Diagnostic accuracy of tomosynthesis-guided vacuum assisted breast biopsy of ultrasound occult lesions

This study aims to evaluate the diagnostic accuracy of digital breast tomosynthesis-guided vacuum assisted breast biopsy (DBT-VABB) of screening detected suspicious mammographic abnormalities comprising of calcifications, asymmetric densities, architectural distortions and spiculated masses. In this institutionally approved study, a total of 170 (n = 170) DBT-VABB were performed, 153 (90%) were for calcifications, 8 (4.7%) for spiculated mass, 5 (2.9%) for asymmetric density and 4 (2.4%) for architectural distortion. All these lesions were not detected on the corresponding ultrasound. Histopathology results revealed 140 (82.4%) benign, 9 (5.3%) borderline and 21 (12.4%) malignant lesions. The total upgrade rate at surgery was 40% for atypical ductal hyperplasia and 5.9% for ductal carcinoma in-situ. 3.6% discordant benign lesions showed no upgrade. DBT-VABB showed 100% specificity, 91.3% sensitivity and 100% positive predictive value (PPV) for detecting malignant lesions. The negative predictive value (NPV) was 80%. 2 (1.2%) patients had mild complications and 1 (0.6%) had severe pain. Our study showed that DBT-VABB was a safe and reliable method, with high sensitivity, specificity, PPV and NPV in the diagnosis of non-palpable benign and malignant breast lesions. Our data also confirmed the accuracy of DBT-VABB in detecting malignant lesions and we suggest further surgical excision in borderline lesions for a more accurate diagnostic evaluation.

www.nature.com/scientificreports/ VABB is a minimally invasive procedure that is able to obtain larger samples without repositioning or reinsertion. It has been demonstrated to be a safe alternative to open surgical biopsy and shows numerous advantages over surgical excision including desirable cosmetic results, fewer complications, less psychological stress, more cost-effective and less hospital stay 11,12 . It can be done either under stereotactic, ultrasound or MRI guidance. Diagnostic underestimation of DCIS and borderline lesions, especially atypical ductal hyperplasia (ADH), are not uncommon in stereotactic guidance percutaneous biopsy. A wide range of percentages have been reported and the majority of published literature concluded that CNB has a higher rate of diagnostic underestimation compared to VABB [13][14][15] .
In Malaysia, DBT-guided stereotactic VABB (DBT-VABB) is not widely used for diagnostic purposes as it is not readily available in all hospitals. It is currently mostly being utilized in tertiary referral hospitals with breast surgery and oncology subspecialties. In this study, we present our experience of utilizing DBT-VABB and evaluate the diagnostic accuracy of this method in non-palpable mammographic breast lesions.

Methodology
Study design. This was a single-center retrospective cross-sectional study, involving all patients with BI-RADS category 4 and 5 mammographic lesions who underwent DBT-VABB from 1st March 2017 to 31st July 2019. These lesions were detected on routine screening mammogram, occult on US and include calcifications, asymmetric densities, spiculated masses and architectural distortions. The study was performed in adherence with the approved guidelines from Medical Ethics Committee of University Malaya Medical Centre (MECID No: 2018724-6517). All experimental protocols were approved by Medical Ethics Committee of University Malaya Medical Centre and all subjects provided written informed consent. Information on the location, morphology and distribution of lesions and patients' clinical data including age, risk factor, number of samples, residual lesion, tissue marker insertion, complications, HPE findings and any subsequent surgery were collected. Pain scoring assessment post-VABB was also recorded. Patients with confirmed breast cancer on the contralateral breast were excluded.

Equipment and technique.
All patients were examined on DBT (Selenia Dimensin Hologic, Bedford, Massachusetts). Patients subsequently had supplementary ultrasound evaluation of the breasts and both axilla using diagnostic B-mode greyscale and colour, medical grade US system (Philips iU22; Philips Healthcare, Bothell, WA, USA) with a high frequency (i.e. 12.5 MHz) linear transducer probe. Ultrasound scans were performed by final year radiology trainee and the findings were verified by radiologists in-charge of the posting. DBT-VABB were performed by consultant radiologists with more than 10 years experience using the Mammotome system (Devicor Medical Products, Leica Biosystems). Few cases were done using EnCore Enspire system (Bard BD). 10G stereotactic biopsy needles were used for both systems.
The biopsy procedure was done in the upright position. Scout images were taken to confirm the presence of lesion within the field of projection and determined the puncture target. In patients with diffuse and regional calcifications, puncture target was made at the site with the densest distribution of calcifications. Pre-fire and post-fire images were taken before DBT-VABB was performed. For calcifications, specimen radiograph was performed to confirm the presence of calcifications. Post-procedure mammogram was also performed to confirm on the complete or partial removal of lesion.
Data collection and analysis. Data were retrieved from local picture archiving and communication system (PACS) and two readers (consensual method), blinded to the pathology results, categorised the findings according to ACR-BIRADS 5th edition (2013) mammogram lexicon; WYC and MTRH (with 5 and 10 years of mammography experience). HPE results were classified as benign, borderline or malignant (DCIS + invasive carcinoma). Final surgical pathology were also documented for patients who underwent subsequent surgical excision. Post-VABB complications were documented. Patients' pain were scored using Numeric Rating Scale (0-10) and categorised as none (0), mild (1-3), moderate (4-6), severe (7-9) and worst pain (10).

