Risk of lung, larynx, pharynx and buccal cavity cancers among carbon electrode manufacturing workers.

I V, COU 1P . Risk of lung, larynx, pharynx and buccal cavity cancers among car bon c1ectrode manufacturing workers. Scand J Work Environ Health 1989;15:30- 37. Among workers employed in factories producing carbon graphite products the risk of cancer due to exposure to polycyclic aromatic hydrocarbons was estimated. In one cohort (plant A), a cancer incidence study was carried out; the number of cases were not significantly increased for lung cancers [7 cases, standardized incidence ra tio (SIR) 79] or for cancers of the upper respiratory and alimentary tract (10 cases, SIR 103). In another cohort (plant B), a mortality study was carried out ; neither the mortality from lung cancer [13 deaths, standardized mortality ratio (SMR) 118] nor that from upper respiratory and alimentary tract cancers (10 deaths, SMR 125) was significantly higher than expected. Within each cohort, a case-referent study was carried out. In plant A the odds ratios were high but nonsignificant for lung cancers (odds ratio 3.42) and upper respiratory and alimentary tract cancers (odd s ratio 2.19) and they showed a nonsignificant relationship with duration of exposure. In plant B, the odds ratios were low for every cancer site.

The incre ased risk of lun g cance r in humans as a result of exposure to chemicals, especially in th e occu pational envi ro nment , is either well known or suspected fo r seve ra l substances (1-4) . T hese exp osures ha ve been inves tigated in ma ny epidemiologic and experimental studies undertaken to co nfirm or clarify the risks of can cer (5). Reviews of these studies are contin ua lly carried o ut by the Internationa l Age ncy for Resea rch on Cancer (IA RC), who se recent publications ha ve been focused on the products derived from coal (6)(7)(8) . Indeed the industrial processes which use coa l and its byproducts generate diverse polycycl ic aromatic hyd rocarbons (P A H), whose carcino geno us po tential is suspec ted (2,6,(9)(10)(11)(12). Th is po llutio n is often o bserved in factories where coal is gasified , in co keworks, in iron a nd stee l fou ndries, and in coal-ta r disti llerie s, as well as in other plan ts which use coa l by-products (2,7,8).
The binding q ua litie s of coal ta r an d coal-tar pitch explain the ir wide use in ind ustry, mainl y for (i) the manufacture of electrodes fo r the production of aluminum, (ii) the manu facture of pitch briquettes for ind ustrial a nd dom estic heating , (iii) the formu lation of pro du cts used in road surfacing and the roofing of flat-I Nation al Institu te for Researc h and Safety , Van 30 topped bu ildings, (iv) the im pregna tion of refracto ry bric ks used in furnaces and blast furnac es, and (v) protection and wa terproo fing trea t ments for pipes a nd tanks . All these activities ha ve in common th e han dling and heat ing of coal-tar pitch and therefore the re lease of PA H.
T he lAR C mo nographs state that there is sufficient evidence that certain exposures in the coke prod uction industry a nd in coal gas ification are ca rci no gen ic to huma ns and that there is limited evidence of carcinogenicity to hum a ns in the aluminum product ion industry an d in iro n a nd stee l fou nding . Among the substances recognized fo r their carcinogenic risk s (grou p I) by lARC are , in pa rt icula r, coal ta r , coal-ta r pitch , a nd soo t (5). However the pr ocesses which use p itch as a pri ma ry material have not yet been prec isely eva luated because the se diverse activities ha ve not yet give n rise to a sufficien t number of epid emiologic st udie s to allow a well-found ed op inio n .
Thi s p ub lica tion presents the results of an epidemiologic study focused on the risk s of lung, larynx, pharynx, and buccal ca vity cancers in rela t ion to the product ion of electro des and other carbon pr od ucts manu factured with the use of coa l-tar pitch as a binder.

Plant description
The stud y included two plant s (factories A an d B) located in so ut hea ste rn France. The manu factu red products were electrodes used in electrometallurgy or in the aluminum industry; the electrodes weighed several tons each. Figure I shows the manufacturing processes, which consist of mixing raw materials, forming the crude paste, and then baking it. In addition, impregnation and graphitization were used in plant A only. During these different steps, handling the products used (coal-tar pitch in particular) and heating them to high temperatures may lead to the release of PAH in the form of dust during stocking and grinding or of vapors during baking, impregnation and graphitization, or of both dust and vapors during mixing and shaping. Plant B was set up in 1897 and plant A in 1895, but the manufacture of graphite electrodes in plant A has been taking place only since 1950.

