PURPOSE To prospectively study the value of PET-CT with fluorine-18 fluorodeoxyglucose (FDG) to predict neoadjuvant chemotherapy (NAC) response of locoregional disease of stages II and III breast cancer patients. MATERIAL AND METHODS A written informed consent and approval were obtained from the Ethics Committee. PET-CT accuracy in the prediction of pathologic complete response (pCR) after NAC was studied in primary tumors and lymph node metastasis in 43 women (mean age: 50 years: range: 27-71 years) with histologically proven breast cancer between December 2009 and January 2011. PET-CT was performed at baseline and after NAC. SUV(max) percentage changes (ΔSUV(max)) were compared with pathology findings at surgery. Receiver-operator characteristic (ROC) analysis was used to discriminate between locoregional pCR and non-pCR. In patients not achieving pCR, it was investigated if ΔSUV(max) could accurately identify the residual cancer burden (RCB) classes: RCB-I (minimal residual disease (MRD)), RCB-II (moderate RD), and RCB-III (extensive RD). RESULTS pCR was obtained in 11 patients (25.6%). Residual disease was found in 32 patients (74.4%): 16 (37.2%) RCB-I, 15 (35.6%) RCB-II and 2 (4.7%) RCB-III. Sensitivity, specificity, and accuracy to predict pCR were 90.9%, 90.6%, and 90.7%, respectively. Specificity was 94.1% in the identification of a subset of patients who had either pCR or MRD. CONCLUSION Accuracy of ΔSUV(max) in the locoregional disease of stages II and III breast cancer patients after NAC is high for the identification of pCR cases. Its specificity is potentially sufficient to identify a subgroup of patients who could be managed with conservative surgery.