Mortality among production workers in pulp and paper mills

FOWLER Mortality among production workers in pulp and paper mills. Scand J Work Environ Health 12 (1986) 552-560. A cohort of 3 572 pulp and paper mill workers employed for at least one year between 1945 and 1955 was followed through 31 March 1977. Vital status was determined for 99 % of the cohort. The 915 deaths observed were 79 0,70 of the number expected on the basis of comparable United States mortality rates. Statistically nonsignificant excesses of deaths due to lymphosarcoma and reticulosarcoma and to stomach cancer were observed. These findings tend to corroborate reports based on state vital statistics, and preliminary case-referent and population based studies of workers in the pulp or paper industries. No deaths due to nasal cancer were observed, but only 0.6 were expected. When process-specific analyses were conducted, the excess risk of lympho sarcoma and reticulosarcoma was increased only for men who worked in sulfate mills. The excess risk of stomach cancer was limited to men who worked in sulfite mills. Process-specific standardized mor tality ratios for these causes were highest after 20 years since first employment in the mills.

However, unevaluated hospital referral patterns may have contributed unknown biases.
In the only published retrospective cohort mortality study of workers in a pulp and paper mill, Ferris et al (6) followed for 10 years a cohort of 271 pulp and paper mill workers identified as actively working in the mill in 1963. Although only 33 deaths occurred among these workers by 1973, the 33 deceased paper and pulp mill workers were reported to have an overall standardized mortality ratio of 126. No cause-specific expected deaths were computed, the period of observation was short, and there was a substantial lack of statistical power.
Because of these associations between various malignancies and pulp and paper mill employment, we conducted a retrospective cohort mortality study of men employed in five pulp and paper mills located in the northwestern part of the United States. Our study was designed to evaluate the chronic health effects of employment in this industry and to test the hypotheses of increased mortality due to malignancies of the buccal cavity, pharynx, stomach, intestine/rectum, larynx, lung, prostate, bladder, kidney, and the hematopoietic and lymphatic systems (table I), as well as that due to coronary heart disease and diseases of the blood-forming organs (20,21).
The sulfate process, which is alkaline-based, depends mainly on pine wood (in the southern part of the United States) or fir (in the northwest mills studied). The acid-based sulfite process most commonl y uses low resin woods such as spruce or hemlock, which contain hydroxyphenolic compounds. Almost any softwood can be used in the groundwood proce ss; one of the sulfite mills in our stud y also ran a groundwood proce ss which used both fir and hemlock (18).
Occupational exposures may arise at any stage in the process from preparation of the raw wood through the production of the final paper or pulp product (table  2). Wood preparation does not differ subst antially for the several proces ses, but there can be significant differences in exposur es in subsequent proce ss steps, particularly in two ar eas, ie, in cooking liquor production and in pulp production, washing, bleaching and recovery, which together convert the wood fibers into cellulose fibers, ready for subsequent processing into paper of vari ous grades or for direct sale as pulp .
Paper makin g often takes place in the same plant which produces the pulp , as was true for four of the five mills studied. The pulp is mixed with water and

