Land use regression modeling of oxidative potential of fine particles, NO2, PM2.5 mass and association to type two diabetes mellitus. (December 2017)

• Record Type: Journal Article
• Authors: Hellack, Bryan; Sugiri, Dorothea; Schins, Roel P.F.; Schikowski, Tamara; Krämer, Ursula; Kuhlbusch, Thomas A.J.; Hoffmann, Barbara
• Is Part Of: Atmospheric environment. Volume 171(2017)
• Page Start: 181
• Page End: 190
• Publication Date: 2017-12
• Subjects: EPR -- Land use regression modeling -- LUR -- PM2.5 -- ROS -- Type 2 diabetes mellitus -- Oxidative potential; Air -- Pollution -- Periodicals; Air -- Pollution -- Meteorological aspects -- Periodicals; 551.51
• DOI: 10.1016/j.atmosenv.2017.10.017
• Languages: English
• ISSNs: 1352-2310

Abstract

Abstract: While land use regression models (LUR) are commonly used, e.g. for the prediction of spatially variable air pollutant mass concentrations, they are scarcely used for predicting the oxidative potential (OP), a suggested unifying predictor of health effects. Therefore a LUR model was developed to examine if long-term OP of fine particulate exposure can be reasonably predicted by LUR modeling and whether it is related to health effects in a study region comprised of urban and rural areas. Four 14-day sampling periods over 1 year at 40 sites in the western Ruhr Area and adjacent northern rural area, Germany, in 2002/2003 were conducted and annual Nitrogen Dioxide (NO2 ), fine particles (PM2.5 ), and OP were calculated. LUR models were developed to estimate spatially-resolved annual OP, NO2 and PM2.5 concentrations. The model performance was checked by leave-one-out cross validation (LOOCV) and cox regression was used to analyze the association of modeled residential OP and NO2 with incident type 2 diabetes mellitus (T2DM) in 1784 elderly women during a mean follow-up of 16 years (baseline 1985–1994). The measured OP and NO2 concentrations were moderately correlated (rSpearman 0.57). The LUR models explained 62% and 92% of the OP and NO2 variance (adjusted LOOCV R 2 57% and 90%). PM10 emission from combustion in a 5000 m buffer was the most important predictor for OP and NO2 . Modeled pollutants were highly correlated (rSpearman 0.87). Model quality for OP was sensitive to the inclusion of a single influential measurement site. For PM2.5 mass only an insufficient model with a low explained variance of 22% (adjusted R 2 ) was developed so no health effects analyses were conducted with estimated PM2.5 . Increases in OP and NO2 were associated with an increase in risk of T2DM by a hazard ratio of 1.38 (95% CI 1.06–1.80) and 1.39 (95% CI 1.07–1.81) per interquartile range of OP and NO2, respectively. We conclude that spatially-resolved OP can be predicted by LUR modeling, but future work is needed to investigate the possibility to increase OP model quality with refined predictors. Highlights: Development of a Land Use Regression model for the prediction of oxidative potential of PM2.5 Results show spatial variation between an urban and a rural area for the oxidative potential of PM2.5 Modeled oxidative potential of PM2.5 can be used in long-term health effects analysis. Modeled long-term oxidative potential and NO2 concentrations were associated with incident type 2 diabetes mellitus.