Clinical significance of the 21-gene signature (Oncotype DX) in hormone receptor-positive early stage primary breast cancer in the Japanese population


  • The following are members of the Japan Breast Cancer Research Group-Translational Research Group: Takayuki Ueno, MD, PhD (Department of Breast Surgery, Graduate School of Medicine, Kyoto University, Kyoto, Japan) and Kenjiro Aogi, MD, PhD (Department of Breast and Endocrine Surgery, National Shikoku Cancer Center, Matsuyama, Japan).



The 21-gene signature has been intensively studied and incorporated into major guidelines for treatment decision in early breast caner. However, it remains to be examined whether this system is applicable to Asian populations.


The authors collected 325 tumor tissues from estrogen receptor (ER)-positive primary breast cancer patients who had undergone surgery and were treated with tamoxifen between 1992 and 1998. The tissues were analyzed for the 21-gene signature, and the patients were classified into groups of low, intermediate, or high risk based on the Recurrence Score.


A total of 280 patients were eligible, with adequate reverse transcription polymerase chain reaction profiles for the Recurrence Score. Of those, 200 and 80 patients had lymph node-negative and lymph node-positive disease, respectively. The proportions of lymph node-negative patients categorized as being at low, intermediate, or high risk were 48%, 20%, and 33%, respectively. In lymph node-negative patients, the Kaplan-Meier estimates of the distant recurrence rate at 10 years were 3.3% (95% confidence interval [95% CI], 1.1-10.0%), 0%, and 24.8% (95% CI, 15.7-37.8%) for those in the low-risk, intermediate-risk, and high-risk groups, respectively. The risk of distant recurrence in the low-risk group was significantly lower than that in the high-risk group when the entire Kaplan-Meier plots were compared (P < .001, log-rank test). There was a significant difference for overall survival between the low-risk and the high-risk groups (P = .008, log-rank test).


This is the first report to show that the 21-gene signature has value in providing prognostic information in Asian populations with ER-positive, lymph node-negative breast cancer. Cancer 2010. © 2010 American Cancer Society.

Breast cancer incidence continues to rise in Japan. 1 Moreover, although the breast cancer mortality rate has improved in Western countries, it is still increasing in Japan and other Asian countries. It is urgent to establish a system to provide efficient treatment options, as well as to propagate the screening system.

Introduction of systemic therapies in the management of operable breast cancer has contributed to improved prognosis of breast cancer patients. An efficient combination of systemic therapies with local therapy is essential to improve survival. It is now well recognized that hormone receptor-positive breast cancer has a large heterogeneity in biological characteristics, which makes the treatment decision, especially application of chemotherapy, complicated in this cancer type.

Gene expression profiles are under intensive study to provide information on the prognosis and the prediction of treatment benefit in early stage breast cancer. The 21-gene signature, Oncotype DX (Genomic Health, Redwood City, Calif), has been developed as a reverse transcription polymerase chain reaction (RT-PCR)-based gene expression assay to predict prognosis and benefits from chemotherapy for hormone receptor-positive breast cancers. 2, 3 On the basis of tumors that were estrogen receptor (ER) positive and lymph node negative, treated with tamoxifen alone in the National Surgical Adjuvant Breast and Bowel Projects (NSABP) Study B-14,4 the Recurrence Score was developed to classify patients into 1 of 3 categories: low, intermediate, and high risk of distant disease recurrence. In a multivariate analysis, the Recurrence Score provided significant predictive power that goes beyond age and tumor size. Studies conducted subsequently have confirmed that the Recurrence Score predicts the likelihood of breast cancer recurrence and have shown that it is predictive of the benefit of chemotherapy, specifically cyclophosphamide, methotrexate, and fluorouracil (CMF).5, 6 American Society of Clinical Oncology (ASCO)7 and National Comprehensive Cancer Network (NCCN)8 included the 21-gene signature assay in their guidelines for lymph node-negative, ER-positive breast cancer in 2007 and 2008, respectively. It has been reported that the Recurrence Score is also prognostic for hormone receptor-positive, postmenopausal tamoxifen-treated patients with positive lymph nodes, and that chemotherapy provides little benefit for patients with a low Recurrence Score, despite the presence of positive lymph nodes.9 A recent study showed that the Recurrence Score was an independent predictor of the risk of distant recurrence in postmenopausal women from a more contemporary population, primarily from the United Kingdom, with lymph node-negative and lymph node-positive hormone receptor-positive disease, who were treated with either tamoxifen or anastrozole,10 showing that the Recurrence Score could be useful for predicting the survival outcome regardless of the type of hormone therapy.

