The present trial was designed to determine the efficacy of the combination of gemcitabine/doxorubicin/paclitaxel (GAT) delivered every other week as first-line therapy in patients with metastatic breast cancer. From February 1998 to September 1999, 41 patients were included in this trial. Doses delivered were doxorubicin 30 mg/m2 on day 1 and paclitaxel 135 mg/m2 plus gemcitabine 2500 mg/m2 both given on day 2, every 14 days. Doses were selected from a previous phase I trial conducted at our institution. Eligibility criteria for the phase II trial included histologically confirmed metastatic breast cancer with bidimensionally measurable lesions; no prior therapy for metastatic disease; adjuvant or neoadjuvant chemotherapy was allowed if given more than 1 year before and cumulative doses of doxorubicin or epirubicin were less than 200 mg/m2 or 360 mg/m2, respectively; Eastern Cooperative Oncology Group (ECOG) performance status of 2 or less; and adequate hematological, hepatic, and renal function. Prophylactic use of granulocyte colony-simulating factor (G-CSF) was allowed if patients were not fully recovered (absolute neutrophil count greater than 1500/microL) from chemotherapy administration before the next dose. Left ventricular ejection fraction was determined initially, at the end of the study, and every 6 months thereafter. The patients' median age was 55 years (range, 33-68 years), and their median ECOG performance status was 0 (range, 0-1). Twenty-eight patients had received adjuvant therapy, 17 with epirubicin (none with doxorubicin). Metastases were present in the bone (19 patients), lung (19 patients), liver (11 patients), and soft tissues (18 patients). Twenty patients had one metastatic site and 21 had two or more sites. Efficacy was assessed on an intent-to-treat basis. A total of 216 cycles of GAT were given. Twenty-two percent of the courses were delayed or given at reduced doses mostly due to neutropenia or thrombocytopenia. G-CSF was required in 58% of the cycles. Grade 3/4 neutropenia was the main toxicity and appeared in 17 patients, one of whom had an episode of febrile neutropenia. Nonhematological toxicities consisted mainly of neurotoxicity and myalgias. A drop of 10%-20% in the left ventricular ejection fraction was detected in two patients and another patient had a decrease greater than 20%, although none developed symptoms of heart failure. Overall response rate was 80.4% (95% confidence interval: 68.3-92.5), with 15 patients (36.6%) achieving a complete response. Median survival time was 27 months and median time to progression was 15 months. The GAT combination is feasible and very active in patients with metastatic breast cancer, with an encouraging response rate including a high rate of complete responses. No congestive heart failure was documented and other toxicities were mild, with the exception of neutropenia.