Pooled reanalysis of cancer mortality among five cohorts of workers in wood-related industries

Pooled reanalysis of cancer mortality among five cohorts of workers in wood-related industries. Scand J Work Environ Health 1995;21:179-90. Objectives To provide more information regarding the risk of cancer associated with wood dust, a pooled reanalysis of data from five cohort studies was performed. Methods The combined cohort consisted of 28 704 persons from five studies: British furniture workers, members of the union representing furniture workers in the United States, two cohorts of plywood workers, and one of wood model makers, among whom 7665 deaths occurred. Pooled analyses were carried out for all of the cohorts combined, the two furniture worker cohorts combined, and the two plywood workers cohorts combined. Results Significaut excesses of nasal [observed 11, standardized mortality ratio (SMR) 3.1, 95% confidence interval (95% CI) 1.6-5.61 and nasopharyngeal (observed 9, SMR 2.4, 95% CI 1.1-4.5) cancer were observed. That for nasal cancer appeared to be associated with exposure to wood dust but was based solely on cases from the British furniture worker cohort, while that of nasopharyngeal cancer was observed for furniture and plywood workers and was associated with both high and low probability of wood dust exposure. Some support for an excess risk of multiple myelorna was also observed but was less clearly associated with wood dust exposure. No excesses of lung, larynx, stomach, or colon cancer were found to be associated with any surrogate indicators of wood dust exposure. C O ~ C ~ U S ~ O ~ S Workers exposed to wood dust may have an excess risk of nasopharyngeal cancer and multiple myeloma in addition to sinonasal cancer. The limitations of this study would tend to obscure relationships, rather than create false positive findings.

tively rare, and most cohort studies may have lacked sufficient power to determine whether or not an excess risk exists. To provide more information regarding the risk of cancer associated with wood dust, we assembled data from recently completed or updated cohort mortality studies of wood workers for a pooled analysis and developed a classification scheme for wood dust exposure for the pooled cohort based on that used in the British study. Although many types of exposure occur within wood-related industries, the primary purpose of these analyses was to assess the relationship between the risk of cancer and exposure to wood dust. The analyses focused on suspected cancer sites of a priori suspicion as indicated by the results of previous studies.

Participating studies
The five recently completed or updated cohort mortality studies included in this pooled analysis are listed in table 1 with the entrance criteria used in the original study and the follow-up period available for our analysis. The entrance criteria used for the original studies were also used for our analysis. The follow-up periods for two cohorts, the British (8) and American furniture workers (10, 1 I), have recently been updated, and the extended follow-up periods were used. A brief description of the cohorts is provided below, but further information regarding the design aspects of the studies can be found in reports published earlier (8,(10)(11)(12)(13)(14)(15).
The cohort of British furniture workers consisted of male workers born prior to 1941 and employed by nine factories in the High-Wycombe area of England prior to 1969. The most recent published follow-up of this cohort was through the end of 1982 (8). Nine additional years of mortality follow-up were performed for our analysis (unpublished). The second cohort consisted of members of the United Furniture Workers of America (UFWA) trade union who began employment between 1946 and 1962 (10). The most recent published follow-up of this cohort was through the end of 1983 (1 1). Two subcohorts, formed according to the products manufactured at the workplace, were used for the purposes of our analysis (11). The first consisted of workers employed at factories producing wood furniture, and the second was composed of workers employed at factories producing other wood products. Workplaces known to produce nonwood products were excluded.
Two cohorts of plywood workers were included in the pooled analysis. The first consisted of workers employed between 1945 and 1955 and for a minimum of one year by four plywood companies in the northwest United States (US) (12). The cohort was enumerated by investigators from the US National Institute for Occupational Safety and Health (NIOSH), and mortality followup was performed for the years 1946 through 1977. The second cohort of plywood workers was enumerated by investigators from the US National Cancer Institute (NCI) as one of 10 cohorts included in a large study of workers exposed to formaldehyde (13,14). The cohort was comprised of workers employed at any time between 1951 and 1965 by a plywood company in the eastern part of the United States. Mortality follow-up was from 1951 through 1979. The last cohort consisted of workers employed for a minimum of one month in wood model shops within the US automobile manufacturing industry between 1940 and 1980 (15 (19). As was the case with the British furniture industry, no industrial hygiene sampling results prior to 1970 were available.
