Mortality from lung cancer in workers exposed to sulfur dioxide in the pulp and paper industry.

Our objective in this study was to evaluate the mortality of workers exposed to sulfur dioxide in the pulp and paper industry. The cohort included 57,613 workers employed for at least 1 year in the pulp and paper industry in 12 countries. We assessed exposure to SO(2) at the level of mill and department, using industrial hygiene measurement data and information from company questionnaires; 40,704 workers were classified as exposed to SO(2). We conducted a standardized mortality ratio (SMR) analysis based on age-specific and calendar period-specific national mortality rates. We also conducted a Poisson regression analysis to determine the dose-response relations between SO(2) exposure and cancer mortality risks and to explore the effect of potential confounding factors. The SMR analysis showed a moderate deficit of all causes of death [SMR = 0.89; 95% confidence interval (CI), 0.87-0.96] among exposed workers. Lung cancer mortality was marginally increased among exposed workers (SMR = 1.08; 95% CI, 0.98-1.18). After adjustment for occupational coexposures, the lung cancer risk was increased compared with unexposed workers (rate ratio = 1.49; 95% CI, 1.14-1.96). There was a suggestion of a positive relationship between weighted cumulative SO(2) exposure and lung cancer mortality (p-value of test for linear trend = 0.009 among all exposed workers; p = 0.3 among workers with high exposure). Neither duration of exposure nor time since first exposure was associated with lung cancer mortality. Mortality from non-Hodgkin lymphoma and from leukemia was increased among workers with high SO(2) exposure; a dose-response relationship with cumulative SO(2) exposure was suggested for non-Hodgkin lymphoma. For the other causes of death, there was no evidence of increased mortality associated with exposure to SO(2). Although residual confounding may have occurred, our results suggest that occupational exposure to SO(2) in the pulp and paper industry may be associated with an increased risk of lung cancer.

Pulp and paper production workers are exposed to a number of hazardous chemicals, and several studies have been conducted that suggest a possible health effect of such exposures. However, previous studies did not usually include an assessment of exposure to specific agents (Toren et al. 1996). Exposure circumstances in pulp and paper mills are complex; it is therefore difficult to identify agents possibly responsible for an adverse health effect from the results of studies based on employment in a given department or mill.
Sulfur dioxide is a common chemical exposure in the pulp production part of the pulp and paper industry, and levels often exceed 2 ppm (Kauppinen et al. 1997). SO 2 is also a major air pollutant suspected to increase mortality from respiratory diseases in the general population (Hoek et al. 2001;Lee et al. 2000;Shinkura et al. 1999;Xu et al. 1994) and to act as a promoter or cocarcinogen (Nisbet et al. 1984). Some early studies (Lee and Fraumeni 1969;Rencher et al. 1977) in SO 2 -exposed workers showed an increased mortality from lung cancer, but in those studies exposure to SO 2 occurred together with exposure to known or suspected carcinogens such as arsenic. Previous studies have reported an increased mortality from lung cancer among workers employed in the pulp and paper industry and, in particular, in sulfite pulp manufacture and maintenance (Band et al. 2001;Langseth and Andersen 2000;Szadkowska-Stanczyk and Szymczak 2001;Toren et al. 1991). It was suggested that asbestos, dust, or chlorinated compounds could be among the agents responsible for the increased lung cancer mortality, but no formal attempt was made in these studies to assess exposure to specific agents.
The International Agency for Research on Cancer has coordinated an international cohort study of workers in the pulp and paper industry to investigate patterns of cancer incidence and mortality. Results based on national components of the multicenter cohort have been reported (Andersson et al. 1998;Fassa et al. 1998;Henneberger and Lax 1998;Henneberger et al. 1989;Jäppinen and Pukkala 1991;Jäppinen and Tola 1986;Andersen 1999, 2000;Rix et al. 1997Rix et al. , 1998Sala-Serra et al. 1996;Szadkowska-Stanczyk and Szymczak 2001;Szadkowska-Stanczyk et al. 1997;Wild et al. 1998). Assessment of exposure to specific agents was conducted by an international panel of industrial hygienists (Kauppinen et al. 1997).
