Association of Breast Tissue Composition on Preoperative Automated Breast Ultrasound With Accuracy of Cancer Multiplicity Evaluation and Recurrence-Free Survival in Patients With Early-Stage Breast Cancer.

INTRODUCTION
In breast ultrasound (US), background echotexture (BE) and glandular tissue component (GTC) are factors used to assess breast tissue characteristics. BE reflects breast tissue composition by indicating the balance between fibroglandular tissue (FGT) and fat, similar to mammographic density [1]. According to the 5th Breast Imaging Reporting and Data System (BI-RADS), documenting BE in radiology reports for screening US is essential because heterogeneous BE can obscure lesions and reduce diagnostic sensitivity [1]. In contrast, GTC evaluates the sonographic appearance of FGT by representing the proportion of glandular to fibrous tissue as reflected by echogenicity [2]. Lee et al. [3] demonstrated that GTC assessed using screening US independently predicted breast cancer risk in women with dense breasts. The recently revised BI-RADS 6th edition recommends documenting BE and GTC in screening and diagnostic US reports [4].
Breast density assessed using digital mammography (DM) is a well-established independent risk factor for breast cancer and has been incorporated into several risk assessment models [567]. Similarly, high background parenchymal enhancement (BPE) on MRI is associated with an increased risk of breast cancer [8]. However, few studies have explored the association between US characteristics of background breast tissue and breast cancer risk, likely because of the inherent limitations of handheld US (HHUS), which cannot image the entire breast and only provides operator-selected images. This limitation hampers the comprehensive evaluation of background breast tissue characteristics. In contrast, automated breast US (ABUS) acquires three-dimensional volume images of the entire breast and generates multi-slice and multiplanar views that are not possible with HHUS. Additionally, ABUS enables structured, standardized image acquisition and facilitates temporal comparisons [91011]. These features facilitate the objective assessment of breast tissue composition, including BE and GTC.
The purpose of this study was to determine whether BE and GTC on preoperative ABUS are associated with the accuracy of assessing additional cancers to determine cancer multiplicity (i.e., unifocal, multifocal/multicentric, or bilateral) and with postoperative recurrence-free survival (RFS) in patients with early-stage breast cancer, based on data from our institution’s participation in a multicenter study evaluating ABUS combined with DM for preoperative staging [12].

MATERIALS AND METHODS

Study Population
This was a retrospective analysis of patients enrolled at Samsung Medical Center as part of a multicenter prospective trial evaluating ABUS for the preoperative staging of early-stage breast cancer (NCT04607473). The current study was separately approved by the Institutional Review Board of the Samsung Medical Center (IRB No. 2024-05-008-001), and the requirement for written informed consent was waived.
A total of 468 women with early-stage breast cancer (clinical Tis, T1–T2/N0) who underwent ABUS and DM for preoperative tumor staging, followed by surgery between October 2019 and April 2021, were identified (Supplementary Material 1, Fig. 1). Early-stage breast cancer was defined as a biopsy-proven index cancer on DM measuring ≤5 cm without palpable axillary lymphadenopathy on clinical examination [12]. The exclusion criteria were as follows: 1) receipt of neoadjuvant chemotherapy (n = 58) and 2) absence of preoperative MRI (n = 1) (Fig. 1). The final study cohort included 409 women with newly diagnosed early-stage breast cancer who underwent ABUS, MRI, and DM.

