BACKGROUND The study was designed to evaluate the efficacy of paclitaxel in metastatic breast cancer patients. The design was motivated by a report from FDA and NCI staff proposing assessment of pre- and post-treatment symptoms as a means of evaluating treatment effectiveness [1]. METHODS Patients with symptomatic and/or measurable metastatic breast cancer with prior treatment received paclitaxel 210 mg/m2 as a 3 hour infusion every three weeks until toxicity or progression. A unique endpoint was subjective symptomatic response, defined as an improvement in the Symptom Distress Scale score by > or = 3 points at two successive evaluations before treatment failure. Patients were also evaluated for objective response and toxicity. RESULTS Of 135 patients registered, 123 were eligible and treated. The subjective symptomatic response rate for 93 symptomatic patients who completed forms was 40%, 95% confidence interval 29-51%. The objective response rate in 77 patients with measurable disease was 19%, 95% confidence interval 11-30%. In patients with both measurable and symptomatic disease, 37% had symptomatic and 13% had objective responses. Median times to treatment failure and death were 4 and 11 months, respectively. Toxicity was greater than anticipated: 12% discontinued treatment due to toxicity, 29% developed at least one Grade 3 neuromuscular toxicity, and two patients died of sepsis while neutropenic. CONCLUSIONS Paclitaxel by 3 hour infusion at a dose of 210 mg/m2 produced excessive neurotoxicity in patients with previously treated metastatic breast cancer. Both sustained subjective symptom reduction and objective responses were demonstrated, but dose reduction for routine practice is recommended.