Exposure measurements
Data on past environmental exposure to benzo(a)pyrene (BaP) were not available. In order to assess the present exposure, an environmental survey was carried out in 1983-1984. Atmospheric samples were collected on glass-fiber filters (Whatmann type GF/C). The PAH are either adsorbed on different sorts of dust or condensed to a solid state and can therefore be collected with a particulate sampling method. Two types of equipment were used. Personal samplers (Rotheroe Mitchell L2C) were carried by the operators for about half a shift (4 h). The filter diameters were 25 mm, and the flow rate was 2 limn. Stationary samplers were used to measure the general ambient atmosphere (Merlin Gerin APA 13 or Chaix Meca). They were equipped with filters 150 mm in diameter, and the flow rate was about 30 limn.
After the sampling, the chromatographic profile and concentrations of several four-to six-cycle P AH, including BaP for which there is a specific recommendation in France (ie, the mean value of BaP in the air must not exceed 150 ng/m'), were determined with high-performance liquid chromatography (HPLC). After benzene extraction with the Soxhlet apparatus, the P AH were separated and quantified with HPLC [CIS reversed-phase column (length 25 cm, internal diameter 4 mm), elution by methanol at a flow rate of I mllmin, detection by fluorimetry (AExc= 365 nm, AEm = 420 nm) and by ultraviolet absorbance (Aab = 284 nm)].

Cohort definition
The data used for the historical follow-up studies were collected by the physicians of the companies according to the availability of the sources of information. Consequently, a cancer incidence study was carried out in plant A over a II-year period from 1975 to 1985, while mortality data were taken into account in plant B throughout a 28-year period from 1957 to 1984. The cohorts included all male workers employed on I Janu-

Sources of data
The personnel office provided the following information for each worker: name, date and place of birth, dates of hire and resignation. Data on tobacco smoking were also supplied for most of the plant A workers. Detailed work histories were not available for all the workers '"f the two cohorts. For the identification of possible links between occupational exposures and respiratory cancers, occupational history and specific job descriptions were recorded only for the men selected for the case-referent study. There was no cancer incidence register in the district where plant A was located. Thus some diagnoses of cancer, such as skin cancers, whose prognosis is usually not too bad might be unknown. For plant B, the vital status of the workers no longer at the factory was ascertained either from the administrative records of the plant or from the registry offices of their birthplaces. Causes of deaths were given by the attending physicians, since death certificates cannot be used for mortality surveys in France.

Statistical methods External comparisons. Each member of the cohorts
contributed to the person-years at risk from the date of hire until the end of the follow-up or until the date of death if deceased. The workers lost to follow-up were assumed to be alive at the end of the study. The expected numbers were computed with two sets of references rates. It has already been mentioned that no cancer incidence rates were available for the general population of the district where plant A was located. Thus expected numbers were calculated with the use of incidence rates of another district, also located in southeastern France, not very far from plant A (13). This register gave mean incidence rates for men for the period 1979-1981. We checked that the age-adjusted cancer mortality rates of these two districts did not differ markedly. We computed the expected numbers of diagnoses for each cancer site by multiplying these rates by the person-years (five-year age groups Case-referent study. In order to take into account specific job exposures to PAH, we carried out a casereferent study. The cases were persons who had died of lung cancer or of upper respiratory and alimentary tract (ie, pharynx, larynx, buccal cavity) cancer in plant B or persons who had been affected by the same types of cancer in plant A. For each case, three referents were chosen at random from within the cohort, matched by sex and year of birth, as well as by smoking habits in plant A (nonsmokers, smokers, smoking habits unknown). The referents were chosen from subjects who had never been affected by respiratory cancer. A few referents who died before the date of diagnosis of the corresponding case in plant A or before the date of death of the corresponding case in plant B were excluded from the statistical analysis. The work histories of the cases and the referents were known since 32 the year of first employment. Each workshop they had worked in was given an exposed or unexposed status. The results of this study are expressed as the odds ratio (OR) values of the exposed workers as compared to those never exposed. The OR estimates and their variances were obtained by the conditional logistic regression fitted by the maximum likelihood method, and thus matching and the variable number of referents were taken into account (14). Smoking habits were not included in the analysis because this confounding factor was accounted for by the matching procedure when it was available (plant A) and was unknown for plant B.
PAH exposure was analyzed in three different ways. In a first analysis, the OR values of the workers who had been exposed for at least one year during their job history (ever exposed) were calculated; this group of workers was compared to never exposed workers (other manual or office workers). A second analysis took into account the length of exposure according to the following three groups: those exposed from 1 to 10 years, those exposed from 11 to 20 years, and those exposed for more than 20 years. The trend test was carried out by fitting a recoded factor based on these four groups (those never exposed and the three exposed groups) and by testing significance using the score statistic (14). A complementary analysis included the latency period by considering as exposed only those subjects exposed for five years or more along with a latency period of 10 years or more (time lapse between the date of the first exposure and the date of diagnosis in plant A and the date of death in plant B). These workers were compared to the never exposed ones, those with exposure of < 5 years and/or a latency period of < 10 years being excluded.
For each referent, the length of exposure to PAH and the length of the latency period from the beginning of exposure were taken into account until the year of diagnosis of the corresponding case in plant A and until the year of death of the corresponding case in plant B.  In the co hort fro m plant B, which includ ed a similar num ber of workers (N = I lI S), the vita l status was un known fo r ISO wo rkers (13.5 % ) who were bo rn abroad . Durin g the survey period, 164 death s were recorded. As sho wn in ta ble 3, the number of dea ths occu rr ing in th is cohort was slightly lower tha n expected (SMR 82). Mort ality from disea ses o f the cir-  b OR z =exposure~5 years and latency tim e;;, 10 years versus never exposed .

