Document gbGN95w31xkdM778Jdo9KKNBL

I Cancer Incidence and Cause Specific Mortality Among Workers in Two Norwegian Aluminum Reduction Plants Pal Romundstad,'."' Tor Haldorsen,' and Aage Andersen' Background Concern about the health hazatds in the aluminum industry has initiated this study where we have investigated associations between exposure to polycyclic aromatic hvdrocarbons (PAH) and juorides. and cancer incidence and cause specific mortaliiy among workers in two Norwegian aluminumplants in operation since I954 and 1957, respectively. Methods The stud? was designed as a historical cohort study and comprised 5627 identified men employed for more than six months. Cancer incidence was investigated from start of employtnent to 1995, and cause specific mortaliry was investigated from I962 to 1995. The observed cases of cancers and observed deaths were compared with expected numbers calculatedfrom national rates. Internal compurisons were made using Poisson regression with age and smoking included in the models. Historical exposure to PAH and fluoride had been estimated previously bv use of statistical modeling on industrial hygiene measurements and process parameters. A job e.rposure matrix was used to investigate possible associations between cumulative exposures, and cancer incidence and cause specific mortality. Smoking habirs were identifred for 92% of the cohort members. Results The study showed a signijcant excess risk for urinary bladder cancer among workers exposed to R4H, but no clear dose-response relationship. When usitig u 3Qyear lag period, a significant excess of bladder cancer in the highest exposure category I > 2000pg/m3 .year PAH) was shown (SIR 4.08).The data also suggested an association between exposure to PAH and pancreatic cancer,but no associationwith lung cancer was seen. The mortality analysis indicated an association between exposure to potroom .emissions Iffitorides)and mortality from chronic bronchitis, emphysema and asthma, bur no associations with cardiovascular diseases. Conclusions The studyjndings are compatible with an excess riskfor bladder cancerfor aluminum plant workers e.rposed to PAH. The increased risk for cancer of the pancreas indicated, should be further evaluated in larger e-rposedpopulations. Am. J . Ind. Med. 31:175- 183, 20oO. Q ZOO0 Wile-Liss. Inr. KEY WORDS: aluminum reduction plants; cancer incidence; mortality; sgderberg; PAH; fluorides; b W e r cancer; pancreatic cancer; COLD !lheCancer Registry of Noway. 090 *UnNersnyHosprtal ot TmdheimDepartment ot OccupatmnalMedlane Comract grant S~WMK The Nordicalummiurn Industrys Secretanatfor Health Emmn- men: anC Safety Contract grant sponsor Confederation of Norwegian Business and industry Correswndew 10 ?Ai A Romundslati. The Cancer Registry of Norway. Institute tor EpdemloiogicalCancer Researcn0310 Montebela Oslo. Norway -ma11pr@ kremegm Accepted 27 Septemoer 1999 INTRODUCTION Several epidemiological studies of aluminum plant workers have shown excess risk of urinary bladder cancer [Trembla).et 11.. 1995; Spinrlli et ai.. 1991: Ranneberg and Andersen. 19951, and some studies have also reported a moderately increased risk of lung cancer [Armstrong et al.. (Q 2000Wiley-Liss.Inc. ' 176 Romundstad et ai. 1994; Spinelli et al., 1991; Andersen et al., 19821. The are baked in furnaces at a separate facility before being excess risk has mainly been related to exposure to fitted into the pots, and they emit only small amounts of carcinogenic substances in coal tar pitch volatiles (CTPV) CTPV during electrolysis. in Splderberg potrooms. Other types of cancers found in The present study concernedtwo Norwegian aluminum excess among workers in aluminum reduction plants in reduction plants: Hydro Aluminium SUM& and Elkem epidemiological studies are cancer of the pancreas [Rock- Aluminium Mosjen. which have been in operation since ette and Arena, 1983; Milham. 1979; Carta et al., 1992; 1954 and 1957, respectively. The plants have mainly been Mur et al., 19871, kidney cancer [Gibbs, 1985; Spinelli operated by the Ssderberg process. The study focused on et al.. 1991; Rockette and Arena, 19831, lymphatic cancer quantitative exposure to PAH and fluorides, and cancer [Gibbs, 1985; Spinelli et al., 1991; Rockette and Arena. incidence and cause specific mortality. 