Cohort Study of Silicon Carbide Production Workers

Transcription

1 American Journal of Epidemiology Vol. 4, No. Copyright 4 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.S.A All rights reserved Cohort Study of Silicon Carbide Production Workers Claire Infante-Rivard, Andre Dufresne, Ben Armstrong, Paul Bouchard, and Gilles Theriault Silicon carbide is produced by a chemical reaction at high temperature between free crystalline silica and petroleum coke. The process generates airborne fibers and fibrogenic dusts such as a-quartz and cristobalite, which are also potentially carcinogenic. The authors report that this is the first cohort study in this industry. The study was carried out among 55 Quebec silicon carbide production workers who had worked at any time from 5 to. Follow-up was to December,, and 67 deaths were observed. The standardized mortality ratio (SMR) for all causes of death was.5 (5% confidence interval (Cl).-.); for nonmalignant respiratory diseases it was. (5% Cl.-.); and for lung cancer it was.6 (5% Cl.-.5). Controlling for smoking status using a Cox regression analysis, the risk for nonmalignant respiratory diseases and for lung cancer increased with exposure to total dust; in the highest exposure category, rate ratios (RR) were 4. (5% Cl.-4.6) for nonmalignant respiratory diseases and.67 (5% Cl.57-4.) for lung cancer. Results were in the expected direction, but the power of the study was low, because of small sample size and use of cumulative total dust as the exposure variable, which may be a poor indicator of lung irritants and other potential carcinogens in this industry, notably silicon carbide ceramic fibers. Am J Epidemiol 4; 4:-5. dust; lung diseases; lung neoplasms; occupational exposure; silica; silicon Silicon carbide is mainly used as an abrasive and refractory material, but it is also used in the production of parts for electronic equipment. Major contaminants in the silicon carbide production industry include crystalline silica, which in the form of islets at the surface of silicon carbide crystals () could contribute to the pathogenesis Received for publication December 7,, and in final form July 5, 4. Abbreviations: Cl, confidence interval; ICD-, International Classification of Diseases, Ninth Revision; RR, rate ratio; SMR, standardized mortality ratio. Department of Occupational Health, Faculty of Medicine, McGill University, Montreal, Quebec, Canada. Centre de Sante Publique de Quebec, Regie Regionale de la Sante et des Services Sociaux de Quebec, Quebec, Canada. Reprint requests to Dr. Claire Infante-Rivard, Department of Occupational Health, McGill University, Pine Avenue West, Montreal, Province de Quebec, Canada HA A. of carborundum pneumoconiosis (, ), and is considered a probable human carcinogen (4), and silicon carbide fibers generated during silicon carbide crystal production (5), which have the characteristics for carcinogenic potency (6). To our knowledge, mortality from nonmalignant respiratory diseases and lung cancer has not been studied in this industry. We carried out a retrospective cohort mortality study in the three Quebec silicon production plants; data collected during a hygiene survey in two of the three plants () were also used to assess the relation between exposure and mortality. MATERIALS AND METHODS Mortality study Employers' lists of workers could not be accessed. However, all employees were

2 Infante-Rivard et al. unionized and unions provided complete seniority lists for the years 56, 6, 6, 7, and for one plant; for the second plant, lists were provided for the years 5, 67, yearly from 7 to 75, 7, and 7; and for the third plant, for the years 5, 56, every second year between 57 and 65, 6, 6, and every second year from 74 to. From these lists, Bouchard (7) identified a total of 7 workers who had worked between January, 5 and December, at one of the three silicon carbide plants operating in Quebec. Of these workers, 66 were not entered into the study because they had worked less than years at a study plant, leaving 6. A telephone interview took place with the study subject if alive or a next-of-kin respondent for the deceased to obtain job history and smoking information. A list of workers who could not be traced was given to local union representatives for consultation; when necessary, persons with similar names in the telephone book were phoned, and neighbors at the last known address were contacted. Finally, 4 workers (6. percent) could not be traced and were not kept in the study; fellow workers remembered most of the untraced workers as short-term workers. Four workers (.6 percent) refused to answer the questionnaire. The mortality analysis thus included 55 workers. In a first unpublished study (7), vital status up to December, was ascertained for all these workers through the Quebec population registry office, where underlying cause of death was coded by nosologists according to the International Classification of Disease, Ninth Revision (ICD-) (). We have now updated the vital status to December,. No effort was made to reach the 4 subjects who could not initially be traced. Using identical methods, vital status was ascertained for all study subjects who were not known dead at the end of. The production process The production process has been described by Smith et al. (). Acheson furnaces are heated by an electric current going through a graphite electrode that lies in a mixture of petroleum coke, pure crystalline silica, and sawdust; the reaction, which lasts about 5 hours at approximately,4 C, produces silicon carbide. Airborne contaminants that are generated in the process include sulfur dioxide, a-quartz, cristobalite, and polycyclic aromatic hydrocarbons. The process starts in the preparation area with the mixing of the raw materials, which are then brought to the furnace and charged by an overhead crane. After the heating cycle, the furnace is cooled; the sides are removed along with silicon carbide in large lumps, the graphite conductor (for reuse), and the unreacted material (old mix) for recycling. The latter operation was done manually prior to 66. Breaking up the lumps of silicon carbide and selecting silicon carbide according to purity is done on the cleaning floor with pneumatically powered chisels. The final product is then crushed and stored until shipment. Available data for exposure assessment From a hygiene survey carried out in 4, total dust samples from personal and fixed station samples were available for all job titles except for four with assumed very low exposure in plant A, and for approximately one-third of jobs in plant B (). None were available in plant C. Data were collected on respirable dust, quartz, cristobalite, and polycyclic aromatic hydrocarbons, but this information was not available for all jobs. It was therefore decided to use only total dust concentrations. Despite the fact that the characteristics of the fibers in this industry have been described by our group (), there were insufficient data to estimate fiber concentration by job. Engineering and work practice changes were described to us by former workers and knowledgeable government inspectors.

