Lactulose H2 breath tests are widely used for quantifying carbohydrate malabsorption, but the validity of the commonly used technique (interval sampling of H2 concentrations) has not been systematically investigated. In eight healthy adults we studied the reproducibility of the technique and the accuracy with which 5 g and 20 g doses of lactulose could be calculated from the H2 excretion after their ingestion by means of a 10 g lactulose standard. The influence of different lengths of the test period, different definitions of the baseline and the significance of standard meals and peak H2 concentrations was also studied. Regardless of baseline definition, estimates of malabsorption were most precise, if areas under the H2 concentration v time curves for four hours or more from the start of the excess H2 excretion were used. The median deviations from the expected values were 20-30% (5-60%, interquartile range). This corresponded to the deviation in reproducibility of the standard dose. We suggest that individual estimates of carbohydrate malabsorption by means of H2 breath tests should be interpreted with caution if tests of reproducibility are not incorporated. Both areas under curves and peak H2 concentrations seem valid for comparison of groups.