Introduction
The extent of the benefit from adjuvant chemotherapy depends on the patient’s individual risk profile [
1]. According to the current recommendations, traditional clinico-pathological parameters such as tumor size, nodal status, histological grading, hormone receptor (HR) status and human epidermal growth factor receptor 2 (HER2) status should be used for risk stratification in women with early stage invasive breast cancer [
2,
3]. In clinical practice, breast cancer is subdivided into different subtypes (i.e., luminal A-like, luminal B-like, HER2-like, triple-negative) [
3]. The development and use of genomic tests using gene expression profiling has the aim to improve prediction of clinical outcome. Especially for HR-positive, HER2-negative, node-negative tumors, the additional use of multigene assays should help to reliably assess the risk profile in order to avoid both under- and overtreatment.
For clinical use, it is crucial that gene expression analyses are assessed according to clear evidence-based criteria [
4]. Recommended commercially available biomarker assays by the St. Gallen International Consensus Conference and the American Society of Clinical Oncology (ASCO) are Oncotype DX
®, EndoPredict
®, MammaPrint
®, PAM50/Prosigna
® and Breast Cancer Index [
2,
3,
5,
6]. All of these gene expression assays are endorsed for guiding the decision on adjuvant chemotherapy in hormone receptor-positive, node-negative tumors if all other criteria do not allow a therapeutic decision. MammaPrint
®, EndoPredict
® and Prosigna
® might also be used in 1–3 involved axillary lymph nodes. Retrospective evidence exists for the use of OncotypeDX
® in node positive disease [
7,
8], although the prospective phase 3 RxSPONDER trial (SWOG S1007,
https://clinicaltrials.gov/ct2/show/NCT01272037) is still ongoing. For the 21-gene recurrence score (TAILORx trial) [
9‐
11] and the 70-gene signature (MINDACT trial) [
12] prospective data are available. There is also a smaller prospective study on EndoPredict
® in 200 patients, which addresses the clinical and psychological impact of this multigene assay [
13]. In addition, there is ample prospective-retrospective evidence for the approved multigene assays.
The RNA-based multigene score EndoPredict
® (EP, Myriad Genetics
®, Salt Lake City, USA) is based on the gene expression analysis of 12 genes, eight cancer-related, three reference genes and one control gene for DNA contamination. EndoPredict
® integrates genomic and clinical information by including clinico-pathological parameters like tumor size and number of involved lymph nodes (EPclin). Integrating these two clinical parameters to the molecular test (EPclin) increased the prognostic value especially for late-distant recurrence for the multigene assay EndoPredict
® [
14]. The test result distinguishes EPclin low-risk and EPclin high-risk patients [
15‐
17]. The clinical benefit of EndoPredict
® to identify a subgroup with an excellent prognosis with endocrine therapy only is well-established [
15,
18‐
20].
With an increasing use of multigene tests, it is of utmost importance to know whether their use changes treatment decisions. This retrospective study in consecutive patients compared pre- and post-test therapy decisions and changes in treatment recommendations. Furthermore, we evaluated clinico-pathological factors influencing the decision for the additional use of EPclin. These real-world data should be of special interest for indication and evaluation of EndoPredict®.
Discussion
In our analysis of 869 consecutive HR-positive, HER2-negative breast cancer patients, gene expression analysis with EPclin was performed in 18.0% for adjuvant treatment decisions. In 33.3% use of EPclin led to a change in treatment recommendations.
Using EPclin, 67 (42.9%) patients were assigned low-risk and 89 (57.1%) high-risk. Remarkably, the distribution between the low and high risk category was different from other studies. In contrast to our results, more low-risk patients were observed in the cohort used for EPclin validation. In this study, 63% of 1702 patients were classified as low-risk and 37% as high-risk [
18,
22]. Similarly, the low-risk group of postmenopausal patients from the TransATAC study (58.8%) was also considerably larger than in the present study [
14]. However, both studies included patients of a prospective study population for whom mainly endocrine therapy was indicated. Therefore the ABCSG-6 /-8 and the TransATAC study cohort represented a completely different study population than the current real-world population in our study. Most likely, a clinical pre-selection to favorable risk might explain this difference. Conversely, another study investigating two different chemotherapy regimens in early breast cancer assigned only 25% to the low-risk group [
16]. Nonetheless, a risk distribution comparable to our results was shown by Müller et al., who evaluated all EndoPredict
® requests in a single pathology institute during one year. 46.4% were classified as low-risk and 53.5% as high-risk [
23]. Only a minority of the patients in our study (
n = 8) was also included in the aforementioned analysis. However, in contrast to their analysis, our study included all eligible HR-positive, HER2-negative patients with 0–3 involved axillary lymph nodes over a period of five years, providing a more unbiased view on the use of EPclin in early breast cancer. The indication to perform a gene expression assay may also be influenced by the experience of the physicians ordering the test, in particular by their individual threshold for administration of chemotherapy. This selection bias could explain the different risk-profiles in different EndoPredict
® cohorts.
