Treatment and monitoring options for patients with metastatic breast cancer (MBC) are increasing, but little is known about variability in care. We sought to improve understanding of MBC care and its correlates by analyzing real-world claims data using a search engine with a novel query language to enable temporal electronic phenotyping.

Using the Advanced Cohort Engine, we identified 6,180 women who met criteria for having estrogen receptor–positive, human epidermal growth factor receptor 2–negative MBC from IBM MarketScan US insurance claims (2007-2014). We characterized treatment, monitoring, and hospice usage, along with clinical and nonclinical factors affecting care.

We observed wide variability in treatment modality and monitoring across patients and geography. Most women received first-recorded therapy with endocrine (67%) versus chemotherapy, underwent more computed tomography (CT) (76%) than positron emission tomography-CT, and were monitored using tumor markers (58%). Nearly half (46%) met criteria for aggressive disease, which were associated with receiving chemotherapy first, monitoring primarily with CT, and more frequent imaging. Older age was associated with endocrine therapy first, less frequent imaging, and less use of tumor markers. After controlling for clinical factors, care strategies varied significantly by nonclinical factors (median regional income with first-recorded therapy and imaging type, geographic region with these and with imaging frequency and use of tumor markers; P < .0001).

Variability in US MBC care is explained by patient and disease factors and by nonclinical factors such as geographic region, suggesting that treatment decisions are influenced by local practice patterns and/or resources. A search engine designed to express complex electronic phenotypes from longitudinal patient records enables the identification of variability in patient care, helping to define disparities and areas for improvement.

Metastatic breast cancer (MBC), of which estrogen receptor (ER)–positive, human epidermal growth factor receptor 2 (HER2)–negative breast cancer is the most common subtype,1 is estimated to affect approximately 150,000 women in the United States2 and is the second leading cause of cancer death among women.3 Although MBC remains incurable, the past 3 decades have seen a remarkable proliferation of therapeutic options for MBC, which together have led to improvements in survival.4 However, the clinical trials that led to the approval of these therapies provided only limited evidence to guide their optimal use in routine practice; for example, there are minimal data on how to best monitor disease progression to guide change in therapy. Expert opinion guidelines are available,5,6 but it is not known how these guidelines are applied to the care of individual patients.


  • Key Objective

  • To classify a cohort of patients with metastatic breast cancer (MBC) according to five specified care events using an expressive temporal query search language applied to a large insurance claims data set.

  • Knowledge Generated

  • A claims data search engine designed to compile fragmented data into a single timeline object enables rapid and intuitive phenotyping of care events. Care events in MBC are associated with both clinical factors (eg, aggressive disease with early chemotherapy) and nonclinical factors (eg, lower median regional income with lower use of positron emission tomography-computed tomography).

  • Relevance

  • Variability in care in MBC across the United States is substantial; analysis of claims data can uncover this variability, including care events that might benefit from greater standardization, and factors influencing this variability, including potential disparities.

Data about how oncologists treat cancer in real-world practice come largely from four sources: institutional retrospective reviews, physician surveys, prospective registries, and insurance claims. Institutional reviews and physician surveys are informative, but small in size and limited to a subset of population.7-17 In the United States, registries such as the SEER database cannot currently be used to examine MBC care, as they collect details on initial therapy only, whereas the majority of patients with MBC develop metastatic disease after initial curative therapy.2 Some prospective registries in other countries include metastatic recurrence, but these are generally limited to academic centers or from a small geographic area.18-22 Claims data, which include diagnosis, procedure, and medication billing codes submitted to third-party payers, contain a wealth of information about a patient's interactions with the health care system and have potential to provide unprecedented insight into variability of care, and the factors that influence it, across a broad population.

