Publication bias

The publication bias, also called publication bias, is the statistically distorted ( bias [ˈbaɪəs]) presentation of the data situation in scientific journals as a result of the preferred publication of studies with “positive” or significant results. It was discovered in 1959 by statistician Theodore Sterling. Positive results are easier to publish than those with “negative”, ie non-significant results and are also more frequently published in specialist journals with a high influencing factor .

description

The term file drawer problem , which the psychologist Robert Rosenthal coined in 1979, is often used synonymously for publication bias . This describes the phenomenon related to publication bias that researchers increasingly no longer submit their insignificant results for publication, but instead let them disappear into a drawer.

Due to the increased frequency of positive results, the effectiveness of therapies can be overestimated in medicine , since studies with proven effectiveness are easier to publish than those that cannot prove effectiveness. This is particularly relevant if therapy recommendations are to be generated based on the already published data based on a meta-analysis . The suspicion of a publication bias can be corroborated by creating a funnel plot .

Conflicts of interest can also lead to publication bias, especially when economic interests play a role. In the medical drug research was conducted at antidepressants found that the published articles in journals had a positive tenor than that in the US regulatory authority FDA submitted registration dossiers. One possible reason here is an economic interest, for example a pharmaceutical company that sponsored a study is significantly more interested in the publication of positive results for its products than in negative results.

For the reasons mentioned, some of the renowned medical journals now require that all studies carried out must be published in advance. Only such announced studies will be accepted for publication. In addition to other aspects, this should enable an overview of the studies carried out on the topic in order to at least be able to estimate the publication bias. For this reason, many specialist journals also require the disclosure of conflicts of interest of the authors, for example by specifying financial support from foundations, research associations, etc.

In addition, there are already specialist journals (primarily on the Internet, see below) that specifically target studies with “negative”, i.e. H. do not publish results that are significant for the purposes of the research question. The Cochrane Collaboration is also very interested in such results in order to be able to use them in their analyzes of the standards in medicine.

Methods for determining publication bias

A so-called funnel plot

A fundamental technique is the creation of a funnel plot (literally "funnel diagram", a common German term does not exist), which was presented in 1984 by statistician Richard J. Light and psychologist David B. Pillemer. On this diagram, the effect observed in the individual study is plotted on the x-axis and the precision (the more test persons or test objects, the smaller the confidence interval and the higher the precision) on the y-axis. If the points in the funnel plot are symmetrically distributed, it can be assumed that both “desired” and “undesirable” study results have been published, i.e. that publication bias is denied.

The funnel plot was created in 1997 by the Swiss epidemiologist Matthias Egger u. a. discussed and supplemented by the Egger test , which either confirms or excludes a statistically significant asymmetry of the diagram.

The relatively new idea that's mark and recapture (Engl. Capture-recapture method ) on publication databases and other bibliographic apply sources: Look up articles on a specific topic on a database and stores the results (capture) . The same search is repeated on a second database (recapture) . This allows the true number of publications on a particular topic to be estimated.

See also

• p-hacking

Web links

• BioMed Central (BMC) , London: Journal of Negative Results in Biomedicine . Open access project supervised by an international team of scientists , which is intended to promote the scientific discussion of unexpected, contradicting, provocative and / or negative research results in the field of biomedicine under peer review (last reviewed on June 6, 2012)
• Journal of Articles in Support of the Null Hypothesis
• CCT - The meta-register for clinical studies
• American Directory of Clinical Studies
• McGauran N, Wieseler B, Kreis J, Schüler YB, Kölsch H, Kaiser T: Reporting bias in medical research - a narrative review . In: Trials . 11, 2010, p. 37. doi : 10.1186 / 1745-6215-11-37 . PMID 20388211 . PMC 2867979 (free full text).
• PsychFileDrawer. Archive of Replication Attempts in Experimental Psychology : online journal published by an international team of scientists that publishes repeated experimental psychological studies regardless of their outcome (last checked on June 6, 2012)
• The All Results Journals : online journal published by an international team of scientists that publishes experimental studies in chemistry, biology, nanotechnology and physics that have not otherwise been accepted for publication because they did not confirm expected results or did not lead to usable results ( last checked on June 6, 2012)

literature

• Hans-Hermann Dubben , Hans-Peter Beck-Bornholdt , Unbalanced Reporting in Medical Science , Hamburg: Institute for General Medicine of the University Medical Center Hamburg-Eppendorf, 2004, ISBN 3-00-014238-X
• Bortz, J. & Döring, N. (2006). Research methods and evaluation for human and social scientists (p. 695). Heidelberg: Springer. ISBN 3-540-33305-3

Individual evidence

1. ^ Arjo Klamer, Robert M. Solow, Donald N. McCloskey: The Consequences of economic rhetoric . Cambridge University Press, 1989, ISBN 978-0-521-34286-5 , pp. 173-74. Archived from the original on May 23, 2011 Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. (Accessed July 8, 2015). 
2. ↑ Jonah Lehrer: The Truth Wears Off . In: The New Yorker , December 13, 2010. Retrieved July 8, 2015. 
3. ^ Theodore D. Sterling: Publication decisions and their possible effects on inferences drawn from tests of significance — or vice versa. Archived from the original on August 7, 2011. In: Journal of the American Statistical Association . 54, No. 285, March 1959, pp. 30-34. doi : 10.2307 / 2282137 . Retrieved July 8, 2015.
4. ↑ PsychFileDrawer: The File Drawer Problem
5. ^ Daniele Fanelli: Negative results are disappearing from most disciplines and countries . Scientometrics, Vol. 90, Number 3 (2012), pp. 891-904, doi : 10.1007 / s11192-011-0494-7 ; Manuela Lenzen: Journal of Second Views. An initiative calls for more experiments to be repeated . Frankfurter Allgemeine Zeitung, June 6, 2012, p. N 5
6. ↑ Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R: Selective publication of antidepressant trials and its influence on apparent efficacy . In: N. Engl. J. Med. . 358, No. 3, January 2008, pp. 252-60. doi : 10.1056 / NEJMsa065779 . PMID 18199864 .
7. ^ Richard J. Light, David B. Pillemer: Summing Up: The Science of Reviewing Research . Harvard University Press, Cambridge, Mass., ISBN 978-0-674-85431-4 , pp. 65 ff .
8. ^ Matthias Egger, George Davey Smith, Martin Schneider, Christoph Minder: Bias in meta-analysis detected by a simple, graphical test . In: BMJ . tape 315 , no. 7109 , September 13, 1997, ISSN  0959-8138 , p. 629-634 , doi : 10.1136 / bmj.315.7109.629 , PMID 9310563 ( bmj.com [accessed December 8, 2016]).
9. ↑ Poorolajal J, Haghdoost AA, Mahmoodi M, Majdzadeh R, Nasseri-Moghaddam S, Fotouhi A. Capture-recapture method for assessing publication bias. Journal of Research in Medical Sciences: The Official Journal of Isfahan University of Medical Sciences. 2010; 15 (2): 107-115. PMC 3082794 (free full text)