The Potential Impact of Digital Breast Tomosynthesis on the Benign Biopsy Rate in Women Recalled within the UK Breast Screening Programme

Introduction

Digital breast tomosynthesis (DBT) is gaining momentum as a screening test internationally (1,2). DBT acquires multiple images over a limited angular range to produce a set of reconstructed images, which can then be viewed individually or sequentially in a cine loop and which provide three-dimensional information. The main advantage of DBT is its potential to improve the sensitivity and specificity of mammography. It provides a series of thin-section images through the breast that reduces the effect of tissue superimposition or overlap, which often leads to cancers being missed or women without breast cancer being recalled for diagnostic work-up after full-field digital mammography (FFDM) (3). The use of DBT should reduce tissue overlap and result in improved sensitivity and specificity relative to FFDM (4). Evidence shows that DBT can improve the cancer detection rate and reduce the false-positive recall rate (4–7).

In the National Health Service Breast Screening Programme (NHSBSP) in the United Kingdom, women are called for routine screening mammography every 3 years between the ages of 50 and 70 years. If an abnormality is detected, they are recalled for triple assessment of the abnormality (clinical examination, additional imaging, and potential biopsy) (8) (Fig 1a). In the United Kingdom, DBT may be used as an additional imaging modality as part of the triple assessment, but it is not used in the routine screening phase.

Figure 1a:

Download as PowerPointOpen in Image Viewer

Figure 1b:

Download as PowerPointOpen in Image Viewer

The NHSBSP also performs high-risk mammographic and MRI surveillance in women who are BRCA carriers or who have equivalent risk to those who are. They are invited annually for breast MRI surveillance from the age of 30 years and for combined mammographic and breast MRI surveillance from the age of 40 years (9,10).

From 2014 to 2015 (11), the NHSBSP screened 2 105 454 women. A total of 87 737 were recalled for screening assessment (4.2%) using FFDM only. Of these, 46.2% (40 509) underwent percutaneous biopsy, with a cancer detection rate of 8.6 per 1000 women screened (18 015 cancers). A total of 22 494 (55.5%) of the biopsies yielded benign results. Any abnormalities found during FFDM and US resulted in biopsy, emphasizing the limitations of FFDM in the differentiation between benign and malignant lesions.

Several studies have compared DBT with FFDM (12–14). These studies focused on the need for additional views when using DBT and the ability of DBT to increase the radiologist’s confidence in determining if abnormalities are benign or malignant. There is very little published material on the effect of DBT in reducing the benign biopsy rate. We hypothesized that using DBT in screening assessment would reduce the number of benign biopsies performed. The purpose of this study was to compare the use of DBT versus no DBT in women recalled for screening assessment after FFDM within the NHSBSP. The primary outcome measures were benign biopsy rate and cancer detection rate.

Materials and Methods

This single institution prospective study was performed between November 13, 2015, and July 29, 2016. Institutional review board approval was obtained (Integrated Research Application System project identification no. 134855). Written informed consent was obtained from all study participants. No industry support was received for this study, and the authors had control of all data used in the study.

Study Participants

Women who had an abnormality detected at screening FFDM were recalled for screening assessment and were invited to take part in our study. An information leaflet informing patients of the study was sent to them at the same time as their appointment letter for screening assessment. When these women attended their screening assessment appointment, they were asked if they would like to take part in our study. The consent form was then signed, and the women received standard care (ie, triple assessment), which consisted of clinical examination of the breast, imaging (additional two-dimensional [2D] mammography and US of the area of abnormality) and biopsy (if required), as well as DBT. This assessment was performed by consultant radiologists (N.S., B.D., I.H., M.M.; 10, 25, 15, and 6 years of experience, respectively). Study participants either proceeded to biopsy or were discharged to routine screening if no abnormality was seen at further imaging. Inclusion criteria included all consecutive women recalled for screening assessment between November 13, 2015, and July 29, 2016. Exclusion criteria included incomplete FFDM or DBT or if a study participant declined needle biopsy, as we would have been unable to determine a definitive diagnosis.

Breast Imaging

A Dimensions 2D/3D Digital Mammography System was used to acquire the 2D digital mammograms and DBT images (Hologic, Bedford, Mass). Only the study participants who consented to take part in the study underwent two-view (mediolateral oblique and craniocaudal) DBT of both breasts at the screening assessment clinic. The DBT images were not available for review by radiologists at the time they performed triple assessment.

