The correlation between various ampicillin/sulbactam in vitro antimicrobial susceptibility test results and the clinical outcome of patients treated with this agent have been examined. A survey of over 29,000 clinical isolates of the family Enterobacteriaceae found that the proportion of susceptible pathogens as assessed by current susceptibility testing interpretive guidelines (NCCLS) for disk diffusion and dilution (MIC) assays was significantly less than the proportion of patients cured or clinically improved in ampicillin/sulbactam clinical trials. Also, the results of two NCCLS methods differ greatly in the perceived percentages of susceptible strains (63.9% versus 72.2%; unacceptable variation). Furthermore, the current interpretive criteria resulted in high false-susceptible (4.2%) and total (19.7%) error rates. When proposed interpretive guidelines were applied, approximately 73 to 87% of the Enterobacteriaceae strains were observed to be susceptible, the variation between methods was minimized, and the error rates were reduced. A retrospective analysis of data from clinical trials with ampicillin/sulbactam indicated that the proportion of patients who were cured or clinically improved and bacterially eradicated was not appreciably different in patients having baseline Enterobacteriaceae pathogens with MICs of 16 or 32 micrograms/ml (ampicillin MIC component) as compared to those with pathogens having MICs of < or = 8 micrograms/ml. Studies in animals, in vitro models, and pharmacokinetic considerations indicate that a change in the MIC breakpoint for ampicillin/sulbactam should be considered. The proposed interpretive guideline revisions for ampicillin/sulbactam susceptibility testing of the Enterobacteriaceae were 1) use current diagnostic reagents with criteria of < or = 16/8 micrograms/ml (> or = 14 mm) as susceptible and > or = 64/32 micrograms/ml (< or = 10 mm) as resistant; e.g., 75.9 to 76.0% spectrum and 1.3% false-susceptible error; 2) use alternative diagnostic reagents (1:1 ratio MIC; 20/20 micrograms disks) with criteria of < or = 8/8 micrograms/ml (> or = 18 mm) as susceptible and > or = 32/32 micrograms/ml (< or = 14 mm) as resistant; e.g., 73.3 to 76.9% spectrum and 1.8% false-susceptible error; or 3) use alternative diagnostic reagents with criteria of < or = 16/16 micrograms/ml (> or = 14 mm) as susceptible and > or = 64/64 micrograms/ml (< or = 10 mm) as resistant; e.g., 84.7 to 86.9% spectrum and 1.3% false-susceptible error. Data from a comprehensive in vitro survey of clinical isolates, retrospective analyses of clinical trials, and studies of animal models support the modification of contemporary interpretive guidelines for ampicillin/sulbactam antimicrobial susceptibility tests. The best short-term criteria would apply current in vitro diagnostic reagents and a modified susceptible breakpoint (< or = 16/8 micrograms/ml as susceptible; option 1 above) until new diagnostic reagents can be qualified by means of studies needed for quality assurance of standardized methods (NCCLS M23-A and FDA procedures). These changes would provide a better in vitro prediction of ampicillin/sulbactam efficacy in clinical practice.