Introduction
Several prognostic multigene tests have been developed for estrogen receptor-positive (ER+) early breast cancer (BC) patients [
1‐
7]. Large clinical validation studies have demonstrated that molecular assays are useful for stratifying patients into risk categories and helpful in making clinical treatment decisions in ER+/node-negative BC patients. Much less is known, however, about the prognostic performance of these tests in patients with axillary lymph node–positive disease. So far, only a few of these assays have been validated in large node-positive BC cohorts treated with endocrine or chemoendocrine treatment. For instance, the 21-gene recurrence score (RS) was initially established and validated in node-negative BC patients [
1,
7,
8]. Later, the SWOG-8814 study demonstrated that RS was able to predict distant metastases in node-positive BC patients [
9]. However, SWOG-8814 and other trials [
10] demonstrated that putative low-risk patients have a considerable, sustained risk for distant metastases. Therefore, the question remains whether multigene assays can be used to (1) identify node-positive BC patients for who can safely be spared from undergoing chemotherapy and (2) tailor more intensive or novel drug-based treatment strategies in clinically high-risk cohorts. Additionally, none of the available tests has yet been validated to predict taxane efficacy [
10‐
12].
The EndoPredict (EP) test has recently been introduced as an RNA-based multigene test to predict the likelihood of distant recurrence in ER-positive/HER2-negative (ER+/HER2−) BC patients treated with adjuvant endocrine therapy. The test is designed to be used in a decentralized setting in molecular pathology laboratories [
6,
13‐
15]. Training in the use of EP was noticeably different compared to other prognostic tests: node-negative and node-positive ER+/HER2− BC patients (
n = 964) were included in the multigene algorithm design, and a combined score of EP, tumor size and nodal status (EPclin) was defined in the large training cohort. EP was subsequently validated in two randomized phase III trials (Austrian Breast and Colorectal Cancer Study Group trials ABCSG6 and ABCSG8;
n > 1,700) that included postmenopausal node-negative and node-positive BC patients treated with endocrine therapy alone [
6]. Subgroup analyses within the ABCSG validation studies indicated that EP and EPclin could be used to identify subgroups showing remarkable differences in 10-year distant recurrence rates in patients with node-negative and node-positive disease. Although the ABCSG6 and ABCSG8 studies demonstrated that EP results enabled the identification of a subgroup of node-positive BC patients with particularly good clinical outcomes, the performance of EP in chemotherapy-treated, node-positive patients has not been evaluated yet.
In this study, we validated the EP score in node-positive ER+/HER2− BC patients in the GEICAM 9906 trial, who were treated with adjuvant chemotherapy followed by hormone therapy. We also evaluated whether EP results could predict the efficacy of incorporating weekly paclitaxel into anthracycline-based regimens.
Discussion
The EP test has recently been validated in two large phase III trials (ABCSG6 and ABCSG8) that included postmenopausal ER+/HER2− BC patients treated with endocrine therapy alone. The ABCSG trials demonstrated that EP adds significant prognostic information to all commonly used clinicopathological parameters (including ER and Ki67) and clinical guidelines [
6,
13]. In line with the ABCSG6 and ABCSG8 clinical validation studies, we conducted a third clinical validation for the EP test by using archived tissue material according to the prospective-retrospective design described by Simon
et al. [
22]. In our present study, we analyzed EP-based data retrospectively in a large FFPE sample set derived from the phase III GEICAM 9906 trial on the basis of prospectively acquired clinical data and prespecified study objectives and laboratory assays.
To the best of our knowledge, this study is the first to show that EP is an independent prognostic parameter for both MFS and OS in node-positive, ER+/HER2− BC patients treated with adjuvant chemotherapy followed by hormone therapy. Also, in this cohort selected by ER+/HER2− status, EP test results add prognostic information to other common clinicopathological variables in this cohort. EP test results provide important information regarding the residual risk of recurrence after a modern anthracycline plus taxane chemotherapy regimen.
The results we report further suggest that EP better captures the tumor-derived intrinsic factors that lead to distant metastasis in node-positive, ER+/HER2− disease compared with some of the clinicopathological markers traditionally used to make treatment decisions (age, grade, nodal status, tumor size, treatment arm, and ER and PR status and Ki67 index).
What are the clinical implications of our results? The EP-/EPclin-based risk classification identifies a subgroup with a particularly low rate of distant metastatic events in a node-positive, high-risk cohort treated with anthracycline with or without a taxane-containing chemotherapy, followed by 5 years of endocrine therapy. In the face of 100% estimated distant MFS in the EPclin-based low-risk group, one might speculate that this patient group does not need an extension of endocrine therapy beyond 5 years. This finding is clinically relevant because the results of several clinical trials suggest that the extension of endocrine treatment should be considered in patients with ER+ BC in order to prevent late metastasis [
23‐
28]. Additionally, EP-based low-risk patients can be sufficiently treated with standard chemotherapy. On the basis of our results, one might even speculate that the EP-based low-risk, node-positive ER+/HER2− patients treated with endocrine therapy might not derive any benefit from chemotherapy.
