Tests for publication bias are unreliable in case of heteroscedasticity.
Aggregated data meta-analysis
Heteroscedastic mixed effects model
Mean difference treatment effect sizes
Journal
Contemporary clinical trials communications
ISSN: 2451-8654
Titre abrégé: Contemp Clin Trials Commun
Pays: Netherlands
ID NLM: 101671157
Informations de publication
Date de publication:
Jun 2021
Jun 2021
Historique:
received:
12
08
2020
revised:
16
04
2021
accepted:
28
04
2021
entrez:
28
6
2021
pubmed:
29
6
2021
medline:
29
6
2021
Statut:
epublish
Résumé
Regression based methods for the detection of publication bias in meta-analysis have been extensively evaluated in literature. When dealing with continuous outcomes, specific hidden factors (e.g., heteroscedasticity) may interfere with the test statistics. In this paper we investigate the influence of residual heteroscedasticity on the performance of four tests for publication bias: the Egger test, the Begg-Mazumdar test and two tests based on weighted regression. In the presence of heteroscedasticity, the Egger test and the weighted regression tests highly inflate the Type I error rate, while the Begg-Mazumdar test deflates the Type I error rate. Although all three tests already have low statistical power, heteroscedasticity typically reduces it further. Our results in combination with earlier discussions on publication bias tests lead us to conclude that application of these tests on continuous treatment effects is not warranted.
Identifiants
pubmed: 34179565
doi: 10.1016/j.conctc.2021.100781
pii: S2451-8654(21)00082-X
pmc: PMC8209747
doi:
Types de publication
Journal Article
Langues
eng
Pagination
100781Informations de copyright
© 2021 The Authors.
Références
Biometrics. 1994 Dec;50(4):1088-101
pubmed: 7786990
Addiction. 2009 Feb;104(2):297-307
pubmed: 19149827
Stat Med. 2006 Oct 30;25(20):3443-57
pubmed: 16345038
Stat Med. 2002 Sep 15;21(17):2465-77
pubmed: 12205693
BMJ. 1997 Sep 13;315(7109):629-34
pubmed: 9310563
Stat Med. 2001 Mar 30;20(6):825-40
pubmed: 11252006
Am Rev Respir Dis. 1985 Apr;131(4):511-20
pubmed: 3873193
JAMA. 2006 Feb 8;295(6):676-80
pubmed: 16467236
Elife. 2017 Sep 08;6:
pubmed: 28884685
BMC Med Res Methodol. 2014 Dec 17;14:132
pubmed: 25516509
Stat Methods Med Res. 2018 Sep;27(9):2722-2741
pubmed: 28024447
J Clin Epidemiol. 2005 Sep;58(9):882-93
pubmed: 16085191
Control Clin Trials. 1986 Sep;7(3):177-88
pubmed: 3802833
Stat Med. 2008 Feb 28;27(5):746-63
pubmed: 17592831
Stat Med. 2014 Dec 30;33(30):5399-412
pubmed: 25316006
Res Synth Methods. 2019 Mar;10(1):57-71
pubmed: 30506832
Lancet. 2010 Mar 13;375(9718):895-905
pubmed: 20226988
Stat Med. 2001 Feb 28;20(4):641-54
pubmed: 11223905
Biom J. 2017 Sep;59(5):1047-1066
pubmed: 28692751
Ann Intern Med. 2014 Feb 18;160(4):267-70
pubmed: 24727843
Behav Res Methods. 2007 Nov;39(4):709-22
pubmed: 18183883
Stat Med. 2007 Feb 20;26(4):721-33
pubmed: 16755545