| Ref ID | 520 |
| First Author | L. Lin |
| Journal | BMJ EVIDENCE-BASED MEDICINE |
| Year Of Publishing | 2020 |
| URL | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6942244/pdf/nihms-1038392.pdf |
| Keywords |
• Publication bias • Statistical • Cochrane • General medical |
| Problem(s) |
• Poor consideration of publication bias |
| Article Type | Empirical |
| Article Subtype | Meta-epidemiological analysis |
| First Author Country | United States |
| Aim | To assess publication bias in a large cohort of Cochrane meta-analyses published between 2008-2013 in order to provide rules of thumb for interpreting small-study effects’ magnitudes based on the measures’ empirical distributions. Popular methods for detecting publication bias were examined including four variations of Egger's regression (the regression intercept and the skewness under both the fixed-effects and random-effects settings; random effects was further considered using the Q test, and with the DerSimonian-Laird method if heterogeneity was detected). Two variations of the trim and fill method were also considered: the proportion of suppressed studies and the relative change of the estimated overall effect size. |
| Level of Investigation | Analytical |
| Summary of Findings | The regression intercepts and skewness indicated substantial "small-study effects" (publication bias) in around 30% to 35% meta-analyses, and considerable small-study effects in around 20% to 40% of meta-analyses. The trim and fill methods (PTF and RTF) implied substantial and considerable small-study effects in less meta-analyses than the other measures. |
| Number of systematic reviews included | 3302 |
| Number of eligible systematic reviews assessed | 29932 |
| Treatment impacted | No |
| Treatment impacted description | |
| Interpretation impacted | Not Applicable |
| Interpretation impacted description |