In the I-SPY 2 TRIAL (NCT01042379), HER2− patients were randomized to receive standard chemotherapy (paclitaxel followed by doxorubicin/cyclophosphamide; T−>AC) or the oral PARP inhibitor veliparib in combination with carboplatin and chemotherapy (VC+T−>AC).23 (link) Pre-treatment samples from patients in the veliparib/carboplatin (VC) and concurrent control arms of the I-SPY 2 trial23 (link) were profiled using Agilent 44 K expression arrays (Agendia, Inc) and reverse phase protein arrays (RPPA). Details of the sample preparation and data processing are provided in the Supplemental Methods . In addition, BRCA1/2 mutation status was assessed (Myriad Genetics Laboratories).
We evaluated six biomarkers as specific predictors of VC response using a pre-specified QBE methodology. Briefly, we follow a three-step methodology, first evaluating the relative performance of the biomarker between arms, followed by assessing biomarker performance in the context of the graduating (TN) signature. Finally, we perform Bayesian analysis to estimate pCR rates in the arms and the predictive probability of VC showing superiority to control in biomarker defined subsets. The six biomarkers evaluated are: (1) BRCA germline mutation; (2) a 7-gene DNA-repair deficiency expression signature (PARPi-7)34 ; (3) a 77-gene BRCA1ness expression signature;40 , 41 (4) the CIN70 chromosomal instability expression signature;42 (link), 43 (link) (5) PARP1 protein levels; and (6) MP1/2 status. Details of the definition and scoring of each biomarker and our evaluation methodology are available in theSupplemental Methods . Each individual biomarker was evaluated separately; and these analyses did not account for multiplicities outside of the study.
In an exploratory analysis, we evaluate the concordance between successful biomarkers using the Kappa statistic. We use a simple voting method to combine the two most successful VC-sensitivity biomarkers and use Bayesian modeling to estimate the pCR rates and predictive probability of phase 3 trial success of biomarker-positive TN and HR+HER2− patients.
We evaluated six biomarkers as specific predictors of VC response using a pre-specified QBE methodology. Briefly, we follow a three-step methodology, first evaluating the relative performance of the biomarker between arms, followed by assessing biomarker performance in the context of the graduating (TN) signature. Finally, we perform Bayesian analysis to estimate pCR rates in the arms and the predictive probability of VC showing superiority to control in biomarker defined subsets. The six biomarkers evaluated are: (1) BRCA germline mutation; (2) a 7-gene DNA-repair deficiency expression signature (PARPi-7)34 ; (3) a 77-gene BRCA1ness expression signature;40 , 41 (4) the CIN70 chromosomal instability expression signature;42 (link), 43 (link) (5) PARP1 protein levels; and (6) MP1/2 status. Details of the definition and scoring of each biomarker and our evaluation methodology are available in the
In an exploratory analysis, we evaluate the concordance between successful biomarkers using the Kappa statistic. We use a simple voting method to combine the two most successful VC-sensitivity biomarkers and use Bayesian modeling to estimate the pCR rates and predictive probability of phase 3 trial success of biomarker-positive TN and HR+HER2− patients.
Free full text: Click here