3 Mortality in the Silicon Carbide Industry Job-exposure matrix In the job-exposure matrix (table ), workers were classified according to job titles grouped under the main production TABLE. Job-exposure matrix* in a cohort of 55 silicon carbide production workers, Quebec, Canada, 5- Exposure area/job Outdoor area Yardman Sawdust Gateman Preparation area Coke preparation Sawdust and sand Mixer Furnace area Crane operator Furnace loader Fumaceman Electrode cleaner Payloader Assistant station operator Product area Carboselector Beltpicker Crusher operator Assistant crusher operator Building Bagger Miscellaneous Maintenance Supervisor Station operator Welder Truck driver Chemist Laborman Boilerman Office work Clerk Janitor Buyer No. of samples Total dust(mg/m ) Before and later A single mean was estimated when plant-specific means differed little; when they did, and they were based on a relatively large number of measures, plant-specific means were retained. process areas. Total dust concentrations from plant A were used for plant C, because dustiness between plants A and C was judged similar by health inspectors, and for plant B when for that plant no data were available. Data from plants A and B were pooled to estimate a single mean when plant-specific means differed little; when they did, and this was based on a relatively large number of measures, plant-specific means were retained. For the four unsampled job titles, a value was assigned from a job judged to have the most similar exposure. Estimates were based on dust samples. The mean exposure to total dust as estimated from samples of workers in each job category was assigned to all workers in that category. Exposure levels of furnacemen and floor cleaners, as measured in, were assigned to unloaders before 66. From 66, measured concentrations for a given job title were weighted according to the actual number of hours worked per day, which was often much less than hours. Total dust cumulative exposure is the sum of the products of exposure concentration and duration for each job held, and is expressed as milligrams per cubic meter years (mg/m -years); a baseline category of less than 5 mg/m -years corresponding to the 5th percentile of the distribution was defined, and the division between the two more highly exposed groups (75 mg/m - years) corresponds to the 75th percentile. Analysis Standardized mortality ratios (SMR) and 5 percent confidence intervals were estimated using the age- and calendar-specific death rates for Quebec males between and 5. Heterogeneity of stratumspecific mortality ratios was tested across age and calendar period strata to verify the assumptions necessary for SMRs to be valid summary ratios. The upper age limit for calculation of person-years and observed deaths was 5 years, at which age they were censored.

