Pre-Treatment CT Radiomics and Clinical Factors Predict Malignant Esophageal Fistula in Patients with Esophageal Cancer

Abstract

Abstract Background Malignant Esophageal fistula (MEF), which occurs in 5–15% of esophageal cancer (EC) patients, has a poor prognosis, and patients eventually die of nutritional failure, chest infection, mediastinal abscess, or great vessel injury in a short period of time. As a result, stratification of the high-risk group and intervention to prevent the occurrence is critical. The goal of this study was to build and validate a model to predict the occurrence of esophageal fistula in EC patients. Methods Fifty-three patients with MEF and 69 controls were randomly assigned to a training cohort (n = 86) and a validation cohort (n = 36). To identify clinically independent predictors, logistic univariate and multivariate regression analyses were used. Radiomic features were extracted from pre-treatment CT, which were then screened using least absolute shrinkage and selection operator (Lasso) regression. A clinical nomogram based on clinical risk factors, a predictive model based on radiomic features, and a nomogram incorporating the radiomic signature and clinical independent predictors were developed. The models were validated and compared by discrimination, calibration, reclassification, and clinical benefit. Results Univariate analysis identified clinical risk factors including stenosis, gender, and T stage. In multivariate analysis, stenosis was found to be an independent predictor (P = 0.023). A clinical nomogram was developed that included stenosis, gender, and T stage. A radiomic signature was created by ten features selected from 851 radiomic features extracted from pre-treatment CT images using Lasso regression. In discrimination, caliberation curve, and decision curve analysis, the joint nomogram incorporating clinical factors and radiomic signature outperformed the clinical nomogram and radiomics predictive model. When compared to the clinical nomogram, the radiomics-clinical prediction nomogram improved NRI by 0.236 (95%CI: 0.153,0.614) and IDI by 0.125 (95%CI: 0.040,0.210), P = 0.004. Conclusion We developed and validated the first radiomics-clinical nomogram for malignant esophageal fistula, which could assist clinicians in identifying patients at high risk of MEF.