Asbestos exposure and the risk of lung cancer in a general urban population.

OBJECTIVES - The aim of the study was to investigate the asbestos-associated risk of lung cancer according to histological type of cancer, lobe of origin, pulmonary concentration, and type of amphi bole fibers and also to estimate the etiologic fraction of asbestos for lung cancer. METHODS - The pulmonary concentration of asbestos fibers in 113 surgically treated male lung can cer patients and 297 autopsy cases among men serving as referents was determined by scanning elec tron microscopy. The age- and smoking-adjusted odds ratios of lung cancer were calculated accord ing to pulmonary fiber concentration for all lung cancer types, sqaumous-cell carcinoma, and adeno carcinoma and for the lower-lobe and the upper- and middle-lobe cancers. RESULTS - The risk of lung cancer was increased according to the pulmonary concentration of as bestos fibers (f) of 1.0 to 4.99 . 10 6 f · s' [odds ratio (OR) 1.7] and ~5.0 . 10· f · g' (OR 5.3). The odds ratios associated with fiber concentrations of ~ 1.0 . 1 D· f · g-I were higher for adenocarcinoma (OR 4.0) than for squamous-cell carcinoma (OR 1.6). The asbestos-associated risk was higher for lower lobe tumors than for upper lobe tumors. The risk estimates for anthophyllite and crocidolite-amosite fibers were similar, except for the risk of squamous-cell carcinoma. An etiologic fraction of 19% was calculated for asbestos among male surgical lung cancer patients in the greater Helsinki area. CONCLUSIONS- Past exposure to asbestos is a significant factor in the etiology of lung cancer in south ern Finland. The asbestos-associated risk seems to be higher for pulmonary adenocarcinoma and low er-lobe tumors than for squamous-cell carcinoma and upper-lobe tumors.

Asbesto s-related lung cancer is believed to be numerically the mo st important occupational cancer in the world (l). The frequencies of histological lung cancer types have varied among asbestos-exposed worker s (2), but there is evidence that the asbesto s-associated risk would be especially high for lun g adenocarcinoma (3,4 ).
Anthophyllite asbestos has be en widely used in Finland due to it s domestic product ion in . About 40 % of all asbesto s used in 1918-1988 consisted of anthophyllite (5). An increased risk of lung cancer has been rep orted among anthoph yllite miners and millers (6,7).
The objecti ve of the pre sent study was to estimate the proportion of lung can cers attributable to past asbestos exposure in Finland and to in vestigate the asbe stos-associated risk by histologi cal type , lobe of ori gin, and amphibole fiber type. J Finnish Institute of Occupational Health, Helsinki, Finland. 2 International Agency for Research on Cancer, Lyon, France.
Reprint requests to: Dr A Karjalainen, Finnish Institute of Occupational Health, Topeliuksenkatu 41 a A, FIN-00250 Helsinki, Finland.

Surgical lung cancer patients
The study population consisted of 135 lung cancer pati ents who underwent surgical lobectomy or pulmectomy in the Department of Th oracic and Cardiovascular Surgery at the Hel sinki Universit y Hospital between August 1988 and Jul y 1993 . Con secutivel y dia gn osed lung cancer cases from two of the three sur gical units of the Dep artment were incl uded . The mean age of the 113 men was 62.8 (range 35-81) year s, and the mean age of the 22 women was 61.3 (ra nge 41-75 ) yea rs.
In each lun g cance r case the histological cell type of lung ca ncer and the lobe of ori gin was determined from the surgica l lung specimen and cl assified acco rding to the 1981 classifi cation of the World Health Or ganization. There were 67 squ amous-cell carcinomas, 49 adenocarcinomas, 9 large-cell carcinomas, 9 small -cell carcinomas, and I adenosq uamo us carc inoma. Th ere we re 71 upper-lobe, 56 lo wer-lobe, and 6 middle-lobe cancers. In two cases the lobe of origin could not be defined due to the large size of the tumor.
Th e presence of histological diffuse interstitial fibrosis compatible with asbestosis was determined from the surg ical lung specimen s. At least three tissue sections per lobe were evaluated .

