Initial Trends in the Use of the 21-Gene Recurrence Score Assay for Patients With Breast Cancer in the Medicare Population, 2005-2009

Importance  In 2006, the Centers for Medicare & Medicaid Services approved coverage for the use of the 21-gene recurrence score (RS) assay in women with early-stage, estrogen receptor–positive, node-negative breast cancers to help guide recommendations for adjuvant chemotherapy. Use of the assay in community settings has not been previously examined in a nationally representative sample of patients.

Objective  To examine trends in the use of the RS assay in routine clinical practice in a nationally representative sample of women with breast cancer.

Design, Setting, and Participants  Retrospective observational study of Medicare beneficiaries diagnosed with incident breast cancer between 2005 and 2009, as recorded in a Surveillance, Epidemiology, and End Results data set with linked Medicare claims through 2010.

Main Outcomes and Measures  Demographic and clinical variables associated with the use of the assay.

Results  A total of 70 802 patients met the study criteria. Use of the RS assay increased from 1.1% in 2005 to 10.1% in 2009 (P < .001). The majority of tests (60.9%) occurred in patients with National Comprehensive Cancer Network–defined intermediate-risk disease (ie, estrogen receptor–positive, node-negative tumors >1 cm). Most patients with other than intermediate-risk disease had borderline indications for testing, including T1b (47.5%) or N1 (26.8%) disease. Testing was associated with younger age, fewer comorbid conditions, higher-grade disease, and being married. Among patients younger than 70 years with intermediate-risk disease, testing rates increased from 7.7% in 2005 to 38.8% in 2009 (P < .001). In multivariable analysis, testing was modestly higher in Northeast than in Western registries (odds ratio, 1.83; 95% CI, 1.49-2.26) but was otherwise not associated with region, local census tract demographic characteristics, black race, location in an urban area, or tumor histologic characteristics.

Conclusions and Relevance  The RS assay was adopted quickly in clinical practice after the Medicare coverage decision in 2006, and use appears to be consistent with guidelines and equitable across geographic and racial groups. Factors influencing adoption of the assay and its impact on adjuvant chemotherapy use in clinical practice remain important areas of study.

Adjuvant chemotherapy following initial surgical management, in addition to hormone therapy, has an overall survival benefit for women with breast cancer, and use of standard clinicopathological factors has not identified groups of women for whom chemotherapy does not confer this benefit.^1,2 Accordingly, guidelines recommend consideration of chemotherapy in estrogen receptor (ER)-positive, lymph node (LN)-negative disease for all but the smallest of tumors.³ Over the past decade, multigene assays have been developed to improve risk stratification and identify patients most likely to benefit from adjuvant chemotherapy.^4,5 Remaining questions include how frequently these tests are used and how they influence use of chemotherapy and patient outcomes.

The Oncotype DX 21-gene recurrence score (RS) assay (Genomic Health Inc) has been commercially available since 2004. It is used to assign patients with ER-positive, LN-negative breast cancer an RS that predicts the risk of developing metastatic disease and the expected benefit of adjuvant chemotherapy.^6,7 Current guidelines recommend use of the assay to identify patients at low risk of developing metastatic disease who may forgo chemotherapy, owing to the marginal benefit, as well as patients at high risk for whom the potential benefit of adjuvant chemotherapy is more substantial.^3,8,9

In 2006, a Medicare local coverage decision approved the use of the RS assay within 6 months of diagnosis in patients with ER-positive, LN-negative stage I or II breast cancer who are considering treatment with tamoxifen or aromatase inhibitors.¹⁰ Nevertheless, it is unclear to what extent the assay has been adopted in routine clinical practice, whether its adoption has been equitable across demographic groups and geographic regions, and whether its use is consistent with current guidelines. We used a Surveillance, Epidemiology, and End Results (SEER) and Medicare linked data set to examine early trends in the use of the 21-gene RS assay between 2005 and 2009.

Box Section Ref ID

The institutional review board of the Duke University Health System approved this retrospective observational study, waiving written informed consent of all participants.

