OBJECTIVE Preclinical data suggest that the efficacy of paclitaxel (Taxol; Bristol-Myers Squibb Co, Princeton, NJ) is schedule-dependent. Schedule dependency is currently under investigation in ongoing randomized trials. METHODS Twelve patients with relapsed non-Hodgkin's lymphoma (NHL) refractory to a 3-hour infusion of 200 mg/m2 Taxol were crossed over to receive a 96-hour infusion of 140 mg/m2 Taxol every 3 weeks in an outpatient setting. Premedication with corticosteroids and antihistamines was not used. Patients who did not achieve at least a partial remission (PR) after two courses or whose disease progressed after one course were removed from the study. RESULTS All 12 patients were assessable for response. Eleven patients received at least two courses and one patient received one course of 96-hour Taxol infusion. None of the 12 patients crossed over to receive 96-hour Taxol infusion achieved a PR or complete response (CR). Eight patients (67%) developed progressive lymphoma, three (25%) had stable disease, and only one (8%) had a minor response. No major hypersensitivity reactions or life-threatening toxicities were observed. CONCLUSIONS Ninety-six-hour Taxol infusion does not produce significant responses in patients with NHL refractory to 3-hour Taxol infusion. Until the results of an ongoing multicenter trial comparing a 3- with a 96-hour infusion are published, the use of 96-hour Taxol infusion in NHL patients should be restricted to investigational programs. Because 48- to 72-hour infusions can produce higher plasma concentrations of Taxol than a 96-hour infusion, these schedules should be investigated to determine if they can induce better clinical responses.