19831, and leukemia [Andersen et al., 1982; Rockette and A description of the plants has been given elsewhere, Arena, 1983; Mur et al., 19871. In an overall summary where the historical exposure to polycyclic aromatic report, the International Agency for Research on Cancer hydrocarbons (PAW and fluoride was estimated objectively (IARC) concluded that there was sufficient evidence by using statistical modeling and industrial hygiene data implying that certain exposures occurring during alumi- describing the relationship between industrial hygiene num production cause cancer and sufficient evidence measurements and process parameters Romundstad et al., from animal carcinogeneity studies implicating several 19991. of the individual components of the CTPV W C , 19871. MATERIAL AND METHODS Epidemiological studies have shown associations between work in potrooms and mortality from cardiovas- Before identification. the study group included 5663 cular diseases [Th&iault et al., 1988; RBnneberg and men who had worked for six months or more at one of the Andersen, 19951, and both mortality studies and morbidity two plants studied (Table I). Information on each employee studies have shown relationshipsbetween work in potrooms was obtained from company records giving name, date of and chronic obstructive lung diseases (COLD) [Milham, birth, departments,jobs, and dates forjob changes. After the 1979; Ranneberg, 1995; Kongerud et al.. 19941. linkage of names. birth dates and available personal Aluminum is produced by electrolysis of alumina identification numbers to data from Statistics Norway, we (A1203) dissolved in molten cryolite (Na3AlF6). The were able to identify 5627 men by a personal identification electrolysis takes place in several carbon-lined steel shells number and the date of deaths and emigrations. (potlines). Aumina dust. solid fluoride salts. and hydrogen The cohort was divided into two subcohorts based on fluoride gas are emitted from the pots.The anode, which is length of employment. Short-term employment was defined situated at the top of the pot, is produced from coke with as less than three years of total employment and long-term coal tar pitch as a binder. The anode is consumed during the employment as three or more years of total employment. All process liberating carbon dioxide (COz), carbon monoxide workers available for follow-up contributed with person- (CO). and sulfur dioxide 602).Two main types of anodes years in the short-term subcohort based on the first three are used for electrolysis: continuous Ssderberg anodes and years of employment. the remaining person-years were prebaked carbon blocks. The Ssderberg anode is regularly allocated to the long-term employment subcohort. The supplied with unbaked anode paste at the top. The heating of splitting of the cohort was performed due to a lack of the uncarbonized paste in the SBderbeq anode leads to accurate job descriptions and to a substantial lack of emissions of coal tar pitch volatiles. The prebaked anodes smoking data for those employed for ~ S thSan three years. TABLE1. The Study Cohort,WorkersinAluminumReductionPlants.Norway Yair employees with at lrast six months rnploymellt Malae r p l o y m witb 1Imst threepan raplaymaat The Sunndalplant not idenbfied(excluded) The Mospxnplant not identified (excluded) Includedinthe cancerincrdencestudy excludedinthe mortalitystudydue to death before01.01.1962 Includedinthe mortalitystudy 2938 14 2725 22 5627 16 5611 2194 3 1867 2 4056 8 4048 1 ! Cancer and Mortality in Aluminum Plants 177 Dose-response analyses were only performed for those lete), asbestos, and electromagnetic fields. The estimation of employed for more than three years. exposure to PAH and fluorides in the Sderberg potrooms is described in another paper [Romundstad et al., 19991. The Cancer Incidence estimation w& based on statistical modeling of personal measurements, stationary measurements and process data in Each individual was observed for the event of cancer order to obtain valid and objective estimates of job specific from the time of hire. Observation continued until 31 exposures from the start of operation to 1995. The December 1995 or the time of death or emigration. For this estimation was performed in several steps. The relationship period, the Cancer Registry of Norway gives a complete between measured area concentrations and process para- coverage of the population for all cancer sites except basal meters were investigated by statistical modeling. Process cell carcinoma of the skin, which was excluded in the parameters and the models were then used to estimate area present analysis. The analysisof cancer incidence was based concentrations in periods lacking area measurement data. on the ICD-7 code (3-digit code. International Classification Next, the relationships between the area measurements and of Diseases, ICD-7) as registered by the Cancer Registry of job specific