Document 5kEk1M6dzyXBVBK7eYxeJnYr0

Title: An investigation of the association between hazardous air pollutants and lymphohematopoietic cancer risk among residents of Harris County, Texas Authors: Kristina M. Walker, Ann L. Coker, Elaine Symanski, Philip J. Lupo Affiliation: University of Texas Health Science at Houston, School of Public Health Abstract While recent reports have commented on the elevated ambient levels of hazardous air pollutants (HAPs) in certain areas of Houston, relative to other cities, few studies have assessed the health effects of HAPs for Houstonians and none have evaluated the association between ambient levels of these pollutants and lymphohematopoietic cancer risk in this population. To begin to address this deficit in the literature, we conducted a population based analysis of ambient levels of HAPs and lymphohematopoietic cancer incidence in Harris County. Cancer cases were obtained from the Texas Cancer Registry (TCR), 1995-2003, and included adult and childhood cancers. We used existing monitoring data collected from 1992-2003 by the Texas Commission on Environmental Quality (TCEQ) to estimate ambient benzene and 1,3 butadiene levels for each census tract, which were then collapsed into 3 or 4 categories, respectively, in the analyses that were carried out. We also estimated risk of lymphohematopoietic cancer associated with residential proximity to the Houston Ship Channel (HSC) as another proxy (albeit a crude measure) of HAP exposure. Poisson regression was used to estimate rate ratios for increasing HAP levels and cancer rates, adjusting for gender, age, and a composite index for socioeconomic status and race / ethnicity. We observed an association among children living within two miles of the HSC and higher rates of acute lymphocytic leukemia (adjusted rate ratio [aRR] = 1.56; p=0.01) compared with those living greater than 10 miles from the HSC. Further, higher 1,3-butadiene levels (>1.15 parts per billion [ppbV] relative to <0.266 ppbV) were associated with acute lymphocytic leukemia, acute myeloid leukemia and all leukemias in children. While we did not observe statistically significant associations for the other two categories (i.e., 0.266 0.381 ppbV or 0.382 - 1.15 pbbV) as compared to the referent group of <0.226 ppbV, a trend of increasing 1,3-butadiene levels and increasing leukemia rates was noted for these cancers. Higher benzene levels were not associated with lymphohematopoietic cancer by types in children. Among adults, neither proximity to the HSC, nor ambient levels of benzene or 1,3-butadiene levels were associated with lymphohematopoietic cancers in a dose dependent manner for either males or females. Additional analyses using more sophisticated methods to assess exposure are planned to confirm our findings. Additionally, future analyses will address many of the limitations of the current analysis. However, observing a specific health effect of HAPs in light of recently documented elevated levels of two known carcinogens, benzene and 1,3 butadiene, in Houston,(1) suggests a need to explore this issue further and to potentially take action to limit potential exposure to HAPs. Introduction In 1997, Texas ranked among the top three states in the number of chemical and allied products manufacturing facilities and in the number of petroleum and coal products manufacturing facilities, with more facilities located in Harris County (Houston) than in any other Texas county. (2) In a recent report on the effects of air pollution, the Mayor's Task Force determined that ambient levels of both benzene and 1,3-butadiene pose a `definite' cancer risk to Houstonians. (1) According to the International Agency for Research on Cancer (IARC), benzene is classified as a known human carcinogen and 1,3-butadiene is classified as a probable human carcinogen. (3) Few studies have assessed the health effects of hazardous air pollutants (HAP) for Houstonians and none have addressed ambient levels of these pollutants and risk of lymphohematopoietic cancer, which includes leukemia, lymphoma and myeloma. The majority of studies that have examined the potential health effects of HAPs have been conducted in occupational workers. However, most have not directly measured individual exposures to pollutants. The HAPs of primary concern in Houston, benzene and 1,3-butadiene, are mainly produced by petrochemical and synthetic rubber and plastics manufacturers. Occupational studies conducted in these settings typically compare worker populations with the population at large and report lower cancer mortality and incidence rates, for all cancers combined. (4-23) Four studies reported more deaths from lymphohematopoietic cancers than expected among refinery/petrochemical plant employees (10, 13) or those exposed to 1,3-butadiene (7) or benzene (24) while findings from eight studies do not support this association.(6, 16, 17, 19, 21, 25-27) When looking at risk of developing cancer, four worker based cohort studies found higher risk of leukemia associated with exposure to benzene, (28, 29) or 1,3-butadiene (5, 30) and ten studies did not find an association. (6, 8-10, 12, 13, 16, 19-21) A nested case-control study (31) reported a significant increased risk of leukemia associated with an average exposure to 1,3-butadiene at levels of 1 ppm and another study reported that workers exposed to benzene at levels below 1 ppm had significantly decreased levels of white blood cells compared with unexposed workers. (32) Few studies have looked at HAPs and risk of cancer in community populations. In these studies, exposure to HAPs is usually measured with a proxy measure such as residential distance from a plant or spill. One study compared actual and expected number of cancers in an area affected by a gasoline spill as a measure of high exposure; residents' leukemia rates were 4.40 (95% CI: 1.09-10.24) higher than expected by chance. (33) Another study found no effect of proximity (three kilometer area radius) to a petrochemical plant on the risk for leukemia (standardized incidence ratio [SIR]=0.99, 95% CI: 0.66-1.51) yet they did note a significant increased risk of multiple myeloma (SIR=2.15, 95% CI: 1.253.67). (34) Three other studies found no increased risk of leukemia in areas thought to be affected by industrial pollution.