Patients with advanced breast cancer were randomised to 3M (mitozantrone 6.5 mg/m2q 21 days, methotrexate 30 mg/m2q 21 days, mitomycin C 6.5 mg/m2q 42 days) or 2M (as 3M but without mitomycin C). The objective response rates of 30% in 51 evaluable patients receiving 3M and 26% of 54 patients receiving 2M were not significantly different. 4/16 patients not responding to 2M responded to 3M on crossover. Both regimes were well tolerated but there was significantly less haematological toxicity and fewer dose reductions and delays with 2M. We conclude that patients should initially be treated with 2M and that non-responding patients should be crossed to 3M.