exposure (personal measurements) were Norway. In the analyses of cancer incidence we u s d the investigated by use of a measurement model. In the last standardized incidence ratio (SIR)and Poisson regression. step, the obtained relationships were used to estimate job SIRS were calculated as ratios between observed and specific exposure. The job specific exposure estimates for expected numbers of all cancers as well as site-specific PAH at the plants ranged from 2 to > 3000pg/m3 and the cancers. The expected numbers were calculated from five- fluoride estimates ranged from 0.1 to 1.7 mg/m3. The expo- year national cancer incidence rates among men by five- sure estimates of PAH and fluorides for jobs in departments years age groups. The 95% confidence intervals (95% CI) other than the potrooms were mainly based on personal were calculated assuming a Poisson distribution of the measurements. observed numbers. Personal sampling did not include the maintenance personnel (mainly mechanics and electricians). Mainte- Cause Specific Mortality nance personnel working in the potrooms were assigned half the average exposure estimated for potmen in the respective Each individual was observed from 1 January 1%2 or potrooms. Maintenance personnel working in the carbon time of hrre if later. Observation continued until 31 plant were assigned half the average exposure estimated for December 1995 or the time of death or emigration. In the the carbon plant workers. The assignment of exposure was mortality analysis, observations during the first six months based on annual reports from the maintenance department, of employment were excluded due to the inclusion criteria where maintenance work in these process departments was of minimum six months employment. Sixteen persons who estimated at about 50% of the total work time. died before the follow-up started in 1962 were excluded Exposure to asbestos (mainly chrysotile) was assessed from the mortality analysis. The mortality analysis was qualitatively based on the frequency and duration of dusty based on the classification code of death (3-digit code. work involving asbestos handling. Workers in the cast International Classification of Diseases, ICD-9)as regis- house, pothers. and metal tappers were assigned as tered by Statistics Norway. The observation period spanned exposed to asbestos. the 7th.8th. and 9th revision of the ICD and the codes in the Exposures to static and time varying magnetic fields 7th and 8th revision were, therefore, transformed to the 9th were estimated quantitatively based on the cell current and revision [RBnneberg, 19951. Cause specific mortality was the cell design. A previous survey of Norwegian aluminum analyzed by calculation of the standardized mortality ratio smelters showed a constant relation of about O.MmT//kA (SMR)and by use of Poisson regression for internal analysis. cell current for static fields. The time varying fields were The standardized mortality ratio (SMR) was calculated between 0.01% and 0.1% of the static fields. and the time as the ratio between the observed and the expected numbers varying fields were therefore calculated as 0.05% of the of deaths. The expected numbers were calculated from the static fields [Thommesen and Bjdseth. 19921. In the biannual national mortality among men by five-years age rectifiers the measurements indicated a level around 10pT groups. The 95% confidence intervals ( 9 5 8 CI) were for time varying fields [Thommesen and Bjdseth, 19921. As calculated assuming a Poisson distribution for the observed our estimates of static fields and time varying fields were numbers. strongly correlated. we only used the latter in the analysis. Exposure Data Smoking Habits The following agents were considered in h s study: Individual smoking habits abstracted from medical files total paniculate PAH. total fluorides (gaseous and particu- at the health departments of the smelters included smoking 178 Romundstad et al. data for about 80% of the workers employed for more than three years. Additional information was gathered for about 12% of the workers through interviews with long time employees at each plant. The data were categorized into never smokers, current smokers, former smokers, and persons with unknown smoking status. In all, we had complete information on smoking habits for 9 2 8 of the workers employed for more than three years. Of these 27% were never smokers, 53% were current smokers, and 20% were ex-smokers. The general distribution of smoking habits was similar at the two plants. Analysis of Dose-Response Relations Cumulative exposure was used as an