Generalizability of Gene-Environment Models By Race/Ethnicity

The field of genetics is currently grappling with the issue of how to identify links between genes and behavior in the context of race/ethnicity (ancestry), given potential differences in the frequency of genetic markers of risk by ancestry. Genetic markers of risk for alcohol dependence, for example, are found at different rates in the various racial/ethnic groups due to differing histories of migration and assortative mating that are reflected in DNA. If the prevalence of target outcome behaviors also varies by race/ethnicity, there is a risk that any links between genetic markers and outcomes may be spurious unless race/ethnicity (genetic ancestry) is adequately modeled. Given this risk for spurious associations, many geneticists use Caucasian-only samples, which clearly is not an acceptable long-term solution. In addition, some genetic risk markers are sex-linked (e.g., MAOA is on the X chromosome), meaning that sex must be appropriately modeled. The social sciences have a long history of exploring and modeling racial/ethnic and gender differences. This presentation gives an example of the use of multiple group modeling to test the generalizability by race/ethnicity and gender of a model that is being used to understand gene-environment interplay in addiction.

Data were drawn from the Seattle Social Development Project (SSDP) a longitudinal study of the etiology of prosocial and antisocial behavior that began in 1985 and has followed a panel of 808 youth from ages 10-33. Prior analyses showed that general positive family environment during adolescence predicted shared variance in problem behavior (indicated by alcohol disorder, nicotine dependence, drug disorder, crime, risky sexual behavior) at age 24. Family smoking and drinking-specific environments predicted unique variance in nicotine dependence and alcohol disorder, respectively. Analyses for this paper build on the earlier model by testing generalizability of the model across Caucasian, African American, and Asian American racial/ethnic groups. Although some differences are observed when racial/ethnic groups and males and females are examined separately, results show that these differences are not statistically significant and that the model is generalizable across race/ethnicity and sex. Implications for testing gene-environment models with diverse samples will be discussed.