223 inflammatory breast cancer patients were diagnosed at the Institut Curie between 1977 and 1987. Patients received chemotherapy and radiation treatment according to three consecutive randomised trials. Five- and 10- year survival rates were 41 and 32%, respectively. Disease-free interval rates were 25.5% at 5 years and 19% at 10 years. Parameters significantly linked with a pejorative prognosis in a multivariate analysis were: diffuse erythema, lymph node involvement, chest wall adherence, and age above 50 years. When therapeutic response parameters were included in the multivariate analysis, the five most important prognostic factors in order of significance were complete tumour regression after completion of induction treatment (at 8 months), complete regression of inflammatory symptoms after 3 months of neoadjuvant chemotherapy, limited erythema at presentation and, less significantly, complete regression of inflammatory symptoms at 8 months and tumour regression at 3 months. In conclusion, patients who achieved a rapid and complete remission had a better prognosis than patients who had an incomplete response to chemotherapy. High-dose chemotherapy and reversal or prevention of drug resistance will be evaluated in future trials. Detailed information on the biology of this disease should allow the design of new strategies aiming to improve patient management.