Analysis

The DBT images were independently read by two of the four previously mentioned radiologists within 6 weeks of the assessment visit. These readers were blinded to the assessment result (ie, whether a biopsy had been performed and the result of this biopsy). If there was disagreement, a third radiologist read the images and a consensus opinion was reached, as per NHSBSP procedures. All four consultants had attended a certified course on DBT image interpretation.

The readers addressed two questions. The first was “Would the screening mammogram have been reported as abnormal with the benefit of DBT?” The DBT images were read along with the synthetic 2D image, and readers were asked to decide whether recall was required. The second question was “Would the abnormality noted on the DBT images have required a biopsy?”

The DBT images were reviewed with screening FFDM and US images and the results of any other tests performed at the time of triple assessment to determine whether the additional DBT information would have influenced the decision to perform biopsy or not. The DBT images received a mammographic score of R1–R5 based on the Royal College of Radiologists Breast Group mammography classification system (R1, normal; R2, benign; R3, indeterminate; R4, suspicious; and R5, malignant) (15). A classification of normal (R1) or benign (R2) is equivalent to returning the case to routine screening, whereas a classification of indeterminate (R3), suspicious (R4), or malignant (R5) is equivalent to recalling that case for screening assessment. Lesions with a score of R3–R5 require biopsy.

The R score of DBT was recorded; the need for a biopsy and the outcome of the biopsy were also documented. Data recorded from the standard assessment included the mammographic abnormality (well-defined mass, spiculate mass, microcalcifications, architectural distortion, or asymmetry), site of the lesion, R score, need for a biopsy (as indicated by mammography or US), and outcome if biopsy (nonblinded) was performed. The DBT images were reviewed in conjunction with current mammograms, without any previous mammograms available for review.

If any readers gained new information from DBT indicating the need for further assessment, biopsy, or both, then the study participant was recalled immediately after the DBT image had been read. This was always done within 6 weeks of the screening assessment visit.

Participants returned to routine screening in 3 years’ time if triple assessment and review of the DBT images yielded normal or benign findings.

With reference to the histologic diagnosis, the rates of true-positive (correct identification of abnormal appearance), true-negative (correct identification of normal appearance), false-positive (incorrect identification of abnormal appearance), and false-negative (incorrect identification of normal appearance) findings were calculated. Sensitivity, specificity, positive and negative predictive values, and accuracy were calculated for each of the two reading conditions, for all studies combined, and for subgroups by mammographic feature type.

Statistical Analysis

Two-tailed McNemar tests were used to test differences in sensitivity and specificity for the two conditions at a 5% significance level. Statistical software (IBM SPSS Statistics for Windows, version 23.0; IBM, Armonk, NY) was used for this analysis.

Results

During the study period, 30 933 women underwent routine screening mammography. A total of 1470 women were recalled for screening assessment (recall rate, 4.8% [1470 of 30 933]). There were 256 cancers identified in this group. Of the 1470 women, 835 (57%) consented to take part in the study (Fig 1b). The average age of the study cohort was 56.7 years ± 7.7 [standard deviation]. There were 827 women included, with eight women excluded because of incomplete data (Table 1). There were 145 breast cancers detected (142 were seen with FFDM, two additional cancers were seen only with DBT, and one was detected with surveillance MRIs).

Thirteen women had bilateral mammographic abnormalities and were recalled for assessment; four of these abnormalities were bilateral cancers. The cancer detection rate was 17.2% (142 of 827) for both groups, with and without the addition of DBT.

Sensitivity and Specificity of DBT versus FFDM as a Screening Test

Of the 827 women recalled after abnormal FFDM results, DBT findings were double reported as normal in 582 women. DBT as a screening tool would have resulted in the recall of only 29.6% (245 of 827) of women. Sixteen women required third-opinion arbitration of the DBT results. DBT would have missed two cancers that were identified with FFDM (Fig 2a); DBT also enabled identification of two cancers that were missed with FFDM (Fig 3). One cancer was occult at mammography and DBT and was identified only with MRI surveillance. The sensitivities of FFDM and DBT as screening tests in this cohort are the same (97.9% [142 of 145]).