The 21-gene signature assay is now widely used for treatment decision in the United States. However, to date, there are few data available on the Recurrence Score for patients in the Japanese (or Asian) population. Given the increasing incidence of early breast cancer in Japan (and other Asian countries), it is valuable to know whether the prediction system for recurrence and chemotherapy impact can be used for that heritage. The objective of this study was to examine the value of the Recurrence Score for prediction of prognosis in a population of Japanese patients.


Study Population

Patients satisfying the following criteria were enrolled in this study: patients who had an early stage, ER-positive breast cancer; received surgery and adjuvant tamoxifen treatment between 1992 and 1998; and had a formalin-fixed paraffin-embedded tissue sample available. Early stage patients were defined as those with a clinical stage of T1-T2 cancer. The pathological method for nodal evaluation followed the Japanese guideline at that time, which is basically the International Union Against Cancer guideline with minor modifications for Japanese patients. 11 Each center participating in this study was required to register all patients during 1992-1998 unless the criteria were not met. The study was approved by the institutional review boards of all participating centers. Clinical data on patients were collected by the Japan Breast Cancer Research Group. The gene assay was performed for fixed paraffin-embedded tissues, which were submitted by the Japan Breast Cancer Research Group to Genomic Health. All laboratory data were obtained masked to clinical outcomes.

Study Endpoints

The associations between the Recurrence Score and distant recurrence-free interval (RFI), RFI, recurrence-free survival (RFS), and overall survival (OS) were evaluated. According to previous studies, 4, 12 distant RFI, the time from surgery to first distant recurrence, was the primary endpoint. Ipsilateral breast recurrence, local chest wall recurrence, and regional recurrences were not considered either as events or as censoring events; contralateral disease, other second primary cancers, and deaths before distant recurrence were considered as censoring events.

RFI was defined as the time from surgery to first locoregional or distant recurrence; contralateral disease, other second primary cancers, and deaths before recurrence were considered as censoring events. RFS was defined as the time from surgery to recurrence or death from any cause. OS was defined as the time from surgery to death from any cause.

Gene Expression Analysis by Oncotype DX

The Recurrence Score was determined from fixed paraffin-embedded tissue by Genomic Health, as previously described. 4 After review of hematoxylin and eosin-stained slides to determine whether sufficient invasive breast cancer was present and whether manual microdissection was indicated, RNA was extracted from 3 or 6 10-μm unstained sections. Total RNA content was measured, and the absence of DNA contamination was verified. Gene expression profiling by standardized quantitative RT-PCR for the assay was performed using the prespecified 21 genes. Reference-normalized expression measurements for each of the 16 cancer-related genes ranged from 2 to 16, where each 1-unit increase reflects approximately a 2-fold increase in RNA. Group scores were developed from individual reference-normalized gene expression measurements for the HER-2 group (HER-2, GRB7), ER group (ER, PR, Bcl2, SCUBE2), proliferation group (Ki-67, STK15, Survivin, Cyclin B1, MYBL2), and invasion group (stromelysin 3, cathepsin L2). The Recurrence Score was derived as a function of these 4 group scores along with the individual reference-normalized expression measurements for CD68, BAG1, and GSTM1. The Recurrence Score is presented on a scale from 0 to 100, and the risk groups for distant recurrence were prospectively defined as low (Recurrence Score <18), intermediate (Recurrence Score 18-30), and high (Recurrence Score ≥31).

