Endoscopic biopsy-based tests are considered to be the reference method for diagnosing Helicobacter pylori infection and monitoring antibiotic treatment, but unbiased data on their diagnostic performance is lacking. In this study we evaluated the diagnostic performance of culture, histology and rapid urease testing of antral biopsies separately and in combination. Antral biopsies were taken from consecutive patients undergoing upper gastrointestinal endoscopies at a single center between January 1995 and May 1997. The biopsies were examined for culture, histology, and CLOtest. The diagnostic performance, i.e., the sensitivity and specificity of the tests was estimated with 7 non-linear equations in 7 unknowns. To determine sources of heterogeneity that may have biased the results, data were stratified for age, gender, and whether they were taken before or after anti-Helicobacter antibiotic treatment. During the study period 631 patients underwent 869 upper gastrointestinal endoscopies. In 122 (14%) of the antral specimens the test results of culture, histology and CLOtest differed. Based on the nonlinear regression techniques we estimated that in 347 tests (40%) H. pylori infection was present. Overall sensitivity, specificity, positive and negative predictive value for each test were as follows: culture 91.4%, 96.3%, 94.2%, 94.4%, respectively; histology 90.3%, 97.8%, 96.4%, 93.8%, respectively; CLOtest 94.9%, 96.7%, 95.0%, 96.6%, respectively. In combination, the three tests provided the definitive diagnosis, either non-infected or infected, in 862 out of the 869 tests. Sensitivity of gastric antral histology was 64.9% (95% CI: 38-86) in females who did and 84.5% (95% CI: 77-90) in females who did not have had recent antibiotic therapy to cure the infection. Approximately 5-10% of H. pylori infected patients, were mis-diagnosed with a single biopsy-based test taken from the gastric antrum. Only a combination of bacterial culture, histological examination and the CLOtest represents an appropriate reference standard for research purposes to identify infected patients.