Incidence of cancer among welders and other workers in a Norwegian shipyard.

among The incidence of cancer among 4778 male shipyard workers, including 783 mild steel welders, was inves tigated in a historical cohort study. The workers had been employed for at least three months between I January 1946and 31 March 1977. The incidence of cancer was observed from I January 1953 through 1986. The loss during follow-up was only0.9 OJ•• There were 53observed cases of lung cancer in the whole cohort versus 31.3 expected on the basis of the national rates for men. There was an increased incidence of lung cancer among the welders, with seven observed cases versus 3.2 expected. Twenty-two cases of bladder cancer were observed versus 15.2 expected. Two malignan t mesoth eliomas had occurred (0.7 ex pected). Smoking was likely to be a confounder in the present study. Due to concomitant exposure to asbestos, the results are inconclusive concerning the possible relationship between exposure to welding fumes and lung cancer.

Th ere has been co ncern for some years ab out a possible excess of cancer , especially lung cancer, among welders. Several epidemiologic studies indicate that welders exhibit a higher incidence of lung cancer th an selected reference populations (1)(2)(3)(4)(5)(6)(7)(8)(9), while others ha ve not reproduced these find ings (10, II). Asbesto s and smoki ng have been possible confounders in most of the studies (12). It has, however, been suggested that welders in general exhibit an approximately 30 0,10 grea ter incidence of lung cancer (than the non welding population) that cannot be ascribed to smo king or other occupa tiona l hazard s (13).
Fum es from stainless steel welding are mut agenic in bact eria tests (14) and cau se tran sformation in mammal ian cell transfo rmatio n assays (15). Fum es from mild steel welding have been found to be inactive in these test systems (15). On the basis of th is backgro und, it has been suggested that the observ ed excess o f lung cancer am ong welder s can be ascribed to highrisk subgro ups who weld stainless steel and are thu s exposed to hexavalent chro mium (16). If th is suggestion is correc t, on e could expect a coh ort of mild steel welders to exhibit only a moder ate or no excess of lung ca ncer.
The purpose of th e present study was to investigate the incidence of cancer amo ng welders and other ship- Reprin t requ ests to : Dr S Langard, Te lemark Centra l Hospita l, Depa rtm ent of Occupat ional Medicine, N-3900 Por sgru nn, Norwa y. yard workers. The welders were all shipyard workers who had mainly welded mild steel. The recognized excess of lung cancer in the shipbuilding industr y is generally att ribut ed to asbestos exposure (9,(17)(18)(19)(20)(21)(22). Hence it was con sidered important to include contemporary workers at the ship yard , not onl y for a po ssibilit y to study their incidence of cancer as such, but also to serve as internal reference groups for the welder s.

Shipbuilding
The stud y was conducted at a ship yard located on the southwestern coast of Norway. The yard was first establi shed in 1896. Initiall y the main products were fishing boats and steamboilers. The company expanded rapidly afte r World War II, partly due to a large amount of repair work . A total of 953 ship s were repaired from 1947 to 1951. Ha ving produced mainl y freighters aft er the war , the shipyard specia lized in building oil tankers from 1950 onwards. From 1971 to 1979 the yard's main product was gas carrier s. Since 1978 the major products have been large offshore stru ctures for the North Sea oil industry.
New ships were built in sections. Steel plates and pipes first entered th e work shop s for cutting, welding, and assembly work . The assemblies were then transported to the dock s for completion, including fitting out and insul ation . Repa ir work was performed at a separate dock, but parts, including asbestos-covered pipes, were transported to the workshops for further handlin g". The main work shops at present are the plate shop covering 10 000 m-, the welding shop (9000 m') , and the pipe shop (3700 m'). The yard has a separate mach ine shop of 2000 m' and sand blasting and painting halls.

Welding
Onl y a few workers welded before 1940, and riveting and welding were performed at the same time during World War II and the subsequent years. Th e number of welders increased substantially dur ing the postwa r period . Metal -arc welding was the method mostly applied in the first years of welding. Rutile and acidiccoa ted electrodes were used in the 1940s. In the earl y 1950s basic-coated electrodes partly replaced the rutile and acidic-coated electrodes and remained the most frequently used electrode. Gas-shielded welding was employed from the late sixties, initially metal inert-gas and metal active-gas welding. Later, from the mid-1970s on, tungsten inert-ga s welding also came into use.
Up until the early I970s welding was performed almost exclusively on mild steel. During the production of gas carriers, two ships delivered in 1973 and 1974 were equipped with tank s consisting of 9 070 nickel steel. Th is steel contained no more than 0.2 % chro mium. Three ships (1975)(1976)(1977) were equ ipped with aluminum tank s. Welding on stainless steel became more and more common during this period , particularly in pipe welding. Stainles s steel welding is now widely applied at the yard . Ho wever, up to 1973-1974, there had been virtually no stainless steel welding at the shipyard.

