Several reasons explained the variation in the results of 22 meta-analyses addressing the same question

• Published in: Journal of clinical epidemiology 2019-09, Vol.113, p.147-158
• Main Authors: Khamis, Assem M, El Moheb, Mohamad, Nicolas, Johny, Iskandarani, Ghida, Refaat, Marwan M, Akl, Elie A
• Format: Article
• Language: English
• Subjects: Adult; Aged; Aged, 80 and over; Cardiomyopathy; Clinical trials; Data Accuracy; Decision analysis; Defibrillators, Implantable; Epidemiology; Female; Heart Diseases - therapy; Humans; Internal Medicine; Male; Meta-analysis; Meta-Analysis as Topic; Meta-research; Middle Aged; Mortality; Odds Ratio; Quality of research; Questions; Statistics methods; Studies; Systematic review; Systematic Reviews as Topic; Time-to-event; Variation
• ISSN: 0895-4356; 1878-5921
• DOI: 10.1016/j.jclinepi.2019.05.023

AbstractObjectiveThe objective of this study was to assess and to investigate the reasons for the variations between the results of meta-analyses addressing the same question. Study Design and SettingWe included systematic reviews, and the trials that they included, on the use of implantable cardiac defibrillator (ICD) in patients with nonischemic cardiomyopathy. We assessed the variation between meta-analyses pooled effect estimates by calculating the percentage of absolute difference. We developed a list of 10 reasons for variations between the results of the meta-analyses clustered in four overarching categories. ResultsWe identified 21 systematic reviews including six trials and reporting on 22 eligible meta-analyses. The percentage of absolute difference between each of the 22 pooled effect estimates (included odds ratio, risk ratio, hazard ratio) and their median value had an average of 3.2%. The number of trials for which the following categories of reasons for variations applied were as follows: (1) different decision to include or exclude trials ( n = 3); (2) differences in analytical approaches ( n = 6); (3) errors in conducting meta-analyses ( n = 5); and (4) unclear reason ( n = 1). ConclusionFew of the observed variations between the results of the 22 meta-analyses could lead clinicians or guideline development organizations to adopt different courses of actions. Variations were most frequently related to both errors and variations in trial eligibility and analytical approaches.