The recently published article by Wang et al1 is a beautiful study, although the presentation and interpretation of the results cause some serious concern. In recent literature there have been 2 other meta-analyses summarizing the use of proton pump inhibitors (PPIs) as a diagnostic test for gastroesophageal reflux disease (GERD), both leading to more attenuated conclusions.2,3 This, in my opinion, reflects reality better than the conclusions of Wang et al1 do. In the case of the PPI test, the choice of the outcome measure on effectiveness of treatment should be as close as possible to complete resolution of symptoms because any abdominal symptom is subject to placebo. However, Wang et al1 chose to consider a 50% symptom reduction enough for testing positive on PPI treatment. Although the authors suggest that the chance of having GERD is increased if more than a 50% symptom reduction in a patient is found, using a higher cutoff point for the test would further affect specificity (which is already low) as well as the sensitivity of the test. Even in these highly selected populations, many patients with functional disease might wrongly be diagnosed as having or not having GERD following PPI treatment. Cremonini et al2 indeed show in their tables that when the higher cutoff point “complete relief of symptoms” would have been chosen in the same studies, the summarized test characteristics would have been reduced significantly (diagnostic odds ratio from 13.83 to 7.80, both with wide confidence intervals).
Numans ME. The Use of Proton Pump Inhibitors as a Diagnostic Test for Gastroesophageal Reflux Disease. Arch Intern Med. 2006;166(2):247-248. doi:10.1001/archinte.166.2.247-b