Defining prognostic subgroups and treatment outcomes in estrogen receptor low-positive de novo metastatic breast cancer.

Introduction
Estrogen receptor (ER) status is a key prognostic factor in breast cancer, with ER-positive patients generally having superior outcomes compared to patients with ER-negative disease [1, 2]. In 2010, the American Society of Clinical Oncology and the College of American Pathologists (ASCO/CAP) guidelines introduced a new ER category, recommending that tumors with 1–10% ER expression by immunohistochemistry be classified as ER “low-positive” [3]. Large clinical cohorts among individuals with early-stage disease have observed that ER low-positive tumors have a prognosis similar to ER-negative tumors [4, 5]. Prognostic factors identified among patients with ER low-positive early-stage disease include age, comorbidities, grade, tumor infiltrating lymphocytes, tumor (T) stage, and nodal (N) stage [6, 7].
ER status is also a predictive marker of endocrine therapy, with therapeutic benefits limited to patients with ER-positive disease [1]. While patients with ER low-positive tumors are eligible for endocrine therapy according to National Comprehensive Cancer Network (NCCN) Guidelines [2], most of these tumors appear to behave similar to ER-negative tumors [8–11]. The NCCN guidelines do not specify which patients with ER-low positive disease could be managed in alignment with ER-negative disease, ER-positive disease, or using a dual strategy [2]. In ER low-positive early-stage patients, studies have shown an improvement in outcomes with the addition of adjuvant endocrine therapy [6, 12, 13] and the addition of immunotherapy to neoadjuvant chemotherapy [14, 15]. These observations suggest both ER-positive and ER-negative treatment strategies may be relevant to consider in ER low-positive patients.
To our knowledge, no published studies have evaluated prognosis or treatment outcomes for patients with ER low-positive tumors in the metastatic setting. Identifying subgroups of patients with a better prognosis compared to ER-negative disease could enable the use of less aggressive initial treatment approaches in some patients with ER low-positive metastatic disease. Our study’s objective was to address these knowledge gaps by comparing outcomes of ER low-positive patients to ER-negative and ER-positive de novo metastatic breast cancer (dnMBC) patients using the National Cancer Database (NCDB) [16]. Additionally, we evaluated for subgroups of ER low-positive patients with improved survival compared to ER-negative patients. Finally, we compared ER-negative, ER-positive, and dual treatment strategies among ER low-positive patients.

Methods

Design
We established a retrospective cohort using data from the NCDB. This clinical cancer database is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society, capturing information on > 70% of newly diagnosed patients with cancer in the United States [16]. The Johns Hopkins University School of Medicine Institutional Review Board deemed this research exempt from review, as all data in the NCDB are de-identified.

Participants
We outline the development of the study cohort in Fig. 1.
We included all adult patients with human epidermal receptor-2 (HER2)-negative, pathologically confirmed breast cancer that had dnMBC diagnosed between 2018 and 2021. This period was selected because ER percentage reporting began in 2018, and 2021 is the most recent year with available survival data. All patients had stage IV disease at diagnosis, i.e., dnMBC. We considered patients with a prior primary tumor (benign or malignant). We excluded patients reported to have had “no treatment or treatment decisions at facility,” positive or unknown HER2 summary status, and unknown ER percentage. We excluded those with a discrepancy between ER percentage and ER summary (which categorizes a patient as ER-positive or ER-negative) [17, 18], in alignment with other NCDB studies [6].

Primary exposure
Our primary exposure was ER status. We defined ER low-positive as 1–10% by immunohistochemistry per 2020 ASCO/CAP guidelines [1]. We defined ER-negative as 0% and ER-positive as 11–100%. The NCDB reported ER as an exact percentage or decile range, based on pretreatment biopsy when available [18]. The NCDB reported ER percentage from nodal or metastatic tissue if primary tumor data were unavailable [17]. If multiple tumors were present, the NCDB reported the ER level from the largest tumor. For cases with multiple biopsies of the same tumor, the NCDB reported the highest ER level [17].

