Meta-analysis involves more than calculating a summary estimate from an amalgamation of studies. The letter and meta-analysis of Drs. Moayyedi, Deeks, and Forman illustrate the need to assess other factors before drawing conclusions. The widely accepted definition of dyspepsia is upper abdominal pain or discomfort, excluding predominant reflux symptoms. Although Moayyedi and colleagues state that they excluded trials including participants with only reflux symptoms, the two positive trials in their meta-analysis included patients with reflux (1, 2). They report that a third trial was positive, but treatment success, the primary end point, did not improve with therapy (3). Analysis of studies in their review using the appropriate definition of dyspepsia shows no benefit (P = 0.5), just as in our meta-analysis.

Significant heterogeneity, as identified for our primary end point, suggests that studies may be too different to aggregate in a meta-analysis. Moayyedi and colleagues reported a P value of 0.53 for heterogeneity. However, using data from their meta-analysis, we calculated a P value of 0.15, which approaches statistical significance (P < 0.1). They state that removal of any article leads to similar results, but removal of a study leads to significant heterogeneity. In addition, Moayyedi and colleagues state that their conclusions were maintained in all sensitivity analyses, but they do not report any of their predefined sensitivity analyses. The sensitivity analysis they included had significant heterogeneity.

Application of Moayyedi and colleagues' methods to our meta-analysis (although not appropriate with heterogeneity) did not produce significance. Additional studies may alter results, but only one of their additional studies showed a significant benefit (8%; P = 0.04) (1). Although this study is more than twice the size of any other trial, it was not peer-reviewed, used a nine-symptom scoring system that includes nondyspepsia symptoms, and has uncertain methods. The other additional studies had differences of –4% to 2%, similar to those in recently published large studies showing no benefit (4).

Ultimately, meta-analysis requires an appropriate examination of study design and patient population. Primary summary estimates from our meta-analysis and the meta-analysis by Moayyedi and colleagues are similar. However, more careful evaluation of both meta-analyses shows no suggestion of benefit when appropriate definitions of dyspepsia are used or when dyspepsia scores are analyzed. Without critical analysis of studies, a larger meta-analysis does not produce a more credible meta-analysis.

Dr. Buckingham suggests screening for H. pylori to prevent gastric cancer. An association of cancer with lifelong H. pylori infection cannot be taken as evidence that eradication decreases cancer risk, and no randomized trial has demonstrated such a benefit. The lack of documented benefit and the high cost (estimated at $1 billion [5]) prevent our recommending widespread screening for H. pylori infection.

Dr. Bozdech raises an intriguing idea, although we know of no documentation showing that giardiasis accounts for a substantial proportion of cases of nonulcer dyspepsia.