(35-37) Five have found a significant association with residence near industrial facilities and specific lymphohematopoietic cancers (38-42) while three others have found no association (34, 42, 43) Only a small number of studies have explored the potential association between HAPs and childhood cancers. In a large California based study including 7,143 childhood cancer cases, Reynolds et al. (15) found a slight but non significant increase in leukemias (aRR=1.15; 95% CI 0.971.37) and proximity to road ways; there was no evidence that cancer rates increase with increasing vehicle or road density. In a follow-up for this study, Reynolds et al. (14) used data from the same 7,143 cases but used a dispersion model developed by the EPA to assign census tract level exposure scores to 25 potentially carcinogenic HAPs. Exposure scores for individual HAPs were summed to create a combined HAP exposure score. Reynolds et al. found a significant (p<0.05) trend of increasing childhood leukemia rates with increasing HAP exposure levels with a 21% (aRR = 1.21; 95% CI: 1.03-1.42) increased risk of childhood leukemia in census tracts with the highest combined exposure score. They further report a 32% increase in leukemia risk (aRR = 1.32; 95% CI: 1.11-1.57) associated with the highest levels of HAP exposure from point sources, compared with non-significant leukemia risks associated with the highest levels of HAP exposures from mobile (aRR=1.18) and area (aRR=1.16) sources. In the United Kingdom, Knox found correlations between residence at birth near geographically defined "hotspots" for several criteria pollutants as well as volatile organic compounds (VOCs), benzene, dioxins, 1,3butadiene, and benz(a)pyrene and childhood leukemias. (11) Additional studies are needed to confirm the findings from these studies. Because there have been few community level studies of HAP and cancer risk and because most of these studies have not utilized data on ambient levels of specific air pollutants, we have conducted a population based analysis of the association between HAPs and lymphohematopoietic cancer incidence in Harris County. We used existing monitoring data collected from 1992-2003 by the Texas Commission on Environmental Quality (TCEQ) to estimate census tract ambient levels of benzene and 1,3-butadiene. Cancer cases were obtained from the Texas Cancer Registry (TCR), 1995-2003, and included cases in adults and children. Methods: Overview and Study Population: Two ecologic analyses utilizing existing data sources were performed to begin to address the question of whether air pollution is associated with increased rates of lymphohematopoietic cancer in Harris County, Texas. Our focus was on two specific hazardous air pollutants, benzene and 1,3-butadiene, identified as posing a `definite' cancer risk in a recent report to the Houston Mayor's Office. (1) The two analyses presented in this report serve as preliminary results for a future planned study incorporating more complex methodology. In the first analysis, we assessed the association between distance from the Houston Ship Channel (HSC) and cancer rates. We then evaluated the association between cancer rates and ambient levels of benzene and 1,3-butadiene. Both analyses were performed at the census tract level. Our study population included all residents of Harris County, Texas between 1995 and 2003. We identified cases of lymphohematopoietic cancer diagnosed and reported to the Texas Cancer Registry (TCR) during the same time period. All incident cases of lymphohematopoietic cancer, independent of age or stage of disease were eligible. The specific lymphohematopoietic cancers we examined include leukemia (ICD-10 codes C91-C95), myeloma (ICD10 codes C88, C90, C96), non-Hodgkin's lymphoma (ICD-10 codes C82-C85) and Hodgkin's disease (7.8% of total) (ICD-10 code C81). We excluded those cancer cases with a prior cancer diagnosis. No primary data were collected for this study and no patients, physicians or hospitals were contacted. Since only cancer registry data was used, no identifying information other than age, race/ethnicity and address at the time of diagnosis was available. Addresses were used solely for the purposes of assignment of cases to census tracts. At the census tract level, rates of cancer were calculated using the Census 2000 estimates for the population. All rate ratios were adjusted for census tract level age, gender, and socioeconomic status and race / ethnicity indicators. Exposure Assessment: Distance from Houston Ship Channel: We evaluated distance of census tracts from the HSC using geographic information system (GIS) methodology. Using ArcGIS(44) we constructed a set of five mutually exclusive buffers two miles apart around the HSC (see figure 1). We then assigned census tracts to a particular buffer based on the buffer in which the centroid of the census tract fell. Consequently, there were five exposure groups corresponding to the five buffers that we will refer to as the following: : 0-2 miles, 2-4 miles, 4-6 miles, 6 -8 miles, 8-10 miles. Census tracts with centroids that fell outside all five buffers (i.e., > 10 miles) served as the referent group. We hypothesized that if distance from the Houston Ship Channel were correlated with and therefore a proxy for industrial air pollution levels and this proxy was related to lymphohematopoietic cancer risk, rates of cancer would be higher in those buffers closest to the HSC. We do recognize that distance from the Houston Ship Channel is only one proxy measure of industrial air pollution and that there are other industrial pollutants outside the HSC. Legend Houston Ship Channel Buffer Distance (Miles) 0-2 2-4 4-6 6-8 8-10 Figure 1. Map of Harris County, Texas depicting five buffer zones around the Houston Ship Channel Compound-Specific Exposure: Notwithstanding the limitations of using ambient air levels as a surrogate of personal exposure (45-47), we utilized data from the Texas Commission on Environmental Quality (TCEQ) to construct an assessment based on the potential for exposure to specific compounds. The TCEQ operates an extensive air monitoring network in the state of Texas, with Houston being one of the most monitored cities in the nation.(48) Although the TCEQ began monitoring air quality in the early 1970's, they did not begin monitoring for HAPs until 1992. Between the years 1992 and 2005, 15 TCEQ monitoring sites were recording data on HAPs in Harris County. However, not all 15 sites were operable for the entire period. Because the cancer data obtained from the TCR extends only to 2003, we limited our use of the TCEQ data through this year as well. One monitoring site was not operable before 2005; therefore, the data used for this assessment comes from a total of 14 monitoring sites noted as red dots in Figure 2. As shown in the figure, these monitors are not geographically distributed throughout the Houston area as many are located within or near the Ship Channel, which poses difficulty in any assessment that attempts to utilize such data to estimate ambient levels at distances remote from the monitoring stations. Legend TCEQ Monitoring Site Houston Ship Channel Figure 2. Map of Harris County, Texas with TCEQ monitoring site locations In order to assess potential exposure to benzene and 1,3-butadiene, we developed a method to estimate ambient levels of these compounds for each census tract, based on the data obtained