Skip Navigation LinksHome > May 2000 - Volume 11 - Issue 3 > Measuring Electromagnetic Fields
Epidemiology:
Commentaries

Measuring Electromagnetic Fields

Jaffa, Kent C.; Kim, Han; Aldrich, Tim E.

Free Access
Article Outline
Collapse Box

Author Information

Submitted November 29, 1999; final version accepted November 29, 1999.

Address correspondence to: Kent C. Jaffa, PacifiCorp, 1407 West North Temple Street., Suite NTO 310, Salt Lake City, UT 84140.

Exposures before diagnosis have been estimated with historical average calculated fields (HACF) that attempt to capture temporal changes in power line fields, but are subject to calculation errors. Contemporary spot measurements (CSM) are not subject to calculation errors, but introduce temporal errors. The commentary discusses the accuracy of these metrics and how inaccuracies influence the effect estimates. Feychting and Ahlbom argue that calculation errors do not have an important influence on the effect estimates in houses, but they do in apartments. They also argue that temporal errors are important and account for an elevated effect estimate. 1 This argument is based primarily on assuming that EMF is a risk factor when examining the specificity and sensitivity of calculations with respect to measurements.

Calculation errors in HACF not only include neglecting local sources, but they also include errors in the calculation of the power line fields themselves. Temporal changes in the Feychting and Ahlbom study are captured by historical power line current changes. The magnitude of both of these errors is not uniform among subjects. It varies with the power line configuration, the distance from the line, the line current, and the attributes of the neglected EMF sources. Based on the Feychting and Ahlbom methodology, we show that CSM are of similar or better accuracy than HACF, even though the measurements are made several years after diagnosis. The commentary does not contest this observation, but argues that temporal errors have an important influence on effect estimates whereas calculation errors do not.

Kaune et al report that the long-term average current increase of the power lines is only 3.8 amperes/year in the Swedish study. 2 Over 16 years (the average period between contemporary and historical times), the average change would be about 61 amperes, which is small compared with their minimum precision of 100 amperes. The year-to-year variability is reported as ±55 amperes (W.T. Kaune, personal communication, 1998). This temporal precision cannot adequately capture these small changes, and it is unknown whether the temporal instability is real or a result of the imprecision. Thus, the ability of HACF to capture temporal changes is uncertain. A positive effect estimate for HACF provides little information about the adequacy of this metric to capture temporal changes.

The application of the specificity/sensitivity argument in their commentary does not fully address our concerns about the importance of calculation errors. Besides the above metric issues, it is unknown which attribute of EMF exposure (time-weighted average, peak, transient, etc.) is related to disease, and it is unknown which subjects are truly exposed. This uncertainty makes it difficult to use this argument. Nevertheless, applying the commentary test to temporal changes shows a similar insignificance for temporal errors as calculation errors. Contemporary average calculated fields (CACF) capture HACF with similar specificities and high-to-low prevalences compared with those for calculations and measurements presented in the commentary. Consequently, this test does not explain the differences in effect estimates for temporal changes or calculation errors, and it appears that this test cannot be applied to this data set in a simplified manner.

As discussed in our paper, calculation errors are important, complex, and do influence the effect estimates. The misclassification from calculation errors acts contrary to the dilution effect. The data clearly show there is better agreement between measurements and calculations in houses, but this does not translate into better agreement between the effect estimates for these two metrics. Consequently, the higher risk estimate for HACF in houses compared with all observations cannot be attributed to less misclassification from calculation errors. Similarly, the best agreement between effect estimates in apartments cannot be explained by more dilution. These observations show that calculation errors do influence the effect estimates by biasing them away from the null. Our paper presents a sound argument based on the available data. CSM are as good as or better than HACF in the Feychting and Ahlbom study.

To judge the validity of a metric by the effect estimate is incorrect, because it is a circular argument that assumes there is an effect. It is imprudent to draw conclusions from EMF effect estimates that are small, based on few exposed cases, and are contradicted by years of research that have failed to show any biologic plausibility for the association. 3 Other EMF studies relying on metrics that are subject to calculation errors, may be subject to similar concerns depending upon the relative importance of calculation and temporal errors in each study. We hope that our paper will increase the attention paid to EMF exposure metrics and we thank Feychting and Ahlbom for their commentary.

Back to Top | Article Outline

References

1. Feychting M, Ahlbom A. With Regard to the Relative Merits of Contemporary Measurements and Historical Calculated Fields in the Swedish Childhood Cancer Study. Epidemiology 2000; 11:357–358.

2. Kaune WT, Feychting M, Ahlbom A, Ulrich RM, Savitz DA. Temporal characteristics of transmission-line loadings in the Swedish childhood cancer study. Biolelectromagnetics 1998; 19:354–365.

3. Olden KO, Health Effects from Exposure to Power-Line Frequency Electric and Magnetic Fields, National Institute of Environmental Health Sciences, NIH 99–4493, May 1999.

Cited By:

This article has been cited 4 time(s).

Epidemiology
A Pooled Analysis of Magnetic Fields, Wire Codes, and Childhood Leukemia
Greenland, S; Sheppard, AR; Kaune, WT; Poole, C; Kelsh, MA; for the Childhood Leukemia-EMF Study Group,
Epidemiology, 11(6): 624-634.

PDF (156)
Epidemiology
Pooled Analysis of Magnetic Fields, Wire Codes, and Childhood Leukemia
Jaffa, KC
Epidemiology, 12(4): 472-473.

Epidemiology
Is There any Evidence for Differential Misclassification or for Bias Away from the Null in the Swedish Childhood Cancer Study?
Mezei, G; Kheifets, L
Epidemiology, 12(6): 750.

Epidemiology
Is There any Evidence for Differential Misclassification or for Bias Away from the Null in the Swedish Childhood Cancer Study?
Jaffa, KC
Epidemiology, 12(6): 750-752.

Back to Top | Article Outline

© 2000 Lippincott Williams & Wilkins, Inc.

Twitter  Facebook

Login

Article Tools

Share