307P - Identifying new biological subgroups of triple-negative breast cancer using next-generation integrative clustering pipeline

Background

Early triple-negative breast cancer (eTNBC) is molecularly heterogenous, with differential prognosis. We aimed to identify new biological subgroups (B-grp) of eTNBC via an integrative clustering pipeline (ICP) and evaluate the clinical value of these B-grp in a large randomised clinical trial.

Methods

TACT2 was a phase 3, randomised trial testing adjuvant accelerated versus standard Epirubicin in early breast cancer. From TACT2, 551 eTNBC tumours were stratified into training (n=367) and testing (n=184) by treatment arm. Expressions of 758 genes (GE) and Triple-Negative Subtypes ( TNS; Burstein CCR 2015 ) of tumour were profiled by nCounter^@ Breast Cancer 360^TM. Immunohistochemistry expressions (IHC) of HER2, EGFR and CK5/6 were assessed. Clinicopathological features (ClinPath) included age, nodal status, tumour size and grade. The ICP included random forest (RF) and 9 unique clustering methods with hypergeometric test to reconcile clustering results to a final output. Survival endpoint was time to recurrence (TTR) for multivariable cox regression (mCox), of which the goodness-of-fit was assessed by chi-square, brier score and integrated calibration index.

Results

The ICP returned 4 B-grp using GE, IHC and ClinPath of the training set. GE was the main factor of clustering, summarising B-grp into immune-enriched (IM), Luminal-AR (LAR), mesenchymal like (MES) and basal like (BL). IM and LAR contained mainly TNS-BLIA and TNS-LAR, respectively. Both MES and BL comprised TNS-BLIA, TNS-BLIS and TNS-MES. The RF was trained using GE as unimodal single-sample predictor to assign the B-grp on the testing set. Adding B-grp to ClinPath yielded more accurate 5-year TTR prediction for test samples when compared to TNS (Table). Table: 307P

Comparing the goodness-of-fit of three mCox models using chi-square, brier score and integrated calibration index

Brier score is the mean square difference between the true classes (0 = no recurrence, 1 = recurrence) and the predicted probabilities at given time point. Integrated calibration index is the weighted absolute difference between the calibration curve and the diagonal line of best fit.

Conclusions

We identified 4 eTNBC B-grp based on distinct biology which provided more accurate 5-year TTR prediction in addition to ClinPath than TNS. The association of the B-grp with survival for other treatments could be further evaluated.

Clinical trial identification

NCT00301925.

Editorial acknowledgement

Legal entity responsible for the study

The Institute of Cancer Research and Lothian Health Board.

Funding

The Institute of Cancer Research; Cancer Research UK; Ontario Institute for Cancer Research with funds from the Ontario Government.