Wood dust exposure potentially occurs during several points in the production of plywood. The highest exposures occur during the sanding, machining, and sawing necessary to finish plywood after the plys (or veneers) have been glued and pressed. Other areas where exposure can occur at lower concentrations are where raw logs are converted into veneer and cut to size. Relatively few data have been collected on exposure to wood dust in US plywood plants. Whitehead (20) reported the results of four personal samples taken in a softwood (fir or cedar) plywood plant in Washington state as part of a NIOSH health hazard evaluation. The mean concentration of wood dust was 0.6 (0.3-1.3) mg . m-3, but the operations evaluated were not specified. The plants included in the NIOSH study of plywood workers principally used softwoods (primarily Douglas fir), while the plant in the east, studied by the NCI, reportedly used hardwoods.
Pattern and model making is a highly skilled trade responsible for the production of prototypes, models, and patterns for mass production industries. However, pattern and model making is not a mass production operation, each piece is made individually starting from blueprints and ending with a finished product that must often meet very close tolerances. The materials used include a variety of hard, soft, tropical, and laminated woods, as well as various plastics, resins, and metals. Soft woods are often used for experimental models, while, for models that need to have more exact and stable dimensions, harder and laminated woods are used. An industrial hygiene survey was conducted at one of the plants studied by Roscoe and his colleagues, and the mean for the 58 wood dust samples was 1.0 (range 0.1-13.9) mg . m-3 (15,21). The average amount of respirable dust was approximately 43% (< 10 pm).
It is likely that the exposure levels in this analysis have changed among the cohorts over time. The increasing mechanization and efficiency of wood-working tools have generally increased the level of dust generated, while the increased use of local exhaust ventilation and respiratory protection have decreased the dust levels. In general, exposures in most industrialized countries are likely to have decreased significantly in the last 20 years due to the increasing knowledge of the hazards of wood dust and the regulatory response to that knowledge. Chemicals used in the wood-related industries vary by type of product and are also likely to have changed over time.

Classification of exposure
The cohorts included in the pooled analysis differed in the amount and detail of data collected exposure. The study of British furniture workers collected job titles for the single "dustiest" job held, which, in most cases, was also the longest job held. The jobs were classified as to dust exposure according to a nine-point relative scale (8). Employment dates were also collected but, unfortunately, the work history follow-up ceased on this cohort after 1968 (while mortality follow-up was continued through 1991). The study of members of the UFWA collected first date of employment and the major product(s) manufactured at the workplace available for all cohort members. In addition, the title of the job at the date of hire was available for 75% of the cohort. Both studies of plywood workers collected information on all jobs held at the factories under study, but the coding of this data varied. The NIOSH plywood study classified jobs into seven work areas differentiated according to qualitative exposure, while the NCI plywood study collected job and work area titles used in the factory. The workers employed in wood model shops in the auto industry had first date of employment in a wood model shop but last date of employment in the auto company.
In order to analyze the risk of cancer associated with wood dust exposure, a special classification scheme was developed incorporating the exposure information available for each study. Wood dust exposure categories were chosen according to the exposure categories used in the British furniture and the NIOSH plywood worker studies and the data available from the other studies. Three categories were defined according to the assumed probability of exposure: possibly exposed, probably exposed, and definitely exposed. British furniture workers were placed in categories according to the "dust" classification in the original study: less dusty (categories 1-3), dusty (categories 4-6), and very dusty (categories 7-9) (8). For the cohorts based upon the membership of the UFWA, job titles were used to place workers in categories similar to those in the British study. The plywood worlters were categorized using the department codes and qualitative exposure information from the NIOSH study and department and job titles hom the NCI study. Workers in plywood finishing operations were placed in the category with the highest probability of wood dust exposure, workers employed in processing fresh logs and making veneer were classified into the middle category, and all others were classified as possibly exposed. All wood model shop workers were placed in the probable category because no information on individual exposure was available. A complete description of the wood dust classification scheme is available in an IARC technical report (40).
Full worlt history information was available for the two plywood worlter cohorts, and therefore analyses could be made by duration of total employment and employment in jobs with potential wood dust exposure. One element that all cohorts in the pooled analysis had in common was year of first employment. This information was used to classify workers according to calendar period of employment as an indicator of potential exposure.
Analyses for standardized mortality ratio (SMR) were carried out on a VAX mainframe computer using the person-years program with standardization by five-year age and calendar periods (22,23). Ninety-five percent confidence intervals (95% CI) were calculated on the assumption of a Poisson distribution. Reference rates for US mortality specific for gender and race (white-nonwhite) were supplied by NIOSH for use with the US studies (24). Mortality rates for England and Wales, obtained from the World Health Organization mortality data bank, were used for the British furniture workers. Parallel analyses were performed for each cohort for cancer sites observed to be in excess in the participating studies and those of a priori interest, based 011 a review of the literature. The cancer sites chosen were the nose, nasopharynx, larynx, lung, lymphatic and heinatopoietic system, colon, and stomach. In addition, the risk of death due to nonmalignant respiratory disease and, more specifically, to bronchitis, emphysema, and asthma were examined. For each of these causes of death, analyses were carried out by the classification of wood dust exposure, decade of first employment, years since first employment, and combinations of these three variables. Analyses by duration of employment were performed for the plywood worker cohorts only because the information on duration of employment was not available for the UFWA study and was incomplete for the British furniture worker study. Lagging of exposure by 5, 10, and 20-year intervals was used to allow for a latency period (25).
