Savitz D A, Ohya T, Loomis D P, Senior R S, Bracken T D, Howard R L
Department of Epidemiology, University of North Carolina School of Public Health, Chapel Hill 27599.
Bioelectromagnetics. 1994;15(3):193-204. doi: 10.1002/bem.2250150304.
Power-frequency electric and magnetic fields are known to exhibit marked temporal variation, yet in the absence of clear biological indications, the most appropriate summary indices for use in epidemiologic studies are unknown. In order to assess the statistical patterns among candidate indices, data on 4383 worker-days for magnetic fields and 2082 worker-days for electric fields collected for the Electric and Magnetic Field Project for Electric Utilities using the EMDEX meter [Bracken (1990): Palo Alto, CA: Electric Power Research Institute] were analyzed. We examined correlations at the individual and job title group levels among indices of exposure to both electric and magnetic fields, including the arithmetic mean, geometric mean, median, 20th and 90th percentiles, time above lower cutoffs of 20 V/m and 0.2 microT, and time above higher cutoffs of 100 V/m and 2.0 microT. For both electric and magnetic fields, the arithmetic mean was highly correlated with the 90th percentile; moderately correlated with the geometric mean, median, and lower and higher cutoff scores; and weakly correlated with the 20th percentile. Electric and magnetic field indices were generally weakly correlated with one another. Rank-order correlation coefficients were consistently greater than product-moment correlation coefficients. Job title group summary scores showed higher correlations among electric field indices and magnetic field indices and between electric and magnetic field indices than was found for individual worker-days, with only the 20th percentile clearly independent of the others. These results suggest that individuals' exposures are adequately characterized by a measure of central tendency for electric and magnetic fields, such as the arithmetic or geometric mean, and an indicator of a lower threshold or cutoff for each field type, such as the 20th percentile or proportion of time above 20 V/m or 0.2 microT. A single measure of central tendency for each type of field appears to be adequate when exposures are assessed at the job title level.
已知工频电场和磁场呈现出明显的时间变化，但在缺乏明确生物学指征的情况下，流行病学研究中使用的最合适的汇总指标尚不清楚。为了评估候选指标之间的统计模式，对使用EMDEX测量仪[布雷肯（1990年）：加利福尼亚州帕洛阿尔托：电力研究所在内的电力公司电场和磁场项目收集的4383个工作日的磁场数据和2082个工作日的电场数据进行了分析。我们研究了电场和磁场暴露指标在个体和职称组层面的相关性，这些指标包括算术平均值、几何平均值、中位数、第20和第90百分位数、高于20 V/m和0.2微特斯拉下限阈值的时间，以及高于100 V/m和2.0微特斯拉上限阈值的时间。对于电场和磁场，算术平均值与第90百分位数高度相关；与几何平均值、中位数以及上下限阈值得分中度相关；与第20百分位数弱相关。电场和磁场指标之间通常弱相关。等级相关系数始终大于积矩相关系数。职称组汇总得分显示，电场指标和磁场指标之间以及电场和磁场指标之间的相关性高于个体工作日的相关性，只有第20百分位数明显与其他指标无关。这些结果表明，通过电场和磁场的集中趋势度量（如算术平均值或几何平均值）以及每种场类型的较低阈值或截止值指标（如第20百分位数或高于20 V/m或0.2微特斯拉的时间比例），可以充分表征个体的暴露情况。当在职称层面评估暴露时，每种场类型的单一集中趋势度量似乎就足够了。