Exposure measurements
culatory system was also signi fica nt ly lower tha n expected (SM R 71, 95 070 C I 51-97), as was mortality from diseases o f th e respira tory syste m (4 observed deaths, SMR 38, 95 0J0 CI 10-98). Th e overall cancer mo rt ality did no t significantly differ fro m that expected , nor did the deaths from lung cancer (SMR 118) or fro m upper respiratory and alim entary tract cancers (SMR 125). The thre e death s from bladder ca ncer were more tha n the expected number, but not significantly so (SMR 194, 95 0J0 C I 40-566).

Case-ref erent study
The subjects considered as never expo sed were essentially manual workers present in works hop areas with 34 no know n exposure to PAH; o nly a few of them (4 0J0) were office wor kers. Table 4 shows the OR values for the ever exposed work ers (O R j ) and for a duration o f exposure of ;::: 5 years along with a latency period of   (table 1). In plant B, the least polluted factor y, the 0 R values were low, while in factory A the OR values were high and increased with the length of exposure. Th is finding will be discussed further.

Discussion
The manufacturing of electrode s involves handling coal-tar pitch; baking it at about 1 OOO°C, which releases almost all the volatile substances; and graphitization at about 2 800°C, which in certain cases is likely to generate PAH. Dur ing this activity, measurements of atmospheric BaP have been carried out in Sweden (2). The levelswere less than I 300 ng/m' near the baking furnaces . Excessive values (19000-40000 ng/ m') were observed in the impregnation workshops, but in the graphitization workshop onl y three values out of 24 exceeded 100 ng/m'. A Russian pub lication (16) reported concentrations between 195 and 4 730 ng/m ' in several factorie s. These values are of the same magnitude as thos e of the present study (tab le 1). Several epidemiologic studies have been carried out in tar distilleries and in industries using coal-tar pitch , mainly aluminum production, and the manufacture of briquettes and roofing. Among tar distillation workers, Henry et al (17) found a non significant excess of bladde r cancer deaths. More recently, unlike a French study (18) whose results were negative for lung cancers and upper respiratory and alimentary tract cancer s, a British investigation (19) concluded that an excess of lung and bladder cancer mortality was work-related. In the aluminum manufacturing industry, several epidemiolo gic studies have been carried out; these studies have been reported in detail by IARC, which concluded that there was limited evidence of lung and bladder cancer risk (7). Another French study was negative (20). In the manufacture of briquettes , several publications have reported clinical cases of skin cancer (21)(22)(23)(24)(25)(26). The study published by Henry in 1946 (22) reported a crude mo rtality rate of 504/10 6 due to scrotal cancer in worker s engaged in briquette production, whereas the national rate was 4.2/10 6 • In French publications (21,(23)(24)(25)(26), no incidence evaluation has been attempted, but it was underlined that the skin cancers were generally situated on the face and occurred after a long duration of exposure (10-30 years) in plants where high pollution levels had been measured (27). Among roofers and slaters (28)(29)(30)(31)(32) and lens polishers (33) significantly increased risks of lung cancer , and to a less extent of upper respiratory tract and digestive cancers, have been obser ved. Finally, several case-referent studies have shown urinary tract tumors in relation to " pitch" exposure, withou t the exposure being clearly stated (34)(35)(36).
Companies working with graphite products have not yet been the subject of numerous epidemiologi c studies. An investigation based on the mutagenic activity of carbon electrode worker s' urine has shown a relationship between exposure and mutagenic effect; smok ing habits were taken into account (37). An Amer ican epidemiologic study carried out in I I carbon products manufacturing plants has recentl y been published (38). The result s do not indicate an adverse effect of coal-tar pitch exposure on mortality. Although the levels of PAH exposure in the workplace atmos pheres were unknown , the authors suggested that these negative results could be explained either by the method' s " low sensitivity to detect mortality excesses that may exist in small subgroups of the study pop ulation [p 349]," or by the " lower concentrations and therefore exposure to suspect carcinogen PAH due to the lower temperature to which coal tar pitch is heated [p 349]" (38).
