To develop and validate a breast cancer risk prediction model using clinical and ultrasound imaging data of the patients with the Breast Imaging-Reporting and Data System (BI-RADS) 4 nodules.

A retrospective analysis was conducted on the clinical data of 377 breast nodules of BI-RADS 4 from 338 patients who were treated in the Second Affiliated Hospital of Chongqing Medical University between January 2017 and December 2018. The nodules were randomly divided into the training group and the validation group at the ratio of 7∶3. Univariate and multivariate logistic step-wise regression analyses were used to identify a combination of variables that were independent predictive factors for breast cancer, and then nomogram prediction was constructed. The performance of the nomogram model was evaluated using the receiver operating characteristic (ROC) and calibration curves. The Hosmer-Lemeshow test was used to assess the goodness-of-fit of the nomogram model, and the clinical decision curve analysis (DCA) was used to evaluate the clinical predictive efficacy of the model.

The study included 377 BI-RADS 4 nodules (202 benign and 175 malignant) from 338 patients. All nodules were divided into two groups: training group (263 nodules) and validation group (114 groups). Age (OR=1.06, 95%CI: 1.03-1.08, P< 0.001), margins (OR=2.22, 95%CI: 1.19-4.13, P=0.012), shape (OR=1.96, 95%CI: 1.01-3.77, P=0.045), calcification (OR=2.43, 95%CI: 1.35-4.36, P=0.003), maximum diameter of the nodule (OR=1.93, 95%CI: 1.38-2.69, P<0.001) and internal blood flow (OR=1.95, 95%CI: 1.08-3.51, P=0.026) were independent predictive factors for breast cancer. The area under the ROC curve for the nomogram was 0.807 (95%CI: 0.755-0.858) in the training group and 0.837 (95%CI: 0.764-0.910) in the validation group. The nomogram prediction model showed a good fit (training group: P=0.656; validation group: P=0.502), and the calibration curve indicated a good consistency between the nomogram and the actual observation. The DCA showed higher net benefit for the model when the threshold probability was greater than 0.1.

The nomogram model based on clinical and ultrasound features can accurately predict the risk of breast cancer in the patients with BI-RADS 4 nodules, thereby reducing unnecessary surgical biopsies.