LIBBY ASBESTOS SITE
LIBBY, LINCOLN COUNTY, MONTANA
Mortality from Asbestosis in Libby, Montana,
The Agency for Toxic Substances and Disease Registry, in cooperation with the Montana Department of Public Health and Human Services, analyzed mortality statistics for Libby, Montana during the 20-year period from 1979 to 1998.
This review of death certificate data was performed to generate an accurate representation of mortality potentially associated with historical asbestos exposure in the Libby community. Underlying causes of death reviewed included nonmalignant respiratory diseases, lung cancer, mesothelioma, digestive cancer, and diseases of pulmonary circulation. Mortality from these underlying causes was compared with mortality statistics for the state of Montana and the U.S. population.
For the 20-year period reviewed in this report, mortality in Libby resulting from asbestosis was approximately 40 to 60 times higher than expected. Mesothelioma mortality was also elevated. However, it was difficult to precisely evaluate the degree to which mesothelioma mortality was elevated because statistics on this extremely rare cancer are not routinely published at the state and national levels. Other underlying causes of mortality including lung cancer, digestive cancer and diseases of pulmonary circulation were not significantly elevated over this period. There were no significant increases in trends for mortality due to all of the underlying causes analyzed.
Because death certificates do not contain many relevant pieces of information including data on exposure, smoking habits, and residential history, the findings of this review require careful interpretation.
In November 1999, the media reported an increase in deaths and illnesses from exposure to asbestos contaminated vermiculite in Libby, Montana . Following these reports, Senator Max Baucus (Montana) requested that the U.S. Department of Health and Human Services (DHHS) provide assistance in addressing health concerns about asbestos exposure among residents in this community. The Agency for Toxic Substances and Disease Registry (ATSDR) was designated as the lead DHHS agency to respond to these concerns. As part of ATSDR's response, a review of mortality data for the Libby community was initiated.
The purpose of this health consultation is to assess mortality in Libby by reviewing death certificate data. This analysis was performed to generate an accurate representation of mortality potentially associated with asbestos exposure in the Libby community. ATSDR staff calculated mortality rates and standardized mortality ratios (SMRs) for underlying causes of death associated with asbestos exposure. The underlying causes of death include nonmalignant respiratory diseases, lung cancer, mesothelioma, and digestive cancer.
Libby, the county seat of Lincoln County, is located in a valley along the Kootenai River in northwestern Montana. The Libby valley runs east and west with lesser valleys extending north and south along Pipe Creek and Libby Creek, respectively. The Libby valley has a vertical relief as high as 4,000 ft. in some areas and is subject to severe temperature inversions during many times of the year. These physical characteristics lead to the trapping of air pollutants and give Libby the distinction of having possibly the worst ventilation of any community in Montana . The population in the valley ranges between 7,000 and 8,000 individuals while the town of Libby has between 2,500 and 3,000 persons. The population of Libby has remained fairly stable over the last 30 years.
Strip mining, transportation, and processing of vermiculite ore containing asbestiform minerals was conducted in the area from 1923 until 1990 (Figure 1). A 1982 analysis of raw vermiculite ore from Libby revealed asbestos contamination in the range of 21% to 26% . The vermiculite deposit in Libby is commingled with geologic formations of tremolite , a naturally occurring mineral falling within a class of hydrated silicate minerals known as amphiboles . Fibrous forms of tremolite, along with other silicate-based amphiboles, are commonly referred to as asbestos. While tremolite is not mined for commercial purposes, it is a common contaminant in other commercially mined minerals including chrysotile, vermiculite, and talc [5-7].
Vermiculite is a naturally occurring mineral and the largest known deposit in the world is in Libby . It has been estimated that before the closing of the mine in 1990, Libby contributed up to 80% of the world's supply of vermiculite . It is a unique mineral with the ability to exfoliate, or expand, upon heating. Exfoliated vermiculite has many commercial uses such as inclusion in concrete aggregates, loose-fill insulation, horticultural applications like soil conditioning, and as a bulk carrier for agricultural chemicals.