4 Infante-Rivard et al. The Cox regression analysis with timedependent covariates was used to assess the relation between cumulative exposure and the hazard of dying of lung cancer (EGRET package, Statistics and Epidemiology Corporation, Seattle, Washington). Survival time in the analysis was age at which a case died. The category of exposure used for comparison within risk sets when there is a failure (e.g., death by lung cancer) is the one in which those still at risk reside at the age at which the case dies. Analyses allowing for a minimum latent period were carried out by ignoring all deaths and years at risk in the first 5 years from starting employment in a study plant. RESULTS Mean age at hire in the cohort was. (standard deviation (SD) = 4.). Mean duration of follow-up was. years (SD =.). A total of,4 person-years were accumulated. There were 67 deaths in the cohort, but deaths were censored in the SMR analysis because they occurred after the age of 5 years. Table shows the SMRs and 5 percent confidence intervals (CI) for major causes of death. SMRs for all causes of death, diseases of the circulatory or digestive system, infectious diseases, and for all malignant diseases were not increased, whereas they were increased for stomach cancer (not significant), nonmalignant respiratory diseases (., 5 percent CI.-.), and lung cancer (.6, 5 percent CI.-.5). Among the 4 workers with lung cancer, were smokers at the time of the interview in 7, and were ex-smokers. Smoking status was unknown for subjects; 74 were reported smokers, were ex-smokers, and were nonsmokers. Table shows the results of the survival analysis for lung cancer and nonmalignant respiratory diseases in relation to cumulative exposure to total dust. Controlling for smoking, the risk of lung cancer increased with level of exposure (rate ratio (RR) =.4 for category in comparison with baseline, and.67 for category ). After a 5-year latency period, although somewhat lower, rate ratios also increased with exposure. There were deaths from all nonmalignant respiratory diseases (ICD- codes 46-5). Risk increased with exposure level and reached statistical significance in the highest category (RR = 4., 5 percent CI.-4.6). Rate ratios were increased after a 5-year latency period. Finally, of the deaths were from chronic obstructive and restrictive respiratory diseases (ICD- codes 4-5). For these, a rate ratio of. (5 percent CI.6-.6) was associated with the second TABLE. Standardized mortality ratios (SMR) and 5 percent confidence intervals (CI) for major causes of death in a cohort of 55 silicon carbide production workers, Quebec, Canada, 5- Cause of death Observed no. Expected no. SMR 5% CI p value (one-sided) All causes Malignant diseases (4-)* Stomach cancer (5) Lung cancer (6) Nonmalignant respiratory diseases (46-5) Diseases of the circulatory system (-45) Diseases of the digestive system (5-57) Infectious diseases (-) > >. >. >. International Classification of Diseases, th Revision (ICD-) code no. in parentheses.

5 Mortality in the Silicon Carbide Industry TABLE. Lung cancer and nonmalignant respiratory disease rate ratios (RR) and 5 percent confidence intervals (Cl) for cumulative exposure to total dust adjusted for smoking in a cohort of 55 silicon carbide production workers, Quebec, Canada, 5- Lung cancer 5-year latency Nonmalignant respiratory diseases 5-year latency Chronic nonmalignant respiratory diseases 5-year latency * Number of subjects, t Reference category. No.* 5 5 <5 RR category of exposure versus baseline, and those in the highest category of exposure had a rate ratio of. (5 percent Cl.6-5.). Again somewhat lower, the rate ratios after a 5-year latency period nevertheless increased with exposure. DISCUSSION No. Cumulative exposure levels (mg/m ) 5-75 RR % Cl No. RR >75 5% Cl The results of this study partly support the hypothesis of an increased risk of nonmalignant respiratory diseases and lung cancer among production workers in the silicon carbide industry. Because there have been no previous studies with which to compare these results, and because of the small size of the study, this conclusion needs to be seen as a first step in the assessment of mortality in this industry. In addition, the fact that only total dust exposure was assessed limits the interpretation of the results. The excess risks measured using external rates are more precise than the estimates obtained from internal comparisons in the Cox analysis. Of concern when using external rates is confounding, and in particular confounding by smoking. However, in this study, the differences in smoking rates between the cohort and the reference population are small: there were 6 percent of ever-smokers in the cohort (if the unknowns are conservatively assumed smokers) versus percent in a similar age cohort of men in the comparison population (). We thus conclude that smoking explains only a very small excess risk. Quantitatively assuming smokers to be times more likely to die from lung cancer than nonsmokers, the lung cancer SMR due to smoking is calculated at.5 according to Axelson's method (), giving a smokingadjusted SMR of.6 (5 percent Cl.4-.4). Although rate ratios increased with exposure in the Cox analysis and the latency analysis did not add to the evidence, the interpretation of results with the cumulative exposure to total dust is more difficult. It is likely that nondifferential misclassification of exposure to total dust leading to a bias toward the null was present. This is a general comment that applies to most retrospective exposure assessments, but it may be compounded in this industry by the hygiene measurement problems, which do not seem completely resolved. The few published surveys show that for similar jobs, there is a difference between surveys of more than one order of magnitude in the measurement of respirable dust (); the surveys have also reported substantial differences in the type and in the airborne concentrations of crystalline silica. In the Quebec plants, the concentrations of quartz and cristobalite measured were much below the ones considered to entail