Figure 2a:

Download as PowerPointOpen in Image Viewer

Figure 2b:

Download as PowerPointOpen in Image Viewer

Figure 2c:

Download as PowerPointOpen in Image Viewer

Figure 3a:

Download as PowerPointOpen in Image Viewer

Figure 3b:

Download as PowerPointOpen in Image Viewer

Figure 3c:

Download as PowerPointOpen in Image Viewer

Positive Predictive Value of DBT versus FFDM in Screening Assessment

The use of screening assessment without the addition of DBT resulted in 571 biopsies being performed and 142 cancers being detected. Of the 571 biopsies performed, 231 were performed with stereotactic guidance; 337, with US guidance; and three, with MRI guidance (Table E1 [online]). The PPV for screening assessment without DBT was 24.9% (142 of 571).

Use of DBT in screening assessment of the same group of women would have resulted in only 298 biopsies because of a mammographic (244 of 298 biopsies) or US (54 of 298 biopsies) abnormality, with 142 cancers identified. Use of DBT in screening assessment resulted in avoidance of 273 biopsies (56 asymmetries, 181 masses, 32 calcifications, and four architectural distortions) and led to a higher PPV of 47.6% (142 of 298).

The specificity significantly differed between the two approaches, with DBT assessment achieving higher specificity (77.5% [538 of 694]) than FFDM assessment (38.2% [265 of 694]) (Table 2). A total of 144 cancers were identified with both techniques. A total of 142 cancers were identified with FFDM (the two missed cancers were detected with DBT), and 142 cancers were identified with DBT (the two missed cancers were identified with FFDM). A further cancer was occult at FFDM and DBT and was identified only with MRI surveillance.

DBT would have generated 298 biopsies in total, of which 142 would have yielded malignant findings. Four biopsies were prompted by DBT alone as a result of the study protocol; the imaging findings had been considered benign at FFDM. Of the four biopsies performed, two yielded malignant findings, and two yielded benign findings. Of the 145 cancers, 92 were invasive and 53 were ductal carcinoma in situ (DCIS) (Table 3).

Of the 92 invasive cancers, 69 were ductal (grade 1, n = 23; grade 2, n = 27; grade 3, n = 19); nine were invasive lobular or pleomorphic invasive lobular (all grade 2), six were mixed ductal lobular (all grade 2), and eight were classified as “other” (mucinous, tubular, papillary, or metaplastic). Of the 53 cases of DCIS, 44 were high-grade DCIS and nine were intermediate-grade DCIS.

DBT and Mammographic Features and Sensitivity in Screening Assessment

DBT and FFDM assessments were equivalent in terms of their sensitivity overall (Table 2) and across the range of mammographic abnormalities present within the malignant cases in the study (10 with asymmetries, 49 with calcifications, 66 with masses, seven with architectural distortion, and 12 with multiple feature types in the same image) (Table 4).

DBT and Mammographic Feature and Specificity

DBT assessments exhibited significantly higher specificity (77.5% [538 of 694]) than did FFDM assessments (265 of 694 [38.2%]) in terms of specificity overall (Table 2) and across the range of mammographic abnormalities present within the nonmalignant cases (239 with asymmetries, 98 with calcifications, 271 with masses, 29 with architectural distortion, and 23 with a mammographic abnormality with multiple feature types) (Table E2 [online]).

Discussion

Our study showed that digital breast tomosynthesis (DBT) in the diagnostic work-up of mammographic abnormalities would have reduced the number of benign biopsies performed at screening assessment. The number of biopsies performed would have been reduced from 69.0% (571 of 827) to 36.0% (298 of 827). In 45 cases, DBT findings were reported as benign; however, because of an abnormal US finding, biopsy was performed as part of the triple assessment. A total of 142 cancers were detected with DBT, and one cancer that was considered benign at DBT was considered indeterminate at US, thus prompting US-guided core biopsy. Importantly, all invasive cancers were identified with DBT.

DBT allows for improved reader accuracy and confidence in determining if a mammographic abnormality is concerning for cancer or not, leading to a reduction in the number of biopsies performed. Our study demonstrates that DBT showed higher specificity across a range of feature types, especially mass lesions and asymmetric densities, which is consistent with the current literature (4,12,13). DBT also improved the specificity of microcalcifications when compared with FFDM. When DBT was used for assessment, it showed a higher sensitivity for mass lesions and asymmetric densities. Our results indicate 273 biopsies could have been avoided if DBT had been used. This suggests that implementing DBT along with triple assessment could potentially reduce the number of false-positive findings for those abnormalities.

