Preoperative core needle biopsy is accurate in determining molecular subtypes in invasive breast cancer

Breast cancer is the most common malignancy affecting women. However, the mortality has decreased in western countries due to earlier diagnosis and more comprehensive treatment [1]. The core needle biopsy (CNB) procedure is almost as accurate as an open excision biopsy (OEB) in the diagnosis of breast diseases, and is now widely taken as the standard procedure for a breast cancer diagnosis [2]. The 2011 European Society of Medical Oncology breast cancer clinical practice guideline required a preoperative disease-related staging, including pathological examination of the CNB with a report on estrogen receptor (ER), progesterone receptor (PgR), and human epidermal growth factor receptor-2 (HER2) status by immunohistochemistry (IHC) or fluorescence in situ hybridization (FISH) [3]. A recent meta-analysis showed that CNB tissue could replace OEB for determining ER, PgR, and HER2 status [4]. Breast cancer is a heterogeneous disease and microarray expression data have demonstrated that there are at least four subtypes of breast cancer, including Luminal A, Luminal B, HER2positive, and basal-like subtypes [5]. Practically, these subtypes can be approximated using clinicopathological markers rather than gene expression array criteria. The 2011 St.Gallen breast cancer consensus also recommended that the IHC status of ER, PgR, HER2, and Ki67 could be used to approximately classify breast cancer into these subtypes, which can guide subsequent systemic treatment [6]. However, due to its relatively smaller sample size and tumor heterogeneity, the biomarker assessment performed on CNB samples may be less reliable than in OEB [7‐9]. Little has been reported on the comparison of molecular breast cancer subtype between CNB and OEB.

Therefore, using IHC and FISH to detect the ER, PgR, HER2, and Ki67 status in CNB and subsequent OEB samples, we then constructed breast cancer molecular subtypes. Our aim was to estimate the concordance between CNB and OEB in evaluating molecular subtypes as well as the receptor status and Ki67 expression levels.

To our knowledge, this is the first study to evaluate the concordance of molecular subtypes between CNB and subsequent OEB samples in large series of breast cancer patients. In the present study, good agreement was demonstrated in evaluating molecular subtypes as well as ER, PgR and HER2 status between CNB and OEB (κ > 0.6). Although, Ki67 expression was found to be slightly higher in the OEB samples.

Concordance rates of 93.6% for ER, and 85.9% for PgR showed a good correlation with these biomarkers between CNB and OEB, similar to other studies, although the ER concordance rate was relatively higher than with PgR [9, 14]. The main explanation may be poorer fixation of OEB compared with CNB specimens, including delayed fixation, under-fixation, and over-fixation with formalin prior to IHC analysis, because the PgR test seems to require a higher preparation quality than an ER test [10, 15]. Another reason could be more heterogeneous distribution within the tumor for PgR compared with ER detection [16].

In terms of HER2 examination, a 96.3% concordance rate after adding FISH testing showed that detecting HER2 on CNB samples was as sensitive in predicting HER2 status as OEB. Previous studies have reported concordance rate between CNB and OEB for HER2 examination to be about 90%. However, one study reported a false positive rate of IHC testing on CNB samples as high as 19.3% [17]. A recent meta-analysis showed that the sensitivity and specificity of HER2 status evaluation of CNB was 81% and 89%, respectively, with the HER2 positivity definition as IHC 2+ or 3+ or FISH+. However, the specificity of HER2 detection in CNB would be improved with a very low false positive rate (specificity 98%) using a HER2 positivity definition as IHC 3+ or FISH + [4]. In our cohort, we carried out all FISH testing in IHC HER2 2+ cases, according to the ASCO/ACP HER2 detection recommendation, most likely explain our high concordance rate.

