APPENDIX D
Exposure Analysis of Selected Studies

When evaluating epidemiologic data for risk assessment, it is important to conduct a meta-analysis of exposures for each study (see discussion in Chapter 2). This is especially important because there are often studies of exposure in an industry, geographic area, or community that can be used together to better understand the nature of exposures during the epidemiologic studies. The studies of trichloroethylene exposures conducted in cardboard manufacturing and small metal cleaning shops in the Ansberg region of Germany provide an excellent example.

COHORT STUDIES

Available studies of trichloroethylene should be rated by type, likely errors, and suitability quantification of dose-risk relationships.

Selected Studies of Aircraft Workers

Spirtas et al. (1991) and Blair et al. (1998)

These studies were conducted on the same maintenance worker cohort at Hill Air Force Base in Utah. Stewart et al. (1998) conducted a detailed exposure assessment. Their work was limited by problems linking subjects with exposures principally because solvent exposures were associated with work in “shops,” but work records listed only broad job titles and administrative units. As a result, exposures were probably substantially misclassified. Trichloroethylene was used principally for vapor degreasing



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues APPENDIX D Exposure Analysis of Selected Studies When evaluating epidemiologic data for risk assessment, it is important to conduct a meta-analysis of exposures for each study (see discussion in Chapter 2). This is especially important because there are often studies of exposure in an industry, geographic area, or community that can be used together to better understand the nature of exposures during the epidemiologic studies. The studies of trichloroethylene exposures conducted in cardboard manufacturing and small metal cleaning shops in the Ansberg region of Germany provide an excellent example. COHORT STUDIES Available studies of trichloroethylene should be rated by type, likely errors, and suitability quantification of dose-risk relationships. Selected Studies of Aircraft Workers Spirtas et al. (1991) and Blair et al. (1998) These studies were conducted on the same maintenance worker cohort at Hill Air Force Base in Utah. Stewart et al. (1998) conducted a detailed exposure assessment. Their work was limited by problems linking subjects with exposures principally because solvent exposures were associated with work in “shops,” but work records listed only broad job titles and administrative units. As a result, exposures were probably substantially misclassified. Trichloroethylene was used principally for vapor degreasing

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues and hand cleaning in some areas during 1955-1968. The investigators determined that 32% had “frequent” exposures to peak concentrations (one or two daily peaks of about 15 minutes to trichloroethylene at 200-600 parts per million [ppm]) during vapor degreasing. Work areas were located in very large buildings with few internal partitions, which aided dispersion of trichloroethylene. (This is different from the Henschler et al. [1995] and Vamvakas et al. [1998] studies, which generally had small enclosed work areas.) However, only a small number of subjects with “high” exposure had long-duration exposures, no more than 16%. Additionally, few workers were exposed only to trichloroethylene; most had mixed exposures to other chlorinated and nonchlorinated solvents. Nonetheless, these modest exposures were associated with some findings suggestive of increases in liver and biliary cancer, multiple myeloma, and non-Hodgkin’s lymphoma that were consistent with studies in animals. Conclusion: A strong exposure assessment was performed, but precision in the exposure assignments was limited by the company’s vague personnel data. The cohort had a modest number of highly exposed (about100 ppm) subjects, but overall most were exposed to low concentrations (about 10 ppm) of trichloroethylene. Garabrant et al. (1988) This study reported on the overall mortality of a cohort of workers in the aircraft manufacturing industry in southern California. The only exposure metric was years of work. An estimated 37% of the cohort was estimated to be potentially exposed to trichloroethylene, but no information was presented on how they were exposed. Given the enormous misclassification on exposure, the effect of exposure would have to be very large to be detected as an overall risk for the population. Negative findings are to be expected. Conclusion: The exposure assignments were insufficient to define exposures of the cohort and the frequency of exposures was likely low. Therefore, this study was not useful for determining whether trichloroethylene is related to increased disease risk. Morgan et al. (1998) This study evaluated a cohort of 20,508 aircraft manufacturing workers in Arizona. The company conducted a limited semiquantitative assessment