indicator of individual dose. This was calculated for each person-year under observation as the product of the estimated exposure intensity and duration summed for all jobs held. Thus, each person could contribute person-time to more than one exposure category in each analysis. `Unexposed' persontime constituted one exposure category and the exposed person-years were divided into two or three cumulative exposure categories. We used two exposure categories if the number of expected cases was less than 25. In all the dose-response analyses involving cancer, we used a minimum lag of three years to exclude exposure in time periods when exposure was not likely to be related to cancer development. We also performed analyses applying lags of 10.20, and 30 years. In this study, lag was defined as the time period from causal action until the date of diagnosis [Rothman. 1981: Checkoway et ai., 19901. Poisson Regression Analysis Poisson regression analysis was used for internal doseresponse analysis and for investigation of potential confounding effects from smoking. The calculation of SMFt and the SIR and the Poisson regression mdyses were performed using the program package 'EPICURE` [Preston et al.. 19931. RESULTS The study cohort comprised 5627 men who contributed 139.554 person-years to the investigation of cancer incidence and 5611 men who contributed 128,020 personyears to the investigation of cause specific mortality. The 4056 men who satisfied the long-term employment criteria Contributed 88.468 person-years in the cancer incidence analysis. In the mortality analysis. 4048 men contributed 85,089 person-years to the long-term employment subcohort. Cancer Incidence As shown in Table II, the cohort's incidence of cancer did not differ significantly from the expected values. except for a low incidence of malignant melanomas. The SIR of 1.97 for multiple myeloma is mainly due to an excess at one of the plants, where we observed 9 cases among the longterm employees vs. 3.6 expected (SIR=2.5). A further evaluation of this unexpected finding showed that the increase was restricted to workers in the unexposed departments. We found an excess risk of bladder cancer among the PAH-exposed subjects, but no dose-response relation (Table UI). The SIR'Sfor bladder cancer showed no increase with increasing lag time. An exception is the highest exposure category (>2000 pg/m3 .year), where 30-years lag period showed a significantexcess SIR of 4.08. Ttus site was further evaluated by Poisson regression with a 30-year lag period in a model where age and smoking were included (Table IV). A non-significant increasing trend was shown ( P =0.06). no confounding from smoking was seen. There appeared to be a relationship between cumulative PAH exposure and cancer of the pancreas (Table V), but when cumulativeexposure to PAH was lagged by 20 and 30 years, the association gradually diminished (not in Table). The result of the Poisson regression analysis of cumulative PAH exposure (lagged by 10year). age and smoking, and the risk of pancreatic cancer is presented in Table VI. We found no indications of an association between cumulative exposure to PAH and the incidence of lung cancer (Tables VII, VIII). We also investigated the potential associations between cumulative exposure to asbestos and the incidence of lung cancer, as well as between cumulative exposure to magnetic fields and lymphatic and hematopoietic cancers and between cumulative PAH exposure and kidney cancer. but found no associations (results not shown). Cause Specific Mortality Table IX shows mortality from selected causes of death by short time employment and long time employment. Neither total mortality nor cause specific mortality differed significantly from the expected value. The all cause mortality (SMR=O.93) for those with more than three years of employment is at the same level as in other Norwegian industrial cohorts [Rsnneberg, 1995; Hobbesland ct al.. 19961. The mortality in the long time employment group was somewhat lower than in the short time employment group. There was no indication of an association between cumulative exposure to fluoride and mortality from cerebrovascular disease or between cumulative exposure to PAH and mortality from ischemic hem disease (results not shown). However, we found a dose-response relation- Cancer and Mortality in Aluminum Plants 179 TABLE II. Observed(Obs)and Expected (Exp)Numberof DifferentTypes of CancerandStandardized IncidenceRatio(SIR)by Lengthof EmploymentAmong 5627 MaleAluminum SmelterWorkersinthe follow-up Period from1954to 1995 Lastthan3 years rf employmelt (51,076 person-years) Momthan 3 years of employmelt (68,468 person-years) Cancer site An sites UP Stomach Colon Rectum parmeas wnx Lung Pleura Prostate Testes Kidney @kidder Mabgnantmelanoma other skin &am. nervoussystem Unspecified NonHcdgkins Leukemia Multiplemyeloma Another sites 140-204 140 151 153 154 157 161 162 163 177 178 180 181 190 191 193 199 201 200.202 204 203 - Obr Exp 83 98.5 0 1.2 7 6.3 10 7.5 4 4.8 2 2.9 0 1.3 15 12.1 0 0.4 6 12.6 6 4.8 3 3.7 5 6.4 3 6.0 3 2.9 3 4.6 1 3.4 3 1.5 1 2.8 2 2.4 2 1.4 7 9.5 SIR 95KCI 0.84 0.67-1.04 0.00 0.00-3.07 1.11 0.45-229 1.33 0.64-2.45 0.83 0.23-2.13 0.70 0.08-2.49 0.00 0.00-2.82 124 0.69-2.04 01x1 0.00-10.3 0.48 0.17-1.04 1.26 0.46-2.72 0.82 0.17-2.37 0.78 0.25-1.82 0.50 0.10-1.46 1.05 0.21-3.02 0.66 0.13-1.91 0.30 0.01 -1.64 200 0.41 -5.84 0.36 0.01-199 0.84 0.10-3.05 1.43 0.17-5.16 0.73 0.30-1.51 O b Exp SIR 95KCI 342 358.4 0.95 0.86- 1.06 2 4.2 0.48 0.06-1.72 25 23.1 1.08 0.70-1.60 20 29.7 0.67 0.41-1.04 20 19.4 1.03 0.63-1.60 13 11.5 1.13 0.60-1.94 7 5.1 1.37 0.55-2.38 46 49.3 0.93 am-1.24 0 1.5 OB0 0.00-2.46 61 57.7 1.06 am -1.36 7 7.8 0.90 0.36-1.85 12 14.0 0.86 au- 1.50 36 26.2 1.37 0.96-1.90 6 16.8 0.36 0.13-0.78 io -11.6 0.86 0.41 1.58 11 11.9 0.92 0.46-1.65 18 12.8 1.41 0.84-2.23 4 2.8 1.43 a39-3.67 5 8.9 0.56 ai8-1.31 9 7.3 1.23 0.56-2.34 11 5.6 1.97 0.98-3.52 19 30.8 0.62 0.37-0.96 ship between cumulative exposure to fluoride and mortality from COLD (Tables X. XI). As we had no cases in the lowest exposure group in the S M R analysis (XI,we had to change the exposure cutpoints for the exposure groups in the Poisson regression analysis*(Table XI). DISCUSSION Our study showed an increased risk of bladder cancer among aluminurn production workers exposed to PAH, indications of an increased risk of cancer of the pancreas and a relationship between mortality from chronic bronchitis, asthma, and emphysema. and exposure to potroom emissions (fluorides). We found no association between exposure to PAH and lung cancer. no association between magnetic field exposure and lymphatic and hematopoietic cancers. and no association between exposure to fluoride or PAH and mortality from cardiovascular diseases. Our finding of an excess risk of bladder cancer among the PAH-exposed workers was in accordance with several other studies, though in the present study the finding is weaker than in the other studies [Tremblay et al., 1995; Spinelli et al.. 19911. A case-control study nested in a large Canadian cohon of aluminum plant workers showed a strong dose-response relationship between exposure to coal tar pitch volatiles and the risk of bladder cancer. and only small changes were seen with increasing lag time [Tremblay et al.. 19951. A possible explanation for these differences might be differences in the length and the magnitude of exposure [Tremblay et al., 19951. In contrast to other studies of aluminum production workers and cancer [Armstrong et al., 1994; Spinelli et ai., 1991; Andersen et al., 19831, we were not able to find an increased risk of lung cancer. The risk of lung cancer among aluminum plant workers has, however. not been consistent among epidemiological studies. Length and magnitude of exposure as well as variations in exposure to other lung carcinogens (asbestos. smoking) might explain some of the discrepancies. But it is interesting that the high exposure to PAH. experienced by the workers in the Ssderberg departments. did not seem to result in an increased risk of . 180 Romundstad et al. TAbLEIII. Obsenred(0bs)and Expected(Exp)NumberofBladderCancer TABLEV. Observed(Obs)andExpectad(Exp)NumberofPancreaficCancer and Standardized Incidence Ratio (SIR) by Cumdative Exposure to PAH and Standardized incidence Ratio (Slfll by Cumutative Exposwe to PAH (Pm3Ye.ar) (Mm3'Y e 4 <so 50-500 500-2000 >2000 Lag = 20yeal <50 50-500 500-2000 >2000 8 9.45 0.85 0.37-l.67 <50 9 3.90 2.31 1.05-4.38 50-500 8 6.55 1.22 0.53-241 >500 11 6.30 1.75 0.87-3.12 Lag =10 year C50 14 12.59 1-11 a61-187 50-500 9 4.04 223 1.02-423 >500 5 5.56 0.90 0.29-2.10 8 4.01 1.99 0.86-393 2 4.21 0.47 0.06-1.72 2 1.71 1.17 0.14-422 9 5.53 1.63 0.74-3.09 2 4.61 a a o.os-i.s7 2 1.76 1.14 0.14-4.11 9 5.09 1.77 0.81-3.36 Lag =3 0 <50 50-500 500-2000 >2000 ~ 24 19.03 1.29 0.83-192 3 2.41 1.04 0.21-3.03 4 399 1.16 0.32-2.97 5 125 4.08 1.32-951 T U L E IV. Poisson R e g m s h Analysis of shdder Cancer R i by CumulativeExposureto PAH(pg/m3.year),Age, and Smoking Age <55 55-65 65-70 70-80 80+ 1.0 618 la69 2204 17.30 - 2.34-16.3 3.46-33.0 795-61.1 2.02- 147 TABLE VI. Paisson Regression Analyss of Pancreatic Cancer Risk by cumuhtive~ ~ p o s utomPAH (pg/m3.year),~ g ea,nd smoking VIllabk mntk I % U halmrbnnlbrt Age c60 60-70 70-80 1.0 - 3.68 0.64-3.56 8.17 1.89-35.2 239 274-208 SMWl Nevarmokers 1.0 - Smokers/formersmokers' 220 0.58-8.30 Unknown 3.77 0.61 -232 cumu~tivaPAHexposure (m/rn3y.ear. ha=10vear) -<50 1.0 50-500 >500 255 0.36-182 6.38 1.33-30.6 0.016 wdane Never smokers Fmnersmokers smokers unknown -1.0 4.56 039-20.9 4.69 1.10-20.0 1.23 0.11 -13.6 CumulativePAHexposure(m/m3.year). lag = 30year <so 1.0 - 50-500 500-2000 <2000 1.07 0.31-3.40 1.13 0.38-3.40 3.71 1.33-10.3 0.06 lung cancer in this study. Our use of national rates in the calculation of SIR instead of regional rates has probably lowered the overall SIR of lung cancer in the present study somewhat due to a low background rate in one of the plants region. This has not affected our conclusions of 00 association, as no increasing trend in risk with exposure was found in the internal comparisons. The indication of an increased risk of pancreatic cancers in this study, has also earlier been noted in epide- miological studies of aluminum workers [Rockette et al.. t Cancer and Mortality in Aluminum Plants 181 TABLE Vli. Observed (Obs) and Expected (Exp) Number of Lung Cancer TABLE VI11 Poiison Regression Analysis of Lung Cancer R i by -and Standardized Incidence Ratio (SIR) by Cumulative Exposure to PAH curr~tative&pasureto PAH(pg/rn3 year), ~ g ea,nd s m d t i (CIB/m3. Year) kmrl?AH apa8un ( p a d . par) Lag = 3year <50 50-500 500-2000 >2m Ob8 Ex) 17 17.51 a 7.45 . 15 12.51 6 11.84 Age SIR %%U <55 55-65 65-70 0.97 0.57- 1.55 70-80 1.07 0.46-1.98 80-c 1.20 0.67-1.90 a51 0.19-1.10 1.0 6.70 14.91 2221 1693 - 3.01-15.0 6.09-36.5 9.47- 521 2.17-132.2 Lag = 20ye.rr <50 50-500 500-2000 >2000 23 24.46 094 0.60-1.41 13 754 1.72 0.92-2.95 8 10.25 0.78 0.34-1.54 2 7116 '0.28 0.03-1.02 1983. Milham, 1979; Carta et al.. 1992; Mur et al., 19871. However, the low incidenceof this type of cancer demands a larger study population at risk to be more conclusive. The mortality analysis indicatedan association between exposun to pot emissions measured by cumulative fluoride exposure and death from chronic bronchitis, emphysema, and asthma. Th~sassociation was also to be expected since both morbidity and mortality studies have shown relationships between pot emissions and obstructive lung diseases [Milham, 1979: RQnneberg,1995; Kongemd et a]., 19941. cumurative PAHexposure(pg/rn3.year), lag = 3year <50 1.0 - 50-500 asl 0.37-2.25 500-2000 l.39 0.70-2.75 >2000 0.61 0.24-1.58 >a5 Recently, a paper has been published showing an increased mortality due to anythmia related conditions and due to acute myocardial infarction among electric utility workers exposed to alternating magnetic fields [Savitz et al., 19991. We, therefore, performed an additional analysis to investigate potential relations between magnetic field exposure and acute myocardial infarction, but we found no indication of an association. TABLE IX. Observed (Ob$ and Expected (Exp) Number of Deaths and Standardized Mortality Ratio (SMA)by Length of Employmentand Causesof Death Among 5611MaleAluminumSmelterWorkersintheFdlaw-up Periodfrom1962101995 krr than 3yean d amplamat (42.ari P-VUS) b t b 8 I 3 l@@Ofnrmpioyarclt (85,019 )mol-yrmm) cmrdlhrtli Ico-9 Godas Alcauses cancer Ischanlcheartd$e2% suddendeath Penpheralartenosclerosis betwovasculardisease Hypertermve disease Kidney disease chmrrcobstructivelungdisease pflemamlosls Extmcauses 1-999 140-209 410-414 798 440.444.785 431-438 401-405 580-587,791 490-493496 501-503.505 EBW-999 oh Exp 207 208.35 46 47.93 62 57.77 6 6.69 2 1.21 18 12.54 2 1.99 1 1.15 2 4.66 0 0.14 33 32.93 WR #Kfl 0.99 0.9-12 0.96 0.7-1.3 1.07 0.8-1.4 0.90 0.3-2.1 1.65 0.2-6.0 1.44 a9-23 1.00 0.1-3.8 0.87 0.03-5.5 0.43 005-1.6 0.00 0.00-27.4 1.00 0.7-1.4 Ob8 Qp an *SKU 712 766.27 0.93 am-1.00 180 189.73 0.95 0.82-1.10 240 240.52 1.00 0.88-1.13 21 25.21 0.83 0.52-1.27 5 5.36 0.93 0.30-2.18 48 53.51 0.90 0.66-1.19 5 8.31 0.60 0.20-1.40 6 4.04 1.49 0.55-3.23 18 20.49 0.88 0.52-1.39 0 0.60 0.00 0.00-6.18 60 69.83 0.86 0.66-1.11 182 Romundstad et al. TABLE X. Observed (Ob4 and Expected (Exp) Number of Deaths from total particulates and the fluoride measurements. The Chronic Obstructive Lunp Disease (Bmnchitis, Emphysem and Asthma) and fluoride exposure estimates might, therefore, be viewed as Standardbed Mortality Ratio (SMR) by Cumulative Exposure to Auoride a surrogate measure also for these agents. (mg/m3fluoride .year) The use of respiratory protection became substantial during the early eighties when about 70% of the workers C U a U k t h r @XpOSUN reported to have used such