Racial Differences in Adjuvant Endocrine Therapy Use and Discontinuation in Association with Mortality among Medicare Breast Cancer Patients by Receptor Status

Blacks and Hispanics have an increased risk of breast cancer death compared with non-Hispanic whites (1–4). These racial/ethnic mortality disparities have been attributed to differences in cancer prognostic (2, 3, 5, 6), sociodemographic factors (4), and the initiation and timing of treatment (2, 4). One way to significantly reduce breast cancer mortality is to improve adherence to effective, recommended treatment (7). Adhering to guidelines for adjuvant endocrine therapy (AET) is associated with improved disease-free survival for women with early-stage breast cancer (8–12); however, adherence rates for recommended treatment remain low. It is estimated that between 55% and 75% of breast cancer patients received recommended AET medication in a 1-year period (13). Discontinuation of AET is associated with the number of other medications prescribed for comorbidities (14), demographic characteristics such as age (15, 16) and the side effects (13, 17–21). Previous studies have examined higher discontinuation for minorities as compared with white women (15, 22).

The National Comprehensive Cancer Network (NCCN) recommends that postmenopausal women with early-stage hormone receptor–positive breast cancer receive an aromatase inhibitor (AI) for 5 years, or tamoxifen for 2 to 3 years, followed by an AI to complete 5 years, or tamoxifen alone for 5 years, if AIs are contraindicated (23). The guidelines suggest that postmenopausal women can begin with tamoxifen or AIs, and understanding whether discontinuation is associated with each drug during the first 12 months of treatment postdiagnosis may provide additional evidence to physicians and patients when deciding one drug over another.

Nearly two-thirds of breast cancer cases in the United States are hormone receptor positive (estrogen or progesterone receptor) and are eligible for AET (23–25). Although recommendations from both the NCCN and the American Society of Clinical Oncology (ASCO) recommend AET for women with stage I–III cancer, none of the organizations recommended AET for women with hormone receptor–negative disease (23, 26). However, women with hormone receptor–negative breast cancer may still receive AET for the following reasons. First, determining hormone receptor–positive status using IHC was initially defined as having ≥10% positive tumor cells (27). However, subsequent studies have identified improved disease-free survival for patients with ≥1% positive tumor cells (28) and even >0% (29). Second, classification of hormone receptor status differed among institutions. Although SEER defined hormone receptor–positive status as having ≥1% positive tumor cells (30), other institutions or providers may consider those patients with any percentage of positive tumor cells (>0%) to be eligible for AET (27, 28). Third, studies have recently shown that even those with hormone receptor–negative breast cancer may experience lower risk of disease-free survival from receiving tamoxifen (31, 32). Therefore, we aimed to examine the patterns of initiation, discontinuation, and mortality associated with AET in these populations. To the best of our knowledge, this would be the first study to examine these research questions in a large, nationwide cohort of elderly women with breast cancer since Medicare Part-D was implemented in 2006 and oral AET was covered under that plan.

Furthermore, what remains unclear is whether there are racial/ethnic differences in initiation and discontinuation by hormone receptor status and whether discontinuing AET is associated with all-cause mortality. We hypothesized that AET discontinuation would be associated with a higher risk of death, regardless of hormone receptor status, and after controlling for AET discontinuation, there would be no significant differences in the risk of mortality among Hispanics, blacks, and non-Hispanic whites.

This study utilized the NCI's SEER (Surveillance, Epidemiology, and End Results) and Medicare-linked database for cases in 2007 to 2009 with Medicare Part-D claims up to December 2010 (33, 34). Briefly, available SEER information includes patient demographics, primary tumor site, tumor morphology, stage at diagnosis, first course of treatment, and vital statistics. Medicare data contain detailed diagnoses, procedures, and billing information for inpatient, outpatient, and pharmacy claims.