from the 14 monitoring sites without any adjustment for meteorological conditions (a limitation that we note later in the report). We first calculated the annual arithmetic mean concentrations of benzene and 1,3butadiene for each monitoring location. We refer to this value as the site-specific annual mean. We then calculated the arithmetic mean concentrations for each of these two compounds over the entire period, 1992-2003, for each site. We refer to this value as the site-specific period mean. Because not all sites were active for all 12 years, we calculated the site-specific period mean as a weighted average of the annual means. Next, we determined, for each census tract, the distance of the centroid to each of the 14 monitoring sites, using ArcGIS(44)software. We were then able to assign each census tract an ambient value for benzene and for 1,3-butadiene based on the average of the site specific period means from the three closest monitoring sites. This value is the tract-specific mean. The tract-specific means for benzene and 1,3-butadiene were categorized based on risk concentrations calculated using the California Office of Environmental Health Hazard Assessment (OEHHA) unit risk estimates (URE) as compiled by the Houston Mayor's Task Force on the Health Effects of Air Pollution. (1) These UREs reflect the excess cancer risk expected to occur as a result from lifetime exposure to 1 g/m3 of the chemical. (1) In order to calculate a risk concentration for a given lifetime cancer risk (1 in a million, for example), the lifetime cancer risk is divided by the URE. (1) Because the UREs are reported by the Task Force in units of (g/m3)-1 and the TCEQ data is reported in units of ppbV, we first calculated the risk concentrations in units of g/m3 and then transformed them based on the following equation: g / m3 = ppbV * MolecularWeight * 273K 22.4 * 298K For example, given a unit risk estimate for benzene of 0.000029, the risk concentration based on 1 in a million risk is 0.034 g/m3 or 0.0107 ppbV. For 1,3-butadiene, the risk concentration based on 1 in a million risk is 0.00266 ppbV. Because the minimum site-specific period means for each benzene and 1,3-butadiene were greater than these concentrations, we chose to base our categorization of the tractspecific means on a lifetime risk of 1 in 10,000 rather than 1 in 1,000,000. The values corresponding to a 1 in 10,000 lifetime cancer risk for benzene and 1,3-butadiene were: 1.07 ppbV and 0.266 ppbV, respectively. For benzene, the tract-specific means were categorized into three groups, <1.07 ppbV, 1.07 - 1.19 ppbV and 1.20 ppbV with the referent group being those values below 1.07 ppbV. For 1,3-butadiene, the tract-specific means were categorized into four groups, with the referent group being those values below 0.266 ppbV. The remaining three categories were: 0.266 - 0.381, 0.382 - 1.14 and 1.15. The benzene categorization resulted in one less category than that for 1,3-butadiene because only 10% of census tracts had tract-specific means greater than 1.07 ppbV; therefore, more than three exposure categories would have resulted in very sparse numbers. We chose to categorize tract-specific means for benzene and 1,3-butadiene in this manner rather than by percentiles because we felt using the UREs to guide the categorization would result in more biologically meaningful categories. The use of UREs were chosen over reference exposure levels (REL) because, unlike the derivation of RELs, the approach to deriving UREs does not assume a threshold level.(49) Therefore, the URE is more appropriate when considering cancer risk. Statistical Analyses: Confounding Factors: For all analyses, we considered age at diagnosis, race / ethnicity and socioeconomic status as potential confounders. Further, because the etiology of childhood cancers (age at diagnosis <20 years) is likely different than adult cancers (age at diagnosis 20 years), we analyzed these two groups separately. To further account for the potential influence of age at diagnosis on cancer incidence within these groups, we adjusted for age in 5 year intervals within the children and adult groups. Additionally, we stratified analyses of adult cancers by gender. For analyses of childhood cancers, we adjusted for gender. Because socioeconomic status is a complex term and involves such attributes as income, education level and household characteristics,(50) and because it is difficult to obtain individual measures of these attributes, we constructed a composite index of socioeconomic status based on census tract level data from the 2000 Census, using principal component analysis to identify the most relevant factors. This variable was then categorized into quartiles based on its distribution in the state of Texas. While we did not include race as an independent variable in the statistical models, we did include a factor representing the percent of the census tract population that is Hispanic and the percent of the census tract population that is African-American in the composite socioeconomic status variable. Including race as an additional independent variable would have produced too fine a stratification for constructing census tract level case counts and population totals, resulting in a loss of power and potentially affecting the ability of the models to converge properly. Analysis #1 Distance from the Houston Ship Channel: To test the association between distance from the HSC and lymphohematopoietic cancer incidence, we used poisson regression. The dependent variable in this regression was case count, indexed by census tract, gender and age group. The buffer zones indicating distance from the HSC was modeled as an ordinal variable with 6 levels, as previously discussed. Other independent variables included in the model were gender (childhood cancers only), age group, and race/ ethnicity and socioeconomic status. An offset equal to the log of the population total for each census tract, gender and age group combination was used to account for the different population sizes in each of these groups. Additionally, generalized estimating equations were used to account for clustering of incidence rates within census tracts. Because the etiology of lymphohematopoietic cancer differs by cancer subtypes, analyses were conducted separately for adult and childhood cancers as well as by the following cancer types: Hodgkin's lymphoma, non-Hodgkin's lymphoma, Myeloma, all leukemias, acute lymphocytic leukemia, acute myeloid leukemia, chronic lymphocytic leukemia, and chronic myeloid leukemia. Analysis #2 Compound-Specific Exposure: We conducted a second poisson regression analysis to test the association between lymphohematopoietic cancer incidence and ambient levels of benzene and 1,3-butadiene, separately. As in the previous regression, the dependent variable for this second analysis was case