Possible causes of increased lung cancer incidence among butchers and slaughterhouse workers.

P, FELLENIUS E, HOGSTEDT C. Possible causes of increased cancer incidence among butchers and slaughterhouse workers. Scand J Work Environ Health 13 (1987) 518-523. An excess of lung cancer among butchersand slaughterhouse workers has been reported in several record-linkage studies. In this case-referent investigation on the possibility of occupational exposures being related to the lung cancer excess, cases and referents were selected from butchers and slaughterhouse workers registered in the Swedish national census of 1960. The case group comprised all men in the study population dying from lung cancer between 1971 and 1982. Two reference groups were formed, ie, all individuals dying from other cancers and a random sample of all dead men in the study population during the same time period. The history of occupations, occupational exposures, and smoking habits was obtained from the next-of-kin by questionnaire. None of the occupational exposures that were studied (work with live ani mal care, in the bleeding area, on the killing floor, or with meat cutting, processing, curing , smoking, chilling and packaging) were associated with an increased lung cancer rate. Tobacco smoking habits may have contributed to the overall excess of lung cancer found previously for this occupational group.

An excess of lung cancer among butchers and slaughterhouse workers has been reported in several studies. Record-linkage investigations from England, Denmark, and Sweden ha ve uniformly shown an excess of lung cancer in this occupational group (7,9,12,17). In the Swedish study, data on smo king habits were available, and a correction for differen ces in smoking habits explained a part , but not all , of the excess (12). A cohort study from the United States (15,16) showed an excess of malignancies o f the hematopoietic system among male slaughterhouse workers and female worker s in meat departments of retail food sto res. An excess of lung cancer was indic ated among female workers in chicken slaughtering plants, but not among male or female slaughterhouse workers in general. A small cohort study from the Federal Republic of Germany indicated a high lung cancer mortality among butchers in comparison with bakers (3).
However, a register study from Washington State (21) and a case-referent study from Buffalo (27) showed no excess of lun g can cer among butchers or slaughterhouse workers . Nevertheless, the abovementioned epidemiologic findings indicate that th ere might be an excess of malignancies, especially lung cancer , among butchers and slaughter house workers. Exposure to virus in the handling of a nimals (15,16,23) and to combustion products from meat smoking (17) Reprint reque sts to : Dr P Gusta vsson , Division of Occupational Medicine, National Institute of Occupational Health, 5-171 84 Solna , Sweden .
518 ha ve been proposed as etiologic factors. The latter theory is supported by the excess of lung cancer in other occupational groups expo sed to combustion products, especially polycyclic aro mat ic hydro carbons (P AH) [chimney sweeps (10,II), co ke and gas workers (4,13,24), and aluminum smeltery workers (8)].
In slaughterhouses the traditional method for smoking meat is to generate the smo ke by slow ly burning wood on the floor of a smokehouse. The work ers go into the smo keho use for loading a nd unloading and for controlling the smo king process. T he wo rk inside the smok eho uses is very du sty, but it is usuall y performed only for a short time. Howe ver , some smo ke co ntinuo usly leaks out of the hou ses, and during unloading larger amounts get out. Therefore the workroom may be fairly smoky during the whole workday. During the last 25 years the smoking pro cess has been modified , and in most plants the smoke is primarily nowadays generated outside the smokehouse , the expo sur e o f the workers being lower (5,22).
The am ount of different PAH was measur ed in Danish smo king plants using the tradition al smo king method (22). Very high levels of total PAH were found immedi ately o utside the smo keho uses in so me plants, the ran ge of 4-h mean values being 2.2-1 487 (mean 208, SO 452) ug/rrr'. The levels were con siderably lower 2 m fro m the smokehouses, up 10 15.1 ug/rn ' . The PAH that were quantitatively most important were of the th ree-rin g type, ie, naphthalene, fluorene , phen anthrene, fluoranthene , and pyrene. Higher PAH, eg, ben zo(a)p yrene [B(a)p ] occurred in smaller am ounts, in a single sa mple as high as 78 ug/ rrr' however. An investigation of two Swedis h smo king plants of diffe rent types showed low levels of B(a)p in a mod ern smo king plant (0.01-0.04 ug/m '), while somewhat higher levels (0.13-0.18 ug /rn ') were found in an older type of plant (5). Another Swed ish investigation showed low levels « 0.0 I ug/rn' ) o f higher PAH in the fumes from meat smokin g, methylphenanthrene being the main P AH con stituent in the smo ke (2). In summary, these results indicate that the exposure levels o f total PAH in smoking plant s of older types may be on the same order as tho se found for gas workers (14) a nd chimney sweeps (6).
