Mezlocillin was used as an initial empiric antibiotic therapy for febrile (> 101 degrees F, ca. 38.33 degrees C) granulocytopenic (< 1,000/microliter) cancer patients. Patients known to be colonized with an organism resistant to 100 micrograms of mezlocillin per mol were excluded. The initial 25 cases (23 patients) received intravenous mezlocillin, 260 mg per kg per day in six divided doses; the mean 1-h-postinfusion serum level was 69 micrograms/ml. Because of the low serum level, the next 25 cases (22 patients) received 450 mg/kg per day, also in six divided doses, resulting in a mean 1-h-postinfusion serum level of 161 micrograms/ml. Both dosage regimens provided similar efficacy. Combined results show that 11 of 21 microbiologically documented infections and 7 of 13 clinically documented infections improved. Instances of bacteremia (number of cases in parentheses) were caused by Pseudomonas aeruginosa (two), Staphylococcus epidermidis (two), Clostridia perfringens (one), and Bacillus species (one); only one case improved. A rise in granulocyte count to > 500/microliters, a serum bactericidal activity of greater than or equal to 1:8 against the infecting pathogen, or both were indicators of a good therapeutic response. Despite exclusion of patients known to be previously colonized with mezlocillin-resistant organisms, 7 of 23 pathogens required a minimal concentration of greater than or equal to 100 micrograms of mezlocillin per ml for inhibition. In addition, surveillance cultures from 18 cases showed resistant organisms colonizing the gingiva, rectum, or both. Side effects of mezlocillin were minimal and included pseudoproteinuria, asymptomatic transient rise in bilirubin, and easily reversible kypokalemia. Mezlocillin, a new semisynthetic penicillin with little toxicity, was found to be inadequate as a single-agent empiric antibiotic therapy for febrile, granulocytopenic cancer patients.