In this study we present an evaluation of the mortality of SO 2 -exposed workers employed in the pulp and paper industry.

Materials and Methods
We included those workers with at least 1 year of employment in the pulp and paper industry. We excluded countries in which no workers were classified as exposed to SO 2 , as well as workers with unknown SO 2 exposure status. A total of 57,613 subjects from Brazil, Denmark, Finland, France, Japan, New Zealand, Norway, Poland, South Africa, Spain, Sweden, and the United States were included in the analysis (51,240 men and 6,373 women); they contributed 1,249,406 personyears of observation from 1945 to 1996. Their distribution by SO 2 exposure status and country is shown in Table 1. Norway provided the largest number of SO 2 -exposed workers (29.8%), followed by Finland (16.3%), Poland (13.1%), and New Zealand (12.7%).
Workers were followed up for mortality according to procedures specific to each country.

Articles
Our objective in this study was to evaluate the mortality of workers exposed to sulfur dioxide in the pulp and paper industry. The cohort included 57,613 workers employed for at least 1 year in the pulp and paper industry in 12 countries. We assessed exposure to SO 2 at the level of mill and department, using industrial hygiene measurement data and information from company questionnaires; 40,704 workers were classified as exposed to SO 2 . We conducted a standardized mortality ratio (SMR) analysis based on age-specific and calendar period-specific national mortality rates. We also conducted a Poisson regression analysis to determine the dose-response relations between SO 2 exposure and cancer mortality risks and to explore the effect of potential confounding factors. The SMR analysis showed a moderate deficit of all causes of death [SMR = 0.89; 95% confidence interval (CI), 0.87-0.96] among exposed workers. Lung cancer mortality was marginally increased among exposed workers (SMR = 1.08; 95% CI, 0.98-1.18). After adjustment for occupational coexposures, the lung cancer risk was increased compared with unexposed workers (rate ratio = 1.49; 95% CI, 1.14-1.96). There was a suggestion of a positive relationship between weighted cumulative SO 2 exposure and lung cancer mortality (p-value of test for linear trend = 0.009 among all exposed workers; p = 0.3 among workers with high exposure). Neither duration of exposure nor time since first exposure was associated with lung cancer mortality. Mortality from non-Hodgkin lymphoma and from leukemia was increased among workers with high SO 2 exposure; a dose-response relationship with cumulative SO 2 exposure was suggested for non-Hodgkin lymphoma. For the other causes of death, there was no evidence of increased mortality associated with exposure to SO 2 . Although residual confounding may have occurred, our results suggest that occupational exposure to SO 2 in the pulp and paper industry may be associated with an increased risk of lung cancer. Key words: epidemiology, lung neoplasms, mortality, pulp and paper industry, among countries, but in most cases it was between the early 1950s and the mid 1990s. In the whole database, 2% of workers were lost to follow-up and 1% emigrated. Causes of death were either abstracted from death certificates or obtained from mortality registries and coded according to the International Classification of Diseases, Revision 9 (ICD-9, 1975).
We reconstructed exposure estimates for SO 2 for each mill and department included in the study and for different time periods, using international industrial hygiene measurement data (both from mills included in the study and from nonparticipating European and North American mills), information from detailed questionnaires about raw materials and production processes submitted by each participating mill, and the experience of the assessment team. SO 2 -exposed workers were employed primarily in the following departments: pulp production (sulfite and kraft processes), pulp production from recycled paper, paper and paperboard production, and nonproduction (e.g., maintenance). Workers involved in the manufacture of paper and paperboard products were not exposed to SO 2 .
