icnirp sci review · icnirp sci review review of the ... ubiquitous part of modern life, and there...

INTERNATIONAL COMMISSION ON NON‐IONIZING RADIATION PROTECTION

ICNIRP SCI PUBLICATION – 2001

ICNIRP SCI REVIEW REVIEW OF THE EPIDEMIOLOGIC LITERATURE ON EMF AND HEALTH PUBLISHED IN: ENVIRON HEALTH PERSPECT 109(6):911‐933; 2001

ICNIRP SCI: 2001 SCI was composed of Anders Ahlbom, Elisabeth Cardis, Adele C Green, Martha Linet, David A Savitz and Anthony J Swerdlow

Environmental Health Perspectives • VOLUME 109 | SUPPLEMENT 6 | December 2001 911

Man has evolved in an environment withextremely low exposure to time-varyingextremely low-frequency electromagneticfields (EMF) from natural sources, resultingfrom the activity of the sun, fields from theearth, and fields emitted by the human body.The advent of residential and industrial use ofelectricity for power, heating, and lighting,however, has brought about far greater andincreasing exposures over the last 120 years,from the generation, transmission, and use ofelectricity (1,2). These exposures are now aubiquitous part of modern life, and there hasbeen concern in some quarters that theymight have adverse health effects.

On initial consideration, it is not obvi-ous that EMF would pose any hazard tohuman health. In particular, this radiationhas insufficient energy to damage DNAdirectly, and therefore in principle shouldnot be capable of initiating cancers. Concern

about a possible danger has arisen in the last20 years, however, and has initially beenbrought to prominence by a report in 1979of an epidemiologic study in Denver byWertheimer and Leeper (3). They found arelation between risk of childhood leukemiaand a proxy measure of degree of exposure toEMF radiation from electricity transmissionlines. Since that study, there has been a bur-geoning of research in this area. The mostintensive epidemiologic effort has concernedchildhood malignancy, especially leukemia,but there has also been considerable researchon possible occupational associations withcancer in adults, on cardiovascular and neu-rological/psychological diseases in adults, andon reproductive outcomes. This research hasbeen accompanied by public apprehensionabout the possibility that exposures to EMF,particularly for children, might be a cause ofmalignancy.

Laboratory research has given no consistentevidence that EMF of the magnitude encoun-tered in every day life for a substantial periodcan affect biological processes or that EMFaffects the risk of cancer in animals. The epi-demiologic literature is therefore particularlyworth careful consideration because it is essen-tially on this evidence alone, at present, thatsuggestions about long-term effects on humanhealth rest. In this review, therefore, we sum-marize and discuss critically the current state ofepidemiologic knowledge and the strengthsand weaknesses of the available evidence onthe relation of EMF exposure in man to risk ofcancer and other adverse outcomes. We havetaken EMF to refer to time-varying electricand/or magnetic fields <300 Hz. Wherestudies have specifically measured electricand/or magnetic fields, we have indicated thetype of field; where they have not, or where itis not clear from the report, we have referred toEMF generically. We have restricted our atten-tion to epidemiology, not experimental humanstudies; and although we have referred to someresearch on physiological effects, these are notreviewed systematically, and the review is pri-marily concerned with pathological end points.Particular attention is paid to methodologicalissues and to exposure measures because thesehave been a contentious and difficult area ofEMF research and are critical to appraisal ofthe existing literature. Finally, we comment onareas where further research is needed.

Exposure Assessment

Common Themes and Difficulties

The challenges in exposure assessment in EMFepidemiology have been discussed ever sincethe first paper was published by Wertheimer

Exposures to extremely low-frequency electric and magnetic fields (EMF) emanating from thegeneration, transmission, and use of electricity are a ubiquitous part of modern life. Concern aboutpotential adverse health effects was initially brought to prominence by an epidemiologic report twodecades ago from Denver on childhood cancer. We reviewed the now voluminous epidemiologicliterature on EMF and risks of chronic disease and conclude the following: a) The quality ofepidemiologic studies on this topic has improved over time and several of the recent studies onchildhood leukemia and on cancer associated with occupational exposure are close to the limit ofwhat can realistically be achieved in terms of size of study and methodological rigor. b) Exposureassessment is a particular difficulty of EMF epidemiology, in several respects: i) The exposure isimperceptible, ubiquitous, has multiple sources, and can vary greatly over time and short distances.ii) The exposure period of relevance is before the date at which measurements can realistically beobtained and of unknown duration and induction period. iii) The appropriate exposure metric is notknown and there are no biological data from which to impute it. c) In the absence of experimentalevidence and given the methodological uncertainties in the epidemiologic literature, there is nochronic disease for which an etiological relation to EMF can be regarded as established. d ) Therehas been a large body of high quality data for childhood cancer, and also for adult leukemia and braintumor in relation to occupational exposure. Among all the outcomes evaluated in epidemiologicstudies of EMF, childhood leukemia in relation to postnatal exposures above 0.4 µT is the one forwhich there is most evidence of an association. The relative risk has been estimated at 2.0 (95%confidence limit: 1.27–3.13) in a large pooled analysis. This is unlikely to be due to chance but, maybe, in part, due to bias. This is difficult to interpret in the absence of a known mechanism orreproducible experimental support. In the large pooled analysis only 0.8% of all children wereexposed above 0.4 µT. Further studies need to be designed to test specific hypotheses such asaspects of selection bias or exposure. On the basis of epidemiologic findings, evidence shows anassociation of amyotrophic lateral sclerosis with occupational EMF exposure although confoundingis a potential explanation. Breast cancer, cardiovascular disease, and suicide and depression remainunresolved. Key words: cancer, chronic disease, epidemiology, extremely low-frequency EMF,review. — Environ Health Perspect 109(suppl 6):911–933 (2001).http://ehpnet1.niehs.nih.gov/docs/2001/suppl-6/911-933ahlbom/abstract.html

Address correspondence to A. Ahlbom, Institute ofEnvironmental Medicine, Karolinska Institutet, Box210, 171 77 Stockholm, Sweden. Telephone: + 46 8728 74 70. Fax: + 46 8 31 39 61. E-mail: [email protected]

We thank L. Kheifets and M. Feychting for review-ing the manuscript, offering comments, and otherhelp. We also thank M. Bittar for invaluable secretarialassistance. Their help was of greatest importance forthe successful completion of this work. We thank theInternational Commission for Non-Ionizing radiationProtection (ICNIRP) for supporting this work.

Received 1 March 2001; accepted 4 June 2001.

Review of the Epidemiologic Literature on EMF and Health

ICNIRP (International Commission for Non-Ionizing Radiation Protection) Standing Committee on Epidemiology:Anders Ahlbom,1 Elisabeth Cardis,2 Adele Green,3 Martha Linet,4 David Savitz,5 and Anthony Swerdlow6

1Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden; 2International Agency for Research on Cancer, Lyon, France;3Epidemiology and Population Health Unit, The Queensland Institute of Medical Research, Brisbane, Australia; 4Division of CancerEpidemiology and Genetics, National Cancer Institute, Bethesda, Maryland, USA; 5Department of Epidemiology, School of Public Health,University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; 6Section of Epidemiology, Institute of Cancer Research, Sutton,Surrey, United Kingdom

Ahlbom

912 VOLUME 109 | SUPPLEMENT 6 | December 2001 • Environmental Health Perspectives

and Leeper (3). A criticism was that thewire-coding scheme Wertheimer and Leeperhad used to classify the subjects’ exposurewould be much too crude to result in a mean-ingful categorization. All subsequent studieshave, to some extent, been criticized for using aless than perfect exposure assessment, althoughthe sources of these problems have been differ-ent across studies depending on their design.

With few exceptions the resulting expo-sure misclassification would be nondifferen-tial and thus be most likely but not certain toresult in a bias towards the null. In effectthese problems in exposure assessment wouldnot result in spurious associations betweenEMF and disease risk; if anything, they wouldmask real associations or lead to underestima-tion of their magnitude. Yet, if a study is pos-itive despite a low correlation between amarker for EMF exposure and the true expo-sure, one could argue that the likelihood ofalternative explanations, such as confounding,would be high. These were in essence thepoints made in relation to the wire codes usedby Wertheimer and Leeper.

Consideration of the extent to which aparticular study was successful in its attemptsto assess EMF is essential when reviewing theliterature. If it turns out that the validity ofthe EMF assessment correlates with the mag-nitude of the observed effect, it would be akey observation in that review.

Three major difficulties with respect toexposure assessment are repeatedly discussedin the EMF literature, namely, the lack ofknowledge about a relevant metric and aboutthe relevant induction period; the retrospec-tive nature of the exposure assessments; andthe incomplete characterization of exposuresources, and the inability to combine expo-sures from different sources into one metric.

Knowledge on relevant metric and rele-vant period of exposures. The exposure iscomplex and multifaceted because of thecyclical nature of exposures from power linesaccording to daily, seasonal, and secular pat-terns; the variation in exposure in a given res-idence from differences in power usage bypersons residing in that home over the courseof a day, a season, and over longer intervals;and the notable variation due to exposuresfrom a wide range of different types of electri-cal appliances, with usage also varying overshort- and longer term intervals. Thus, anyeffort, no matter how comprehensive, to cap-ture retrospectively the variation by time ofday, season, and longer periods will undoubt-edly fall far short in capturing the complexityand multifaceted nature of the exposure.Regardless of the numbers, types, andrepeated nature of any measurements ofEMF, there will be incomplete characteriza-tion of exposures from all sources if the goalis to integrate exposure over long periods.

Because there is no known biologicalmechanism by which EMF can increase therisk of cancer or other diseases, the relevantexposure metric is unknown. Indeed, if sucha metric were known, it would imply thatimportant aspects of the mechanism wereunderstood and that a health effect exists.Similarly, the induction period of any poten-tial etiology is unknown, and therefore so isthe period of exposure that should be exam-ined as relevant to risk. The only knowninteraction between EMF and the humanbody is the induction of an electric current,which is proportional to the magnetic field(flux density). The magnetic field in its turnis proportional, among other things, to theelectric current by which it is generated. Themagnetic field is not easily shielded by vegeta-tion or buildings. For these reasons the mag-netic field rather than the electric field hasbeen studied in most of EMF epidemiology.

A major issue has been how to handle thetime variations in the magnetic field. In manystudy designs, the time-weighted average wasused implicitly. This holds for all the studiesbased on a characterization of homes or jobs,such as wire-code and job-title studies. It hasbeen argued that because the levels of currentsencountered in the environment are orders ofmagnitude below the levels for which biologi-cal effects are seen, one should be looking atrapid changes in the fields or at shortmoments of highly elevated fields. Sufficientlyrapid changes, called “transients,” may indeedinduce currents of a sufficient magnitude forbiological effects to occur, although presum-ably not for a sufficiently long time for thecells to react; there are currently no epidemio-logic data on this (4). Several of the studiesthat used sophisticated magnetic field meters,such as the EMDEX, have been able to lookat various patterns of time changes, inaddition to time-weighted averages (5).

Retrospective exposure assessments. All epi-demiologic studies to date have been based ona retrospective assessment of the exposure; it isunlikely that prospective studies will ever bedone, given the rarity of the outcomes of inter-est. In some studies the retrospective exposureassessment is explicit, such as when historicalfields are calculated or when wire codes or jobtitles are determined for the etiologically rele-vant period. But studies that use actual mea-surements of the fields are also retrospectivebecause it is often inferred that those fieldswould also apply retrospectively. Therefore, ithas been a topic of discussion whether carefullyassessed contemporaneous fields or morecrudely assessed historical fields offer the bestestimate of exposure during the relevant timeperiod. To address this issue, several studieshave examined the amount of change in themagnetic field from one period to another andto what extent a contemporaneous field can be

used to predict the field at some historicalpoint in time (discussed below).

Completeness of exposure characteriza-tion. The first epidemiologic study on EMFand chronic disease risk was based on a char-acterization of the homes of children withrespect to potential magnetic field levels gen-erated by nearby power lines (3). Obviously,this approach neglects magnetic field expo-sure encountered outside of the home andmagnetic field exposure in the home fromsources other than the power line. Similarly,the first study on occupational exposure, pub-lished a few years later, was based on job titlesclassified without the benefit of measure-ments and ignored all exposure outside ofwork (6). By taking measurements it is, inprinciple, possible to incorporate all in-homefields regardless of their source. A few studieshave also combined exposure at work andexposure at home (7). Two studies haveattempted to capture the complete exposureregardless of where it is experienced, byputting portable meters on children incase–control studies (8,9). However, thisassumes that the behavior of the case childrenhas not changed from that in the etiologicallyrelevant period prior to diagnosis. Anotherattempt to capture the complete exposurewould be to ask questions about use of appli-ances and other EMF sources, as was done inthe U.S. National Cancer Institute (NCI)study (10). However, the questionnairefocused on selected appliances used by preg-nant women, and on their offspring; and thusthe results underascertained mothers’ andsubjects’ exposures to magnetic fields fromelectric appliances. Furthermore, it is difficultto combine such answers into a single indexthat reflects the complete EMF exposure.

Residential We describe here five types of measurementsused in the majority of published epidemio-logic studies of residential EMF exposure andfocus on some of the difficulties associatedwith assessment of residential EMF expo-sures. Relatively few methodological studieshave evaluated reproducibility of exposuremeasurements within residences, by data col-lector, and over time. In the absence of aclear “gold standard,” only limited considera-tion has been given to the validity of theexposure assessment approaches undertakento date. A further methodological issue com-plicating residential EMF (and other formsof residential) exposure assessment in chil-dren and adults is the problem of residentialmobility. In the absence of data identifyingthe relevant timing for potentially carcino-genic or other exposures that may be etiolog-ically related to occurrence of chronic diseaseoutcomes, it is virtually impossible topinpoint the timing of exposure that is to be

EMF and health

retrospectively assessed; this issue is alsodiscussed in more detail below.

Types of exposure measurement. WIRE

CODES. Wire codes, a proxy measure of thepotential for exposure to residential magneticfields produced by electric current flow innearby power lines, is a method for estimat-ing magnetic field levels from visual inspec-tion of the characteristic features (size ofwires, closeness to the origin of electric cur-rent, etc.) and distance of power lines adja-cent to residences. The first wire-codingclassification, developed by Wertheimer andLeeper (3), categorized homes as havingeither high (HCC) or low-current configura-tion (LCC). Wertheimer and Leeper (11)subsequently expanded the wire-codingscheme to include four categories: very highcurrent configuration (VHCC), ordinaryhigh current configuration (OHCC), ordi-nary low current configuration (OLCC), andvery low current configuration (VLCC).Savitz and colleagues (12) later added a cate-gory for homes with adjacent power linesburied underground (UG). If two or morepower lines are adjacent to a residence, theWertheimer–Leeper classification assigns awire-code category to a residence on the basisof the shortest distance between a residenceand the nearest transmission line, three-phaseprimary distribution line, first-span secondarydistribution line, short first-span secondarydistribution line, or second span secondarydistribution line. The three-phase primarydistribution lines are further classified asthick or thin according to the diameter oftheir conductors. Average measurements ofresidential magnetic fields have been shownto rise with increasing category of wire codein Seattle, Washington (13), Denver,Colorado (14), Los Angeles, California (15),nine Midwestern and mid-Atlantic states(16), and five Canadian provinces (9).Kheifets et al. (17) examined the distributionof wire-code categories according to spotmagnetic field measurements using data fromseven studies and found that the percent ofhomes included within the VHCC categoryvaried from 3 to 12%, with the highest per-centages observed for studies in Los Angeles.The distribution of spot-measured magneticfields within each wire-code category wasevaluated for four of the studies, with allshowing a monotonic trend for increasingmedian field with increasing wire code inOLCC, OHCC, and VHCC categories, butthe 10–90 percentile ranges in each categoryoverlapped widely (17). Data from the1,000-home study (18) and the nine-stateNCI study (16) demonstrated a similar rangein magnetic field levels within each wire-codecategory as was observed in Denver (12) butincluded markedly higher values than thoseseen in Los Angeles (15,19).

Practically, it can be difficult to visuallydistinguish between different types of sec-ondary distribution lines or to estimate theconductor diameter, thus potentially leadingto error. To minimize possible misclassifica-tion from such errors, a simplified wire-codescheme was developed by Kaune and Savitz(20). The modified wire-code classificationincludes three categories: high wire codes(HWC), medium wire codes (MWC), andlow wire codes (LWC). The Kaune–Savitzclassification was tested on data from the sec-ond case–control study in Denver (12) andthe NCI study of nine Midwestern and mid-Atlantic states (21) and yielded similar butmore precise risk estimates of the relationbetween residential wire-code level and child-hood cancer (16,22). Data from the NCIstudy revealed that the difference in magneticfield measurements between extreme wire-codecategories was greater for the Wertheimer–Leeper classification than for the Kaune–Savitz scheme, although cross-classification ofresidential magnetic field measurements byboth wire-coding schemes demonstrated thatthe Kaune–Savitz modified code providedadditional discrimination. In addition, theKaune–Savitz code resulted in almost twice asmany homes being assigned to the highestcategory compared with the Wertheimer–Leeper code, without an appreciable decreasein measured magnetic fields in homes in thehighest category (16).

DISTANCE BETWEEN POWER LINES AND

RESIDENCES. While early residential studies ofEMF in the United States used the wire-codeclassification developed by Wertheimer andLeeper (3,11), some of the initial Europeaninvestigations examined risk of cancer in rela-tion to distance of subjects’ residences fromelectric generating or transmission equip-ment, including high-voltage power lines,overhead power lines, substations, transform-ers, electric railroads, or subways (23–25).Subsequently, studies in the Nordic countriesevaluated risk according to distance betweenresidences and power lines (26,27) orbetween residences and overhead lines,underground cables or substations (28). Inthe NCI study, risk of childhood leukemiawas evaluated according to distance of resi-dences from transmission and three-phase,primary distribution power lines along withseparate evaluation of other components ofwire codes (29).

CALCULATED HISTORICAL MAGNETIC FIELD

LEVELS. The availability of longstanding popu-lation registry databases (including computer-ized real estate data, population registryinformation, national cancer registry data andmortality registry data) in conjunction withassignment of a unique personal registrationnumber to each individual at or close to birthhas enabled unique types of population-based,

linked registry cohort or nested case–controlstudies of cancer to be carried out within theNordic countries (30,31). Detailed historicalinformation from power companies on elec-tric structures (including detailed maps andspecifications of overhead high-voltage powerlines, underground cables, towers, electricsubstations), distances (between towers,phases, etc.), the ordering of phases, and loadon the power lines could be linked with pop-ulation registry data to estimate residentialmagnetic field levels generated by power lines,using special computer programs. Variationsof this type of approach, termed calculatedhistorical magnetic fields, were used to esti-mate residential magnetic field levels in popu-lation-based epidemiologic studies carried outin Sweden (32), Denmark (28), Finland (27),and Norway (26). In effect, utilization of cal-culated historical magnetic field levels wascloser in strategy to the exposure assessmentapproach used to assign wire codes to homesthan to methods employing contemporane-ous direct measurement of magnetic field lev-els in homes to estimate retrospectively pastresidential exposures.

RESIDENTIAL AREA MEASUREMENTS. In theabsence of population registry data (whichincludes detailed information on the distancebetween transmission or distribution linesand residences) and historical informationfrom power companies on structural andrelated characteristics of power lines and loaddata, the unique types of linked registrystudies that are possible in Nordic countriesare not feasible in most other countries.Elsewhere, direct contemporaneous magnetic(and sometimes electric field) measurementshave been the most common approach usedto estimate historical residential magnetic fieldlevels. Initial studies characterized field levelsusing short-term, or “spot,” measurementstaken immediately outside (23) or within resi-dences (12), the latter obtained in the child’sand parents’ bedrooms. Subsequently, 24-hrmeasurements were obtained in rooms inwhich subjects spent a substantial propor-tion of time, based on interview data(9,15,19,21,33–37). Such measurements aremade after diagnosis, but unlike measure-ments based only on power lines, these in-home measurements reflect all sources ofmagnetic fields in the residence (38). Studiesexamining the relationship of children’s per-sonal magnetic field exposures with residen-tial and school area measurements havedemonstrated good correlation, particularlybetween 24-hr personal dosimetry and the24-hr bedroom measurements of youngerchildren in nine Midwestern and mid-Atlantic states in the United States (39,40). Astudy comparing personal and residential areameasurements of children in the UnitedKingdom also demonstrated that a 90-min


Ahlbom


measurement within the child’s home couldclassify children into the lowest 90% ofexposure with acceptable sensitivity andspecificity (37,41).

PERSONAL MAGNETIC FIELD MEASURE-MENTS. Two Canadian case–control studiesutilized personal exposure measurement(8,9). In each of these studies, children worean EMF meter in a small backpack or waistpouch (the dosimeters were placed in closeproximity to infants) for 48 hr; dosimetermeasurements were evaluated in relation toinformation obtained from an activity diary(listing times and locations of the subject’sactivities) that parents were asked to com-plete. The rationale put forth by the investi-gators for using personal measurements wasto ascertain children’s EMF exposure from allsources, including residential, school, andother away-from-home exposures (9) and toprovide more detailed information aboutcharacteristics of individual spatial and tem-poral variation in exposure (8). Experience islimited when this exposure assessmentapproach is used. The criticism is that anycase–control differences observed might sim-ply reflect changed activity patterns of casesfollowing a diagnosis of leukemia. One of thetwo groups of Canadian investigators evalu-ated this issue and found that cases spentmore time at home and less time at schoolthan controls at the time of the personal mea-surement, but these differences accounted foronly about 3% of the total time (9). Theseinvestigators also used alternative measures tomodel historical exposures (including 24-hrchildren’s bedroom measurements, wire cod-ing using the Wertheimer–Leeper and theKaune–Savitz wire-coding schemes, andperimeter measurements of childrens’ resi-dences) and compared the risk estimates forleukemia associated with the modeled histori-cal exposure estimates with the risk estimatesfor leukemia associated with those using per-sonal dosimetry (9). (See below.)

