Ambient air exposures to arsenic and cadmium and overall and prostate cancer–specific survival among prostate cancer cases in Pennsylvania, 2004 to 2014
We especially thank James Rubertone for data extraction and other inquiries related to the Pennsylvania Cancer Registry.
Abstract
Background
Exposures to arsenic (As) and cadmium (Cd) have been associated with higher prostate cancer (PC) mortality; however, these associations have been inconsistent. The authors investigated whether higher ambient air concentrations of As and Cd are associated with lower overall and PC-specific survival among PC cases in Pennsylvania.
Methods
Incident PC cases of patients, aged 40 years or older, with a clinical diagnosis and nonmetastatic disease were identified in the 2004 to 2014 Pennsylvania Cancer Registry. Demographic, clinical, and pathologic information were extracted from the Pennsylvania Cancer Registry. The 3- and 5-year average and cumulative air concentrations of As and Cd were extracted from the Environmental Protection Agency's Toxics Release Inventory database. Spatial-temporal hierarchical accelerated failure time models were used to examine the associations between air concentrations of As and Cd and overall and PC-specific survival for the total population and stratified by geographical region defined by rurality and Appalachia status, after adjusting for confounders.
Results
There were 78,914 PC cases included. Increasing 3- and 5-year average and cumulative air concentrations of As and Cd were significantly associated with lower overall and PC-specific survival among cases, after adjusting for confounders, for the total population, and stratified by geographical region for most of the estimates.
Conclusions
Data suggest that increasing ambient air exposures to As and Cd may play a role in overall and PC-specific mortality risk among PC cases. Exposures to As and Cd are modifiable and may provide insight into potential strategies to improve PC health outcomes.
Lay Summary
- Arsenic and cadmium exposures linked to increased prostate cancer deaths remain unclear.
- We investigated whether air levels of arsenic and cadmium reported to be released from industries decrease overall and prostate cancer–specific survival among prostate cancer cases identified in the 2004 to 2014 Pennsylvania Cancer Registry.
- Among the 78,914 prostate cancer cases, increasing air levels of arsenic and cadmium are found to be associated with lower overall and prostate cancer–specific survival for the total population and within rural and urban Appalachia and urban non-Appalachia counties in Pennsylvania.
- Reducing exposures to arsenic and cadmium have the potential to decrease prostate cancer deaths.
Introduction
Prostate cancer (PC) is the second leading cause of cancer-related death in men in the United States, with 34,130 deaths estimated in 2021.1 Many PC cases are characterized with indolent low-grade PC; however, some men will develop more aggressive PC that could lead to poorer health outcomes such as metastasis or PC-related death. Black race, obesity, and smoking are risk factors associated with more aggressive PC.2-4 However, environmental exposures such as toxic exposures to heavy metals as possible risk factors for aggressive PC leading to PC-related death remain unknown.
Human and animal studies suggest that environmental exposures to pollutants such as heavy metals may play a role in carcinogenesis.5, 6 According to the International Agency for Research on Cancer, heavy metals such as arsenic (As) and cadmium (Cd) are considered as carcinogenic to humans.7 Studies have shown exposures to As and Cd influencing cancer cellular processes such as cellular proliferation and differentiation, apoptosis, and/or angiogenesis as well as inhibiting DNA repair and inducing oxidative stress in cell-based models.7-10 Furthermore, epidemiologic studies have found associations between As and Cd exposures, either through inhalation or ingestion, and PC mortality, especially in occupational populations where exposures to these heavy metals are more common6, 7, 11, 12; however, these study findings have been inconsistent.6, 7, 13 Few studies have examined the association between As and Cd exposures and PC mortality in the US general population.
In Pennsylvania, a study reported high levels of Cd and As around abandoned, discharging coal mines contaminating groundwater and surface water in Pennsylvania.14 The US Environmental Protection Agency (EPA) requires certain industrial facilities to report released quantities of various hazardous chemicals that are released in air, water, and soil above certain thresholds that are recorded in EPA's Toxics Release Inventory (TRI) database.15 Exposures to As and Cd and their effects on PC mortality, the second leading cause of cancer-related deaths among Pennsylvania men,16 are unknown in Pennsylvania.
