Main points

• High-risk breast lesions are a heteroge­neous group of proliferative lesions that are precursors to breast carcinogenesis and are associated with a higher risk of future breast cancer.

• Clinical management of high-risk lesions using image-guided biopsy is challenging.

• High-risk breast lesions should be managed in a case-based manner after discussion among a multidisciplinary team.

High-risk breast lesions comprise a heterogeneous group of proliferative lesions that are precursors of breast carcinogenesis and are associated with a higher risk of future breast cancer development.^1,2 These lesions include atypical ductal hyperplasia (ADH), lobular neoplasia (LN) (a term encompassing both atypical lobular hyperplasia [(ALH) and lobular carcinoma in situ (LCIS)], intraductal papilloma (IP) with/without atypia, flat epithelial atypia (FEA), and radial scars (RSs)/complex sclerosing lesions.^3,4 Other terms used in the literature to describe these entities are “lesions of the breast with uncertain malignant potential”, “borderline lesions”, or “B3 lesions”. High-risk lesions are commonly detected due to the increased use of core or vacuum biopsy techniques for screen-detected lesions. High-risk lesions are found in about 3%–9% of cases of percutaneous image-guided breast biopsies performed following a suspicious imaging finding.^5,6

Percutaneous image-guided needle biopsy has become a standard approach for the tissue diagnosis of suspicious breast lesions. It is performed using either the core needle biopsy (CNB) or vacuum-assisted biopsy (VAB) technique. CNBs are usually performed under ultrasound (US) guidance, while stereotactic system or magnetic resonance imaging (MRI) guidance is used for VABs. When the CNB/VAB detects a high-risk lesion, the possibility of missing the malignant component of the lesion exists; ductal carcinoma in situ (DCIS) and invasive carcinoma can only be detected when surgical excision is performed. The positive predictive value (PPV) for malignancy for CNB/VAB is about 10%–30%.^5,7,8

The appropriate management of a high-risk lesion following diagnosis using image-guided biopsy is controversial, and recommendations, including surgical excision and follow-up, vary for different lesions.⁹ Surgical excision has been used as a general approach to avoid missing an underlying malignancy.¹⁰ However, in recent studies, risk parameters have been identified to upgrade a lesion to suspected malignancy, allowing more conservative approaches in selected cases. To achieve consensus about managing high-risk lesions, the International Consensus Conference was held in Zurich, Switzerland, in 2016 and 2019. This conference introduced second-line VAB as an alternative to open surgical excision in most lesions, and several guidelines were introduced for managing these lesions.¹¹

This multicenter study aimed to document the excisional biopsy or follow-up results of high-risk lesions diagnosed on image-guided CNB/VAB and evaluate the clinical, imaging, and histologic features for associated malignancy risk. The possibility of an upgrade related to histologic subtype, tissue sampling, and other variables was also evaluated.

Methods

This retrospective multicenter study included 1.343 patients from 30 centers diagnosed with a high-risk lesion on image-guided CNB or VAB. The ethics committee of the Ege University Faculty of Medicine approved the study (approval number: 20-6T/41; date of approval: June 10, 2020). The study reviewed existing data, so patient consent was not required.

Patients

Radiology records for the 12 years between 2008 and 2020 were reviewed for all image-guided biopsies and pathology reports. Patients diagnosed with ADH, LN (ALH/LCIS), papilloma (with or without atypia), RSs, or FEA on image-guided CNB/VAB were included. The needle biopsy could be performed using either US or stereotactic guidance and tru-cut or vacuum biopsy needles. In addition, patients managed with an excisional biopsy or those with at least one year of documented radiological follow-up after diagnosis with one of the above-mentioned high-risk lesions were included in the study. FEA, seen in conjunction with LN or ADH, was categorized under either LN or ADH as appropriate. Pleomorphic LCIS, fibroepithelial lesions, and mucocele-like tumors were excluded. Further, lesions associated with in situ or invasive carcinoma on CNB were excluded.

All patients were discussed at a multidisciplinary meeting, and decisions were made based on biopsy and radiology findings. Histopathologic diagnoses were made according to the current guidelines.¹²

Data analysis

The images used in this study were retrieved from the Picture Archiving and Communication System digital archive, and the results of mammograms, US images, and MRI scans were re-evaluated. The findings were categorized by imaging method, as follows: microcalcifications, a mass/nodular opacity, or suspicious non-mass findings (i.e., architectural distortion and asymmetry) on a mammogram; a mass/nodular lesion or non-mass lesion (i.e., distortion or echogenicity changes) on a US examination; and a mass or non-mass contrast-enhancing lesion on an MRI scan. The patients without imaging findings were also noted. In each case, the lesion’s largest diameter was recorded and categorized into two groups: under and above 15 mm. The final Breast Imaging Reporting and Data System (BI-RADS) category on the imaging reports was also documented.

Other data collected from the records included the needle biopsy type and biopsy sampling method. The imaging-guidance modality was chosen based on the findings and visibility of the lesion on imaging. Stereotactic guidance was usually preferred for microcalcifications, while US was used for lesions that were visible on the US. Stereotactically guided biopsies were performed in 182/1.343 (13.6%) patients on a prone table or an add-on unit using a 9 to 12 G directional vacuum-assisted needle. US-guided biopsies were performed in 1.161/1.343 (86.4%) patients with an automated tru-cut system using 14 to 16 G needles with a 2-cm cutting surface. The number of cores used was another analysis point, and this information was categorized as n <4 and n ≥4.

The final pathology reports of the patients who had surgical excision or were stable for at least one year of follow-up were documented. The follow-up period varied between 12 and 180 months (median: 30 months). In total, 929 patients underwent surgical excisional biopsy, and 414 were followed up with at six-month intervals. Recommendations for excision or surveillance were made in a multidisciplinary meeting. The pathological results of excisional biopsies were recorded as either no change in the primary diagnosis by needle biopsy or upgraded to indicate malignancy. The presence of DCIS or invasive carcinoma on histologic examination after an excisional biopsy was regarded as necessitating an upgrade of the high-risk lesion. In follow-up patients, the diagnosis on needle biopsy was considered a compatible result if there was no change in the findings; however, a biopsy was recommended if any suspicious change was observed on follow-up.

The diagnosis on CNB/VAB and the outcomes were compared, and the upgrade rate and PPV for malignancy [i.e., (number of malignant cases/total number of participants) × 100] were calculated. The association between the parameters described above and the upgrade rate or PPV for malignancy were evaluated.

Statistical analysis

All patient information was anonymously submitted to a medium in an anonymized manner via software from all included centers. Patient characteristics were reported as frequencies and percentages (%) for categorical variables, and descriptive statistics (mean, median, minimum, and maximum values) were calculated for continuous variables. If the variables had a normal distribution, the mean and standard deviation (SD) were given; otherwise, the median and range were given. Pearson’s chi-squared test, the Fisher–Freeman–Halton test, and Fisher’s exact test were used to analyze the categorical variables in groups. A value of P < 0.05 was considered statistically significant. The data were analyzed using the IBM SPSS Statistics version 25.0 statistical software package.

Results

In total, 1.343 patients met our criteria, of which 375 (27.9%) had ADH, 55 (4.1%) had LN (ALH/LCIS), 611 (45.5%) had IP without atypia, 123 (9.2%) had IP with atypia, 114 (8.5%) had FEA, and 65 (4.8%) had RSs. The patients were between 17 and 86 years of age, with a mean age of 47.45 years (SD: 0.459).

In all cases, the upgrade rate to malignancy was 20.6% (Table 1). Of these, 52% (144/277) were cases of ADH. Almost half (47.2%) of all upgrades were to invasive cancer. Upgrades to low-grade DCIS were more common than upgrades to high-grade DCIS (32.9% vs. 19.9%, respectively). According to the pathologic subtypes, IP with atypia was the most common type, with an upgrade rate of 44.7% (55 of 123 cases), followed by ADH (38.4%, 144 of 375 patients). The upgrade rate was 12.7% in LN, 8.7% in FEA, 9.4% in IP without atypia, and 4.6% in RSs. The pathological subtype had a statistically significant relationship with malignant vs. benign diagnosis P < 0.001).

In about half the cases (49%), the lesions were categorized as BI-RADS 4A lesions, followed by BI-RADS 4B (26.7%), BI-RADS 4C (13.1%), and BI-RADS 5 (3.3%). These categories were most common in the subtypes ADH and IP with atypia. A biopsy was still recommended in 7.9% of patients, although the lesions were categorized as BI-RADS 3. The malignancy upgrade rate had a statistically significant relationship with the BI-RADS category (P < 0.05). In the imaging findings, microcalcifications on mammograms were common in the LN, FEA, and ADH subtypes (70.3%, 64%, and 54.8%, respectively). The mass lesions detected on mammography, US, or MRI were most frequently IPs with or without atypia. Lesions presenting as non-mass lesions on mammography, US, or MRI were seen in all subtypes. There was no statistically significant relationship between the imaging findings and upgrade rates.

Table 2 summarizes the outcomes of biopsy or follow-up and malignancy upgrade rate according to the BI-RADS category, needle size, number of samples, and lesion size. There was no statistically significant relationship between malignancy upgrade rate and needle size (P > 0.05). However, the relationship between the malignancy upgrade rate and the number of samples and lesion size was statistically significant (P = 0.008 and P < 0.001, respectively). Lesion sizes were between 4 and 135 mm (mean: 13.8; median: 10), which was the most predictive factor for an upgrade when all the patients were analyzed together.

The results varied among the lesion subtypes. In the ADH group, a statistically significant change in the upgrade rate was recorded according to BI-RADS category, needle thickness, number of samples, and lesion diameter (Table 3). The malignancy rate was higher in the BI-RADS 4C and 5-category lesions. The upgrade rate varied according to needle type, 41% vs. 28% for tru-cut vs. vacuum biopsies. The number of samples and lesion size correlated with upgrade rates, being 49.0% when there were fewer than four samples vs. 34.3% when there were four samples or more, and 29.1% vs. 51.6% for lesions ≤15 mm and >15 mm, respectively. In the multivariate analysis, the lesion diameter and needle size were the most predictive of a lower upgrade rate (20% when 9 to 12 G needles were used for lesions ≤15 mm in size) (Table 4).

The only statistically significant variable in the LN subtype was the BI-RADS category. There was no statistically significant correlation between the needle thickness, number of samples, lesion size, and upgrade at the final diagnosis. In the multivariate analysis, the results did not predict malignancy when larger needle sizes were used in lesions ≤15 mm (Table 4).

In the IP without atypia and FEA subtypes, the upgrade rate changed with the lesion diameter and BI-RADS category (Table 3). In the multivariate analysis, the patients who had IP without atypia, diagnosed using a 9 to 12 G needle and a lesion ≤15 mm, were all in the benign group (n = 34; Table 4). In IP without atypia, the upgrade rate was 6% in patients with lesions ≤15 mm, in whom 14 to 16 G needles were used for sampling, vs. 18% in patients with lesions >15 mm, in whom samples were taken using the same needle size. The difference between these results was statistically significant. The percentage of upgrade to malignancy did not change even when 9 to 12 G needles were used in lesions >15 mm (18%). In the FEA subtype, there was no malignancy at final diagnosis when 14 to 16 G needles were used to sample lesions ≤15 mm in 44 patients; however, the upgrade rate was 20% with the same needle size when the lesion size was >15 mm. The upgrade rate also varied with lesion size when 9 to 12 G needles were used (4% for lesions ≤15 mm, and 17% for lesions >15 mm) (P = 0.002; Table 4).

In patients with atypical IP, a statistically significant difference was found between lesion size and an upgrade to malignancy (Table 3). Of the 123 atypical IP lesions, 121 were sampled with 14 to 16 G needles, and the malignancy rate was 38% for lesions ≤15 mm in size and 57% for those >15 mm, which was statistically significant (P = 0.043; Table 4).

In the RS subtype, a statistically significant difference was found between needle size and upgrade to malignancy (P = 0.038; Table 3). Fourteen to 16 G needles were used for all these lesions, and the 42 lesions that were ≤15 mm in size were all benign at the final diagnosis. The other predictive factor for the upgrade was lesion size, with only one malignant of the 46 RS lesions ≤15 mm (Table 4).

Discussion

This multicenter study reviewed the outcomes of surgical biopsies and long-term follow-up visits of high-risk lesions (B3 lesions) diagnosed using a CNB. More than two-thirds of the patients (n = 929) underwent a surgical biopsy, and 20.6% were found to have a breast malignancy. At final diagnosis, more than half of all malignancies were DCIS (52.8%), and low-grade DCIS was more common than high-grade DCIS (32.9% vs. 19.9%, respectively). In the literature, the PPVs ranged from 9.9% to 35.1% when all subtypes of B3 lesions were included.^7,13,14 Bianchi et al.⁷ reviewed 3.107 cases and reported a 21.2% upgrade rate, similar to this study’s results. In addition, the number of DCIS cases was higher than that of invasive cancers in studies by Houssami et al.¹⁴ and Strachan et al.¹⁵