We estimated prevalence and level of exposure for each department in each of the 52 participating mills for every time period in which relevant production conditions appeared to remain constant. The prevalence of exposure referred to the proportion of workers in the department exposed to the agent on an average workday and was categorized as very low (< 5% of workers in the department exposed, coded as 0.025), low (5-50%, coded as 0.25), high (51-95%, coded as 0.75), and very high (> 95%, coded as 0.975). The level of exposure referred to the mean level of exposure at work averaged over the work year among the exposed workers and was categorized as 0 (mean value = 0.15 ppm), 1 (0.5 ppm), 2 (1.5 ppm), and 3 (5 ppm). We defined a group with high exposure as including workers in the two upper categories of both level and prevalence of SO 2 exposure; the high-exposure group included 2,495 workers, providing 56902.6 person-years of observation. We constructed several SO 2 exposure variables, all categorized into quartiles for the statistical analysis: duration of exposure (< 4, 4-12, 13-24, ≥ 25 years), time since first exposure to SO 2 (< 18, 18-28, 29-38, ≥ 39 years), cumulative exposure [Σ level (mean value) × duration; < 23, 23-61, 62-127, ≥ 128 ppm-years], and weighted cumulative exposure [Σ prevalence (code) × level × duration: < 14, 14-38, 39-90, ≥ 91 ppm-years]. Cumulative and weighted cumulative exposures were calculated for all exposed workers as well as for workers in the highexposure group. Workers with potential exposure but unknown prevalence or level were considered exposed but excluded from the calculation of cumulative exposure.
Standardized mortality ratios (SMRs) were calculated as the ratio of observed to expected deaths. We computed expected deaths by multiplying the person-years in each sex-specific, age-specific, and 5-year calendar periodspecific stratum by the national reference rates using the Person Years program (Coleman et al. 1986). National rates were derived from the World Health Organization (WHO) Mortality Database (WHO 2001). Ninety-five percent confidence intervals (CIs) of the SMRs were calculated under the assumption that the observed numbers of deaths follow a Poisson distribution. We performed tests for linear trend in SMRs using a method described by Breslow and Day (1987).
Expected deaths were not available for the South African cohort. In preliminary analyses, the overall SMR in the Brazilian cohort was < 0.5, suggesting possible underascertainment of deaths. These two national components were excluded from the SMR analysis.
We used Poisson regression analysis to examine internal dose-response relations and to explore the effect of potential confounding factors. Rate ratios (RRs) and 95% CIs derived from the analysis were adjusted for P-Y, person-years. a High exposure is a subset of ever exposed. Obs, observed. Cohorts from Brazil and South Africa were excluded from the SMR analysis. a High exposure is a subset of ever exposed.
country, sex, age, calendar period, and employment status (i.e., whether person-years accumulated while workers were employed in the companies included in the study or not). The reference group for each RR was the first level of each variable. Because there was potential confounding among occupational exposures, we also adjusted the RRs for coexposure to other potential carcinogens (acid, asbestos, volatile organochlorine compounds, combustion products, organic dyes, epichlorohydrine derivatives, formaldehyde, pulp and paper dust, reduced organic sulfur compounds, talc, welding fumes, wood dust). In the final models, however, we retained a shorter list of potential confounders. Adjustment for lifestyle factors such as tobacco smoking was not possible.

Results
In this cohort 7,613 deaths occurred among SO 2 -exposed workers, including 488 lung cancer deaths. The SMR analysis, based on 7,508 deaths from 10 countries, yielded a deficit (SMR = 0.89; 95% CI, 0.87-0.91) in mortality compared with national rates (Table 2). We also observed a deficit in mortality for all malignant neoplasms (SMR = 0.91; 95% CI, 0.87-0.96; based on 1,756 deaths) and for cancers of the esophagus (SMR = 0.57; 95% CI, 0.38-0.83; 27 deaths) and stomach (SMR = 0.81; 95% CI, 0.69-0.94; 172 deaths). Lung cancer mortality was slightly increased (SMR = 1.08; 95% CI, 0.98-1.18; 482 deaths). The results of the analysis of gender-specific mortality did not reveal any particular feature, with the results among women being based on a relatively small number of deaths (not shown in detail). Among workers classified in the high-SO 2 -exposure group, there was an increased mortality from lung cancer and, although not statistically significant, from non-Hodgkin lymphoma and leukemia (Table 2). Table 3 shows the results of the comparisons of SO 2 -exposed and unexposed workers without and with adjustment for estimated exposure to known or suspected occupational carcinogens. The RRs of bronchitis, emphysema, and asthma were also decreased, whereas those of lung cancer (RR = 1.49; 95% CI, 1.14-1.96) and non-Hodgkin lymphoma (RR = 2.55; 95% CI, 1.06-6.13) increased. Results after further adjustment for exposure to other agents, such as formaldehyde, organochlorine compounds, and pulp and paper dust, were similar to those presented for RR2 in Table 3, although the precision of the RRs was decreased (not shown in detail).