devices. Our exposure estimates flaoridr-mgh' .year Obs Exp SMR 95% Cl have not been adjusted for the use of respiratory protection. No kl as we had no data on individual use. This has probably not affected the results since the exposure was dramatically <a5 0 5.46 0.00 o.oo-a68 reduced during the same period of time Pomundstad et al., a5-4.9 6 5.89 1.02 0.37-222 19991. >5 12 9.14 1.31 0.68-229 We are reasonably sure that the exposure estimates used in this study described the actual exposure experienced by most of the workers [Romundstad et al., 19991. A potential limitation of the exposure evaluation of PAH is the possible TABLE XI. Poisson Regression Analysis of Mortality from Chronic qualitative differences in exposure to various PAHs or Obstructive Lung Diseases (Bronchitis, Emphysem and Asthma) by various components of the CTPV in different departments CumulativeExposure!o fluoride (mg/m3R u d e . year), Age, and Smoking and jobs. There is probably an important difference in the bio- Variablr Rateratlo 95KU Pralurtrendtest availability of PAH in various industries and settings due to differences in the bonding properties of P.4H to various Age particles. Dissolution or desorption of crystalline PAHs <60 1.0 - from a particle is normally a rapid reaction [Gerde et al., 60-70 70-80 6.29 1.66-23.8 27.49 7.994-95.2 19911. However, when the binding energies for adsorbed PAHs are sufficiently high. the hydrocarbon will be tied 80+ 80.55 14.7-442 almost irreversibly to the surface [Rivin and Atkins, 19871. High binding energies has been found for carbonaceous Smoking particles in particular [Rivin and Atkins, 19871. Recently, it Never smokers/ former smokers/ unknownsmokinghabits Smokers 1.0 17.3 - 2.8-130 Cumtdatiefluoride exposure(mg/rn3 year, nolag) -<2.5' 1.0 has been shown that particulate PAWat aluminum smelters are more strongly adsorbed in soot carbon matrix in both effluent and recipient waters of aluminum reduction plants than was earlier believed Waes et al., 19981. Compared to petroleum derived PAHs,they are less available to desorp- tion processes and,presurnably less bioavailable PESand Oug, 19971. The low solubility of the carcinogenic PAH relative to the greater solubility of less carcinogenic PAHs, 25-7.5 >7.5 3.26 a66-16.2 (Loo4 5.78 125-287 and a strong adsorbance of carcinogenic PAH to carbon soot and alumina, might only give a negligible mass flux of carcinogenic PAHs to the lung tissue. If this is correct, most of the potentially hannful PAHs might pass through the body, or at least the target tissue in the lung, without being metabolized. This might explain why an apparently massive The results from this study are not likely to have been exposure to P A H have not resulted in an increased risk of seriously confounded by smoking as the internal analyses lung cancer in this study. were only slightly affected after the inclusion and adjust- The strength of this study lies in the exposure ment for smoking habits. Although the smoking variable assessment, the available smoking data, and the possibility was crude, well known associations between smoking and of a reliable computation of cancer morbidity. The Cancer bladder cancer, lung cancer. pancreatic cancer, and COLD Registry of Norway gives a near 100%complete coverage of were obtained. all cancers in Norway based on information from patholo- We had difficulties in specifically disentangling a risk gical laboratories. compulsory reporting from physicians posed by an individual agent as several of the exposure and clinical departments. and death certificates from variables were correlated or appeared in mixtures (CTPV). Statistics Norway. Since 1953. about 85% of the cancers Industrial hygiene measurements that had been performed have been histologically verified and <2% are based on the indicated a strong correlation between exposure to SO?, death certificate alone. ! Cancer and Mortality in Aluminum Plants 183 A complete cohort with few unidentified persons should also give confidence to the results of this study. Further- more,the findings in this study are strengthened by the use of both internal and external groups for comparisons and by the fact that the findings were similar at the two plants investigated. ACKNOWLEDGMENTS Lyon,France: International Agency for Research on Cancer, 1987;8991 IARC scientific publications no. 8296-7, Kongerud J, B B J~. Seyseth V. 1994. Aluminum potroom asthma: the NowCgian CX@CXC. EUKRespir J 7(1): 165-172. Milham S. Mortality in aluminum reduction plant workers. 