In a randomized, evaluator-blind, multicenter trial, we compared cefepime (2 g three times a day) with imipenem-cilastatin (500 mg four times a day) for the treatment of nosocomial pneumonia in 281 intensive care unit patients from 13 centers in six European countries. Of 209 patients eligible for per-protocol analysis of efficacy, favorable clinical responses were achieved in 76 of 108 (70%) patients treated with cefepime and 75 of 101 (74%) patients treated with imipenem-cilastatin. The 95% confidence interval (CI) for the difference between these response rates (-16 to 8%) failed to exclude the predefined lower limit for noninferiority of -15%. In addition, therapy of pneumonia caused by an organism producing an extended-spectrum beta-lactamase (ESBL) failed in 4 of 13 patients in the cefepime group but in none of 10 patients in the imipenem group. However, the clinical efficacies of both treatments appeared to be similar in a secondary intent-to-treat analysis (95% CI for difference, -9 to 14%) and a multivariate analysis (95% CI for odds ratio, 0.47 to 1.75). Furthermore, the all-cause 30-day mortality rates were 28 of 108 (26%) patients in the cefepime group and 19 of 101 (19%) patients in the imipenem group (P = 0.25). Rates of documented or presumed microbiological eradication of the causative organism were similar with cefepime (61%) and imipenem-cilastatin (54%) (95% CI, -23 to 8%). Primary or secondary resistance of Pseudomonas aeruginosa was detected in 19% of the patients treated with cefepime and 44% of the patients treated with imipenem-cilastatin (P = 0.05). Adverse events were reported in 71 of 138 (51%) and 62 of 141 (44%) patients eligible for safety analysis in the cefepime and imipenem groups, respectively (P = 0.23). Although the primary end point for this study does not exclude the possibility that cefepime was inferior to imipenem, some secondary analyses showed that the two regimens had comparable clinical and microbiological efficacies. Cefepime appeared to be less active against organisms producing an ESBL, but primary and secondary resistance to imipenem was more common for P. aeruginosa. Selection of a single agent for therapy of nosocomial pneumonia should be guided by local resistance patterns.