Structural Equation Modeling in a Near-Highway Exposure Study: Ultrafine Particles and Association with Cardiovascular Disease Risk

Abstract

Background: Structural equation modeling (SEM) is a statistical method that has not been used widely in epidemiological studies of ambient air pollution. Aims: Assess the association between individually-calculated exposure to UFP (<100 nm, estimated as particle number concentration or PNC) and cardiovascular disease risk. Methods: Air pollution data based on 55 days of mobile monitoring was used to build a regression model for assigning residential ambient hourly PNC for participants of the Community Assessment of Freeway Exposure and Health (CAFEH) study (N=140, a subset of CAFEH). Participants completed questionnaires on residential and occupational exposure, income, education, smoking, diet, physical activity and stress. Blood samples were analyzed for C-reactive protein (CRP), interleukin-6 (IL-6), tumor necrosis factor-? receptor II (TNF?); fibrinogen. Ankle brachial pressure index (ABI) as well as systolic, diastolic and pulse pressures were also measured. SEM was conducted using Proc Calis in SAS to determine path loadings, p-values between latent variables, and the individual variables contributing to the latent variables. Results: CRP, IL-6 and TNF? were not normally distributed and had population median values of 1.62mg/L for CRP; 1.61pg/mL for IL-6; and 2256pg/mL for TNF?. Fibrinogen was normally distributed with a mean of 475mg/dL. Initial SEM model development for the dependent latent variable construct cardiovascular disease risk produced the largest standardized estimates with the indicator variables CRP, TNF?, BMI, hypertension, and IL-6. PNC and the independent latent variable construct other combustion exposures displayed a positive association with cardiovascular disease risk, while and socio-economic status displayed a negative association. Conclusion: Initial analysis indicates that our model may have one endogenous latent variable, cardiovascular disease risk, and two or more exogenous latent variables related to air pollution and lifestyle.