count, indexed by census tract, gender and age group. The independent variables were gender, age group, socioeconomic status and the tract-specific mean for either benzene or 1,3-butadiene, categorized as already discussed. An offset equal to the log of the population total for each census tract, gender and age group combination was used to account for the different population sizes in each of these groups. Generalized estimating equations were also used to account for clustering of incidence rates within census tracts. This analysis was conducted separately for adult and childhood cancers as well as by the following cancer types: Hodgkin's lymphoma, non-Hodgkin's lymphoma, myeloma, all leukemias, acute lymphocytic leukemia, acute myeloid leukemia, chronic lymphocytic leukemia, and chronic myeloid leukemia Results Analysis among children, adjusting for gender, age, and socio-economic status: Based on the buffer analysis (table 1), no trend toward increasing risk of any lymphohematopoietic cancer by distance from the HSC in 2 mile intervals was noted. However, rates of acute lymphocytic leukemia were significantly higher for those living 0-2 miles from the HSC (aRR = 1.56; p=0.0105) compared with those living greater than 10 miles from the HSC. Additionally, higher 1,3-butadiene levels (>1.15 ppbV relative to <0.266 ppbV) were associated with acute lymphocytic leukemia (p=0.05), acute myeloid leukemia (p 0.03) and all leukemias (p = 0.02) and a trend of increasing 1,3butadiene levels and increasing leukemia rates was noted for these cancers (table 2). Higher benzene levels (1.07-1.19 ppbV or 1.20 ppbV relative to <1.07 ppbV) were not associated with any lymphohematopoietic cancer in children (table 3). Analysis among adults, stratifying by gender and adjusting for age and socio-economic status: No significant trends of increasing lymphohematopoietic cancer risk with increasing proximity to the HSC was observed for men or women for any specific cancer type (table 4). Although we did observe several significant associations between Hodgkin's disease, non-Hodgkin's Lymphoma and myeloma and specific distances from the HSC, no consistent pattern of risk emerged for either gender (table 4). Finding an association for both genders would provide greater support for an environmental versus an occupational route of exposure. Among male adults, higher 1,3-butadiene levels were associated with non-Hodgkin's lymphoma and a significant trend was also noted (p=0.010) (table 5). A similar pattern was not observed for females. Although we also found a 57% increased risk of Hodgkin's disease among females associated with 1,3-butadiene levels of 0.382 - 1.14 ppbV relative to the lowest level of <0.226 ppbV, there was not a significant increasing trend (table 5). We did also observe an increased risk of chronic myeloid leukemia in men associated with benzene levels between 1.07 -1.19 ppbV (table 6). However, higher benzene levels were not associated with a significant trend toward increasing lymphohematopoietic cancer rates for men or women for any specific cancer type (table 6). Discussion Our findings and methods are similar to those reported by Reynolds et al. (14); they report a 21% increase in childhood leukemia associated with those census tracts with higher modeled ambient levels of HAPs. On the basis of our assessment using residential proximity to the HSC, those living within 2 miles had a 56% higher childhood lymphocytic leukemia rate than did those living greater than 10 miles from the HSC (p=0.01). When we evaluated risks using monitored ambient levels of specific HAPs, those living in areas with the highest estimated 1,3-butadiene levels had significantly higher childhood acute lymphocytic leukemia, acute myeloid leukemia and all leukemias. Neither proximity to the HSC nor levels of benzene or 1,3-butadiene were associated with adult lymphohematopoietic cancer. Our finding that higher 1,3-butadiene levels were associated with higher non-Hodgkin's lymphoma rates (among adult males only) may be due to an occupational exposure confounding the association. Had we found that non-Hodgkin's lymphoma rates were associated with higher 1,3butadiene levels in both males and females, we would have better evidence of an unconfounded association. Our team has taken the following efforts to limit bias and improve study validity. We limited selection bias by using population based TCR data. To the extent that TCR is complete and accurate for lymphohematopoietic cancer from years 1995-2003, we have included all cases in the population. We used Census 2000 as the data source for the denominator at the census tract level which is the smallest unit of analysis possible for rare cancers while maintaining sufficient study power. One limitation of ecologic studies is that we do not have individual data on exposure, disease, and confounding factors. Because we used the smallest geographic unit possible, census tracts, we got closer to having individual level data and therefore have reduced the impact of ecological misclassification bias on study findings. Our use of census data to additionally control for important confounders including age, gender and socioeconomic status also limits confounding bias common in ecological studies. Because cancer rates vary by age and socioeconomic status and because there is considerable variation in the distributions of these factors within census tracts across Houston, we adjusted our analyses for both factors. Smoking may also confound the putative association between exposure to ambient benzene and 1,3-butadiene since smoking may provide another source of exposure to these chemicals. It has been shown in numerous studies that aggregate measures of socioeconomic status may be viewed as a proxy for individual level smoking status. (51-54) By adjusting rate ratios for socioeconomic status we are also attempting to control for confounding due to occupational exposures, although residual confounding is likely still present. We attempted to address this residual confounding in adults through stratification by gender, as males are more likely to be affected by this type of exposure. At the suggestion of several environmental scientists, we repeated our analyses for childhood leukemias using the USEPA's 1999 National Air Toxics Assessment (NATA) modeled ambient 1,3butadiene and benzene levels at the census tract level. In general we saw a similar pattern to that reported in table 2 based on our TCEQ estimates of ambient levels. From table 2, for all leukemias in children, the rate ratio for comparing the highest grouped ambient level to the lowest was 1.40 and the p-value was 0.02. Using the NATA data for childhood leukemias, again comparing the highest with the lowest 1,3-butadiene levels, this rate ratio was 1.32 (p=0.09). The latter was of borderline statistical significance. Note