S16. PASCAL GUENEL,* JAVIER NICOLAU,* ELLEN IMBERNON, t GERARD WARRET:f: AND MARCEL GOLDBERG*

Transcription

1 International Journal of Epidemiology International Epidemiological Association 1993 Vol. 22, No.6 (Suppl. 2) Printed in Great Britain Design of a Job Exposure Matrix on Electric and Magnetic Fields: Selection of an Efficient Job Classification for Workers in Thermoelectric Power Production Plants PASCAL GUENEL,* JAVIER NICOLAU,* ELLEN IMBERNON, t GERARD WARRET:f: AND MARCEL GOLDBERG* Guenel P (INSERM U. 88, 91 Boulevard de l'hopital. F Paris, France), Nicolau J, Imbernon E, Warret G and Goldberg M. Design of a job exposure matrix on electric and magnetic fields: Selection of an efficient job classification for workers in thermoelectric power production plants. International Journal of Epidemiology 1993; 22 (Suppl. 2): S16-S21. Occupational exposure to 50 Hz electric and magnetic fields (EMF) was measured among 184 workers in thermoelectric power production plants using an individual portable dosimeter. A job exposure matrix (JEM) is elaborated from these data to be used in an epidemiological study on the potential carcinogenic effects of EMF. To reduce the range of exposure misclassification in the study, groups of workers with high exposure homogeneity must be identified. Classifying the workers by type of plant yielded homogeneous exposure groups, especially for workers in non-nuclear power plants. Workers in nuclear plants had higher mean exposure to magnetic fields, but the homogeneity of exposure was smaller. The exposure also differed between occupations, but the occupational title did not produce a uniform increase of exposure homogeneity within subgroups. It is concluded that the place of work is the most important determinant of exposure to magnetic fields for workers in thermoelectric power plants to be included in the JEM. In recent years concern has grown about the potential carcinogenic risks of Hz electric and magnetic fields (EMF) produced by electric distribution network and electric devices. Population-based studies have shown increased risk of leukaemia and brain tumours in children living in homes with high current configuration as assessed by the distance and type of overhead power lines around the houses. 1-3 Studies of occupational groups working with electrical equipment have shown increased risks of leukaemia, brain tumours, malignant melanoma, and male breast cancers." Recent reports from Sweden seem to confirm for leukaemia the associations observed in the earlier investigations. 5,6 In most occupational studies, job title * INSERM U.88, 91 Boulevard de I'H6pital, F-750l3 Paris, France. t Division Epidemiologic, Service General de Medecine du Travail, EDF-GDF, 22 Avenue de Wagram, F Paris, France. t Service de Medecine du Travail, EDF-GDF, Boulevard de la Marquette, F Toulouse, France. is used as a surrogate for exposure to EMF. Only a few studies included exposure measurements, and it was not proven that the occupations studied had higher exposures than background. Other occupational exposures that may be responsible for the excess risk were generally not accounted for. To clarify these results, new epidemiological studies on the EMF problem include assessment of occupational exposures based on actual exposure measurements. In an epidemiological study currently underway, the exposure is not assessed individually for the study subjects, but predicted from a job exposure matrix (JEM) elaborated from exposure measurements taken in other workers. The exposure data were collected among French utility workers (Electricite de France [EDF]), to be used in a large case-control study among the 150 ()()() workers of the utility scattered throughout the country. These data constitute one of the most impor- S16

2 JEM ON ELECTRIC AND MAGNETIC FIELDS IN POWER PLANTS S17 tant series of measurements on EMF available to date among utility workers. Separate JEM were.elaborated for the different sectors of the utility (hydroelectric and thermoelectric power plants, power transmission and distribution, central offices, etc.). The development of the JEM for thermoelectric power plants is described in this paper. Many different procedures for grouping the workers in the JEM are possible a priori, e.g. according to job title or to place of work. In the epidemiological study, a case or a control will be assigned the mean exposure of the group that he belongs to, but important departures from this can occur for individual workers. These errors in exposure assignment result in non-differential exposure misclassification that tend to reduce the strength of the associations between exposure and disease. Thus, the matrix is more efficient when the job classification leads to high exposure contrast between the groups and high exposure homogeneity within the groups. The data are therefore examined to look for the most appropriate classification, using several indices of exposure variability calculated from the exposure data collected in the survey. 