Real-world outcomes in metastatic HR+/HER2-, HER2+ and triple negative breast cancer after start of first-line therapy

Breast cancer is the second leading cause of cancer death among women, with approximately 44,000 deaths expected in USA in 2022 [1]. Advances in the treatment of metastatic breast cancer (mBC) have improved the prognosis to some extent. For example, the median overall survival (OS) for mBC in 2010 was reported to be 20.5 months in phase III trials [2], whereas more recent studies indicate the current median OS to be closer to 3 years [3,4]. However, outcomes vary depending on the human epidermal growth factor receptor 2 (HER2) status and hormone receptor (HR) status, and as shown in the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines^®) [5], the array of systemic treatments available for metastatic (stage IV) breast cancer is extensive.

HER2 positive (HER2+) patients can be treated with targeted therapies alone or combined with other agents. In contrast, hormonal therapy is considered the standard initial treatment for mBC that is HR positive (HR+; estrogen receptor positive [ER+] or progesterone receptor positive [PR+]) and HER2 negative (HER2-). Recent improvement in the survival of mBC may be attributable to the approval of several cyclin-dependent kinase 4/6 (CDK4/6) inhibitors for treatment of disease that is HR+ in recent years, adding to other therapy classes for this indication (e.g., aromatase inhibitors [AI], selective estrogen receptor modulators [SERM] or degraders [SERD]), or to new therapies approved for HER2+ breast cancer, (e.g., neratinib, fam-trastuzumab deruxtecan-nxki), complementing a range of existing therapies in this class (trastuzumab, pertuzumab, trastuzumab emtansine).

Patients with HR negative (HR-) and HER2- disease are classified as having triple negative breast cancer (TNBC). TNBC accounts for 10–15% of all breast cancers [6] and is an aggressive subtype of breast cancer that has poor prognosis compared with HR+/HER2- and HER2+ patients [7,8]. TNBC accounts for approximately 5% of all cancer-related deaths annually, and has lower 5-year relative survival rates than other breast cancer subtypes [6,9]. Metastatic TNBC is difficult to treat due to the lack of molecular targets [10]. Chemotherapy remains the primary first-line (1L) option (e.g., anthracyclines, taxanes, anti-metabolites and microtubule inhibitors), particularly for patients without evidence of programmed death-ligand 1 (PD-L1) positive disease [5]. Sacituzumab govitecan-hziy was approved by the US FDA for use in metastatic TNBC in 2020, but is currently indicated only for use in later lines of therapy [11].

Many recent summaries of outcomes in mBC were drawn from patients who initiated care before 2010, and in some cases, much earlier, so an updated review of this population is needed [3,4,10]. This study was designed to evaluate real-world clinical outcomes, including real-world progression-free survival (rwPFS) and OS, by biomarker subtype, among a geographically and racially diverse sample of patients who initiated 1L therapy for treatment of mBC within the past 5 to 10 years, during a time of transition in testing methodologies and thresholds and entrance of new agents into the market.

Methods

Study design & data source

This was a retrospective, observational cohort study using de-identified data from ConcertAI's Patient360™ dataset, a repository of electronic medical record (EMR) data available through data-sharing agreements with practices and other data providers. The dataset contains data that originated as structured EMR fields, but it also includes data that were originally unstructured, such as physician progress notes. These were deeply curated through manual review of the medical records during creation of the Patient360 dataset. The source EMR data were drawn from more than one hundred oncology treatment sites, principally community-based, in contrast to academic settings. Practice sizes vary and include small community practices, major health systems, and large, multi-site oncology groups, in both rural and urban settings throughout the USA. Key variables were abstracted from progress notes and other documents within the unstructured medical records by trained Clinical Research Nurses. Data were mapped to a standard data model, the Quality Data Model from the Centers for Medicare and Medicaid Services [12]. The study was exempt from review by an institutional review board because it was retrospective and used de-identified secondary data.