Image Analysis
For preoperative local tumor staging, two-view full-field DM, ABUS, and breast MRI were performed for all patients. The imaging techniques are described in detail in Supplementary Material 2.
For the evaluation of BE and GTC, three board-certified radiologists (radiologists with 3–22 years of experience in breast imaging, M.K.K., H.J.K., and E.Y.K.) retrospectively reviewed the preoperative ABUS images. All image reviews were performed using anonymized ABUS data from a standalone viewer that was not connected to PACS. Image review was conducted in two sequential sessions: BE was first assessed in all patients, followed by GTC evaluation in women with mammographically dense breasts (heterogeneously or extremely dense) after a 1-month interval. Before the review, three radiologists participated in a calibration session with an expert (J.S.C.) using 50 representative cases that were not included in the study data to align their interpretations of BE and GTC.
BE, defined as the overall pattern of breast tissue composition on ABUS reflecting the balance between FGT and fat, was classified as homogeneous-fat (mainly consisting of fat lobules and echogenic supporting bands), homogeneous-fibroglandular (thick echogenic fibroglandular layer beneath thin hypoechoic fat), or heterogeneous (focal or diffuse heterogeneity seen as multiple small areas of increased and decreased echogenicity, with posterior shadowing at fat–parenchyma interfaces), according to BI-RADS for all patients [1]. In women with mammographically dense breasts, glandular tissue for GTC classification was defined as gray areas within the FGT that were distinct from fat lobules on ABUS. The proportion of gray areas representing glandular tissue relative to white areas representing fibrous stroma within the FGT was assessed. In entirely fatty or scattered fibroglandular breasts, the glandular component is too limited or too intermingled with fat for a reliable GTC assessment. Therefore, our analysis focused on women with dense breasts, in whom GTC can be evaluated more consistently. GTC was qualitatively classified based on the amount of glandular tissue in FGT as follows: minimal (<25% of the FGT), mild (25%–49% of the FGT), moderate (50%–74% of the FGT), or marked (≥75% of the FGT) [23].
To assess inter-observer agreement, the radiologists independently recorded the BE and GTC in each session. During the review, the readers were informed that ABUS was performed for the preoperative staging of newly diagnosed breast cancer. However, they were blinded to any information that could identify the patients (e.g., name and age), preoperative imaging findings on DM and MRI, final surgical pathology, and post-treatment outcomes, including postoperative surveillance imaging and recurrence status. Because the reviewers were blinded to both cancer multiplicity and the specific locations of any cancer, the readers were instructed to assess BE and GTC using the contralateral breast when a suspicious lesion was present in only one breast and to use images from quadrants without suspicious lesions when both breasts contained suspicious lesions. Specifically, when the four quadrants within the same breast showed variable tissue characteristics, the representative BE and GTC category was determined based on the pattern observed in the majority of quadrants; in cases of a 2:2 tie, the category showing more heterogeneous features (i.e., heterogeneous over homogeneous for BE, or high over low for GTC) was selected. For each case, the final BE and GTC categories were determined by majority vote (agreement by at least two of the three reviewers) and used for all main analyses.

Data Curation, Including Postoperative Follow-Up
The following data were collected from electronic medical records: age at breast cancer diagnosis; menopausal status; family history of breast cancer in first-degree relatives; BRCA1/2 genetic testing results; breast density on DM; BPE on MRI; clinical and pathologic TNM stage; histopathological characteristics of the index and additional cancers detected by preoperative imaging; hormone receptor and HER2 expression status of index cancers; surgical methods; and adjuvant treatments including radiation therapy, chemotherapy, and endocrine therapy (tamoxifen and/or aromatase inhibitor) (Supplementary Material 3). Among imaging variables, GTC was dichotomized as “low” (minimal or mild GTC) and “high” (moderate or marked GTC) using a threshold of 50% GTC within the breast FGT. Breast density on DM was classified as “nondense” (almost entirely fatty and scattered fibroglandular tissue) and “dense” (heterogeneous and extremely dense). BPE on MRI was dichotomized as “low” (minimal and mild BPE) and “high” (moderate and marked BPE).
Postoperative surveillance with combined DM and whole-breast US was initiated between 6 months and 1 year after surgery and was subsequently performed annually for 5 years at our institution. Breast MRI was performed in patients with a high risk of recurrence or upon patient preference [13]. US examinations were performed by dedicated breast radiologists with 1–24 years of experience. Distant metastasis was evaluated using chest radiography, contrast-enhanced CT, bone scintigraphy, or whole-body 18F-fluorodeoxyglucose PET, as clinically indicated. The follow-up duration was measured from the date of surgery to the date of recurrence, death, or last follow-up.

