What Is the Real Impact of Estrogen Receptor Status on the Prognosis and Treatment of HER2-Positive Early Breast Cancer?

It is well known that HER2⁺ breast cancer is a heterogeneous disease (1, 2). Besides HER2 status itself, there are no validated biomarkers to identify patients likely to benefit from anti-HER2 treatments (3). It has been proposed, however, that the expression of the estrogen receptor (ER) drives variable degrees of benefit from existing anti-HER2 treatment strategies. An increasing number of trials are being developed targeting specifically HER2⁺/ER-positive (ER⁺) or HER2⁺/ER-negative (ER-neg) early breast cancer, and we believe that, in the future, these may be regarded as distinct diseases or we may need additional tools to identify the varying biology within HER2⁺ disease that is represented by ER status.

Distinct biological characteristics may justify the different clinical behavior observed in HER2⁺ early breast cancer according to ER status. The ER and the HER2 pathways drive breast cancer proliferation by a complex interaction of cellular signaling processes (4). Due to the intracellular cross-talk between HER2 and ER, treatments interfering with one of the two pathways can have an effect on the other, with reactivation of ER signaling being a known mechanism of resistance to anti-HER2 agents in HER2⁺/ER⁺ breast cancer (5).

The distribution of intrinsic (PAM50) breast cancer subtypes within HER2⁺ tumors also differs according to ER status (6): around 75% of HER2⁺/ER-neg tumors and 30% of HER2⁺/ER⁺ are “HER2-enriched,” which is the subtype associated with HER2/EGFR pathway activation (2), high proliferation rates, and an immune-activated stroma with elevated tumor-infiltrating lymphocytes (TIL) concentrations (Fig. 1). On the other hand, around 70% of HER2⁺/ER⁺ tumors are luminal A/B, which are ER-dependent tumors, with low HER2/EGFR pathway activation and a high rate of PIK3CA mutations (1), that are associated with lower response to anti-HER2 treatment (7).

In the first 5 years after diagnosis, HER2⁺/ER-neg patients have worse survival compared with the HER2⁺/ER⁺ group. Among 42,887 patients treated between 2010 and 2015, 5-year overall survival (OS) rates were 88.2% in HER2⁺/ER⁺ versus 83.9% in HER2⁺/ER-neg subgroup (P < 0.001; ref. 8). Yet, after 5 years, recurrence risk lowers among HER2⁺/ER-neg patients, but is maintained among those with HER2⁺/ER⁺ tumors (9–11). This leads to similar long-term prognosis: in the HERA trial, 12-year OS in the trastuzumab arms was 81% in the HER2⁺/ER⁺ and 78%–79% in the HER2⁺/ER-neg subgroups (9).

On the other hand, achievement of a pathologic complete response (pCR) after neoadjuvant treatment is a prognostic factor in both groups, but it has a stronger impact on event-free survival (EFS) in the HER2⁺/ER-neg [hazard ratio (HR), 0.25] compared with the HER2⁺/ER⁺ subgroups (HR, 0.58; ref. 12). However, the overall median follow-up for EFS in this meta-analysis was just 5.4 years; thus, the prognostic impact of pCR on late recurrences is currently unknown.

It is well documented that HER2⁺/ER-neg patients treated with chemotherapy and dual anti-HER2 therapy present higher pCR rates (61%–91%) compared with patients with HER2⁺/ER⁺ early breast cancer (26%–56%; refs. 13–20); and, compared with HER2⁺/ER⁺ tumors, HER2⁺/ER-neg tumors have consistently demonstrated greater increases in pCR rates with dual anti-HER2 therapy plus chemotherapy compared with chemotherapy/trastuzumab alone (13, 18–20). Nonetheless, duration of anti-HER2 treatment seems to matter in HER2⁺/ER⁺ early breast cancer: in TBCRC023, increasing treatment length with trastuzumab/lapatinib and letrozole from 12 to 24 weeks led to an increase in pCR rates from 9% to 33%, whereas no significant difference was observed in HER2⁺/ER-neg tumors (21). This suggests that in HER2⁺/ER⁺ tumors, in the absence of chemotherapy, prolonged inhibition of the HER2 and ER pathways may be necessary not only to decrease proliferation, but to induce apoptosis as well. Yet, adding concomitant endocrine therapy (ET) to anti-HER2 therapy and/or chemotherapy did not yield any advantage in the NSABP B-52 or ADAPT HER2⁺/HR⁺ studies (22, 23). This is intriguing, as one would expect that the ER/HER2 pathway blockade would suppress their cross-talk and lead to higher tumor response.