This was a retrospective cohort study of women ≥65 years with stage I–III breast cancer who were enrolled in Medicare for ≥12 months before and after the date they filled their first AET prescription (either tamoxifen or AI). Women were excluded if they were not enrolled in Medicare Part-D, did not have both Parts A and B, or were enrolled with an HMO from the year of diagnosis to the last follow-up. We stratified the analysis into three cohorts to reflect NCCN recommendations. NCCN strongly recommends AET for women with stage I–III, hormone receptor–positive breast cancer (Group 1) but does not recommend for hormone receptor–negative (Group 2) or unknown status for stage I–III (Group 3; ref. 23).

We identified women who belonged to four racial/ethnic groups: non-Hispanic white, non-Hispanic black/African American, Hispanic, and Asian, which were described elsewhere (30).

Women with breast cancer were categorized in the SEER registry as hormone receptor positive, negative, or unknown based on tumor specimen values (30). If, at a minimum, one tumor specimen assay was noted to be positive for estrogen or progesterone receptors, then the patient was considered to have hormone receptor–positive disease (30). The SEER, as of 2010, considered hormone receptor–positive breast cancer if ≥1% cells stained positive (30).

Demographic information included age (median 75) and marital status. Socioeconomic status information obtained from the 2000 U.S. Census included tertiles of the percentage of residents living below the federal poverty level (FPL) at the census tract level (<5.4%, 5.4%–11.8%, and ≥11.8%) and whether they lived in a metropolitan region (4, 34). Tumor characteristics included AJCC tumor stage, size, grade, and lymph node status. Radiotherapy and surgery were identified from SEER data or through procedure codes in Medicare claims made <6 months of diagnosis, and chemotherapy use was identified only from Medicare claims <6 months of diagnosis using the methods described elsewhere (4). Comorbidity scores were then generated and described in detail elsewhere (35–37). We calculated average 30-day out-of-pocket payments for AET medication (38, 39).

Initiation of AET (tamoxifen or AI) was considered a single prescription based on the drug name up to 1 year after the date of diagnosis (40). AIs were defined as anastrozole, exemestane, or letrozole. We created a binary variable (yes or no) for initiation if a woman filled a prescription for any AET medication and an indicator for the type of AET (AI or tamoxifen) at initiation. A woman could only initiate AET therapy with AI or tamoxifen based on the first prescription following cancer diagnosis. We categorized patients as having discontinued therapy (yes or no) if they initiated therapy, and the calculated drug supply based on the last prescription date plus any surplus from a prior prescription indicated a minimum of 120-day supply gap with no AET medication on hand. All patients for the discontinuation measure were followed from their initial prescription date until the end of the study period (December 31, 2010). We censored other patients at the end of the study period or until they went >2 consecutive months without Medicare Part-D coverage or died. The median and mean length of follow-up from initiation to discontinuation was 523 days and 580 days, respectively (range, 5–1,460 days). All-cause mortality and breast cancer–specific mortality were defined separately as dead (of any causes or breast cancer specific) or alive at the last follow-up of vital statistics (December 31, 2011) to be censored. The median and mean length of follow-up from diagnosis to death was 1,205 and 1,214 days, respectively (range, 320–1,845 days).

Differences in the distribution of sociodemographic and tumor characteristics were first examined across racial/ethnic groups and then by AET initiation. Multivariable logistic regression models were performed to assess the association of race/ethnicity and AET initiation, controlling for tumor prognostic, sociodemographic, and clinical treatment factors. We performed Cox proportional hazards regressions to estimate the association between race/ethnicity and AET discontinuation and to determine the risk of all-cause and cancer-specific mortality in association with AET discontinuation. For this model, we included AET discontinuation within 12 months as every patient in the cohort was alive and had at least 12 consecutive months of continuous enrollment in Part-D after their first AET prescription. All analyses were stratified by hormone receptor status and controlled for prognostic and treatment factors.