The observed and expected numbers from these parallel analyses were summed for the pooled analysis. In this paper the results are presented for all of the cohorts combined, the two furniture worlter cohorts combined, and the two plywood worlter cohorts combined. The results of all of the pooled analyses were inspected to determine if the results were due to cxcesses or deficits from individual studies. In such cases, individual study results were also reported. Gender and race (white-non- Total all wood workers white) specific results were also examined to determine if they differed from the results for the full cohort. The results of the parallel analyses of the individual cohorts and the complete results of the pooled analyses are available in an IARC technical report (40).

Results
The size and vital status at the end of follow-up of each of the cohorts are presented in table 2. The pooled cohort consisted of 28 704 persons, among whoin 7665 deaths had occurred. The NIOSH plywood and wood model shop cohorts were composed of white males and the British furniture workers were all men (race unspecified, but assumed to be white). Thus the cohort was primarily male (92%) and white (89%). However, 3165 cohort members were nonwhite and composed 12% (N = 1460), 16% (N = 733) and 42% (N = 972) of the US wood furniture, other wood products, and NCI plywood cohorts, respectively. The cohort also included 2198 women who composed 15% (N = 1778) and 9% (N = 420) of the US wood furniture and the other wood product cohorts, respectively. Eighty-three women and 53 persons of other and unknown race from the NCI plywood study were excluded because of their small numbers. Other differences in the relation to numbers in published reports were due to the exclusion of three members of the NIOSH plywood worker cohort due to missing birth dates and the deletion of one duplicate record from the British furniture worker cohort which was discovered in the course of extending the follow-up. Table 3 displays the summary mortality results for all cohorts combined. Overall mortality was significantly less than expected when compared with the general population rates. Statistically significant excesses were observed only for sinonasal and nasopharyngeal cancer. Among the other cancers, a greater number of multiple myeloma deaths was observed than expected, but for all other sites the relative risks observed were less than or close to one, and statistically significant deficits were observed for cancers of the buccal cavity and pharynx, intestines, liver, lung, skin, bladder, and leukemia. The standardized mortality ratios for nonmalignant respiratory disease and bronchitis, emphysema, and asthma, as well as for all other noncancer causes of death, were less than one.
In table 4 the summary results are presented for the furniture and plywood worker cohorts. The standardized mortality ratios observed for most cancers and nonmalignant causes of death were less than or close to one. All 11 nasal cancer deaths and seven of the nine nasopharyngeal cancer deaths occurred among the furniture workers. Two nasopharyngeal cancers occurred among ply- wood workers. A greater than expected number of deaths from Hodgkin's disease and multiple myeloma were observed among the plywood workers, but the confidence intervals were wide and included one. Tables 5, 6, and 7 present the stratified results by likelihood of wood dust exposure, decade of first employment, and years since first employment, respectively. Among study subjects identified as having the highest probability of wood dust exposure, the relative risks for both nasopharyngeal and sinonasal cancer were elevated (        wood dust exposure category was 9.9 (95% CI 4.5-18.8). The risk of multiple myeloma also increased with the probability of assumed exposure to wood dust. No other trends with probability of exposure were apparent.
