Realities of diagnosing Helicobacter pylori infection in clinical practice: a case for non-invasive indirect methodologies.

BACKGROUND: The current, arbitrarily defined gold standard for the diagnosis of H. pylori infection requires histologic examination of two specially stained antral biopsy specimens. However, routine histology is potentially limited in general clinical practice by both sampling and observer error. The current study was designed to examine the diagnostic performance of invasive and non-invasive H. pylori detection methods that would likely be available in general clinical practice. METHODS: The diagnostic performance of rotating clinical pathology faculty using thiazine staining was compared with that of an expert gastrointestinal pathologist in 38 patients. In situ hybridization stains of adjacent biopsy cuts were also examined by the expert pathologist for further comparison. Receiver operator characteristic (ROC) analysis was performed to evaluate whether the diagnostic performance of the expert pathologist differed depending upon the histologic method employed. A similar analysis was made to evaluate the diagnostic performance of pathology trainees relative to the expert. In the absence of an established invasive gold standard, non-invasive testing methods (rapid serum antibodies, formal Elisa antibodies and carbon-14 urea breath testing) were evaluated in 74 patients by comparison with a gold standard defined using a combination of diagnostic tests. RESULTS: Using either rapid urease testing of biopsy specimens or urea breath testing as the gold standard for comparison, the diagnostic performance of the rotating clinical pathology faculty was inferior to that of the expert gastrointestinal pathologist especially with regard to specificity (e.g., 69 percent for the former versus 88 percent, with the latter relative to rapid urease testing). Although interpretation of in situ hybridization staining by the expert appeared to have an even higher specificity, ROC analysis failed to show a difference. The mean ROC areas for thiazine and in situ hybridization staining for trainee pathologists relative to the expert were 0.88 and 0.94, respectively. In untreated patients, urea breath testing had a sensitivity and specificity of 100 percent as compared with thiazine staining with a sensitivity of 83 percent and a specificity of 97 percent. Post-therapy, breath testing had a sensitivity of 100 percent but a specificity of only 86 percent as compared with invasive testing with a sensitivity and specificity of 100 percent. Rapid serum antibody testing and formal Elisa antibody testing agreed in 93 percent of cases (Kappa 0.78) with the rapid test being correct in three of the four disagreements. CONCLUSIONS: The current study illustrates a number of realities regarding H. pylori diagnosis. There is no diagnostic gold standard in general clinical practice. Accurate interpretation of specially stained slides is a learned activity with a tendency towards overdiagnosis early on. Urea breath testing is likely to be the diagnostic method of choice for untreated patients in general clinical practice although antibody testing is almost as accurate. Rapid antibody tests are at least as accurate as formal Elisa antibody tests. Urea breath testing is useful for confirming cure after therapy, but false-positive results may occur in some patients.