Document p08dwBXo21bwdOjreBQm7Ymd

08/06/02 12:59 FAX 310 208 5971 INFO 2 XI002 American JoUrnai of Epidemiology .. I AW? copyright C 1988 byTho Johns IJopVins university School of Hygiano and Public Haanh ah rights reserved Vol. 148, No. 2 Prinred In U.S.A. ORIGINAL CONTRIBUTIONS Pleural Mesothelioma: Dose-Response Relation at Low Levels of Asbestos Exposure i'1 a French Population-based Case-Control Study Y. Iwatsubo," J. C. Pairon,'-3 C. Boutin,' 0. Menard,5 N. Massin," Q. Caillaud.7 E Orlowski,'-2 F. Galateau-Salle.e J. Bignon,',3 and P. Brocharda A hospital-based case-control study of the association between past occupational exposure to asbestos and pleural mesotheliomg was carried out in live regions of France. Between 1987 and 1993,405 cases and 387 controls were interviewed. The job histories of these subjects were evaluated by a group of experts for exposure to asbestos fibers according to probability, intensity, and frequency. A cumulative exposure index was calculated as the product of these three parameters and the duration of the exposed job, summed over the entire working life. Among men, the odds ratio increased with the probability of exposure and was 1.2 (959'x, confidence interval (Cl) 0.8--1.9) for possible exposure and 3.6 (95 (fa Cl 2.4-5.3) for definite exposure, a dose-response relation was observed with the cumulative exposure index: The odds ratio Increased from 1.2 (95% Cl 0.8-1.8) for the lowest exposure category to 8.7 (95% Cl 4.1-18.5) for the highest. Among women, the odds ratio for possible or definite exposure was is.s (95/a C14.1-86.2). We found a clear dose-response relation between cumulative asbestos exposure and pleural mesotheliomg In a population-based case-contra( study with retrospective assessment of exposure. A significant excess of mesgtheliomg was observed for levels of cumulative exposure that were probably far below the limits adopted in most industrial countries during the 1980s. Am J Epidemiol 1998;148:133-42. asbestos; case-control studies; mesotheliomg; occupational exposure Mesotheioma is a rare cancer that is mainly due to occupational or nonoccupational asbestos exposure. The background level is assumed to be as low as 1-2 per million inhabitants (1). During recent decades, however, its prevalence has been increasing in the general populations of most industrialized countries (2.3). Received for publication. October 30,1996, and accepted for publication April 7,1997. Abbreviations: CFI, Cumulative exposure index; Cl, confidence inir rv4; I SCO, international Standard Classification of Oecupatons (1968 edition); iSiC, International Standard Industrial Classification of Ali Economic Activities: OR, odds ratio. 1 INSERM Unit4139, EA2345, Crttell, France. 2 insiitut interuniversitaire cie ivlddecine du Travail de Paris lla de France, Paris, France. 3 CHI de Creteil, Cr6teil, Franca. HGpitW de la Conception, Marseille, France. s HBOtal de Brabofs, Vandoauvre, France. e Institut National de Recherche et de S4curlt6, Vandoauvre, France. 7 HSOtal Sabourin. Clomiont Ferrand, France s Colfe national franrais d'anatomopathologistas sp6cia15tes du mesotheliome (group MESQPATH), Caw, France, e Universitrs Bordeaux II, Bordeaux, France. Reprint requests to prY. Iwalsubo, INSERM UnitA 139-EA2346, f M3, Facutte de m6decine, 8 rue du General Sarraii, 94010 Creteil Codex. FmcCL In the cohorts of workers occupationally exposed to asbestos that have been followed since the 1960s, the risk of mesotheliomg has increased with the level or duration of exposure or both (4-14). The absence o accurate measurements for low exposure levels limits the reliability of any current quantitative assessments of the risk they carry. Furthermore, since only a few subjects in these cohorts were exposed to low levels of asbestos, there is not enough statistical power to show any significant association with mesotheliomg. Case-control studies among the general population and its variety of occupational categories exposed to different asbestos levels are more likely to include subjects whose exposure was low. Despite recent developments in.retrospective assessment of exposure (15, 16), the quanti tative assessment of low levels remains difficult, since measurements of dust cclncentration during the relevant periods are not often available. Previous studies of mesothelioma (4-14) have ex amined exposure parameters, including cumulative ex posure and such time-related variables as time since or age at first exposure. Other exposure parameters, in