Clinical outcomes in ER+ HER2 -node-positive breast cancer patients who were treated according to the Recurrence Score results: evidence from a large prospectively designed registry

The Recurrence Score® is increasingly used in node-positive ER+ HER2-negative breast cancer. This retrospective analysis of a prospectively designed registry evaluated treatments/outcomes in node-positive breast cancer patients who were Recurrence Score-tested through Clalit Health Services from 1/2006 through 12/2011 (N = 709). Medical records were reviewed to verify treatments/recurrences/survival. Median follow-up, 5.9 years; median age, 62 years; 53.9% grade 2; 69.8% tumors ≤ 2 cm; 84.5% invasive ductal carcinoma; 42.0% N1mi, and 37.2%/15.5%/5.2% with 1/2/3 positive nodes; 53.4% Recurrence Score < 18, 36.4% Recurrence Score 18–30, and 10.2% Recurrence Score ≥ 31. Overall, 26.9% received adjuvant chemotherapy: 7.1%, 39.5%, and 86.1% in the Recurrence Score < 18, 18–30, and ≥ 31 group, respectively. The 5-year Kaplan–Meier estimates for distant recurrence were 3.2%, 6.3%, and 16.9% for these respective groups and the corresponding 5-year breast cancer death estimates were 0.5%, 3.4%, and 5.7%. In Recurrence Score < 18 patients, 5-year distant-recurrence rates for N1mi/1 positive node/2–3 positive nodes were 1.2%/4.4%/5.4%. As patients were not randomized to treatment and treatment decision is heavily influenced by Recurrence Score, analysis of 5-year distant recurrence by chemotherapy use was exploratory and should be interpreted cautiously: In Recurrence Score < 18, recurrence rate was 7.7% in chemotherapy-treated (n = 27) and 2.9% in chemotherapy-untreated patients (n = 352); P = 0.245. In Recurrence Score 18–30, recurrence rate in chemotherapy-treated patients (n = 102) was significantly lower than in untreated patients (n = 156) (1.0% vs. 9.7% P = 0.019); in Recurrence Score ≤ 25 (the RxPONDER study cutoff), recurrence rate was 2.3% in chemotherapy-treated (n = 89) and 4.4% in chemotherapy-untreated patients (n = 488); P = 0.521. In conclusion, our findings support using endocrine therapy alone in ER+ HER2-negative breast cancer patients with micrometastases/1–3 positive nodes and Recurrence Score < 18.


