Background

The 21-gene Oncotype DX® Breast Recurrence Score™ (RS) assay (Genomic Health, Inc., Redwood City, CA, USA) aids in adjuvant chemotherapy decision-making for those with early-stage hormone receptor-positive (HR+), HER2-negative breast cancer. The assay was initially validated to predict recurrence risk and chemotherapy benefit among those with lymph node-negative disease [1, 2]. In this context, the RS assay provides additional information to the chemotherapy decision-making process, and has been shown to impact chemotherapy recommendations [3–6], reduce overall chemotherapy overuse [7], and be cost-effective [8].

Following its commercial availability in 2004, the test became covered by insurers beginning in 2006 and has been incorporated into current clinical guidelines beginning in 2007 for use among those with lymph node-negative disease [9]. Retrospective evidence accrued thereafter suggested that the test is also predictive among select patients with minimal lymph node involvement (1–3 positive nodes) [10, 11]. RS results in select lymph node-positive (LN+) patients have been validated to be prognostic for 10-year distant recurrence and breast cancer-specific survival (BCSS) [10, 11] and predictive of chemotherapy benefit in prospectively designed studies of archival tissue [10]. As a result, the 2015 National Comprehensive Cancer Network Clinical Practice Guidelines in Breast Cancer have incorporated Oncotype DX testing into clinical guidelines for LN+ patients, specifically noting that Oncotype DX testing can be considered for patients with 1–3 involved ipsilateral axillary lymph nodes [12].

Currently, the RxPONDER trial is underway to evaluate the use of the 21-gene assay for breast cancer patients with 1–3 positive lymph nodes (HR+, HER2-negative), by randomizing research participants with a RS ≤ 25 to receive hormonal therapy alone or hormonal therapy with adjuvant chemotherapy [13]. Awaiting results from RxPONDER, limited information remains available about outcomes of contemporary patients with LN+ disease who were treated based on RS results. As such, the primary objective of this study was to assess 5-year BCSS among those with LN+ breast cancer who have oncotype DX RS results.

Methods

Data source

The National Cancer Institute’s Surveillance, Epidemiology, and End Results Program (SEER) breast cancer cases (diagnosed 2004–2012) were linked to 21-gene Oncotype DX Breast Recurrence Score assay results from the Genomic Health Clinical Laboratory (2004–2013) by Information Management Services (IMS, Inc.; Calverton, MD, USA). IMS released the de-identified, linked dataset to the study team for analyses following SEER approval.

Study population

The study population included men and women with estrogen receptor (ER)- and/or progesterone receptor (PR)-positive invasive breast cancer as defined by the SEER-reported ER and PR (positive or borderline) immunohistochemistry results. We include men because treatment guidelines for male breast cancer are similar to those for female breast cancer, and evidence indicates that the distribution of Recurrence Score results among men is similar to that among women [14]. We excluded those (1) with distant metastases, (2) affected by Hurricane Katrina (incident cases in Louisiana from July to December 2005), (3) with breast cancer diagnosis over the age of 99, (4) with zero years of survival time, (5) with autopsy data only, and (6) without a state death certificate. The cohort was stratified by lymph node status: lymph node-negative (N0, including isolated tumor cells), micrometastases only (N1mi), 1 positive lymph node (1LN), 2 positive lymph nodes (2LN), 3 positive lymph nodes (3LN), or 4–90 positive lymph nodes (≥4LN). We examined uptake of the 21-gene assay from 2004 to 2013 using these inclusion/exclusion criteria (n = 352,468).

For the remaining analyses, we further restricted the study population. We included only those who had a 21-gene assay Recurrence Score result within 12 months following breast cancer diagnosis among those diagnosed between 2004 and 2012. Furthermore, we excluded patients with HER2-positive disease. HER2 status was determined by the RT-PCR single-gene HER2 scores (≤11.5 = HER2-negative) performed by Genomic Health, Inc., as HER2 status was unavailable in SEER prior to 2010.

Measures

Uptake was defined as the proportion of patients in the study cohort who had a 21-gene assay ordered within 12 months following breast cancer diagnosis.

BCSS was calculated using cause of death data reported by the SEER registries, which link with state death certificates and the National Center for Health Statistics’ National Death Index [15]. These analyses utilized a variable derived by the SEER program that uses a mapping to dichotomize causes of death as breast cancer-specific or other-cause specific [16]. In the estimates of BCSS, follow-up is censored at the time of other-cause deaths or last known follow-up. In the estimates of other-cause survival, follow-up is censored at the time of breast cancer-specific death or last known follow-up. Overall survival was estimated by including all deaths as events regardless of cause.