Other expo sure s also occur in slaughterhouses, the contact with viruses from animal bodie s has been mentioned . Nitrit e is used for the cur ing of meat and the form at ion of nitrosamines in the meat has been pr oposed as a po ssible so urce of ca rcinogenic expo sure in the diet (25), but it should not result in exposure by inh alation among th e slau ghterhou se workers. Pla stic film is used for meat wrapping, and the fum es formed when the film is heated ma y give rise to irrit ation of the ai rways , ie, meat-wr apper' s asthma (26). There is no indication, however, that this exposure is car cino genic.
Th e present study was performed to identify occupational exposur es possibly relat ed to the lung can cer excess found earli er among but chers and slaughterhouse workers.

Subjects and methods
A case-referent study was performed. The study population was all Swedish men registered as a butcher or slaughterhouse wor ker in the national census of 1960. Thi s census ha s been cro ss-linked with the nat ion al register of deaths at the Swedish central bureau of statistics (Sta tistics Sweden). Only deceased persons were available for the study, since census data rega rding living person s are strictly confidential.
The case group comprised all men in the study population dying from lung cancer between 1971 and 1982. The lung can cer diagno sis was checked again st the Swedish Can cer Register, which a lso co ntains inf ormation about histological tumor type. Only those cases that were histolo gically verified as prim ary lung cancer of the squ amous-cell carcinoma type or th e small cell/ und ifferentiated carcino ma type were included among the cases (58 person s).
Two reference groups were selected . Referen ce group A (129 persons) comprised a ll men in the study popul ation who died from any type of ca ncer except those suspected to have an etiologic relationship to chemical exposures, ie, malignan cies of the oral cavity , pharynx and esopha gus [eighth revision of the International Cla ssification of Diseases (ICD) 140-151], liver tumors (ICD 155), respiratory malignanci es (ICD 160-163), malignant skin tumors (ICD) 172-173), bladder canc er (ICD 188), and malignancies of the hematopoietic syste m (ICD 200-209). Reference gro up B (151 person s) comprised a random sa mple o f all deceased men in th e study population, with the same excepti o ns as for group A . In add itio n, deaths du e to violence or poisoning were excluded for ethical reason s (ICD E800-E999).
For each person in the case and refer ence groups a next-o f-kin (widow, child , or sibling) was traced via population register s held by the clerical pari shes. A questionnaire was distributed that inquired about work histor y, work tasks, a nd tob acco smoking hab its. Two reminders were sent to tho se not responding , and in some cases the que st ionnaires were co mpleted by telephone intervi ew. Th e results of the tracing and the response to the questionnaire are present ed in table I.
In all , 287 qu estion nai res were sent out and 267 returned, the nonresponse rate being 7 0/ 0. In so me insta nces the next-of-kin had no knowledge of the work history of th e study subject. For a few persons it was found that the coding of the 1960 census was erroneou s and the study subject had actuall y never worked as a butcher o r slaug hterho use worker. For 230 persons an adequate wor k history was obtained . The answers were proces sed and coded without knowledge of the case-referent status of th e ind ividual s.
Th e statistical ana lysis was per formed in two steps. First, a n a na lysis st ratified for smo king ha bits was carr ied out (18). Th e confidence interval for the approximat e relati ve risk (RR) was ca lculated according to the test-ba sed method (19). Second, in order to include several job types simultaneo usly in the model , and also adj ust for di ffer ences in smo king habits and age at death, an unc onditional logistic regression was per form ed (I ). The wo rk tasks were introduced in the model as binary indicator variables (0/ 1), and smo king habits were introduced as two indicator variables, "ex-smoker" and "current smoker," versu s "neversmoker" as a base line. The category "current smoker" also included those who had given up smoking less than a year before death. Age was introduced in the model as a continuous variable, although one model was also tested in which age was introduced as an indicator variable after division was made into age classes.

