drug chemotherapy combined with bevacizumab.

[CONCLUSION] The prognostic model and the corresponding nomogram constructed in this study can be used to conveniently evaluate the 1- to 3-year survival probability of patients with RAS-mutant mCRC. Exploratory analysis using real-world data revealed that two-drug chemotherapy combined with bevacizumab can significantly prolong the PFS of this population, providing a reference for the selection of first-line treatment regimens for this group. Both oxaliplatin- and irinotecan-based two-drug regimens combined with bevacizumab can be used as options for first-line treatment, and clinical selection can be made comprehensively on the basis of the individual conditions of patients. This study has several limitations, such as its retrospective design, unbalanced sample size in efficacy subgroups, and lack of external validation of the model, and the conclusions need to be further verified by large-sample prospective multicenter studies.