however that the actual levels of 1,3-butadiene differ for the two analyses yet the relative grouping based on quartiles are similar. Further and more sophisticated analyses are planned to compare the NATA modeled estimates with the actual TCEQ monitoring data to determine reliability of our categorizations and the extent of exposure misclassification. Additionally, techniques known as kriging will be used to better estimate ambient benzene and 1,3-butadiene levels by census tracts based on TCEQ monitoring data. This is the one of the first health effects studies to use monitoring data as a measure of ambient air levels of HAPs in Houston. However, there are challenges to using these data. The strategy that we used to estimate ambient levels for each census tract was limited by several factors. First, the TCEQ monitoring sites are not distributed randomly across Houston and not all monitors have been active over the entire period of interest (1995-2003). This presented a particular challenge for those census tracts in the presumed low exposed area where few monitors exist (see Figure 2). The distribution of monitors raises a question related to the appropriateness of using monitoring data collected for one purpose (i.e., assessing compliance) for another (i.e., assessing an exposure-response relationship). Note however that these data are the only existing source of potential ambient HAP exposures collected over time. Secondly, the methods used in this analysis to estimate ambient levels of pollutants for each census tract did not account for meteorological conditions, which affect the transport and fate of air pollutants. Analyses are planned that will improve upon the methods used to estimate ambient levels by census tract and thereby limit the biases that may have been introduced in the current investigation. Finally, we opted to estimate ambient levels for each census tract in this investigation by averaging data collected over the entire period (along with cancer incidence). While long-term HAP exposures are more likely to be etiologically related to cancer development, this approach raises questions regarding temporality of the potential exposure-disease relationship. We did exclude those cancers that occurred prior to the recorded years of HAP monitoring data reported by TCEQ as a means to address this issue of temporal sequence. In subsequent analyses we will attempt a more comprehensive assessment of lagged exposures relative to cancer incidence. However, these analyses will be limited by study power as the annual number of cancer cases are relatively small. We opted to explore HAP ambient levels with these specific cancers because of the relatively short latency period from exposure to cancer development. Because the TCEQ monitoring data and the TCR cancer data have only been collecting data over the past decade, this preliminary report represents the first opportunity to correlate ambient HAP levels with those cancers for which we might expect to see an association given the shorter latency periods. Additionally, because the California OEHHA unit risk estimates (URE) are not the same as URE obtained from other sources, such as the United States Environment Protection Agency, had we used another source, our resulting classification would have been different which could then have affected our final results. We used distance of census tracts from the HSC as a crude surrogate for proximity to higher ambient levels of air pollutants given the concentration of petroleum and chemical manufacturing facilities in the HSC. However, we recognize that there is variability in ambient air levels within the HSC, that there may be industrial facilities that emit benzene and 1,3-butadiene located outside the HSC, and that mobile sources contribute to air pollution as well. Nonetheless, our preliminary results of an association between childhood leukemia rates and both proximity to industrial sources (defined as distance from the HSC) and ambient levels of HAPs, derived from TCEQ monitoring data, suggest that a true association may exist. Given the increased sensitivity of children to environmental toxicants including carcinogens(55) and that children metabolize toxins differently than adults with behaviors which present a unique opportunity for exposure(56), it is plausible that we might observe an association between HAPs and lymphohematopoietic cancer in children but not adults. Other than the increased sensitivity of children, studying childhood cancer development presents advantages over that of adults since there are few known risk factors for childhood cancer and because the latency of childhood cancers is much shorter. Additional studies in child and adult populations will be necessary to allow for more definitive statements to be made regarding whether increased cancer risk is associated with ambient air pollution. In conclusion, the primary limitation of this study lies in the `exposure assessment' and the potential for biases that may have affected study validity. In fact, we do not have individual exposure data nor were we able to assess the validity of using monitoring data as a surrogate measure for personal exposure. Thus, the results presented here should be viewed as preliminary. Nonetheless, additional analytic studies with more refined methodology are warranted to further evaluate the association between increased lymphohematopoietic cancer risk and ambient levels of hazardous air pollutants, the results of which might inform policies intended to mitigate health risks of Houstonians in the future. Table 1. Crude annual incidence rates and adjusted rate ratios for association between distance from the Houston Ship Channel (HSC) and lymphohematopoietic cancer incidence among children (aged <20) Cancer Type Hodgkin's Disease Distance from HSC (Miles) >10 8-10 6-8 4-6 # Census Tracts 344 60 62 71 Crude Annual # Cases Incidence Rate (per Rate Ratio* p-value 100,000) 64 1.168 1.00 10 1.444 1.17 0.6169 9 1.452 1.15 0.6917 11 1.348 1.06 0.8794 p-value for trend 2-4 77 7 0-2 35 4 0.609 0.867 0.47 0.0995 0.61 0.3708 0.1522 >10 344 46 0.839 1.00 non-Hodgkin's Lymphoma 8-10 6-8 4-6 2-4 0-2 >10 60 7 62 5 71 8 77 10 35 3 344 274 1.011 0.807 0.981 0.870 0.650 5.000 1.28 0.5825 1.05 0.9076 1.30 0.4597 1.14 0.7044 0.91 0.8947 0.7031 1.00 All Leukemia Acute Lymphocytic Leukemia 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 60 38 62 36 71 44 77 55 35 29 344 207 60 32 62 23 71 34 77 40 5.487 5.809 5.393 4.783 6.285 3.777 4.620 3.712 4.167 3.479 1.09 0.6412 1.17 0.4335 1.11 0.5282 0.96 0.8322 1.22 0.2688 0.6189 1.00 1.21 0.2974 0.98 0.9917 1.16 0.4405 0.96 0.8280 0-2 35 27 >10 344 40 8-10 60 4 Acute Myeloid 6-8 62 8 Leukemia 4-6 71 5 2-4 77 9 0-2 35 1 5.852 0.730 0.578 1.291 0.613 0.783 0.217 1.56 0.0105 0.3274 1.00 0.71 0.5215 1.58 0.3301 0.74 0.5577 0.87 0.7434 0.20 0.1290 0.3470 *Rate Ratios Adjusted for Age, Gender, Ethnicity, and Socioeconomic Status Note: There were no childhood Myeloma cases; Results are not shown for chronic lymphocytic leukemia or chronic myeloid leukemia due to small numbers (4 and 16 total cases, respectively) Table 2. Crude annual incidence rates and adjusted rate ratios for association between ambient 1,3- Butadiene levels and cancer Incidence among children aged 0-19 Cancer Type 1,3-Butadiene Level* 1 # Census Tracts # Cases 148 22 Crude Annual Incidence Rate (per 100,000) 0.86 Rate Ratio** 1.00 p-value p-value for trend Hodgkin's 2 158 25 1.21 1.53 0.171 Disease 3 137 22 1.23 1.67 0.152 4 206 36 1.29 1.80 0.071 0.099 non-Hodgkin's 1 2 148 20 158 20 0.78 0.96 1.00 1.44 0.264 Lymphoma 3 137 18 1.01 1.73 0.128 4 206 21 0.75 1.25 0.510 0.559 1 148 116 4.51 1.00 All Leukemia Acute Lymphocytic Leukamia 2 3 4 1 2 3 4 1 158 99 137 93 206 168 148 89 158 76 137 68 206 130 148 12 4.77 5.21 6.02 3.46 3.67 3.81 4.66 0.47 1.10 1.19 1.40 1.00 1.09 1.11 1.38 1.00 0.530 0.292 0.024 0.629 0.528 0.051 0.017 0.041 Acute Myeloid 2 158 13 0.63 1.50 0.359 Leukemia 3 137 14 0.78 1.79 0.217 4 206 28 1.00 2.53 0.033 0.026 *1,3-Butadiene levels correspond to: level 1 =<0.266 ppbV; level 2 = 0.266 - 0.381 ppbV; level 3 = 0.382-1.14 ppbV; and level 4 = >1.15 ppbV **Rate Ratios Adjusted for Age, Gender, Ethnicity and Socioeconomic Status Note: There were no childhood Myeloma cases; Results are not shown for chronic lymphocytic leukemia or chronic myeloid leukemia due to small numbers (4 and 16 total cases, respectively) Table 3. Crude annual incidence rates and adjusted rate ratios for association between ambient Benzene levels and cancer Incidence among children aged 0-19 Cancer Type Benzene Level* # Census Tracts # Cases Crude Annual Incidence Rate (per 100,000) Rate Ratio** p-value p-value for trend Hodgkin's Disease non-Hodgkin's Lymphoma All Leukemia Acute Lymphoctyic Leukemia Acute Myeloid Leukamia 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 543 85 79 14 27 6 543 66 79 8 27 5 543 399 79 62 27 15 543 308 79 42 27 13 543 55 79 11 27 1 1.11 1.27 1.27 0.86 0.73 1.06 5.22 5.63 3.17 4.03 3.81 2.75 0.72 1.00 0.21 1.00 1.14 0.721 1.16 0.745 0.673 1.00 0.98 0.965 1.36 0.487 0.587 1.00 1.01 0.943 0.60 0.078 0.164 1.00 0.86 0.450 0.66 0.183 0.147 1.00 1.17 0.733 0.27 0.186 0.372 *Benzene Levels correspond to: level 1 = <1.07 ppbV; level 2 = 1.07 - 1.19 ppbV; and level 3= 1.20 ppbV **Rate Ratios Adjusted for Age, Gender, Ethncity, and Socioeconomic Status Note: There were no childhood Myeloma cases; Results are not shown for chronic lymphocytic leukemia or chronic myeloid leukemia due to small numbers (4 and 16 total cases, respectively) Table 4. Crude annual incidence rates and adjusted rate ratios for association between distance from the Houston Ship Channel (HSC) and lymphohematopoietic cancer incidence among adults (aged 20) Cancer Type Gender Distance from HSC (Miles) # Census Tracts # Cases Crude Annual Incidence (per 100,000) Rate Ratio* p-value p-value for trend Male Acute Myeloid Leukemia Female Male Chronic Lymphocytic Leukemia Female >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 344 169 60 37 62 277 71 37 77 37 35 16 344 188 60 27 62 31 71 37 77 34 35 11 344 233 60 37 62 41 71 45 77 35 35 15 344 162 60 31 62 22 71 36 77 39 35 7 2.79 4.38 33.97 3.84 3.26 3.78 2.94 3.02 3.66 3.77 3.00 2.71 3.85 4.38 5.03 4.67 3.08 3.54 2.53 3.47 2.60 3.67 3.44 1.72 1.00 1.37 1.14 1.34 1.25 1.49 1.00 0.89 1.19 1.22 1.07 1.00 1.00 0.92 1.14 1.04 0.76 0.82 1.00 1.07 0.95 1.29 1.37 0.69 0.1272 0.4942 0.0956 0.2300 0.1376 0.5832 0.3712 0.2758 0.7170 0.9966 0.6582 0.4452 0.8375 0.1467 0.4916 0.7148 0.8166 0.1922 0.1091 0.3529 0.052 0.431 0.327 0.303 Chronic Myeloid Leukemia Male Female >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 344 106 60 18 62 24 71 20 77 16 35 8 344 87 60 9 62 8 71 19 77 11 35 6 1.75 2.13 2.94 2.08 1.41 1.89 1.36 1.01 0.94 1.94 0.97 1.48 1.00 1.04 0.8925 1.50 0.1996 1.04 0.8713 0.73 0.2530 0.88 0.7650 0.613 1.00 0.65 0.1769 0.65 0.2090 1.25 0.3999 0.67 0.2202 1.03 0.9505 0.608 Male Myeloma Female >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 344 304 60 53 62 62 71 84 77 69 35 29 344 310 60 43 62 58 71 83 77 66 35 18 5.03 6.28 7.60 8.72 6.08 6.85 4.84 4.81 6.85 8.46 5.83 4.43 1.00 1.04 0.8132 1.29 0.0733 1.40 0.0105 1.07 0.6517 1.16 0.4786 0.111 1.00 0.77 0.1122 1.09 0.5622 1.19 0.1884 0.89 0.4139 0.62 0.0608 0.483 Table 4. (cont.) Crude annual incidence rates and adjusted rate ratios for association between distance from the Houston Ship Channel (HSC) and lymphohematopoietic cancer incidence among adults (aged 20) Cancer Type Gender Male Acute Myeloid Leukemia Female Male Chronic Lymphocytic Leukemia Female Chronic Myeloid Leukemia Male Female Male Myeloma Female Distance from HSC (Miles) >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 >10 8-10 6-8 4-6 2-4 0-2 # Census Tracts # Cases 344 169 60 37 62 277 71 37 77 37 35 16 344 188 60 27 62 31 71 37 77 34 35 11 344 233 60 37 62 41 71 45 77 35 35 15 344 162 60 31 62 22 71 36 77 39 35 7 344 106 60 18 62 24 71 20 77 16 35 8 344 87 60 9 62 8 71 19 77 11 35 6 344 304 60 53 62 62 71 84 77 69 35 29 344 310 60 43 62 58 71 83 77 66 35 18 Crude Annual Incidence (per 100,000) 2.79 4.38 33.97 3.84 3.26 3.78 2.94 3.02 3.66 3.77 3.00 2.71 3.85 4.38 5.03 4.67 3.08 3.54 2.53 3.47 2.60 3.67 3.44 1.72 1.75 2.13 2.94 2.08 1.41 1.89 1.36 1.01 0.94 1.94 0.97 1.48 5.03 6.28 7.60 8.72 6.08 6.85 4.84 4.81 6.85 8.46 5.83 4.43 Rate Ratio* 1.00 1.37 1.14 1.34 1.25 1.49 1.00 0.89 1.19 1.22 1.07 1.00 1.00 0.92 1.14 1.04 0.76 0.82 1.00 1.07 0.95 1.29 1.37 0.69 1.00 1.04 1.50 1.04 0.73 0.88 1.00 0.65 0.65 1.25 0.67 1.03 1.00 1.04 1.29 1.40 1.07 1.16 1.00 0.77 1.09 1.19 0.89 0.62 p-value 0.1272 0.4942 0.0956 0.2300 0.1376 0.5832 0.3712 0.2758 0.7170 0.9966 0.6582 0.4452 0.8375 0.1467 0.4916 0.7148 0.8166 0.1922 0.1091 0.3529 0.8925 0.1996 0.8713 0.2530 0.7650 0.1769 0.2090 0.3999 0.2202 0.9505 0.8132 0.0733 0.0105 0.6517 0.4786 0.1122 0.5622 0.1884 0.4139 0.0608 p-value for trend 0.052 0.431 0.327 0.303 0.613 0.608 0.111 0.483 *Rate Ratios adjusted for age, ethnicity, and socioeconomic status Table 5. Crude annual incidence rates and adjusted rate ratios for association between ambient 1,3- Butadiene levels and cancer incidence among adults (aged 20) Cancer Type Gender 1,3Butadiene Level* # Census Tracts # Cases Crude IR (per 100,000) Rate Ratio** p-value p-value for trend 1 148 89 3.31 1.00 Male 2 3 158 81 3.52 1.12 0.475 137 64 3.12 1.06 0.741 Hodgkin's 4 206 81 2.54 0.84 0.307 0.222 Disease 1 148 55 1.92 1.00 Female 2 3 158 48 1.95 1.04 0.842 137 53 2.67 1.57 0.025 4 206 75 2.23 1.19 0.369 0.238 1 148 487 18.11 1.00 Male 2 3 158 464 137 413 20.14 20.14 1.08 0.250 1.29 0.015 non-Hodgkin's 4 206 618 19.41 1.20 0.014 0.010 Lymphoma 1 148 444 15.54 1.00 Female 2 3 158 435 137 364 17.71 18.36 