Publications Search

Abstract

Title: Do confounding or selection factors of residential wiring codes and magnetic fields distort findings of electromagnetic fields studies?
Author: Hatch EE, Kleinerman RA, Linet MS, Tarone RE, Kaune WT, Auvinen A, Baris D, Robison LL, Wacholder S
Journal: Epidemiology 11(2):189-198
Year: 2000
Month: March

Abstract: In contrast with several previous studies, our recent large case-control study found little association between childhood acute lymphoblastic leukemia (ALL) and electric-power-line wire codes. Here we examine internal evidence from our study to assess the possibility that selection bias and/or confounding may have affected the findings. We compared the relation between childhood ALL and wire codes and direct measurements of magnetic fields in subjects who participated in all phases of the study with the relation in all subjects, including those who declined to allow access inside the home. We found that the odds ratio for ALL among those living in homes with very high current configurations increased by 23% when 107 'partial participants' were excluded. We found similar, but slightly smaller, increases in the odds ratios when we performed the same comparisons using direct measurements of magnetic fields, excluding subjects who allowed only a measurement outside the front door. 'Partial participants' tended to be characterized by lower socioeconomic status than subjects who participated fully, suggesting possible selection bias. We also examined the relation between a large number of potential confounding variables and both proxy and direct measurements of magnetic fields. Univariate adjustment for individual variables changed the odds ratio for ALL by less than 8%, while simultaneous adjustment for several factors reduced the estimate by a maximum of 15%. We conclude that while confounding alone is unlikely to be an important source of bias in our own and previous studies of magnetic fields, selection bias may be more of a concern, particularly in light of the generally low response rates among controls in case-control studies.