7-9 MATERIAL AND METHODS Measurements of exposure to 50 Hz EMF were performed with 40 portable individual dosimeters POSITRON.lO The devices were used to record each minute the EMF in the environment. The records are ordered in the dosimeter into one of 16 predefined exposure intervals (0-0.61, , ,..., , > V1m (volts per metre) for electric fields and , , ,..., , >200 mg (milligauss) for magnetic fields). Each measure is assigned the value of the midpoint of the interval. Dosimeters were tested and calibrated before each use. Measurements were performed for one full week per worker, with the dosimeter placed in the shirt pocket or at the belt. Among approximately 50 thermoelectric power plants in the utility, eight were selected at random. These eight plants included four classical thermoelectric power plants (CTPP) generating power from coal or petroleum, and four nuclear power plants (NPP). All workers in the same plant were measured during the same week. Jobs in power plants were first grouped a priori on the basis of work tasks performed and of electrical equipment used. Six different occupations were defined in this way: sheet-metal workers, mechanics, electricians, production control workers, instrumentationlcalibration workers, and general maintenance workers. For each plant workers, representative of these job categories, were randomly selected for participation. In total, a minimum of 20 workers per job group was required. Additionally office workers working outside power plants and gas workers of the utility were included in the survey as a reference group with background exposure to EMF. At the end of the measurement period the records were entered into a microcomputer for analysis. Each worker's exposure profile was carefully examined and a few were excluded because of dosimeter malfunction. Only work days with at least 6 hours measurement were included in the analysis. The time-weighted average exposure was calculated for each worker and for each working day. Examination of the exposure distributions showed that the exposures were approximately log-normally distributed, and the analysis was performed on log-transformed data. ANOVA models were tested using NESTED or GLM procedures of the SAS statistical software. Two types of analyses were carried out. In the analysis at group level, the withinworker variance (or day-to-day variation) and the between-worker variance were calculated separately for each group of workers. The results are expressed as geometric standard deviations «( isd). The ratio A w of the within-worker variance to the between-worker variance is a measure of the homogeneity of exposure in the group of workers. An elevated ratio indicates that the exposure homogeneity is high. In a second type of analysis, the different procedures for grouping the workers are compared to each other. In the model, the between-worker variance is further divided into within-group and between-group variance. The ratio A G of the within-group variance to the between-group variance is a measure of the contrast between the exposure distributions of the groups. This ratio is low when the contrast is high. RESULTS Exposure to EMF was measured among 184 thermoelectric power plant workers for a total of 776 work days, and among 113 electric utility workers with background exposure for a total of 483 work days. Table 1 shows that the geometric mean «( im) exposure in power plants was 7.24 V1m for electric fields, a value that is only slightly higher than the background exposure (6.50 V1m, r-test P = 0.05). Work in power plants leads on the average to higher exposure to magnetic fields (2.06 mg) than work in offices (0.86 rrrg, z-test P < 0.(01). Table 1 also shows the withinand the between-worker geometric standard deviations (GSD) in each category. In power plants, the value of the within-worker variance represents approximately 40% of the total exposure variability for electric fields and for magnetic fields. The variance of exposure

3 SI8 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY (SUPPLEMENT 2) TABLE I Geometric mean and geometric standard deviations ofthe exposure to electric and to magneticfields among powerplant workers and among workers with background exposure Occupation N N GM GSD ww GSD BW A W GM GSD ww GSD BW A W workers days (Vim) (mg) Power plant workers Background exposure group GM: geometric mean; GSD ww = geometric standard deviations within worker; GSD BW = geometric standard deviation between worker; A W = ratio of the within-worker variance to the between-worker variance (0 2 WW/ 0 2BW)' between workers was greater than the day-to-day variation, with a ratio A w of 0.68 for electric fields and 0.90 for magnetic fields, indicating poor exposure homogeneity in this occupational category. The same observation applies to the background exposure