INTRODUCTION
The 21-gene Recurrence Score®(RS) assay (Oncotype DX® Breast Recurrence Score assay, Genomic Health Inc), is incorporated in major international guidelines, [1][2][3][4] and is widely used to guide treatment decisions in estrogen receptor (ER)+ human epidermal growth factor receptor 2 (HER2)-negative node-negative breast cancer. The prognostic/predictive abilities of the RS assay in this population were first validated using a prospective-retrospective study design. [5][6][7] More recently, the RS was further validated as a prognostic indicator in the TAILORx and West German Study Group (WSG) PlanB phase 3 trials in which patients were treated based on their RS results, and in the prospectively designed analysis of the Surveillance, Epidemiology, and End Results (SEER) registry, which was complemented with the RS results. [8][9][10] Furthermore, we previously reported findings from an analysis of a prospectively designed registry including 2028 N0/N1mi ER+ breast cancer patients who were tested through Clalit Health Services (CHS, the largest health maintenance organization in Israel), 11 which complemented the TAILORx, WSG PlanB, and SEER studies by providing both rate of recurrence and breast cancerspecific mortality in a cohort of patients where decision making incorporated the RS in real-life clinical practice.
Withholding chemotherapy in ER+ HER2-negative node-positive breast cancer patients remains controversial even in the presence of data from such studies as the ATAC, ABCSG 6, and ABCSG 8 trials, which have demonstrated that many patients with 1-3 positive nodes do well on endocrine therapy alone. [12][13][14] To further help identify patients who can safely forgo chemotherapy, RS testing is increasingly used in patients with up to three positive axillary lymph nodes. Similar to the approach used for the nodenegative population, the initial validation of the prognostic/ predictive abilities of the RS assay in the node-positive population employed a prospective-retrospective study design. 7,15 However, unlike the node-negative population, at present, the additional evidence supporting RS testing in the node-positive population is limited to the aforementioned WSG PlanB study in which all patients with RS > 11 received chemotherapy, and to a secondary analysis of the SEER registry in which only breast cancer-specific mortality has been reported. 9,10 Prospective evidence from the phase 3 RxPONDER (SWOG S1007) trial evaluating endocrine therapy alone vs. chemoendocrine therapy in patients with 1-3 positive nodes and RS ≤ 25 is anticipated in the future. 16 In Israel, CHS approved reimbursement of the RS assay for nodenegative patients in 1/2006 and extended its coverage to include node-positive patients in 1/2008. Since its initial reimbursement approval by CHS and through November 2016, 5483 nodenegative, and 1850 patients with N1mi/1-3 positive nodes have been tested through CHS. The focus of the current CHS registry analysis is on the latter population. This analysis was designed to evaluate, in real-life clinical practice, the relationship between the RS results, adjuvant treatments received, and clinical outcomes in patients with N1mi or 1-3 positive nodes who were RS-tested more than 5 years ago and had their RS results incorporated into clinical decision making. Notably, this study complements another study focusing on node-negative patients who were RS-tested through CHS. 17

Patient characteristics
Between 1/2006 and 12/2011, 755 CHS members with breast cancer and N1mi or 1-3 positive lymph nodes were RS-tested through CHS. Forty patients were excluded for various reasons (received neoadjuvant treatment, were metastatic at diagnosis, etc). For four additional patients (0.5%) recurrence data were not available or they were lost to follow-up, and two more patients were excluded because they recurred within 6 months of testing ( Supplementary Fig. S1). The final cohort included 709 patients and the median follow-up was 5.9 (range, 0.3-10.3) years.