To better understand As and Cd exposures on PC health outcomes, we conducted a population-based study to investigate whether increasing ambient air concentrations of As and Cd are associated with lower overall and PC-specific survival among newly diagnosed PC cases in Pennsylvania. We also examined whether these associations differ by geographical region defined by rurality and Appalachia status. We hypothesized that increasing air concentrations of As and Cd are associated with lower overall and PC-specific survival among PC cases.
Materials and Methods
Study Population
We identified incident PC cases of patients, aged 40 years or older, with a clinical diagnosis of PC (Gleason grade group [GGG] 1) from the Pennsylvania Cancer Registry (PCR) from 2004 to 2014. Cases who had an unknown GGG but a known tumor stage ≥T3 were included. Patients diagnosed with distant metastasis were excluded. Demographic, clinical, and pathologic information from the PCR were extracted.17-19
Geographical region was defined by rurality and Appalachia status using the US Department of Agriculture's Rural-Urban Continuum Code20 where counties coded with a Rural-Urban Continuum Code less than 421 and using the Appalachian Regional Commission definition,22 respectively. There are 3 geographical strata by rurality and Appalachia status in Pennsylvania: rural Appalachia, urban Appalachia, and urban non-Appalachia (Fig. 1). There are not any rural non-Appalachian counties in Pennsylvania. Definitive treatment was defined as primary site surgery and/or radiation to capture PC treatments with curative intent. Cases were categorized in 2 disease groups defined by pathologic GGG and pathologic tumor stage or GGG at biopsy and clinical tumor stage if pathologic results were missing: 1) less aggressive PC (GGG 1 or 2 and tumor stage T1-T2) and 2) more aggressive PC (GGG ≥3 or tumor stage ≥T3).
Because cancer is a reportable disease in Pennsylvania, informed consent of PC cases was waived. This study was approved by the Pennsylvania Department of Health and the Institutional Review Board of the Pennsylvania State University College of Medicine.
Exposures and Outcomes
The exposure variables were 3- and 5-year average and cumulative air concentrations of As and Cd based on reported releases included in the EPA's TRI database. The reported releases to air of As and As compounds in pounds were added together to form the source quantity for the air dispersion model; the same procedure was used for Cd and Cd compounds. Both direct (stack) and fugitive air emissions were included. In the TRI database, the terms As or Cd compounds include the set of oxides, sulfates, chlorides, nitrides, organic compounds, and other less common formulations that form with the heavy metals.
From TRI-reported releases by Pennsylvania industries, we created an air dispersion model to estimate the 3- and 5-year mean and cumulative daily exposure at each patient location to Cd and As air concentrations. Using a Gaussian plume model implemented with MATLAB 2018a, concentrations surrounding release points were estimated on an hourly basis with the relevant meteorological conditions from the nearest National Weather Service automated surface observation point. Dispersion coefficients were based on the stability class inferred from wind speed and cloud cover during the hour. All concentrations were evaluated on a 10 × 10 m grid to enable precise divisions between different geographical regions. Each release point and patient location was assessed to be at the closest grid node based on the longitude and latitude determined by the facility address in the TRI database or by the patient's home address at the time of diagnosis. For each patient, the hourly concentrations over the 3- and 5-calendar-year periods preceding the calendar year of diagnosis were evaluated, such that total TRI emissions between 1999 and 2014 were included in the model.
TRI-reported releases of As and Cd are annual totals that are modeled as continuous hourly releases during the entire calendar year. Although the facilities reporting these releases are commonly electric utilities (53%), primary metals (14%), and wood products (8%), which would tend to operate on a mostly continuous basis, there is no available data to indicate whether any facility ceased emissions for a period of time during the year. Using hourly wind data, the model builds a probabilistic distribution of the pollutant concentrations around the emission points that minimizes the effect of temporary cessations of emissions at particular sites. Given typical wind speeds in Pennsylvania (mean = 2.8 meters/second, median = 2.6 meters/second), a 1-hour plume can disperse over a mean distance of 10 km from the point of release.