0.97 0.645 1.09 0.327 4 206 508 15.10 0.91 0.168 0.226 1 148 312 11.60 1.00 Male 2 3 158 303 137 248 13.15 12.09 0.97 0.702 0.96 0.669 All Leukemia 4 206 360 11.31 0.91 0.296 0.296 1 148 236 8.26 1.00 Female 2 3 158 254 137 212 10.34 10.70 1.06 0.583 1.21 0.112 4 206 320 9.51 1.09 0.400 0.358 1 148 23 0.86 1.00 Acute Lymphocytic Leukemia Male Female 2 3 4 1 2 3 158 19 0.82 0.88 0.662 137 15 0.73 0.76 0.445 206 32 1.01 1.03 0.922 0.843 148 14 0.49 1.00 158 15 0.61 1.09 0.813 137 13 0.66 1.10 0.831 4 206 23 0.68 1.22 0.597 0.600 Table 5. (cont.) Crude annual incidence rates and adjusted rate ratios for association between ambient 1,3-Butadiene levels and cancer incidence among adults (aged 20) Cancer Type Gender Male Acute Myeloid Leukemia Female Chronic Lymphocytic Leukemia Male Female 1,3- Butadiene Level* 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 # Census Tracts # Cases Crude IR (per 100,000) 148 74 158 89 137 54 206 106 148 81 158 81 137 68 206 98 148 105 158 110 137 85 206 106 148 61 158 78 137 63 206 95 2.75 3.86 2.63 3.33 2.83 3.30 3.43 2.91 3.90 4.77 4.15 3.33 2.13 3.17 3.18 2.82 Rate Ratio** 1.00 1.24 0.92 1.20 1.00 1.03 1.23 1.01 1.00 1.01 0.95 0.79 1.00 1.24 1.47 1.31 p-value p-value for trend 0.143 0.670 0.240 0.849 0.243 0.955 0.962 0.754 0.119 0.214 0.054 0.117 0.470 0.884 0.085 0.136 Chronic Myeloid Leukemia Male Female Myeloma Male Female 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 148 53 158 34 137 48 206 57 148 39 158 34 137 28 206 39 148 125 158 161 137 128 206 187 148 119 158 132 137 118 206 209 1.97 1.48 2.34 1.79 1.36 1.38 1.41 1.16 4.65 6.99 6.24 5.87 4.16 5.37 5.95 6.21 1.00 0.68 1.07 0.88 1.00 0.90 0.99 0.83 1.00 1.18 1.08 1.06 1.00 0.92 0.99 1.11 0.093 0.791 0.601 0.650 0.983 0.410 0.177 0.582 0.641 0.521 0.921 0.405 0.956 0.458 0.988 0.200 *1,3-Butadiene levels correspond to: level 1 = <0.266 ppbV; level 2 = 0.266 - 0.381 ppbV; level 3 = 0.382-1.14 ppbV; and level 4 >1.15 ppbV **Rate Ratios Adjusted for Age, Ethnicity, and Socioeconomic Status Table 6. Crude annual incidence rates and adjusted rate ratios for association between ambient Benzene levels and cancer Incidence among adults (aged 20) Cancer Type Gender Benzene Level* # Census Tracts # Cases 1 543 280 Crude Incidence Rate (per 100,000) 3.20 Rate Ratio** 1.00 p-value p-value for trend Male 2 79 25 2.41 0.85 0.518 Hodgkin's Disease Female 3 1 2 27 10 543 199 79 21 2.33 2.17 2.01 0.81 1.00 1.05 0.492 0.841 0.392 3 27 11 2.47 1.22 0.551 0.560 nonHodgkin's Lymphoma Male Female 1 2 3 1 2 543 1728 79 185 27 69 543 1487 79 191 19.73 17.80 16.04 16.22 18.25 1.00 0.98 0.93 1.00 1.11 0.826 0.588 0.178 0.585 3 27 73 16.37 1.16 0.528 0.285 1 543 1038 11.85 1.00 Male 2 79 137 13.18 1.05 0.665 All Leukemia 3 27 48 1 543 884 11.16 9.64 1.00 1.000 0.861 1.00 Female 2 79 102 9.75 0.98 0.871 3 27 36 8.07 0.92 0.566 0.609 1 543 75 0.86 1.00 Acute Lymphoctyic Leukemia Male Female 2 3 1 2 79 12 27 2 543 55 79 8 1.15 0.47 0.60 0.76 1.34 0.51 1.00 1.02 0.454 0.284 0.957 0.650 3 27 2 0.45 0.66 0.566 0.658 1 543 287 3.28 1.00 Male 2 79 25 2.41 0.65 0.052 Acute Myeloid Leukemia Female 3 1 2 27 11 543 284 79 31 2.56 3.10 2.96 0.80 1.00 1.04 0.402 0.844 0.118 3 27 13 2.92 1.12 0.660 0.656 Chronic Lymphocytic Leukemia Male Female 1 2 3 1 2 3 543 345 79 44 27 17 543 262 79 28 27 7 3.94 4.23 3.95 2.86 2.68 1.57 1.00 0.99 1.08 1.00 0.93 0.64 0.940 0.763 0.769 0.256 0.839 0.293 Chronic Myeloid Leukemia Male Female 1 2 3 1 2 543 155 79 32 27 5 543 120 79 14 1.77 3.08 1.16 1.31 1.34 1.00 1.71 0.71 1.00 1.04 0.037 0.500 0.895 0.589 3 27 6 1.35 1.12 0.759 0.761 Myeloma Male Female 1 2 3 1 2 3 543 501 79 70 27 30 543 483 79 71 27 24 5.72 6.73 6.98 5.27 6.78 5.38 1.00 0.91 1.11 1.00 0.94 0.89 0.516 0.595 0.675 0.584 0.908 0.525 *Benzene Levels correspond to: level 1 = <1.07 ppbV; level 2 = 1.07- 1.19 ppbV and level 3 = 1.20 ppbV **Rate Ratios Adjusted for Age , Ethnicity, and Socioeconomic Status References 1. A closer look at air pollution in Houston. Identifying priority health risks: Report of the Mayor's task force on the health effects of air pollution. Houston, Texas: University of Texas, Health Science Center. Institute for Health Policy.; 2006. 2. United States Census Bureau. Censtats Databases. County Business Patterns. . In; Retrieved On: 4/5/06. http://www.census.gov/epcd/cbp/view/cbpview.html. 3. Lists of IARC Evaluations. World Health Organization, International Agency for Research on Cancer, Lyon, France, 1993. 4. Beall C, Delzell E, Rodu B, Sathiakumar N, Myers S. Cancer and benign tumor incidence among employees in a polymers research complex. . J Occup Environ Med 2001;43:914-924. 5. Delzell E, Macaluso M, Sathiakumar N, Matthews R. Leukemia and exposure to 1,3-butadiene, styrene and dimethyldithiocarbamate among workers in the synthetic rubber industry. . Chem Biol Interact 2001;135::515-34. 6. Divine BJ, Hartman CM. Update of a study of crude oil production workers 1946-94. Occup Environ Med 2000;57:411-417. 7. Divine BJ, Hartman CM. A cohort mortality study among workers at a 1,3 butadiene facility. . Chem Biol Interact 2001;135-136:535-53. 8. Divine BJ, Hartman CM, Wendt JK. Update of the Texaco mortality study 1947-93: Part I. Analysis of overall patterns of mortality among refining, research, and petrochemical workers. . Occup Environ Med 1999;56:167-173. 9. Gun RT, Pratt NL, Griffith EC, Adams GG, Bisby JA, Robinson KL. Update of a prospective study of mortality and cancer incidence in the Australian petroleum industry. Occup Environ Med 2004;61:150-156. 10. Huebner WW, Wojcik NC, Rosamilia K, Jorgensen G, Milano CA. Mortality updates (19701997) of two refinery/petrochemical plant cohorts at Baton Rouge, Louisiana, and Baytown, Texas. . J Occup Environ Med 2004;46:1229-1245. 11. Knox EG. Childhood cancers and atmospheric carcinogens. J Epidemiol Community Health 2005;59:101105. 12. Lewis RJ, Schnatter AR, Drummond. I, et a. Mortality and cancer morbidity in a cohort of Canadian petroleum workers. Occup Environ Med 2003;60:918-928. 13. Raabe GK, Collingwood KW, Wong O. An updated mortality study of workers at a petroleum refinery in Beaumont, Texas. Am J Ind Med 1998;33:61-81. 14. Reynolds P, Von Behren J, Gunier RB, Goldberg DE, Hertz A, Smith DF. Childhood Cancer Incidence Rates and Hazardous Air Pollutants in California: An Exploratory Analysis. Environ Health Perspect 2006;111:663-668. 15. Reynolds R, Von Behren J, Gunier RB, Goldberg DE, Hertz A, Smith D. Traffic patterns and childhood cancer incidence rates in California, United States. Cancer Causes and Control 2002;13:665673. 16. Rodu B, Delzell E, Beall C, Sathiakumar N. Mortality among employees at a petrochemical research facility. Am J Ind Med 2001;39:29-41. 17. Sathiakumar N, Delzell E, Hovinga M, et a. Mortality from cancer and other causes of death among synthetic rubber workers. Occup Environ Med 1998;55:230-235. 