particular, the time-related pattern of exposure, might be useful. Although the current asbestos exposure pro- 133 PAGE W11-RCVD AT 816120023:54:34 PM (EastTM Daylight Timey SVR:MGHTFAX,CLT15, DNiS;13071 C810:310 208 5971 i DURATION Ws):06.22 09/00/02 12:59 FAX 310 208 5971 INFO 2 003 134 Iwatsubo et si. file involves mostly intermittent exposure, the data now available do not allow any conclusion about whether asbestos inhalation at intermittent peaks con tributes to the risk of mesothelioma- The aims of this study were to examine the doseresponse relation by using several types of exposure parameters and to study the role of time-related expo sure patterns (intermittent compared with continuous) in a large case-control study conducted in France since 1987. MATERIALS AND METHODS This report is based on data collected is a hospitalbased case-control study of pleural malignant me sothelioma (hereafter referred to as mesothelioma). This study is ongoing, and the present analysis is limited to data collected between January 1, 1987, and December 31,1993. Five administrative regions of France are currently participating- The study began in the Paris metropolitan area in 1987 and was extended in 1989 to the region ofProvence-Apes-C8te-d'Azur and to Corsica and, in 1992, to Lorraine and Au vergne. The respiratory disease, chest surgery, and oncology, departments of all public hospitals and the main private clinics were informed o the study and invited to participate. Mesothelioma patients in this study met the follow ing criteria-1) consultation, at any stage of the disease, in a participating hospital; 2) histologically confirmed diagnosis; 3) resident in a participating region at di agnosis; and 4) alive at the time of interview. The diagnosis of mesothetioma was confirmed by the French Mesothelioma Panel (1.7, 18). The panel excluded 46 (10 percent) of the subjects initially con sidered eligible (for whom, after pathology review, the principal diagnosis was adenocarcinoma), tu 125 sub jects (31 percent of the remaining 405 cases), how ever, the panel could not reach a conclusion because the histologic sample was insufficient or because the slides had not been sent to the panel. The likelihood of diagnosis was then determined by reviewing clinical data (clinical history, radi.ologic data), laboratory test reports, and the histologic conclusions of the local pathologists. Hospital controls were individually matched for sex, age (5 years), place of residence (administrative department), and racial or ethnic origin (black, white, North African, Asian, or other) and were selected in the departments of internal medicine, oph thalmology, and surgery. Patients with a medical his tory of maiignaait tumors or asbestos-related diseases (i.e., asbestosis and lung cancer) were excluded as controls. To the extent possible, controls were chosen in the same hospital as their matching classes. Data collection An experienced interviewer questioned patients dur ing their hospitalization. In a few cases, the subject was interviewed at home. A standardized question naire was used to collect information on work history: work periods, including the starting and ending dates of each job that lasted at least 6 months; the compa ny's economic branch of activity; and a description of the tasks performed by the subject. This information allowed us to classify the subject's job according to the International Standard Classification of Occupa tions (TSCO) code for occupations (19) and the Inter national Standard Industrial Classification of All Eco nomic Activities (ISIC) code (20) for industrial activities. For each job pedod, the subjects were asked five specific questions about direct (handled) and in direct (working in the immediate vicinity ofcolleagues who handled) asbestos exposure. Exposure assessment A panel of five experts in industrial hygiene evalu ated occupational exposure to asbestos, as follows: 1) all job periods of all subjects (cases and controls) were sorted by economic branch of activity (ISIC codes) and occupation (ISCO codes); 2) the job periods were selected for review according to the likelihood of exposure of the job titles, classified by ISZC and ISCO codes; 3) the job periods for which subjects reported exposure were selected; and 4) occupational exposure to asbestos was evaluated for all job periods selected in either step 2 or step 3, in sequential order of both the ISIC and ISCO codes. Eachjob period for each subject was thus evaluated independently. This procedure was