Published ahead of print on June 23, 2005, doi:10.1164/rccm.200412-1731OC
© 2005 American Thoracic Society doi: 10.1164/rccm.200412-1731OC Residential Proximity to Naturally Occurring Asbestos and Mesothelioma Risk in CaliforniaDepartment of Public Health Sciences, Department of Geology, and Department of Statistics, University of California at Davis, Davis, California Correspondence and requests for reprints should be addressed to Marc B. Schenker, M.D., M.P.H., Division of Environmental & Occupational Health, Department of Public Health Sciences, University of California, Davis, One Shields Avenue, TB168 Davis, CA 95616-8638. E-mail: mbschenker{at}ucdavis.edu
Rationale: Little is known about environmental exposure to low levels of naturally occurring asbestos (NOA) and malignant mesothelioma (MM) risk. Objectives: To conduct a cancer registry-based case control study of residential proximity to NOA with MM in California. Methods: Incident MM cases (n = 2,908) aged 35 yr or more, diagnosed between 1988 and 1997, were selected from the California Cancer Registry and frequency matched to control subjects with pancreatic cancer (n = 2,908) by 5-yr age group and sex. Control subjects were selected by stratified random sampling from 28,123 incident pancreatic cancers in the same time period. We located 93.7% of subjects at the house or street level at initial diagnosis. Individual occupational exposure to asbestos was derived from the longest held occupation, available for 74% of MM cases and 63% of pancreatic cancers. Occupational exposure to asbestos was determined by a priori classification and confirmed by association with mesothelioma. Main Results: The adjusted odds ratios and 95% confidence interval for low, medium, and high probabilities of occupational exposures to asbestos were 1.71 (1.32–2.21), 2.51 (1.91–3.30), and 14.94 (8.37–26.67), respectively. Logistic regression analysis from a subset of 1,133 mesothelioma cases and 890 control subjects with pancreatic cancer showed that the odds of mesothelioma decreased approximately 6.3% for every 10 km farther from the nearest asbestos source, an odds ratio of 0.937 (95% confidence interval = 0.895–0.982), adjusted for age, sex, and occupational exposure to asbestos. Conclusions: These data support the hypothesis that residential proximity to NOA is significantly associated with increased risk of MM in California.
Key Words: Cancer Registry case-control study • GIS • malignant mesothelioma • naturally occurring asbestos • occupational exposure to asbestos Epidemiologic studies have confirmed that occupational exposure to asbestos causes mesothelioma (1–7). However, almost all population-based studies have found that many mesothelioma cases had no known occupational exposure to asbestos. Some of these cases could be due to domestic and neighborhood exposure to asbestos, or even environmental exposure to naturally occurring asbestos (NOA). An association with domestic or neighborhood exposure to asbestos or other mineral fibers and an increased risk of mesothelioma has been found in several locations including Australia (8), South Africa (9), Italy (10), and New Jersey (11). Studies among residents of two Anatolian villages in Turkey showed that the astounding high incidence of malignant mesothelioma results from environmental exposure to carcinogenic tremolite and erionite, a fibrous zeolite, which is present in the volcanic tuffs that are used as building stone (12). An association between environmental exposure to NOA and mesothelioma has been observed in Cyprus (13), Greece (14), New Caledonia (15), Corsica (16), China (17), and Italy (18), but no research has studied the association of residential distance from environmental (nonoccupational) asbestos and mesothelioma risk (19). It remains unknown whether environmental exposure to low levels of NOA also causes mesothelioma. Research has failed to find a threshold level below which there is no mesothelioma risk, and low-dose exposure to asbestos can cause mesothelioma (20). Most epidemiologic studies for mesothelioma have failed to find evidence of occupational, domestic, or neighborhood exposure to asbestos in a sizable proportion of mesothelioma cases (7). For example, 25% of mesothelioma cases in a Spanish case–control study had no significant occupational, domestic, or neighborhood exposure to asbestos (3). It is plausible that exposure to NOA may account for a portion of those mesothelioma cases. California has large amounts of serpentinite and other ultramafic rocks that are the predominant but not exclusive source of NOA and are distributed mostly in the Sierra Nevada, Coast Ranges, and Klamath Mountains in Northern and Central California (Figure 1). The most common type of asbestos is chrysotile, but other types including tremolite are also found in California. (21). There have been increasing concerns about whether an increased risk of mesothelioma is associated with environmental exposures to NOA in these areas. Some of these areas include new housing developments and potential exposure to a large number of people.
