Background
The addition of pertuzumab, a monoclonal antibody that inhibits HER2-HER3 dimerization and activates antibody-dependent cellular cytotoxicity (ADCC), to neoadjuvant treatment with chemotherapy and trastuzumab has significantly improved pathological complete response (pCR) rates in patients with HER2-positive breast cancer [
1‐
3]. However, effects on long-term survival outcomes have been modest, and pertuzumab treatment is associated with increased toxicity and costs [
4‐
8]. Results of previous studies indicate that a subset of patients is highly responsive to neoadjuvant treatment consisting of dual HER2-blockade with trastuzumab and pertuzumab without chemotherapy [
1,
9‐
11]. In contrast, another subset of patients seems to be non-responsive despite dual HER2-blockade. Biomarkers that differentiate patients with high likelihood of response and excellent prognosis from those with poor outcomes may assist in selecting some patients for chemotherapy-free regimens and others for intensified or novel regimens. Although hormone receptor (HR) status has been established as a predictor of pCR in patients with HER2-positive breast cancer who are treated with neoadjuvant HER2-blockade, it does not accurately differentiate between non-responders and responders, as both patients with HR-negative disease and patients with HR-positive disease benefit from pertuzumab [
1‐
3,
5]. While several other potential biomarkers for predicting response to HER2-blocking agents have been investigated, a reliable marker has not been established [
12]. Thus, additional prognostic and predictive biomarkers are required to improve treatment decision making.
It is plausible that tumors which rely heavily on the HER2-pathway for their survival and proliferation, so-called HER2-driven or HER2-addicted tumors, are highly sensitive to HER2-blockade. This hypothesis has been supported by studies that show a strong association between the HER2-enriched intrinsic subtype and pCR in patients receiving neoadjuvant dual HER2-blockade [
13]. However, an interaction with pertuzumab treatment has not been investigated, and associations with long-term outcomes have not been described.
Intrinsic tumor subtypes can be determined using BluePrint, an 80-gene molecular subtyping test that classifies breast tumors as Basal-, Luminal- or HER2-type based on gene expression analysis [
14]. Recent data shows that, although the majority of tumors exhibit a high signature score for one single subtype, in some tumors equally high gene expression scores are observed for more than one subtype, which indicates that multiple pathways are activated [
15]. Analysis of tumors that were assigned a ‘dual subtype’ has shown that their biology differs from tumors with a single dominant subtype, which may have implications for treatment response and prognosis [
15]. Indeed, secondary analyses from the APHINITY trial (NCT01358877) suggest that pertuzumab benefit is largely restricted to patients with single-activated HER2-type tumors and less pronounced in patients with other single- or dual-activated subtypes [
16]. In this study, we evaluated BluePrint standard and dual subtypes as a biomarker for predicting response to trastuzumab-containing neoadjuvant chemotherapy with or without pertuzumab in a large cohort of patients with HER2-positive breast cancer.
Discussion
The results of our study show that molecular subtypes as determined by BluePrint are associated with response to neoadjuvant pertuzumab in stage II and III HER2-positive breast cancer, independent of other clinical variables such as hormone receptor status and HER2 immunohistochemistry score. Patients with tumors that are classified as HER2-type according to the standard BluePrint or further classified as HER2-single-type according to the dual-subtype readout have a high chance of reaching pathological complete response after neoadjuvant treatment with chemotherapy, trastuzumab and pertuzumab. In addition, a clear benefit of pertuzumab for overall survival and breast cancer-specific survival was seen in the patients with HER2-type or HER2-single-type tumors, which was not seen for patients with other subtypes.