Outcome
Our primary outcome was overall survival. Participating registries annually report follow-up to the NCDB [19]. The CoC accreditation standards require an “90% follow-up rate for all living, eligible, analytic patients diagnosed within the last 5 years or program’s first accreditation data (whichever is shorter) and an 80% follow-up for all eligible analytic cases from the most current year of completed cases through 15 years before or the cancer registry’s reference date” [19]. Cause of death data were not available.

Covariates
Sociodemographic factors included age at diagnosis (exact unless 90 years or older), sex, and self-identified race/ethnicity (White, Black, Asian/Asian Pacific Islander, and other/unknown (“American Indian, Aleutian, or Eskimo”, “Other” or “Unknown”) [20]. We categorized patients as with or without additional comorbidities using the Charlson-Deyo Score, which is based on ICD-10 codes, excluding cancer history [19].
We defined progesterone receptor (PR) status as negative (< 1%) or positive (1–100%) [21]. We categorized histology based on the International Classification of Diseases for Oncology, Third Edition [20]. We created a variable based on sites of metastatic disease to delineate bone-only metastases, given its prognostic significance [22]. Grade/differentiation was defined based on pathology report as low, intermediate, or high or, if grade was not available, well-, moderately-, or poorly differentiated/undifferentiated) [23].

First-course treatments
Information was available on first-course endocrine and/or chemotherapy treatment receipt and timing (days from diagnosis to start of treatment) using the same codes as the SEER Rx database [20, 24]. We defined “no treatment receipt” as a lack of receipt of chemotherapy or endocrine therapy.
Several assumptions were made to define first-course treatments, which were used in an exploratory analysis restricted to ER low-positive patients. “Chemotherapy” in the NCDB can include cytotoxic chemotherapy and/or cyclin-dependent kinases 4 and 6 inhibitors (CDK4/6i). Given that CDK4/6i are not prescribed as monotherapy in first-line dnMBC treatment [25], we presumed patients who received “chemotherapy” without endocrine therapy received cytotoxic chemotherapy as their first-course treatment. We used the time from diagnosis to treatment to define two types of presumed treatment plans for patients recorded as receiving first-line “chemotherapy” and endocrine therapy. If a patient’s “chemotherapy” start date was before their endocrine therapy start date, they were presumed to have received cytotoxic chemotherapy followed by maintenance endocrine therapy (± CDK4/6i). We presumed that patients who received “chemotherapy” on or after their endocrine therapy start date received endocrine therapy + CDK4/6i (no cytotoxic chemotherapy). These assumptions are supported by real-world data showing nearly all metastatic breast cancer patients receive cytotoxic chemotherapy without concurrent endocrine therapy [26]. We classified immunotherapy as received or not in an additional exploratory analysis, given that in 2020, the Food and Drug Administration approved immunotherapy (pembrolizumab) for metastatic triple-negative (HER2-negative, ER-negative, PR-negative) breast cancer patients with PD-L1 expression [27]. In summary, among ER low-positive patients, we classified patients as receiving cytotoxic chemotherapy, cytotoxic chemotherapy followed by endocrine-based maintenance treatment (± CDK4/6 inhibitor), endocrine therapy + CDK4/6 inhibitor, endocrine monotherapy, and no systemic treatment.

Statistical analysis
Descriptive statistics were reported using mean and standard deviation for normally distributed continuous variables and median with interquartile range for skewed distributions. Categorical variables were reported using frequency and percentages.
We defined overall survival as the number of months from diagnosis to date of death or censoring (date of last contact or end of follow-up), whichever occurred first. We used the Kaplan–Meier method to assess crude cumulative survival over the follow-up period stratified by ER status (negative, low-positive, positive). In our primary analysis, we used Cox proportional hazards regression to estimate hazard ratios (HR) with 95% confidence intervals (95% CI) for the association between ER status and overall survival. The proportional hazards assumption was met using Schoenfeld residual tests and complementary log–log plots.
To build our multivariable model, we evaluated clinical features, sociodemographics, comorbidities, and treatment receipt. We did not add detailed treatments due to different treatment recommendations for ER-negative and ER-positive patients [2]. Covariates were eligible for inclusion if they had < 5% missing data. We kept covariates in the model if they were known prognostic factors based on the published literature, and/or significantly added to the model based on the Wald test. Patients with missing covariate information were excluded from multivariable analysis. Subgroup analyses were performed by age and race.
Multivariable analyses were repeated among ER low-positive patients, additionally adjusting for detailed treatments to identify potential prognostic subgroups. Next, based on these results, we conducted analyses among the entire cohort, with ER low-positive patients stratified by PR status.

