This study was designed to examine the relationship between different methodologies for response evaluation and long-term survival estimation in patients underwent neoadjuvant chemotherapy (NCT) for breast cancer. We retrospectively analyzed 569 patients who were diagnosed with LABC and received NCT followed by breast and axilla surgery. The RECIST 1.1 criteria and Miller-Payne (MP) grading scale were used to evaluate patient responses to NCT. Univariate and multivariate survival analyses were performed to investigate the correlation between treatment response and long-term patient survival. Clinical response (RFS [P<0.001]; OS [P=0.003]), pathological response evaluated by pCR (RFS [P<0.001]; OS [P<0.001]), and MP grade (RFS [P<0.001]; OS [P<0.001]) were significant predictors of risks of relapse and survival. However, in hormone receptor-positive (ER and/or PR+) subtypes, the clinical response (P=0.004 for Luminal-A and P=0.038 for Luminal-B) and MP grade (P=0.002 for Luminal-A and P<0.001 for Luminal-B) significantly predicted RFS independently according to multivariate Cox regression model. MP grade (P=0.015 for Luminal-A and P=0.009 for Luminal-B) also was an independent predictor of patients' OS. However, these two methods failed to predict patient survival in hormone receptor-negative (ER and PR-) subtypes. Our findings indicate that the value of response evaluation methods varies for different breast cancer subtypes. Conceiving of further prospective approaches for new individualized response-evaluation models are needed in the neoadjuvant setting.