The objective of this multicenter study was to compare the therapeutic index of two different doses of paclitaxel given as a 3-hour infusion in patients with metastatic breast cancer (MBC), who had failed to respond to previous chemotherapy. A total of 471 patients with MBC were randomized to receive intravenous paclitaxel at a dose of 175 or 135 mg/m2 every 3 weeks. Better treatment results were achieved with high-dose (HD) versus low-dose (LD) paclitaxel: overall response rate, 29% versus 22% (P = .108); complete response (CR) rate, 5% versus 2% (P = .088); median time to disease progression, 4.2 versus 3.0 months (P = .027); and median survival time, 11.7 versus 10.5 months (P = .321). Patients previously exposed or resistant to anthracyclines were as likely to respond as those without such prior exposure. Treatment was well tolerated, as documented by the number of administered treatment courses (median, six v five; range, one 17 v one to 18), the low frequency of dose reductions (14% v 7%, P = .024), and the small number of patients (n = 9 or 4% vn = 5 or 2%) who required treatment discontinuation for adverse reactions. The incidence and severity of neutropenia and peripheral neuropathy were dose-related. After quality-of-life-adjusted time-to-progression analysis, the HD arm (175 mg/m2) retained its advantage over the LD arm (135 mg/m2). The results of this trial substantiate the activity of paclitaxel in the treatment of MBC. The observed superior efficacy with a dose of 175 mg/m2 over 135 mg/m2 suggests a dose-effect relationship. The clinical activity in anthracycline-resistant patients is particularly noteworthy. Paclitaxel in breast cancer needs further evaluation in large trials that use combination chemotherapy and involve earlier disease stages.