Our aim was to review systematically the diagnostic accuracy of the Helicobacter pylori stool antigen test. Bibliographical searches were performed in several electronic databases and abstracts from congresses up to May 2003. Eighty-nine studies (10,858 patients) evaluated the stool antigen test in untreated patients. Mean sensitivity, specificity, positive predictive value and negative predictive value were 91%, 93%, 92% and 87%, respectively. Analysis of the eight studies (1399 patients) in which pretreatment evaluation of the monoclonal stool antigen test was performed showed better (p < .001) results (96%, 97%, 96% and 97%, respectively), with a clearer distinction between positive and negative results. Thirty-nine studies (3147 patients) evaluated the stool antigen test for the confirmation of H. pylori eradication 4-8 weeks after therapy, with accuracies of 86%, 92%, 76% and 93% for mean sensitivity, specificity, positive predictive value and negative predictive value, respectively. Results were similar when a gold standard based on at least two methods was used. Relatively low accuracy was reported in some posttreatment studies with the polyclonal stool antigen test. However, excellent results (p < .001) were achieved in all the six studies evaluating the monoclonal stool antigen test 4-8 weeks posttreatment. Results evaluating the stool antigen test < 4 weeks posttreatment are contradictory. Proton-pump inhibitors seem to affect the accuracy of the stool antigen test. Sensitivity and/or specificity in patients with gastrointestinal bleeding may be suboptimal. The stool antigen test performs well in children. Finally, the stool antigen test seems to be a cost-effective method.