Research on the biomarkers in BI-RADS risk stratification to assist in the diagnosis of breast cancer

We investigated the efficacy of a diagnostic model that combines circulating tumor DNA (ctDNA) in peripheral blood, serum tumor markers, and the Breast Imaging Reporting and Data System (BI-RADS) based on Breast Ultrasound (US) in differentiating benign from malignant BI-RADS category 4 breast nodules, to provide evidence to improve discrimination between benign and malignant BI-RADS category 4 breast nodules. In this prospective cohort study conducted at Beijing Tongren Hospital (March 2022–May 2024), 116 patients with BI-RADS 4 breast nodules were enrolled. Pathological diagnoses were available for 104 patients (69 malignant and 35 benign), whereas 12 patients remained under follow-up without pathological confirmation. Three samples (1 malignant, 1 benign, and 1 without pathological confirmation) failed sequencing quality control; therefore, the final model included 102 patients with both qualified sequencing and pathological data (68 malignant and 34 benign). Binary logistic regression combined SMC, CA15-3, and BI-RADS classification. Discrimination, calibration, decision curves, and pairwise DeLong tests were assessed. While BI-RADS 4a, 4b, and 4c subcategories exhibited malignancy rates of 42.9%, 51.5%, and 80.7% ( P = 0.001). The malignant group demonstrated significantly higher median SMC (3.00 vs. 1.00, P < 0.001) and mean CA15-3 levels (2.21 vs. 0.76 kU/L, P = 0.039), with SMC positively correlating with Ki-67 expression ( P < 0.001). The combined model achieved superior diagnostic accuracy (AUC = 0.798 (95% CI: 0.703–0.892), sensitivity of 86.8%, and specificity of 70.6%. Its AUC exceeded that of BI-RADS alone (0.692; ΔAUC = 0.105, 95% CI: 0.021–0.189; P = 0.014). The diagnostic model that integrates SMC, CA15-3 levels, and BI-RADS could significantly improve the differentiation of benign from malignant BI-RADS category 4 breast nodules.