Statistical analyses

For the 21-gene assay uptake evaluation, the proportion tested was calculated among those with HR+, non-metastatic, invasive breast cancer as the denominator. The proportions were calculated and plotted by year of breast cancer diagnosis (from 2004 to 2013). Descriptive statistics of the analytic sample were tested using ANOVA and Chi square to compare sample characteristics by number of positive lymph nodes. The survival analysis used the actuarial method using SAS version 9.4 (SAS Institute Inc.; Cary, NC) and was stratified by 21-gene assay RS cutpoints (RS < 18, RS 18–30 or RS ≥ 31) and by number of positive lymph nodes. A two-degrees-of-freedom log-rank test was used to determine if there was evidence of differences across the three categorical RS groups. Survival curves are presented through 5 years, but all observed events and follow-up were included for statistical testing purposes. Confidence intervals were computed using the log–log transformation. As a sensitivity analysis, we also examined BCSS by 21-gene assay RS cutpoints that correspond to those used in the RxPONDER trial (RS < 11, RS 11–25, RS ≥ 26) [13].

Exploratory analysis

Chemotherapy use among those with RS results was examined as an exploratory aim. Chemotherapy was assessed using SEER-reported data from 2004 to 2012. The proportion of those for whom the registries reported chemotherapy use (versus those reported as no or unknown) was plotted by the standard RS cutpoints (low < 18, intermediate 18–30 or high ≥ 31) and by continuous RS value (0–100). This aim was exploratory, as chemotherapy is under-reported in SEER [17], and as such the results must be interpreted with caution.

Results

Of the 352,468 HR+ and non-metastatic breast cancer cases, 74,207 patients had a 21-gene breast cancer assay ordered (Supplemental Fig. 1). Overall, the uptake of the 21-gene breast cancer assay increased over time (Fig. 1), with lower uptake among patients with greater lymph node involvement. For example, during 2004–2013, only 1.6% of those with ≥4LN received the 21-gene breast cancer assay compared with 21.6% of those with micrometastases. Uptake of the 21-gene breast cancer assay increased from 0.7% in 2004 to 44.4% in 2013 among those with micrometastases.

Fig. 1
figure 1

The proportion of individuals receiving Recurrence Score testing who had HR+, non-metastatic breast cancer, by year and number of positive lymph nodes. N0 lymph node-negative, N1mi micrometastases only, 1LN 1 positive lymph node, 2LN 2 positive lymph nodes, 3LN 3 positive lymph nodes, ≥4LN 4, or more positive lymph nodes

Full size image

Patient characteristics of those with LN+ disease diagnosed between 2004 and 2012 who received a Recurrence Score result (n = 6768) varied by lymph node involvement. More specifically, patients with greater lymph node involvement tended to be older, to have tumors that were larger and higher in grade, to have higher mean RS results, and to live within a lower socioeconomic status (SES) area, compared with those with lower lymph node involvement (Table 1). There were no significant differences across lymph node status by race, year of diagnosis, or ER/PR status.

Table 1 Descriptive statistics for patients with HR+, non-metastatic breast cancer who had Oncotype DX Recurrence Score results, by number of positive lymph nodes, 2004–2012 (n = 6768)
Full size table

Chemotherapy reported as “yes” increased as lymph node involvement and RS value increased (Fig. 2). For example, among those with RS < 18, the registries reported 19% of patients with micrometastases as “yes” for chemotherapy use, compared with 56% of those with ≥4LN. Among those with ≥4LN, the proportion for whom the registries reported as “yes” for chemotherapy use increased from 56% among those with low RS results (RS < 18) to 77% among those with high RS results (RS ≥ 31). Reported chemotherapy was also modestly higher for patients with larger tumors or poorly differentiated tumors; however, age was the only variable with an effect of comparable magnitude to nodal status or RS results.

