Background: The PARP inhibitor veliparib in combination with carboplatin (VC) was one of the experimental regimens evaluated in the phase 2 neoadjuvant I-SPY 2 standing trial for high risk breast cancer patients. VC graduated in the triple negative (TN) signature. However, not all TN patients achieved pathologic complete response (pCR) and some HR+HER2- patients responded. The I-SPY 2 biomarker component provides a platform for rigorous evaluation of mechanism-of-action-based markers in the context of established biomarkers (HR, HER2, MammaPrint) within the trial. Here, we report results from 5 investigator-submitted biomarker proposals and the MammaPrint High1/High 2 (MP1/2) classification as specific predictors of VC response.

Methods: Data from 116 HER2- patients (VC: 72 and concurrent controls: 44) were available. BRCA1/2 germline mutation was assessed by Myriad Genetics. 3 expression signatures relating to DNA damage repair deficiency (PARPi-7, BRCAness and CIN70) and MP1/2 classification were evaluated on Agilent 44K arrays. PARP1 levels were measured using reverse phase protein arrays. We used logistic modeling to assess biomarker performance. A biomarker is considered a specific predictor of VC response if it associates with response in the V/C arm but not the control arm, and if the biomarker x treatment interaction is significant (likelihood ratio test, p<0.05). In an exploratory analysis, we combined successful biomarkers to refine the 'predicted sensitive' group and used Bayesian modeling to estimate the pCR rates of 'predicted sensitive' TN and HR+HER2- patients in each arm.

Results: BRCA1/2 germline mutation status associates with VC response, but its low prevalence in the control arm precludes further evaluation. Of the biomarkers evaluated, three (PARPi-7, BRCAness, and MP1/2) associate with response in the VC arm but not the control arm, and have biomarker x treatment interactions with p < 0.05 that retains significance upon adjusting for HR status. These signatures are only moderately correlated. When we combined the PARPi-7 and MP1/2 classifications using a simple voting scheme, the 40% of TN patients who are PARPi7-high and MP2 have an estimated pCR rate of 79% in the VC arm. In contrast, TN patients negative for at least one of these signatures (PARPi7-low and/or MP1) only have an estimated pCR rate of 35%. Only 9% of HR+/HER2- patients are PARPi7-high and MP2; but these patients also appear more responsive to VC with estimated pCR rates of 49%, compared to 13% in 'biomarker-negative' HR+HER2- patients.

Conclusion: If verified in a larger trial, PARPi7, BRCAness and MP1/2 signatures may help refine predictions of VC response, thereby improving patient care. Evaluation of the combined signature for patients treated with platinum-based chemotherapy is ongoing.