Ref erents
A series of autopsy cases served as the male referents. Lung tissue sample s were collected as a part of a large study on sudden death s among men. This series comprised all sudden, unexpected deaths of men aged 35 to 69 years who had died in Helsinki and were autopsied between 15 January 1991 and 30 l anuary 1992 at the Department of Forensic Medicine, University of Helsinki. Cases in which the body was combu sted or macerated were exclud ed. Two men aged 33 years were inadvertently included. The autopsy series comprised 30% of all death s among men in this age group in Helsinki.
The distribution of the causes of death among the referent s is shown in table I. About 60% of the deaths were due to a disease, including ischemic heart disease (85 cases), other cardiovascular diseases (37 cases ), respiratory diseases (13 cases), cerebrovascular diseases (10 cases), hepatic diseases (8 cases), malignant neoplasms (5 cases), gastrointestinal diseases (6 cases), neur ologic diseases (5 cases), deaths due to chronic alcoholism (4 cases), metabolic disease (2 cases) , and urologic disease (1 case). There were no deaths due to mesothelioma or asbestosis. Two of the deaths were due to lung cancer, and in one addition al case lung cancer was diagno sed but was not the primary cause of death. These lung cancer cases were excl uded from the original series of 300 autopsy cases ; the reference group thus consisted of 297 men. We performed analyses both using all 297 referents and using only those 176 cases in which the death was due to a disease. As the result s were similar, only the results for all 297 referents are prese nted. No reference group was available for the female lung cancer patients.

Random lung cance r patients
As the results of the study are based on the lung burden analyses of the aforementioned surgical lung cancer serie s, they may not be generalizable to non-

Electron microscopy
The tissue pieces for the fiber analysis by electron microscopy were taken from the peripheral part of the lung, not including pleural or tumor tissue. The lung tissue sample s of the referents were taken from the left upper lobe. The samples of the lung cancer patients, in cases of lobecto my, were taken from the lobe where the tumor was situated. In cases of bilobectomy or pulmectomy, the sample was taken from the lobe which appeared to be closest to normal (with the least amount of emphysema or pneumonia).
A tissue piece of about 100 mg (wet weight) was taken for the fiber analysis. A low-temper ature ashing technique was used to remove organic tissue. Fibers (f) were detected with a JEOL 100 CX-ASID4D electron microscope in the scanning mode (8). A length-to-width ratio of> 3 and roughly parallel sides were used as the fiber criteria . A magnification of 5000 x was used in the counting. Fiber s longer than I IJ.m could be detected . A minimum of 200 viewing field s were evaluated to find at least 4 to 30 fibers per sample, depending on the density. With this procedure an analytica l sensitivi ty (one fiber per sample) of about 0.07 . 10 6 f . g dry tissue:' could be reached. According to Poisson statistics , this value corresponds to a detection limit of <0 .3 . 10 6 f · s' (four times the analytical sensitivity).
An energy disper sive X-ray microanalyzer (Tracor TN 5500) was used to determine the fiber type by comparing peak ratios to standard spectra. Amosite and crocidolite have almost similar X-ray spectra and are distinguished poorly. Therefore their data have not been presented separately. In a study on Finnish lung cancer patients, crocidolite fibers accounted for the great majority of amosite-cro cidolite fiber s identified with transmiss ion electron microscopy (9). Chrysotile fibers are poorly detected by scanning electron microscopy, and consequently the results represent the concentration of amph ibole fibe rs. Tremolite concentrations exceeding 0.3· 10 6 f . g -I were detected in nine samples. Chry sot ilc and tremolite fibers have been included in the number of total asbestos fibers, but their data have not been reported separately. The fiber analyse s were carried out by a person who was unaware of the casereferent status of the samples.

