Document a2Gj8L1BVe43M4kpG5r3zwmX

Ann, otesp, ffyg., VO. 4$, Nk 4,pp. 309-326.2004 2004 British OeeMpaijonsl Hygime Society Published by Oxford University Press DO!: t0.i093Mthyg/meh022 Mesothelioma and Lung Cancer Among Motor Vehicle Mechanics: a Meta-analysis MICHAEL GOODMAN1*, M. JANE TETA2, PATRICK A. HESSEL3, DAVE) H. GARABRANT4, VALERIE A. CRAVEN5, CAROLYN G. SCRAFFORD6 and MICHAEL A. KELSB7 'Emory University Rollins School ofPublic Health, 1518 Clifton Road NE, Atlanta, GA 30322, USA; 2Exponent Health Group, 1730 Rhode IslandAvenue NW, Suite 1100, Washington, D.C 20036, USA; 'Exponent Health Group, 2 North Riverside Plaza, Suite 1400, Chicago, 1L 60606, USA; 'University of Michigan School ofPublic Health, 1420 Washington Heights, Room 6529, Ann Arbor, MI48109-2029, USA; 'Exponent Health Risk Group, 631 First Street, Santa Rosa, CA 95404, USA; 'Exponent Food and Chemicals Group, 1730 Rhode IslandAvenue NW, Suite 1100, Washington, DC 20036, USA; 'Exponent Health Group, 149 Commonwealth Drive, Menlo Park, CA 94025, USA Received 26 August 2003; in final form 13 November 2003; published online on 17 May 2004 We conducted a systematic review and analysis of the epidemiological literature that examines the risk of lung cancer and mesothelioma among motor vehicle mechanics who may have been engaged in brake repair and, thus, were potentially exposed to asbestos. All relevant studies were classified into three tiers according to their quality. Tier ID (lowest quality) studies were cited for completeness, but were not included in the meta-analysis. Meta relative risks (metaRRs) were calculated for mesothelioma and lung cancer using both fixed and random effects models for Tiers I and DC, separately, followed by stratified analyses based on study design or exposure characterization (garage workers versus brake workers) and, for lung cancer studies, based on adequate adjustment for smoking. The meta-analysis for Tier I (higher quality) and Tier D (lower quality) studies of mesothelioma yielded RR estimates of 0.92 (95% Cl 0255-1.56) and 0.81 (95% Cl 032-1.28), respectively, farther stratification according to exposure charac terization did not affect the results. The meta-analysis for lung cancer produced RR estimates of 1.07 (95% Cl 0.88-1Jl) for Tier I and 1.17 (95% Cl 1.01-136) for Tier H. When the hmg cancer analysis was limited to studies that used adequate control for smoking, the resulting RR estimate was 1.09 (95% Cl 0.92-138). Based on these findings, we conclude that employment as a motor vehicle mechanic does not increase the risk of developing mesothelioma. Although some studies showed a small increase in risk oflung cancer among motor vehicle mechanics, the data on balance do not support a conclusion that lung cancer risk in this occupational group is related to asbestos exposure. Keywords: asbestos; brakes; epidemiology; lung cancer; mesothelioma; meta-analysis; motor vehicle mechanics INTRODUCTION The causal association between inhaled asbestos fibers and the development of lung cancer is well established. While the causal role of amphibole asbestos in the development of mesothelioma is also clear, there is still disagreement regarding the doseresponse relationship between chrysotile asbestos "Author to whom correspondence should be addressed. Tel: +1 404 727 2734; Fax: +1 404 727 8737; E-mail: mgoodm2@sph.emory.edu and mesothelioma and the role of amphibole contam inants in that relationship (Hodgson and Damton, 2000). Excess risk of these two cancers has not been found in all settings where there is potential exposure to asbestos. Recent attention has shifted from highly exposed occupational groups such as insulators and shipyard workers to those with asbestos exposures that could be both qualitatively and quantitatively different One such occupation is motor vehicle repair, where exposure to short chrysotile fibers can occur during installation and repair of asbestoscontaining brakes. (In North America, automobile 309 ALL-STATE LEGAL4 HWBUI0011484 310 M. Goodman ei aL brakes typically contained chrysotile asbestos embedded in a solid binder.) The process of brake replacement involves two potential opportunities for release of asbestos fibers: (i) small amounts of chrysotile asbestos (usually less than 1%) that may be present in the brake wear debris and (ii) asbestos that can be released during grinding and beveling of new asbestos brake linings or pads. Some authors (Lorimer et aL, 1976), regulatory agencies (EPA, 1986c) and trade organizations (World Trade Organization, 2000) have opined in the past that motor vehicle mechanics are likely to be at increased risk ofdeveloping asbestos-related disease, most notably mesothelioma. These opinions have been based primarily on the fact that asbestos expo sures can occur during brake work and cases of mesothelioma have been reported among workers who had done brake repair (EPA, 1986a,b,c). ' When the EPA conducted its evaluation (EPA, 1986b.c) the epidemiological information on mesothelioma among vehicle mechanics was limited to only three studies (McDonald and McDonald, 1980; Teta et aL, 1983; Spirtas et aL, 1985). However, in more recent years a number of addi tional epidemiological studies have examined the risk. of mesothelioma and/or lung cancer among motor vehicle mechanics or specifically among brake workers. These studies are preferable to case reports and case series in assessing associations between exposure and disease. We conducted a systematic review of the epidemi ological literature examining the relative risks of mesothelioma and lung cancer among workers engaged in motor vehicle repair and, when possible, among workers occupationally exposed to brake dust. A previous review of the literature examined six case-control studies of mesothelioma among garage mechanics (Wong, 2001). However, we felt that these analyses could be enhanced by including additional published and unpublished studies and by including a re-analysis of one of the original data sets. In addi tion, we expanded the scope of our review beyond mesothelioma to include studies of lung cancer. This issue is of growing scientific, public health and societal importance (Schneider and Smith, 2000; Truby, 2002). Large numbers of people have been exposed to brake dust over the last several decades (Lorimer et aL, 1976; Nicholson et aL, 1984; Huncharek, 1990) and an increased risk of asbestosrelated cancers among these workers could translate into a substantial burden of disease. METHODS Study selection A number of electronic literature databases were searched using a variety of search strategies and multiple combinations of keywords such as `asbestos', `brakes', `mesothelioma', `lung cancer', `cancer', 'garage mechanics', `automobile mechanics', "motor mechanics', `mechanics', etc. Copies of the articles were obtained, including those from foreign language journals, which were trans lated into English. Reference lists of identified arti cles were examined to locate additional studies. Internet and literature searches were also conducted to identify relevant studies that were not published in the peer-reviewed literature. Of partic ular interest were government documents and book chapters. When information was missing from published reports, attempts were made to contact the authors to obtain the missing information. In order to be included in the review, studies were required to meet all of the following criteria: outcomes of interest included mesothelioma and/or lung cancer; relative risk estimates and associated variance measures were either reported by the authors or could be calculated based on the data obtained from the authors or reported in the papers; the exposed population was involved in motor vehicle repair, excluding general mechanics. After this initial study selection, the meta-analysis included two steps: (i) a review and quality scoring of each study and (ii) a quantitative analysis of the pooled measures of association from studies that met the inclusion criteria. Review ofthe literature AH potentially relevant studies underwent a formal evaluation and were assigned a quality score according to their methodological strengths and weaknesses. The general approach involved awarding each study a point (+1) for each methodo logical strength and penalizing with a negative score (-1) for each evident shortcoming. The quality scoring was conducted according to the following criteria. Overall study design; proportionate mortality/ incidence ratio (PMR/PIR) studies or death cer tificate-based standardized