PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Environ Res. Author manuscript; available in PMC 2011 January 1.
Published in final edited form as:
PMCID: PMC2795078
NIHMSID: NIHMS153621

Residential Proximity to Industrial Facilities and Risk of Non-Hodgkin Lymphoma

Abstract

Industrial pollution has been suspected as a cause of non-Hodgkin lymphoma (NHL), based on associations with chemical exposures in occupational studies. We conducted a case-control study of NHL in four SEER regions of the United States, in which residential locations of 864 cases and 684 controls during the 10 years before recruitment were used to characterize proximity to industrial facilities reporting chemical releases to the Environmental Protection Agency's Toxics Release Inventory (TRI). For each of 15 types of industry (by 2-digit SIC code), we evaluated the risk of NHL associated with having lived within 2 miles of a facility, the distance to the nearest facility (categories of ≤0.5-mile, >0.5-1.0, >1.0-2.0, >2 [referent]), and the duration of residence within 2 miles (10 years, 1-9, 0 [referent]), using logistic regression. Increased risk of NHL was observed in relation to lumber and wood products facilities (SIC 24) for the shortest distance of residential proximity (≤0.5-mile: odds ratio [OR]=2.2, 95% confidence interval [CI]: 0.4-11.8) or longest duration (10 years: OR=1.9, 95% CI: 0.8-4.8); the association with lumber facilities was more apparent for diffuse large B-cell lymphoma (lived within 2 miles: OR=1.7, 95% CI: 1.0-3.0) than for follicular lymphoma (OR=1.1, 95% CI: 0.5-2.2). We also observed elevated ORs for the chemical (SIC 28, 10 years: OR=1.5, 95% CI: 1.1-2.0), petroleum (SIC 29, 10 years: OR=1.9, 95% CI: 1.0-3.6), rubber/miscellaneous plastics products (SIC 30, ≤0.5-mile: OR=2.7, 95% CI: 1.0-7.4), and primary metal (SIC 33, lived within 2 miles: OR=1.3, 95% CI: 1.0-1.6) industries; however, patterns of risk were inconsistent between distance and duration metrics. This study does not provide strong evidence that living near manufacturing industries increases NHL risk. However, future studies designed to include greater numbers of persons living near specific types of industries, along with fate-transport modeling of chemical releases would be informative.

Keywords: non-Hodgkin lymphoma (NHL), geographic information system (GIS), Toxics Release Inventory, Risk Screening Environmental Indicators (RSEI), industrial pollution

Introduction

Industrial pollution has been suspected as a cause of non-Hodgkin lymphoma (NHL), based in part on findings from occupational studies in which chemical exposures were related to increased risk of NHL (Hartge et al., 2006), and because NHL rates increased dramatically during the latter half of the 20th century (Clarke and Glaser, 2002), lagging slightly behind expanded industrial production in the United States (US) (Alexander, 1952).

Several previous studies have investigated the risk of NHL associated with living near industrial facilities. Self-reported residence within 0.5-mile of any type of industrial facility was associated with increased risk of NHL (OR=1.5, 95% CI: 1.1-1.9) in a case-control study conducted in Iowa and Minnesota by Linos et al. (Linos et al., 1991). Specific types of industry associated with increased NHL risk (ORs>1.5) for residence within 2 miles were paper (OR=1.5, 95% CI: 0.9-2.4), petroleum (OR=1.5, 95% CI: 0.7-3.2), and stone, clay and glass industries (OR=1.6, 95% CI: 1.0-2.7). Johnson et al. (Johnson et al., 2003) conducted a case-control study in Canada in which residential proximity to industry was determined through linkage of residential postal codes to public data on the location of facilities from 1960 to the early 1990s. Increased risk associated with residential proximity to industry was most apparent for follicular lymphoma among women (OR=1.5, 95% CI: 1.1-2.0), and for residence within 0.5-mile of copper smelters (OR=10.8, 95% CI: 1.2-97.5) and sulfite pulp mills (OR=3.7, 95% CI: 1.5-9.4). In a recent study based on Spanish mortality records, in which residential proximity to industry was based on a linkage of industry locations to the centroid of the deceased's municipality of residence, only the paper/pulp industry was associated with increased risk of NHL mortality (RR=1.2, 95% CI: 1.1-1.4) (Ramis et al., 2009).