We identified 14,902 Medicare patients with stage I–III breast cancer enrolled in Part-D (Table 1). Greater than four fifths (81.8%) had hormone receptor–positive breast cancer, and the remaining were hormone receptor negative (13.3%) or unknown status (6.0%). The cohort consisted predominately of non-Hispanic whites (81.1%), followed by non-Hispanic blacks (7.2%), Hispanics (6.1%), and Asian (4.8%). A larger proportion of blacks (71.2%) and Hispanics (60.8%) compared with non-Hispanic whites (29.6%) lived in areas where >11.8% lived below the FPL. A greater proportion of non-Hispanic whites (83.0%) than blacks (71.9%) and Hispanics (77.6%) was diagnosed with hormone receptor–positive breast cancer. A larger proportion of blacks and Hispanics compared with non-Hispanic whites had stage III breast cancer (15.5% and 15.3% vs. 9.3%).

A total of 64.5% initiated AET regardless of hormone receptor status (Table 2). Among hormone receptor–positive breast cancer patients, 74.8% initiated therapy. Notably, more women with hormone receptor–positive breast cancer initiated therapy with AIs (61.8%) than with tamoxifen (12.9%). Among hormone receptor–negative and unknown status breast cancer patients, 5.6% and 54.0% initiated AET, respectively. Among hormone receptor–positive and hormone receptor–unknown patients, a greater proportion of Hispanics initiated AET than any other racial/ethnic group (79.2% and 63.5%, respectively). A smaller proportion of women over the age of 80 and those with 3 or more comorbidities initiated AET compared with younger women or those with fewer comorbidities regardless of hormone receptor status. Women in areas where ≥11.8% of the population lived below the FPL initiated AET regardless of hormone receptor status.

Blacks had lower odds of AET initiation within 12 months of diagnosis compared with non-Hispanic whites [OR, 0.76; 95% confidence interval (CI), 0.66–0.88] after controlling for all demographic, treatment, and prognostic factors (Table 3). Older women (over 80 years) compared with younger women (65–69) had lower odds of AET initiation among hormone receptor–positive patients (OR, 0.69; 95% CI, 0.61–0.79). Also, women with more comorbidities, compared with none, had lower odds of AET initiation. Women who lived in areas where ≥11.8% of the population lived below the FPL had higher odds of initiating AET than those who lived in areas where <5.4% were below the FPL (OR, 1.16; 95% CI, 1.05–1.28). Initiation of tamoxifen was less likely among non-Hispanic blacks compared with whites (OR, 0.70; 95% CI, 0.55–0.89) and less likely among women with three or more comorbidities compared with 0 (OR, 0.79; 95% CI, 0.64–0.97), whereas Hispanics had higher odds of AI initiation compared with whites with hormone receptor–positive breast cancer (OR, 1.30, 95% CI, 1.09–1.55). Older women (age ≥80) were also less likely to initiate AET with AIs compared with younger women (age 70–74; OR, 0.51; 95% CI, 0.45–0.56).

A total of 80.6% of women discontinued therapy during the study period. Among women with hormone receptor–positive breast cancer who initiated AET, non-Hispanic blacks had a lower risk of AET discontinuation than did non-Hispanic whites (HR, 0.89; 95% CI, 0.80–0.98) during the study period after controlling for all other factors (Table 4). The risk of discontinuing AET was greater for women with hormone receptor–positive cancer who initiated therapy with AIs than with tamoxifen (HR, 1.12; 95% CI, 1.05–1.20).

AET discontinuation within 12 months of initiation was associated with a significantly higher risk of all-cause mortality among women with hormone receptor–positive (HR, 1.70; 95% CI, 1.48–1.95) and negative (HR, 11.65; 95% CI, 2.33–58.39) breast cancer, but not among those with unknown hormone receptor status (Table 5). Discontinuation of AET within 12 months of diagnosis was also associated with a 2.7-fold increased risk of breast cancer–specific mortality compared with those who did not discontinue therapy (HR, 2.76; 95% CI, 1.74–4.38) after controlling for all other factors. Women from areas where ≥11.8% live below the FPL had an increased risk of all-cause death than women in areas with <5.4% living below the FPL (HR, 1.25; 95% CI, 1.05–1.47). In the age-adjusted analysis, blacks had a 92% increased risk of death compared with non-Hispanic whites (HR, 1.92; 95% CI, 1.59–2.30). No significant differences in the risk of death were observed between blacks and Hispanics compared with non-Hispanic whites after controlling for discontinuation, poverty status, and all other treatment and prognostic factors.