The SMR values for sinonasal cancer were greatest among workers first employed prior to 1940 (table 6). For furniture workers first employed before 1940 the SMR was 13.4 (95% CI 6.1-25.5). The SMR values for workers employed prior to 1950 were highly elevated for nasopharyngeal cancer and also appeared to be elevated for multiple myeloma. No other trends with time period of first employment were apparent. In the analysis by years since first employment, the SMR values were highest for sinonasal cancer and nasopharyngeal cancer for the workers at least 30 years after first employment, while the SMR for multiple myeloma was highest for the workers 20 to 29 years after first employment (table 7). Table 8 summarizes the stratified analyses of nasopharyngeal cancer and multiple myeloma among the furniture and plywood workers separately. For the furniture workers, the pattern was similar to that of sinonasal cancer with the highest risk observed for persons with the highest probability of wood dust exposure, those employed prior to 1940, and for persons at least 30 years after first employment. However, nasopharyngeal cancers also occurred in two furniture workers and two plywood workers who were not known to have been employed in operations exposed to wood dust. The trends for multiple myeloma with probability of exposure, years since first exposure, and decade of first exposure among the full pooled cohort were also apparent among the pooled furniture workers and also, although based on small numbers, among the plywood workers. Analyses were also performed with stratification by two-way combinations of probability of exposure, years since first exposure, and decade of first exposure. Eight sinonasal cancers occurred among furniture workers employed in the highest wood dust category prior to 1940 (SMR 18.6, 95% CI 8.0-36.6). Five nasopharyngeal cancers occurred among furniture workers employed in the highest wood-dust category prior to 1950 (SMR 13.9, 95% CI 4.4-32.4). In the two-dimensional analysis no new associations appeared that were not readily apparent in tables 5 through 8. There was no evidence of excess risks of lung, laryngeal, colon, or stomach cancer among the entire pooled cohort, or the pooled furniture or plywood workers alone, when SMR values were examined in relation to the probability of wood-dust exposure, time period of first employment, years since first ernployment, or combinations of these parameters.
The results of the analyses of plywood workers by total durqtion of employment and duration of employment in jobs exposed to wood dust (probable or definite) are presented in table 9 for the lymphatic and hematopoietic cancers. All of the exposures were lagged by five years, a latency period thus being allowed for. The risk of death due to all lymphatic and hematopoietic neoplasms was elevated among the workers employed for more than 10 years in any job and for those employed for 10 years or more in wood-exposed jobs. The excess of Hodgkin's disease occurred among the workers employed in wood-exposed jobs while multiple myeloma was elevated among the workers employed 10 or more  "Allowing for a five-year latency period using lagging years in any job. Allowing for no latency period or lagging by 10 or 20 years decreased the apparent associations. Both nasopharyngeal cancers among plywood workers occurred in workers employed for less than five years.
The results for sinonasal cancer for the pooled cohort were primarily due to 10 cases (expected 1.4) that occussed among the British furniture workers. The remaining sinonasal cancer death occurred among white male US furniture workers (expected 0.9). The results from parallel analyses of the British and US furniture workers were inspected for the apparent discrepancy between the results of the two cohorts. While 9 of the 10 sinonasal cancers among the British furniture workers occurred among men first employed prior to 1940, the entrance criterion for the US furniture workers was first employment between 1946 and 1962. The expected number of sinonasal cancers was very small in both groups, but fewer were expected in the US versus the British among workers in the highest probability of exposure category (0.2 versus 0.7) and the 30 or more years since first employment category (0.1 versus 0.8).
No sinonasal cancer deaths were observed among nonfurniture workers, whereas approximately one would have been expected. Of the nine nasopharyngeal cancers, five occurred among the British furniture workers (expected 0.9), two among US furniture workers (expected 1.6), and one each in the NIOSH (expected 0.3) and NCI plywood (expected 0.2) cohorts. The SMR values for multiple myeloma ranged from 1.1 to 2.3 in the individual cohorts. In general, the pooled cohort results were consistent with those of the individual cohorts for common causes of death. For example, the SMR values ranged from 0.7 to 1.0 for all cancers and from 0.8 to 1.0 for lung cancer. The one exception was the previously reported excess of stomach cancer in one cohort (15).

Discussion
The primary strength of this analysis is the increased power to examine rare causes of death afforded by the pooling of data. In addition, the classification of probability of wood-dust exposure, although crude, represents an improvement over simple summary SMR values for the entire cohort, and only the British furniture worker cohort had previously been examined in this manner. The large numbers also provided the ability to examine trends in the data in relationship to time period of exposure and the number of years since exposure began.
The chief limitation of this analysis was the lack of accurate information regarding exposure to wood dust and other agents. Although it was possible to create some crude exposure indices, misclassification undoubtedly occurred. Within studies, misclassification will have occurred because persons were inaccurately assigned to exposure categories due to the limited exposure information available. Since the exposure information used was obtained prior to the occurrence of disease, the misclassification would be nondifferential and would tend to obscure potential relationships between wood dust and disease and could distort dose-response relationships (26,27). Further misclassification may have occurred between studies because differences in the detail of the exposure information available could have caused the accuracy of exposure assignment to vary by study. The effects of this type of exposure misclassification are difficult to predict.
Another limitation of this study was the reliance on mortality data as an indicator of disease incidence. Death certificates are a poor source for information regarding nonmalignant respiratory disease and may also lack details on the anatomic subsite of origin and histology information for fatal cancers (28), which are particularly important for nasopharynx and sinonasal cancer. In addition, national rates, which were used in all of the analyses, may not accurately reflect the base-line disease risk in this population due to the healthy worlzer effect (29), social class, and ethnic and regional differences. The pooled cohort had a significantly depressed SMR for total mortality, and therefore the SMR values for some diseases may have been artificially low.
