Get access

Methods for the bias adjustment of meta-analyses of published observational studies

Authors

  • Suhail A. R. Doi,

    Associate Professor of Clinical Epidemiology, Consultant, Corresponding author
    1. Department of Endocrinology, Princess Alexandra Hospital, Brisbane, Queensland, Australia
    • School of Population Health, University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Jan J. Barendregt,

    Associate Professor of Epidemiologic Modelling
    1. School of Population Health, University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Adedayo A. Onitilo

    PhD scholar, Chairman, MemberBoard of Trustees
    1. School of Population Health, University of Queensland, Brisbane, Queensland, Australia
    2. Department of Hematology/Oncology, Marshfield Clinic Weston Center, Weston, WI, USA
    3. Marshfield Clinic Research Foundation, Marshfield, WI, USA
    Search for more papers by this author

  • Conflict of interest: JJB is founder and owner of Epigear International Pty Ltd, which makes the MetaXL software.
  • This article was published online on July 29, 2012. Errors in two equations in the appendix were subsequently identified. This notice is included in the online and print versions to indicate that both have been corrected [February 26, 2013].

Correspondence

Dr Suhail Doi

Clinical Epidemiology Unit

School of Population Health

University of Queensland

Herston Road

Brisbane, Qld 4006

Australia

E-mail: sardoi@gmx.net

Abstract

Background

A unique challenge in meta-analysis of observational studies is bias adjustment. Two different approaches have been proposed for doing this – using summary scores versus component scores. The prevailing view on this matter is that summary quality scores are inaccurate because information from its components can cancel each other out.

Methods

A head-to-head comparison of the component score adjustment with our method using summary scores is undertaken, using data reported by the authors of the component method.

Results

It is demonstrated that the consideration of components or of aggregate scores does indeed lead to the same conclusions. Yet, the latter does not require imputation of the direction and magnitude of changes to effect sizes.

Conclusions

The summary quality score used for bias adjustment within the context of an appropriate model may be most expedient. Implications for the bias adjustment of meta-analyses of observational studies are discussed.

Ancillary