There is growing evidence regarding the role of DBT as a screening test. Friedwald et al (1) published a study that looked at the effect of DBT in breast screening programs within the United States. They showed that positive predictive values went from 4.3% to 6.4% (difference, 2.1%; 95% confidence interval: 1.7, 2.5; P < .001) for recall and from 24.2% to 29.2% (difference, 5.0%; 95% confidence interval: 3.0, 7.0; P < .001) for biopsy when DBT was included alongside FFDM. These findings are further supported by the results of two prospective European studies (16,17). Skaane et al (16) reported a 40% improvement in the detection of invasive cancers (an additional 25 invasive cancers were detected) and a 15% reduction in the number of false-positive results (before arbitration). The false-positive rates were 61.1 per 1000 examinations with FFDM alone and 53.1 per 1000 examinations with FFDM and DBT in a study of 12 621 screening examinations with FFDM and tomosynthesis compared with FFDM alone.

This is one of a few studies to specifically look at the role of DBT in reducing benign biopsy rate at diagnostic work-up after abnormal findings were seen on a screening mammogram (18–22). Several studies have looked at the role of DBT in the diagnostic work-up. These studies have mainly focused on comparing FFDM with DBT in interpreting lesions recalled after abnormal findings were seen at screening mammography. The largest prospective study to look at the role of DBT in diagnostic work-up compared DBT with FFDM and screen-film mammography (12). In that study, 738 women were included, and assessment determined that 204 (26.8%) had a malignant diagnosis (147 invasive cancers, 57 in situ cancers), 286 (37.7%) had a benign diagnosis, and 269 (35.4%) had normal findings. The improvement in diagnostic accuracy with the addition of DBT was for soft-tissue lesions. Michell et al concluded that DBT improved reader accuracy in determining if a mammographic abnormality was malignant or benign or if findings were normal, thereby reducing the need for additional mammographic views. Bernardi et al (13) performed a prospective simulation study where if DBT had been used as a screening tool then 74.4% of women (102 of 137) would not have been recalled, as all these women were false-positive recalls with either benign or normal results. DBT was correct in advising assessment of the 21 patients with breast cancer (no false-negative findings). In our study, DBT as a screening tool would have reduced the number of women recalled for diagnostic work-up by 71.4%. In doing so, we would have missed two of 145 cancers that were screen detected with FFDM alone, but we would have identified two of 145 cancers that were not detected with FFDM. All four were cases of noninvasive cancer.

Studies have shown that women who underwent percutaneous needle biopsy at second-stage screening were less likely to attend the subsequent screening round (23). In addition, the longer the period between assessment at second-stage screening and final confirmation of a nonmalignant diagnosis, the less likely women were to return for the next screening round (24). This finding is probably related to the longer associated period of anxiety. It is often stated that it is the not knowing that women find most difficult.

Our study had some limitations. First, DBT images were reviewed at a separate workstation, without any prior digital mammography studies being available. Previous images might have reduced the recall rate even further. Second, the radiologists reading the studies specialized in breast imaging and received additional training in the interpretation of DBT images; however, all four readers had less than 1 year of experience reading DBT images. Third, only 835 of 1470 women consented to take part in the study, which introduced bias in the study results.

In conclusion, DBT has an important role in screening assessment. The key benefit is the reduction in biopsy rate from 69.0% (571 of 827) to 36.0% (298 of 827), without any reduction in the cancer detection rate.

Disclosures of Conflicts of Interest: N.S. disclosed no relevant relationships. M.M. disclosed no relevant relationships. I.H. disclosed no relevant relationships. Y.C. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: institution received a research grant from Hologic. Other relationships: disclosed no relevant relationships. B.J.G.D. disclosed no relevant relationships.

Author Contributions

Author contributions: Guarantors of integrity of entire study, N.S., B.J.G.D.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, N.S., I.H.; clinical studies, I.H., B.J.G.D.; statistical analysis, I.H., Y.C.; and manuscript editing, all authors.