Exploratory analyses
The variable for grade/differentiation was not included in our primary analysis as it had a substantial amount of missing data (Table 1). However, we conducted an additional analysis assessing the association of ER and ER-PR status with survival, adjusting for grade/differentiation. The analysis was conducted given prior literature demonstrating that PR-negative tumors tend to be higher grade than PR-positive tumors [28, 29], and grade has been shown to be a prognostic marker in ER-low positive early-stage disease [6]. We kept those with missing grade in the model as a separate level of the category so that we could compare the results to the primary analysis. We conducted a second analysis where we stratified the ER-negative patients by PR status. While rare, ER-negative/PR-positive tumors can also be considered for endocrine therapies per NCCN guidelines [2].

Exploratory treatment analysis
We assessed overall survival by first-course treatments among patients with ER low-positive dnMBC who received systemic treatment. We assessed the statistical significance of the interaction between treatments and PR status using the Wald test.

Sensitivity analyses
For analyses assessing the prognosis of ER status, we (a) excluded those with discordant pathological and clinical stage (pathological stage IV and clinical stage 0–III), and (b) changed the PR status cut point to be 0–10% vs. 11–100% [30] among the ER low-positive patients. Next, for analyses assessing treatment outcomes, we (a) excluded those with a history of another primary tumor, and (b) additionally adjusted for time from diagnosis to systemic therapy.
We used STATA (Version 18.0, StataCorp, 2023. College Station, TX [31]) for our analyses. We considered p-values < 0.05 to be statistically significant.

Results

Descriptive statistics
Our cohort included 27,672 patients with HER2-negative dnMBC, of whom 5,760 (21%) had ER-negative, 699 (3%) had ER low-positive, and 21,213 (77%) had ER-positive disease (Table 1). The average age of diagnosis was 62.2 (standard deviation = 14.4) years for patients with ER low-positive dnMBC, compared to 62.0 (14.3) and 63.7 (14.0) years for individuals with ER-negative and ER-positive disease, respectively. Twenty-three percent of ER low-positive patients self-identified as Black, higher than ER-positive (15%), and lower than ER-negative patients (27%). PR was positive (1–100%) in 32% of ER low-positive patients, rare in ER-negative patients (7%), and common in ER-positive patients (83%). Among ER low-positive patients, 19% had bone-only metastases, compared with 39% of ER-positive patients and 14% of ER-negative patients. The majority of ER-negative and ER low-positive tumors were high-grade or poorly/undifferentiated (65 and 58%, respectively), whereas only 24% of ER-positive were high-grade or poorly/undifferentiated (with ~ 20% missing in each group). Additional characteristics of the cohort are reported in the Online Supplement (eTable 1).

Overall survival by ER status
A greater proportion of deaths occurred among ER low-positive (71%) and ER-negative patients (74%) compared to ER-positive patients (46%) (Table 1). Patients with ER-positive dnMBC had significantly longer survival than both ER low-positive and ER-negative dnMBC patients (Fig. 2A).
In age-adjusted analysis (Table 2), we observed a borderline significant difference in overall survival for individuals with ER low-positive disease compared to ER-negative disease (HR = 0.91, 95% CI 0.83–1.00) (Table 2). The association was not statistically significant in fully-adjusted multivariable analysis (HR = 1.04, 95% CI 0.94–1.14) (Table 2) and in analyses stratified by age and race (eTable 2).