Metrics evaluated. In most epidemiologicstudies reported to date, residential magneticfield measurement data have been evaluatedusing spot measurements or time-weightedaverage levels or medians of longer-termmeasures (both of the latter representingmeasures of central tendency). Thresholdlevels, generally considered as exposures≥0.2, 0.3, or 0.4 µT have been used (0.1 µT= 1 mG). Yet, other alternative metrics havebeen proposed (42–44), including othermeasures of central tendency (such as 30th,40th, 60th, or 70th percentiles), peak expo-sures (defined as the highest measured values(e.g., 90th, 95th, or 100th percentiles), andmeasures of short-term variability, includingthe rate-of-change metric proposed byWilson et al. (45), the modified rate-of-change metric proposed by Burch et al. (46),

the number of consecutive values taken 30sec apart that differ by a minimum absolutevalues of 0.03, 0.05, or 0.10 µT (47), andother measures of rapid change, such as tran-sients (5,48). To date, only the data from theNCI study have been evaluated in anexploratory analysis of alternative metrics.The available measurements taken in theNCI study did not permit transients to beexamined, but overall the measures thatshowed the strongest association with risk ofleukemia were those of central tendency(49); the results of the exploratory analysisdid not change the fundamental conclusionfrom the earlier report of the results of thenine-state U.S. study (21). A case–controlstudy in Germany first described strongerassociations of leukemia risk with night-timemeasurements (33,34); this finding was con-firmed in the NCI study (49). Althoughsome investigators have suggested the possi-bility of windows in the dose–response rela-tion, for example, intervals of field strengththat exclusively increase risk (50), data fromthe NCI study revealed no evidence fordeparture from linearity for any of themagnetic field strength indices (49).

Time period(s) evaluated. Some data sug-gest that one potentially important period ofexposure for childhood cancer is during theprenatal period (51,52). However, etiologi-cally relevant time windows for most cancersor other chronic diseases in adults are poorlyunderstood. Even though other time periods(such as the preconception period, or perhapsan interval in early infancy) may also be etio-logically important, there are only very limiteddata implicating any agents in these time peri-ods in the etiology of childhood cancer orother childhood chronic diseases. The initialstudy assessing the relationship of EMF withchildhood cancer estimated EMF in resi-dences in which cases and controls resided atbirth and death (3). Subsequently, investiga-tors focused on homes in which cases resided:within a short interval prior to or at diagnosis(12,37,53–55); at birth (25); during preg-nancy (21,55); at birth and/or diagnosis (23);closest to diagnosis or resided in longest (15);a varying length of time dependent upon thechild’s age at diagnosis (8,9); continuouslyduring the 4-year interval prior to diagnosis(56); in the 5 years prior to diagnosis, regard-less of number of homes (21); from concep-tion to diagnosis (19,28); or during aparticular year or period prior to diagnosisthat a subject resided in a county with a highpower line (26,27,32). For most residentialstudies of EMF in adults, the period evaluatedgenerally included a specified interval (rangingfrom 4 to 15 years) prior to or at diagnosis(11,57) or during all the time a subject residedwithin a defined distance of a designated high-tension power line prior to diagnosis (58,59).

Retrospective exposure assessment limita-tions. Many assumptions must be consideredin evaluating epidemiologic studies using ret-rospective exposure assessment. Theseassumptions have been described in manyepidemiology texts and in previous reviews ofthe epidemiologic studies of EMF (43,44). Inaddition to other shortcomings described inmore detail in other parts of this section onresidential studies, one of the key issues is theextent to which contemporaneous area mea-surements (which include a comprehensiveset of carefully performed measurements)provide an accurate estimate of past expo-sures. The literature on this topic is limited inthe scope of the measurements, the numberof residences evaluated, a relatively shortinterval between initial and subsequent mea-surements or other aspects (41,60,61). Theresults are discussed below.

Incomplete characterization of sources.With the exception of the two Canadianstudies using personal dosimetry, none of thechildhood or adult residential measurementstudies attempted to include comprehensiveassessment of all sources of exposure to indi-viduals. The studies focusing exclusively onwire codes limited consideration of potentialsources of exposure to residentially proximatepower lines. Similarly, measurements focus-ing on the distance between a subject’s resi-dence and nearby power lines also restrictedevaluation of EMF to nearby power lines.Those studies incorporating area measure-ments taken within residences would partiallycapture not only EMF exposures from nearbypower lines to the specific site where the mea-surement was taken, but also the contributionof EMF exposures from nearby electric appli-ances. Yet, such area measurements were usu-ally restricted to a limited number of placeswithin a residence, thus capturing a limitednumber of sources of exposure within the res-idence. In addition, area measurements takeninside a home were often restricted to homesresided in at the time of measurement.Generally, most studies evaluated one resi-dence per subject; sometimes studies focusedonly on residentially stable subjects or resi-dentially stable controls in a case–controlstudy. Subjects with substantial residentialmobility were incompletely evaluated orsometimes excluded from studies focusing onarea measurements. In general, historical cal-culated field measurements include not only aone-time estimate of an individual’s exposurefrom nearby high power lines, but also alonger term temporal component of an indi-vidual’s exposure. However, historical calcu-lated fields do not include the contribution ofEMF exposures from electric appliances orother sources. On the other hand, residen-tially mobile as well as residentially stablesubjects are included in studies using this type

EMF and health

of measurement if the residentially mobilesubjects move to different homes within thecorridor based on distance from specifiedpower lines that define the target population.

Reliability and reproducibility of EMFexposure measurements. Given the problem-atic nature of retrospective exposure assess-ment and absence of knowledge about therelevant metric and biologically meaningfultime period of exposure, a gold standard tocompare with the extensive number of expo-sure assessment approaches used is not avail-able. Although results of analyticalepidemiologic studies are sometimes com-pared with large cross-sectional studies (18),the latter also include measurements obtainedat a single point in time and employ differentselection factors than those used in U.S. ana-lytical epidemiologic studies. In addition, theU.S. power frequency characteristics as well asthe transmission and distribution lines differfrom those in many other countries.However, several studies described compar-isons between two independent types of mea-surement. For example, the Swedish study ofchildhood cancer compared contemporane-ous spot measurements within homes to thehistorical calculated fields for those homes(32), the five-province Canadian study com-pared a construct of area in-home measure-ments plus assigned wire-code levels topersonal dosimetry (9), and several studiesevaluated the distribution of one type of mea-surement stratified by a second type(9,12,15,19,32,34,56) or evaluated the corre-lation of different metrics for magnetic fields(MF) (12,16,49).

Few studies have examined reproducibil-ity of assignment of wire codes to residences.In a study of 81 homes in Colorado, only 8were assigned wire codes in 1990 that differedfrom the wire-code category determined in1985 (60), and there was 92% agreement inwire-code assignments of 187 residences thatwere independently wire coded twice in theNCI study (16). For both studies, coding dif-ferences were due to differing distance mea-surements, differing characterization ofprimary distribution line-conductor sizes as“thick” or “thin,” and differing classificationof secondary wires as “first-span” versus“second-span.”

The coefficients of correlation betweenresidential area magnetic field spot measure-ments of 81 Colorado homes, despite differ-ences in the time of day of the twomeasurements taken in the same home,ranged from 0.70 to 0.90, thus indicatinggood to very good correlation even thoughthe two sets of measurements were taken 6years apart (60). Repeat long-term measure-ments (e.g., 24 hr for all but one measure-ment, the latter taken over a 2-week period)taken every 2 months over a year in 51 homes

in Detroit, Michigan, and Minneapolis–St.Paul, Minnesota showed good correlation ofrepeated measurements within a given resi-dence over time, although a small but statisti-cally significant seasonal effect was found(61). Nevertheless, considerable unexplainedvariability characterized measurements inabout one third of the homes. The resultssupport the need for at least one 24-hr mea-surement, but the likely improvement inexposure classification and decrease in mis-classification that would result from suchadditional measurements must be balancedby the added intrusiveness and cost (61).Good correlation (correlation coefficient =0.76) was seen for measurements taken lessthan 1 year apart in 607 residences in anationwide study in the United Kingdom,whereas the correlation coefficient was 0.66for the 182 repeated residential measurementstaken 2 or more years apart (41). Only oneinvestigation has reported the reproducibilityof exposure measurement among data collec-tors assessing wire-code configurations, andthe results showed good reproducibility (16).The results of the Swedish study, demon-strated a good correlation between contempo-raneously calculated fields and spotmeasurements but a weaker correlationbetween historically calculated fields and spotmeasurements (32).

The validity of the residential area mea-surements (and school area measurementstaken in the study in the United Kingdom) tocapture childrens’ personal magnetic fieldexposures was evaluated in two substudiescarried out for 24-hr each among 29 volun-teers (20) and 64 control children (40) in thenine Midwestern and mid-Atlantic states inthe United States, and during three separateweeks among 100 healthy children in theUnited Kingdom (41). Children under 9years of age in the United States spent40–44% of a typical 24-hr school day in theirbedroom; at home, personal dosimetry levelswere highly correlated with total 24-hr mag-netic field exposure levels and with 24-hr areameasurements taken in their bedrooms(39,40). Detailed results from the UnitedKingdom validation study will be publishedin the near future, but overall good correla-tion was seen between mean annual personalexposure and both the 90-min and 24-hr resi-dential area measurements (37).

Occupational ExposureExposure assessment in studies of occupa-tional EMF exposure and health outcomeshas been a central concern since the earliestreports on neurobehavioral changes in high-voltage substation workers (62) and leukemiain electrical workers (6). Although one canreadily determine an individual’s job title, oreven the environment in which the worker

spends time, determining the actual exposureto various forms of EMF is a major challenge.Before discussing the strategies used in paststudies, the conceptual challenges to charac-terizing occupational EMF accurately shouldbe noted.

Exposure metrics and period of relevance.As noted above, the specific exposure metricof interest is not known with certainty. Theoccupational environment has even moreextreme variability than the residential envi-ronment, both temporally and spatially. Inaddition, exposure to electric fields, whilemostly shielded in residential environments,might be important in occupational environ-ments. Consider as an example the magni-tude of exposure incurred by electric powercompany linemen in line work (often over100 µT) compared with the exposure whilein transit to the next work location (oftenclose to zero). In the occupational environ-ment, the selection of an index is likely tomatter, and correlation across indices will notnecessarily be high enough for alternatives toyield similar results (63).

Exposure assessment methods. Given therarity of most of the diseases of interest, suchas leukemia and brain cancer, it is impossibleto measure directly the exposures of all theindividuals of interest over the relevant etio-logic period. For studying rare outcomes suchas cancer, exposure of the thousands or myr-iad workers of interest is estimated on thebasis of either a generic assignment of expo-sure or detailed assessment of a relativelysmall number of workers with extrapolationto the larger group of interest.

JOB TITLES. The earliest research concern-ing potential occupational health effects ofEMF blurred the distinction between“exposed to EMF” and simply “working in anelectrical occupation.” The modern era ofresearch on occupational EMF exposurebegan with Milham (6), who compiled a listof jobs that were presumed, without empiri-cal evidence, to incur elevated exposures toelectric and/or magnetic fields, as they werethought to involve frequent or prolongedwork in proximity to energized electric equip-ment. This list served as the basis for a multi-tude of epidemiologic studies that followed.

The notable advantage of reliance on jobtitles as the basis for assigning EMF exposureis the widespread accessibility of such infor-mation. Occupation at the level of a job titleis readily available both in public records andin epidemiologic studies not focused onEMF. People can report their occupationdirectly, and even proxy respondents can do arespectable job reporting for their parent orspouse, as long as the expectation is at thelevel of a job title and does not requiredetailed information on work environment orwork practices (64); cancer registration (65)


Ahlbom


allows the study of very large occupationalcohorts. To address such rare diseases asleukemia and brain cancer, large populationsare essential. Case–control studies that gatheroccupational histories can be evaluated forinformation on associations with work inelectrical occupations (66).

Another important strength of examiningjob titles is the simplicity and ease of under-standing how the exposure index was con-structed. Evaluation of years of employmentin a particular job is much more readilyunderstood (and scrutinized) than complexindices integrating grouped jobs andimputed exposures, resulting in indices withsuch units as “microtesla-years.” Job titles areuniquely transparent and direct in describingwhat was evaluated—everything that extendsbeyond the job title is an inference that issusceptible to error.

As an exposure marker, there is also a sub-stantial disadvantage to job titles. The rela-tion between the job title and actualworkplace EMF exposure is not very strongor predictable (67). Some jobs that seem toinvolve EMF exposure may in fact not typi-cally produce elevated exposure, and eventhose that do are tremendously heterogeneousacross individuals and time (68,69). Specialchallenges arise in community-based studies,namely, those not limited to a specific com-pany or industry, with attempts to assignexposure to very broad occupational groups(70) As the job title becomes more specific inits implications for work setting and activities(71), the value of job title as a marker ofexposure is enhanced but still very poor withonly 5% of variance explained (67).

Because job titles constitute nominal or atbest ordinal indicators of exposure, there is nodirect way to combine exposures over timewithout additional quantification andassumptions. Another major challenge inusing job titles alone as a marker of EMFexposure is that the job simultaneously servesas a marker of many other exposures. Jobsconstitute a package of exposures, and theycannot necessarily be isolated from oneanother unless there is an array of jobs withassociated differing exposures. Even beyondcorrelated workplace exposures to chemicalsor physical agents that might confound theassociation between EMF and disease, jobsare not chosen randomly, and socioeconomic,behavior, and other correlates of occupationcould be pertinent to disease risk.

JOB EXPOSURE MATRICES. As the applica-tion of job titles to assignment of EMF expo-sure becomes more formal and sophisticated,it crosses the boundary into the realm ofjob–exposure matrices. A job–exposurematrix is most easily conceptualized as a tablewith jobs constituting the rows and assign-ment of exposure indices in the columns. In a

sense, even an algorithm as simple as statingthat workers in certain sectors (e.g., electricutility, electronic equipment repair) areexposed to EMF and others are not is alreadya crude job–exposure matrix, with 0’s for theunexposed jobs and 1’s for those thought tohave exposure. The rows of that matrix corre-sponding to the level of detail in the jobs canbe subdivided into increasingly specificadministrative units and work locations.Similarly, the assignment of exposure scorescan extend well beyond the dichotomy ofexposed versus unexposed. There are a num-ber of incentives to formalize the use of jobtitles in the form of such matrices.

The job–exposure matrix is a means ofcharacterizing exposure for the many personsof interest whose occupational exposure can-not possibly be measured or even scrutinizedin detail to assess potential exposure. Usingjobs as the unit for aggregation, some but notall individuals holding that job can be evalu-ated through expert assessment or measure-ment and a score assigned to all those whohold or previously held the job.

The assignment of exposure can be basedon informal assessment of the work locationand activities. The next level of evaluationinvolves expert assessment through observa-tion or background knowledge of the relevantindustries. An expert panel, for example,might evaluate a list of jobs and determinewhether there is likely to be elevated work-place EMF exposure associated with each.The most sophisticated approach requires acombination of expert evaluation andmeasurement for a sample of workers.

A number of studies have developed quan-titative exposure matrices using this approach(72–74). The strategy starts with the selectionof reasonably homogeneous job groups forassignment, sampling workers in those groupsfor direct measurement of workplace EMFexposure, using statistical approaches to assignexposure to the job group, and finally applyingthat information to all individuals in the study.The opportunity to develop a detailed, empiri-cally driven job–exposure matrix is muchgreater within an industry than across manyindustries, in part because of the reduced diver-sity in types of jobs to evaluate but also becauseof ease of workplace access for measurement.

Providing quantitative exposure estimatesfor the jobs of interest offers the opportunityto quantify the variation in exposure withinand between job groups (67,72). Moreover,when assigning exposures to time intervalsthat include multiple jobs, only quantitativeindices can be integrated to produce a sum-mary score. Quantification also allows formore direct comparisons across work settings(67) and helps to relate the literature onoccupational EMF to studies of residentialexposures and electric appliances.

The decision about the proper unit foranalysis, namely, the rows of the matrix, iscritical. Some argue, for example, that it isnecessary to consider the specific power plantin making such assignments in the electricutility industry (75), not just the job title.The trade-off between the homogeneity ofnarrowly constituted groups and the limitednumber of measurements per group must bereconciled as well. Just as in the case of resi-dential measurements, the incorporation ofthe many quirks of the specific person’s activ-ities on the particular day of measurementcontributes to the exposure assignment. If thelineman’s truck breaks down and he spendsthe day by the side of the road, that is part ofwhat determines his exposure for the day. Inprinciple, those events are part of the linemanexperience, and with a large enough sample,those events should be part of what makes thesampled exposure representative of linemen.

Incompleteness of characterization ofsources. Even at best, occupational EMFexposure characterization will be incomplete,given the failure to incorporate exposureencountered in the residence and through useof electric appliances. Also within the workenvironment, some incidental exposures suchas those encountered by driving near over-head power lines or having an office locatednear electric conductors are nearly impossibleto capture. Instead, occupational exposureassessment focuses on specific, observablesources of exposure that are distinctive to thejob of interest.

Reliability and reproducibility. Althoughmeasurement of workplace exposure has beenexamined rather extensively to address day-to-day variability, the overall approach to assign-ing exposure in occupational studies has notbeen generally evaluated (76). That is,whether another set of investigators assignedthe task of characterizing exposure would endup with the same scheme is open to question.For the simplest of job–exposure matrices, forexample, dividing workers into operationsversus office, the reliability would likely bequite good, whereas for the more detaileddecisions on the job groups and the numberand methods of measurement, reliabilitywould likely be much lower.

Cancer

Childhood Cancer

Magnetic field exposures from power lines.OVERVIEW. Since Wertheimer and Leeper in1979 (3) hypothesized that magnetic fieldsfrom residentially proximate high-tensionpower lines and electric power substationswere associated with increased risks of child-hood cancer, more than 18 additional epi-demiologic studies in at least nine countries(Table 1) have used a spectrum of exposure

EMF and health


Table 1. Characteristics of studies and results on the relation between EMF exposure and childhood cancer.