The SEER-Medicare linked data set used for this analysis was obtained in May 2013. SEER-Medicare is a collaborative effort between the National Cancer Institute and the Centers for Medicare & Medicaid Services that links routinely collected population-based data from SEER cancer registries to Medicare administrative claims data. SEER data represent approximately 25% of the US population with cancer. SEER-Medicare data have been used to examine factors that affect the quality of cancer care.^6-8 Medicare provides health insurance for 97% of persons 65 years or older in the United States, and Medicare claims data reflect health care services used and comorbid conditions. For this analysis, we used SEER data from 2005 to 2009 to identify the study cohort and SEER-linked Medicare claims from 2004 to 2010 to confirm Medicare enrollment and identify use of the RS assay. We used claims from the year before diagnosis to identify comorbid conditions.

From the 12 SEER registries that were continuously active from 2000 onward, we identified all patients diagnosed with breast cancer between 2005 and 2009. To identify patients likely to have complete Medicare claims related to breast cancer management, we required that patients have a primary diagnosis of breast cancer (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 174 or 175) recorded on an inpatient, outpatient, durable medical equipment, hospice, home health, or insurance-based Medicare claim between 2 months before and 4 months after the SEER-reported diagnosis. We included patients 66 years or older at diagnosis. We excluded patients who were diagnosed at autopsy. We also excluded patients who had discontinuous Medicare Part A and Part B coverage (ie, fee-for-service Medicare) or were enrolled in Medicare Part C (managed care) from the year before diagnosis through the end of the study period or death.¹¹

To examine use of the RS assay among patients for whom it was clinically indicated, we performed a prespecified subgroup analysis limited to patients who would be recommended to receive the test according to studies by Paik and colleagues^6,7 and guidelines of the American Society of Clinical Oncology.⁸ These studies and guidelines recommend use of the assay in patients with invasive, ER-positive, LN-negative tumors larger than 1 cm or larger than 0.5 cm with either poor differentiation or adverse histologic features (ie, angiolymphatic invasion, high nuclear grade, or high histologic grade).⁸ Patients with tumors 5 cm or larger were also excluded from the intermediate-risk category. Human epidermal growth factor receptor 2 status was not available for the analysis.

Use of the RS assay was the primary study outcome and was detected with an algorithm developed for this study. We collected insurance file claims from 2 months before to 6 months after the SEER-reported month of diagnosis to check for use of the assay. For each patient, we identified each instance of Current Procedural Terminology (CPT) code 84999 (ie, unlisted chemistry procedure), sorted by payment, and searched for the dollar amount, provider identification (ID), and zip code. We observed that the most expensive test for many patients shared the identical performing provider ID, corresponding to Genomic Health Inc, the only provider of the RS assay. More than 95% of these claims had an exact payment of $3414. We used the first date of service in the event of multiple claims, and we assumed that zero-payment claims represented tests that were performed but not reimbursed by Medicare. Similar methods identifying use of the RS assay in Medicare claims data have been published recently.¹² We identified receipt of chemotherapy using relevant CPT codes (Q0083-Q0085, G0355-G0363, J8510-J9999, 96400-96549) reported on any outpatient or carrier file claim.¹³

We used carrier claims to determine the unique physician ID number of the referring physician. Each physician’s practice zip code was assigned using the zip code most frequently associated with claims filed by that physician. Although we initially proposed to study use of the RS assay at the level of hospital referral region, overall use was too low to provide adequate statistical power to detect differences across regions. Instead, we aggregated analyses at the geographic region level to examine geographic variation. We obtained demographic variables from the SEER Patient Entitlement and Diagnosis Summary File, including age at diagnosis, sex, race, ethnicity, marital status, and local census tract characteristics (ie, urban-rural classification; proportion of the population that did not finish high school; proportion of the population below the poverty line; and proportion of the population reporting black race). We grouped the 12 SEER registries by census region: Northeast (Connecticut), Midwest (Detroit and Iowa), South (Atlanta and rural Georgia), and West (Hawaii, Los Angeles, New Mexico, San Francisco, San Jose, Seattle, and Utah). We identified National Cancer Institute–defined comorbid conditions using inpatient, outpatient, and carrier Medicare claims¹⁰ in the year before diagnosis. We examined the distribution of common comorbid conditions (ie, those with greater than 2% frequency) at 1 year before diagnosis.