Vermiculite production at the Libby mine was initially low. Production increased from an estimated 20,000 tons in 1940 to 150,000 tons in 1950 and 200,000 tons in 1970 . Before 1954, the vermiculite ore was processed through dry milling. In 1955 a second mill using a wet screening process was brought on-line to assist with the increasing volume of material being handled. In 1974, both mills were shut down for "hygienic and technical reasons" and a new mill was built nearby that operated exclusively with a wet milling process . Drilling at the mine was another activity that led to heavy dust production. In 1970, drills with baghouses were introduced to decrease dust problems . In 1990, the vermiculite mine and associated processing facilities in Libby closed.
The association between asbestos exposure and adverse health effects is well documented. In 1924, Cooke first described nonmalignant health effects associated with occupational asbestos exposure . Later descriptions of the resultant pulmonary fibrosis led to a new term for this disease, "asbestosis" . A decade later, case reports of occupational asbestos exposure and lung cancer began to appear in the public health literature . Selikoff et al., in their seminal papers on asbestos related mortality in insulation workers from 1943 to1976, found marked increases in deaths due to lung cancer, mesothelioma, and asbestosis in this occupational cohort. They also reported weak associations with digestive cancer [14,15].
In 1978, a cluster of 12 cases of pleural effusion was identified in employees at a facility in Ohio using vermiculite mined and processed in Libby . These cases led two investigators to study the occupational cohorts at the mine and processing facilities in Libby during the 1980s [5,10,16-18]. Both studies revealed significant increases in mortality due to nonmalignant respiratory disease and lung cancer [10,16].
Asbestos-related lung diseases have been observed in Libby residents [1,10,16]. The potential exposure scenarios in this community include individuals with occupational exposure, family members of workers exposed through worker take-home, community members exposed through ambient environmental levels, and residents exposed through the use of vermiculite as insulation in their home or as a soil amendment in their garden. Workers were occupationally exposed to high concentrations of asbestos fibers released during the mining and milling of vermiculite [5,8]. Household members of workers were potentially exposed to asbestos fibers adhering to hair, shoes, and work clothes brought home. Exposure and resulting health effects from worker take-home have been thoroughly documented in other cohorts . Community members could have been exposed to ambient environmental levels as a result of exfoliation and packaging of contaminated vermiculite in Libby. Air sampling in downtown Libby in 1975 as well as in the 1980s detected levels well above the Occupational Safety and Health Administration's occupational limit of 0.1 fibers per cubic centimeter . Also, there were anecdotal reports of children playing in piles of vermiculite at the exfoliation facility as well as reports of excessive dust emanating from these piles that interrupted activities in the downtown area. There have been other occurrences of community exposure similar to the scenario in Libby, including documented adverse health effects in Turkey from environmental exposure to mineral deposits containing fibrous forms of tremolite [20, 21].
This review was designed to use death certificate data to assess mortality potentially associated with asbestos exposure in Libby. No additional information on decedents was used. The Montana Department of Public Health and Human Services (MDPHHS) provided most of the mortality data referenced in this report.
Underlying causes of mortality are coded using the International Classification of Diseases (ICD) system. The review of mortality in Libby included all underlying causes of death associated with exposure to asbestos (Appendix 1). ICD codes used in this review included malignant neoplasms of digestive organs and peritoneum (150-159); malignant neoplasms of respiratory and intrathoracic organs excluding nasal cavities and accessory sinuses (161-165); malignancy without specification to site (199); diseases of pulmonary circulation (415-417); and nonmalignant respiratory diseases, excluding upper respiratory tract infections (480-519). Additionally, some referent underlying causes were included in the analysis for their lack of association with asbestos exposure. These were used to check for temporal changes in mortality reporting and recording. The referent underlying causes of mortality included malignant neoplasms of female breast (174), malignant neoplasms of prostate (185), and cerebrovascular diseases (430-438).
MDPHHS, Vital Statistics Section, generated an electronic file of mortality for 1979-1998 using state death certificate data (Appendix 2). This information included all individuals with selected underlying causes of mortality who claimed Libby as their residence at time of death. MDPHHS does not document some of the data on death certificates in its electronic database of decedents, such as residential street address at time of death, occupation, and business/industry. Therefore, paper copies of death certificates were retrieved to gather this information, which was then added to the database. MDPHHS provided copies of death certificates for all deaths that occurred in Montana. For the remaining deaths that occurred outside of Montana, certificates were sought from the respective states (Idaho, Kansas, Nebraska, New York, North Dakota, Oregon, Tennessee, Washington, and Wyoming).