Results
In general, the individuals had worked for a long time as a butcher or slaughterhouse worker, both the mean and median of the duration of employment being more than 30 years (table 2). Most of the individuals had performed several types of work (table 3). Two more distinct work-task patterns were discerned in a correlation analysis (not shown). One corresponded with butcher work with the handling of animals and slaughtering, and the other with the preparation of meat (CUlling, processing, smoking, and curing) . The mixing of work tasks was typical for the industry during the first half of this century, increasing specialization having occurred during the last 25 years. Very few persons in the study had had meat smoking as a single main task for a long period . The age distributions were nearly equal in the case and reference groups (table 4). As expected, tobacco smoking was more common among the lung cancer cases than among the referents (table 5). No lung cancer cases appeared among the never smokers. The difference in smoking habits was small between reference groups A and B.
A stratified analysis was performed for all types of work with tobacco smoking as the stratification variable . The adjusted summary relati ve risk (Mantel-Haenszel estimator) and confidence inter vals are presented for all types of work in table 6. The relative risk for lung cancer was not increased for any of the studied types of work, but was lowered for work with meat cutting. The numbers of exposed and unexposed cases and referent s in each stratum is shown for work with meat smoking in table 7.
In the regression analysis, the first step was to test the influence of tobacco smoking in a logistic regression model in which smoking habits were introduced as a factor represented by indicator variables. As expected, the log-likelihood ratio statistic (I) for this model was highly significant lX' [2 degrees of freedom (df)] = 23.27, P = <O.OO Il due to the strong association between tobacco smoking and lung cancer. The inclusion of age at death as a continuous variable improved the model only marginally, the log-likelihood ratio statistic being 2.36 [X' (I df) , P = 0.12). The alternative of including age as a factor after division into four age classes also yielded a nonsignificant improvement [X' (3 df) = 5.25, P = 0.15). A comparison of the relative risk estimates for lung cancer for 520 the different work tasks before and after correction for age and smoking habits indicated that neither of these factors acted as important confounders. However, since both factors were con sidered a priori as potential confounders and a minor confounding effect could not be excluded, they were retained in the model, and in all of the subsequent analyses smoking habit was included as a factor and age as a continuous variable.
The work tasks were first introduced in the model one at a time. The results (not shown) were nearly identical to those from the stratified analysis (table 6), both with regard to point estimates and confidence intervals. In a second model the work tasks were aggregated into animal contact (live animal care, bleeding, or work on the killing floor), meat handling (cutting, curing, smoking, processing, or chilling), packaging, and other work (table 8). Neither did this analysis show a lung cancer excess for any of the aggregated work types.

Discussion
None of the work tasks for butchers and slaughterhouse workers were associated with an elevated lung cancer risk . The question must be asked of whether this absence of positive findings be can interpreted as negative eviden ce against an occupational origin of lung cancer excess among butchers and slaughterhouse workers or only as inconclusive results. The procedures for identifiying and tracing the study subject s and nextof-kin generally worked well. The nonresponse rate was low, and there was no reason to believe that the non response was selective and would thereby introduce bias. In general, the subjects in the study had worked for a long time in the occupation in question, and in this respect they seem to be an appropriate target populat ion.
Then , there is the possibility that the negative findings were caused by a misclassification of the exposure . However, the point estimate for the relative risk for lung cancer among the persons who had worked with meat smoking was 0.84 with group A as reference . If a true relationship existed between work in smo kehouses and an increa sed risk for lung cancer, th is result could only be explained by an almost complete misclassification, which seems rather unlikely.
