Prognosis of breast cancer molecular subtypes in routine clinical care: A large prospective cohort study

Background In Germany, most breast cancer patients are treated in specialized breast cancer units (BCU), which are certified, and routinely monitored. Herein, we evaluate up-to-date oncological outcome of breast cancer (BC) molecular subtypes in routine clinical care of a specialized BCU. Methods The study was a prospectively single-center cohort study of 4102 female cases with primary, unilateral, non-metastatic breast cancer treated between 01 January 2003 and 31 December 2012. The five routinely used molecular subtypes (Luminal A-like, Luminal B/HER2 negative-like, Luminal B/HER2 positive-like, HER2-type, Triple negative) were analyzed. The median follow-up time of the whole cohort was 55 months. We calculated estimates for local control rate (LCR), disease-free survival (DFS), distant disease-free survival (DDFS), overall survival (OS), and relative overall survival (ROS). Results Luminal A-like tumors were the most frequent (44.7 %) and showed the best outcome with LCR of 99.1 % (95 % CI 98.5; 99.7), OS of 95.1 % (95 % CI 93.7; 96.5), and ROS of 100.0 % (95 % CI 98.5; 101.5). Triple negative tumors (12.3 %) presented the poorest outcome with LCR of 89.6 % (95 % CI 85.8; 93.4), OS of 78.5 % (95 % CI 73.8; 83.3), and ROS of 80.1 % (95 % CI 73.8; 83.2). Conclusions Patients with a favorable subtype can expect an OS above 95 % and an LCR of almost 100 % over 5 years. On the other hand the outcome of patients with HER2 and Triple negative subtypes remains poor, thus necessitating more intensified research and care. Electronic supplementary material The online version of this article (doi:10.1186/s12885-016-2766-3) contains supplementary material, which is available to authorized users.


Background
Breast cancer (BC) mortality has declined over the past decade in most developed countries, due to new developments in screening, diagnostics, surgery, radiotherapy, and (neo) adjuvant systemic therapy, in conjunction with structural improvements (multidisciplinarity, implementation of specialized breast cancer units) and target agreements (evidence-based guidelines, certification processes) [1,2]. Over the past decade, increasing molecular and genetic knowledge [3][4][5][6] has provided a new understanding of breast cancer as a heterogeneous, systemic disease that can be classified into different subtypes with different clinical and pathological features, different therapeutic response patterns, and different outcomes [7,8]. The main molecular classification of breast cancer have been distinguished by gene expression profiling into intrinsic subtypes by Peru et at [5]. These modern microarray-based gene expression profiles (GEP) are the best way to visualize the heterogeneity of breast cancer, but lacking gene expression profiling in clinical routine due to cost and practicability made a surrogate classification necessary [9]. The molecular subtypes of breast cancer correspond reasonably well to a clinical characterization on the basis of hormone-and HER2 status, as well as proliferation markers or histological grade [10]. So the classification based on immunohistochemistry (IHC) markers was recommended by the St. Gallen Expert Consensus in 2011 [11] and confirmed again in 2013 [12]. It has become the accepted standard in routine clinical patient care. Classification into five molecular subtypes (Luminal A-like, Luminal B/HER2 negative-like, Luminal B/HER2 positive-like, HER2-type, Triple negative) helps to sort patients into groups with divergent prognoses and different response patterns to specific Every-day-routine outcome assessment of specialized breast cancer unit (BCU) must validate guideline-based care of BC patients in order to optimize the therapy of every individual case. This paper reports the outcome data of a prospective cohort of 4102 patients with primary, unilateral, non-metastatic BC treated at a specialized BCU according to routinely used molecular subtype definitions based on immunohistochemistry markers.

Patients
Since 01 January 2003 the medical history and the demographic, diagnostic, therapeutic, and follow-up data of all breast cancer patients referred to the BCU at Heidelberg University have been prospectively entered into our database. This register is routinely used for certification purposes and is monitored.
Patients from the registry were included in the present analysis if they had invasive or carcinoma-in-situ cancer of the breast and were newly diagnosed or treated between 01 January 2003 and 31 December 2012.
Patients were excluded from this analysis for any of the following reasons: male sex (n = 38), distant metastasis at the intake visit (M1, n = 296) or bilateral tumors (n = 619).
Patients with incomplete immunohistological information (149, i.e. 4.1 % of 3603) were included in the overall analysis, but they could not be considered in the subgroup outcome analysis.

Histology and stage
Tumors were defined according to the World Health Organization [13], graded along Elston and Ellis [14], and grouped into stages according to the TNM classification [15]. The expression of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki-67 were assessed with an IHC assay of formalin-fixed, paraffin-embedded tumor tissue according to international standards.