En vironmental monitoring
In 1973 a surve y of welding fume exposure was undertaken, initiated by concern about the untoward health effects of expo sure to nickel-containing welding fumes. The chromium and nickel content of the deposit was not known, but the welding wire that was most frequently used contained 70 % nickel and 13 % chromium. Nine welders using metal-arc or metal inertgas techniques on steel with a content of 9 % nickel were monitored for one week. The total welding fume concentr ation and the nickel and chromium content of the fumes was measured with personal samples . Welding was performed at the docks (in gas carrier tank s), in the plate and pipe shops, and in open and confined spaces.
Total fume concentrations ranged from 3.6 to 23.6 mg/m ', taken as weekly mean s, with 7.3 mg/m' as the median value for all the welders. No correlation was found between the fume concentration and the place of work . Th e concentra tions of nickel were between 0.11 and 1.97 (median 0.43) mg/m ' , and the total chro mium content ranged from 0.03 to 0.65 (median 0.12) mg/m' . An anal ysis fo r hexavalent chromium was not carried ou t. No measurements were available for earlier periods.

388
Ventiiation and personal protection In the 1940s there was a system for lead ing fresh air passively into conf ined spaces during welding at the docks. This system was replaced with fans in the 1950s. In the beginn ing the welders used du st masks made of felt and later switched to carbon filters. Aft er exhaust ventilation was pro vided at the dock s from the mid-1960s on, the conditions impro ved markedly. Personal respira tory prot ection and exhau st ventilation were provided in the shops in the early 1970s.
Wor kers who regularly wor ked in the vicinity of a welder were also exposed to welding fumes, but at lower concentrations . These worker s mainly included pipe fitters (who performed some welding themselves), shipwrights, burners (also exposed to fumes from acetylene cutt ing), and their helpers and apprentices.

Use of asbestos
Asbestos was used at the yard until the early 1970s for insulat ing in machine rooms and for person al protection . Chrysotile was almost exclusively used. All insulating work with asbestos was carried out by the yard 's own employees until the late 1960s, when an outside firm was engaged to insulate the new ships. In ships und er repair asbesto s was removed by the b oilerm akers and , to some extent , by pipe fitters, mechani cs, welders, shipwright s, and by their helpers and apprentices. Some asbestos was removed in the workshops , especially in the pipe shop. Insulation was mainl y performed by the boilermakers and , to some extent, by pipe fitters. Boilers and pipes were covered with asbestos plates, and the joints were filled with asbestos powder mixed with a bind ing agent. Th is mixture was made on the spot and genera ted substantial amo unts of dust.
Measurements of asbestos dust concentrations from the period of interest were not available . Pr esumably the docks have been the area most contaminated by asb estos, followed by the pipe shop .

Use of primers
In general the steel was initially unprimed . Howe ver, lead (not lead chro mate) was applied to some extent. Zinc chromate was introduced for priming ship shells at the docks in the mid-1950s, and the a pplication was performed by painters and dock workers. Some welding was performed on these primed shells. Iron -based primers were introduced in the 1960s and coincided with the primin g being per form ed befo re the steel plates entered the shops for cutting and assembly work . Zinc has also been a fume constituent in the occasional welding of galvanized steel.

Subjects and methods
The personn el register was used for compiling the list o f names of the employees at the shipya rd . The regis- Table 1. Format ion of the cohort.