Magnetic field measure-Primary exposure Study Cancers (numbers Wire codes RR ments RR (95% CI)

Reference Study population metric(s) design cases/controls) (95% CI) (high category) (high category)

Wertheimer and Denver residents born in Colorado. Wire code of diagnosis/ CC* All cancers (328/328) 2.25 (HCC) —Leeper, 1979 (3) Cases: <19 yr, deaths (1950–1973). death home Leukemia (155/155) 2.98 (1.78–4.98) (HCC) —

Controls: birth certificates Brain tumors (66/66) 2.40 (1.03–5.41) (HCC) —

Fulton et al., 1980 Rhode Island residents. Wire code. Cases: all CC Leukemia (119/240) 1.00 (HCC) —(53) Cases: <20 yr. Controls: birth lifetime homes.

certificates Controls: birth homes

Tomenius, 1986 Stockholm County, Sweden Front door measurement CC All cancers (1,033/890) — 1.8 (≥0.3 µT)(23) residents. Cases: <19 yr (1958– birth and diagnosis Leukemia (243/212) — 0.3 (≥0.3 µT)

1973). Controls: birth certificates residences Brain tumors (294/253) — 3.7 (≥0.3 µT)

Savitz et al., 1988 Denver residents. Cases: <15 yr Wire-code spot MF CC WC MF(12) (1976–1983). Controls: random measurements child’s All cancers 320 128 2.20 (0.98–5.21) (VHCC) 1.35 (0.63–2.90) (≥0.25 µT)

digit dialing bedroom, low power Leukemia 97 36 2.75 (0.94–8.04) (VHCC) 1.93 (0.67–5.56) (≥0.25 µT)Brain tumors 59 25 1.94 (0.47–7.95) (VHCC) 1.04 (0.22–4.82) (≥0.25 µT)Controls 259 207

Myers et al., 1990 Yorkshire, England residents. Distance of home to CC All cancers (374/588) 1.10 (0.47–2.57) (<25 m —(25) Cases: <15 yr (1970–1979). nearest overhead line; distance)

Controls: birth register estimated MF strength 0.4 (0.04–4.33) (≥0.1 µT)

London et al., 1991 Los Angeles County residents. Wire-code and 24-hr CC WC MF(15) Case: <10 yr (1980–1987). child’s bedroom MF meas- Leukemia 211 162 2.15 (1.08–4.26) (VHCC) 1.22 (0.52–2.82) (≥0.125 µT)

Controls: friends and random urement in home lived in Controls 205 143digit dialing longest, low power

Feychting and Sweden residents within 300 m Historically calculated Nested All cancers (141) 1.3 (0.6–2.7) (≥0.3 µT) —Ahlbom, 1993 (32) of 220 or 400 kV power line. fields CC Leukemia (38) 3.8 (1.4–9.3) (≥0.3 µT)

Cases: <15 yr (1960–1985). Brain tumors (33) 1.0 (0.2–3.9) (≥0.3 µT)Controls: selected at random Controls (554)from cohort to match cases

Olsen et al., 1993 Denmark residents. Cases: Historically calculated CC All cancers (1,707/4,788) 5.6 (1.6–19) (≥0.4 µT) —(28) <15 yr (1960–1986). Controls: fields Leukemia (833/1,666) 6.0 (0.8–44) (≥0.4 µT)

Central Population Registry Brain tumors (624/1,872) 6.0 (0.8–44) (≥0.4 µT)

Verkasalo et al., Finland residents within 500 m Historically calculated Cohort All cancers (140) 1.5 (0.74–2.7) (≥0.2 µT) —1993 (27) of 110–400 kV power line. fields Leukemia (35) 1.6 (0.32–4.5) (≥0.2 µT)

Cases: <17 yr (1974–1990) Brain tumors (39) 2.3 (0.75–5.4) (≥0.2 µT)

Preston-Martin Los Angeles County residents. Wire code at diagnosis, CC WC MFet al., 1996 (19) Cases: <20 yr (1984–1991). first, and longest Brain tumors 281 106 1.2 (0.6–2.2) (VHCC) 1.7 (0.6–5.0) (≥0.3 µT)

Controls: random digit dialing residence Controls 250 99

Gurney et al., 1996 Seattle and surrounding western Wire code of diagnosis CC Brain tumors (120/240) 0.5 (0.2–1.6) (VHCC) —(94) Washington State residents. home

Cases: <20 yr (1984–1990).Controls: random digit dialing

Tynes and Norway residents in census Historically calculated Nested All cancers (532/2,112) 0.9 (0.5–1.8) (≥0.14 µT) —Haldorsen, 1997 (26) ward with high-voltage power fields CC Leukemia (139/546) 0.3 (0.0–2.1) (≥0.14 µT) —

lines. Cases: <15 yr (1965–1989). Brain tumors (144/599) 0.7 (0.2–2.1) (≥0.14 µT) —Controls: selected at randomfrom cohort to match cases

Linet et al., 1997 U.S., residents of 9 mid-Atlantic Wire-code residences CC WC MF(21) and Midwestern States. >70% 5 yr before Acute lym- 402 624 0.88 (0.48–1.63) (VHCC) 1.24 (0.86–1.79) (≥0.3 µT)

Cases: <15 yr (1989–1993). diagnosis; TWA MF phoblastic Controls: random digit dialing measurements all leukemia

residences combined Controls 402 615 1.72 (1.03–2.86) (≥0.3 µT)>70% 5 yr before diagnosis

Michaelis et al., Northwest Germany (Lower 24-hr child’s bedroom CC Leukemia (176/414) — 2.3 (0.8–6.7) (≥0.2 µT)1997 (33) Saxony) and Berlin residents. MF measurement

Cases: <15 yr (1991–1995).Controls: government office residents’ registry

Dockerty et al., New Zealand residents. 24-hr child’s bedroom CC Leukemia (115/117) — 15.5 (0.3–7.6) (≥0.2 µT)1998 (35) Cases: <15 yr (1990–1993). MF measurement

Controls: birth certificate

(Continued)

Ahlbom


assessment methods to evaluate the relation-ship. Over time, the epidemiologic studieshave also generally enrolled larger numbers ofsubjects; focused increasingly on childhoodleukemia and, to a lesser extent, brain andnervous system tumors; addressed method-ological shortcomings of earlier investiga-tions; and increasingly collected data on abroad range of other suspected confoundingfactors. Descriptions of the epidemiologicstudies of EMF and childhood cancer can befound in the original reports. A brief sum-mary is presented in Table 1. The reader isalso referred to comprehensive reviews andsummaries of the literature by expert commit-tees appointed by the National RadiologicalProtection Board in the United Kingdom(77,78), the National Research Council ofthe U.S. National Academy of Sciences (43),and the National Institute of theEnvironmental Health Sciences (part of theU.S. National Institutes of Health) (44). Inthis section of the review we provide a histori-cal synthesis of the epidemiologic studies ofchildhood cancer risk in relation to magneticfield exposures from power lines and fromelectric appliances. Among the emphases arethe evolution of the childhood cancer out-comes evaluated, the growing sophisticationof the exposure assessment strategies used,and the increasing understanding of themethodological issues.

TOTAL CHILDHOOD CANCER. Wertheimerand Leeper (3) reported significantly elevatedrisks for total childhood cancer (relative risk[RR] = 2.25) in Denver due to excess risks forchildhood leukemia (RR = 2.98), brain and

nervous system tumors (RR = 2.40), andlymphomas among nonoverlapping cases andcontrols (RR = 2.08). Comparing subjectsresiding in homes with high current configu-rations to those living in homes with low cur-rent configuration, a subsequent study inDenver found similar, albeit slightly lower,risks for all cancers combined (12). Althoughrisks for all childhood cancers combined werealso evaluated in one U.K. (25), and fiveNordic studies (23,26–28,32) (Table 1), thebiological and etiological interpretation ofresults for a grouping of disparate childhoodmalignancies is unclear.

LYMPHOMAS. Subsequent to the twostudies in Denver, which reported elevatedrisks of lymphoma on the basis of 18 casesresiding in HCC homes (3) and 3 cases inVHCC homes (12), results of later investi-gations have not supported a link betweenchildren’s estimated residential magneticfield exposures and childhood lymphomas(except for a 5-fold, nonsignificantly ele-vated risk reported by Olsen et al. (28) onthe basis of a single case). The studies reveallittle evidence of a relationship betweenchildhood lymphoma and MF exposurefrom residentially proximate power lines,but the data include very small numbers ofhighly exposed cases (Table 1).

BRAIN AND NERVOUS SYSTEM TUMORS.Significantly increased relative risks of braintumors were reported in the first (RR = 2.4)(3) and second (RR = 1.9) (12) Denverstudies among children residing in homescharacterized by HCC and VHCC, respec-tively. However, later studies in the United

States (19,55) generally have not found excessrisks of brain and nervous system tumorsassociated with high residential wire-codeconfigurations. Direct spot measurements inDenver (12) and 24–48-hr residential mag-netic field measurements were also not linkedwith increased risks in Los Angeles (19) or theUnited Kingdom (37). Calculated magneticfield levels were not linked with increasedrisks of childhood brain and nervous systemtumors in Sweden (32) or Norway (26),whereas nonsignificantly increased risks inDenmark (28) and a smaller risk in Finland(27) were based on two and three cases ofbrain and nervous system tumors, respec-tively. The absence of a relationship betweenresidential EMF exposures and childhoodbrain tumors in the large and methodologi-cally rigorous Los Angeles study (19) and inthe nationwide U.K. (37), Swedish (32),Danish (28), and Norwegian (26) studiesfocusing on childhood brain tumors do notshow that childhood brain tumors are etio-logically linked with exposure to residentialsources of EMF (Table 1).

LEUKEMIA. Studies of EMF have increas-ingly focused on childhood leukemia.Increasingly sophisticated exposure assess-ment approaches have been used in morerecent studies.

Wire-code classification. The evolution ofthe wire-code configuration classificationscheme, originally created by Wertheimer andLeeper (3) and further refined by Wertheimerand Leeper (11), Savitz et al. (12), and Kauneand Savitz (20), is described above. All studiesexamining the relationship of wire-code con-

Table 1. Continued.

Magnetic field measure-Primary exposure Study Cancers (numbers Wire codes RR ments RR (95% CI)

Reference Study population metric(s) design cases/controls) (95% CI) (high category) (high category)

McBride et al., Canada, residents of 5 Wire code of home 2 yr CC Leukemia 0.77 (0.37–1.60) (VHCC)1999 (9) provinces. Cases: <15 yr (1990– before diagnosis Wire code (303/309)

1994). Controls: province 48-hr personal 48-hr personal 1.04 (0.69–1.57) (≥0.2 µT)health insurance rolls measurement dosimetry monitoring

(293/339)24-hr child’s bedroom 24-hr child’s 1.27 (0.69–2.33) (≥0.2 µT)

2 yr before diagnosis bedroom (272/304)Green et al., 1999 Southern Ontario Canada Wire code spot MF CC Leukemia(8) residents. Cases: <15 yr (1985– measurements; 48-hr Wire code (79/125) 1.5 (0.3–8.7) 1.13 (0.31–4.06) (≥0.4 µT)

1993). Controls: telephone personal monitoring Spot meas- (OHCC + VHCC)marketing lists urements (21/46)

48-hr personal 4.5 (1.3–15.9) (≥0.14 µT)monitoring (88/133)

UKCCS, 1999 England, Wales, Scotland In-home MF measure- CC All cancers (2,265/2,270) — 0.89 (0.34–2.32) (≥0.4 µT)(37) residents. Cases: <15 yr (1992– ments. Phase I: 90-min Leukemia (1,094/1,096) — 1.68 (0.40–7.10) (≥0.4 µT)

1995). Controls: Family Health measurement in family Brain tumors (390/393) — 0 cases/2 controls (≥0.4 µT)Services Authorities register room and spot measure-

ments in child’s bedroom. Phase II (highest 10%): 48-hr measurement in child’s bedroom. School: spot measurements

Abbreviations: CC, case–control; TWA, time-weighted average; UKCCS, United Kingdom Childhood Cancer Study; WC, wire code; yr, years.

EMF and health

figuration and risk of childhood leukemiaemployed the case–control design. The rela-tion between wire-code configuration andmeasured magnetic field levels may be influ-enced by in-home electric wiring, grounding,electric appliances, and other nearby sourcesof EMF (12). Wire-code levels predict mea-sured magnetic fields in all areas of theUnited States, although the correlation is notvery strong (see above). The significantly ele-vated risks estimated for childhood leukemiain relation to high wire-code configurationsin Denver (3,12) and Los Angeles (15), werenot replicated in Rhode Island (53), in ninemid-Atlantic and Midwestern states (21) orin five provinces in Canada (9) (Table 1).

Distance between power lines and resi-dences. Several investigations evaluated therelation between distance of residences frompower lines or other sources of high magneticfields and risk of childhood leukemia(24,26,29,32). One study used a measure ofdistance but reported results only as a mea-sure of voltage (of the two closest transmis-sion or distribution lines) divided by thedistance in meters, the square of the distanceor the cube of the distance (56), a type ofmeasurement not used in other studies, andthus difficult to evaluate or compare withother studies. Elevated risks (OR = 1.45, 2.0,1.3) of childhood leukemia were reported forthe small fraction (0.6%) of children residingwithin 100 m or 50 m of an overhead powerline or within 25 m of a substation, respec-tively, in southeast England (24). An excessrisk of leukemia was observed among childrenresiding 50 m or less from 220 or 400 kVpower lines in Sweden (based on 6 cases)(32). However, risk of childhood leukemiawas not increased among children residingless than 51 m from high-voltage lines inNorway (based on 9 cases) (26). Risk of acutelymphoblastic leukemia was not increasedamong children residing within 40 m oftransmission lines (based on 10 cases) orthree-phase primary distribution lines (basedon 105 cases) in nine Midwestern and mid-Atlantic states in the United States, nor wasrisk increased according to the contributionof all transmission lines and three-phase pri-mary distribution power lines near a child’sresidence (based on 108 cases) (Table 1) (29).

Calculated historical magnetic field levels.The novel exposure assessment approach usedin the Nordic countries (see above) linkeddata from various registries with long-termpower line load data and specifications forpower lines and associated structures obtainedfrom the utility industry (32). The Nordicstudies, although varying somewhat in studydesign, were all population based. Twostudies defined cohorts residing within a spec-ified distance of high-tension power lines,then ascertained childhood cancer cases

within the cohorts during specified periods (anested case–control approach) (26,32). Athird study used a similar cohort method,reporting results from a cohort analysis (27).The Danish study identified incident child-hood cancer cases during a specified periodand selected matched controls from the cen-tral population register; proximity to high-voltage facilities was assessed using maps ofhigh-tension overhead lines or undergroundcables, and residential magnetic field levelsestimated from the distance of the subject’sresidence from the line or cable, the charac-teristics of nearby power lines, and electricityload data (28). Among the leukemia caseswith estimated residential magnetic fieldexposure levels ≥0.2 µT (7, 3, 3, and 2 inSweden, Denmark, Finland, and Norway,respectively), a 3.8-fold increased risk ofleukemia was reported in Sweden (32), a 6-fold increase in Denmark (28), a 1.6-foldincrease in Finland (27), and no excess risk inNorway (Table 1) (26).

Residential measurements. In residentialstudies assessing exposure using spot and/or24-hr or longer area magnetic field measure-ments, increases in leukemia, ranging from1.3- to 1.5-fold elevated, were reported forchildren with average magnetic field exposures≥0.2 µT in Denver (based on 3 cases) (12),Los Angeles (based on 20 cases with exposures≥0.268 µT) (15), Lower Saxony and Berlin,Germany (based on 4 cases) (33,34), nineMidwestern and mid-Atlantic states in theUnited States (based on 58 cases) (21), fiveprovinces in Canada (based on 54 cases) (9),and the United Kingdom (based on 21 cases)(including England, Wales, and Scotland)(37). A 3.3-fold increase (95% confidenceinterval [CI] = 0.5–23.7) of leukemia waslinked with 24-hr children’s bedroom time-weighted average measurements ≥0.2 µT in astudy in New Zealand (based on 5 cases)(34,79), and an odds ratio of 1.1 (95% CI =0.31–4.06) was linked with point-in-timemeasurements ≥0.13 µT taken in the child’sbedroom in a study in southern Ontario,Canada (based on 21 cases) (Table 1) (36).The latest study is from Germany and showeda relative risk of 1.6 (0.7–3.7) for 0.2 µT and3.2 (1.3–7.8) for nighttime exposure (80).

Personal magnetic field measurements. TwoCanadian studies employed personal exposuremeasurements as the primary direct measureof children’s exposure to magnetic field levels.Unfortunately, it is difficult to compare resultsbetween the two Canadian studies or betweenthe southern Ontario study and those con-ducted elsewhere because results of the studyby Green et al. (8) are not reported using thesame categorical cut point of ≥0.2 µT pro-vided in most reports, despite an adequatenumber of cases (n = 20, according to Table1) with average magnetic field exposures

≥0.2 µT. McBride et al. (9) reported only asmall difference between cases and controls inactivity patterns, but the results from personaldosimetry measurements are difficult to inter-pret in the absence of more widespread use ofthis measurement approach.

Summary of results of individual studies,meta-analysis, and pooled analysis. Greatestweight should be given to results of themethodologically more rigorous studies withlarger numbers of subjects with high MFexposure levels (9,21) and to population-based studies with few methodological short-comings (26–28,32,37). Extensive effortshave been undertaken to summarize quanti-tatively the individual studies in meta-analy-ses (43,44,77,78,81–84) and pooled analyses(85,86). Pooled analysis offers the availabilityof raw data as a special advantage, but, simi-lar to meta-analysis, requires great care in themethodological approach used and interpre-tation of results (87–89). Using data fromstudies in six European countries (26–28,32–34,37), nine Midwestern and mid-Atlantic states in the United States (21), fiveprovinces in Canada (9), and New Zealand(35,79). Ahlbom et al. (85) found risk to benear the no-effect level among the 3,203 chil-dren with leukemia and 10,338 control chil-dren with summary residential MF exposurelevels <0.4 µT, whereas a 2-fold leukemia risk(RR = 2.0, 95% CI = 1.27–3.13) wasobserved among the 44 leukemia cases (ofwhom 24.2 represented the expected numberand 19.8 the excess number) and 62 controlchildren with estimated residential MF expo-sures ≥0.4 µT. Thus, fewer than 20 childrenamong 3,203 with leukemia represent theexcess over expected numbers among childrenresiding in homes with magnetic field expo-sure levels >0.4 µT. Adjustment for potentialconfounding variables did not appreciablyaffect the results.

Magnetic field exposures from electricappliances. Five studies have evaluated risksof childhood leukemia (15,35,90,91) or brainand nervous system tumors (19,35,90) associ-ated with use of electric appliances. All thestudies employed interviews of subjects’mothers to help assess exposure information.Overall, the small number of studies and theabsence of measurement data within thestudies preclude straightforward interpreta-tion of results. The results based on interviewdata are summarized briefly below.

LEUKEMIA. A few associations wereobserved in two or three studies. Two investi-gations (12,91) reported small increases inrisk associated with prenatal use of electricblankets, but only one of these (12) found adose–response effect. There was little evi-dence of elevated risk of leukemia in offspringassociated with mothers’ prenatal use of othertypes of electric appliances. Postnatal use of


Ahlbom


electric blankets (12,35,91) and hair dryers(15,91) was linked with modestly elevatedrisks in more than one study, but there wasno evidence of dose–response relationships.Risk of leukemia was increased overall, but nodose–response effect was found, among chil-dren watching black-and-white television inLos Angeles (15), whereas leukemia rose withincreasing number of hours children watchedtelevision (mostly color televisions, as fewblack and white televisions were used),regardless of the child’s distance from thetelevision set in the nine Midwestern andmid-Atlantic states (91). An MF measure-ment study of more than 70 televisions ofvolunteer families in the greater Washington,DC, area concluded that MF exposures werenot substantially greater than ambient levelsat typical distances that children sit whilewatching television or playing video games ontelevision screens (92). Risks were increasedfor postnatal exposure to a few other appli-ances in a single study (91), but overall thefindings were not consistent among the fourstudies, nor was there generally evidence ofdose–response relationships.

BRAIN TUMORS. There was little consis-tency among the results of the three studiesthat have evaluated risk of childhood braintumors associated with prenatal and postnatalexposures to electric appliances. The firststudy (90) reported a dose–response relationfor increasing number of night-time hours ofmaternal use of electric blankets and risk ofbrain tumors in offspring. This finding wasnot replicated in the other studies. However,Preston-Martin and colleagues (19) describedsmall increases in risk of brain tumors amongthe offspring of mothers who used waterbedsduring pregnancy. Dockerty et al. (35) foundno associations of childhood brain tumorswith maternal prenatal use of electric appli-ances, but noted nonsignificantly elevated riskof childhood brain tumors linked with post-natal use of electric blankets, waterbeds, andcurling irons.

Overall, only limited data are available onelectric appliances and risk of childhoodleukemia or brain tumors. There is little con-vincing evidence that EMF exposures frommaternal prenatal or children’s postnatal use ofelectric appliances is associated with increasedrisk of childhood leukemia or brain tumors.

Methodologic issues. SELECTION BIAS AND

CONFOUNDING. Important methodologicalconsiderations in the design, conduct, andinterpretation of every epidemiologic studyinclude the potential for selection biases.Although the possible role and the effect ofeach of these biases have been discussed inmost of the summaries of the relation ofEMF and childhood cancer (43,44,77), rela-tively few studies have attempted to evaluateor quantify their relative importance.

Selection bias. Nonparticipants often differfrom participants, and participation rates tendto be lower for controls than cases in case–con-trol studies. The design and methods used inthe Nordic studies do not require individualsubjects to be approached, but rely on infor-mation available in various registries. Thus,selection bias is not an issue in the Nordicstudies but is a concern in other studies. Toevaluate the possible role of selection bias,Hatch et al. (93) compared the relationbetween childhood leukemia and wire codesand direct measurements of magnetic fields inhomes of subjects who participated in allphases of the study with the relation in all sub-jects, including those who declined to allowaccess inside the home or on the property, inthe U.S. study conducted in nine Midwesternand mid-Atlantic states. The results revealedsomewhat higher odds ratios for childhoodleukemia when partial participants wereexcluded. Similar but slightly smaller increasesin the odds ratios were observed, compared tothose based on all subjects, when subjects whoallowed a measurement only outside the frontdoor were excluded. Because partial partici-pants tended to be characterized by lowersocioeconomic status than subjects who partic-ipated fully, these findings suggested selectionbias. Like almost all of the other case–controlstudies of childhood cancer and EMF, thecase–control investigation in nine Midwesternand mid-Atlantic states was characterized bygreater nonparticipation by controls than cases,and higher socioeconomic status among con-trols than cases. The investigators of the studyin five provinces in Canada (9) and the nation-wide study in the United Kingdom (37) alsonoted a somewhat higher socioeconomic statusand lower participation among controls thancases in those studies. Selection bias due tononparticipation or differential restrictionsplaced upon cases and controls may haveaffected the results. Differential residential sta-bility requirements were placed on cases versuscontrols in Denver (12), and cases were morelikely than controls to have resided in theirhome for their entire lifetime in Los Angeles(15). Subjects in the Los Angeles study whorefused to participate at either the randomdigit dialing or interview stages did not havetheir homes wire coded (15). The case–controlstudy in New Zealand also reported differen-tial levels of participation between cases andcontrols and evidence of higher socioeconomicstatus among controls than cases (35). If resi-dentially stable controls were also more likelyto reside in neighborhoods with low residentialEMF exposure or wire-code levels, a spuriousrelation may have resulted between residentialEMF and childhood cancer. In contrast, selec-tion bias (for wire codes but not for measure-ments) may have been reduced in thenine-state Midwestern and mid-Atlantic study

compared to earlier studies in the UnitedStates. This was because wire codes wereassessed for subjects who refused to participatein the second interview or to allow access tothe home or property and magnetic field mea-surements were obtained immediately outsidethe front door for all residences eligible formeasurement regardless of whether the datacollector was permitted to take measurementsinside the residences (91). Savitz et al. (12)also wire coded a higher proportion of sub-jects than the proportions included in theinterview and in-home measurement compo-nents, because eligible homes were wire codedfor subjects refusing to participate since accessto the home or property was not needed forwire coding.