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues of exposure based on the judgment of long-term employees. No details were provided on the protocol for processing the jobs in the work histories into job classifications; no examples were provided. Exposure categories were assigned to job classifications: high = worked on degreasers (industrial hygiene reported exposures were >50 ppm); medium = worked near degreasers; and low = work location was away from degreasers but “occasional contact with [trichloroethylene].” There was also a “no exposure” catgory. No data were provided on the frequency of exposure-related tasks. Without more information, it is not possible to determine the quality of some of these assignments. Only the high category is an unambiguous setting. Depending on how the degreasers were operated, which likely changed over time, operator exposure to trichloroethylene might have been substantially greater than 50 ppm. There are too many possible situations in which an exposure category of medium or low might be assigned to determine whether the ranking is useful. Therefore, the medium and low rankings are likely to be highly misclassified. This study had limited ability to detect exposure-related effects. Conclusion: The development of exposure assignments this study was insufficient to define exposures of the cohort. Therefore, this study was not useful for determining whether trichloroethylene might cause increased risk of disease. Costa et al. (1989) A small cohort of 8,626 aircraft manufacturing workers in Italy was studied. No exposure assessment was used. Conclusion: This study was not useful for determining whether trichloroethylene may cause increased risk of disease. Cancer Incidence Studies Using Biological Monitoring Databases Finland and Denmark historically have maintained national databases of biological monitoring data obtained from workers in industries where toxic exposures are a concern. Legislation required that employers provide workers exposed to toxic hazards with regular health examinations, which must include biological monitoring to assess the uptake of toxic chemicals, including trichloroethylene. In Sweden, the only local producer of trichloroethylene operated a free exposure-surveillance program for its customers, measuring urinary trichloroacetic acid (U-TCA). These programs used the linear relationship found for average inhaled trichloroethylene versus U-

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues TCA: trichloroethylene (mg/m3) = 1.96; U-TCA (mg/L) = 0.7 for exposures lower than 375 mg/m3 (69.8 ppm) (Ikeda et al. 1972). This relationship shows considerable variability among individuals, which reflects variation in urinary output and activity of metabolic enzymes. Therefore, the estimated inhalation exposures are only approximate for individuals but can provide reasonable estimates of group exposures. There is evidence of nonlinear formation of U-TCA above about 400 mg/m3 or 75 ppm of trichloroethylene. The half-life of U-TCA is about 100 hours. Therefore, the U-TCA value represents roughly the weekly average of exposure from all sources, including skin absorption. The Ikeda et al. relationship can be used to convert urinary values into approximate airborne concentration, which can lead to misclassification if tetrachlorethylene and 1,1,1-trichloroethane are also being used because they also produce U-TCA. In most cases, the Ikeda et al. relationship provides a rough upper boundary of exposure to trichloroethylene. Anttila et al. (1995) This Finnish study evaluated cancer risk in a small cohort of individuals (2,050 males and 1,924 females) who had been monitored between 1965 and 1982 for exposures to trichloroethylene by measuring their U-TCA. The main source of exposure was identified as degreasing or cleaning metal surfaces. Some workplaces identified rubber work, gluing, and dry cleaning. There were an average of 2.7 measurements per person. Using the Ikeda et al. (1972) conversion relationship, the exposure for trichloroethylene was approximately 7 ppm in 1965, which declined to approximately 2 ppm in 1982; the 75th percentiles for these dates were 14 and 7 ppm, respectively. The maximum values for males were approximately 380 ppm during 1965 to 1974 and approximately 96 ppm during 1974 to 1982. Females showed a similar pattern over time but had somewhat higher exposures during the 1970s (approximately 4 ppm). Duration of exposure was counted from the first measurement of U-TCA, which might underestimate the length of exposure. Without job histories, the length of exposure is uncertain. Another concern is the sampling strategy; it was not reported how the workers were chosen for monitoring. Therefore, it is not clear what biases might be present in the data, especially the possibility of undersampling highly exposed workers. Conclusion: This study had a small cohort drawn from a wide variety of industries, predominantly degreasing and metal cleaning. Exposures to trichloroethylene were generally low, most less than 14 ppm. The

