Background

Invasive lobular carcinoma (ILC) accounts for approximately 15% of all breast cancer cases. Most of the correlations between multigene assays and patient outcome were derived from studies based on mixed cohorts with IDC and ILC. Here, we specifically investigate the prognostic value of Endopredict (EPclin) in a large cohort of ILCs pooled from large phase III randomised trials.

Methods

A total of 2630 breast cancer patients with ER-positive, HER2-negative disease were included in this analysis. The primary objective of this analysis was to determine the prognostic value of EPclin for distant recurrence (DR) in years 0-10 in women with ILC. Secondary objectives included subgroup analysis and comparison of the prognostic value of EPclin in ILC vs. IDC.

Results

470 women (17.9%) presented with an ILC, 1944 (73.9%) with IDC, and 215 (8.2%) had other histological types. EPclin was highly prognostic in women with ILC (HR = 3.32 (2.54-4.34), P < 0.0001) and provided more prognostic value than clinicopathological parameters alone (HR = 2.17 (1.73-2.72), P < 0.0001) (added value: ΔLR-χ²=17.6). 298 women (63.4%) were categorised into the low EPclin risk group and they had a 10-year DR of 4.8% (2.7-8.4) compared to those 172 (36.6%) women in the high risk group with a 10-year DR risk of 26.6% (20.0-35.0) (HR = 6.33 (3.31-12.12), P < 0.0001). EPclin provided highly prognostic information in LN- (N = 326; HR = 2.56 (1.63-4.02)) and LN+ disease (N = 144; HR = 3.70 (2.49-5.50)). For women with IDC, EPclin categorised 59.1% of women into the low risk group and 40.9% into the high EPclin risk group (HR = 5.04 (3.67-6.90)). 17% fewer women were deemed high risk by EPclin if they had ILC vs. IDC.

Conclusions

We present gene expression data and clinical outcome from the largest cohort of patients with ILC treated with adjuvant endocrine therapy in the absence of chemotherapy. EPclin provided highly significant prognostic value and significant risk stratification for women with ILC. 10-year DR risk in the EPclin low risk groups were similar between ILC and IDC, suggesting that chemotherapy in these low risk groups is not indicated, irrespective of tumour type. Our results suggest that histological subtyping is not indicated when interpreting EndoPredict.

Legal entity responsible for the study

TransATAC and ABCSG Research Groups.

Funding

Cancer Research UK.

Disclosure

I. Sestak: Lecture fees: Myriad Genetics, NanoString. M. Knauer: Advisory board: Agendia, Genomic Health; Lecture fees: Agendia. R. Kronenwett: Inventor on the Endopredict patent; Employee, previous shareholder: Sividon. J. Cuzick: Research grants: AstraZeneca, Myriad Genetics, Memorial Sloan Kettering, Qiagen, Beckton Dickinson, Genera, Aventis Pharma; Honoraria: Merck, Roche, Qiagen, Myriad Genetics; Speaker’s bureau: Beckton Dickinson/Hologic. M.I. Gnant: Honoraria: Amgen, AstraZeneca, Celgene, Eli Lilly, Invectys, Pfizer, NanoString, Novartis, Roche; Advisor board: AstraZeneca, Eli Lilly; Research grants: AstraZeneca, Novartis, Roche, Pfizer. M. Dowsett: Honoraria: Pfizer, Myriad Genetics, Roche; Consulting or advisory role: Roche/Genentech, GTx, Radius Health; Research funding (Inst): AstraZeneca, Novartis, Pfizer, Radius Health. P. Dubsky: Grant support: Agendia, Sividon, Myriad Genetics, NanoString; Advisory role: Amgen; Speaker’s fee: Myriad Genetics, Sividon, Amgen. All other authors have declared no conflicts of interest.

Background

Breast conserving surgery (BCS) followed by radiotherapy is the standard treatment for early stage breast cancer. As a result of uncertainty about the absolute benefit of radiotherapy in older patients, there is variation in the proportion of patients receiving radiotherapy after BCS. In this study, we assessed the variation between hospitals in the Netherlands among patients aged 75 years or older and evaluated its impact on locoregional recurrence risk.

Methods

Eligible patients diagnosed between 2003 and 2009 were selected from the Netherlands Cancer Registry. Since directly comparing treated and untreated patients would lead to bias due to confounding by indication, the impact of omission of radiotherapy on locoregional recurrence risk was assessed using individual patient data with hospital as instrumental variable. To this end, hospitals were sorted according to the proportion of radiotherapy given and subsequently the cohort was divided into three equally sized groups (radiotherapy high, middle and low) which were analysed using Fine and Gray regression models. This procedure was repeated stratified by age and comorbidity.

