Spatial environmental modeling of autoantibody outcomes among an African American population.

In this study of autoimmunity among a population of Gullah African Americans in South Carolina, the links between environmental exposures and autoimmunity (presence of antinuclear antibodies (ANA)) have been assessed. The study population included patients with systemic lupus erythematosus (n = 10), their first degree relatives (n = 61), and unrelated controls (n = 9) where 47.5% (n = 38) were ANA positive. This paper presents the methodology used to model ANA status as a function of individual environmental influences, both self-reported and measured, while controlling for known autoimmunity risk factors. We have examined variable dimension reduction and selection methods in our approach. Following the dimension reduction and selection methods, we fit logistic spatial Bayesian models to explore the relationship between our outcome of interest and environmental exposures adjusting for personal variables. Our analysis also includes a validation "strip" where we have interpolated information from a specific geographic area for a subset of the study population that lives in that vicinity. Our results demonstrate that residential proximity to exposure site is important in this form of analysis. The use of a validation strip network demonstrated that even with small sample numbers some significant exposure-outcome relationships can be detected.