Occupational history
The lung cancer patients were interviewed personally about their complete, chronological occupational history, including past occupational, dome stic, and environmental exposure to asbestos. The interview was carried out during their stay in the hospit al. Special attention was focu sed on the detailed description , occurrence, duration, and years of work and task s with definite or probable exposure to asbestos or other occupational carcinogens. The surgical and random lung cancer patients were interviewed by the same interviewer.
The probabil ity of past occupational exposure to asbestos was evaluated by two occupational hygienists by consensus and with out any knowledge of the asbestos counts from the tissue samples . An exposure time of one month was regarded as minimum. The exposure was classified into four categories according to the following guidelines: Definit e exposure (group I): emplo yment in mining asbestos, manufacturing asbestos products, insulating with asbestos, or demoli shing old buildings; probable exposure (group 2): employment in shipyards, the con struction industry, or metal workshops ; possible exposure (group 3): employment in various trades with exposure to dust, such as mining, power plants, tran sportation, or the pulp and paper industry; unlik ely or unknown expos ure (group 4): employment in occupations with no known exposure to asbestos. In addition , the frequency and duration of tasks with at least probable exposure to asbestos were considered in the classification .

Smokin g habits
The information on the smoking habits of the lung cancer patients came from a personal interview. The smoking habits of the referent s were recorded in an interview of one of the relati ves. A relative suitable and willing for the interview was available for onl y 166 of the 297 referents. For 94% of these 166 cases the widow, a child, one of the parents, or one of the sisters or brothers was inter viewed. Two of the male and two of the female lung cancer patients were nonsmokers; the rest were either current or ex-smokers. About 18% of the referents were nonsmokers.

Stati stical analyses
The odds ratio s (OR) and their confidence intervals (CI) were calculated with logistic regression. Adjust-Scand J Work Environ Health 1994, yo120 , no 4 ment for age (four classes) and smoking-years (five classes) were used. The etiologic fraction (population attributable risk ) was calculated by multiplying the proportion of exposed cases of all the cases by the term (RR-I )/RR. The age-adjusted odds ratios were used as estimates of the risk ratio (RR). Table 2 show s the age-adjusted odds ratio s of lung cancer according to asbestos exposure as est imated from the pulmonary fiber counts. The risk estimate was greater for high fiber counts (<::5 . 10 6 f . g') than for interm ediate fiber counts ( 1.0-4.99 . 10 6 f . g '), The risk estimates wer e similar for anthophyllite and crocidolite-amosite fibers . Wh en the two exposed categories were combined, an age-adju sted odds ratio of 2.3 (95 % CI 1.3-3.9, P =0.004) was calculated for asbestos fiber counts of <:: 1.0 . 10 6 f. g-I. Thi s value gives an etiologic fraction of 19% for asbestos among male lung cancer patients. The risk estimates of table 2 give an etiologic fract ion of 9.8% for 1-4.99· 10 6 f· s' and 9.3% for <::5.0·

