Supplemental breast screening ultrasound (US) increases the detection rate of node-negative cancers in women with dense breasts compared with mammography alone [
1,
2]. However, the increase in cancer detection rate also results in an increase in false-positive findings [
1–
3]. In the ACRIN 6666 study [
1,
2], the addition of supplemental US screening prompted 7.8% (207/2659) of women to undergo biopsy and another 10.7% (284/2659) to undergo short-term follow-up in the initial year. In years 2 and 3 combined, the number of biopsies prompted by US was 5.0% (242/4814), and 3.7% (180/4814) underwent short-term interval follow-up [
3]. BI-RADS 3 lesions account for a large fraction of the short-term follow-up studies and biopsies resulting from supplemental US screening, significantly increasing health care costs [
3,
4]. Though US BI-RADS 3 lesions are common, they have a very low rate of malignancy (0.4–0.8%) [
3,
5–
7]. In the absence of a mammographic correlate, the malignancy rate of US BI RADS 3 lesions is particularly low [
6].
Most cancers in US BI-RADS 3 lesions are not diagnosed until at least 12 months of follow-up is conducted [
3,
5,
8]. In the ACRIN 6666 study, US BI-RADS 3 lesions that were ultimately diagnosed as malignant on follow-up remained node negative at 1 year, and only one cancer had a metastatic node at 24-month follow-up [
1,
3]. A retrospective analysis of prospectively collected data from the ACRIN 6666 study suggested that routine annual follow-up for US BI-RADS 3 lesions may be warranted, though further validation remains needed [
3].
Automated breast ultrasound (ABUS) represents an alternative to physician-performed handheld US. ABUS is less time-consuming for the physician, provides 3D breast volumes, and can be performed by trained sonographers or mammographers with consistent performance [
9]. However, when performing ABUS, color or power Doppler ultrasound cannot be performed, and manual setting adjustments during the examination are not possible. Previous studies have shown that ABUS has high sensitivity and fair interobserver concordance [
10–
12].
The aim of this study was to evaluate outcomes in patients who were prospectively recommended for routine annual follow-up for lesions assessed as BI-RADS 3 on supplemental screening ABUS.
Methods
Patients
This prospective study (NCT02650778) was approved by the institutional review board of Western Reserve Health Education and was HIPAA compliant. The study received equipment support from Siemens Healthineers. The authors had control of the data and the information submitted for publication. All participants gave written informed consent. No participants were previously reported.
Participants were enrolled from August 2013 to December 2016 at a single outpatient imaging center. Women with a BI-RADS 1 or 2 result on consecutive screening mammograms and a breast density of C or D were invited to participate. Individuals agreeing to participate underwent an additional ABUS examination for research purposes within 30 days of the screening mammogram. This ABUS examination was the first (prevalence) screening US examination for all participants. The Gail risk score was calculated for each participant to provide an estimate of the number of participants with elevated risk. Participants with Gail risk score indicating greater than a 1.6% 5-year risk of breast cancer or with a personal history of breast cancer were considered to be at elevated risk [
13].
Screening Mammography
Participants underwent standard mediolateral oblique and craniocaudal view screening of both breasts using film-screen mammography, digital mammography, or digital breast tomosynthesis. Mammograms were interpreted according to BI-RADS 4th edition [
14] by one of six radiologists who had from 15 to 30 years of experience in interpreting mammograms. Breast density was assessed by the radiologist interpreting the mammogram according to BI-RADS classifications.
Supplemental Screening by Automated Breast Ultrasound
Participants underwent a single prevalence ABUS examination using a Siemens S2000 system and a 15-cm L14–5MHz transducer (Siemens Healthineers). Parameters were set according to patient's breast cup size. Each patient underwent three to five scans of each breast depending on the size of the breast. Overlapping scans were obtained to confirm complete coverage of the breast. The examination time ranged from approximately 15–30 minutes depending on the size of the breasts. Examinations were performed by one of six sonographers or mammography technologists who participated in a training course on ABUS and who satisfactorily completed an ABUS examination in at least 10 women before the start of this study [
9].
The ABUS examinations were interpreted by a single radiologist (R.G.B.) with 20 years of experience in breast US and 2 years of experience in interpreting ABUS. Examinations were interpreted on an independent workstation with 3D reconstruction of the 2D scanned clips (
Fig. 1). Each detected lesion was documented, evaluated in all three planes, and assigned a BI-RADS score of 1, 2, 3, or 0. The examination's overall ABUS BI-RADS score was the highest of the BI-RADS scores across all lesions identified in both breasts, with 0 being the highest score.
