Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Jun 1.
Published in final edited form as: Cancer Causes Control. 2018 Apr 25;29(6):563–572. doi: 10.1007/s10552-018-1036-x

Ambient PM2.5 air pollution exposure and hepatocellular carcinoma incidence in the United States

Trang VoPham 1,2,*, Kimberly A Bertrand 3, Rulla M Tamimi 1,2, Francine Laden 1,2,4, Jaime E Hart 2,4
PMCID: PMC5940508  NIHMSID: NIHMS963355  PMID: 29696510

Abstract

Purpose

To conduct the first epidemiologic study prospectively examining the association between particulate matter air pollution <2.5 μm in diameter (PM2.5) exposure and hepatocellular carcinoma (HCC) risk in the U.S.

Methods

Surveillance, Epidemiology, and End Results (SEER) provided information on HCC cases diagnosed between 2000 and 2014 from 16 population-based cancer registries across the U.S. Ambient PM2.5 exposure was estimated by linking the SEER county with a spatial PM2.5 model using a geographic information system (GIS). Poisson regression with robust variance estimation was used to calculate incidence rate ratios (IRRs) and 95% confidence intervals (CIs) for the association between ambient PM2.5 exposure per 10 μg/m3 increase and HCC risk adjusting for individual-level age at diagnosis, sex, race, year of diagnosis, SEER registry, and county-level information on health conditions, lifestyle, demographic, socioeconomic, and environmental factors.

Results

Higher levels of ambient PM2.5 exposure were associated with a statistically significant increased risk for HCC (n = 56,245 cases; adjusted IRR per 10 μg/m3 increase = 1.26, 95% CI 1.08, 1.47; p < 0.01).

Conclusions

If confirmed in studies with individual-level PM2.5 exposure and risk factor information, these results suggest that ambient PM2.5 exposure may be a risk factor for HCC in the U.S.

Keywords: PM2.5, particulate matter, air pollution, liver cancer, hepatocellular carcinoma, geographic information system

1. Introduction

Hepatocellular carcinoma (HCC) is the most common histological type of primary liver cancer, accounting for 85–90% of primary liver cancer cases [1, 2]. Risk factors for HCC vary by geography and include chronic hepatitis B virus (HBV) infection, chronic hepatitis C virus (HCV) infection, aflatoxin exposure, heavy alcohol consumption, smoking, obesity, and diabetes [3]. Although liver cancer incidence and mortality have been increasing in many regions around the world including the U.S. [4–6], an estimated 40.5% of HCC cases in the U.S. are unexplained by known risk factors including HCV, HBV, alcohol consumption, diabetes, and obesity [7]. Primary prevention of HCC is essential as the five-year relative survival rate remains low (<12%) [8].

Recent evidence suggests that exposure to fine particulate matter air pollution <2.5 μm in diameter (PM2.5) may increase the risk of liver cancer [9, 10]. PM2.5 is a ubiquitous environmental exposure produced from combustion sources such as motor vehicles and power plants [11]. PM, and outdoor air pollution in general, is classified as an International Agency for Research on Cancer (IARC) Group 1 human carcinogen largely based on evidence of positive associations with lung cancer in epidemiologic and experimental studies [12]. Humans are primarily exposed to PM2.5 via inhalation; its relatively finer particle size fraction allows for deposition deep in the lung (e.g., alveoli) through sedimentation and diffusion processes [13]. In addition, PM2.5 has been shown to induce oxidative damage, inflammation, and genotoxicity in the liver [14], promote HCC cell invasion and migration [15], and promote collagen deposition in the liver by activating TGF-β signaling [16]. Higher PM2.5 exposure has also been associated with reduced liver cancer survival [17] as well as higher prevalence of hepatic steatosis, a risk factor for HCC [18]. Two epidemiologic studies have examined the association between PM2.5 exposure and HCC incidence in Taiwan and Europe showing positive associations, although there were temporal mismatches as the date of case diagnoses preceded the exposure time periods [9, 10]. To date, no studies have been conducted in the U.S.

Although PM2.5 levels have decreased in the U.S. over the past two decades, PM2.5 remains an important environmental concern as urbanization continues to rise in parts of the U.S. and around the world such as in China [19–21]. An estimated 13.6% of the U.S. population resides in areas where PM2.5 concentrations exceed the 24-hour Environmental Protection Agency (EPA) National Ambient Air Quality Standards of 35 μg/m3 [22, 23]. Further, in 2017, an estimated 19.9 million (6.2% of the population) reside in locations with unhealthy levels of year-round particle pollution in the U.S. [24]. PM2.5 exposure is a modifiable risk factor that can be mitigated through reduced time spent outdoors at and during high-traffic locations and times [25]. The objective of this study was to prospectively examine the association between ambient PM2.5 exposure and HCC incidence in the U.S.

2. Methods

2.1 Study population

The Surveillance, Epidemiology, and End Results (SEER) database is a U.S. National Cancer Institute program collecting individual-level information on cancer incidence, survival, and treatment from population-based cancer registries covering 28% of the U.S. population [26]. The following 16 registries were included in the analysis: Atlanta (metropolitan); Greater California; Connecticut; Detroit (metropolitan); Greater Georgia; Iowa; Kentucky; Los Angeles; Louisiana (excluding July–December 2005 cases due to Hurricanes Katrina and Rita); New Jersey; New Mexico; Rural Georgia; San Francisco-Oakland; San Jose-Monterey; Seattle (Puget Sound); and Utah. The study area included all 607 counties located in the catchment areas captured by these 16 SEER registries that did not restrict coverage to specific populations and that were located in the 48 contiguous U.S. states with available PM2.5 exposure data. To protect patient confidentiality, the SEER database does not include personal identifiers. This study was exempt from Institutional Review Board review.

2.2 Outcomes

The following criteria were used to define HCC cases: International Classification of Diseases for Oncology, Third Edition (ICD-O-3) topography code C22.0 for primary liver cancer and ICD-O-3 histology codes 8170 to 8175 [27]; diagnostic confirmation (e.g., positive histology) excluding clinical diagnosis only [28]; sequence number of one primary only; diagnosis between 2000 and 2014; and not reported via autopsy or death certificate only [29]. As conducted in previous SEER-based epidemiologic studies, counts of HCC cases were stratified by age at diagnosis (<65 years, ≥65); sex (male, female); race (white, black, Asian/Pacific Islander/American Indian/Alaska Native); and year of diagnosis (2000–2007, 2008–2014) for each county [30–32]. Each county was associated with one SEER registry.