Subgroups
According to the St. Gallen International Expert Consensus recommendations 2011 [11], five molecular subtypes of invasive breast cancer have been differentiated by their expression of the IHC markers ER, PR, HER2, and Ki-67: The classification of 2011 was used because it corresponded best to the way we had categorized patients during the time period covered in this report [11].
Positivity for ER and PR was defined as an Immunoreactive Score [16] of at least 1 out of 12 or a Total Score [17] of at least 1 out of 8. All cases of non-invasive carcinomain-situ (CIS, regardless of specific subtype) have been defined as an additional subgroup for a separate analysis.
For invasive BC, the cell proliferation marker Ki-67 was available in the majority of our cohort (3004/3603, 83.4 %), while grading, either 1 or 3, was used in 599 of the 3603 cases (16.6 %) for subgroup classification (Table 1). For the differentiation of Luminal-like tumors, cases with a negative HER2 receptor status in combination with a positive ER or PR receptor and a grading of 1 led to the attribution of the Luminal A-like subgroup. In contrast a grading of G3 was assigned to the subgroup of Luminal B/HER2 negative-like tumors.

Treatment
The Heidelberg University BCU was fully certified on 10 October 2003, by the German certification board of the German Cancer Society and the German Society for Senology on the basis of the management of cases in 2002 and 2003. Thus all the cases included in this study were managed under certified conditions, which were confirmed by an annual re-certification process [18,19].

Endpoints and outcome assessment
The outcome from the time of diagnosis was assessed for the whole cohort, the five BC subtypes, and the CIS cases for several outcome parameters. The endpoints were local control rate (LCR), disease-free survival (DFS), distant disease-free survival (DDFS), overall survival (OS), and relative overall survival (ROS). Relative survival was   defined as the ratio of the observed survival to the survival expected in the general West German population of the same age and sex during the same period of time [20]. Outcome was assessed as follows. First, hospital records were reviewed to obtain information with regard to survival, local and regional relapse, and distant metastasis. If outcome information was not available in the hospital record, the patient's family doctor or gynecologist was contacted by mail or phone. If the required information could not be obtained by this approach either, an inquiry about the patient's survival status was made at the responsible residents' registration office. If the patient was still alive, she was contacted by mail and asked whether she had developed local or distant relapse with a detailed questionnaire. Follow-up was performed for cases diagnosed until 31 December 2012. Within this study period (starting 01 January 2003) n = 2322 patients had a complete follow-up information (i.e. could be followed until either death or study end. Of the remaining 1780 patients, 140 were lost to follow-up during the years 2003-2011, i.e. in total 140/4102 (3.4 %). The median time of follow-up was 45 months among those who were lost to follow-up, slightly shorter than among the whole cohort (55 months).

Statistical analysis
The data were analyzed using SAS software version 9.3 (SAS Institute Inc.; Cary, NC, USA) and SPSS software version 22 (IBM; Armonk, NY, USA). The proportions of patients experiencing events at 5 years, the corresponding 95 % confidence intervals (95 % CI), and all survival plots are based on Kaplan-Meier estimates using PROC LIFETEST with the actuarial approach. Relative survival rates at 5 years were also calculated. The expected survival of the general population was calculated according to the Ederer II method [21], based on life tables for Germany for the years 2002 to 2010.