mortality odds ratio (SMOR) studies = -1; else (cohort or case-con trol studies) = 0. Asbestos exposure: non-specific = 0 (e.g. `car mechanic'); specific = 1 [e.g. `brake repairmen' or industrial hygiene (IH) based]. Age adjustment no = -1; yes = 0. Confounding by other occupational exposure: likely = -1 (e.g. studies where motor vehicle mechanics with other multiple occupations were compared with persons with no history of any atrisk occupations); possible but not clearly evi dent = 0; unlikely/addressed = I (e.g. studies that Mesothelioma and lung cancer among motor vehicle mechanics 311 accounted for other known at-risk occupations in the analysis). Exposure-response analysis: no = 0; yes = 1. Analysis by latency: no = 0; yes = 1. For case-control studies: response rate <80% or not reported = -1; >90% = 1; else = 0. For cohort studies: follow-up: <10 years = -I; >20 years = I; else = 0. Reporting bias: likely = -I (e.g. interview-based case-control studies with dear differences in terras of sources of information between cases and controls): possible but not clearly evident = 0; unlikely/addressed = l (e.g. in case-control studies using recorded occupational histories). Selection bias: likely = -1 (e.g. due to reliance on referral of cases to a clinic or using inappropriate controls); possible but not clearly evident = 0 (e.g. in hospital-based case-control studies); unlikely/addressed = 1 (e.g. in cohort studies or population-based case-control studies). In addition to the above criteria, mesothelioma studies were evaluated based on whether or not the diagnoses were confirmed by a pathology review (i.e. +1 if yes, 0 ifno). Lung cancer studies were also eval uated based on their ability to adjust results for smoking habit. Studies that did not control for smoking received a negative score (-1), studies that adequately controlled for both smoking status (current, former or never) and intensity (duration and/or number of cigarettes per day) received a posi tive score (+1) and studies that had only partial control for smoking (e.g. using only ever-never cate gories, only pack-years or using blue collar/mtemal reference groups) were neither penalized nor rewarded. The scoring was used as a formal approach to clas sify all studies into three tiers. Tier III included studies that had an overall negative score (i.e. less than 0) and were considered unreliable. These studies were only mentioned for completeness and were not included in the meta-analysis. Those studies that had scores ofzero or above were divided into two approx imately equal groups. Studies with the higher (above median) total score were included in Tier I and considered most informative. Tier II included the remaining studies that received a total score of >0 but were considered less useful due to methodological shortcomings. Statistical analysis We calculated a meta relative risk (meta-RR) for Tiers I and U separately, followed by stratified anal yses of both tiers combined, based on study design or exposure characterization and, for lung cancer studies, based on adequate adjustment for smoking. The necessary input from each study included (i) an RR estimate and (ii) the associated measure of variance, which usually can be derived from the 95% confidence interval (95% Cl). When case-control studies did not report the results in terms of RRs but did provide information necessary to reconstruct the two-by-two tables, the odds ratios (ORs) and 95% CIs were calculated using Epi Info software (CDC, 2001). Some cohort studies did not report 95% CIs, but did provide information on the numbers of observed and expected cases. In those instances, 95% CIs were calculated based on the Poisson distribution, as recommended by Bresiow and Day (1987). In one study (Morabia era/., 1992), the information needed to calculate 95% CIs was not provided. However, we were able to calculate the standard deviation based on the information that there were 39 exposed controls (1.2% of all controls) and the statis tical power to detect an OR of 1.5 was 0.41, as reported by the authors. There are two general approaches for combining the data in a meta-analysis: a fixed effects model or a random effects model. The fixed effects method assumes no heterogeneity among studies and attributes all observed variations among results to sampling error alone (Sutton et al, 1998). The random effects model assumes that the study-specific effect sizes arise from a random distribution of effect sizes with a certain mean and variance. All analyses involved a test for heterogeneity. However, the interpretation of the test for heteroge neity is problematic because of the wide variation in study designs, study populations and reference groups. Therefore, we used both the fixed and random effects models for each analysis. Where the variability among studies was negligible (high level of homogeneity), the random effects model reduced to a fixed effects model (Sutton et aL, 1998). The details of calculations for both models are provided in the Appendix. Mesothelioma RESULTS Overview ofthe literature. Relative risk estimates could not be calculated for three cohort studies of motor vehicle mechanics (two from Sweden and one from Denmark) that provided information on mesothelioma (Jarvholm and Brisman, 1988; Hansen, 1989; Gustavsson et at, 1990). For this reason, these studies could not be included in the meta-analysis. These three cohort studies combined reported three observed cases of mesothelioma and one case of `pleural cancer'. All three mesothelioma cases had other potential occupational asbestos expo sures. For the remaining case of pleural cancer, infor mation regarding other exposures was not provided. 312 M. Goodman etal There were 11 studies that reported (or permitted of mesothelioma. The use of non-pulmonary cancers calculations of) the relative risk estimates for mesothe as controls has both advantages and disadvantages. lioma. These studies underwent formal evaluation Hospital-based cancer cases may not be representa and scoring. The results of scoring for each study are tive of the general population (i.e. possible selection presented in Table 1. Four studies were included in bias). However, the choice of controls with other Tier III (Coggon et al, 1995; Hodgson et al., 1997; cancers may have reduced recall bias. The use of the Milham and Ossiander, 2001; NIOSH, personal highest asbestos exposure to characterize each partic communication, 2002). Of the seven remaining studies, four studies with scores between 3 and 5 were included in Tier I (Table 2A) and three studies with scores between 0 and 2 were included in Tier II (Table 2B). All relevant studies were published in ipant's occupational history helped decrease poten tial confounding by other asbestos exposures. However, the occupational category defined as `garage' is insufficiently specific. English. Two studies were conducted exclusively in the USA, one study was conducted in Canada, one combined US and Canadian data and three took place in Europe (one in Germany, one in Denmark and one in Spain). The years of publication ranged from 1980 to 2004. A more detailed discussion of each Tier I and Tier study follows. Teta et al (1983). After a pathology review, cases from the Connecticut Tumor Registry and from a large Veterans Administration hospital were compared with controls selected from the death certificate files of the Connecticut State Department of Health Services. Occupational histories for cases and controls were obtained from death certificates TterL McDonald and McDonald (1980). The study here compared histologically confirmed mesothelioma cases to matched controls who had pulmonary metastases from non-pulmonary malignancies. Occupa tional histories obtained through interviews with relatives were ranked according to their potential for asbestos exposure. Of the 156 cases and 156 controls and from city directories. The OR for subjects employed in `automobile repair and related service' was 0.65 (95% Cl 0.08-5.53). Unlike other case-control studies, Teta et al. (1983) relied on objective historical employment information rather than subjective reports from inter views. The choice of population controls, the high without a recognized increase in mesothelioma risk, response rate and the histological confirmation of the the occupation `garage' was reported for 11 cases and mesothelioma diagnoses have to be considered as i 12 controls, from which we calculated an OR of 0.91 methodological strengths. The main shortcomings (95% a 0.35-2.34). are non-specific exposure characterization and the This study had a large sample size, a high response inability to eliminate other exposures due to its small rate and used pathologists to establish the diagnosis size. Table 1. Quality scores of studies evaluating the association between mesothelioma risk and employment as a motor vehicle i mechanic i Quality criteria Study i 1 2 3 4 5 6 7 3 9 10 n Overall study design; PMR/PIR/SMOR = -1; else = 0 0 0 0 0 0 0 0 -1 -1 -1 -l I Asbestos exposure; non-specific (e.g. 