Confounding. An evaluation of therelation between a large number of potentialconfounding variables and wire-code levelsand direct measurements in the nine stateMidwestern and mid-Atlantic study (22)revealed that univariate adjustment for indi-vidual variables changed the odds ratios foracute lymphoblastic leukemia by less than 8%and simultaneous adjustment reduced the riskestimates by a maximum of 15% (93).Categories of potential confounding factorsthat were evaluated but found to demonstrateno effect or only a very small effect includesocioeconomic factors (mother’s and father’seducation and occupation, family income,racial/ethnic group, home ownership), resi-dential features (urbanicity, primary source ofheat, type of air conditioning), lifestyle factors(maternal or paternal smoking, breast feed-ing, maternal use of a sewing machine, timespent watching television), residential mobil-ity, reproductive factors (mother’s or father’sage at first birth, total number of live birthsprior to the index diagnosis/reference date),and use of selected electric appliances (electricblankets, waterbeds, hair dryers, and others)(93). A comparison of the potential effects ofconfounding versus selection bias in the nine-state U.S. study suggested that confoundingalone was unlikely to be an important sourceof bias. The conclusion that selection biasmay be more of a concern than confoundingin most studies of residential magnetic fieldexposures and childhood cancer risk (93) isfurther underscored by the inconsistencyamong studies in the relation between incomeand wire codes. Studies in Seattle (94) andColumbus, Ohio (95), reported inverse asso-ciations between income and wire-code levels,but no evidence of such a relationship wasobserved in the nine Midwestern and mid-Atlantic states study (93). In evaluation ofrisks associated with the use of electric appli-ances, the relevant exposure has been assumedto be magnetic fields. Yet, other features alsocharacterize users of such electric appliances.For example, families in which children

EMF and health

spend many hours watching television arelikely to differ behaviorally and in other waysfrom families in which little television iswatched. In the U.S. National HealthExamination Survey, time spent watchingtelevision was reported to be a strong predictorof obesity during adolescence (96).

MEASUREMENT ERROR. As discussed in“Retrospective Exposure AssessmentLimitations” a single, time-weighted averagemeasurement taken after diagnosis may notrepresent typical levels or even the propermetric for the period or residential area that isrelevant. Because elevated risk appears to berestricted to only a very small fraction of chil-dren who are highly exposed and becausethere is no basis for determining the patternof measurement errors in each study, it is notpossible to assess the extent of measurementerror in a given study nor is it possible to cor-rect for such unknown errors.

In the study by Savitz et al. (12) and thestudy by Feychting and Ahlbom (32) therewas evidence of an association between trafficdensity and leukemia, but without adjust-ment for traffic density having an effect onthe EMF and cancer relation (97,98).

REPORTING BIAS. Reports about one’s ownor one’s child’s typical behavior during yearsprior to an interview are prone to error, partic-ularly because behavior patterns change rapidlywith age. The respondent’s report may reflecthabits from another year or another child inthe family. Nondifferential forms of error, forexample, those affecting cases and controlsequally, tend to reduce an apparent associationbetween exposures and a disease (99) and mayminimize true dose–response patterns. Incase–control studies of childhood cancer,errors may be more likely to be differential,thus potentially exaggerating true case–control differences. Such differential errorscan arise in several ways. When asked aboutprediagnosis behavior, mothers may actuallyreport postdiagnosis behavior. Another typeof problem that can result in differential mis-classification is recall bias, in which themother of a case may be more likely to recallminor exposures occurring several years pre-viously, whereas a mother of a healthy childis more likely to forget such exposures.Another possibility is that mothers of casesmay exaggerate the duration or frequency ofearlier exposures, whereas mothers of con-trols may report such exposures more accu-rately. Exposures that have been linkedrepeatedly with increased cancer risk by themedia may be more likely to be mentionedby mothers of cases than mothers of controls.It is possible that some of the associationsreported for various electric appliances andchildhood cancer may be due to recall bias,although attempts to evaluate this have notshown evidence of bias (91,94).

RANDOM VARIATION AND RANDOM

ERROR. When several types of measurementor a battery of questions are applied to assessa single hypothesis, as in many of the studiesof childhood cancer and EMF (includingelectric appliances), individual elementsshould not be overinterpreted. Random varia-tion or random error increases the likelihoodof a positive finding for at least one individualmeasurement or question within the group ofmeasurements or battery of questions.

Summary. Following the original reportby Wertheimer and Leeper (3) linking thethree most common forms of childhoodcancer with a proxy measure of residentialEMF (wire codes), more than 18 studies innine countries have shown no convincing evi-dence of a relationship of childhood braintumors or lymphoma with residential expo-sure to EMF from nearby power lines. Thereis no clear evidence of a relationship betweenchildhood leukemia and residential EMFexposures among children with estimatedexposure levels under 0.4 µT. A 2-foldincrease in relative risk of childhoodleukemia, confined to a very tiny fraction ofchildren (estimated as 0.8% in one largepooled analysis) with residential EMF expo-sures ≥0.4 µT, is difficult to interpret in theabsence of a known biological mechanism orreproducible experimental support of carcino-genesis. There is also some evidence to sug-gest that selection bias may account for someof the increase in risk among the proportionof children with high residential EMF expo-sure. In the absence of new and convincingexperimental evidence linking EMF with car-cinogenesis, additional epidemiologic studiesare unlikely to provide further clarification ofthe relationship unless large numbers of caseswith exposures ≥0.4 µT can be accrued, andmethodological shortcomings, particularlyselection bias, can be minimized.

Adult CancerThe literature on occupational EMF andcancer is voluminous, particularly forleukemia and brain cancer, whereas researchon residential or appliance exposure in rela-tion to those and other cancers in adults hasbeen quite limited. The recent concern withpossible effects of EMF on breast cancer,largely driven by the hypothesized effect onmelatonin (100,101), has generated limitedfindings, which we discuss, but there are sev-eral major ongoing studies in the UnitedStates that have not yet been published. Thebulk of epidemiologic evidence is onleukemia, brain cancer, and breast cancer.

Meta-analyses of the occupational EMFliterature by Kheifets and co-workers (81,100)identified 38 pertinent studies of leukemiaand 29 studies of brain cancer after truncatingthe list to those suitable for meta-analysis, and

the literature has continued to grow. Becauseothers have summarized the vast array ofstudies and because the more recent ones areso far superior to those that preceded them,the focus in this review is on the smaller num-ber of studies with sophisticated approaches toexposure assessment. Those that rely solely onjob titles will be summarized in the aggregateon the basis of previous reviews.

Leukemia. REVIEW OF OCCUPATIONAL

STUDIES. The literature that began in the early1980s consists of reports linking routinelycollected information on job titles withcancer incidence or mortality in large popula-tions. The exposure inferences were basedsolely on general knowledge of the exposuresassociated with those jobs, whether extrapo-lated from other studies or based on expertevaluation. In the aggregate (81,102), certainpatterns emerge. There is a small increasedrisk of leukemia associated with work in elec-tric occupations, with a relative risk the orderof 1.2 across the many studies (81). Withinthe range of the 38 studies evaluated byKheifets et al. (81), there was little differencein risk associated with various measures ofstudy quality, but the range available for con-sideration was limited. Furthermore, therewas no indication that jobs thought to havehigher exposure (welders, electricians, line-men, and power plant operators) had higherrisks than electric workers generally found tohave lower exposures (installers, engineers,and television or radio repairmen). Acrossleukemia subtypes, where there have beenstriking differences in individual studies, inthe aggregate, the differences are modest.Pooled relative risk estimates calculated byKheifets et al. (81) ranged from 1.2 forchronic myeloid leukemia (CML) to 1.4 forchronic lymphocytic leukemia (CLL).

One other pooling effort is noted, namelythe aggregation of the studies of electric util-ity cohort studies in the United States,Canada, and France (104). Previously pub-lished studies of roughly comparable design(73,105,106) were analyzed using commonmethods to juxtapose and ultimately pool theresults. Despite what appeared to be ratherimpressive differences in leukemia resultsacross studies, with no association found insouthern California Edison workers (105) orin an aggregation of U.S. utility workers(106), and mixed but generally positiveresults for the Canada–France study (73), theresults were broadly compatible within therange of random variation. That is, despitethe large size of these studies, random erroralone could well account for the spectrum ofresults that were obtained once a common setof statistical tools was applied. Beyond theapplication to these specific studies, thisobservation is an important reminder aboutthe challenges of interpreting ostensibly


Ahlbom


contradictory findings where the results donot differ dramatically and precision of all thestudies is limited. Including results fromOntario Hydro, the pooled relative risk esti-mate for leukemia was 1.09 per 10 µT-year(95% CI = 0.98–1.21).

The major studies of occupational electricor magnetic field exposure and leukemia thatrelied on measurement-based job-exposurematrices are summarized in Table 2. Wheredata were adequate, results for majorleukemia subtypes are presented as well, butsummaries of results were necessarily selec-tive. Several studies are readily described as

showing no indication of increased risk ofleukemia in association with occupationalmagnetic field exposure based on the pub-lished analyses (105–108). In contrast, anequal number of studies did show indicationsof increased risk with greater estimated mag-netic field exposure (71,73,75,109,110). Inmost of the supportive studies, the relativerisk estimate in the uppermost category fortotal leukemia was between 1.5 and 2.0, butfor some leukemia subtypes, the estimateswere larger and less precise. Acute lympho-cytic leukemia (AML) was more substantiallyelevated in two studies (73,110) and CLL in

the study by Floderus et al. (71). Electricfields have received less attention, with onestudy suggesting a strong association (75),one an inverse association (111), and two noassociation (107,112).

Whether we examine a large number ofstudies on the basis of job title or a smallernumber of studies using relatively advancedexposure assessment technology, the infer-ences tend to be similar. Some individualstudies show notably positive associationsbetween measures of EMF and leukemia, withdose–response gradients and reasonable preci-sion, whereas other studies broadly similar in

Table 2. Summary of the principal studies of occupational EMF exposure and leukemia and brain cancer using measurement-based job–exposure matrices.

Comments on Comments on Reference Setting, industry Leukemia results, RR (95% CI) leukemia results Brain cancer results, RR (95% CI) brain cancer results

Matanoski et al., U.S., telephone workers >Median (mean): 2.5 (0.7–8.6) Increases association Not available —1993 (109) with longer latency

Floderus et al., Sweden, general 2nd quartile: 0.9 (0.6–1.4) Weaker association for 2nd quartile: 1.0 (0.7–1.6) Slightly stronger 1993 (71) population 3rd quartile: 1.2 (0.8–1.9) median exposure; no 3rd quartile: 1.5 (1.0–2.2) gradient for median

4th quartile: 1.6 (1.1–2.4) association with AML 4th quartile: 1.4 (0.9–2.1) fields, time above CLL/2nd quartile: 1.1 (0.5–2.3) 0.2 µTCLL/3rd quartile: 2.2 (1.1–4.3)CLL/4th quartile: 3.0 (1.6–5.8)

Sahl et al., 1993 California, electric utility >Median: 1.0 (0.8–1.4) Slight association for >Median: 1.0 (0.6–1.5) —(105) >99th percentile: 1.1 (0.8–1.4) fraction >5.0 µT >99th percentile: 0.8 (0.5–1.3)

Theriault et al., Canada–France, electric >Median: 1.5 (0.9–2.6) Association primarily at > Median: 1.5 (0.9–2.8) Association 1994 (73) utility >90th percentile: 1.8 (0.8–4.0) Ontario Hydro >90th percentile: 2.0 (0.8–5.0) consistent across

CLL/> median: 1.5 (0.5–4.0) Astrocytoma/> median: 1.5 (0.9–2.8) three companiesAML/> median: 3.2 (1.2–8.3) Glioblastoma/> median: 1.3 (0.5–3.8)

Benign tumors/> median: 2.3 (0.8–6.7)

Tynes et al., 1994 Norway, railway Low: 1.0 (0.4–2.2) — Low: 0.8 (0.3–2.0) —(107) High: 0.6 (0.2–1.3) High: 0.9 (0.4–2.3)

Electric field—low: 0.4 (0.2–1.1) Electric field—low: 0.7 (0.3–1.7)Electric field—high: 1.0 (0.5–2.2) Electric field—high: 1.2 (0.5–2.8)

Savitz and Loomis, U.S., electric utility 30–<50th percentile: 1.0 (0.7–1.6) Association with work 30–<50th percentile: 1.6 (1.0–2.6) Weaker association 1995 (106) 50–<70th percentile: 1.1 (0.7–1.8) as electrician; little dif- 50–<70th percentile: 1.5 (0.8–2.6) with work in

70–<90th percentile: 1.0 (0.6–1.6) ference for AML, CLL 70–<90th percentile: 1.7 (0.9–3.0) individual electrical ≥90th percentile: 1.1 (0.6–2.1) ≥90th percentile: 2.3 (1.2–4.6) occupations

Guenel et al., 1996 France, electric utility Electric fields No confounding by Electric fields No confounding by (111) >50–75th percentile: 1.0 (0.5–2.0) magnetic fields, SES. >50–75th percentile: 2.5 (1.0–6.2) magnetic fields,

>75–90th percentile: 0.7 (0.3–1.9) Similar for AML, >75–90th percentile: 1.4 (0.5–4.5) SES>90th percentile: 0.4 (0.1–1.3) non-AML >90th percentile: 3.1 (1.1–8.7)

Miller et al., 1996 Ontario, Canada, electric Electric: >33–67th percentile: 2.1 Stronger association for Not available —(75) utility (0.6–7.2) AML, weaker for CLL.

Electric: >67th percentile: 4.5 Slightly stronger for (1.0–19.7) AML

>33–67th percentile: 1.7 (0.6–4.8)>67th percentile: 1.6 (0.5–5.1)

Feychting et al., Sweden, general 0.13–0.19 µT: 1.4 (1.0–2.2) Strong interaction with 0.13–0.19 µT: 1.0 (0.7–1.6) —1997 (110) population ≤0.20 µT: 1.7 (1.1–2.7) residential magnetic ≤0.20 µT: 1.0 (0.6–1.7)

AML/0.13–0.19 µT: 2.1 (0.9–5.0) field exposureAML/≤0.20 µT: 2.7 (0.9–7.9)CLL/0.13–0.19 µT: 1.4 (0.7–2.5)CLL/≤0.20 µT: 1.9 (1.0–3.8)

Harrington et al., England, electric utility Not available — >33–67th percentile: 1.1 (0.6–2.0) No effect with 1997 (74) >67th percentile: 1.0 (0.5–1.9) latency, adjustment

for confounders

Rodvall et al., Sweden, general Not available — Glioma/0.2–0.4 µT: 1.1 (0.4–2.7) Weaker association1998 (119) population Glioma/>0.4 µT: 1.9 (0.8–5.0) for median than

mean

Johansen and Denmark, electric utility Background: 1.0, low: 1.0, — Background: 0.5, low: 0.9, —Olsen, 1999 (108) medium: 0.9, high: 1.1 medium: 0.7, high: 0.7

SES, socioecomonic status.

EMF and health

design and quality, do not. The comparativeanalysis by Kheifets (76) points out how sus-ceptible study findings are to subtleties of sta-tistical methods and to random error.Without a formal meta-analysis, the results inTable 2 are likely to be consistent with a smallgradient of increasing risk with increasingexposure that varies largely by chance acrossstudies. Although individual studies may sug-gest that a stronger effect is found for electricfields (75), for specific subtypes of leukemia(71), or in conjunction with residential expo-sures (110), replication is required to drawconclusions about such patterns.

REVIEW OF RESIDENTIAL STUDIES. Theeffect of exposure from transmission lines hasbeen studied in four case–control studies(57,59,113,114). No information, however,was collected in those studies either on othersources of residential exposures [except bySeverson et al. (57)], or on occupationalexposures [except by Feychting et al. (110)].This may have resulted in substantial expo-sure misclassification. A small increased riskfor all leukemia was seen in only one (113) ofthe four studies, in association with calculatedmagnetic fields of more than 0.1 µT in theyear preceding diagnosis. Results of analysesof specific subtypes of leukemia are inconsis-tent across studies and difficult to interpretbecause of small numbers of exposed cases.An increased risk was seen for AML andCML but not for CLL in the Swedish study(59). The odds ratio for AML was reduced,however, and the risk of CML disappearedwhen analyses were restricted to subjects withno or very little occupational exposure,whereas the odds ratio for subjects with bothhigh occupational and residential exposuresincreased (6.3, 95% CI 1.5–27 for both AMLand CML, based on only 3 exposed cases). Inthe Finnish study, a significant increase wasseen for CLL only, for exposures over 10years before diagnosis and for durations ofexposures of 12 years or more, based on 3exposed cases (114).

The risk of leukemia from the use of elec-tric appliances was considered in twocase–control studies (57,115–117). Neitherof these studies provides information aboutsuch risk, however, because of limitations ofstudy design and exposure assessment.

CONCLUSIONS. The research on the riskof adult leukemia in relation to occupationaland residential magnetic field exposureincludes a number of large studies of varyingquality, with the most research by far address-ing occupational exposures. Some of thesestudies are excellent (7,71,73,106,110);applying sophisticated epidemiologic meth-ods to the evaluation of the role of magneticfields, though a few studies have attempted toaddress electric fields as well. Results fromthese studies have ranged from null to rather

strong positive associations, with relative risksin the upper exposure categories above 2.0.Unfortunately, there is not a clear pattern inwhich the better studies are more or less likelyto produce positive associations. In the aggre-gate, assuming random error accounts for dif-ferences among studies, the results are mostconsistent with a weak positive association,with relative risks for the more highly exposedgroups of the order of 1.1–1.3. Relative risksof this magnitude are below the level at whichepidemiologic methods can effectively assesscausal relations. Nevertheless, the evidence atpresent supporting a role for EMF in the eti-ology of adult leukemia is weak. The stan-dards for future epidemiologic studies tomake a notable difference in the totality ofevidence are extremely high. An exceptionalopportunity to study very large populationswith well-characterized, relatively high expo-sure and detailed cancer incidence data wouldbe required to provide a significant advance-ment in our knowledge on this topic.

Nervous system tumors. REVIEW OF OCCU-PATIONAL STUDIES. Completely analogous tothe literature on electric occupations andleukemia, there is a sizeable literature on elec-tric occupations and brain cancer. Interest inbrain cancer as a potential consequence ofEMF exposure began slightly later than theinterest in leukemia, with an influential paperby Lin et al. (118) linking electric occupa-tions to brain cancer using death certificatedata. At the time of the meta-analysis byKheifets et al. (102), 29 relevant reports hadbeen published, most of which assessed expo-sure on the basis of job title alone. Moststudies tended to show a small increase in riskof brain cancer among electric workers, witha pooled relative risk estimate of 1.2. Somestudies showed no association, and the riskestimates were highly imprecise in manystudies, reflecting the rarity of brain cancer.The association was stronger for studies thatpresented results restricted to gliomas (RR =1.4) and was stronger for electrical engineers(RR = 1.7) but similar across the other spe-cific occupational categories. There was notendency either for jobs thought to havehigher exposure or for studies with moresophisticated exposure assessments to showstronger associations. The pooling effortdescribed above in which results from utilityworker studies in France, Canada, and theUnited States were combined yielded an esti-mated relative risk of 1.12 per 10 µT-years(95% CI = 0.98–1.28), virtually identical tothat found for leukemia (104). Once again,what appeared to be heterogeneity acrossstudies was compatible with random variationaround a common small effect.

Ten studies that provided risk estimatesfor electric or magnetic fields using measure-ment-based job-exposure matrices and brain

cancer are summarized in Table 2. Notsurprisingly, the study findings are mixed,with suggestions of positive associations infive (71,73,106,111,119) and the remaindershowing no indication of an association. Evenamong the studies designated as positive,there were rarely monotonic dose–responsegradients and the largest relative risk estimatesrarely exceeded 2.0. No pattern could beidentified on the basis of the type of studypopulation (electric utility, general popula-tion). Too few studies presented results forhistologic subtypes of brain cancer to drawconclusions about heterogeneity of risk. Theevidence at present for supporting a role forEMF in the etiology of brain cancer is weak.Results are most compatible with a smallassociation, with some studies finding noassociation and some finding a stronger effect.There are insufficient data to identify particu-lar exposure sources or patterns or diseasesubtypes associated with larger relative risks.

REVIEW OF RESIDENTIAL STUDIES. Thestudies of residential exposures, once again,provide little additional information. Fourstudies have considered the risk of brain andCNS tumors in relation to residential expo-sures from high voltage transmission lines(58,59,113,120,121). No clear associationwas seen in any of these studies. Occupationalexposure was taken into account in one study(110) but did not affect the results. None ofthese studies collected information on othersources of residential exposure.

CONCLUSIONS. The conclusions providedfor EMF and adult leukemia are essentiallyapplicable to the brain cancer literature aswell. A large number of studies, mostlyaddressing occupational exposure, have gener-ated measures of association ranging fromnull to rather strongly positive, but in theaggregate, relative risk estimates would be inthe range of 1.1–1.3, a level at which a mean-ingful discussion of causality is not possible.