Study Endpoints
The primary endpoint was the accuracy of preoperative patient categorization regarding cancer multiplicity—unifocal (a single index cancer), multifocal or multicentric (additional ipsilateral cancer), or bilateral (contralateral cancer)—using ABUS in combination with DM. ABUS findings were compared with the histopathological findings to determine the presence of additional cancer lesions beyond the biopsy-confirmed index lesions. An accurate assessment was considered when patient categorization was consistent with histopathological findings. If the ABUS classification did not match the histopathological results, the classification was considered inaccurate (Supplementary Material 4).
The secondary endpoint was RFS, defined as the interval from the date of surgery to the first recurrence or death from any cause. Recurrence events were classified as local-regional (including the ipsilateral breast or regional lymph nodes), contralateral breast, or distant metastases, depending on the site of the first recurrence. Local, regional, and contralateral breast recurrences were confirmed as breast cancer by biopsy, whereas distant metastases were included when biopsy was not feasible but imaging findings were unequivocally consistent with metastatic disease. The follow-up duration was measured from surgery until the last follow-up visit, date of recurrence, or death.

Statistical Analysis
Associations between clinicopathological and imaging variables and the accuracy of patient categorization of cancer multiplicity by ABUS combined with DM were evaluated using univariable and multivariable logistic regression analyses to estimate odds ratios (ORs) for inaccurate categorization. From the collected dataset, only variables that could be assessed preoperatively were included as independent variables. Variables with P < 0.05 in univariable analyses were entered into the multivariable model, which was constructed using a bidirectional stepwise selection procedure based on the score chi-square statistic (entry and stay criteria P-value: slentry = 0.05, slstay = 0.05).
To evaluate the factors associated with RFS, hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards regression. Variables with P < 0.05 in univariable analyses were included in the multivariable Cox model using the same bidirectional stepwise selection procedure. Firth correction was applied to reduce small-sample bias due to the limited number of recurrence events.
The analyses described above were performed for all patients (excluding GTC) and for a subgroup of patients with dense breasts on DM (including GTC).
Interobserver agreement in assessing BE for the overall cohort and GTC for the dense breast group was computed using the Fleiss κ statistic with its 95% CIs. The lower bound of the 95% CI was used to determine the degree of interobserver agreement: marginal, 0–0.39; good, 0.40–0.74; and excellent, 0.75–1.00 [14].
A two-sided P-value of <0.05 was considered statistically significant. All statistical analyses were performed using R Statistical Software (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Characteristics of Study Population
Table 1 presents the characteristics of the overall cohort (n = 409) and the dense breast group (n = 323). The median age was 48 years (interquartile range, 43–53 years) in the overall cohort and the dense breast group. BE on ABUS was most commonly classified as homogeneous-fibroglandular, which was observed in 88.3% of the overall cohort and 97.5% of the dense breast group. Mild GTC were the most frequent (59.2%) GTCs on ABUS in the dense breast group. Premenopausal women were more prevalent in the dense breast group (60.4%) than in the overall cohort (51.1%). The proportions of other variables were comparable between the overall and dense breast groups.

Factors Associated With Inaccurate Cancer Multiplicity Categorization by ABUS Combined With DM
Preoperative cancer multiplicity categorization using ABUS with DM was accurate in 368 patients (90.0%) overall and in 289 patients (89.5%) in the dense breast group. Inaccurate categorization occurred in 41 patients (10.0%) overall and in 34 patients (10.5%) in the dense group (Supplementary Material 5, Supplementary Table 1).
Table 2 summarizes the factors associated with inaccurate preoperative cancer multiplicity categorization using ABUS with DM. BE on ABUS showed no association with categorization accuracy in either the overall cohort (P = 0.64) or the dense breast group (P = 0.20). GTC on ABUS also showed no significant association with categorization accuracy in the dense breast group (P = 0.55) (Fig. 2). In contrast, high BPE on MRI was associated with inaccurate categorization in both the overall cohort (OR 2.13 [1.10–4.10]; P = 0.02) and the dense breast group (OR 2.75 [1.32–5.71]; P = 0.007). Age, menopausal status, BRCA1/2 mutation status, and family history of breast cancer were not associated with categorization accuracy (all Ps > 0.05).