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues maximum values were generally less than 400 ppm. The duration of exposure was uncertain. Axelson et al. (1994) This Swedish study evaluated cancer risk in a small cohort of individuals (1,421 males and 249 females), who were monitored for U-TCA as part of a surveillance system by the trichloroethylene producer during 1955 to 1975. Eighty-one percent of the cases had low exposures (<50 mg/L), corresponding to an airborne concentration of trichloroethylene of approximately 20 ppm. There was uncertainty about the beginning and end of exposure. Exposure was assumed to begin with the first urine sample and to end in 1979 (the reason for this date is unclear). Because the investigators did not have job histories, there is considerable uncertainty about the duration of exposure. Most subjects appear to have had short durations of exposure, but these might have been underestimated. Another concern is the sampling strategy. It was not reported how the workers were chosen for monitoring. Therefore, it is not clear what biases could be present in the data, especially the possibility of undersampling highly exposed workers. Conclusion: This study had a small cohort drawn from a wide variety of industries, predominantly degreasing and metal cleaning. Exposure to trichloroethylene was generally low (most less than 20 ppm). The duration of exposure was uncertain. Hansen et al. (2001) This Danish study evaluated cancer risk in a small cohort of individuals (n = 803) who had been monitored for trichloroethylene exposures in a national surveillance program. The retirement and measurement records contained general information about the type of employer and the subject’s job. The subjects in this study came predominantly from the iron and metal industry with jobs such as metal-product cleaner. Each subject had 1 to 27 U-TCA measurements, going back to 1947. Using the linear relationship from Ikeda et al. (1972), the historic median exposures estimated from the U-TCA concentrations were rather low: 9 ppm for 1947 to 1964, 5 ppm for 1965 to 1973, 4 ppm for 1974 to 1979, and 0.7 ppm for 1980 to 1989. However, the distributions were highly skewed, with coefficients of variation of 160% to 370% or estimated geometric standard deviations of 1.9 to 3.7. This is clear evidence that, in general, workers in a wide variety of industry and job groups and identified as “exposed” have low exposures.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues Conclusion: This study had a small cohort drawn from a wide variety of industries, predominantly degreasing and metal cleaning, who had generally low exposures (most less than 20 ppm). Studies of Other Cohorts Sinks et al. (1992) An epidemiologic study was conducted of renal cancer among paper board printing workers. The process involved producing food containers from waxed paperboard. Exposures were poorly described. Trichloroethylene was mentioned in material-safety data sheets for one or more materials used by the process but no information was provided about where or how the material was used. Some benzidine-containing materials were used in the process as inks for printing colored forms on the food containers. It was not possible to assess the degree of contact with trichloroethylene or the printing inks. Conclusion: This study was not useful for assessing risks associated with exposures to trichloroethylene. Raaschou-Nielsen et al. (2003) A cohort of 40,049 blue-collar workers was drawn from 347 companies with documented trichloroethylene use. A separate exposure assessment was conducted using regulatory agency data from 1947 to 1989 (Raaschou-Nielsen et al. 2002). The percentage of exposed workers was found to decrease as company size increased: 81% for <50 workers, 51% for 50-100 workers, 19% for 100-200 workers, and 10% for >200 workers. About 40% of the workers in the cohort were exposed (working in a room where trichloroethylene was used). Smaller companies had higher exposures. Median exposures to trichloroethylene were 40-60 ppm for the years before 1970, 10-20 ppm for 1970 to 1979, and approximately 4 ppm for 1980 to 1989. Conclusion: Only a small fraction of the cohort were exposed to trichloroethylene. The highest exposures occurred before 1970, and the iron and metal industry doing degreasing and cleaning with trichloroethylene had the highest exposures, with a median concentration of 60 ppm and a range up to about 600 ppm.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues Henschler et al. (1995) This was a cohort study of workers in a cardboard factory in the area of Arnsberg, Germany. Trichloroethylene was used in this area until 1975 for degreasing and solvent needs. Plant records indicated that 2,800-23,000 L per year was used. Small amounts of tetrachloroethylene and 1,1,1-trichloroethane were used occasionally, but in much smaller quantities than trichloroethylene. Trichloroethylene was used in three main areas: cardboard machine, locksmith’s area, and electrical workshop. Cleaning the felts and sieves and cleaning machine parts of grease were done regularly every 2 weeks, in a job that required 4-5 hours, plus whatever additional cleaning was needed. Trichloroethylene was available in open barrels and rags soaked in it were used for cleaning. The machines ran hot (80-120°C) and the cardboard machine rooms were poorly ventilated and warm (about 50°C), which would strongly enhance evaporation. This would lead to very high concentrations of airborne trichloroethylene. Cherrie et al. (2001) estimated that the machine-cleaning exposures to trichloroethylene were greater than 2,000 ppm. Workers reported frequent strong odors and a sweet taste in their mouths. The odor threshold for trichloroethylene is listed as 100 ppm (ATSDR 1997a). Workers often left the work area for short breaks “to get fresh air and to recover from drowsiness and headaches.” Based on reports of anesthetic effects, it is likely that concentrations of trichloroethylene exceeded 200 ppm (Stopps and McLaughlin 1967). Those reports, the work setting