Incidence of cancer among workers in Norwegian hydroelectric power companies.

OBJECTIVES - The goal of this study was to examine whether exposure to electric or magnetic fields is related to cancer. METHODS - The study cohort consisted of 5088 men who had worked for at least one year between 1920 and 1991 for any of eight participating companies which produce and distribute hydroelectric power in Norway. The occupational exposure of these workers included extremely low-frequency elec tromagnetic fields. Incident cancer cases identified from the Cancer Registry of Norway were ana lyzed on the basis of the standardized incidence ratio with the Norwegian male population as refer ence. RESULTS - The incidence of cancer was close to unity for the cohort. The standardized incidence ratio for lymphoma was below unity, whereas those for leukemia and brain tumors were similar to those expected. Calculated cumulative exposure to electric or magnetic fields was not associated with the incidence of leukemia or brain tumors, but an excess of malignant melanoma was shown for the highest category of magnetic field exposure. An analysis of combined possible exposure to oils con taining polychlorinated biphenyls and exposures to magnetic fields or possible exposure to electric sparks gave standardized incidence ratios of 265 and 280, respectively, for the higher exposure category. CONCLUSIONS - These results do not support the assumption of a possible association between expo sure to electromagnetic fields and leukemia and brain tumors. The possible association between ex posure to polychlorinated biphenyls or magnetic fields and risk of malignant melanoma should be further evaluated in future studies.

Public interest in a possible association between expo sure to extremely low-frequency (ELF) electromagnetic (EM) fields and cancer has increased over the last decade. Several epidemiologic studies have suggested an increased risk for cancer, particularly leukemia and brain tumors, in various categories of "electrical occupations." Interpreting studies of workers exposed to EM fields has been especially difficult because of the lack of historical exposure data and biological support for the hypothesis that occupational exposure to ELF fields in the 50-60 Hz range can enhance cancer development. Furthermore, the respective roles of electric and magnetic fields have not been established, the results show no homogeneous pattern, and reviewers have suggested confounding factors and publication bias as explanations for the observed excess risks (l, 2). Currently, there is no clear experimental evidence to support any single interaction mechanism through which ELF fields could influence I Cancer Registry of Norway, Institute of Epidemiological Cancer Research , Oslo, Norway . Norwegian Radiation Protect ion Authority, 0stenl s, Norway.
Reprint requests to: Dr T Tynes, Cancer Registry of Norway, Montebello , N-0310 Oslo, Norway. cell membrane properties (3 , 4). It is not known whether peak field strength, cumulative exposure, or relative orientation of the field is important.
Two earlier Norwegian studies on cancer and exposure to EM fields have been carried out (5,6). A study on cancer incidence among workers with potential exposure to EM fields showed an excess risk for leukemia for those who had been employed for long periods, but no excess risk for brain tumors was seen (5). A nested case-referent study of railroad workers gave no support to the hypothesis that EM fields can promote cancer (6). The purpose of the present study was to usc calculated cumulative exposure to magnetic fields to determine whether exposure to EM fields during the production and distribution of hydroelectric power is associated with cancer.

Subjects and methods
The study cohort con sisted of 5088 male workers in eight large Norwegian hydroelectric power companies. Their job title indicated exposure to electric and magnetic fields , they had been employed for at least one year, and they had been first employed between 1 January 1920 and 31 December 1985. The cohort was establi shed from employment records available Scand J Work Environ Health 1994, vo l 20, no 5 at each company. The completeness was not guaranteed for employees dead or retired before 1950. The two largest companies contributed more than 50 % of the cohort. The records varied greatly from company to company but generally included the follow ing information: name, personal identification number, gender, date of birth , date of leaving, and summary of work history. For some employees information on jobs held before hire at the pres ent company were also available. Information on whether a man had ever smoked was also collected, although this information was not available for 12% of the cohort. Among the cohort members with such information available, 70 % were smokers or exsmoker s and 30% had never smoked.
All new cases of cancer in Norway hav e been recorded by the Cancer Reg istry since 1953 (7 ). The system is based on compulsory reporting by hospital departments and histopathological laboratories. All cause s of death on death certificates are coded by the Central Bureau of Statistic s and reported regularl y to the Cancer Registry. Can cers are coded accord ing to the Intern ational Classification of Dise ases, seventh revision. Site 193 (brain tumors ) includes all tumors of the central nervous system and mali gnant tumors of the peripheral ner vous system.
The cohort was followed from 1953 to the end of 1991 by use of the personal identification numb er given to all citizens alive at the census in 1960 or who were born in or immigrated to Norway later. The date of death or emigration and details of any cancer diagno sis were obtained for each individual. Only five (0. 1%) of the workers were lost to follow-up; six workers emigrated during the follow-up peri od.