Several studies have examined claims data to describe treatment sequences in MBC,23-26 and other work has analyzed claims and electronic health record data to characterize treatment patterns and comorbidities in cancer broadly.27 However, to our knowledge, claims data have not been used to systematically examine national variation in care patterns for MBC nor to understand the factors that contribute to this variation. Efforts to do so are hindered by challenges in searching large-scale, longitudinal patient records to identify specified electronic phenotypes. We overcome these challenges by using a novel clinical search engine, the Advanced Cohort Engine (ACE), which applies a temporal query language (TQL) to a patient object datastore to allow rapid search for defined clinical patterns and end points.28

In this study, we applied ACE to IBM MarketScan claims to identify a cohort of women who had claims consistent with estrogen receptor–positive, HER2-negative MBC. We defined electronic phenotypes for five events specific to the care of MBC: one related to treatment selection, three related to monitoring of disease progression, and one related to end-of-life care. Our objective was to identify patient clinical factors, such as age and disease aggressiveness, and nonclinical factors, such as geographic region and median income of the metropolitan area where claims were submitted, which contributed to the observed variability in these care events.

Data Source

We analyzed de-identified insurance claims from the IBM MarketScan Research Database29 from 2007-2014, including more than 125 million unique individuals in the United States with insurance through their employers. Because death information is unavailable in the MarketScan database, we also analyzed the smaller Optum Clinformatics claims data set,30 consisting of claims records linked to death information derived from the electronic health record, in an exploratory analysis related to end of life.

Building Patient Cohorts

Consistent with previous studies,24,31-35 we adopted the validated approach of using at least two instances of a secondary malignant neoplasm code occurring after at least two instances of a breast cancer code to indicate the presence of MBC (Appendix Table A1). We did not use whole-body imaging and chemotherapy codes to identify MBC31,32 as the primary focus of our work was to evaluate variability in monitoring and treatment. We required that there is no code for HER2-targeted therapy (to exclude presumptive HER2-positive breast cancer), at least one code for an endocrine therapy (to exclude presumptive triple-negative breast cancer), at least 1 year of follow-up after first secondary malignant neoplasm code, and nonmissing data for metropolitan statistical area.

We used the Stanford ACE, a search engine developed to enable scalable and rapid search of longitudinal patient data,36,37 to construct the MBC cohort using the above definition. ACE natively supports temporal search criteria that enable complex cohort definitions to be executed over an in-memory datastore of patient records.28 ACE manages the complexity of large claims databases by compiling claims data fragmented across multiple tables into a single timeline object for each enrollee and using a custom in-memory datastore. The resulting subsecond query response times (even over claims data for tens of millions of enrollees28) combined with ACE's expressive TQL allow users to integrate electronic phenotyping (the process of defining the necessary and sufficient data elements to identify a clinical condition or event) with cohort building into a single iterative process for phenotype validation, data retrieval, and analysis. ACE's TQL uses a schema-agnostic query algebra that natively handles operations over event durations and time intervals (Fig 1).

Exposure Variable Definitions

We defined aggressive MBC as the presence of at least one code for visceral (lung or liver) or CNS metastasis within one year of the first MBC code (Appendix Table A1). Age and calendar year of MBC diagnosis were directly available. We calculated the Charlson Comorbidity Index (CCI) for each patient using the comorbidity R package.38 We used geographic region, defined as the Census Bureau–designated division39 of the provider who recorded the first MBC code, as an available proxy for other unmeasured nonclinical factors. Median regional income category was determined by linking metropolitan statistical area of the provider to 2010 Census Bureau40 income data, thus corresponding to the median regional income where the care was received.

Outcome Definitions

We defined electronic phenotypes for five care events: (1) first-recorded treatment after MBC, defined as the first code for an endocrine or chemotherapy after first secondary malignancy code; (2) majority imaging modality, defined as whether there were more computed tomography (CT) or PET-CT codes after first secondary malignancy code; (3) imaging frequency, defined as whether the average interval between CT and PET-CT scans was below or above the cohort's median between-scan interval; (4) serum tumor marker usage, defined as any code for CA 15-3 or CA 27-29 testing after first secondary malignancy code; and (5) hospice utilization, defined as the presence of one of the several hospice services, provider type, revenue, or Current Procedural Terminology (CPT) codes41 (Appendix Table A1). Because patients in IBM MarketScan might have been alive at end of record (effectively censored from further follow-up), the observed hospice utilization may represent a lower bound on the true utilization. We therefore also ascertained hospice utilization in the smaller Optum cohort of patients meeting criteria for MBC, where death information is available, to assess whether the rate of hospice utilization observed in IBM MarketScan was an underestimate.