description, and the large volume of trichloroethylene used are all consistent with very high concentrations of airborne trichloroethylene. The workers in the locksmith’s area and the electrical workshop also had continuous exposures to trichloroethylene associated with degreasing activities; parts were cleaned in cold dip baths and left on tables to dry. Trichloroethylene was regularly used to clean floors, work clothes, and hands of grease, in addition to the intense exposures during specific cleaning exercises, which would produce a background concentration of trichloroethylene in the facility. Cherrie et al. (2001) estimated the long-term exposure to trichloroethylene was approximately 100 ppm. Conclusion: The subjects in this study clearly had substantial peak exposures to trichloroethylene that exceeded 2,000 ppm and probably sustained long-term exposures greater than 100 ppm, which are unconfounded by concurrent exposures to other chlorinated organic solvents.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues CASE-CONTROL STUDIES Population level job-exposure matrices (JEMs) have been developed for the United States, Canada, United Kingdom, Germany, and some Scandinavian countries. Community level JEMs have been developed for some cities and cancer registries (e.g., Montreal). If investigators do not develop their own local JEM, local variations across broad regions and countries can add misclassification. Consequently, the use of multiple exposure assignment strategies, such as multiple JEMs and self-assessment by questionnaires, will not give the same results when applied to a given population. For example, the study of renal cell cancer and trichloroethylene exposures by Brüning et al. (2003) applied the CAREX database information developed by the International Agency for Research on Cancer and the British JEM develop by Pannett et al. (1985) to assign exposures to jobs in the work histories of a German set of cases and controls. These two schemes were then compared with a self-assessment of exposures obtained from a questionnaire. In this case, the workers had more knowledge of their personal exposures than usual because they had been told or had observed that they were using either trichloroethylene or Perc (tetrachloroethylene) for degreasing. The schemes gave related but different results. Comparisons were difficult because they did not use the same groupings of industries, which would result in differences in misclassification. For example, the most consistent finding for the CAREX data was a significant increase in renal cell cancer for jobs in “land, air, and sea transport” and “cleaning and waste disposal” and no elevation for “iron, steel, and nonferrous industries” and “occupations with contact to metals,” which are difficult to interpret. Pesch et al. (2000a,b) noted several important general points about exposure assessment for case-control studies. First, the chemical information available to many workers is insufficient for valid recall of agent-specific exposures in general population studies. Second, risk cannot be determined for low-prevalence jobs or industries when studying a regional population. Third, associations with exposure can be highly nonlinear because categoric assignments of duration (“short,” “medium,” “long,” or “very long”) and intensity (“low,” “medium,” “high,” or “substantial”) are arbitrary and do not necessarily represent a linear dose relationship. Fourth, large numbers of job titles and low prevalence of long-duration, intense exposure within job title groups greatly reduce power. Job tasks using specific materials have a higher specificity for exposure, but data gathering is limited by interviewer knowledge of the exposures. Fifth, duration and cumulative exposure variables do not consider age at first exposure, which also affects cancer risk.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues TABLE D-1 Trichloroethylene Exposure Summary for the Arnsberg Area Studies Study Peak Exposures Long-Term Exposures Notes Henschler et al. 1995 >2,000 ppm, machine cleaning with neurological symptoms; about 100 ppm continuous cold cleaning 100 ppm 100 ppm Cherrie et al. (2001) estimates Vamvakas et al. 1998 400-600 ppm hot cleaning with neurological symptoms 100 ppm Cherrie et al. (2001) estimates Brüning et al. 2003 400-600 ppm hot cleaning, with neurological symptoms 100 ppm   Studies in the Arnsberg Region of Germany A series of studies (including Henschler et al. [1995]) have been conducted in an area with a long history of trichloroethylene use in several industries. The main importance of these studies is that there is considerable detail on the nature of exposures, which made it possible to estimate the order of magnitude of exposure, even though there were no direct measurements (see Table D-1). Vamvakas et al. (1998) In a follow-up to the Henschler et al. (1995) study, a case-control study was conducted in the Arnsberg region of Germany where there has long been a high prevalence of small enterprises manufacturing small metal parts and goods, such as nuts, lamps, screws, and bolts. Exposures to trichloroethylene resulted from dipping metal pieces into vats, with room temperatures up to 60°C, and placing the wet parts on tables to dry. Some work rooms were noted to be small and poorly ventilated. These conditions are likely to result in high inhalation exposure to trichloroethylene (100-500 ppm). Cherrie et al. (2001) estimated the long-term exposures to be approximately 100 ppm. Some of the cases included in this study were also pending legal compensation. As a result, there had been considerable investigation of the exposure situation by occupational hygienists from the Employer’s Liability Insurance Association and occupational physicians, including walk-through visits and interviews of long-term employees. The legal action could introduce a bias, a tendency to overreport some of the subjective reports by the subjects. However, the objective working conditions were assessed by knowledgeable professionals, who corroborated the presence of the poorly controlled hot

