A Generalized Weighted Quantile Sums Approach That Accounts for Interactions Between Highly Correlated Exposures and Other Factors

• Activity Number: 636 - Statistical Methods of Air Quality and Exposure
• Type: Contributed
• Date/Time: Thursday, August 2, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): MinJae Lee* and Maureen Samms-Vaughan and Jan Bressler and MacKinsey Christian and Manouchehr Hessabi and Megan Grove and Sydonnie Shakespeare-Pellington and Charlene Coore Desai and Jody-Ann Reece and Katherine Loveland and Eric Boerwinkle and Mohammad H. Rahbar
• Companies: University of Texas McGovern Medical School and The University of the West Indies and University of Texas School of Public Health at Houston and University of Texas School of Public Health at Houston and University of Texas Health Science Center at Houston and University of Texas School of Public Health at Houston and The University of the West Indies and The University of the West Indies and The University of the West Indies and University of Texas McGovern Medical School and University of Texas School of Public Health at Houston and University of Texas McGovern Medical School
• Keywords: Weighted Quantile Sum; Interaction; Correlated environmental exposures; Autism Spectrum Disorder (ASD)

Abstract

It is well known that the associations between environmental chemical exposures and health outcomes are complex multi-dimensional problems that are of great interest to public health researchers. To account for potentially high correlation among exposures that occur together, a weighted quantile sum (WQS) regression is used to assess the effects of mixtures of exposures on health outcomes. However, analysis of such environmental exposures data without consideration of synergistic interactions among exposures and with other factors could result in misleading findings. We propose a generalized WQS regression approach for handling correlated data by estimating a weighted index in which the weights of each exposure that may differ by level of other variables are simultaneously determined. Findings from our simulation study indicate that the proposed method estimates a weighted index that successfully identifies interactions under certain scenarios of data structures, while a traditional WQS method does not. We also demonstrate application of our method to real data from the Epidemiological Research on Autism Spectrum Disorder (ASD) in Jamaica (ERAJ) study.