Statistical Analysis

We used Cox proportional hazards regression to examine the correlation of aggressive disease with time to end of record. For univariate analyses of associations between characteristics and care events, we used chi-squared tests for categorical variables and t-tests for continuous variables, accounting for multiple hypothesis testing with the Bonferroni approach (reporting as statistically significant those associations with P < .002). For multivariate analyses, we created logistic regression models with care events as outcome variables and aggressive disease status, CCI, age at MBC, year of MBC, and either geographic division or median regional income (analyzed separately as each derived from provider location) as predictor variables, reporting odds ratios and 95% CIs. We assessed for each event whether a logistic regression including geographic region or median regional income explained a significantly greater proportion of its variance than a logistic regression without these covariates, reporting the analysis of deviance chi-squared P value and using the Bonferroni approach (reporting as statistically significant P values < 0.005). Statistical analyses were performed using R version 3.5.2.

Cohort Characteristics

The analyzed cohort was n = 6,180 (Fig 2), with a median follow-up from first MBC code to end of record 2 years. Each patient's ACE-defined MBC timeline (Fig 3) consisted of demographics as well as diagnoses, procedures, and medications. The median age at presumptive diagnosis of MBC was 59 years, and nearly half of patients (46%) had presumptive aggressive disease, which correlated with shorter time to end of record (hazard ratio, 1.29; 95% CI, 1.22 to 1.35).

Care Events

First-recorded therapy after MBC was endocrine for most (67% v 33% chemotherapy-first) (Table 1). We observed 719 unique paths through treatment in the endocrine-first group and 550 unique paths in the chemotherapy-first group (Fig 4). The most common first-recorded endocrine therapy was an aromatase inhibitor (38%), followed by fulvestrant (13%); the most common first-recorded chemotherapy was a taxane (10%), followed by combination chemotherapy (7.3%) and single-agent capecitabine (6.9%). Nineteen percent received only one documented treatment after metastasis, most commonly an aromatase inhibitor (17%).


TABLE 1. Univariate Analyses of the Associations Between Each Care Event and Various Clinical and Nonclinical Factors

For monitoring, most patients received more CT scans than PET-CT scans (76% majority CT scans v 24% majority PET-CT scans) (Table 1), with a median between-scan interval of 81 days. Serum tumor markers were used for 58% (Table 1). As the percentage with a hospice code in the MarketScan cohort (28%), the Optum cohort of all MBC patients (n = 9,991) (25%), and the Optum cohort of MBC patients with confirmed death (n = 3,647) (25%) were very similar, we used the larger MarketScan cohort to assess factors contributing to hospice utilization.

Factors Correlated with Care Events

We examined the patient-specific (CCI and age), disease (aggressiveness), nonclinical (geographic region and median regional income), and other (year of diagnosis) factors associated with each of the five care events. In univariate analyses (Table 1, all P < .0001), chemotherapy as first-recorded therapy was more common in patients who were younger or had aggressive disease. PET-CT was more often used for patients without aggressive disease and usage increased with later calendar year of diagnosis. Imaging was more frequent (relative to the median) in patients who were younger or who had aggressive disease and in later years. Tumor marker use was more common in patients with fewer comorbidities and younger age. Hospice usage was higher in patients with aggressive disease and in younger patients.

With the exception of hospice utilization (P = .0044, not significant after adjustment for multiple hypothesis testing), all care events were significantly associated with geographic region (P < .0001). We next examined all patient and disease factors in a multivariate model for each care event (Fig 5; Appendix Table A2), in particular to understand whether geographic region contributed to the observed variability in care after controlling for patient-specific factors. We found that in the multivariate models, geography explained a statistically significant proportion of the variance in the use of first-recorded therapy (chemotherapy v endocrine therapy), use of PET-CT v CT, imaging frequency, and use of tumor markers (analysis of deviance chi-squared P < .0001 for all; Appendix Table A2). We also examined median regional income on the basis of metropolitan statistical area: this value, like geographic region, is derived from provider location but has greater granularity (Appendix Table A3). Median regional income also explained a significant amount of variance in the use of first-recorded therapy, with lower-income areas associated with greater use of chemotherapy first (P = .0043), and in imaging modality, with lower-income areas associated with the lower use of PET-CT (P = .00049).

