ENTRAR            

ORIGINALS

Publication bias: does it also exist in meta-analysis studies?

Mª Carmen Rosa Garrido
Fundación Pública Andaluza para la Investigación Biosanitaria de Andalucía Oriental (FIBAO). Complejo Hospitalario de Jaén, Servicio Andaluz de Salud. Jaén, España

Index de Enfermería [Index Enferm] 2016; 25(1-2): 7-8

How to cite this document

    It was around the middle of last century that Sterling defined publication bias, indicating that studies with statistically significant results have a higher possibility of being published. On the other hand, studies with results that are not significant are published to a lesser extent. Sterling made the problems that could cause this bias clear: researchers that repeat studies with negative results, posible errors in the studies with positive results which will not be discussed in other studies with negative results, etc.¹
    There are various factors which influence publication bias: (a) Author's decision to not publish results that are not statistically significant. (b) Journal editor's rejection to publish studies with negative results, even though the methodological quality² is adequate to provide reliable results. (c) The exemption of these type of studies in the bibliographical searches of other researchers.
    The issue which is being addressed in this text is: does publication bias affect all researchers in the same way or does it depend on the scientific evidence of their design? Does publication bias particularly exist in meta-analysis studies?
    Glass coined the term 'meta-analysis' for the first time in 1976, and he defines it as 'the statistical analysis of a collection of results from individual studies in order to evaluate the findings'.³ Meta-analysis is understood as a systematic examination which concludes with a quantitative study of the results from the different selected articles. The objective is to provide a structured, explicit and systematic response to a specific question. Due to being considered as research studies they have to be structured as such including the following sections: hypothesis of the study, information search, inclusion criteria, evaluation of studies, publication bias, statistical methods, heterogeneity analysis.⁴
    The rise of the movement named 'Evidence-Based Medicine' which seeks to provide medicine practice with the most objective rationale possible, has resulted in the hierarchy of epidemiological design according to it's scientific evidence level as a direct consequence. This classification indicates that the systematic reviews present a higher level of evidence.⁵ Furthermore, the implementation of meta-analysis has expanded beyond clinical trials and has also been employed in observational studies, dose-response relationship and diagnostic test evaluation studies.
    All of this has facilitated the exponential rise of the number of meta-analysis studies which are anually published. The issue which needs addressing pertains to if all meta-analysis articles published effectively incorporate adequate methodological quality.
    Some authors have already pinpointed deficiencies, in 1999 the QUOROMdeclaration was published, through a Delphi technique with 30 experts, a list of 18 items and a flow chart to describe the process of meta-analysis was produced.⁶ The aforementioned declaration was orientated towards both authors and editors of journals to aid the critical evaluation of this type of analysis. The QUOROMdeclaration has been updated with the PRISMAdeclaration⁷ incorporating conceptual and methodological innovations. Parallel to the PRISMAdeclaration the latest update has been devised by Manuel Cochrane of systematic reviews of 2011 interventions.⁸ The manual serves as a guideline to the authors in which definitions and useful templates to carry out the review can be found. Another of the tools to be created on the topic of systematic reviews is PROSPERO. It is an international database of systematic reviews registered in health and social care. The main protocol characteristics of the review are registerd there and they are retained as a permanent record (PROSPERO, University of York. https://www.crd.york.ac.uk/PROSPERO/). PROSPERO aims to reduce the non-planned duplication and the increase of transparency, helping to safeguard against the selective report making it possible for editors and reviewers to compare methods with the final report.
    My personal reflection is despite the multiple tools and efforts of the scientific community to ensure meta-analysis publications are of adequate quality, meta-analysis is not subject to publication bias in any of its approaches, given that we are immersed in the 'meta-analysis fever'. Unfortunately few experts on the matter exist in relation to the number of publications demonstrating that the studies carried out are of a low scientific quality.
 

References

1. Sterling RJ. Publication decisions and their possible effects on inferences drawn from test of significance or viceversa. J Am Stat Assoc. 1959; 54: 30-4.
2. Mahoney MJ. Publication perjudices: an experiment study of confirmatory bias in the peer review system. Cog Ther Res. 1977; 1: 161-175.
3. Glass GV. Primary, secondary and meta-analysis of research. Educ Res. 1976; 5: 3-8.
4. Delgado M, Sillero M, Gálvez R. Metaanálisis en epidemiología (primera parte): características generales. Gac Sanit. 1991; 27: 265-72.
5. Revisiones sistemáticas y metaanálisis: ¿son la mejor evidencia? Rev Méd Chile. 2005; 133: 246-249.
6. Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF, et al. Improving the quality of reports of meta-analyses of randomised controlled trials: The QUOROM statement. Lancet 1999; 354: 1896-900.
7. Urrútia, Gerard; Bonfill, Xavier. Declaración PRISMA: una propuesta para mejorar la publicación de revisiones sistemáticas y metanálisis. Med Clin (Barc). 2010; 135(11): 507-511.
8. Centro Cochrane Iberoamericano, traductores. Manual Cochrane de Revisiones Sistemáticas de Intervenciones, versión 5.1.0 [actualizada en marzo de 2011]. Barcelona: Centro Cochrane Iberoamericano; 2012 [acceso: 06/11/2015]. Disponible en https://www.cochrane.es/?q=es/node/269.