1979. J &CUP Med 21~475-480. Mur JM.Moulin JJ, Meyer-Biscb C, Massin N, CoulonJP.Loulergue J. 1987.Mortality of aluminum reduction plant workers in France. Int J Epiderniol 16257-264. We acknowledge the support from the project committee members Eirik Nordheim, Erle Grieg Astrup. Amid Bastiansen, Bj0m Bergan Skar, Per Iver Oksne, Sverre Laag&d and Bj0m Hilt. We thank Jahn Ivar Martinsen for valuable help in programming and data analysis, and H&on Lasse Leira for coostructive comments. Nas IC, Axelman J, Naf C, Broman D. 1998.Role of soot carbon and other carbon matrices in the distribution of PAHs among particles, DOC. and the dissolved phase in the effluents and recipient waters of an aluminumreduction plant. Environ Sci Technol32(12): 1786-1792. Nars K,Oug E. 1997.Multivariate approach to distribution patternsand fate of polycyclic aromatic hydrocarbons in sediments b m smeltcr- affected Norkgian fjords ad coastal waters. Environ Sci Techno1 31(5):1253- 1258. REFERENCES Anderscn A. Dahlbcrg BE, Magnus K. Wannag A. 1982.Risk of cancer in the Norwegian aluminum industry. Int J Cancer 29295-298. Andcrsen E,Potter JD,Mack TM. 1996. Pancrcatic cancer. In. Schonenfield D. Fraumcni JF Jr. editors. Cancer epidemiology and preveotion. New York.Oxford University Press. p 725-771. Armstrong B. Tremblay C. Baris D. Thiriault G.1994.Lung cancer mortality and polynuclear aromatic hydrocarbons: a case-cohort study of aluminum productiolr workers in Arvida. Quebec, Canada. Am J Epidcmiol 139250-262. Bmlow NE,Day NE. 1987.Statisticalmethods in cancer mearch: vol [I (Tk design and analysis of cohort studies). Lyoo: International Agency for Research on Cancer (IARC). Carta P. Cocco PL. Ron C, Pau M, Gmsu M, Cherchi P. 1992 Mortality in workers of a primary aluminum foundry in Portovesme in Sardinia. Mcd Lav 83(5):530-535. Checkoway H. Peam N. Hickey JLS. Dement JM. 1990. Latency aodysis in occupational epidemiology. Arch Environ Health 45:95- 100. Gerde P,Mcdinsky MA. Bond JA. 1991.Particleassociated polycyclic aromatic hydrocarbons-a reappraisal of their possible role in pulmoapry carcinogewis. Toxicol Appl Pharmwol 15:108 (1):l-13. Gibbs GW. 1985. M o d i t y of aluminum reduction plant workers. 1950 thrwgh 1977.J &CUP Med 27:761-770. Reston DL,Lubin JH. Pierce DA, McConney ME. 1993. Epicure. Seattle. Washington: Hirosoft International Corporation. 1993. Rivin D.Atkins DH. 1987.Sorption of bioactive material on carbon. Carbon 25:135-145. Rockette HE, Arena VC. 1983. Mortality studies of aluminum d u c t i o n plant worlrers: potroom and carbon depamnent J Occup MIA 25549-557. Romundstad P R Haldorsen T. Rsnneberg A. 1999. Exposm to PAH and fluoride in aluminum reduction plants in Noway: historical estimationof exposure using process parameters and industrialhygiene ~ Y I C a s U r e ~ fAs .m J Ind Med 35:164-174. Rothmao KJ. Induction aod latent periods. 1981. Am J Epidemiol 114253-259. Rsnneberg A, Andersen A. 1995. Mortality and cancer morbidity in workers from an aluminurn smelter with prebaked carbon anodespart 11: cancer morbidity. Occup Environ Med 52250-2.54. Rsnneberg A. 1995.Mortality and cancer morbidity in workers from an aluminum smelter with prcbaked carbon modes-part Ill: mortality from circulatory and respiratory diseases. OccG Environ Med 52255-261. Savitz DA. Liao D.Sasnc A. Kleckner RC. Kavet R. 1999. Magnetic field exposure and cardiovascular disease monality among elecaic utility workers. Am J Epidemiol 149135-142. Spimlli JJ, Band PR. Svirchev L. Gallagher W.1991. Mortality and cancer incidence in aluminum reduction plant workers. J @cup Med 33:1150-1155. Grandjcan P. Olsen JH. Juel K. 1993.Excess cancer incidence among Thcriault GP, Tremblay CG. Armstrong BG.1988. Risk of ischemic workers exposed to fluoride. Scand J Work Environ Health (19Suppl.) heart diseasc among primary alununum production workers. Am J Ind 1: 108-109. Med 131659466. Hobbesland A, Kjuus H.TheUe DS. 1996.A study of mortality among 14,730 male worircrs in 12 Norwegian Fmoalloy plants: Cohort characteristics and the main cause of death. Occup Environ Med 53(8):540-546. IARC. 1987.IARC Monographs on the evaluation of the carcinogenic risk of chemicals to humans. Supplement 7: Overall evaluations of carcinogenicity: an updating of LARC. Monographs volumes 1 to 42. Thommesen G. Bjelseth P. 1992.Static and low fnquency magnetic fields in Norwegian alloy and elecaolysis plants. Report 1992.1.Oslo: National Institute of Radiation Hygiene, 1992 (in Norwegian). Trcmblay C. Annstmng 5,Thiriault G.Brodeur J. 1995.Estimation of risk of developing bladder cancer among workers exposed to coal tar pitch volatiles in the primary aluminum industry. .4rn J Ind Med 27~335-348.