This review examines the hypothesis that excess lung cancer risk in worker cohorts exposed to asbestos occurs only among those with asbestosis. The adequately designed studies in the literature support this hypothesis. The summary relative risk for lung cancer was 1.00 in seven cohorts with no deaths from asbestosis. In addition, there is a high correlation between asbestosis rates and lung cancer rates in 38 cohorts in contrast to a poor correlation between cumulative exposure data and lung cancer relative risks in eight cohorts with adequate data. The evidence indicates that asbestosis is a much better predictor of excess lung cancer risk than measures of exposure and serves as a marker for attributable cases. (CHEST 1999; 115:536-549)
Key words: asbestos; asbestosis; cohort studies; lung cancer; relative and attributable risk
Abbreviations: CI = confidence interval; f/mL = fibers per milliliter; mppcf = million particles (dust) per cubic foot; RR = relative risk; SIO = small irregular opacities; SMR = standardized mortality (or morbidity) ratio; SMSA = standard metropolitan statistical area
While the primary cause of lung cancer in this century is cigarette smoking, occupational causes contribute to the problem. Among these, asbestos has received the most attention.
In 1955, Doll published an important cohort study showing that the risk of lung cancer was substantially elevated among asbestos textile workers in London. The cohort consisted of 113 men exposed for [is greater than or equal to] 20 years. Follow-up from 1922 to 1953 revealed 11 deaths from lung cancer compared with 0.8 expected based on lung cancer mortality rates in England and Wales. All 11 cases were confirmed by autopsy and all had asbestosis.
Most subsequent cohort studies have not included specific information on the relationship of asbestosis to the increased risk of lung cancer. Consequently, some investigators have espoused the concept that increased lung cancer risk can occur in the absence of asbestosis. Much debate has resulted over the hypothesis that excess lung cancer risk occurs only among those workers who develop asbestosis, some favoring the hypothesis[3-6] and some opposing it.[7-9] In this debate, there also has been a difference of opinion as to whether the issue is one of causation or one in which asbestosis should be considered a marker that predicts an increased risk of lung cancer and implies attributability.
This review will focus on the cohort studies that provide evidence bearing on the hypothesis. The restriction to cohort studies is based on the principle that risk can be measured only by the incidence or mortality rate. The literature reviewed was limited to articles published through January 1997.
Case-control studies were not considered because there are only a few and they are prone to more opportunities for bias than are cohort studies. They usually do not meet the requirement of two important assumptions inherent in the method: the cases should be representative of all cases of the disease under investigation and the controls should be representative of the population that gave rise to the cases. An example of this problem is a recent case-control study dealing with the issue at hand. Wilkinson et al studied a series of 9,71 patients with confirmed lung cancer admitted to a London chest disease hospital and 678 control subjects in the same hospital with other respiratory disease or cardiac disease. Such hospital-based series are not representative samples of lung cancer cases or the population giving rise to the cases. Other substantial flaws in this study were described by Jones et al in a thorough review of the literature bearing on the hypothesis being considered herein. Consequently, the conclusion by the authors that asbestos is associated with lung cancer even in the absence of radiologic evidence of pulmonary fibrosis is not warranted. Autopsy-based studies are subject to similar bias.
While observational cohort studies have some inherent potential for methodologic flaws, they represent the best available current technique for estimating risk of disease. Bias can be a problem, but the major concern is the control of potential confounders. Age and era are generally taken into account but smoking, the most common cause of lung cancer, has been evaluated in only a few studies and in a less than satisfactory way.
It has been suggested that asbestosis is an outcome independent of lung cancer in the response to exposure and that both diseases are associated only because both are dose related to exposure. This interpretation of the available data is speculative until it can be confirmed by valid investigation. In any case, the hypothesis considered herein implies only that asbestosis is a marker for increased risk of lung cancer.
COHORT STUDIES WITH RELEVANT DATA
Data published in a 1960 book in German by Bohlig et al are presented in Table 1. They reported a study of living German asbestos workers in 1951 to 1959 with regard to the lung cancer incidence rate per 10,000. The rate was 5.0 among those without asbestosis, almost the same as the rate in the general population aged 50 to 79 years. The rate increased dramatically to 50.0 in workers with recorded asbestosis.
Table 1--Observed and Expected Lung Cancer Incidence in German Asbestos Workers 1951 to 1959(*)
(*) Source: Bohlig et al.
Five years later, Jacob and Anspach examined the standardized mortality ratios (SMRs) for lung cancer in a cohort of all 2,636 asbestos workers in Dresden from 1959, to 1964 by sex and the presence or absence of asbestosis. Table 4 in their report presented the data for all workers and a subcategory of those with asbestosis in each sex. It is necessary to calculate the data for those without asbestosis. The results are shown in Table 9, after subtracting six cases of pleural mesothelioma occurring only in women. The risk of lung cancer in each gender group was elevated only in workers with asbestosis defined by a chest radiographic classification previously devised by Saupe.
Table 4--Lung Cancer SMRs Among Quebec Miners and Millers by Chest Radiograph Readings(*)
(*) Source of data: Liddell and McDonald, 1980.
([dagger]) Based on the Poisson distribution.
([double dagger]) SMRs are calculated from RRs in the authors' Table 6.
Table 9--Cumulative Asbestosis and Excess or Deficit Lung Cancer Rates in 38 Cohorts
(*) I = inception cohort; C = cross-sectional cohort.
([dagger]) Incidence study.
([double dagger]) Cohort limited to 15,999 workers employed before 1969.
In 1968, Knox et al reported a further study of 256 male workers in the London textile plant originally investigated by Doll. The men had worked at least 20 years (their groups 1 to 3). Their Table 7 showed 18 deaths attributed to lung cancer on the death certificate. The data with respect to the presence or absence of asbestosis are presented in my Table 3. Eleven of the 18 died with asbestosis on the death certificate. Of the seven without such a diagnosis, three had asbestosis at autopsy (see Table 7 in the article by Knox et al). It is well known that death certificates may sometimes omit clinical diagnoses other than the underlying cause of death. If this was true for these 3 cases of lung cancer, then the remaining 4 cases would almost exactly equal the 4.10 cases expected. There is no information on radiographic evidence of asbestosis in this study.
