Statistical Methods for Environmental Genetic Research

University Of Southern California, Los Angeles CA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): There is a substantial body of evidence indicating that air pollution is significantly associated with adverse health outcomes, mostly focusing on adults. There is also a growing interest in assessing the role of genetics in modulating the impact of the environment. In the CHS, one of the few longterm studies with a focus on children's health, the investigators adopted the multilevel modeling paradigm to show that air pollution leads to significant deficits in lung growth, increases in respiratory symptoms and school absenteeism, increases in asthma incidence in heavily exercising children in high ozone communities, and interacts with genetic factors. These results were obtained by using existing analytic techniques and by developing new methods that address the scientific questions of interest, where none existed. But important analyses still remain to be done due to lack of appropriate statistical techniques that handle exposure measurement error multicollinearity among pollutants, proliferation of possibly related outcomes and susceptible subgroups, and analysis of gene-environment interactions in this multilevel setup. Moreover, the new studies raise additional meteorologic issues that need to be addressed in order to examine lung growth trends into young adulthood (Project 1), effects of air pollution and genetics on asthma via a new cohort (Project 2), and effects of genes and geneenvironment interactions on children's respiratory health (Project 3). The main focus of this Project is to develop novel statistical methods that will help in integrating inferences across outcomes and methods for examining cause-and-effect relationships between lung function and respiratory symptom outcomes. In this way, a coherent story on the effects and interrelationships of genetic and environmental factors on children's health can be derived. Multilevel models that account for exposure measurement error will be developed to avoid possible attenuation of study results. Bayesian model averaging techniques will be developed in the multilevel setting to handle the multicolinearity problem in the highly correlated mix of pollutants in Southern California and also to enable summarization of evidence across subgroups in a systematic way. New techniques will be developed for genetic analysis of longitudinal data to explore main effects of candidate genes (along with gene-environment and gene-gene interactions) and toxicokinetic models for complex oxidative stress pathways that may mediate the effect of air pollution.