In this study an excess risk of sinonasal cancer was observed among the British furniture workers with the highest wood-dust exposure but not among US furniture workers. It has been noted previously that furniture worlzers in the United States may not have as high a relative risk of sinonasal cancer as their British counterparts (10,30). The reasons for this inconsistency are not known, but in this analysis at least part of the discrepancy may have been due to differences in study design and exposure between the two groups. In Britain, the elevated risks were observed among the furniture worlzers first employed prior to 1940, while entry into the US cohort began in 1946. The available industrial hygiene measurements would seem to indicate that wood-dust exposures may have been higher in the British industry. However, it is not possible to make a direct comparison because little is known about the workplaces where the US workers were employed, and no measurements are available from earlier periods. It may also be of significance that the British furniture industry in the High Wycombe region produced furniture primarily of hardwood, which has been associated with higher risks of sinonasal cancer than softwoods (1,7,8). No information on species of wood was available for the US furniture worlcers.
This reanalysis also observed an excess of nasopharyngeal cancer among furniture and plywood workers. Among the furniture workers, a large excess was observed for the men employed in jobs with the greatest likelihood of high wood-dust exposure. We also found an excess of nasopharyngeal cancer for workers holding jobs for which exposure to wood dust was less certain. This observation may be due to a inisclassification of exposure, separate effects of different exposures, such as formaldehyde among the plywood workers, or, given the small numbers, chance. Some case-referent studies have reported an association between nasopharyngeal cancer and wood work occupations (4,(31)(32)(33).
Our analyses suggest that the risk of some of the lymphatic and hematopoietic neoplasms may be elevated among wood workers, particularly multiple myeloma. In the full pooled cohort, the risk of multiple myeloma was greatest among workers with the highest probability of wood-dust exposure, but among furniture workers (who may have had the highest exposures) the trend was less pronounced. Three out of the four multiple myelomas among the plywood workers occurred among persons in the lowest probability of exposure category. From these analyses the specific factors that may be associated with excesses of multiple myeloma among wood workers are unclear, but the overall pattern suggests that both chemical agents and wood dust may play a role. Some studies have found multiple myeloma to be related to chemical exposures that may occur within wood-related industries, such as paints, solvents, and pesticides (34)(35)(36)(37). Further assessment of exposures in plywood and other wood-related industries is warranted.
No evidence was found for an excess risk of death due to lung cancer and nonmalignant respiratory disease in any of the stratified analyses, nor were any trends suggestive of a relationship apparent. Among plywood workers, a s~nall excess of bronchitis, emphysema, and asthma was observed, but it did not appear to be associated with any of the surrogate measures of wood dust exposure which were used. An important issue to consider is the respirability of wood dust. The particle size of dust generated by wood-working operations varies somewhat according to the process used and the type and dryness of the wood. Industrial hygiene sampling, however, has generally found the majority of suspended wood dust to have a mass median aerodynamic diameter of greater than 10 pm, and this is generally above the respirable size (38,39). In this analysis, exposure to finer wood dust would have been expected to occur during furniture and plywood finishing operations. The lack of an excess of disease among workers in these categories could be a true indicator of the absence of effect or could be due to a combination of a poor measurement of exposure to respirable dust or the healthy worker effect. It has been noted that wood workers may smoke less because smoking is prohibited in some workplaces due to the fire hazard posed by wood dust (8). If true, negative confounding by cigarette smoking may be possible. Unfortunately, no data on smoking or other potentially confounding factors were available for this analysis.
In conclusion, this reanalysis of data from five cohorts of workers in wood-related industries confirmed the association between sinonasal cancer and wood-dust exposure; this conclusion is based primarily on the strength of the British furniture worker study. In addition, a similar pattern was observed for nasopharyngeal cancer; this result provides further evidence that it may also be related to wood-dust exposure. Some support for an excess risk of multiple myeloma was also observed. No excesses of lung, larynx, stomach, or colon cancer were found to be associated with any of the suirogate indicators of wood-dust exposure used. Although the studies included in this analysis lacked detailed information on exposures and potential confounders, these limitations and the healthy worker effect would tend to obscure relationships between wood dust and cancer rather than create false positive findings. Large studies with better indicators of exposure are needed to examine the relationship between some of the relatively rare cancers found in excess in this study and exposures in woodrelated industries. The use of cancer incidence data, rather than mortality, could also increase study power and allow the examination of the risks associated with specific histological types.