Breast cancer. REVIEW OF OCCUPATIONAL

STUDIES. An interest in breast cancer as a pos-sible consequence of electric and magneticfield exposure arose largely from a hypothe-sized mechanism proposed by Stevens and co-workers (100,101). It was hypothesized thatelectric and magnetic fields suppress the pro-duction of nighttime melatonin, analogous tolight exposure at night, and that reduction inmelatonin increases the risk of developingbreast cancer. Over the past decade a fairlysizable body of research has addressed theinfluence of EMF on melatonin production.The question of an effect of EMF on mela-tonin lends itself to both human experimentalstudies (122) and observational studies ofhumans outside the laboratory. The literaturefrom human experimental studies is generallynegative regarding an effect of nighttimeEMF on melatonin production (123–125).


Ahlbom


Several observational studies of environmentalexposures to EMF and melatonin, in con-trast, have suggested effects in humans, butthe pattern of findings is not persuasive. Inthe study of electric blanket users (126), only7 of the 28 volunteers were affected, and inthe studies of electric utility workers, an alter-ation in melatonin metabolite was found onlyin association with a rather unusual magneticfield metric (standardized rate of change) (46)or only among workers with low occupationalsunlight exposure (127). At present, the the-ory regarding a melatonin pathway gets weaksupport from the empirical data.

The initial epidemiologic reportsconcerned male breast cancer, starting withtwo letters to the editor at The Lancet(128,129) that reported increased risks associ-ated with electric occupations and electro-magnetic field exposure, respectively. A largepopulation-based case–control study in theUnited States provided much stronger sup-port for an association, with an odds ratio of6.0 (95% CI = 1.7–21) among electricians,telephone linemen, and electric power work-ers (130). Another reasonably large study wasreported and it did not support an association(OR = 0.7, 95% CI = 0.3–1.9) (131). Largestudies of electric utility workers did not findincreased risks of male breast cancer associ-ated with magnetic field exposure (73,106),though statistical power was quite limitedbecause of the rarity of male breast cancer.

Research on breast cancer among women,a much more common disease, has beeninhibited by the rarity of electric occupationsamong women. Analyses of a large databaseon occupation and mortality in the UnitedStates yielded an indication of a modestlyincreased risk of breast cancer mortalityamong female electric workers (OR = 1.4,95% CI = 1.0–1.8) (132). Reanalyses of thesame data set using slightly different methodsto classify exposure indicated an associationonly among black women, not among whitewomen (133). The limitations of relyingsolely on job title and cause of death are sub-stantial, including a complete lack of infor-mation on potentially critical confoundingvariables. The most detailed study to dateconcerning electromagnetic fields and femalebreast cancer utilized a multistate case–control study combined with a systematiceffort to classify jobs by exposure potential(134). On the basis of an analysis of the 5,223cases and 7,236 controls who had worked out-side the home, an increased risk was found forthe highest potential for occupational expo-sure to electromagnetic fields (OR = 1.4, 95%CI = 1.0–2.1). The association was somewhatstronger among premenopausal women thanamong women overall (OR = 2.0, 95% CI =1.0–3.8). Both Forssén et al. (7) andKliukiene et al. (135) find some support for

an association between EMF and breastcancer risk in women below 50 years of age;in the Forssén study this is particularly truefor estrogen-receptor–positive breast cancer.The Forssén study is particularly interestingbecause it includes information on residentialand occupational exposure combined.

REVIEW OF RESIDENTIAL STUDIES. Becauseof the rarity of occupational exposures to ELFamong women, population-based studies ofresidential exposures have the potential of pro-viding valuable information on risk of breastcancer related to ELF. The evidence fromsuch studies is limited, however. The risk ofbreast cancer in women in relation to residen-tial exposures from transmission lines has beenconsidered in three studies (113,130,136). Noassociation was seen in two studies (113,120),but in the third (136) a nonsignificantlyincreased risk was seen for exposure in the 6years preceding the diagnosis, as well as inyoung women (under 50 years of age) and inwomen whose breast cancers were estrogen-receptor positive. Among women with estro-gen-receptor–positive breast cancers and lessthan 50 years of age, the odds ratio was 7.4(1.0–178) on the basis of only 6 exposedcases. No information, however, was availableon other sources of exposure to ELF or onsome important risk factors for breast cancer(such as parity and age at first pregnancy),which could confound the association.

The effects of electric blanket use wereconsidered in one case–control study each ofpostmenopausal (137) and premenopausalbreast cancer (138). A small, nonsignificantincreased risk was seen in both pre- and post-menopausal women for continuous use ofelectric blanket throughout the night com-pared to never use. The increase reached sta-tistical significance (OR = 1.5, 95% CI =1.1–1.9) when the results of both studies werecombined, although there was no associationwith duration of use. The results of thesestudies are difficult to interpret because of verylow response rates and lack of information ontype and age of the electric blankets or onother sources of ELF exposures (139–141).

The risk of male breast cancer in relationto transmission lines was considered in onlyone study (136). Only 9 cases were includedin the study. A 2-fold, nonsignificantlyincreased risk was seen.

CONCLUSIONS. The totality of evidencelinking EMFs to breast cancer, in men orwomen, remains weak. Nevertheless, givenhow common female breast cancer is and themultitude of studies seeking information onrisk factors, further evaluation of occupa-tional EMF exposure is desirable and shouldbe feasible (142). The major limitation is inexposure prevalence and the opportunity toassess female occupational exposure morecarefully. As the findings of three major

studies of residential exposure to magneticfields and breast cancer have not yet been dis-seminated, future research plans should awaitthat information before deciding on the needfor and direction of any new initiatives.

Other cancers. Brief mention should bemade of several other cancers that have beeninvestigated in relation to occupational EMFexposure. A marked association betweenpulsed EMF exposure and lung cancer wasfound in the Canada–France electric utilityworker study (143), with a monotonicdose–response gradient culminating in anodds ratio of 6.7 (95% CI = 2.7–16.6) in thehighest exposure stratum. Unfortunately,lack of comparable data and uncertaintyabout the nature of the exposure inhibitedattempts at replication. The one effort to re-address this association was in U.S. electricutility workers and within the limitations ofextrapolating a job–exposure matrix fromone study to another, the findings were notcorroborated (144).

Limited attention has been focused onnon-Hodgkin’s lymphoma (145,146), withsome support for a possible association.Colon cancer was associated with electric fieldexposure in a French utility worker study(111), illustrating a number of sporadic eleva-tions in cancer risk found across the series ofstudies in which the design permitted exami-nation of all cancer types (73,75).

A particularly intriguing line of researchhas been the possibility of a relation betweenchildhood cancer and parental occupationalEMF exposure. However, results have beeninconsistent and unconvincing (147–149).

Other End Points

Neurodegenerative Disease

Concerns about possible psychiatric or psy-chological effects of EMF exposure wereraised by investigators from the Soviet Unionin the late 1960s and early 1970s on the basisof anecdotal reports of symptoms such asinsomnia, memory loss, and headache (150).However, these and other early reports havebasically remained unconfirmed (151).Relatively recently, however, hypotheses relat-ing EMF to neurodegenerative disorders haveattracted a new interest. For a number ofmethodological reasons, these diseases aremore difficult to study than cancer. The mostobvious difficulty is that they are notrecorded in registries in the same way as can-cers and that mortality registries are less reli-able as sources of cases. These and otherdifficulties are reflected in the literature.Unfortunately, the studies that have bestavoided these problems suffered instead fromsmall numbers. The overwhelming focus hasbeen on amyotrophic lateral sclerosis (ALS)and Alzheimer’s disease (AD) and there are

EMF and health

only some scattered data on other diagnoseswithin this group of diseases (152,153).

Amyotrophic lateral sclerosis. Sevenstudies on ALS have been published(154–160). Certain characteristics of thesestudies are displayed in Table 3. All thestudies are based on occupational exposure toEMF. Some used job title on the death cer-tificate or a census record as a proxy for expo-sure and others used job history accompaniedwith a job–exposure matrix or some otherexposure index to assess EMF exposure. Themethods for diagnosis and case ascertainmentvaried across studies. Some studies used deathcertificate information, whereas others usedcases from specialized neurological clinics.

The seven studies may be divided intothree groups according to design features(Table 3). One group consists of the threestudies that did not use mortality registries toascertain the cases but instead identified themfrom neurological clinics or, in one instance,from an ALS society. Two of the three studiesare clinically based and lack specified popula-tion bases from which the cases were gener-ated and they used friends and relatives as thesources for controls (154,157). Thus, thesetwo studies are susceptible to selection bias,the direction or magnitude of which cannotbe predicted with any certainty. Therefore,despite other assets, such as specific diagnosesand careful exposure assessment in one ofthem, the overall contribution is limited. Thethird study in this group has a clearly defined

study population from which in principle allprevalent and diagnosed cases were identifiedand the controls constituted a random samplefrom that population (156). Exposure assess-ment in this study, however, was based on aquestionnaire with rather crude questionsregarding electricity work and occupationalexposure to EMF and the results were some-what inconsistent (Table 4).

The next group consists of two studies thatare both based on death certificates for theidentification of cases and on job titles for theassessment of exposure (158,159), in one casefrom death certificates (158) and in the otherfrom a census (159) (Table 3). The strengthsof these two studies include minimization ofselection and recall bias as a consequence of thereliance upon registry information. Also, thelarge numbers of subjects, reflected by the nar-row confidence intervals, are considerableassets. The major weakness is the crude infor-mation on which exposure assessment is based.It is based only on job title at one point intime without any measurement or other datato back it up (Table 4).

The third group comprises the two lateststudies based on cohorts of utility workers,one in the United States and one in Denmark(159,160). Both studies are designed suchthat the risk of selection bias is small, becausethey each start with a well-defined cohort andbecause deaths are searched for in mortalityregistries. Both studies also have employeddetailed procedures for exposure assessment

that involved classification of jobs on thebasis of measurements. The duration of eachjob was another strength. Despite the largenominal sizes of the cohorts, however, theeffective numbers of exposed cases are mod-est. These two studies are by far those thatcarry the most weight in overall assessment.The designs of the two studies are relativelysimilar and so are the findings. The combinedresults from these two studies is a relative riskof 2.7 (1.4–5.0) (Table 4).

The combined results from the two utilityworker studies (159,160) show a clearincrease in ALS mortality. The combinedconfidence interval suggests that the riskincrease is unlikely to be due to chance.There is no obvious bias in design, such asexposure or diagnosis misclassification, thatcould explain the elevated risk. If anything,such a bias would have been expected toresult in an attenuation of the relative risk.


Table 3. Certain characteristics and findings of studies on the relation between EMF exposure and ALS.

Definition and Study Result RR Reference Study population and subject identification estimation of exposure design Numbers (95% CI)

Deapen and Study population: not specified. Cases: Questionnaire: electrical CC 518 cases (19 electrical occupation) 3.8Hendersen, 1986 ALS Society, U.S. in 1979. Controls: friends occupation 3 yr prior to diagnosis 518 controls (5 electrical occupation) (1.4–13.0)(154)

Gunnarsson Male population of Sweden 1970–1983. Job title in census 1960: electrical CC 1,067 cases (32 exposed) 1,005 controls 1.5et al., 1991 (155) Cases: deaths with ALS as underlying worker (0.9–2.6)

or contributing cause in mortality registry. Controls: random sample from population

Gunnarsson Male population of central and southern Questionnaire: electrical work and CC 58 cases (4 MF exposure) 0.6 (MF exp)et al., 1992 (156) Sweden in 1990. Cases: patients with MND exposure to MF 189 controls (0.2–2.0)

in neurologic departments. Controls: random sample from population

Davanipour Study base: not specified. Cases: ALS Questionnaire about occupational CC 28 cases 32 controls cutoff: 75th 2.3et al., 1997 (157) patients at outpatient clinic in southern history: EMF exposure assessed by percentile, of case distribution (0.8–6.6)

California. Controls: relatives hygienist. Cumulative (E1) and average (E2)average (E2) exposure

Savitz et al., Male population in 25 U.S. states, Job title on death certificate: elec- CC 114 cases in electrical occupation in 1.31998 (158) 1985–1991. Cases: deaths from ALS. trical occupation in aggregate and aggregate (1.1–1.6)

Controls: deaths from other causes individual jobsSavitz et al., Male employees at 5 U.S. utility companies, Measurements and employment Cohort 9 cases with >20 years in exposed 2.41998 (159) 1950–1988. Cases: deaths with ALS noted records. Combination of duration occupations (0.8–6.7)

on death certificate, identified through and EMF indexmultiple tracking sources

Johansen and Male employees in Danish utility companies, Employment records and job– Cohort 21,236 males in cohort. 9 exposed 2.5Olsen, 1998 (160) observed 1974–1993. Cases: deaths from exposure matrix: estimated cases (1.1–4.8)

ALS in mortality registry average exposure level

MND, motor neurone disease.

Table 4. Pooling across groups of studies on EMF expo-sure and ALS.

Number of Pooled studies studies RR 95% CI

All 7 1.5 1.2–1.7Clinically and ALS 3 3.3 1.7–6.7society-based studiesMortality registry and 2 1.3 1.1–1.6census-based studiesUtility cohort studies 2 2.7 1.4–5.0

Ahlbom


Thus, the two utility worker studiescombined provide relatively strong evidencethat work with EMF exposure in the utilityindustry is indeed related to increased ALSmortality (Table 4). This result is reinforcedby the results of the other studies on ALS dis-cussed above, even though the five otherstudies have to be given less weight.

Alzheimer’s disease. Five studies on ADwere found (158,159,161–163) (Table 5).

The first two studies shown in Table 5were clinic-based, case–control studies. Thefirst combined three series of AD patients,one from the United States and two fromFinland (161). These series came from neuro-logical centers that specialized in diagnosisand treatment of AD and can therefore beassumed to be based on high-quality diag-noses. For one series of AD patients vasculardementia patients were used as controls; forthe second series, controls were other patientswithout neurological disease, and for thethird, neighborhood controls. The secondstudy comes partly from the same group ofinvestigators and was an attempt to confirmthe findings from their first publication(162). It was also based on patients from aspecialized clinic in the United States andused another group of patients as controls.Both studies based exposure classification onjobs as reported by the patient or a relative.The major weakness is the lack of a specifiedstudy population and thus the potential forselection bias.

Of the three remaining AD studies, onewas based on the Swedish Twin Registry. Theinvestigators evaluated twins included in theregistry, which was set up for the purpose ofconducting genetic studies of dementia intwins (163). Exposure to EMF was assessed

through interviews that included job history.Diagnostic quality in the study was good, aswas the detail in which EMF exposure wasassessed. Another strength was the definedpopulation base for the study. The main prob-lem with this study was its small size, asreflected by the relatively wide confidenceintervals. It also had a contradiction in itsfindings depending on whether primary or lastoccupation was used as the basis for analysis.

The last two studies were discussed in theALS section above, because they provide dataon both diseases (Table 5). These are thedeath certificate study and the utility workerstudy, both in the United States (158,159).As discussed in the ALS section, these areboth reliable studies, but the death certificatestudy used a crude measure for the EMFexposure assessment. The utility worker studyis less suited for AD because of the limitedusefulness of death certificate as a source ofdisease classifications. However, the investiga-tors report results both for underlying causesof mortality and for contributing causes, andthere is a difference between those results.When contributing causes are used, there islittle support for an association between EMFand AD, while the use of underlying causegives some support for such an association.Because of the nature of this disease, it seemsmore logical to look at contributing causes.

Interpretation. Even if the studies onALS consistently suggest an increased risk inEMF-exposed subjects, one would like con-firmatory results from additional studies, instudies specifically designed for the purpose.Assuming that the observed risk elevation isaccurate, it still remains to be explained.Aside from the hypothesis that EMF expo-sure increases ALS risk, one must consider

alternative explanations. One such alternativewould be confounding from electric shockexposure. It is conceivable that exposure toelectric shocks increases ALS risk and, also,that work in the utility industry carries a riskof experiencing electric shocks. Some of thereviewed studies did report analyses thatindeed linked electric shocks to ALS(154,156,160), but none of the studies pro-vided an analysis in which the relationbetween EMF and ALS was studied with con-trol for electric shocks. A crude calculationcan be made from data provided by Deapenand Hendersen (154), and this seems to indi-cate that the EMF association holds up evenafter control for electric shock experience.

As for AD, when evaluated across all thestudies, there appears to be an associationbetween estimated EMF exposure and diseaserisk (Table 6). However, this result is mainlyconfined to the first two studies in the UnitedStates, and it is not clearly confirmed by thelater studies (153,154,158,161,162). The twostudies that show excess (161,162) may havebeen affected by selection bias. Because thestudy populations are undefined, there is noway to determine the extent to which thecontrols are representative with respect toexposure of the population from which thecases originated.

Conclusion. For reasons discussed in thepreceding sections, the ALS results areintriguing and point toward a possible riskincrease in subjects with EMF exposure.However, confirmatory studies are needed, asis an appropriate consideration of confound-ing, for example, from electric shocks, as aconceivable explanation. As for AD, itappears that the excess risk is constrained tostudies with weaker designs; thus support for

Table 5. Certain characteristics and findings of studies on the relation between EMF exposure and Alzheimer’s disease.

Definition and Study Resulting RR Reference Study population and subject identification estimation of exposure design Numbers (95% CI)

Sobel et al., 1995 Study population: not specified. Cases: three sets Interview data on primary CC 386 cases (36 exposed) 3.0(161) of AD patients examined, 77–93 years of age, at one occupation. Classification into 475 controls (16 exposed) (1.6–5.4)

neurologic clinic in the U.S. and two in Finland. Controls: high/medium vs low EMFthree sets—vascular dementia patients, patients exposurewithout neurologic disease, and neighborhood controls.

Sobel et al., 1996 (168) Study population not specified. Cases: patients with Statewide data form information CC 326 cases 3.9probable or definite AD treated at AD medical center on primary occupation. Classification 152 controls (1.5–10.6)in California, USA. Controls: patients who were into high/medium vs low EMF cognitively impaired or demented exposure

Feychting et al., 1998 Study population: subsample of the Swedish Twin Interviews. Primary and last CC 55 cases 0.9 (primary)(163) Registry. Cases: identified through a screening and occupation. Classification into 228 and 238 controls (0.3–2.8)

evaluation procedure. Controls: intact twins with one three levels, based on JEM, (similar with twin in each of two control groups when two twins highest >0.2 µT other control were eligible group)

Savitz et al., 1998 (158) Male population in 25 U.S. states, 1985–1991. Job title on death certificate: CC 256 cases in electrical 1.2Cases: deaths from AD. Controls: deaths from other electrical occupation in aggregate occupations, in aggregate (1.0–1.4)causes and individual jobs

Savitz et al., 1998 (159) Male employees at five U.S. utility companies, 1950– Measurements and employment Cohort 16 cases with >20 years in 1.41988. Cases: deaths with AD mentioned on death records. Combination of duration exposed occupations (0.7–3.1)certificate identified from multiple tracking sources and EMF index

Abbreviations: JEM, job exposure matrix.

EMF and health

the hypothesis of a link between EMF andAD is weak.

Suicide and DepressionPsychiatric disorders were discussed early inthe literature about possible chronic healtheffects of EMF exposure, but researchstopped, perhaps because the original findingswere not replicated. However, more recentlythis research area has been revived, at leastpartly as a consequence of the hypothesis thatEMF may affect melatonin levels.

Suicide. The studies on EMF and suicideare summarized in Table 7. The first of thesewas published in 1979 and was followed byfive more studies, the latest published in2000. The first study, in England and basedon 589 suicide cases and controls, was carriedout in two steps. In the first, EMF levels wereestimated based on nearby power lines. In thesecond, measurements were taken in thehomes of the study subjects (164,165). Thestudy found higher fields at case homes thanin control homes. However, the study ismethodologically limited and has been criti-cized both for the ways subjects were selectedand for the statistical analyses. The subse-quent studies have used a range of differentapproaches to assess exposure varying fromcrude techniques based on distance betweenhome and power lines, or on job titles, tomore sophisticated approaches based ondetailed information about cohorts of utilityworkers (160,166–170). Only the mostrecent study provides some support for theoriginal findings.

Depressive symptoms. The next set ofstudies addresses depressive symptomsdirectly (Table 8). The first two are difficultto interpret because of methodological limita-tions related to the procedures for selection ofstudy subjects because they did not use vali-

dated scales for identification of depressivesymptoms (171,172). In addition, the studyby Perry et al. (172) also reported unusuallyhigh average EMF levels that remain unex-plained. The remaining studies used validateddepression scales. One of these studiesshowed a clear association between proximityto power line and depression (173), whereasthe other three provided little evidence forsuch an association (174–176). The study byPoole et al. (173) is well designed; it com-pares subjects on properties abutting a powerline right-of-way to subjects further away,and the results appear internally consistent.The investigators report a relative risk of 2.8(95% CI = 1.6–5.1). McMahan et al. (175)employed a similar design and measurementsto confirm that the homes close to the linehave considerably higher EMF levels thanhomes further away. This study also appearsvalid but yields a relative risk of 0.9(0.5–1.9). McMahan et al. offer a number ofpossible explanations for the lack of consis-tency between these two studies but none ofthe explanations is convincing.

Interpretation and conclusion. Whenassessing the overall literature on EMF andsuicide, it is necessary to consider the relativeweights of the available studies together withtheir results. In doing so the original studymust be given a relatively light weight in rela-tion to the later studies because of method-ological limitations. Nevertheless, the lateststudy also suggests that an excess risk mayindeed exist.