Results
There were 716 deaths from all causes in the cohort versus 681.3 expected, and 252 cases of cancer were observed versus 244.1 expected (table 2).
Fifty-three cases of lung cancer were observed in the whole cohort against 31.3 expected, which is statistically significant at the 99 % level (table 2). The incidence of bladder cancer was also elevated, 22 cases versus 15.2 expected. The number of observed cases of colon cancer was significantly lower than expected (observed 8, expected 17.9), but the difference was less apparent when cancers of the colon and rectum were merged (observed 24, expected 30.2). There was also a deficit of stomach cancer (observed 14, expected 22.1). operators, foremen, tool makers, welding supervisors, electricians).
The Cancer Registry of Norway has recorded all new cases of cancer since I January 1953. The registration is more than 99 % complete. In the present study the incidence of cancer among the national male population serves as a reference entity. The expected number of selected cancers was estimated with the use of national five-year age-specific incidence rates for each year from 1953 to 1986. Events occurring between 1946 to 1952 were not available. Each worker was under observation from the beginning of 1953 or, if he was first employed at a later date, from the middle of the year he was first employed until the occurrence of death, emigration, or the end of follow-up.
Several workers had held more than one job at the yard . All calculations were made with each occupational category being viewed as a separate entity. Thus one worker may have contributed an observed cancer to more than one occupational category, but not more than once within each category or when the incidence of cancer of the total cohort was considered.
Confidence intervals were estimated on the assumption of a Poisson distribution of events with the use of a two -sided test of significance.  Information was obtained on 6239 workers. Five were excluded due to missing names or date of birth. The study was restricted to men who were first employed at the shipyard between I January 1946 and 31 December 1976, leaving 4887 eligible for further analysis. The starting date of I January 1946 was chosen because it coincided both with the establishment of an informative personnel register, which is currently in use, and the introduction of welding at the yard. The closing date of 31 December 1976 allows for a minimum assumed development time for cancer of at least 10 years for all participants.
The 153 female workers were excluded , as were 41 foreigners no longer resident in the country, and 24 workers who were deceased or had emigrated before the establishment of the Cancer Registry in 1953. Those 126 who had emigrated after I January 1953 were considered to be under follow-up until their date of emigration. Only 44 (0.9 0,10) of the eligible subjects could not be traced . The formation of the cohort is shown in table I.
Retired workers, including a former administrator from the personnel office, were temporarily engaged to evaluate the quality of the personnel register with respect to job titles and approximate dates of employment. Job titles had changed during the study period. With the assistance of these retirees' thorough knowledge of previous work conditions, the diversity of job title s in the register could be limited to 32. The inclusion criteria were set and the assignment of workers to the final job titles was completed before the data on health status were collected.
On the basis of common exposures at the study plant , all job titles were merged into IO different occupational categories. These categories contribute more person-years than the single job titles and thus provide a more robust basis for estimating an expected incidence of cancer. The occupational categories and job titles were as follows: welders; metal workers, occasionally welding (caulker burners, pipe fitters, coppersmiths); metal workers, seldom or never welding (shipwrights, riveters , boilermakers, grinders , blacksmiths); machine shop workers (mechanics, turners, filers) ; carpenters (carpenters, joiners, pattern makers); dock workers (laborers, scaffolders, dock workers); sandblasters and painters; assistants and apprentices; office and administrative personnel; and others (transport workers, guards, dustmen, storemen, crane Two cases of malignant mesothelioma had occurred (expected 0.7), one in a turner emplo yed in 1946 for less than one year and one in a shipwright emplo yed from 1950to 1965. The time between first employment at the yard and the occurrence of cancer was 34 and 35 years, respecti vely.
Lung cancer was in excess in several occupational categories (table 3). The excess was statistically significant among the machine shop workers (observed 8, expected 2.9) and among the assistants and apprentices (observed 38, expected 19.2). As most of the workers started out as unskilled , this latter group comprises thr ee-fourths of the entire cohort. Therefore, for all types of cancer, the results of this group closely resern-  bled those of the total cohort. High standardized incidence ratios for lung cancer were also found among the welders (observed 7, expected 3.2), the metal workers welding only rarely, and the dock workers. The excess incidence of lung cancer was the most pronounced among the shipyard workers first employed in [1950][1951][1952][1953][1954][1955][1956][1957][1958][1959] (observed 27, expected 13.0, P<O.OI) and in [1960][1961][1962][1963][1964][1965][1966][1967][1968][1969] (observed 8, expected 3.1, P < 0.05). The eight cases of lung cancer found among the machine shop workers had all occurred in workers first employed before 1960. There was no cluster of lung cancer cases by period of employment in the subcohort of welders. Table 4 shows the incidence of cancer at selected sites by occupational category. There was no statistical1y significant variation in risk for any cancer site between occupational groups, with the exception of a statisti-cal1y significant excess of cancer of the rectum among dock workers. However, when cancer of the rectum was merged with cancer of the colon, the persistent excess was no longer significant (observed 7, expected 3.3). A deficit of colon cancer was only seen among the assistants and apprentices. For cancer of the blad-der, the standardized incidence ratio (SIR) was nonsignificantly increased among the metal workers rarely welding (observed 4, expected 1.3). No statistical1y significant excess or deficit of cancer in al1 sites was found in any occupational category.
To allow for different presumed development times for cancer, analyses were performed according to time since first employment. The incidence of lung cancer and cancer of the bladder in the whole cohort was significantly increased 10 years and more after first employment in the yard. Seven of the eight lung cancers among the machine shop workers occurred 20 years or more after first employment.
In table 5 the cohort has been stratified according to total length of employment, and table 6 shows the incidence of lung cancer by duration of employment in selected occupational categories. A presumed IO-year development time is simulated by the exclusion of person-years until 10 years after first employment. There was no significant trend in risk with duration of employment for any cancer site except lung cancer (table 5). The excess incidence of lung cancer was the most pronounced for workers employed for • P<O.Ol. less than one year (observed 21, expected 7.5, SIR 280), and the standardized incidence ratio decreased to I 19 for workers employed for more than five years (table  5). The majority of these short-time employees had been engaged as assistants or apprentices (observed 18, expected 6.3), and the excess was also significantly elevated among the short-time dock workers (observed 4, expected 0.8) (table 6). For employees with a duration of employment of one to five years there was a statistically significant excess of lung cancer among the welders (observed 5, expected 0.9) and the machine shop workers (observed 4, expected 0.8). The incidence of cancers of the colon and rectum was marginally increased among the workers employed for more than five years (observed 14, expected 12.2) (table 5).