On the opposite end of the spectrum, the high chemosensitivity of HER2⁺/ER-neg tumors was demonstrated in the ADAPT HER2⁺/HR⁻ study, in which the addition of paclitaxel to trastuzumab/pertuzumab increased pCR rates from 36% to 91% (17). Similarly, the KRISTINE trial showed a significantly higher pCR rate in this group with docetaxel/carboplatin plus trastuzumab/pertuzumab compared with T-DM1/pertuzumab (73% vs. 54%, respectively; ref. 16).

In the adjuvant setting, HER2⁺/ER⁺ and HER2⁺/ER-neg subgroups have a similar degree of proportional risk reduction with trastuzumab (9, 24, 25). Yet, in contrast with the neoadjuvant setting, duration of adjuvant trastuzumab treatment might matter more for HER2⁺/ER-neg early breast cancer: in a recent meta-analysis, patients receiving 9 weeks to 6 months of trastuzumab had worse EFS compared with patients treated for 12 months (26). Therefore, in HER2⁺/ER-neg disease, longer trastuzumab treatment is probably more effective than a shorter course. However, the trials included in this meta-analysis had a median follow-up of less than 5 years, so long-term data on HER2⁺/ER⁺ patients are needed.

Challenging this idea, the ExteNET trial showed that sequential anti-HER2 blockade (summing 2 years of treatment) was beneficial for HER2⁺/ER⁺ patients, but not for HER2⁺/ER-neg patients (27). On the other hand, updated survival results from APHINITY show that early dual HER2 blockade with trastuzumab/pertuzumab has an effect both in HER2⁺/ER-neg and HER2⁺/ER⁺ diseases (28). Reasons for this difference may lie in these drugs' mechanisms of action, timing of administration, and length of follow-up. After stopping adjuvant trastuzumab, HER2⁺/ER⁺ patients continue ET, but the single blocking of the ER pathway may lead to reactivation of the HER2 pathway on residual trastuzumab-resistant malignant cells (4). As neratinib was administered after trastuzumab, it may have prevented the reactivation of the HER2 pathway and, hence, benefited HER2⁺/ER⁺ patients, but not those with HER2⁺/ER-neg tumors. As neratinib irreversibly inhibits several HER-family members (EGFR, HER2, HER4), it may have prevented this cross-talk in a more effective way than administration of trastuzumab for 2 years (which only inhibits HER2; ref. 9). In APHINITY, the early dual blockade has probably eliminated trastuzumab-resistant cells in part of the patients, but the benefit was seen later in HER2⁺/ER⁺ patients, as these tend to have later relapses, as compared with HER2⁺/ER-neg patients.

Recently, the KATHERINE trial showed that in patients with residual disease after neoadjuvant treatment, 1 year of adjuvant T-DM1 was superior to trastuzumab, independently of ER status (29). It should be noted that this trial included selected high-risk patients, for whom suboptimal response to neoadjuvant trastuzumab could have influenced the benefit from a salvage treatment with a different anti-HER2 drug, potentially confounding the effect of baseline ER status.

The biggest question nowadays is how to select patients for escalation/deescalation approaches, maintaining or increasing pCR rates and survival, with minimal toxicity. Therefore, the quest for finding prognostic and predictive biomarkers has been intense (3). One attempt was to quantify gene-expression levels of ESR1 and ERBB2, showing that high levels of ERBB2 (20, 30–32) and low levels of ESR1 (31) are associated with higher pCR rates. Expression levels of genes involved in HER2 and ER pathways were also combined in a predictive model in an analysis of the NSABP B-31 trial. Patients were classified into three groups: (i) “high ESR1”/“low-intermediate ERBB2”: no benefit from trastuzumab (HR, 1.58); (ii) “low ESR1”/“low-intermediate ERBB2”: moderate benefit (HR, 0.60); (iii) “low-intermediate ESR1”/“high ERBB2”: large benefit (HR, 0.28; ref. 33). This eight-gene signature, showing no linear correlation between ERBB2 and ESR1 mRNA levels and benefit from adjuvant trastuzumab, was later validated in the N9831 and FB-7 trials (34, 35). Nonetheless, these intriguing findings have not yet been tested in prospective studies or applied in the clinic.

Other biomarker studies focused on the correlation between PAM50 subtypes and benefit from anti-HER2 treatment. A consistent finding is that HER2-enriched tumors present higher pCR rates (between 28% and 72%) when treated with anti-HER2 therapies (with/without chemotherapy), compared with other subtypes (20, 32, 36, 37), showing that HER2-enriched subtype predicts sensitivity to anti-HER2 treatment. In fact, several studies show that ER status loses its association with pCR when subtype is taken into account, indicating that subtype captures most of the information displayed by ER status and also brings additional information (20, 32, 36, 37). One important aspect to take into account when designing deescalation trials is that HER2-enriched subtype is also chemosensitive, with overall pCR rates in HER2⁺ disease of around 60% with versus 30% without chemotherapy, respectively (20, 32, 36, 37).

