To determine the incidence and the prognostic value of ipsilateral breast tumor recurrence (IBTR) in patients treated with primary chemotherapy and breast-conserving surgery. Between January 1985 and December 1994, 257 patients with invasive T1 to T3 breast carcinoma were treated with primary chemotherapy, lumpectomy, and radiation therapy. The median follow-up time was 93 months. To evaluate the role of IBTR in metastase-free survival, a Cox regression multivariate analysis was performed using IBTR as a time-dependent covariate. The IBTR rates were 16% (+/- 2.4%) at 5 years and 21.5% (+/- 3.2%) at 10 years. Multivariate analysis showed that the probability of local control was decreased by the following independent factors: age < or = 40 years, excision margin < or = 2 mm, S-phase fraction more than 4%, and clinical tumor size more than 2 cm at the time of surgery. In patients with excision margins of more than 2 mm, the IBTR rates were 12.7% at 5 years and 17% at 10 years. Nodal status, age < or = 40 years, and negative estrogen receptor status were predictors of distant disease in the Cox multivariate model with fixed covariates. The contribution of IBTR was highly significant (relative risk = 5.34) when added to the model, whereas age < or = 40 years was no longer significant. After IBTR, 31.4% (+/- 7.0%) of patients developed metastases at 2 years and 59.7% (+/- 8.1%) at 5 years. Skin involvement, size at initial surgery, and estrogen receptor status were predictors of metastases after IBTR. IBTR is a strong predictor for distant metastases. There are implications for conservative surgery after downstaging of the tumor and therapy at the time of IBTR.