Respiratory cancer mortality among workers employed in thermoelectric power plants.

Respiratory can cer mortality among workers employed in thermoelectricpower plants. Scand J Work Environ Health 1989;15:383-386. A historicalcohort studyof 406 workers wasundertaken to evaluatethe possiblehealth hazards of employment in traditional electric power plants with particular regard to cancer mortality. The total mortality was below that expected [observed40, standardized mortality ratio (SMR) 87] on the basis of national rates. Cancer mortality was slightly increased (SMR 112), mainly becauseof an excess of respiratory cancer (observed 9, SMR 171, 90 010 CI 89- 299). Lung cancer deaths were higher than expectedamong thoseless than 60 years of age and among those witha duration of exposureand a laten cy period since first employment of over 10years (observed 7, SMR 184). Furthermore, the increasewas more pronounced among maintenance workers and workers in general services. Previous data indicated that there had been past exposure to some known respiratory carcinogens (ie, asbestos, polycyclic aro matic hydrocarbons). This studysuggests that a possible lungcancer risk from traditional electricitygener ating plants should be considered and further studied.

In discu ssions on the future energy supply concern is often expressed about possible health hazards in nuclear po wer plants, whereas less attention is paid to the health implications of con ventional thermoelectric energy production . However, in a criti cal cros s-comparison of the risks of the two types of production system s, Cohen & Pritchard (I) had to conclude that no data were available on occupational can cer risks in oilor coal-burning power stations. More recently an increased risk of cancer among workers in conventional the rmoelectric power plants has been reported (2). In pa rti cular, pa st exposure to several known or suspected carcinogen s (asbestos , polycyclic aromatic hydrocarbons, hydrazine , chromium, nickel, and ber yllium) was con sidered a putative factor in the observed increase in total can cer mortality. Furthermore, there is another recent hypothesis which also seems to call for additional studies of power plant workers, namely, the possibility that electromagneti c field s, as present in pow er plants (as well as elsewhere), could ha ve some impact on human health (3). Since th e operating sta ff of a power station (both con ventional and nucle ar) is limited in size, it is difficult to produce large studies of these workers, but the health hazards may nevertheless be evaluated from the aggregated data of several studies.
In view of these circumstances the pr esent historical cohort stu dy was undertaken to clarify furthe r the I Epidemiologic Unit, Latium Regional Health Authority, Rome, Italy. 2 Local Health Unit RMI2I, Civitavecchia, Italy.
Reprint requeststo: Dr F Forastiere, Regional Epidemiologic Unit, Via Santa Costanza 53, 1-00198 Rome, Italy. pos sible health hazards connected with employment in two traditional thermoelectric power plants , with particular regard to can cer mortality.

Subjects and methods
Two plants located in Civitavecchia (Latium , italy) were included in the study. The first started operation in 1953 as a coal-fired plant producing a power of 70 MW ; in 1959, th e power was raised to 220 MW. In 1962 the fuel was changed to oil and, in 1969, the power was again increased, to 460 MW . The second plant ha s been op erating with oil since 1964 and produces a po wer of 200 MW; in the period 1969-1973 three new generating units (320 MW each) were added.
Historical measurements of airborne dust are not available, even though a possible expo sure to asbestos, especi ally during maintenance work, is likely to have occurred (4). In 1982, a special hazard control program was set up to avoid air contamination from asbestos fibers during insulating operations, ie, the local exhaust ventilation was increased, wetting was initiated , and control of waste through the use of pla stic sheeting was begun. Despite this program, in a survey conducted in 1985 during a routine maintenance outage, a range of 1.8-9.5 asbestos fibers (chryso tile asbestosj/cm" was fo und in four samples collected inside the protected area and 3.5 fibers/em' in a sample collected in th e surrounding atmosphere. It could be hypothesized that pa st exposure to asbestos had been even higher and more widespread in the general work environment, given the former absence of asbestos control measures.
Personnel records relati ve to all workers active on 1 January 1968 or tho se who were subsequently employed up to 31 December 1970 were made available by the company. A retrie val of all sub jects ever employed was impossible because of changes in the recording system. From these records we obtained the