Overall survival by ER-PR status
We identified PR positivity as an independent prognostic factor among ER low-positive patients after adjusting for age, sites of disease, race, ethnicity, insurance, comorbidities, and treatments [PR-positive (1–100%) vs. PR-negative (< 1%): HR = 0.74, 95% CI 0.59–0.92] (eTable 3). Therefore, we performed an analysis that included further stratification of ER low-positive patients by PR status.
Figure 2B depicts overall survival by ER status, with ER low-positive patients stratified by PR status. Compared to ER-negative patients, ER low-positive/PR-positive patients had improved overall survival (Fig. 2B). Median survival was longest among ER-positive patients [43.6 months (95% CI 42.7–44.7)], followed by ER low-positive/PR-positive patients [19.8 months (95% CI 14.8–24.8) months], which was longer compared to both ER low-positive/PR-negative [11.8 (10.6–13.5) months] and ER-negative patients [12.9 (12.5–13.6) months] (p < 0.01, Fig. 2B). In multivariable analysis, ER low-positive/PR-positive dnMBC patients experienced a 16% lower hazard of death relative to ER-negative patients (HR = 0.84, 95% CI 0.71–1.00) (Table 2). In contrast, survival was not statistically significantly different in ER low-positive/PR-negative patients when compared to ER-negative patients in age-adjusted analysis (Table 2). However, ER low-positive/PR-negative patients had significantly worse survival compared to ER-negative patients in multivariable analysis (HR = 1.14, 95% CI 1.02–1.27) (Table 2).
In an exploratory analysis that additionally adjusted for grade, findings were similar (eTable 4). Furthermore, when ER-negative patients were stratified by PR status, ER-negative/PR-positive patients had a similar survival compared to ER-negative/PR-negative patients (adjusted HR = 1.05, 95% CI 0.93–1.19).

Treatment outcomes among ER low-positive dnMBC patients
Among ER low-positive patients, 564 (81%) received systemic treatment (eTable 5). In age-adjusted analysis, ER low-positive patients who received either chemotherapy followed by endocrine therapy (± CDK4/6i) or endocrine therapy + CDK4/6i (no chemotherapy) had improved survival compared to ER low-positive patients who received chemotherapy alone (eTable 5). In fully-adjusted analysis, findings were similar (chemotherapy with endocrine therapy ± CDK4/6i: HR = 0.62, 95% CI 0.44–0.88; endocrine therapy and CDK4/6i: HR = 0.70, 95% CI 0.49–1.01) (Table 3). We present findings stratified by PR status because we observed a significant interaction between PR status and treatments (p = 0.01). We observed worse survival among ER low-positive/PR-negative patients who received endocrine monotherapy compared to chemotherapy alone (HR = 2.20, 95% CI 1.44–3.35), which was not seen among ER low-positive/PR-positive patients (HR = 0.82, 95% CI 0.46–1.46) (Table 3). The hazard ratios of endocrine therapy plus CDK4/6i versus chemotherapy alone were similar in PR-positive and PR-negative disease (Table 3). Similar results were observed when stratifying patients who received cytotoxic chemotherapy without endocrine therapy by immunotherapy receipt (eTable 6).

Sensitivity analyses
When excluding participants with a discordant clinical stage [N = 954 (3% of our cohort)], our findings of survival by ER status and survival by ER-PR status were not meaningfully changed (eTable 7). To assess if a PR percentage threshold of 10% provides better prognostic discrimination than a threshold of 1%, we also conducted an analysis where ER low-positive patients were stratified by PR 0–10% versus 11–100% (eTable 8). Neither group had significantly improved overall survival to ER-negative patients in multivariable analysis (eTable 8).
We conducted two sensitivity analyses of treatment outcomes in ER low-positive dnMBC. First, we excluded those with a history of prior tumor (benign or malignant) (N = 100), because survival may be impacted by prior cancer treatments (eTable 9). Second, we performed a survival analysis additionally adjusting for time from diagnosis to systemic therapy (which had > 5% missing data) to account for potential survival differences in those with a treatment delay (eTable 10). In both analyses, results were not meaningfully changed.