Table 3--SMRs for Lung Cancer Among 256 Men Employed [is greater than or equal to] 20 Years in an Asbestos Textile Factory, 1916 to 1966, Based on Death Certificates(*)
(*) Source of data: Knox et al.
([dagger]) Based on national age-, sex-, and period-specific death rates.
([double dagger]) Based on the Poisson distribution.
([sections]) Three had asbestosis at autopsy not mentioned on the death certificate.
Table 7--Other Characteristics of Asbestos-Exposed Cohorts Without Deaths due to Asbestosis
(*) NA = not available.
Liddell and McDonald reported in 1980 a cohort study of Quebec miners and millers on radiologic findings as predictors of mortality. Of a cohort of 11,379 workers born in 1891 to 1920, a subcohort of 4,559 men was established with the following requirement: there was at least one chest radiograph before leaving the industry up to 1966. Tracing was carried out for causes of death until the end of 1975; only 1.3% were untraced. The SMRs for lung cancer are shown in Table 4 by chest radiograph reading. In the group with normal chest radiographs, there was no increased risk for lung cancer, but when small parenchymal opacities were found (according to a system that later became the UICC/Cincinnati classification), the SMR was statistically significantly elevated to 3.11. Of a total of 119 lung cancer deaths, 37 patients (31%) had abnormal chest radiographs that did not include small opacities; I calculated the SMR to be 3.30 based on data in the report. This elevation was also statistically significant.
Thus, there was a substantially elevated risk of lung cancer in the group with chest radiograph abnormalities other than small opacities. This may be due to the fact that Liddell and McDonald did not allocate individuals with abnormal radiographs to mutually exclusive categories. All these categories showed some elevation of risk and this may be due to contamination from two of the categories. First, 47 workers had large opacities; these are not characteristic of people exposed to asbestos and at least some of the large opacities represented lung cancer since the relative risk of lung cancer was very high, 20.7. Second, 218 workers had films classified under "additional symbols" in the International Labor Organization (ILO) form for classification of chest radiographs for dust diseases of the lung; this is a miscellaneous category likely to include obvious or suspected lung cancer.
Furthermore, the interval between termination and death from lung cancer varied markedly. Only 41% had an interval of [is less than] 5 years, in 29% it was [is greater than or equal to] 10 years, and in 17% it was [is greater than or equal to] 20 years. Becklake et al reported a study of 277 Quebec miners and millers followed from a survey, which occurred in 1967 and 1968 and was followed-up in 1974. Chest radiographs were evaluated independently by three readers. Of those with no parenchymal abnormality in the first radiograph, 17 to 38% were read as showing parenchymal abnormality in 1974 (interreader variation). Consequently, in the mortality study, there was ample opportunity for the workers to develop parenchymal disease prior to death from lung cancer.
In 1991, Hughes and Weill published the results of a cohort study in which 642 men employed in two asbestos cement plants were followed up from 1969, when chest radiographs were taken, through 1983. The analysis was controlled for age, cigarette smoking, and asbestos exposure. Mortality was examined [is greater than or equal to] 20 years after hire. The radiograph assessment was made independently by three experienced readers who were blind to the exposure history. They used the 1971 ILO classification. The median reading was used in the analysis.
Their findings are presented in Table 5. The only group with a statistically significant increase in the risk of lung cancer was composed of 77 workers with small irregular opacities (SIOs) in a profusion of [is greater than or equal to] 1/0:SMR of 4.29.
Table 5---SMRs for Lung Cancer by Radiograph Reading in 1969 in a Cohort of Asbestos Cement Workers 20+ Years After Hire, Followed-up Through 1983(*)
(*) Source of data: Hughes and Weill.
The numbers of lung cancer cases in each category are small so it is possible that the statistically insignificant but slightly elevated SMRs for pleural disease and the category for small irregular opacities with profusion of 0/1 could represent a small excess risk, but the probability of this is small. It is important to note that the data on asbestos exposure (concentration and cumulative exposure) were similar for the 77 men with 1/0+ SIO and the 211 men having no radiographic abnormalities employed [is greater than] 21.5 years. Among the workers with normal radiographs, there was no significant trend in the lung cancer risk by cumulative exposure. Although there was a gap between the chest radiograph and death from lung cancer, as in the Quebec miner and miller study, it did not exceed 15 years.
LUNG CANCER RISK IN COHORTS WITH NO DEATHS DUE TO ASBESTOSIS
An ongoing review of articles on asbestos-related disease has been maintained through a periodic Medline search and bibliographies in publications. If there is a close association between asbestosis and lung cancer, one would expect that those cohorts with no deaths from asbestosis would show no excess risk of lung cancer. My search has provided a group of seven cohort studies having no deaths due to asbestosis.[18-24] Information on these has been assembled in Tables 6-8.
Table 6 Industry and Gender Of Asbestos-Exposed Cohorts Without Deaths due to Asbestosis
Table 7--Other Characteristics of Asbestos-Exposed Cohorts Without Deaths due to Asbestosis
(*) NA = not available.
Table 8--SMRs in Asbestos-Exposed Cohorts Without Deaths due to Asbestosis(*)
(*) Note: no study had adequate adjustment for smoking habits.
([dagger]) 1.01 (95% CI, 0.65 to 1.49) for latency of > 20 years.
The industries represented in Table 6 varied from miners to factory workers and maintenance people. Table 7 shows that latency period was taken into account in six of the seven studies, mortality being examined after 15 years from onset of exposure in two reports and after 20 years in four. In the article by Gardner et al, the results are given for latencies of 1 to 43 years, but the SMR of 0.97 in my Table 8 was shown in the text to be essentially the same for lateneies of [is greater than] 20 years: 1.01 (95% confidence interval [CI], 0.65 to 1.49). Follow-up periods extended to [is greater than] 40 years from onset of exposure for some cohort members in five of the studies so there was ample time to detect an increase in lung cancer risk.
