Magnetic fields and cancer.

In 1979 Wertheimer and Leeper developed their W-L wire code, based on power line construction, size, and distance from a residence, as a surrogate for residential magnetic field exposure (1). This W-L wire code was applied to data from a 1988 Denver study, and an association was reported between "high" W-L wire code and childhood cancer (2). The authors of "Childhood Cancer in Relation to a Modified Residential Wire Code" (3) use the 1988 Denver data and report a stronger association between their newly defined "high" wire code and childhood cancer. However, they report that this new wire code is not more strongly correlated with measured magnetic fields. (The authors state that both the W-L wire code and the new wire code can explain only 20% of the magnetic field variability.) A stronger association with childhood cancer without a correspondingly stronger correlation with magnetic fields suggests that factors other than magnetic fields may be involved in the reported wire code association with childhood cancer. The authors suggest that their new wire code is less susceptible to classification error than was present with the W-L wire code and that, had all of the 1988 Denver study homes been classified correctly, the W-L wire code would have produced a stronger association with childhood cancer. However, a "Back-to-Denver" (BTD) study, based on a subset of the 1988 Denver homes, found the W-L wire code classification error rate to be "quite small" (4). This finding appears to weaken the authors' argument that classification error resulted in an appreciable reduction in the association between W-L high wire code and childhood cancer in the 1988 Denver study. The authors of the BTD study conclude that factors other than magnetic fields may be responsible for the reported wire code association with childhood cancer (4). Explanations exist that suggest the reported stronger childhood cancer association with the new wire code may be false. For example, the selection criteria used in the original study (2) caused the case and control populations to be noncomparable with respect to residential mobility. Specifically, controls were required to be residentially stable from case diagnosis to selection as a control but cases were not. Thus, an artificial association between residential mobility and cancer was created by the subject selection procedure. This artificial association was shown to be a possible explanation for at least part of the reported association between W-L high wire code and childhood cancer (5). An even stronger bias from differential residential mobility might exist with the new wire code. This possibility could be tested by restricting the analysis to only those cases and controls that were residentially stable from diagnosis to matched control selection. Children that lived in homes in whlch magnetic field measurements were taken were residentially stable from diagnosis to control selection (2) and could provide the basis for such a comparison. Another possible explanation for a false association between wire codes and cancer is that individual cases may not have been matched with a control from the same neighborhood. In the 1988 Denver study children with the same telephone exchange as the case were eligible for selection as controls (2). The area corresponding to a particular telephone exchange likely does not correspond to a single neighborhood. For example, in Columbus, Ohio, the telephone exchange for the oldest neighborhood (developed in the 1800s) includes modern residential areas that were annexed to the city after 1970. Neighborhood differences between the cases and controls have the potential to significantly influence wire code associations because wire codes have been found to be strongly associated with the age of a neighborhood. For instance, 50% of the homes in one of the oldest neighborhoods (inner city) in Columbus were classified as W-L high wire code, while less than 25% of the homes in