We describe the baseline characteristics of the study population by year of diagnosis using frequencies and percentages for all variables and Cochran-Mantel-Haenszel tests of nonzero correlation for temporal trends. We also separately describe the baseline characteristics of patients with RS-defined intermediate risk. We used χ² tests to compare patients who received the RS assay with those who did not. Our multivariable modeling strategy was similar to the strategy our research group used in a previous analysis of the SEER-Medicare data.^14,15 Specifically, we used generalized multivariable regression models with a binary distribution and a logit link to examine factors associated with the primary outcome of interest (ie, use of the RS assay). We used adjusted standard errors (SEs) to account for overdispersion due to clustering of patients by SEER registry. We included all nonoverlapping clinical and demographic patient characteristics described herein to explore factors associated with use of the RS assay to characterize which patient groups were most likely to receive the assay. Because of the number of hypotheses being tested and the large sample size, we set α = .001 to determine statistical significance and used 2-sided tests for all comparisons. We plotted assay use by year of diagnosis for the overall study population and for patients with intermediate-risk disease for whom testing was recommended, then stratified by age group. We used SAS software, version 9.2, for all analyses (SAS Institute Inc).

We identified 70 802 Medicare beneficiaries diagnosed with breast cancer in the SEER registries between 2005 and 2009 (eTable in the Supplement). The RS assay was used for 3684 patients (5.2%). During the same period, the proportion of patients diagnosed with ER- and progesterone receptor (PR)-positive tumors increased from 70.7% to 77.6% and 57.5% to 65.5%, respectively, and the number of patients with ambiguous or unreported receptor status declined. Other demographic and clinical characteristics remained relatively stable and showed similar trends in a sensitivity analysis of patients with stage I to II, non–ER-negative, non–PR-negative disease.

In the overall study population, use of the RS assay increased from 1.1% to 10.1% between 2005 and 2009. Sixty-one percent of assays were performed for 25.6% of patients in whom the test was recommended by National Comprehensive Cancer Network (NCCN) guidelines (ER-positive, LN-negative, and invasive tumors larger than 1 cm), with 26% of all patients with intermediate-risk disease receiving the assay by 2009, compared with 10.1% of the overall study population and 2.7% of patients without intermediate-risk disease (Figure, A). Among the 1442 tests performed for patients without intermediate-risk disease, 685 (47.5%) had T1b disease (0.5-1.0 cm), and 386 (26.8%) had N1 disease (1 to 3 axillary lymph nodes). Few patients had ER-negative (5.2%), T1a (6.4%), Tis (1.0%), N2 or N3 (1.5%), or distant metastatic (≤0.7%) disease. Among patients with invasive, non–intermediate-risk, localized, ER-positive disease, relatively few patients with T1a (2.9% [86 of 2991]), T1b (6.7% [566 of 8441]), or N1 (4.9% [369 of 7487]) disease received the assay. Use of the assay in patients with T1b tumors was significantly higher in the South (9.4%) than in the Midwest (5.1%), West (5.5%), and Northeast (7.1%) (P < .001). In contrast, rates of testing among patients with T1a or N1 disease did not vary by geographic region (P = .52 and P = .41, respectively). Use of the assay among patients with T1b disease was more frequent for patients with higher-grade tumors or LN-positive disease. Among patients with N1 disease, lower stage, smaller tumor size, and lower tumor grade were associated with use of the assay (P < .001). Among patients with N1 disease for whom the assay was used, 95% had stage II disease.

Because we observed use of the RS assay predominantly among patients with intermediate-risk disease, we limited further analysis to this population (n = 18 128; Table 1). Compared with patients who did not receive the test, patients who did receive it were younger (49.8% vs 20.9% aged 66-70 years; P < .001), had fewer comorbid conditions (9.6% vs 16.3% with multiple comorbid conditions; P < .001), lived in census tracts with lower rates of poverty (20.4% vs 23.7% in the highest quartile of poverty; P < .001), and were more likely to be married (55.7% vs 40.9%; P < .001). They also were more likely to reside in the Northeast (24.0% vs 20.5%; P < .001). There was no difference in the proportion of black patients among those who received the test (5.7%) and those who did not (5.9%; P = .69). Use of the assay was associated with higher-grade disease (P < .001) but was otherwise not associated with differences in overall disease stage, tumor size, or PR status.

