Effects of Age and Immune Landscape on Outcome in HER2-Positive Breast Cancer in the NCCTG N9831 (Alliance) and NSABP B-31 (NRG) Trials

Breast cancer is among the most common malignancies afflicting women worldwide (1). Although breast cancer is generally a disease of aging, a number of patients develop breast cancer at a young age. In developed countries, breast cancer is the leading cause of cancer-related death among women younger than 40 years of age (2), and approximately 6.6% of breast cancer occurs in patients younger than 40 (3). Several studies consistently demonstrated that young age is an independent factor associated with adverse outcome (4–7), which can only partly be attributed to aggressive phenotypes. Breast cancers in young patients are more likely to have hormone receptor (HR) negativity, lymphovascular invasion, and high histologic grade compared with those in older patients (8). Nevertheless, after adjustment for these known aggressive phenotypes, young age continues to be an independent predictor of recurrence and poor outcome (7).

Early genomic studies established that breast cancers arising in younger patients are different from those occurring in older patients. One of the gene sets that differs in breast cancers occurring in younger patients is the immune function–related gene set (4). Still, the immune architecture of breast cancers arising in younger patients compared with older patients has not been elucidated to date. Furthermore, given that breast cancer is a heterogeneous disease comprised of multiple subtypes, it is critical to discern the immune architecture for each intrinsic subtype of breast cancer. HER2-positive breast cancer is a unique subset, accounting for approximately 15% to 30% of breast cancers (9, 10). Historically, HER2 overexpression conferred poor outcome, with high rates of disease recurrence, metastasis, and death. However, with the advent of treatments targeting HER2, particularly trastuzumab, the outcome of patients with HER2-positive breast cancer has drastically improved in the past two decades (11).

Trastuzumab, an mAb against the extracellular domain of HER2, is thought to exert its activity via several mechanisms, including inhibition of downstream signaling of HER2 (12) as well as immune-mediated tumor cell killing, particularly by antibody-dependent cellular cytotoxicity (13, 14). To date, the effects of patients’ preexisting immune response according to age and benefit of trastuzumab have not been elucidated. In the previous study from the Herceptin Adjuvant (HERA) trial, age of 40 years or younger at study entry was not associated with risk of early recurrence (15). However, the median follow-up of this study was only 2 years. Therefore, we sought to evaluate the long-term outcome and immune landscape of adjuvant trastuzumab-treated HER2-positive breast cancer patients aged 40 years or younger compared with older patients. Patients in this study were accrued to the N9831 trial from North Central Cancer Treatment Group (NCCTG), now part of Alliance for Clinical Trials in Oncology, and the B-31 trial from National Surgical Adjuvant Breast and Bowel Project (NSABP), now part of NRG.

The NCCTG N9831 trial is a randomized, multicenter, phase III trial that enrolled women with primary, operable, node-positive or high-risk node-negative HER2-positive breast cancer who were eligible for adjuvant chemotherapy. Patients were randomly assigned to 1 of 3 arms: arm A received only standard adjuvant chemotherapy with doxorubicin and cyclophosphamide followed by weekly paclitaxel (AC-T); arm B patients received AC-T followed by trastuzumab after completion of chemotherapy; and arm C patients received AC-T concurrently with trastuzumab (AC-TH). NSABP B-31 trial is a two-arm, randomized, phase III trial that enrolled patients with operable, node-positive, HER2-positive breast cancer. Eligible patients were randomly assigned to AC-T, with paclitaxel either every 3 weeks or weekly, or to AC-TH. The study has been performed in accordance with the Declaration of Helsinki and good clinical practice guidelines. Each participant signed an Institutional Review Board–approved, protocol-specific informed consent document in accordance with federal and institutional guidelines. The primary results of these two trials were previously published in 2005 (16), 2011 (17, 18), and 2014 (19).

Data regarding pathologic quantification of stromal tumor-infiltrating lymphocytes (sTIL) were available in a subset of patient samples from AC-T and AC-TH arms of the NCCTG N9831 trial, as previously reported by Perez and colleagues (20). sTIL quantification was not available in the NSABP B-31 trial. Histopathologic analysis of sTILs was performed with a single hematoxylin and eosin–stained section using the method previously described by Loi and colleagues (21), Denkert and colleagues (22), Adams and colleagues (23), and Salgado and colleagues (24). The degree of sTILs was collected in deciles, and similar to the previous publication by Loi and colleagues (25), lymphocyte-predominant breast cancer (LPBC) was defined as tumors with 50% or more sTILs.