6 4 Infante-Rivard et al. health risks; quartz concentrations ranged from zero to micrograms per cubic meter (/xg/m ) and cristobalite concentrations from zero to 6 /Ag/m (). However, it was observed that at the surface of silicon carbide crystals there were islets of crystalline silica, heterogeneously distributed and covering a portion of the silicon carbide particles (). Although silicon carbide particles are not readily soluble in pulmonary fluids, the cell membrane of the alveolar macrophage is placed in contact with surface layers that include silica. Using the sheep model, Perrault et al. (4) reported that silicon carbide covered with cristobalite increased fibronectin production and total phospholipids considered as indicators of fibrogenicity that could lead to pneumoconiosis or cancer. Other hygiene measurement problems in this industry relate to the fibers that are generated in the process of silicon carbide production (5,, ). According to our data (), the geometric mean of length to diameter ratio of silicon carbide fibers lie between those for amosite and tremolite asbestos fibers, and are of greatest carcinogenic potency (6). Moreover, preliminary data show that these fibers have the potential of remaining in the lung a very long time (5). However, fiber production does not appear to be related to total dust concentration. Scansetti et al. () found that peak fiber concentration was reached during removal of unreacted material already shown to produce the highest quartz concentrations, but this is not the dustiest operation. In the three Norwegian plants studied by Bye et al. (5), the highest fiber concentration was reached when mixing the raw material in plant C, at the furnace in plant A, and during removal of the raw material in plant B. As for other carcinogens of concern, we could not measure asbestos fibers in this industry () nor in the lungs of deceased workers (6). In addition, polycyclic aromatic hydrocarbons could only be detected at the head of one of the Acheson furnaces but their concentration away from the head of the furnace and in the ambient air rapidly decreased, possibly by dilution and/or adsorption to graphite dust particles. For all dead subjects, smoking and job histories were obtained from next-of-kin; we believe that the latter reported smoking status as correctly as the workers themselves. However, misclassification in the reporting of job title by the next-of-kin is more likely, although the majority of workers in the study spent a lifetime in this industry where there was only a limited number of distinct jobs. In conclusion, all results in this study were in the expected direction, but because of limited sample size, firm conclusions could not be reached. There is still incomplete understanding of the contaminants in this industry. However, some results from our investigation and from other studies suggest that traditional indicators of exposure may not fully account for toxicity of the silicon carbide industry. This could have led to an underestimation of the risks in this study. Further study of longterm effects of silicon carbide production will likely contribute to our understanding of the health effects of manmade mineral fibers. ACKNOWLEDGMENTS This study was supported by a grant from the Institut de Recherche en Sante et Securite du Travail du Quebec (IRSST). REFERENCES. Dufresne A, Lesage J, Perrault G. Evaluation of occupational exposure to mixed dusts and polycyclic aromatic hydrocarbons in silicon carbide plants. Am Ind Hyg Assoc J 7;4:6-6.. Masse S, Begin R, Cantin A. Pathology of silicon carbide pneumoconiosis. Mod Pathol ; :4-.. Durand P, Bdgin R, Samson L, et al. Silicon carbide pneumoconiosis: a radiographic assessment. Am J Ind Med ;: International Agency for Research on Cancer. Silica and some silicates. IARC monographs on the evaluation of the carcinogenic risk of chemicals in humans. Lyon: International Agency for Research on Cancer, 6;4:. 5. Bye E, Eduard W, Gjonnes J, et al. Occurrence of airborne silicon carbide fibers during indus-

7 Mortality in the Silicon Carbide Industry 5 trial production of silicon carbide. Scand J Work Environ Health 5;: Stanton MF, Layard M, Tegeris A, et al. Relation of particle dimension to carcinogenicity in araphibole asbestoses and other fibrous materials. J Natl Cancer Inst ;67: Bouchard P. Etude de la mortalite des travailleurs du carbure de silicium du Quebec, 5-. MSc dissertation. Quebec: Universite Laval, 7.. International Classification of Diseases: th revision. US Department of Health and Human Services, Public Health Service, Health Care Financing Administration. (DHHS publication no. (PHS) -6).. Smith TJ, Hammond SK, Laidlaw F, et al. Respiratory exposures associated with silicon carbide production: estimation of cumulative exposures for an epidemiological study. Br J Ind Med 4;4:-.. Dufresne A, Perrault G, Sebastien P, et al. Morphology and surface characteristics of particulates from silicon carbide industries. Am Ind Hyg Assoc J 7;4:77-.. Millar W. The smoking behavior of Canadians 6. Ottawa: Ministry of Supplies and Services, Health and Health and Welfare Canada,. (Catalogue H-66/E).. Axelson O, Steenland K. Indirect methods of assessing the effects of tobacco use in occupational studies. Am J Ind Med ;:5-.. Scansetti G, Piolatto G, Botta GC. Airborne fibrous and non-fibrous particles in a silicon carbide manufacturing plant. Ann Occup Hyg ; 6: Perrault G, Dufresne A, Sebastien P, et al. Experimental pulmonary activity of crystalline silica at the surface of inert silicon carbide. Preliminary results. Paper presented at the th International Conference on Occupational Lung Diseases, Prague, Czechoslovakia, September. 5. Dufresne A, Loosereewanich P, Harrigan M, et al. Pulmonary dust retention in a silicon carbide worker. Am Ind Hyg Assoc J ;54: Dufresne A, Loosereewanich P, Armstrong B, et al. Pulmonary retention of ceramic fibers in silicon carbide workers. Paper presented at the American Industrial Hygiene Conference & Exposition, Anaheim, CA, May 4.