Meta-analysis of capsule endoscopy in patients diagnosed or suspected with esophageal varices

Abstract

The PillCam ESO (Given Imaging, Israel) or esophageal capsule endoscopy (ECE) is a novel technique used in the diagnostic evaluation of esophagus. Many studies have been performed to compare the accuracy of ECE against the current gold standard esophago-gastro-duodenoscopy and a meta-analysis recently published by Lu et al suggests that ECE may have an acceptable sensitivity and specificity in detecting esophageal varices. We would like to discuss the importance and implication of publication bias in this meta-analysis.

Key Words: Capsule endoscopy; Screening varices

TO THE EDITOR

In March 2009 issue of World Journal of Gastroenterology, Lu et al[1] published their interesting findings regarding the accuracy of esophageal capsule endoscopy (ECE) in detecting esophageal varices and its utility in screening and surveillance of esophageal varices. We would like to add that it is vital to comment on the presence or absence of any bias when reporting a meta-analysis as this will allow the readers to assess the strengths and weaknesses of the recommendation made. A recently published statement on preferred reporting items for systematic reviews and meta-analyses (PRISMA)[2], an evolution of the original quality of reporting of meta-analyses (QUOROM) guidelines, suggests that publication bias should be assessed while reporting meta-analyses and systematic reviews. Since publication bias was not reported in the above meta-analysis, we analyzed the data for the presence or absence of publication bias.

We assessed the publication bias using funnel plot. Funnel plots were plotted using log odds ratio vs standard error (Figure 1A) and log odds ratio vs precision (Figure 1B), both showed no evidence of publication bias. We complemented the funnel plots with Eggers test[3] and rank correlation analysis[4] and both also showed no evidence of publication bias (P > 0.05).

Figure 1 Funnel plot. A: Log odds ratio vs standard error; B: Log odds ratio vs precision (Precision = 1/standard error).

We hope that this information about the absence of publication bias in this meta-analysis will add more value to the conclusion reported in the above meta-analysis.