'car mechanic') = 0; 0 1 1 00 0 1 0 0 0 0 specific (e.g. brake repairman, or IH-based) = 1 Age adjustment: no = -1; yes = 0 0 0 0 -l 0 0 -1 0 0 0 0 Confounding by other occupational exposure: likely - -1; possible = 0; unlikely/addressed - 1 1 11 00 0 0000 0 Exposure-response analysis: no = 0; yes ~ 1 0 1 0 00 0 0 0 0 0 0 ! Analysis by latency: no ~ 0: yes = 1 0 01 00 0 0000 0 For case-control studies: response rate: <80% or not reported = -I; >90% = 1; else = 0 1 -1 0 11 1 1 000 0 For cohort studies: follow-up <10 yr = -1; >20 yr= 1; else = 0 0 00 00 0 0000 0 Reporting bias: likely = -1; possible = 0; unlikely/addressed = 1 0 0 1 -1 1 0 0000 0 Selection bias: likely = -1; possible = 0; unlikely/addressed = 1 0 10 1I 0 0000 0 Diagnosis of mesothelioma by pathology review: no = 0; yes = 1 1 11 11 0 1 000 0 Total score 3 45 14 1 2 -1 -1 -1 -1 1, McDonald and McDonald (1980); 2, Hessel et al (2004); 3, Teschke et al. (1997); 4, Agudo et al. (2000); 5, Teta el at (1983); 6. Hansen (personal comumcation, 2003); 7, Woitowira and Rijdelsperger (1994); 8, Coggon et al. (1995); 9, Hodgson el al. (1997); 10, MHham and Ossiander (2001); 11, NIOSH (personal comunication, 2002). HWBUI0011487 Mesothelioma and long cancer among motor vehicle mechanics 7m w \g> es ~ m 3a <s1 4s 4s M 66 rt If oooo II 3 5 32 no n 03 M DCS 3Q o f S' o2 3g t1 g O-- I 1% O 60 Jj Is n 82 H I a S II S fI i* u sg .s g Q 2 f X 313 s JiSt =5= M.ao8 s| 2 O > eo > c g S3 3 O* <3\ O I1 I 5 I oS3 gl HWBUI0011488 314 M. Goodman e! al HWBUI0011489 Mesothelioma and lung cancer among motor vehicle mechanics 315 National Cancer Institute study (Spirtas et al, Although fairly small in size, this study is strong 1994,1985; Hessel et al., 2004, in press). The 1994 report of a case-control study by the National Cancer Institute (NCI) identified mesothelioma cases in the Los Angeles County Cancer Surveillance Program, the New York State Cancer Registry (excluding New York City) and 39 Veterans Administration hospitals. A pathology review was conducted in most cases. Controls included patients who died of causes other methodologically because it considered other asbestos exposures, specified exposure as 'brake lining installation or repair', relied on histologically confirmed diagnoses of mesothelioma and frequency matched next-of-kin interviews among cases and controls. Unlike other studies, the analyses in Teschke et al. (1997) considered latency. than cancer, respiratory disease, suicide or violence. The authors did not calculate ORs by occupational category because of multiple confounding exposures, preferring to discuss attributable risk of asbestos exposure in general. Among subjects engaged in `brake lining installation or repair1, 33% also had shipbuilding or shipyard work and 55% had performed insulation work. In an earlier analysis of the same data, but with cases and controls matched for age, the authors did calculate the ORs for different occupational groups (Spirtas et aL, 1985). Brake lining installation or repair had an OR of 1.0 (95% Cl 0.6-1.6). . A re-analysis of these data (Hessel et al, 2004, in press) for the occupational category `brake instal lation or repair' yielded an OR of 1.04 (95% Cl 0.462.22). After adjusting for any of eight occupations with potential asbestos exposure, the OR was 0.82 (95% Cl 0.36-1.80). When cases and controls with a history of employment in any of the eight occupa tions were removed from the analysis, the OR for Tier II studies. Olsen and Jensen (1987) and Hansen (personal communication, 2003). In 1987, Olsen and Jensen published a proportionate inci dence ratio (PIR) surveillance study that linked cases from the Danish Cancer Registry with occu pational histories (Olsen and Jensen, 1987). There were no cases of mesothelioma (pleural or perito neal) for the occupational category 'repair of motor vehicles and motorcycles' and for the industry category `garage'. We contacted the authors of this study and learned that the data had been updated in a case-control study (Hansen, personal communication, 2003). For the category `repair of motor vehicles and motorcycles' the OR was 0.8 (95% Cl 0.4-1.5), based on 10 cases. The weaknesses of this study are lack of histological confirmation of mesothelioma diagnosis and inability to obtain a complete work history. Its strengths include analysis by latency and a relatively large sample size. occupational brake installation and repair was 0.62 Woitowitz and Rodelsperger (1994). This German (95% C3 0.01-4.71). These NCI data have several case-control study compared occupational histories important features: (i) exposure was defined as brake of 324 pathology confirmed mesothelioma with two installation and repair, (ii) confounding by other groups of controls: 315 hospital controls selected occupational exposures could be addressed, (ill) among patients who underwent lung resection and information on duration of employment allowed an 182 population controls. Sixteen cases. 16 hospital exposure-response analysis and (iv) and the majority controls and 12 population controls were listed as ! of the mesotheliomas underwent pathology review. A `motor vehicle repair workers'. Calculations based limitation of this data set is the relatively low (<80%) on these data produced an OR of 0.97 (95% Cl 0,45- response rate. 2.12) using hospital controls and 0.74 (95% Cl 0.32-- ! Teschke et al. (1997). Teschke and colleagues 1,75) using population controls. For people definitely compared pathology-confirmed mesothelioma cases engaged in brake service, the OR was 0.75 (95% Cl from the British Columbia Cancer Agency data to 0.25-2.23) using hospital controls and 1.32 (95% Cl matched controls selected among provincial voters. 0.30-6.51) using population controls. When the two Occupational and exposure histories were obtained, types of controls were combined, the OR was 0.87 whenever possible, directly from cases and controls. (95% Cl 0.43-1.70) for motor vehicle mechanics and The OR for `vehicle mechanics' considered as an a 0.89 (95% Cl 0.31-2.47) for persons definitely i priori suspect occupational group was 0.8 (95% Cl engaged in brake servicing. (These calculations use 0.2-2.3). The OR for the category `brake lining updated numbers provided by the authors. Prof. H.-J. installation or repair' was 0.3 (95% O 0.0-1.4). Woitowitz and Dr K. Rodelsperger, of Justus-Liebig After removing cases and controls with at-risk occu University, Germany.) pational asbestos exposures, the OR for 'vehicle The strengths of this study are its ability to examine mechanics' was 0.4 (95% Cl 0.0--3.2). The results did the association with brake repair, its high response not change after 20 years of latency. The authors also rate and the pathology review of cases. The most reported that, after removal of at-risk exposures, important shortcomings include the lack of adjust `brake installation and repair did not appear to be ment for age and an inadequate description of subject associated with mesothelioma'. selection. i i HWBUI0011490 316 M Goodman el al. Agudo et aL (2000). This hospital-based casecontrol study compared pathology-confirmed cases of mesothelioma to population/hospital controls selected using a two-step procedure (Agudo and Gonzalez, 1999). The authors reported that there were three cases and 14 controls in the category `mechanics, motor vehicle'. The non-exposed cate gory included 51 cases and 148 controls that had never worked in any of the at-risk occupations. Based on this information, the crude OR for `mechanics, motor vehicle' was 0.62 (95% Cl 0.11-2.36). The comparison of motor vehicle mechanics who may have had other potential asbestos exposure to persons without any exposure is problematical. However, the result indicates that substantial confounding by other exposures in this study population is unlikely. Because 44% of cases and less than 1% of controls had next of kin interviews, information bias needs to be considered. The study's high response rate, novel methods of control selection and confirmation of diagnosis by pathology review are