The study was based on a comparison of lung cancer risk for different work ta sks among decea sed person s only , and it thereby has the character of a proportional mortality study . Inference to the original study population can be made only insofar as the exposure under study is not associated with the reference disea se(s) (20). If this would be the case , a true excess risk could be obscured by an excess risk also for the reference disease(s) . In thi s study, two different reference groups were used, based on different sub sets of causes of death. Diseases that could be suspected a priori to be related to chemical exposures were ex-eluded from the referen ce groups, and, in additi on, both reference groups gave reasonably comparable estimates of the exposure frequency in the study population . Against this background it seems impro bable that such a mechan ism should have biased the results of the study.   Table 3. Percentag e of cases and refer ents having worked in the lis ted type s of work or wor k areas. (Case group = 45, reterenee grou p A =99, refe rence group B =100) Table 7. Lung cancer and meat smoki ng -Analys is stratified by tob acco sm oking habits and group A us ed as the refer ence group.
Exp osed Unexpo sed " Unadju sted relat ive risk .  Logit P(X) = a + i~, b j X j, wh ere P(X) =probability of death from lung can cer.
Variables: X, = age, X 2 = ex-smoker , X 3 = smo ker, X. =ani- Referen ce group B a Aggregat ed ty pes of wor k inc luded si mult aneously in t he model:  If systematic errors cannot explain the negative findings, the next step is to consider the precision of the study . At least two aspects of precision have to be considered, ie, first , the precision when trying to explain the previous finding of increased lung cancer incidence among Swedish butchers and slaughterhouse workers and, second, the precision regarding inference to butchers and slaughterhouse workers in general. Regarding the first, th is study was based on the same population that gave rise to the initial observation, that is, the cohort of butchers and slaughterhouse workers in the Swedish census of 1960. In this context, the study had the character of a total investigation , and it can simply be concluded from the point estimates of the relative risks of different work tasks (table 6) that, since all of them are smaller than the excess risk found in the record -linkage study (standardized mortality ratio 175), the total excess risk cannot be explained by an isolated excess risk in any of these subgroups.
Regard ing precision in relation to butchers and slaughterhouse workers in general, it might be of interest to look at the results independently from the record-linkage studies. It must then be assumed, very hypothetically, that the Swedish record-linkage finding was caused by chance or something other than the exposures under study. The risk increase among butchers and slaughterhouse workers in general, compared with the general population, was 1.20 and 2.50 in the English and Danish studies, respectively. The frequency of, for example, work with meat smoking was around 45 070 (estimated from the current study) . Suppose that the lung cancer risk among butchers and slaughterhouse workers not working in smokehouses is equal to the national average. If a risk of 1.5 (somewhere between the Danish and English excesses) is to be explained, it can be shown that the lung cancer risk among the smokehouse workers must be 1 + (1.5 -1)/0.45 = 2.11 times that of the unexposed. The upper limit of the 95 0J0 confidence interval for the relative risk for lung cancer among men working as meat smokers was 1.77 with group A as reference, and a true excess risk of 2.11 is therefore very unlikely . .If analogous calculations are performed for the other work tasks, the result s are the same for all the tasks except meat curing, for which the evidence is not as strong. Thus, also when the finding s of the Swedish record-linkage study are ignored, the result s seem to represent some evidence again st an occupationally related excess of lung cancer among butchers and slaughterhouse workers in general.
The precision of the study could also be discussed in an even more generalized context, ie, in regard to whether exposure to smokehouse dust represents a lung can cer hazard in general. The current study cannot be taken as evidence again st such a relationship, since the exposure to smoke is comparatively dispersed for smokehouse workers in slaughterhouses. The conclusions from this study are limited to smokehouse work performed by this occupational category only.
It has been shown that tobacco smoking habits have contributed to at least a part of the excess risk among Swedish butchers and slaughterhouse workers (12), and the present results indicate that only a rather small part of the excess risk is explained by occupational factor s. Ther efore , it seems reasonable to believe that tobacco smoking habits are the predominant factor for the excess amon g Swedish slaughterhouse workers.
Occupational hygiene investigations made in meat smoking plant s indicate that comparatively high levels of PAH are generated with the traditional method of smok ing meat and that the smoke contains cancerogenic PAH compounds. The reason for the negative findings in this study might be the comparatively short duration of the exposure because of the mixed work tasks prevalent earlier in slaughterhouses.