Abstract
BACKGROUND: Primary small bowel cancer (SBC) is an infrequent tumor recognized internationally, but lacks prognostic prediction models. This study aims to develop and validate prognostic nomograms for overall and specific mortality in SBC patients based on a cohort of SBC patients. METHODS: Patients with SBC between 2010 and 2015 were selected from the Surveillance, Epidemiology, and End Results (SEER) database. These patients were randomized into training and verification sets at a ratio of 7:3. Univariate and multivariate analyses were performed using Cox proportional hazards and competing risk models to screen independent predictors for overall and specific mortality in these patients. Based on these predictors, nomograms were constructed to predict the risks of overall and specific mortality in SBC patients. The accuracy and reliability of the nomograms were assessed utilizing the concordance index (C-index), area under the receiver operating characteristic curve (AUC), and calibration curve. RESULTS: This study included 6,863 patients with SBC, who were randomly split into a training set (70%, N=4,804) and a verification set (30%, N=2,059). In the training set, there were 1,630 all-cause deaths and 1,057 cancer-specific deaths, while in the verification set, 708 patients died from all causes and 431 died from cancer-specific causes. Univariate and multivariate analyses identified 12 independent predictors for overall mortality, including age, race, marital status, primary tumor location, pathological type, histological grade, tumor (T) stage, node (N) stage, metastasis (M) stage, and surgical procedure. There were also 12 independent predictors for specific mortality, encompassing age, marital status, primary tumor location, pathological type, histological grade, T stage, N stage, M stage, surgical procedure, and radiotherapy. Based on these factors, Cox proportional hazards and competing risk models were established to predict overall and specific mortality at 1, 3, and 5 years for SBC patients. The calibration curves suggested that the predicted values of the models aligned with the observed values, indicating good accuracy of the models. In the training and verification groups, the C-index values for overall survival (OS) rate were 0.789 and 0.785, respectively, and for cancer-specific mortality were 0.878 and 0.851, respectively. The AUC values for predicting OS rate at 1, 3, and 5 years were 0.827, 0.803, and 0.787 in the training set and 0.794, 0.800, and 0.807 in the verification set, respectively. These results consistently indicate that the model has good discriminatory power and predictive performance. CONCLUSIONS: This study constructed and validated risk predictive nomograms for overall and specific mortality risks in patients with primary SBC. These models are accurate and reliable and can assist clinicians in predicting survival rates in individuals with SBC.