among its strengths. Meta-analysis. The results of the meta-analysis for mesothelioma are presented in Table 4. All tests for heterogeneity produced non-significant results regardless-of stratification and in all analyses the random effects model reduced to a fixed effects model. The meta-RR estimates for Tier I and Tier II. ' studies were 0.92 (95% a 0.55-1.56) and 0.81 (95% Cl0.52-1.28), respectively. Analysis limited to studies evaluating the associ ation between mesothelioma and brake repair, as opposed to motor vehicle repair, resulted in a metaRR of 0.86 (95% Cl 0.48-1.56). Inclusion of only those studies that considered - other asbestos exposures resulted in a meta-RR of 0.80 (95% Cl 0.46-1.40). Lung cancer Overview of the literature. Twenty-nine studies were identified initially. Of these, seven were elimi nated because they did not meet the three initial inclusion criteria or they presented data that over lapped with other studies. The remaining 22 studies were evaluated and scored (Table 5). Thirteen studies were further excluded from the meta-analysis because they had an overall negative score. Nine studies remained, of which four (scores of 3 or 4) were included in Tier I (Table 6A) and five (scores 0-2) were included in Tier II (Table 6B). Tier I included two cohort studies (Gustavsson et aL, 1990; Hrubec el at, 1992) and two case-control studies (Lerchen et al., 1987; Benhamou et al, 1988). Tier II included two cohort studies (Jarvholm and Brisman, 1988; Hansen, 1989) and three case-control studies (Williams et al, 1977; Vineis et al, 1988; Morabia etal, 1992). Only six studies adequately controlled for smoking; of these, three were included in Tier I. Despite adequate control for smoking, three studies (Williams etal, 1977; Vineis etal, 1988; Morabia etal, 1992) were included in Tier II due to other limitations. The earliest of the studies included in the meta-analysis was published in 1977 (Williams etal., 1977) and the most recent were published in 1992 (Hrubec et al., 1992; Morabia et at., 1992). Notably, none of the Tier I or II studies was published in the last 10 yr. Three of the four cohort studies included in the meta-analysis were published in Europe: two in Sweden (Jarvholm and Brisman, 1988; Gustavsson et al, 1990) and one in Denmark (Hansen, 1989). The fourth cohort study was published in the USA (Hrubec et al, 1992, 1995) and presented in two reports, one evaluating cancer risk by occupation and one evaluating cancer risk by industry. All but one (Benhamou et al, 1988) of the case-control studies were conducted in the USA. Tier l Lerchen et al (1987). This case-control study compared lung cancer patients from the New Mexico tumor registry to matched controls selected either through random digit dialing or from the Health Care Financing Administration records. The information for roughly half of all cases and 2% of controls was available from the next of kin inter views. Smoking variables used for adjustment included smoking status (current, former or never), number of cigarettes/day and smoking duration. The non-exposed group included subjects never employed in the industry or occupation of interest For auto mechanics the adjusted OR was 0.9 (95% Cl 0-5-1.9). Table 4. Meta-analysis results for mesothelioma Analyses Her I studies Tier II studies Studies that eliminated other exposures* Studies that defined exposure as `brake work'b K Meta-RR 95% a 0-test 4 0.92 0.55-1.56 0.97 3 0.81 0.52-1.28 0.92 4 0.80 0.46-1.40 0.94 3 0.86 0.48-1.56 0.46 K, number of studies: Q-test, P-value of the test for heterogeneity: RS. an estimate of relative risk; Cl. confidence interval. Agudo era! (2000), McDonald and McDonald (1980), Teschke et al (1997) and Hessel el al (2004, in press). bWoitowitz ami Rbdebperger (1994), Teschke el al (1997) and Hessel et at (2004, in press). i HWBUI0011491 Mesothelioma and lung cancer among motor vehicle mechanics o o oo -- o o ooo O-- Q O N - 0--0 -- -- 7 oo oo oo -- oo-- -- o-- -- O -- -- OOf^> O OOO ---- Q---- OO OO OOO ---- o---- OO OO OOO -- *" OO OO OOO o-- ~ 7 oo oo oo o --I oo oo . o o OOO ---- O-- o-- oo O-- O-- OOO OOO o - o- ooo oo oo -- O f"3 o o o ooo o -- -- -- m o oo ooo -- -- -- oo oo oo oo OOO OOO o ---- oo oo --oo OOO o-- -- cn - 2 317 sl s sa s >% 1 O' HWBUI0011492 318 M. Goodman et aL Table 6. Summary of lung cancer studies and conesponding RR estimates included in the meta-analysis First author Year Design (4) Tier I studies Lerchen 1987 Case-control Benhamou 1988 Case-control Gustavsson 1990 Cohort Exposure definition Source of cases Comparison group RR estimate (95% CD Auto mechanics New Mexico Tumor Registry Randomly selected residential telephone numbers or Medicare roster (age 65+) 0.9 (0.5-I.9? Motor vehicle mechanics French hospitals French hospital patients 1.06 (0.73-1.54)* without tobacco-related disease Bus garage mechanics, Swedish Cancer Occupationally active 1.22 (0.71-1.96) servicemen or hostlers Registry population of greater Hrubec Nested casecontrol 1992 Cohort 1995 <B) Tier II studies Williams 1977 Case-control Jarvholm 1988 Cohort Vincis 1988 Case-control Hansen Moiabia 1989 Cohort 1992 Case-control General population of greater Stockholm Highest asbestos exposure categoiy Lowest asbestos exposure categoiy AutomobilemechanicsUS veterans* life All other occupations and repairmen insurance records (occupation) Automobile repair services and garages (industry). 1.15 (0.67-1.84) 1.20 (0.26-5.64? 1.1 (0.85-1.41?<* 0.9(0.64-1.23)" Car repair services Third National Other cancer cases Cancer Survey 0.85 (NR)" Car repair mechanic Swedish Cancer General Swedish Register population 1.27(1.03-1.56) Automobile brake workers Cancer registries. Other hospital patients, 1.2 (0.9-1.7? hospitals, or death death certificates, and certificates drivers license records Auto mechanics Death certificates Other skilled workers 1.01 (0.72-1.37) Mechanics and US hospitals automobile repairmen U.S. hospital patients 0.7 (NR? without tobacco-related Q, confidence interval; RR, estimate of relative risk. Adequately controlled for smoking. ''Values used in meta-analysis. 95% Cl recalculated, authors only reported 90% Cl. The strengths of this study are the populationbased design and adequate adjustment for smoking. Conscious of the disparity in next of kin interviews, the authors evaluated the number ofjobs reported for cases and controls by interview method. Surrogate interviews generally identified fewer jobs, but the difference between cases and controls was small. Benhamou etaL (1988). In this French casecontrol study lung cancer cases were matched to controls selected from hospital patients with diseases not related to tobacco exposure. All subjects had completed a questionnaire that included a full occu pational history. The non-exposed group consisted of people who had never been engaged in the particular occupation under study. Smoking variables included smoking status, age when started smoking, cigarettes per day and duration of smoking. Analyses for `motor vehicle mechanics' yielded an adjusted OR of 1.06 (95% Cl 0.73-1.54). The major strengths of this study are its very detailed adjustment for smoking and an apparently high response rate (Lubin et aL, 1984; Benhamou et aL, 1985). However, the use of hospital-based controls could be considered a limita tion. Gustavsson et ai. (1990). Gustavsson and col leagues followed a cohort of workers who were employed in one of five Stockholm bus garages between 1945 and 1970. Exposure to asbestos was estimated by industrial hygienists based on personal sampling results. Observed mortality for the period from 1952 to 1986 was compared with mortality of the general and the `occupationally active' popula tion of Stockholm. The SMRs for lung cancer were 1.22 (95% Cl 0.71-1.96) using occupationally active population rates and 1.15 (95% Cl 0.67-1.84) using general population rates. The lung cancer SMRs by asbestos exposure index (exposure intensity x duration in years) were 0.86 for index 0-20,1.97 for 20-40 and 1.18 for >40. Reason fo r collision N o control for smoking, fo llo w up 1 yr; total score <0 No control for smoking, follow up 5 yr; total score <0 95% C l NR NR RR estimate 1.28 1.46 Table 7. Summary o f lung cancer studies and corresponding RR estimates not included in the meta-analysis Mesothelioma and lung cancer among motor vehicle mechanics 319 *9 hi 'I. r. SR 2 tx VCu .1,,2 1a* 1 p ci as - .5 .5 3|| ill III S 8 asi 8 i I 2R= ~ - r- 7 4 SS Si *7 o *A oo ---- 1 i if A ee _ d P-. ---- o o -- ir- o o4 d 'Z. Z I Si E El ' I III 12 1 si