Study characteristics
The systematic search process yielded 117 eligible studies (Figure 1) that included 509 environmental factor- and childhood cancer risk- specific meta-analyses. A total of 107 studies were excluded, among which duplicate studies, namely duplicate meta-analyses on the same exposure and outcome pair were identified for 10 associations (Supplementary Table 2). The 117 eligible systematic reviews and meta-analyses summarized evidence from 3873 individual study estimates stemming from 1140 primary studies. The vast majority (84.6%) of original studies were of case-control design (n=964), and the remaining 15.4% were cohort studies (n=176).
The 509 individual meta-analyses of the 117 eligible studies examined associations between environmental factors and 10 different cancer types (all cancer sites, hematological malignancies including overall acute leukemia, acute lymphoblastic [ALL] and acute myeloid [AML] leukemia, CNS tumors, neural tumors/neuroblastoma, sarcomas, bone tumors, retinoblastoma, testicular tumors, Wilms tumors and other solid tumors). The majority of studied outcomes were childhood hematological malignancies (n=374 associations, 73.5%), followed by CNS tumors (n=91 associations, 17.9%) and neural tumors/neuroblastoma (n=18 associations, 3.5%; Table 1). The potential risk factors were assessed during all relevant exposure windows, namely during preconception, pregnancy, delivery and childhood. A median of 11 (range 2-39) study estimates were combined for each meta-analysis. The average number of cancer cases in each meta-analysis was 3688 (range 19-31,610; Table 1). Around two-thirds of the 509 meta-analytical associations (62.9%) were assessed through random effects models, whereas 23.8% of these associations were evaluated only through fixed effects models.