Comparison of two nomograms to predict pathologic complete responses to neoadjuvant chemotherapy for breast cancer: evidence that HER2-positive tumors need specific predictors.

The aim of this study is to compare two published nomograms, the "Institut Gustave Roussy/M.D. Anderson Cancer Center" (IGR/MDACC) and the Colleoni nomograms, in predicting pathologic complete responses (pCR) to preoperative chemotherapy in an independent cohort and to assess the impact of HER2 status. Data from 200 patients with breast carcinoma treated with preoperative chemotherapy were collected. We calculated pCR rate predictions with the two nomograms and compared the predictions with the outcomes. Sixty percent of the patients with HER2-positive tumors received trastuzumab concomitantly with taxanes. Model performances were quantified with respect to discrimination and calibration. In the whole population, the area under the ROC curve (AUC) for the IGR/MDACC nomogram and the Colleoni nomogram were 0.74 and 0.75, respectively. Both of them underestimated the pCR rate (P = 0.026 and 0.0005). When patients treated with trastuzumab were excluded, the AUC were excellent: 0.78 for both nomograms with no significant difference between the predicted and the observed pCR (P = 0.14 and 0.15). When the specific population treated with trastuzumab preoperatively was analyzed, the AUC for the IGR/MDACC nomogram and the Colleoni nomogram were poor, 0.52 and 0.53, respectively. The IGR/MDACC and the Colleoni nomograms were accurate in predicting the probability of pCR after preoperative chemotherapy in the HER2-negative population but did not correctly predict pCR in the HER2-positive patients who received trastuzumab. This suggests that responses to preoperative chemotherapy, including trastuzumab, are biologically driven and that a specific nomogram or predictor for HER2-positive patients has to be developed.