It has also been proposed that high scores of a gene-expression signature of loss of function of RB1 may identify HER2⁺ tumors more likely to respond to chemotherapy, and this concept has been validated in HER2⁺/ER⁺ tumors using a refined 87-gene signature (RBsig; refs. 38, 39). Moreover, higher TIL levels are associated with improved pCR rates and survival in both HER2⁺/ER-neg and HER2⁺/ER⁺ early breast cancer (40, 41), and evidence of immune activation by RNA signatures is independently associated with higher pCR (20, 31). It has also been demonstrated that high TIL levels (>75%) predicted benefit from adjuvant trastuzumab/pertuzumab versus trastuzumab alone (42). Of note, TILs levels tend to be lower in HER2⁺/ER⁺ compared with HER2⁺/ER-neg tumors (40).

A decrease in Ki67 during neoadjuvant treatment is a surrogate for response and prognosis in HER2-negative/ER⁺ early breast cancer, so its role is being evaluated in HER2⁺/ER⁺ disease (43). In ADAPT HER2⁺/HR⁺, patients with an early tumor response (Ki67 decrease ≥30% or low-cellular tumor at 3 weeks) had a higher pCR rate compared with nonresponders (36% vs. 20%, respectively; ref. 23). In the PerELISA trial, HER2⁺/ER⁺ patients with a Ki67 drop after 2 weeks of letrozole (molecular responders) continued on letrozole, and trastuzumab/pertuzumab were added for another 12 weeks. Nonresponders were switched to paclitaxel with trastuzumab/pertuzumab. The pCR rate was 21% among molecular responders, but even more remarkable was the 81% pCR rate among nonresponders (44). Even though the correlation between an early Ki67 drop and long-term survival is not yet validated in patients with HER2⁺ tumors, it could become a biomarker for selecting patients with HER2⁺/ER⁺ who do not respond to ET and would be candidates to chemotherapy, given the impressive pCR rate achieved in this subgroup.

Thus, a dilemma for future clinical trials is how to select patients: should one be guided by HER2/ER immunohistochemistry status, by molecular subtype or by a combination of both? Based on current evidence, tumors with the highest anti-HER2 sensitivity are those HER2⁺ by immunohistochemistry/ISH and the PAM50 HER2 enriched. Whether additional biomarkers will help identify higher anti-HER2 sensitivity is currently unknown, although TILs and PIK3CA mutations arise as promising candidates (Fig. 2). Indeed, tumors with high HER2 amplification and an intact PI3K pathway are especially sensitive to HER2-targeted therapies (30). TILs could also potentially be used to select patients, but there are no standardized cutoff points to define “high” versus “low” TIL infiltration (45). Another question is if an early biopsy to assess Ki67 levels should be performed in HER2⁺/ER⁺ patients receiving neoadjuvant treatment. This is limited by Ki67 low reproducibility (46) and absence of validation as a surrogate for long-term survival benefit in HER2⁺ early breast cancer.

Despite distinct biological and clinical characteristics between HER2⁺/ER-neg and HER2⁺/ER⁺ early breast cancer, they are still treated similarly, except for ET. Nevertheless, different deescalation strategies for HER2⁺ early breast cancer might be explored according to ER status: (i) to decrease anti-HER2 treatment duration, (ii) to avoid anthracyclines, and (iii) to design chemotherapy-free regimens.

As detailed before, a meta-analysis showed a worse outcome with a shorter treatment duration of adjuvant trastuzumab in patients with HER2⁺/ER-neg disease, but this comparison was nonsignificant in patients with HER2⁺/ER⁺ tumors. Moreover, in the PERSEPHONE trial (the only trial so far proving its noninferiority hypothesis of 6 months vs 1 year of trastuzumab), the HR for the comparison of the two strategies was very close to 1 in the HER2⁺/ER⁺ subgroup (HR, 0.91; 95% confidence interval, 0.68–1.21; ref. 47). Therefore, in patients with HER2⁺/ER⁺ tumors, a shorter duration of trastuzumab treatment may be seen as a safe option by some—nonetheless, due to the late relapse profile of these tumors, longer follow-up data are needed before adopting this strategy.