While molecular subtypes have been previously evaluated in HER2-positive breast cancer, we are the first to compare the effects of neoadjuvant chemotherapy combined with trastuzumab and pertuzumab versus trastuzumab only within different molecular subtypes, in relation to both pCR as well as survival outcomes, in such a large sample size. Previously, it was shown in the NBRST study that patients with HER2-type breast tumors had the highest pCR rates, in particular when treated with dual HER2-blockade. However, the study was underpowered for evaluation of long-term survival and treatment interactions were not reported. The study also did not include subgroup analyses and evaluation of the dual subtypes [
28]. Recently, results of the APHINITY trial suggested that patients with HER2-single-type tumors might have a greater benefit of adjuvant pertuzumab than patients with other subtypes [
16]. Of note, due to the differences in inclusion criteria the population of the APHINITY trial may have a different prognosis than our population. In addition, subtype evaluation in the APHINITY study was performed in a nested case–control set, resulting in a much higher proportion of Luminal- and Basal-type tumors compared to our study. Therefore, associations were analyzed by inverse probability weight corrected Cox regression, which showed a trend toward greater pertuzumab benefit in the patients with HER2-type tumors, similar to our findings. Additionally, a systematic review analyzing 16 studies in early-stage HER2-positive disease for which PAM50 subtyping was performed, found that the HER2-enriched subtype was significantly associated with pathological complete response independent of hormone receptor status [
13]. The effect of dual HER2-blockade versus single HER2-blockade within different subtypes was not evaluated, and long-term outcomes were not assessed. The design and sample size of our study allowed us to not only evaluate molecular subtypes in relation to both overall as well as breast cancers specific subtypes, but also study the effect of dual versus single HER2-blockade in relation to molecular subtypes through interaction tests and subgroup analyses. However, given that the majority of HER2-positive tumors in our study population exhibits a HER2-activated subtype, other subtypes were grouped together in most analyses. The absence of a significant interaction in our study may indicate that molecular subtype is not specific to pertuzumab benefit. Indeed, molecular subtype seems predictive of response to HER2-targeting therapy in general, since the patients who received single HER2-blockade treatment with trastuzumab only and have a HER2-type tumor also have better pCR rates than patients with other tumor subtypes. However, since we did observe a clear benefit of additional treatment with pertuzumab in HER2-type tumors and not in non-HER2-type tumors based on subgroup analyses, the lack of interaction may also be due to the smaller sample size of the non-HER2-type group. Further studies are warranted to validate our findings and confirm the lack of pertuzumab benefit in patients with clinically HER2-positive but genomically non-HER2-type tumors. Given that dual HER2-blockade is currently standard of care for patients with stage II-III HER2-positive breast cancer, this could be analyzed by extending the current analyses to include patients treated after 2016.
HER2-targeting agents are known to exert their effect by both blocking the HER2-pathway and through activation of natural killer cell-mediated antibody-dependent cellular cytotoxicity (ADCC) [
29,
30]. Results of several clinical trials indicate that a subgroup of patients responds well to treatment with only dual HER2-blockade, without chemotherapy [
4,
9‐
11]. It is hypothesized that these patients’ tumors rely heavily on the HER2-pathway for survival and thus might be particularly sensitive to the combination of HER2-pathway blockade and ADCC. Finding biomarkers that identify this group of patients with ‘HER2-driven’ tumors is highly relevant for future studies evaluating de-escalation of chemotherapy. Given that BluePrint was developed through supervised analysis based on ER, PR and HER2 status (IHC and mRNA expression), it might be able to capture subtype-specific pathways better than previously used methods, such as PAM50 subtyping [
14,
31]. In addition, the recently developed BluePrint dual-subtype readout has shown that some tumors display multiple activated pathways and appear to be biologically different from the true single-subtype tumors, which may be valuable for further distinction of truly HER2-driven tumors and relevant for the probability of response to HER2-targeted treatment [
16,
24,
25]. Our results show that the BluePrint HER2-single-subtype indeed seems to be a stronger prognostic factor than the standard readout HER2-type. Many patients that were initially classified as HER2-type by the standard readout are classified as Luminal-HER2-type upon dual readout, which is associated with lower odds of pCR and worse prognosis in both univariable and multivariable analyses. Interestingly, we found that among patients with a high HER2 immunohistochemistry score presence of a subtype other than HER2 or HER2-single is associated with a severely diminished chance of reaching pathological complete response. This suggests that HER2 IHC scoring may not fully account for tumor heterogeneity and that molecular subtyping may have additional value to HER2 IHC scoring for prediction of pCR after neoadjuvant treatment with chemotherapy and HER2-blockade. Thus, BluePrint molecular subtypes may be valuable in conjunction to other biomarkers to identify HER2-driven tumors. Since HER2 IHC scoring was not performed centrally and that detailed information on HER2 evaluation was not available for further analysis, it cannot be ruled out some tumors may have been falsely classified as HER2 positive. In addition, we found the non-HER2-single subtype to be associated with a lower chance of pCR in both hormone receptor-negative and hormone receptor-positive diseases. However, the heterogeneity of molecular subtypes is substantially smaller among patients with hormone receptor-negative breast cancer, and further research is needed to investigate clinical utility of subtyping in this group of patients.
