Associations of mammographic and ultrasonographic features with immunohistochemistry-defined surrogate subtypes of breast cancer: a single-center retrospective study

Abstract Background This study aimed to evaluate the associations of mammographic and ultrasonographic features with histopathological characteristics and immunohistochemistry-defined surrogate subtypes of breast cancer, with particular attention to the triple-negative phenotype. Earlier reports have been inconsistent, possibly because imaging descriptors are rarely re-read under a single lexicon or tested for reproducibility, and multivariable models often include postoperative variables. Methods Patients with operable breast carcinoma treated at a single center between 2018 and 2023 were retrospectively studied, excluding those who received neoadjuvant therapy. The lesion was the unit of analysis. All examinations were re-read under the Breast Imaging Reporting and Data System (BI-RADS) 5th edition lexicon by a radiologist blinded to pathology, and 50 lesions were re-read by a second, blinded radiologist. Surrogate subtypes were assigned based on estrogen receptor, progesterone receptor, human epidermal growth factor receptor 2 (HER2), and Ki-67. Mammographic variables were analyzed only when mammography was available. Subtypes were compared using Benjamini–Hochberg-adjusted tests, and associations with the triple-negative phenotype were assessed by Firth penalized logistic regression restricted to pre-biopsy variables, with bootstrap internal validation. Results The cohort comprised 184 patients (185 lesions); 173 invasive carcinomas in women formed the analysis set, of which 52 (30.1%) underwent mammography. Margin (adjusted P = 0.006), echo pattern (adjusted P = 0.006), and BI-RADS category (adjusted P = 0.025) differed between subtypes. After adjustment, circumscribed margin (odds ratio 3.57, 95% confidence interval 1.17–10.67) and Doppler vascularity (3.07, 1.13–9.20) remained associated with the triple-negative phenotype, whereas isoechogenicity did not (1.44, 0.42–4.71; optimism-corrected area under the curve 0.71). Inter-reader agreement was almost perfect for circumscribed margin (κ = 0.90) and Doppler vascularity (κ = 0.96) but only slight for the echo pattern (κ = 0.19). Axillary ultrasonography was 97.3% specific and 75.8% sensitive for nodal metastasis. Mammographic margin only differed before adjustment ( P = 0.025). Conclusions In this exploratory single-center cohort, a circumscribed margin and Doppler vascularity were associated with the triple-negative phenotype, describing a benign-appearing sonographic presentation of this subtype. Isoechogenicity was neither independent of the margin nor reproducible between readers. These associations require confirmation in larger, prospective, independently read cohorts.