Discussion
To our knowledge, this is the first nationwide study to report on survival in HER2-negative, ER low-positive de novo metastatic breast cancer patients. Overall, patients with ER low-positive and ER-negative breast cancer had similar survival, and their survival was worse than the survival observed in patients with ER-positive disease. However, after further stratification, ER low-positive/PR-positive patients had improved survival compared to ER-negative patients. Interestingly, ER low-positive/PR-negative patients had worse survival compared to ER-negative patients in multivariable analysis. These findings support the heterogeneity of ER low-positive disease and identify PR positivity as a potential prognostic biomarker among ER low-positive dnMBC patients. Further, exploratory analyses suggested patients with ER low-positive dnMBC had similar or improved outcomes with initial treatments incorporating endocrine therapy, either with CDK4/6i and/or following induction chemotherapy, compared to chemotherapy without endocrine therapy.
Our findings that patients with ER low-positive dnMBC had a worse prognosis than ER-positive patients but a similar prognosis to ER-negative patients are consistent with observations reported in early-stage disease [4, 32–35]. One study, however, did report ER low-positive patients to have superior relapse-free and disease-specific survival compared to ER-negative patients [36].
We identified PR as a prognostic marker in patients with ER low-positive HER2-negative dnMBC, which has not been previously reported. PR is a known prognostic marker in ER-positive metastatic breast cancer [37, 38] and recently shown to be prognostic in an NCDB study of early-stage ER low-positive patients [6]. While mechanistic studies are needed to elucidate the underlying biology, it is established that PR expression is reflective of endogenous estrogen signaling activity, as it is an ER-induced gene target as well as an ER-associated protein that modulates ER behavior [39]. PR is also involved in regulating cellular proliferation and differentiation [40]. We hypothesize that PR expression, even in the setting of low-positive ER expression, results in a more endocrine-driven cancer phenotype, resulting in a relative improvement in outcomes.
We observed similar or improved survival in ER low-positive dnMBC patients receiving multimodal therapy incorporating endocrine therapy compared to those receiving chemotherapy alone. These results align with findings in early-stage ER low-positive disease demonstrating the potential benefit of adjuvant endocrine therapy [6, 12, 13]. In stratified analyses by PR status, the relative hazards of death for multi-modality treatment versus chemotherapy alone were similar. This finding is consistent with findings from a study in the neoadjuvant setting among patients with early-stage ER low-positive disease [10]. In that observational study, rates of pathologic complete response to neoadjuvant chemotherapy were similar regardless of PR status [10]. An interesting observation in our study was that ER low-positive/PR-positive (but not ER low-positive/PR-negative) patients had similar survival with endocrine monotherapy compared to chemotherapy alone. Of note, endocrine monotherapy is not the current standard of care for the first-line treatment of metastatic ER-positive breast cancer [2].
Strengths of this analysis include the use of real-world data from the NCDB, a large, nationwide cohort, which enhances generalizability and enables the study of a rare breast cancer subtype. Additionally, the NCDB has an active follow-up program for vital status, baseline information on sociodemographics, tumor characteristics, and comorbidities, and low levels of missingness. By limiting the study population to de novo metastatic disease, we eliminated potential confounding from exposure to prior breast cancer treatments for which there was no detailed information.
Limitations of this database include a lack of central review of ER status and absence of data regarding the anatomic site of biopsy. The NCDB is limited to those who received treatment or treatment decisions at a CoC-accredited facility (Fig. 1). The database provides treatment data only for first-course treatments and does not delineate chemotherapy from CDK4/6i. We used “chemotherapy” and endocrine therapy timing to best approximate treatments received, but studies are needed in cohorts with more specific treatment details. Furthermore, during the study period, immunotherapy became a standard of care for some patients with ER-negative metastatic breast cancer [41], so we cannot meaningfully interpret the association of immunotherapy with survival in this cohort. Moreover, the NCDB lacks cause of death data. However, in a metastatic cohort, most deaths are likely cancer-related. Finally, due to the observational nature of this study, there may be unmeasured confounders impacting treatment outcomes.

Conclusions
In this large cohort study, patients with ER low-positive dnMBC had a prognosis inferior to patients with ER-positive dnMBC but similar to patients with ER-negative dnMBC. Notably, ER low-positive/PR-positive dnMBC patients had superior survival compared to patients with ER-negative and ER low-positive/PR-negative tumors. An exploratory analysis suggested a benefit of incorporating endocrine therapies as part of the first-line treatment of ER low-positive dnMBC. Future research is needed to validate our treatment findings and test whether our observations apply to all patients with ER low-positive metastatic disease, including those with recurrent metastatic and HER2-positive breast cancer.

Supplementary Information
Below is the link to the electronic supplementary material.