The percentage of participants with disease-free survival at 3 and 5 years. Disease-free survival (DFS) is measured from the time between randomization and the date of first documented disease recurrence, or death from any cause.

The pCR was defined as no histologic evidence of residual invasive adenocarcinoma in the breast and axillary lymph nodes, with or without the presence of ductal carcinoma in situ (DCIS) in the breast.

Beta III tubulin positivity determined by cross-validation method. Optimal cutoff: ≥46% tumor cells staining at 2 plus or 3 plus intensity (corresponding Beta III tubulin positivity=39.4%). Pre-specified cutoff of Beta III tubulin positivity: ≥50% 2plus or 3plus cells (corresponding prevalence=38.5%). Optimal cutoffs for TACC3 and CAPG positivity determined by cross-validation method: 6.889 and 6.844 \[log2 normalized intensity units\], respectively (corresponding to prevalence rates of 43.3% and 44.3%).

For each of the 2 biomarker sets (20-gene or 26-gene), a multi-gene model was built using penalized logistic regression on all pharmacogenomic evaluable subjects for each treatment arm separately. Receiver Operating Characteristic (ROC) plots for separate arm using 5 fold cross validation were generated. ROC for separate arms using cross over were also added. Further analysis on the multiple gene models (as mentioned in the SAP) was planned only based on the initial findings from the 2 ROC plots. For 20- and 26-gene models, ROC curves generated for each study arm did not indicate that these multi-gene models differentially predicted for pCR between the treatment arms, so further analyses to estimate the optimal cut-off and the pCR rates were not conducted.