Statewide cancer reporting in California has been mandated by law since 1985, and beginning in January 1988, the California Cancer Registry (CCR) has covered the entire population in California through nine regional population-based cancer registries. The CCR is one of the largest comprehensive population-based statewide cancer registries in the world, covering more than one tenth of the U.S. population. About 300 incident cases of malignant mesothelioma per year are reported in California. It is estimated that more than 98.9% of all mesothelioma cases diagnosed in California are reported to the California Cancer Registry (22). This study used a unique opportunity to evaluate the association between exposure to potential sources of indigenous environmental asbestos and incidence of mesothelioma. California has more serpentinite and ultramafic rocks than other states in the United States, and its distribution is patchy, with exposed areas separated from unexposed areas (23). Additionally, California's large population provided us with an adequate sample size to study this rare cancer.
Case Ascertainment and Control Selection Incident cases of malignant mesothelioma diagnosed from 1988 through 1997 in California were identified from the CCR. The International Classification of Diseases for Oncology, Second edition (ICD-O-2, 1990), was used by the CCR for the period covered by the study (22). The case series consisted of all reported patients ages 35 yr or older diagnosed with malignant mesothelioma (histology ICD-O-2 codes 9050–9053) from 1988 through 1997 in California. For each case, one control frequency-matched by 5-yr age group and sex was selected by stratified random sampling from patients diagnosed with incident malignant pancreatic cancer (primary site ICD-O-2 codes C250–259) reported to the CCR in the same period.
Assessment of Residential Proximity to NOA
Assessment of Occupational Exposure to Asbestos
Statistical Analysis
A total of 2,966 incident cases diagnosed with malignant mesothelioma were reported to the California Cancer Registry in the period between 1988 and 1997. There were 17 duplicate records deleted and 41 cases younger than 34 years were removed. The study included 2,908 malignant mesothelioma cases (2,354 men and 554 women). We selected 2,908 control subjects frequency matched by 5-yr age group and sex by stratified random sampling from the 28,123 incident cases of malignant pancreatic cancer reported to the CCR in the same period. Among 5,816 subjects, mesothelioma cases and control subjects with pancreatic cancer in our study were balanced in vital status (93% versus 97%), percentages of data abstracts from inpatient or outpatient records (96% versus 92%), race/ethnicity, and marital status by sex (Table 1). The percentage of histologic diagnosis of mesothelioma cases was significantly higher than for control subjects with pancreatic cancer (82% versus 56%). Data on the longest-held occupation and industry were available for 74% of mesothelioma cases and 63% of control subjects with pancreatic cancer, respectively. All cases and control subjects were located based on their residential addresses at initial diagnosis. Among these, 93% of geocodes were completed at a house or street level (the most precise), 6% at a geographic centroid of a five-digit zip code, and only 0.3% at a geographic centroid of a city or county (the least precise). Geographic coding rates were comparable in cases and control subjects (Table 1).
Logistic regression analyses were restricted to the subset of 1,133 mesothelioma cases and 890 control subjects with pancreatic cancer who had complete data on residence, age, sex, and occupation. Univariate analyses in this data set showed nearly identical age distributions and proportion of males. The cases had substantially higher frequency of occupational exposures to asbestos (Table 2).
