A problem in identifying risk factors for disease using surrogate exposure variables that are under genetic control

The use of a surrogate exposure variable to represent a complex of genetic and/or nongenetic factors is commonplace in epidemiologic studies. The authors present an hypothetical example in which a surrogate exposure results from underlying unknown genetic and nongenetic factors, yet only the genetic component predisposes to disease. The results demonstrate how risk may be incorrectly attributed to the nongenetic component of exposure and suggest a possible explanation for the identification of a risk factor in one case-control study in one population, but not in another study of the same disease conducted in a different population.