Cancer cells induce proliferation and local accumulation of immunosuppressive cells, such as FOXP3-positive cells known as regulatory T cells (Tregs), leading to tumor-induced immune tolerance. Although cancer chemotherapy is usually considered immunosuppressive, some chemotherapeutic agents activate an anticancer immune response. Therefore, we postulated that the number of tumor-infiltrating FOXP3-positive cells during primary systemic chemotherapy (PSC) correlates with therapeutic outcomes in patients with breast cancer. Between September 2000 and January 2005, we examined 93 patients with breast cancer diagnosed by core-needle biopsy and treated with PSC. Core-needle biopsy (CNB) and surgical resected specimens were stained with a FOXP3 mouse monoclonal antibody to compare the numbers of FOXP3-positive cells in the tumors before and after PSC. A median cut-off value of >16.3/high power field (HPF) and >6.6/HPF defined high numbers of Tregs in CNB and in surgical specimens, respectively. We then assigned the patients into 4 groups (HH, high number of FOXP3-positive cells in both CNB and surgical specimen; LL, low number in both specimens; HL, high in CNB and low in the surgical specimen; LH, low in CNB and high in surgical specimen). Lymph vessel invasion-positive, clinically non-responder and ER-negative tumors contained significantly more FOXP3-positive cells after PSC (p=0.04, p=0.03 and p=0.04, respectively). Prognosis was better among patients with low numbers than high numbers of FOXP3-positive cells both in CNB and in surgically resected specimens. In multivariate analysis, LL group demonstrated significantly better recurrence-free survival with risk ratio of 5.81 (95%CI, 1.09-107.5; p=0.04) rather than that of non-LL group (LH, HL and HH). These findings suggest that the number of FOXP3-positive cells identified during PSC represents a promising predictive factor that might also be an important therapeutic target for breast cancer.