2.3 Exposure assessment

Ambient PM2.5 exposure was estimated for each county in the study area using a spatial PM2.5 exposure model. The model was created by applying inverse distance weighting (IDW) spatial interpolation to PM2.5 concentrations (μg/m3) in 2000 measured at 1,082 monitoring sites located across the contiguous U.S. provided by the U.S. Environmental Protection Agency (EPA) Air Quality System (AQS) database annual summary file [33, 34]. PM2.5 data from 2000 were selected as the first year cases were diagnosed was 2000; further, few PM2.5 monitoring data were available prior to 1999 [35]. IDW was used to create a spatial raster prediction surface of PM2.5 exposure levels for the entire contiguous U.S., where the predicted value for PM2.5 at any given location is the distance-weighted average of sample points (i.e., monitors) in a surrounding neighborhood [36]. The IDW neighborhood was defined using the 12 nearest monitors, resulting in an approximately 10 x 10 km spatial model, similar in spatial resolution to PM2.5 exposure models used in previous studies [37, 38]. In sensitivity analyses, visual inspection of prediction surfaces created using neighborhoods including between 5 and 20 monitors yielded similar results. Monitors outside of the contiguous U.S. were excluded from IDW modeling as they exert relatively minimal influence on PM2.5 concentrations at prediction points in the contiguous U.S. (the closest non-contiguous U.S. monitor was 899 km in distance to the nearest contiguous U.S. boundary) [39]. Using 2000 U.S. county boundaries [40], the PM2.5 model was aggregated to the county level in a geographic information system (GIS) (i.e., PM2.5 raster cell centroids intersecting a county were averaged to calculate a mean county PM2.5 value). Annual average ambient PM2.5 values were linked with each county in the study area, using the county at diagnosis that was available for each case from SEER. In secondary analyses, ambient PM2.5 exposure was estimated using the EPA AQS Annual Air Quality Statistics Report, which provided county-level PM2.5 estimates in 2000 for the weighted annual mean (weighted by the calendar quarter) [41]. PM2.5 estimates from the secondary analyses are available for a total of 687 counties (150 of the 607 counties in the study area; Supplementary Figure 1), while PM2.5 estimates from the primary IDW exposure metric are available for 3,109 counties (all 607 counties in the study area). All spatial analyses were conducted in ArcGIS 10.5.1 (Esri, Redlands, CA) using the contiguous U.S. Albers equal area conic coordinate system (NAD83 datum; USGS version).

2.4 Additional covariates

The following individual- and county-level information on known and suspected HCC risk factors and variables known to be associated with the exposure were evaluated as potential confounders. From the SEER database, we acquired individual-level data on age at diagnosis, sex, race, year of diagnosis, and SEER registry. We acquired the following county-level socioeconomic and demographic information from the 2000 U.S. Census Bureau Summary Files (SFs) that were available through SEER: educational attainment (percentage with a Bachelor’s degree or higher), poverty (percentage of individuals below the poverty level), percentage unemployed, median household income, and percentage foreign born (a proxy for HBV prevalence as HBV is endemic in parts of Asia and Africa [1]). We also acquired information on county-level urbanicity (a proxy for HCV prevalence as rural-urban differences in HCV have been observed [42]) using U.S. Department of Agriculture Rural-Urban Continuum Codes [43].

The following county-level data were acquired from the Institute for Health Metrics and Evaluation (IHME), created by applying small area models to data from the U.S. Behavioral Risk Factor Surveillance System and U.S. National Health and Nutrition Examination Survey: sex-specific age-adjusted prevalence of heavy alcohol consumption in 2005 (average >1 drink per day for women or >2 drinks per day for men in the past 30 days) [44]; diabetes in 2000 (percentage of adults aged ≥20 years who reported a previous diabetes diagnosis and/or have a fasting plasma glucose ≥126 mg/dL and/or hemoglobin A1c ≥6.5%) [45, 46]; physical activity in 2001 (participation during the past month in any physical activities/exercises outside of work); obesity in 2001 (body mass index [BMI] ≥30 kg/m2) [47]; and current smoking in 2000 (currently smoking daily or nondaily cigarettes) [48].

County-level age-adjusted drug poisoning-related mortality rates, defined using ICD-10 underlying cause-of-death codes X40-X44, X60-X64, X85, or Y10-Y14, were estimated using two-stage hierarchical models applied to the U.S. National Vital Statistics System multiple cause-of-death mortality files [49, 50]. Drug poisoning mortality was considered as a proxy for HCV prevalence as a substantial proportion of drug poisoning deaths are due to injection drug use, which is the predominant route of HCV transmission in the U.S. [51]. County-level population density (population per mi2), which is associated with higher traffic concentrations and thus other sources and types of air pollution, was downloaded from the 2000 U.S. Census Bureau SF1 [52, 53]. We created a variable for region of residence, which is associated with differences in PM2.5 chemical composition, by grouping all counties into the four U.S. Census Bureau-defined regions: Northeast, Midwest, South, and West [54, 55]. Vitamin D has been shown to be associated with a reduced risk for HCC [56]. Ultraviolet (UV) radiation exposure is the primary source of vitamin D for most individuals [57, 58]. We estimated ambient UV exposure using a spatiotemporal exposure model created by applying geostatistical methods to known predictors of UV including ozone, aerosol optical depth, cloud cover, and elevation [59]. UV raster cell centroids were intersected with county boundaries and aggregated to the county level using GIS. County-level data were compiled using unique U.S. Federal Information Processing Standard (FIPS) codes.

2.5 Statistical analysis

Poisson regression with robust variance estimation was used to calculate incidence rate ratios (IRRs) and 95% confidence intervals (CIs) for the association between ambient PM2.5 exposure and HCC risk. PM2.5 exposure was examined continuously per 10 μg/m3 increase. Restricted cubic regression splines were used to test for deviations from linearity. All models were a priori-determined to include the following variables on known and suspected HCC risk factors: age, sex, race, year, SEER registry, urbanicity, heavy alcohol consumption, smoking, obesity, diabetes, socioeconomic status (median household income, Bachelor’s degree education or higher, unemployment, and poverty), foreign born, and ambient UV exposure. Population density and SEER registry (used as a more granular variable for region) were also included in the final model to account for other types of outdoor air pollution and differences in PM2.5 constituents. We evaluated potential confounding by physical activity and drug poisoning mortality. As adjustment for these variables did not substantially change the IRR for PM2.5 exposure and HCC risk, they were not included in the final model. The natural logarithm of the population size, stratified by county, age (<65 years, ≥65), sex (male, female), race (white, black, Asian/Pacific Islander/American Indian/Alaska Native), and year (annual average from 2000–2007; annual average from 2008–2014) was used as the offset in all models. Population data were acquired from the National Center for Health Statistics (NCHS) Bridged-Race Resident Population Estimates 1990–2014 online database downloaded from the Centers for Disease Control and Prevention (CDC) WONDER [60].

