Evaluation of the Quality of Adjuvant Endocrine Therapy Delivery for Breast Cancer Care in the United States

Key Points

Question  What are the patterns and factors associated with receipt of adjuvant endocrine therapy for breast cancer?

Findings  This cohort study of 981 729 women with stages I to III breast cancer in the National Cancer Database found that receipt of adjuvant endocrine therapy increased from 69.8% in 2004 to 82.4% in 2013 among patients with hormone receptor–positive cancer. Surgery and radiotherapy were the factors most significantly associated with receipt of adjuvant endocrine therapy.

Meaning  Optimal receipt of adjuvant endocrine therapy has not been achieved, but team-based care may support guideline-concordant care.

Importance  Randomized trials in breast cancer have demonstrated the clinical benefits of adjuvant endocrine therapy (AET) in preventing recurrence and death. The examination of concordance with AET guidelines at a national level as a measure of quality of care is important.

Objective  To investigate temporal trends and factors related to receipt of AET for breast cancer.

Design, Setting, and Participants  This retrospective cohort study included 981 729 women with breast cancer in the National Cancer Database from January 1, 2004, to December 31, 2013. Women with stages I to III breast cancer who received all or part of their treatment at the reporting institution were included in the analysis.

Main Outcomes and Measures  Temporal changes in AET receipt (estimating the annual percentage change) and AET practice patterns (using logistic regression) and the effect of AET guideline concordance on survival of women with hormone receptor–positive (HR+) breast cancer (using the multivariable Cox proportional hazards model).

Results  Of the 981 729 eligible patients (mean [SD] age, 60.8 [13.3] years), 818 435 had HR+ and 163 294 had HR-negative (HR−) cancer. Among the patients with HR+ cancer, receipt of AET increased over time, from 69.8% in 2004 to 82.4% in 2013. Among patients with HR− cancer, receipt decreased from 5.2% in 2004 to 3.4% in 2013. Hospital-level adherence (≥80% of patients with HR+ cancer received AET) increased from 40.2% in 2004 to 69.2% in 2013. Receipt of AET varied significantly by age (lower in patients ≥80 years), race (lower in African American and Hispanic participants), geographic location (lower in West South Central, Mountain, and Pacific census regions), and receptor status (lower in patients with estrogen receptor–negative and progesterone receptor–positive cancer). Surgery and radiotherapy were the factors most significantly associated with appropriate AET receipt (only 45.0% in patients who received lumpectomy without radiotherapy). Receipt of AET was associated with a 29% relative risk reduction in mortality. Based on this effectiveness estimate, if all patients with HR+ cancer received AET, approximately 14 630 lives would have been saved over 10 years.

Conclusions and Relevance  From 2004 to 2013, underuse and misuse of AET have decreased for patients with breast cancer, but optimal use has not been achieved, and significant variation in care remains. The involvement of surgery and radiotherapy were among the most significant factors associated with optimal use, which underscores the benefits of team-based care to support guideline-concordant therapy.

Since the 1983 publication of the landmark Nolvadex Adjuvant Trial Organisation results, adjuvant endocrine therapy (AET) has been an integral treatment in breast cancer.^1,2 The Early Breast Cancer Trialists’ Collaborative Group (EBCTCG)³ meta-analysis showed that 5-year results for treatment with tamoxifen citrate reduced recurrence by 39% and 15-year breast cancer mortality by approximately one-third in estrogen receptor–positive (ER+) disease. Aromatase inhibitors have become the standard of care for postmenopausal women.⁴ An EBCTCG⁵ meta-analysis found that compared with 5-year findings for tamoxifen, 5 years of treatment with an aromatase inhibitor reduced recurrence and mortality.

Previous guidelines recommended treatment of premenopausal women with hormone receptor–positive (HR+) disease with tamoxifen for 5 years and postmenopausal women for 5 years with an aromatase inhibitor or tamoxifen or both in sequence. Based on recent trials,^6-9 the American Society of Clinical Oncology issued a 2016 guideline update¹⁰ to recommend ovarian suppression for 5 years in combination with AET for high-risk premenopausal women and AET alone for women with stage I breast cancers not warranting chemotherapy and node-negative cancers of 1.0 cm or less. Similarly, the National Quality Forum¹¹ (cancer measure 0220) endorsed tamoxifen or a third-generation aromatase inhibitor (considered or administered) within 1 year of diagnosis as a marker of quality care for patients with HR+ American Joint Committee on Cancer stage T1cN0M0, II, or III disease. Studies have demonstrated that even patients with node-negative cancers of 1.0 cm or less benefit from AET.^12-14 These guidelines highlight the importance of AET in the treatment of nonmetastatic breast cancer.

