The prognostic significance of molecular phenotype in breast cancer is well established in the literature. Recent studies have demonstrated that subgross lesion distribution (unifocal, multifocal, and diffuse) and disease extent also carry prognostic significance in this disease. However, the correlation of molecular phenotypes with subgross parameters has not yet been investigated in detail.
In total, 444 consecutive invasive breast cancers that were documented in large-format histology slides and worked up with detailed radiologic-pathologic correlation were sampled into tissue microarray blocks and stained immunohistochemically to delineate the molecular subtypes.
Diffuse or multifocal distribution of the invasive component of breast carcinomas in this series was associated with a 4.14-fold respectively 2.75-fold risk of cancer-related death compared with unifocal tumors irrespective of molecular phenotype. Patients who had human epidermal growth factor receptor 2 (HER2)-positive cancers; estrogen receptor-negative, progesterone receptor-negative, and HER2-negative (triple-negative) cancers; or basal-like cancers had a 2.18-fold, 2.33-fold, and 4.07-fold risk of dying of disease, respectively, compared with patients who had luminal A carcinomas. Unifocal luminal A, HER2-positive, and basal-like cancers were associated with significantly better long-term survival outcomes than their multifocal or diffuse counterparts; luminal B and triple-negative tumors also had the same tendency. In multivariate analysis, patient age, tumor size category, lymph node status, lesion distribution, and molecular phenotypes remained significant.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Breast cancer mortality has decreased significantly in countries that have organized, population-based mammography screening1; however, despite using the most recent diagnostic tools and modern therapeutic regimens, women are still dying from this disease. Therefore, exploring the prognostic information generated by radiologic, histologic, and molecular tumor characteristics is essential to facilitate individualized treatment and further decrease disease-specific mortality.
Patient age, tumor size, histologic grade, and lymph node status still remain powerful prognostic indicators in patients with breast cancer.2, 3 On the other hand, many second-generation prognostic parameters have emerged and have been introduced into routine practice, and the most prominent are estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression status, and proliferative activity.
The description of lesion distribution (unifocal, multifocal, or diffuse growth patterns) and disease extent assessed on large-format histologic sections provide valuable additional prognostic information to the classic first-generation and second-generation prognostic factors.4-6 These parameters are assessed preoperatively using modern multimodality radiology imaging techniques. The high accuracy of these techniques has been evidenced by detailed postoperative radiologic-pathologic correlation.7 This kind of correlation is essential for the correct histologic assessment of lesion distribution and disease extent in breast cancer. Several authors have reported increased lymph node metastatic rates in multifocal cancers compared with unifocal cancers,8, 9 and an increased local recurrence rate in extensive disease also is well evidenced in the literature.10, 11 In addition to some other studies, we previously reported a higher risk of breast cancer-related death in patients who had multifocal/diffuse or extensive disease compared with those who had unifocal tumors or limited disease extent.12
Recent advances in human genome research and high-throughput molecular technologies have revealed the molecular complexity of breast cancer and have contributed to the recognition of its biologic heterogeneity and implications for treatment.13, 14 Routine use of the attractive but expensive and less available method of gene expression profiling is significantly limited.15, 16 Protein expression profiling using traditional immunohistochemistry is widely available and correlates very well with intrinsic genetic expression microarray assessment in delineating the following molecular subtypes: luminal A; luminal B; HER2-positive; ER-negative, PR-negative, HER2-negative (triple-negative); and basal-like.17, 18 These breast cancer subtypes quickly have become a new standard in classification of breast cancer and a basis for therapeutic decision-making.19 Their relation to traditional morphologic and clinical parameters has been highlighted in some recent studies,20-24 but the association with subgross morphology parameters only rarely has been analyzed.25, 26 To our knowledge, the long-term outcome of patients with unifocal, multifocal, and diffuse invasive carcinomas of different molecular phenotypes has not been previously studied.
