4.5 Epidemiological studies of non-cancer health effects in humans

4.5.1 Occupational exposure

This review includes only the studies that are the strongest methodologically, on the basis of exposure assessment, on occupational exposure to EMF and neurodegenerative and psychological diseases, reproductive disorders, and cardiovascular disease (see Table 4.35).

Table 4.35. Minimal exposure assessment required for a study to be included in this review

4.5.1.1 Reproductive effects

(a) Adverse birth outcomes and maternal exposure to electromagnetic fields

Schnorr et al. (Schnorr et al., 1991) conducted a retrospective study of spontaneous abortions in women employed as telephone operators by two large telephone companies in 1983-86. The study base consisted of 4246 married women aged 18-33 who were identified from company records and agreed to an interview about their reproductive, medical, and occupational histories (76.6% of the 5544 woman identified). The 136 cases were pregnancies resulting in spontaneous abortions while the respondent worked as a telephone operator. The 737 controls were pregnancies among operators who had live births. Exposure to EMF was assessed by job title and validated by field measurements. Women considered to be exposed were directory assistance operators who had always used conventional VDTs with cathode ray tubes during the study period. Those considered to be unexposed were general telephone operators who had worked at monitors with light-emitting diode or neon glow tube displays up to 1986; these monitors emitted no VLF fields and lower ELF magnetic fields than the VDTs (Bracken, 1991; NIOSH, 1990). The EMF fields from the VDTs had rough saw tooth waveforms with principal frequencies of 45 Hz for one model and 60 Hz for another; the non-VDT monitors had 60 Hz sinusoidal waveforms. The exposure of the operators to EMF at the level of their abdomen varied considerably with the worker's location and the type of monitor. In measurements taken at a sample of work sites with Holiday Industries model HI 3600-01 and 3600-02 meters, the geometric mean operator exposure to ELF magnetic fields was 0.078 µT for one model and 0.073 µT for the other. The exposure from VDTs was the same as from the light-emitting diode monitors (geometric mean = 0.078 µT) and higher on average than that from the neon glow tube monitors (0.041 µT). In comparison with those in other office environments, this ELF exposure of the telephone operators is not elevated, although VLF fields were rare before the introduction of computers.

The spontaneous abortion rate was similar for the VDT-exposed and the general telephone operators (14.8 and 15.9%, respectively). No associations were found with VDT use at any period of pregnancy, either when the analyses were unadjusted or when they were adjusted for such risk factors as smoking, thyroid disorders, alcohol consumption, and spontaneous abortions before the study period. [As the comparison of VDTs with non-VDT monitors produced a clear differential in exposure to VLF (but not ELF) EMF, this study does not allow any conclusions about the risks for spontaneous abortion from exposure to ELF magnetic fields. Derivation of information on spontaneous abortions from questionnaires has limitations.]

Grajewski et al. (Grajewski et al., 1997) examined the risk for preterm births and reduced birthweights among VDT operators from the study population of Schnorr et al. (Schnorr et al., 1991). They evaluated 304 singleton pregnancies in 284 women who had worked as VDT telephone operators during the first 28 weeks of pregnancy and 403 among 363 unexposed operators. Birthweight, gestational age, birth defects, and the reproductive risk factors studied by Schnorr et al. were determined during interviews. The self-reported data were validated from birth certificates and medical records and found to be within ± 100 g for birth weight and ± two weeks for gestational age for more than 80% of the subjects. The number of hours that the subject had been exposed to VDTs was estimated from the time she had spent working as a directory assistance operator in the telephone company's payroll records, which also reported the number of hours of work per day, leave, and vacations. The following potential confounders were included in the analysis: diabetes, maternal weight gain during pregnancy, thyroid conditions, hypertension medications, pre-eclampsia, toxemia, interpregnancy interval, and adverse pregnancy outcomes (including previous reduced-weight or preterm births) before the study period.

No association was found between length of work with VDTs at any period of the pregnancy and either reduced birth weight (< 2800 g) or preterm birth (21-37 weeks gestation). Seven cases of major birth defects (2.3%) were reported by the VDT-exposed operators and four (1.0%) by the unexposed; however, only three of the 11 self-reported birth defects could be confirmed from medical records. [See summary of Schnorr et al. for comments on limitations of the study. Some validation of outcome was possible from medical records.]

Lindbohm et al. (Lindbohm et al., 1992) performed a retrospective study of spontaneous abortions among women using VDTs. The study base consisted of women employed as bank clerks and clerical workers by three large companies during the years 1975-85. The 368 cases of spontaneous abortion among study base members aged 20-35 identified from the Finnish pregnancy registry had to be confirmed by questionnaire for the subject to be included in the analysis (8.5% of miscarriages unconfirmed). The 1069 controls were woman who had given birth to liveborn infants in the registry. Job histories, VDT use, reproductive risk factors, and ergonomic factors were requested on the questionnaires, which were returned by 191 cases (52%) and 394 controls (36%). Exposure to VDT fields was assessed from the questionnaires, company records on the VDT models used by different work groups, and laboratory measurements of VLF and ELF magnetic fields at a fixed location (approximating the fetus' position) near the various VDT models. Exposure to magnetic fields was expressed as the time derivative (dB/dt) for VLF fields and the peak-to-peak magnitude for ELF fields. [In order to compare peak-to-peak values in this study with the root-square-squared (rms) magnitudes used in other studies, a range of conversion factors (0.266-0.010) was derived empirically from the Finnish VDT measurements reported by Jokela et al. Because these are saw tooth waveforms, the empirical conversion factor is less than the better-known value: 1/22 = 0.354 derived for sinusoidal waves.] (Jokela et al., 1989)