384
following information for the employees: name , birth date, place of residence, dates of first and last employment at the facilities, and job titles. For each subject the vital status for the period 1 January 1968-31 December 1986 was ascertained either through the company records of workers still active or through the registry office of the last municipality of residence. Causes of death , retrieved from the registry office of the place where death occurred, were coded according to the eighth revision of the International Classification of Diseases.
A modified life-table anal ysis program was used to calculate the person-years for at risk of dying (5). Each subject was considered from I January 1968, or from the subsequent date of first employment, through the end of the study date (31 December 1986) or the date of death. Workers lost to follow-up were considered alive at the end of the follow-up. National causespecific death rates for men, stratified by age (five-year classes) and calendar period (five-year classes), for the period 1969-1983 were mult iplied by the corresponding person-years of the cohort. Mortality figures for the period 1969-1973 were also used for the year 1968, and the 1979-1983 national death rates were applied to the remaining years, 1984-1986.
The standardized mortality ratio (SMR) was calculated as the ratio between the ob served and expected deaths multiplied by 100. The 90 % confidence interval (90 % CI) of the SMR, based on the assumption of a Poisson distribution of the observed cases, was calculated with the use of Fisher exact limits. They correspond to confidence inter vals computed with a onetailed P-value of 0.05 .

Results
A total of 406 workers were ever employed at the plant s at some point between 1968 and 1970. Table I presents the frequency distribution of the study cohort by vital status at the end of the follow-up period, age at first employment, and duration of employment. Only a small proportion of the subjects was untraceable for vital status; the majority of workers were under 35 years of age when hired. Short-term emplo yees with a total of less than one year of employment have not been included in the study group, while 37.9 % of the subjects had worked for more than 20 years. The person-year s of observation were 7251; table 2 shows their distribution by the age of the employees.
Fewer than the expected number of deaths occurred for all causes (observed 40, expected 45.95, SMR 87, 90 % CI 66-113) mainly because of a "healthy worker effect" among those with a short latency time from first employment (latency -s 10 yea rs: observed 4, expected 8.50, SMR 47; latency> 10 years: ob served 36, expected 37.45, SMR 96). For most causes the observed mortality was not different from the expected (table  3). The total cancer mortality was slightly increased (observed 16, expected 14.31, SMR 112), mainly becau se of an excess in respiratory can cer (observed 9, expected 5.26, SMR 171), alth ough th e confidence intervals of the SMR (90 % CI 89-299) included th e null value .
Eight lung cancer death s were observed , while 4.50 were expected on the basis of nat ional mortal ity rates (SMR 178, 90070 CI 88-321 ). The SMR for lung cancer was raised among tho se less than 60 years of age (observed 6, expected 2.43, SMR 247, 90 % CI 108-487), while no increase was found amon g the older subjects (observed 2, expected 2.07, SMR 97). Ta ble 4 presents the relationship between lun g cancer, dur ation of emplo yment, and time since first employment. Onl y a few dea ths are pre sent in each of these catego ries, and a clear trend is not evident. Nevertheless, th e SMR values appear to be increased when both durat ion of exposure and lat ency period from first employment exceed 10 years. Workers with a latency time of 10 to 20 years or greater than 20 years had SMR values of 200 and 152, resp ectively. Furthermor e, the worker classified as having a short expo sure period and latency had been wor king for an additional eight years as an electrician (on contrac ted wor k) in one plant.
Finally, th e analy sis of lung cance r according to job titl e, as obtained from the company record s, showed that increased risk occurred ma inly among maintenan ce worke rs (mecha nics, insulator s, welders, electri cian s) (o bserved 4, expected 1.85 , SM R 216, 90 010 CI 74-495) and amo ng tho se empl oyed in general services (including cleaners) (observed 2, expected 0.74, SMR 270,90070 CI 48-851 ), whereas no cases were fo und among th e technicians (expecte d 0.30) and two cases were recor ded among the genera l ope rat or s (expected 1.58, SM R 127, 90 % CI 22-398).

