Fig. 4

The performance evaluation of the CTCP model for predicting RFS in the discovery cohort. A The predictive model is visualized into a five-factor nomogram to predict RFS in patients with stage I-III CRC after curative surgery. B The calibration curves show the concordance between model prediction and actual distribution for 2 year RFS. Gray line: reference line. Red line: the prediction curve generated by this model. C The DCA curves show the net clinical benefit of the application of the CTCP model to predict 2 year RFS. D ROC analysis is used to compare the CTCP model with the ctDNA alone to predict 2 year RFS. E The recurrence rate for CRC patients stratified by the CTCP model. F Kaplan–Meier curves of RFS for CRC patients stratified by the CTCP model. CTCP ctDNA and clinicopathological risk factors, CRC colorectal cancer, RFS recurrence-free survival, ctDNA circulating tumor DNA, ROC receiver operating characteristic, DCA decision curve analysis