Research-based PAM50 predicts risk of relapse in residual disease after anti-HER2 therapies

Background: In the NeoSphere trial, we assessed research-based PAM50 subtype predictors at baseline in all patients (pts), and at surgery (SX) in pts with residual disease (RD) to investigate the changes after treatment and the association with pCR and distant event free survival (DEFS). Methods: 417 HER2+ pts were randomized to neoadjuvant TD, TPD, TP or PD (T=trastuzumab, P=pertuzumab, D=docetaxel), and received FAC/FEC and T after SX. 296 pts had RD. We assessed the intrinsic subtypes, ROR-P and ROR-S in 350 (83.9%) and 191 (64.5%) pts at baseline and SX respectively (166 paired samples). We evaluated the association with pCR and DEFS in the overall population and by ER status (161 (46.0%) ER+; 189 (54%) ER-). Results: The intrinsic subtype distribution was different by ER status. After adjustment for ER, no significant association with pCR was found. None of the baseline biomarkers was associated with DEFS in the overall population. In ER+, Her2-enriched (Her2E) subtype was associated with worst prognosis (HR 3.00 [1.10-8.22]; p = 0.032). Higher ROR-P and ROR-S were significantly associated with higher risk of recurrence in ER+ only (test for interaction p = 0.035 and p = 0.012, respectively). At SX, there was a significant increase in LumA and a decrease in Her2E and LumB subtypes. ROR scores also decreased. In ER+, post-treatment Her2E (HR 7.50 [1.86-30.1]; p = 0.044) and LumB (HR 3.12 [1.03-9.94]; p = 0.004) subtypes were associated with worst outcome than LumA. ROR-S and ROR-P were associated with outcome in ER+ tumors only (significant test for interaction). Compared to low-ROR-S group, high (HR 10.5 [2.79-39.3]; p = 0.0005) and median (HR 3.52 [1.11-11.4]; p = 0.032) groups had higher recurrence risk. Five year-DEFS was 94%, 78% and 50% in the low, median and high ROR-S groups, respectively. In paired samples, the post-treatment biomarker assessment was more informative than at baseline. Conclusions: In the NeoSphere study, PAM50 provides different prognostic information in ER+ and ER- tumors. Pre-treatment biomarkers were less informative than post-treatment, reflecting their dynamic modulation. In ER+ tumors with RD, PAM50-derived scores assessed on surgical samples might identify patients who need treatment escalation. This warrants independent validation. Legal entity responsible for the study: Fondazione Michelangelo. Funding: Has not received any funding. Disclosure: G. Bianchini: Consultant or advisory board member: Roche, EISAI, Pfizer, AstraZeneca, Novartis, MSD, Lilly, Amgen. C.M. Perou: Equity stock holder, consultant, and Board of Director Member: BioClassifier LLC; Listed an inventor on patent applications on the Breast PAM50 assay. A. Prat: Consultancy: Pfizer, Eli Lilly, Novartis, Nanostring Technologies Research funding Novartis, Nanostring Technologies; Scientific advisory board: Oncolytics Biotech. L. Gianni: Consultant or advisor board member: Roche, Pfizer, Boehringer Ingelheim, Celgene, Tahio Pharmaceutical, Synthon, AstraZeneca, Genomic Health, Merck Sharp & Dohme, Synaffix, Eli Lilly, Odonate Therapeutics, Sandoz, Onkaido, Oncolytics Biotech, ADC Therapeutics, Seattle Genetics. All other authors have declared no conflicts of interest.