Discussion
To our knowledge, this is the largest registry-based study focused on the relationship between utilization of the RS assay, RS, and receipt of chemotherapy, in chemotherapy-eligible, early-stage breast cancer patients. In addition to another year of data collection, our study provides a focused propensity matched analysis on chemotherapy receipt in contrast to a recently published study also evaluating NCDB data which provided a broad survey of therapeutic implications and disparities associated with the RS assay [
12]. We found that both performing the test, compared to not, and the RS itself was strongly associated with chemotherapy receipt. Additionally, within the low and intermediate RS groups a higher numerical value of the RS was associated with chemotherapy receipt.
Before the introduction of MGST testing, evidence-based guidelines for decision-making regarding chemotherapy in breast cancer patients began after the completion of multiple randomized cooperative studies, which were later included in the Early Breast Cancer Trialists Collaborative Group meta-analysis in 2005. Individual trials and the meta-analysis demonstrated a benefit of chemotherapy for all women with early-stage disease, based on clinical–pathologic criteria [
13]. Subsequently, other investigators provided evidence of both the prognostic (2004) and predictive (2006) value of the RS assay [
2,
14]. As a consequence, care guidelines changed, acknowledging the potential utility of the RS assay for decision-making. As early as 2007, the American Society of Clinical Oncology (ASCO) described appropriate use of the RS assay. Soon thereafter, the NCCN and others incorporated the RS assay into their guidelines for chemotherapy [
3]. Recognizing the importance of the RS assay and other types of MGST, the NCDB began to capture data for them in 2010. There were at least four reasons to do so. The first is to monitor for “appropriateness” of ordering the test; i.e., are providers compliant with the NCCN and other guidelines for ordering the test? Is there under- or overutilization of the test itself? Recent examples of inappropriate and unnecessary testing of different test types have been highlighted by ASCO in their contributions to the Choosing Wisely campaign [
15]. Second, if the test is ordered, do the results change provider's behavior—and the patient–provider decision—to use chemotherapy? Without monitoring, there is no way to know if test results are changing practice. New tests are of no value if their results do not influence the treatment decisions for which the test was developed. Third, validation of the predictive function of the RS assay to determine “response” or “benefit” from chemotherapy requires tracking of long-term cancer outcomes; i.e., is the overall survival of patients—matched for clinical–pathologic factors—the same or different when patients with low RS and no chemotherapy are compared to patients with low RS who received chemotherapy? The NCDB is critical for such a validation, due to its robust patient numbers and demographic diversity. Lastly, a national dataset that includes RS is necessary to search for inequities and disparities for access to this important test. Jasem et al. evaluated racial disparities in their analysis of the 2010–2012 NCDB [
12].
We found that 58.5 % (77,664/132,651) of patients who had data available and were NCCN criteria eligible for testing were tested. We are unaware of any professional organization that has yet endorsed a “benchmark” or “target” for the proportion of eligible patients that should have the test ordered. Future investigations searching for variability of use of the RS assay by provider, facility, and other characteristics are warranted.
In a large meta-analysis, RS assay categorized 49, 39, and 12 % of patients into low-, intermediate-, and high-risk groups, respectively [
16]. We found a significantly higher rate of low-risk patients at 59.5 %, and lower rates of intermediate- and high-risk patients at 32 and 8.5 %, respectively, suggesting patients enrolled in clinical trials may have higher risk features than those in standard US clinical practice.
After adjustment for all other relevant factors, we observed that patients who did not have RS assay testing had significantly increased odds of receiving chemotherapy (OR 1.21). Other studies have supported the ability of the RS assay to change physician treatment recommendations compared with those made on the basis of clinical pathologic characteristics alone, with a shift in recommendations from chemotherapy and hormonal therapy to hormone therapy alone. Reported reductions in chemotherapy use range from 10 to 20 % with use of the assay [
17‐
24], although not all studies have reported declines in chemotherapy use [
22]. Using matched analysis, we found an absolute reduction in chemotherapy receipt of 3 % when RS assay was utilized. This corresponds to a number needed to test of 33 to forego chemotherapy use in one patient, suggesting a significant impact in stewardship of limited health care resources and cost of care. Over the 4-year study period, we saw an overall decrease in chemotherapy use with a corresponding increase in utilization of RS assay.