Fig. 2
figure 2

Chemotherapy use reported as “yes” versus “no/unknown,” by Recurrence Score results and number of positive lymph nodes. Overall chemotherapy use was reported as “yes” in 32.1% with N1mi, 34.6% with 1LN, and 43.9% with 2–3LN. For comparative purposes, chemotherapy use reported for lymph node-negative patients (N = 49,681), by Recurrence Score results, are shown in Supplemental Fig. 2

Full size image

Five-year BCSS outcomes for those with RS < 18 ranged from 98.9% (95% CI 97.4–99.6) in those with micrometastases to 92.8% (95% CI 73.4–98.2) among those with ≥4LN. Similar patterns of worsening BCSS with increasing number of positive lymph nodes were found for those with RS results 18–30 and ≥31 (Table 2). The RS result was strongly predictive of BCSS among the group with micrometastases or 1–3LN, and the group with micrometastases or 1LN (both p < 0.001; Fig. 3). RS results were not significantly associated with other non-BCSS (Fig. 3). In a multivariable Cox model adjusting for grade, tumor size, age, and race, both the number of positive lymph nodes and the RS category remained strongly associated with BCSS (p < 0.001) (Supplemental Table 1).

Table 2 Five-year breast cancer-specific survival in patients with lymph node-positive HR+ invasive breast cancer and Recurrence Score results in the SEER database; N = 6768
Full size table
Fig. 3
figure 3

Five-year breast cancer-specific survival (a, c, e) and other-cause (non-breast cancer-specific) survival (b, d, f), by Recurrence Score group among patients with micrometastases and 1 positive lymph node (N1mi, 1LN; N = 5440; a, b), patients with 2–3 positive lymph nodes (2–3LN; N = 1043; c, d), and patients with ≥4 positive lymph nodes (≥4LN; N = 285; e, f). For comparative purposes, 5-year BCSS and other-cause (non-BCSS) among patients with micrometastases up to three positive lymph nodes, by Recurrence Score group, are shown in Supplemental Fig. 3. Also, 5-year overall survival, by Recurrence Score group and number of positive lymph nodes, is shown in Supplemental Fig. 4

Full size image

Discussion

Overall, 5-year BCSS is high in patients with RS < 18, particularly among those with micrometasteses and fewer positive lymph nodes. Five-year survival worsened both as the number of involved lymph nodes increased, and as RS result increased. Not all patients with LN+ disease had aggressive disease. Patients with RS < 18 had good survival and a low proportion for whom the registries reported as “yes” for chemotherapy use, suggesting that the addition of chemotherapy to adjuvant treatment in this RS group may have relatively little impact on BCSS. Furthermore, our findings support evidence that those with LN+ disease who receive 21-gene assay testing tend to be older and to have lower grade, smaller tumors, compared with those who do not receive such testing [18], and that rates of test uptake have increased over time among LN+ patients [19].

Future analyses that account for treatment are warranted; in particular, a comparison of survival among those who did and did not receive chemotherapy is of interest. Longitudinal data updates through the SEER and Genomic Health Clinical Laboratory linkage will allow for better characterization of how the 21-gene breast cancer assay is used, as well as survival outcomes among LN+ patients. Studying health services use of the 21-gene breast cancer assay and associated health outcomes in LN+ patients are particularly important, as our findings indicate increasing uptake of the assay.

Limitations

This study has several limitations. First, follow-up time has been relatively short to date; as such, additional analyses with longer follow-up are warranted. Of note, most chemotherapy benefit, if any, is observed in the first 5 years, and 5-year BCSS predicts 10- and 15-year BCSS [4]. Thus, study findings are informative for the decision to treat with chemotherapy and for longer-term survival outcomes. Second, future studies should incorporate treatment data into survival analyses; specifically, to examine survival by chemotherapy use and to account for surgery and/or radiation therapy. Chemotherapy use is known to be under-reported to SEER, and therefore such analyses were suboptimal in this study. Moreover, estimates of chemotherapy use should be interpreted with caution, as they likely under-report chemotherapy uptake. Third, the SEER registries do not collect information on distant recurrence, and this outcome could not be evaluated. Finally, we were unable to restrict the 21-gene assay uptake analyses by HER2-negative status in Fig. 1, as SEER did not begin collecting HER2 data until 2010. This may have led to an underestimate of 21-gene assay uptake prior to 2010 and to challenges to the comparison of survival between tested and untested patients.

Conclusions

This SEER study adds to the current understanding of survival outcomes among LN+ patients who receive the 21-gene breast cancer assay RS results. Our findings demonstrate high survival outcomes among those with low RS results, and lesser lymph node involvement. Future studies should continue to examine longer-term survival outcomes and incorporate treatment variables into analyses.