Age-adjusted logistic regression analysis for occupational exposure to asbestos showed that occupations with a high probability of occupational exposure to asbestos in men such as boilermaker, insulator, plumber, pipefitter, steamfitter, sheet metal worker, electrician, and painter were strongly associated with mesothelioma risk, whereas few female cases and control subjects had occupations with a high probability of occupational asbestos exposure. This supported the hypothesis that the difference of mesothelioma incidences between men and women mainly resulted from different occupational exposures to asbestos. Industries with high probability of occupational exposure to asbestos in men such as shipyard and Navy were significantly associated with the elevated mesothelioma risk (Table 2). In the primary analysis of the association of NOA and mesothelioma, a multivariate logistic regression model was fitted, assuming a constant percent change in the odds of mesothelioma with a 10-km increase in nearest distance, adjusted for the effects of age, sex, and occupational exposure (Table 3). The estimated effect was a 6.3% decrease (computed from the odds ratio as [(1 – 0.937) x 100%] in the odds of a mesothelioma case for every 10 km farther away from a NOA source (95% CI, 1.8–10.5%; p = 0.006). Occupational exposure, as expected, was the strongest predictor of an increased odds of mesothelioma, with a significant dose–response relationship. Age and sex as major risk factors of mesothelioma and matching factors in this study were included in the final multivariate model to adjust for the impact of imbalance in age and sex from exclusion of cases and controls with unknown occupational exposure and more than 100-km nearest distances.
An analysis excluding the cases with a moderate or high probability of occupational asbestos exposure gave very similar results. Among all subjects the odds ratio for mesothelioma for a 10-km increase in nearest distance was 0.943 (95% CI, 0.896–0.922). The observed odds ratios for men and women were 0.939 (95% CI, 0.887–0.994) and 0.961 (95% CI, 0.861–1.073), respectively. Figure 2 illustrates the shift in the log odds of mesothelioma versus pancreatic cancer as the distance from NOA sources increases among men and women. People who lived closer to an asbestos source had a greater chance of having mesothelioma, and the chance decreased steadily as the distance increased. In stratified analyses of men and women, we saw comparable findings, although the results in women were based on small samples and were not statistically significant (results not shown). Fractional polynomial analysis suggested that the nearest distance was reasonably included in the multivariate logistic regression model as a linear continuous independent variable for residential proximity to sources of NOA, which is the assumption for including the distance as a continuous variable in the logistic regression model. Logistic regression diagnostics supported the use of this model, with no evidence for model failure or interactions (results not shown).
We obtained similar results when we repeated the analysis using only subjects with a nearest distance of 50 km or less. When we repeated the analysis using either subjects with 150 km or less or all subjects, using the log-transformed nearest distance rather than the nearest distance in the multivariate logistic regression models, we also found that there was an inverse relationship between the log-transformed nearest distance and risk of mesothelioma (data not shown). Pleural mesotheliomas accounted for 91% of the total cases. When only pleural mesothelioma cases were used, proximity to NOA was significantly and positively associated with case status for both men and women. On the contrary, when only peritoneal mesothelioma cases were used, proximity to NOA was not significantly associated with case status and did not show a trend with distance.
Results of this cancer registry–based case control study for mesothelioma in California demonstrated that residential proximity to geologic sources of NOA in California was independently and positively associated with case status. The odds of being a mesothelioma case are 6.3% lower than the odds of being a control subject for every 10 km farther from the nearest asbestos source. These findings were adjusted for occupational exposure to asbestos, age at initial diagnosis, and sex. The association was observed in both men and women, although the association was statistically significant in men only. Our study confirmed that some occupations such as shipyard worker, boilermaker, insulator, plumber, pipefitter, and steamfitter, and industries such as shipping, construction, and Navy had higher occupational exposure to asbestos and were strongly associated with an increased risk of malignant mesothelioma, consistent with results from previous studies of mesothelioma (1–6, 30). Although it is likely that occupational exposure misclassification exists, our study did see the expected dose–response relationship between occupational exposure to asbestos and mesothelioma. The method defined by asbestosis PMR for occupations based on the longest held occupation and industry available in CCR data appeared reasonable to classify occupational exposure to asbestos. A major strength of this study was the very large number of incident mesothelioma cases used to assess the potentially weak association between exposure to NOA and mesothelioma incidence. The large size of this study provided us with a unique opportunity to detect relatively low-dose exposures to NOA in the presence of significant occupational exposure to asbestos. Another strength of this study was that GIS approaches were used to characterize potential individual exposure to NOA in a very efficient way. Geocoding of residential addresses provided more accurate assessment of potential environmental exposure to NOA at the individual level than methods using less precise geographic classification. Selection bias