Between January 1977 and June 1983, 64 consecutive patients were treated for unilateral inflammatory nonmetastatic breast cancer. Our protocol included three or four courses of induction chemotherapy, then locoregional irradiation therapy with Co-60, followed by maintenance chemotherapy only if induction chemotherapy had proven effective. Eight patients with a residual tumor after radiotherapy underwent a modified radical mastectomy. Actuarial 3-year overall survival for the whole group was 38%, and the median disease-free survival time was 19 months. The effect of 17 factors on overall survival or disease-free survival was analyzed. With univariate analysis, eight factors were found to affect overall survival or disease-free survival: extent of initial erythema, size of initial edema, lymph node involvement, erythema present at the end of initial chemotherapy, erythema present at the end of radiotherapy, tumor size at the end of induction chemotherapy, residual breast tumor at the end of maintenance chemotherapy, and performance of a radical mastectomy. Age at diagnosis, menopausal status, type of chemotherapy, and date of appearance of inflammatory signs did not influence prognosis. Multivariate analysis using the Cox proportional hazard model isolated three bad prognosis factors: erythema involving the whole breast at initial diagnosis, erythema present at the end of initial chemotherapy, and lymph node involvement.