M >0 f*S do -- 8 s 5 6 pOOgOSu Stys3OSOSOSOSOSO8OSOS KO &C3\ 0\ gCu II 9 II -S 8 1 I 1 1 1 Q S 5 il u ^ a 1I ! E ai ! : ; j 1 Design Cohort Cohort Year 1963 1976 First author Enterline Menck HWBUI0011494 320 M. Goodman et al The standardized incidence ratio for lung cancer compared with the general population rates was 1.61 (95% Cl 0.94-2.57). However, in this analysis two cases of mesothelioma and one case of alveolar cell cancer were counted as Tung cancers'. A nested case-control analysis using logistic regression reported the following RRs: index 0-20 = 1.0 (reference); 20-40 = 1.67 (95% Cl 0.50-5.60); 40-60 = 1.26 (95% Cl 0.32-5.00); >60 = 1.20 (95% Cl 0.26-5.64). The limitations of this study are the potential inclusion of workers not involved in motor vehicle repair and the lack of smoking information. However, the use of internal comparisons in the nested case-control analysis potentially attenuated the confounding effect of smoking. A particularly important feature that sets this study apart from other studies is its ability to conduct IH-based doseresponse analyses. Hmbec et al. (1992,1995). Hiubec and co-workers conducted a cohort study of 248 046 US veterans followed from 1954 through 1980. In addition to occupational history, the cohort members responded to questionnaires providing information on smoking habits. The response rate was 84%. The underlying cause of death was identified for 95% of the decedents. Cause-specific mortality by occupation was adjusted for smoking using information on smoking status and amount of smoking. For cancers of the respiratory system, the smoking-adjusted KR was 1.1 (90% Cl 0.89-1.36) in the occupational group `automobile mechanics and repairmen' and 0.9 (90% Cl 0.69-1.17) in the industry type `automobile repair services and garages'. This study's strengths included its large sample size, ability to control for smoking (unusual for a cohort study) and a long follow-up period. However, its weakness was the use of the category `respiratory cancer', which is less specific than Tung cancer'. Tier IL Williams et al. (1977). Using the data from the Third National Cancer Survey, Williams et al. conducted inter-cancer case-control analyses for various occupations and industries while controlling for age, sex, race, education, smoking, alcohol use and geographic location. Only 57% of the cases approached for interviews participated. The nonexposed category consisted of persons in any other known job. For the industry category `car repair serv ices' the lung cancer analysis showed an OR of 0.85 (confidence interval not reported). Although this study adequately controlled for tobacco, alcohol and socio-economic status in all analyses, its main weak ness was the poor response rate and the use of all other cancers as controls. Jarvholm and Brisman (1988). These authors used I960 Swedish census records to identify men employed as `mechanics' in the `car repair' industry. This information was linked to the Swedish Death Register (1961--1973) and the Swedish Cancer Registry (1961-1979). There were 39 deaths from lung cancer versus 23 expected, yielding an SMR of 1.70. Ninety-three lung cancers occurred among car mechanics from 1961 to 1979, while 73.0 were expected, resulting in a standardized incidence ratio (SIR) of 1.27 (95% Cl 1.03-1.56). These results are limited by the lack of adjustment for smoking and by the absence of asbestos-specific exposure infor mation. Vineis et al. (1988). Occupational data from five case-control studies from five US states were combined to determine the risk oflung cancer associ ated with different occupations. Cases were identi fied from cancer registries, hospitals or death certificates. Controls for one of the individual studies were population based and matched on vital status; the remaining studies used hospital controls or deceased controls from death certificates. The nonexposed group consisted of people without any history of exposure to established and suspected lung carcinogens. Odds ratios were calculated for over employment in the selected occupations and were adjusted for age, birth cohort and cigarette use. Ninety-eight cases and 90 controls were ever employed as 'automobile brake workers' (OR = 1.2, 95% Cl 0.9-1.7). Among the limitations of this study is the lack of a uniform job classification scheme. The comparison of automobile brake workers who may have had other potential asbestos exposure to persons without any exposure potentially biased the results. The combined response rate was <80%. The study's advantages include a relatively large sample size, adequate control for smoking and information' specific to brake repair workers. Hansen (1989). A cohort of 21 800 male `auto mechanics' and 52 000 male skilled workers identi fied from the 1970 Danish Census was followed for 10 yr. Deaths were identified through the Danish National Bureau of Statistics. The reference cohort of skilled workers included carpenters, electricians, instrument makers, dairymen, upholsterers and glaziers. There were 41 lung cancer deaths among the motor vehicle mechanics, compared to 40.7 deaths expected based on the reference rates (SMR = 1.01, 95% Cl 0.72-137). Limitations of the study included short follow-up, inclusion of workers with potential exposure to asbestos (e.g. carpenters and electricians) in the comparison group and lack of control for smoking. However, the latter limitation may have been partially offset by the use of a comparison group of manual workers. Mesothelioma and lung cancer among motor vehicle mechanics 321 Morabia et al. (1992). This hospital-based casecontrol study was conducted in nine US metropolitan areas. Cases were matched (age, race, hospital, year of interview and smoking) to two types of controls (cancer and non-cancer). Participants completed standardized, in-person questionnaires. Adjustment for smoking was performed using smoking status and amount of smoking categories. The reference group (non-exposed) consisted of people who were never employed in an occupation with exposure to confirmed or suspected lung carcinogens. For the usual occupation of 'mechanics and repairmenautomobile' the adjusted OR was 0.7 (confidence interval not reported). The main strength of this study is adjustment for smoking. The comparison of auto mobile mechanics who may have had other potential asbestos exposure to persons without any history of at-risk exposures is a potential source of bias. Use of hospital controls and the fact that the authors did not report a response rate should also he considered as weaknesses. Meta-analysis. The results of the summary ana lysis for lung cancer are presented in Table 8. All tests for heterogeneity produced non-significant results regardless of stratification, and in all analyses the random effects model reduced to a fixed effects model. Analyses of the Tier I and Tier U studies pro duced meta-KR estimates of 1.07 (95% Cl 0.88-1.31) and 1.17 (95% Cl 1.01-1.36), respectively. When the analysis was limited to studies that used adequate control for smoking regardless of tier, the resulting meta-RR estimate was 1.09 (95% Cl 0.92-1.28). The analysis that was limited to case-control studies (including the nested case-control results from Gustavsson et al., 1990) resulted in a meta-RR esti mate of 1.08 (95%CI 0.87-1.34), while the analysis of the cohort studies produced a meta-RR of 1.16 (95% Cl 1.00-1.34). DISCUSSION None of the individual studies that examined risk of mesothelioma among motor vehicle mechanics demonstrated increased risk and, thus, the results of the meta-analysis also showed no increased risk of mesothelioma. It is also noteworthy that the meta analysis results remained virtually the same after the analysis was limited to studies specifically mentioning brake mechanics and did not change after the removal of those with other occupations poten tially involving asbestos exposures. Although the three potentially relevant cohort studies (Jarvholm and Brisman, 1988; Hansen, 1989; Gustavsson et aL, 1990) could not calculate the RR, their findings largely support the case-control and PMR/PIR surveillance studies and are consistent with the results of the meta-analysis. There were four cases in a total of -600 000 person-yr of follow-up. For reference, the reported annual rates of mesothe lioma among males in Denmark and Sweden are 1.47 and 1.1 per 100 000, respectively (Ferguson et aL, 1987; Parkin etaL, 1997). In a Swedish study of bus garage mechanics Gustavsson et aL (1990) reported two cases of mesothelioma. One of these was