Metronidazole susceptibility testing by E test was compared to that by disk diffusion for 263 Helicobacter pylori isolates and to that by breakpoint agar dilution for 90 H. pylori isolates. In 5% and 6% of the cases, respectively, results were discrepant. For each of 52 clinical isolates an E test was performed on 10 separate colonies. Subpopulations of resistant and susceptible bacteria were found in five cases. From three isolates, each colony was subcultured and tested up to 10 times. All but 1 of 292 tests showed the same result. We conclude that the E test is reliable and that subpopulations are responsible for discordant results.