Results

Overall, 2390 patients were included. In the “radiotherapy high” group 96% of the patients underwent radiotherapy compared to 88.0% in the “radiotherapy middle” group and 72.2% in the “radiotherapy low” group. Compared to the “radiotherapy high” group, no significant difference in locoregional recurrence risk was observed for the “radiotherapy middle” group (adjusted sHR 1.58, 95% CI: 0.74-3.40, p = 0.240) or “radiotherapy low” group (adjusted sHR 1.76, 95%: 0.81-3.81, p = 0.150). Variation in the proportion of patients undergoing radiotherapy increased with increasing age and comorbidity, however, no significant difference in locoregional recurrence risk was seen between the constructed radiotherapy groups stratified by age and comorbidity.

Conclusions

This study shows that omission of radiotherapy after BCS among patients aged 75 years or older is not uncommon in the Netherlands. However, the observed variation in proportion of patients receiving radiotherapy between hospitals was not reflected in the locoregional risk.

Legal entity responsible for the study

Leiden University Medical Center.

Funding

ZonMw.

Disclosure

All authors have declared no conflicts of interest.

Background

The immune system is critical in modulating therapeutic sensitivity and tumor progression in breast cancer patients. A plethora of methods exists to evaluate the amount and composition of tumor immune infiltrates and are used interchangeably, based on the assumption that they provide similar information. A systematic comparison of these methods, including microscopy evaluations and algorithms based on transcriptomic and methylation data, is still lacking.

Methods

We characterized stromal (str) and intra-tumoral tumor infiltrating lymphocytes (TIL) and 6 immune cell-types (CD3⁺, CD4⁺, CD8⁺, CD20⁺, CD68⁺, FOXP3⁺) using immunohistochemistry in the 560 genomes cohort from the International Cancer Genomics Consortium (Nik-Zainal Nature 2016). The same traits were computed using deconvolution methods (CIBERSORT, methylCIBERSORT, quanTIseq, EPIC), as well as published transcriptomic or methylation-based immune signatures. We first studied the associations of immune cells as continuous variables and further categorized tumors as hot (strTIL ≥60%) or cold (strTIL ≤10%).

Results

The immune infiltrate was reproducibly assessed by pathologists for all cell types and concordance correlation coefficients ranged from 0.63 (for strCD4) to 0.84 (for strTIL). The correlations between all methods to assess global immune infiltration were extremely variable, ranging from 0.08 to 0.95 (Table). The correlations between methylation or transcriptomic estimates and histopathology were weak to moderate and did not exceed 0.56. Several transcriptomic estimates were strongly correlated with each other (>0.85). ROC analyses showed that most methods can more accurately identify hot tumors as compared to cold tumors. Comparison regarding the specific immune cell types further highlighted heterogeneity between the different methods. Table. Spearman rank (r) between methods to assess absolute presence of tumor infiltrating lymphocytes.

Conclusions

This study highlights important differences between the currently existing methods in the assessment of global immune infiltration in breast cancer and raises therefore extreme caution when assessing immune infiltrates in the clinical context.

Legal entity responsible for the study

The International Cancer Genome Consortium (ICGC).

Funding

The Dutch Cancer Society (DCS) Fondation Lambeau-Marteaux Les Amis de Bordet Fondation Cancer (Luxemburg).

Disclosure

All authors have declared no conflicts of interest.

Table: 3O

Background

Higher stromal TILs (sTILs) are associated with higher pathologic complete response (pCR) rates with CT in early-stage TNBC. We evaluated the association between sTILs and PD-L1 expression with response to pembro+CT as NAT for TNBC in KEYNOTE-173 (NCT02622074).

Background

Higher stromal TILs (sTILs) are associated with higher pathologic complete response (pCR) rates with CT in early-stage TNBC. We evaluated the association between sTILs and PD-L1 expression with response to pembro+CT as NAT for TNBC in KEYNOTE-173 (NCT02622074).

Trial design

sTILs were quantified pretreatment (PT) and on-treatment (OT; during first 3 weeks of pembro monotherapy). PT PD-L1 expression was reported as combined positive score (CPS). Endpoints were pCR rate (ypT0 ypN0; ypT0/Tis ypN0) and objective response rate (ORR) after the first 4 cycles of NAT (taxane±carboplatin+pembro). Correlation between PD-L1 and sTILs was assessed by Spearman's rank correlation coefficient.

