A.C. Heath1, P.A.F. Madden,1 A. Todorov,1 & N.G. Martin2
Results from national twin studies in Sweden, Finland, Australia and the U.S.suggest that genetic factors have an important influence on the onset and course of smoking behavior. We review structural equation model-fitting (SEM) methods that have been used to test for genetic influences on smoking behavior, with illustrations using data from the 1981 survey of the Australian twin panel. The hypothesis that genetic and environmental factors that determine probability of becoming a regular smoker ( initiation') are statistically independent of genetic and environmental influences on probability of becoming a persistent long-term smoker ( persistence') can be rejected; but so also can the hypothesis of a single heritable liability dimension with persistent smokers higher in liability than successful quitters. Relaxation of the assumption of orthogonal liability dimensions can be achieved either via a mediating variable' model that allows for a partial regression of the persistence dimension on the initiation dimension, or via a combined' model under which some successful quitters are assumed to have low liability on the initiation' dimension. We show how tests for genetic effects using model-fitting methods can also be accomplished using a logistic regression model-fitting approach, using dummy variables to model cotwin's smoking status and zygosity. Indeed, when there is no zygosity difference in prevalence, the likelihood-ratio chi-square tests of the hypothesis of no genetic influence are identical in the two approaches. Such a regression approach is readily extended to a survival analysis framework, using Cox regression to model time to successful smoking cessation, and can be applied more easily than SEM to test for mediators of genetic influences on smoking persistence. Using data from the 1981 survey of the Australian twin panel, we show that high heritability of smoking persistence in men (71%) is not explained by associations with personality, education or other sociodemographic variables.
Address: 40 N. Kingshighway, Suite One, St Louis, MO 63108., Phone: 1-314-286-2206; Fax: 1-314-286-2213; Email: andrew@wupsych1.wuSTL.edu
1Dept. of Psychiatry, Washington University School of Medicine, St Louis, U.S.A. 2Division of Epidemiology and Population Health, Queensland Institute of Medical Research, Australia. 3Supported by NIH grants AA07728, CA75581 and DA00272.