Small study effects and excess significance bias
Indication for small study effects was evident in 65 meta-analyses
(13%) based on the Egger’s regression asymmetry test (p ≤0.10), of
which only 29 meta-analyses included 10 or more original studies, namely
enough power for the Egger’s test to identify the presence of small
study effects (Supplementary Table 5). The estimation of small study
effects was not feasible in 68 meta-analyses (13%) due to the small
number of included studies (n=2).
Thirty-seven meta-analyses (7%) had evidence of excess significance
bias based on the largest study effect size as the plausible effect
size. Of these 37 studies, 29 examined the risk of hematological
malignancies and the remaining 8 studies the risk of CNS tumors in
relation to various environmental exposures during preconception,
pregnancy or childhood (Supplementary Table 5).