Multiple epidemiological studies imply that alcohol dependence is a complex behavioral disorder influenced by both genetic and environmental factors. Hence, to add information about environmental exposure to its genetic association designs is the appropriate strategy recommended for the following reasons. First, the knowledge of the interplay between genetic and environmental factors is essential for a complete understanding of the etiology of alcohol dependence. Second, accounting for heterogeneity of genetic effects due to interaction can be vital for the validity and enhanced power of the primary hypothesis in association testing. The most commonly cited environmental contributor to risk of alcohol dependence is age at first drink. This association replicated in many epidemiological studies suggests that underage drinking is linked to increased alcohol involvement and dependence. Twin studies have recently demonstrated that age at first drink moderates genetic influences on alcohol dependence. Therefore, age at first drink is an ideal candidate for an environmental modifier of genetic vulnerability in association studies. In the presence of a continuous environmental exposure, the investigator needs to carefully select the functional form of the association between the factor and the logarithm of the odds ratio of alcohol dependence. Otherwise, the misspecified main effect would lead to a dramatic inflation of Type 1 error rates in standard logistic regression tess of genetic associations. There is also overwhelming evidence that confirms the role of familial influences on age at first drink, which suggests the presence of gene-environment association. Hence tests that exploit gene-environment independence (i.e. case-only) and do not require that the main effect of the environment be specified, are susceptible to biased results and cannot be used. To relax the assumption about a parametric relationship between risk and exposure as well as reduce the dramatic effects of its misspecification, a nonparametric term is introduced into a model. Score testing approach is used to avoid numerical difficulty associated with parameter estimation under general nonparametric models. The resulting tests are straightforward to implement, they maintain the desired Type 1 error level. Moreover, the proposed tests gain substantial power when there is an interaction between genetic and environmental factors and lose little power when there is none. The proposed methodology will be applied to the data collected as part of Study of Addiction: Genetics and Environment (SAGE) available through the database of Genotypes and Phenotypes (dbGaP). However, flexibility of the proposed methodology as well as availability of the software will result in easy adaptation to other genetic association studies with a continuous environmental exposure.
The study is motivated by the problems of testing for genetic effects on alcohol dependence in the presence of gene-environment interaction when environmental exposure, such as age at first drink, is continuous. To relax the assumption about a parametric relationship between risk and exposure as well as reduce the dramatic effects of its misspecification, the nonparametric term is introduced into a model. Efficiency and flexibility of the proposed methodology along with the availability of the software compatible with popular genetic analysis packages will result in easy adaptation to other genetic association studies with a continuous environmental exposure.
