Meta-analyses of studies on the association between electromagnetic fields and childhood cancer.

During the last 15 years several studies have investigated a possible relationship between exposure to electromagnetic fields (EMF) and childhood cancer. There is considerable variation between these studies with respect to methods of exposure assessment and reported results. Methods of exposure assessment range from simple visual criteria to costly and time consuming measurements or estimations of electric flux density. Additional individual refinements further hinder the comparability of results. We carried out several meta-analyses of data published so far taking into account the heterogeneity between studies as far as possible. Our particular interest was to investigate a potential dose-response-like relationship by comparing analyses for different cut-off points of exposure. Our meta-analyses suggest a marginal association between all cancer diagnoses combined and EMF exposure assessed by the two-level wire code (odds ratio, OR = 1.37, 95% confidence interval, CI: 0.94-2.00). Based on this criterion a significant effect was found for bases of leukemia (OR = 1.66, CI: 1.11-2.49) but not for central nervous system (CNS) tumors (OR = 1.5, CI: 0.69-3.26) or lymphomas (OR = 1.32, CI: 0.52-3.37). A significant increase in overall cancer risk with increasing stages of the four-level wire code (P = 0.003) could not be confirmed when data of the initial study performed by Wertheimer and Leeper were excluded (P = 0.17). When the exposure criterion was based on distance to the transmission line, estimated ORs for all cancers combined and for leukemias increased with distances decreasing from 100 to 25 m. Those analyses incorporating data on measured or calculated EMFs demonstrated also an increase of overall cancer risk with higher cut-off points. However, regarding individual diagnoses, this finding was reflected only in the group of brain tumors. One possible explanation for the high degree of heterogeneity between studies--especially with respect to methods of exposure assessment and choice of the respective cutpoint relevant for an increase in cancer risk--could be that published cut-off points were not always chosen in advance, but were selected because in exploratory analyses the most striking results were obtained with these specific cut-off values. Should this speculation be true at least partially, any meta-analysis will yield a false-positive finding. Further results of comparable studies with strictly a priori planned analyses are necessary to properly investigate a possible link between EMF and childhood cancer.