Rank-based predictors for response and prognosis of neoadjuvant taxane-anthracycline-based chemotherapy in breast cancer

For neoadjuvant taxane and anthracycline-based chemotherapy for breast cancer, patients with pathological complete response (pCR) have a favorable prognosis compared with patients with residual disease (RD). Although a number of pCR predictors based on microarray profiles have been proposed to guide neoadjuvant chemotherapy, most of these have not been independently validated in inter-laboratory datasets, possibly owing to the fact that microarray measurements are sensitive to experimental batch effects and inter-array data normalizations. In this study, we developed a rank-based method to tackle this difficulty. First, we extracted from two datasets a combination of gene pairs, each of which had opposing relative expression orders in patients with pCR and those with RD, and used these to build a pCR predictor. This pCR predictor was found to have sensitivities of 74 and 86 % and specificities of 71 and 68 % in another two independent datasets from multiple laboratories, and these results were better than the performances of three previously reported predictors. Considering that patients with minimal RD also tend to have a good prognosis, we then developed a prognosis predictor for RD as a complement to the pCR predictor, in order to identify a group of patients likely to have a good prognosis, taking into account both the RD level and the intrinsic risk factors. In the independent validation, there was a significant difference ( P = 0.001) in distant relapse-free survival between the patients likely to and those not likely to have a good prognosis according to our prognosis predictor. In conclusion, the rank-based predictors for response and prognosis can accurately and robustly predict patients with improved prognosis who might benefit from neoadjuvant taxane and anthracycline-based chemotherapy.