Results
Two of our patients had been previously diagnosed as suffering from asb estosis, and in seven additional cases mild histological fibrosis compatible with asbe stosis was detected in the surgical lung specimen . When the cases with clinical asbe stosis or histological indication of fibrosis compatible with asbestosis were excluded, an increased risk of lung cancer was still associated with asbe stos fiber concentrations of <::5 .0 . 10 6 f · g' (age-adjusted OR 2.8, 95% CI 0.9-8.7, P =0.07) and asbestos fiber concentrations of 1.0-4.99 . 10 6 f . g:' (OR 1.5,95% CI 0.8-2.9, P = 0.19).
The risk was ele vated for both adeno carcinoma and squamous-cell carcinoma, but the risk estimate was greater and stati stically signifi cant only for adenocarcinoma (table 2). For the men with a pulmonary fiber concentration exceeding or equaling I . 10 6 f . g', about 44 % of the lung cancers were adenocarcinomas versu s 27% of adenocarcinomas among those with less than I . 10 6 f . g' in lung tissue. Among the 22 fema le lung cancer patients the distribution pattern was different. None of them showed an elevated pulmonary fiber concentration (<:: 1 . 10 6 f · s'): yet 55% of the lung cancers were adenocarcinomas (table 3). The asbestos-associated odds ratios were higher for lower-lobe cancer than for cancers of the upper-middle lobe (table 4).
Smoking habits were known for 166 of the referents. The smoking-adjusted risk estimates for asbestos expo sure (tables 2 and 4), acquired for only referents with known smoking habits, were similar to the smoking-unadj usted risk est imates.
Lung tissue samples were not available for the random lung cancer patients. Table 5 shows the distribution into exposure categories according to the eval-245 Scand J Work Environ Health 1994, vol 20, no 4  b Adjusted for age and pu lmonary co ncentration of crocldol ite-amost te fibe rs. C Adjusted for age and pu lmonary con cent rat ion of anthophyll ite fib ers. Table 3. Distribution of histological lung cancer types accord ing to pulmonary concentration of asbestos fibers among 113 male and 22 female surg ically treated lung cancer patients.
Histolog ica l ce ll type of lung cance r Fib er concentr ation in lung t issue were detected in 79% of those with definite, 41% of those with probable, 25% of those with possible, and 7% of those with unlikely exposure (table 6). Among the 39 patients with a concentration of at least 1 . 10 6 f . g-I there were one asbestos factory worker and four men who had done insulation work. The rest had • OR 2 adjusted for age and smoking years with the use of 166 referents with known smo king habits. been exposed in construction, shipyard , or maintenance work.