The mortality review encompassed a 20-year period from January 1, 1979 to December 31, 1998. This period was selected for two reasons. First, the highest exposures to asbestos in Libby are thought to have occurred from the 1950s through the early 1970s. This is based on reports indicating that the peak production volumes occurred during this period and engineering controls were being implemented during or after this time. If the highest exposures occurred then, the 20-year period under analysis provided time for cancer and other nonmalignant diseases to progress through their clinical course and lead to death. The second reason for this time frame deals with the classification and coding of mortality data. Underlying causes of death for these years were coded using ICD-9. Before 1979 and after 1998, different ICD coding schemes were used, making direct comparisons of disease outcomes difficult.
Various geographic boundaries were defined prior to analyzing death certificate data. These areas were used to assess the mortality experience within different sized regions. Using small geographic areas of analysis eliminated populations in Lincoln County not impacted by asbestos contamination. Had these groups been included in the analysis, unexposed individuals may have diluted any potential associations at the local level. The use of multiple geographic boundaries was also helpful in determining the stability of crude mortality rates and SMRs. By relying on different study areas, it was possible to evaluate the effects of imposing artificial study boundaries on mortality calculations.
Six geographic boundaries were used for the analyses (Figure 2). These regions varied widely in size and were derived from different sources. They are described below beginning with the smallest, most restrictive area, and proceeding to the largest.
1. Libby city limits - This 1.1-square mile area encompasses downtown Libby. The boundary file describing this area was obtained from the U.S. Census Bureau's TIGER/Line files.
2. Extended Libby boundary - This 2.2-square mile area includes downtown Libby as well as most of the major subdivisions surrounding the downtown area. This boundary was created through the use of aerial photography to digitize a border around the most developed areas adjacent to, and including, downtown.
3. Air modeling - This 16-square mile area describes the region where computer modeling has indicated the highest probability for asbestos fiber distribution associated with exfoliation, mining, and milling processes. The modeling that generated this boundary integrated local meteorological data and contaminant dispersion information into a predictive, grid-based output. Environmental sampling data were not used.
4. Medical screening - ATSDR's Division of Health Studies (DHS) established this 25-square mile area for the purpose of prompting individuals who were most likely impacted by contamination to enter the medical screening program. This area includes a concentric area around downtown Libby (within 2½ miles of the center of town) and an area along the Kootenai River (½ mile on either side) to approximately 7 miles upstream towards the Libby dam.
5. Libby valley - This 65-square mile area includes the extended Libby community living within the valley up to an elevation of 800 ft. It was derived from a digital elevation model for Lincoln County obtained from the U.S. Geologic Survey.
6. Central Lincoln County - This 314-square mile catchment area describes a concentric region with a 10-mile radius. It is centered in downtown Libby at the intersection of California Ave. and U.S. Highway 2. This is the same point used to center the area initially defined for medical screening recruitment by DHS.
Many residents in the region identified Libby as their town of residence even though they lived well outside city limits. This is because Libby is the county seat as well as the largest city in a sparsely populated area. Death certificate information reflected this fact, as street addresses located many miles outside of town designated Libby as the city of residence. After an initial inspection of death certificate data, it was decided that a more focused method was required for analyzing mortality in Libby.
Georeferencing of Decedents
All cases were georeferenced to a point location using the street address documented on individual death certificates. For decedents with P.O. Boxes or Route number addresses (8%, 32/419), the local postmaster was able to provide physical street addresses to accurately locate most of those residences. Many decedents had a street address for the Libby Care Center, an elderly care facility in downtown Libby. Several of these decedents did not live in Libby prior to entering the Care Center, had not been living at the facility for an extended period, or were not living there during a time when they could potentially have been exposed. Therefore, efforts were made to gather more accurate residential histories on these individuals to find out where they lived before entering the Care Center, including consulting patient records at the facility, historical phone books for Lincoln County, and voter registration and sanitation records.
Baseline Population Estimates
Two methods were used to estimate population sizes within the six geographic regions under analysis (Table 1). Baseline population data for these regions were derived using 1990 U.S. Census Bureau data and confirmed with a private vendor database. Initially, the six geographic boundaries were superimposed onto a boundary file of U.S. Census blocks using a geographic information system (GIS). Data for census blocks falling completely within the respective areas of analysis were aggregated. An area proportionment technique was applied to generate population estimates within census blocks bisected by the geographic areas of analysis. These methods produced total population estimates for all six regions as well as information on the age of the populations to generate SMRs.