Exposure assessment
No personal dosimeter measurements were available. The procedure for calculating exp osure to magnetic field s can be described as follo ws: a job title list and Table 1. Exposure to magnetic fields du ring an 8·h workday (typical value) fo r various worksites in power production and distribution, based on results of spot measurements (root mean square). job description was drawn up for each power company by long-time employees who could recall the time spent at different work sites and relevant exposure s in the job. All job titles with possible exposure to ELF EM field s were included. Spot measurements (root mean square) of magnetic fields were performed by two experienced electrical engineers at current worksites in the two larger power companies, and the results were used to determine a typical value (table 1). This informati on was combined with the job descriptions to establi sh current and past time-weighted-average exposure to magnetic field s (in microteslas) for each job title on the list . A category value (four levels ) for electric field s was also included according to maximum voltage in each job title . Other exposures, such as solvents, herbicides, asbestos , and cable oils , were also included in the job title list with the use of four levels according to frequency of exposure (never, monthly, weekly, and dail y).

Work history
In each company retired or current workers who had good knowl edge of historical data examined the employment records from company files to provide work historie s (all jobs held by each coh ort member ). The work historie s were then coded according to the job title list and merged with the exposure data. The expo sure of each cohort member was accumulated from the first year of employment in an exposed j ob up to the year of retirement or end of follow-up to give each subject's cumulative exposure to magnetic field s (microtes la-years) . Exposure to electric fields was calcul ated by multiplying the maximum voltage in the actu al job by the number of years employed. Cumulative exposure s to solvents, herbicides , asbe stos , and cable oil were obta ined by multiplying the exposure levels (0, I, 2, or 3) with the number of years employed . Three categories of expo sure to magnetic and electric field s were formed by allocating similar numbers of per son-years to each group . The average duration of employment was 22 years. Onl y I % of the co hort had worked less than two yea rs, and 48% had worked more than 20 years.

Statistics
The basic unit of statistica l computation was the numb er of years each employee was followed from the date of fir st employment to the end of the study period or death . Each year contributed by each cohort member was classified by age and calendar year , and the person-years of all workers were then summed by age and calendar year. In the cohort analysis, the standardized incidence ratio (SIR) for total cancer is given with the national male population as the referen ce entity. The expe cted numb er of cancer cases was calculated from the five-year, age-spec ific incidence rates for the reference ent ity for each year offollow-up (1953-1991). Ninety-five percent con-fidence interva ls (95% CI) were determined on the assumption of a Poisson distribution of the observed number of cancer cases and the use of a two-sided test of significance; a result was regarded as statistically significant if the 95% confi dence interval did not include 100. The actual computation was performed with a standard computer program (8).

Total cohort
During the follow-up period 1953-I99 1 ( 116 930 person-years), 486 new cases of cancer were observed (table 2). No significan t change in risk was seen for cancer at any site. The standardized incidence ratio was lower than unity for malignant lymphoma and greater than unity for stomach cancer and nonmelanoma skin cancer.