Disclosure

J. Bayani: Other, Personal, Other, Gene Signature of Residual Risk Following Endocrine Treatment in Early Breast Cancer (Patent Title), National Phase Application, Patent number: 2016368696 (Country: Australia, Date: Mar. 10, 2022); 2016800813945 (China, Mar. 18, 2022); 3387168 (Europe, May 12, 2022); 7043404 (Japan, Mar. 18, 2022); 11,566,292 (United States, Jan.31, 2023): Patent - Granted. H. Tovey: Financial Interests, Institutional, Other, Salary funded by grants to her institution: AstraZeneca UK Ltd, Eli Lilly and Company Limited, Pfizer Inc, Aventis Pharma Limited. J. Bliss: Financial Interests, Institutional, Research Funding: AstraZeneca, Merck Sharp & Dohme, Puma Biotechnology, Pfizer, Roche, Novartis (previously GSK), Eli Lilly, Janssen-Cilag, Clovis Oncology; Financial Interests, Personal, Other, Travel, accommodations, expenses: Pfizer. D.A. Cameron: Financial Interests, Institutional, Funding: Aptitude Health, Roche Sweden, Pfizer Limited, Celldex Therapeutics Inc, Carnall Farrar, San Antonio Breast Cancer Consortium, Highfield Communication, AstraZeneca Global Commercial Organisation - Egypt, Celgene Corporation, Chief Scientists Office, Cancer Research UK, HTA, Samsung Bioepis Co Ltd (Korea), Eli Lilly & Company, Costello Medical Consulting Ltd, prIME Oncology, AstraZeneca UK Limited, Roche Products Ltd, Novartis Pharma AG, Novartis Pharmaceuticals Corporation, Pfizer Limited, PFS Ltd, Novartis Pharmaceuticals UK Limited, Merck Sharp Dohme Limited, PUMA Biotechnology, Inc., Pfizer Limited, F. Hoffmann-La Roche AG, Clovis Oncology, Breast International Group (BIG), Breast Cancer Institute, Daiichi Sankyo, USA, Eisai, Elsevier Ltd, European Cancer Organisation, Exact Therapeutics, G1 Therapeutics, Galapagos NV, Genentech Inc, GSK, ICON Clinical Research, Prima BioMED, RTI Health Solutions, Seattle Genetics, Erytech Pharma, Succinct Medical Communications, Seagen, Sapience Therapeutics Ltd , Bexon / Zymeworks Biopharmaceuticals Inc., Make Seconds Count, Next Gen Healthcare Communications, Clarity Pharmaceuticals, Immutep, Oncolytics Biotech (U.S), Grail UK, Byondis, Gilead Sciences Ltd UK. J.M.S. Bartlett: Financial Interests, Personal, Advisory Role, Consultancy: BioNTech AG, Biotheranostics, Inc., RNA Diagnostics, Inc., oncoXchange/MedcomXchange Communications Inc, OncoCyte Corporation, Ontario Institute for Cancer Research, AstraZeneca, Cerca Biotech; Financial Interests, Personal, Advisory Board: MedcomXchange Communications Inc; Financial Interests, Personal, Other, Honoraria: MedcomXchange Communications Inc; Financial Interests, Personal, Research Funding: Thermo Fisher Scientific, Genoptix, Agendia, NanoString Technologies, Inc., Stratifyer GmbH, Biotheranostics, Inc., Exact Sciences; Other, Personal, Other, CIN4 predicts benefit from anthracycline, National Phase Application, (Canada, Jan. 11, 2017): Patent - Applied; Other, Personal, Other, Systems, Devices and Methods for Constructing and Using a Biomarker, National Phase Application, 15/328,108 (United States, Jan. 23, 2017); 15824751.0 (Europe, Jan. 12, 2017); (Canada, Jan. 12, 2017): Patent - Applied; Other, Personal, Other, Immune Gene Signature Predicts Anthracycline Benefit, PCT (international application), PCT/CA2016/000305, Filing date: Dec. 7, 2016: Patent - Applied; Other, Personal, Other, Gene Signature of Residual Risk Following Endocrine Treatment in Early Breast Cancer (Patent Title), National Phase Application, Patent application number 3007118 (Country of Filing: Canada – Patent Application Date: 06/01/18): Patent - Applied; Other, Personal, Other, A Molecular Classifier for Personalized Risk Stratification for Patients with Prostate Cancer (Invention Title), PCT International Application No.: PCT/CA2021/050837, International Filing Date: June 18, 2021: Patent - Applied; Other, Personal, Other, CIN4 predicts benefit from anthracycline (Invention Title), National Phase Application, Patent number: 11214836 (Country: United States, Date: Jan. 4, 2022); 3169815 (Europe, Dec. 23, 2020): Patent - Granted; Other, Personal, Other, Targeting the Histone Pathway to Detect and Overcome Anthracycline Resistance (IP Title), National Phase Application, Patent number: 2016800728463 (Country: China, Date: Aug. 27, 2021); 11,015,226 (United States, May 25, 2021); 3359508 (Europe, Sept. 9, 2020); 3,000,858 (Canada, Sept. 27, 2022): Patent - Granted; Other, Personal, Other, Gene Signature of Residual Risk Following Endocrine Treatment in Early Breast Cancer (Patent Title), National Phase Application, Patent number: 2016368696 (Country: Australia, Date: Mar. 10, 2022); 2016800813945 (China, Mar. 18, 2022); 3387168 (Europe, May 12, 2022); 7043404 (Japan, Mar. 18, 2022); 11,566,292 (United States, Jan.31, 2023): Patent - Granted; Other, Personal, Other, Disclosure Name: A Molecular Classifier for Personalized Risk Stratification for Patients with Prostate Cancer, Date: 21/08/2019: Invention Disclosure. M.C.U. Cheang: Financial Interests, Personal, Funding: NanoString Technologies; Financial Interests, Personal, Advisory Role: Veracyte; Other, Personal, Licencing Fees or royalty for IP, Breast Cancer Classifier: US Patent No. 9,631,239.: Patent. All other authors have declared no conflicts of interest.