OBJECTIVE To test the hypothesis that comprehensive broad-spectrum empirical antimicrobial therapy is superior to limited-spectrum empirical antimicrobial therapy in intra-abdominal infections. METHODS Prospective, randomized, double-blinded study. METHODS University-affiliated hospitals in Canada. METHODS Two hundred thirteen patients with intra-abdominal infections and planned operative or percutaneous drainage. METHODS Limited-spectrum empirical antimicrobial therapy consisted of cefoxitin sodium, 2 g, intravenously, every 6 hours (n = 109). Comprehensive broad-spectrum empirical antimicrobial therapy consisted of a combination of imipenem and cilastatin sodium, 500 mg, intravenously, every 6 hours (n = 104). METHODS Failure to cure the intra-abdominal infection (persistence of infection or death). RESULTS Of initial isolates, 98% were sensitive to imipenem plus cilastin sodium compared with 72% for cefoxitin. No difference was found in the failure rate between treatment groups. Among various reasons for failure (including technical), 12 of 80 patients in the limited-spectrum empirical antimicrobial therapy group had resistant organisms at a second intervention compared with 1 of 74 in the comprehensive broad-spectrum empirical antimicrobial therapy group (P < .003, chi 2). One death in the limited-spectrum empirical antimicrobial therapy group was due to autopsy-proved disseminated Pseudomonas aeruginosa (blood, peritoneum, lung, and pleural fluid) that was resistant to cefoxitin, and the other was associated with peritonitis due to cefoxitin-resistant Enterobacter cloacae. One death in the comprehensive broad-spectrum empirical antimicrobial therapy group was associated with peritonitis from Clostridium perfringens that was sensitive to imipenem plus cilastin sodium, and the other was associated with peritonitis from Pseudomonas aeruginosa that was resistant to imipenem plus cilastin sodium. CONCLUSIONS Treatment failure of intra-abdominal infection may be due, in part, to the presence of resistant pathogens at the site of infection. Therefore, routine culture of these sites seems worthwhile and empirical therapy should be as comprehensive as possible and should cover all potential pathogens.