The Second School of Clinical Medicine, Southern Medical University, No.253 Gongye Middle Avenue, Guangzhou, Guangdong, People's Republic of China.
Department of Ultrasound, Affiliated Hospital of Guangdong Medical University, No.57 People's Avenue South, Zhanjiang, Guangdong, People's Republic of China.
J Ultrasound Med. 2021 Oct;40(10):2189-2200. doi: 10.1002/jum.15612. Epub 2021 Jan 13.
Nodular sclerosing adenoses (NSAs) and malignant tumors (MTs) may coexist and are often classified into the same Breast Imaging Reporting and Data System (BI-RADS) category. We aimed to build and validate an ultrasound-based nomogram to distinguish MT from NSA for building a precise sequence of biopsies.
The training cohort included 156 patients (156 masses) with NSA or MT at one study institution. We used best subset regression to determine the predictors for building a nomogram from ultrasonic characteristics and patients' age. Model performance and clinical utility were evaluated using Brier score, concordance (C)-index, calibration curve, and decision curve analysis. The independent validation cohort consisted of 162 patients (162 masses) from a separate institution.
Through best subset regression, we selected 6 predictors to develop nomogram: age, calcification, echogenic rim, vascularity distribution, tumor size, and thickness of breast parenchyma. Brier score and C-index of the nomogram in the training cohort were 0.068 and 0.967 (95% confidence interval [CI]: 0.941-0.993), respectively. In addition, calibration curve demonstrated good agreement between prediction and pathological result. In the validation cohort, the nomogram still obtained a favorable C-index score of 0.951 (95% CI: 0.919-0.983) and fine calibration. Decision curve analysis showed that the model was clinically useful.
If multiple NSA and MT masses are present in the same patient and are classified into the same BI-RADS category, our nomogram can be used as a supplement to the BI-RADS category for accurate biopsy of the mass most likely to be MT.
结节性硬化性腺病（NSA）和恶性肿瘤（MT）可能共存，通常被归入同一乳腺影像报告和数据系统（BI-RADS）类别。我们旨在建立和验证一种基于超声的列线图，以区分 MT 与 NSA，从而为建立精确的活检序列提供依据。
训练队列包括来自一家研究机构的 156 例 NSA 或 MT 患者（156 个肿块）。我们使用最佳子集回归确定从超声特征和患者年龄中构建列线图的预测因子。使用 Brier 评分、一致性（C）指数、校准曲线和决策曲线分析评估模型性能和临床实用性。独立验证队列由另一家机构的 162 例患者（162 个肿块）组成。
通过最佳子集回归，我们选择了 6 个预测因子来开发列线图：年龄、钙化、回声边缘、血管分布、肿瘤大小和乳腺实质厚度。在训练队列中，列线图的 Brier 评分和 C 指数分别为 0.068 和 0.967（95%置信区间：0.941-0.993）。此外，校准曲线表明预测结果与病理结果具有良好的一致性。在验证队列中，该列线图仍获得了良好的 C 指数评分 0.951（95%置信区间：0.919-0.983）和精细校准。决策曲线分析表明该模型具有临床实用性。
如果同一患者存在多个 NSA 和 MT 肿块且被归入同一 BI-RADS 类别，我们的列线图可作为 BI-RADS 类别的补充，用于准确活检最有可能为 MT 的肿块。