should be limited in our study because the data on mesothelioma and pancreatic cancer were both obtained by the CCR in a comparable way. The data included all registered cases, and the control patients were selected using stratified random sampling from patients with malignant pancreatic cancer in the same time period. A limitation of the registry based study was lack of accurate and complete lifetime occupational histories on individual cases and controls. Incomplete occupational data and lack of duration and intensity of occupational exposure to asbestos in the past made it difficult for us to evaluate occupational exposure to asbestos using common occupational exposure assessments such as job-exposure matrices and expert assessment for exposures. Although the CCR was supposed to record the patient's longest-held occupation and industry, it often listed only current or most recent occupation. Occupational exposure is the predominant cause of asbestos-induced mesothelioma. We should be cautious to explain our results because our assessment for occupational exposure to asbestos was based on incomplete occupational history on individual cases and control subjects. Another limitation of the study was the lack of a reliable and acceptable job-exposure matrix for asbestos to characterize the potential of individual occupational exposure to asbestos based on occupation or job titles. Therefore we had to derive occupational exposure to asbestos based on incomplete information on occupations and industries. Misclassification for occupational exposure was unavoidable in this study. Although the National Institute for Occupational Safety and Health has reported asbestosis PMR for selected occupations and industries to reflect strength of occupational exposure to asbestos, it included only limited occupations and industries with higher occupational exposure to asbestos and more asbestosis cases (26). The job-exposure matrix for asbestos developed by Cocco and Dosemeci in 1998 lacked validation of occupational exposure to asbestos (27), and too many occupations were defined as exposed in a study of mesothelioma in Australia (31). The study was also limited by the lack of complete lifetime residential history on individual cases and controls. When we calculated the distance to the nearest NOA from subject residential address at initial diagnosis, we had to assume that subject's residential address at initial diagnosis was the same as the longest residential address in his or her life. Exposure misclassification in our study could either decrease or increase the observed dose–response relationship. Incomplete residential data and lack of residential duration in one's lifetime made it difficult to accurately assess environmental exposure to naturally occurring and other asbestos sources. Reliable assessment of exposure to NOA using GIS approaches ideally depends on accurate and complete residential histories. A field study is needed to examine the relationship between actual asbestos exposure potential and the distance to the nearest asbestos sources. In additionally, consideration should be given to the impact of age of potential exposure, deposit size, residence duration, nature of human disturbance activity, meteorologic conditions, and other geologic/demographic attributes on the actual or potential exposure to NOA. Our study is based on ultramafic rock-specific GIS map, but not asbestos-specific GIS map for distribution of asbestos sources in California. A new digital map with more accurate and specific information about asbestos sources in California needs to be developed for further studies. It is essential for public health efforts to assess more accurately potential exposure to NOA. Some efforts should be increased to characterize the location of asbestos deposits using imaging spectroscopy to map ultramafic rocks, serpentinites, and tremolite-actinolite–bearing rocks in California (32). Frequency matching was used to minimize confounding from age and sex. We were unable to control for potential confounding from domestic or residential exposure to asbestos because these data were not available for this study. Selection bias was limited because the data included all eligible cases from the population-based CCR and the control subjects were selected by stratified random sampling from the same population-based cancer registry. The data were collected in a similar manner for cases and control subjects; therefore, information bias from the data collection was limited, although some information bias cannot be avoided. For example, data on the longest occupation and industry were available for 74% of mesothelioma cases and 63% of pancreatic cancer cases in the CCR records. Because patients diagnosed with mesothelioma may be more likely to have been asked about occupational history and asbestos exposure, this may have led to overestimating occupational exposure to asbestos in this study. We have found that residential proximity to NOA shows an independent and dose–response association with mesothelioma risk. The findings are biologically plausible in view of the known strong association of occupational asbestos exposure and mesothelioma, and the observation of an association of NOA and mesothelioma in other areas of the world (33). A prospective case–control study based on asbestos-specific GIS map in California should be done with incident mesothelioma cases. This would facilitate obtaining accurate and complete lifetime residential and occupational histories. Further research also is needed to develop a model to evaluate relationship between actual environmental exposure to NOA and major influencing factors, and to develop public health strategies to reduce exposure to asbestos from environmental sources.
Supported by the National Cancer Institute (Project Grant IR03CA81615-01). Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form December 22, 2004; accepted in final form June 19, 2005
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
