Typically, in EMF epidemiological studies, no direct information was provided indicating that the two study groups were not comparable. It is simply expected that this will be the case, because the groups were not chosen using randomization. Occasionally, however, investigators collected information that explicitly showed the non-comparability of the groups. For example, in a study involving powerline EMFs and acute non-lymphocytic leukemia, the EMF and control groups differed with regard to the percentage of smokers, non-whites, and poor people. See R.K. Severson, R.G. Stevens, W.T. Kaune, D.B. Thomas, L. Heuser, S. Davis and L.E. Server: Am. J. Epidemiol. 128: 10, 1988. In a study of the link between powerline EMFs and acute lymphoblastic leukemia in children, a difference in economic status was reported. See M. Linet, E. Hotch, R.A. Kleinerman, L.L. Robison, W.T. Kaune, D.R. Friedman, R.K. Severson, C.M. Haines, C.T. Hartsock, S. Niwa, S. Wacholder and R.E. Tarone: Residential exposure to magnetic fields and acute lymphoblastic leukemia in children, N. Eng. J. Med. 337:1, 1997.

How can the absence of randomization perhaps explain finding an association? David Savitz, an epidemiologist at the University of North Carolina, reported in 1988 that exposure to powerline EMFs was associated with an increased risk of leukemia and brain cancer in children. The controls in that study were identified at random, based on telephone numbers. Charles Poole, a statistician at Boston University, theorized that the control group was defective because poor people were unlikely to have telephones. When Poole parsed Savitz’s data, Poole found associations indicating that his theory might be valid. If so, this would suggest that something other than EMFs might have resulted in the observed increase in risk for cancer. See D. Taubes: Epidemiology faces its limits, Science 269:164-169, 1995.