In our analysis, the decision to perform EPclin was multivariate associated with younger age, smaller tumor size, positive nodal status, low/intermediate histological grade and intermediate Ki-67 in multivariate testing. The younger median age in the EPclin-tested cohort (57 years) was similar to the observations in other studies [
23‐
25]. The younger age in breast cancer patients is an unfavorable prognostic factor. More aggressive tumor therapy, however, is associated with the increased probability of long-term side effects. Therefore, a reason for the additional performance of a multigene assay in younger patients is primarily whether chemotherapy can be safely avoided. Conversely if a luminal-like tumor is present in elderly patients, endocrine therapy alone is more likely. Histological grade and the proliferative marker Ki-67 are important variables for the risk classification of early breast cancer. Both variables, however, have considerable inter-laboratory and -observer variability [
26,
27]. Therefore, it is not surprising that, similar to other reports, the additional gene expression analysis is most often used in tumors of intermediate histological grade [
23] and intermediate Ki-67. Median Ki-67 was 20% in the EP cohort v. 10% in the untested patient group. Accordingly, in a large study by Nitz and co-workers, patients with intermediate Ki-67 (> 10% to < 40%) were most likely to benefit from a gene expression test [
28]. Histological grade as an influencing factor for use of multigene assays has also been observed in other settings with an increased percentage of G2 tumors in cohorts tested with Oncotype DX
® [
25,
29]. In the assessment of the relationship between EndoPredict
® and clinico-pathological variables only T2 tumor size (
P = 0.007) and positive nodal status (
P = 0.031) showed a higher chance for EPclin risk category (Table
3). Among the multigene tests, the EPclin test is the only one that integrates both clinical parameters (tumor size and nodal status) in addition to molecular genetic parameters. Thus, our results underline the strong influence of these two clinical factors on the patient’s individual risk profile with breast cancer.
We have shown that in 33.3% EPclin led to changes in the therapy recommendation. Comparable rates of therapy changes have been observed in other studies. Penault-Llorca et al. showed in a prospective study that EPclin resulted in a change of therapy recommendation in one third of patients [
13]. Similarly, Müller et al. reported a change in 37.7% [
23]. In our study, EPclin led in 19.2% to therapy escalation and in 14.1% to de-escalation. Interestingly, both studies described therapy de-escalation considerably more often than escalation, presumably reflecting both differences in the patient population as well as in the willingness to recommend adjuvant chemotherapy in intermediate-risk HR-positive, HER2-negative early breast cancer. In line with the latter argument is the fact that in our patients without gene expression profiling the rate of chemotherapy recommendation was only 25.5%. Differences of 30–50% between pre- and post-test treatment recommendations were also shown in studies using Oncotype DX
® and the 70-gene signature (MammaPrint
®) [
12,
30‐
32].
Our study has some strengths and limitations. A potential weakness is that our study was retrospective and performed in a single certified breast cancer center. Furthermore, patient preferences were not addressed. An extension of our study endpoints to patient-related questions (e.g., expectations and wishes on a gene signature test, impact of the test result for the patient), could have led to valuable additional information. However, a major strength of our study is the consecutive inclusion of patients which enables the assessment of EPclin in the context of well-established clinic-pathological factors. To the best of our knowledge, this is the first study dealing with this topic using real-world data. In our study post-test results of EndoPredict® were the reason for a change in clinical decisions in more than 1/3 of tested patients. This is a valuable finding in a "real life" use of this multigene test. With an increasing use of multigene assays it is of particular interest for the indication and evaluation of the EPclin results to know if the clinical treatment decision changes.
In conclusion, EPclin was only performed in a subset of consecutive HR-positive, HER2-negative early breast cancer patients. However, EPclin resulted in a considerable change in therapy recommendations in one third of patients, potentially reducing over- and under-treatment in early breast cancer patients.
Compliance with ethical standards
Conflict of interest
K. Almstedt received speaker honoraria from Roche Pharma AG, Pfizer Pharma GmbH and AstraZeneca. M. Schmidt received honoraria for speaker or consultancy role from AMGEN, AstraZeneca, Eisai, Lilly, Myelo Therapeutics, Novartis, Pantarhei Bioscience, Pfizer, and Roche. He received research funding from AstraZeneca, BioNTech, Eisai, Genentech, Myelo Therapeutics, Novartis, Pantarhei Bioscience, Pfizer, Pierre Fabre, and Roche. He received travel reimbursement from Pfizer and Roche. M. Otto received honoraria for speaker or consultancy role from AstraZeneca, Boehringer-Ingelheim, Roche and Sividon. He received travel reimbursement from AstraZeneca, and Boehringer-Ingelheim. MJ. Battista received honoraria for speaker or consultancy role from AstraZenca, MSD, PharmaMar, Roche Pharma AG, TEVA, Tesaro. He received travel reimbursement from Celgene, PharmaMar and Pierre Fabre. S. Krajnak received speaker honoraria from Roche. He received research funding from Novartis. He received travel reimbursement from PharmaMar. A. Hasenburg received honoraria from AstraZeneca, Celegen, MedConcept Gm, Med update GmbH, Medicultus, Pfizer, Promedicis GmbH, Pierre Fabre, Softconsult, Roche Pharma AG, Streamedup!GmbH and Tesaro Bio Germany GmbH. She is a member of the advisory board of PharmaMar, Promedicis GmbH, Pierre Fabre Pharma GmbH, Roche Pharma AG and Tesaro Bio Germany GmbH. She received research funding from Celgene. C. Denkert has been cofounder and shareholder of Sividon Diagnostics (now Myriad), has received speaker honoraria from Teva, Novartis, Pfizer, Roche, Amgen and has been consultant for MSD Oncology, Amgen, Roche and Daiichi-Sankyo. All other authors declare that they have no conflicts of interest.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.