Several variables should be considered when deciding what the next steps should be following a diagnosis of a high-risk lesion of the breast through an image-guided needle biopsy. In this study, upgrades to malignancy were associated with the BI-RADS category, the number of samples taken, and the lesion size. The lesion size was the most predictive factor for an upgrade, the rate being 14% vs. 33% for tumors that were ≤15 mm vs. those that were >15 mm in size, respectively. The underestimation rate was reduced with more sampling, 18.7% for four samples or more and 25% for fewer than four samples. In all lesions, 16.4% were categorized as BI-RADS 4C or 5, and the malignancy rate in this group was 47.2% vs. 15.5% for BI-RADS 3 or 4A–B lesions. These results show the importance of the BI-RADS classification and the radiologic–pathologic concordance of the lesions. There was no statistically significant relationship between radiologic findings (i.e., calcification and mass or non-mass lesion) and upgrade rate.

In managing high-risk lesions, surveys have shown significant variation in the recommendations of radiologists, pathologists, and surgeons.^9,16,17 Surgical excision has traditionally been performed as a safe option to exclude any associated adjacent malignancy that could have been missed when performing a CNB on high-risk lesions. However, high-risk breast lesions comprise a variety of lesion subtypes, showing different radiologic and histologic features and levels of malignancy risk.¹¹ Malignancy diagnosed at surgical excision is more frequent in lesions with atypia than in those without. In the present study, an associated DCIS or invasive cancer malignancy was most commonly seen in cases of atypical IP (upgrade rate of 44.7%), followed by ADH (38.4%). Accordingly, this study’s results indicate that the surgical excision of these two categories of lesions is warranted. Strachan et al.¹⁵, Rakha et al.¹⁸, and de Beça et al.¹⁹ also reported that the underestimation of malignancy is much higher in lesions with atypia than in those without atypia. Moreover, distinguishing ADH from a low-grade DCIS through a pathology review can be difficult. As such, the management of these two lesion types was consistent, with excision still being recommended in the guidelines for both. However, alternatives such as further sampling or surveillance can be considered for lesions without atypia.^11,20

The reported rates of underlying co-existing malignancy for ADH diagnosed by needle biopsy varied between 4% and 54%, with a pooled median diagnosis upgrade rate of 25%.^1,21 There have been efforts to identify indicators of ADH lesions with a low risk of being upgraded to malignancy. Several histopathologic criteria, including the extent of the ADH and percentage of lesion removal, were found to be predictive factors of the upgrade rate.^22,23,24 While the present study did not evaluate any histopathologic parameters, the variables of biopsy type, needle size, number of samples, lesion size, and BI-RADS category of the lesion all showed a statistically significant correlation with the upgrade rate. ADH lesions ≤15 mm sampled with larger core needle sizes had a lower upgrade rate. However, this rate was too high to avoid excisional biopsy, which was done in 11 of 56 cases. Schiaffino et al.²⁵ proposed the conservative management of ADH only in a highly selective group of patients diagnosed using a stereotactic VAB for a single group of microcalcifications, without residual findings, and without a high percentage of hyperplasia at histological assessment.

In the current study, the upgrade rate within the subtype of LN was 12.7%. Pleomorphic types and variants of LCIS were excluded from this study, as there is already a consensus that excision is necessary for such types to ensure no underlying cancer is missed.²⁶ The BI-RADS score, lesion size, and needle type predicted an upgrade to LN carcinoma. Lesions measuring ≤15 mm and sampled with 9 to 12 G VAB needles were benign at surgical excision; however, the upgrade rate was 10% when 14 to 16 G needles were used for the same lesion sizes. The malignancy rate for lesions >15 mm also changed depending on the needle type, 40% for vacuum and 20% for core biopsy needles. Although the upgrade rate in LN has been reported to range between 0% and 50% in the literature.^4,27,28,29 Recent studies have shown that the upgrade rates decrease significantly when the BI-RADS score and pathologic results are concordant.^30,31,32,33 Mooney et al.⁶ reported a 5% upgrade rate upon excision for LN diagnosed incidentally vs. a 39% upgrade rate for targeted lesions. Therefore, routine excision is no longer required in all ALH or LCIS cases.³⁴

In the current study, the upgrade rate for IP with atypia was 44.7%, and for IP without atypia, 9.4%. A meta-analysis demonstrated a 15.7% pooled underestimation for non-malignant papillary breast lesions, with higher rates among atypical lesions (i.e., 36.9% for atypical lesions vs. 7% for benign IPs).³⁵ Therefore, there is no debate that surgical excision should be done after diagnosing atypical papillary breast lesions on core biopsy. However, there is no consensus on how best to manage benign IPs. In the current study, the lesion diameter and BI-RADS score correlated with an upgrade to malignancy in IPs without atypia. No malignancy was found in lesions ≤15 mm, sampled with larger needle sizes. When sampling lesions of the same size with 14 to 16 G needles, the upgrade rate was 6.1%. It has previously been demonstrated that lesions >15 mm in a peripheral location, with image-pathology discordance, are associated with a significant risk of upgrade to malignancy.³⁶