Results from the previous studies, while intriguing, may suffer from recall bias due to self-reported residential proximity to facilities (Linos et al., 1991), or may be affected by non-differential exposure misclassification due to proximity assignment based on residential location as the centroid of a postal code (Johnson et al., 2003) or municipality (Ramis et al., 2009). Our aim was to further investigate whether there is any increase in NHL incidence associated with living near industrial facilities with reported releases of chemicals to the environment. To this end, we conducted a case-control study of the risk of NHL associated with residential proximity to industry in the decade before diagnosis. In our study, we used a more objective source of industry location than that based on recall, and we identified residential location to the street segment or nearest intersection of the participant. Industries of a priori interest were those with elevated risk estimates in the previous studies, in addition to industries that emit agents which are known or suspected to cause lymphohematopoietic cancers, such as petroleum processing for potential releases of benzene and other solvents, and pulp and paper mills for dioxins releases. Most of the suspicion surrounding these agents derives from studies of exposures in occupation, and the question remains whether environmental exposures, either in the ambient environment or from point sources, cause NHL.

Materials and Methods

Study population

The study included participants in a case-control study of NHL, conducted by the National Cancer Institute (NCI) through collaborating centers (Chatterjee et al., 2004). Between July 1998 and June 2000, participants were enrolled from four US Surveillance, Epidemiology, and End Results (SEER) registry areas: the state of Iowa, Los Angeles County, and the metropolitan areas of Detroit, MI and Seattle, WA. The four study sites were chosen to meet the broad aim of the parent study to investigate potential environmental causes of NHL.

Cases included 1,321 patients with newly diagnosed NHL of ages 20 to 74 who did not report HIV infection. Data on case histology were obtained from each local SEER registry and were based on abstracted reports of the diagnosing pathologist. Population controls (n=1,057) were identified by random digit dialing (under age 65) and from Medicare eligibility files (65 years and older) and were frequency matched to cases by age, sex, and race. Overall response percentages were 59% and 44% for cases and controls, respectively. Among eligible participants we attempted to contact, 76% of cases and 52% of controls participated in the study. Written informed consent was obtained from each participant prior to interview. A computer-assisted personal interview was administered that contained questions about demographic characteristics, hair coloring, occupational history, pesticide use history, and other exposures. Human subjects review boards approved the study at the NCI and at all participating institutions.

Residential locations

Global positioning system (GPS) readings were taken outside of the current residence for nearly all participants (99%). Interviewers took the measurements 6.1 m (20 ft) away from the home using a 12-channel handheld Garmin GPS12 Personal Navigator (Garmin International, Inc., Olathe, KS). Because approximately 72% of GPS coordinates were collected before the end of selective availability (deliberate corruption of GPS satellite signals by the US Department of Defense resulting in errors of 100 meters or more) on May 1, 2000 (Office of Science and Technology Policy, 2000), GPS coordinates that were discrepant from the geocoded interview address by more than 200 meters were corrected using a combination of digital orthophotography, Census Bureau street files, road maps, and driving to the residential location to collect new GPS coordinates (Seattle, Los Angeles, parts of Iowa).

Historic addresses were sought in a residential history section of an interviewer-administered questionnaire. Participants were sent a residential calendar in advance of the interview and were asked to provide the complete address of every home they lived in from birth to the current year, indicating the years they moved in and out. They were also asked to provide information about temporary or summer homes where they lived for a total of 2 years or longer.

Residential street addresses were geocoded using the TeleAtlas' (Lebanon, NH) MatchMaker SDK Professional version 4.3 (October 2002) spatial database of roads and a modified version of a Microsoft Visual Basic version 6.0 program issued by TeleAtlas to match input addresses to the spatial database using an offset of 25 feet from the street centerline. Addresses that were not successfully geocoded were checked for errors using interactive geocoding techniques. Where only a street intersection was available for the residential location (1.3% of residences), we assigned the geographic location of the residence to the middle of the intersection.