The results on mortality from selected causes of death according to weighted cumulative SO 2 exposure are reported in Table 4.
The overall cancer mortality increased significantly with weighted cumulative exposure, no matter whether the analysis included all workers or only those classified in the highexposure group. A trend was also suggested for mortality from stomach cancer, lung cancer, and non-Hodgkin lymphoma, although it was significant only for the latter two neoplasms when all workers were retained in the analysis. Results on mortality from all cancers combined were driven by the increased mortality from lung cancer. When the latter neoplasm was excluded, the RRs for increasing levels of weighted cumulative exposure were (for the categories reported in Table 4) 1.02 (95% CI, 0.82-1.28), 1.23 (95% CI, 0.98-1.55), and 1.24 (95% CI, 0.98-1.58). Mortality from nonneoplastic respiratory diseases decreased-although not significantly so-with increasing weighted cumulative exposure to SO 2 . A similar analysis for other causes of death did not suggest any association. An analysis that did not consider estimated prevalence of exposure (i.e., based on cumulative exposure instead of weighted cumulative exposure) yielded results very similar to those reported in Table 4.
We found no trend between either duration of exposure or time since first exposure and mortality from the causes reported in Table 4 (not shown in detail).
The analyses of the effect of combined exposure between SO 2 and other occupational agents on lung cancer mortality are presented in Table 5. There was a suggestion of an interaction between SO 2 and welding fumes but not between SO 2 and either asbestos or combustion products.

Discussion
The main result of this cohort study was an association between SO 2 exposure and mortality from all neoplastic diseases and lung cancer. In the case of lung cancer, a marginally increased mortality compared with unexposed workers was significantly increased after Articles • Sulfur dioxide and lung cancer Environmental Health Perspectives • VOLUME 110 | NUMBER 10 | October 2002  adjustment for exposures to lung carcinogens. In addition, internal comparisons showed lung cancer mortality elevated 2-fold among workers in the highest category of cumulative SO 2 exposure compared with workers in the lowest exposure category. The lack of an association between lung cancer mortality and duration of SO 2 exposure can be explained by variability in exposure levels across time and country, making duration of exposure a poor indicator of total dose. These findings suggest that SO 2 exposure in the pulp and paper industry may contribute to lung carcinogenesis. The evidence of a genotoxic effect of SO 2 in experimental systems is limited (IARC 1992). Groups of workers exposed to SO 2 in Sweden (Nordenson et al. 1980) and in China (Meng and Zhang 1990) have been shown to have significantly increased frequency of chromosomal aberrations. Additional nongenotoxic mechanisms through which SO 2 might exert a carcinogenic effect on the lung include slowing of mucociliary clearance, impairment of alveolar macrophage function, and other effects on the immune response such as increased epithelial permeability, which would facilitate absorption of carcinogenic components of particulate matter (Beeson et al. 1998). Even though the molecular basis of SO 2 carcinogenicity is unclear, Leung et al. (1985) and Menzel et al. (1986) suggested that SO 2 may affect the detoxification of xenobiotic compounds by inhibiting the enzymatic conjugation of glutathione and reactive electrophiles. Because glutathione conjugation represents the major pathway of elimination of benzopyrene epoxides in the lung, their results offered a possible explanation for the cocarcinogenicity of SO 2 in combination with polycyclic aromatic hydrocarbons.
Following early observations by Peacock and Spence (1967) of an increased incidence of lung cancer in mice, Ohyama et al. (1999) reported an increased incidence of lung cancer in rats exposed to SO 2 . In a study of chemical workers exposed to SO 2 , Bond et al. (1986) reported a significant association between lung cancer mortality and SO 2 exposure, for which there was a significant dose-response relationship. Results of two general population studies, the American Cancer Society Study (Pope et al. 1995) and the Adventist Health Study (Abbey et al. 1999;Beeson et al. 1998) suggested a positive association between SO 2 exposure as an air pollutant and increased lung cancer mortality.