SMRs ranged from 0.79 to 1.37 (Table 8). The only SMR that was statistically significantly elevated above 1.00 was reported by McDonald et al in 1984. This dealt with a chrysotile friction products plant. I excluded short-term workers with [is less than] 1 year of employment because the authors were puzzled by "the fact that the only subcohort with SMRs dearly above expectation comprises men employed for less than one year." The SMR for this group was 1.80, higher than any other subcohort stratified by duration of employment. This phenomenon has been noted by others. Furthermore, there was no gradient of the SMRs in the remaining subcohorts nor was there a dose-response trend with increasing cumulative dust exposure.
The summary SMR for all seven cohorts (Table 8) was exactly 1.00. This finding implies that if the asbestos exposure was not sufficient to cause any deaths from asbestosis, there is no increased risk of lung cancer. A further implication is that there is an exposure threshold for asbestos-related lung cancer as discussed by Browne, a point that will be considered later in this review.
There are also cohort studies of people who have pleural plaques but no asbestosis with regard to lung cancer risk. In 1993, I reviewed six such studies. Two of them reported overlapping data from the same shipbuilding city, Barrow in England. These two were seriously flawed because the authors could not obtain permission to carry out a cohort study of all employees with pleural plaques. None of the remaining four studies had a statistically significant increase in the SMRs for lung cancer. The summary SMR for these four studies was 1.01 (95% CI, 0.75 to 1.33).
In 1994, Hillerdal reported an additional study of people with only pleural plaques. The cohort was derived from periodic chest radiographic surveys of a Swedish community with a population of about 250,000. The surveys were done in 1970 to 1985 and follow-up was maintained to the end of 1991. The participation rate was 70% in the age group of 40 to 69 years. Unfortunately, the authors included in the cohort of 1,596 men a group of 246 (15.4%) men in whom plaques were discovered by accident during "investigation for some other reason" so there was a potential source of bias. After excluding those men who developed radiographic evidence of asbestosis, the SMR was 1.4 (95% CI, 1.04 to 1.97) after adjustment for smoking habits. In the age group 40 to 69 years, the relative risk (RR) was 1.29 (95% CI, 0.88 to 1.68), not statistically significant. In view of the potential bias mentioned above, interpretation of the relative risk is uncertain.
Tola et al studied cancer incidence in 12,693 shipyard and machine shop male workers who were employed at least 1 year during the period 1945 to 1960 in Finland. Almost all shipyards were first built in 1945. Follow-up was carried out through the national cancer registry for 1953 to 1981 and 99.7% of the workers were traced. In a group of 4,582 shipyard welders and platers, none was diagnosed as having asbestosis. Indirect evidence indicated that asbestos exposure had not been high in Finnish shipyards. The standardized incidence ratios were not statistically significantly elevated for lung cancer among workers in these jobs: 1.15 (95% CI, 0.76 to 1.67) for welders and 1.16 (95% CI, 0.95 to 1.41) for platers.
ASSOCIATION BETWEEN ASBESTOSIS RATES AND EXCESS (OR DEFICIT) LUNG CANCER RATES
If asbestosis is a marker for the increased risk of lung cancer in people exposed to asbestos, then there should be a high degree of association between asbestosis rates and excess (or deficit) lung cancer rates in various cohorts. A search of the literature through 1996 was made for cohorts in which it is possible to estimate the rates for each disease, using the number of individuals at the start of observation as the denominator and calculating the cumulative rates, because person-year denominators were not available in some of the reports.
The analysis was restricted to the latest report on each cohort with the requirement that at least 90% of members be traced. When there were data by sex and occupation, each group was tabulated as a separate cohort, eg, men, women, construction work, mining, etc. All studies used mortality as the outcome measure except for the incidence study by Tola et al. In the mortality studies, asbestosis was the underlying cause of death. In the incidence study by Tola et al, the authors stated that there were no diagnoses of asbestosis. There was a total of 38 cohorts analyzed in 30 reports,[18-24,29-51]
Table 9 shows the data for the 38 cohorts. The first eight are those in which there were no deaths from or diagnosis of asbestosis and they are listed alphabetically by last name of the first author. The remaining cohort studies are also listed alphabetically by last name of the first author. It should be noted that the article by Peto et al is listed with those cohorts having deaths from asbestosis in Table 9 while a subcohort of workers employed [is less than] 10 years appears in Tables 6-8 because there were no deaths from asbestosis in the subcohort. In addition, the male and female cohorts studied by Gardner et al were studied separately in Table 9 but combined in Tables 6-8.
Twenty-nine of the 38 cohorts included [is greater than] 1,000 subjects and as many as 15,999. Each cohort in Table 9 is classified by the type of worker registration: an inception cohort registers all workers at the beginning of employment while a cross-sectional cohort registers workers at one point in time without regard to previous employment history. This is of some importance because cross-sectional registration introduces some bias that tends to increase the RR for lung cancer. The bias results from excess representation of workers with moderate-to-heavy cumulative exposure.
Figure 1 is a scattergraph of data for each cohort recorded in Table 9. The cumulative excess (or deficit) lung cancer rate is regressed on cumulative asbestosis rate for all 38 cohorts. The linear correlation coefficient was 0.74 and this was statistically highly significant. The intercept was not statistically significantly different from a zero excess lung cancer rate.
[Figure 1 ILLUSTRATION OMITTED]
Linear regression for the 20 inception cohorts gave a higher correlation coefficient of 0.93, highly statistically significant, and again the intercept was not significantly different from zero. The analysis for the 18 cross-sectional cohorts gave similar results but the correlation coefficient was lower, 0.69, also statistically highly significant.
The study by Newhouse et al is of particular interest because there were six subcohorts numbering 512 to 1,369 workers (Table 10). Four subcohorts of male factory workers were stratified semiquantitatively by degree (light and severe) and duration ([is less than or equal to] 2 years and [is greater than] 2 years) of asbestos exposure. It should be noted that the small excess lung cancer death rate in the male factory workers with light/ moderate exposure for [is less than or equal to] 2 years is not statistically significantly different from no excess (see third footnote). The other two categories consisted of laggers and female factory workers without information on degree and duration of exposure. Figure 2 is a scattergraph of cumulative excess lung cancer death rate plotted against cumulative asbestosis death rate for the six groups. The four male factory subcohorts are represented, from left to right, by the first three points and the last one. For these four points, the linear correlation coefficient is 0.99. The points for the laggers and female factory workers (fourth and fifth points) were close to the regression line for all six sets of data; the correlation coefficient was 0.98, statistically highly significant, and the intercept was not significantly different from a zero excess lung cancer rate.