In this report, we present long-term follow-up results from a large, representative, consecutive series of women with breast cancer, correlating the different molecular subtypes (assessed with immunohistochemistry) with lesion distribution and disease extent (assessed on detailed radiologic-pathologic correlation using large-format histology slides).
MATERIALS AND METHODS
This study was a retrospective analysis of surgical pathology specimens from 574 consecutive breast cancer cases diagnosed at the Department of Pathology and Clinical Cytology, Falun Central Hospital, Falun, Sweden, during the period from January 1996 to December 1998. Patients with recurrent breast cancers who were diagnosed initially before the study period, those with purely in situ carcinomas, and those with microinvasive (<1 mm) carcinomas were excluded, as were those who received neoadjuvant therapy that led to tumor regression. The remaining 444 women comprised the study population. ER, PR, and HER2 status and the expression of basal markers (cytokeratin 5/6 [CK5/6], CK14, and/or epidermal growth factor receptor [EGFR]) were established retrospectively for the purposes of this study. All patients were treated according to the regularly updated protocol of the Breast Unit of the Regional Oncology Centrum (Uppsala/Örebro region), which follows both local and international guidelines. Chemotherapy regimes were based on 5-fluorouracil (5FU), epirubicin, and cyclophosphamide (FEC) and cyclophosphamide, methotrexate, and 5FU (CMF) protocols; and hormone therapy was based on tamoxifen and aromatase inhibitors. The patients in this study population did not receive adjuvant trastuzumab, and they did not undergo sentinel lymph node biopsy. This study was approved by the Regional Ethical Committee of Uppsala/Örebro.
Large-Section Histology and Tissue Microarray
All specimens were prepared using large-format histopathology, a method that has been described previously in detail.27, 28 Briefly, a preoperative tumor board discussed each case, and the radiologic appearance was registered, including the tumor distribution and extent. With the help of this information and the whole-specimen radiograph, the most appropriate sections were obtained to represent the proper extent of the breast lesion. The postoperative tumor board discussed all cases again and checked the concordance of radiologic and histologic findings. A team of 4 pathologists (T.T., G.P., M.G., and M.T.) reviewed all large-format histology slides for the current study. The immunohistochemical assessment was made by using a tissue microarray (TMA). One of the authors (G.P.) reviewed sections microscopically from the original paraffin blocks and marked the most representative areas. Two-millimeter punch biopsies were taken from the large blocks corresponding to the marked area and merged into TMA paraffin blocks that contained 24 to 30 punch biopsy cores. In unifocal carcinomas with heterogeneous histologic appearance as well as multifocal and diffuse cancers, multiple punch biopsies were obtained, as illustrated in Figure 1. Each TMA block also included 1 control and 1 empty square in order to proper identification.
Immunohistochemistry and Scoring
The following antibodies (supplier, type, dilution) were used for staining the TMA slides: ER (clone SP1; Ventana Medical Systems, Tucson, Ariz; 1:200 dilution), PR (clone PgR 636; Dako, Glostrup, Denmark; 1:50 dilution), CK5/6 (clone D5/16 B4; Dako; 1:100 dilution), CK14 (clone LL002; Novocastra, Newcastle upon Tyne, United Kingdom; 1:20 dilution), EGFR (clone E30; Dako; 1:25 dilution), and HER2 (code A 0485; Dako; 1:250 dilution). Hormone receptor assessment was based on the percentage of positive cell nuclei independent of staining intensity; and the cutoff value was set at 10% following the Swedish guidelines (available at: http://www.svfp.se/node/214; accessed June 30, 2012). Cytoplasmic staining in >10% of tumor cells was the criterion for CK5/6 and CK14 positivity, and membranous staining in >10% of the tumor cells was the criterion for EGFR positivity. HER2 positivity was assessed in accordance with the manufacturer's criteria. In this study, only samples with 3+ staining were regarded as HER2-positive. HER2 fluorescence in situ hybridization was not available for this material.
