In an international, multicenter, open-label, randomized comparative study, adult patients in intensive care units were enrolled to receive cefpirome intravenously at 2 g twice daily or ceftazidime intravenously at 2 g three times daily for the empiric treatment of pneumonia. Randomization was performed after a double stratification according to the investigator's initial choice of monotherapy or combination therapy and then on the basis of the severity of disease. The primary endpoint was the clinical response at the end of treatment in the intent-to-treat population. Data for all patients were reviewed by a blinded observer. Of the 400 enrolled patients, 201 received cefpirome (monotherapy, 56%) and 199 received ceftazidime (monotherapy, 51%). Pneumonia was hospital acquired for 75% of the patients. Clinical failures rates were 34 versus 36% (odds ratio = 0.922; upper bound of 90% confidence interval = 1.301) in the intent-to-treat analysis for cefpirome and ceftazidime, respectively. For the cefpirome and ceftazidime groups, there were 35 versus 30% clinical failures among monotherapy-stratified patients, respectively, and 34 versus 42% clinical failures among combination therapy-stratified patients, respectively. The rates of clinical failures in the per-protocol analysis were 38 and 42%, respectively. In the population of patients evaluable for bacteriologic efficacy, eradication or presumed eradication was obtained for 71% (172 of 241) and 70% (162 of 230) of the pathogens isolated from the patients receiving cefpirome and ceftazidime, respectively. The mortality rates within 2 weeks after the end of treatment were similar (cefpirome group, 31%; ceftazidime group, 26%), as were the percentages of patients with at least one treatment-related adverse event (17 and 19%, respectively). An empiric treatment strategy with cefpirome at 2 g twice daily is equivalent in terms of efficacy and tolerance to ceftazidime at 2 g three times daily for the treatment of pneumonia in patients in intensive care units.