group. To explain the differences between workers and to improve exposure homogeneity, workers in power plants were further divided according to various occupational characteristics. The following classification criteria were examined: the plant where measurements were carried out; the type of plant, i.e. CTPP or NPP; the levels of power production during the week of measurement «200 MW, MW; MW; >900 MW); the occupation (six groups as listed above); the sociooccupational level (engineers/technicians, foremen, skilled/semi-skilled workers). Three criteria turned out to be significant predictors of exposure variation after accounting for the others in multiway analyses of variance: the plant, the type of plant, and the occupation. The multiway analysis of variance combining these three factors is shown in Table 2. The effects of both plants and type of plant are strongly significant. Occupation, however, is not significant for electric fields, and is at the limit of statistical significance for magnetic fields. Simplified job classifications were also examined, but the effect of occupation was only marginally affected. It should be noted that the plant itself cannot be used as a classification criterion in the JEM as only eight plants of the utility were measured among many others. As a result two JEM are possible for workers in thermoelectric power plants. The first would include only two categories, CTPP workers and NPP workers. The other would be based on the combination of the type of plant and of the occupation, with 12 (6 x 2) occupational categories. These two JEM are analysed below. The components of the variance calculated with these job classifications and with the classification by plant are shown in Table 3. As expected, the withinworker variance was similar to that in Table 1 «( isd ww = 1.98 for magnetic fields and GSD ww = 1.90 for electric fields) and was not affected by the different job classifications. In Table 3, the between-worker variance is split up in between-group and within-group variance. The proportion of the variance between workers explained by the variance between groups was expected to be higher in the classification by plant and occupation combined than in the classification by plant alone. This proportion is only 21ltJo for electric TABLE2 fields Analysis of variance for the combined effects ofplant, type ofplant and occupation on the exposure to electric fields and to magnetic Factor df Mean square F value P Mean square F value P Plant Type of plant Occupation Residual

4 JEM ON ELECTRIC AND MAGNETIC FIELDS IN POWER PLANTS S19 TABLE 3 Comparison of different classification procedures for workers in thermoelectric power plants GSD BW GSD BW Job N GSD 02 ww GSD WG GSD BG BG / A G GSD ww GSD WG GSD BG 02 BG / classification groups BW BW A G Type of plant Occupation x type of plant Plant GSD WW = geometric standard deviations within worker; GSD BW = geometric standard deviation between worker; GSD WG = geometric standard deviation within group; GSD BG = geometric standard deviation between group; 2 0 BG = between-group variance; 2 0 BW = between-worker variance; A G = ratio of the within-group variance to the between-group variance (0 2 WG/ 0 2BG)' fields and for magnetic fields. Higher proportions are obtained with the classification by plant, especially for magnetic fields where more than halfthe between-worker variance is accounted for. The ratio A G shows that the within-group variance is more than twice the between-group variance in the proposed JEM, indicating strong overlap of the exposure distributions between the groups. In the classification by plant, this ratio decreases to < 1 for magnetic fields. The detailed results of the classification by occupation and by type of plant are presented in Table 4. The mean exposures to EMF are higher in NPP than in CTPP, and the same is true for any occupation with the exception of production workers. For electric fields, most groups had exposures lower than or close to background exposure (Table 1). When the workers in each type of plant were classified by occupation, the exposure homogeneity was decreased or only marginally increased. For magnetic fields, the highest exposures were found among electricians and instrumentation workers in NPP. Overall, the exposures to magnetic fields were relatively homogeneous among CTPP workers as the within-worker variance was 1.36 TABLE 4 Geometric means and geometric standard deviations by type ofplant and by occupation for workers in thermoelectric power plants N N GM GSD ww GSD BW A W GM GSD ww GSD BW A W workers days (Vim) (mg) CTPP, all Sheet-metal workers Mechanics Maintenance workers Production workers Electricians Instrumentation workers NPP, all Sheet-metal workers Mechanics Maintenance workers Production workers Electricians Instrumentation workers CTPP = classical thermoelectric power plants. NPP = nuclear power plants. Notes: see Table 1.

