Imaging Features and Diagnostic Performance of US in Nonmass Lesions with Varying Clinical Indications

Abstract

Background: The diagnostic performance of breast US for nonmass lesions (NML) is not well validated. Purpose: To evaluate and compare the imaging features and diagnostic performance of US for NMLs across clinical indications. Materials and Methods: This retrospective study included NMLs that underwent US-guided biopsy from January 2014 to February 2021. Lesions were also classified by their clinical indications as screening, diagnostic: work-up (DWU), or diagnostic: current breast cancer (DCBC). Logistic regression was used to identify factors associated with malignancy. The positive predictive values (PPVs) of imaging features and areas under the receiver operating characteristic curve (AUCs) of the clinical Breast Imaging Reporting and Data System (BI-RADS)-based assessments were calculated and compared between clinical indication groups using the Fisher exact or chi(2) test and 2000 bootstrap samples, respectively. Results: A total of 1152 NMLs in 1152 women (mean age, 47.7 years +/- 11.1 [SD]) were analyzed. The malignancy rates of the screening, DWU, and DCBC subgroups were 10.4% (26 of 251), 43.4% (295 of 679), and 40.1% (89 of 222), respectively. Lesion size, hypoechogenicity, hyperechogenicity, segmental distribution, abnormal duct changes, calcifications, posterior shadowing, and the absence of multiple small cysts were the features on US images that were associated with malignancy (all P < .05), with the lowest PPVs in the screening subgroup (PPVs of screening, DWU, and DCBC subgroups: 0%-60%, 36.4%-70.7%, and 11%-100%, respectively). Clinical BI-RADS assessments showed a lower AUC in the DCBC subgroup (AUC, 0.72) compared with the screening (AUC, 0.89) and DWU (AUC, 0.88) subgroups (both P < .001). In the screening subgroup, NMLs without US features suspicious for cancer and mammographic findings had malignancy rates of 2.5%-2.7%. Within the screening, DWU, and DCBC subgroups, PPVs of US features varied by findings at mammography, symptoms, and lesion laterality, respectively (P < .05). Conclusion: The PPVs of the US features of NMLs differed across clinical indications and were influenced by additional factors (mammographic findings, symptoms, and lesion laterality). AUCs of clinical BI-RADS assessments were lowest in women undergoing preoperative US for newly diagnosed current breast cancer.