Abstract PD2-07: Impact of using cross-platform gene expression profiling technologies and computational methods for intrinsic breast cancer subtyping in PALOMA-2 and PALLET

Background Intrinsic breast cancer molecular subtyping (IBCMS) provides significant prognostic information for patients (pts) with breast cancer (BC) treated with chemotherapy, human epidermal growth factor receptor 2 (HER2) targeted therapies, and endocrine therapies (ETs). Classifying tumors into intrinsic subtypes to determine optimal treatment is often applied using PAM50, commercially known as Prosigna. Meanwhile, Absolute Assignment of Breast Cancer Intrinsic Molecular Subtypes (AIMS) computational method was trained to predict PAM50-based IBCMS. As the PAM50 algorithm was developed to capture the major subtypes in a general pt population, clinicopathologic distribution of the study cohort and technology platform calibration should be considered in IBCMS analyses. This study compared different next-generation sequencing technologies and methodologic approaches of PAM50 on tumor samples from 2 randomized trials of postmenopausal women with estrogen receptor-positive (ER+)/HER2-negative (HER2-) BC. Methods PALOMA-2 is a double-blind, randomized study of first-line palbociclib (PAL) + letrozole (LET) for ER+/HER2- advanced BC (ABC). Tumor samples from consented pts were subtyped using the validated RUO PAM50 assay (ruoProsigna, NanoString); results were compared with published subtype results using AIMS on EdgeSeq Oncology Biomarker Panel (HTG Molecular Diagnostics). PALLET is a phase 2, randomized trial of PAL+LET as neoadjuvant therapy in pts with ER+ HER2- BC. Baseline frozen tumor biopsies underwent whole transcriptome mRNA-sequencing (mRNA-seq). IBCMS was performed using AIMS; PAM50 subtyping was performed on data normalised with subgroup-specific gene centering and microarray-RNA-sequencing calibration. Results In PALOMA-2, 222 pts had both ruoProsigna and AIMS data; an overall 54% agreement rate between methods was observed, with 46% (56/121) of Luminal B (LumB) subtype by ruoProsigna assigned as Luminal A (LumA) by AIMS and 67% (6/9) of basal-like by ruoProsigna as HER2-enriched (HER2-E) by AIMS (Table 1). In PALLET, 224 pts had mRNA-seq data; a 69% agreement between the two approaches (AIMS and PAM50) was observed, with only 4% (2/49) of LumB assigned as LumA by AIMS but 17% (26/156) and 16% (25/156) of LumA considered LumB or normal-like by AIMS, respectively. Progression-free survival (PFS) by ruoProsigna-derived subtype in PALOMA-2 showed that PAL+LET benefited all pts but those with a basal-like subtype (Table 2). With AIMS, PAL+LET provided a PFS benefit in pts with LumA and LumB subtypes, but was less effective in the HER2-E subtype. Conclusion Intrinsic subtyping has potential clinical utility. PAL+ET should be considered for ER+/HER2- ABC, except possibly in pts with a basal-like tumor, consistent with previous reports. A standardized clinical PAM50 assay and bioinformatics approach should be used as discrepancies in gene expression platforms and algorithms lead to different results and could misguide treatment decisions. Clinical trial identification: Pfizer (NCT01740427)

Citation Format: Maggie Cheang, Mitch Dowsett, Mothaffar Rimawi, Stephen Johnston, Samuel Jacobs, Judith Bliss, Katherine Pogue-Geile, Lucy Kilburn, Zhou Zhu, Eugene F. Schuster, Hui Xiao, Lisa Swaim, Shibing Deng, Dongrui R. Lu, Eric Gauthier, Jennifer Tursi, Dennis J. Slamon, Hope S. Rugo, Richard S. Finn, Yuan Liu. Impact of using cross-platform gene expression profiling technologies and computational methods for intrinsic breast cancer subtyping in PALOMA-2 and PALLET [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr PD2-07.