[Figure 2 ILLUSTRATION OMITTED]
Table 10--The Association Between Cumulative Asbestosis Mortality Rate and Cumulative Excess Lung Cancer Mortality Rate in One Study(*)
(*) Newhouse et al, 1985.
([dagger]) Number per 1,000 at start of observation.
([double dagger]) Based on an observed number of 24 and expected number of 16.5 in 884 workers with an SMR of 1.45; 95% CI, 0.93 to 2.16.
Thus, the cumulative asbestosis mortality rate is an excellent predictor of the cumulative excess lung cancer mortality rate.
EXPOSURE-RESPONSE RELATIONSHIP FOR LUNG CANCER
Opponents of the hypothesis that excess lung cancer risk occurs only among workers who develop asbestosis have the following argument: asbestos is fibrogenic and carcinogenic but these two effects are assumed to be independent; since both the risk of asbestosis and the risk of lung cancer are exposure related, the diseases are associated but attributable lung cancer is not dependent on asbestosis.
Steenland et al have pointed out that in recent years there has been increasing study of internal markers (biomarkers) as indexes of internal dose of disease-causing agents. Not everyone who is exposed to a carcinogen develops cancer attributable to the carcinogen. An internal marker may be a better predictor of cancer risk than a measure of external exposure. Steenland et al noted the well-known observation that cohorts of asbestotics usually have higher lung cancer risks than cohorts of asbestos-exposed workers for which there is no stratification by the presence or absence of asbestosis. They reviewed some of the relevant literature and concluded that the presence or absence of asbestosis is a better predictor of lung cancer risk than the level of asbestos exposure.
In the previous section of this review, analysis of the cohort literature showed a strong association between asbestosis and lung cancer on a group basis in 38 cohorts. To compare the strength of the association between external exposure to asbestos and lung cancer risk, the literature is much more limited.
Lash et al recently published a meta-analysis of the relation between cumulative asbestos exposure and the relative risk of lung cancer in 15 cohorts. They found substantial heterogeneity in the exposure-response relationship. Sources of heterogeneity included not only exposure measurement but also industry category, smoking habits (where available), and standardization procedures.
The cumulative exposure units used in the various cohorts were based by the original authors on dust counts in million particles per cubic foot (mppcf) in some reports and on fiber counts (f/mL) in other reports. Lash et al used three conversion factors to convert mppcf to f/mL; the conversion factors differed by type of industry.
For the purpose of plotting exposure-response curves, I limited the cohorts in the review by Lash et al to eight that provided data on at least five exposure strata in the latest publication available. There were seven articles for the eight cohorts.[20,21,38,40,56-58]
For a textile plant in Charleston, SC, Lash et al used a 1994 report by Dement et al. so I used this one also in this analysis; in Table 9 I used a 1994 article by Brown et al presenting essentially the same data. For the Quebec miner and miller cohort, Lash et al used two nonoverlapping reports by McDonald et al.[18,41] I also used these in Table 9 that I prepared before publication of the latest comprehensive report on this cohort by Liddell et al. The latter article was used for analysis in this section. Thus, the seven reports utilized in this section include my references 20, 21, 38, 40, and 56 to 58, as noted above.
Three articles used by Lash et al were excluded for the following reasons: Henderson and Enterline dealt only with retirees and two reports by Seidman et al[49,60] on an amosite factory cohort relied on estimates of exposure from another amosite factory owned by the same company but established elsewhere in the United States after the factory reported on was closed.
The eight cohorts included in this analysis provided 42 pairs of data. Lung cancer SMR was plotted against cumulative asbestos exposure in fiber/mL-years provided by Lash et al. The resulting scattergraph is not shown here. There was no correlation between the two variables. The correlation coefficient was 0.036 (p [is less than] 0.80).
The individual exposure-response curves are shown in Figures 3-5. Figure 3 shows the curves for four cohorts of workers engaged in manufacture of asbestos cement and friction products. In two plants reported by Hughes et al, plant 1 showed no exposure-response relationship while in plant 2 the curve indicated a small elevation of risk. The cohort reported by A. D. McDonald et also showed no exposure-response relationship. In sharp contrast, the cohort reported by Finkelstein showed a curve with a very strange shape. This deserves further comment.
[Figures 3-5 ILLUSTRATION OMITTED]
The Finkelstein curve is based on a 1984 report in which the lung cancer risks presented were not SMRs. They were rate ratios based on mortality rates per 1,000 man-years among 535 exposed male employees in an Ontario asbestos-cement factory run by the same company as those plants covered in the article by Hughes et al. The comparison was to rates for Ontario men and the various rates were standardized to the age and latency distribution of the entire cohort. Flaws in the design of the study are as follows: tracing was only 84% complete, the Ontario male rate was based on statistics for 1970 to 1974 whereas the period during which person-years of exposed men were accumulated is not given, the numbers of lung cancer deaths in each stratum of exposure were small, and exposure estimates in the early years of exposure were uncertain. Interpretation of this strange exposure-response curve is impossible.
Figure 4 shows the exposure-response curves for three asbestos textile plants.[21,40,57] All curves indicate an exposure-response relationship but there is substantial variation in the slopes. The SMR for the South Carolina plant reached a height of 8.33 at a median cumulative exposure of 215 f/mL-years in contrast to 4.16 at 168 f/mL-years at the Pennsylvania plant and 2.22 at 259 f/mL-years at the London plant.
Figure 5 shows the lung cancer SMR curve for the 1997 report on the Quebec miner and miller cohort. It should be noted that the exposure levels are much higher for this cohort than for the other seven cohorts: the maximum SMR was only 2.97 at a cumulative exposure of 4,200 f/mL-years. This curve is of particular interest because the cohort was one of the largest reported in the literature. Of special note is the observation that the lung cancer SMRs at the three lowest cumulative exposure levels are low and irregular: 1.13 at 45 f/mL-years, 1.38 at 195 f/mL-years, and 1.21 at 450 f/mL-years. For these three strata combined, the SMR was 1.21. Liddell et al said that this small elevation in the low exposure range was probably due to cigarette smoking.