Statistical analysis.
Statistical analysis was performed with a statistic software, using Statistical Package for the Social Sciences (SPSS) for Windows Version 23.0 (IBM corporation, Armonk, NY). For quantitative data, range and median were reported. Descriptive statistics were used and data were presented as frequency (n) and per cent distribution (%). Diagnostic underestimation was calculated by dividing the number of upgrade in surgical HPE by the number of ADH or DCIS in VABB, respectively. Differences between proportions were tested using a Pearson's chi-square test. Fisher exact tests were used when any assumptions of Pearson's chi-square test were violated. A p-value of < 0.05 was applied as a threshold for statistical differences. Figure 1 outline the flow of the lesion from the initial detection to surgical management or follow-up. All carcinoma in-situ and invasive cancer will proceed to surgery. All benign and borderline lesions will be discusssed in multidisciplinary team (MDT) meeting and radiology-pathology discordant lesions will proceed to surgery.
Informed consent. Written informed consent was obtained from the patients.
Tissue marker clip was inserted in all lesions with clip migration occurring in 11.8% of patients. The distance of clip migration from the biopsy site ranged from 0.4 cm to 4 cm.
Overall, only 2 patients developed a small haematoma confined to the biopsy site, measuring less than 2 cm. No severe complication was recorded.
For pain assessment, most patients (n = 94, 55.4%) only experienced mild pain. 1 (0.6%) patient had severe pain and this was associated with high number of specimens sampled, n = 24 and the resultant development of a small haematoma. These data were presented in Table 1.
Comparison of VABB HPE with mammographic findings. VABB HPE results in relation to mammographic findings were presented in Table 3. The majority of VABB patients were in BI-RADS 4a category. All BI-RADS 5 lesions were malignant. Fisher Exact test demonstrated statistically significant differences in VABB HPE among BI-RADS category (p < 0.01). This study demonstrated that positive predictive value (PPV) in detecting malignant lesion increases in tandem with BI-RADS category.
With regards to morphology, 87.6% calcifications and all asymmetric densities were benign. Malignancy were seen in 1 of 4 patients with architectural distortion and 1 of 8 patients with spiculated mass. Figure 2 demonstrates morphology of mass-like lesions on mammogram. www.nature.com/scientificreports/ For calcifications' morphology, the probability of malignancy (PPV) was higher in fine linear (25%). For calcifications' distribution, the probability of malignancy was 26.7% in linear category and 15.8% in segmental category. Figure 3 showed the examples of benign calcifications and Fig. 4 showed malignant calcifications.  Table 4). Underestimation of VABB was when there was an upgrade of benign to malignant, in-situ malignancy to invasive carcinoma and borderline lesions to in-situ malignancy or invasive carcinoma. Any upgrade from low-grade DCIS to moderate or high-grade DCIS and from moderate-grade DCIS to high-grade DCIS was also considered as an upgrade. In our study, diagnostic underestimation cases were only observed in ADH and high-grade DCIS. There were 5 (3.6%) benign discordant lesions with no change in HPE after surgery. 6 out of 9 borderline lesions proceeded with surgery. 2 ADH were upgraded to DCIS representing 33.3% of diagnostic underestimation.The images of these cases were shown in Figs. 5 and 6. The remaining 3 borderline lesions did not proceed with surgery. 1 patient with FEA was lost to follow-up, 1 patient with ADH was not fit for surgery due Atypical ductal hyperplasia** 7 Flat epithelial atypia** 1 LCIS** 1 Table 3. VABB HPE with mammographic findings including BI-RADS category, lesions morphology and distribution and morphology of calcifications. * p-value of < 0.05 was applied as a threshold for statistical differences.  In the malignant group, only one high-grade DCIS was upgraded to invasive lobular carcinoma at surgery representing 5.9% of diagnostic underestimation. The mammographic findings of this patient were illustrated in Fig. 7.