In the present study, some limitations must be discussed before conclusions are drawn. The validity of the external reference used to calculate the expected numbers of cases in plant A cannot be assessed. In the same way, the SIR values might have been underestimated because of the lack of a local cancer register and because of the persons lost to follow-up. However, the SIR values did not differ markedly from 100, except for cancers of the testes, which could be a random cluster, since no occupational factor has yet been pointed out. Both the low overall mortality observed in plant B and the significantly low SMR values for disease of the circulatory and respiratory system could be due to the "healthy worker effect." Furthermore, these low values could be partly explained by the lack of precision of causes of deaths ascertained after a long period and by the high percentage of subjects of unknown vital status.
The case-referent study carried out in plant B failed to reveal an association between cancer mortality and occupational exposure to PAH. On the contrary, in plant A, the OR values were high and related to the duration of exposure, although not significantly. This result contrasts with the lower SIR values and raises the question of whether the data collection could have been biased (more exhaustive among the exposed workers than among the unexposed ones). We checked that this was not the case by carrying out a case-referent study using, as cases, cancers other than lung and upper respiratory and alimentary tract cancers. The ORr was 0.75 (13 exposed cases out of 21; 41 exposed referents out of 59), the OR z was 0.61 (7 exposed cases, 22 exposed referents), and the OR values were not related to the duration of exposure. In the matching procedure, smoking habits were taken into account for plant A only, while they were unknown for plant B. One can regret this lack of information because cigarette smoke contains PAH (39). Nevertheless smoking habits are unlikely to be a confounding factor in this study, since nothing seems to indicate a connection with occupational exposure to PAH.
These results are not in conflict with those of present atmospheric pollution (personal samplings). The means and the ranges of the BaP concentrations were the highest in plant A (mean value 2 700 ng/m'), while the corresponding data from factory B were much lower (mean value 170 ng/m'). However these pollution rates only reflect recent work conditions (the years [1983][1984], whereas the cancer cases may be due to exposure dating from much further back, 1934 in plant A and 1911 in plant B. Thus it cannot be asserted that the classification of plants according to recent pollution measurements has remained the same throughout the last five decades. As a consequence their link with the observed risks must be considered with caution as it could be attributed to chance. 36 The low values of the OR estimates for plant Bare unlikely to be attributable to differential smoking or exposure misclassification. It must therefore be considered as a random value. One can notice nevertheless that the 95 070 confidence intervals of the overall risk ratio contain values as high as two.
In conclusion, this study was focused on the risks of lung, larynx, pharynx, and buccal cavity cancers in relation to exposure to coal-tar pitch or to P AH. The levels of risk were determined with both the duration of exposure and the latency period taken into account.
The OR values and trend tests suggest a relationship between exposure and the occurrence of these cancers in one plant, but the results are negative for the other one. A definite conclusion about the risk of respiratory cancer for workers involved in the manufacture of carbon electrodes cannot therefore be reached. A further follow-up study is planned.