Development and validation of nomograms for predicting the prognosis of colorectal cancer patients.

BACKGROUND

Accurate prognosis prediction is essential in colorectal cancer (CRC) for guiding treatment decisions, yet the traditional tumor-node-metastasis (TNM) staging system often lacks precision. This study aimed to develop improved prognostic tools for CRC patients.

METHODS

Prognostic nomogram models were developed using data from 2,435 CRC patients who underwent curative resection. Parameters were selected via least absolute shrinkage and selection operator (LASSO) regression to include overall survival (OS) and disease-free survival (DFS) nomograms. The performance of these nomograms was evaluated against the TNM staging system using ROC analysis, calibration curves, and decision curve analysis (DCA).

RESULTS

Critical prognostic factors identified included tumor invasion depth, distant metastasis, tumordifferentiation grade, extranodal tumor deposits (ENTD), R1 resection, and log odds of positive lymph nodes (LODDS). The OS nomogram demonstrated area under the curve (AUC) values of 0.786, 0.776, and 0.803 for predicting 1-, 3-, and 5-year survival, respectively, compared to 0.768, 0.750, and 0.782 for TNM staging. The DFS nomogram predicted 1-, 3-, and 5-year DFS with AUCs of 0.764, 0.777, and 0.789, respectively, compared to 0.762, 0.761, and 0.770 for TNM staging. Calibration plots indicated strong predictive capabilities, and DCA confirmed greater net benefits over TNM staging.

CONCLUSIONS

Our developed prognostic nomogram models offer enhanced accuracy over traditional TNM staging in predicting CRC prognosis. Integrating these models into clinical practice can potentially improve personalized treatment strategies for postoperative CRC patients, enhancing overall clinical outcomes.