Classification of ABUS BI-RADS 3 lesions strictly followed the ACRIN 6666 definitions [
1,
3]. Lesions considered probably benign at US could not be palpable or exhibit suspicious US features. Lesions considered probably benign when identified at baseline screening included oval masses parallel to the skin and hypoechoic to fat with circumscribed borders and no posterior features or minimal posterior enhancement; hyperechoic masses with central hypoechoic to anechoic components suggestive of fat necrosis; hypoechoic oval masses with homogeneous low-level internal echoes that otherwise met the criteria for simple cysts (circumscribed acoustic enhancement); microlobulated or oval masses composed entirely of clustered microcysts with or without layering microcalcifications; probable artifactual posterior shadowing at the interface of fat lobules without any associated mass and that changed appearance with different angles of insonation; and architectural distortion thought to be a result of postsurgical scarring, which could be classified as probably benign or benign at the investigator's discretion [
15]. The presence of multiple bilateral circumscribed masses (at least three masses total and at least one mass in each breast) was considered to represent an ABUS BI-RADS 2 lesion [
16,
17],
Patients with an overall ABUS BI-RADS score of 1, 2, or 3 were provided a recommendation to return to routine mammographic screening. Specifically, patients with an overall ABUS BI-RADS 3 score were reported as “no sonographic evidence of malignancy, return to routine screening” to limit patients or referring physicians from pursuing an alternate intervention (e.g., short-term follow-up or biopsy). Thus, follow-up for negative supplemental ABUS screening was based on the patient or physician preference. Patients with an overall ABUS BI-RADS score of 0 were requested to undergo a handheld US examination for further evaluation.
HandHeld Ultrasound Examinations
Follow-up handheld US examinations for ABUS BI-RADS 0 lesions were performed using either an Aixplorer (SuperSonic Imagine) system with a 15-MHz transducer, a Siemens S3000 (Siemens Healthineers) system with a 14-MHz or 9-MHz transducer, or a Philips Epiq 7 (Philips Healthcare) system with a 12-MHz transducer. Handheld US examinations were performed by one of five sonographers with more than 5 years of experience in breast US and breast elastography. The examination was a targeted evaluation of the area of concern on ABUS. If the lesion was suspicious for malignancy, then whole-breast scanning with evaluation of the axilla was performed. All lesions were evaluated using power Doppler US and with both strain and shear wave elastography. The elastography findings were interpreted according to the investigators' prior experience and the results were recorded; however, the elastography results were not used to downgrade lesions in the study, and otherwise suspicious lesions underwent biopsy despite elastography findings [
18–
20]. Follow-up handheld US examinations were interpreted using BI-RADS by one of four physicians with from 10 to 20 years of experience in interpreting breast US. All US BI-RADS 4 and 5 lesions underwent biopsy. Additional US BI-RADS 3 lesions detected on handheld US were followed at 6 months and then at 1 year; these lesions were not included in the ABUS BI-RADS 3 follow-up analysis. If an ABUS BI-RADS 0 lesion could not be reproduced on handheld US and the handheld US examination was classified as BI-RADS 1 or BI-RADS 2, then the ABUS lesion was considered to have been artifactual.
Patient Follow-Up
Patients with ABUS BI-RADS 1, 2, or 3 were assessed for mammography, US, or clinical follow-up. Patients were not required to return to our center for follow-up of ABUS BI-RADS 1, 2, or 3 examinations. If the patient did not return, then the follow-up was performed by contacting the patient's referring physician to determine if the patient had undergone subsequent breast imaging at other facilities. Any subsequent diagnosis of breast cancer after the initial ABUS examination was recorded. Patients without 2 years of imaging or clinical follow-up were considered lost to follow-up. ABUS BI-RADS 3 lesions were considered benign if cancer was not diagnosed in the same quadrant as the ABUS BI-RADS 3 lesion within 2 years.
Statistics
Quantitative variables were described as mean, SD, and range. Categoric variables were described as counts and percentages. Prevalence of cancers and recall and biopsy rates were presented using the binomial 95% CIs. Fisher exact test was used to compare the ABUS recall rate with the ABUS recall rate obtained if all ABUS BI-RADS 3 lesions were recommended for 6-month interval follow-up. Fisher exact test was also used to compare the cancer detection rate with the rate reported in the ACRIN 6666 study [
1–
3]. Fisher exact test was implemented for r × c (in this case, 2 × 3) tables using the algorithm by Mehta and Patel [
21] that extends Fisher exact test from 2 × 2 contingency tables. The analyses were performed using Stata v16.1 (StataCorp). All
p values less than .05 were considered statistically significant.