chosen to minimize errors in the exposure assessment due to knowledge of the subjects' lifetime exposure. The experts were blinded to the case-control status of each job period, and decisions were made by con sensus. The experts had access to all information from the questionnaire, such as job history, tasks performed, and self-report of. direct or indirect exposure to asbestos. This evaluation of eachjob allowed eachjob period to be classified according to the probability, intensity, and frequency of exposure. Categories ofintensity and frequency were established by die experts before the evaluation began by using the following semiquantitative scale: probability of exposure: not exposed, pos sible, definite; frequency, sporadic (less than 5 percent of work time); irregular (5-50 percent of work time); continuous (more than 50 percent of work time); in tensity: low (less than I fiber/ml); medium (1-2 fibers/ ml); high (2-10 fibers/ml); very high: (>10 fibers/ml). We attributed weighting factors to each exposure category to calculate an exposure index; probability: Am J EpidenVol VOL 148, No. 2.1998 PAGE 311,rRCVD AT 8(6120(12 3:64:34 PM PastemDaylight Time" SVR:RIGHTFAULTI? DNI81307x C81D;310 208 5971=DURATION (mm-ss):06-22 08/00/02 13:00 FAX 310208 5971 INFO 2 Eih 004 Pleural Mesothelioma and Asbestos Exposure 135 null -- 0, possible = 0.5, definite - 1', frequency: sporadic - 0.025, irregular = 0.25; continuous 0.75; intensity: low = 0.1 fiber/ml, medium = 1 fiber/ml, high = 10 fiben/ml, very high= 100 fibers/ml. Because the latency period of the disease is so long, we did not analyze asbestos exposure during the 20 years before the mesothelioma diagnosis (1,21,22). We used the following exposure parameters for each subject. Highest probability. Intensity, and frequency. Each subject's highest probability of exposure was deter mined by the highest probability of any job period during lifetime work history. Highest intensity and frequency were determined the same way. Duration of exposed jobs. Duration of exposed jobs (years) is defined as the total duration of job periods involving possible or definite exposure. Cumulative axposure index (CEI). CEI is the life time sum of the products of probability, frequency, intensity, and duration for each job period. Because no measurements of airborne asbestos levels were avail able, all estimations of exposure parameters were based on die experts' subjectivity, that is, setniquantification, to which we subsequently assigned weight ing factors. This index of cumulative exposure was expressed in terms of fibers/ml-years inside quotation marks ("f/m.l-years"). Pattern of exposure in time. We examined the rel ative risks associated with the pattern of exposure by distinguishing subjects who had undergone only intermitterlt exposure from those whose exposure was con sidered continuous. Subjects' exposure was classified as intermittent if it was sporadic or irregular and if they had never worked at a job with continuous expo sure. The continuous category was reserved for sub jects who had been employed in at least one job with continuous exposure. In addition to these composite variables, age at first exposure and time since first exposure were also ex amined. Statistical analysis We calculated the odds ratio by using logistic re gression and the unconditional maximum likelihood method, with the aid of )3MDf software (23), This technique allowed us to include the cases who had no controls. The analysis took the matching variables into account. The relation between asbestos exposure and mesothelioma was examined separately for men and women. Quantitative parameters were categorized by per centile points. To allow us to consider the effect of some previously used cutoff points, we used additional categories for studying cumulative exposure (5 and 10 ff^T'd-years,,}. The effect of the time-related exposure pattern (that is, intermittent vs. continuous) was analyzed after tak ing into account cumulative exposure. RESULTS The study included 405 cases and 387 controls (ta ble 1). The largest group of cases (69.9 percent) came from the Paris metropolitan area. Cases and controls did not differ significantly by sex (82 percent and 81 percent men, respectively) or age at interview (63.5 and 63.9 years, respectively). Since almost the entire sample was white (96.8 percent of cases and 97.7 percent of controls), we did not adjust for race or ethnic origin. The socioeconomic category ofthe sub ject was determined by the last occupation held before the interview and coded using the major groups of the ISCO (table 2). Cases and controls differed signifi- with more blue-collar workers among the cases. Thus, for all comparisons, the odds ratios were adjusted for socioeconomic category. 