Through analysis of insurance claims data that cover nearly half the US population, we discovered substantial variability in the care of patients who met claims-based criteria for estrogen receptor–positive, HER2–negative MBC diagnosis, with > 1,200 unique treatment sequences observed across the cohort. Using the search engine ACE, we classified the cohort according to five electronic phenotypes, uncovering that sociodemographic factors explain a significant amount of variability in MBC care. To our knowledge, this is the first study to characterize correlates of care variability at the national level and provides the most complete picture to date of determinants of MBC care variability in the United States.

Some of the variability in MBC care was related to patient characteristics in expected ways. For example, early use of chemotherapy was more common in patients who were younger or who met criteria for aggressive disease (early visceral or CNS metastasis)9-12,17,19-21; imaging was more frequent in these populations as well. PET-CT was more commonly used when disease did not meet aggressiveness criteria, perhaps because of its potential utility in detecting progression in bone-only metastasis, although no guidelines suggest the use of one imaging technology over another in this setting.42,43 We also observed changes in MBC care over time, with more frequent imaging and more PET-CT usage in later years, pointing toward increasing cost of care. These findings fit with previously observed trends of increasing number of imaging studies per patient in early- and advanced-stage cancer over time44 and increasing use of PET-CT over time in the Medicare population.45-47 The lower hospice utilization in patients older than 65 may reflect incomplete follow-up, as their end of record may in some cases reflect withdrawal from private insurance and sole usage of Medicare.

Importantly, some of the observed variability in care appeared to be associated with nonclinical characteristics. Four of five care events were significantly associated with geographic region, used as a proxy for unmeasured health care system and sociodemographic factors, including local practice patterns, insurance contracting patterns, health care system structure (such as community v academic practices), racial or ethnic composition of patients served by hospitals, and socioeconomic status. We found in multivariable analyses that early chemotherapy use was more common and PET-CT usage was less common in lower-income areas. Previous work has also pointed to the uneven use of PET-CT in the Medicare population, including greater usage among White patients and higher socioeconomic groups.48,49 Given the large size of the geographic regions, with only nine covering the entire United States, between-region variability in unmeasured nonclinical factors was likely muted by within-region averaging effects; thus, the contribution of nonclinical health care system and sociodemographic factors to variability in MBC care is likely greater than we report here. Recent work has suggested that improvements in survival of de novo stage IV MBC have benefited patients in some geographic regions more than others.50 It is plausible that the differential application of therapies and monitoring approaches as described here may contribute to disparities in MBC survival.51

Our study's characterization of MBC care variability and the factors that influence it was enabled by ACE, a clinical search engine designed to enable fast and scalable execution of electronic phenotypes from longitudinal patient records.28 This tool allowed us to define patient cohorts on the basis of care events; these cohorts could then be compared against multiple clinical and nonclinical variables to learn what factors explained the identified variability in each care event. The unique intersection of informatics, data science, and cancer care research using national-level claims data was facilitated by ACE's ability to express and execute complex temporal search criteria using a query language designed for that purpose and the efficiency of data retrieval made possible by its in-memory data store. Clinical research questions often require following a patient's trajectory, and ACE's ability to piece together disparate fragments across time to allow simple clinical queries should have broad applicability to efforts to characterize variability in clinical care across the population, to identify disparities, and to understand aspects of care that could benefit from standardization and optimization.

This study has several limitations. First, claims data lack information on some key factors that may contribute to care variability, such as patient and tumor-specific prognostic factors, sociodemographic factors, and health care system factors. Our approach to classify aggressive disease could not be validated with the current data, but we note that its strong correlation with time to end of record suggests that it does capture disease biology and that the proportion of patients identified as having aggressive disease (46%) was similar to prospective studies where 50%-60% of patients had visceral metastasis.20,21 Second, incomplete information both before start of record and after end of record is a challenge of claims data and contributes to the large drop in cohort size from the initial data set to the analytic cohort (nearly 60% of our cohort was excluded because of under one year of follow-up). Although uncertainty of the exact date of diagnosis of MBC meant that the first-recorded therapy after the first MBC code might not have always been the first-line metastatic therapy, we note that the results were broadly consistent with those reported from registries and institutional reviews: in this cohort, first-recorded therapy was endocrine (as recommended by guidelines5,6) for 67%, similar to what has been reported from retrospective reviews and prospective registries.9-12,14,16,17,19-21 Third, we analyzed the time period from 2007 to 2014, which was before the introduction of CDK4/6 inhibitors for first-line treatment of MBC52; we therefore cannot comment on how these treatments have changed the landscape of care. Finally, we lacked data on what factors might influence care decision making, and although no single or set of geographic regions consistently drove the observed variability, it remains possible that regional differences between billing or coding practices could explain some of the regional variation. These study limitations are offset by considerable strengths, notably the temporal integration of detailed treatment and monitoring data for a large national sample of patients.