The primary time-to-event outcome was PC-specific survival; and, the secondary outcome was overall survival.
Statistical Analysis
Summary statistics for demographic and clinical characteristics were provided by means with standard deviations for continuous variables and frequencies with percentages for categorical variables. Analysis of variance tests or Kruskal-Wallis tests were applied for comparisons of continuous variables. Pearson χ2 or Fisher exact tests were applied for comparisons of categorical variables among 3 geographical regions, as appropriate. PC-specific and overall survival rates were calculated based on Kaplan-Meier curves. The group comparison on survival was based on log-rank tests. ArcGIS10.6.1 software, using the Web Mercator coordinate system (EPSG: 3857), was used to map the facilities of 2004 to 2014 stack and fugitive air emissions in which the median emissions were calculated for the total time period (Fig. 1).
Spatial-temporal hierarchical accelerated failure time (AFT) models were used, after adjusting for confounders, for the total population and stratified by geographical region. There were 2 models in which the following confounders were included: 1) age at diagnosis (40-49, 50-59, 60-69, ≥70 years), continuous year of diagnosis (2004-2014), race (Black, White, other, unknown), and PC aggressiveness (less aggressive, more aggressive, unknown); and 2) all variables from model 1 plus continuous serum PSA, geographical region, insurance status (insured, not insured, unknown), and definitive treatment received (yes, no, unknown). Additional analyses were conducted in selected models in which more aggressive PC was redefined including serum PSA ≥20 ng/mL; this re-defined PC and lymph node positivity was adjusted for in these models. In all of these models, a Weibull AFT regression was considered, where correlations between counties and diagnosis year were captured by multivariate random effects. Bayesian techniques for parameter estimation with 95% credible interval (CI) and inference were used by adopting noninformative multivariate conditional autoregressive priors.
The statistical analyses were conducted in software R (version 4.0.3) and SAS (SAS, Cary, North Carolina). All tests were 2-sided with a significance level of .05.
Results
There were 78,914 patients with PC, aged 40 to 105 years, included. Table 1 describes the study characteristics of these cases for the total study population and stratified by geographical region. For the total study population, 2.3% and 14.4% of them had PC-specific and all-cause deaths, respectively. The 5-year overall and PC-specific survival rates were 56.4% (95% CI, 55.6%-57.2%) and 90.6% (95% CI, 90.0%-91.1%), respectively. When stratified by geographical region, rural Appalachia had statistically signficantly lower PC-specific and overall survival rates compared to urban non-Appalachia (P ≤ .03) but no difference when compared to urban Appalachia (P > .05). For the magnitude of the reportable 2004 to 2014 total stack and fugitive air emissions released by the 59 industries throughout Pennsylvania, they ranged from 0 to 6061.00 lb (median, 94.39 lb) for the 42 As facilities and from 0 to 4526.00 lb (median, 10.00 lb) for the 17 Cd facilities in Pennsylvania (Fig. 1).
