Article Text

Clinical trials in Helicobacter pyloriinfection
  1. X CALVET,
  2. R COMET
  1. Internal Medicine Service,
  2. Consorci Hospitalari del Parc Taulí,
  3. Parc Taulí, s/n,
  4. 08208 Sabadell, Spain

    Statistics from

    Editor,—We read with great interest the excellent guidelines for clinical trials in Helicobacter pyloriinfection published last September (Gut1997;41(suppl 2):S1–23). The recommendations almost without exception, are extremely useful. We feel, however, that one or to points might be expanded, or deserve extra comment.

    As the working party states, most of the studies published on eradication therapy are non-comparative or are not large enough to reach a definite conclusion. This is because the sample size required to rule out a 10% difference between treatment groups for a two-branch comparative study is nearby 600 patients.1 To our knowledge no studies of this size have been published to date and, therefore, the length and composition of ideal eradication therapy has to be determined in large multicentre trials.

    Inclusion criteria, initial evaluation, and follow up in these large eradication studies should be as simple as possible in order to allow inclusion of such a large number of subjects with minimal disturbances for both investigators and patients.

    In more than one place, the guidelines call for two positive diagnostic tests for a patient to be considered reliably infected byH pylori and included in a treatment study. However, the reliability of diagnostic tests largely depends on the prevalence ofH pylori infection in the population studied. When studying a population known to harbour a very high prevalence of infection, such as patients with duodenal ulcer, two diagnostic tests are not necessary. A simple statistic calculation applying the Bayes theorem will show that in these patients a single diagnostic test—even serology—will be enough to confirm the diagnosis of H pylori infection. In brief, if we assume that the prevalence of infection is at least 90–95% in patients with duodenal ulcer, even assuming an exceptionally low sensitivity and specificity of 80% for the diagnostic technique, the predictive value of a positive result ranges between 97 and 99%. This is shown graphically in fig 1. Using more sensitive and specific tests such as histology or the urea breath test (UBT), the predictive value of a positive result will rise to almost 100%. Therefore, a second test will only add cost to the study, and will lead to the exclusion of a valuable percentage of patients with false negative results in the second test.

    Figure 1

    Predictive value of a positive serological test for H pylori in patients with duodenal ulcer. Positive predictive value = 72/(72+2) × 100 = 97.2%. Patients with duodenal ulcer are assumed to have a 90% prevalence of H pylori infection. Data have been calculated for a sensitivity and specificity of 80% for the serological test.

    Similar reasoning can be applied to the post-treatment controls. If we compare treatments with an effectivity of over 80% and perform a six week UBT (assuming a sensitivity and specificity of 90% for the technique)2 the predictive value of a negative test result is over 97% (fig 2). This predictive value increases when the efficacy of the treatment and the specificity of the test increase. Therefore, a second test is probably unnecessary. We agree that patients with a positive UBT after treatment should undergo endoscopy, but for different reasons: the predictive value of a positive UBT after treatment is low, and nearly one third of patients will be cured despite a positive post-treatment UBT (fig2).

    Figure 2

    Value of a negative UBT to evaluate cure of H pylori infection after treatment. Negative predictive value = 72/(72+2) × 100 = 97.2%. The eradication rate is assumed to be 80%. Data have been calculated for sensitivity and specificity of 90% for the UBT.

    Finally, the time for control is not clearly defined. Four weeks may be too short a period for control. Studies of eradication follow up have shown a 5% “recrudescence” rate after apparently effective treatment. However, these late failures are rare beyond 4–6 months after treatment.3 Therefore, the longer the period between the treatment and the UBT the more reliable the results will be. The possibility of performing a single test after 3–4 months should therefore be considered.


    View Abstract

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.