A systematic review on the impact of studies that are not fully published on results of meta-analyses

Article type
Authors
Schmucker C1, Schell L1, Blümle A1, Briel M2, Schwarzer G3, Basler D4, Von Elm E5, Meerpohl J1
1German Cochrane Center, University Medical Center Freiburg, Germany
2Institute for Clinical Epidemiology and Biostatistics, University Hospital Basel, Switzerland
3Institute of Medical Biometry and Medical Informatics, University Medical Center Freiburg, Germany
4Department of Neonatology, University Hospital Zuerich, Switzerland
5Cochrane Switzerland, University Hospital Lausanne, Switzerland
Abstract
Introduction:
As part of a systematic review a meta-analysis aims to provide an unbiased summary of data from the literature. However, potentially important studies could be missing because of selective publication and inadequate dissemination of results. If results of missing studies differ systematically from published ones, a meta-analysis will be biased with an inaccurate assessment of the intervention’s effects.

Objectives:
As part of the OPEN project (www.open-project.eu) we conducted a systematic review of methodological research projects (MRPs) that assessed whether inclusion of studies that are not published or published in the grey literature impacts on pooled effect estimates in meta-analyses (quantitative measure) and leads to different conclusions in meta-analyses (qualitative measure).

Methods:
Four bibliographic databases were searched with no limit to publication year or language. MRPs were considered eligible for inclusion if they reviewed a cohort of meta-analyses which 1) compared pooled effect estimates of meta-analyses of health care interventions according to publication status or 2) examined whether the inclusion of unpublished/grey studies impacts the overall finding of a meta-analysis.

Results:
Seven MRPs including 1866 studies (from 1348 published studies) that compared pooled treatment effect estimates between published and unpublished/grey studies were identified. One MRP showed that published studies had larger pooled treatment effects in favour of the intervention than unpublished studies (Ratio of ORs 1.15, 95% CI 1.04-1.28). In the remaining six MRPs pooled effect estimates and overall findings were not changed by the inclusion of unpublished/grey studies.

Conclusion: There is only limited evidence from MRPs for an impact of studies that are not fully published on results of meta-analyses or systematic reviews. However, in several individual examples (e.g., a meta-analysis of reboxetine by Eyding et al 2010 published in BMJ) it could be shown that the inclusion of unpublished studies led to findings that differed from those based on published data.