Latent class analysis to model multiple chemical exposures among children

Abstract

BackgroundChildren are exposed to multiple potentially harmful chemicals simultaneously. Efforts to understand the patterns and consequences of these exposures have been hampered by statistical limitations in estimations of higher order interactions. ObjectivesThe current study uses latent class analysis, a form of person-centered modeling to identify unobservable subgroups within populations and examine relationships between latent classes and measures of immune function. MethodsData from the National Health and Nutrition Examination Survey 2011–2012 were analyzed. A sample of 721 children aged 6–19 years were included who provided data on 47 chemicals of interest representing six chemical classes. Groups were identified using latent class analysis controlling for race/ethnicity, age, sex and poverty status. ResultsTwo alternative approaches to identifying latent classes each resulted in similar three class solutions, including one group of children characterized by low co-exposures across chemicals, a group with moderate co-exposure levels, and a group characterized by high co-occurring levels of polycyclic aromatic hydrocarbons, volatile organic compounds, phenols and phthalates. Under one of the approaches, latent classes were significantly associated with immune function as measured by lymphocyte and neutrophil counts. ConclusionsLatent class analysis offers a potential approach to measuring and understanding interactions among multiple co-occurring chemical stressors. However, additional work is needed to test the ability of latent classes to predict health variables.