Using stratified analyses, we explored potential effect modification by factors that may be associated with differential time spent outdoors and thus PM2.5 exposure, or associated with disparities in HCC incidence rates: age, sex, race, urbanicity, physical activity, obesity, diabetes, heavy alcohol consumption, and median household income. We further stratified by population density and region of residence, which influence ambient levels and/or chemical constituents of PM2.5, and by smoking, which has been associated with decreased lung deposition of PM2.5 and competition for metabolic activation [61]. To examine potential exposure misclassification due to migration, we stratified by residential mobility using data from the SEER-provided 2000 U.S. Census Bureau SF1 on the percentage of the county population that stayed in the same house (no migration from 1995–2000). Movers were defined as those residing in counties in which at least 51.9% (20th percentile of all counties) of the population did not migrate. We stratified by year of diagnosis to examine the effect of a potential exposure lag. Tests for interaction were conducted by adding an interaction term to the model and using likelihood ratio tests to determine statistical significance (p < 0.05). We performed sensitivity analyses using the secondary ambient PM2.5 exposure measure from the EPA AQS Annual Air Quality Statistics Report; using Poisson models with a random intercept for county to determine if there was potential county-level clustering; and using scaled Poisson models applying the Pearson and deviance methods to account for overdispersion [62]. All statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC).

3. Results

A total of n = 56,245 HCC cases diagnosed between 2000 and 2014 were included in the analysis. HCC cases were on average 62.4 years of age at diagnosis, mostly male (77.1%), white (68.5%), and/or resided in the Western region of the U.S. (61.5%) (Table 1). Using county-level data from the underlying population from which HCC cases were sampled, HCC cases at the time of diagnosis resided in counties where annual average ambient PM2.5 levels were 14.6 ± 3.1 μg/m3 (Table 1). HCC cases resided in counties where 23.9% of the population smoked cigarettes, 25.7% were obese (BMI≥30 kg/m2), 8.3% consumed a heavy amount of alcohol (i.e., an average of more than 1 drink per day for women and more than 2 drinks per day for men in the past 30 days), and 11.4% had diabetes. HCC cases resided in densely populated counties (average 1,750.9 population/mi2) that were characterized by an average median household income of $47,100, where less than 30% of the population had a Bachelor’s degree education or higher, 13.1% were living below the poverty level, and 17.9% were foreign born. Figure 1 shows annual average ambient PM2.5 exposure in 2000 categorized by quintiles calculated using all 607 counties included in the study. Annual average ambient PM2.5 levels ranged between 5.5 and 19.8 μg/m3. Higher PM2.5 concentrations were observed in the Southern U.S. (counties in the Kentucky and Georgia registries), as well as parts of the Northeast (New Jersey) and West (California), consistent with the higher levels typically observed in these regions during the 2000s [35].

Table 1.

Population characteristics of n = 56,245 hepatocellular carcinoma cases in the U.S. (SEER 2000–2014).

Characteristic Cases n (%)
Individual level
Age at diagnosis (years) (mean ± SD) 62.4 ± 11.6
Sex
 Male 43,357 (77.1)
 Female 12,888 (22.9)
Race
 White 38,546 (68.5)
 Black 7,737 (13.8)
 Asian or Pacific Islander 9,305 (16.5)
 American Indian or Alaskan Native 657 (1.2)
Region of residence at diagnosis
 Northeast 7,596 (13.5)
 South 9,995 (17.8)
 Midwest 4,084 (7.3)
 West 34,570 (61.5)
Year of diagnosis
 2000–2007 23,589 (41.9)
 2008–2014 32,656 (58.1)
County level
PM2.5 (μg/m3) (mean ± SD)a 14.6 ± 3.1
UV (mW/m2) (mean ± SD)a 214.4 ± 36.1
Percent heavy alcohol consumption (mean ± SD)a 8.3 ± 2.2
Percent smoking status (mean ± SD)a 23.9 ± 4.8
Percent physical activity (mean ± SD)a,b 76.9 ± 5.8
Percent obese (mean ± SD)a,b 25.7 ± 4.1
Percent diabetes (mean ± SD)a 11.4 ± 1.7
Median household income ($10,000) (mean ± SD)a 47.1 ± 11.1
Percent bachelor’s degree or higher (mean ± SD)a 26.1 ± 9.2
Percent unemployed (mean ± SD)a 6.5 ± 2.3
Percent poverty (mean ± SD)a 13.1 ± 5.3
Population density (population/mi2) (mean ± SD)a 1,750.9 ± 2,883.5
Percent foreign born (mean ± SD)a 17.9 ± 12.1
Urbanicitya
 Rural 460 (0.8)
 Urban 55,785 (99.2)
Open in a new tab

Abbreviations: HCC, hepatocellular carcinoma; PM2.5, particulate matter <2.5 μm; SD, standard deviation; SEER, Surveillance, Epidemiology, and End Results; UV, ultraviolet radiation.

a

County-level information was based on the county at diagnosis for cases from SEER.

b

Sex-specific physical activity and obesity prevalence rates were averaged to estimate a total prevalence.

Fig. 1.

Fig. 1

Open in a new tab

Ambient PM2.5 exposure (μg/m3) in 2000 by quintiles across 607 counties in the U.S. Higher PM2.5 levels are shown in darker colors. The PM2.5 quintiles were calculated using all 607 counties captured by the 16 SEER population-based cancer registries included in the study.

Higher ambient PM2.5 exposure was not associated with HCC risk in basic models adjusting for age, sex, race, year, and SEER registry (IRR = 0.99 per 10 μg/m3 increase, 95% CI 0.87, 1.14; p = 0.93) (Table 2). After additional adjustment for county-level heavy alcohol consumption, smoking, obesity, diabetes, population density, median household income, Bachelor’s degree education, unemployment, poverty, foreign born, urbanicity, and ambient UV exposure, we observed a statistically significant positive association between ambient PM2.5 exposure and HCC risk. A 10 μg/m3 increase in ambient PM2.5 exposure was associated with a 26% higher risk of HCC (adjusted IRR = 1.26, 95% CI 1.08, 1.47; p < 0.01) (Table 2). Model building is shown in Supplementary Table 1. The strongest confounders were population density, socioeconomic factors, and ambient UV.

Table 2.

