The longitudinal stability of the urea breath test (UBT), which measures urease as a biomarker for infection with Helicobacter pylori (a major risk factor for gastric cancer), was evaluated in the environs of Tsukuba, Japan. 13C-UBT measurements were monitored at four time points in 46 free-living, H. pylori-infected, asymptomatic volunteers over a period of 7 weeks. Subjects were asked to refrain from eating cruciferous vegetables, which might confound interpretation of results. Their compliance was monitored using both dietary records and direct biochemical testing of overnight urine. There was large between-subject UBT variation in this population (logUBT mean, 3.34; SD, 0.67). Within-subject (longitudinal) UBT values were remarkably stable in about one-quarter of the subjects (coefficients of variations for these individuals were <21%), whereas coefficients of variations in the highest quartile of variability ranged from 40% to 80%. About half of the sequential UBTs (63 of 138 such measurement pairs) changed >10 per thousand "delta over baseline" between measurements. This study provides the elements to optimize the design of a clinical trial in this population to examine the efficacy of a dietary intervention to reduce H. pylori infection. The number of subjects required to detect a 30% difference in average UBT value is highly dependent on the baseline stability of UBT measurements. For the least variable quartile, as few as 12 subjects would be needed; for the most variable quartile, at least 147 subjects would be required in each arm.