Methodologic Issues in the Studies of Childhood Leukemia and Overhead Power Lines

Aims: While studies have consistently found an association between childhood leukemia risk and magnetic fields, similar associations between childhood leukemia and distance to overhead power lines suggest that other factors associated with magnetic fields and proximity to overhead power lines may be responsible for observed associations including bias, confounding, or other methodologic challenges, particularly when it comes to residential mobility and dwelling type. Methods: First, we pooled data from multiple studies to assess the association with distance and evaluate whether it is due to magnetic fields or other factors associated with distance from lines. We then analyzed a single study from California to assess predictors of residential mobility between birth and diagnosis, and account for potential confounding due to residential mobility. Next, we simulated a synthetic dataset based on that study and used it to assess the sensitivity of electromagnetic field (EMF)-leukemia associations to different scenarios of uncontrolled confounding by mobility under two major hypotheses of the infectious etiology of childhood leukemia; then used the findings to conduct sensitivity analysis and empirically offset the potential bias due to unmeasured mobility in the actual California study. Finally, we assessed whether dwelling type is a risk factor for childhood leukemia, what covariates are related to dwelling type, whether dwelling type behaves as a confounder or as a potential effect measure modifier in the EMF-leukemia relationship.Results: Although we found no material association between childhood leukemia and distance to nearest overhead power line of any voltage, there was a slight increase in risk of leukemia among children living <50 m from 200+ kilovolt power lines, consistent with some previous findings. There was no association found with calculated magnetic fields in this set of studies, however, and odds ratios (ORs) remained unchanged with adjustment for potential confounders in the pooled analysis.In the California study, we found that mobility was strongly associated with age, dwelling type, and SES. Both EMF-leukemia associations were stronger in the stratum of non-movers, too, but adjustment for proxy variables had no effect. In the hybrid-simulation study, as expected, the stronger the assumed relationship between mobility and exposure and outcome, the greater the potential bias. However, no scenario created a bias strong enough to completely explain away previously observed associations. In all mobility analyses, only dwelling type seemed to affect the relationship based on a small subset of subjects, However, when expanded to a larger subset, dwelling type was neither associated with childhood leukemia risk, nor functioned as a confounder. Stratification revealed potential effect measure modification by dwelling type only.Conclusion: Although uncontrolled confounding by residential mobility had some impact on the estimated effect of EMF exposures on childhood leukemia, it is unlikely to be the primary explanation for the associations observed between power lines exposure and childhood leukemia. Similarly, dwelling type does not appear to play a significant role as either a risk factor or confounder. Future research should explore the role of dwelling and mobility as an effect measure modifier and potential interaction effects