Discussion
An elevated risk of lung cancer , with an indication of a dose-response, was assoc iated with increased pulmonary fiber concentrations. The comparisons of smoking-adj usted and unadju sted risk estimates did not indicate that the results were biased by differences in the smoking habits between the exposed and unexposed subjects. The smoking habit s were not known for all the referents , however, and a complete adjustment for smoking could thus not be done. Further difficulti es in the adjustment of the effect of tobacco smoking were caused by the small number (two cases) of nonsmokers among the male lung cancer patient s. Hence, the lowest smoking category used in the analyses not only contained nonsmokers, but also smokers with less than 20 smoking-years. The accuracy of the informat ion given by the relatives of the referents on smoking-years may have also been less than that given by the lung cancer patients in their personal interview. Yet it is unlikely that these methodological inaccuracies would explain the risk estimates associated with past asbestos exposure. As estimated from the occupational histories, the past exposure to asbestos was similar between the surgical and random lung cancer patients. This finding suggests that there were no major differences in the exposure between the surgical and random lung cancer patients . Among the surgical lung cancer patients there was, however, an overlap in the pulmonary fiber concentrations between the exposure categories, especially between groups of probabl e and possible exposure. This finding implies that the pulmonary fiber concentration and the occupational history classification are not identical exposure indicators. The differences in the distribut ion of histological types between the two cancer series also indicate that the result s of the surgical lung cancer series may not be generalizable to apply to all lung cancers, especially in respect to small-cell cancer. In previous studies the etiologic fraction of asbestos in a random lung cance r series was 6% in Glasgow and the west of Scotland, 20% in Trieste, Italy, 23% in Telemark, Norway, and 16% in Goteborg, Sweden (10)(11)(12)(13). It must be emphasized that, when the eti-ologic fraction is equated with the expected change in disea se load following the removal of a risk factor, it is supposed that the relationship between the factor and the disease is causal and not only statistical. In the case of asbestos and lung cancer a causal relationship is reasonably well establi shed , and the etiologic fraction is a relevant measure when prevention strategies are evaluated. When etiologic fractions are interpreted, it must, however, be emphasized that, due to the combined effect of smoking and occupational carcinogens, the sum of the epidemiologically calculated etiologic fraction s of different agent s in a given population usually exceeds 100%. The use of the etiologic fraction also involves terminological and conceptual problems and usually requires specific assumptions about exposure action and interactions (14).
In previous studies of several asbestos-exposed cohorts, the ratio of excess cases of lung cancer to mesothelioma cases has varied from I: I to 10:1 (1, 15). During the five-year period from 1987 to 199I, there were 34 cases of mesothelioma among men in Helsinki according to the statistics of the Finnish Cancer Registry. Th is value would correspond to 34 to 340 excess cases of lung cancer, which is 3.9 to 39% of the 872 lung cancer cases among men during the same period of time in Helsinki. Our estimate of 19% among male surgical lung cancer patients is within this range. The corresponding proportion for other parts of Finland is probably lower, as the use of asbestos was more extensive and the number of expo sed construction, shipyard, and industry workers was higher in Helsinki and the surrounding area than in other parts of Finland. This result is reflected by the age-adjusted incidence rate of mesothelioma among men in 1987-1991, which, according to statistics of the Finnish Cancer Registry , was 9 per million men in Finl and as compared with 23 per million men in Helsinki. If we assume that the regional differences in the diagnostics of mesothelioma are minor, these figures suggest that the asbestos-attributable fraction of all Finnish lung cancers among men may be less than half of our estimate for the greater Helsinki area .
Most of the exposed lung cancer patients had been exposed in construction, shipyard, or maintenance work. Exposure during the use of asbestos products in construction and shipyard work and indirect exposure during insulation work and asbestos spraying are thus not to be neglected when the health hazards of asbestos and their prevention are evaluated. In such cases it is often difficult to assess the intensity of past exposure on the basis of the work history and exposure interview, and crucial inform ation of the individual exposure can be gained from a pulmonary fiber analysis. When pulmonary fiber analyses are used at the individual level, it may be preferable to use several samples from different parts of the lung to improve the analytical reliability. A wide varia-248 tion in the fibe r counts has been observed in interlaboratory compar isons , and laboratory-bound reference values should be used (16). Despite the obvious advantages of the pulmonary fiber analysis. it should be stressed that a complete work history with special empha sis on jobs and tasks with probable or definite exposure to asbesto s is the most import ant exposure indicator, and actually the only way to estimate latency.
Chrysotile is cleared more rapidl y from the lungs than amphiboles, and, even if transmission electron microscopy is used, the chrysotile content of lung tissue is not an equall y representative measure of past cumulative chrysotile exposures as is the amphibole content for amphibole exposures (17, 18). As about 40 % of all asbestos used in Finland durin g 1918-1988 consisted of amphiboles (5) and a mixed exposure to chrysotile and amphiboles took place in most of the industrial applications, the amphibole content in the lung tissue is probably a reasonably representative indicator of past exposure to asbestos in Finland . In cases with main exposure to chrysotile these methodologi cal problems must, however, be recognized . The use of occupational history as an exposure indicator in the risk analyses would not suffer from such problems. In our study the differences in the quality of occupational history data between the cases and the referents, however, did not allow us to do such analyses.