'fable 3 presents the main occupations and indus tries that entailed asbestos exposure among the 3,498 job periods for men- We consider in this table only activities and professions that contained at least 50 job periods and for which at least 25 percent of the job TABLE 1. Main characteristics of cases and controls by study area. French Masothefioma Case-Control study, 1987_1993 8(uuy area and yeWS a study Paris rnwtreporltan area (14187-1993) Provenoo-Alpes-Cite-'Azur (1989-1993) CDriiei <1909-1993) LOrrzino (1992m.1993) Auvargn.* <1952- .1993) Mo, Pas 82 B 20 4 Of iulaa 78 92 75 B9 00 Cuai WAn (BE) 62.9 64.5 S7 .5 64. B Bfl . 0 (10.8) (8.9) (7.0) U.33) (4.1) Ranw L5-00 44-5 52 -B6 63-73 CeurlroE Wa, Q Mean' <sD) Range 279 7a $s_4 (11.2) 29-W 73 39 652 (94) 43-$4 7 96 64 .7 f4 - 9) 56--71 35 39 65 . B (10.7) 47-07 3 7 5 66.0 (4,0) B2 70 Total - SD, standard deviation. 105 152 63.5 110.5) 25-68 307 B1 63.5 (10.7) 2.9-99 ii-I i-eo nr_ 0 1Q0A PAGE Oil RCVDAT UQ4023:54:34 PM [Eastern DothtTime)z SVR:MGHTFAX-CLT1510NIS:1307tCSID:310206 > DURATION(mwss):06-22 08/06/0 2 13:00 FAX 310 208 5971 INFO 2 005 136 Iwatsubc ct al. TABLE 2 Distribution of mcsntIKelloma Gazes and controls according to aocloeaonormio category,* French Mesothelioma Cast study, iaaZsl99ji________ ISCO CAM (maw groups} Men Cases cannrs No. % Na_ ' % womfia Las" COMrrols No, Vu No. % Professional, technical, and related worfters Ml 14.3 14.5 7 9.6 15 20.3 Administrative and managerial workers Clerical and related workers Sales workers Service workers 2 22 6.7 17 5.s 1 1.4 0 O 3 35 10.0 40 12.9 28 38.4 25 33.7 4 SO 9.1 24 7.7 5 6.9 3 4.1 5 14 4.3 32 10.B 13 17.8 20 27.0 Agricultural, animal husbandry, and forestry workers; Fishermen, and hunters 6 2 0.5 8 2,6 0 3 4.1 Production and related workers, transport equipment operators, and laborers 71814 179 54.4 145 19 26.0 s 10.8 No occupational activity I 12 Socioeconomic category corresponding to the international Standard Classification of Occupations (ISCO) code of the last job field by the subject before interview. periods were evaluated as possibly or definitely ex posed. In the industries and occupations in which we had anticipated asbestos exposure, the proportions of exposure were high. For example, exposure was likely to have occurred in 264 of the 487 (54 percent) men's job periods in the construction industry and in 55 of the 70 (79 percent) men's job periods in the shipbuild ing industry. In some occupations, exposure was fre quent, e.g., 82 percent among motor vehicle mechan ics and 85 percent among plumbers and pipe fitters. The proportion of exposed job periods in the catego ries of other industrial activities and occupations was low (16 and 19 percent, respectively). Table 4 indicates the distribution ofjob periods of male cases and controls according to starting date and exposure intensity for possibly and definitely exposed job periods. Very few job periods were considered as very highly exposed, and those were found mainly among cases (18 job periods in cases vs. four among controls). These were observed after 1950 when the industrial use of asbestos had developed. Table 5 reports the distribution of male cases and controls according to various exposure parameters. The exposure measures in this table have not been adjusted for the other exposure parameters. Mesothefoma risk increased with exposure probability, inten sity: and frequency. The odds ratio for possible expo sure was 1.2 (not significant), and for definite exposure, it was 3.6. Risk increased with frequency of exposure, but subjects with sporadic exposure were not at greater risk of mesothelioma than were controls_ Risk also increased with the total duration of exposed jobs: The odds ratio for subjects exposed for at least 20 years was 5,4. The odds ratio for the relation between pleural me sothelioma and asbestos exposure parameters did not increase with time since first exposure, nor was any consistent trend observed with age at first exposure. As determined by the experts' evaluations and the weighting factors, the cumulative exposure of our pop ulation was rather low. Twenty-three percent of the cases and 35 percent of the controls had been exposed to less than 0.5 "f/ml-years." A gradient was observed with the CEI; the odds