In this retrospective cohort study of Medicare patients, a substantial proportion of women with hormone receptor–unknown breast cancer initiated AET, even though these patients were not expected to receive AET based on ASCO and NCCN clinical guidelines. Also, the odds of AET initiation and the risk of AET discontinuation differed by race/ethnicity and AET drug type, regardless of hormone receptor status. Those who discontinued AET, regardless of hormone receptor status positive or negative, had a significantly higher risk of all-cause mortality.

We previously reported AET initiation among older women with hormone receptor–positive, stage I–III breast cancer and found that 74.8% initiated therapy within 12 months (40). Other studies have found initiation between 68% and 70%, but those women were younger and commercially insured (41, 42) and may have better initiation rates than our study population. We also found that blacks, compared with non-Hispanic whites, had lower odds of initiating AET, but after stratifying by hormone receptor status, there was no association. Previous studies showed that Hispanics (41, 43) and blacks (43, 44) have lower odds of AET initiation compared with whites. These previous studies were conducted among different study populations and used AET initiation self-report, a younger cohort, or both.

We observed racial/ethnic differences in discontinuation of AET. We have reported AET adherence for stage I–III, hormone receptor–positive breast cancer patients but did not find a significant difference among Hispanic and black patients over non-Hispanic whites, after controlling for sociodemographic, prognostic, and treatment factors (45). What is even more interesting is that we found that the risk of AET discontinuation was greater for women who initiated therapy with the AIs than with tamoxifen among those with hormone receptor–positive breast cancer. This may be influenced by other factors, such as AI side effects, which we did not control for in this study (10, 11, 46, 47). Similar to other studies, we did find that discontinuing AET was associated with medication cost (12, 48), comorbidities (49), and age (49). Previous studies have examined lower adherence rates for non-whites, a finding that may contribute to the disparities in breast cancer mortality observed between minorities and whites (15, 22, 50). However, after controlling for AET discontinuation, we did not observe racial/ethnic differences in the risk of all-cause mortality. We found that discontinuation of AET was independently associated with a higher risk of all-cause mortality, which is corroborated by Hershman and colleagues (12), who studied a younger population but did not examine discontinuation among hormone receptor–negative or unknown cancer. However, they found that Hispanics had a lower risk of death than non-Hispanic whites (12). We did not observe this association in older Medicare women with breast cancer.

No studies have examined AET initiation and discontinuation in women with hormone receptor–negative or unknown breast cancer. In our study, 5.6% of hormone receptor–negative and 54% of unknown status breast cancer patients initiated AET. Although the sample size was limited among the hormone receptor–negative patients who initiated AET, still, women with hormone receptor–negative breast cancer who discontinued AET had significantly increased risk of death. There are a few important points to consider when discussing the results of this finding. Previous studies based on medical claims data only for AET adherence or discontinuation included any women who filled a prescription for AET and might have included women with hormone receptor–negative cancer because those data do not allow for differentiation of hormone receptor status (12, 22, 39). Women may be on AET despite having hormone receptor–negative breast cancer because they may have IHC assay >0% with positive tumor cells but were classified as negative based on a changing cut-off point of >1%, which may vary based on institutional guidelines (27, 28). Next, women with hormone receptor–negative breast cancer may be on AET for primary prevention of metachronous disease as it may reduce the risk of contralateral primary breast cancer (51, 52). Tamoxifen use can significantly improve 5-year disease-free survival in hormone receptor–positive breast cancer using IHC assay >0% cutoff (29). Recently, a study found that the benefit of tamoxifen response in hormone receptor–negative breast cancer (using <1% tumor positive as cutoff) was mediated by androgen receptor expression (31). Another study found that estrogen receptor (ER) β1 status is a significant predictor of disease-free survival in ERα-negative breast cancers treated with tamoxifen (32). Both of these receptor expressions (androgen receptor and ERβ1) are not currently assessed in the SEER dataset but could explain our study results that AET discontinuation in hormone receptor–negative breast cancer is, at least partially, associated with increased mortality risk.