Discussion
In the present study an observed case of cancer was ascribed to an occupational category when a worker had at least three months of employment in one of the jobs constituting the category. Assignment to one category was independent of employment in another job at a different period of time at the yard. One worker may thus have contributed with a cancer case in more than one occupational category. With the use of this method the observations within the categories were not diluted by the pooling of the worker's total experience into one category only, and the observed number of cases in each category is a true representation of the cancers that may have occurred as a result of work exposure in the category at issue. The personnel register was the only source of information about each worker. Current use of the register and the fact that it was established comparatively late gives us reason to trust its completeness. The files on former employees have been kept fully independent of the reasons each one had for terminating work, and this practice implies that a possible loss of personnel cards has not been dependent on outcome of illness or death. As the register contained a substantial number of files on workers employed for less than three months, we have no reason to believe that there is any deficit of short-time employees in the cohort.
As a result of the control and corrections of the individual personnel cards by former employees, misclassification of workers by the job title given in the personnel register seems unlikely.
Only 44 (0.9 010) of the workers were lost to followup. Twenty-nine of these were assistants or apprentices. Only 15 skilled workers were not traced. These IS unidentified workers were scattered among several occupational categories; hence a possible outcome dependency of the loss will have had very little influence on the results. With respect to the emigrants it is unlikely that the outcome under study has had any bearing on the decision to go abroad.
A primary positive selection of workers may have affected the cohort at its inception , and this phenomenon could have concealed a possible association between work-related exposure and illness. However, its impact on the incidence of cancer is assumed to be limited. In some of the analyses a presumed "latency period " of 10 years from first date of employment in an occupational category was accounted for. Consequently, these effect estimates have not been attenuated by work experience etiologically unrelated to exposure in the job or jobs at issue (23,24).
The main finding of the present study was an excess incidence of lung cancer in the total cohort of shipyard workers (SIR 169). The magnitude of the excess was higher than that reported in some epidemiologic studies on shipyard employees (9,17,20), while other authors have demonstrated an approximately twofold risk of lung cancer for shipyard workers (19,21,22).
Exposure to asbestos is likely to explain part of the excess of lung cancer found in this cohort. The demonstration of an increased incidence of lung cancer in several occupational categories at the shipyard is in accordance with the results of other investigations (9,17,18,20,21). However, the results from these studies are inconsistent as to which occupational groups had an excess of asbestos-related cancers, probably due to diverse routines for asbestos handling or possibly to methodological shortcomings.
At the shipyard, asbestos had been handled mainly by the boilermakers, but also to some extent by the pipe fitters, mechanics, welders, and shipwrights. Bystander exposure was common, and the docks was probably the area where most of the asbestos was handled . As might be expected, an increased incidence of lung cancer was found among the machine shop workers, dock workers, metal workers who rarely welded (induding boilermakers and shipwrights), and welders. Exposure to zinc chromate primer might have been a minor additional causal factor among the dock workers (25).
The overall standardized incidence ratio for lung cancer among the welders did not exceed the corresponding standardized incidence ratio of the other occupational categories. The mean age at employment and the secular distribution of person-years were not so different as to invalidate tentative internal comparisons . Five lung cancers versus 0.9 expected occurred in the subcohort with 1-5 years of employment (table 6). Although this excess was highly significant, the result is difficult to interpret as there was no excess of lung cancer among the welders employed at the shipyard for more than five years. The extent to which possible previous welding experience at other shipyards or in other industries has influenced the results is not known for the different welder subcohorts.
