BackgroundCarcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9) are the predominant serum tumour markers (STMs) for predicting the prognosis of colorectal cancer (CRC). The objective of this research is to develop clinical prediction models based on preoperative and postoperative CEA and CA19-9 levels.Methods1,452 consecutive participants with stage I-III colorectal cancer were included. Kaplan-Meier method, log-rank test, and multivariate COX regression were used to evaluate the significance of preoperative and postoperative STMs. Patients were grouped into a discovery cohort (70%) and a validation cohort (30%). Variables for the nomograms were selected according to the Akaike information criterion (AIC). Subsequently, two clinical predictive models were constructed, evaluated, validated, and then compared with the AJCC 8th TNM stage.ResultsThe overall survival (OS) rate and disease-free survival(DFS) rate declined progressively as the number of positive tumour markers(NPTMs) before and after surgery increased. For both OS and DFS, age, sex, pN stage, and NPTMs before and after surgery were independent prognostic factors, and then clinical prediction models were developed. The Concordance index (C-index), Receiver operating characteristic (ROC) curve, calibration curve, Decision curve analysis (DCA), and risk score stratification all indicated that the models possessed robust predictive efficacy and clinical applicability. The Net reclassification index (NRI) and Integrated discrimination improvement (IDI) indicated that the performance of models was significantly superior to the TNM stage.ConclusionNomograms based on pre-and postoperative CEA and CA19-9 can accurately predict survival and recurrence for stage I-III CRC patients after radical surgery, and were significantly better than the AJCC 8th TNM stage.