Patients

Eligible patients met the following inclusion criteria: females (≥18 years old) with diagnosis of mBC (International Classification of Diseases [ICD]-9 code of 174.x or ICD-10 code of C50.x) at any point on or after 1 January 2010, with diagnosis of mBC defined by the presence of stage IV disease or diagnosis of secondary malignant neoplasms confirmed during human review of the medical record; documentation of HR+/HER2- breast cancer, HER2+ breast cancer, or TNBC at metastatic diagnosis; and initiation of 1L treatment in the metastatic setting. Completion of 1L therapy was not required, but patients who received no systemic therapy in the mBC setting were excluded. Patients who presented initially after 31 December 2019 were excluded as having insufficient follow-up. For tumor marker documentation, HR+ included either ER+ or PR+ breast cancers. HER2- was defined as immunohistochemistry negative (IHC 0 to 1+) or HER2 negative not otherwise specified (NOS). Tumors were considered HER2+ if they met one of the three following definitions: immunohistochemistry (IHC) equivocal (2+) and FISH positive; IHC positive (3+); or HER2+ NOS. TNBC was defined as tumors negative for all three biomarkers (i.e., ER-, PR- and HER-). If multiple indications of a given tumor marker were available, the indication closest to the date of metastatic diagnosis was recorded. Patients enrolled in a clinical trial during the study period and those diagnosed with other primary cancers were excluded.

Study end points & assessments

The primary end points for analysis reported in this paper were rwPFS and OS. We defined rwPFS as the interval from the start of 1L in the metastatic setting until disease progression or death, whichever happened first. The dates and occurrence of disease progression were based on provider notes from the unstructured medical records. OS was defined as the interval from the start of 1L in the metastatic setting until death as documented in structured EMR data, in provider notes, or in third-party sources including Social Security Death Index records and obituary records. Therapy was defined as 1L if it was the first hormonal or non-hormonal systemic breast cancer therapy, either monotherapy or combination therapy, given on or after the initial metastatic diagnosis date. Systemic treatments initiated within 30 days of the start of 1L therapy were considered to have been administered as part of combination therapy. However, when one agent was being discontinued and another started (e.g., replacement of one drug with another), oral agents were allowed to overlap with other drugs up to 3 days without considering this to represent combination therapy. Additional study variables that were collected included patient characteristics at initial diagnosis (race/ethnicity, cancer stage, histology, and tumor grade), and at mBC diagnosis (age, sites of distant metastasis, Eastern Cooperative Oncology Group Oncology [ECOG] performance score, menopause status, smoking status, alcohol use status and receipt of radiation treatment).

Statistical analysis

The analysis was grouped by biomarker-based classifications, namely HR+/HER2- versus TNBC and HER2+ versus TNBC. Patient characteristics were assessed using standard descriptive methods. Mean, standard deviation, and median values were reported for continuous variables, and categorical variables were described using frequencies and percentages. Fisher's exact or χ2 tests, and analysis of variance, were used to compare categorical and continuous patient characteristics, respectively, between biomarker-based groups. Kaplan–Meier survival analyses were used to descriptively characterize rwPFS and OS for HR+/HER2- versus TNBC groups and HER2+ versus TNBC, separately. Cox regression analysis was used to assess differences in rwPFS and in OS between biomarker groups, controlling for baseline demographic and clinical characteristics. A stepwise selection at p < 0.10 level was used to determine patient characteristics for inclusion in the final model. Biomarker classification, age, race/ethnicity, and stage at initial diagnosis were included regardless of statistical significance. All analyses were preformed using SAS (SAS/STAT User's Guide, Version 9.4; SAS Institute, Cary, NC). Results were interpreted as significant at alpha = 0.05, two tailed.

Results

A total of 1518 patients with mBC were included in the study sample. The median start date of treatment in the mBC setting was November 2015, and the median interval from mBC diagnosis to start of 1L treatment was 36 days. The predominant mBC subtype was HR+/HER2- (62.5%), followed by HER2+ (23.4%) and TNBC (14.1%). Table 1 shows the demographic and clinical characteristics of the sample by biomarker status. Patients with HR+/HER2- were older than patients with HER2+ and TNBC, with mean ages of 59.4 (±13.6) versus 55.7 (±13.8) and 54.8 (±12.5) years, respectively (both p < 0.001). Although the majority of the sample was White, a higher proportion of patients with TNBC were Black/African–American compared with the HR+/HER2- and HER2+ groups (p < 0.001 for both comparisons).

Invasive ductal carcinoma was the predominant histology at initial presentation with breast cancer for all biomarker groups, although it was significantly more prevalent in patients with TNBC (70.1%) compared with patients with HR+/HER2- mBC (56.9%). Patients with TNBC were less likely to have had stage IV disease at initial diagnosis (21.0%) compared with patients with HR+/HER2- and HER2+ (28.9 and 28.5%, respectively; both p < 0.001). However, patients with TNBC were more likely to have had high-grade tumors at initial diagnosis than patients with HER2+ and HR+/HER2-. Grade 3 tumors were identified in 65.9% of patients with TNBC compared with 45.4% and 25.5% in the HER2+ and HR+/HER2- groups, respectively (p < 0.001 for both comparisons).