Subjects and methods
All 37 pulp and paper mills in the states of Washington, Oregon and California that were thought to have been in operation prior to 1950 were surveyed. Five pulp and paper mills were selected for study on the basis of the age of the mill, the completeness of the personnel records, and the type of proce ss used by the mill. Of the five pulp and paper mills selected, three produced Kraft sulfate or sulfate pulp (which are essentially the same proce ss) and two produced sulfite pulp. One of the two sulfite pulp mills also used the groundwood process, which is purely mechanical and involves no chemical treatment of the pulp. All five mills began pulp operations between 1918 and 1930. The end products of the study mills were newsprint, purified chemical cellulose, bleached and unbleached pulp, paper, board and boxboard. A different bleach process was used by each of the four mills which used bleach.
Before the pulp and paper mills were selected, the study cohort was defined to include only white males who worked at least one year between 1945-1955 inclusive. These criteria allowed the investigation to focus on persons with a minimum latency and whose followup would be facilitated due to their occupational stability and their having worked after World War II. Salaried office workers, females , and nonwhite males were excluded from the study due to their minimal exposures or small numbers. Employment records of the remaining men were microfilmed. The demographic data and the daily detailed work histories of each worker, including departments , job titles and dates, including layoffs, were coded for the computer from these records.
The 3 572 pulp and paper workers who met the study cohort criteria were followed from their last date of employment through 31 March 1977 ( (31)was used to obtain person-years at risk of dying by race and sex in five-year calendar-time periods, five-year age groups, five-year duration of employment periods, and fiveyear latency (time since first employment at the plant) periods. Person-years were multiplied by the appropriate US mortality rates to obtain the expected number of deaths.' Since the cohort was limited to white male production employees who had at least one year of employment between 1945 and 1955, person-years at risk began either at one year after first date of employment or on I January 1946: whichever came later. Workers were included in a process subcohort analysis if they worked for one year or more in the appropriate process category. The sulfite process subcohort consisted of I 779 men, and the sulfate subcohort of 1 796 men . It was possible for a person to be included in more than one process sub cohort if he worked in more than one study mill; however only a few persons had worked longer than one year in each of the processes studied .
Because of the widespread interest in the health effects of formaldehyde, an additional subcohort of I 262 men was examined separately. These men had worked I d or more in the areas of the mills believed to have had formaldehyde exposures (table 2). These areas were the paper-making portion of the mills, where urea and melamine formaldehyde resins were used (15). These resins were used in some of the sizing agents and pigments added to the slurry. They were also contained in some of the coatings applied to finished paper.
The observed numbers of deaths in the study cohort and subcohorts were compared with the numbers expected as derived from sex-, age-, race-, calendar-time-, and cause-specific US mortality rates. SMR values were calculated by dividing the number of observed deaths by the number of expected deaths and multiplying by 100. The SMR values were not calculated when the observed and expected numbers of deaths were both less than three. On the basis of our literature review, we developed a priori hypotheses for excess risks of death due to the malignancies listed in table I, coronary heart disease and diseases of the blood-forming organs. Thus we decided to use one-sided testing of the se causes of death at the 0.05 alpha level. We also wanted estimates of upper limits of risk. Consequently, we calculated two-sided 90 % confidence interval (90 0,10 CI) values for the SMR values . An evaluation of the lower limit of these is equivalent to evaluation of the one-sided p = 0.05 test of excess risk. For consistency , similar .1 Because appro pri ate comparison rates for lymphosarcoma and reticulosarcoma were not available before 1950, the rates from 1950-1954 were used for 1946-1949. confidence inter vals are presented for all SMR values in the paper. The Fisher exact method was used to calculate the intervals if the observed deaths were less than eight, and an approximate method was used if the observed numb er of death s was eight or more (27).

Results
For the entire cohort (table 3), 90 961 person-years of obser vation were accrued . A statisti cally significant deficit of deaths was observed, 915 versus I 150.3 expected (SMR 79) (table 4).
Mortality due to circulatory system disease (SMR 81) and all malignant neoplasms (SMR 76) was also statistically significantly below that expected. The SMR values for the following five site-specific malignancies were nonsignificantly elevated: stomach cancer (17 observed, 13.8 expected, SMR 123), prostat e cancer (17 observed, 14.8 expected , SMR 115), kidney cancer (6 observed, 5.2 expected, SMR 115), bladd er cancer (8 observed , 6.9 expected, SMR 116), and lymphosarcoma and reticulosarcoma (10 observed, 5.9 expected, SMR 169). All intestinal malignancies were of the colon; none were of the small intestine. Of the 10 deaths in the lymphosarcoma and reticulosarcoma category, nine (90 0J0) were due to lymphosarcoma, as opposed to 61 070 observed in the US population for 1970 (29). There was a slight excess of death s due to diseases of the blood-forming organs (5 observed, 3.0 expected, SMR 167). No deaths due to nasal cancers were found. In addition , a carefu l review of all 903 death certificates revealed no mention of sinonasal cancer as a contributory cause of death or other significant condition.
A detailed examination of the work histories of the 10 lymphosarcoma and reticulosarcoma cases revealed that four of the men worked in pulp production , two worked in raw wood preparation, two worked in maintenance , and two worked in paper-making areas. The bimodal age distribution of deaths was not unu sual.
Only causes of death with elevated SMR values or tho se causes of death which were of interest because an excess risk had been observed in previous studies are presented in tables 6 and 7 for the process subcohorts. Among the subcohort of sulfite/groundwood process workers (table 6), significantly fewer deaths occurred than were expected , resulting in an SMR of 86. The SMR values were elevated, but were not statistically significant, for two site-specific malignancies, ie, stomach cancer (11 observed, 7.4 expected, SMR 149)and kidney cancer (4 observed, 2.7 expected, SMR 148). The few deaths due to diseases of the blood and blood-forming organs were more than expected (3 observed, 1.6 expected).
There were 392 deaths among the subcohort of men employed in pulp and paper mills (table 7) which used Kraft or sulfate pulping processes, whereas 544.1 deaths were expected, resulting in a significantly decreased SMR of 72. The SMR for all malignant neoplasms was also 72 0J0 of that expected. However, there was a twofold excess (90 0J0 CI 90-408) due to lym- 3 555 Table 5. Detailed work histories of workers who died of lymphosarcoma and reticulosarcoma by pulp process.
phosarcoma and reticulosarcoma (6 observed, 2.9 expected, SMR 207). Only lymphosarcoma was found among the sulfate workers; no reticulosarcoma deaths occurred. A slight excessmortality occurred due to cancer ofthe prostate (8 observed, 6.7 expected, SMR 119) and cancer of the bladder (4 observed, 3.2 expected, SMR 125). Mortality due to lymphosarcoma among the sulfate process workers and mortality due to stomach cancer among the sulfite process workers, as well as mortality due to prostate and urinary tract cancer among the 16 Pulp production area 9 Raw wood preparation area 2 No information on record