Although previous studies of AET patterns and outcomes used only medical claims or pharmacy data without details on tumor characteristics (39, 48, 53), we were able to examine initiation, discontinuation, and all-cause mortality by hormone receptor status, stage at diagnosis, and other prognostic factors because Medicare claims data were linked to the SEER registry database. Our measure of discontinuation was defined as patients with 120 consecutive days without AET coverage. Therefore, we observed a slightly lower rate of discontinuation than previous studies that used 45 or 90 days without coverage as a measure (12, 22). Even with a conservative discontinuation measure, women who discontinued treatment within 12 months of initiation had a greater risk of death than those who did not, even after controlling for all other factors. Furthermore, because we had complete medical claims, pharmacy, and SEER registry data, this is one of the most comprehensive studies examining racial/ethnic differences in AET patterns and outcomes.

Our study was limited, however, by the population, which only included women ≥65 years and enrolled in Medicare Part-D. Therefore, results may not be generalizable to younger patients or those without comprehensive prescription drug coverage. Second, unmeasured confounding factors, such as psychosocial factors, related to the quality of care women receive (e.g., physician–patient communication) may influence their AET continuation but could not be captured in this study (54). Third, calculating discontinuation using prescription claims assumes that patients take their medications as often as they refill prescriptions. However, pharmacy records are considered the most accurate and valid estimation of actual medication use in large populations over time (55, 56). Fourth, 4.9% of the cohort had unknown and 13.4% had hormone receptor–negative status. Although we combined these groups in stratified analyses, those with unknown receptor status could be hormone receptor positive and hence misclassified, leading to biased estimates. Also, as previously discussed, women were categorized as having hormone receptor–negative status (<1% positive IHC) but may have actually had hormone receptor–positive disease according to the less stringent definition used in other studies (>0% positive IHC).

About three fourths of patients with hormone receptor–positive breast cancer initiated AET, and over 50% of women with hormone receptor–unknown and more than 5% of those with hormone receptor–negative status received AET. AET discontinuation was associated with a significantly higher risk of all-cause mortality, regardless of hormone receptor status and tumor stage. Overall, our study underscores the importance of continuing AET for all breast cancer patients who initiate treatment. Future studies are needed to explore the survival benefits of receiving AET for hormone receptor–negative breast cancer in other populations and to study the factors that influence AET discontinuation.

No potential conflicts of interest were disclosed.

The interpretation and reporting of these data are the sole responsibilities of the authors.

Conception and design: A.J. Farias, X.L. Du

Development of methodology: A.J. Farias, X.L. Du

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X.L. Du

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.J. Farias, X.L. Du

Writing, review, and/or revision of the manuscript: A.J. Farias, X.L. Du

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Farias, X.L. Du

Study supervision: X.L. Du

We acknowledge the efforts of the National Cancer Institute, Center for Medicare and Medicaid Services, Information Management Services, Inc., and the Surveillance, Epidemiology, and End Results Program tumor registries in the creation of this database.

A.J. Farias is a postdoctoral fellow supported by a University of Texas Health Science Centerat Houston School of Public Health's Cancer Education and Career Development Program grant from the NCI (R25-CA57712). This study was also supported, in part, by grants to X.L. Du from the Agency for Healthcare Research and Quality (R01-HS018956) and the Cancer Prevention and Research Institute of Texas (RP130051).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.