Industrial facility locations

Locations of industrial facilities were obtained from the Risk Screening Environmental Indicators (RSEI) model, version 2.1.2 (USEPA, 2008a). RSEI contains data from the EPA Toxics Release Inventory (TRI) (USEPA, 2008b), a program to provide the public with the information on releases of toxic chemicals in their communities. TRI reporting requirements began in 1987 and apply to the manufacturing sector (Division D of the Standardized Industrial Classification [SIC] codes: SICs 20-39) (OSHA, 2008), plus seven sectors that were not included in our study. Facilities are required to report if they have ten or more full-time employees and either 1) manufacture or process over 25,000 pounds of listed TRI chemicals; or 2) use more than 10,000 pounds of any listed chemical. Facilities report the amount of each type of chemical released or moved off-site and information about the facility including location (latitude and longitude). The RSEI database contains the set of latitude/longitude coordinates deemed to be ‘best’ – chosen through a series of tests and checks comparing the facility-reported coordinates, geocoded addresses, results of a major 1996 EPA quality assurance effort, and the EPA's Locational Reference Tables (LRT) (USEPA, 2004). RSEI also performs analyses in order to designate a single ‘primary’ SIC code for each facility (the company's main line of business), where it was not reported as such in the facility's TRI submission.

We included all manufacturing sector facilities (SIC 20-38; we excluded SIC 39 – miscellaneous manufacturing industries) reporting on-site chemical releases; off-site facilities which only received chemical transfers were excluded. Each facility was mapped in ArcGIS using the latitude-longitude coordinates provided in RSEI.

Exposure coding

We constructed metrics for residential proximity to industrial facilities anywhere in the US during a 10-year period prior to each participant's reference year (the diagnosis year for cases or the corresponding reference year for controls). We classified manufacturing facilities according to the primary 2-digit SIC codes (OSHA, 2008). We defined ‘proximity’ as distance within 2 miles, in order to allow comparison with results from previous studies (Johnson et al., 2003; Linos et al., 1991). Annual exposure status of each participant was determined by measuring the distance between the residence and the nearest facility within a given SIC; the participant was coded as unexposed for that SIC in that year if no facility was present within a 2-mile radius. If a participant lived in more than one residence during the year, the minimum distance to a given SIC was calculated as an average of the distances to the nearest facility from each residence, weighted by the proportion of the year spent at each residence.

Several proximity variables were considered in relation to NHL risk. An indicator variable was created for ever residing within 2 miles of each SIC during the 10-year study period. Indicator variables for ever residing within certain distance increments from each SIC were created, for >0 to 0.5-mile, >0.5-1.0 mile, >1.0-2.0 miles, and >2.0 miles (with distances chosen to allow comparison to the previous studies (Johnson et al., 2003; Linos et al., 1991); each person was counted only in the category of the shortest distance they had lived from the SIC. Variables were also created to indicate years of residence within 2 miles of each SIC (0 years, 1-9 years, and 10 years). We excluded SICs with zero cells for any of the exposure variables; namely, tobacco products (SIC 21), textile mill products (SIC 22), apparel and other finished products made from fabrics and similar materials (SIC 23), and leather and leather products (SIC 31), leaving 15 SICs for inclusion in our analyses.

Statistical analyses

Risk analyses were limited to participants for whom we had matched street address or nearest intersection within the US for >70% of their person years during the 10-year exposure period (e.g., >7 years; N=864 cases [65.4% of total], 684 controls [64.7%]). We estimated odds ratios (OR) and 95% confidence intervals (CI) for each exposure variable using unconditional logistic regression. All risk estimates were adjusted using indicator variables for the study design factors: age (<35, 35-44, 45-54, 55-64, 65+), gender, race (white, black, other/unknown race), and study site (Iowa State, Los Angeles County, Detroit, Seattle), in addition to education (in years: <12, 12-15, 16+). Because of concern about overmatching, we also conducted analyses without adjustment for study site. The results were essentially the same with- or without adjustment for study site: risk estimates differed by <10% and in both directions (i.e., estimates were not systematically biased towards the null with adjustment for study site). We therefore retained study site in our analyses based on the fact that we decided to include it a priori as a study design factor.