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues dip tanks, extensive use of trichloroethylene for all types of cleaning, and the process descriptions. Conclusion: These workers had substantial, sustained exposures to high concentrations of trichloroethylene at 400-600 ppm during hot dip cleaning and greater than 100 ppm overall. Cherrie et al. (2001) concluded that the exposures in the aircraft industry evaluated by Stewart et al. (1991) were likely similar to those of Vamvakas et al. (1998). However, Cherrie et al. overlooked the important difference in room size between these settings: the small- and medium-sized businesses would be using much smaller work rooms than the aircraft-hangar-sized work rooms of the aircraft manufacturers, which would allow the local emissions to dissipate to a greater degree than in the German settings. Cherrie et al. focused on the durations of exposure, without also considering the substantial differences in the types of exposures, which would substantially affect exposure intensity and associated symptoms. Brüning et al. (2003) This study is a second case-control follow-up of renal cell cancer in the Arnsberg area of Germany, which was intended to deal with some of the methodological issues present in the two earlier studies. The major advantage of studies in the Arnsberg area is the high prevalence of exposure to trichloroethylene because of the large number of companies doing the same kind of industrial work. An interview questionnaire procedure for self-assessment of exposures similar to the one used by Vamvakas et al. (1998) was used to obtain detailed information about solvents used, job tasks, and working conditions, as well as the occurrence of neurological symptoms. The industry and job title information in the subjects’ job histories were also analyzed by two schemes of expert-rated exposure assignments for broad groups of jobs. The CAREX database from the European Union and the British JEM developed by Pannett et al. (1985) were used. This was done to obtain a potentially less biased assessment of the exposures. However, both of these rating approaches are very broad and they have potentially high rates of misclassification of exposure intensity in job groupings and industry groupings. High-intensity exposures tend to be a small fraction of any broad grouping with exposure, such as topside workers among all cokeoven workers (Lloyd 1971). This dilution of risk increases with the breadth of the group. The substantially increased risk of lung cancer in coke-oven workers is nearly undetectable among all steel workers, whose standardized mortality ratio shows only a small elevation (Lloyd and Ciocco 1969).

