The clinical efficacy and safety of ampicillin/sulbactam versus metronidazole-gentamicin were evaluated in a comparative, randomized, prospective study. Forty-four patients were enrolled: 22 received the ampicillin/sulbactam regimen, and 22 received the metronidazole-gentamicin combination. There were 33 cases of severe acute pelvic inflammatory disease, two tuboovarian abscesses, five cases of endomyometritis, and two cases of posthysterectomy pelvic cellulitis. Aerobic and anaerobic cultures from the infection sites yielded 447 microorganisms from 44 patients (an average of 10 bacteria per infection; 6.4 anaerobes and 3.7 aerobes). The most frequent isolates were Bacteroides sp., 54; Bacteroides bivius, 17; black-pigmented Bacteroides, 12; Bacteroides disiens, 11; Fusobacterium, 13; Peptostreptococcus anaerobius, 24; Peptostreptococcus asaccharolyticus, 21; anaerobic gram-positive cocci, 34; Gardnerella vaginalis, 29; Neisseria gonorrhoeae, 17; alpha-hemolytic streptococci, 15; and Escherichia coli, five. Clinical cure was noted in 19 of 20 patients treated with ampicillin/sulbactam and 18 of 21 patients treated with metronidazole-gentamicin. One treatment failure occurred in the ampicillin/sulbactam group in a patient who required antichlamydial therapy and had a complex left adnexal mass consistent with an abscess. The cases of metronidazole-gentamicin failure included two patients initially diagnosed as having tuboovarian abscesses who required a change in antibiotic therapy to control the infections. The third patient had postabortion endomyometritis that did not respond to metronidazole-gentamicin therapy within 48 hours, and required a change of medication. No adverse hematologic, renal, or hepatic effects were noted in either group of patients.