Two hundred eleven patients with advanced breast cancer were randomized to receive either epirubicin (E) 50 mg/m2 and prednisolone (LEP) or E 100 mg/m2 and prednisolone (HEP). The intended treatment consisted of 16 courses of LEP or eight courses of HEP given at 3-weekly intervals. Reasons for stopping treatment early included progressive disease, stable disease without symptomatic improvement, or severe toxicity deemed intolerable by either the patient or physician. Toxicity was recorded at 3-weekly and response at 9-weekly intervals using the World Health Organization (WHO) criteria of response and toxicity. Two hundred nine patients were eligible for analysis, 98% of whom have been followed for more than a year. One hundred four patients received LEP and 105 HEP. Significantly worse myelosuppression, alopecia, nausea and vomiting, and mucositis were seen in the high-dose arm (P less than or equal to .001). More patients in the LEP arm stopped treatment before the fourth course than in the HEP arm, and the commonest reason for stopping was progressive disease. A similar median number of courses was given in each arm. There was a significantly higher response in the HEP arm (HEP - complete response [CR] + partial response [PR] = 41%, LEP - CR + PR = 23%). Despite this, no statistically significant differences was seen in overall survival or progression-free interval. The median survival for HEP and LEP was 44 and 46 weeks, respectively.