Ki67 antigen has been used to evaluate the proliferative activity of breast cancer for several decades, and a meta-analysis has shown that high Ki67 expression confers a higher risk of relapse and a worse survival [18]. There was an increasing debate about the lack of standardization of Ki67 pathological interpretation and standard cutoff value for Ki67 high expression. Published studies have used various Ki67 cutoff value such as mean, median, the optimal cut-off value or arbitrary values [18]. In the current study, we used 20% (mean value in ER+/HER2- tumors and median value for the whole patients on CNB) as another Ki67 cutoff value (14%) for determining Luminal A and B subtypes in HR+/HER2- diseases. Breast Cancer International Research Group (BCIRG) 001 trial subgroup analysis showed that Ki67 IHC results, whose cutoff value was 14%, can define which ER+/HER2- tumors can get more benefit from adjuvant docetaxel treatment [19]. Moreover, in PACS 01 trial, ER positive breast cancer patients with Ki67 ≥ 20% were more sensitive to docetaxel treatment in the adjuvant setting [20]. In patients with advanced breast cancer, higher Ki67 levels have been significantly associated with decreased time to aromatase inhibitor treatment failure [21]. The comparison of baseline Ki67 labeling index and post-treatment level would enhance the informative value of the test in patients received preoperative endocrine therapy [22]. Gene expression profiling has revealed that the Ki-67 gene seems to play an important role in several “proliferation signatures” and can be assessed by the Ki67 index [23]. Furthermore, Ki67 is a key selected gene in the Oncotype DX^TM assay, which can be used to predict the outcome and chemotherapy sensitivity in ER+/HER2- tumors [24, 25]. Thus, the 2011 St. Gallen breast cancer consensus recommended that proliferation markers, such as Ki67, can be applied to classify breast cancer into different Luminal subtypes, guiding further treatment [6]. In our study, we found moderate agreement (κ = 0.545) and a slightly higher Ki67 expression in OEB samples compared with CNB samples, with the mean Ki67 expression values of 29.3% and 26.8%, respectively. A major reason for this Ki67 expression difference may be due to sampling error and tumor heterogeneity, as CNB might not reflect the true status of the entire tumor [12]. However, we found no improvement in Ki67 evaluation in smaller compared with larger tumors. There is literature suggesting that four cores can provide sufficient tumor for biomarker testing and good diagnostic accuracy, meaning that Ki67 evaluation might improve with increasing number of cores [26]. However, the level of concordance between the CNB and OEB improved only slightly with increasing number of core passes, but reaching a plateau after 6 or more core passes [27]. In addition, the concordance rates were much higher for ER, PgR, and HER2 than Ki67, which again can be explained by more heterogeneous distribution within the tumor for Ki67, especially in ER/PR positive or grade 1–2 tumors. A comparison of immunocytochemical assays for Ki67 and other biologic variables on preoperative fine-needle aspirates with OBE results showed the concordance between cytology and histology was the lowest for Ki67 evaluation: 89% for ER, 78% for PgR, 79% for p53, and 70% for Ki67, respectively [28]. In one study, Ki67 did not discriminate different biological subtypes of disease with distinct clinical courses rather describe the composition of the mixture of cells in the tumor, which also reflected that the heterogeneity of breast cancer might contribute to Ki67 scoring inconsistence [29]. Greer et al. compared the Ki67 expression between CNB and OEB by IHC and showed a concordance rate of 73% with a κ value of 0.48, similar with our result [27]. Our data indicate that, due to heterogeneous distribution of the Ki67 antigen, CNB may not adequately represent its true biologic profile and Ki67 should be detected on both CNB and OEB in order to avoid misclassifying tumor subtypes and omission of life-saving systemic therapy, especially on HR+/HER2- tumors.

Breast cancer can no longer be considered as a single disease [5, 30]. Molecular subtypes can be defined by microarray testing and this classification approximated using IHC results of ER, PgR, HER2 and other proliferation biomarkers. In order to make an appropriate individualized therapeutic strategy, management of breast cancer according to these distinct subtypes is required. CNB is being increasingly used for breast cancer diagnosis and translational research. However, there has been no large published data about the agreement of these molecular subtypes between CNB and OEB in breast cancer. In the current study, with a large series of breast cancer patients, we found good agreement in evaluating molecular subtypes on CNB compared with those on OEB samples (κ = 0.658). Furthermore, a high concordance rate was also detected if by subdividing Luminal B subtype into Luminal B-HER2- and Luminal-HER2+ subtypes. However, approximately 14% of HR+/HER2- specimens would be classified as Luminal A on CNB and Luminal B on OEB, thus depriving these patients of potentially helpful chemotherapy. In summary, our results show a high concordance rate and good agreement between CNB and OEB in the distinction between Luminal and non-Luminal molecular subtypes. However, the differentiation of Luminal A from Luminal B in HR+/HER2- patients is less accurate due to intra-tumoral heterogeneity of Ki67.

CNB had good agreement in evaluating molecular subtypes as well as ER, PgR, and HER2 status in breast cancer. Due to inherent Ki67 heterogeneity ER+/HER2- tumors, distinguishing Luminal A from Luminal B is less accurate; thus Ki67 ought to be examined both on CNB and OEB samples, especially in HR+/HER2- tumors. Our findings support the recommendation that CNBis considered the initial procedure to assess molecular subtypes and receptor status in invasive breast cancer.