Statistical Analyses

The prespecified primary analysis was to test the difference of the Kaplan-Meier curves for distant RFI between the low Recurrence Score group and the high Recurrence Score group by log-rank test. The secondary analysis was to determine the relationship between distant RFI and the Recurrence Score using a Cox proportional hazards model. The likelihood ratio test compared the competing full model, based on age, clinical tumor size, and the Recurrence Score, with the reduced model that excluded the Recurrence Score. The Recurrence Score was used as a continuous variable in the Cox model, with hazard ratio (HR) for distant RFI calculated relative to an increment of 50 units to facilitate comparison with previous results. 4 The 10-year rates of distant RFI and other endpoints were assessed using the Kaplan-Meier method, and the 95% confidence intervals (CIs) were calculated based on the Greenwood bounds on the Kaplan-Meier estimate. A P value of <.05 was considered to indicate a significant result, and all P values were reported as 2-sided. Statistical analyses were performed using SAS for Windows (version 9.1.3). All data analyses were performed independently by the Japan Breast Cancer Research Group, with assay results provided by Genomic Health.


Characteristics of Study Population

Eight centers participated in this study, and they had 325 patients in total who were diagnosed as ER-positive, early stage breast cancer, had surgery, and received adjuvant tamoxifen treatment between 1992 and 1998. Of these, 12 subjects had little or no clinical data available. Twenty-seven samples were deemed ineligible by pathology evaluation, 1 had insufficient RNA (<275 ng), and 5 failed RT-PCR. Therefore, there were 280 eligible patients with adequate RT-PCR profiles. Of these, 200 were lymph node negative, 61 had 1 to 3 positive lymph nodes, and 19 had 4+ positive lymph nodes. The remainder of this report focuses on the 200 lymph node-negative patients; that is, hereafter, we study patients with lymph node-negative, ER-positive breast cancer who received surgery and adjuvant tamoxifen.

The distributions of patients by age, clinical tumor size, nuclear grade, and Recurrence Score group are displayed in Table 1. Approximately two-thirds of the patients were aged ≥50 years, and 54% had tumors >2 cm in diameter. The distribution of nuclear grade was 15% grade 1, 40% grade 2, and 18% grade 3, with grade unavailable for the remaining 27% of patients. The distribution by Recurrence Score group was 48% low, 20% intermediate, and 33% high. Mastectomy was performed for 143 patients. Two patients had radiation after mastectomy. Of the 57 patients with breast conservation surgery, 44 had radiation therapy. The indication of radiation for breast-conserving surgery had been determined institutionally, based on factors such as margin-status and age.

Table 1. Patient Characteristics by Recurrence Score Group
CharacteristicsLow [RS <18], No. (%)Intermediate [RS 18-30], No. (%)High [RS ≥31], No. (%)All, No. (%)
  1. RS indicates Recurrence Score; RT, radiotherapy.

Total95 (48)40 (20)65 (33)200
Age, y    
 <5031 (33%)15 (38%)22 (34%)68 (34%)
 ≥5064 (67%)25 (63%)43 (66%)132 (66%)
Clinical tumor size, cm    
 ≤251 (54%)17 (43%)24 (37%)92 (46%)
 >244 (46%)23 (58%)41 (63%)108 (54%)
Nuclear grade    
 118 (19%)6 (15%)6 (9%)30 (15%)
 240 (42%)18 (45%)22 (34%)80 (40%)
 37 (7%)8 (20%)21 (32%)36 (18%)
 Unavailable30 (32%)8 (20%)16 (25%)54 (27%)
 Mastectomy61 (64%)31 (78%)51 (78%)143 (72%)
 Breast conservation34 (36%)9 (23%)14 (22%)57 (29%)
 Without RT66 (69%)34 (85%)54 (83%)154 (77%)
 With RT29 (31%)6 (15%)11 (17%)46 (23%)

Recurrence Score and Risk of Distant Recurrence

The Kaplan-Meier plots for distant RFI by Recurrence Score group are presented in Figure 1. In the primary analysis, patients in the Low Recurrence Score group had a significantly lower risk of distant recurrence than patients in the High Recurrence Score group (P < .001, log-rank test). No distant recurrences were observed among the 40 patients in the Intermediate Recurrence Score group; this group did not appear to be at higher risk of distant recurrence than the Low Recurrence Score group.