Analysis of different populations has shown a range in guideline concordance. In studies using the Surveillance, Epidemiology and End Results (SEER) Medicare-linked data for 2006 to 2007¹⁵ and 2006 to 2009,¹⁶ the rate of AET use was approximately 70%. Kimmick et al¹⁷ examined North Carolina Medicaid-linked data and found a prescription fill rate of 70%. In the Breast Cancer Quality of Care Study (B-QUAL),¹⁸ a prospective cohort study at 3 US sites, the investigators found that 12.0% to 17.9% of participants did not initiate AET. Flanagan et al¹⁹ examined AET in patients with ductal carcinoma in situ using the National Cancer Database (NCDB). The authors found that fewer than half of patients with ER+ disease received AET.

The objectives of this study were to describe the temporal changes and AET practice patterns and to evaluate the effect of AET guideline concordance on survival of women with HR+ breast cancer. In the context of the changing landscape of cancer care, an increased emphasis on tying reimbursement to guideline-concordant care, and recent data²⁰ showing some patient populations being left behind, identification of factors associated with quality in breast cancer treatment is important.

The NCDB data are annually reported from about 1500 Commission on Cancer–accredited hospitals and capture approximately 70% of newly diagnosed malignant cancers in the United States. Aspects of the NCDB data have been described elsewhere,^21,22 and data were coded and reported using established protocols. This study used deidentified data and was deemed exempt from human protection oversight and informed consent by the institutional review board of the University of Chicago, Chicago, Illinois.

Data from 2 032 209 patients diagnosed with breast cancer from January 1, 2004, to December 31, 2013, were queried. The study was limited to women 18 years or older with a diagnosis of American Joint Committee on Cancer stages I to III breast cancer who received all or part of their first course of treatment in the reporting facility. Patients undergoing surgery but no neoadjuvant systemic treatment were eligible. The study excluded patients with prior cancer diagnoses. Patients with missing data for HR status or AET status were also excluded. The type of endocrine therapy is not specified in the NCDB. After these selections, 981 729 patients were included (eFigure 1 in the Supplement).

In the NCDB, AET recommended and administered as first-course treatment was recorded, and the reason for nonadministration was recorded for a subset of patients. Based on HR status, the population was divided into the following 2 groups: those with HR+ (ER+ or progesterone receptor–positive [PR+]) cancer and those with HR negative (HR− [ER− and PR−]) cancer for investigating AET underuse and misuse, respectively. Because the American Society of Clinical Oncology and College of American Pathologists lowered the ER and PR cutoff point positivity from 10% to 1% positive nuclei in 2010,²³ 10% may have been used before and 1% may have been used after 2010. For the analysis of underuse, we calculated patient-level and hospital-level adherence. Consistent with prior studies, adherence was benchmarked at 80%, based on the Commission on Cancer adherence standard.²⁴

Treatment modalities examined included surgery, radiotherapy, and chemotherapy. Patient demographics, including age at diagnosis, race, and insurance status, were analyzed as categorical variables. Patient comorbidity status was represented by the adoption by Deyo et al²⁵ of the Charlson Comorbidity Index, which is a cumulative score of 15 health conditions. Tumor characteristics, including tumor size, lymph node involvement, grade, histologic type, and ER and PR status, were analyzed as categorical variables. Tumor size was also analyzed as a continuous variable using restricted cubic spline functions, a less biased method of modeling continuous variables. Area-based indicators of income and educational attainment were derived at the zip code level from 2012 US Census data and included as quartiles in the general US population. Geographic region was based on the reported residence at diagnosis. The 2013 rural-urban continuum codes were assigned according to the patient’s county of residence. Facility-level characteristics included facility type (community, comprehensive community, and academic cancer center), annual patient volume, and distance to the facility.²⁶

We examined the temporal trend of AET receipt from 2004 to 2013 and estimated the annual percentage change using a generalized linear model for binomial distribution with identity link. Then we investigated demographic, clinical, and facility factors that were associated with AET receipt using logistic regression models. Odds ratios (ORs) and 95% CIs were estimated in the regression models. Given the large sample size, P < .001 was considered statistically significant, and the χ² statistic was reported as an indicator of significance strength.