We explored additional potential confounding by evaluating the change in risk estimates for the proximity variables when including the following covariates in the model: 1) participant's occupational history (ever/never worked in the same industry as that being evaluated for proximity, denoted by 2-digit SIC); 2) other factors previously associated with risk of NHL or NHL subtypes in our study that could potentially be associated with location of residence, including family history of NHL (yes/no), body mass index (<20, 20 to <25, 25 to 30, >30), smoking status (never, former, current), alcohol consumption (in grams ethanol per week: <1, 1 to <15, 15 to <60, 60+), total vegetable intake (servings/week), and total fruit intake (servings/week); and 3) US 2000 census block group-level characteristics of the current residence, including percent Hispanics in the population, percent whites in the population, median household income in 1999, median years of education among men 45-64 years old, median years of education among women 45-64 years old, percent population living in single family housing units, and percent of persons living in urban areas.

We further examined associations in analyses stratified by study site, gender, and age (<50 and ≥50 years), to explore consistency of our findings. We also assessed the risk associated with proximity metrics for the NHL subtypes diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma in polytomous logistic regression models.

We explored the risk of NHL associated with residential proximity to industry longer than 10 years by conducting a reanalysis of the 10-year proximity variables limited to the participants who had lived in their current home for longer than 20 years (363 cases [27.5% of total], 291 controls [27.5% of total]), under the presumption that industrial facilities present during the period of TRI reporting had likely existed at the same site for some years before mandatory reporting began (1998 is the first year TRI data are available from RSEI), and long-term residents would therefore have likely been exposed prior to the 10-year exposure period.

We evaluated the possibility for selection bias by comparing participants to nonparticipants (persons who were eligible for the study but did not participate either because they could not be reached or they refused) in terms of the proximity of their current residences to industrial facilities, for those subjects with street address or matched intersection. Specifically, we compared the frequency of residence within 2 miles of an industrial facility (separately for any type of facility and for each 2-digit SIC), and tested whether differences in exposure by participant status were confounded by demographic characteristics of the individual or neighborhood. Furthermore, we tested whether the association between proximity of the current home to industrial facilities and NHL risk differed between the ‘complete’ study population (participants + nonparticipants combined) and participants only, by inclusion in logistic regression models (one for each 2-digit SIC) of a term for the multiplicative interaction between proximity and population (complete study population vs. participants only).

Results

Within the subset of participants included in our analysis of residential proximity to industrial facilities (864 cases and 684 controls), both cases and controls were more likely than those in the overall study to be older and white, and less likely to be from the Los Angeles County study site (Table 1). Of participants included in our analysis, 65.9% had lived in residences within 2 miles of industrial facilities during the 10-year exposure period. Residential proximity within 0.5-mile of any type of industrial facility was less frequent (13.5% of participants).

Table 1
Characteristics of cases and controls in NCI-SEER non-Hodgkin lymphoma study and in the substudy of residential proximity to industrial facilities (frequency [percentage], except where indicated)

