Hypothesis: Antibiotic regimens containing aminoglycosides result in a similar outcome compared with nonaminoglycoside regimens in the treatment of gram-negative infections in surgical patients. Design: An inception cohort study of hospitalized surgical patients from December 1, 1996, through September 30, 1998. Patients were observed from the time of diagnosis of infection to discharge. Setting: University hospital. Patients: Two hundred fifty-eight consecutive gram-negative infections occurring in general surgical and trauma patients and patients undergoing transplantation. Sixty- six patients received aminoglycosides as a component of their treatment regimen, whereas 192 received other agents. Results: Patients treated with aminoglycosides were younger (mean ± SEM age, 48 ± 2 vs 53 ± 1 years; P = .04 by univariate analysis) and had a similar APACHE II (Acute Physiology and Chronic Health Evaluation II) score (mean ± SEM, 17 ± 1 vs 15 ± 1; P = .10), yet had a significantly higher mortality vs patients treated with other agents (29% vs 14%; P = .02). A larger proportion of patients requiring hemodialysis were treated with aminoglycosides (33% vs 13%; P = .001). Although there was no difference in the sites of infection between groups, surgical patients with gram-negative pneumonia had a higher mortality when treated with aminoglycosides (37% vs 18%; P = .04), despite similar APACHE II scores (mean ± SEM, 20 ± 1 vs 18 ± 1; P = .40). Conclusions: Despite a younger age and similar severity of illness, patients with gram-negative infections treated with aminoglycosides were associated with a higher mortality rate, although this may be related to selection bias in the use of aminoglycoside agents. The mortality rate associated with gram-negative pneumonia was also higher in patients treated with aminoglycosides, despite a similar severity of illness. Future randomized studies are necessary to reanalyze the role of aminoglycosides in treating surgical patients with gram-negative infections, particularly pneumonia.