Abstract
BACKGROUND: Most existing models for predicting liver metastasis primarily rely on single clinical indicators or traditional imaging features, which, while useful, offer limited accuracy and reliability. In recent years, spectral computed tomography (CT) has emerged as a dual-energy imaging technology that provides detailed quantitative analyses of the blood supply characteristics and metabolic activity of tumors. The integration of serum biomarkers such as carcinoembryonic antigen (CEA) and cancer antigen 19-9 (CA19-9) with spectral CT features holds great potential for significantly enhancing the accuracy of liver metastasis predictions. This study aims to develops a novel nomogram prediction model based on spectral CT, CEA, and CA19-9 to predict the risk of liver metastasis after colorectal cancer (CRC) surgery. METHODS: This study recruited 100 patients diagnosed with CRC and receiving initial treatment at Jiangsu Cancer Hospital between June 2020 and June 2022. All patients underwent preoperative spectral CT examination. Patients were categorized into two groups based on the occurrence of liver metastasis within two years post-surgery: the liver metastasis group (n=60) and the non-metastasis group (n=40). A comparison was made between the two groups regarding the clinical and pathological characteristics and the changes in spectral CT parameters of the primary lesion before surgery. The predictive efficacy of preoperative spectral CT parameters of the primary lesion for liver metastasis was assessed. The risk factors for liver metastasis following CRC surgery were determined using multivariable logistic regression analysis, and a nomogram prediction model was established. A 7:3 ratio was used to randomly divide the dataset into a training set (n=70) and a validation set (n=30). The model's performance was evaluated using the receiver operating characteristic (ROC) curves, calibration curves, and decision curves analysis (DCA). RESULTS: Following CRC surgery, liver metastasis was found to be independently associated with CEA, cancer antigen 19-9 (CA19-9), and preoperative spectral CT characteristics of the original lesion during the venous phase, including lesion iodine concentration (IClesion), spectral slope in Hounsfield units (λ(HU)), and the CT values of the tumor lesions on the 40-keV (CT(40 keV)). The nomogram developed from these predictors demonstrated high discriminative ability, with area under the curve (AUC) of 0.9078 [95% confidence interval (CI): 0.8419-0.9738] in the training cohort and 0.9502 (95% CI: 0.8792-1.0000) in the internal validation cohort at the optimal cutoff of 0.6460. The calibration curve showed that the observed and expected values agreed well. According to DCA, the nomogram model had good clinical value. CONCLUSIONS: The nomogram model constructed based on spectral CT parameters, CEA, and CA19-9 demonstrates potential in predicting postoperative liver metastasis in CRC, providing a reference for preoperative personalized treatment. However, its generalizability needs to be further confirmed through multi-center external validation.