Predictive model for pancreatic steatosis after neoadjuvant chemotherapy in breast cancer: Construction and ultrasound validation

Abstract

ObjectiveTo develop and verify a nomogram model for predicting occurrence of pancreatic steatosis (PS) after chemotherapy in breast cancer patients. MethodsA total of 215 breast cancer (BC) patients who underwent neoadjuvant chemotherapy in the hospital from February 2018 to November 2018 were prospectively collected for imaging and clinical data. Subjects were randomly divided into a training set (172 cases) and a validation set (43 cases) according to a random number table in a 4:1 ratio, and independent risk factors associated with PS were screened and modeled in a nomogram model after using single-factor and multifactorial logistic regression analyses on the training set. The model's performance was evaluated through various analyses, including subject operating characteristic curve(ROC), calibration curves, decision curve (DCA), and clinical impact curve (CIC), assessing its predictive performance, precision, and practical value in a clinical setting. ResultsThe results of univariate and multifactorial binary logistic regression analyses showed that age, triglycerides, and whether or not neoadjuvant chemotherapy was an independent risk factor for the development of PS, and these three independent risk factors were used to further construct a nomogram model. The model demonstrated strong predictive efficacy with an area under the ROC curve of 0.817 (95% CI = 0.753–0.880) in the training set and 0.762 (95% CI = 0.615–0.901) in the validation set. Calibration curve analysis confirmed accurate prediction of PS, and both DCA and CIC confirmed the model's clinical utility and validity. ConclusionThe nomogram model developed in this study shows strong prediction effect, discriminatory power, and clinical applicability, serving as an intuitive and visual tool for assessing the risk of PS in breast cancer chemotherapy patients.