Sulfate mills 34
Paper machine area Laboratory

31
Paper machine area

31
Pulp production area No information on record Unexposed area

Discussion
The site-specific malignancies with the highest SMR values in this study included lymphosarcoma and re-entire cohort, were further analyzed by duration of employment and latency. The excess risk of mortality due to lymphosarcoma among the sulfate process workers was the most evident only after 20 years' latency and employment (4 observed, 1.1 expected, SMR 364) (table 8). The stomach cancer risk among the employees of the sulfite pulp mills (table 9) was also the most elevated after a 20-year latency period (9 observed, 5.1 expected, SMR 176), but during that latency period did not increase with duration of employment. Among the pulp and paper workers employed over 20 years, excesses of prostate cancer (11 observed, 7.0 expected, SMR 157) and urinary tract cancer (9 observed, 5.5 expected, SMR 164) were evident. Among those with greater than 30 years' latency, excesseswere seen for prostate cancer (12 observed versus 6.9 expected) and urinary tract cancer (8 observed versus 4.4 expected). These excesses were independent of process.

26
No information on record Pulp production area    ticulosarcoma and cancer of the stomach and the urinary tract, all of which have been elevated in previous studies of workers in paper-related industries. Although none of these elevated values were statistically significant , this lack of significance could be due to the smaIl numbers of deaths. Both the low background death rates of the causes of death of a priori interest and the smaIl cohort size contributed to the problem of the smaIl numbers of expected deaths for these causes. This problem is reflected in the wide confidence intervals observed for most of the SMR values in table 4. These intervals aIlow us to rule out any excess risk of death due to malignancies of the buccal cavity and pharynx, intestines, pancreas, and lung.
However , we can only exclude the possibility of SMR values greater than 200-300 for all other malignancies of any a priori interest from table I . We were surprised by the strikingly low overaIl SMR of 79 in this group of pulp and paper workers, all of whom entered the cohort at least 22 years before the end of follow-up (table 4). This low SMR does not appear to be due to inadequate foIlow-up (l 010). To explore the effect of persons lost to foIlow-up on the calculation of expected deaths, we carried out an alternative analysis in which person-years at risk were terminated on the day the workers lost to foIlow-up quit work rather than assuming them to be alive at the end of the study. None of the SMR values calculated by 557 this alternative method differed by more than 1 0/0 from those in table 4.
The cohort did not appear to be diluted by persons with limited duration of exposure or latency. To investigate this possibility, analyses were conducted for both process subcohorts which were stratified by latency and duration of employment. All the SMR values for the long latency and duration of exposure in the all-causes and all-malignant-neoplasms categories were within 5 % of the overall SMR values presented in tables 6 and 7. None of the SMR values was greater than 100.
The excess risk of death due to lymphosarcoma was greater among the workers in the sulfate process mills (SMR 207) than among those in the sulfite process mills (SMR 133). Among the workers in sulfate process mills, the increased SMR values for lymphosarcoma with long durations of employment and after 20 years since first employment were consistent with an hypothesized occupational etiology. A recent report of provincial mortality (8) based on death-certificate occupation reported statistically significant increased deaths due to lymphosarcoma and reticulosarcoma, as well as to prostate cancer, among Canadian pulp mill workers. However, industrial hygiene information on the work environment and type of pulping process was not reported.
While the excess risk in our study appeared to be restricted to sulfate mills, the lymphosarcoma cases do not seem to cluster in any particular work area. Despite the limitations of the environmental data, it remains noteworthy that many of the lymphosarcoma cases worked in areas with presumable exposure to wood volatiles, which may contain either neoabietic acids (16) or pentachlorophenols used for slime control (14). The latter have been linked to lymphomas by Hardell et al (11). The twofold elevated risk of death from lymphosarcoma among the sulfate workers was independent of the slightly elevated relative risks for lymphosarcoma reported by Milham (21) in a proportionate mortality analysis of pulp mill workers in the same geographic area of the United States. Only three of the ten deaths due to lymphosarcoma or reticulosarcoma in our study cohort were included in his analysis of state vital statistics. Because of the known relationship between immunosuppressive drug therapy and the subsequent development of lymphatic system cancers (16), the nine lymphosarcoma and one reticulosarcoma death certificates were examined for all contributory causes of death. However, no conditions which might require immunosuppressive drug therapy were mentioned on their death certificates.