| Variables | Total (n = 78,914) | Urban Non-Appalachia (n = 42,301) | Urban Appalachia (n = 26,375) | Rural Appalachia (n = 10,238) |
|---|---|---|---|---|
| Age, mean (SD), y | 65.8 (9.03) | 65.5 (9.01) | 66.0 (9.06) | 66.7 (8.96) |
| Age group, No. (%) | ||||
| Age 40-49 y | 2236 (2.8) | 1353 (3.2) | 660 (2.5) | 223 (2.2) |
| Age 50-59 y | 18,129 (23.0) | 10,036 (23.7) | 6025 (22.8) | 2068 (20.2) |
| Age 60-69 y | 31,469 (39.9) | 16,918 (40.0) | 10,475 (39.7) | 4076 (39.8) |
| Age ≥70 y | 27,080 (34.3) | 13,994 (33.1) | 9215 (34.9) | 3871 (37.8) |
| Race, No. (%) | ||||
| White | 66,769 (84.6) | 32,689 (77.3) | 24,132 (91.5) | 9948 (97.2) |
| Black | 8409 (10.7) | 6655 (15.7) | 1627 (6.2) | 127 (1.2) |
| Other | 873 (1.1) | 678 (1.6) | 156 (0.6) | 39 (0.4) |
| Unknown | 2863 (3.6) | 2279 (5.4) | 460 (1.7) | 124 (1.2) |
| Insurance, No. (%) | ||||
| Insured | 65,277 (82.7) | 34,370 (81.3) | 22,333 (84.7) | 8574 (83.7) |
| Uninsured | 330 (0.4) | 178 (0.4) | 106 (0.4) | 46 (0.4) |
| Unknown | 13,307 (16.9) | 7753 (18.3) | 3936 (14.9) | 1618 (15.8) |
| Serum PSA, mean (SD), ng/mL | 10.3 (15.0) | 10.2 (15.3) | 10.3 (14.4) | 10.6 (15.1) |
| Gleason grade group, No. (%) | ||||
| 1 | 34,408 (43.6) | 18,581 (43.9) | 11,095 (42.1) | 4732 (46.2) |
| 2 | 32,957 (41.8) | 17,827 (42.1) | 11,111 (42.1) | 4019 (39.3) |
| 3-5 | 11,447 (14.5) | 5839 (13.8) | 4132 (15.7) | 1476 (14.4) |
| Unknown | 102 (0.1) | 54 (0.1) | 37 (0.1) | 11 (0.1) |
| Gleason score, mean (SD) | 6.78 (0.862) | 6.76 (0.840) | 6.81 (0.890) | 6.75 (0.878) |
| Tumor stage, No. (%) | ||||
| T1 | 32,687 (41.4) | 16,848 (39.8) | 11,345 (43.0) | 4494 (43.9) |
| T2 | 35,499 (45.0) | 19,494 (46.1) | 11,506 (43.6) | 4499 (43.9) |
| T3 | 7613 (9.6) | 4121 (9.7) | 2622 (9.9) | 870 (8.5) |
| T4 | 490 (0.6) | 254 (0.6) | 164 (0.6) | 72 (0.7) |
| Unknown | 2625 (3.3) | 1584 (3.7) | 738 (2.8) | 303 (3.0) |
| Aggressiveness, No. (%) | ||||
| Less aggressive | 50,122 (63.5) | 27,110 (64.1) | 16,413 (62.2) | 6599 (64.5) |
| More aggressive | 24,390 (30.9) | 12,925 (30.6) | 8540 (32.4) | 2925 (28.6) |
| Unknown | 4402 (5.6) | 2266 (5.4) | 1422 (5.4) | 714 (7.0) |
| Definitive treatment received, No. (%) | ||||
| No | 14,882 (18.9) | 6743 (15.9) | 5509 (20.9) | 2630 (25.7) |
| Yes | 63,199 (80.1) | 35,104 (83.0) | 20,587 (78.1) | 7508 (73.3) |
| Unknown | 833 (1.1) | 454 (1.1) | 279 (1.1) | 100 (1.0) |
| Definitive treatment regimen, No. (%) | ||||
| Primary site surgery only | 28,147 (35.7) | 15,462 (36.6) | 9399 (35.6) | 3286 (32.1) |
| Radiation only | 33,177 (42.0) | 18,519 (43.8) | 10,656 (40.4) | 4002 (39.1) |
| Both treatments received | 1875 (2.4) | 1123 (2.7) | 523 (2.0) | 220 (2.1) |
| Neither treatment received | 14,882 (18.9) | 6743 (15.9) | 5509 (20.9) | 2630 (25.7) |
| Unknown | 833 (1.1) | 454 (1.1) | 279 (1.1) | 100 (1.0) |
- Abbreviation: PSA, prostate-specific antigen; y, year; SD, standard deviation.
- All reported percentages are column percentages. Pennsylvania Cancer Registry documentation was top-coded at 98.0 and bottom-coded at 0.1. Tumor stage was based on the TNM staging system. Definitive treatment was defined as primary site surgery and/or radiation. Primary site surgery refers only to total organ resection (radical prostatectomy not otherwise specified). A statistical significance difference was observed for all study variables (except for insurance) among urban non-Appalachia, urban Appalachia, and rural Appalachia (P < .05).