Figure 1.

Kaplan-Meier plot of distant recurrence-free interval is shown by Recurrence Score group. The numbers of patients at risk are shown at Years 0, 5, and 10. The numbers of distant recurrences occurring during the 5-year intervals are shown in parentheses. Int. indicates intermediate.

In a univariate Cox proportional hazards model of distant RFI, the continuous Recurrence Score was significantly associated with the risk of distant recurrence (HR, 6.20 for a 50-point increase; 95% CI, 2.27-17.0). When adjusted for age (<50 vs ≥50 years) and clinical tumor size (≤2 vs >2 cm) in a multivariate Cox model, the continuous Recurrence Score maintained its statistical significance (HR, 6.03; 95% CI, 2.17-16.7) (Table 2).

Table 2. Multivariate Cox Proportional Hazards Analysis of Age, Clinical Tumor Size, and Recurrence Score in Relation to the Likelihood of Distant Recurrence
ModelHR95% CIP
  1. HR indicates hazard ratio; CI, confidence interval; RS, Recurrence Score.

Analysis without RS   
Age at surgery (≥50 years vs <50 years)0.72(0.29-1.76).469
Clinical tumor size (>2 cm vs ≤2 cm)1.65(0.65-4.19).285
Analysis with RS   
Age at surgery (≥50 years vs <50 years)0.69(0.28-1.70).428
Clinical tumor size (>2 cm vs ≤2 cm)1.38(0.54-3.53).490

Recurrence Score and Other Endpoints

Table 3 contains the Kaplan-Meier estimates (and 95% CI) for the 10-year rates of distant recurrence, recurrence, recurrence or death, and death by Recurrence Score group. The estimated rates of distant recurrence were 3.3%, 0%, and 24.8% at 10 years in the low, intermediate, and high Recurrence Score groups, respectively. The results for other endpoints RFI, RFS, and OS were similar to those for distant RFI; patients in the low Recurrence Score group had significantly lower risks of recurrence, of recurrence or death, and of death from any cause than patients in the high Recurrence Score group. Patients in the low and intermediate Recurrence Score groups experienced relatively low event rates. Cox proportional hazards models, adjusted for age and clinical tumor size, showed that HRs with respect to the continuous Recurrence Score were 3.38 (95% CI, 1.32-8.69) for the risk of recurrence, 2.09 (95% CI, 0.84-5.20) for the risk of recurrence or death, and 2.67 (95% CI, 0.93-7.62) for the risk of death.

Table 3. Kaplan-Meier Estimates of 10-Year Event Rates by RS Group
EndpointEventLow [RS <18], n=95, % (95% CI)Intermediate [RS, 18-30], n=40, % (95% CI)High [RS ≥31], n=65, % (95% CI)
  1. RS indicates Recurrence Score; CI, confidence interval; DRFI, distant recurrence-free interval; NA, not available; RFI, recurrence-free interval; RFS, recurrence-free survival; OS, overall survival.

DRFIDistant recurrence3.3 (1.1-10.0)0 (NA)24.8 (15.7-37.8)
RFIRecurrence5.5 (2.3-12.8)2.5 (0.4-16.5)24.6 (15.6-37.6)
RFSRecurrence or death9.6 (5.1-17.6)5.1 (1.3-18.8)23.4 (14.8-35.9)
OSDeath6.4 (2.9-13.6)2.6 (0.4-16.8)19.1 (11.3-31.3)

Individual Genes, Gene Groups, and Risk of Distant Recurrence

Figure 2 presents a Forest plot of the HRs for all 16 cancer-related genes, estimated from separate univariate Cox models fit to distant RFI in which gene expression, expressed in units of cycle threshold, was included as a single continuous covariate. Each unit of cycle threshold represents approximately a 2-fold change in gene expression. Figure 3 is a Forest plot of the HRs for the gene groups specified in the Materials and Methods section. The proliferation gene group and the HER-2 gene group were both treated nonlinearly as described previously. 4 We found that the 2 plots match those given in the NSABP B-14 clinical validation study.4 The confidence intervals for the 2 studies overlap for all genes and gene groups, indicating consistency between the results of the 2 studies with regard to the influence of the components of Recurrence Score on distant recurrence.