The literature on depressive symptomsand EMF is difficult to interpret because thefindings are not consistent. This complexitycannot easily be resolved by suggesting thatone type of result can be confined to a groupof studies with methodological problems orsome other limitation.

Cardiovascular DiseasesConcerns about cardiovascular changesresulting from exposure to EMFs originatedfrom the same sources as concerns about neu-rological effects, namely, descriptions in the1960s and early 1970s of the symptomsamong Russian high-voltage switchyard oper-ators and workers (62,150). Although thesereports remain unconfirmed (177), morerecent investigations suggest that there maybe some direct cardiac effects of EMF expo-sure, mostly related to heart rate. Theseeffects, however, appear to occur only undercertain conditions (178). No known substan-tive changes occur in other parameters of car-diac function, such as the shape ofelectrocardiogram or blood pressure, in rela-tion to EMF exposure (179).

Several recent occupational cohort studieshave examined mortality from cardiovasculardisease (CVD) among electric utility workers.The first study (168) was carried out on acohort of over 20,000 workers employed inan electric company in Quebec. Exposure to60-Hz electric and magnetic fields wasassessed principally through a job-exposurematrix. Among those exposed (who were allblue-collar workers), mortality rates were gen-erally lower than those in the unexposedgroups, including overall cardiovascular mor-tality. No analyses of mortality by CVD sub-type were reported. In contrast, Savitz et al.(180) investigated risk for each subgroup offatal cardiovascular disease in a cohort of


Table 6. Pooling across groups of studies on EMF expo-sure and Alzheimer’s disease.

Number of Pooled studies studies RR 95% CI

All 5 2.2 1.5–3.2Clinical-based studies 2 3.2 1.9–5.4Population-based studies 3 1.2 0.7–2.3

Table 7. Certain characteristics and findings of studies on the relation between EMF exposure and suicide.

Definition and StudyReference Study population and subject identification estimation of exposure design Numbers Result RR (95% CI)

Reichmanis et al., 1979 Suicide cases and controls in England Estimates of residential exposure CC 589 suicide cases Higher estimated and (161); Perry et al., from power lines. Measurements at measured fields at 1981 (165) subjects’ homes case homesMcDowall, 1986 (166) Persons residing in the vicinity of Home within 50 m of substation or SMR 8 cases 0.75

transmission facilities in specified areas 30 m of overhead line (nonsignificant)in the U.K. at the time of 1971 census

Baris and Armstrong, Deaths in England and Wales, Job titles on death certificates. PMR 495 suicide cases in No increase for1990 (167) 1970–1972 and 1979–1983 Electrical workers in aggregate electrical occupations electrical workers

and specific jobsBaris et al., 1996 (168) Male utility workers, Quebec, Canada, Job exposure matrix based on positron CC 49 cases of suicide No evidence for

1970–1988. Cases: deaths from suicide, measurements. E- and B- and pulsed 215 controls magnetic fields. noted in mortality registry. Controls: fields from average and geometric Some support for 1% random sample from the cohort means and from cumulative and some electric field

current exposure indices.Johansen and Male employees in Danish utility Employment records and JEM: SMR 21,236 males in cohort- 1.4Olsen, 1998 (108) companies observed 1974–1993. estimated average exposure level. exposed cases (nonsignificant)

Cases: deaths from suicide, noted in Medium and high exposuremortality registry

Abbreviations: PMR, proportional mortality ratio; SMR, standardized mortality ration.

Ahlbom


approximately 139,000 male utility workers(180). In this study it was hypothesised a pri-ori that long-term exposure to magnetic fieldsleads to an increased risk of death due to car-diac arrhythmias and acute myocardial infarc-tion. Primary cause of death was taken fromthe death certificate; exposure was assessedaccording to the duration of employment inoccupations with high exposure to magneticfields, and by cumulative exposure, buildingin various lag periods. Although overall car-diovascular disease and ischemic mortalitieswere lower in the study cohort than in theU.S. population, deaths from arrhythmia-related conditions and acute myocardialinfarction were related to increasing exposure5–20 years before death, using both indices.The specificity of the study hypothesis,which was crucial to the findings, arose outof evidence (although inconsistent) fromhuman laboratory studies that a pattern ofreduced heart rate variability occurred imme-diately after exposure to power-frequencymagnetic fields (181). Reduction in heart ratevariability is reported to be predictive of car-diovascular disease and death in adults inpopulation-based studies (182–184).Changed heart rate variability reflectschanged cardiac autonomic control(185,186), suggesting that this is a possiblemechanism of action of EMF exposure on theheart. The limitations of speculating aboutcausal mechanisms of types of CVD as coded

on death certificates of uncertain validity andreliability have been pointed out (187). Alsothere are difficulties in explaining how themechanism underlying the transient changesin heart rate variability seen in healthy youngmen after EMF exposure in controlled set-tings (181,188) can also explain deaths fromarrhythmia and infarction many years afterlong-term occupational exposure to ELFEMFs. Indeed, a recent large study conductedin Sweden has shown no effect of EMFexposure on myocardial infarction (189).

Interpretation and conclusion. In sum-mary, evidence of cardiovascular effects dueto elevated exposure to magnetic fields isweak, and whether a specific association existsbetween exposure and altered autonomic con-trol of the heart remains speculative until cor-roborating evidence from further largeepidemiologic studies becomes available.

Reproductive EffectsIn the 1980s, laboratory findings werereported showing that weak (approximately1 µT) magnetic fields may adversely affectchick embryogenesis (190,191). In addi-tion, clusters of adverse pregnancy out-comes were reported among users of videodisplay terminals (VDTs ) (192), and epi-demiologic data were published suggestingthat maternal use of electric blankets andwater beds may influence fetal development(193). Subsequently, several studies of the

effects of EMF exposure on reproductivehealth have been conducted (194).

Residential exposure. Studies investigatingthe reproductive effects of residential expo-sure to ELF magnetic fields have evaluatedeither exposures to general residential mag-netic fields or to specific sources, namelyheated waterbeds, electric blankets, andceiling heating coils.

Several studies have been conducted ofvarious reproductive end points in relation togeneral residential exposure. With regard tospontaneous abortion, high-intensity mag-netic fields measured at the front doors ofhomes of volunteers’ homes in a “work andfertility” cohort study in Finland, were associ-ated with a marginally significant, 5-foldincreased risk (based on fewer than 10 casesand adjusted only for smoking status) (195).Two later studies, Savitz and Anath (196) andBelanger et al. (197), found no increase in riskof spontaneous abortion however. An investi-gation arising out of a case–control study ofchildhood cancer, found pregnancies inhomes with a magnetic field intensity >0.2 µTwere no more likely than others to end inspontaneous abortion (196) (again smallnumbers of cases and design limitations weak-ened the results). Similarly in a prospectivestudy of nearly 3,000 women in New Haven,Connecticut, intrauterine growth rate (IUGR)and spontaneous abortion were unrelated towire code of maternal residence (classified as

Table 8. Certain characteristics and findings of studies on the relation between EMF exposure and depression.

Definition and StudyReference Study base and subject identification estimation of exposure design Numbers Result RR (95% CI)

Dowson et al., 1988 Persons in England who lived near 132-kV Distance between home Cross- 132 near power line, 9 with Strong association (171) power line and persons who lived 3 miles and overhead power line. sectional depression; 94 away from between depression

away. Questionnaire asking about depression. power line,1 with depression and proximity to overhead power line

Perry et al., 1989 Persons with depression discharged Measurements at front CC 359 patients discharged with Average measurement:(172) from hospital in England; controls from doors. Average for case and diagnosed depressive illnesses Cases: 2.3 mG

electoral list. control groups compared. Controls: 2.1 mG Poole et al., 1993 Residents in 8 towns along a trans- Distance from power line: Cross- 382 persons interviewed 2.8 (1.6–5.1)(173) mission line right-of-way in the U.S., 1987. near vs far. Near: properties sectional

A sample was interviewed. Depressive abutting right-of-way or visiblesymptoms were identified by CES-D. towers.Cutoff for depression was median of score.

Savitz et al., 1994 Male veterans who served in the U.S. Army Present job identified in inter- Cross- 183 electrical workers, 13 with 1.0 (0.5–1.7)(174) for the first time, 1965–1971. Two diagnostic view together with duration. sectional lifetime depression; 3,861

inventories were used: the Diagnostic Electrical worker. nonelectrical workersInterview Schedule and the Minnesota Personality Inventory. Lifetime depression used for report here.

McMahan et al., Population of neighborhood near a trans- Average EMDEX measure- Cross- Total of 152 women 0.9 (0.5–1.9)1994 (175) mission line in Orange County, California, ments at the front door: sectional

USA, 1992. Sample of homes near and one Homes on easement: 4.86 mGblock away from power line. Depressive One block away: 0.68 mGsymptoms identified through questionnaire and CES-D scale.

Verkasalo et al., Finnish twins who answered the BDI Residential magnetic field Cross- 12,063 persons BDI scores not 1997 (176) in 1990. estimated from power lines sectional related to exposure

near the homes.

Abbreviations: BDI, Beck Depression Inventory; CES-D scale, Center for Epidemiologic Studies–Depression scale.

EMF and health

having HCC or LCC) (197). The secondstudy also found no increased risk of low birthweight or premature delivery in relation tohigh residential EMF exposure (196).Meanwhile, birth defects were the outcome ofinterest in a study conducted in southwesternFrance to explore whether women livingwithin 100 m of high-voltage power lines atthe time of birth had children at increased riskof congenital anomalies (198). There was nosuch increase, though too few patients livedwithin 25 m of the power lines (i.e., actuallyexperienced increased EMF exposure) to testthe association properly.

Wertheimer and Leeper (193) f irstraised the possibility of a more specific asso-ciation between maternal use of electricallyheated beds and adverse pregnancy out-come. These investigators examined sea-sonal patterns of fetal growth and abortionamong users of heated beds in Denver andreported that more abortions and morebabies of low birth weight were conceivedin winter than in summer months. Theeffects of heat could not be disentangledfrom those of EMFs however. Subsequentlythey showed a similar correlation betweenseasonality of spontaneous abortions occur-ring within a year prior to conception of aliveborn infant and exposure to ceiling cableheat (199). The data have been criticizedbecause of biased ascertainment of birthsand abortions and because the rate of con-genital malformations in the unexposedgroup was abnormally low (200).

Subsequently, four case–control studiesexamining the effects of electrically heatedbeds have been reported. No association wasseen between recalled periconceptual electricblanket or heated waterbed use and neuraltube and oral cleft defects identified in theNew York State Congenital MalformationsRegistry (201). In a study of similar design,cases of congenital urinary tract anomalieswithout chromosomal abnormalities wereidentified through the Washington BirthDefects Registry and risk was calculated inrelation to prenatal use of electric blanketsand heated waterbeds. No increase in riskwas seen among all cases and controls, but anincrease was seen in the subgroup of womenwith infertility. Low response rates amongcases and controls and the small number ofexposed cases (five) in a subgroup analysis,detract from the reliability of these data(202). More recently, two case–control datasets have been analyzed to assess risk ofneural tube defects and orofacial clefts inrelation to periconceptual use of electricblankets, bed warmers, and heated waterbeds(203). A study based on medical recordsincluding autopsy and ultrasonographyreports in clinics in various California urbanareas found no clear evidence of increased

risk of defects in relation to high frequencyor duration of use of electrically heated beds.

Two prospective studies have also beenconducted. In one the use of electricallyheated beds by nearly 3,000 women receivingcare at centers in the New Haven area wasmonitored. Time-weighted EMF exposurefrom beds was calculated based on bed-type–specific measurements multiplied bynightly hours of use reported at prenatal inter-view. No association was found between lowbirth weight or intrauterine growth rate andelectrically heated bed use (204). Althoughelectric blanket use at conception was weaklyassociated with spontaneous abortion, corre-sponding use of heated waterbeds was not. Nomeasures of dose–response were associatedwith increased risk of abortion. The otherstudy, of over 5,000 pregnant women, foundthat users of electric bed heaters had lowerrates of spontaneous abortion than nonusers,and no increase in risk with increasing inten-sity of use was seen (205).

Occupational exposure. Studies of repro-ductive outcomes in relation to maternaloccupational exposure to magnetic fields havemostly investigated pregnant women workingwith VDTs. Magnetic fields experienced byoperators of most VDTs (and certainly mod-ern VDTs are not materially higher thanthose experienced in the general environment(207–210), however. Thus the hypothesisthat increased risk of reproductive outcomesis related to increased EMF exposure logicallycannot be tested in studies where VDTs arethe sources of EMF exposure. Moreover, instudies to date, possible confounding factorssuch as stress and other work-related factorshave largely gone unaddressed (192,208).These problems notwithstanding, magneticfield exposure of VDT operators has largelybeen estimated by assessing time spent work-ing at the terminal (208), and more than adozen studies have addressed the question ofthe possible harm to pregnant women fromVDT use (192,194,207,208,211,212), withno consistent evidence of an effect. Of these aminority of studies have measured magneticfields emitted by VDTs directly, such as twolarge studies conducted in the United Statesand Finland, respectively (206,210). In thefirst (206,213), telephone operators who usedVDTs in the first trimester of pregnancy hadno excess risks of spontaneous abortions(206), low birth weight, or premature deliv-ery (213). In the second, women employed asclerks in Finland in the period 1975–1985who were selected from a national pregnancydatabase showed no overall increase in spon-taneous abortions in relation to use of VDTs,though in a very highly exposed subgroup (20exposed cases), a 3-fold increase in risk wasseen after adjusting for ergonomic factors andmental stress. The possibility of recall bias,

both of VDT use and mental stress, exists inthis study; and response rate among cases andcontrols was relatively poor. Further, only5–10% of VDT users in this study and nousers in the previous study (206) were in thehighest exposure category.

A few studies have investigated the repro-ductive health of groups besides VDT opera-tors who have been occupationally exposed toEMFs. The increase in congenital malforma-tions observed in the offspring of some 370married men employed by a Swedish powercompany (214), was not observed in morerecent studies (215), suggesting that the for-mer may have been a chance result.Moreover, no plausible biological explanationfor paternal transmission of risk is known(192,216). Similarly, little support has beenfound for the theories that either fertility ofexposed workers (208,217) or the sex ratio oftheir offspring (192,218) are perturbed byexposure to low-level ELF magnetic fields.

Conclusion. Until the recent cohortstudies of pregnancy outcome following resi-dential and electric blanket EMF exposure(197,204,205), little evidence has been avail-able on the effect of EMF exposure on overallreproductive health (204,219). Investigationsaddressing the diversity of reproductive out-comes are notoriously difficult, with assess-ment of spontaneous abortions beingparticularly so (216). Not only has the accu-racy of pregnancy outcome assessment beenquestionable in many studies, but also expo-sure measurement has been of variable valueand this is especially true of the vast majorityof studies addressing reproductive health inrelation to VDT use, which offer little infor-mation on EMF exposure.

Although there may be some relationsamong reproductive outcomes either throughshared determinants or because one eventprecludes the occurrence of another (e.g.,infertility and spontaneous abortion), themost realistic and promising strategy is tofocus on specific, narrowly defined reproduc-tive outcomes. When relevant studies are sub-divided in that way, only spontaneousabortion has been examined in several studiesof reasonable quality, and the evidence fromthose studies cumulatively suggests no associ-ation with EMF exposure is present.

Thus fundamental methodologic limita-tions preclude firm conclusions about repro-ductive outcomes. Studies with refinedmeasurements of exposure and outcomecould yield different results than thosereported to date. However, on the basis oftheoretical considerations and both experi-mental and epidemiologic studies (43,103),there is very little encouragement for pursu-ing research on EMF and reproductivehealth. Existing evidence does not supportthe hypothesis that maternal exposure to


Ahlbom


EMF through residential, including heatedbed, exposure or through the workplace isassociated with adverse pregnancy outcomes.

Discussion

Epidemiologic investigation of possible asso-ciations of EMF exposure with risk of chronicdisease is an unusually difficult enterprise.Certain conclusions can be drawn however:

a) The epidemiologic studies conductedon possible health effects of EMF haveimproved over time in sophistication of expo-sure assessment and in methodology. Severalof the recent studies on childhood leukemiaand on occupational exposures in relation toadult cancer are close to the limit of what canrealistically be achieved by epidemiology, interms of size of study and methodologicalrigor, using presently available measurementmethods.

b) Exposure measurement is a particulardifficulty of EMF epidemiology, in severalrespects:• The exposure of interest is imperceptible,

ubiquitous, originates from multiplesources, and can vary greatly over timeand over relatively short distances.

• The relevant exposure period, for cancersat least, is before the date at which mea-surements can realistically be obtainedand is of unknown duration and induc-tion period.

• The appropriate exposure metric isunknown, and there is no substantiatedbiological mechanism or animal modelfrom which to impute it.c) In the absence of evidence from cellular

or animal studies, and given the methodolog-ical uncertainties and in many cases inconsis-tencies of the existing epidemiologicliterature, there is no chronic disease outcomefor which an etiological relation to EMFexposure can be regarded as established.

d ) A large body of high-quality dataexists, with measurements of exposure,strong methodology, and large study sizes,for childhood leukemia and brain tumorsand for occupational exposure in relation toadult leukemia and brain tumors. Among allthe outcomes evaluated in epidemiologicstudies of EMF, childhood leukemia in rela-tion to postnatal exposures above 0.4 µT isthe one for which there is most evidence ofan association. The relative risk has been esti-mated at 2.0 (95% confidence limits (CL) =1.27–3.13) in a large pooled analysis. This isunlikely to be due to chance but may bepartly due to bias. This is difficult to inter-pret in the absence of a known mechanismor reproducible experimental support. In thelarge pooled analysis, only 0.8% of all chil-dren were exposed above 0.4 µT. Furtherstudies need to be designed to test specifichypotheses such as aspects of selection bias or

exposure. On the basis of epidemiologicfindings, there is evidence for an associationof ALS with occupational EMF exposurealthough confounding is a potential explana-tion. Whether there are associations withbreast cancer, cardiovascular disease, andsuicide and depression remains unresolved.

Overall, despite 20 years of extensive epi-demiologic investigation of the relation ofEMF to risk of chronic disease, there are stillepidemiologic questions that need to beresolved. To be of value, however, futurestudies of these questions must be of highmethodological quality, of sufficient size andwith sufficient numbers of highly exposedsubjects, and must include appropriate expo-sure groups and sophisticated exposure assess-ment. Especially for childhood leukemia,little is to be gained from further repetition ofinvestigation of risks at moderate and lowexposure levels, unless such studies can bedesigned to test specific hypotheses, such asselection bias or aspects of exposure not previ-ously captured. In addition there is a need forstudies in humans of possible physiologicaleffects of EMF that might relate to risks ofchronic disease.

REFERENCES AND NOTES

1. Jackson JD. Are the stray 60-Hz electromagnetic fields associ-ated with distribution and use of electric power a significantcause of cancer? Proc Natl Acad Sci U S A 89:3508–3510(1992).

2. Swanson J. Long-term variations in the exposure of the popula-tion of England and Wales to power-frequency magnetic fields.J Radiol Prot 16:287–301 (1996).

3. Wertheimer N, Leeper E. Electrical wiring configurations andchildhood cancer. Am J Epidemiol 109:273–284 (1979).

4. Repacholi M, Ahlbom A. Commentary. Lancet 354:1918–1919(1999).

5. Kaune WT, Davis S, Stevens RG, Mirick DK, Kheifets L.Measuring temporal variability of residential magnetic fieldexposures. Bioelectromagnetics 22(4):232–245 (2001).

6. Milham S. Mortality from leukemia in workers exposed to elec-trical and magnetic fields [Letter to the Editor]. N Engl J Med307:249 (1982).

7. Forssén UM, Feychting M, Rutqvist LE, Floderus B, Ahlbom A.Occupational and residential magnetic field exposure andbreast cancer in females. Epidemiology 11:24–29 (2000).

8. Green LM, Miller AB, Agnew DA, Greenberg ML, Li J,Villeneuve PJ, Tibshirani R. Childhood leukemia and personalmonitoring of residential exposures to electric and magneticfields in Ontario, Canada. Cancer Causes Control 10:233–243(1999).

9. McBride ML, Gallagher RP, Theriault G, Armstrong BG, TamaroS, Spinelli JJ, Deadman JE, Fincham S, Robson D, Choi W.Power-frequency electric and magnetic fields and risk of child-hood leukemia in Canada. Am J Epidemiol 149:831–842 (1999).

10. Kleinerman RA, Linet MS, Hatch EE, Wacholder S, Tarone RE,Severson RK, Kaune WT, Friedman DR, Haines CM, MuirheadCR, et al. Magnetic field exposure assessment in a case-controlstudy of childhood leukemia. Epidemiology 8:575–583 (1997).

11. Wertheimer N, Leeper E. Adult cancer related to electricalwires near home. Am J Epidemiol 11:345–355 (1982); 273–284(1979).

12. Savitz DA, Wachtel H, Barnes FA, John EM, Tvrdik JG. Case-control study of childhood cancer and exposure to 60-hertzmagnetic fields. Am J Epidemiol 128:21–38 (1988).

13. Kaune WT, Stevens RG, Callahan NJ, Severson RK, Thomas DE.Residential magnetic and electric fields. Bioelectromagnetics8:315–335 (1987).

14. Barnes F, Wachtel H, Savitz D, Fuller J. Use of wiring configura-tion and wiring codes for estimating externally generated elec-tric and magnetic fields. Bioelectromagnetics 10:13–21 (1989).