We saw no overall association of residential proximity to industrial facilities of any type (all types combined) within a 10-year period and risk of NHL (OR=1.0), nor were there elevated ORs with proximity within 0.5-mile (vs. >2 miles, OR=1.1) or duration of 10 years (vs. 0 years, OR=1.0). Increased risk estimates were observed for a few types of industries; however, power was limited due to infrequent residential proximity to specific SICs (Table 2). Lumber and wood products, except furniture (SIC 24) was the only industry type for which NHL risk was associated with each of the three proximity metrics and with some indication of an exposure-response gradient, i.e., for having lived within 2 miles of a facility (OR=1.4, 95% CI: 0.9-2.1), within 0.5-mile of a facility (OR=2.2, 95% CI: 0.4-11.8; p-trend with decreasing distance = 0.15), or within 2 miles of a facility for 10 years (OR=1.9, 95% CI: 0.8-4.8; p-trend with increasing duration = 0.10). These associations were not confounded by occupational history of work in the lumber and wood products industry, as there was only one participant with exposure to SIC 24 in both their residential and occupational histories. Increased NHL risk associated with proximity within 2 miles to lumber facilities (SIC 24) was observed to some degree within each study site: Detroit (OR=7.4, 95% CI: 0.7-78); Iowa (OR=1.4, 95% CI: 0.6-3.6); Los Angeles County (1.3, 95% CI: 0.5-3.2); and Seattle (1.5, 95% CI: 0.8-2.8), and among men (OR=1.4, 95% CI: 0.8-2.7) and women (OR=1.4, 95% CI: 0.7-2.6). However, the association was observed only among older participants (age <55 years: OR=0.8, 95% CI: 0.4-1.6; age 55+: OR=2.2, 95% CI: 1.2-4.0).

Table 2
Associations of non-Hodgkin lymphoma with residential proximity to industrial facilitiesa during a 10-year period before diagnosis (or reference date for controls)

Several other types of industry were associated with NHL but with little consistency across proximity metrics and no indication of exposure-response gradients (Table 2). NHL risk was increased by 50% in association with 10 years proximity to chemical facilities (chemicals and allied products, SIC 28), and by 60% for residence between 0.5- to 1 mile. There was 90% increased NHL risk associated with 10 years of proximity to petroleum refineries (petroleum refining and related industries, SIC 29), 170% increased risk associated with residence within 0.5-mile of rubber facilities (rubber and miscellaneous plastics products, SIC 30), and 30% increased risk associated with proximity to primary metal industries (SIC 33). An inverse association was observed with furniture and fixtures facilities (SIC 25), as indicated by a trend of decreasing NHL risk with increasing number of years living in proximity of a facility. There was no evidence of important confounding of any of the associations by occupation, other risk factors from our case-control study, or census block group-level characteristics.

Analyses limited to participants who had lived in their current home for over 20 years (363 cases, 291 controls) generally produced diminished ORs compared to analyses based on 10 years of residence (results not shown). For example, among this subgroup there were no notable associations for residential proximity to lumber facilities (SIC 24) for having lived within 2 miles (OR=1.2, 95% CI=0.5-2.5), lived ≤0.5-mile (OR=0.8, 95% CI=0.04-14.1, p-trend with decreasing distance=0.09), or residence within 2 miles for 10 years (OR=1.5, 95% CI=0.5-4.2, p-trend with increasing duration=0.87).

There were no statistically significant differences between DLBCL and follicular lymphoma in the risk associated with residential proximity to industrial facilities (Table 3 shows results for selected SICs). However, any increased risk from residential proximity to lumber and wood product facilities was more apparent for DLBCL than follicular lymphoma, for having lived within 2 miles (DLBCL: OR=1.7; follicular: OR=1.1), lived ≤0.5 mile (DLBCL: OR=2.5; follicular: OR<0.001), the trend with decreasing distance (DLBCL: p=0.09; follicular: p=0.84), residence for 10 years within 2 miles (DLBCL: OR=2.5; follicular: OR=1.5), and the trend with increasing duration (DLBCL: p=0.05; follicular: p=0.70).

Table 3
Subtype-specific associations of non-Hodgkin lymphoma with residential proximity to industrial facilities during a 10-year period before diagnosis year (or reference year for controls)a