Smoking habits were collected for cohort members who were alive in 1976. Dr. Liddell kindly supplied the age distribution of the cohort in that year and the percent distribution by smoking habit in each age group (personal communication, May 1997). The prevalence of current cigarette smoking was 66.2% in the age group 56 to 65 years and 46.5% in the age group [is greater than or equal to] 66 years. Somewhat similar data were obtained for Quebec men in 1975, giving prevalences of 55.8% for age group 45 to 64 and 37.4% for age group [is greater than or equal to] 65 years. These figures were used to calculate the ratio of observed to expected prevalences in the Quebec cohort: 1.21, identical to the SMR of 1.21 described above and thus confirming the opinion of Liddell et al that cigarette smoking accounts for the small increase in lung cancer risk in the first three exposure categories.
This observation supports the concept of a threshold for lung cancer risk, especially since 72% of the 587 lung cancer deaths occurred at these low cumulative asbestos exposure levels.
The exposure-response curve for the asbestosis mortality rate was plotted in Figure 5 in terms of the number per 100,000 person-years. The curve rises very smoothly from the very lowest to the highest exposure category, suggesting that there is a much lower threshold, if any, for asbestosis deaths than for lung cancer deaths. The rates rise from 12 to 399 per 100,000 person-years. Above a median exposure of 450 f/mL-years, the curves for lung cancer and asbestosis are almost identical. The two curves are based on disease measures that are not the same for lung cancer as for asbestosis. Nevertheless, the correlation is remarkable.
Only a few cohort studies have addressed directly the issue of asbestosis as a marker for increased risk of lung cancer among workers exposed to asbestos. What evidence exists supports the hypothesis that asbestosis is such a marker as reviewed in the first section above. Additional circumstantial evidence has been described in subsequent sections: (1) there is no excess risk of lung cancer in cohorts with no deaths from asbestosis; (2) workers with pleural plaques but no asbestosis have no increased risk of lung cancer in well-designed studies; and (3) the association between asbestosis and excess lung cancer rates is much stronger than the association between cumulative asbestos exposure and the relative risk of lung cancer.
The literature also contributes support for the hypothesis in two other lines of investigation: animal research and epidemiologic studies of lung cancer risk in other diseases characterized by diffuse pulmonary fibrosis.
To my knowledge, there is only one animal study that provides direct evidence on the, issue: an extensive investigation by Wagner et al. They exposed rats to various types of asbestos by inhalation. Animals are ordinarily exposed to asbestos in high concentrations, but Wagner et al varied the duration of exposure. One group of 201 rats was exposed for only 1 day and examined after survival for at least 600 days. Forty-four (22%) developed asbestosis and 11 (25%) of these 44 were found to have pulmonary tumors. The remaining rats without asbestosis had a tumor rate of 3.8% (6 of 157). A group of 126 control rats who survived at least 300 days had a tumor rate of 5.6%. While the data supplied specified different criteria for minimal survival time, the mean survival time was identical in exposed and control rats: 784 days in the 201 exposed rats and a total of 154 unexposed rats.
Among the 17 tumors in the exposed rats, 3 were adenocarcinomas and 14 were adenomas. The International Agency for Research on Cancer has established the principle that benign tumors are the result of exposure to carcinogens when they occur with malignant tumors of the same cell type in the same organ. In the Wagner et al study, exposures of 1 day produced three adenocarcinomas (their Table 7). In addition, there were two rats with mesothelioma.
An increased risk of lung cancer has been observed in humans with other diseases manifesting diffuse pulmonary fibrosis. Restricting the literature to cohort studies with estimates of RR for lung cancer leaves only a few reports to consider.
In 1980, Turner-Warwick et al. reported on 205 cases of cryptogenic fibrosing alveolitis seen at the Brompton Hospital in London between 1955 and 1973. These were followed-up for 4 to 21 years and 20 (9.8%) died from lung cancer; all but 2 of the 20 were smokers. The expected number of lung cancer cases was calculated from general population death rates by age and sex. The RR was 14.1, statistically highly significant. Four of the lung cancers were suspected by chest radiograph at the time of first hospital attendance (prevalence cases) while the rest were incidence cases; subtracting the prevalence cases gives an RR of 11.3, still statistically highly significant.
Two cohort studies reported in 1985 showed increased lung cancer risk among patients with systemic sclerosis (scleroderma). Peters-Golden et al followed-up 71 patients, diagnosed in 1972 to 1979 at a Baltimore hospital, until 1982 to 1983 and found 3 cases of lung cancer. The expected number was derived from the Surveillance, Epidemiology, and End Results (SEER) cancer registry rates for 1973 to 1977. The RR was 16.5, statistically highly significant. Two were nonsmokers. Two had radiographic evidence of pulmonary fibrosis and the other, a nonsmoker, had restrictive disease by pulmonary function testing.
Roumm and Medsger followed-up 262 patients diagnosed at the University of Pittsburgh in 1971 to 1982 and living in the Pittsburgh standard metropolitan statistical area (SMSA). The mean duration of follow-up was 4.3 years and this institution ascertained approximately 80% of the cases in the SMSA. Expected numbers of lung cancer were calculated from the Third National Cancer Survey (1975) adjusted by age and sex to the SMSA population. There was a 23% loss to follow-up but the lost ones were similar to those with complete follow-up. There was a statistically significant RR for lung cancer of 4.4 based on four cases; all were smokers and had pulmonary fibrosis by radiograph.
While the numbers of lung cancers in these two studies are small, the RRs are high and consistent with the findings at the Brompton Hospital. The possibility of selection bias, however, cannot be ruled out in hospital-based cohorts.