This study represents the first application of ACE, a recently developed clinical search engine,28 to understand national variability of cancer care from insurance claims data, and its results emphasize the power of this method in defining electronic phenotypes from large cohorts of longitudinal patient data. Although treatment and monitoring variability are appropriate when based on patient- and disease-specific factors that benefit from personalized approaches, the existence of considerable geographic and income-related variability strongly suggests that nonpertinent factors contribute to unwarranted variability in MBC care. Additional investigation is needed to improve the tailoring of care for all patients with MBC.

© 2021 by American Society of Clinical Oncology

Presented in part at the ASCO 2020 Annual Meeting, May 29-30, 2020.


Supported by the Breast Cancer Research Foundation (G.W.S. and A.W.K.), the Suzanne Pride Bryan Fund for Breast Cancer Research (A.W.K.), the Jan Weimer Faculty Chair in Breast Oncology (A.W.K.), the BRCA Foundation (A.W.K.), and the National Institutes of Health/National Cancer Institute (K08CA252457 to J.L.C.-J.).

Conception and design: Jennifer L. Caswell-Jin, Alison Callahan, Natasha Purington, Haruka Itakura, Nigam H. Shah, Allison W. Kurian

Financial support: Allison W. Kurian

Provision of study materials or patients: Alison Callahan, Nigam H. Shah

Collection and assembly of data: Alison Callahan

Data analysis and interpretation: Jennifer L. Caswell-Jin, Alison Callahan, Natasha Purington, Summer S. Han, Haruka Itakura, Esther M. John, Douglas W. Blayney, George W. Sledge, Allison W. Kurian

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

The following represents disclosure information provided by the authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to or

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

Jennifer L. Caswell-Jin

Research Funding: QED Therapeutics

Uncompensated Relationships: Natera

Uncompensated Relationships: Color Genomics

Alison Callahan

Consulting or Advisory Role: Atropos Health LLC

Patents, Royalties, Other Intellectual Property: I am named as an inventor on a patent pending for a search engine for patient data

Douglas W. Blayney

Leadership: Artelo Biosciences

Stock and Other Ownership Interests: Artelo Biosciences, Madorra

Consulting or Advisory Role: Creare, Daiichi Sankyo, Embold Health, Lilly, Google, G1 Therapeutics, Merck

Research Funding: BeyondSpring Pharmaceuticals, Creare

Open Payments Link:

George W. Sledge

Leadership: Syndax, Tessa Therapeutics

Stock and Other Ownership Interests: Syndax, Tessa Therapeutics, Pionyr

Consulting or Advisory Role: Symphogen, Synaffix, Syndax, Verseau Therapeutics, Grail, AstraZeneca

Research Funding: Genentech/Roche, Pfizer

Travel, Accommodations, Expenses: Verseau Therapeutics, Tessa Therapeutics

Nigam H. Shah

Leadership: Prealize, Atropos Health

Consulting or Advisory Role: Apixio

Research Funding: Google

Allison W. Kurian

Research Funding: Myriad Genetics

Other Relationship: Ambry Genetics, Color Genomics, GeneDx/BioReference, Invitae, Genentech

No other potential conflicts of interest were reported.