The definition of breast cancer size and distribution evaluated in large-format histology sections was published previously.27, 28 A tumors was defined as unifocal if only 1 invasive focus could be observed in a large section, which may or may not have contained an in situ component. A breast lesion was regarded as multifocal if more than one well demarcated, invasive tumor focus was detected, with such foci separated from each other by normal breast tissue, benign lesions, or in situ carcinoma, regardless of the distance between foci. When tumors were widespread over a large-section area with no distinct tumor mass, like a spider's web, they were defined as diffuse. We analyzed the largest invasive focus for molecular classification in multifocal breast cancers, because individual foci rarely vary in their molecular subtype.29 The molecular phenotypes of tumors were determined according to the criteria of Nielsen et al modified as follows26: luminal A (ER-positive/PR-positive or PR-negative/HER2 negative/negative basal markers); luminal B (ER-positive/PR-positive or PR-negative/HER2-positive/negataive basal markers); HER2 positive (ER-negative/PR-negative/HER2-positive/negative basal markers); triple negative (ER-negative/PR-negative/HER2 negative/negative basal markers); and basal-like (positive for CK5/6 or CK14 and/or EGFR-positive in at least 1 of the invasive tumor foci irrespective of ER/PR/HER2 status).
Data describing the imaging and histologic findings were collected prospectively in a separate database in the Departments of Mammography and Clinical Pathology at Falun Central Hospital. Additional data were collected from patients' clinical files. Patients were followed for an average (±standard deviation) of 9.86 ± 4.07 years (range, 0.02-14.15 years) and for a median of 11.70 years. Cause of death was obtained from the official death registry of the Swedish Board of Welfare (Socialstyrelsen). Survival analysis was carried out for the time to death from breast cancer, with censoring at the end of follow-up (December 31, 2009) or death from other causes. We used Kaplan-Meier survival curves to compare cumulative survival in 3 pathologic categories (unifocal, multifocal, and diffuse), applying the log-rank test to indicate statistical significance, and in 5 categories for the breast cancer molecular phenotypes. If the sample size was small for each stratum, then the log-likelihood method was used to calculate P values. We used Cox proportional-hazards regression to estimate the hazard ratio of dying from breast cancer in multifocal and diffuse breast cancer versus unifocal and luminal A breast cancer versus other molecular phenotypes, with (multivariate analysis) and without (univariate analysis) adjustment for age and tumor attributes (tumor size, lymph node status, and histologic grade). Wald tests and likelihood-ratio tests also were used to calculate P values for each variable in multivariable Cox proportional-hazards regression model. The level of statistical significance was set at 5%.
In this series of 444 consecutive, newly diagnosed breast cancers, 316 of 444 samples (71.2%) were classified as luminal A, 40 of 444 samples (9%) were classified as luminal B, 18 of 444 samples (4.1%) were classified as HER2-positive, 44 of 444 samples (9.9%) were classified as basal-like, and 26 of 444 samples (5.8%) were classified as triple-negative. The luminal A type tumors were unifocal in 212 of 316 samples (67.1%), multifocal in 79 of 316 samples (25%), and diffuse in 25 of 316 samples (7.9%). The luminal B group tumors were unifocal in 24 of 40 samples (60%), multifocal in 9 of 40 samples (22.5%), and diffuse in 7 of 40 samples (17.5%). The HER2-positive tumors were unifocal in 7 of 18 samples (38.9%), multifocal in 6 of 18 samples (33.3%), and diffuse in 5 of 18 samples (27.8%). The basal-like cancers were unifocal in 29 of 44 samples (65.9%), multifocal in 13 of 44 samples (29.5%), and diffuse in 1 of 44 samples (2.3%). In 1 basal-like tumor, the focality of the lesions could not be assessed. The triple-negative tumors were predominantly unifocal (18 of 26 samples; 69.2%) and had multifocality and a diffuse appearance in 4 of 26 samples (15.4%) (Table 1).
Table 1. The Distribution of 444 Invasive Breast Carcinomas According to Molecular Phenotype and Subgross Histologic Growth Pattern
Molecular Phenotype of Invasive Breast Carcinoma: No. of Patients/Total No. (%)
Subgross Histologic Growth Pattern
Abbreviations: HER2, human epidermal growth factor receptor 2.