To estimate the effectiveness of AET guideline concordance on overall survival, we fit multivariable Cox proportional hazards models among patients with HR+ cancer diagnosed from 2004 to 2012. Patients diagnosed in 2013 were excluded because the duration of follow-up was too short. To eliminate survivor bias, we removed survival time of the first 8 months from diagnosis (termed landmark analysis)²⁷ because only survivors who live long enough can receive AET. We chose the 8-month cutoff point because approximately 80% received AET within 8 months, and most deaths in the first 8 months are likely not related to AET. As a measure of relative benefit, we calculated hazard ratios and 95% CIs. To facilitate clinical interpretation, we estimated marginal mortality risks from the fitted Cox proportional hazards models. The marginal mortality risk can be interpreted as the population mean risk if all patients in a specific subgroup received or did not receive AET, while keeping all other prognostic factor profiles unchanged. Furthermore, we estimated the number of lives lost owing to nonadherence with AET guidelines by applying the difference in 10-year marginal mortality risks between AET users and nonusers to the total number of untreated patients with invasive breast cancer. Herein, we assumed that the NCDB captured 72.5% of US breast cancer cases.²⁸ All statistical analyses were conducted with STATA software (version 14.0; StataCorp).

Of the 981 729 eligible patients (mean [SD] age, 60.8 [13.3] years), 818 435 had HR+ cancer and 163 294 had HR− cancer. In the patient-level adherence analysis, the receipt of AET among patients with HR+ cancer increased from 69.8% in 2004 to 82.4% in 2013 (Figure 1A). This change represents an annual percentage change of 1.51% (95% CI, 1.48%-1.54%; P < .001). Among patients with HR− cancer, the AET receipt decreased from 5.2% in 2004 to 3.4% in 2013, representing an annual percentage change of −0.17% (95% CI, −0.14% to −0.21%; P < .001). In hospital-level adherence analysis, 40.2% of hospitals met the 80% threshold of guideline concordant care in 2004, which increased to 69.2% in 2013 (Figure 1B).

eTable 1 in the Supplement presents AET receipt among patients with HR+ breast cancer by factors that were statistically significant in multivariable logistic regression. Receipt of AET varied by age at diagnosis; the rate was highest in women aged 50 to 69 years (>80%) and particularly low among those younger than 40 years (79.1%) and patients 80 years or older (60.5%). Compared with non-Hispanic white patients (79.0%), black (76.4%) and Hispanic (75.9%) patients were less likely to receive AET. The combination of ER and PR status was strongly associated with AET receipt; only 51.2% of the patients with ER−/PR+ cancer received AET. The association between tumor size and AET receipt was nonlinear for patients with node-positive and node-negative disease, with receipt of AET highest for patients with tumors measuring 1.0 to 2.0 cm (eFigure 2 in the Supplement). Receipt of AET dramatically differed by local treatment as follows: postmastectomy radiotherapy (90.1%), postlumpectomy radiotherapy (85.5%), mastectomy alone (73.2%), and no radiotherapy after lumpectomy (45.0%). Patients receiving chemotherapy were more likely to use AET (86.4%) compared with those who did not (74.6%). Receipt of AET varied by geographic regions, with New England (83.3%) and the West North Central states (83.9%) having the highest proportions and West South Central states (71.2%) having the lowest proportion of AET use (eFigure 3 in the Supplement). Interestingly, facilities in rural areas and those treating a large volume of cases were more likely to be concordant with AET guidelines.

As expected, time to initiation of AET strongly depended on whether radiotherapy and chemotherapy were given, but only weakly depended on surgery type (Figure 2). For patients not receiving radiotherapy and chemotherapy, the median duration from diagnosis to AET was 75 days, compared with 112 days for radiotherapy alone, 202 days for chemotherapy alone, and 241 days for radiotherapy and chemotherapy (P < .001).