ratio rose from 1.2 for the subjects with less than 0.5 "f/ml-years" to 8.7 for the category with more than 10 "fhnl-years." Among women, a significant risk of mesothelioma was observed among those possibly and definitely exposed to asbestos: considered together (odds ratio (OR) = 18.8, 95 percent confidence interval (Cl) 4.1-86.2). Because of the small number of women exposed to asbestos, especially among controls (25 cases and two controls, for 33 and 3 percent of their respective categories), we did not analyze the doseresponse relation among women any farther. The results about the tinge-related pattern of expo sure reveal a significantly elevated odds ratio among workers whose exposure to asbestos was intermittent (OR = 1.8, 95 percent Cr 1.3-2.6). The oddsratio was much greater, however, for continuous exposure (OR = 5.7, 95 percent Cl 3.4-9.7). The median CEI within each category considered, i.e., <0.5,0.5-0-99, 1-9.99, and :f-10 "f/ml-years;1 was similar atnong intermittent and continuous exposure cases, except in the highest class of CEI (? 10 "frill-years") (0.1, 0.65, 3.5, and 38.7 "foal-years" for the intermittmt expo sure groups and 0.075, 0.65,3.1, and71.3 "Td-years" for the continuous groups, respectively). We attempted PAGE 5111x RCVDAT 81601002 3;54,34 PM (Eastern DaylOtTime)1 SYR,RIGHTFAX.CLT15 Am j t;oo6Mv Vol- 148, W 2,1998 DNIS;1307 f CSID.310208 59111 DURA10R On*ojl622 08/00/02 13. 01 FAX 310 208 5971 INFO 2 000 Pleural Mesotheiiorna and Asbestos Exposure 137 TABLE 9. Setectod principal industrial activities and occupations entailing asbestos exposure among men Frennh Mesothelioma r`.s',nnH-nl Study. tail;7-1993 Ua. Proportion olexposed TdlR Iql, penods job perlodst t Industrial activities (4-digit !S!C$ co(:e) 5000 Construction 3843 Manufacture of motor vehicles 7111 Railway transport 3341 Shipbuilding end repairing . 3813 Manufacture of stNCtural metal products 3511 Manufacture of basic industrial chemicals, except fertilizers 9513 Repair of motor vehicles and motorcycles 3823 Manufacture of metal and wood working machinery 3829 Manufacture of machinery and equipment, except electrical not elsewhere das:.ified 3945 Manufacture of aircraft 3710 Iron and Matt basic industries Otherindustries and industries not spcdfied (n .., , 487 113 76 7o 65 52 62 58 54 51 51 2,359 54 27 30 79 49 64 71 26 3o ll SI 18 Occupations (3-dgit ISCO$ Code) 9-99 Laborers not elsewhere classiflod 3-41 Machinery fitters and machine assemblers 8-55 Electrical wiremen e-49 Machinery fitters, machine assemblers, and precision instrument makers (except electrical) not eisewhero dassitied 5-73 Sheet-metal workers 9-64 Carpenters, joiners, and parquetry workers 8-71 Plumbers and pipe fitters 9-51 Bricldayem, stonemasons, and the setters 8-43 Motorvetdde mechanics 3-91 Stock clerks Other professions 152 110 107 99 85 75 73 69 67 50 %603 26 38 54 57 49 37 85 58 as 28 19 Total job periods 3,498 27 * in these tetbies, aniy activities and professions that contained at lacist 50 job periods and for which at least 25 percent of the job periods were evaluated as possibly or definitely exposed were considered, f, Possible or definite exposure to asbestos without taking into account the 20-year latency period, t IS10, Intemationel Standard industrial Classification of all Economic Activities: |SCO, international Standard Classification of Occupations. to separate the possible effect of the exposure delivery pattern from that of cumulative exposure by a strati fied analysis. The odds ratios increased with the CEI among subjects with intermittent and with continuous exposures (table 5). The amplitude of the odds ratio differed, however, between these categories. When we examined the odds ratios for subjects within each of our CEI categories, they were almost twice as high for subjects with continuous exposure as for those inter mittently exposed, except for the CEI category of 0.5-1 "flmi-years." DISCUSSION This study, one of the larger population-based casecontrol studies published (24-37) sheds light on sev eral important aspects ofmesothelioma and asbestos. As far as we know, our study is the first conducted in a general population that uses a semiquantitative assessment of exposure to examine the dose-response relation between asbestos exposure and mesothelioma. The mesothelioma cases of this study were identi fied in hospitals that had agreed to participate in the case-control survey. Cases seen in other hospitals and those who were not followed within a hospital struc ture were