Patient characteristics
The final cohort comprised 4102 patients, of which, 3603 (87.8 %) had invasive carcinoma and 499 (12.2 %) had CIS. Most invasive carcinoma cases were hormone receptor positive (ER: 79.9 %, PR: 72.1 %), HER2 negative (86.5 %), and had a grading of 2 (53.4 %). Most of the patients had a maximum tumor size of 2 cm (pT1: 62.2 %) without axillary lymph node involvement (pN0 68.6 %). Median age of the whole cohort was 57 years and most patients were postmenopausal (60.9 %). Breast conservation surgery was performed in 73.1 % of the study cohort and a mastectomy in 26.9 %. Concerning surgical management of the axilla sentinel lymph node biopsy alone (SLND) was performed in 40.7 % and axillary lymph node dissection in 39.0 % of the patients. Detailed patient characteristics of the cohort are shown in Table 1.
The UICC stage distribution (Additional file 1: Table S4) as well as the frequency of age, menopausal status and laterality (Additional file 2:   Table 2). This confirms the favorable prognosis of primary non-metastatic breast cancer receiving adequate treatment. In this study, we focused on a well-defined and homogenous patient cohort. The outcomes seen here can be expected at any specialized BCU. Most of the outcomes statistics published in the literature derive from clinical trials with the exclusion of certain types of patients commonly seen in routine care, e.g. the elderly patients with comorbid conditions. Thus, it is important to assess outcomes among a complete, unselected patient population seen in a routine clinical setting. On the basis of the favorable outcome results reported here, additional quality-of-life aspects might be brought more into focus for outcome quality for specific subgroups in the future.
In the face of unavailable gene expression profiles in clinical routine, the BC surrogate classification according to the St. Gallen Consensus 2011 [11] allows a differentiation of five molecular subtypes with distinct prognoses. Although the management of BC patients according to these subtypes has gained importance, it is beyond controversy that the traditionally assessed tumor characteristics, e.g. nodal status and tumor size, still have independent prognostic impact [22]. Because the St. Gallen subtype classification is widely accepted as a surrogate for subtyping according to intrinsic signatures [9], we used this classification for subtype-specific outcome analysis as they are distinct and well applicable in the context of outcome assessment. Standard pathological assessments seem adequate to define useful groups such as TN, HER2-type, and LumB/HER2 pos.-like tumors, for which treatment recommendations are seldom controversial [23]. In contrast to other studies (e.g. [24]), the Ki-67 score was available for the vast majority of cases, enabling us to differentiate the Luminal-like HER2 negative tumors. Nevertheless, the validity and robustness of Ki-67 is still controversial, although it has been widely accepted as a cell proliferation marker that is widely available [25]. Especially the St. Gallen 2011 cut-off recommendation of 14 % for Ki-67 (which was proposed and validated by  Cheang et al. [26]) has been viewed critically due to a substantial inter-observer and intra-observer variability, especially for mid-range Ki-67 scores [27][28][29]. This discordance is highly problematic because a recommendation for or against chemotherapy for hormone receptor positive, HER2 negative, grade 2 tumors depends mainly on the Ki-67 threshold in the St. Gallen Consensus 2011. Because of this ambiguity in defining exact surrogate subtypes it might be difficult to compare subtype outcome results with other studies that used different surrogate definitions. Despite this difficulty in comparison with other study designs the general trend concerning distribution (at least for clear defined subtypes like TN) and outcome in our cohort is in approximate accordance with other results e.g. from Canada [30], USA [31,32], South Korea [33], Belgium [24], Spain [34,35], Italy [36] and France [37]: LumA and LumB tumors were the most frequent (LumA was 44.7 %, LumB/HER2 neg. was 31.8 %, and LumB/HER2 pos. was 6.2 %), followed by TN cancers (12.3 %) and HER2 type (5.0 %). For the majority of patients with a Luminal A type a very favorable OS over 5 years of 95.1 % (95 % CI 93.7; 95.5) and an excellent LCR of 99.1 % (95 % CI 98.5; 99.7) was possible. But it becomes also evident that outcome possibilities for HER-2 type and TN cases are still much poorer even in times of more effective systemic treatment (Table 3). Two exemplary studies with large cohorts-the single-hospital report from Broukhaert et al.
in Belgium [24] and the population-based report from Minicozzi  In our cohort the median age of early breast cancer patients was 57 years compared to 64 years in Germany. Concerning surgical procedures mastectomy was performed in 26.9 % of all patients, which is lower than in a current published report from the SEER database with a mastectomy rate of 34 % showing an increase in the United States especially in women with node-negative and in-situ disease (Table 1) [38]. The widely use of specific anti-HER2 therapy in this cohort could increase outcome for patients with HER2 positive breast cancer [39].  Table 3). The poor survival of triple-negative tumors reflects the lack of effective and specific therapy for this subgroup of patients.

Strengths and limitations
This study adds recent up-to-date outcomes for the 5 different molecular subtypes from a large prospective cohort with a broad usage of Ki-67 for subtype definition (with a cut-off at 14 %). Unfortunately the subgroups in this study are too small with a favorable outcome and the analysis is underpowered to exhibit further differences in stage distribution as well as surgical and systemic management of primary breast cancer on outcome. The effective disentangling of these and further possible effect would require far larger sample sizes than are present in our data base, and should be pursued within cooperative research projects [40].
In our study, the cohort mirrors the typical, unselected cohort of breast cancer patients treated at a specialized breast cancer unit, as there were no specific exclusion criteria. Unfortunately, we did not systematically document a performance status describing comorbidities. As a prospective, single center study, we cannot exclude any potential center effects that may have confounded the results. Further outcome studies from other clinical settings may help to identify if and to what extent outcomes may vary between different breast units.

Conclusion
If primary BC is managed at a specialized BCU under guideline-adherent conditions, patients with a favorable subtype can expect an OS above 95 % and an LCR of almost 100 % over 5 years. On the other hand the outcome of patients with HER2 and TN subtypes remains poor, thus necessitating more intensified research and care.

Additional files
Additional file 1: