Ultrasound features of non-circumscribed margin associates with favorable prognosis in breast cancer patients in China: a retrospective cohort study

2.1 Study population

The data of 601 BC patients who underwent lumpectomy or mastectomy between January 2007 and June 2015 were retrieved and assessed. Among them, 386 patients were from Tianjin Medical University Cancer Institute and Hospital, and 215 were from the First Affiliated Hospital of Tsinghua University. This study was conducted in accordance with the ethical standards outlined by the Institutional Ethics Committee and the Helsinki Declaration of 1975 (revised in 1983), and ethical approval was obtained from the Research Ethics Committee of Tianjin Medical University Cancer Institute and Hospital and the Institutional Review Board of Tsinghua University. The patients were included based on the following criteria: (1) availability of complete clinical, pathological, ultrasound imaging and follow-up data; (2) no prior treatment, including radiotherapy or adjuvant chemotherapy, before surgery; (3) absence of distant metastasis at the time of surgery; (4) receipt of surgical tumor resection; (5) adherence to standardized post-surgical treatment protocols; and (6) absence of concurrent malignant diseases. Tumor stage and clinicopathological diagnosis were determined according to the 7th edition of the Tumor Node Metastasis (TNM) classification system of the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) [18].

2.2 Ultrasound analysis

Ultrasound imaging was performed using the LOGIQ 7 or LOGIQ 9 ultrasound system (GE Healthcare) equipped with a linear transducer operating at a frequency of 9–12 MHz. All real-time ultrasound scans were conducted by one of two experienced breast sonographers using standardized protocols. The acquired images were stored in the Picture Archiving and Communication System (PACS) for subsequent review. Ultrasound features, including tumor margin, posterior acoustic shadowing and microcalcifications, were retrospectively analyzed by two trained breast imagers, Lihong Liu and Hongjuan Han. Both sonographers had received fellowship training in breast imaging, with one having 25 years of experience and the other possessing extensive expertise in the field. To minimize bias, they were blinded to patients’ clinical histories and pathological diagnoses. In cases of discordance, consensus was reached through mutual discussion. Tumor margins were categorized as circumscribed or non-circumscribed, with the latter including angular, spiculated, microlobulated or indistinct margins. Posterior acoustic features were classified as either with or without shadowing. Microcalcifications were defined as positive (<0.5 mm) or negative (≥0.5 mm) based on their size within the mass (Fig. 1).

Fig. 1.Representative ultrasound images illustrating different tumor margin characteristics and acoustic features in breast cancer (BC) patients. The arrows indicate (A) indistinct margin, (B) microlobulated margin, (C) angular margin, (D) spiculated margin, (E) posterior shadowing and (F) microcalcifications.

2.3 Pathologic and biological analyses

ER and PR status were considered positive if nuclear staining was observed in ≥1% of tumor cell nuclei and negative if staining was present in <1% of nuclei. Immunohistochemical (IHC) staining (Hercep Test, Dako) was performed to assess HER2 expression. Staining intensity was classified as follows: 0 (0–10% membrane staining of invasive tumor cells), 1+ (weak, >10% incomplete membrane staining), 2+ (moderate, >10% partial or complete membrane staining) and 3+ (strong, >30% complete membrane staining). Cases rated as 0 or 1+ were considered unamplified, while those rated as 3+ were classified as HER2-positive. Equivocal (2+) cases underwent further evaluation using fluorescence in situ hybridization (FISH). A high Ki-67 index was defined as nuclear staining in ≥14% of tumor cells.

2.4 Follow-up

Progression-free survival (PFS) was defined as the time from the initial surgical procedure to tumor recurrence or distant metastasis. Patients who remained progression-free at the final follow-up were considered censored in the analysis. Overall survival (OS) was defined as the time from surgery to death or last follow-up, with patients who were alive at the final follow-up also treated as censored events. Survival data were obtained through clinical visits or telephone interviews with patients and their relatives. The last follow-up date was March 2021.

2.5 Nomogram development and validation

Hazard ratios (HRs) and 95% confidence intervals (CIs) for potential prognostic factors were estimated using the Cox proportional hazards (PH) regression model. Independent risk factors were identified through stepwise backward selection in the Cox PH model. In this study, the patients were divided into a training set, comprising 386 patients from Tianjin Medical University Cancer Institute and Hospital, and a validation set, consisting of 215 patients from the First Affiliated Hospital of Tsinghua University. The nomogram for predicting 5-, 7- and 10-year PFS was constructed based on the training cohort, incorporating all identified independent prognostic factors. The model’s predictive performance was evaluated using internal validation (training cohort) and external validation (validation cohort).

2.6 Statistical analysis

All statistical analyses were performed using SPSS version 24.0 (SPSS Inc., Chicago, IL, USA). Categorical variables were compared using Pearson’s chi-square test. Univariate survival analysis was conducted using the Kaplan-Meier method, while independent prognostic factors were identified through Cox regression analysis. A two-sided p-value < 0.05 was considered statistically significant. The nomogram was developed and validated using R software version 3.6.3.

3.1 Patient characteristics

The baseline clinical and biological characteristics of the study population are shown in Table 1. A total of 601 patients met the inclusion criteria, and the mean age was 51.4 ± 13.0 years (range, 22–88 years). All patients were female and of Chinese ethnicity. Pathological diagnoses included carcinoma in situ (n = 15, 2.5%), infiltrating ductal carcinoma (n = 515, 85.7%), and other invasive carcinomas (n = 71, 11.8%). Tumor grading based on the World Health Organization (WHO) classification identified 89 patients (14.8%) as grade I, 376 (62.6%) as grade II and 136 (22.6%) as grade III. LVI was observed in 25 patients (4.2%), while 454 patients (75.6%) had no axillary lymph node metastasis. Tumor size distribution included 388 patients (64.6%) classified as T1, 196 (32.6%) as T2 and 17 (2.8%) as T3. Based on tumor staging, 14 patients (2.3%) were classified as stage 0, 315 (52.4%) as stage I, 213 (35.5%) as stage II and 59 (9.8%) as stage III. Regarding molecular biomarker expression, 70.7% of patients were ER-positive, 65.7% were PR-positive and 15.6% were HER2-positive. Molecular subtypes [19] based on immunohistochemistry were classified as luminal A (19.5%), luminal B (57.6%), HER2-positive (6.3%) and basal-like (16.6%).

Ultrasound examination revealed that 437 patients (72.7%) exhibited non-circumscribed margins, 159 (26.5%) demonstrated posterior shadowing and 242 (40.3%) had microcalcifications.

3.2 Associations between ultrasonic features and clinicopathological factors

Table 2 summarizes the relationships between ultrasound features and clinicopathological characteristics. Non-circumscribed tumor margins were significantly associated with invasive ductal carcinoma (p = 0.004), smaller tumor size (p = 0.024), and higher ER and PR positivity (both p <0.001). In contrast, circumscribed margins were predominantly observed in basal-like carcinoma (p < 0.001). Posterior shadowing was associated with N3 lymph node status (p = 0.002) and a higher PR-positive rate (p = 0.025). Additionally, the presence of microcalcifications correlated with higher histological grade (p = 0.015).

3.3 Univariate and multivariate survival analyses of ultrasonic and clinicopathological characteristics for PFS and OS in breast cancer patients

The median follow-up duration for the entire cohort was 136 months (range, 7–168 months).

In regard to PFS, patients with non-circumscribed margins exhibited a significantly higher PFS rate (90.4%) compared to those with circumscribed margins (61.0%) (p < 0.001, Fig. 2a), while no significant differences in PFS were observed between patients with and without posterior shadowing (81.8% vs. 82.6%, p = 0.735, Fig. 2b) or between those with and without microcalcifications (83.1% vs. 81.9%, p = 0.664, Fig. 2c). Univariate regression analysis identified several factors significantly associated with PFS, including tumor margin (p < 0.001), tumor size (p < 0.001), lymph node status (p < 0.001), tumor stage (p < 0.001), histological grade (p = 0.012), LVI (p = 0.003), molecular subtype (p = 0.023) and HER2 expression (p = 0.004) (Table 3). Multivariate Cox regression analysis based on these eight variables identified tumor margin p < 0.001), tumor size (p = 0.011), lymph node status (p < 0.001), and molecular subtype (p = 0.007) as independent predictors of PFS in BC patients (Table 3).