Recent studies have suggested that imaging follow-up may be reasonable in selected cases, including radiologically concordant or incidentally detected benign IPs of ≤15 mm diagnosed using large-gauge core biopsy needles.^37,38,39 Pareja et al.⁴⁰ found an upgrade rate of 2.3% in their evaluation of 171 radiologic–pathologic concordant IPs without atypia. In a study by Menes et al.#^*#ref41, upgrades to cancer occurred in 2% of asymptomatic women diagnosed with a benign papillary lesion using a needle biopsy following a mammogram showing a lesion classified as BI-RADS 4. From these results, imaging follow-up seems reasonable for benign papillomas found to be small upon core biopsy, adequately sampled, and radiologically concordant.

For the pure FEA subtype, the reported malignancy upgrade rates vary widely; most published studies have recommended excision. However, in several recent studies, imaging follow-up has been proposed for patients without residual calcifications.^42,43 In the present study, these upgrade rates were statistically significantly correlated with lesion diameter and BI-RADS score. The rate was 1.5% for lesions ≤15 mm. In relation to needle sizes, the upgrade rate was 0% (0/44 cases) for 14 to 16 G and 4% (1/23 cases) for 9 to 12 G needle sizes. A sampling error may have caused the only positive case in this study. In a recent systematic review and meta-analysis including 2.482 cases across 42 studies, this rate was 5%; however, when more than 90% of the calcifications were removed, no cancer was found at excision, and close imaging follow-up was recommended for such patients.⁴⁴ Schiaffino et al.⁴⁵ found a malignancy rate of less than 2% in patients diagnosed using VAB with no residual microcalcifications in concordant findings. The World Health Organization Working Group proposed observation as an acceptable management strategy for radiological–pathological correlated pure FEA.⁴⁶

The upgrade rate was 4.6% for RSs in this study. When the lesion was ≤15 mm, the rate of associated malignancy at surgical excision was 2.2% (1/46 cases). There are variable results in the literature, ranging between 0% and 40%.⁴⁷ In a meta-analysis of 49 studies, including 3.163 RS cases with surgical outcomes, the pooled upgrade rate was 7%; yet in the subtype assessed with an 8 to 11 G VAB needle and lacking atypia, this rate was 1%.⁴⁸ Li et al.⁴⁹ and Conlon et al.⁵⁰ found 0.9% and 2% upgrade rates in patients without atypia. Accordingly, imaging surveillance seems to be a reasonable option for selected patients.

The large sample size of high-risk lesions in each subtype and the multicenter design are major strengths of the current study. This makes the statistical analysis more valuable. However, this study has several limitations, including its retrospective design, which could have resulted in missing data, and the potential differences in the clinical practices used for selecting and managing patients in different centers. Last, this study did not look at long-term follow-up results for the participants.

In conclusion, high-risk lesions identified by needle biopsy do not follow similar patterns; routine excision is unnecessary for every lesion. This study’s upgrade rates to malignancy were related to the subtype, presence of atypia, and other variables, such as the BI-RADS score, lesion size, biopsy method used, and sampling adequacy. Because of these variables, there cannot be a general recommendation for all high-risk lesions of the breast. Clinical, radiologic, and pathologic features should all be reviewed before deciding whether surgical excision or close follow-up is most appropriate for a lesion. For ADH, although current guidelines recommend surveillance for small-volume lesions that are entirely removed through core biopsy, the recommendation remains typical management of surgical excision. IP with atypia also requires excision following CNB/VAB because of the high rate of associated malignancy. For LN, IP without atypia, pure FEA and RSs, underestimation rates were related to the BI-RADS score and, therefore, to radiology–pathology concordance, sampling adequacy, and lesion size. The upgrade rates increased with higher BI-RADS scores and lesion size in conjunction with insufficient tissue sampling. Boateng et al.¹⁰ reported lower rates when large core needles (i.e., 9 to 11 G) were used and higher rates when 14 G needles were used. Therefore, all cases should be managed case-wise after a multidisciplinary team discussion.

Conflict of interest disclosure

The authors declared no conflicts of interest.