Of 174 786 patients who did not receive AET, 34 314 (19.6%) were recommended but refused, 7957 (4.6%) were recommended but not administered without a reason, 9473 (5.4%) were not recommended owing to patient risk factors. For the remainder of AET nonusers, reasons were not given. We explored factors associated with refusal (vs nonrefusal in women not receiving AET) and found that older women, those living in communities with higher educational levels, and those living in rural or small urban areas were more likely to refuse, whereas Hispanic women were less likely to refuse AET.

eTable 2 in the Supplement presents AET receipt among patients with HR− cancer by factors that were statistically significant in multivariable logistic regression. Among patients with HR− breast cancer, 4.2% received AET, although it lacks efficacy and risks adverse effects. Patients treated in low-volume hospitals (4.6%) were more likely to be inappropriately treated. Geographically, patients living in New England (3.7%) and West North Central states (3.5%) were less likely while those living in West South Central states (5.4%) were more likely to be inappropriately treated.

A total of 708 393 patients were included in the survival analysis (eFigure 1 in the Supplement). During a median follow-up of 4.9 years, 76 565 patients died. In the multivariable Cox proportional hazards model adjusted for age, race, insurance status, comorbidity index, year of diagnosis, ER and PR status, histologic type, tumor grade, tumor size, node status, surgical margin, type of surgery, radiotherapy, chemotherapy, facility type, facility location, and facility volume, receipt of AET was associated with a 29% relative reduction in mortality (adjusted hazard ratio, 0.71; 95% CI, 0.70-0.73; P < .001), with an absolute difference of 2.8% at year 5 and 6.1% at year 10 (Figure 3). Based on this effectiveness estimate of overall survival, if all patients with HR+ cancer were treated in concordance with guidelines, approximately 14 630 lives would have been saved during the 10 years of the study period. We explored the heterogeneity effect of AET by several clinical factors and found the effect of AET was significant in all subgroups (Table) (P < .001). The effect of AET was similar across subgroups defined by age, node status, or chemotherapy, whereas it was weaker among patients treated with lumpectomy but without radiotherapy (hazard ratio, 0.84; 95% CI, 0.80-0.87). More than half these patients were older than 70 years (57.1%), and we have shown that the effect of AET is smaller for older patients (hazard ratio, 0.75; 95% CI, 0.74-0.77), possibly owing to limited life expectancy.

We investigated temporal changes and relevant factors associated with AET receipt in the United States from 2004 to 2013. Factors associated with AET receipt varied, and we found local treatments, including surgery and radiotherapy, to be the most important factor. In 2013, a considerable proportion of patients with HR+ cancer who did not receive AET (17.6%) and approximately 3% of patients with HR− cancer were inappropriately treated. Optimistically, from 2004 to 2013, AET underuse and misuse decreased. These findings demonstrate a significant improvement in AET receipt for patients with invasive breast cancer compared with a similar study using the NCDB for patients with ductal carcinoma in situ¹⁹ that found fewer than half of patients received AET in 2012. In addition, the rate receiving AET is similar to or better than the rates for prior studies examining AET initiation through other means.^15-18,29

We also demonstrated the importance of concordance with AET guidelines and found that AET use was associated with a 29% relative risk reduction in mortality. This finding expands that of prior literature regarding mortality and disease-free survival associated with discontinuation and nonadherence by examining AET initiation.^30,31 This estimate of AET effectiveness is in line with relative risks of 0.66 to 0.71 from randomized clinical trials.³

The NCDB also differs from SEER data in that the former is a convenience sample of Commission on Cancer–accredited hospitals and does not capture all cases in a defined region. Accredited hospitals have more general and cancer-related resources than nonaccredited hospitals.³² However, given its scope and larger number of cases, as compared with the SEER database, the NCDB can be used to assess national practice patterns.³² For example, we found that exact HR status was strongly associated with AET receipt, highest among patients with ER+/PR+ cancer, followed by those with ER+/PR− cancer, and lowest among those with ER−/PR+ cancer, suggesting that clinicians place more value on the ER. Patients with the ER−/PR+ tumors are being undertreated if not given AET.³³

Age also played a role in AET decision making. The low AET rate in younger patients has been found in prior studies of adherence and persistence.^34-36 Premenopausal women may be more likely to make a risk-benefit assessment rejecting AET because these therapies are associated with adverse effects. In addition, tamoxifen is a category D drug, and young women may prioritize fertility issues.³⁷ However, the significant benefit of AET may be underrecognized in these women who are increasingly choosing bilateral mastectomy.³⁸ In the group 80 years or older, this could reflect a consideration of comorbidities, life expectancy, and quality of life.^39,40

