We designed a multicenter study to compare tobramycin/clindamycin to imipenem/cilastatin for intra-abdominal infections. We included the Acute Physiology and Chronic Health Evaluation (APACHE II) index of severity and excluded patients without established infection. Two hundred ninety patients were enrolled, of whom 162 were evaluable. Using logistic regression to analyze both outcome at the abdominal site of infection and outcome as mortality, we found a significant correlation for both with APACHE II score (p less than 0.0001 for both). Next we analyzed the residual effect of treatment assignment and found a significant improvement in outcome for imipenem/cilastatin-treated patients (p = 0.043). The differences in outcome were explained by a higher failure rate for patients with gram-negative organisms for tobramycin/clindamycin-treated patients (p = 0.018). This was reflected in a significantly higher incidence of fasciitis requiring reoperation and prosthetic fascial replacement. Maximum peak tobramycin levels were analyzed for 63 tobramycin/clindamycin patients harboring gram-negative organisms. For failures the maximum peak was 6.4 +/- 1.9 micrograms/mL, and time to maximum peak was 4.6 +/- 5.2 days. For successes the maximum peak was 6.1 +/- 1.7 micrograms/mL, occurring at 3.8 +/- 2.6 days. This study supports inclusion of severity scoring in statistical analyses of outcome results and supports the notion that imipenem/cilastatin therapy improves outcome at the intra-abdominal site of infection as compared to a conventionally prescribed amino-glycoside-based regimen.