The elevated SMR for stomach cancer may have been due to life-style factors, which were not controlled in the analyses. Studies have detected an increased incidence among men of lower socioeconomic status. In spite of possible confounding by socioeconomic status, evidence for a possible occupational associa-558 tion with pulp or paper mill employment is also provided by two other studies (4,21).
The increased risk of dying from stomach cancer among the pulp and paper workers corroborates other studies of workers of such wood-related industries as the furniture, plywood, and sawmill and logging industries (10,11,20,21,24,25), and among carpenters (20,21). According to a recent report, the excess in paper and pulp workers appears to have declined over the last three decades (20). The possibility that these excessrisks of stomach cancer are related to wood dusts either directly or indirectly as carriers of carcinogens needs to be evaluated by measurements of particle-size distributions and the modeling of respiratory and digestive system deposition patterns (5,13). Six of the 11 men with stomach cancer in the sulfite/ groundwood workers' subcohort had worked in areas of wood preparation (table 10), where log debarking and chipping into small pieces takes place (tables 2 and 10).
The fourfold excess of urinary tract cancer after 30 years' latency among those who worked in paperdrying areas with potential exposure to formaldehyde appears to warrant further investigation. Although at least one mill was believed to have used benzidenebased dyes, known urinary tract carcinogens, no exposure data were available at the mills. Levels of up to 2.5 ppm of formaldehyde (Stanford Research Institute project no 5847 report) have been found in these pulp mills. It is noteworthy that an excess risk of kidney cancer (a proportional mortality ratio of 256, based on 6 cases) has been reported among embalmers exposed to formaldehyde (30).
Two (2,4) of several epidemiologic studies of pulp and paper mill workers have reported excess risks for dying of cancer of the buccal cavity, but a large deficit (1 observed, 6.6. expected) was observed in our study . A fourfold relative risk for oral and laryngeal cancer was observed by DeCoufie et al (4) among men who had worked as operatives in the paper industry for five years or more. In addition a significantly increased mortality rate for oral cancer was observed among white males by Blot & Fraumeni (2) for US counties with more than I % of the population employed in the paper industry.
The elevated rates we observed for lymphosarcoma, stomach cancer, and urinary tract cancer, while not statistically significant, were provocative because of the associations observed in earlier epidemiologic studies. These excesses do not appear to be due to differences between local and US mortality rates. In Washington State, where the mills are located, the mortality rates for lymphosarcoma, other lymphomas and mycosis fungoides; bladder cancer; and stomach cancer are all within 1 l1Jo of the respective US mortality rates (17).
A deficit of risk due to lung cancer was observed in this study (50 observed, 62.1 expected, SMR 81). Lung cancer has been associated with pulp and paper mill employment in some studies (1, 12,19) but not in others (3,9). In the only study which considered cigarette smoking, a case -referent study of 535 male residents of coastal Georgia, no association was found with pulp or paper mills (3).

Conclusion
Statistically nonsignificant excesses of mortality due to lympho-and reticulosarcoma and stomach cancer were found in a cohort of pulp and paper mill workers. These findings tend to corroborate reports ba sed on state vital statistics, as well as preliminary studies of pulp or paper mill workers which observed associatio ns with lymphatic cancer. In our study, the lymphosarcoma excess was restricted to those who had worked in the sulfate mills, while the stomach cancer excess was limited to those who had worked in the sulfite mills . These process-specific SMR values were highest after 20 years since first employment in the mills. Although no etiologic agents were distinguishable, further research, including the continuing followup of this cohort and the characterization of exposureresponse relationships, should be conducted. Future research should focus on the potential carcinogenic roles of exposure to wood dust, formaldehyde, sulfur compounds , and other chemicals in this industry.