15. London SJ, Thomas DC, Bowman JD, Sobel E, Cheng T-C,Peters JM. Exposure to residential electric and magnetic fieldsand risk of childhood leukemia. Am J Epidemiol 134:923–937(1991).

16. Tarone RE, Kaune WT, Linet MS, Hatch EE, Kleinerman RA,Robison LL, Boice JD Jr, Wacholder, S. Residential wire codes:reproducibility and relation with measured magnetic fields.Occup Environ Med 55:333–339 (1998).

17. Kheifets LI, Kavet R, Sussman SS. Wire codes, magnetic fields,and childhood cancer. Bioelectromagnetics 18:99–110 (1997).

18. Kaune WT, Zaffanella LE. Assessing historical exposures ofchildren to power-frequency magnetic fields. J Expo AnalEnviron Epidemiol 4(2):149–170 (1994).

19. Preston-Martin S, Navidi W, Thomas D, Lee P-J, Bowman J,Pogoda J. Los Angeles study of residential magnetic fields andchildhood brain tumors. Am J Epidemiol 143:105–119 (1996).

20. Kaune WT, Savitz DA. Simplification of the Wertheimer-Leeperwire code. Bioelectromagnetics 15:275–282 (1994).

21. Linet MS, Hatch EE, Kleinerman RA, Robison LL, Kaune WT,Friedman DR, Severson RK, Haines CM, Hartsock CT, Niwa S, etal. Residential exposure to magnetic fields and acute lym-phoblastic leukemia in children. N Engl J Med 337:1–7 (1997).

22. Savitz DA, Kaune WT. Childhood cancer in relation to a modi-fied residential wire code. Environ Health Perspect 101:76–80(1993).

23. Tomenius L. 50-Hz electromagnetic environment and the inci-dence of childhood tumors in Stockholm County.Bioelectromagnetics 7:191–207 (1986).

24. Coleman MP, Bell CMJ, Taylor HL, Primic-Zakelj M. Leukemiaand residence near electricity transmission equipment: a case-control study. Br J Cancer 60:793–798 (1989).

25. Myers A, Clayden AD, Cartwright RA, Cartwright SC. Childhoodcancer and overhead power lines. A case-control study. Br JCancer 62:1008–1011 (1990).

26. Tynes T, Haldorsen T. Electromagnetic fields and cancer in chil-dren residing near Norwegian high-voltage power lines. Am JEpidemiol 145:219–226 (1997).

27. Verkasalo PK, Pukkala E, Hongstro MY, Valjus JE, Jarvinen PJ,Heikkila KV, Koskenvuo M. Risk of cancer in Finnish children liv-ing close to power lines. Br Med J 307:895–899 (1993).

28. Olsen JH, Nielsen A, Schulgen G. Residence near high voltagefacilities and risk of cancer in children. Br Med J 307:891–895(1993).

29. Kleinerman RA, Kaune WT, Hatch EE, Wacholder S, Linet MS,Robison LL, Niwa S, Tarone RE. Are children living near high-voltage power lines at increased risk of acute lymphoblasticleukemia? Am J Epidemiol 151:512–515 (2000).

30. Ericson A, Kallen B. An epidemiological study of work withvideo screens and pregnancy outcome. II: A case-control study.Am J Ind Med 9:459–475 (1986).

31. Lynge E, Kurppa K, Kristofersen L, Malker H, Sauli H. Silicadust and lung cancer: results from the Nordic occupationalmortality and cancer incidence registers. J Natl Cancer Inst77:883–889 (1986).

32. Feychting M, Ahlbom A. Magnetic fields and cancer in childrenresiding near Swedish high-voltage power lines. Am JEpidemiol 138:467–481 (1993).

33. Michaelis J, Schuz J, Meinert R, Menger M, Grigat J-P,Kaatsch P, Kaletsch U, Miesner A, Stamm A, Brinkmann K, etal. Childhood leukemia and electromagnetic fields: results of apopulation-based case-control study in Germany. CancerCauses Control 8:167–174 (1997).

34. Michaelis J, Schuz J, Meinert R, Zemann E, Grigat J-P, KaatschP, Kaletsch U, Miesner A, Brinkmann K, Kalkner W, et al.Combined risk estimates for two German population-basedcase-control studies on residential magnetic fields and child-hood acute leukemia. Epidemiology 9:92–94 (1997b).

35. Dockerty JD, Elwood JM, Skegg DCG, Herbison GP.Electromagnetic field exposures and childhood cancers in NewZealand. Cancer Causes Control 9:299–309 (1998).

36. Green LM, Miller AB, Villeneuve PJ, Agnew DA, Greenberg ML,Li J, Donnelly KE. A case-control study of childhood leukemia insouthern Ontario, Canada, and exposure to magnetic fields inresidences. Int J Cancer 82:161–170 (1999b).

37. United Kingdom Childhood Cancer Study Investigators.Exposure to power-frequency magnetic fields and the risk ofchildhood cancer. Lancet 354:1925–1931 (1999).

38. Jaffe KC, Kim H, Aldrich TE. The relative merits of contemporarymeasurements and historical calculated fields in the Swedishchildhood cancer study. Epidemiology 11:353–356 (2000).

39. Kaune WT, Darby SD, Gardner SN, Hrubec Z, Iriye RN, LinetMS. Development of a protocol for assessing time-weighted-average exposures of young children to power-frequency mag-netic fields. Bioelectromagnetics 15:33–51 (1994).

EMF and health

40. Friedman DR, Hatch EE, Tarone RE, Kaune WT, Kleinerman RA,Wacholder, S, Boice JD Jr, Linet MS. Childhood exposure tomagnetic fields: residential area measurements compared topersonal dosimetry. Epidemiology 7:151–155 (1996).

41. United Kingdom Childhood Cancer Study Investigators. UnitedKingdom Childhood Cancer Study: objectives, materials, andmethods. Br J Cancer 82:1073–1102 (2000).

42. Morgan MG, Nair I. Alternative functional relations betweenELF field exposure and possible health effects: report on anexpert workshop. Bioelectromagnetics 13:335–350 (1992).

43. National Research Council. Possible Health Effects of Exposureto Residential Electric and Magnetic Fields. WashingtonDC:National Academy Press, 1997.

44. Portier CJ, Wolfe MS, eds. Assessment of Health Effects fromExposure to Power-line Frequency Electric and Magnetic Fields.Working Group Report. NIH Publ no. 98-3981. Research TrianglePark, NC:National Institute of Environmental Health Sciences,1998;10.

45. Wilson BW, Lee GM, Yost MG, Davis KC, Heimbigner T,Buschbom RL. Magnetic field characteristics of electric bed-heating devices. Bioelectromagnetics 17:174–179 (1996).

46. Burch JB, Reif JS, Yost MG, Keefe TJ, Pitrat CA. Nocturnalexcretion of a urinary melatonin metabolite among electric util-ity workers. Scand J Work Environ Health 24:183–189 (1998).

47. Neter J, Wasserman W, Kutner JH. Applied Linear StatisticalModels, 3rd ed. Homewood, IL:Irwin Publishing, 1990;502–504.

48. Electric Power Research Institute. Residential TransientMagnetic Field Research: Interim Report. Project RP2966-07.Report TR-103470. Palo Alto, CA:EPRI, 1994.

49. Auvinen A, Linet MS, Hatch EE, Kleinerman RA, Robison LL,Kaune WT, Misakian M, Niwa S, Wacholder S, Tarone RE.Extremely low-frequency magnetic fields and childhood acutelymphoblastic leukemia: An exploratory analysis of alternativeexposure metrics. Am J Epidemiol 152:20–31 (2000).

50. Blackman DF. ELF effects on calcium homeostasis. In: ExtremelyLow Frequency Electromagnetic Fields: The Question of Cancer(Wilson BW, Stevens RG, Anderson LE, eds). Columbus, OH:Battelle Press, 1990.

51. Little J. Epidemiology of childhood cancer. IARC Sci Publ149:1–386 (1999).

52. Wiemels JL, Cazzaniga G, Daniotti M, Eden OB, Addison GM,Masera G, Saha V, Biondi A, Greaves MF. Prenatal origin ofacute lymphoblastic leukaemia in children. Lancet354:1499–1503 (1999).

53. Fulton JP, Cobb S, Preble L, Leone L, Forman E. Electrical wiringconfigurations and childhood leukemia in Rhode Island. Am JEpidemiol 111:292–296 (1980).

54. Bell J, Coleman MP. Extremely low frequency (ELF) electromag-netic fields and leukaemia in children. Br J Cancer 62:331–332(1990).

55. Gurney JG, Mueller BA, Davis S, Schwartz SM, Stevens RG,Kopecky KJ. Childhood brain tumor occurrence in relation toresidential power line configurations, electric heating sources,and electric appliance use. Am J Epidemiol 143:120–128(1996).

56. Petridou E, Trichopoulos D, Kravaritis A, Pourtsidis A, DessyprisN, Skalkidis Y, Kogevinas M, Kalmanti M, Koliouskas D,Kosmidis H, et al. Electrical power lines and childhoodleukemia: a study from Greece. Int J Cancer 73:345–348 (1997).

57. Severson RK, Stevens RG, Kaune WT, Thomas DB, Heuser L,Davis S, Sever LE. Acute nonlymphocytic leukemia and residen-tial exposure to power frequency magnetic fields. Am JEpidemiol 128(suppl 1):10–20 (1988).

58. Feychting M, Ahlbom A. Magnetic Fields and Cancer in PeopleResiding near Swedish High Voltage Power Lines. Stockholm:Karolinska Institute, 1992.

59. Feychting M, Ahlbom A. Magnetic fields, leukemia, and centralnervous system tumors in Swedish adults residing near high-voltage power lines. Epidemiology 5:501–509 (1994).

60. Dovan T, Kaune WT, Savitz D. Repeatability of measurementsof residential magnetic fields and wire codes. Bioelectro-magnetics 14:145–159 (1993).

61. Banks RS, Thomask W, Mandel JS, Kaune WT, Wacholder, S,Tarone RE, Linet MS. Temporal trends and misclassification inresidential 60-Hz magnetic field measurements.Bioelectromagnetics (in press).

62. Asanova TP, Rakov AM. Health conditions of workers exposedto electric fields on open switchboard installations of 400–500kV. Cig Tr Zabol 10:50–52 (1966).

63. Armstrong BG, Deadman JE, Thériault G. Comparison of indicesof ambient exposure to 60-Hertz electric and magnetic fields.Bioelectromagnetics 11:337–347 (1990).

64. Loomis DP, Savitz DA. Mortality from brain cancer and leukemiaamong electrical workers. Br J Ind Med 47:633–638 (1990).

65. Tynes T, Andersen A, Langmark F. Incidence of cancer inNorwegian workers potentially exposed to electromagneticfields. Am J Epidemiol 136:81–88 (1992).

66. Pearce N, Reif J, Fraser J. Case-control studies of cancer inNew Zealand electrical workers. Int J Epidemiol 18:55–59(1989).

67. Kelsh M, Kheifets L, Smith R. Assessing the impact of workenvironment, utility and sampling design on occupational expo-sure summaries: a case study of magnetic field exposuresamong electric utility workers. Am Ind Hyg Assoc J 61:174–182(2000).

68. Floderus B. Is job title an adequate surrogate to measure mag-netic field exposure? Epidemiology 7:115–116 (1996).

69. Floderus B, Persson T, Stenlund C. Magnetic field exposures inthe workplace: reference distribution and exposures in occupa-tional groups. Int J Occup Med Environ Health 2:226–238(1996).

70. London SJ, Bowman JD, Sobel E, Thomas DC, Garabrant DH,Pearce N, Bernstein L, Peters JM. Exposure to magnetic fieldsamong electrical workers in relation to leukemia risk in LosAngeles County. Am J Ind Med 26:47–60 (1994).

71. Floderus B, Persson T, Stenlund C, Wennberg A, Ost A, KnaveB. Occupational exposure to electromagnetic fields in relationto leukemia and brain tumors: a case-control study in Sweden.Cancer Causes Control 4:465–476 (1993).

72. Kromhout H, Loomis DP, Mihlan GJ, Peipins LA, Kleckner RC,Iriye R, Savitz DA. Assessment and grouping of occupationalmagnetic field exposure in five electric utility companies. ScandJ Work Environ Health 21:43–50 (1995).

73. Thériault G, Goldberg M, Miller AB, Armstrong B, Guenel P,Deadman J, Imbernon E, To T, Chevalier A, Cyr D, et al. Cancerrisks associated with occupational exposure to magnetic fieldsamong electric utility workers in Ontario and Quebec, Canada,and France: 1970–1989. Am J Epidemiol 139:550–572 (1994).

74. Harrington JM, McBride DI, Sorahan T, Paddle GM, vanTongeren M. Occupational exposure to magnetic fields in rela-tion to mortality from brain cancer among electricity generationand transmission workers. Occup Environ Med 54:7–13 (1997).

75. Miller AB, To T, Agnew DA, Wall C, Green LM. Leukemia fol-lowing occupational exposure to 60-Hz electric and magneticfields among Ontario electric utility workers. Am J Epidemiol144:150–160 (1996).

76. Kheifets L. Occupational exposure assessment in epidemiologicstudies of EMF. Radiat Prot Dosim 83:61–69 (1999).

77. National Radiation Protection Board. Electromagnetic fields andthe risk of cancer. Report of an advisory group on non-ionisingradiation. NRPB 3:1–138 (1992).

78. National Radiation Protection Board. ELF electromagnetic fieldsand the risk of cancer. report of an advisory group on non-ionis-ing radiation. NRPB 12(1) (2001).

79. Dockerty JD, Elwood JM, Skegg DCG, Herbison GP.Electromagnetic field exposures and childhood leukaemia inNew Zealand. Lancet 354:1967–1968 (1999).

80. Schüz J, Grigat JP, Brinkmann K, Michaelis J. Childhood acuteleukemia and residential 16.7 Hz magnetic fields in Germany. BrJ Cancer 84(5):697–699 (2001)

81. Kheifets LI, Afifi AA, Buffler PA, Zhang ZW, Matkin CC.Occupational electric and magnetic field exposure andleukemia. A meta-analysis. J Occup Environ Med 39:1074–1091(1997).

82. Wartenberg D. Residential magnetic fields and childhoodleukemia: meta-analysis. Am J Public Health 88:1787–1794(1998).

83. Loomis D, Lagorio S, Salvan A, Comba P. Update of evidence onthe association of childhood leukemia and 50/60 Hz magneticfield exposure. J Expo Anal Environ Epidemiol 9:99–105 (1999).

84. Angelillo IF, Villari P. Residential exposure to electromagneticfields and childhood leukaemia: a meta-analysis. Bull WorldHealth Org 77:906–915 (1999).

85. Ahlbom A, Day N, Feychting M, Roman E, Skinner J, Dockerty J,Linet M, McBride M, Michaelis J, Olsen JH, et al. A pooledanalysis of magnetic fields and childhood leukaemia. Br JCancer 83:692–698 (2000).

86. Greenland S, Sheppard AR, Kaune WT, Poole C, Kelsh. A pooledanalysis of magnetic fields, wire codes, and childhoodleukemia. Epidemiology 11:624–634 (2000).

87. Greenland S. Can meta-analysis be salvaged? Am J Epidemiol140:783–787 (1994).

88. Blair A, Burg J, Foran J, Gibb H, Greenland S, Morris R, RaabeG, Savitz D, Teta J, Wartenberg D. Guidelines for application ofmeta-analysis in environmental epidemiology. ISLI Risk ScienceInstitute. Regul Toxicol Pharmacol 22:189–197 (1995).

89. Poole C, Greenland S. Random-effects meta-analyses are notalways conservative. Am J Epidemiol 150:469–475 (1999).

90. Savitz DA, John EM, Kleckner RC. Magnetic field exposure fromelectric appliances and childhood cancer. Am J Epidemiol131:763–773 (1990).

91. Hatch EE, Linet MS, Kleinerman RA, Tarone RE, Severson RK,Hartsock CT, Haines C, Kaune WT, Friedman D, Robison LL, etal. Association between childhood acute lymphoblasticleukemia and use of electrical appliances during pregnancy andchildhood. Epidemiology 9:234–245 (1998).

92. Kaune WT, Miller MC, Linet MS, Hatch EE, Kleinerman RA,Wacholder, S, Mohr AH, Tarone RE, Haines C. Children’s expo-sure to magnetic fields produced by U.S. television sets usedfor viewing programs and playing video games.Bioelectromagnetics 21:214–227 (2000).

93. Hatch EE, Kleinerman RA, Linet MS, Tarone RE, Kaune WT,Auvinen A, Baris D, Robison LL, Wacholder, S. Do confoundingor selection factors of residential wiring codes and magneticfields distort findings of electromagnetic fields studies?Epidemiology 11:189–198 (2000).

94. Gurney JG, Davis S, Schwartz SM, Mueller BA, Kaune WT,Stevens RG. Childhood cancer occurrence in relation to powerline configurations: a study of potential selection bias in case-control studies. Epidemiology 6(suppl 1):31–35 (1995).

95. Jones TL, Shih CH, Thurston DH, Ware BJ, Cole P. Selectionbias from differential residential mobility as an explanation forassociation of wire codes with childhood cancer. J ClinEpidemiol 46:545–548 (1993).

96. Dietz WH, Gortmaker SL. Do we fatten our children at the tele-vision set? Obesity and television viewing in children and ado-lescence. Pediatrics 75:807–812 (1985).

97. Savitz DA, Feingold L. Association of childhood cancer with res-idential traffic density. Scand J Work Environ Health15(5):360–363 (1989).

98. Feychting M, Svensson D, Ahlbom A. Exposure to motor vehicleexhaust and childhood cancer. Scand J Work Environ Health24:8–11 (1998).

99. Copeland KT, Checkoway H, McMichael AF, Holbrook RH. Biasdue to misclassification in the estimation of relative risk. Am JEpidemiol 105:488–495 (1977).

100. Stevens RG. Electric power use and breast cancer: a hypothe-sis. Am J Epidemiol 125:556–561 (1987).

101. Stevens RG, Davis S, Thomas DB, Anderson LE, Wilson BW.Electric power, pineal function, and the risk of breast cancer.FASEB J 6:853–860 (1992).

102. Kheifets LI, Afifi AA, Buffler PA, Zhang ZW. Occupational elec-tric and magnetic field exposure and brain cancer: a meta-analysis. J Occup Environ Med 37:1327–1341 (1995).

103. Portier CJ, Wolfe MS (eds). Assessment of Health Effects fromExposure to Power-Line Frequency Electric and Magnetic Fields.NIH Publ no 98–3981. Research Triangle Park, NC:NationalInstitute of Environmental Health Sciences, 1998.

104. Kheifets LI, Gilbert ES, Sussman SS, Guenel P, Sahl JD, SavitzDA, Thériault G. Comparative analyses of the studies of mag-netic fields and cancer in electric utility workers: studies fromFrance, Canada, and the United States. Occup Environ Med56:567–574 (1999).

105. Sahl JD, Kelsh Ma, Greenland S. Cohort and nested case-con-trol studies of hematopoietic cancers and brain cancer amongelectric utility workers. Epidemiology 4:104–114 (1993).

106. Savitz DA, Loomis DP. Magnetic field exposure in relation toleukemia and brain cancer mortality among electric utility work-ers. Am J Epidemiol 141:123–134 (1995).

107. Tynes T, Jynge H, Vistnes AI. Leukemia and brain tumors inNorwegian railway workers, a nested case-control study. Am JEpidemiol 139:643–653 (1994).

108. Johansen C, Olsen JH. Risk of cancer among Danish utilityworkers—a nationwide cohort study. Am J Epidemiol147:548–555 (1998).

109. Matanoski GM, Elliott EA, Breysse PN, Lynberg MC. Leukemiain telephone linemen. Am J Epidemiol 137:609–619 (1993).

110. Feychting M, Forssen U, Floderus B. Occupational and residen-tial magnetic field exposure and leukemia and central nervoussystem tumors. Epidemiology 8:384–389 (1997).

111. Guenel P, Nicolau J, Imbernon E, Chevalier A, Goldberg M.Exposure to 50-Hz electric field and incidence of leukemia, braintumors, and other cancers among French electric utility work-ers. Am J Epidemiol 144:1107–1121 (1996).

112. Kheifets L, London S, Peters J. Leukemia risk and occupationalelectric field exposure in Los Angeles county. Am J Epidemiol146(suppl 1):87–90 (1997).

113. Li CY, Thériault G, Lin RS. Residential exposure to 60-Hz magneticfields and adult cancers in Taiwan. Epidemiology 8:25–30 (1997).

114. Verkasalo PK, Pukkala E, Kaprio J. Magnetic fields andleukemia: risk for adults living close to power lines (thesis).Scand J Environ Health 22(suppl 2):7–55 (1998).


Ahlbom


115. Preston-Martin S, Peters JM, Garabrant DH, Bowman JD.Myelogenous leukemia and electric blanket use.Bioelectromagnetics 9(suppl 3):207–213 (1988).

116. Lovely RH, Buschborn RL, Slavich AL, Anderson LE, Hansen NH,Wilson BW. Adult leukemia and personal appliance use: a pre-liminary study. Am J Epidemiol 140:510–517 (1994).

117. Sussman SS, Kheifets L. Adult leukemia risk and personal appli-ance use: a preliminary study (letter). Am J Epidemiol 143(suppl 7):743–745 (1996).