The disparity between asbestosis and cumulative asbestos exposure as predictors of excess lung cancer risk is striking. The poor consistency of fiber counts as predictors in various cohort studies may be due to a number of uncertainties. The fiber counts were generally derived from area samples rather than from individual personal samples. Conversion factors used to estimate historical fiber counts from total dust counts prior to the 1970s are of questionable validity. Variation in fiber dimensions, geometry, chemistry, integrity, and durability makes fiber counts problematic in the causation of disease as the only measure of exposure in different industrial environments. For these reasons, there is doubt about the relationship between exposure measurements and the effective dose actually delivered to the target tissues. In addition, variation in host factors and other environmental exposures may be confounding in the exposure-response relationship.
Another problem lies in the use of cumulative exposure estimates. Such estimates are derived from combining measures of intensity and duration of exposure. In the exposure-response curves shown in Figures 3-5 for the individual cohorts, some of the curves do show increase in lung cancer SMRs with increasing cumulative exposure. However, McDonald et al pointed out that "It has become increasingly evident that the linear relations that have been found between SMRs and cumulative exposure are an oversimplification." They indicated that there is a need to assess "the separate and combined effects of duration and intensity of exposure to asbestos, with appropriate allowance for a number of time related variables, and with due regard to cigarette smoking."
In particular, it should be recognized that duration of exposure overlaps the latency period of asbestos-related disease and this may create artificial linearity in the exposure-response relationship. Vacek and McDonald used multivariate RR models, which do not assume that exposure intensity and duration have equal effects on risk, to study lung cancer data from a cohort of vermiculite miners exposed to tremolite. The exposure-response pattern was S-shaped and differed substantially from the pattern using a cumulative exposure index.
It is well established that lung cancer among asbestos-exposed workers is unusual in the absence of cigarette smoking and that the exposure to both agents has a more than additive effect on the risk of lung cancer. This may be partially explained by increased retention of asbestos fibers in the airways of smokers. Cigarette smoking provides an additional link between asbestosis and lung cancer. There is evidence that smoking alone causes pulmonary interstitial fibrosis at a microscopic level and some evidence that this may be discernible as SIOs radiographically. The latter idea has been disputed with the suggestion that the SIOs associated with smoking may represent other pathologic findings such as bronchiolar wall thickening or an appearance created by emphysematous changes. The lack of adequate radiographic-pathologic studies makes interpretation of the radiographic changes uncertain.[71-74] In any case, the SIOs in smokers are generally of minimal degree, largely limited to profusions of 0/1 and 1/0.
However, there is more general agreement in the literature that smokers exposed to asbestos are at higher risk of developing radiographic evidence of asbestosis than are nonsmokers.[74,75] Thus, smoking is a factor in determining which asbestos-exposed individuals are diagnosed as having asbestosis and therefore is an effect modifier of the relationship between asbestosis and lung cancer risk.
However, there are more basic mechanisms that link the two diseases together. These have been elucidated by tissue, cellular, and molecular investigations.[76-78] Although the validity of extrapolation from such experimental research to explain human disease is uncertain, pathogenetic models may be constructed to improve our understanding.
Recent investigations indicate that inflammation is linked to neoplasia through several mechanisms after exposure to asbestos. Partial ingestion of long fibers by macrophages activates these cells to release substances such as lymphokines, growth factors, active oxidants, and proteases. Some of these may be genotoxic and others may cause cell proliferation. The latter increase opportunities for errors to occur during DNA replication, leading to malignant change and limiting repair of DNA damage induced by mutagens. Tissue culture studies have shown that bronchial epithelial cells ingest asbestos fibers. Asbestos fibers may adsorb polycyclic aromatic hydrocarbons (eg, those in cigarette smoke) and thus induce the aryl hydrocarbon hydroxylase system to produce metabolites that can interact with DNA. In addition, asbestos causes chromosomal abnormalities in mammalian cell systems.
Mossman has suggested two caveats concerning the interpretation of the above experimental results: (1) there is a lack of dose-response studies, and (2) comparative studies using both positive and negative controls are limited. Nevertheless, the results suggest a close link between bronchial cancer and the preceding inflammatory reaction to asbestos. This link is consistent with the hypothesis that lung cancer risk is elevated only in humans exposed to asbestos when there is asbestosis. That the increased risk is limited to those with radiologic evidence of asbestosis is supported by the available good epidemiologic evidence summarized in this review. More studies of greater magnitude are desirable, including cohorts with serial chest radiographic surveillance and histologic confirmation if possible.
The importance of the issue is both academic and medicolegal. Assuming that the resources available for the compensation of asbestos-exposed workers with lung cancer are limited, fairness would be enhanced by diverting funds from those with no increased risk of lung cancer to those who are at increased risk. The RR for lung cancer is so variable in the cohort studies reported that the attributable risk ranges from zero to a substantial proportion of the workers. A reliable marker for increased risk is therefore needed and asbestosis seems to satisfy this need.
Table 2--SMRs for Lung Cancer Among Dresden Asbestos Workers, 1952 to 1964, by Sex and Presence of Radiologic Asbestosis(*)
(*) Source of data: Jacob and Anspach.12
([dagger]) Based on Dresden age-, sex-, and period-specific death rates.
([double dagger]) Based on the Poisson distribution.