Gene expression data involving immune-related genes were analyzed in 1,378 samples from the NCCTG N9831 trial. There were no data regarding immune-related gene expression in the NSABP B-31 samples. NanoString (NanoString Technologies, Inc.) was used to quantify mRNA in RNA from paraffin-embedded tumor samples from NCCTG N9831. NanoString custom CodeSets were constructed to comprise 1,252 genes, including 5 housekeeping genes (B2M, GAPDH, POLR2A, UBC, YWHAZ) for normalization purposes. Background variation was removed by correcting for the geometric mean of positive spike-in RNAs, and by subtracting the mean plus 2 SDs of negative spike-ins. Normalization proceeded by multiplying a patient's expression profile by a factor scaled to the geometric mean of their housekeeping genes, via the R package NanoStringNorm (26). Molecular tumor-infiltrating lymphocyte (mTIL) signatures were calculated using normalized and log₂ transformed data as previously published by Danaher and colleagues (27). The geometric mean across relevant genes (Supplementary Table S1) was calculated to generate the composite score for each immune subset signature, namely CD45, B cell, CD8 T cell, cytotoxic cell, exhausted CD8, immature dendritic cell (iDC), macrophage, mast cell, neutrophil, natural killer (NK) CD56^dim cell, T cell, and regulatory T-cell signatures. The score was mean-centered and standardized in all analyses.

For outcome analysis, recurrence-free survival (RFS) was defined as the time from random assignment to breast cancer recurrence (local, regional, or distant recurrence of breast cancer or breast cancer–related death). The time to event for patients who died without recurrence was considered censored at the time of death. Log-rank P values were employed to assess survival differences among groups with more than two categories. Cox proportional hazards models were used to generate point estimates of HRs and corresponding 95% confidence intervals (CI) to assess the benefit of trastuzumab for RFS in each age group, with age treated as both a continuous and dichotomous variable (defined as age ≤40 vs. >40 years by Partridge and colleagues; ref. 15). The effect of age was also assessed in multivariate HRs controlling for hormonal status, breast cancer subtype, and immune subtype scores. Gene Ontology Enrichment Analysis (Gene Ontology Consortium) was used to calculate fold enrichment of biologic processes of genes of interest. All statistical analyses were carried out in R version 3.2.3 (The R Foundation) on a dataset locked on November 14, 2017.

There were 4,547 patients with available data from both trials, with 3,132 from NCCTG N9831 and 1,415 from NSABP B-31 (Supplementary Fig. S1A and S1B). The median follow-up was 10.3 years in NCCTG N9831 and 8 years in NSABP B-31. The median age was similar in both trials, with 49 years (range, 19–82) in NCCTG N9831 and 49.6 years (range, 26–77) in NSABP B-31. Across both trials, 3,625 (79.72%) patients were older than 40 years and 922 (20.28%) were 40 years or younger. Patients’ baseline characteristics for both NCCTG N9831 and NSABP B-31 are stratified by age in Table 1, respectively. Similar to the previous report by Partridge and colleagues (15), younger age (≤40) was associated with significantly higher HR positivity in NCCTG N9831, but not in NSABP B-31 (χ² P = 0.02 and P = 0.78, respectively). Tumor grade was also higher in younger patients (P = 0.04). However, there was no significant difference in tumor size or nodal status among younger versus older patients (Table 1).

Comparing younger versus older patients, similar survival benefit of trastuzumab was observed in both age groups (age by treatment interaction P = 0.1; P = 0.27 for NCCTG N9831 and P = 0.82 for NSABP B-31). In the combined analysis of both NSAPB B-31 and NCCTG N9831 trials, age was a significant predictor of poor outcome only in patients who received chemotherapy alone (P < 0.001) and not in patients who received trastuzumab-based chemotherapy (P = 0.12). In NCCTG N9831, 8-year RFS rate was 63.5% with AC-T versus 83.1% with AC-TH in younger patients (HR, 0.43; 95% CI, 0.28–0.66; P < 0.001; Fig. 1A). In older patients, 8-year RFS rate was 71.1% in the AC-T arm versus 83.4% in the AC-TH arm (HR, 0.56; 95% CI, 0.45–0.69; P < 0.001; Fig. 1B). In NSABP B-31, 8-year RFS rate was 46.3% with AC-T versus 72.0% with AC-TH in younger patients (HR, 0.45; 95% CI, 0.29–0.68; P < 0.001; Fig. 1C) and 60.3% and 80.3%, respectively, in older patients (HR, 0.42; 95% CI, 0.33–0.54; P < 0.001; Fig. 1D).

