OBJECTIVE The aim of this study was to determine the numerical and categorical reproducibility of the 14C urea breath test in a routine clinical setting. BACKGROUND The 14C urea breath test is used for diagnosis of Helicobacter pylori infection and is regarded by some centers as the gold standard. METHODS In 140 referrals, duplicate tests were performed 1 wk apart. Eighteen patients were excluded because they had taken antibiotics or had changed medication that could influence the gastric urease activity between the tests. RESULTS The limits of agreement between the repeated tests were 45-234%. Hence, the numerical result of a repeated test was 95% likely to differ from the first with a factor of approximately two or less. However, when patients were classified as "positive" or "negative" according to our previously determined cut-off limit, only three patients were discordantly classified by the two tests. The observed agreement between the tests was 98%, and the chance-corrected proportional agreement (kappa) was 95%, which is far better than that reported for the histological diagnosis of gastric H. pylori infection. CONCLUSIONS The diagnostic reproducibility of the urea breath test was very good.