Assessment of publication bias required improvement in oral health systematic reviews

ObjectivesTo examine whether the assessment of publication bias in a broad cross-section of oral health systematic reviews (SRs) is in accordance with established methodology.Study Design and SettingThe electronic databases of 15 dental journals and the Cochrane Database of Systematic Reviews (Oral Health Group) were searched between January 2008 and December 2014 to identify eligible SRs and meta-analyses. The method of publication bias assessment and a range of study characteristics at the SR and the meta-analysis level were recorded.ResultsFour hundred fifty-eight systematic reviews were identified. Quantitative synthesis was undertaken in 162. MEDLINE (N = 454, 99%), Cochrane (N = 343, 75%), and EMBASE (N = 283, 62%) were the most frequently searched databases, whereas search for unpublished literature was used in 40% of the reviews. Publication bias was assessed in 46 meta-analyses with only 46% having appropriately used established methodology, such as funnel plots or statistical tests. Of the 38 meta-analyses including at least 10 studies, only 21 (55%) performed the assessment of publication bias.ConclusionThis empirical study highlights the shortcomings related to publication bias assessment in SRs within the field of oral health with publication bias either not assessed or done so inappropriately in more than half of the meta-analyses.