Discussion
Non-palpable mammographic lesions are one of the major challenges in diagnosis of breast cancer. Calcifications are the most frequent mammographic finding for early diagnosis of breast cancer 7,10-12 . In our study, we found that the distribution of malignancies in BI-RADS category is similar with the standard guideline by American College of Radiology 16 . Our results corroborate with the findings of other studies which showed that the likelihood of malignancy increases with BI-RADS category 7,11,12,17,18 . Although calcifications play a crucial role in early breast cancer diagnosis, they are also associated with 63.2% of benign pathology 7 . The prediction of benign or malignant calcifications are possible by following ACR BI-RADS mammography lexicon guidelines. Previous study by Jun Liu et al. showed that the probability of malignancy in calcifications' morphology was higher in fine pleomorphic (44.7%) while with respect to distribution, the highest percentage were in segmental distribution (78%) and 68% in linear distribution 7 . These findings were similar to our study.
Majority of previous literatures confirmed that VABB is a safe procedure, highly accurate technique and comparable to surgical biopsy for histopathological diagnosis in non-palpable breast lesions. A study performed by Penco et al. reported that the sensitivity of VABB ranged from 99.7% to 100% and false negative rate of 1.7% to 7.1% in 4086 patients within 10 years duration 19 . Tsai et al. reported that stereotactic VABB has a sensitivity of 95.24%, false-negative rate of 4.76% and NPV of 99.61% in 817 patients within 5 years duration 11 . Similarly, Kettritz et al.achieved sensitivity and NPV of more than 99% 17 . We postulate that our study produced a higher false-negative rate of 20% due to a smaller sample size (2/10) as compared to other studies. However, in line with previous literatures, we found that VABB shows high specificity (100.0%) and sensitivity (91.3%). We also proved that VABB has high PPV (100.00%) for malignancy and high NPV (81.8%) for benignity. This study also demonstrates that DBT-guided VABB is safe and reliable despite not having prior experience with the procedure.
As it is generally known, diagnostic underestimation and false-negative rate is higher in core needle biopsy (CNB) compared to VABB. According to previous literature, range of diagnostic underestimation by CNB for DCIS was from 27 to 59% and approximately 11.5% to 88% for ADH 13,14 . Unfortunately, stereotactic core needle biopsy is still the main choice for obtaining histological diagnosis in majority of hospitals in Malaysia due to cost. However, in the long run, VABB will reduce the need for surgery and reduce the incidence of delayed in diagnosis of cancer due to underestimation, which will then further burden the healthcare system. www.nature.com/scientificreports/ We found that the DCIS and ADH underestimation rate of our VABB were in line with previous studies. In a previous study by Tsai et al. 11 , the DCIS underestimation rate was 16.7% while another study by Kettritz et al. 17 revealed that underestimation of ADH and DCIS were 24% and 12%, respectively. Another recent study by Badan et al. reported the underestimation rate of ADH was 25% and DCIS was 14.28% 13 . According to Inyoung Youn et al., the underestimation rate of ADH was 33.3% on VABB although all microcalcifications were completely removed 20 . Our data confirmed the accuracy of VABB in detecting malignant lesions and we suggest for further surgical excision in borderline lesions for a more accurate diagnostic evaluation. This has become a normal practice in our centre. However, according to the international consensus for borderline lesions, surveillance instead of surgical excision is possible in older age groups since most of the invasive cancers that develop after ADH progress slowly 21 .
All benign lesions in our study were mostly followed-up until 48 months and were found to be true negative. This can obviate unnecessary operations for benign lesions diagnosed by VABB. The diagnostic accuracy does not depend on obtaining a large number of specimens, instead an adequate number of correctly targeted www.nature.com/scientificreports/ specimens is essential. Previous studies showed that even with standardized retrieval of 20 specimens per lesion, underestimation can still occur 22 . In our study, the number of specimens ranged from 5 to 27 specimens with mean of 6.8. Similarly, the mean number of core was 8.5 in previous study by Esen Gul et al. but they have proven that although the average core number was low, their false negative rate was 0% with the total excision rate of almost 44% 22 . www.nature.com/scientificreports/ The complication rate of stereotactic VABB is low, approximately 3.7% 18 . The commonest complication of stereotactic VABB was haematoma which was well-controlled by manual compression and none of the previous study had reported patient requiring any surgical intervention for management of the complications 7,12,18,23 . In our study, 98.8% of our patients had no complications and 1.2% have a small haematoma of less than 2 cm.
VABB is a well-tolerated procedure in our study. The majority of our patients only experienced mild pain (55.4%) and 25.9% of them did not experienced any pain during the procedure. Only 0.6% of the patients in our study complained of severe pain which was associated with larger sample size. This finding is in line with previous study conducted by Seely et al. which revealed an average pain score of 3.1 with stereotactic VABB which is in mild category 24 .

Limitations
There are several limitations in our study. Firstly, due to the nature of the retrospective database at a single institution, some useful patient data were not well recorded. Our study has small sample size and may be insufficient in providing an accurate overall picture. Further clinical studies with multicenter collaborations may provide more accurate conclusion. Secondly, the benign lesions underestimation was mainly evaluated based on clinical followup at a maximum of 2 years duration in our study. A long-term follow-up is necessary for more accurate results.

Conclusion
Our study has shown that DBT-VABB is reliable in the diagnosis of non-palpable breast lesions. DBT_VABB is also a tolerable procedure without major complication. It has high sensitivity, specificity, PPV and NPV in detecting malignant as well as benign lesions. Benign lesions diagnosed by VABB can be safely followed up without any surgical interventions and malignant diagnosis on HPE should undergo definitive surgical treatment. Subsequent surgical excision for all borderline lesion, especially ADH should be recommended regardless of complete removal of calcifications in VABB. Our study also confirmed that BI-RADS category assessment is effective to anticipate the probability of malignancy.
The promising results in our study supported the implementation of VABB as a standard tool for diagnosis of non-palpable mammographic lesions in all tertiary referral hospital.  www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.