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues In an attempt to avoid reporting biases associated with the legal proceeding for compensation, analyses were conducted on self-reported exposure to selected agents (yes or no). The regional use of trichloroethylene and Perc (tetrachloroethylene) were so widespread that most individuals recognized the local abbreviations. If individuals claimed to be exposed when they were not, it would reduce the finding of a relationship if one existed. Similarly, subjects were grouped by frequency of perceived symptoms (any, less than daily, and daily). Overreporting would also introduce misclassification and reduce evidence of any relationship. Self-reporting of exposure to chemicals in case-control studies is generally considered unreliable because, within the broad population, workers rarely know which specific chemicals they are exposed to. However, in cohort studies and case-control studies in which one industry dominates a local population, this is less likely because the numbers of possible industries and job titles are much smaller than in a broad population. The Arnsberg area studies focused on a small area where one type of industry was very prevalent, and that industry used primarily just two solvents: trichloroethylene and tetrachloroethylene. As a result, it was common knowledge among the workers what solvent an individual was using, and, for most, it was trichloroethylene. If the base population was enlarged to include more areas, which did not have the same focus of industry, the ability to detect this association would be expected to decrease, as was found in the study by Pesch et al. (2000a), which included the Arnsberg area plus four other areas (reviewed below). Conclusion: Again, this is the same basic Arnsberg population studied by Vamvakas et al. (1998), so the exposures will be the same: substantial, sustained high exposures to trichloroethylene at 400-600 ppm during hot dip cleaning and greater than 100 ppm overall. Pesch et al. (2000a) This was a multicenter study of renal cell carcinoma in Germany, which included the Arnsberg region plus four others. Two general JEMs, British and German, were used to assign exposures based on subjects’ job histories reported in an interview. Researchers also asked about job tasks associated with exposure, such as metal degreasing and cleaning, and use of specific agents (organic solvents, chlorinated solvents, including specific questions about carbon tetrachloride, trichloroethylene, and tetrachloroethylene). A category of “any use of a solvent” mixes the large number with infrequent slight contact with the few noted earlier who have high-intensity and prolonged contact.

OCR for page 406
Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues Conclusion: While this case-control study includes the Arnsberg area, several other regions are included as well, where the source of the trichloroethylene and chlorinated solvent exposures are much less well defined. As a result, most subjects identified as exposed to trichloroethylene probably had minimal contact, averaging concentrations of about 10 ppm or less. Brauch et al. (1999, 2004) This pair of studies evaluated mutations in the von Hippel-Lindau gene in subjects drawn from the Vamvakas et al. (1998) cohort. The findings were analyzed with the combination-exposure index that uses both exposure duration and severity of neurological symptoms. Conclusion: This is an appropriate use of the Vamvakas et al. (1998) exposure assignments for the individual cases evaluated. These workers had substantial, sustained high exposures to trichloroethylene at concentrations of 400-600 ppm during hot dip cleaning and greater than100 ppm overall.