Several prognostic tests have been validated for BC patients in the past decade. The Onco
type DX Breast Cancer Assay (Genomic Health, Redwood City, CA, USA) and the MammaPrint diagnostic test (Agendia, Irvine, CA, USA) were the first commercially available gene expression tests to predict the risk of recurrence in early-stage BC [
1‐
3]. Whereas researchers in decision impact studies have demonstrated that both tests reduce health-care costs and spare patients from unnecessary chemotherapy [
29‐
31]. Onco
type DX has a higher level of supporting evidence than MammaPrint on the basis of large prospective-retrospective clinical validation studies. However, a recent biomarker substudy of the Arimidex, Tamoxifen Alone or in Combination trial (ATAC; ClinicalTrials.gov Identifier: NCT00849030) suggested that IHC4 provides prognostic information similar to that garnered from the Onco
type DX assay [
32]. Furthermore, IHC4 and the Clarient InsightDx Mammostrat test (Clarient Diagnostic Services, Aliso Viejo, CA, USA)—expanded immunohistochemical tests—can be used to determine the expression levels of diverse protein panels [
33]. Although both tests are valuable as predictors for risk of recurrence, further statistical validation is still needed to ascertain whether these IHC tests can be standardized for everyday clinical use [
34]. EP test results, together with other second-generation multigene tests (Predictor Analysis of Microarray 50-gene test (the PAM50 Breast Cancer Intrinsic Classifier) and Breast Cancer Index assay), have recently been introduced. In contrast to first-generation multigene algorithms and IHC4, these tests can also be used to predict late metastases [
35‐
37]. Additionally, the PAM50 and EP tests can be carried out in a decentralized setting [
15].
Most of the prognostic tests mentioned herein were initially developed to improve decision-making in the clinical management of node-negative BC patients [
1,
4,
5,
7]. Although some prognostic tests have been evaluated in heterogeneous patient populations, including patients with node-positive tumors, data regarding node-positive disease are scarce. Onco
type DX was the first prognostic multigene test validated in a large clinical trial of node-positive patients, the SWOG-8814 study, which demonstrated that Onco
type DX was effective in identifying subgroups with fair prognoses, although the likelihood of distant relapse still exceeded 30% in the low-risk group [
9]. Our present study contributes to the literature by providing additional evidence suggesting that multigene tests can be used in patients with node-positive disease. Furthermore, in contrast to the Onco
type DX–derived data mentioned above, our EP-based results strongly suggest that the EP-based low-risk group has a considerably low residual risk of distant recurrence (<10%) after standard chemotherapy. In contrast, high-risk patients have a high probability of exhibiting residual disease after conventional chemotherapy and should be considered for a tailored extension or intensification of adjuvant treatment, as well as for registration in clinical trials testing novel treatment strategies.
As a secondary aim in this study, we analyzed whether EP could be used to identify patients who might benefit from the addition of weekly paclitaxel to anthracycline-based chemotherapy treatment. Taxanes, although they are more toxic than conventional anthracycline-based regimens, are one of the most active cytotoxic agents and are widely used in standard chemotherapy. However, the absolute benefit of taxane-based treatment is small (3% to 7% absolute OS benefit within 5 years after treatment) [
16,
38‐
41] and needs to be balanced against serious side effects. Thus, there is an urgent need for novel predictors regarding which subgroups of patients stand to benefit substantially from treatment with taxanes. Of the several molecular markers analyzed as predictors of taxane efficacy [
42‐
44], none of the markers or prognostic multigene assays have indicated that taxane benefits are maintained across relevant patient subgroups [
10,
11]. Similarly, our present results show that EP test results were not predictive of the efficacy of taxanes, because patients receiving weekly paclitaxel in addition to anthracycline-based chemotherapy did not experience any significant benefit. However, these results should be interpreted in the context of the main limitation of our study: a small sample size. Sample size is especially important in the validation of a predictive marker in patients with ER+/HER2− tumors, which are generally less chemoresponsive [
45] and less effective in taxane-based treatment [
16] compared with patients with other tumors.
The mechanistic determinants of taxane treatment response are poorly understood and still a matter of debate. Recently, we reported that a low proliferation score (11 cell-cycle genes) was predictive of weekly paclitaxel efficacy in the GEICAM 9906 trial [
11]. Although the question whether this association also exists in patients with ER+/HER2− tumors was not examined, the association might be relevant in this context because proliferation is one of the strongest variables in calculating the EP test score, as well as in other available prognostic multigene tests. In contrast, the proliferation marker Ki67 was identified as predictive in luminal tumors treated with chemotherapy that included docetaxel [
44], and patients with luminal B tumors benefited from taxane-based treatment [
44,
46]. The aforementioned conflicting results suggest that there are still a number of key processes in need of further study to elucidate the interaction between markers, taxane agents (paclitaxel and docetaxel), doses applied, administration schedules and treatment strategy duration.
Competing interests
JCB declares receiving salary from Sividon Diagnostics GmbH and holding a patent application related to the content of this article. KK and KF declare receiving salary from Sividon Diagnostics GmbH. KEW, RK and CP declare receiving salary from Sividon Diagnostics GmbH, holding shares in Sividon Diagnostics GmbH and holding a patent application related to the content of this article. The rest of the authors declare that they have no competing interests.
Authors’ contributions
MM was involved in the conception and design of the study; acquisition, assembly analysis and interpretation of data; and drafting the manuscript. JCB participated in the conception and design of the study, designed the gene expression experiments, participated in the analysis and interpretation of data as well as statistical analysis and drafted the manuscript. KK designed and carried out the gene expression experiments, participated in the analysis and interpretation of data and was involved in drafting the manuscript. KF participated in the design of the study and in the statistical analysis and revised the manuscript critically. KEW participated in the design of the study and in the statistical analysis and revised the manuscript critically. RK participated in the conception and design of the study, participated in the analysis and interpretation of data and helped with drafting the manuscript. CP participated in the conception and design of the study, participated in the interpretation of data and helped with drafting the manuscript. LC, MRB, AR, BM, CR, CC, EA and ARL were involved in the acquisition and assembly of data and in the critical revision of the manuscript. MC completed statistical analyses and interpretation of data and critically revised the manuscript. EC and RC participated in the design and coordination of the study and helped with drafting the manuscript. All authors read and approved the final manuscript.