Figure 2.

The estimated hazard ratios and 95% confidence intervals for individual genes in 200 lymph node-negative patients are shown. ER indicates estrogen receptor; PR, progesterone receptor.

Figure 3.

The estimated hazard ratios and 95% confidence intervals for gene groups in 200 lymph node-negative patients are shown. Both linear and nonlinear forms are shown for the proliferation gene group and the HER-2 gene group; the nonlinear forms are used in the calculation of the Recurrence Score as explained in Paik et al. 4 ER indicates estrogen receptor.


To date, the validity of the 21-gene signature assay had not been assessed in Asian populations. This is the first study to confirm the clinical value of the assay in an Asian population. The increase in incidence and mortality of breast cancer in Asian countries necessitates systems to optimize treatment strategies for breast cancer. Thus, it is of great importance to validate the multiple gene assay in Asian countries, because the assay system has been incorporated into Western guidelines such as NCCN, ASCO, and St. Gallen consensus.

The patients in the study were registered from 8 high-volume centers in Japan spread geographically across the country. A standardized quantitative assay was performed, where the methods were prospectively defined, with all laboratory data obtained masked to study endpoints. The results of the study indicate that the assay quantifies the likelihood of distant recurrence in Japanese patients with lymph node-negative, ER-positive breast cancer who were treated with tamoxifen, and may identify certain proportions of women who are at low and high risks of distant recurrence. The distribution of Recurrence Score group in this study (48% low, 20% intermediate, 33% high) is comparable to that of the NSABP B-14 clinical validation study (51%, 22%, and 27%, respectively). 4 The 10-year rates of distant recurrence (and 95% CI) in this study were 3.3% (1.1%-10.0%), 0%, and 24.8% (15.7%-37.8%) in the low, intermediate, and high Recurrence Score groups, respectively, as compared with 6.8% (4.0%-9.6%), 14.3% (8.3%-20.3%), and 30.5% (23.6%-37.4%) in the NSABP B-14 study.

It is notable in the Forest plots of our study that the expression profiles of individual genes and gene groups for Japanese patients were very similar to those for patients in the NSABP B-14 study. 4 The confidence intervals for the 2 studies overlap for all genes and gene groups, indicating consistency between the results of the 2 studies with regard to the influence of the components of the Recurrence Score on the risk of distant recurrence. Thus, our results provide an important finding in support of the clinical utility of the 21-gene signature assay for patients with early stage breast cancer across different ethnic groups.

Regardless of the similar Forest plot patterns between the 2 studies, we did not show a significant difference between intermediate Recurrence Score group and low Recurrence Score group. The low rate of distant recurrence observed among the 40 intermediate Recurrence Score patients in this study is interesting, but the relatively small number of patients precludes deeper investigation in this study. We note that a lack of available clinical and pathology data, including, for example, nuclear grade, margin status, or lymphovascular invasion, is a major limitation in interpreting our results.

In conclusion, it was shown that the 21-gene signature assay identified patients with different risks of distant recurrence in a population of Japanese women with lymph node-negative, ER-positive breast cancer who received surgery and adjuvant tamoxifen. These results indicate that the Recurrence Score is useful in decision making for adjuvant therapy outside of the largely Caucasian populations in which it was originally validated.


We thank the women who participated in these studies and their families, and Kiyomi Kashiwa and Akiko Shiomi for their administrative assistance.


This study was funded by a research grant from the Japanese Ministry of Health, Labor, and Welfare, for “Development of Multidisciplinary Treatment Algorithm With Biomarkers and Modeling of the Decision-making Process With Artificial Intelligence for Primary Breast Cancer” (H18-3JIGAN-IPPAN-007).