An important limitation of the present study is the lack of information on potential lifestyle confounders, chiefly tobacco smoking. Smoking is a well-known potential confounder in studies of lung cancer. Although smoking habits in the cohort are not known, there are indirect approaches to consider whether smoking might be an important confounder in our study. Jäppinen and Tola (1986) surveyed smoking habits in the Finnish component of this study and reported that smoking habits did not differ substantially from those of the national population. According to Axelson (1978), smoking habits in various industrial populations rarely diverge so much that the confounding effect of smoking distorts the risk ratios of lung cancer outside the range of 0.5-1.5. The simple comparisons of risk between SO 2 -exposed and unexposed workers were in this range, but the analyses of cumulative exposure gave relative risks above 1.5. In addition, we did not find an increased mortality from smoking-related diseases other than lung cancer, such as chronic bronchitis and bladder cancer. Case-control studies conducted within the pulp and paper industry provided evidence against a confounding effect of smoking (Henneberger and Lax 1998). A further argument against substantial confounding by smoking is the presence of dose-response relationship within the group of workers exposed to SO 2 .
We attempted to control for the possible effect of other occupational exposures, such as asbestos. However, similar to SO 2 exposures, these exposures were assessed at the level of department and therefore were likely to be subject to substantial misclassification, leading to possible residual confounding.
The assessment of exposure was carried out by industrial hygienists who were familiar with the pulp and paper industry, although not with all of the mills included in our study. It is likely, therefore, that some misclassification of exposure occurred. Furthermore, work histories were available only at the department level and for the period of employment in the mills under study. If the exposure among workers in a department is not homogeneous, then unexposed workers are classified as potentially exposed (and vice versa), resulting in a tendency to underestimate the risk, if there is one. We addressed the potential misclassification of exposure by repeating the doseresponse analysis after restriction of the study population to workers with high exposure.
As in most cohort studies of industrial workers, a deficit in overall mortality was found in our study in the SMR comparisons with the national populations. This is a common occurrence in occupational investigations known as the "healthy worker effect," a combination of several factors associated with employment such as selection of the work force and changes in lifestyle accompanying employment (Monson 1986;Wen et al. 1983).
Death from nonneoplastic respiratory disease was not increased in SO 2 -exposed workers. It is possible that the SO 2 exposure is not sufficiently high to cause nonmalignant respiratory diseases that are severe enough to lead to death. In addition, susceptible persons with respiratory disease may not seek employment at the mills or may quit employment because of possible symptoms or disease. Such selection procedures tend to underestimate the risk of nonneoplastic respiratory disease mortality in industrial cohorts. The possible decreased trend with increasing estimated exposure suggests a possible depletion of susceptible individuals from the groups with highest exposures.
Mortality from stomach cancer was nonsignificantly increased among workers with high cumulative SO 2 exposure in the highexposure group: the lack of a corresponding increase in the SMR analysis suggests a noncausal interpretation (e.g., confounding by another carcinogenic exposure). Because stomach cancer mortality shows important geographical variations, we looked at countryspecific SMRs: we could not find an indication of an association with SO 2 exposure in either high-risk countries (e.g., Japan, Spain) or low-risk countries (e.g., United States, Sweden). Mortality from non-Hodgkin lymphoma was elevated among workers classified in the high-SO 2 -exposure group but not among other exposed workers. Although the excesses of mortality from non-Hodgkin lymphoma and stomach cancer seem less convincingly related to SO 2 exposure than that of lung cancer, they suggest that further studies are warranted. Stomach cancer risk was increased in previous studies of pulp and  paper production workers (Rix et al. 1997;Robinson et al. 1986;Wingren et al. 1991). In summary, our findings are compatible with the hypothesis that exposure to SO 2 in the pulp and paper industry is associated with an increased risk of lung cancer, especially in high-exposure groups. Although confounding, particularly from smoking, may have been occurred, our results are compatible with the notion that SO 2 may have a cancerpromoting effect when it occurs in combination with other carcinogens in the pulp and paper industry.