Given that previous studies of HER2-positive breast cancer demonstrated differential complete pathologic response to neoadjuvant chemotherapy among patients with HR⁺ versus HR⁻ disease (28, 29), we evaluated outcome as a function of HR status. Furthermore, to match NSABP B-31 trial, which only included patients with lymph node involvement, we only analyzed patients with lymph node involvement in the NCCTG N9831 trial (N = 2,716). We observed a significant difference in outcome among younger versus older patients with HR⁺ disease receiving AC-T (P = 0.04, Fig. 2Ai), although HR⁻ patients did not reach this threshold (P = 0.09, Fig. 2Aii; P_interaction > 0.1). Similar results were observed in NSABP B-31 (P = 0.01, Fig. 2Bi and P = 0.60, Fig. 2Bii; P_interaction > 0.1). However, there was no significant difference in outcome between younger and older patients who received AC-TH in either trial, regardless of their HR status (Fig. 2Aiii, 2Aiv, 2Biii, and 2Biv).

Pathologic quantification of sTILs was available in 831 patients from the NCCTG N9831 trial with 435 patient samples from the AC-T arm and 396 from the AC-TH arm. In this cohort, 159 (19.13%) patients were 40 years old or younger and 672 (80.87%) patients were older than 40. One hundred twenty-seven (15.3%) patient samples had sTILs of 50% or greater and were categorized as LPBC. Overall, there was no significant association between the amount of sTIL by deciles and age in both continuous (P = 0.19; Fig. 3A) and dichotomized analyses (χ² P = 0.75). Among younger patients, 14.4% (23/159 patients) had LPBC. Similarly, 15.5% (104/672) of older patients had LPBC. More comprehensive analysis was carried out using gene expression signatures to identify individual immune infiltrates. mTILs used CD45 to estimate total leukocyte infiltration. There was a small but statistically significant increase in mTIL CD45 expression with older age (linear regression, P = 0.01; Fig. 3B). Similar trends were observed in other mTIL subset signatures (Table 2), including cytotoxic cells (P = 0.03), exhausted CD8 (P = 0.02), macrophages (P < 0.001), neutrophils (P < 0.001), NK CD56^dim (P < 0.001), T cells (P = 0.02), and regulatory T cells (P = 0.04). However, there was no statistically significant difference in B cells (P = 0.09), CD8 T cells (P = 0.14), and iDCs (P = 0.17) among younger and older subgroups.

Patient outcomes related to sTILs and age were further evaluated in the 831-patient cohort with available sTIL data in the NCCTG N9831 trial. In patients with LPBC who received AC-T, there was statistically significant improvement in RFS (HR, 0.18; 95% CI, 0.05–0.7; P = 0.01; Fig. 4A) in patients older than 40 compared with younger than 40. The older patients in this group had excellent outcomes, with 95% remaining recurrence free at 15 years, despite not receiving adjuvant trastuzumab. However, younger patients had poor RFS when receiving AC-T, regardless of LPBC status. In contrast, among patients who received AC-TH, there was no significant difference in RFS (P = 0.26; Fig. 4B) among younger and older patients stratified by LPBC status. Similar findings were observed using age and sTILs as continuous variables (Supplementary Table S2). In patients who received AC-T, there was no significant difference in RFS when age and sTILs were analyzed as continuous variable (HR, 1.00; 95% CI, 0.97–1.02; P = 0.74 and HR, 1.20; 95% CI, 0.75–1.92; P = 0.44, respectively). Similar findings are observed in patients who received AC-TH (HR, 1.01; 95% CI, 0.98–1.04; P = 0.52 and HR, 1.25; 95% CI, 0.72–2.18; P = 0.43). Furthermore, the age by sTIL interaction was significant with an age dichotomization of 40 years, but was not for 50 or 60. As shown in Supplementary Fig. S2 and Supplementary Table S3, younger LPBC patients have significantly higher levels of genes enriched in proliferation (P = 0.02), but lower levels of genes enriched in chemokines related to leukocyte migration and function (P < 0.001). In contrast, older patients have significantly higher levels of genes involved in immunosuppression and exhaustion, including FOXP3, HAVCR2, LAG3, and ICOS.