The RS assay in previous work has been associated with chemotherapy receipt in 10 % (low RS), 36 % (intermediate RS), and 72 % (high RS), although no clinical pathologic correlates were available in these patients [
25]. In our study, we observed grade 1 tumors had a high RS in only 1.4 % of cases showing that a low grade can preclude the need for RS testing. In comparison, the rates of chemotherapy receipt in our analysis were significantly lower in the low-risk group at 5 %, slightly higher rates in intermediate-risk patients at 40 %, and an increased rate of high-risk patients receiving chemotherapy at 85 %. These results are similar to a prior investigation that captured patients and testing from an insurance-claims database linked to state cancer registries [
6]. In this study, Potosky et al. reported on 2362 patients less than 65 years of age with invasive cancer. They found that 51, 39, and 10 % of these women had low, intermediate, and high RS, of which 11, 47, and 88 %, respectively, received chemotherapy, demonstrating a statistically significant direct relationship between the RS and receipt of chemotherapy. On the other hand, our findings differ from a recent analysis utilizing a Surveillance, Epidemiology, and End Results (SEER) dataset. Dinan et al. in a study limited to Medicare beneficiaries observed no overall association between use of the RS assay and receipt of chemotherapy; however, in an unadjusted subset analysis, they did find a correlation between the RS assay and lesser chemotherapy in patients aged 66–70 years [
7].
We observed that patients with low and intermediate risk scores receiving chemotherapy more often if additional clinical pathologic factors were present. We also found concordance with a linear model of increasing receipt of adjuvant chemotherapy as the absolute RS rose in both low- and intermediate-risk patients after adjustments for all other factors, supporting the previous observations of Potosky et al. [
6]. For example, we found that in the intermediate group when the RS score increases from 20 to 30, the odds of a patient receiving chemotherapy increase by a factor of 12.7. These results suggest that oncologists are using the entire range of scores within the low- and intermediate-risk group combined with clinical pathologic risk factors to guide treatment decisions. As expected, the low RS group had the highest rate of declining chemotherapy when offered (56 %), likely based on perceived lack of benefit of adjuvant chemotherapy—a finding recently supported in a prospective clinical trial of node-negative women [Trial Assigning Individualized Options for Treatment (TAILORx)], wherein a similar low-risk group had very low rates of recurrence at 5 years with endocrine therapy alone [
5]. Interestingly, 24 % of the intermediate RS group declined chemotherapy after it was offered, likely representing unclear benefits. Planned analysis of intermediate group RS in the TAILORx trial is designed to address the efficacy of chemotherapy in women with Recurrence Score values from 11 to 25, but results are still pending.
A strength of this study includes the sample size afforded by the NCDB, allowing the majority of patients with early-stage chemotherapy-eligible breast cancer in the United States to be captured. Given this cohort, we were able to adjust for the known confounding patient, tumor, and facility factors that affect receipt of chemotherapy as well as match similar patient groups.
Our study has limitations. It has a retrospective study design in which numerous patients had missing values for RS. However, given the sample size of the NCDB, it is unlikely that our findings would change, even with an uneven distribution of the patients with missing RS values between the low-, intermediate-, and high-risk groups. In addition, unobserved confounding factors can limit interpretation of study outcomes derived from observational data. Some misclassification and treatment underreporting are unavoidable in a large registry-based dataset like the NCDB. The NCDB database does not provide information regarding the intensity of ER expression (such as weakly positive, 1+, 2+, 3+, or percentage ER positive) but rather just categorizes ER as “positive,” “negative,” or “borderline.” This is a limitation as many clinicians may have used these intensity data to aid in deciding whether or not to order a RS assay or recommend chemotherapy. Additionally, we could not assess the chemotherapy regimen recommended or compare breast cancer recurrences, progression-free survival, or overall survival given the short duration of MGST data in the NCDB and lack of outcome data for less than 5 years in the NCDB. Earlier endpoints of chemotherapy benefit, such as local or regional recurrence rates, are not captured in the NCDB. Lastly, we are unable to ascertain reasons for treating providers not obtaining a RS assay or recommending for or against chemotherapy. For instance, inherent bias may exist if oncologists are more inclined to test early-stage breast cancer patients whom they were considering for chemotherapy, such as younger patients or those with higher grade tumors, and not order tests for patients whom they intend to treat with hormone therapy alone on the basis of more favorable clinical pathologic features or patient unfitness or aversion to chemotherapy. Alternatively, perhaps oncologists are less likely to order a RS assay in younger patients who have high-grade tumors that are only weakly ER positive as they are uncomfortable foregoing chemotherapy in this population. In fact, a bias among oncologists against ordering a RS assay in these patients perceived to be at higher risk based on pathologic characteristics may be suggested by our finding of a lower percentage of patients falling in the high-risk RS group as compared to previous studies.
In conclusion, utilization of RS assay and RS was strongly associated with chemotherapy receipt in a nationally representative cohort of chemotherapy-eligible women with early-stage breast cancer. Future investigations are warranted to clarify the optimal proportion of eligible patients that should undergo RS assay testing. Until then, the baseline percentage identified in the NCDB study herein is 47 %.