Ana lysis by time since fi rst empl oyment and duration of employment
An analysis of all cancers, brain tumors, and leukemia by time since first employment (data not shown) showed no change in incidence. For stomach cancer an excess risk was shown for workers with more than 30 years since first employment, and a rise in the standardized incide nce ratio was detected for malignant melanoma (the SIR rose from 73 in the category <20 years since first employme nt to 131 in the category >30 years since first employment). An analysis by duration of employment (table 3) showed a rise in the standardized incidence ratio for all cancers combined, cancer of the pancreas, lung (significant linear trend, Poisson trend statistics) and kidney, malignant melanoma, nonmelanoma, and lymphoma. Table 4 shows the number of cases of all cancer and the number for selected sites for groups of job categories. The excess risk for cancer at all sites for installation electricians was not assoc iated with duration of employme nt (data not shown). Two of the three cases of leukemia among the installation electricians occurred in workers who had been employed for less than 20 years. An elevated risk for nonmelanoma skin cancer was seen for installation electricians (data not shown, 3 observed, SIR 197) and for the "other worker" group (7 observed , SIR \75).

Analysis by cumu lative exposure
Tab le 5 shows the number of observed cases of cancer and the standardi zed incidence ratios for selected cancer sites by calculated levels of exposure to cumulative magnetic fields. An excess risk was seen for malign ant melanoma at cum ulative exposures with magnetic fields was seen for leukemia, and brain tumor showed a tendency towards a negative correlation. An analysis by calculated levels of exposure to electric fields showed no assoc iation for any site (data not shown) . The cumulative categorized data on workers possibly exposed to electric discharges showed no association for either leukemia or brain tumors (data not shown). The evaluation of risk for malignant melanoma and combined exposures to magnetic fields and possible exposure to electric discharges or to oils contaminated with polychlorinated biphenyls (PCB) showed a tendency towards an effect (table 6). The standardized incide nce ratio for lung cancer was analyzed   for combined expos ure to asbestos and oil-impregnated cables. No significant association was seen , although there was an increased standardized incidence ratio for the comb inatio n more than 20 years 342 of expos ure to asbestos and more than 30 years of expos ure to oil-impregnate d cab le (20 observed, SIR 163). The observed number of cases of lung cance r amon g eve r smokers was 55 (SIR 137, 95% CI 103-178). Neit her bladder cancer nor leukemia showed any association with smok ing.