Results
Of 23,426 women who underwent screening mammography, 8542 (36.5%) women were eligible for study participation (i.e., BI-RADS 1 or 2 on screening mammography and breast density C or D), and 2257 (26.4%) agreed to participate. The mean (± SD) patient age was 58.0 ± 11.2 years (range, 31–94 years). Screening mammography was performed by film-screen mammography in 182 (8.1%), digital mammography in 1922 (85.2%), and digital breast tomosynthesis in 153 (6.8%). The supplemental ABUS examinations were scored as BI-RADS 1 in 1186 (52.5%) women, BI-RADS 2 in 591 (26.2%), BI-RADS 3 in 395 (17.5%), and BI-RADS 0 in 85 (3.8%). Two-year follow-up was available for 1117 (94.2%) patients with ABUS BI-RADS 1, 556 (94.1%) with ABUS BI-RADS 2, and 394 (99.7%) with ABUS BI-RADS 3. Demographics for patients grouped by ABUS BI-RADS score are provided in
Table 1. A flow diagram is presented in
Figure 2.
A total of 572 ABUS BI-RADS 3 lesions were reported in the 395 patients with an overall ABUS BI-RADS score of 3. Of these lesions, the mean (± SD) size was 7.6 ± 3.5 mm (range, 3–25 mm; median 7.0 mm). These included 216 (37.8%) complicated cystic lesions, 338 (59.1%) solid circumscribed lesions with or without gentle lobulations, and 18 (3.1%) surgical scars (based on history). Of the 394 patients with an ABUS BI-RADS score of 3 and with 2-year follow-up, 359 (91.1%) had both imaging and clinical follow-up, and 35 (8.9%) had only clinical follow-up. No breast cancer (0%; 95% CI, 0.0–0.9%) was diagnosed in the same quadrant as the ABUS BI-RADS 3 lesion during the follow-up period. Therefore, the 394 ABUS BI-RADS 3 lesions with follow-up were all deemed benign.
Figures 3 and
4 show ABUS BI-RADS 3 lesions that were benign on follow-up.
The ABUS recall rate was 3.8% (85/2257; 95% CI, 3.6–4.0%). If short-term follow-up had been recommended for all 395 women with ABUS BI-RADS score of 3, then the recall rate would have been significantly different (p < .001) at 21.3% (480/2257, 95% CI, 19.6–23.0%). Among the 85 patients with ABUS BI-RADS 0, subsequent handheld US examination resulted in BI-RADS 1 in eight (i.e., deemed to have been artifactual on ABUS), BI-RADS 2 in 63, BI-RADS 3 in two, BI-RADS 4A in five, BI-RADS 4B in one, BI-RADS 4C in two, and BI-RADS 5 in four patients. Of these 85 patients, 12 underwent biopsy, of which the five BI-RADS 4A lesions were benign (intraductal papilloma in one, fibrocystic change in two, fibroadenoma in two) and the seven BI-RADS 4B, 4C, and 5 lesions were positive for cancer. Thus, the biopsy rate in the overall study sample was 0.4% (12/2257; 95% CI, 0.3–0.9%) and the positive biopsy rate was 58.3% (7/12). All five BI-RADS 4A lesions were classified as benign on elastography; if these lesions had been downgraded as a result of elastography, then the biopsy rate would have been 0.3% (7/2257) and the positive biopsy rate would have been 100.0% (7/7).
Over the 2-year follow-up period, breast cancer was diagnosed in eight additional patients: 4/1117 (0.4%) patients with ABUS BIRADS 1, 2/556 (0.4%) patients with ABUS BI-RADS 2, and 2/394 (0.5%) patients with ABUS BI-RADS 3 (other quadrants than the quadrant with the ABUS BI-RADS 3 lesion). The malignancy rate was not significantly different between patients with ABUS BI-RADS scores of 1, 2, and 3 (
p = .28).
Table 2 provides additional information regarding the 15 patients in the study sample with a diagnosis of breast cancer (seven cancers diagnosed by the initial screening study and eight cancers diagnosed over the 2-year follow-up). One of the seven cancers diagnosed by the initial supplemental ABUS (i.e., patients referred for handheld US after ABUS BI-RADS 0) and none of the eight cancers diagnosed during subsequent follow-up (i.e., patients recommended for return to routine screening after ABUS BI-RADS 1, 2, or 3) were node-positive breast cancer.
The study sample included 2013/2257 (89.2%) patients with average risk and 244/2257 (10.8%) patients with elevated risk. Among patients with ABUS BI-RADS 3, 48/395 (12.2%) were elevated risk, and 347/395 (87.5%) were average risk. Cancer was diagnosed on initial supplemental ABUS screening in 0.2% (5/2013) of patients with average risk and 0.8% (2/244) of patients with elevated risk.