Associations between ambient PM2.5 exposure and HCC incidence (SEER 2000–2014)

PM2.5 exposure (per 10 μg/m3 increase) Cases (n) Basica
IRR (95% CI)		 p Fully adjustedb
IRR (95% CI)		 p
Overall analysis
PM2.5 56,245 0.99 (0.87, 1.14) 0.93 1.26 (1.08, 1.47) <0.01
Stratified analyses
Population densityc,d 0.02
 Low 1,850 0.87 (0.60, 1.28)
 Medium 4,483 0.78 (0.58, 1.03)
 High 49,912 1.32 (1.11, 1.58)
Smokingc,e 0.03
 Low 43,209 1.23 (1.01, 1.51)
 Medium 7,031 0.97 (0.67, 1.40)
 High 6,005 1.19 (0.73, 1.96)
Residential mobilityf 0.02
 Non-mover 31,039 1.20 (1.01, 1.43)
 Mover 25,206 1.25 (0.96, 1.63)
Open in a new tab

Abbreviations: CI, confidence interval; HCC, hepatocellular carcinoma; IRR, incidence rate ratio; PM2.5, particulate matter <2.5 μm; SEER, Surveillance, Epidemiology, and End Results.

a

Adjusted for age at diagnosis, sex, race, year of diagnosis, and SEER registry.

b

Additionally adjusted for the following county-level variables: prevalence of heavy alcohol consumption, smoking, obesity, diabetes; population density; median household income; percentage with a Bachelor’s degree or higher; percentage unemployed; percentage of individuals below the poverty level; percentage foreign born; urbanicity; ambient UV exposure.

c

We stratified by tertiles of population density and sex-specific smoking prevalence across all 607 counties included in the analysis.

d

Low population density refers to <31.9 population/mi2. Medium population density refers to 31.9–83.3 population/mi2. High population density refers to >83.3 population/mi2.

e

Low smoking prevalence refers to cases residing in a county where <22.5% of the female population or <27.7% of the male population smoked cigarettes. Medium smoking prevalence refers to cases residing in a county where 22.5–25.7% of the female population or 27.7–31.7% of the male population smoked cigarettes. High smoking prevalence refers to cases residing in a county where >25.7% of the female population or >31.7% of the male population smoked cigarettes.

f

Non-movers were defined as individuals who resided in a county where ≥51.9% (20th percentile of all 607 counties) of the population stayed in the same home. Movers resided in a county where <51.9% of the population stayed in the same home.

We observed statistically significant interactions between ambient PM2.5 exposure and population density (p for interaction = 0.02), smoking (p = 0.03), and residential mobility (p = 0.02) (Table 2). PM2.5 exposure was positively associated with HCC risk in areas with high population density (adjusted IRR = 1.32, 95% CI 1.11, 1.58), but was not associated with HCC risk in areas with low/medium population density. PM2.5 exposure was positively associated with HCC risk in areas with low smoking prevalence (adjusted IRR = 1.23, 95% CI 1.01, 1.51). However, the sample size in areas with low/medium population density or low/medium smoking prevalence were relatively smaller than in high population density or high smoking prevalence areas. In general, the association between PM2.5 and HCC risk was positive across all strata defined by residential mobility. The association between PM2.5 and HCC risk did not vary by age, sex, race, urbanicity, physical activity, obesity, diabetes, heavy alcohol consumption, median household income, and region (p > 0.05). In sensitivity analyses among n = 48,187 HCC cases with PM2.5 exposure information available from the EPA AQS Annual Air Quality Statistics Report, we did not observe an association between PM2.5 exposure and HCC risk (adjusted IRR = 1.01, 95% 0.88, 1.16) (Supplementary Table 2). We observed a suggestive positive association between PM2.5 exposure and HCC risk among cases diagnosed between 2000 and 2007 (adjusted IRR = 1.21, 95% CI 0.98, 1.48); and a statistically significant positive association among cases diagnosed between 2008 and 2014 (adjusted IRR = 1.30, 95% CI 1.13, 1.50) (Supplementary Table 3). Similar results were observed when using Poisson regression with a random intercept for county and scaled Poisson models applying either the Pearson and deviance methods (results not shown).

4. Discussion

We observed a statistically significant positive association between county-level ambient PM2.5 exposure and HCC risk in the U.S. after adjustment for individual-level age at diagnosis, sex, race, year of diagnosis, and SEER registry, and county-level information on health conditions, lifestyle, demographic, socioeconomic, and environmental factors. To the best of our knowledge, this is the first study examining the association between ambient PM2.5 exposure and HCC risk in the U.S.

The results of this study are consistent with the growing body of literature suggesting that exposure to PM2.5 impacts hepatocarcinogenesis. Two epidemiologic studies examining individual-level PM2.5 exposure (based on geocoded residential addresses) and liver cancer risk have shown generally positive associations. In a prospective cohort study of 464 HCC cases in Taiwan, higher levels of PM2.5 exposure were associated with a statistically significant increased risk of HCC in the Taiwan Penghu Islands (HR per 13.1 μg/m3 increase = 1.22, 95% CI 1.02, 1.47) after adjusting for age, sex, hepatitis B surface antigen, hepatitis C antibody, alanine transaminase, alcohol consumption, and smoking [9]. In a prospective cohort study as part of the European Study of Cohorts for Air Pollution Effects (ESCAPE) project that included 279 primary liver cancer cases, higher PM2.5 exposure was associated with a suggestive but not statistically significant positive association with liver cancer (HR per 5 μg/m3 increase = 1.34, 95% CI 0.76, 2.35) after adjusting for age, sex, smoking, alcohol consumption, occupational exposure, employment status, education, and area-level socioeconomic status [10]. However, neither study considered potential confounding by diabetes, which has been associated with both PM2.5 and liver cancer [63, 64]. The ESCAPE study also did not adjust for chronic HBV or HCV infection as potential confounders, although the authors noted that confounding by hepatitis may be unlikely as correlates of HBV and HCV were adjusted for in the models (e.g., education), and prevalence of these viruses is low in the study area. Further, there was a temporal mismatch as PM2.5 exposure information was available from 2006–2009 for HCC cases diagnosed between 1991 and 2009 in Pan et al. [9], and from 2009–2011 for liver cancer cases diagnosed between 1985 and 2012 in Pedersen et al. [10]. The PM2.5 exposure assessments used in these studies may not represent long-term exposure relevant to hepatocarcinogenesis, making the interpretation of findings challenging.

In this study, we observed a statistically significant positive association between ambient PM2.5 exposure and HCC risk. These results are consistent, in direction and magnitude, with previous research showing positive or suggestively positive associations with liver cancer [9, 10]. We developed an objective measure of ambient PM2.5 exposure, linking a spatial PM2.5 model, created using PM2.5 concentrations measured at EPA monitors spanning the contiguous U.S., with the SEER county at diagnosis for each HCC case. We estimated PM2.5 exposure in 2000 as few PM2.5 monitoring data were available prior to 1999 [35], which would have provided relatively fewer data points for spatial interpolation of PM2.5 before 1999. In addition, 2000 was the first year cases were diagnosed in our study. HCC has been associated with latency periods of up to 20 years; thus, we were primarily interested in estimating historical exposure that may be more relevant to hepatocarcinogenesis compared to recent exposure. We assumed that the county at diagnosis represented the location where cases resided during the period relevant to the development of liver cancer (prior to diagnosis), which is supported by observing that cases in our study resided in counties in which the majority of residents did not move from 1995–2000. As expected, compared to cases diagnosed between 2000 and 2007, we observed a stronger positive association between PM2.5 exposure and HCC risk among cases diagnosed between 2008 and 2014 (characterized by a potential exposure lag of between 8 and 14 years). Although results in a sensitivity analysis using a PM2.5 exposure measure from the EPA AQS Annual Air Quality Statistics Report were null, 457 of the 607 counties in the study were missing these exposure data, resulting in excluding over 14% of HCC cases (many of whom resided in high-exposure areas), and reduced exposure variability.

