To evaluate the prognostic performance of two lymph node-based staging systems, tumor-ratio-metastasis (TRM) and tumor-log odds-metastasis (TSM), developed from the lymph node ratio (LNR) and log odds of positive lymph nodes (LODDS), and to compare them with the American Joint Committee on Cancer (AJCC) TNM staging system in colorectal cancer.

The training cohort consisted of postoperative colorectal cancer patients with pathologically confirmed lymph node metastasis registered in the Surveillance, Epidemiology, and End Results (SEER) database between 2010 and 2015 (n=13 469). The external validation cohort was a two-center cohort composed of patients treated at Chuzhou First People's Hospital and Shanghai General Hospital between 2018 and 2022 (n=402). X-tile software was used to determine the optimal cut-off values for LNR and LODDS (0.20/0.57 and −0.58/0.17, respectively), on the basis of which the TRM and TSM staging systems were constructed. Prognostic performance was assessed using multivariable Cox regression, Harrell's concordance index, time-dependent receiver operating characteristic curves and the corresponding area under the receiver operating characteristic curve at 36 and 60 months, calibration plots, and decision curve analysis. Model fit and comparative performance were further evaluated using the Akaike information criterion, Bayesian information criterion, and likelihood ratio test.

In the training cohort, both TRM and TSM were independent prognostic factors for overall survival (both P＜0.05). The 3-and 5-year area under the receiver operating characteristic curve values were 0.829 and 0.828 for TRM, 0.827 and 0.823 for TSM, and 0.823 and 0.822 for TNM, respectively. Harrell's C-index was also slightly higher for TRM (0.757) and TSM (0.755) than for TNM (0.751). The Akaike information criterion, Bayesian information criterion, and likelihood ratio test consistently favored TRM. In the external validation cohort, the 3-year area under the receiver operating characteristic curve values were 0.757 for TNM, 0.827 for TRM, and 0.829 for TSM, while the corresponding 5-year values were 0.650, 0.710, and 0.709. TRM showed the highest C-index overall and outperformed TNM and TSM in overall predictive performance. Calibration plots showed acceptable agreement between predicted and observed survival. Decision curve analysis indicated greater net clinical benefit for TRM and TSM than for TNM across most clinically relevant threshold probabilities, with TRM showing the most consistent overall performance.

The TRM and TSM systems, derived from LNR and LODDS, provided better prognostic discrimination and potential clinical utility than the conventional AJCC TNM staging system in colorectal cancer. External validation in a two-center cohort further supported their robustness, and TRM showed the most favorable overall performance. Further large-scale prospective multicenter studies are needed to confirm their generalizability and support clinical implementation.