The APT trial (in which patients with ≤3 cm, node-negative tumors received 12 weeks of paclitaxel and 1 year of trastuzumab) showed a 7-year invasive disease-free survival of 94.6% for HER2⁺/ER⁺ disease and of 90.7% for HER2⁺/ER-neg disease, with just four distant recurrences (1%; ref. 48). Thus, this treatment deescalation seems very safe for patients with low-burden tumors. Given the high chemosensitivity of HER2⁺/ER-neg tumors, leading to a higher rate of pCR, a possible strategy for avoiding anthracyclines in patients with somewhat larger tumors would be to assess the tumor's sensitivity upfront with neoadjuvant double HER2-blockade plus a taxane. The DeCrescendo trial will test this hypothesis in patients with cT1c-T2N0 HER2⁺/ER-neg tumors, who will receive 12 weeks of taxanes plus trastuzumab/pertuzumab. Those who achieve pCR will be spared from adjuvant anthracyclines and will only receive double HER2-blockade until completion of 1 year of treatment. The study is powered to test 3-year relapse-free survival (RFS) rate, with a narrow confidence interval, primarily in patients with PAM50 HER2-enriched tumors and then in the overall population.

The reduced chemosensitivity of HER2⁺/ER⁺ early breast cancer fuels the interest in alternative, targeted approaches without chemotherapy. One is the use of CDK4/6 inhibitors, which synergize with trastuzumab in HER2⁺/ER⁺ tumor models (49). In the small NA-PHER2 study, 20 weeks of palbociclib/ET/trastuzumab/pertuzumab provided a significant reduction in tumor proliferation, but led to a modest pCR rate of 27% (50). Long-term survival results and efficacy according to molecular subtypes are awaited. There are ongoing trials [PALTAN (NCT02907918); TOUCH (NCT03644186)] exploring this approach. In particular, TOUCH is assessing RBsig (39) as a potential biomarker to identify HER2⁺/ER⁺ patients who could be spared from chemotherapy and treated with ET/anti-HER2/palbociclib. On the other hand, a recent joint analysis of neoadjuvant “chemotherapy-free” trials (SOLTI-PAMELA, TBCRC023/TBCRC006, and PerELISA) has shown that the combination of the HER2-enriched subtype with a high mRNA expression of ERBB2 leads to the identification of a subgroup of patients exquisitely sensitive to double HER2 blockade—with pCR rates ranging from 44.5% to 66.7% (51). Among HER2⁺/ER⁺ patients, only 32% were “HER2-enriched/ERBB2-high,” whereas among the HER2⁺/ER-neg subgroup, there were 68% of patients with this profile. Hence, as the HER2⁺/ER-neg subgroup is clearly enriched with these HER2-addicted tumors, one could envision a trial restricted to HER2⁺/ER-neg patients: they would be screened upfront and those with a HER2-enriched/ERBB2-high expression tumor would be proposed to receive neoadjuvant trastuzumab/pertuzumab only. Like in the DeCrescendo trial, patients achieving pCR would be spared of any adjuvant chemotherapy and the primary endpoint would be a high RFS rate with a narrow confidence interval.

Given the predictive value of ER status, it would be reasonable for the next wave of clinical trials to target separately HER2⁺/ER⁺ and HER2⁺/ER-neg patients. This would lead to a better tailoring of deescalation treatment strategies, and hence to a greater chance of “success.” In addition, this could greatly reduce the morbidity and costs associated with the long and intense systemic regimens that patients with HER2⁺ tumors still receive, despite years of biomarker research. Thus, besides the use of ER status, we feel that an integrated classification of HER2⁺ tumors, based on their DNA, RNA, protein expression, and immune microenvironment, would also be useful for further treatment refinement in the future.

M. Brandão reports receiving other commercial research support and speakers bureau honoraria from Roche/GNE. R. Caparica reports receiving speakers bureau honoraria from AstraZeneca, Boehringer Ingelheim, and Janssen, and is an unpaid consultant/advisory board member for Pfizer and AstraZeneca. L. Malorni reports receiving commercial research grants from Pfizer and Novartis, and reports receiving speakers bureau honoraria from Novartis. A. Prat is a paid consultant for Roche, Oncolytics Biotech, Daiichi Sankyo, Novartis, Pfizer, and Nanostring Technologies, and reports receiving commercial research grants from Novartis, Nanostring Technologies, Incyte, and Puma. L.A. Carey is an unpaid consultant/advisory board member for Roche/Genentech, Novartis, and Seattle Genetics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Brandão, R. Caparica, M. Piccart

Development of methodology: Not applicable

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Not applicable

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Not applicable

Writing, review, and/or revision of the manuscript: M. Brandão, R. Caparica, L. Malorni, A. Prat, L.A. Carey, M. Piccart

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Not applicable

Study supervision: M. Piccart

M. Brandão and R. Caparica acknowledge the support of Institut Jules Bordet, which funded their research fellowship.