Discussion
The present study showed no association between the occurrence of leuke mia or brain tumors and exposure to electric or magnetic fields in hydroelectric power companies. No exposure-response trend was shown for cancer at any site in relation to cumu lative exposures to magnetic fields, although there was an excess risk for malignant melanoma in the highest expos ure category. For brai n tumors, a tendency was seen towards a negative corre lation with magnetic fields. This study had the following strengths: (i) complete data with detai led job histories from employment records from the eight participating power companies and (ii) cancer incidence data from a highqua lity natio nal cance r register. Exposure was dete rmined on the basis of job title and department, however, and not from direct recordings for each individual; historical reco rds do not give inform ation on differe nces among workers within the same job category . The calculated cumulative exposure to magnetic and electric fields is, at best, a qualified est imate and may have introd uced misclass ification that is nondifferential (ie, independen t of the end po int) and in practice may have a magnitude and direction such that a linear, true exposure-response relationship would be disto rted so as to appear inconsistent with causation (ie, to bias the risk estimates towards unity) (9).
Mag netic fields near high-voltage power distr ibution systems change less over time than those associated with other electric insta llations (eg, railroads and some electric appliances). Very few of the workers (less than 10%) in our study had exposure to magnetic fields greater than 2 JlT. An earlie r study of leukemia and brain tumors in Norwegian railroad line workers, who experience an average 8-h timeweighted exposure to magnetic field s of 20 J.lT (frequency 16 2/3 Hz), showed no association (6). Workers in power comp anies are expo sed to a varying degree to electrical shocks or electrical spark discharg es. Discharges have been suggested to be associated with an increased frequency of chrom osome aberrations (10,II). In our study, however, no association was seen for either leukemia or brain tumor with categorized data on the possibility of such exposure. Although we controlled for categorized exposures to solvent s, herbicides, and cable oils in our analys is, a potential confounding effect of these parameters cannot be excluded. As the exposure data were based on information from experienced current and retired workers and no historical measurements were available, misclas sification may have occurred.
The selection of reference rate s in occupational epidemiologic cohort studies may be problematic when cancer incidence at selected sites has a geographic gradient. The incidence of cancer in urban area s in Norw ay is 20% higher than in rural areas ( 12). The fraction of urban residency in this cohort was 10% higher than that of the general male Norwegian population. A deficit of total cancer would therefore appear in our cohort if residency were taken into consideration. The null effect in our study may refle ct a possible " healthy worker effect" for selection into exposed occupations, for an incidence study on cancer this is, however, not very likely , especially not for leukemia and brain tumo rs.
Previou s epidemiologic studies on leukem ia risk among electri cal worker s have shown an association with such work, but most studies have not provided any evidence of a quantitative exposure-response relationship between risk and the level of EM field exposure (13)(14)(15)(16). An exception is a Swedish casereferent study of leukemia and brain tumor, which showed an association between magnetic field measurements (dosimeter) and leukem ia; for chronic lymphocytic leukemia in particular, a clear association with exposure intensity was observed (17). A recent cohort and nested-c ase referent study of electric utility workers, however, showed no consistent association betw een such work and leukemia, brain cancer, or lymphoma (18). The Norwegian census study of the incidenc e of cancer among workers potentially expo sed to EM fields showed a 40 % excess risk for leukemia in workers with a long duration of emplo yment (5). The finding that electricians in power supply and electri c line workers show elevated risks was not confirmed in the present study . Our cohort was relatively young, and at the end of the observation period (3 1 December 1991) 4099 (80.5%) co hort memb ers were still alive. The census cohort was, on the average, older (72% still alive at the end of the observation period) , but the excess leukem ia risk was not restricted to the older cohort members. Our cohort covers the largest Norwegian hydroelectric power producers and distributors, presumably with a more specialized work force than the census cohort. In additi on only a smaller fraction of the workers in our cohort was present in the census cohort, which had only 20% of the power supply electricians, 15% of the power plant operators, and 5% of the installation electricians and elect ric line workers. The excess risk seen previously for leukemia may therefore have occurred in similar job s in smaller companies where there is possibly higher exposure to other factors that contribute to leukemia risk , such as, for example , solvents.
An association between electrical work and brain tumors has also been reported in previous epidemiologic studies (14)(15)(16)(19)(20)(21) although negati ve results have also emerged (22,23). The Norwegian census cohort study (5) showed no exces s risk for brain tumors among electrical workers, and our findings do not support the hypothesis that exposure to electric or magnetic field s in power companies enhances the risk for brain tumor.
Previous studies have drawn attention to the possibility that melanom a might be related to EM field s. Studies in telecommunications and the electronic and electrical manufacturing industry have shown excess risks for malignant melanoma (24--27). Exposure to magneti c field s has been associated with changes in the diurn al pattern of the pineal hormone melatonin (4). The pattern s for malignant melanom a in our data may indicate a tendency toward s an interaction between exposure to polychlorinated biphenyls and other exposures, in particular exposure to electric sparks and magnetic fields (table 6). Nine of 11 cases among electricians in power supply were possibly exposed to polychlorinated biphen yls. As polychlorinated biphenyls were used as dielectric fluids in electrical capacitors until 1977, accidental exposure may have occurred owing to explosions or the destruction of old equipment. An association bet ween exposure to polychlorinated biphenyls and mortality from malignant melanoma has been sugge sted previously (28). The creation of polychlorinated dibenzofurans or other organochlorine compounds from combined exposure to polychlorinated biphenyls and electric sparks may be a mechani sm of interest (table 6).
An association has been reported between lung cancer and exposure to mineral oils (29). Our data gave no opportunity to confirm this finding , although a tenden cy towards an association was seen.

Concluding remarks
In summary, the results of this study give no support to previous findings of a possible association between exposure to electrical and magneti c field s and leukemia and brain tumor s. The results for malignant melanoma should be evaluated in future studies . Our results are based on a combination of data from a high-quality national cancer regi ster and detailed job histories obtained from employment records. These advantages must be we ighed against the limitations of lack of hi storical do simeter data for relevant exposures and limited knowledge about the biological effects of EM fields. As the classification of cumulative exposure to magnetic fields in this study may be an imprecise surro gate for a true biologically effective exposure, the results should be interpreted with caution.