Twenty-two patients with advanced breast cancer participated in a randomized cross-over study of one cycle each of doxorubicin followed 3 weeks later by mitozantrone or vice versa. Before further treatment, patients selected which drug they wished to continue. Of 18 patients completing the study, 13 chose to continue mitozantrone, 2 doxorubicin and 3 had no preference (P = 0.007). Patients were told to assume similar efficacy of the two drugs and drug preference was based primarily on side-effects. Patient self-assessment of quality of life and physician assigned toxicity scores both indicated that nausea and vomiting, appetite and alopecia were significantly worse following doxorubicin than after mitozantrone. Except for alopecia, no significant period or carry-over effects were noted although the power of the study to detect such interactions was low. This study design may prove useful in enabling patients to select their preference between two treatments of similar efficacy.