This study aimed to investigate the correlation between baseline MRI features and baseline carcinoembryonic antigen (CEA) expression status in rectal cancer patients. A training cohort of 168 rectal cancer patients from Center 1 and an external validation cohort of 75 rectal cancer patients from Center 2 were collected. A nomogram was constructed based on the training cohort and validated using the external validation cohort to predict high baseline CEA expression in rectal cancer patients. The nomogram's discriminative ability and clinical utility were tested using the receiver operating characteristic (ROC) curve, and decision curve analysis (DCA). The baseline CEA high-expression group had significantly higher MRI-detected metastatic lymph node (mLN), MRI-detected extramural vascular invasion (mEMVI), infiltrating tumor border configuration (iTBC), peritoneal invasion, annular infiltration, maximum extramural depth (MED), and tumor length than the normal CEA group (P < 0.05). Among them, MED [odds ratio (OR):1.19 (1.03-1.38), P = 0.016] and annular infiltration [OR:2.36 (1.06-5.25), P = 0.036] were independently predicting factors for high baseline CEA expression. The trained and validated model for predicting high baseline CEA expression in the training and external validation cohorts had the area under the curves (AUC) of 0.787 (95% CI 0.716-0.859) and 0.799 (95% CI 0.698-0.899), respectively. The calibration curves of both cohorts demonstrated good agreement between predicted and observed outcomes. Decision curve analysis indicated the clinical value of the nomogram. We developed a visual nomogram to predict high baseline CEA expression for patients with rectal cancer, enabling clinicians to conduct a personalized risk assessment and therapy.
Keywords: CEA; MRI; Maximum extramural depth; Rectal cancer.
© 2024. The Author(s).