In a prospective, randomized, blinded study, 826 patients undergoing clean neurosurgical procedures received single intravenous doses of ceftizoxime (2 g) (n = 422) or a combination of vancomycin (1 g) and gentamicin (80 mg) (n = 404) 1 hour before an incision was made. Patients with infected or contaminated wounds and those receiving shunts or other implants were excluded. Primary wound infections occurred within 30 days in five patients in each group and were most common after spinal surgery and procedures through previous incisions. Secondary infections (pneumonias, urinary tract infections, and intravenous line-related bacteremia) occurred in 24 patients in the ceftizoxime group and 25 in the vancomycin/gentamicin group. The infection rates after transsphenoidal procedures (n = 129) were remarkably low in both groups. Ceftizoxime caused no adverse drug reactions, but six patients in the vancomycin/gentamicin group had clinically significant infusion-related hypotension or flushing. Placement of a temporary external drain, use of an operating microscope, preoperative steroids, and diabetes were not associated with increased infection rates. Analysis of routinely encountered ventricular cerebrospinal fluid and simultaneously obtained peripheral blood showed low but detectable levels of all three antibiotics within 2 hours; only ceftizoxime, however, achieved cerebrospinal fluid levels sufficient to inhibit the staphylococcus and Gram-negative bacilli most often associated with postneurosurgical infections. We conclude that ceftizoxime is as effective as vancomycin and gentamicin in neurosurgical prophylaxis but is less toxic and penetrates cerebrospinal fluid better.