In this analysis, the strongest confounders were population density, socioeconomic factors, and UV, which have been associated with liver cancer. Higher population density has been associated with increased liver cancer risk, and higher socioeconomic status and UV and vitamin D levels have been associated with decreased liver cancer risk [32, 56, 65, 66]. PM2.5 concentrations are higher in areas with higher population density, related to sources of PM2.5 including combustion from motor vehicles and other anthropogenic sources [12]. Higher socioeconomic status and the ability to choose whether or not to reside near highways and/or improve air quality has generally been associated with lower PM2.5 levels [67]. UV is inversely associated with PM2.5 as PM2.5 particles absorb and scatter UV wavelengths, reducing surface UV levels [68].

We observed statistically significant interactions between PM2.5 exposure and population density, smoking, and residential mobility. However, the sample size in low/medium population density areas and low/medium smoking prevalence areas is smaller compared to areas with high population density or high smoking prevalence, and the stratified estimates for residential mobility were generally similar and positive. PM2.5 exposure was positively associated with HCC risk in areas with high population density, which have been characterized by higher traffic intensity, decreased driving speeds, and increased emissions, suggesting that densely populated areas may be associated with higher emissions per vehicle [69]. PM2.5 exposure was positively associated with HCC risk in areas with low smoking prevalence. Although the mechanisms of action may differ between lung vs. liver cancer, several studies have demonstrated stronger adverse effects of PM2.5 on lung cancer risk among never-smokers and former smokers compared to current smokers, possibly due to competition for metabolic activation [61, 70].

Limitations of this study include lack of personal PM2.5 exposure information and exposure misclassification associated with using the county at diagnosis provided by SEER. We also lacked information on individual-level factors affecting PM2.5 exposure including time spent outdoors. Further, the ecological fallacy cannot be ruled out, as the association between area-level PM2.5 and HCC may not represent the individual-level association between PM2.5 and HCC risk. However, we applied established spatial interpolation methods to estimate a PM2.5 exposure model using a nationwide network of EPA monitors. Ambient PM2.5 exposure was assessed using the same methods across all counties in the study. Further, spatially interpolated PM2.5 prediction surfaces of similar spatial resolutions, as well as spatiotemporal exposure models created using the EPA monitors used in our study, have been predictive of cancer risk in previous epidemiologic studies [35, 71–73]. Similar positive associations for PM2.5 and liver cancer were observed in individual-level epidemiologic studies [9, 10]. Although we did not have information on residential history and could not estimate long-term historical PM2.5 exposure, cases lived in counties where the majority of residents stayed in the same home between 1995 and 2000, and results were similarly positive after stratifying by county-level residential mobility. Residual confounding due to lack of information on individual-level risk factors for HCC, such as alcohol consumption and chronic HCV infection, is possible. However, we were able to adjust for individual-level age, sex, race, year of diagnosis, and SEER registry, as well as county-level information on known and suspected HCC risk factors, including heavy alcohol consumption, smoking, obesity, diabetes, and socioeconomic factors. We also adjusted for urbanicity as a proxy for HCV prevalence and percentage of foreign-born individuals as a proxy for HBV prevalence in our analyses. HBV may not be a strong confounder of the association as it was not associated with PM2.5 in previous research [9]. Although areas with a high prevalence of HCV, which is the major risk factor for liver cancer in the U.S., have coincided with areas characterized by high PM2.5 levels, our results were adjusted for population density and socioeconomic factors, which are associated with HCV prevalence [9, 42]. We also adjusted for county-level obesity and diabetes, major risk factors for HCC in the U.S. that have exhibited rural/urban variations and associations with PM2.5 [74, 75].

Strengths of our study include the large sample size of confirmed HCC cases from SEER population-based cancer registries covering a substantial proportion of the U.S. population. We conducted an objective location-based exposure assessment utilizing a spatial PM2.5 model incorporating PM2.5 concentrations measured at over 1,000 U.S. EPA AQS monitors. The study area includes counties located across the contiguous U.S. characterized by a wide range of PM2.5 levels. We also evaluated potential confounding and effect modification using individual- and county-level information from many objective data sources including SEER and the U.S. Census Bureau.

In conclusion, results from the first prospective analysis in the U.S. suggest that higher ambient PM2.5 exposure may be an important risk factor for HCC in this country. Future research examining this association using long-term personal PM2.5 exposure measures and individual-level HCC risk factors is warranted to confirm these findings.

Supplementary Material

Supplementary Material

Acknowledgments

The authors would like to acknowledge the U.S. National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) program, Institute for Health Metrics and Evaluation (IHME), Environmental Protection Agency (EPA), Census Bureau, Minnesota Population Center, Centers for Disease Control and Prevention (CDC), and National Center for Health Statistics (NCHS) for providing the publicly available data used in this study. This research was supported by the U.S. National Institutes of Health (NIH) NCI Training Program in Cancer Epidemiology (T32 CA009001).

Footnotes

Compliance with ethical standards

Conflict of interest: The authors declare that they have no conflicts of interest.

Ethical approval: To protect patient confidentiality, the SEER database does not include personal identifiers. This study was exempt from Institutional Review Board review. For this type of study, formal consent is not required.