Fig. 2.Kaplan-Meier survival curves depicting progression-free survival (PFS) in BC patients stratified by ultrasound features. (A) PFS according to tumor margin, (B) PFS according to posterior shadowing, and (C) PFS according to microcalcifications. p-values were calculated using the log-rank test, with p < 0.05 considered statistically significant.

For OS, patients with non-circumscribed margins were found to have a significantly higher OS rate (92.7%) compared to those with circumscribed margins (68.3%) (p < 0.001, Fig. 3a). The OS rates were comparable between patients with and without posterior shadowing (83.6% vs. 86.9%, p = 0.207, Fig. 3b) and between those with and without microcalcifications (83.9% vs. 87.5%, p = 0.216, Fig. 3c). Univariate regression analysis revealed that tumor margin (p < 0.001), age (p = 0.009), tumor size (p < 0.001), lymph node status (p < 0.001), tumor stage (p < 0.001), histological grade (p = 0.013), LVI (p = 0.074), molecular subtype (p = 0.014), and HER2 expression (p = 0.005) were significantly associated with OS (Table 4). In the subsequent multivariate regression analysis, tumor margin (p < 0.001), age (p = 0.020), tumor size (p = 0.026), lymph node status (p < 0.001), and molecular subtype (p = 0.002) were identified as independent prognostic factors for OS (Table 4).

Fig. 3.Kaplan-Meier survival curves depicting overall survival (OS) in BC patients stratified by ultrasound features. (A) OS according to tumor margin, (B) OS according to posterior shadowing, and (C) OS according to microcalcifications. p-values were calculated using the log-rank test, with p < 0.05 considered statistically significant.

3.4 Nomogram construction

A nomogram was developed to identify high-risk BC patients with poor prognoses and potential metastatic lesions. Risk factors associated with PFS were initially evaluated using univariate and multivariate regression analyses (Table 3). Although tumor stage, HER2 expression, histological grade and LVI were significantly associated with PFS in univariate analysis, they were not retained as independent predictors in multivariate analysis. Based on multivariate regression findings, four independent prognostic factors, including non-circumscribed margin (p < 0.001), tumor size (p = 0.011), lymph node status (p < 0.001) and molecular subtype (p = 0.007), were selected for nomogram construction. Using these variables, a predictive model was developed to estimate 5-, 7- and 10-year PFS in BC patients (Fig. 4).

Fig. 4.A nomogram for predicting 5-, 7- and 10-year PFS in BC patients based on four independent prognostic factors: non-circumscribed margin, tumor size, lymph node status and molecular subtype. HER-2: human epidermal growth factor receptor 2.

3.5 Nomogram validation

The predictive performance of the nomogram was assessed through both internal and external validation. In the training cohort, the concordance index (C-index) for PFS prediction was 0.752 (95% CI [0.690–0.815]), demonstrating good discriminative ability. External validation using an independent cohort yielded a C-index of 0.772 (95% CI [0.705–0.840]), further confirming the model’s robustness. Calibration curve analysis showed strong concordance between the nomogram-predicted and observed survival probabilities in both the training and validation cohorts (Fig. 5). The predictive accuracy of the nomogram was further evaluated using receiver operating characteristic (ROC) curve analysis (Fig. 6). The area under the curve (AUC) values for 5-year PFS were 0.729 (95% CI [0.636–0.820]) in the training cohort and 0.774 (95% CI [0.700–0.852]) in the validation cohort. For 7-year PFS, the AUC values were 0.759 (95% CI [0.687–0.830]) in the training cohort and 0.757 (95% CI [0.691–0.824]) in the validation cohort. Similarly, the AUC values for 10-year PFS were 0.775 (95% CI [0.707–0.842]) and 0.775 (95% CI [0.701–0.849]) in the training and validation cohorts, respectively. Taken together, these findings indicate that the nomogram provides reliable and accurate predictions of PFS in BC patients.

Fig. 5.The calibration plots for predicting 5-, 7- and 10-year PFS in the training and validation cohorts. The x-axis represents predicted PFS, while the y-axis indicates observed PFS. (A,C,E) Calibration plots for the training cohort; (B,D,F) Calibration plots for the validation cohort. PFS: progression-free survival.

Fig. 6.Receiver operating characteristic (ROC) curves evaluating the discriminatory accuracy of the nomogram for predicting PFS in the training and validation cohorts for (A) 5-year PFS, (B) 7-year PFS and (C) 10-year PFS. AUC: area under the curve.