TABLE A1. Insurance Claims Codes Used for Variable Definitions


TABLE A2. Multivariate Analyses by Geographic Region


TABLE A3. Multivariate Analyses by Median Regional Income

1. Tao L, Chu L, Wang LI, et al: Occurrence and outcome of de novo metastatic breast cancer by subtype in a large, diverse population. Cancer Causes Control 27:1127-1138, 2016 Crossref, MedlineGoogle Scholar
2. Mariotto AB, Etzioni R, Hurlbert M, et al: Estimation of the number of women living with metastatic breast cancer in the United States. Cancer Epidemiol Biomarkers Prev 26:809-815, 2017 Crossref, MedlineGoogle Scholar
3. Siegel RL, Miller KD, Jemal A: Cancer statistics, 2020. CA Cancer J Clin 70:7-30, 2020 Crossref, MedlineGoogle Scholar
4. Caswell-Jin JL, Plevritis SK, Tian L, et al: Change in survival in metastatic breast cancer with treatment advances: Meta-analysis and systematic review. JNCI Cancer Spectr 2:pky062, 2018 Crossref, MedlineGoogle Scholar
5. Rugo HS, Rumble RB, Macrae E, et al: Endocrine therapy for hormone receptor-positive metastatic breast cancer: American Society of Clinical Oncology guideline. J Clin Oncol 34:3069-3103, 2016 LinkGoogle Scholar
6. Gradishar WJ, Anderson BO, Abraham J, et al: Breast cancer, version 3.2020, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 18:452-478, 2020 Crossref, MedlineGoogle Scholar
7. Seah DS, Luis IV, Macrae E, et al: Use and duration of chemotherapy in patients with metastatic breast cancer according to tumor subtype and line of therapy. J Natl Compr Canc Netw 12:71-80, 2014 Crossref, MedlineGoogle Scholar
8. Macalalad AR, Hao Y, Lin PL, et al: Treatment patterns and duration in post-menopausal women with HR+/HER2− metastatic breast cancer in the US: A retrospective chart review in community oncology practices (2004-2010). Curr Med Res Opin 31:263-273, 2015 Crossref, MedlineGoogle Scholar
9. Bonotto M, Gerratana L, Di Maio M, et al: Chemotherapy versus endocrine therapy as first-line treatment in patients with luminal-like HER2-negative metastatic breast cancer: A propensity score analysis. Breast 31:114-120, 2017 Crossref, MedlineGoogle Scholar
10. Lobbezoo DJ, van Kampen RJ, Voogd AC, et al: In real life, one-quarter of patients with hormone receptor-positive metastatic breast cancer receive chemotherapy as initial palliative therapy: A study of the Southeast Netherlands Breast Cancer Consortium. Ann Oncol 27:256-262, 2016 Crossref, MedlineGoogle Scholar
11. Andre F, Neven P, Marinsek N, et al: Disease management patterns for postmenopausal women in Europe with hormone-receptor-positive, human epidermal growth factor receptor-2 negative advanced breast cancer. Curr Med Res Opin 30:1007-1016, 2014 Crossref, MedlineGoogle Scholar
12. Watanabe J, Hayashi T, Tadokoro Y, et al: Clinical pattern of primary systemic therapy and outcomes of estrogen receptor-positive, HER2-negative metastatic breast cancer: A review of a single institution. Breast Cancer Res Treat 166:911-917, 2017 Crossref, MedlineGoogle Scholar
13. Zanotti G, Hunger M, Perkins JJ, et al: Treatment patterns and real world clinical outcomes in ER+/HER2- post-menopausal metastatic breast cancer patients in the United States. BMC Cancer 17:393, 2017 Crossref, MedlineGoogle Scholar
14. Kurosky SK, Mitra D, Zanotti G, et al: Treatment patterns and outcomes of patients with metastatic ER(+)/HER-2(−) breast cancer: A multicountry retrospective medical record review. Clin Breast Cancer 18:e529-e538, 2018 Crossref, MedlineGoogle Scholar
15. Goldschmidt D, Dalal AA, Romdhani H, et al: Current treatment patterns among postmenopausal women with HR+/HER2− metastatic breast cancer in US Community Oncology Practices: An observational study. Adv Ther 35:482-493, 2018 Crossref, MedlineGoogle Scholar
