The optimal method for the determination of the minimum inhibitory concentration (MIC) of antimicrobials against Helicobacter pylori has not been established. The epsilometer agar diffusion gradient test (E-Test; AB Biodisk, Solna, Sweden) was compared with broth microdilution, the reference method, and disk diffusion for the antimicrobial susceptibility testing of 122 clinical isolates of H. pylori to ampicillin, clarithromycin, and metronidazole. Isolates were considered to be resistant when the MIC values was > 8 micrograms/ml for either ampicillin or metronidazole and > 2 micrograms/ml for clarithromycin. For an individual isolate, the MICs for ampicillin and clarithromycin determined by broth microdilution and the E-test were highly reproducible, with replicate results being within +/- 1 log2 dilution. The correlation between the MICs determined by E-test and broth microdilution was excellent for both ampicillin and clarithromycin (90.1% and 88.5% were within +/- log2 dilution, and 98.3% and 96.7% of the values were within +/- 2 log2 dilution, respectively). In no instance did the interpretation of "sensitive" or "resistant" differ. Conversely, only 70.5% of the E-test results of metronidazole were within +/- 1 log2 dilution of the broth microdilution results. In addition, 15 (12.3%) of the H. pylori isolates interpreted as resistant by the E-test were sensitive by the broth microdilution method. All discrepancies occurred when the E-test MIC values fell between 8 and 32 micrograms/ml. The results of the ampicillin and clarithromycin disk diffusion assay correlated 100% with the results of the broth microdilution. However, these data suggest that when the E-test MIC results of metronidazole yield values between 8 and 32 micrograms/ml, the MIC should be reevaluated by another method.