Discussion
While a total cancer mortality similar to tha t expected from the genera l population was ob served in th e study group, some increase in lung cancer was found. Even though chance alon e could explain the observed increase, it is worth noting th at ther e are som e factors that seem to suggest a causa l role o f the workplace. Th e increase among younger ages, for instance, when certificatio n of dea th should be of high qu alit y, is characteristic and similar to what has been describ ed for some well-known lung carcinogens. Furthermore, the increased risk remained when a longer du rat ion of expo sure and lat ency period from first employment was considered. Interpret ation of the findings from the analysis by job titles is difficult because only a few lung cancer dea ths occurred in each category. On the oth er hand, the excess found among maintenance work ers and cleaners makes the association with the workplace more plau sible because such workers have presumabl y been heavily exposed to airborne pollutants, especially asbestos fibers. One of the lung cancer cases, in fact , had been working as a full-time insulator. With regard to the whole cohort , ho wever , the number of workers directly involved in hand ling asbestos was low (around 10), since all the insulating operations had gradually been transferred to outside contractors.
The increase found with regard to lung cancer is consistent with that determined in two other epidemiolo gic studies conducted on wor kers emplo yed in thermoelectric plants in nor thern Italy (6,7). Five and six lung cancer cases were observed in Tu rb igo (6) and nea r Venice (7), respectively, while 2.83 and 4.39, resp ectively, were expected on th e basis of general population mortality rates. The confidence intervals of the SMR values included the null value, but bo th studies had a low power to det ect a statistically significant associatio n.
With regard to the internal validity of the study some points should be considered . First, smoking represen ts a pot ential confounder in this study, but complete infor mation on this aspect was not available. On the oth er hand , we examined th e smoking habits of 235 active worke rs at the plants as reported du ring per iodic health examinations in the years 1986 and 1987. Reference figure s for smoking for the gener al populat ion were available from a national interview survey carri ed out in 1983 (8); in a comparison of the estimates of the prevalence of smoking, an age adju stment was appl ied with the age distribut ion of the working population at the plants as the standard. The prevalence of non smokers at the plants (32.3 070) was similar to tha t of the general population (34.8 070); curre nt smoke rs in the sample were less represented in com - Second, the Latium region, where the plants ar e located, has mean annual certificat ion rate s fo r lung cancer de aths a mo ng men that a re very simila r to th ose of Italy as a who le. Du ring the period 1969-1978, a directly standardized lung cancer death rate o f 56.5 (per 100 000 inha bita nts) was observed for Latium, whereas 53.9 was the rate recorded for Ital y. Th erefo re, the use of national refer ence rates instead of the Latium regional data could not explain the increased risk o f lung cancer.
Third , due to th e inco m pleteness o f th e reco rd s of subjects ever emp loye d , th e co ho rt co mpositio n was mainl y of a cross-sectio nal type (350 of the study member s wer e ac tive on I J anuary 1968), and o nly a sma ll proportion of th e wor ker s (13.8 % ) were hired during the three subseq uent years. Th erefore, the mortality experience of those retired before 1968 was missing . Since the o riginal populations co uld have been subjected to different degrees of selectio n pressur es, a possible selectio n bias co uld ha ve been int roduced. Co nsidering, how ever , th at the elapsed time from the sta rt o f plant o peratio n to the beginn ing of ob servat ion was not long , the extent o f such a bias sho uld be trivial.
Finally, it is worth noting th at the power of th is study to detect an increased risk of "lung cancer was lo w. Gi ven a n alpha o f 0.1 (two-s ided) a nd a n SMR of 150 a nd 200, th e probability o f detecting a significantly increas ed risk (I-beta) was onl y 0.27 a nd 0. 57, respect ively.
In co nclusio n , workers at two thermoelectric power plants seemed to incur some additio na l risk o f lu ng cancer, but it is di ffi cult to establish the precise eti ologic factor. Asbestos appears to be the most plausible agent , even if othe r subs ta nces pre sent in the work enviro nment and du ring mainten ance wo rk could have acted alon e o r in co mbina tion with asbe stos. In evalu-386 a ting the potential health hazards ass ocia ted wit h different electric ity production systems, a possible cancer risk from traditional electricity generating plants should be consid ered and further studied .