B3 lesions represent a heterogeneous group of breast lesions diagnosed by interventional biopsy as CNB or VAB. Due to their uncertain malignant potential, there is a lack of conclusive recommendations after diagnosis of B3 lesions. Since different B3 lesions can be present in one interventional biopsy, we decided to group the different lesions. Therefore, we subsumed biopsies that had various B3 lesions containing cellular atypia as ADH, FEA, and LN1-2 were named as “Risk of associated invasive BC and DCIS”. CFL lesions were grouped as “Fibroepithelial lesions”, while papillomas and RS/CSL were combined as “Papillomas/RS”. We hypothesized, that the group “Risk of associated invasive BC and DCIS” had the highest malignant upgrade rate.

This study of 192 patients investigated the malignant upgrade rate of B3 lesions after open surgical excision in a German population. The highest rate of invasive BC and/or DCIS was observed in the group “Risk of associated invasive BC and DCIS” containing the B3 lesions FEA, ADH, and LN1-2, respectively, with 17.5%. By contrast, the upgrade rates in the "Fibroepithelial lesions" group were 0%, and for "Papillomas/RS", it was 4.3%. Overall, we found an upgrade rate of 10.9%.

The group “Risk of associated invasive BC and DCIS”—containing ADH, FEA, and LN1-2—with 103 lesions represented the largest subgroup (53.6%) in our database. Compared to the two groups “Fibroepithelial lesions” and “Papillomas/RS”, we found a significant higher malignant upgrade rate of 17.5% (95% CI 10.7–26.2%).

For the B3 lesions ADH, FEA, and LN1-2, malignant upgrade rates of 28%, 11%, and 17%, respectively, were reported in a recent meta-analysis [18]. However, for the different lesions a large range of upgrade rates are published. In another review and meta-analysis of 93 studies, a comparable upgrade rate of 29% (95% CI 26–32%) for ADH alone after surgical excision is reported. Nevertheless, upgrade rates differed also for the distinct biopsy methods, with pooled upgrade rates of 42% (95% CI 31–53%), 23% (95% CI 19–27%), 32% (95% CI 22–43%) for ultrasound guidance, stereotactic guidance, or MRI guidance, respectively [19]. FEA shows a malignant upgrade rate in other meta-analyses between 5% (95% CI 3–6%), 8.8%, and 11.1%, respectively [20,21,22]. Identically, malignant upgrade in LN1-2 is published between 17% (95% CI 13–21%) in a recent meta-analysis and 3.7—28% in smaller study populations [18, 23,24,25]. Our combined malignant upgrade rate of 17.5% (95% CI 10.7–26.2%) for ADH, FEA, and LN1-2 fits into the wide range of different upgrade rates with various different study populations, types of breast biopsies, and other numerous preconditions. In contrast to the prementioned meta-analysis, we report lower malignant upgrade rates for ADH (21.1%) and FEA (7.1%), but a higher upgrade rate for LN1-2 (66.7%) [18].

The second largest group in our data are papillomas and RS/CSL with 69 out of 192 lesions (35.9%). For this group, we report 4.3% malignant upgrade. A limitation of our work is that our data lack information regarding the presence or absence of cellular atypia in papillary lesions. Atypical papillary lesions having the highest risk of upgrading to malignant after surgical excision in a meta-analysis [26]. In another systematic review and meta-analysis, an upgrade rate of 12% (95% CI 10–15) for papillomas, while for RS/CSL, an upgrade rate of 8% (95% CI 6–11) was reported [18]. In contrast, we report an upgrade rate of merely 4.3% (95% CI 0.9–12.2). One explanation for our relatively low upgrade rate could be, that we excluded papillary lesions or RS/CSL containing larger cellular atypia from the group “Papillomas/RS” and put them into the group with ADH, FEA, and LN1-2, since cellular atypia is the clinically dominant lesion.

Only including papillary lesions without atypia the prementioned meta-analysis presents a upgrade rate of 7% (95% CI 4–10)[18]. In line with our results, others found also a lower proportion of malignancies of 3.4% in non-atypical papillary lesions after surgical excision [27].

B3 lesions with fibroepithelial lesions represent the smallest group in our dataset comprising 20 lesions (10.4%). In line with other studies, we found no upgrade of fibroepithelial lesions to invasive BC and DCIS [8]. However, the focus in B3 lesions containing fibroepithelial lesions is not to rule out primary BC but malignant phyllodes tumors. In a retrospective analysis of 51 B3 lesions diagnosed as phyllodes tumors, only five were malignant phyllodes tumors after open excision [28]. In our data set, we could not observe an upgrade to malignant phyllodes tumors in the 20 fibroepithelial lesions. Another review found an upgrade rate of less than 2% for fibroepithelial lesions [29]. In a cohort of 215 patients with fibroepithelial B3 lesions, the upgrade rate to borderline or malignant phyllodes tumors was 2%, while LCIS was found in 1% [30]. Therefore, guidelines do not generally recommend the excision of fibroepithelial lesions [2]. A more refined approach for deciding whether to proceed with open excision is to consider the growth rate of the fibroepithelial lesion and suspicion for phyllodes tumor [31, 32].

Missing data on previous breast biopsies or previous diagnosis of invasive BC or DCIS are possible limitations of our work. Since earlier breast malignancy and breast biopsies are risk factors for BC and DCIS, an impact on the previous work cannot be excluded [33, 34]. Furthermore, our data lack information on symptoms prior to interventional breast biopsy. The presented retrospective data originate from a tertiary referral center in a university hospital maybe influencing patient acquisition. Since our work includes data until June 2019, our data lack information regarding the latest classification update on the WHO Classification of Tumours—Breast Tumours—in particular the new classification of lobular neoplasia [35]. Furthermore, our data lack information regard 70 cases that were not followed-up by open surgical excision. Some patients for example refused an open excision or were treated elsewhere so we cannot report upgrade rates for these cases.

In brief, the advantage of this study is the large number of patients included. The case only study design with the absence of specific inclusion criteria (e.g., type of biopsy or lesion type), the large study population of 192 patients, and grouping of B3 lesions separates this study from others. Since we identify statistically significant differences between the three B3 lesion groups, our results suggest distinct therapeutic implications. As proposed by other research groups, a surgical excision can be avoided for lesions with an upgrade rate of less than 2%. This is extrapolated analogously to the BI-RADS category 3, where a possible diagnostic delay is not associated with a worse prognosis [36]. Therefore, we suggest open surgical excision should be recommended for ADH, FEA, LN1-2, papillomas, and RS/CSL. Fibroepithelial lesions without growth tendency and without diagnostic uncertainties do not need to be excised.