Table 2 lists the results of the multivariable analysis of the use of the RS assay among patients with intermediate-risk disease. Compared with patients aged 66 to 70 years, older patients had significantly lower odds of receiving the test (P < .001). The presence of multiple comorbid conditions was associated with 36% lower odds of receiving the test. Patients with PR-positive tumors had 18% lower odds. Factors associated with greater odds of testing included being married (odds ratio [OR], 1.19; 95% CI, 1.08-1.32), living in the Northeast compared with the West (OR, 1.83; 95% CI, 1.49-2.26), tumor larger than 2 cm (OR, 1.29; 95% CI, 1.16-1.43), and disease of histologic grade G2 (OR, 1.39; 95% CI, 1.23-1.57) or G3 (OR, 1.28; 95% CI, 1.16-1.41) compared with G1 (P < .001 for all comparisons). Use of the assay was not associated with local census tract characteristics, race, location within an urban area, or tumor histologic type. Among patients aged 66 to 70 years, testing became more common over the study period, increasing from 7.7% in 2005 to 38.8% in 2009 (Figure, B). Among patients older than 80 years, testing remained infrequent, reaching only 4.0% by 2009.

In the overall study population (N = 70 802), rates of chemotherapy use remained similar between 2005 (16.2%) and 2009 (15.9%). In patients with intermediate-risk disease (n = 18 218), we observed a nonsignificant increase in chemotherapy use between 2005 (8.2%) and 2009 (10.0%).

To our knowledge, our study is the first nationally representative analysis of the adoption of the 21-gene RS assay in a community setting within the Medicare breast cancer population. We found that use of the assay was largely restricted to the populations for which it was initially approved in 2005, namely women with ER-positive, LN-negative, stage I or II breast cancer for which the test is most informative regarding the potential benefit of adjuvant chemotherapy.^3,8,9 Consistent with expectations,^8-10 we observed that younger patients, patients with fewer comorbid conditions, and patients with larger tumors had greater odds of receiving the test. Conversely, use of the assay was least likely among older patients with multiple comorbid conditions, a population with lower overall life expectancy in which the potential benefit of chemotherapy is considerably lower.¹ Among patients without NCCN intermediate-risk disease, most patients had borderline indications for testing, including N1 or T1b disease, suggesting that testing was largely consistent with guidelines and clinical practice recommendations.

However, a recent retrospective analysis of the Southwest Oncology Group phase 3 trial SWOG-8814 suggests that the predictive and prognostic value of the RS assay may also apply to patients with ER-positive, LN-positive disease.¹⁶ German case series have also reported use of the assay in women with LN-positive disease.¹⁷ Together, these results suggest the potential for the assay to be recommended in patients with NCCN high-risk disease. At present, whether a favorable RS would or should forgo chemotherapy in patients with LN-positive disease remains an area of ongoing debate.

In the present study, we did not observe evidence of differential receipt of testing with respect to black race or local socioeconomic factors in adjusted analyses, which suggests relatively homogeneous adoption of the technology. Both in the overall study population and in patients with intermediate risk, we did not observe changes in rates of chemotherapy use over the study period. However, further investigation is warranted to examine the association between use of the RS assay and chemotherapy in more detail.

Although use of the RS assay was highest in the Northeast, it was otherwise relatively uniform with regard to patient demographic and clinical characteristics. We propose 2 potential explanations for why nonclinical (ie, racial or geographic) variation in assay use has been less prominent than in the use and adoption of other emerging medical technologies.^18-20 First, approval of the assay included precise clinical indications, including ER-positive, LN-negative tumors in women considering adjuvant chemotherapy. This contrasts with receipt of positron emission tomography (PET) imaging, which by 2005 was approved for virtually any cancer indication.²¹ Second, the assay is a “send-off” test, in contrast to advanced imaging, robotic surgery, and other procedure-based technologies that require significant up-front investment in infrastructure. Such expensive technologies would be expected to pose a more significant barrier to adoption in resource-limited settings and may be affected by profit incentives such as self-referral, which have been demonstrated in the use of magnetic resonance imaging, PET imaging,²² and intensity-modulated radiation therapy for prostate cancer.²³ We did, however, observe higher rates of assay use in the South in patients with non–intermediate-risk T1b disease, suggesting that use of the assay that is not concordant with guidelines may be subject to greater geographic or other nonclinical variation. Greater use of the assay among patients with low-risk disease in the South and overall in the Northeast warrants further investigation.