Survival Outcomes and Factors Associated With RFS
Median postoperative follow-up duration was 3.5 years (range, 0.7–4.5 years). Recurrence was observed in 11 patients (2.7%) in the overall cohort, including 9 patients (2.8%) in the dense breast group (Supplementary Material 6, Supplementary Table 2). Five women had ipsilateral breast cancer, four had contralateral breast cancer, one had ipsilateral axillary lymph node recurrence, and one had distant metastasis. No deaths occurred, and the 3-year RFS was 98.0% (95% CI, 96.6–99.4).
Table 3 presents the factors associated with RFS after initial breast cancer treatment. In the multivariable analysis, heterogeneous BE on ABUS was independently associated with worse RFS compared with homogeneous-fibroglandular BE in the overall cohort (HR 11.24 [2.82–44.92]; P = 0.001) and in the dense breast group (HR 14.17 [2.69–74.60]; P = 0.002) (Fig. 3). In the dense breast group, high GTC on ABUS was associated with worse RFS compared with low GTC (HR 10.32 [2.35–45.28]; P = 0.002) (Fig. 4). BRCA1/2 mutation positivity was associated with worse RFS in both the overall cohort (HR 15.94 [2.47–102.97]; P = 0.004) and the dense breast group (HR 24.34 [2.75–215.06]; P = 0.004), compared with those who did not undergo BRCA mutation testing. Larger pathologic index cancer size was associated with RFS in both the overall cohort (HR per 1-cm increase 1.91 [1.13–3.23]; P = 0.02) and dense breast group (HR 2.62 [1.50–4.59]; P = 0.001).

Interobserver Agreement in ABUS BE and GTC Evaluation
Interobserver agreement for BE (κ = 0.679 [95% CI: 0.636–0.722]) and GTC (κ = 0.539 [95% CI: 0.496–0.582]) on ABUS among the three radiologists was good.

