Supplementary Material for: Bias in Peripheral Depression Biomarkers
datasetposted on 26.01.2016 by Carvalho A.F., Köhler C.A., Brunoni A.R., Miskowiak K.W., Herrmann N., Lanctôt K.L., Hyphantis T.N., Quevedo J., Fernandes B.S., Berk M.
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Background: To aid in the differentiation of individuals with major depressive disorder (MDD) from healthy controls, numerous peripheral biomarkers have been proposed. To date, no comprehensive evaluation of the existence of bias favoring the publication of significant results or inflating effect sizes has been conducted. Methods: Here, we performed a comprehensive review of meta-analyses of peripheral nongenetic biomarkers that could discriminate individuals with MDD from nondepressed controls. PubMed/MEDLINE, EMBASE, and PsycINFO databases were searched through April 10, 2015. Results: From 15 references, we obtained 31 eligible meta-analyses evaluating biomarkers in MDD (21,201 cases and 78,363 controls). Twenty meta-analyses reported statistically significant effect size estimates. Heterogeneity was high (I2 ≥50%) in 29 meta-analyses. We plausibly assumed that the true effect size for a meta-analysis would equal the one of its largest study. A significant summary effect size estimate was observed for 20 biomarkers. We observed an excess of statistically significant studies in 21 meta-analyses. The summary effect size of the meta-analysis was higher than the effect of its largest study in 25 meta-analyses, while 11 meta-analyses had evidence of small-study effects. Conclusions: Our findings suggest that there is an excess of studies with statistically significant results in the literature of peripheral biomarkers for MDD. The selective publication of ‘positive studies' and the selective reporting of outcomes are possible mechanisms. Effect size estimates of meta-analyses may be inflated in this literature.