References

  • 1.Carr BI. Hepatocellular Carcinoma: Diagnosis and Treatment. 3. Springer International Publishing; Cham, Switzerland: 2016. [Google Scholar]
  • 2.El–Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology and molecular carcinogenesis. Gastroenterol. 2007;132(7):2557–2576. doi: 10.1053/j.gastro.2007.04.061. [DOI] [PubMed] [Google Scholar]
  • 3.McGlynn KA, Petrick JL, London WT. Global epidemiology of hepatocellular carcinoma: an emphasis on demographic and regional variability. Clin Liver Dis. 2015;19(2):223–238. doi: 10.1016/j.cld.2015.01.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Petrick JL, Braunlin M, Laversanne M, Valery PC, Bray F, McGlynn KA. International trends in liver cancer incidence, overall and by histologic subtype, 1978–2007. Int J Cancer. 2016;139(7):1534–1545. doi: 10.1002/ijc.30211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Wong MC, Jiang JY, Goggins WB, Liang M, Fang Y, Fung FD, Leung C, Wang HH, Wong GL, Wong VW, Chan HL. International incidence and mortality trends of liver cancer: a global profile. Sci Rep. 2017;7:45846. doi: 10.1038/srep45846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Global Burden of Disease Liver Cancer C. Akinyemiju T, Abera S, Ahmed M, Alam N, Alemayohu MA, Allen C, Al-Raddadi R, Alvis-Guzman N, Amoako Y, Artaman A, Ayele TA, Barac A, Bensenor I, Berhane A, Bhutta Z, Castillo-Rivas J, Chitheer A, Choi JY, Cowie B, Dandona L, Dandona R, Dey S, Dicker D, Phuc H, Ekwueme DU, Zaki MES, Fischer F, Furst T, Hancock J, Hay SI, Hotez P, Jee SH, Kasaeian A, Khader Y, Khang YH, Kumar GA, Kutz M, Larson H, Lopez A, Lunevicius R, Malekzadeh R, McAlinden C, Meier T, Mendoza W, Mokdad A, Moradi-Lakeh M, Nagel G, Nguyen Q, Nguyen G, Ogbo F, Patton G, Pereira DM, Pourmalek F, Qorbani M, Radfar A, Roshandel G, Salomon JA, Sanabria J, Sartorius B, Satpathy M, Sawhney M, Sepanlou S, Shackelford K, Shore H, Sun J, Mengistu DT, Topor-Madry R, Tran B, Ukwaja KN, Vlassov V, Vollset SE, Vos T, Wakayo T, Weiderpass E, Werdecker A, Yonemoto N, Younis M, Yu C, Zaidi Z, Zhu L, Murray CJL, Naghavi M, Fitzmaurice C. The Burden of Primary Liver Cancer and Underlying Etiologies From 1990 to 2015 at the Global, Regional, and National Level: Results From the Global Burden of Disease Study 2015. JAMA oncology. 2017 doi: 10.1001/jamaoncol.2017.3055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Makarova-Rusher OV, Altekruse SF, McNeel TS, Ulahannan S, Duffy AG, Graubard BI, Greten TF, McGlynn KA. Population attributable fractions of risk factors for hepatocellular carcinoma in the United States. Cancer. 2016;122(11):1757–1765. doi: 10.1002/cncr.29971. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.El-Serag HB. Hepatocellular carcinoma. The New England journal of medicine. 2011;365(12):1118–1127. doi: 10.1056/NEJMra1001683. [DOI] [PubMed] [Google Scholar]
  • 9.Pan WC, Wu CD, Chen MJ, Huang YT, Chen CJ, Su HJ, Yang HI. Fine Particle Pollution, Alanine Transaminase, and Liver Cancer: A Taiwanese Prospective Cohort Study (REVEAL-HBV) J Natl Cancer Inst. 2016;108(3) doi: 10.1093/jnci/djv341. [DOI] [PubMed] [Google Scholar]
  • 10.Pedersen M, Andersen ZJ, Stafoggia M, Weinmayr G, Galassi C, Sørensen M, Eriksen KT, Tjønneland A, Loft S, Jaensch A. Ambient air pollution and primary liver cancer incidence in four European cohorts within the ESCAPE project. Environ Res. 2017;154:226–233. doi: 10.1016/j.envres.2017.01.006. [DOI] [PubMed] [Google Scholar]
  • 11.Environmental Protection Agency. [Accessed 5 Jun 2017];AirNow. 2017 https://airnow.gov/index.cfm?action=airnow.mapcenter&mapcenter=1#tabs-6.
  • 12.International Agency for Research on Cancer. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Vol. 109. WHO Press; Geneva, Switzerland: 2016. Outdoor Air Pollution. [Google Scholar]
  • 13.Falcon-Rodriguez CI, Osornio-Vargas AR, Sada-Ovalle I, Segura-Medina P. Aeroparticles, Composition, and Lung Diseases. Front Immunol. 2016;7:3. doi: 10.3389/fimmu.2016.00003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kim JW, Park S, Lim CW, Lee K, Kim B. The role of air pollutants in initiating liver disease. Toxicol Res. 2014;30(2):65. doi: 10.5487/TR.2014.30.2.065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Zhang Q, Luo Q, Yuan X, Chai L, Li D, Liu J, Lv Z. Atmospheric particulate matter2. 5 promotes the migration and invasion of hepatocellular carcinoma cells. Oncol Lett. 2017;13(5):3445–3450. doi: 10.3892/ol.2017.5947. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Zheng Z, Zhang X, Wang J, Dandekar A, Kim H, Qiu Y, Xu X, Cui Y, Wang A, Chen LC. Exposure to fine airborne particulate matters induces hepatic fibrosis in murine models. J Hepatol. 2015;63(6):1397–1404. doi: 10.1016/j.jhep.2015.07.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Deng H, Eckel SP, Liu L, Lurmann FW, Cockburn MG, Gilliland FD. Particulate matter air pollution and liver cancer survival. Int J Cancer. 2017;141(4):744–749. doi: 10.1002/ijc.30779. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Li W, Dorans KS, Wilker EH, Rice MB, Long MT, Schwartz J, Coull BA, Koutrakis P, Gold DR, Fox CS, Mittleman MA. Residential Proximity to Major Roadways, Fine Particulate Matter, and Hepatic Steatosis: The Framingham Heart Study. Am J Epidemiol. 2017;186(7):857–865. doi: 10.1093/aje/kwx127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.World Health Organization. Global Urban Ambient Air Pollution Database. 2016. [Google Scholar]
  • 20.Straif K, Cohen A, Samet J. IARC scientific publication. 2013. Air pollution and cancer; p. 161. [Google Scholar]
  • 21.Han L, Zhou W, Li W. Fine particulate (PM2.5) dynamics during rapid urbanization in Beijing, 1973–2013. Sci Rep. 2016;6:23604. doi: 10.1038/srep23604. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Yip FY, Pearcy JN, Garbe PL, Truman BI Control CfD, Prevention. Unhealthy air quality—United States, 2006–2009. MMWR Surveill Summ. 2011;60(Suppl):28–32. [PubMed] [Google Scholar]
  • 23.Correia AW, Pope CA, 3rd, Dockery DW, Wang Y, Ezzati M, Dominici F. Effect of air pollution control on life expectancy in the United States: an analysis of 545 U.S. counties for the period from 2000 to 2007. Epidemiology. 2013;24(1):23–31. doi: 10.1097/EDE.0b013e3182770237. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.American Lung Association. [Accessed 6 Jun 2017];State of the Air 2017. 2017 http://www.lung.org/assets/documents/healthy-air/state-of-the-air/state-of-the-air-2017.pdf.