Similar to other studies (3, 4, 30), our study demonstrates that breast cancers arising in younger patients have distinct characteristics and gene expression compared with older counterparts. Overall, breast cancer in younger patients is associated with more aggressive phenotypes. When combining all breast cancer subtypes, previous studies showed younger patients were more likely to have HR⁻ disease, higher HER2 protein expression, larger tumors, pathologic grade 3, lymphovascular invasion, and lymph node involvement compared with older patients (8). In HER2-positive patients, previous HERA trial reported that younger patients are more likely to have estrogen receptor–positive and progesterone receptor–positive disease (15). However, in our study, younger age was associated with higher HR positivity compared with older age in NCCTG N9831 trial but not in NSABP B-31 trial. These findings signify the importance of evaluating biologic process related to age separately in each breast cancer subtype.

Among patients who received AC-T, our analysis in the NCCTG N9831 and NSABP B-31 trials demonstrated that younger age confers significantly poorer outcomes among patients with HR⁺ HER2-positive, but not HR⁻ HER2-positive breast cancer. This is in contrast to the HERA trial report (15), which found no significant difference in disease-free survival (HR, 1.16; 95% CI, 0.89–1.51; P = 0.27) among patients treated with AC-T when comparing between younger and older patients. However, outcome based on HR status in the context of age was not reported. Furthermore, the median follow-up in that particular report was relatively short, with only 2 years of median follow-up. Nevertheless, the poor outcomes associated with younger age were not observed among patients who received AC-TH. This finding is similar to the HERA trial report, which also showed no significant difference in disease-free survival among patients treated with AC-TH (HR, 1.18; 95% CI, 0.87–1.61; P = 0.30; ref. 15). To our knowledge, our analysis represents the first study to demonstrate that trastuzumab may mitigate the poor prognosis associated with younger age in patients with early stage HER2⁺ breast cancer as age was not significantly associated with poor outcome in our combined analysis in patients treated with adjuvant trastuzumab-based chemotherapy. Nevertheless, among patients received trastuzumab in the NSABP B-31 trial, there was numerically more number of older patients without recurrence compared with younger patients (80.3% vs. 72%, respectively). Given the fact that patients in the NSABP B-31 had higher risk disease with more lymph node involvement than patients in the NCCTG N9831, it is possible that the detrimental effect of age may exist in patients with higher risk disease. In addition, our study also shows the difference in outcomes among HR⁺ versus HR⁻ within HER2-positive disease based on age. However, beneficial effects of trastuzumab on age persist, regardless of HR status.

One potential explanation for the poorer outcome observed only in younger patients with HR⁺ HER2⁺ breast cancer but not HR⁻ HER2⁺ breast cancer may be due to the type of adjuvant endocrine therapy used in these trials. The accrual period of our two trials preceded the information from the SOFT and TEXT trials (31), which demonstrated superiority of ovarian suppression in combination with tamoxifen or aromatase inhibitor in premenopausal patient with HR⁺ disease. The majority of younger premenopausal patients in our two trials were mainly treated with adjuvant tamoxifen and older postmenopausal patients were treated with adjuvant aromatase inhibitors. Giving the fact that aromatase inhibitor confers approximately 30% improvement in the outcome compared with tamoxifen (32), this may explain the poorer outcome observed in younger patients only in the HR⁺ HER2⁺ subset.