5 S20 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY (SUPPLEMENT 2) times greater than the between-worker variance. When workers in CTPP were further classified by occupation, the exposure homogeneity increased in four occupations out of six. The two exceptions were maintenance and production workers, for whom the ratio A. w was equal to or close to 1. In the group of NPP workers, the between-worker variance was higher than the day-to-day variation. Splitting by occupation also showed increased homogeneity of exposure in four groups out of six. Sheet-metal workers and mechanics, however, had a high between-worker variance above 2.4, and decreased A. w ' DISCUSSION Previous data on the exposure to EMFofelectric utility workers are very sparse.ui'? It was important to provide new information on this group of workers, since it is part of the set of 'electrical' occupations for which increased cancer incidence or mortality has been reported in several epidemiological studies. The present data show that the average exposure to magnetic fields is more than twice the background for workers in power production plants, but they do not show particularly high exposures to electric fields. The quantitative exposure estimates presented in this paper constitute an important part of the on-going epidemiological study. They will be used for identifying exposure groups and for analysing the relationship between increasing EMF exposure levels and cancer risk. Since the study includes more than 10 ()()() cases and controls selected among French utility workers, exposure measurements could not be obtained for each individual. The average exposure intensity measured in a sample of workers classified according to the type of plant and/or occupation will be used to estimate the exposure levels of all subjects. The main objective of this study was therefore to identify an efficient criterion for classifying thermoelectric power plant workers into a few subgroups with well-contrasted exposure levels. The data first showed that the plant where exposure values were recorded is the most important determinant of exposure. It explains most of the between-worker variance for magnetic fields. It has been pointed out, however, that a classification by plant cannot be used in the JEM since the exposure data were collected from eight plants only in the utility. In addition, measurements were performed at the same time for all workers in a given plani. To know whether the particular phase of the production cycle during the week of measurements may have influenced the exposure level, information about the production of the plant was obtained. The examination of the mean exposure per plant did not show any clear pattern of variation with the levels of power production (from 0 to 250 MW in CTPP and from 400 to 1300 MW in NPP). To explore this question further, additional measurements collected in the same plant at different points in time of the production cycle would be particularly useful. However, because the levels of power production represent a wide variety of situations, the present data can be considered as good estimates of mean exposure over the entire production cycle. Independent of the plant itself, the type of plant is also an important predictor of exposure to EMF, with work in NPP entailing exposure to magnetic fields much higher on average than work in CTPP. This difference in exposure may result from differences in plant design and equipment associated with differences in maximum power production capacities «400 MW for CTPP and generally >900 MW for NPP). Overall, workers in CTPP constituted a group with more homogeneous exposures than workers in NPP. The range of power production capacities of NPP is large and several types of nuclear plants exist. Additional data should be collected on this type of plant to examine further the characteristics that may account for the relatively high variability of exposure. In total, because the exposures of workers in CTPP are relatively homogeneous, and because exposure homogeneity could be improved for workers in NPP, type of plant is an important criterion that should be included in the JEM. The role of occupation is less clear. It had no significant effect on exposure to electric fields when accounting for the place of work. Exposure to electric fields was close to background levels, and the homogeneity of exposure was not clearly increased when workers were classified by occupation. Therefore, for electric fields, occupation can be omitted from the JEM. For magnetic fields, the effect of occupation on the exposure was at the limit of statistical significance. Overall, the use of occupation to classify workers, in addition to type of plant, increased slightly the exposure homogeneity within groups (A. w ) and the exposure contrast between groups {A.J. Thus, a JEM for magnetic fields including both types of plant and occupation can be recommended. Long-term (e.g. one year) mean exposure per worker is the relevant parameter to be used in the epidemiological study, and the mean exposures calculated in this survey are reasonable estimates of average exposures over long periods. The betweenworker variance shown in Table 4, however, includes other components of exposure variation that could not be estimated from the data, i.e. exposure variation that