In our study the sample site for the fiber analysis among the lung cancer cases depended on which lobe or lobes were resected, wherea s the tissue samples of the referent s were all taken from the upper lobe. We assume that these fiber concentrations are comparable and equally repre sentati ve of cumulative exposure to asbestos. In previous electron microscop y studies counting fibers of all sizes together, either no systematic differences or an indication of higher concentrations in the upper lobes than in the lower lobes has been found (8,(19)(20)(21). If the fiber concentration is higher in the upper lobes, our method would have underestimated the exposure of lower-lobe lung cancer cases and thus also the asbesto s-associated risk estim ates.
The absence of elevated pulmonary fiber concentration s and expo sure histories indicating probable or definite exposure to asbestos among the 22 female surgical and 30 random lung cancer patients suggests that the proportion of lung cancers attributable to asbestos is much lower among women than among men. Thi s observation reflects the low proport ion of femal e worker s in the construction. shipyard, and asbestos industries. Thi s result is also in accordance with the statistics of the Finnish Registry of Occupational Diseases; in 1987-1991 129 and 4 cases of asbestos-related lung cancer were reported for the men and women, respectively. This ratio of men to women among the cases (about 32: I) is much higher than the about 4.5: I ratio for men to women amon g all lung cancers in Finland during 1987-199 1 (860 I men and 1905 women according to the statistics of the Finnish Cancer Registry).
Ou r results indicate that the asbestos-associated risk of lung cancer is higher for adenocarcinoma than for squamous-cell carcinoma. Because of the small number of patients, we were unable to draw any conclusions about the association between asbestos and small-cell or large-cell carcinoma . The higher risk of adenocarcinoma, as compared with squamous-cell carci noma, is in agreement with the findings of previous studies among Danish and Swedish asbestos cement workers (3,4), but also opposite result s or report s with no difference in the risk have been published (2,11,22). Different selection criteria between exposed and referen ce lung cancer cases (surgery, autopsy, bronchoscopy) and variation between pathologists in the typing of lung tumors may have confo unded some of these studies (23-25). It thus remains controversial whether the asbestos-associated risk of lung cancer is greater for some histological cancer type or wheth er there are no differences in the risk. A differen ce in the risk between these cancer types would raise the question of a possible difference in the carcinogenic mechanisms. Both asbestos and smoking seem to be compl ex carcinogens which can affec t more than one stage of lung carcinogenesis, and very limited information is available on the interaction between asbe stos and smoking in causing specific histological types of lung cancer (26). It is interesting that in our study the proportion of adenocarcinomas was especiall y high among the female lung cancer patients (all unexposed to asbestos). Th is find ing prob ably indicates that women have a lower incidence of tobacco-associated squamous-cell carcinomas rather than that they would have a higher incidence of adenocarcinoma. The differen ce in the asbestos-associated risk between lower-and upper-lobe tumors is compatible with the findings of our previous report (9).
There was an elevated risk of lung cancer associated with high pulmonary concentrations of asbestos fibers even after cases with histological indication s of mild asbestosis were excluded. These findings support the view that asbestos increases the risk of lung cancer even in the absence of asbestosis. As for the cases with asbestosis , it is impossible to conclude whether the fibrosis and the lung cancer were independently caused by asbestos or whether the cancer was caused by the fibrosis. Most of the asbestosis cases were mild and could be detected in the histological examination only. It is doubtful whether such a mild case of fibro sis could have caused the cancer. Some of the exposed referents might have also had mild histological fibrosis compatible with asbestosis and should thus have been excluded from the aforementioned analysis. Thi s exclusion would have increased the risk estimate. As the presence of histological fibrosis among the referents was not evaluated, no such exclu sion could be done . It is notew orthy that six of the nine lung cancer patients Scand J Work Envi ron Health 1994, vol 20. no 4 with histological fib rosis (and heavy exposure) had an adenocarcinoma, two a squamous-cell carcinoma, and one a small-cell carci noma.
Because of the small number of nonsmokers among the male lung cancer patient s, it was not possible to compare the asbestos-associated risk between the nonsmokers and smokers. The two male nonsmokers in the lung cancer series were both unexposed to asbestos and had adenocarcinoma.
The use of pulmonary fiber concentrations as exposure indicators resulted in similar lung cancer risk estimates for anthophyllite and crocidolite-amosite fibers. This finding is in agreement with the lung cancer risks observed in previous studies among anthophyllite and crocidolite miners (6,27,28). It is noteworthy that there is a clear difference in the risk of mesothelioma between these cohorts of anthophyllite and crocidolite miners. This difference could be due to the differen ce in the dimensional characteristics between these fiber types. It has been concluded that mesothelioma is the most closely assoc iated with numbers of fibers longer than 5 urn and thinner than 0.1 IJ.m, whereas lung ca ncer is the most closely associated with numbers of fibers longer than 10 IJ.m and thicker than 0.15 um (29). In the present study the risk estimates of squamous-cell carcinoma were elevated for elevated anthophyllite concentrations but not for elevated crocidolite-amosite concentrations. This observation was based on a small number of cases and was not statistically significant. The distributi ons of histological lung cancer types in the cohorts of anth ophyllite or crocidolite miners have not been report ed.