The distribution of the lesions in 1 basal-like tumor could not be assessed.
Table 2 summarizes the basic clinical and morphologic characteristics of the cases. Most of the luminal A, luminal B, and triple-negative breast cancers were lymph node-negative tumors at the time of operation, but approximately 50% of HER2-positive and basal-like breast cancers had metastases. The luminal A tumors were diagnosed as grade 1 and 2 invasive ductal cancers and as lobular carcinomas in 225 of 316 patients (71.5%). The basal-like and triple-negative cancers were grade 3 ductal carcinomas in 29 of 44 patients (65.9%) and in 12 of 16 patients (46.2%), respectively.
Table 2. Basic Characteristics of Patients With Breast Carcinoma in the Current Series According to Molecular Phenotype
Molecular Phenotype: No. of Patients/Total No. (%)
Highly significant differences in cumulative breast cancer-specific survival were observed among the different molecular subgroups. Patients who had HER2-positive, basal-like, and triple-negative cancers had a shorter cumulative survival compared with those who had tumors with luminal A and B phenotypes (Fig. 2). The cumulative survival in the luminal A, HER2-positive, and basal-like subgroups was altered according to the subgross distribution of lesions, indicating that patients who had unifocal cancers with these phenotypes had significantly better long-term survival outcomes than those who had multifocal or diffuse tumors (Fig. 3). Such an alteration was not statistically significant in the patients who had luminal B or triple-negative tumors, but the tendency was evident. The number of patients in these subgroups was relatively small.
Tumors ≥15 mm (5.31-fold risk), those with positive lymph nodes (4.05-fold risk), or diffuse or multifocal tumors (4.14-fold and 2.75-fold risk, respectively) carried an increased risk of dying from breast cancer irrespective of their molecular phenotype, as indicated in Table 3. Patients who had breast cancer with the basal-like phenotype had a 4.07-fold risk of dying of the disease (95% confidence interval, 2.22-7.47; P < .0001) compared with those who had luminal A carcinomas. In multivariate analysis, patient age, tumor size category, lymph node status, distribution of the lesions, and molecular phenotype remained significant.
Table 3. Univariate and Multivariate Analyses of the Association Between Breast Cancer Molecular Phenotypes, Subgross Lesion Distribution (Unifocal, Multifocal, or Diffuse), and Disease Extent With the Risk of Breast Cancer Death (Adjusted for Age, Tumor Size, Lymph Node Status, and Histologic Grade)
Breast cancer represents a heterogeneous group of diseases with regard to clinical manifestation, radiologic appearance, tumor morphology, and outcome. Pathologists have recognized the histologic heterogeneity of breast cancer for a long time, but this concept was recently brought to the forefront by microarray-based gene expression profiling analyses.31-33 Therapeutic decision-making is still based on a handful of clinicopathologic features that do not necessarily reflect this heterogeneity in every case. There is a clear need for additional parameters to refine prognostication and prediction of therapy response to better individualize treatment.