After determining baseline population counts using census data, estimates were generated using an alternative method. A private vendor database of phone numbers and corresponding residential addresses for the Libby area was georeferenced . Since this database describes households and not individuals, each record successfully georeferenced was multiplied by 2.5, a number derived from 1999 census projections that estimated the population of Lincoln county (18,679) and the number of households in the county (7,464).
Population estimates from georeferencing the private vendor database and using corresponding conversion methods were within 95% of the estimates produced with 1990 census data. The estimates generated with 1990 census block data were used in calculating all SMRs in this analysis.
In order to quantitatively evaluate asbestos-related mortality in Libby, the number of observed cases was converted to ratios. The use of ratios enabled the comparison of the number of cases in Libby with a reference population to determine if there was an excess in mortality in the Libby study population. The observed number of deaths was compared with "expected" numbers using ratios. The expected number of deaths was based on mortality numbers in reference populations. The references for this analysis included Montana and the entire U.S. population. The ratios of observed to expected deaths, called SMRs, were standardized to eliminate effects due to age differences among Libby residents, the state of Montana, and the United States as a whole. SMRs were calculated by dividing the number of observed cases in Libby by the expected number of cases in the general population.
SMR = Observed cases ÷ Expected cases
The expected number of cases for each underlying cause of death was generated using both Montana and U.S. age-specific mortality data. The expected number of cases using the Montana reference was generated from information provided by MDPHHS on the number and age distribution of deaths due to the selected underlying causes of death in Montana from 1979 to 1998. The expected number of cases using the U.S. reference was generated from data on U.S. mortality from 1979 to 1998 provided by the Centers for Disease Control and Prevention (CDC), National Center for Health Statistics. Age categorization by 5-year intervals (0-4, 5-9, 10-14, … 80-84, 85+) was used in computing SMRs. Following is the formula used in calculating the expected number of cases.
When the observed number of cases equals the expected number of cases, the SMR equals one (1.0). However, chance variation can influence SMR calculations, particularly when looking at mortality in small populations like Libby. Therefore, statisticians have developed methods to take this chance variation into account. Statistically significant deviations from 1.0 are evaluated using a 95% confidence interval (CI). The 95% CI determines the probability that an SMR greater or smaller than 1.0 stems from chance alone. If the 95% CI includes 1.0, then the SMR is not considered to be significant and may be due solely to chance. Following are formulas for calculating the upper and lower 95% CIs .
Lower 95% CI:
Obs x [1
- ( 1 ÷ ( 9 x
Obs)) - (1.96 ÷ (
3 x Obs0.5))]3
Upper 95% CI:
(Obs + 1)
- ( 1 ÷ ( 9 x
(Obs + 1))) + (1.96 ÷ (
3 x (Obs + 1)0.5))]3
SMRs and corresponding 95% CIs were generated for the ICD-9 groupings, sub groupings, and individual underlying causes of death listed in Appendix 1. SMRs were calculated for the entire 20-year period as well as at 5-year intervals to look for temporal trends in mortality.
MDPHHS provided paper copies of death certificates for 419 decedents documenting Libby as their place of residence at time of death and dying of one of the selected underlying causes listed previously. The majority of these deaths occurred in Montana (87%, 366/419). Some Libby residents died outside of Montana (13%, 53/419). The respective states where the deaths occurred all--except for one certificate in New York, which was not located--provided paper copies of death certificates. Washington accounted for 81% (43/53) of the out-of-state deaths in the Libby study population.
The 20-year study population of 419 decedents was primarily white (99%, 415/419) with a small number of Native Americans (1%, 4/419). There were slightly more men (53%, 221/419) than women (47%, 198/419). Age at time of death ranged from 17 to 105 years with a mean of 73 years.
Georeferencing was successfully completed on 99% (413/419) of the decedents. Six decedents were not georeferenced because they had a P.O. Box that could not be matched to a physical street address (5) or because there was no valid death certificate to abstract address information (1). Because they were distributed among different underlying causes of death, these six decedents did not significantly alter SMR calculations. They were not included in the final SMR calculations.
