Data synthesis and analysis
1. Summary effect estimates and 95% CI for each association between environmental factors and childhood cancer risk were calculated through fixed and random effects models .
2. Heterogeneity was assessed with the Cochran’s Q test and the I2 statistic (ranging from 0% to 100%), defined as the ratio of between-study variance over the sum of the within-study and between-study variances . We further calculated the 95% CIs to assess the uncertainty around heterogeneity estimates .
3. Ninety-five percent prediction intervals for the summary random effect estimates were calculated to further assess heterogeneity and to estimate the effect that would be expected in future studies investigating the same association .
4. Small study effects, namely whether smaller studies tend to contribute higher effect estimates compared to larger studies were also examined; such differences between small and large studies may indicate publication bias or other reporting biases, genuine heterogeneity or chance . To account for small study effects, we used the Egger’s regression asymmetry test (p ≤0.10) and we also assessed whether the random effects summary estimate was larger than the point estimate of the largest-most precise study, namely the study with the smallest standard error included in each meta-analysis.
5. Excess significance bias (set for individual meta-analyses atp ≤0.10) were assessed exploring whether the observed number of studies with nominally statistically significant results (“positive” studies, p <0.05) within each meta-analysis was greater than the expected number of studies with statistically significant results. Specifically, we calculated the expected number of statistically significant studies in each meta-analysis from the sum of the statistical power estimates for each component study using an algorithm from a non-central t distribution . The power estimates of each component study depend on the plausible effect size for the tested association, which was assumed to be the smallest standard error, namely the effect of the largest study in each meta-analysis .