DISCUSSION
This is the first analysis of a large prospectively designed registry investigating both distant recurrence and breast cancer death rates in patients with N1mi (42% in our cohort) or 1-3 positive nodes (58% in our cohort) where the RS has been incorporated into real-life clinical decision making. We found that chemotherapy use was consistent with the RS, and that clinical outcomes were very good in patients with RS < 18 (5-year distant recurrence rate, 3.2%; 5-year breast cancer death rate, 0.5%) and also excellent in the 90.2% of patients with RS < 18 selected to receive endocrine therapy alone (5-year distant recurrence rate, 2.7%; 5year breast cancer death rate, 0.6%). Chemotherapy use was aligned with the RS results. This finding is consistent with the CHS N0/N1mi analysis and with decision impact studies conducted in the N1mi/N1 population worldwide. [18][19][20][21][22][23][24][25][26][27] Within the intermediate RS category (RS 18-30), chemotherapy use increased with higher RS results, consistent with prior Israeli studies including the analysis of the CHS N0/N1mi cohort and with the SEER registry analysis for node-negative and node-positive patients. 11,28,29 This observation suggests that clinicians assess the RS result as a continuous parameter.
Our findings are consistent with those found in the original prospective-retrospective validation studies in node-positive patients, with the analysis of node-positive patients in the SEER registry, and with the results of the prospective WSG PlanB study, which included node positive and high-risk node-negative patients, and as such constitutes the only prospective evidence for the clinical outcomes following RS-based treatment in nodepositive patients (41.2% of the study patients with available tissue within the tumor bank). 7,9,10,15 In WSG PlanB, the RS risk groups were RS ≤ 11, 12-25, and>25 (similar to those in TAILORx in which the lower range was RS ≤ 10). 8,9 Patients in the lower RS range were recommended to omit chemotherapy, whereas all patients with RS ≥ 12 received chemotherapy. After median follow-up of 35 months, 3-year disease-free survival in patients with RS ≤ 11 (chemotherapy untreated, n = 348) was 98% vs. 98% and 92% in those with RS 12-25 and RS > 25, respectively. 9 Distant recurrence rates in WSG PlanB patients have not been reported yet. 9 Our findings are also consistent with the previously reported analysis of the CHS N0/N1mi cohort, 11 as in both cases, chemotherapy use was aligned with the RS results and the risk of recurrence was significantly higher in patients with higher RS results. Not surprisingly, in the RS < 18 and RS 18-30 groups, chemotherapy use was higher in the current cohort than in the corresponding RS groups in the CHS N0/N1mi cohort, and the 5-year risk of distant recurrence was also higher in each of the RS groups (RS < 18, 0.8% vs. 3.2%; RS 18-30, 3.2% vs. 6.3%; RS ≥ 31, 10.3% vs. 16.9%). Notably, in the original validation study in node-positive patients (SWOG 8814), all patients received anthracycline-based chemotherapy; whereas, in WSG Plan B, patients received anthracycline-based or taxane-based chemotherapy. 9,15 In our analysis, as in the previously reported analysis of the N0/N1mi CHS cohort, 11 the chemotherapy used included predominantly anthracycline-based or taxane-based regimens.
Thus far, MINDACT has been the only prospective phase 3 trial reporting findings from randomized arms in which a genomic assay (the 70-gene signature, MammaPrint® by Agendia, Inc) was investigated as an aid to treatment decisions in early-stage breast a b  cancer. 30 This trial enrolled 6693 patients with early-stage breast cancer (79% node-negative and 21% node-positive), of which 644 qualified for the primary analysis. The trial met its primary endpoint as patients who had high risk by clinical characteristics (using Adjuvant!Online) and low risk by the 70-gene signature and were randomized to the non-chemotherapy arm had a 5-year rate of distant recurrence-free survival of 94.7% (the lower boundary of the 95% CI was 92.5%, which is above the predefined boundary of 92%). 30 Notably, in patients (per-protocol population) with high clinical risk and low risk by the 70-gene signature, disease-free survival was significantly better with chemotherapy (93.3% vs. 90.3%, P = 0.03). In patients (per-protocol population) with low clinical risk and high risk by the 70-gene signature, no significant chemotherapy benefit was observed. 30 In contrast to the MINDACT findings, our results (consistent with the prospective-retrospective RS validation studies showing minimal chemotherapy benefit in low RS patients 15 ) strongly support forgoing chemotherapy in patients with ER+ HER2negative breast cancer with micrometastases/1-3 positive nodes and RS < 18 and treating them with endocrine therapy alone. Unlike our previously reported findings for N0 patients, 11 the current analysis of node-positive patients does not support sparing adjuvant chemotherapy in patients with RS 18-25 and suggests that chemotherapy treatment should be recommendedto all patients with RS > 18.

5-year distant recurrence rate
Notably, in this analysis of a real-life, national level large registry, no exclusion criteria were applied with respect to age, gender, location, socioeconomic status or comorbidities. Nevertheless, the registry analysis is limited by its nonrandomized design and the potential for selection bias with respect to patients being selected for RS testing; in Israel the test is less frequently used in the nodepositive population relative to the node-negative population and we estimate that 40% of eligible N1mi/N1 patients are being tested. Another limitation is the generalizability of our findings, as the population in our analysis was mostly Caucasian. As chemotherapy treatment decisions at CHS rely heavily on the RS result, particularly within the RS < 18 and RS ≥ 31 groups, analyses of outcomes by chemotherapy use should be interpreted cautiously. The subgroup analyses should also be interpreted cautiously as some of the groups (e.g., patients with 2 or 3 positive nodes) are small and the event rate is very low. Notably, if the event rate is this low in the RxPONDER trial, where the analysis is event-driven, it may take longer than expected until the data become available. Chemotherapy benefit is expected to be seen in the first years after treatment and our cohort has a median followup of 5.9 years. Further analysis is planned when the cohort has longer median follow-up to assess risk of late recurrences.
In conclusion, in this analysis of prospectively-designed registry assessing distant recurrence and breast cancer death in ER+ HER2negative breast cancer patients with N1mi or 1-3 positive nodes where the RS result has been incorporated in treatment decision making, chemotherapy-untreated patients with RS < 18 had very good clinical outcomes supporting the use of endocrine therapy alone in this population.