  • 25.Laumbach R, Meng Q, Kipen H. What can individuals do to reduce personal health risks from air pollution? J Thorac Dis. 2015;7(1):96–107. doi: 10.3978/j.issn.2072-1439.2014.12.21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.National Cancer Institute. [Accessed 3 Apr 2017];Surveillance, Epidemiology, and End Results Program. 2017 https://seer.cancer.gov/
  • 27.Fritz A, Percy C, Shanmugaratnam K, Sobin L, Parkin DM, Whelan S. International Classification of Diseases for Oncology: ICD-O. 3. World Health Organization; Geneva, Switzerland: 2000. [Google Scholar]
  • 28.Davila JA, Morgan RO, Shaib Y, McGlynn KA, El-Serag HB. Diabetes increases the risk of hepatocellular carcinoma in the United States: a population based case control study. Gut. 2005;54(4):533–539. doi: 10.1136/gut.2004.052167. doi: http://dx.doi.org/10.1136/gut.2004.052167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.VoPham T, Brooks MM, Yuan JM, Talbott EO, Ruddell D, Hart JE, Chang CC, Weissfeld JL. Pesticide exposure and hepatocellular carcinoma risk: A case-control study using a geographic information system (GIS) to link SEER-Medicare and California pesticide data. Environ Res. 2015;143(Pt A):68–82. doi: 10.1016/j.envres.2015.09.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Cahoon EK, Pfeiffer RM, Wheeler DC, Arhancet J, Lin SW, Alexander BH, Linet MS, Freedman DM. Relationship between ambient ultraviolet radiation and non-Hodgkin lymphoma subtypes: A US population-based study of racial and ethnic groups. Int J Cancer. 2015;136(5):E432–E441. doi: 10.1002/ijc.29237. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Bowen EM, Pfeiffer RM, Linet MS, Liu WT, Weisenburger DD, Freedman DM, Cahoon EK. Relationship between ambient ultraviolet radiation and Hodgkin lymphoma subtypes in the United States. Br J Cancer. 2016;114(7):826–831. doi: 10.1038/bjc.2015.383. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.VoPham T, Bertrand KA, Yuan JM, Tamimi RM, Hart JE, Laden F. Ambient ultraviolet radiation exposure and hepatocellular carcinoma incidence in the United States. Environ Health. 2017;16(1):89. doi: 10.1186/s12940-017-0299-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Environmental Protection Agency. [Accessed 7 May 2017];AirData Download Files Documentation. 2017 http://aqsdr1.epa.gov/aqsweb/aqstmp/airdata/FileFormats.html.
  • 34.Roberts JD, Voss JD, Knight B. The association of ambient air pollution and physical inactivity in the United States. PLoS One. 2014;9(3):e90143. doi: 10.1371/journal.pone.0090143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Yanosky JD, Paciorek CJ, Laden F, Hart JE, Puett RC, Liao D, Suh HH. Spatio-temporal modeling of particulate air pollution in the conterminous United States using geographic and meteorological predictors. Environ Health. 2014;13(1):63. doi: 10.1186/1476-069X-13-63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.O’Sullivan D, Unwin D. Geographic information analysis. 1. John Wiley & Sons; Hoboken, NJ: 2014. [Google Scholar]
  • 37.Kloog I, Koutrakis P, Coull BA, Lee HJ, Schwartz J. Assessing temporally and spatially resolved PM 2.5 exposures for epidemiological studies using satellite aerosol optical depth measurements. Atmos Environ. 2011;45(35):6267–6275. [Google Scholar]
  • 38.Kloog I, Ridgway B, Koutrakis P, Coull BA, Schwartz JD. Long- and short-term exposure to PM2.5 and mortality: using novel exposure models. Epidemiology. 2013;24(4):555–561. doi: 10.1097/EDE.0b013e318294beaa. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Li L, Losser T, Yorke C, Piltner R. Fast inverse distance weighting-based spatiotemporal interpolation: a web-based application of interpolating daily fine particulate matter pm2. 5 in the contiguous us using parallel programming and kd tree. Int J Environ Res Public Health. 2014;11(9):9101–9141. doi: 10.3390/ijerph110909101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Minnesota Population Center. [Accessed 15 Feb 2017];National Historical Geographic Information System: Version 11.0 [Database] 2016 http://doi.org/10.18128/D050.V11.0.
  • 41.Environmental Protection Agency. [Accessed 2 Jun 2017];Outdoor Air Quality Data. 2017 https://www.epa.gov/outdoor-air-quality-data/about-air-data-reports.
  • 42.Kauhl B, Heil J, Hoebe CJ, Schweikart J, Krafft T, Dukers-Muijrers NH. The spatial distribution of hepatitis C virus infections and associated determinants—An application of a geographically weighted poisson regression for evidence-based screening interventions in hotspots. PloS One. 2015;10(9):e0135656. doi: 10.1371/journal.pone.0135656. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.National Cancer Institute. [Accessed 3 Apr 2017];Surveillance, Epidemiology, and End Results: County Attributes. 2017 https://seer.cancer.gov/seerstat/variables/countyattribs/
  • 44.Dwyer-Lindgren L, Flaxman AD, Ng M, Hansen GM, Murray CJ, Mokdad AH. Drinking patterns in US counties from 2002 to 2012. Am J Public Health. 2015;105(6):1120–1127. doi: 10.2105/AJPH.2014.302313. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Institute for Health Metrics and Evaluation. Diagnosed and Undiagnosed Diabetes Prevalence by County in the U.S., 1999–2012. Institute for Health Metrics and Evaluation; 2016. [Google Scholar]
  • 46.Dwyer-Lindgren L, Mackenbach JP, van Lenthe FJ, Flaxman AD, Mokdad AH. Diagnosed and Undiagnosed Diabetes Prevalence by County in the US, 1999–2012. Diabetes Care. 2016;39(9):1556–1562. doi: 10.2337/dc16-0678. [DOI] [PubMed] [Google Scholar]
  • 47.Dwyer-Lindgren L, Freedman G, Engell RE, Fleming TD, Lim SS, Murray CJ, Mokdad AH. Prevalence of physical activity and obesity in US counties, 2001–2011: a road map for action. Popul Health Metr. 2013;11(1):7. doi: 10.1186/1478-7954-11-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Dwyer-Lindgren L, Mokdad AH, Srebotnjak T, Flaxman AD, Hansen GM, Murray CJ. Cigarette smoking prevalence in US counties: 1996–2012. Popul Health Metr. 2014;12(1):5. doi: 10.1186/1478-7954-12-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Centers for Disease Control and Prevention. [Accessed 23 Feb 2017];Drug Poisoning Mortality: United States, 1999–2014. 2016 https://blogs.cdc.gov/nchs-data-visualization/drug-poisoning-mortality/
  • 50.Rossen LM, Khan D, Warner M. Trends and geographic patterns in drug-poisoning death rates in the US, 1999–2009. Am J Prev Med. 2013;45(6):e19–e25. doi: 10.1016/j.amepre.2013.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Rudd RA, Aleshire N, Zibbell JE, Matthew Gladden R. Increases in drug and opioid overdose deaths—United States, 2000–2014. Am J Transplant. 2016;16(4):1323–1327. doi: 10.15585/mmwr.mm6450a3. [DOI] [PubMed] [Google Scholar]