Our current study confirmed the relevance of the traditional morphologic parameters that have a high impact on breast cancer outcome.34-36 Tumor size and lymph node status were confirmed as significant prognostic parameters in both univariate and multivariate analysis, as was histologic grade in univariate analysis. Furthermore, we also confirmed our previous observations and those of some other groups that subgross morphology, especially multifocality and diffuse distribution of invasive cancer, are related significantly to survival.12, 37 Compared with patients who had unifocal tumors, patients who had diffuse invasive tumors had a 4.14-fold increased risk of dying from disease, whereas those who had multifocal tumors had a 2.75-fold increased risk. These observations deviate from other published results that did not demonstrate an impact of multifocality on survival.38 The discrepancies may be related to methodological issues or differences in defining breast cancer multifocality. The lesion distribution is best evaluated with multimodal imaging studies (mammography, ultrasound, and magnetic resonance imaging) correlated with large-format histopathology slides rather than with macroscopic or traditional histology assessment.39
The seminal studies by Perou et al and Sorlie et al demonstrated that breast cancers can be discriminated according to the similarity between their transcriptome profiles.13, 15 Those authors developed an intrinsic gene set and used hierarchical clustering analysis to classify tumors into 4 main groups: 1) luminal, which includes tumors that express ER and genes that belong to the ER pathway and reflects a profile that, to some extent, recapitulates that of normal luminal epithelial cells; 2) basal-like cancers, which are hormone receptor-negative/low tumors that express genes usually present in basal/myoepithelial cells; 3) HER2-positive tumors, which over express HER2 and genes belonging to the HER2 amplicon; and 4) the normal breast-like group, which consistently clusters together with normal breast samples and fibroadenomas. The main results of those studies were that breast cancers could be separated by hierarchical clustering analysis into ER-positive and ER-negative/low tumors. More recently, slightly different new genetic subtypes have been delineated, such as an interferon-rich group, a molecular apocrine subtype, and a claudin-low subtype; whereas it appears that the normal breast-like subtype is likely to constitute an artifact of the frozen-sample procedure.30, 40
Several attempts have been made to substitute the expensive and complicated procedure of genetic profiling and cluster analysis with simpler immunohistochemistry methods, resulting in phenotyping classifications that deviate slightly from each other. The luminal group, which is regularly characterized with ER expression, was divided into the luminal A subgroup, in which patients have an excellent prognosis, and the luminal B subgroup, in which patients have an unfavorable prognosis. This separation was made on the basis of HER2 expression,41 tumor grade,18 or the Ki-67 proliferation index.19 Despite the finding that patients with basal-like cancers have a significantly worse prognosis than patients with other types of breast cancers, there is still no internationally accepted definition for these tumors. They are most often characterized with expression of myoepithelial markers, such as CK5/6, CK14, CK17, and EGFR. However, to our knowledge, no consensus exists regarding whether or not these tumors should be classified triple-negative.22, 42 Therefore, we tested separately the triple-negative tumors and those expressing myoepithelial markers as well as hormone receptors and/or HER2. In addition, we chose to define the luminal B groups with ER and HER2 positivity, because proliferation and grade are not phenotypic parameters. Despite the differences in criteria, retrospective analyses have revealed that this taxonomy was associated with the outcome of breast cancer patients, with basal-like and HER2-positive tumors displaying more aggressive clinical behavior, whereas luminal A carcinomas were associated with an excellent prognosis.40, 41
Our current results indicate that molecular phenotype influences breast cancer-related survival. Patients who had basal-like phenotype cancer had a much worse prognosis than those who had luminal A or luminal B type cancer. Patients with basal-like cancer had a 4.07-fold risk of cancer-related death, which was almost equal to the risk of having metastatic deposits in axillary lymph nodes (4.05-fold risk). The triple-negative phenotype itself was associated with higher risk of cancer-related death (2.33-fold) compared with luminal A type cancers in univariate analysis, which reached the level of risk of all grade 3 carcinomas but was not statistically significant in the current series, probably because of the small number of events in this subgroup. Our results are similar to those published by Dawood et al18 and other groups.40, 42
In conclusion, the main objective of the current study was to examine the relation between breast cancer subgross morphology and molecular phenotype with regard to prognosis and long-term survival. Multifocality of the invasive carcinomas was identified as a significant, independent morphologic prognostic parameter in luminal A, HER2-positive, and basal-like tumors. This tendency also was evident in the remaining 2 molecular tumor categories but without reaching statistical significance, probably because of the small number of such cases. Diffuse distribution of the invasive cancers also was identified as an independent unfavorable prognostic parameter in our material, although >50% of these tumors belonged to the luminal A molecular subgroup. Furthermore, most patients who died of diffuse disease also belonged to the luminal A group. This result indicates that molecular classification of breast cancers is a powerful tool but gains in power if combined with conventional and subgross morphologic parameters.
Dr. Pekar received support from the Dalarna Center for Clinical Research.