There was no excess of lung cancer among the workers assigned to the category of metal workers occasionally welding (table 4). The pipe fitters, who had also been exposed to some extent to asbestos, constituted a great proportion of the se workers. The result cannot be explained in term s of variation in age at first employment or by the secula r distribution of personyea rs. With onl y 1.8 cases expected, the result is hampered by con siderable uncertainty.
The excess incidence of lung cancer in the total cohort decrea sed as the length of employment increa sed and was signi ficant only for the workers employed for less than one yea r (table 6). Th e excess of lung can cer among the short-term workers was th e most evident among th e assistants and apprentices, who formed a homogen eou s gro up befo re the educat ion of apprentices was properly organized in the early 1960s. A substantial number of these short-time employees had not been engaged in other jobs at the shipyard . The se workers thu s contributed several lun g can cer cases to the overall excess. It is likely that this group of unskilled worker s exhibi ts an increa sed lung cancer risk which ma y be related both to ind ividual disease determinants and to exposure factors at other workpla ces. For skilled workers the excess of lung cancer was the most prominent for the workers with a duration of employment o f one to five years. Th ese result s focus on the possibilit y of smoking acting as an important confounder in thi s study. However, no information was available on the smo king habits of the present co hort.
In 1964-1 965 a sur vey of smoking habits was performed in the 1893-1927 Norwegian birth cohort (26). Th e percentage of current plus pre viou s smokers was 77.6 a/o. Industrial workers exhibited a prevalence of 82 % versus 85 a/o for offi ce and ad ministrative emplo yees. According to the method de scribed by Axelson (27), the degree of confounding determined on the basis of the se numbers with regard to lung can cer caused by smoking alone amounts to a maximum standardi zed incidence ratio o f 110. A standardized incide nce ratio of about 170 would correspond to a percentage of non smokers among the referen ce population and the shipyard employees o f 50 and 20, respectively. In the present study such a difference seems unlikely.
The Axelson method (27) does not account for durat ion of smoking and lifetim e con sumption of tobacco . The uncertainties in applying the 1964-1965 survey of smo king (26) to the present shipya rd cohort also preclude firm conclusio ns concern ing the contribution of smo king to the lung cancer excess in th e different su bgroups.
Lung ca ncer rat es in some other Eur opean co untries (eg, Great Britain) are three times as high as in Norway . The lung cancer rate of wo rkers employed fo r less th an one year was thu s similar to that of an average British man. Differ ences in smo king could in fact account for larger differences in lun g ca ncer rates in Norw ay th an in most oth er western European countrie s.
Two cases o f mesothelioma were ob served again st 0.7 expected . Wh en j udging the expect ed figur es, one 2 has to take into considera tion that th e figure s fo r the general population are to a great extent due to asbestos exposur e. Th ere is co nsistent evidence that expo sur e to chrysot ile, which may ha ve occurred in the pre sent cohort, is associated with a smaller risk of mesothelioma than is exposure to amphiboles (28). As the estimated mean development time is about 30 to 35 yea rs, the observation period in the present study ma y be too short to reveal a po ssible excess. Th e small numbers preclude fur ther interpretation.
In relation to the observed defi cit of kidne y cancer (tabl e 4), it should be noted that the regional figur es for the incidence of thi s tumor are low. Th e part of the country where the shipyard is located had a deficit of 29 % of kidney can cer in the period 1966-1975 compa red with the national incidence (29).
The deficit of stomach cancer ma y also partly be expla ined by region al va riations in the incidence of cancer. Th e incidence of stomach cancer in the part of the country where the ship yard is located was 14 % lower than the national figures in the period 1966-1975 (29).
The excess incidence of cancer of the bladder (tabl e 4) is difficult to interpret in terms of occupatio nal exposure, and is suggestive of smoking acting as a confounder in this study. Exposure to asbesto s and tobacco smo king may explain the excess of laryn x can cer (table 5), but with only six cases observed the result may be due to chance.
It is unl ikely th at expo sur e to welding fume s constitutes a major lung cancer ha zard in the pre sent st udy, but a small excess of lung cancer caused by weld ing fume exposure ma y well be present in the figures. Further follow-up with identi fication of those exposed to fumes from stainless steel welding fro m the early 1970s onw ards, when also asbestos ceased to be used at the shipya rd , is needed to determine whether the excess of lung cancer amo ng the welders is related to sta inless steel weldin g.