Laparoscopic Surgery for Crohn's Disease: A Meta-Analysis

Excerpt

To the Editor—I have with great interest read meta-analysis by Tan and Tjandra.1 The authors have performed a substantial and extensive literature search, and they describe and evaluate the individual included studies with correct methodology. They used the software RevMan 4.2, which was developed by the Cochrane Collaboration, when they performed their meta-analyses, but they have not used it in a proper way. In their analyses, they mix randomized, controlled studies with different kinds of nonrandomized and retrospective studies. This is known to enhance the risk of bias substantially and is absolutely not recommended by the Cochrane Collaboration.2 Although the authors discuss the problems of heterogeneity, they only mention this potential bias when compiling data from randomized and nonrandomized studies briefly. A way to solve this problem could have been to perform sensitivity analysis leaving out all data from the nonrandomized studies, thus investigating whether the results differ from the overall analysis. …