 Weiss W. Cigarette smoking and lung cancer trends; a light at the end of the tunnel? Chest 1997; 111:1414-1416
 Doll R. Mortality from lung cancer in asbestos workers. Br J Ind Med 1955; 12:81-86
 Browne K. Is asbestos or asbestosis the cause of the increased risk of lung cancer in asbestos workers? Br J Ind Med 1986; 43:145-149
 Churg A. Asbestos, asbestosis, and lung cancer. Mod Pathol 1993; 6:509-511
 Weiss W. Asbestos-related pleural plaques and lung cancer. Chest 1993; 103:1854-1859
 Jones RN, Hughes JM, Weill H. Asbestos exposure, asbestosis, and asbestos-attributable lung cancer. Thorax 1996; 51(suppl 2):S9-S15
 Roggli VL, Hammar SP, Pratt PC, et al. Does asbestos or asbestosis cause carcinoma of the lung? Am J Ind Med 1994; 26:835-838
 Abraham JL. Asbestos inhalation, not asbestosis, causes lung cancer. Am J Ind Med 1994; 26:839-842
 Hillerdal G, Henderson DW. Asbestos, asbestosis, pleural plaques and lung cancer. Scand J Work Environ Health 1997; 23:93-103
 Wilkinson P, Hansell DM, Janssens J, et al. Is lung cancer associated with asbestos exposure when there are no small opacities on the chest radiograph? Lancet 1995; 345:1074-1078
 Bohlig H, Jacob G, Muller H. Die asbestose der lunger. Stuttgart, Germany: Georg Thieme Verlag 1960:60. Cited by Sluis-Cremer GK. The relationship between asbestosis and bronchial cancer. Chest 1980; 78(suppl):380-381
 Jacob G, Anspach M. Pulmonary neoplasia among Dresden asbestos workers. Ann NY Acad Sci 1965; 132:536-548
 Saupe E. Roentgenatlas des asbestose der lungen. Leipzig: Thieme, 1938
 Knox JF, Holmes S, Doll R, et al. Mortality from lung cancer and other causes among workers in an asbestos textile factory. Br J Ind Med 1968; 25:293-303
 Liddell FDK, McDonald JC. Radiological findings as predictors of mortality in Quebec asbestos workers. Br J Ind Med 1980; 37:257-267
 Becklake MR, Thomas D, Liddell FDK, et al. Follow-up of respiratory measurements in Quebec chrysotile asbestos miners and millers. Scand J Work Environ Health 1982; 8(suppl 1):105-110
 Hughes JM, Weill H. Asbestosis as a precursor of asbestos related lung cancer: results of a prospective mortality study. Br J Ind Med 1991; 48:229-233
 McDonald JC, Liddell FDK, Gibbs GW, et al. Dust exposure and mortality in chrysotile mining, 1910-75. Br J Ind Med 1980; 37:11-24
 Thomas HF, Benjamin IT, Elwood PC, et al. Further follow-up study of workers from an asbestos cement factory. Br J Ind Med 1982; 39:273-276
 McDonald AD, Fry JS, Wooley AJ, et al. Dust exposure and mortality in an American chrysotile asbestos friction products plant. Br J Ind Med 1984; 41:151-157
 Peto J, Doll R, Hermon C, et al. Relationship of mortality to measures of environmental asbestos pollution in an asbestos textile factory. Ann Occup Hyg 1985; 29:305-355
 Gardner MJ, Winter PD, Pannett B, et al. Follow-up study of workers manufacturing chrysotile asbestos-cement products. Br J Ind Med 1986; 43:726-732
 McDonald JC, McDonald AD, Sebastien P, et al. Health of vermiculite miners exposed to trace amounts of fibrous tremolite. Br J Ind Med 1988; 45:630-634
 Tsai SP, Waddell LC Jr, Gilstrap EL, et al. Mortality among maintenance employees potentially exposed to asbestos in a refinery and petrochemical plant. Am J Ind Med 1996; 29:89-98
 Stewart PA, Schairer C, Blair A. Comparison of jobs, exposures, and mortality risks for short-term and long-term workers. J Occup Med 1990; 32:703-708
 Browne K. A threshold for asbestos-related lung cancer. Br J Ind Med 1986; 43:556-558
 Weiss W. Asbestos-related pleural plaques and lung cancer. Chest 1994; 105:1854-1859
 Hillerdal G. Pleural plaques, and risk for bronchial carcinoma and mesothelioma: a prospective study. Chest 1994; 105:144-150
 Tola S, Kalliomaki P-L, Pukkala E, et al. Incidence of cancer among welders, platers, machinists and pipe fitters in shipyards and machine shops. Br J Ind Med 1988; 45:209-218
 Acheson ED, Gardner MJ, Pippard EC, et al. Mortality of two groups of women who manufactured gas masks from chrysotile and crocidolite asbestos: a 40-year follow-up. Br J Ind Med 1982; 39:344-348
 Acheson ED, Gardner MJ, Winter PD, et al. Cancer in a factory using amosite asbestos. Int J Epidemiol 1984; 13:3-10
 Botta M, Magnani C, Tarracini B, et al. Mortality from respiratory and digestive cancers among asbestos cement workers in Italy. Cancer Detect Prev 1991; 15:445-447
 Brown DP, Dement JM, Okun A. Mortality patterns among female and male chrysotile asbestos textile workers. J Occup Med 1994; 38:882-888
 Enterline PE, Kendrick MA. Asbestos-dust exposures at various levels and mortality. Arch Environ Health 1967; 15:181-186
 Enterline PE, Hartley J, Henderson V. Asbestos and cancer: a cohort followed up to death. Br J Ind Med 1987; 44:396-401
 Giaroli C, Belli S, Bruno C, et al. Mortality study of asbestos cement workers. Int Arch Occup Environ Health 1994; 66:7-11
 Hodgson JT, Jones RD. Mortality of asbestos workers in England and Wales. Br J Ind Med 1986; 43:158-164
 Hughes JM, Weill H, Hammad YY. Mortality of workers employed in two asbestos cement manufacturing plants. Br J Ind Med 1987; 44:161-174
 Huilan Z, Zhiming W. Study of occupational lung cancer in asbestos factories in China. Br J Ind Med 1993; 50:1039-1042
 McDonald AD, Fry. JS, Wooley AJ, et al. Dust exposure and mortality in an American factory using chrysotile, amosite, and crocidolite in mainly textile manufacturing. Br J Ind Med 1983; 40:368-374
 McDonald JC, Liddell FDK, Dufresne A, et al. The 1891-1920 birth cohort of Quebec chrysotile miners and millers: mortality 1976-88. Br J Ind Med 1993; 50:1073-1081
 McDonald JC, McDonald AD, Armstrong B, et al. Cohort study of mortality of vermiculite miners exposed to tremolite. Br J Ind Med 1986; 43:436-444