Patterns of metastasis also varied across the mBC biomarker subtypes. The most common site of distant metastasis for HR+/HER2- and HER2+ mBC was bone (65.2% and 49.6%, respectively), whereas the most common site of distant metastasis for TNBC was lung (40.2%) followed by bone (36.9%). These rates of bone involvement differed significantly, with TNBC patients less likely to have bone involvement than patients with HR+/HER2- and HER2+ (p < 0.001 and p = 0.003, respectively). In addition, TNBC patients were significantly more likely to have lung and distant lymph node involvement than the other groups, which did not differ from each other. TNBC patients also had significantly more brain involvement than HR+/HER2- patients, as did HER2+ patients. Other clinical characteristics did not differ significantly between HR+/HER2- versus TNBC, and HER2+ versus TNBC groups.

The most common 1L metastatic treatments for HR+/HER2- patients during the overall study period were hormone directed monotherapies: 29.7% AIs, 8.2% SERMs, and 7.2% CDK4/6 inhibitors, SERDs or gonadotropin-releasing hormone agonists. Another 29.4% received these agents as part of combination therapy. An additional 18.6% received a taxane or anthracycline-containing therapy. Among HER2+ patients, two-thirds (67.0%) received HER2-directed therapies during 1L, and these were mainly combination therapies including trastuzumab with a taxane. Most of the remaining HER2+ patients (24.8% of the total) received hormone-directed therapies during 1L, and 79.2% received a HER2-directed therapy either in 1L or a later line See Figure 1 for additional description of the sequencing of HER2-targeted therapies across the first three lines of therapy. Most TNBC patients (62.1%) treated during the study period received capecitabine or taxane-based therapies in 1L, and about two-thirds of these (42.1% of the total) were capecitabine or paclitaxel monotherapy. Seven in ten of the remaining TNBC patients received platinum or anthracycline-based therapies. See Table 2 for additional detail regarding the treatments administered by biomarker subtype. Overall, 63.6% of patients received second-line (2L) treatment, and 43.4% received third-line (3L) treatment, but the biomarker groups varied in the rate at which they had a record of later lines of therapy. Patients with TNBC received 2L and 3L treatment at a lower rate than the other groups, with just 57.0 and 33.6% of TNBC patients receiving treatment in 2L and 3L, respectively (p = 0.039 for 2L and p = 0.003 for 3L). The differences in the rates were nonsignificant when TNBC was excluded.

Median rwPFS from start of 1L treatment differed by biomarker status, with longer rwPFS among patients with HR+/HER2- (17.8 months; 95% CI, 15.5–21.7) and HER2+ (14.1 months; 95% CI, 11.5–18.8), compared with patients with TNBC (6.5 months; 95% CI, 5.2–8.4; p < 0.001 for both comparisons; Figure 2). As can be seen in the figures, the rwPFS survival curve for TNBC diverged from the other groups within about 2 months after start of 1L therapy. The difference in rwPFS between biomarker groups remained after controlling for baseline demographic and clinical characteristics. Cox regression analysis showed that the TNBC group had significantly shorter rwPFS after start of 1L therapy in the metastatic setting, compared with HR+/HER2- (adjusted hazard ratio [aHR], 1.630; 95% CI, 1.341–1.981; p < 0.001) and compared with HER2+ (aHR, 1.760; 95% CI, 1.424–2.174; p < 0.001; Table 4). There was no difference in the risk of a rwPFS event observed between the HR+/HER2- and HER2+ groups. As expected, age ≥65 years, grade 3 tumor at initial diagnosis, major sites of metastasis (bone, brain, liver and lung), and poorer performance status (ECOG ≥2) were independently associated with a shorter rwPFS (Table 4). There was no difference in rwPFS between patients reporting White race, and those reporting Black race or Hispanic ethnicity (Table 4).