not a mechanic but an electrician and both cases `may have been exposed to asbestos during previous employments'. In another Swedish study Jarvholm and Brisman (1988) found `one case of mesothelioma in the cancer register and none in the death register'. The authors noted that the person diagnosed as having mesothelioma also worked in the construction industry and concluded that their data `indicate no increased risk of mesothe lioma in car mechanics ...' (Jarvholm and Brisman, 1988) . The Danish study of garage mechanics also found only one death due to cancer of the pleura (Hansen, 1989) . The author interpreted the finding of a single case of pleural cancer as an indication that exposure to asbestos via brake repair `was not negligible'. Unlike Gustavsson et al. (1990) and Jarvholm and Brisman (1988), Hansen (1989) did not evaluate other sources of asbestos exposure for the individual who developed pleural cancer. The conclusion by Hansen is inconsistent with the results of another record linkage study based on the data from the Danish Cancer Registry and from the Supplementary Pension Fund and the Central Popula tion Registry (Olsen and Jensen, 1987). In this study Table 8. Meta-analysis results for lung cancer Analyses Her I studies Tier 0 studies Studies with adequate adjustment for smoking2 Case-control studies Cohort studies K Meta-RR 95% a Q-tesl 4 1.07 0.88-1.31 0.96 5 1.17 1.01-06 0.64 6 1.09 0.92-1.28 0.92 6 1.08 0.87-1.34 0.92 3 1.16 1.00-1.34 0.44 Ct confidence interval; K, number of studies; 2-test, P-value of the test for heterogeneity; RR, estimate of relative risk. Htubeceroi (1992), Letcben etaL (1987), Benhamou etaL (1988), Vineis et aL (1988), Williams et aL (1977) and Morabia et aL (1992). iI I HWBUI0011496 ! 322 M. Goodman et aL covering nearly the same period of time as the and mesothelioma (ERG, 2003). Secondly, asbestos Hansen (1989) study, there were no cases of fibers in brake pads are embedded in resin (Weir and mesothelioma (pleural or peritoneal) in the category Meraz, 2001) and therefore are less likely to become `garages' and in the category `repair of motor airborne. Thirdly, much of the chrysotile in brake vehicles and motorcycles'. Because cancer registries pads is transformed to forsterite during the braking typically report higher numbers of cases than are process (Anderson et a!., 1973). Forsterite does not recorded on death certificates, the discrepancy appear to have asbestos fiber properties and is not between one case of pleural cancer and no cases of considered carcinogenic in humans (Wong, 1992). mesothelioma could be explained by the fact that not More importandy, in his recent publication Langer all pleural cancers are mesotheliomas. Another demonstrated that chrysotile's biological activity potential explanation for the discrepancy between becomes virtually nil hundreds of degrees below the Hansen (1989) and Olsen and Jensen (1987) is the forsterite transformation temperature. Thus, difference in evaluating occupational exposure. complete transformation of the mineral is not Hansen relied on occupation on the day of the 1970 required to result in loss of activity (Langer, 2003). census, while Olsen and Jensen allocated cancer Further evidence in support of the epidemiological cases to the occupational group in which the person findings comes from a recent pathology study evalu was employed for the longest period of time and with ating asbestos content in lung tissue of -1500 at least 10 yr of latency. mesothelioma cases (Roggli et a!, 2002). Fifty-one It is important to note that automobile repair is a of those cases were employed as motor vehicle common occupation. For example, according to the mechanics, of whom >50% had other sources of 1984 NIOSH'report, -5 000 000 persons in the USA asbestos exposure. The authors reported that lung had a history of having been formally involved in burden analyses in this occupational group either automobile repair (Nicholson etal., 1984). Assuming reflected background levels or, when asbestos was a background rate of 1-3 cases/million/yr found above background levels, the fibers were (McDonald, 1985; Parkin eta!., 1997), as reported to commercial and non-commercial amphiboles (Roggli apply to persons without occupational asbestos et aL, 2002). These observations led the authors to exposure in North America, some cases of mesothe conclude that `brake dust is unlikely to cause lioma are expected to be found among motor vehicle mesothelioma'. mechanics, even in the absence of any increase in risk Unlike mesothelioma, lung cancer has several due to this occupation. clearly identifiable occupational and lifestyle causes The pitfalls of relying on case reports/case series that, when not taken into account, could explain some for causal inference were illustrated by two reports of the observed associations. Motor vehicle that appeared in the early 1990s. In 1991, Woitowitz mechanics are among the highest ranked occupa and Rddelsperger described cases of mesothelioma tional groups with respect to smoking prevalence in among their clinic patients who had worked at one the USA (Leigh, 1996; Bang and Kim, 2001). Thus, time as garage mechanics (Woitowitz and Rodel- adequate adjustment for smoking is particularly I sperger, 1991). These patients were part of an important in studies of lung cancer among motor ongoing case-control study. Based on their clinical vehicle mechanics because small to moderate unad experience, the authors concluded that there was an justed increases in risk can be attributable to the high increased incidence of mesothelioma among car prevalence of smoking habits in this occupational I mechanics. This conclusion was criticized by Wong, group. based on the absence of a comparison group (Wong, Another important source of confounding is the 1992). Woitowitz and Rodelsperger published the history of occupational exposures other than those results of their case-control study in 1994. The associated with motor vehicle repair. For example, in results of this study led its authors to conclude that the case-control study by Vineis et aL (1988), the employment as a garage mechanic, and specifically category `automobile brake repair workers' included as a brake repairman, was not associated with an all individuals who ever belonged to this occupa increased risk of mesothelioma (Woitowitz and tional group, regardless of other jobs held. In Rodelsperger, 1994). contrast, the unexposed group included only cases The lack of association between work as a motor and controls without any history of known or vehicle mechanic or brake mechanic and meso suspected at-risk occupations. Assuming that 50% of thelioma may result from several factors. First, brake workers in the Vineis study held other asbestos the asbestos fibers found in airborne samples are exposure-related jobs (e.g. insulators, pipefitters or generally short (<5 (im) chrysotile fibers welders) and that the average RR of lung cancer (Rodelsperger et of, 1986). Although still the focus among these occupations is 1.6 (Goodman et aL, of debate, there appears to be increasing consensus 1999), confounding by other asbestos exposures that short fibers, particularly those <5 pm in length, would result in an OR of 1.3. Because asbestos is are associated with little (if any) risk of lung cancer not the only occupational carcinogen, additional HWBUI0011497 Mesothelioma and lung cancer among motor vehicle mechanics 323 confounding by other occupational exposures may further affect the results. It is important to point out that most asbestosexposed occupations showing a substantia] increase in risk of asbestos-related lung cancer are expected to show an even higher elevated risk of mesothelioma. For this reason, it would be difficult to conclude that motor vehicle mechanics may be at risk for devel oping lung cancer from asbestos given the absence of any increase in risk of mesothelioma. Importantly, the only epidemiological study that used an IH-bascd exposure assessment (Guscavsson et a!., 1990), among persons employed at as bus garage, found no dose-response relationship between the level of asbestos exposure and the risk of lung cancer. The role of meta-analyses in observational epide miology is the focus of ongoing discussion. By combining several studies, meta-analyses have an inherent ability to show relatively small statistically significant departures from null. However, these rela tively precise meta-RR estimates may not accurately reflect the magnitude of the association unless the analyses take into consideration potential sources of systematic error. It is perhaps more useful to view meta-analysis as a formal way of understanding and quantitatively describing the level of consistency and inconsistency among studies and to identify potential sources of error that may affect the result One source of error that warrants consideration in a meta-analysis is publication bias, which tends to occur because studies with statistically significant positive findings are more likely to be published than studies with null results (Easterbrook et at, 1991). Publication bias can be evaluated through identifica tion of unpublished research or by calculating the fail-safe N, defined as the number of studies with a non-significant result that would bring a statistically significant meta-RR estimate to non-significant levels (Rosenthal. 