18. Sathiakumar N, Delzell E, Rodu B, Beall C, Myers S. Cancer incidence among employees at a petrochemical research facility. J Occup Environ Med 2001;43:166-74. 19. Satin KP, Bailey WJ, Newton KL, Ross AY, Wong O. Updated epidemiological study of workers at two California petroleum refineries, 1950-95. . Occup Environ Med 2002;59:248256. 20. Tsai SP, Chen VW, Fox EE, et a. Cancer incidence among refinery and petrochemical employees in Louisiana, 1983-1999. Ann Epidemiol 2004;14:722-30. 21. Wong O, Harris F, Rosamilia K, Raabe GK. An updated mortality study of workers at a petroleum refinery in Beaumont, Texas, 1945 to 1996. J Occup Environ Med 2001;43:384-401. 22. Matanoski GM, Schwartz L. Mortality of workers in styrene-butadiene polymer production. J Occup Med 1987;29(8):675-80. 23. Matanoski GM, Santos-Burgoa C, Schwartz L. Mortality of a cohort of workers in the styrenebutadiene polymer manufacturing industry (1943-1982). Environ Health Perspect 1990;86:10717. 24. Rinsky RA, Hornung RW, Silver SR, Tseng CY. Benzene exposure and hematopoietic mortality: A long-term epidemiologic risk assessment. Am J Ind Med 2002;42:474-480. 25. Bloemen LJ, Youk A, Bradley TD, Bodner KM, Marsh G. Lymphohaematopoietic cancer risk among chemical workers exposed to benzene. Occup Environ Med 2004;61:270-274. 26. Ruder AM, Ward EM, Dong M, Okun AH, Davis-King K. Mortality patterns among workers exposed to styrene in the reinforced plastic boatbuilding industry: an update. Am J Ind Med 2004;45:165-176. 27. Tsai SP, Wendt JK, Ransdell JD. A mortality, morbidity, and hematology study of petrochemical employees potentially exposed to 1,3-butadiene monomer. Chem Biol Interact 2001;135-136:555-67. 28. Adegoke OJ, Blair A, Shu XO, et a. Occupational history and exposure and the risk of adult leukemia in Shanghai. Ann Epidemiol 2003;13:485-494. 29. Glass DC, Gray CN, Jolley DJ, et a. Leukemia risk associated with low-level benzene exposure. Epidemiology 2003;14:256-577. 30. Graff JJ, Sathiakumar N, Macaluso M, Maldonado G, Matthews R, Delzell E. Chemical exposures in the synthetic rubber industry and lymphohematopoietic cancer mortality. J Occup Environ Med 2005;47:916-932. 31. Matanoski G, Elliott E, Tao X, Francis M, Correa-Villasenor A, Santos-Burgoa C. Lymphohematopoietic cancers and butadiene and styrene exposure in synthetic rubber manufacture. Ann N Y Acad Sci 1997;837:157-69. 32. Lan Q, Zhang L, Li G, et a. Hematotoxicity in workers exposed to low levels of benzene. Science 2004;306:1774-1776. 33. Patel AS, Talbott EO, Zborowski JV, et a. Risk of cancer as a result of community exposure to gasoline vapors. Arch Environ Health 2004;59:497-503. 34. Sans S, Elliott P, Kleinschmidt I, et al. Cancer incidence and mortality near the Baglan Bay petrochemical works, South Wales. Occup Environ Med 1995;52:217-224. 35. Pan BJ, Hong YJ, Chang GC, Wang MT, Cinkotai FF, Ko YC. Excess cancer mortality among children and adolescents in residential districts polluted by petrochemical manufacturing plants in Taiwan. J Toxicol Environ Health 1994;43:117-129. 36. Yang CY, Chiu HF, Chiu JF, Kao WY, Tsai SS, Lan SJ. Cancer mortality and residence near petrochemical industries in Taiwan. J Toxicol Environ Health 1997;50:265-273. 37. Pekkanen J, Pukkala E, Vahteristo M, Vartiainen T. Cancer incidence around an oil refinery as an example of a small area study based on map coordinates. Environ Res 1995;71:128-134. 38. Belli S, Benedetti M, Comba P, et a. Case-control study on cancer risk associated to residence in the neighbourhood of a petrochemical plant. Eur J Epidemiol 2004;19:49-54. 39. Linos A, Blair A, Gibson RW, et a. Leukemia and non-Hodgkin's lymphoma and residential proximity to industrial plants. Arch Environ Health 1991;46:70-74. 40. Shore DL, Sandler DP, Davey FR, McIntyre OR, Bloomfield CD. Acute leukemia and residential proximity to potential sources of environmental pollutants. Arch Environ Health 1993;48:414-420. 41. Yu CL, Wang SF, Pan PC, et a. Residential Exposure to Petrochemicals and the Risk of Leukemia: Using Geographic Information System Tools to Estimate Individual-Level Residential Exposure. Am J Epidemiol 2006;164:200-7 42. Wilkinson P, Thakrar B, Walls P, et a. Lymphohaematopoietic malignancy around all industrial complexes that include major oil refineries in Great Britain. Occup Environ Med 1999;56:577580. 43. Mannes T, Emmett K, Willmore A, Churches T, Sheppeard V, Kaldor J. Effect of a chemical manufacturing plant on community cancer rates. BMC Public Health 2005;5:34. 44. Environmental Systems Research Institute, ArcGIS Version 9.1 In. 45. Wallace LA, Pellizzari E, Hartwell TD, et al. Personal exposures, indoor-outdoor relationship and breath levels of toxic air pollutants measured for 355 persons in New Jersey. Atmos Environ 1985;19:1651-1661. 46. Wallace LA, Pellizzari E, Hartwell TD, Et al. The California Team study: Breath concentrations and personal exposures to 26 volatile compounds in air and drinking water of 188 residents of Los Angeles, Antioch, and Pittsburg, CA. Atmos Environ 1988;22:2141-2163. 47. Kim YM, Harrad S, Harrison RM. Levels and sources of personal inhalation exposure to volatile organic compounds. Environ Sci Technol 2002;36(24):5405-10. 48. Texas C, on, Environmental, Quality. . The Houston Air Quality Challenge. Rapid economic and population growth create a potent blend for the region's environment. Natural Outlook. In; 2005. 49. Clements AL, Flatt VB, Fraser MP, Hamilton WJ, Ledvina PS, Mathur SK, et al. The control of air toxics: Toxicology motivation and Houston implications. Houston, Texas: Rice University; 2006. 50. Greenwald HP, Borgatta EF, McCorkle R, Polissar N. Explaining reduced cancer survival among the disadvantaged. Milbank Quarterly 1996;74:215-238. 51. Chuang YC, Cubbin C, Ahn D, Winkleby MA. Effects of neighbourhood socioeconomic status and convenience store concentration on individual level smoking. J Epidemiol Community Health 2005;59:568-573. 52. Duncan C, Jones K, Moon G. Smoking and deprivation: are there neighbourhood effects? Soc Sci Med 1999;48:497-505. 53. Kleinschmidt I, Hills M, Elliott P. Smoking behaviour can be predicted by neighbourhood deprivation measures. J Epidemiol Community Health 1995;49 Suppl:S72-S77. 54. Shohaimi S, Luben R, Wareham N, et a. Residential area deprivation predicts smoking habit independently of individual educational level and occupational social class. A cross sectional study in the Norfolk cohort of the European Investigation into Cancer (EPIC-Norfolk). J Epidemiol Community Health 2003;57:270-276. 55. Rice JM, Ward JM. Age dependence of susceptibility to carcinogenesis in the nervous system. Ann N Y Acad Sci 1982;381:274-289. 56. Simcox NJ, Fenske RA, Wolz SA, Lee IC, Kalman DA. Pesticides in household dust and soil: exposure pathways for children of agricultural families. Environ Health Perspect 1995;103:1126-1134.