Additional information was obtained on 52% (47/91) of the decedents from the Libby Care Center. This subgroup contributed a large number of deaths to the analysis, yet many were not Libby residents long enough to have been potentially exposed to asbestos. Some of these decedents (34%, 16/47) were eliminated from the analysis after review of their residential history because they did not live in Libby before the closing of the mine and milling facilities in 1990. Decedents for which no additional residential information was obtained (48%, 44/91) were included in the analysis. Therefore, of the original 91 decedents from the Libby Care Center, 82% (75/91) were used in the final SMR calculations (Table 2).
Crude Mortality Rates
The crude mortality rates revealed unusual trends for all underlying causes of death (Table 3). As the geographic area of analysis decreased, crude rates increased. This was true even for referent outcomes, chosen specifically for their lack of association with the exposure of concern. Because these rates were not age-adjusted, the unusual trending may have been related to spatial patterns in the age distribution of the community.
SMRs discussed below are based on specific underlying causes of mortality. SMRs were generated for the six geographic areas of analysis and are presented in tabular form (Tables 4, 5, and 6). SMRs for asbestosis are also displayed graphically (Figure 3). The number of observed cases for Central Lincoln County is reported below for each disease-specific SMR.
All 11 cases of pneumoconiosis in the Libby study population resulted from asbestosis. Mortality resulting from asbestosis was elevated over the 20-year period. SMRs for asbestosis using the Montana population as a reference ranged from 36 to 47 depending upon the geographic area of analysis (Table 4). SMRs for asbestosis using the U.S. as a reference population ranged from 60 to 75 (Table 5). Asbestosis SMRs calculated using both the Montana and U.S. population standards were statistically significant, regardless of the geographic area of analysis. Using either Montana or U.S. age-specific mortality as a reference, one would expect Libby to have had no more than one case over the 20-year period. However, depending upon the geographic region considered, there were between five and 11 observed cases.
Of the 11 observed cases, the majority (91%, 10/11) were men. Gender-specific SMRs reflected this, as men had statistically significant elevated SMRs at all levels of geographic analysis (Tables 6 and 7). Women did not display any elevated SMRs.
No statistically significant temporal trends were detected when looking at asbestosis mortality by 5-year increments. However, if the study population was split into two 10-year increments, 73% (8/11) died in the last half of the study period, between 1989 and 1998.
There were 82 observed cases of lung cancer in the Libby study population. There was no increase in lung cancer mortality over the 20-year period. There were no temporal or gender-based trends in lung cancer mortality.
Four mesothelioma deaths were identified during the review of death certificates provided by Montana and Washington. Reviewing paper copies of death certificates was critical in identifying these deaths in the Libby area. The International Classification of Diseases, 9th Revision (ICD-9) does not contain a specific code for mesothelioma because this condition is defined by cell morphology and ICD-9 is an anatomically based system. Because mesotheliomas are found at several sites, there are multiple codes in the ICD-9 system that potentially describe this disease, including 158 (peritoneal malignancy), 163 (pleural malignancy), and 199 (malignancy without specification to site). The mesotheliomas in the Libby study population were classified as "malignancy without specification to site" (ICD-9 199) even though the death certificates clearly stated that the underlying cause was malignant mesothelioma. This discrepancy emphasizes the difficulties in trying to assess mesothelioma mortality using death certificate data. These four deaths were not used in the calculation of crude mortality rates or SMRs because it is difficult to generate accurate disease counts at the state and national levels.
There were 53 observed cases of digestive cancer in the Libby study population. There was no increase in mortality due to digestive cancer over the 20-year period. There were no temporal or gender-based trends in digestive cancer mortality.
Diseases of pulmonary circulation
There were 14 observed cases of diseases of pulmonary circulation. There was a slight increase in mortality when compared with Montana and U.S. data. SMRs generated with the Montana and U.S. population standards were consistently above the value of 1.0, yet only the SMR for the smallest geographic region was statistically significant. There were no temporal or gender-based trends in mortality due to diseases of pulmonary circulation.
There were 22 cases of breast cancer, 14 cases of prostate cancer, and 94 cases of cerebrovascular disease in the Libby study population. SMRs generated with the Montana and U.S. standard populations for all of these referent underlying causes fell near 1.0. These results support the validity of the analysis since the referent underlying causes were not associated with asbestos exposure and, as expected, were not elevated.