Grading the evidence
We graded the evidence regarding the association of environmental risk factors and childhood cancer accounting for the above criteria, namely the p -value of the significant associations, the presence and extent of heterogeneity, as well as the presence of small study effects and excess significance bias (Table 2). Overall, 40% of the 509 meta-analyses were nominally statistically significant, and were thereafter evaluated for strong, highly suggestive, suggestive or weak evidence. A hundred sixty-six of the 509 reported meta-analyses (32.5%) presented weak evidence (p <0.05 for the summary random effects).
We found no association supported by strong evidence. Of note is that the association of isolated cryptorchidism with testicular cancer fulfilled all criteria of strong evidence with the exception of considerable heterogeneity (I2>50%), and was thus considered as highly suggestive (Table 2).
Overall, 12 meta-analyses (2.4%) were supported by highly suggestive evidence (Figure 2). Among these meta-analyses, two showed a decreased risk, by approximately 20%, of ALL in relation to maternal vitamin supplementation during pregnancy (summary OR: 0.81, 95% CI: 0.74-0.88) and breastfeeding for more than 6 months (summary OR: 0.76, 95% CI: 0.68-0.84). Four meta-analyses supported by highly suggestive evidence found an increased risk for overall leukemia, especially ALL, in relation to paternal smoking during pregnancy , as well as residential exposure to pesticides, especially insecticides during pregnancy or childhood . There was also highly suggestive evidence for the association between high or increased birthweight and overall leukemia, particularly ALL . The remaining 4 out of 12 highly suggestive associations showed increased risk, by 14-22%, for childhood CNS tumors, and especially astrocytoma in relation to high birthweight (>4000 grams), as well as increased risk for testicular cancer in relation to testicular microlithiasis (summary OR: 15.46, 95% CI: 6.93-34.47) and isolated cryptorchidism (summary OR: 2.90, 95% CI: 2.21-3.82; Figure 2).
A total of 26 meta-analyses (5.1%) were supported by suggestive evidence (Figure 2). Three associations showed a significant inverse association of maternal folic acid supplementation during pregnancy and daycare attendance with ALL, as well as between breastfeeding for more than 6 months and overall acute leukemia. The remaining 23 associations showed increased risk for childhood hematological malignancies related to various exposures (use of assisted reproductive technologies, alcohol and coffee consumption during pregnancy, home exposure to pesticides, especially herbicides and insecticides during childhood, high exposure to traffic density, high benzene exposure, petrol station/repair garage proximity, fetal loss history, paternal ever smoking, birthweight increase and preterm birth) , as well as increased risk for CNS tumors in relation to birthweight increase and indoor pesticide exposure during preconception or childhood .
Sensitivity analyses including only systematic reviews and meta-analyses with ≥4 cohort or nested-case-control studies (n=28 meta-analyses) yielded statistically significant results for nine (32.1%) associations which were all supported by weak evidence. These associations concerned medically assisted reproduction, high birthweight and advanced paternal age in relation to increased risk of hematological and other solid cancers (results not shown).