Blood flow on ultrasound imaging is a predictor of lump margin status in breast-conserving patients: a retrospective matching study

With Institutional Review Board approval, we identified all patients with invasive breast cancer confirmed by core needle biopsy who had undergone breast-conserving surgery at Shanxi Bethune Hospital between 1 January 2015 and 31 June 2022 (all were of Han Chinese descent). Patients who had neoadjuvant therapy and the biopsy tissue presented intraductal components or lobular histology or absence of margin status were excluded. The case group included patients with positive margins, and the control group included other patients. Bilateral BCS was recorded as two independent cases.

Pathological and demographic information was obtained from the hospital database. Clinical examination (palpability), location, and imaging features were retrieved from the electronic medical records. Imaging features were collected from ultrasonography, mammography, and breast magnetic resonance imaging (MRI) reports routinely performed before breast-conserving surgery.

Mammography provides information about microcalcifications, and because MRI is a highly sensitive examination, it is mainly used for evaluating multiple lesions. Multiple lesions included multifocal (multiple areas of tumor in one quadrant) and multicentric (multiple areas of tumor affecting more than one quadrant) [19] and were confirmed pathologically. If breast magnetic resonance imaging is considered to involve multiple lesions and is subsequently confirmed by postoperative pathology, it is categorized as the "multiple lesions" group. If breast magnetic resonance imaging is considered to involve multiple lesions, but postoperative pathology confirms some of these lesions to be benign, then the patient is classified as the "single lesion" group.

Ultrasound is useful for lump edge assessment, which is the borderline between tumor and normal tissue, and blood flow assessment. Typically, the tumor margins are not smooth and classified into three types: arcade-like structure, crab sign, and spiculated margin. A single lesion could present three manifestations at the same time. Referring to the Adler grading criteria [20], blood flow can be divided into four levels. Level 0, no blood flow signals detected within the mass. Level I, slight blood flow present, with 1 to 2 punctate or slender rod-like tumor vessels visible and rod-like blood flow does not exceed half the diameter of the lesion. Level II, moderate blood flow present, with 3 to 4 punctate vessels visible or a longer vessel entering the lesion; the length of the vessel may be close to or exceed the radius of the mass. Level III, abundant blood flow present, with ≥ 5 punctate vessels visible or 2 longer vessels visible. The schematic diagram for the four blood flow signal grades can be found in Fig. 1. In this study, "increased flow" refers to levels II and III. Crab sign, spiculated margin, and increased blood flow were recorded for those factors which may increase the possibility of a positive margin. Two sonographers (at least one with > 10 years of experience) independently issued the ultrasonography report. If there were discrepancies, they were adjudicated by a third sonographer. Ultrasound examinations were performed using a Philips iU22 equipped with an L12–5 probe (Philips Medical Systems, Bothel, WA).

Each surgery is performed by an experienced breast surgeon. The primary surgeon determines the incision location, the use of localization techniques, and other operative details based on factors, such as the tumor's position, size, and the condition of the breast.

It is standard practice in our institution that all breast cancer lumps are evaluated by frozen section examination. When the tissue is received in the laboratory, it is "annotated" to accurately represent the in vivo position. Sections for margin evaluation are perpendicular to the inked surface, and the distance between carcinoma and the inked margin can be measured microscopically. Two breast pathologists with > 10 years of experience independently assessed the specimens. If the margins are positive, they are usually re-excised, and the new margin is evaluated again by intraoperative frozen section analysis. All specimens are paraffin-embedded and tested after BCS. Reoperation is recommended if the permanent margin is unclear (except for the positive frozen margin with re-excision). In our study, the margin result was based on the postoperative pathological result. If intraoperative re-excision was performed, the slide of the primary excision was used as the final result. The positive definition is consistent with current guidelines [21, 22]: no inked margins for invasive cancer.

Matching ensures that the distributions of confounding variables are as close to identical as possible. CCM (case–control matching) involves selecting subjects according to the matching factors, and subjects in the control group are completely consistent with subjects in the case group [23]. In PSM (propensity score matching), the control selection for each case is based on the propensity score rather than matching factors [24]. To balance the baseline characteristics between groups, CCM and PSM were used to match patients with positive and negative margins at a 1:1 ratio. The matching parameters included age, BMI, menopausal status, tumor location, palpability, multiple lesions, tumor size, and microcalcifications. In CCM, the maximum allowance of difference (MAD) of BMI was set to 1 kg/m², the MAD of the tumor size was 1.0 cm, and an exact match was used for age, menopausal status, tumor location, palpability, multiple lesions, and microcalcifications. In PSM, the matching tolerance score between the case group and the control group was set at 0.02.

Normality was tested with the Kolmogorov–Smirnov test for continuous variables, which were expressed as mean ± standard deviation (SD) and analyzed by analysis of variance (ANOVA). Categorical variables were expressed as a proportion, and the paired chi-square test (χ²) was used to compare differences between groups. Missing values were handled using multiple imputations. All statistics were performed in SPSS 26.0, and all tests were two-sided. A P value < 0.05 was considered statistically significant. The standardized mean difference (SMD) was calculated to assess the balance after matching, and the threshold was set for a mean difference of 0.1 [25]. To control confounding variables, conditional logistic regression models were built and used to estimate the odds ratios (OR) and their associated 95% confidence intervals (CIs).