For the total study population, the 3-year compared to the 5-year average air concentrations were higher for both As and Cd (Table 2). The 5-year compared to the 3-year cumulative air concentrations were higher for both As and Cd. Cases in urban non-Appalachia had higher 3- and 5-year average and cumulative air concentrations of As compared to cases in urban Appalachia and rural Appalachia. Cases in rural Appalachia had higher 3- and 5-year average and cumulative air concentrations of Cd compared to cases in urban Appalachia and urban non-Appalachia.
| Concentrations | Total (n = 78,914) | Urban Non-Appalachia (n = 42,301) | Urban Appalachia (n = 26,375) | Rural Appalachian (n = 10,238) |
|---|---|---|---|---|
| 3-y average (SD), μg/m3 | ||||
| Arsenic | 0.0000739 (0.00126) | 0.0000886 (0.00167) | 0.0000499 (0.000527) | 0.0000748 (0.000357) |
| Cadmium | 0.00000524 (0.0000805) | 0.000000253 (0.00000368) | 0.0000108 (0.0000820) | 0.0000115 (0.000180) |
| 5-y average (SD), μg/m3 | ||||
| Arsenic | 0.0000706 (0.00104) | 0.0000878 (0.00136) | 0.0000448 (0.000490) | 0.0000661 (0.000318) |
| Cadmium | 0.00000448 (0.0000686) | 0.000000228 (0.00000250) | 0.00000900 (0.0000678) | 0.0000104 (0.000156) |
| 3-y cumulative (SD), μg-h/m3 | ||||
| Arsenic | 1.94 (33.2) | 2.33 (43.8) | 1.31 (13.8) | 1.97 (9.39) |
| Cadmium | 0.138 (2.12) | 0.00664 (0.0968) | 0.284 (2.15) | 0.303 (4.73) |
| 5-y cumulative (SD), μg-h/m3 | ||||
| Arsenic | 3.09 (45.6) | 3.85 (59.5) | 1.96 (21.5) | 2.90 (13.9) |
| Cadmium | 0.196 (3.00) | 0.00997 (0.109) | 0.394 (2.97) | 0.457 (6.81) |
- Abbreviation: SD, standard deviation; y, year.
Table 3 summarizes the regression results for the fixed effect parameters in AFT models. The estimates are directly associated with the natural logarithm of time, with a negative value indicating a decrease in survival time and a positive value for an increase in survival time. From these models, increasing 3- and 5-year average and cumulative air concentrations of As and Cd were statistically significantly associated with lower overall and PC-specific survival among PC cases, after adjusting for confounders in models 1 and 2, except for some estimates for As with PC-specific survival in models 1 and 2.
| Overall Survival | PC-Specific Survival | |||
|---|---|---|---|---|
| Estimate (95% CI) | Estimate (95% CI) | |||
| Model 1 | Model 2 | Model 1 | Model 2 | |
| Arsenic Concentrations | ||||
| Average, μg/m3 | ||||
| 3 y | –0.024 (–0.027 to –0.023) | –0.067 (–0.069 to –0.065) | –0.011 (–0.027 to –0.004) | –0.010 (–0.015 to –0.006) |
| 5 y | –0.029 (–0.030 to –0.027) | –0.074 (–0.077 to –0.072) | –0.011 (–0.028, 0.001)a | –0.023 (–0.033 to –0.012) |
| Cumulative, μg-h/m3 | ||||
| 3 y | –0.007 (–0.013 to –0.003) | –0.019 (–0.023 to –0.015) | –0.0004 (–0.003 to 0.002)a,b, a,b | –0.003 (–0.004 to –0.0004)a |
| 5 y | –0.006 (–0.011 to –0.003) | –0.022 (–0.024 to –0.019) | –0.002 (–0.005, 0.0005)a,b, a,b | –0.003 (–0.004, 0.002)b |
| Cadmium Concentrations | ||||
| Average, μg/m3 | ||||
| 3 y | –0.023 (–0.025 to –0.023) | –0.067 (–0.069 to –0.065) | –0.012 (–0.018 to –0.009) | –0.044 (–0.055 to –0.034) |
| 5 y | –0.024 (–0.026 to –0.022) | –0.075 (–0.078 to –0.074) | –0.010 (–0.014 to –0.006) | –0.047(–0.054 to –0.041) |
| Cumulative, μg-h/m3 | ||||
| 3 y | –0.003 (–0.004 to –0.003) | –0.037 (–0.043 to –0.032) | –0.005 (–0.010 to –0.002) | –0.005 (–0.008 to –0.003) |
| 5 y | –0.004 (–0.006 to –0.004) | –0.048 (–0.051 to –0.045) | –0.004 (–0.008 to –0.001) | –0.003 (–0.006, 0.001)b |
- Abbreviations: CI, credible interval; PC, prostate cancer; y, year.