  • 52.U.S. Census Bureau. 2000 Census Summary File 1-United States [machine-readable data files]/prepared by the U.S. Census Bureau. 2000 [Google Scholar]
  • 53.Brauer M, Hoek G, van Vliet P, Meliefste K, Fischer P, Gehring U, Heinrich J, Cyrys J, Bellander T, Lewne M. Estimating long-term average particulate air pollution concentrations: application of traffic indicators and geographic information systems. Epidemiol. 2003;14(2):228–239. doi: 10.1097/01.EDE.0000041910.49046.9B. [DOI] [PubMed] [Google Scholar]
  • 54.U.S. Census Bureau. [Accessed 1 May 2017];Census Regions and Divisions of the United States. 2017 https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf.
  • 55.Bell ML, Dominici F, Ebisu K, Zeger SL, Samet JM. Spatial and temporal variation in PM(2.5) chemical composition in the United States for health effects studies. Environ Health Perspect. 2007;115(7):989–995. doi: 10.1289/ehp.9621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Fedirko V, Duarte-Salles T, Bamia C, Trichopoulou A, Aleksandrova K, Trichopoulos D, Trepo E, Tjønneland A, Olsen A, Overvad K. Prediagnostic circulating vitamin D levels and risk of hepatocellular carcinoma in European populations: A nested case-control study. Hepatology. 2014;60(4):1222–1230. doi: 10.1002/hep.27079. [DOI] [PubMed] [Google Scholar]
  • 57.Holick MF. Vitamin D deficiency. The New England journal of medicine. 2007;357(3):266–281. doi: 10.1056/NEJMra070553. [DOI] [PubMed] [Google Scholar]
  • 58.Millen AE, Bodnar LM. Vitamin D assessment in population-based studies: a review of the issues. Am J Clin Nutr. 2008;87(4):1102S–1105S. doi: 10.1093/ajcn/87.4.1102S. [DOI] [PubMed] [Google Scholar]
  • 59.VoPham T, Hart JE, Bertrand KA, Sun Z, Tamimi RM, Laden F. Spatiotemporal exposure modeling of ambient erythemal ultraviolet radiation. Environ Health. 2016;15(1):111. doi: 10.1186/s12940-016-0197-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.National Center for Health Statistics. Centers for Disease Control and Prevention (CDC) WONDER. [Accessed 7 Feb 2018];Bridged-Race Population Estimates 1990–2014. 2018 https://wonder.cdc.gov/controller/datarequest/D116.
  • 61.Hamra GB, Guha N, Cohen A, Laden F, Raaschou-Nielsen O, Samet JM, Vineis P, Forastiere F, Saldiva P, Yorifuji T, Loomis D. Outdoor particulate matter exposure and lung cancer: a systematic review and meta-analysis. Environ Health Perspect. 2014;122(9):906–911. doi: 10.1289/ehp.1408092. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Lee J-H, Han G, Fulp W, Giuliano A. Analysis of overdispersed count data: application to the Human Papillomavirus Infection in Men (HIM) Study. Epidemiol Infect. 2012;140(06):1087–1094. doi: 10.1017/S095026881100166X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Chen CL, Yang HI, Yang WS, Liu CJ, Chen PJ, You SL, Wang LY, Sun CA, Lu SN, Chen DS. Metabolic factors and risk of hepatocellular carcinoma by chronic hepatitis B/C infection: a follow-up study in Taiwan. Gastroenterol. 2008;135(1):111–121. doi: 10.1053/j.gastro.2008.03.073. [DOI] [PubMed] [Google Scholar]
  • 64.Pearson JF, Bachireddy C, Shyamprasad S, Goldfine AB, Brownstein JS. Association between fine particulate matter and diabetes prevalence in the US. Diabetes Care. 2010;33(10):2196–2201. doi: 10.2337/dc10-0698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Nasca PC, Burnett WS, Greenwald P, Brennan K, Wolfgang P, Carlton K. Population density as an indicator of urban-rural differences in cancer incidence, upstate New York, 1968–1972. Am J Epidemiol. 1980;112(3):362–375. doi: 10.1093/oxfordjournals.aje.a113002. [DOI] [PubMed] [Google Scholar]
  • 66.Yun EH, Lim MK, Oh JK, Park JH, Shin A, Sung J, Park EC. Combined effect of socioeconomic status, viral hepatitis, and lifestyles on hepatocelluar carcinoma risk in Korea. Br J Cancer. 2010;103(5):741–746. doi: 10.1038/sj.bjc.6605803. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Hajat A, Diez-Roux AV, Adar SD, Auchincloss AH, Lovasi GS, O’Neill MS, Sheppard L, Kaufman JD. Air pollution and individual and neighborhood socioeconomic status: evidence from the Multi-Ethnic Study of Atherosclerosis (MESA) Environ Health Perspect. 2013;121(11–12):1325–1333. doi: 10.1289/ehp.1206337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Kerr J, Fioletov V. Surface ultraviolet radiation. Atmos-Ocean. 2008;46(1):159–184. [Google Scholar]
  • 69.Brauer M, Hoek G, van Vliet P, Meliefste K, Fischer P, Gehring U, Heinrich J, Cyrys J, Bellander T, Lewne M, Brunekreef B. Estimating long-term average particulate air pollution concentrations: application of traffic indicators and geographic information systems. Epidemiology. 2003;14(2):228–239. doi: 10.1097/01.EDE.0000041910.49046.9B. [DOI] [PubMed] [Google Scholar]
  • 70.Hoek G, Raaschou-Nielsen O. Impact of fine particles in ambient air on lung cancer. Chin J Cancer. 2014;33(4):197–203. doi: 10.5732/cjc.014.10039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 71.Hystad P, Demers PA, Johnson KC, Brook J, van Donkelaar A, Lamsal L, Martin R, Brauer M. Spatiotemporal air pollution exposure assessment for a Canadian population-based lung cancer case-control study. Environ Health. 2012;11(1):22. doi: 10.1186/1476-069X-11-22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 72.Hystad P, Demers PA, Johnson KC, Carpiano RM, Brauer M. Long-term residential exposure to air pollution and lung cancer risk. Epidemiol. 2013;24(5):762–772. doi: 10.1097/EDE.0b013e3182949ae7. [DOI] [PubMed] [Google Scholar]
  • 73.Puett RC, Hart JE, Yanosky JD, Spiegelman D, Wang M, Fisher JA, Hong B, Laden F. Particulate matter air pollution exposure, distance to road, and incident lung cancer in the Nurses’ Health Study cohort. Environ Health Perspect. 2014;122(9):926–932. doi: 10.1289/ehp.1307490. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 74.Befort CA, Nazir N, Perri MG. Prevalence of obesity among adults from rural and urban areas of the United States: findings from NHANES (2005–2008) J Rural Health. 2012;28(4):392–397. doi: 10.1111/j.1748-0361.2012.00411.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.He D, Wu S, Zhao H, Qiu H, Fu Y, Li X, He Y. Association between particulate matter 2.5 and diabetes mellitus: A meta-analysis of cohort studies. J Diabetes Investig. 2017;8(5):687–696. doi: 10.1111/jdi.12631. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material
NIHMS963355-supplement-Supplementary_Material.docx (246.8KB, docx)

RESOURCES