Similar to the findings observed for rwPFS, median OS after the start of 1L therapy also differed by biomarker status. Patients with HR+/HER2- and HER2+ had longer median OS compared with patients with TNBC (60.1 months; 95% CI, 51.6–68.9, and 56.6 months; 95% CI, 49.4–76.0 vs 14.5 months; 95% CI, 11.6–18.9, respectively; p < 0.001 for both comparisons; Figure 3). Consistent with our observations for rwPFS, the difference in OS for patients with TNBC emerged within about two months after the start of 1L treatment, and only a minority of patients with TNBC (36.6%) survived through 24 months after the start of 1L therapy, compared with 73.4% and 73.3% in the HR+/HER2- and HER2+ groups, respectively (Table 3, Figure 3). After controlling for baseline demographic and clinical characteristics, the difference in OS between biomarker groups remained. Cox regression analysis showed that the TNBC group had significantly shorter OS after start of 1L therapy in the metastatic setting compared with the HR+/HER2- group (aHR, 2.198; 95% CI, 1.783–2.711; p < 0.001) and HER2+ group (aHR, 2.442; 95% CI, 1.933–3.083; p < 0.001; Table 5). There was no difference in OS between patients with HR+/HER2- versus HER2+ mBC. Similar to findings for rwPFS, age ≥65 years, higher tumor grade at initial diagnosis (grades 2 and 3 tumor), major sites of distant metastasis (brain, liver, and lung), ECOG ≥2, and a history of smoking were independently associated with a shorter OS (Table 5). There was no difference in OS between patients reporting White and Black race, although Hispanic patients tended to be at lower risk of OS compared with White patients (p = 0.056; Table 5).

Discussion

This retrospective, observational study examined real-world clinical outcomes in mBC across several biomarker subtypes (HR+/HER2-, HER2+, and TNBC) in patients diagnosed with mBC in 2010 or later, with median mBC treatment start date of November 2015. Among the 1518 eligible patients in this predominantly community-based sample, TNBC was the least frequent breast cancer subtype (14.1%). TNBC patients were younger than the HR+/HER2- biomarker group. They were also more likely to be Black/African–American, to have been diagnosed at an earlier stage, and to have higher tumor grades at initial diagnosis compared with the other biomarker groups. Similar to earlier reports, TNBC in this sample was associated with poorer clinical outcomes, a general pattern this study shows has not changed in recent years. Patients with TNBC had shorter rwPFS in 1L (6.5 months vs 17.8 and 14.1 for HR+/HER2- and HER2+, respectively) and shorter OS from start of 1L (14.5 months vs 60.1 and 56.6 for HR+/HER2- and for HER2+, respectively).

This study adds to the literature by reporting effectiveness outcomes across the major biomarker-defined mBC subtypes in a US-based sample that is more recent, that is geographically diverse, and that has representation of racial and ethnic minorities that align with the underlying population. Other recent studies reporting outcomes by biomarker subtype do not cite a race/ethnicity distribution [2,3,6,13–17], or may be narrow geographically [7,18], or focus mainly on a specific biomarker subtype (e.g. [7,16], for TNBC). Others that report race may indicate very low representation of Black patients [19]. In contrast, the current study accrued a geographically diverse patient sample in which more than 20% of patients were Black, nearly 12% were Hispanic, and more than 10% were other or unknown race.

Findings from this study were generally consistent with previous reports as to the distribution of patients by biomarker subtype [7,9,18,20], the age and race distribution [9,18,21–23], the distribution of tumor grade [7], and the clinical outcomes of TNBC compared with other mBC biomarker groups [13–17,19]. The study reported that 43.7% of patients received radiation therapy in the mBC setting, a figure that aligns with previous studies of palliative radiation therapy use in the Medicare setting [24]. Although the current study showed that Black patients were disproportionately likely to have TNBC, multivariate analysis of outcomes for rwPFS and OS showed that Black patients were not at higher overall risk for these outcomes. However, the study did find a trend toward lower OS risk in Hispanic patients.

Two recent papers report findings from a study of more than 22,000 patients diagnosed with mBC from 2008 to 2016 at French cancer centers [20,25]. Median OS for patients with TNBC in that study was similar to findings in the current study (14.8 vs 14.5 months) [25], but OS was shorter in the French study for patients with HER2+ MBC (50.1 vs 56.6) and HR+ MBC (43.3 vs 60.1). The better outcomes we report for patients with HER2+ disease may be attributable to uptake of newer therapies, a general longitudinal trend reported by Grinda and colleagues in their own data [20]. However, the reason for the larger difference in patients with HR+ mBC is unclear. Several other recent studies report outcomes by biomarker subtype, but these tend to be limited geographically, to have smaller samples or earlier data, or to focus on de novo patients, for example [26–29]. And among these, only File and colleagues examined the intermediate end point of PFS [29]. Considering recency of the data, coverage of all the major biomarker subtypes, inclusiveness of the sample, and the examination of both OS and rwPFS outcomes, this study adds significantly to the understanding of the current outcomes of patients with mBC.

