Integration of molecular subtyping and immune profiling to predict pathologic complete response in HER2+ breast cancer treated with neoadjuvant dual HER2 therapy.

621 Background: Neoadjuvant outcomes are variable in human epidermal growth factor receptor 2-positive (HER2+) early breast cancer (EBC), highlighting the unmet need for predictive biomarkers. The BluePrint (BP) assay classifies tumors by molecular intrinsic subtype, whereas the ImPrint hormone receptor-positive (HR+) and ImPrint triple-negative genomic signatures predicted immune sensitivity in the ISPY2 trial. Because HER2+ EBCs are genomically heterogeneous, and because anti-HER2 has an immunologic mechanism of action, we hypothesized that BP and ImPrint could be used to predict outcomes (pathologic complete response, pCR) following neoadjuvant chemotherapy + anti-HER2 (trastuzumab + pertuzumab, HP). Methods: We evaluated n=252 HER2+ EBCs receiving neoadjuvant chemotherapy + HP in the prospective, observational FLEX study. Tumors were classified by BP subtype (HER2, Luminal A/B, or Basal) and ImPrint (+ or -). Fisher’s exact test was used to compare pCR rates across BP/ImPrint results. Multivariable logistic regression was used to evaluate independent associations with pCR, adjusting for nodal status and tumor size. Results: 72% of HER2+ EBCs were HR+ and 28% were HR–. In HR+/HER2+ EBC, 52% were classified by BP as non-HER2 (Luminal A 6%, Luminal B 44%, Basal 2%), whereas in HR-/HER2+ EBC, 14% were classified genomically as non-HER2, all of which were Basal. In the HR+/HER2+ subgroup, pCR rates differed significantly by BP subtype and ImPrint, with the highest pCR observed in BP-HER2/ImPrint+ tumors (Table). Similar significant differences were observed in HR-/HER2+ EBC. In a multivariable model of the HR+/HER2+ subgroup, BP and ImPrint independently predicted pCR after controlling for node status and tumor size (BP-HER2 odds ratio/OR: 8.23 [95% CI: 3.47–21.44]; p<0.001; ImPrint+ OR: 4.48 [95% CI: 1.48–14.89]; p<0.05). Conclusions: An integrated genomic approach that combines BP and ImPrint maximizes prediction of pCR outcome in this dataset, particularly for the HR+/HER2+ subgroup. Further evaluation is warranted, as these data could be useful for guiding clinical decision-making, including selection of chemotherapy backbone and adjuvant therapy. Importantly, this approach may help identify patients who are overtreated or undertreated with current strategies, including those with exceptional treatment sensitivity who may be candidates for chemotherapy de-escalation. Clinical trial information: NCT03053193 . pCR rates by genomic subtypes within HR+/HER2– EBC (Fisher’s exact test). Biomarker Subtype* N pCR rate (%) p-value BluePrint HER2 87 55 < 0.001 Luminal B 80 20 ImPrint Positive 30 60 0.006 Negative 151 32 BluePrint + ImPrint HER2, ImPrint+ 14 79 < 0.001 HER2, ImPrint- 73 51 Luminal B, ImPrint+ 14 43 Luminal B, ImPrint- 66 15 *Excluded Luminal A and Basal due to small N.