Patient population
This retrospective analysis of a prospectively designed registry investigated the relationship between the RS, adjuvant treatments received, and clinical outcomes in all CHS patients with ER+ HER2-negative breast cancer and N1mi or 1-3 positive nodes who were RS-tested between 1/2006 (CHS approval of the assay) and 12/2011. The primary endpoint was distant recurrence and the secondary endpoint was breast cancer death. Exclusion criteria included: ER-negativity by immunohistochemistry (IHC; cutoff: 0.5) and reverse transcription polymerase chain reaction (RT-PCR) performed as part of the 21-gene assay (cutoff: 6.5 units); HER2 positivity by IHC (cutoff: 3), fluorescence in situ hybridization (ratio cutoff: 2.2), or RT-PCR performed as part of the 21-gene assay (cutoff: 10.7 units); receiving trastuzumab in the adjuvant setting; receiving neoadjuvant treatment; having metastatic disease at the time of testing or within 6 months of testing (as the latter were assumed to be metastatic at diagnosis); and receiving adjuvant therapy for another malignancy within 6 months of testing.
This registry analysis was approved by the institutional review boards of the CHS Community Division and participating medical centers, and was granted a waiver for obtaining patient consent. It was conducted in accordance with the Declaration of Helsinki.

Data source
The following data sources were used: The Teva Pharmaceutical Industries Oncotest database (for RS results and patient/tumor characteristics); medical records (for treatments received and recurrence/death status); and the CHS claims arm (for treatments received and death status). Those extracting the clinical information were unaware of the RS status.

Statistical analysis
The statistical analysis plan was pre-specified. Descriptive statistics were used to summarize clinicopathological characteristics and adjuvant treatment decision. The primary endpoint was 5-year KM distant recurrence estimates and 95% CI by RS risk categories as defined by Paik et al. 5 (<18, 18-30,≥31). Patients without recurrence were censored at the time of last follow-up, date of medical records review, or at time of death (due to any cause). A secondary endpoint was 5-year KM breast cancer death estimates and 95% CI. Patients with metastatic disease at the time of death were considered events, patients were censored at the time of last follow-up or death from other causes, and recurrences were ignored for purposes of this endpoint. Analysis was also performed on clinicopathological subgroups, as well as within the intermediate RS group (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30), using a cutoff of 25 as in the RxPONDER trial. 16 As exploratory analysis, 5-year KM distant recurrence estimates and 95% CI by chemotherapy use within RS risk categories as defined by Paik et al. 5 and within the intermediate group (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) were examined. A 2degrees of freedom log-rank test was used to compare distant recurrence rates across RS groups. Survival curves are presented through 7-years, though all observed events and follow-up are utilized for statistical tests. Cox proportional hazards regression models were used to evaluate the association of RS group (<18, 18-30, ≥31), age (50-69 vs. <50; ≥70 vs. <50 years), size (<2, ≥2 cm), grade (1-3), and nodal status (N1mi vs. 1-3 positive nodes) with distant recurrence. RS and tumor size were also included as continuous variables. Univariate analysis identified a set of prognostic baseline factors, and a full multivariable model was fit. Non-significant covariates were removed from the multivariate analysis. Proportional hazards assumptions were assessed and met in the final model. SAS 9.4 (SAS Institute Inc., Cary, NC) was used for the analysis. P < 0.05 was considered statistically significant.

Data availability
All relevant data are available upon request from the corresponding author.