Of 60 pts, 53 had tumors evaluated for PT sTILs, 50 for OT sTILs, 52 for PD-L1 CPS, and 51 for both sTILs and CPS. Overall pCR rates were 56.7% (ypT0 ypN0) and 60% (ypT0/Tis ypN0); ORR was 78.3%. There was significant correlation between PT sTILs and PD-L1 CPS (ρ=0.63; P<0.001) and between PT sTILs and OT sTILs (ρ=0.63; P<0.001). Higher sTILs were significantly associated with response (for PT: ypT0 ypN0 P=0.033; ypT0/Tis ypN0 P=0.02; ORR P=0.028; for OT: ypT0 ypN0 P=0.01; ypT0/Tis ypN0 P=0.008; ORR P=0.005). PT PD-L1 CPS was significantly associated with response: ypT0 ypN0 P=0.073; ypT0/is ypN0 P=0.030; and ORR P=0.021. Responders had higher median PT sTILs vs nonresponders: 40% [7.5, 72.5] vs 10% [5, 27.5] for pCR rate by ypT0 ypN0 and 42.5% [8.8, 71.2] vs 10% [5, 20] for pCR rate by ypT0/Tis ypN0; 20% [5, 70] vs 10% [4, 22.5] for ORR. Responders had higher median OT sTILs vs nonresponders: 65% [11.2, 85] vs 22.5% [4.5,41.2] for pCR rate by ypT0 ypN0 and 65% [5, 86.2] vs 22.5% [3.5, 35] for pCR rate by ypT0/Tis ypN0; 60% [10, 85] vs 10% [1, 30] for ORR. Likelihood ratio tests demonstrated that for both pCR endpoints, PD-L1 CPS (P=0.936/P=0.492) and PT sTILs (P=0.746/P=0.715) did not significantly add value to OT sTILs when predicting pCR.

Higher quantities of PT sTILs and PD-L1 CPS and OT sTILs were significantly associated with higher pCR rates and ORR in primary TNBC treated with pembro and NAT.

Methods

sTILs were quantified pretreatment (PT) and on-treatment (OT; during first 3 weeks of pembro monotherapy). PT PD-L1 expression was reported as combined positive score (CPS). Endpoints were pCR rate (ypT0 ypN0; ypT0/Tis ypN0) and objective response rate (ORR) after the first 4 cycles of NAT (taxane±carboplatin+pembro). Correlation between PD-L1 and sTILs was assessed by Spearman’s rank correlation coefficient.

Results

Of 60 pts, 53 had tumors evaluated for PT sTILs, 50 for OT sTILs, 52 for PD-L1 CPS, and 51 for both sTILs and CPS. Overall pCR rates were 56.7% (ypT0 ypN0) and 60% (ypT0/Tis ypN0); ORR was 78.3%. There was significant correlation between PT sTILs and PD-L1 CPS (ρ = 0.63; P < 0.001) and between PT sTILs and OT sTILs (ρ = 0.63; P < 0.001). Higher sTILs were significantly associated with response (for PT: ypT0 ypN0 P = 0.033; ypT0/Tis ypN0 P = 0.02; ORR P = 0.028; for OT: ypT0 ypN0 P = 0.01; ypT0/Tis ypN0 P = 0.008; ORR P = 0.005). PT PD-L1 CPS was significantly associated with response: ypT0 ypN0 P = 0.073; ypT0/is ypN0 P = 0.030; and ORR P = 0.021. Responders had higher median PT sTILs vs nonresponders: 40% [7.5, 72.5] vs 10% [5, 27.5] for pCR rate by ypT0 ypN0 and 42.5% [8.8, 71.2] vs 10% [5, 20] for pCR rate by ypT0/Tis ypN0; 20% [5, 70] vs 10% [4, 22.5] for ORR. Responders had higher median OT sTILs vs nonresponders: 65% [11.2, 85] vs 22.5% [4.5,41.2] for pCR rate by ypT0 ypN0 and 65% [5, 86.2] vs 22.5% [3.5, 35] for pCR rate by ypT0/Tis ypN0; 60% [10, 85] vs 10% [1, 30] for ORR. Likelihood ratio tests demonstrated that for both pCR endpoints, PD-L1 CPS (P = 0.936/P=0.492) and PT sTILs (P = 0.746/P=0.715) did not significantly add value to OT sTILs when predicting pCR.

Conclusions

Higher quantities of PT sTILs and PD-L1 CPS and OT sTILs were significantly associated with higher pCR rates and ORR in primary TNBC treated with pembro and NAT.

Editorial acknowledgement

Rajni Parthasarathy, PhD, and Diane Neer, ELS, MedThink SciCom.

Clinical trial identification

NCT02622074.

Legal entity responsible for the study

Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Funding

Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Disclosure

V. Karantza: Emplyee: Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, may hold stock/stock options: Merck & Co., Inc., Kenilworth, NJ, USA. R. Salgado: Research support: Roche, Merck. All other authors have declared no conflicts of interest.