Medical discussion at hospital with patient

Selecting Studies for Meta-Analysis: Publication Bias

Meta-analysis is a statistical method of combining the results of individual studies. Meta-analysis may allow a more precise estimate of treatment effects, and may explain differences between the results of individual studies. Although meta-analysis can be a powerful and useful technique, they must be designed carefully to ensure that the results are not misleading.
The identification and selection of individual studies used in a meta-analysis is critical. Publication bias has a direct impact on this selection.
Publication bias refers to the greater likelihood of papers showing positive results to be published than negative results. If the sample of studies selected for meta-analysis is biased, the conclusions drawn from the analysis may be invalid. Turner(2) analyzed the publication status of antidepressant drugs. He found that 97% of the positive studies were published versus only 12% of negative studies. The inclusion of negative studies in the analysis reduced the positive effects of the drugs.
Statistical tests have been developed to identify and assess the impact of publication bias in meta-analysis.
Attempting to locate unpublished studies is time consuming, difficult, and can use methodology hard to assess.(1)
In evaluating the results of meta-analyses (as well as for individual studies) it is important to be aware of who is sponsoring the studies and whether the authors have any vested interests.
Is it possible to overcome the problem of publication bias? The National Institutes of Health keeps a registry of all studies it supports, and the FDA keeps a registry and database in which drug companies must resister all trials they intend to use in applying for marketing approval or for changes in labeling.
Prospective meta-analysis looks forward to meta-analysis in the planning stage of individual trials. Researchers of these trials agree, prior to knowing the results of their studies, to combine their findings when the trials are complete.(3) The researchers agree on trial design as well as outcome measures which facilitates the analysis of results.

References
1. Rothstein, Hannah, Alexander Sutton & Michael Borenstein, eds. Publication Bias in Meta-Analysis: Prevention, Assessment and Adjustments. 2005. John Wiley & Sons

http://www.wiley.com/WileyCDA/WileyTitle/productCd-0470870141.html
2. Turner, Erick, et al. Selective Publication of Antidepressant Trials and Its Influence on Apparent Efficacy. N Engl J Med. 2008; 358: 252-60. Trials (BioMed Central) 2011;12:104

http://www.nejm.org/doi/full/10.1056/NEJMsa065779
3. Turok, David et al. The methodology for developing a prospective meta-analysis in the family planning community.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3103448/
4. Walker, Esteban, et al. Meta-analysis: Its strengths and limitations. Clev Clin J Med. June 2008; 75(6): 431-439.

http://www.ccjm.org/index.php?id=107937&tx_ttnews[tt_news]=360745&cHash=c4f41096946e87f82f539928262b817f