- The total number of study participants is 78,914 for each model. Each concentration was adjusted by the following variables: age at diagnosis (40-49, 50-59, 60-69, and ≥70 years), year of diagnosis–continuous variable, race [Black, White, other, and unknown], and prostate cancer aggressiveness [less, more, and unknown] for model 1 and model 1 variables, serum prostate-specific antigen–continuous variable, rurality-Appalachia status (urban Appalachia, rural Appalachia, and urban Non-Appalachia), insurance status (insured, not insured, and unknown), and definitive treatment regimen (yes, no, and unknown) for model 2.
- a Rounding occurred at 5 places behind the decimal because of a rounding error.
- b The 95% CI includes 0, which means nonsignificant results.
When we examined the association between geographical region and overall and PC-specific survival, cases in urban and rural Appalachia had statistically significantly higher overall survival compared urban non-Appalachia but no significant associations with PC-specific survival, after adjusting for 3- and 5-year average and cumulative air concentrations of As and Cd, individually, in most of the models (Supporting Table 1). Within each geographical region, increasing 3- and 5-year average and cumulative air concentrations of As and Cd were associated with statistically significantly lower overall and PC-specific survival among cases for most the estimates (Supporting Tables 2-4). The statistically significance associations remained when PC aggressiveness was redefined, and lymph node positivity was adjusted for in the selected models (Supporting Table 5).
Discussion
To our knowledge, this is the first study that examined whether ambient air concentrations of As and Cd were associated with overall and PC-specific survival in the Pennsylvania general population stratified by rurality and Appalachia status. We found increasing 3- and 5-year average and cumulative concentrations of As and Cd associated with lower overall and PC-specific survival for the total study population and stratified by geographical region, after adjusting for confounding factors. Our study findings suggest that increasing ambient air exposures to As and Cd may play a role in the overall and PC mortality risk among PC cases in Pennsylvania and possibly in other populations that are exposed to these heavy metals.
Previous studies have linked exposures to As and Cd to PC mortality.6, 7, 11 For example, higher occupational exposure to Cd6 and Cd and As levels in well water7 were found to be associated PC mortality. In addition, dose responses between increasing Cd and As exposures and the risk of PC death were found.6, 7, 11 However, associations between these heavy metal exposures and PC mortality have been inconsistent,6, 7, 13 possibly because of differences in sample size, assessment of exposure, and study populations. Exposures to As and Cd have shown to have biological influences on the prostate such as high levels of Cd accumulating in the prostate8 and As exposures inducing malignant transformation in prostate epithelial cells,11 contributing to aggressive tumor growth,23 and decreasing androgen dependence in human prostate cells.23 In our study, we demonstrated that increasing ambient air exposures may influence overall and PC-specific mortality risk among PC cases identified in the Pennsylvania general population. Our study findings may be applicable to other US populations being that there are industrial facilities throughout the United States that report emissions to EPA and therefore are included in the EPA's TRI database. In addition, our study finding is consistent with our previous study that used EPA's Environmental Quality Index, which identified several environmental factors such as air pollutants (ie, ortho-toluidine) associated with lower PC-specific survival in this same cohort of PCR men.17 Further evaluation of the association between ambient air exposures to As and Cd and PC health outcomes is warranted where environmental factors such as other air pollutants like particulate matter (ie, PM2.5) are considered.