To our knowledge, only 1 prior study has examined overall use of the RS assay. Hassett et al²⁴ analyzed data from 17 medical centers (14 academic cancer centers across the United States and 3 community hospitals in Michigan) and observed increases in testing rates from 15% in 2006 to 28% in 2008 in a large cohort of more than 7000 patients. These rates were higher than those observed in our study, likely because of differences in patient selection at predominantly academic medical centers and younger patient age (62% of patients were younger than 60 years). The study found that black race and lower education were associated with lower rates of assay use after adjustment for insurance status, education level, and other clinical and demographic factors similar to those used in our study. However, in our study, we did not observe either variable to be significantly associated with lower use among women with intermediate-risk disease in whom testing was indicated. In comparison with our larger, nationally representative series, most smaller clinical case series targeted patients who were 10 to 20 years younger than the patients in our study, although otherwise clinically similar, with predominantly ER-positive, LN-negative, T1b to T1c, or T2 disease.^4,17,25-28

Although we did not examine costs associated with use of the RS assay, the potential impact of both appropriate and inappropriate testing on costs is a critical component of understanding and optimizing cancer care in real-world practice. By 2009, the assay was being used in approximately 10% of patients. Although the clinical value of the assay in these patients is beyond the scope of the present study, the direct payments associated with these studies averaged over all patients with breast cancer was $341 per patient, or upwards of $75 million nationally, providing an initial rough estimate of the financial implications of testing.

Our study has limitations. Only tests submitted to Medicare could be detected in the analysis. It is unlikely, but unknown, whether many tests among Medicare beneficiaries would be paid by third-party insurers or by patients themselves out of pocket. Patients in the SEER registry overall are more likely to be nonwhite, to live in high-poverty areas, and to live in urban areas,¹¹ and the examined population was limited to age 66 or older. All of these factors may impact the generalizability of our findings. Finally, the years of data available represent the initial adoption of the RS assay; the test only became available in 2005, and the SEER-Medicare data were available only through 2009 at the time of the analysis.

In a nationally representative sample of Medicare beneficiaries, clinical factors associated with use of the 21-gene RS assay were consistent with guidelines and smaller single or multi-institutional clinical series reported for the same time period. We observed only modest regional variation and no significant racial variation in testing. However, testing rates varied significantly by age and disease stage, suggesting that optimal testing rates may vary by patient subgroups. Current guidelines recommend testing as an option for patients considering adjuvant chemotherapy. However, because the guidelines do not mandate or uniformly recommend the assay for all such patients, it is difficult to evaluate whether the testing rates we observed represent adequate utilization. Further study is warranted in patients with breast cancer who are not included in the SEER-Medicare database, particularly younger women for whom the factors affecting chemotherapy use and assay use may differ from those observed in our study. Evolving clinical paradigms of clinical management and testing indications, including the use of the assay in node-positive disease, and their impact on costs, chemotherapy use, and outcomes at the national level remain important areas of study.

Accepted for Publication: January 6, 2015.

Corresponding Author: Michaela A. Dinan, PhD, Duke Clinical Research Institute, Duke University School of Medicine, PO Box 17969, Durham, NC 27715 (michaela.dinan@duke.edu).

Published Online: March 5, 2015. doi:10.1001/jamaoncol.2015.43.

Author Contributions: Dr Dinan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Dinan, Reed, Hirsch, Lyman.

Acquisition, analysis, or interpretation of data: Dinan, Mi, Reed, Lyman, Curtis.

Drafting of the manuscript: Dinan, Lyman.

Critical revision of the manuscript for important intellectual content: Dinan, Mi, Reed, Hirsch, Lyman, Curtis.

Statistical analysis: Dinan, Mi, Reed.

Obtained funding: Dinan.

Administrative, technical, or material support: Curtis.

Study supervision: Dinan, Reed.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by grant K99HS022189 from the Agency for Healthcare Research and Quality (AHRQ).

Role of the Funder/Sponsor: The AHRQ had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Damon M. Seils, MA, Duke University, provided editorial assistance and prepared the manuscript for this article. Mr Seils did not receive compensation for his assistance beyond that received in the normal course of his employment. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare & Medicaid Services; Information Management Services Inc; and the SEER Program tumor registries in the creation of the SEER-Medicare database.

Disclaimer: While this study used the linked SEER-Medicare database, the interpretation and reporting of these data are the sole responsibility of the authors.