118. Lin RS, Dischinger PC, Conde J, Farrell KP. Occupational expo-sure to electromagnetic fields and the occurrence of braintumors. J Occup Med 27:413–419 (1985).

119. Rodvall Y, Ahlbom A, Stenlund C, Preston-Martin S, Lindh T,Spannare B. Occupational exposure to magnetic fields and braintumours in central Sweden. Eur J Epidemiol 14:563–569 (1998).

120. Verkasalo PK, Pukkala E, Kaprio J, Heikkila KV, Koskenvuo M.Magnetic fields of high voltage power lines and risk of cancerin Finnish adults: nationwide cohort study. Br Med J313:1047–1051 (1996).

121. Wrensch M, Yost M, Miike R, Lee G, Touchstone J. Adultglioma in relation to residential power frequency electromag-netic field exposures in the San Francisco Bay area.Epidemiology 10:523–527 (1999).

122. Brainard GC, Kavet R, Kheifets LI. The relationship betweenelectromagnetic field and light exposures to melatonin andbreast cancer risk: a review of the relevant literature. J PinealRes 26:65–100 (1999).

123. Selmaoui B, Lambrozo J, Touitou Y. Magnetic fields and pinealfunction in humans: evaluation of nocturnal acute exposure toextremely low frequency magnetic fields on serum melatonin anurinary 6-sulfatoxymelatonin circadian rythms. LifeSci58(18):1539–1549 (1996).

124. Graham C, Cook MR, Riffle DW, Gerkovich MM, Cohen HD.Nocturnal melatonin levels in human volunteers exposed tointermittent 60 Hz magnetic fields. Bioelectromagnetics17:263–273 (1996).

125. Akerstedt T, Arnetz B, Ficca G, Paulsson LE, Kallner A. A 50-Hzelectromagnetic field impairs sleep. J Sleep Res 8:77–81(1999).

126. Wilson BW, Wright CW, Morris JE, Buschbom RL, Brown DP,Miller DL, Sommers-Flannigan R, Anderson LE. Evidence for aneffect of ELF electromagnetic fields on human pineal glandfunction. J Pineal Res 9(4):259–269 (1990).

127. Burch JB, Reif JS, Yost MG, Keefe TJ, Pitrat CA. Reducedexcretion of a melatonin metabolite in workers exposed to 60Hz magnetic fields. Am J Epidemiol 150:27–36 (1999).

128. Tynes T, Andersen A. Electromagneic fields and male breastcancer [Letter]. Lancet 336:1596 (1990).

129. Matanoski GM, Breysse PN, Elliott EA. Electromagnetic fieldexposure and male breast cancer [Letter]. Lancet 337:737(1991).

130. Demers PA, Thomas DB, Rosenblatt KA, Jimenez LM,McTiernan A, Stalsberg H, Stemhagen A, Thompson WD,Curnen MG, Satariano W, et al. Occupational exposure to elec-tromagnetic fields and breast cancer in man. Am J Epidemiol134:340–347 (1991).

131. Rosenbaum PF, Vena JE, Zielezny MA, Michalek AM.Occupational exposures associated with male breast cancer.Am J Epidemiol 139:30–36 (1994).

132. Loomis DP, Savitz DA, Ananth CV. Breast cancer mortalityamong female electrical workers in the United States. J NatlCancer Inst 86:921–925 (1994).

133. Cantor KP, Dosemeci M, Brinton LA, Stewart PA. Re: Breastcancer mortality among female electrical workers in the UnitedStates [Letter]. J Natl Cancer Inst 87:3 (1995).

134. Coogan PF, Clapp RW, Newcomb PA, Wenzl TB, Bogdan G,Mittendorf R, Baron JA, Longnecker MP. Occupational exposureto 60-hertz magnetic fields and risk of breast cancer in women.Epidemiology 7:459–464 (1996).

135. Kliukiene J, Tynes T, Martinsen JI, Blaasaas KG, Andersen A.Incidence of breast cancer in a Norwegian cohort of womenwith potential workplace exposure to 50 Hz magnetic fields. AmJ Ind Med 36:147–154 (1999).

136. Feychting M, Forssén U, Rutqvist LE, Ahlbom A. Magnetic fieldsand breast cancer in Swedish adults residing near high-voltagepower lines. Epidemiology 9:392–397 (1998).

137. Vena JE, Graham S, Hellmann R, Swanson M, Brasure J. Use ofelectric blankets and risk of postmenopausal breast cancer. AmJ Epidemiol 134(suppl 2):180–185 (1991).

138. Vena JE, Freudenheim J.L., Marshall J.R., Laughlin R, SwansonM, Graham S. Risk of premenopausal breast cancer and use ofelectric blankets. Am J Epidemiol 140(suppl 11):974–979(1994).

139. Gammon MD, Schoenberg JB, Britton JA, Kelsey JL, Stanford

JL, Malone KE, Coates RJ, Brogan DJ, Potischman N, SwansonCA, et al. Electric blanket use and breast cancer risk amongyounger women. Am J Epidemiol 148(suppl 6):556–563 (1998).

140. Zheng T, Holford TR, Mayne ST, Owens PH, Zhang B, Boyle P,Carter D, Ward B, Zhang Y, Zahm SH. Exposure to electromag-netic fields from use of electric blankets and other in-homeelectrical appliances and breast cancer risk. Am J Epidemiol151(suppl 11):1103–1111 (2000).

141. Laden F, Neas LM, Tolbert PE, Holmes MD, Hankinson SE,Spiegelman D, Speizer FE. Hunter DJ. Electric blanket use andbreast cancer in the Nurses Health Study. Am J Epidemiol152(suppl 1):41–49 (2000).

142. Kheifets L, Matkin C. Industrialization, electromagnetic fields,and breast cancer risk. Environ Health Perspect 107:145–154(1999).

143. Armstrong B, Thériault G, Guenel P, Deadman J, Goldberg M,Heroux P. Association between exposure to pulsed electro-magnetic fields and cancer in electric utility workers in Quebec,Canada, and France. Am J Epidemiol 140:805–820 (1994).

144. Savitz DA, Dufort V, Armstrong B, Theriault G. Lung cancer inrelation to employment in the electrical utility industry andexposure to magnetic fields. Occup Environ Med 54:396–402(1997).

145. Milham S Jr. Mortality in workers exposed to electro-magneticfields. Environ Health Perspect 62:297–300 (1985).

146. Schroeder JC, Savitz DA. Lymphoma and multiple myelomamortality in relation to magnetic field exposure among electricutility workers. Am J Ind Med 32:392–402 (1997).

147. Wilkins JR III, Wellage LC. Brain tumor risk in offspring of menoccupationally exposed to electric and magnetic fields. Scand JWork Environ Health 22:339–345 (1996).

148. Sorahan T, Hamilton L, Gardiner K, Hodgson JT, Harrington JM.Maternal occupational exposure to electromagnetic fieldsbefore, during, and after pregnancy in relation to risks of child-hood cancers: findings from the Oxford Survey of ChildhoodCancers, 1953-1981 deaths. Am J Ind Med 35:348–357 (1999).

149. Feychting M, Floderus B, Ahlbom A. Parental occupationalexposure to magnetic fields and childhood cancer (Sweden).Cancer Causes Control 11:151–156 (2000).

150. Asanova TP, Rakov AN. The State of Health Of Persons Workingin the Electric Field of Outdoor 400 kV and 500 kV Switchyards.Piscataway, NJ:Institute of Electrical and Electronic EngineersPower Engineering Society, 1972;10.

151. Knave B, Gamberale F, Bergström S, Birke E, Iregren A,Kolmodin-Hedman B, Wennberg A. Long-term exposure to elec-tric fields—a cross-sectional epidemiological investigation ofoccupationally exposed workers in high-voltage substations.Scand J Work Environ Health 5:115–125 (1979).

152. Johansen C, Koch-Henriksen N, Rasmussen S, Olsen JH.Multiple sclerosis among utility workers. Neurology 52(suppl6):1279–1282 (1999).

153. Wechsler LS, Checkoway H, Franklin GM, Costa LG. A pilotstudy of occupational and environmental risk factors forParkinson’s disease. Neurotoxicology 12:387–392 (1991).

154. Deapen DM, Henderson BE. A case-control study of amy-otrophic lateral sclerosis. Am J Epidemiol 123:790–798 (1986).

155. Gunnarsson L-G, Lindberg G, Söderfeldt B, Axelson O.Amyotrophic lateral sclerosis in Sweden in relation to occupa-tion. Acta Neurol Scand 83:394–398 (1991).

156. Gunnarsson L-G, Bodin L, Söderfeldt B, Axelson O. A case-con-trol study of motor neurone disease: its relation to heritability,and occupational exposures, particularly to solvents. Br J IndMed 49:791–798 (1992).

157. Davanipour Z, Sobel E, Bowman JD,Qian Z, Will AD.Amyotrophic lateral sclerosis and occupational exposure toelectromagnetic fields. Bioelectromagnetics 18:28–35 (1997).

158. Savitz DA, Loomis DP, Tse C-K J. Electrical occupations andneurodegenerative disease: Analysis of U.S. mortality data.Arch Environ Health 53:1–3 (1998).

159. Savitz DA, Checkoway H, Loomis DP. Magnetic field exposureand neurodegenerative disease mortality among electric utilityworkers. Epidemiology 9:398–404 (1998).

160. Johansen C, Olsen J. Mortality from amyotrophic lateral sclero-sis, other chronic disorders, and electric shocks among utilityworkers. Am J Epidemiol 148:362–368 (1998).

161. Sobel E, Davanipour Z, Sulkava R, Erkinjuntti T, Wikström J,Henderson VW, Guckwalter G, Bowman JD, Lee P-J.Occupations with exposure to electromagnetic fields: A possi-ble risk factor for Alzheimer’s disease. Am J Epidemiol142:515–523 (1995).

162. Sobel E, Dunn MS, Davanipour Z, Qian Z, Chui HC. Elevated riskof Alzheimer’s disease among workers with likely electromag-netic field exposure. Neurology 47:1477–1481 (1996).

163. Feychting M, Pedersen NL, Svedberg P, Floderus B, Gatz M.

Dementia and occupational exposure to magnetic fields. ScandJ Work. Environ Health 24:46–53 (1998).

164. Reichmanis M, Perry FS, Marino AA, Becker RO. Relationbetween suicide and the electromagnetic field of overheadpower lines. Physiol Chem Phys 11:395–403 (1979).

165. Perry FS, Reichmanis M, Marino AA. Environmental power fre-quency magnetic fields and suicide. Health Physics 41:267–277(1981).

166. McDowall ME. Mortality of persons resident in the vicinity ofelectric transmission facilities. Br J Cancer 53:271–279 (1986).

167. Baris D, Armstrong B. Suicide among electric utility workers inEngland and Wales [Letter]. Br J Ind Med 47:788–792 (1990).

168. Baris D, Armstrong BG, Deadman J, Theriault G. A case cohortstudy of suicide in relation to exposure to electric and magneticfields among electrical utility workers. Occup Environ Med53:17–24 (1996).

169. Baris D, Armstrong BG, Deadman J, Theriault G. A mortalitystudy of electrical utility workers in Quebec. Occup Environ Med53(suppl 1):25–31 (1996).

170. van Wijngaarden E, Savitz DA, Kleckner RC, Cai J, Loomis D.Exposure to electromagnetic fields and suicide among electricutility workers: a nested case-control study. West J Med173(suppl 2):94–100 (1999).

171. Dowson DI, Lewith GT. Overhead high voltage cables and recur-rent headache and depression. Practitioner 232:22 (1988).

172. Perry FS, Pearl L, Binns R. Power frequency magnetic field:depressive illness and myocardial infarction. Public Health103:177–180 (1989).

173. Poole C, Kavet R, Funch DP, Donelan K, Charry JM, Dreyer N.Depressive symptoms and headaches in relation to proximity ofresidence to an alternating-current transmission line right-of-way. Am J Epidemiol 137:328–330 (1993).

174. Savitz DA, Boyle CA, Holmgreen P. Prevalence of depressionamong electrical workers. Am J Ind Med 25:165–176 (1994).

175. McMahan S, Ericson J, Meyer J. Depressive symptomatology inwomen and residential proximity to high voltage transmissionlines. Am J Epidemiol 139:58–63 (1994).

176. Verkasalo PK, Kaprio J, Varjonen J, Romanov K, Heikkilä,Koskenvuo M. Magnetic fields of transmission lines anddepression. Am J Epidemiol 146:1037–1045 (1997).

177. Baroncelli P, Battisti S, Checcucci A, Comba P, Grandolfo M,Serio A, Vecchia P. A health examination of railway high-volt-age substation workers exposed to ELF electromagnetic fields.Am J Ind Med 10:45–55 (1986).

178. Sastre A, Graham C, Cook MR. Brain frequency magnetic fieldsalter cardiac autonomic control mechanisms. Neurophysiol Clin111:1942–1948 (2000).

179. Jauchem JR. Exposure to extremely-low-frequency electromag-netic fields and radiofrequency radiation: cardiovascular effectsin humans. Int Arch Occup Environ Health 170:9–21 (1997).

180. Savitz DA, Liao D, Sastre A, Kleckner RC, Kavet R. Magneticfield exposure and cardiovascular disease mortality and electricutility workers. Am J Epidemiol 149(suppl 2):135–142 (1999).

181. Sastre A. Cook MR. Graham C. Nocturnal exposure to intermit-tent 60 Hz magnetic fields alters human cardiac rhythm.Bioelectromagnetics 19(2):98–106 (1998).

182. Tsuji H, Larson MG, Venditti FJ Jr, Manders ES, Evans JC,Feldman CL, Levy D. Impact of reduced heart rate variability onrisk for cardiac events. The Framingham Heart Study.Circulation 94(suppl 11): 2850–2855 (1996).

183. Liao D, Cai J, Rosamond WD, Barnes RW, Hutchinson RG,Whitsel EA, Rautaharju P, Heiss G. Cardiac autonomic functionand incident coronary heart disease: A population-base case-cohort study. The ARIC study. Am J Epidemiol 145:696–706(1997).

184. Dekker JM, Schouten EG, Klootwijk P, Pool J, Swenne CA,Kromhout D. Heart rate variability from short electrocardio-graphic recordings predicts mortality from all causes in middleaged and elderly men. The Zutphen study. Am J Epidemiol145(10):899–908 (1997).

185. Akselrod S, Gordon D, Ubel FA, Shannon DC, Berger AC, CohenRJ. Power spectrum analysis of heart rate fluctuation: a quanti-tative probe of beat-to-beat cardiovascular control. Science213:220–222 (1981).

186. Willich SN, Maclure M, Mittleman M, Arntz HR, Muller JE.Sudden cardiac death. Support for a role of triggering in causa-tion. Circulation 87(suppl 5):1442–1450 (1993).

187. Finkelstein MM. Magnetic field exposure and cardiovasculardisease mortality among electric utility workers. Am JEpidemiol 150:1258–1259 (1999).

188. Graham C, Sastre A, Cook MR, Kavet R. Exposure to strong ELFfields does not alter cardiac autonomic control mechanisms.Bioelectromagnetics 21(suppl 6):413–412 (2000).

189. Ahlbom A, Feychting M. Unpublished data.

EMF and health

190. Delgado JMR, Leal J, Monteagudo JL, Gracia MG.Embryological changes induced by weak, extremely low-fre-quency electromagnetic fields. J Anat 134:187–220 (1982).

191. Ubeda A, Trillo MA, Chacon L, Blanco MJ, Leal J. Chick embryodevelopment can be irreversibly altered by early exposure toweak extremely-low-frequency magnetic fields. Bioelectro-magnetics 15(suppl 5):385–398 (1994).

192. Robert E. Intrauterine effects of electromagnetic fields (low fre-quency, mid frequency RF and microwave): review of epidemio-logic studies. Teratology 59:292–298 (1999).

193. Wertheimer N, Leeper E. Possible effects of electric blanketsand heated waterbeds on fetal development. Bioelectro-magnetics 7:13–22 (1986).

194. Shaw GM, Croen LA. Human adverse reproductive outcomesand electromagnetic field exposures: review of epidemiologicstudies. Environ Health Perspect 101(suppl 4):107–119 (1993).

195. Juutilainen J, Matilainen P, Saarikoski S, Laara E, Suonio S.Early pregnancy loss and exposure to 50-Hz magnetic fields.Bioelectromagnetics 14:229–236 (1993).

196. Savitz DA, Ananth CV. Residential magnetic fields, wire codesand pregnancy outcomes. Bioelectromagnetics 15:271–273(1994).

197. Belanger K, Leaderer B, Hellenbrand K, Holford TR, McSharry J,Power ME, Bracken MB. Spontaneous abortion and exposure toelectric blankets and heated water beds. Epidemiology 9:36–42(1998).

198. Robert E, Harris JA, Robert O, Selvin S. Case-control studymaternal residential proximity to high voltage power lines con-genital malformation in France. Paediatr Perinat Epidemiol10:32–38 (1996).

199. Wertheimer N, Leeper E. Fetal loss associated with two sea-sonal sources of electro-magnetic field exposure. Am JEpidemiol 129:220–224 (1989).

200. Chernoff N, Rogers JM, Kavet R. A review of the literature onpotential reproductive and developmental toxicity of electricand magnetic fields. Toxicology 74:91–126 (1992).

201. Dlugosz L, Vena J, Byers T, Sever L, Bracken M, Marshall E.Congenital defects and electric bed heating in New York State:a register-base-case-cohort study. Am J Epidemiol135:1000–1011 (1992).

202. Li D-K, Checkoway H, Mueller BA. Electric blanket use duringpregnancy in relation to the risk of congenital urinary tractanomalies among women with a history of subfertility.Epidemiology 6(suppl 5):485–489 (1995).

203. Shaw GM, Nelson V, Todoroff K, Wasserman CR, Neutra RR.Maternal periconceptional use of electric bed-heating devicesand risk for neural tube defects and orofacial clefts. Teratology60:124–129 (1999).

204. Bracken MB, Belanger K, Hellenbrand K, Dlugosz L, Holford TR,McSharry JE, Addesso K, Leaderer B. Exposure to electromag-netic fields during pregnancy with emphasis on electricallyheated beds: association with birthweight and intrauterinegrowth retardation. Epidemiology 6(suppl 3):263–270 (1995).

205. Lee GM, Neutra RR, Hristova L, Yost M, Hiatt RA. The use ofelectric bed heaters and the risk of clinically recognized sponta-neous abortion. Epidemiology 11(suppl 4):406–415 (2000).

206. Schnorr TM, Grajewski BA, Hornung RW, Thun MJ, EgelandGM, Murray WE, Conover DL, Halperin WE. Video display termi-nals and the risk of spontaneous abortion. N Engl J Med324:727–733 (1991).

207. Delpizzo V. Epidemiological studies of work with video displayterminals and adverse pregnancy outcomes (1984–1992). Am JInd Med 26(4):465–480 (1994).

208. Huuskonen H, Lindbohm ML, Juutilainen J. Teratogenic andreproductive effects of low-frequency magnetic fields. MutatRes 410:167–183 (1998).

209. Mezei G, Kheifets LI, Nelson LM, Mills KM, Iriaye R, Kelsey JL.Household appliance use and residential exposures to 60-Hzmagnetic fields. J Expo Anal Environ Epidemiol 11(1):41–49(2001).

210. Lindbohm ML, Hietanen M, Kyyronen P, Sallmen M, vonNandelstadh P, Taskinen H, Pekkarinen M, Ylikoski M,

Hemminki K. Magnetic fields of video display terminals andspontaneous abortion. Am J Epidemiol 136(9):1041–1051(1992).

211. Rephacholi MH, Greenebaum B. Interaction of static andextremely low frequency electric and magnetic fields with livingsystems: health effects and research needs. Bioelectro-magnetics 20:133–160 (1999).

212. Marcus M, McChesney R, Golden A, Landrigan P. Video displayterminals and miscarriage. J Am Med Women Assoc 55(suppl2):84–88, 105 (2000).

213. Grajewski B, Schnorr TM, Reefhuis J, Roeleveld N, Salvan A,Mueller CA, Conover DL, Murray WE. Work with video displayterminals and the risk of reduced birthweight and preterm birth.Am J Ind Med 32(suppl 6):681–688 (1997).

214. Nordstrom S, Birke E, Gustavsson L. Reproductive hazardsamong workers at high voltage substations. Bioelectro-magnetics 4:91–101 (1983).

215. Tornqvist S. Paternal work in the power industry: effects onchildren at delivery. J Occup Environ Med 40(suppl 2):111–117(1998).

216. Brent RL. Reproductive and teratologic effects of low-frequencyelectromagnetic fields: a review of in vivo and in vitro studiesusing animal models. Teratology 59:261–286 (1999).

217. Hjollund NH, Skotte JH, Kolstad HA. Bonde JP. Extremely lowfrequency magnetic fields and fertility: a follow up study of cou-ples planning first pregnancies. Occup Environ Med 56:253–255(1999).

218. Irgens A, Kruger K, Skorve AH, Irgens LM. Male proportion inoffspring of parents exposed to strong static and extremelylow-frequency in electromagnetic fields in Norway. Am J IndMed 32(5):557–561 (1997).

219. Hatch M. The epidemiology of electric and magnetic field expo-sures in the power frequency range and reproductive outcomes.Paediatr Perinatal Epidemiol 6:198–214 (1992).


icnirp sci review · icnirp sci review review of the ... ubiquitous part of modern life, and there...

Documents