DISCUSSION
To date, few studies have investigated whether breast tissue characteristics assessed using US, such as BE and GTC, are associated with future cancer risk or influence the accuracy of cancer diagnosis. In this study of patients with early-stage breast cancer, BE and GTC on ABUS were not associated with the accuracy of preoperative assessment of additional cancers for determining cancer multiplicity when ABUS was used in combination with DM in both the overall and dense breast groups. However, in the same cohort, heterogeneous BE on ABUS, along with pathologic cancer size and BRCA mutation positivity, was associated with worse RFS in both the overall and dense breast groups. In addition, high GTC on ABUS was associated with worse RFS in the dense breast group.
Heterogeneous BE is characterized by multiple small areas of increased and decreased echogenicity, often accompanied by posterior shadowing at the fat-parenchyma interface [1]. These shadowing areas may reflect a more extensive fat-glandular interface, defined as the boundary separating FGT from the surrounding fat, compared with homogeneous BE. According to previous studies [1516], cancers tend to arise in areas of the fat-glandular interface, and an extensive fat-glandular interface is associated with a higher likelihood of cancer development. These interfaces are also where major blood vessels and lymphatic channels supplying the breast are located, and may contribute to carcinogenesis [1516]. This observation may explain the association between heterogeneous BE and worse RFS observed in our study, although the precise mechanisms remain unclear and require further investigation. In addition, high (moderate or marked) GTC within FGT was associated with worse RFS than low (no or mild) GTC in the dense breast group. This finding is consistent with a previous study reporting that high GTC on screening breast US was associated with an increased future cancer risk in women with dense breasts [3]. However, this association should be interpreted with caution because only 11 recurrence events occurred among the 409 patients, resulting in wide CIs and unstable estimates. Moreover, the markedly skewed distribution of BE (88% homogeneous-fibroglandular) may have limited the reliability of the analysis. Therefore, further studies with larger cohorts are required to validate our results.
In addition, our overall results suggest that tissue characteristics assessed using different imaging modalities may capture distinct biological or structural features of the breast rather than reflecting the same construct. DM density represents the total amount of FGT relative to fat, whereas MRI BPE reflects functional characteristics influenced by hormonal and vascular factors [1718]. On US, BE reflects the internal heterogeneity of the FGT, including variations in the fat–glandular interface, whereas GTC represents the proportion of glandular to fibrous components, both derived from echogenicity differences [123]. Collectively, our findings demonstrate that both BE and GTC on ABUS are independently associated with RFS in early-stage breast cancer, further supporting their potential utility as complementary imaging biomarkers for cancer risk stratification, similar to DM density and MRI BPE.
Previous studies have reported that a larger cancer size and BRCA mutation positivity are associated with an increased risk of breast cancer recurrence after treatment [1920212223]. Consistent with prior findings, our results demonstrated that both larger pathologic cancer size and BRCA mutation positivity were independently associated with worse RFS in patients with early-stage breast cancer. However, the prognostic impact of BRCA mutation in our study should be interpreted with caution. As BRCA testing was selectively performed, only a small subset of patients ultimately underwent testing. Moreover, recurrent events were rare. These factors resulted in large HRs with wide CIs, leading to statistically unstable estimates. Therefore, this finding requires validation in larger and more uniformly tested cohorts. In contrast, several previously reported factors associated with breast cancer recurrence after treatment, including N stage, younger age, and human epidermal growth factor receptor 2 positivity, were not statistically significant in our analysis [192021222324]. This may be attributed to the inclusion of only early-stage breast cancers, the consistent application of curative surgery and adjuvant treatments according to the NCCN guidelines, and the relatively short follow-up period.
Although it remains unclear whether and how BE or GTC affects the sensitivity of breast US, we hypothesized that heterogeneous BE or high GTC may hinder the detection of small or subtle cancers. However, our findings showed that neither BE nor GTC affected the evaluation of additional cancers when used in combination with DM, suggesting that ABUS can be reliably used for the preoperative staging of patients with early-stage breast cancer. In a related finding, high (moderate or marked) BPE on MRI was significantly associated with inaccurate preoperative evaluation of additional cancers on ABUS in both the overall and dense breast groups. BPE on MRI reflects functional, hormone-, and vascular-driven enhancement [18], whereas BE on ABUS represents the structural distribution of fibrous and glandular tissues. In our cohort, only one patient with high BPE demonstrated heterogeneous BE, and BE itself was not associated with ABUS accuracy. We speculate that a high volume of hormonally active FGT, manifesting as a high BPE on MRI, may negatively affect the evaluation of additional small cancers on ABUS, although this mechanism could not be confirmed from our data. Further research is required to better understand this association.
ABUS is actively used for screening to overcome the limitations of HHUS, which can be time-consuming and operator-dependent. Recently, research has expanded the role of ABUS beyond screening to diagnostic areas such as preoperative staging and evaluation of the response to neoadjuvant chemotherapy in patients with breast cancer [122526]. ABUS acquires three-dimensional volumetric images of the entire breast and reconstructs multi-slice and multiplanar views that cannot be achieved with HHUS. This allows for the objective evaluation of BE and GTC. Thus, expanding the use of ABUS may encourage further research on the role of BE and GTC as potential imaging biomarkers of future breast cancer risk.
Our study had several limitations. First, its retrospective design and single-center setting may have introduced a potential selection bias. Second, the follow-up period was relatively short, which may have been insufficient to fully assess the long-term recurrence outcomes. Further large-scale, long-term studies are needed to validate the effects of BE and GTC on the risk of breast cancer. Third, BE and GTC were evaluated based on visual assessments. Therefore, there is a potential for interobserver variability. However, considering that DM density and MRI BPE are also evaluated visually based on standardized criteria, we believe that the use of BE and GTC as imaging markers would not pose a significant issue as long as the readers are well acquainted with the reference standards. Finally, although the readers were blinded to DM, MRI, pathology, and clinical outcomes, they were aware that ABUS was performed for preoperative staging in patients with confirmed breast cancer. While this awareness may be regarded as a potential source of bias, it reflects the real-world clinical setting in which radiologists interpret preoperative images based on their knowledge of cancer diagnosis.
In conclusion, in patients with early-stage breast cancer, heterogeneous BE and high GTC on preoperative ABUS, together with larger cancer size and BRCA mutation positivity, were associated with worse RFS. However, BE and GTC did not affect cancer multiplicity evaluation when ABUS was used in combination with DM.