1979). However, in this case the results are essentially null and thus publication bias would not be expected to affect our findings. The search for sources of error inevitably leads to evaluation of study quality. Several authors recom mend formalizing such evaluations by using quantita tive scoring of individual studies (Jenicek, 1989; Downs and Black, 1998). Others view the use of scoring schemes as somewhat arbitrary and advise against using them (Juni et aL, 1999). We would agree that both points of view have merit. Using a quality score as a method of weighting study results or as a variable in a regression model may introduce a subjective element into an analysis (Greenland, 1998). On the other hand, it is important to consider and lake into account methodological strengths and weaknesses that are likely to affect the results. For these reasons, we feel that the use of quality tiers is justified as long as the methodology of assigning studies to a particular tier is transparent to the readers. It would be difficult to ensure such transpar ency without some kind of formal scoring approach. The particular scoring method used in this study reflects the consensus of its authors, but we realize that other approaches may also exist. It is important to point out that if one were to compare the meta-analyses of all studies combined with those of individual tiers, the results would appear somewhat different The meta-RRs for mesothelioma based on all three tiers is 0.67, with a 95% a between 0.53 and 0.84, while the meta-RR for Tier 1 is 0.92 (95% Cl 0.55-1.56). The corres ponding results for lung cancer are 1.16 (95% Cl 1.13-1.19) and 1.07 (95% Cl 0.88-1.31). These comparisons indicate that the statistically significant departures from null (down for mesothelioma and up for lung cancer) in a meta-analysis of all studies combined could be explained by methodological problems of individual studies. CONCLUSIONS The available epidemiological data show that employment as a motor vehicle mechanic does not increase the risk of developing mesothelioma. Although some studies showed a small increase in risk of lung cancer among motor vehicle mechanics, the epidemiological data on balance do not support a conclusion that lung cancer in this occupational group is related to asbestos exposure from vehicle repair. Acknowledgements--This research was funded primarily by Ford Motor Co, Daimler-Chiysler Corp. and General Motors Corp. Some of the authors have testified as expert witnesses in litigation regarding the potential health effects associated with brake repair. APPENDIX Meta-analysis calculations Using the fixed effects assumption, the general formula for the weighted average effect size of k studies is: T-tsh- 2>. f ss ! (l) where Ti is the effect size estimate of the idi study and wi is the weight associated with it. The weights that minimize the variance of T. are given by: (2) V; 124 M. Goodman elal where vt is the variance in each study. The average Bureau of Census. (1970) 1970 Census of population: classifi effect size T. has a conditional variance v. given by: cation index of industries and occupations. Washington, DC: Department of Commerce. Bums P, Swanson GM. (1991) The occupational cancer inci dence surveillance study (OCISS): risk of lung cancer by usual occupation and industry in the Detroit metropolitan area. Am I Ind Med; 19: 655-71. Caistenscn J, Pershagen G, Eklund G. (1988) Smoking- Equations for variance change for the random effects assumption. The total variance of an effect size esti mate is given by: adjusted incidence of lung cancer among Swedish men in different occupations. Int J Epidemiol: 17; 753-8. CDC. (2001) Epi Info 2000, Version 1.1.2.2 November 2001. Centers for Disease Control and Prevention. Coggon D, Inskip H, Winter P, Pannett B. (1995) Differences /,*= a2 + v( (4) in occupational mortality from pleural cancer, peritoneal cancer, and asbestosis. Occup Environ Med; 52:775-7. where o2 is the random effects variance and v, is the De Stefani E. Kogevinas M, Boffetta P, Ronco A, Mendilaharsu M. (1996) Occupation and the risk of lung cancer in conditional variance given above. The random Uruguay. Scand J Work Environ Health; 22:346-52. effects variance a1 calculated is based on a weighted sample estimate Q of the unconditional variance of Ts. In this method, the random variance is estimated by: Decoufle P, Stanislawczyk K, Houten L, Brass DJ, Viadana E. (1977) A retrospective survey of cancer in relation to occu pation. US Department of Health, Education, and Welfare. Department of Labor. (1977) Dictionary of occupational titles. Washington, DC: US Department of Labor. 02 - re-(*-n) Downs SH, Black N. (1998) The feasibility of creating a check (5) list for the assessment of the methodological quality both of randomized and non-randomized studies of health care inter ventions. J Epidemiol Community Health; 52:377-84. Dubrow R, Wegman D. (1984) Occupational characteristics of cancer victims in Massachusetts 1971-1973. US Department of Health and Human Services. Enteriine PE. McKiever MF. (1963) Differential mortality L; = l from lung cancer by occupation. J Occup Med; 5; 283-90. Easterbrook PJ, Berlin JA, Gopaian R, Matthews DR. (1991) In the random effects model,- the average effect size 7. and its variance v. are calculated using equations 1--3 above, however, V* is substituted for vf. The Publication bias in clinical research. Lancet; 337(8746); 867-72. Eastern Research Group Inc. (2003) Report of the Expert Panel on beairh effects of asbestos and synthetic vitreous fibers: random effects model will give a non-zero estimate only when Q is greater than its expected value. Other wise, it is assumed to be zero, and the fixed effects model applies. influence of fiber length, 17 March 2003. Available at www. alsdr.edc.gov/HAOasbestospanelfasbestostocJmnl. EPA. (1986a) Asbestos; proposed mining and import restric tions and proposed manufacturing, importation, and process ing restrictions, 40 CFR Par' 763. Fed Reg; 51:3738. EPA. (1986b) Controlling brake dust to protect your health. What every auto mechanic should know. Washington, DC: REFERENCES US Environmental Protection Agency. EPA. (1986c) Guidance for preventing asbestos disease among Agudo A, Gonzalez C. (1999) Secondary matching: a method for selecting controls in case-control studies on environ mental risk factors, int J Epidemiol; 28: 1130-3. Agudo A, Gonzalez CA, Bleda MJ a al (2000) Occupation and risk of malignant pleural mesothelioma: a case-control study in Spain. Am l led Med; 37:139-68. Anderson AE, Dealer RL, McCune RC, Sprys TW. (1973) Asbestos emissions from brake dynamometer tests. Ford Motor Co. Bang KM, Km JH. (2001) Prevalence of cigarette smoking by occupation and industry in the United States. Am J Ind Med; 40:233-9. Benhamou S, Bcnhamou E, Tiimarche M, Flamaat R. (1985) Lung cancer and use of cigarettes: a French case-control study. J Natl Cancer Inst; 74:1169-75. Benhamou S, Benhamou E, Flamont R. (1988) Occupational auto mechanics. Washington, DC: US Environmental Protection Agency. Ferguson DA, Berry G, JeHhovsky T el al (1987) The Australian Mesothelioma Surveillance Program 1979-1985. Med J Aust; 147: 166-72. Fmkelstein MM. (1995) Occupational associations with lung cancer in two Ontario cities. Am J Ind Med; 27: 127-136. Goodman M, Morgan RW, Ray R. Malloy CD, Zhao K. (1999) Cancer in asbestos-exposed occupational cohorts: a meta analysis. Cancer Causes Control; 10:453-65. Greenland S. (1998) Meta-analysis. In Rothman KJ and Greenland S, editors. Modem epidemiology, 2nd Edo. Phila delphia, PA: Lippincott-Raven. Gustavason P, Plato N, Lidstrom E-B, Hogsiedt C. (1990) Lung cancer and exposure to diesel exhaust among bus garage workers. Scand J Work Environ Health; 16: 348-54. risk factors of lung cancer in a French case-control study. Br J Ind Med; 45:231-3. Blair A, Walrath J, Rogot E. (1985) Mortality patterns among U.S. veterans by occupation. I. Cancer. I Natl Cancer Inst; 75: J 039-47. Hansen ES. (1989) Mortality of auto mechanics. A ten-year follow-up. Scand J Work Environ Health; 15; 43-6. Hesse! PA, Tela MI. Goodman M, Laa E. (2004) Meso thelioma among brake mechanics: an expanded analysis of a case-control study. Risk Anal; in press. : Biestow NE, Day NE. (1987) Statistical methods in cancer Hodgson JT, Damton A. (2000) The quantitative risks of research. Vol. H. The design and analysis of cohort studies, mesothelioma and lung cancer in relation to asbestos expos IARC Scientific Publication no. 82. Lyon, France; IARC. ure. Ann Occup Hyg; 44: 565-601. HWBUI0011499 Mesothelioma and lung cancer among motor vehicle mechanics 325 Hodgson JT, Peto I, Jones JR, Matthews PE, (1997) Meso thelioma mortality in Britain: patterns by birth cohort and occupation. Ann Occup Hyg: 41(suppl 1): 129-33. Hrubec Z, Blair AE, Rogot E, Vaught J. (1992) Mortality risk by occupation among U.S. veterans of known smoking status. 