Local treatments appear to be key factors associated with likelihood of appropriate AET use. One possible explanation is that patients who did undergo radiotherapy may be more likely to receive standard-of-care therapy in general. In addition, with more physicians involved in a patient’s care, these patients would be more likely to be recommended for guideline-concordant care. Our results add to other studies of multidisciplinary cancer care teams that have shown improvement in adherence to recommended standards and best practices with a team-based approach.^41-45

We also investigated sociodemographic factors related to AET use. Black women were less likely to receive AET. This important finding could be contributing to the disparity in racial breast cancer survival, especially in the setting of an increasing incidence of HR+ tumors in this population.^20,46,47 Receipt of AET was also less likely in Hispanic and more likely in Asian women, a finding that is similar to those of other studies that have examined ethnic and racial differences in AET initiation and could contribute to the survival variation in these groups as well.^48-50 Recent evidence suggests that economic interventions aimed at lowering out-of-pocket costs could potentially help lessen racial disparities in AET use.⁵¹

Facility factors also played a role in AET receipt, for which the rate varied substantially by facility volume and geographic location. We found that facilities in West South Central states and low-volume institutions were more likely to misuse and underuse AET. These quality-of-care deficits could affect breast cancer mortality, because geographic differences have been well documented.^52-54

The study has limitations. Use of AET is known to be underreported in registry data because outpatient treatments are not always reported in hospital records. A calibration study in the SEER database suggested that the AET use was approximately 10% underreported,^55,56 and the NCDB may experience a similar magnitude of underreporting. The database also does not provide information on adherence, persistence, or whether AET was used as directed. In addition, other known NCDB limitations that could affect patients’ treatment decisions, including absence of the specific type of AET recommended, family history, and genomic profile.¹⁹ We are also not privy to the physician-patient discussion that occurred, and suboptimal communication affects AET patterns of care.^18,57,58 For example, in the B-QUAL study,¹⁸ noninitiation was less likely in those who found the quality of patient-physician communication to be higher. Friese et al⁵⁹ used SEER data and found that inadequate information about adverse effects detrimentally affects AET initiation.

This study reported nationwide data on AET practice patterns from 2004 to 2013. The secular patterns showed that appropriate AET use has been increasing, but optimal use has not been achieved. Our findings suggest that local treatment, including surgery and radiotherapy, may play a substantial role in AET decision making. Facilitation of multidisciplinary team–based care may help optimize guideline-concordant treatment by ensuring patients are not lost to follow-up and are recommended for evidence-based care.⁴⁵ The data also show that certain populations are being deprived of this life-saving therapy, including minorities and those on either end of the age spectrum. We hope that, with the coming launch of the Medicare Access and CHIP Reauthorization Act⁶⁰ and value-based reimbursement, the oncology community will take steps to close this quality gap and prevent the potential loss of lives, especially among young women who are undertreated with AET. Advances in cancer delivery, including the Oncology Care Model and oncology clinical pathways, could be the first step by linking incentive payments to guideline-concordant care.^61,62

Corresponding Author: Dezheng Huo, MD, PhD, Department of Public Health Sciences, University of Chicago, 5841 S Maryland Ave, MC 2000, Chicago, IL 60637 (dhuo@health.bsd.uchicago.edu).

Accepted for Publication: November 17, 2016.

Published Online: February 2, 2017. doi:10.1001/jamaoncol.2016.6380

Author Contributions: Dr Huo had full access to all 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: All authors.

Acquisition, analysis, or interpretation of data: Daly, Olopade, Hou, Yao, Huo.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: Daly, Olopade, Hou, Winchester, Huo.

Statistical analysis: Daly, Hou, Huo.

Obtained funding: Olopade.

Study supervision: Olopade, Yao, Huo.

Conflict of Interest Disclosures: Dr Daly reports serving in a leadership role in Quadrant Holdings; owning stock or other assets in Quadrant Holdings, CVS Health, Johnson & Johnson, McKesson, and Walgreens Boots Alliance; and receiving travel, accommodations, and expenses from Quadrant Holdings. Dr Olopade reports owning stock or other assets in CancerIQ. No other disclosures were reported.

Funding/Support: This study was supported in part by grants MRSG-13-063-01-TBG (Dr Huo) and CRP-10-119-01-CCE (Dr Olopade) from the American Cancer Society, a pilot grant from the University of Chicago (Dr Huo), and a T-32 basic research training grant in medical oncology from the National Institutes of Health/National Cancer Institute (Dr Daly).

Role of the Funder/Sponsor: The funding sources 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.