The seven additional cancers detected by the initial supplemental ABUS examination among the 2257 patients represented detection of 3.1/1000 additional cancers (95% CI, 1.2–7.0%). This cancer detection rate was not significantly different (
p = .81) from the rate in the ACRIN 6666 study of 3.7/1000 (95% CI, 2.1–5.8%) [
1].
Discussion
In this prospective single-institution study, recommending return to routine screening for ABUS BI-RADS 3 lesions resulted in a recall rate of 3.8%. In contrast, recommending a 6-month follow-up for all ABUS BI-RADS 3 lesions would have resulted in a recall rate of 21.3%. The biopsy rate was 0.4%. The positive biopsy rate was 58.3% and would have been 100.0% had elastography been used to downgrade and avoid biopsy of BI-RADS 4A lesions on handheld US. Supplemental ABUS detected 3.1/1000 additional cancers. Over the 2-year follow-up period, no cancer was detected in the same quadrant as an ABUS BI-RADS 3 lesion, and cancer was later diagnosed after the initial prevalence ABUS examination in 0.4% of patients with ABUS BI-RADS 1, 0.4% with ABUS BI-RADS 2, and 0.5% with ABUS BI-RADS 3 (other quadrants from the lesion). No patient who received a diagnosis of breast cancer at follow-up after the initial screening had nodal metastasis. The findings indicate that a return to routine screening for ABUS BI-RADS 3 lesions results in a substantial decrease in recall rate and is unlikely to result in an adverse outcome.
The observation that no ABUS BI-RADS 3 lesion was malignant is consistent with an expected low malignancy rate for BI-RADS 3 lesions on US according to numerous earlier studies [
1,
15,
22–
24]. An important corollary to this low expected malignancy rate is that the small number of cancers diagnosed at follow-up should have a prognosis matching that of cancers detected at screening [
25]. Indeed, in our study, the occurrence of a cancer detected in the 2-year follow-up period was similar for patients with ABUS BIRADS scores of 1, 2, and 3 on the initial screening. Therefore, it is appropriate that ABUS BI-RADS 3 lesions be recommended for routine annual follow-up, similar to patients with ABUS BI-RADS 1 or 2. Avoiding earlier follow-up for these probably benign lesions can be associated with substantial cost savings [
4,
26–
30].
In comparison with the ACRIN 6666 study [
1,
2] that only performed supplemental screening US in patients with elevated risk, our study included patients with both average and elevated risk. Nonetheless, the two studies had similar frequencies of BI-RADS 3 lesions on supplemental US screening and similar cancer detection rates, with the supplemental screening identifying node-negative cancers not identified on screening mammography in both cohorts. The present findings support a current common recommendation to perform supplemental screening US in all women with breast density of C or D [
31].
This study has several limitations. It is a single-center study performed by investigators with considerable clinical and research experience in breast US and ABUS; the findings therefore may not be generalizable to all practice settings. In particular, the initial ABUS examinations were interpreted by a single radiologist with extensive training in breast US, and reproducibility was not evaluated. However, the use of ABUS provides the advantage of standardization in the documentation, recording, and archiving of US images from the entire breast excluding the axilla, which can help mitigate variability in image acquisition. It has been shown that with adequate training, ABUS images of similar quality can be obtained from ultrasonographers and mammographers [
9]. Adopting the ACRIN 6666 definitions of BI-RADS 3 lesions will further enhance standardization in the classification of BI-RADS 3 lesions. Also, breast density was determined qualitatively by the interpreting radiologist without quantitative assessment. Additionally, ABUS BI-RADS 3 lesions were not imaged longitudinally by US to assess change over time. However, US follow-up of ABUS BI-RADS 3 lesions would carry the risk of additional follow-up and biopsy of benign lesions. Because most patients would be followed either yearly or every other year for screening even if a malignant BI-RADS 3 lesion identified on supplemental US screening was not followed at 6 months, the cancer would most likely still be node negative at the next screening examination [
15]. An additional limitation is that most patients had 2D digital mammography; the detection rate of supplemental screening may have been less if tomosynthesis was used. Finally, the Gail risk score was used to assess patient breast cancer risk, though it has known limitations [
11].
In conclusion, this prospective study provides data from women with dense breasts who underwent supplemental ABUS screening with a recommendation for return to routine screening for BI-RADS 3 lesions detected at ABUS. No ABUS BI-RADS 3 lesion (among 394 with 2-year follow-up) was malignant. Further, cancer was diagnosed during later follow-up at a similar rate for patients with ABUS BI-RADS 1, 2, and 3. In the overall study sample, the biopsy rate was low, and the positive biopsy rate was high. No cancer diagnosed at follow-up after initial screening was node positive. The findings provide increasing support for the appropriateness of a recommendation for return to routine annual screening for BI-RADS 3 lesions detected at supplemental ABUS.