In our evaluation of potential selection bias, current residences of participants were less likely to be located within 2 miles of an industrial facility than were those of nonparticipants, for any type of manufacturing facility (52.1% of participants vs. 59.5% of nonparticipants, p<0.001) and for most of the specific 2-digit SIC codes we evaluated. The differences in proximity to industry by participation were fully accounted for by variation in the census block group-level demographic variables. The association between proximity of current residence to industry and NHL risk did not differ by participation. In logistic regression analyses including an interaction term between residential proximity and participation for their relation with NHL (with separate models for any SIC and for each 2-digit SIC), most of the interaction ORs were close to 1.0 and none was statistically significant. Nevertheless, there were non-significantly elevated interaction ORs for lumber and wood products (SIC 24) (OR=1.5, 95% CI: 0.6-3.9), primary metal industry (SIC 33) (OR=1.2, 95% CI: 0.9-1.7), and furniture and fixtures (SIC 25) (OR=0.6, 95% CI: 0.3-1.2). These ORs indicate that participants were more likely than nonparticipants to have overrepresentation of exposed cases and/or unexposed controls for SIC 24 and SIC 33 and thus selective participation may have biased the estimated risks to approximately the full extent of the magnitude of risks we observed for the lumber industry (OR=1.4) and primary metal industry (OR=1.3). Similarly, the inverse interaction OR for furniture and fixtures facilities (SIC 25) indicates that selection bias may have fully caused the decreased NHL risk we observed associated with SIC 25 (OR=0.6). However, as these comparisons between participants and nonparticipants included location of the current residence only, we could not completely evaluate the extent to which selection bias may have affected our results for the full 10-year exposure period.

We compared the results of our analysis of residential proximity to industrial facilities to findings from our previously published analysis of occupation in the same case-control study (Schenk et al., 2009). In our study of occupation, we found no statistically significant increases in NHL risk associated with having ever worked in any of the industries we evaluated in this study of residential proximity. Specifically, for industries with elevated risk estimates with regard to residential proximity, there was no increased risk associated with work in the lumber industry (SIC 24) (OR=0.8, 95% CI: 0.4-1.4), and a significantly decreased risk associated with work in the chemical industry (SIC 28) (OR=0.5, 95% CI: 0.3-0.9). There were non-significantly increased risks associated with work in the petroleum (SIC 29) (OR=2.1, 95% CI: 0.7-6.1), rubber (SIC 30) (OR=1.6, 95% CI: 0.8-3.3), and primary metal (SIC 33) (OR=1.3, 95% CI: 0.7-2.3) industries.

Discussion

We examined residential proximity to industrial facilities and risk of NHL using an objective measure of facility location. Although we observed increased risks of NHL associated with proximity metrics for certain types of manufacturing facilities including lumber (SIC 24), chemicals (SIC 28), petroleum (SIC 29), rubber (SIC 30), and primary metal (SIC 33), the associations were generally imprecise due to infrequent exposure, were inconsistent across different proximity metrics, and did not display notable exposure-response trends. There was also some indication that increased risks estimated for lumber and wood products facilities (SIC 24) and primary metal industries (SIC 33) may have been biased by selective participation in the study by case-control and exposure status. Furthermore, the elevated risk estimates we observed for residential proximity to lumber and chemical facilities were not supported by increased risks associated with having worked in those industries. Thus, although our findings are not entirely negative, we conclude that these data do not provide strong evidence of an association between living near manufacturing facilities and risk of NHL.

Our finding of no association between residential proximity to manufacturing facilities of any type (all types combined) and risk of NHL differs from previous reports which found 50-90% increased risk of follicular lymphoma (Johnson et al., 2003; Linos et al., 1991) and 60% increased risk of diffuse lymphoma (Linos et al., 1991) associated with residence within 2 miles of a facility. While it is possible that the observed associations of Linos et al. (Linos et al., 1991) were biased by selective recall of proximity to industry, the Johnson et al. (Johnson et al., 2003) study used an external database on industry locations linked to participants' residences – similar to our approach – and recall bias would not be a likely explanation. However, our approach differed from both of the previous studies in the exposure periods evaluated. The Canadian Environmental Quality Database (Johnson et al., 2003) contained data spanning over 30 years and the Linos study inquired about residential proximity over the subject's lifetime, in contrast to the 10-year exposure period that was evaluated in our study using the TRI/RSEI data. Long-term exposure or exposure in the distant past may be more relevant than recent exposure to NHL etiology; however, the relevant timing of exposure is likely to differ by chemical. For example, there is some evidence that polychlorinated biphenyls, which are suspected lymphomagens, operate as etiologic agents primarily in the decade before diagnosis (Engel et al., 2007).

