We commend Nomura and coworkers [1] for using an existing cohort to increase our understanding of the relation between Helicobacter pylori and peptic ulcer disease. However, the nested case–control design that was used has a few methodologic problems. First, the case–control design was used in the context of a cohort design, and principles that apply to the cohort design should be respected. Specifically, the disease status (duodenal or gastric ulcer) at study entry is unknown. Consequently, the population at risk may not have been free of disease when the blood samples were collected. If persons in whom peptic ulcer disease develops are more likely to have a history of peptic ulcer disease, even if asymptomatic, they will be more likely to have antibodies to H. pylori when blood samples are measured at baseline. Therefore, the reported association will be overestimated. Second, the authors excluded the controls who developed peptic ulcer disease. In such a design, controls who develop disease should be included as case-patients. By excluding them, the reported association is again overestimated because several case-patients who probably did not have antibodies will not be counted. At the very least, the antibody status of these persons should be reported. Because of these methodologic issues, the reported positive association may be overestimated, and persons who develop peptic ulcer disease may not be more likely to have had previous high levels of antibodies to H. pylori. Nested case–control studies are an efficient way to answer questions of cause and effect from previously collected data. The many design issues should be handled with care to prevent erroneous conclusions.