 Meurman LO, Kiviluoto R, Hakama M. Mortality and morbidity among the working population of anthophyllite asbestos miners in Finland. Br J Ind Med 1974; 31:105-112
 Neuberger MM, Kundi M. Individual asbestos exposure; smoking and mortality--a cohort study in the asbestos cement industry. Br J Ind Med 1990; 47:615-620
 Newhouse ML, Berry G, Wagner JC. Mortality of factory workers in East London. Br J Ind Med 1985; 42:4-11
 Newhouse ML, Sullivan KR. A mortality study of workers manufacturing friction materials. Br J Ind Med 1989; 46:176-179
 Piolatto G, Negri E, LaVecchia C, et al. An update of cancer mortality, among chrysotile asbestos miners in Balangero, Northern Italy. Br J Ind Med 1990; 47:810-814
 Rossiter CE, Coles RM, HM Dockyard, Devonport: 1947 mortality study. In: Wagner JC, Davis W, eds. Biological effects of mineral fibres (vol 2): Proceedings of a Symposium, September 1979, Lyon, France. IARC Scientific Publication No 30. World Health Organization, International Agency for Research on Cancer, 1980; 713-721
 Seidman H, Selikoff IJ, Gelb SK. Mortality experience of amosite factory workers: dose-response relationships 5 to 40 years after onset of short-term work exposure. Am J Ind Med 1986; 10:479-514
 Sluis-Cremer GK, Liddell FDK, Logan WPD, et al. The mortality of amphibole miners in South Africa, 1946-80. Br J Ind Med 1992; 49:566-575
 Wignall BK, Fox AJ. Mortality of female gas mask assemblers. Br J Ind Med 1982; 39:34-38
 Weiss W. Heterogeneity in historical cohort studies: a source of bias in assessing lung cancer risk. J Occup Med 1983; 25:290-294
 Weiss W. Lung cancer due to chloromethyl ethers: bias in cohort definition. J Occup Med 1989; 31:102-105
 Steenland K, Tucker J, Salvan A. Problems in assessing the relative predictive value of internal markers versus external exposure in chronic disease epidemiology. Cancer Epidemiol Biomarkers Prev 1993; 2:487-491
 Lash TL, Crouch EAC, Green LC. A meta-analysis of the relation between cumulative exposure to asbestos and relative risk of lung cancer. Occup Environ Med 1997; 54:254-263
 Finkelstein MM. Mortality among employees of an Ontario asbestos-cement factory. Am Rev Respir Dis 1984; 129:754-761
 Dement JM, Brown DP, Okun A. Follow-up study of chrysotile asbestos textile workers: cohort mortality and case-control analyses. Am J Ind Med 1994; 26:431-447
 Liddell FDK, McDonald AD, McDonald JC. The 1891-1920 birth cohort of chrysotile miners and millers: development from 1904 and mortality to 1992. Ann Occup Hyg 1997; 41:13-36
 Henderson VL, Enterline PE. Asbestos exposure: factors associated with excess cancer and respiratory disease mortality. Ann NY Acad Sci 1979; 330:117-126
 Seidman H, Selikoff IJ, Hammond EC. Short-term asbestos work exposure and long-term observation. Ann NY Acad Sci 1979; 330:61-89
 Wheatley G. Smoking habits of Canadians, 1965 to 1979. Technical Report Series No. 9, December 1980. Health Promotion Directorate, Health Services and Promotion Branch, Health and Welfare Canada
 Wagner JC, Berry G, Skidmore JW, et al. The effects of inhalation of asbestos in rats. Br J Cancer 1974; 29:252-269
 International Agency for Research on Cancer. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Overall Evaluations of Carcinogenicity: An Updating of IARC Monographs Volumes 1 to 42, suppl 7. Lyon, France, 1987; 23
 Turuer-Warwick M, Lebowitz M, Burrows B, et al. Cryptogenic fibrosing alveolitis and lung cancer. Thorax 1980; 35:496-499
 Peters-Golden M, Wise RA, Hochberg M, et al. Incidence of lung cancer in systemic sclerosis. J Rheumatol 1985; 12:1136-1139
 US Dept of Health and Human Service: Surveillance, Epidemiology, and End Results: Incidence and Mortality Data 1973-77; National Cancer Institute Monograph 57, 1981
 Roumm AD, Medsger TA Jr. Cancer and systemic sclerosis. Arthritis Rheum 1985; 28:1336-1340
 Vacek PM, McDonald JC. Risk assessment using exposure intensity: an application to vermiculite mining. Br J Ind Med 1991; 48:543-547
 Vainio H, Boffetta P. Mechanisms of the combined effect of asbestos and smoking in the etiology of lung cancer. Scand J Work Environ Health 1994; 20:235-242
 Churg A, Stevens B. Enhanced retention of asbestos fibers in the airways of human smokers. Am J Respir Crit Care Med 1995; 151:1409-1413
 Weiss W. Cigarette smoke, asbestos, and small irregular opacities. Am Rev Respir Dis 1984; 130:293-301
 Meyer JD, Islam SS, Ducatman AM, et al. Prevalence of small lung opacities in populations unexposed to dusts. Chest 1997; 111:404-410
 Weiss W. Smoking and pulmonary fibrosis. J Occup Med 1988; 30:33-39
 Dick JA, Morgan WKC, Muir DFC, et al. The significance of irregular opacities on the chest roentgenogram. Chest 1992; 102:251-260
 Ducatman AM, Withers BF, Yang WN. Smoking and roentgenographic opacities in US Navy asbestos workers. Chest 1990; 97:810-813
 Rom WN, Travis WD, Brody AR. Cellular and molecular basis of the asbestos-related diseases. Am Rev Respir Dis 1991; 143:408-422
 Hesterberg TW, Hart GA, Bunn WB. In vitro toxicology of fibers: mechanistic studies and possible use for screening assays. In: Warheit DB, ed. Fiber toxicology. New York, NY: Academic Press, 1993; chap 7; 139-170
 Mossman BT. Carcinogenesis and related cell and tissue responses to asbestos: a review. Ann Occup Hyg 1994; 38:617-624
 Preston-Martin S, Pike MC, Ross RK, et al. Increased cell division as a cause of human cancer. Cancer Res 1990; 50:7415-7421
(*) From the MCP Hahnemann School of Medicine, Allegheny University of the Health Sciences, Philadelphia, PA.
Manuscript received December 23, 1997; revision accepted August 5, 1998.
Correspondence to: William Weiss, MD, 3912 Netherfield Rd, Philadelphia, PA 19129-1014
COPYRIGHT 1999 American College of Chest Physicians
COPYRIGHT 2000 Gale Group