TS3-06 Abstract: Although there is a substantial body of literature demonstrating the acute adverse health effects associated with air pollution, evidence regarding the effects of chronic exposure has been mainly limited to a few large cohort studies of mortality. The current study not only evaluates the relationship between particulate exposure and mortality, but also prospectively evaluates associations between chronic particulate exposures and cause-specific incident disease on a nationwide basis. These relationships are being examined using data from the Nurses’ Health Study, an ongoing prospective cohort study of 121,700 women residing throughout the United States. Incident cases of cardiovascular disease, respiratory disease, and cancers diagnosed during the study period are identified routinely through the biennial self-administered questionnaire. Cases are confirmed by supplemental questionnaire and review of medical records. Mortality is reported by next of kin and obtained by regular searches of the National Death Index. Information on potential confounders and effect modifiers is collected on the biennial questionnaire. Nurses’ residential addresses from 1988 through 2002 were geocoded and used as the basis for assigning exposure estimates. Exposure estimates were developed from predictive models of particulate matter less than 10 μm in diameter (PM10) and of PM2.5 for the years 1988 through 2002. Our air pollution prediction models for monthly average PM10 use data by monitoring site from U.S. EPA's Air Quality System (AQS) and various other sources, including the IMPROVE network and Harvard research studies. The modeling process also includes GIS-generated variables for block group, tract, and county-level population density; distance to nearest road, elevation, land use/cover, and meteorologic variables. All data were imported into generalized additive statistical models (GAMs) with smooth terms of space, time (separate surfaces for each month), and the GIS variables. The modeling approach combines a monthly pollution surface (the spatial component) and constant (time-invariant) effects of predictors derived from the GIS (the nonspatial component). As nationwide PM2.5 monitoring data are not available until 1999, our prediction models for PM2.5 use available PM2.5 monitoring data, corrected observations of horizontal visual range, and information from the PM10 predictive models to estimate monthly averages for 1988 through 2002. Currently, Cox proportional hazards models are being used to examine the associations of PM10 and PM2.5 with fatal and nonfatal myocardial infarction, strokes, and all-cause mortality. Potential confounders and effect modifiers such as smoking, physical activity, body mass index (BMI), physician-diagnosed hypertension and diabetes, menopausal status have been explored in the modeling process.