A notable strength of our study is that we were able to adjust for many potential confounders, including education, race, body mass index, smoking, occupational history, and neighborhood characteristics. Nevertheless, there was limited power in our analysis to detect with precision most risks associated with certain types of facilities, because residential proximity within 2 miles was uncommon, and was even less frequent within 0.5 mile. For example, 76 (11%) controls lived within 2 miles of petroleum facilities (SIC 29), and only 8 (1%) lived within 0.5 mile. In addition, our study is dependent on the locations of current and historic residences of study participants, which were measured by GPS or geocoding. Because increased risks associated with living near industry are likely to be more prominent at distances close to the facility, even small errors in the accuracy of location may have affected our results. We would expect any errors to be nondifferential by case or control status, likely biasing results toward the null. Nevertheless, our approach based on accurate geolocations to the street segment/intersection level (Ward et al., 2005) would generally be subject to less non-differential misclassification than the approach in previous studies based on the centroid of the postal code (Johnson et al., 2003) or the municipality (Ramis et al. 2009).

Our study relied on proximity to industrial facilities as a proxy for exposure to chemicals. Implicit within this simple metric is the assumption that pollution spreads uniformly away from its source. This is clearly not always the case. The fate and transport of released chemicals depend on the media to which the chemical is emitted (e.g., air, surface water), stack height and wind direction for air releases, drainage patterns for water, and other factors – in addition to the properties of the chemical and the amount released. Another limitation of proximity metrics is that they do not take into account specific types of chemicals released from facilities. Assessment of the types of chemicals released – which vary between and within different types of industry – would provide a more specific measure of exposure that could potentially identify a causal agent for NHL. For example, lumber and wood products facilities (SIC 24) may or may not release pentachlorophenol which has been used widely for wood preservation since the 1950s and contains impurities such as dioxins and furans which have themselves been associated with risk of NHL (De Roos et al., 2005). SIC 24 facilities likely also vary considerably in other agents historically used or released, such as dusts, coal tar creosote, polycyclic aromatic hydrocarbons, benzene, and copper chromated arsenic (IARC, 1981). Our proximity metric simply provides an indicator of the mix of exposures across SIC 24 facilities in the US, in addition to other factors associated with proximity. We pursued this approach in order to more strategically focus our future research on specific exposures within industries. However, more specific and detailed measures of exposure may be needed to detect dose-response patterns of NHL risk associated with living near industrial facilities.

In conclusion, while our study does not provide strong evidence that living near manufacturing facilities increases NHL risk, we cannot rule out potential risks associated with specific types of industry. Our findings for the lumber industry were strongest for diffuse type lymphomas and we observed increased risks with both distance and duration proximity metrics. Associations we observed with certain types of industry – including lumber (SIC 24), chemicals (SIC 28), petroleum (SIC 29), rubber (SIC 30), and primary metal (SIC 33) – may be truly causal or may be due to unmeasured bias or chance. Future studies on this topic that are designed to capture a larger population living in proximity to a specific type of industry could improve the statistical power to detect risk gradients by distance. Additionally, studies that account for chemical transport in the environment and employ biomarkers of exposure and biologic effects in residents surrounding a particular type of facility may be useful in establishing dose-response relations that could lend credibility to a causal association.

Acknowledgments

We gratefully acknowledge Irish Lonn (Information Management Services, Inc., Silver Spring, MD) for assistance in data processing and preparation, Adrienne Katner (Louisiana Department of Health and Hospitals, Metairie, LA) for her assistance in obtaining the RSEI database, and Richard Engler of the EPA for guidance in extracting the relevant RSEI data for our analysis. We also thank Laura Gold, Robert Mathes, Hozefa Divan, and Jim Giglierano and his staff at the Iowa Geologic Survey for their efforts in ground-checking residential locations.

