Abstract
BACKGROUND: Air pollution contains known human carcinogens. Prior studies evaluating an association between pollution exposure and central nervous system tumors (BT) are inconclusive. Using multiscale geographically-weighted linear regression (MGWR), we investigated the association between average county-level air pollution and geographic variation of glioblastoma and tumors of the meninges (ToM) incidence in the United States (US). MATERIAL AND METHODS: Annual county-level air pollution from 2000-2016 was obtained from the Environmental Protection Agency’s Air Quality Index (AQI) which analyzes daily ambient air concentration for ozone and particulate pollution in participating US counties. County-level annual age-adjusted incidence rates (AAAIRs) of glioblastoma and ToM per 100,000 population (>20 years old) between 2004-2020 were estimated using data from the Central Brain Tumor Registry of the United States, which aggregates data from the Centers for Disease Control and Prevention’s National Program of Cancer Registries and the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program. Counties with missing data were excluded. The 2016-2020 American Community Survey data were used to identify county sociodemographic and average exposure indoor/workplace pollution. Spatial patterns were visualized using Moran’s I measure and the global G-statistic analyses computed in R. Variables were mean normalized and regressed in a MGWR model using county centroids in the Spatial Analysis Research Center’s MGWR 2.2 program to assess geographically-varying effects of the independent variables. RESULTS: Data was available from 1,109 counties. Significant spatial clusters and outlier counties were identified for both tumor types (glioblastoma Moran’s I=0.126, p <0.001, G-statistic=0.003; ToM Moran’s I=0.487, p <0.001, G-statistic=0.003). When restricted to significantly high/low counties (p <0.05), low-rate clustering for glioblastoma was observed in Alaska, Arizona, Virginia, and parts of the central mid-west. High-rate clustering was observed in parts of the north, mid-west, and northeast. ToM showed low-rate clustering in northwest, Virginia, Kansas, and parts of the north mid-west, with high-rate clustering in Washington, Utah, Colorado, and Wisconsin. MGWR model showed that in both tumor types, increased AQI (glioblastoma mean βMGWR 0.06; MoT mean βMGWR 0.12) was associated with increased tumor incidence. This association was present in 45% and 59% of counties, respectively. CONCLUSION: This is the first study applying a MGWR model to assess geospatial associations between ambient air pollution and BT incidence, allowing for magnitude, direction, and significance of coefficients to change across geographic space. and allows for more accurate identification of geographic regions affected by potential BT health disparities.