not included, 'there is no reaSOxl to suppose, however, that the type ofhealth care facility depended on the level of asbestos exposure. We ought to point out another source of selection bias. Mesothelioma diagnosis remains difficult. The patient who has a known history of asbestos exposure is more likely to be diagnosed with mesothelioma ihazt a patient with similar symptoms but no known history of asbestos exposure. This bias could have heightened the doseresponse relation between asbestos exposure and me sothelioma. There are probably few cases erroneously diagnosed as mesothelioxna, since the French Me- RAGE 6111 RCVD AT 80023:54^1 (EastemDaylightTjme)"sVR:RIGHTFAUm5 r DNISH3011 CS1D:31Q 208 5971 DURATION (mm-3s):0622 08/00/02 13: 0 1 FAX 310 208 5971 INFO 2 007 138 Iwatsubu et m. TABLE 4. Distribution of job periods among men, awarding to the intensity of exposure* and decade of beginning, French Masnthftlintna Case-Control Studv.1987-1Q3 Prah.UTHv of oipos.ro Eelore 1H0 Ktl. % 1930-1939 NOi % 1940.1549 No. % Distributionjib periods 1950-1559 .1960-1369 NO, % N4. % 1970 and after No. % Total Ho. % cases Not exposed LOW Med turn Very hjgh S3 10 12 2 0 77.6 9.4 11.2 1-9 172 25 25 6 0 75.1 109 11.4 26 392 65 74 23 2 66,9 113.1 14.9 4.6 0.4 7A0 55 64 40 7 62.8 12.3 14.4 9.0 1.6 202 54 SO 20 4 61.2 16.4 15.2 6.J 12 216 38 29 21 5 $0.9 123 9.4 Q.A 1.0 1284 247 255 112 18 $7.9 12.9 13.3 5.9 0.9 total 107 100 220 100 496 100 446 100 330 100 soo 100 11,9161 100 controls Mot pxPosoo LAW Medium High Very high 65 2 7 4 ) 02.3 25 0.9 6.1 1.2 192 21 13 2 0 84.2 92 5,7 09 341 46 31 4 1 90.6 10.9 7.3 1.0 0.2 282 33 27 G 1 sole 95 7.7 1.7 0.3 218 29 26 1 1 79.3 10.6 9.5 0.4 0.4 183 17 Is 3 0 83.9 7.0 6.9 1.4 1,281 148 119 20 4 81.5 9A 7.6 1.3 0.3 Total 79 100 220 100 423 100 340 100 275 100 21S 100 1,572$ 100 * IrdensilyForposlbfa ordetkvlo Bxfxreure to asbartos without tafdng Into account the 20-year latOHey pefiod, t job perbds for wfll^n the year of begtnVrsg Is mi^ng - g. $job Perlodsfor wtih the yoar of beglming is missing = 4. sothelioma Panel excluded, after pathology review, 10 percent of subjects initially considered eligible and confirmed the diagnosis for 62 percent of the cases on the basis of pathology reports. We accepted the re maining 28 percent after reviewing available histo logic data and hospital records. The use of hospital controls could have entailed some bias. In particular, cases and controls differed in socioeconomic status, with the latter group containing fewer blue-collar workers. This difference could have arisen from a selection bias and might reflect the controls' failure to represent adequately the population from which the cases were drawn, or it might be due to a particularly high rate of pleural mesothelioma among blue-collar workers because of their high prev alence of asbestos exposure. In the latter case, taking socioeconomic status into account could have led to overadjustment of the relation between asbestos expo sure and pleural mesothelioma. The crude odds ratios, however, were of same order of magnitude as the adjusted values. The validity of the information about asbestos ex posure depends on how well we have avoided three types of errors: difference in the quality of interview data according to disease status (recall bias or inter viewer bias), errors by the experts in classifying the subjects into defined categories, and errors related to the accuracy of the weighting factors subsequently assigned to each category. As recently stated in an International Agency for Research on Cancer meeting on retrospective assessment of occupational exposure in epidemiology (38), the validity of expert judgment, which relies on both the l'-nowledge and the experience of industrial hygienists, has rarely been evaluated. Indeed, when no objective method of measuring ex posure is available, their judgment is most often con sidered the gold standard. Our study assessed frequency and intensity of ex posure by using ordinal categories with specific boundaries. This procedure should have minimized the misclassification of subjects between extreme expo sure categories. The experts themselves, however, re ported sometimes encountering difficulties in distin guishing between sporadic and irregular exposure and between low and moderate exposure. Moreover, they suggested that the quality of their assessment for the periods under consideration (20 or more years ago) might not be as good as for more recent years because of the lack of published data for these periods. These errors could have led to the nondifferentiai misclassification o subjects into expo sure categories and the possible underestimation of the odds ratios (39). To avoid the exposure suspicion bias, the experts were blinded to case-control status when they evalu ated exposure. Recall bias could have iuliluenced the duality of the answer to the questionnaire and, subse quently, the expert judgment. To test this potential bias, we compared the experts' assessment with results from an asbestos job exposure matrix (dd). We found no difference between cases and controls (data not shown), suggesting that it was unhlcely that a substan tial recall bias had affected the experts' judgment- The interviewers, however, were aware of case-control sta tus and thus might have conducted the interviews of the case subjects more thoroughly than those of con trols. Since the experts considered all of the informa tion available, they might have been able to evaluate Am J Ebtdemial Val. 148, No. 2,1905 PAGE 1111' RGYDAT 81511002 3:54,34 PM [Eastern Dapght Timer ^SYR;RIGHTFAULT15* DNS1307' =3-1 O JOB 5871 1 DURAION (msS):06-22 08/08/0 2 23:02, FAX 310 208 5971 INFO 2 008 Pleural Mesothelioma and Asbestos Exposure 139 FABLE 5. Odds ratios for relations between pleural mesothelioma and asbestos exposure paramote4, with a lfltenrv nerind pf-2n-veara Asbestos exposure paranlfitera - ___ ___ No. of No. of codrots OR* 95%a ol; Hghest probability of exposure Not exposed 99 154 1.0 Possible 51 71 1.2 0.8-1.9 Definite 164 87 3.5 2.4-6.3 Highest inton6ity of exposure Low Madtzn High ' 55 106 74 74 12 0.8-1.9 65 28 1.6-4.3 16 7.1 3.9-12.9 Highestfrequency of exposure Sporadic irregular Continuous 56 94 85 as 1.0 46 3.3 26 5.7 0.7-1,6 2.1+.1 3.4- ,9.7 Duration of exposed job (years) 17 8-19 x2Q 63 74 98 64 1.7 1.1-2.8 ea 2.0 34 5.4 3.2-8.9 Tune since first exposure (years) 20-37 88-48 Z49 77 83 75 53 23 47 2.8 58 2.2 1.4-3.8 1.9-4.6 1.43.6 Age atfirst exposure (years) <16 16-22 >.23 66 96 73 56 1.9 1.2-3.1 52 3.0 1.9-4.6 51 2.3 1.5-3.7 Cumulative exposure ("flml-yead#) 0.001-0,49 0.5-10.99 1-9.9 710 Temporal exposure pattern^ Intermittent <0.5 0.5-0.99 1-9.99 XI0 77 1D9 29 12 80 27 49 in oumumve exposure CVM-Ycar) 66 93 19 8 48 21 17 5 1,2 4.2 5.2 JLZ 1-1 4.0 4.0 6.9 2.0-8.8 3.1-8.8 41=16.5 O.B-1.7 1.7-9.7 22-7.2 2.1-16.7 Continuous i0_6 0.5-0.99, 1-9.99 5-/0 11 11 1,9 0.8-q.S 10 4 4.8 1.4*15.4 32 6 9.2 3.7-23.1 32 5 11.3 4.110.7 Odds ratios {Opts) edjustad for age and sodoeconornio oatag*ry, t Ch confidence interval. $ Cumulative exposure index was based on subjective assesanant, that is, so niquantification of exposure by the experts and seleolad weighting factors assigned to each category of exposure, with no objective measurement of airborne asbestos ieveis. Thus, the exposure unit, flml-years, is expmesed In quotation mraka. ' Subjects' exposure was classified as intermittent it ifwas sporadic or irregular and if they had never worked at a job with continuous exposure. The continuous category was re arvad forsubjects who had been employed in at least one job with continuous exposure. n- r r=r,irlorninf Vol. 148. No. 2, 1998 PAM 8111"RCVDAT816010013:54:34PM FastemDayYohtTime)" sVR:MGHTFAX-ClT15r M1307 < CSID:310 80S 5971 tDURAIIQN #ws};0622 08/08/02 13'02 FAX, 310 208 5971 INFO 2 |a009 140 Iwatsubo et al. exposure more precisely for the cases than for the controls. The frequency of the exposure category "possible," used when the experts could not reach a definite conclusion, was higher among controls than among cases, so that this type of error cannot be excluded. We thus undertook a supplementary analy sis to examine, at least in part, the effect of this bias. First, we considered all of the jobs in the possible category as nonexposed. The pattern of dose-response relation was very similar to that observed. Tin signifi cant risk for subjects in the category of less than 0.5 "f/ml-years" and an odds ratio of 7.8 (95 percent Cl 3.8-16.2) for those in the category of more than 10 "flml-years." Classifying all of the poaaibiy exposed subjects as definitely exposed did not change the dose- response relation pattern very much either (OR = 1.0, 95 percent Cl 0.7-1.6 for the lowest category and OR = 7.7, 95 percent Cl 3.8-15.7 for the highest). The validity of the dose-specific risks in our study also depends greatly on the values of the weighting factors selected for each exposure category. For this purpose, we-attempted to retain the intervals used by the experts. Although this