Funding sources: The work of the lead author was supported by an R03 grant from the National Cancer Institute (R03 CA115183). The parent case-control study was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics of the National Cancer Institute, National Institutes of Health, Department of Health and Human Services, through contracts with four centers of the Surveillance, Epidemiology, and End Results (SEER) program in Iowa state (contract #N01-CN-67008), Los Angeles County (#N01-CN-67010), and the metropolitan areas of Seattle (#N01-PC-67009) and Detroit (#N01-PC-65064). Support for Dr. Nuckols was provided through an IPA agreement between NCI-DCEG and Colorado State University. The research activities in the study did not begin until after informed consent was obtained from the subject. All study procedures were approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center (FHCRC Institutional Review file #6330).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • Alexander JW. Industrial expansion in the United States, 1939-1947. Economic Geography. 1952;28:128–142.
  • Chatterjee N, Hartge P, et al. Risk of non-Hodgkin's lymphoma and family history of lymphatic, hematologic, and other cancers. Cancer Epidemiol Biomarkers Prev. 2004;13:1415–1421. [PubMed]
  • Clarke CA, Glaser SL. Changing incidence of non-Hodgkin lymphomas in the United States. Cancer. 2002;94:2015–2023. [PubMed]
  • De Roos AJ, Hartge P, et al. Organochlorines in plasma and risk of non-Hodgkin lymphoma. Cancer Research 2005
  • Department of Labor, Occupatonal Safety and Health Administration (OSHA) SIC Division Structure. 2008. [April 19, 2006]. < http://www.osha.gov/pls/imis/sic_manual.html>.
  • Engel LS, Laden F, et al. Polychlorinated biphenyl levels in peripheral blood and non-Hodgkin's lymphoma: a report from three cohorts. Cancer Res. 2007;67:5545–5552. [PubMed]
  • Hartge P, Wang SS, Bracci PM, Devesa SS, Holly EA. Non-Hodgkin Lymphoma. In: Schottenfeld D, Fraumeni JF Jr, editors. Cancer Epidemiology and Prevention. New York: Oxford University Press Inc.; 2006. pp. 898–918.
  • International Agency for Research on Cancer (IARC) The lumber and sawmill industries (including logging) IARC Monogr Eval Carcinog Risk Chem Hum. 1981;25:49–197. [PubMed]
  • Johnson KC, Pan S, et al. Residential proximity to industrial plants and non-Hodgkin lymphoma. Epidemiology. 2003;14:687–693. [PubMed]
  • Linos A, Blair A, et al. Leukemia and non-Hodgkin's lymphoma and residential proximity to industrial plants. Arch Environ Health. 1991;46:70–74. [PubMed]
  • Office of Science and Technology Policy. Statement by the president regarding the United States' decision to stop degrading global positioning system accuracy. 2000. [July 30, 2004]. < http://www.igeb.gov/sa/>.
  • Ramis R, Vidal E, et al. Study of non-Hodgkin's lymphoma mortality associated with industrial pollution in Spain, using Poisson models. BMC Public Health. 2009;9:26. [PMC free article] [PubMed]
  • Schenk M, Purdue MP, et al. Occupation/industry and risk of non-Hodgkin's lymphoma in the United States. Occup Environ Med. 2009;66:23–31. [PMC free article] [PubMed]
  • Ward MH, Nuckols JR, et al. Positional accuracy of two methods of geocoding. Epidemiology. 2005;16:542–547. [PubMed]
  • U.S. Environmental Protection Agency (USEPA) RSEI Technical Appendix D. Locational Data for TRI Reporting Facilities and Off-site Facilities. 2004. [May 5, 2008]. < http://www.epa.gov/oppt/rsei/pubs/tech_app_d_v212.pdf>.
  • U.S. Environmental Protection Agency (USEPA) Risk-Screening Environmental Indicators (RSEI) 2008a. [December 1, 2007]. < http://www.epa.gov/oppt/rsei/>.
  • U.S. Environmental Protection Agency (USEPA) Toxics Release Inventory (TRI) Program. 2008b. [December 1, 2007]. http://www.epa.gov/tri/