Given that HER2-type and HER2-single-type tumors had excellent prognosis after treatment with trastuzumab and pertuzumab, molecular subtyping may be informative for the selection of patients who could be potential candidates for treatment de-escalation. The TRAIN-3 study is a de-escalation trial in which patients with stage II-III HER2-positive disease are referred to surgery once they reach pCR during neoadjuvant treatment with chemotherapy and dual HER2-blockade; our data may be validated by retrospective analysis of molecular subtype in these patients. In addition, our results could be confirmed further in future trials by randomizing patients with genomically HER2-type tumors between standard treatment with dual neoadjuvant chemotherapy and dual HER2-blockade, versus treatment with dual HER2-blockade only. Besides its potential use in de-escalation, molecular subtyping may identify patients with low chances of response or poor prognosis, who could benefit of treatment strategies other than HER2-blockade. Current adjuvant treatment decisions are based on presence of pCR following neoadjuvant treatment, where patients with no pCR are treated with trastuzumab-emtansine (T-DM1) [
32,
33]. Results of recent studies indicate that trastuzumab-deruxtecan (T-DXd) may benefit both patients with low HER2-expression and HER2-positive metastatic patients who acquired resistance to T-DM1 [
34,
35]. Given that T-DXd is thought to have a more potent bystander killing effect than T-DM1 by penetrating cells adjacent to HER2-positive cells, it may be a highly interesting treatment option in the early-stage setting for patients with tumors that are HER2 positive based on immunohistochemistry (score of 3 +), but of a non-HER2 subtype as determined by gene expression analysis.
Our study has a few limitations. Despite taking measures to ensure adequate matching, cases and controls differed with respect to chemotherapy regimens, tumor grade and nodal stage. Given that anthracycline-based treatment was preferred as a chemotherapy regimen in the Netherlands during the study period, the majority of control patients had received anthracyclines, which made perfect matching impossible, and also affected the matching of the other clinical variables [
36]. In addition, tumor grade was only available after case–control selection and thus not accounted for during matching. Tumor stage (II vs. III) was included as a matching factor but could not prevent an imbalance in nodal stage. Given that several studies have shown that anthracycline-containing and anthracycline-free chemotherapy regimens lead to comparable outcomes in HER2-positive breast cancer, we consider it unlikely that this has substantially affected our results [
6,
8,
17,
37]. In addition, since the imbalances are not severe and all analyses have been corrected for these variables, we are confident that they have not impacted our results significantly, although residual confounding cannot be fully excluded. The majority of patients that received pertuzumab underwent their treatment as part of the TRAIN-2 clinical trial, whereas patients that did not receive pertuzumab were not trial participants. Given that our study is based on data from the Netherlands Cancer Registry, which does not provide detailed information on comorbidity, we cannot fully exclude the possibility that the non-pertuzumab cohort is overall less healthy than the pertuzumab cohort. However, since some hospitals did and some did not participate in the TRAIN-2 trial, and inclusion rates were high among participating hospitals, study participation was largely based on in which hospital the diagnosis was made and thus mostly arbitrary. In addition, since all of the patients in our study cohort were fit for chemotherapy and were of a similar age, we have no reason to assume substantial health differences between the pertuzumab and the non-pertuzumab groups. Due to the imbalances in the matched groups and the potential selection bias, further validation of our results in independent datasets is warranted.
In conclusion, our results indicate that in patients with stage II-III HER2-positive breast tumors that are classified as HER2-type or HER2-single-type upon molecular characterization, the addition of pertuzumab to neoadjuvant chemotherapy and trastuzumab may improve pathological complete response and may decrease the risk of death due to breast cancer. Given the excellent long-term outcomes after treatment with dual HER2-blockade in patients with HER2-type or HER2-single-type tumors, molecular subtyping might be a valuable biomarker for candidate selection in future trials investigating either de-escalation of neoadjuvant chemotherapy, or alternative or intensified treatment strategies. Prospective validation of our findings is needed to confirm the role of BluePrint in patient selection for dual HER2-blockade.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.