procedure is assumed to provide more precise exposure evaluation than would a relative ranking of subjects by an ordinal scale without specified boundaries, some nnisclassification of subjects according to dose-specific exposure prob ably occurred. Indeed, all jobs classified in the same exposure category were assigned the same weighting value without consideration of the variability of expo sure within the category. Such nondifferential misclas- sification of the subjects usually attenuates the relation between exposure and disease and flattens the dose- response curve (39). We should note that the intervals used by the experts for the categories of intensity were rather dissymetric-narrow for medium exposure and large'for very high exposure. There were few job periods with very high exposure, however, so that errors due to the variability in this category should have had little effect on the dose-response relation observed. - We observed a dose-response relation with cumula tive exposure. Because, as stated, the exposure assess ment for the earliest periods might have been under estimated and because of the imprecision o intensity weighting factors, we tested two models using two other series of coefficients for weighting intensity of exposure: 1) second model: 0.5, 1.5, 6, and 550 fibers/ ml, for low, medium, high, and very high exposures, respectively (midpoints of boundaries), and 2) third model: 0.5, 5, 50, and 500 fibers/m.l for low, medium, high, and very high exposures, respectively. These models showed a dose-response relation with the CEI similar to that in the first model, but they did not show as clear a dose-response trend as the first model. In the second model, the odds ratio was 1.0 (95 percent Cl 0.7-1.6) for the lowest dose and 6.4 (95 percent Cl 3.4-12.2) for the highest. The corresponding odds ratios for the third model are 0.9 (95 percent Cl 0.51.4) for the lowest and 7.1 (95 percent Cl 4.2 11.9) for the highest. The pattern of the dose-response curve could have depended on the length of latency period selected. We have used a 20-year latency period, as suggested by McDonald and McDonald (1,21), who concluded that latency is seldom less than 20 years and usually 30-40 years. We also examined the effects of 10- and 30-year latency periods. The results obtained with the former were very similar to those we found with the 20-year latency period. A 30-year latency period re sulted in a lower odds ratio and a less clear doseresponse relation, suggesting that exposure misclassification occurred using such a long latency period. Because no objective measurement was available to testthe validity of the experts1 evaluation, we express the cumulative exposure using units of eimi year? in quotation marks. Even in cohort studies, however, precise measurement of exposure is difficult (2,41). In this study, we used several surrogate parameters for dose to examine dose-response relation, as sug gested by Blair and Stewart (42) and Suarez-Ahaazor et al. (43). We considered separately the intensity, frequency, and duration of exposure, and each was significantly related to mesothelioma. The relative risk increased along with each parameter. In addition, when each of these parameters was adjusted for the others, the relative risk of each, although lower, re mained significant. These results suggest that each exposure parameter contributed to some extent to the Occurrence of mesothelioma, although the doseresponse relation seemed to be describedbest by the CEL The existence of a causal association between as bestos exposure and mesothelioma was fast demon strated in 1960 (44). Both cohort (6-9,11-14,45) and case-control (32, 34-37, 46, 47) studies focusing on mesothelioma and examining surrogate parameters for dose have reported a dose-respouse relation. However, because of the rarity of mesothelioma, even among asbestos workers, little quantitative infor mation is available from which the dose-response re lation can be precisely estimated (1,41,48). Peto et al. (49), using mathematical models, ob served that the risk of mesothelioma in one occupa tionally exposed cohort (North American insulators) was best described by a model in which the risk increases with the third or fourth power of time since first exposure_ They also concluded that their data were compatible with a linear dose response relation Am J t pidemlol Vnl. 148 Nrv 2.199& PAGE 9f 11 "RCVD AT 81601002 3:54,34PM (Eastern DayfghtTimer SYR:RIGHTFAULTI5'' DNISM07"CSffi:MD 208 5971 ' DURATION <mm-s4*22