Summary effect size
The summary fixed effects estimates were significant in 257 of 509
meta-analyses (50%), while the summary random effects estimates were
significant in 207 associations (41%) at a threshold ofp <0.05 (Supplementary Table 4). At a stricter threshold
(p <0.001), 114 (22%) and 77 (15%) meta-analyses were
statistically significant using fixed and random effects models,
respectively. When the p -value was set at 10-6,
43 (8%) fixed effects estimates and 21 (4%) random effects estimates
were statistically significant.
Of the 21 statistically significant summary random effect sizes atp <10-6 (Supplementary Table 4), 13
showed a significant association between high birthweight, paternal
smoking, maternal alcohol consumption, exposure to benzene, rural
population mixing, as well as residential and occupational exposure to
pesticides, especially insecticides and childhood leukemia risk. The
magnitude of these effect sizes ranged between 1.19 for high birthweight
and 3.30 for maternal occupational pesticide exposure. In addition, four
summary random effects estimates found a decreased risk for childhood
leukemia at p <10-6 in relation to
maternal dietary vitamin intake during pregnancy (summary OR: 0.81,
95%CI: 0.74, 0.88), maternal multivitamin supplementation during
pregnancy (summary OR: 0.64, 95%CI: 0.52, 0.80), breastfeeding for more
than 6 months (summary OR: 0.76, 95%CI: 0.68, 0.84), and high maternal
education (summary OR: 0.81, 95%CI: 0.76, 0.87). The remaining 4 out of
21 significant effect sizes found an increased risk for CNS tumors, and
especially astrocytomas in relation to high birthweight, as well as for
testicular cancer in relation to testicular microlithiasis and isolated
cryptorchidism.
The largest study in each meta-analysis was statistically significant in
186 of 509 associations (37%), albeit the effect sizes of the largest
studies were in general more conservative than the respective summary
random effects estimates (Supplementary Table 4).