We initiated a phase II study to determine the efficacy of carboplatin and paclitaxel (Taxol; Bristol-Myers Squibb Company, Princeton, NJ) in patients with advanced esophageal or gastric cancer. Objective tumor responses, duration of response, time to disease progression, overall survival, and toxicity profile are the end points evaluated in this study. The study includes patients with locally advanced metastatic or recurrent esophageal or gastric cancer with no history of prior chemotherapy or radiation treatment. Patients are required to have a performance status of 0 to 2 and no organ failure. Staging diagnostics include computed tomography scans and endoscopic ultrasound when appropriate. Starting dose of paclitaxel is 200 mg/m2, and carboplatin is given to achieve an area under the concentration-time curve of 5.0. Dose escalation is attempted if nadir counts permit. Seventeen patients have registered so far. Of those, three are not evaluable for efficacy analysis: one has not been on study long enough to be assessed for response and the other two are not assessable for objective tumor response. Three patients with gastric cancer have achieved partial responses. Two of five patients with esophageal cancer responded (partial response) to this treatment. Dose escalation by one step was feasible in approximately half the patients. There were no episodes of neutropenic fever. Carboplatin and paclitaxel given on an outpatient basis is well tolerated and an interesting regimen for the treatment of patients with advanced gastric or esophageal cancer.