6 JEM ON ELECTRIC AND MAGNETIC FIELDS IN POWER PLANTS S21 was accounted for variability between worker, but that is actually due to variability from one period of the year to another. Unlike point estimates of average exposure, the data may therefore overestimate the between-worker variance that would have been calculated over long periods, or similarly underestimate the homogeneity of exposure groups. Thus, the procedure for grouping the workers in the JEM might be more efficient for epidemiological purposes than what appears from this analysis. It should also be pointed out that unlike many other occupational exposures, the exposure to EMF does not occur exclusively in occupational situations. Because of the noise introduced in the data by the background exposure of any worker, the identification of well contrasted exposure groups might be particularly difficult for EMF. The epidemiological study has a retrospective design. Thus past exposures have to be estimated from knowledge about changes in working conditions and in electrical equipment that occurred in the past. The analysis of the exposure data is useful to derive estimates of past exposure. In thermoelectric power plants, it was shown that the place of work is particularly important for predicting the exposure. It seems reasonable to assume that there have been few changes in the exposure of workers in a given type of plant, as long as there have been no major changes in the technical design of the plant. In summary, this paper showed that the exposure to magnetic fields among workers in thermoelectric power plants was clearly higher than background. More heterogeneity of exposure was observed for magnetic fields than for electric fields. The type of plant was found to be the most important characteristic to be included in a JEM, but the role of the particular occupations within a plant is less clear. Similar analyses in the other sectors of the company are performed. One can expect from this validated JEM that more appropriate exposure groups will be identified and compared in the epidemiological study. Improvements of the JEM for power production plants could be obtained if more measurements were available. ACKNOWLEDGEMENTS The authors are indebted to Bernard Hutzler for his skilful assistance on the technical aspects of the exposure measurements and to Alice Gueguen for her help in the statistical analysis. REFERENCES 1 Wertheimer N, Leeper E. Electrical wiring configurations and childhood cancer. Am J Epidemiol1979; 109: Savitz D A, Watchel H, Barnes F A, John E M, Tvrdik J G. Casecontrol study of childhood cancer and exposure to 60-Hz magnetic fields. Am J Epidemiol1988; 128: London S J, Thomas D C, Bowman J D, Sobel E, Cheng T C, Peters J M. Exposure to residential electric and magnetic fields and risk of childhood leukemia. Am J Epidemiol 1991; 134: Advisory Group on Non-Ionising Radiation. Electromagnetic Fields and the Risk oj Cancer. Documents of the NRPB Vol. 3, No. I Didcot, UK: National Radiological Protection Board, Floderus B, Persson T, Stenlund C et al. Occupational Exposure to Electromagnetic Fields in Relation to Leukemia and Brain Tumors. A Case-Control Study. Solna, Sweden: National Institute of Occupational Health, 1992, pp Feychting M, Ahlborn A. Magnetic Fields and Cancer in People Residing near Swedish High Voltage Power Lines. IMMrapport 6/92. Stockholm: Institutet for miljomedicin. Karolinska institutet, 1992, pp Rappaport S M. Assessment of long-term exposures to toxic substances in air. Ann Occup Hyg 1991; 35: Kromhout H, Oostendorp Y, Heederick D, Boleij J S M. Agreement between qualitative exposure estimates and quantitative exposure measurements. Am J Ind Med 1987; 12: Heederick D, Boleij J S M, Kromhout H, Smid T. Use and analysis of exposure monitoring data in occupational epidemiology: An example of an epidemiological study in the Dutch animal food industry. Appl Occup Environ Hyg 1991; 6: Heroux P. A dosimeter for assessment of exposures to ELF fields. Bioelectromagnetics 1991; 12: The EMDEX project: Technology Transfer and Occupational Measurements. Volume 1: Summary. Palo Alto, California: Electric Power Research Institute, EPRI EN-7048, Deadman J E, Camus M, Armstrong B G et al. Occupational and residential exposures to 60 Hz electric and magnetic fields and high-frequency electric transients: exposure assessment using a new dosimeter. Am Ind Hyg Assoc J 1988; 49: