Tobacco Documents Online

BRIEF REPORTS Sensitivity of the Relation between Cumulative Magnetic Field Exposure and Brain Cancer Mortality to Choice of Monitoring Data Grouping Scheme Hans Kromhout,t Dana P. Loomis,ta Robert C. Kleckner,2 and David A. Savitz2 We examined the effectiveness of alternative grouping strate- gies with respect to cumulative exposure to magnetic fields and brain cancer mortality among electric utility workers. We applied a statistically optimal job-exposure matrix to calculate cumulative exposure over full work histories. We studied the sensitivity of the exposure-disease relation by assigning an array of different quantitative exposure estimates based on six schemes for grouping exposure measurements. The quantita- tive relation between cumulative magnetic field exposure and brain cancer mortality appeared to be sensitive to the choice of grouping scheme, with the optimized grouping scheme indicating stronger relations than standard schemes. (Epidemiology 1997;8:442-445) Keywords: EMF, sensitivity, exposure-response relation, exposure assessment, brain cancer, workers. Currently, two approaches are available to develop esti- mates of individual workers" quantitative exposure.1-4 The first is comparable with approaches generally em- ployed in nutritional epidemiology and utilizes personal estimates of historical exposure. The best example of this approach in occupational epidemiology is in studies of ionizing radiation exposure, in which each worker's ex- posure is monitored continuously during the entire pe- riod of employment. In most occupational studies, how- ever, large temporal variation in exposure intensity, lack of historical data, and complicated logistics of data col- lection discourage application of the individual ap- proach. More common is a group-based approach, in which monitoring data are used to assign exposure scores to workers who share the same environment, for exam- ple, department, job, function, or occupation. Until recently, there was no formal method available to determine the optimal scheme for grouping workers when using the latter approach. A default grouping by job title was therefore typically applied. A simple ratio of the between-worker variance component and the sum of the between- and within-worker components has been From the tDepartment of Air Quality, Wageningen Agricultural University, Wageningen, The Netherlands; and ZDepartment of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC. Address reprint requests to: Hans Kromhout, Department of Air Quality, Wageningen Agricultural University, P.O. Box 8129, 6700 EV Wageningen, The Netherlands. Supported by Contract RP-2964-05 from the Electric Power Research Institute, Palo Alto, CA. Submitted July 8, 1996; final version accepted February 3, 1997. 1997 by Epidemiology Resources Inc. proposed as a measure of between-group contrast in average exposure.4 Estimating this ratio for different grouping schemes provides an opportunity to choose objectively the best-performing option for exposure as- sessment, using statistical criteria. The optimal grouping scheme can differ between industries, and even between agents within an industry3 Moreover, analyses to iden- tify the optimal grouping strategy can be done indepen- dently of the assignment of individual exposure scores and analysis of the exposure-response relation. Despite the theoretical advantages of using objective methods to aggregate workers in a group-based exposure, the effects of these procedures have not been empirically evaluated in studies with cumulative exposure data. Here, we examine the sensitivity of a previously ob- served exposure-disease relation to the choice of schemes for grouping exposure measurements• We reas- signed exposure scores using an array of different quan- titative exposure estimates based on different occupa- tional grouping schemes, and we present the effect on the observed cumulative exposure-response relation. Methods In an earlier paper,5 we observed a relative risk for brain cancer of 1.07 [95% confidence interval (CI) = 1.01- 1.14] per/,T-year of exposure to mggnetic fields, using a company-specific job-exposure matrix optimized by methods described above to calculate cumulative mag- netic field exposure over the work histories of 138,905 men employed at five electric power companies in the United States.6 Detailed information on the design of the retrospective cohort mortality study can be found elsewhere.5-s Only the measurement strategy used to estimate exposure to magnetic fields, the different group- 442