Factors predicting the risk of breast cancer: construction and validation of a nomogram model

A simple and practicable strategy for predicting the risk of breast cancer (BC) urgently needs to be established. This study aimed to construct a nomogram model based on age and ultrasound (US) features to predict BC.

Consecutive adult females with a breast mass who underwent breast US followed by biopsy or surgery (Breast Imaging Reporting and Data System (BI-RADS) categories III-V) or who received follow-up US more than 6 months after the initial US (BI-RADS category II) from August 2020 to November 2023 were enrolled in this prospective multicenter study. The participants were allocated to three groups (the training set, internal validation set, and external validation set). A logistic regression analysis of the training set was performed to identify the independent variables associated with BC, based on which the breast nomogram model (B-NM) was constructed. The performance of the B-NM was evaluated using the area under the curve (AUC) of the receiver operating characteristic curve and calibration diagrams.

The training set comprised 306 females (43.0±11.9 years), the internal validation set comprised 45 females (46.0±12.6 years), and the external validation set comprised 114 females (40.6±11.4 years). Age and four US variables (mass size, orientation, margin, and vascularity) were found to be independently associated with BC. The B-NM combining these variables demonstrated relatively good performance {AUC [95% confidence interval (CI)]: 0.914 (0.882-0.946) vs. 0.905 (0.827-0.982) vs. 0.846 (0.750-0.940)} in the training, internal, and external validation sets, respectively. The calibration diagram showed that the nomogram's predicted probabilities were highly consistent with the observed values.

The B-NM incorporating age and US variables demonstrated favorable discrimination and calibration in predicting the risk of BC.