1954-1980, NIH publication no. 92-3407. US Department of Health and Human Services. Hiubec Z, Blair AE, Vaught J. (1995) Mortality risk by indus try among U.S. veterans of known smoking status. 1954-- 19SO, NIH publication no. 95-2747. VS Department of Health and Human Services. Huncfcaick M. (1990) Brake mechanics, asbestos, and disease risk. Am J Forensic Med Pathol; 11: 236-40. JarvholmB, Brisman J. (1988) Asbestos associated tumours in car mechanics. BrJInd Med; 45: 645-6. Jenicek M. (1989) Meta-analysis in medicine. Where we are and where we want to go. J Clin Epidemiol; 42: 35-44. Juni P, Witschi A, Bloch R, Egger M. (1999) The hazards of scoring the quality of clinical trials for meta-analysis. J Am Med Assoc; 282: 1054-60. Langer AM. (2003) Reduction of the biological potential of chrysolite asbestos arising from conditions of service on brake pads. Regul Toxicol Pharmacol; 38:71-7. Leigh JP. (1996) Occupations, cigarette smoking, and lung cancer in the epidemiological follow-up to the NHANES I and the California Occupational Mortality Study. Bull N Y Acad Med; 73: 370-97. Lerchen ML. Wiggins CL, Samel JM. (1987) Lung cancer and occupation in New Mexico. J Nad Cancer Inst; 79: 639-45. Lorimer W, Rohl A, Millet A, Nicholson W, Selikoff I. (1976) Asbestos exposure of brake repair workers in the United States. Ml Sinai J Med: 43:207-18. Lubin JH, Blot WJ, Berrino F, Flamam, R, Gillis CR, Kunzc M, Schmahl D, Visco G. (1984) Patterns of lung cancer risk according to type of cigarette smoked.lot J Cancer; 33: 569-76. Malker HS, McLaughlin JK, Malker BKtt al. (1985) Occupa tional risks for pleural mesothelioma in Sweden, 1961-79. J Natl Cancer Inst; 74:61-6. McDonald JC. (1985) Health implications of environmental exposure to asbestos. Environ Health Perepect; 62: 319-28. McDonald AD, McDonald JC (1980) Malignant mesothelioma in North America. Cancer; 46: 1650-6. Menck HR, Henderson BE. (1976) Occupational differences in rates of lung cancer. J Occup Med; 18:797-801. Militant S, Ossiander E.. (2001) Occupational mortality in Washington State 1950-1999. Washington State Depart ments Health. Milne JCL, Sandler DP, Everson RB. Brown SM. (1983) Lung cancer and occupation in Alameda County: a death- certiiieate case-control study. Am J Ind Med; 4:565-75. Morabia A, Markowitz S, Garibaldi K, Wynder EL. (1992) Lung cancer and occupation: results of a multicentre casecontrol study. Br J Ind Med; 49:721--7. Nicholson WJ, Daunt SM, LorimerWV etal. (1984) Investiga tion of health hazards in brake lining repair and maintenance workers occupationally exposed to asbestos. Cincinnati, OH: National Institute for Occupational Safety and Health (NIOSH). Office of Management and Budget. (1972) Standard industrial classification manual. Washington, DC: OMB Statistical Policy Division. Olsen JH, Jensen OM. (1987) Occupation and risk of cancer in Denmark: an analysis of 93,810 cancer cases, 1970-1979. Scand J Work Environ Health; 13(suppl 1): 1-91. Parkin DM, Whelan SL, Ferlay J, Raymond L. Young J. (1997) Cancer incidence in five continents. 1ARC Scientific Publications no. 143. Lyon, France: IARC, p. 317. Petersen G, Milham S. (1980) Occupational mortality in the state of California 1959-61. NOISH publication no. 80-104. USDHEW. Pezzolto SM, Poletto L. (1999) Occupation and histopathology of lung cancer: a case-control study in Rosario, Argentina. Am J Ind Med; 36: 437-43. Rodelsperger K, Jaha H, Bruckel B, Manlce J, Paur R, Woitowitz HJ. (1986) Asbestos dust exposure during brake repair. Am J Ind Med; 10; 63-72. Roggli VL, Sharma A, Butnor KJ. Spom T, Vollmer RT(2002) Malignant mesothelioma and occupational exposure to asbestos: a clinicopathological correlation of 1445 cases. Ultrastrue! Pathol: 26: 55-65. Rosenthal R. (1979) The `File Drawer Problem' and tolerance for null results. Psychol Bull; 86; 638-41. Rushton L, Alderson MR, Nagarajah CR. (1983) Epidemiolog ical survey of maintenance workers in London Transport Executive bus garages and Chiswick Works. Br J Ind Med; 40: 340-5. Schiffman MH, PickleLW, Fontham Eel al (1988) Case-con trol study of diet and mesothelioma in Louisiana. Cancer Res; 48: 2911-5. Schneider AS, Smith C. (2000) Nation's mechanics at risk from asbestos: deadly fibers are found in brakes, but officials have kept silent. Seattle, WA: Seattle PostIntelligencer. Available at http://seattlepi.nwsource.com/ uncivilaction/brksl6.shtnil. Schoenberg JB. Slemhagen A, Mason TJ, Patterson I, Bill J, Altman R. (1987) Occupation and lung cancer risk among New Jersey white males. J Nad Cancer Inst; 79: 13-21. Schwartz E. (1987) Proportionate mortality ratio analysis of automobile mechanics and gasoline service station workers in New Hampshire. Am Had Med; 12:91-9. Spirtas R, Keehn R, Wright W, Stark A, Beebe G, Dickson E. (1985) Mesothelioma risk related to occupational or other asbestos exposure: preliminary results from a case-control study. Am J Epidemiol; 122; 518. Spirtas R, Heaneman EF, Bernstein L el al. (1994) Malignant mesothelioma: attributable risk of asbestos exposure. Occup Environ Med; 51: 804-11. Swanson GM. Lin C, Bums PB. (1993) Diversity in the associ ation between occupation and lung cancer among black and white men. Cancer Epidemiol Biomarkers Prev; 2:313-20. Sutton A, Abrams K. Jones D. (1998) Systematic reviews of trials and other studies. Health Technol Assess; 2: 1-276. Teschke K, Morgan MS, Chcckoway H, Franklin G, Spinelli JJ, van Belle G, Weiss NS. (1997) MesotheliomasurveiUance to locate sources of exposure to asbestos. Can J Public Health; 88:163-8. Teta MJ, Lewinsohn HC, Meigs JW, Vidone RA, Mowad LZ Flannery JT. (1983) Mesothelioma in Connecticut, 19551977. Occupational and geographic associations. J Occup Med; 25:749-56. Truby M. (2002) False sense of security puts mechanics at risk. Detroit, IL: The Detroit News. Available at http:// www.detnews.com/specialreport/2002/asbestos/a2013453346.htm. Vineis P, Thomas T, Hayes RBei aL (1988) Proportion of lung cancers in males, due to occupation, in different areas of the USA. Ini J Cancer, 42; 851-6. Weir FW, Metaz LB. (2001) Morphological characteristics of asbestos fibers released during grinding and drilling of fric tion products. Appl Occup Environ Hyg; 16: 1147-9. Williams RR, Stegens NL, Goldsmith JR. (1977) Associations of cancer site and type with occupation and industry from the Third National Cancer Survey Interview. J Natl Cancer Inst; 59:1147-85. Woitowitz HJ, Rodelsperger K. (1991) Chrysotile asbestos and mesothelioma. Am J Ind Med; 19: 551-3. Woitowitz HJ, Rodelsperger K. (1994) Mesothelioma among car mechanics? Arm Occup Hyg; 38:635-8. Wong O. (1992) Chrysotile asbestos, mesothelioma, and garage mechanics. Am J Ind Med; 21:449-55. HWBUI0011500 326 M. Goodman etaL Wong O. (1993) Considerations of specificity in assessing the relationship between asbestos and cancer. Am J lad Med; 23: 521-3. Wong O. (2001) Malignant mesothelioma and asbestos expos* tire among auto mechanics: appraisal of scientific evidence. Regul Toxicol Pharmacol; 34: 170-7. World Trade Organization. (2000) European Communities-- measures affecting asbestos and asbestos-containing prod ucts. Report of the panel. Geneva, Switzerland: WTO. Zahm SH, Browason RC, Chang JC, Davis JR, (1989) Study of lung cancer histologic types, occupation, and smoking in Missouri. Am J led Med: 15:565-78. j \ i t ; i HWBUI0011501