Using gene set enrichment analysis, a previous large-scale genomic study demonstrated that distinct gene sets that are different in younger patients involve immune function, hypoxia, BRCA1, stem cells, apoptosis, histone deacetylase, and multiple other oncogenic signaling pathways, including MYC, E2F, RAS, and mTOR (4). Although this previous study showed that immune function genes are among the gene sets distinctly different with age, the actual difference in immune-related genes in tumors arising in younger versus older patients remains largely unknown. In general, aging of the immune system is associated with a decline in both humoral and cellular adaptive immune response. This process is termed immune senescence (33). Due to involution of the thymic gland with aging, the number and portion of naïve T cells considerably decrease with aging process (34). Furthermore, there is an increase in the proportion of memory T cells, which may cause defects in cytokine production (35) and accumulation of terminally differentiated T cells that are dysfunctional and have limited T-cell receptor repertoire diversity (36). Nevertheless, the immune landscape of younger versus older patients with breast cancer, specifically with HER2⁺ subtype, is not defined. In contrast to a decline in the immune response expected from immune senescence, we observed no statistically significant difference in the amount of sTILs measured by pathologic quantification. However, with more in-depth analysis using gene expression signatures, there were small, but significant increases in mTIL signatures for total lymphocytes, cytotoxic cells, macrophages, neutrophils, NK CD56^dim cells, and total T cells. Older patients appeared to have slight increases in exhausted CD8 T cells and regulatory T cells, but there was no statistically significant difference in B cells, CD8 T cells, and iDCs.

In our analysis, older patients with LPBC appeared to have excellent outcomes with AC-T, despite not receiving trastuzumab. However, the number of older patients with LPBC was too small with only 55 patients receiving AC-T and 49 patients receiving AC-TH. Therefore, definitive conclusion about the added benefit of trastuzumab cannot be determined. When comparing gene expression among younger and older patients with LPBC, it appears younger patients have lower levels of chemokines critical for immune function. In contrast, older patients appear to have more immunosuppression and exhaustion. These findings highlight that immune escape mechanisms may differ between younger and older patients. Therefore, different treatment strategies may be needed to modulate immune response to improve outcomes in these 2 distinct groups of patients.

However, our study has several limitations. One of the limitations is the lack of predetermined statistical analysis plan to evaluate specific association between age and outcome in both of the trials. Furthermore, the data of sTIL and gene expression analysis of immune-related genes were not available in both trials. Currently, the data of sTIL and gene expression analysis of immune-related gene were only available in the NCCTG-N9831 trial and not the NSABP B-31 trial. Therefore, we were unable to further validate these findings from the NCCTG-N9831 trial in the NSABP B-31 trial. Therefore, the results of this analysis should be viewed as hypothesis generation.

S. Chumsri reports receiving commercial research grants from Merck & Co. S. E. Warren holds ownership interest (including patents) in NanoString Technologies. G. Colon-Otero reports receiving commercial research grants from Novartis. E.A. Perez is an employee of Genentech through May 2018 and the Mayo Clinic; holds ownership interest (including patents) in Roche; and is a consultant/advisory board member for Puma Oncology and Saichii Dankyo. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Chumsri, K.L. Knutson, E.A. Perez, E.A. Thompson

Development of methodology: E.A. Perez, E.A. Thompson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Chumsri, K.L. Pogue-Geile, A.E. Soyano-Muller, E.A. Perez, A. Moreno-Aspitia, E.A. Thompson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Chumsri, D.J. Serie, Z. Li, A.E. Soyano-Muller, A. Mashadi-Hossein, S. Warren, Y. Lou, E.A. Perez, A. Moreno-Aspitia, E.A. Thompson

Writing, review, and/or revision of the manuscript: S. Chumsri, D.J. Serie, K.L. Pogue-Geile, A.E. Soyano-Muller, A. Mashadi-Hossein, Y. Lou, G. Colon-Otero, K.L. Knutson, E.A. Perez, A. Moreno-Aspitia, E.A. Thompson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Chumsri, E.A. Thompson

Study supervision: S. Chumsri, G. Colon-Otero, E.A. Perez, E.A. Thompson

Research reported in this publication was supported by the NCI of the NIH under Award Numbers U10CA180821 and U10CA180882 (to the Alliance for Clinical Trials in Oncology); U10CA180868 (NCTN), UG1CA189867 (NRG NCORP), U10CA180822 [NRG Oncology SDMC (Biostats)], U24CA196067 (BSB; Lab; NRG); and U24CA196171. This work was also supported in part by funds from the Breast Cancer Research Foundation (BCRF-17-161), Bankhead-Coley Research Program (6BC05), the DONNA Foundation, the Pennsylvania Department of Health and Genentech. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

NRG LAB Disclaimer: "Under a grant from the Pennsylvania Department of Health. The Department specifically disclaims responsibility for any analysis, interpretations, or conclusions."

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.