Studies that examined geographical differences in As and Cd exposures among PC cases in the United States are limited, in particular, by Appalachia status. Most Pennsylvania counties (52 out of 67 counties) are located in the Appalachia region where the majority of the facilities that had detectable As and Cd air concentrations between 2004 and 2014 were located. However, these facilities were widespread in the Appalachia counties compared to being more concentrated in non-Appalachia counties (Fig. 1). In a previous study, we found PC cases who lived in urban Appalachia Pennsylvania were more likely to have aggressive PC at diagnosis compared to cases who reside in urban, non-Appalachia PA18; however, in this study, we were unable to determine whether this association was because of environmental influences. In the present study, we found no statistically significant difference in PC-specific survival for majority of the estimates between urban and rural Appalachia compared to urban non-Appalachia, after adjusting for air concentrations of As and Cd and other confounders. Reasons for this finding warrant further investigation.
Besides As and Cd, there are other pollutants that are emitted from the EPA's TRI industries that could possibly play a role in disease risk. For example, TRI-reported pollutant releases to the air from this same set of emitters include hydrochloric acid, sulfuric acid, hydrogen fluoride, and zinc compounds. To our knowledge, this set of air pollutants is not known to be associated with PC. Of course, other nonreported pollutants may also be emitted from these and other nearby industries such as particulate matter or nitrogen oxides. However, we lack the data with which to test their effects, and these pollutants are much more ubiquitous resulting from a wide array of human activities.
There were limitations to this study. First, the air concentrations of As and Cd were estimates, and we did not test biological samples to measure As and Cd levels or examine dietary and other environmental As and Cd exposures to assess total exposure. The PCR had limitations such as the lack of information on other factors that may contribute to the As and Cd exposure levels and/or affect survival (ie, lifestyle behaviors, occupation, residential history, socieconomic factors, and comorbidities), missing data on some study variables such as insurance status (~17%), and inconsistencies with some clinical variables in a few cases (ie, the sum of the Gleason pattern not equaling to the total Gleason score). Another limitiation was the inability to control for migration of PC cases in and out of Pennsylvania. We used 3- and 5-year air concentation data of As and Cd before PC diagnosis as opposed to ≥10 years of exposure data that may provide a more accurate evaluation of their influence on prostate carcinogenesis. Finally, individual-level estimates of As and Cd concentrations of the PC cases were sparse and had a skewed distribution; we log-transformed these estimates to achieve a normal distribution.
In conclusion, ambient air exposures to As and Cd are modifiable. The majority of these emissions come from coal-fired power plants, where Cd and As are trace elements in the coal that is burned. Large proportions of the remaining As and Cd emissions come from metal manufacturing, wood products manufacturing, electrical equipment manufacturing, and chemicals manufacturing. Engineering controls are available to reduce or eliminate Cd and As emissions in many instances, but a lack of mandates means that the costs of these process changes are not justified business expenses. Therefore, regulatory policies for industries who produce these emissions could be 1 possible way to induce changes to reduce their concentrations in the environment. Reducing these exposures to As and Cd may be potential strategies to improve PC health outcomes.
Funding Support
This study was funded by the Highmark Foundation's Human Health and the Environment Seed Grant Program at the Pennsylvania State University.
Conflict of Interest Disclosures
The authors declare no conflict of interest.
Author Contributions
Alicia C. McDonald: Conceptualization, funding acquisition, methodology, writing–original draft, and writing–reviewing and editing. Jeremy Gernand: Data curation, analysis, methodology, and writing–reviewing and editing. Nathaniel R. Geyer: Data curation, analysis, and writing–reviewing and editing. Hongke Wu: Analysis and writing–reviewing and editing. Yanxu Yang: Analysis and writing–reviewing and editing. Ming Wang: Conceptualization, funding acquisition, methodology, data curation, analysis, and writing–reviewing and editing.