16. Walbaum B, Acevedo F, Medina L, et al: First-line endocrine therapy for advanced breast cancer. A real-world study at a Latin American University Health Institution. Curr Med Res Opin 36:1195-1199, 2020 Crossref, MedlineGoogle Scholar
17. Twelves C, Cheeseman S, Sopwith W, et al: Systemic treatment of hormone receptor positive, human epidermal growth factor 2 negative metastatic breast cancer: Retrospective analysis from Leeds Cancer Centre. BMC Cancer 20:53, 2020 Crossref, MedlineGoogle Scholar
18. Deluche E, Antoine A, Bachelot T, et al: Contemporary outcomes of metastatic breast cancer among 22,000 women from the multicentre ESME cohort 2008-2016. Eur J Cancer 129:60-70, 2020 Crossref, MedlineGoogle Scholar
19. Hartkopf AD, Huober J, Volz B, et al: Treatment landscape of advanced breast cancer patients with hormone receptor positive HER2 negative tumors—Data from the German PRAEGNANT breast cancer registry. Breast 37:42-51, 2018 Crossref, MedlineGoogle Scholar
20. Fietz T, Tesch H, Rauh J, et al: Palliative systemic therapy and overall survival of 1,395 patients with advanced breast cancer—Results from the prospective German TMK cohort study. Breast 34:122-130, 2017 Crossref, MedlineGoogle Scholar
21. Jacquet E, Lardy-Cleaud A, Pistilli B, et al: Endocrine therapy or chemotherapy as first-line therapy in hormone receptor-positive HER2-negative metastatic breast cancer patients. Eur J Cancer 95:93-101, 2018 Crossref, MedlineGoogle Scholar
22. Simon J, Chaix M, Billa O, et al: Survival in patients with HR+/HER2− metastatic breast cancer treated with initial endocrine therapy versus initial chemotherapy. A French population-based study. Br J Cancer 123:1071-1077, 2020 Crossref, MedlineGoogle Scholar
23. Rocque GB, Kandhare PG, Williams CP, et al: Visualization of sequential treatments in metastatic breast cancer. JCO Clin Cancer Inform 3:1-8, 2019 MedlineGoogle Scholar
24. Rocque GB, Gilbert A, Williams CP, et al: Prior treatment time affects survival outcomes in metastatic breast cancer. JCO Clin Cancer Inform 4:500-513, 2020 LinkGoogle Scholar
25. Swallow E, Zhang J, Thomason D, et al: Real-world patterns of endocrine therapy for metastatic hormone-receptor-positive (HR+)/human epidermal growth factor receptor-2-negative (HER2−) breast cancer patients in the United States: 2002-2012. Curr Med Res Opin 30:1537-1545, 2014 Crossref, MedlineGoogle Scholar
26. Goyal RK, Carter GC, Nagar SP, et al: Treatment patterns, survival and economic outcomes in Medicare-enrolled, older patients with HR+/HER2− metastatic breast cancer. Curr Med Res Opin 35:1699-1710, 2019 Crossref, MedlineGoogle Scholar
27. Chen R, Ryan P, Natarajan K, et al: Treatment patterns for chronic comorbid conditions in patients with cancer using a large-scale observational data network. JCO Clin Cancer Inform 4:171-183, 2020 LinkGoogle Scholar
28. Callahan A, Polony V, Posada JD, et al: ACE: An Advanced Cohort Engine for searching longitudinal patient records. J Am Med Inform Assoc, 13:ocab27, 2021 Google Scholar
29. IBM: Google Scholar
30. Optum: Clinformatics® Data Mart. Google Scholar
31. Ritzwoller DP, Hassett MJ, Uno H, et al: Development, validation, and dissemination of a breast cancer recurrence detection and timing informatics algorithm. J Natl Cancer Inst 110:273-281, 2018 Crossref, MedlineGoogle Scholar
32. Uno H, Ritzwoller DP, Cronin AM, et al: Determining the time of cancer recurrence using claims or electronic medical record data. JCO Clin Cancer Inform 2:1-10, 2018 LinkGoogle Scholar
33. Chubak J, Onega T, Zhu W, et al: An electronic health record-based algorithm to ascertain the date of second breast cancer events. Med Care 55:e81-e87, 2017 Crossref, MedlineGoogle Scholar