Findings regarding stage at initial diagnosis for metastatic TNBC versus other biomarker groups in this study are different from registry reports that are not limited to mBC [7,18]. The registry studies reported a higher rate of initial presentation at stage IV for patients with TNBC compared with other biomarker subtypes, whereas this study reported a lower rate. This pattern of higher rates of de novo metastatic disease for TNBC overall, but lower rates for TNBC among those who ever develop metastatic disease, can only occur as a result of a higher rate of conversion from early to metastatic disease in TNBC compared with other biomarker subtypes.

Many treatments were approved during the period covered by this study, including CDK4/6 inhibitors for treatment of HR+ disease starting in 2015, and pertuzumab and other HER2-targeted therapies in 2012 [30]. These treatments would have been unavailable to patients starting 1L during the early part of the study period, and outcomes should therefore be interpreted in light of that. Overall, 67.0% of patients with HER2+ disease received a HER2-targeted therapy during 1L. Although others went on to receive a HER2-directed therapy in later lines, this rate appears lower than what is reported from other real-world studies. Prospective registry studies have reported rates of 1L HER2-targeted therapy in this population around 90% [31,32], although 1L HER2-targeted therapy use older patients in the ESME study was somewhat lower, at 76.5% [33]. The lower rates in the current study may reflect broader treatment differences in retrospective and prospective data sources over the last decade. Review from other retrospective sources in the USA will be needed to further clarify this pattern. Further research that directly assesses the outcomes of patients who receive CDK4/6 inhibitors for the treatment of HR+/HER2- mBC, or HER2-targeted therapies for HER2+ mBC, will also help clarify the impact of these treatments in real-world populations.

These findings underscore the continuing unmet need of patients diagnosed with TNBC. They are more likely to be diagnosed with metastatic disease initially, more likely to go on to develop metastatic disease if they are not, and they have fewer effective treatment options if they become metastatic. This study and other studies also highlight the differential impact this unmet need has in patients of color, and the importance of continued development of novel therapies for treatment of TNBC, such as sacituzumab govitecan-hziy, which was approved in April 2020 for treatment of TNBC in later lines of therapy. Continued improvement in the outcomes of mBC will depend on molecular testing to assess actionable targets, such as BRCA1/2 and PD-L1, as well as advancing the identification of other molecular pathways that can be targeted. These include PI3 Kinase (PI3K), CDK4/6, Poly ADP Ribose Polymerase (PARP), and low expression of HER2. Finally, development of novel therapies to target these pathways is essential to continued progress.

This study has several limitations. Most notably, the study was retrospective, and relied on information collected as part of routine clinical care, and our sample may therefore differ from other study samples in ways that introduce some degree of bias into the conclusions. Although the sample size was moderately large, within-strata samples in certain demographic or clinical subsets were small and may have limited the types of follow-up analyses that could be conducted. The study was limited to patients who received systemic therapy in the metastatic setting, and findings should therefore not be generalized to patients who do not receive systemic treatment after metastatic diagnosis. Finally, these data were drawn from principally community oncology treatment settings. Although the source practices are geographically diverse within the USA, findings may not generalize to treatment settings outside the USA, or to academic medical centers or tertiary care treatment settings within the USA.

This study also has several important strengths. The sample was geographically diverse, with excellent representation of Black and Hispanic patients. Second, study patients were not accrued from practice sites that are members of any one group purchasing organization, and did not require participation in any registry, enrollment in any insurance program, or membership in any healthcare system. Treatment patterns and outcomes should therefore reflect what occurs in the underlying real-world population. Finally, much of the study data were generated through in-depth human curation of the unstructured medical record. Determination of biomarker subtype, race/ethnicity, stage at initial diagnosis, histology at initial diagnosis, tumor grade, sites of distant metastasis, ECOG performance status, menopausal status, smoking status and occurrence and date of disease progression, were all verified by human review of the medical record, rather than through use of proxy indicators in structured data. This ensures the highest level of quality and completeness available outside of prospective data collection, and allows for increased confidence in the findings from this real-world sample.

Conclusion

This real-world, observational study showed that mBC patients with TNBC continue to have a poor clinical prognosis compared with patients with HR+/HER2- and HER2+ mBC. These findings underscore the need for new therapies to improve clinical outcomes in patients with metastatic TNBC.