S1328 Examining Associations Between Arsenic Pollution and Hepatocellular Carcinoma in Texas

Abstract

Introduction: Alcohol is the major chemical risk factor for hepatocellular carcinoma (HCC) around the world, however other toxins including Arsenic have been shown to promote hepatocarcinogenesis in animals, though the exact mechanism is poorly understood. This ecological study assesses neighborhood-level HCC burden in Texas relative to arsenic exposure. Methods: Using data from the Texas Cancer Registry, we identified a cohort of individuals diagnosed with HCC between 2011 and 2015. The primary exposure of interest is Arsenic pollution as reported in the 2011 National Air Toxics Assessment (NATA) inventory, this national screening assessment by the Environmental Protection Agency (EPA) uses emissions data to estimate health risks from toxic air pollutants. NATA calculates the concentrations of toxic air pollutants at the census tract (neighborhood) level; the inhalation exposure concentrations of Arsenic are in units of micrograms per cubic meter, however, for analysis, exposure concentrations were divided into deciles. Arsenic concentrations, demographic data, and the Area Deprivation Index (composite measure of neighborhood socioeconomic disadvantage that relies on 17 census variables drawn from these categories: poverty, housing, employment, and education) were included in multivariable Poisson-based modeling using negative binomial regression to evaluate the association between Arsenic exposure and HCC incidence in Texas. Results: In a univariable model, the association between Arsenic inhalation exposure concentrations and HCC was not significant (IRR = 1.06 [95% CI, 0.98-1.16]). Whereas, in a multivariable model that included selected demographic and socioeconomic factors, results show that variation in census tract HCC incidence across Texas is significantly associated with the inhalation exposure concentrations of Arsenic. Based on our findings, in a typical Texas census tract, a 10-unit increase in decile classification of Arsenic inhalation exposure concentration increases the risk of HCC incidence by a factor of 1.30, while holding other explanatory variables constant (IRR = 1.30 [95% CI, 1.19-1.42]) (Table). Conclusion: Variation in HCC incidence across Texas’s census tracts is significantly associated with the inhalation exposure concentrations of Arsenic at the census tract level, higher Arsenic concentrations are associated with an increased incidence. This ecological finding needs to be further examined in direct association studies. Table 1. - Relationships between HCC incidence in Texas (2011 to 2015) and inhalation exposure concentrations of Arsenic (2011 estimate). Texas census tracts; N = 5,205 Univariable Models a Multivariable Model b IRR 95% CI p value IRR 95% CI p value Arsenic concentrations c 1.06 0.98-1.16 0.159 1.302 1.194-1.419 0.001 Area Deprivation Index 1.17 1.16-1.20 <0.001 1.618 1.444-1.813 <0.001 % Hispanic or Latino (NH) 1.08 1.07-1.09 <0.001 1.092 1.080-1.105 <0.001 % Non-Hispanic Asians 0.68 0.65-0.71 <0.001 0.892 0.853-0.934 <0.001 % Non-Hispanic African American 1.06 1.05-1.07 <0.001 1.098 1.080-1.116 <0.001 % Others; 2+ races d 0.58 0.54-0.63 <0.001 - % Population ≥ 60 y.o. 1.19 1.15-1.22 <0.001 1.406 1.359-1.455 <0.001 % Population male 1.05 1.00-1.11 0.064 1.171 1.117-1.226 <0.001 % Non-Hispanic White e 0.91 0.90-0.92 <0.001 aUnivariable models where Arsenic exposure concentrations was regressed on the HCC incidence separately. Also, each covariate was regressed on the HCC incidence separately.bMultivariable model where Arsenic concentrations and the covariates (P value ≤ 0.10) were regressed on the HCC incidence simultaneously.cThe 2011 National Air Toxics Assessment (NATA) inhalation exposure concentrations for Arsenic are in units of micrograms per cubic meter; for analysis, concentrations were divided into deciles.dVariable dropped from the multivariable model run because p value > 0.10.eTo avoid model overfitting from multicollinearity among race/ethnicity, Non-Hispanic White excluded from multivariable model.