34. Danese MD, Lindquist K, Doan J, et al: Effect of central nervous system metastases on treatment discontinuation and survival in older women receiving trastuzumab for metastatic breast cancer. J Cancer Epidemiol 2012:819210, 2012 Crossref, MedlineGoogle Scholar
35. Stokes ME, Thompson D, Montoya EL, et al: Ten-year survival and cost following breast cancer recurrence: Estimates from SEER-medicare data. Value Health 11:213-220, 2008 Crossref, MedlineGoogle Scholar
36. Schuler A, Callahan A, Jung K, et al: Performing an informatics consult: Methods and challenges. J Am Coll Radiol 15:563-568, 2018 Crossref, MedlineGoogle Scholar
37. Gombar S, Callahan A, Califf R, et al: It is time to learn from patients like mine. NPJ Digit Med 2:16, 2019 Crossref, MedlineGoogle Scholar
38. Gasparini A: comorbidity: An R package for computing comorbidity scores. J Open Source Softw 3:648, 2018 CrossrefGoogle Scholar
39. United States Census Bureau: Census Regions and Divisions of the United States. Google Scholar
40. Census Bureau: Google Scholar
41. Lindley LC, Cohrs AC, Keim-Malpass J, et al: Children enrolled in hospice care under commercial insurance: A comparison of different age groups. Am J Hosp Palliat Care 36:123-129, 2019 Crossref, MedlineGoogle Scholar
42. Morris PG, Lynch C, Feeney JN, et al: Integrated positron emission tomography/computed tomography may render bone scintigraphy unnecessary to investigate suspected metastatic breast cancer. J Clin Oncol 28:3154-3159, 2010 LinkGoogle Scholar
43. Woolf DK, Padhani AR, Makris A: Assessing response to treatment of bone metastases from breast cancer: What should be the standard of care? Ann Oncol 26:1048-1057, 2015 Crossref, MedlineGoogle Scholar
44. Loggers ET, Fishman PA, Peterson D, et al: Advanced imaging among health maintenance organization enrollees with cancer. JCO Oncol Pract 10:231-238, 2014 LinkGoogle Scholar
45. Hillner BE, Tosteson AN, Song Y, et al: Growth in the use of PET for six cancer types after coverage by medicare: Additive or replacement? J Am Coll Radiol 9:33-41, 2012 Crossref, MedlineGoogle Scholar
46. Veenstra CM, Vachani A, Ciunci CA, et al: Trends in the use of (18)F-fluorodeoxyglucose PET imaging in surveillance of non-small-cell lung and colorectal cancer. J Am Coll Radiol 13:491-496, 2016 Crossref, MedlineGoogle Scholar
47. Dinan MA, Curtis LH, Hammill BG, et al: Changes in the use and costs of diagnostic imaging among Medicare beneficiaries with cancer, 1999-2006. JAMA 303:1625-1631, 2010 Crossref, MedlineGoogle Scholar
48. Onega T, Tosteson TD, Wang Q, et al: Geographic and sociodemographic disparities in PET use by Medicare beneficiaries with cancer. J Am Coll Radiol 9:635-642, 2012 Crossref, MedlineGoogle Scholar
49. Dinan MA, Curtis LH, Carpenter WR, et al: Variations in use of PET among Medicare beneficiaries with non-small cell lung cancer, 1998-2007. Radiology 267:807-817, 2013 Crossref, MedlineGoogle Scholar
50. Malmgren JA, Calip GS, Atwood MK, et al: Metastatic breast cancer survival improvement restricted by regional disparity: Surveillance, Epidemiology, and End Results and institutional analysis: 1990 to 2011. Cancer 126:390-399, 2020 Crossref, MedlineGoogle Scholar
51. Dawood S, Broglio K, Gonzalez-Angulo AM, et al: Trends in survival over the past two decades among white and black patients with newly diagnosed stage IV breast cancer. J Clin Oncol 26:4891-4898, 2008 LinkGoogle Scholar
52. Finn RS, Martin M, Rugo HS, et al: Palbociclib and letrozole in advanced breast cancer. N Engl J Med 375:1925-1936, 2016 Crossref, MedlineGoogle Scholar


No companion articles


DOI: 10.1200/CCI.21.00031 JCO Clinical Cancer Informatics no. 5 (2021) 600-614. Published online May 27, 2021.

PMID: 34043432

ASCO Career Center