An increased risk was reported in association with exposure to ELF magnetic fields from VDTs, expressed both as rms magnitude ³ 0.24 µT and cumulative exposure per week; the increase was significant for the highest exposure category (see Table 4.36). [There appeared to be exposure-response relationships, although they were not tested statistically]. The relative risk for high exposure was markedly greater for late spontaneous abortions, ³ 12 weeks into the pregnancy (8 cases; OR = 9.5; 95% CI = 1.8-52). No association was seen with VLF VDT fields, but a significant association was seen with the time-weighted average exposure per week (which incorporated time spent at VDTs, calculated from company records) in the group with high exposure (2.7; 1.2-6.1). The relative risks were adjusted for self-reported work load, exposure to solvents, number of previous births, use of intrauterine devices, and frequency of breakdowns in automatic data processing machines. [The response rate was very poor, particularly among controls, and subjects for whom all VDT models could not be identified were excluded. The magnetic fields were measured in the laboratory, with high background ELF fields, rather than in workplaces.]

Bracken et al. (Bracken et al., 1995a) performed a prospective cohort study of fetal growth and exposure to EMF (see section 4.5.2.1 for a full description of the study). Although the focus was residential sources of EMF, such as electrically heated beds and wire codes, occupational exposures were also assessed by seven-day monitoring of magnetic fields and questions about VDT use. The study population consisted of women with a uterine pregnancy and no diabetes who had received prenatal care in the New Haven, Connecticut, area over a four-year period. The participating subjects (83% of eligible women) were interviewed before 16 weeks of estimated gestational age and took part in a program of EMF measurements that included the AMEX-2 wrist monitor. The monitoring encompassed much occupational exposure, because 81.3% of the subjects worked outside the home. The outcomes were low birthweight (< 2500 g) and intrauterine growth retardation (< 10th percentile of weight for gestational age). Neither outcome was associated with VDT use or TWA exposure to magnetic fields. [The AMEX-2 has a magnetic field sensor in only one direction and therefore provides very inaccurate measures in comparison with three-axis monitors. See Section 4.5.2.1 for comments on the limitations of this study.]

(b) Adverse reproductive effects from paternal exposure to EMF

Tornqvist (Tornqvist, 1998) studied reproductive effects among electric workers in the utility industry, in a large retrospective study and a small prospective study of perinatal survival, multiple births, low birthweight, sex ratio, congenital malformations, and childhood cancers (also discussed in section 4.2.1). In the retrospective study, Swedish census data were used to establish a cohort of 2077 male electrical workers and to obtain information on their work histories and spouses or cohabitants for a 25-year period. Information on 3350 children born to these couples was obtained from the Medical Birth Registry, and any cancers that the children incurred were ascertained from the National Cancer Registry. For each child, the father's exposure to EMF was determined from job titles reported in the five-year censuses before and after the year of the birth. The father was considered to have been exposed if he had held electrical occupations at both censuses spanning the prenatal year; otherwise, he was considered to be unexposed. Since the unexposed fathers held electrical occupations either before or after the child was conceived, they are assumed to belong to a socioeconomic class similar to that of the exposed fathers. Relative risks were calculated for the various reproductive outcomes relative to those in the total population of births, adjusted for maternal age, parity, and year of birth. Relative risks were also calculated for the exposed vs. the unexposed fathers.

In the prospective study, the health of a cohort of 460 young electrical workers whose first job was in the power industry was followed. Questionnaires covering job histories and reproductive issues were part of the health examinations administered every three years. Over a 10-year period, these men fathered 364 children. Exposure to EMF during the prenatal year were derived from the job histories and a JEM based on measurements taken in the Swedish power industries (Lindh et al., 1997). Workers considered to be highly exposed were those with exposure to both TWA magnetic fields > 0.5 µT and electric fields > 30 V/m for at least 10 min per shift; low exposure was defined as exposure to both electric and magnetic fields below these thresholds.

Neither study showed any significant association with multiple births, low birth weight, sex ratio, or congenital malformations. The proportion of malformations was slightly increased among the offspring of highly exposed fathers in both the retrospective study (OR = 1.1; 95% CI, 0.82-1.5) and the prospective study (1.6; 0.43-14.8). [The paper gives 1.48 as the upper confidence limit, which must be a typographical error.] The author noted that the retrospective study involved large numbers and crude information on exposure, while the prospective study involved small numbers but good information on exposure.

4.5.1.2 Neurodegenerative diseases

Alzheimer disease is a progressive, irreversible degenerative disease of the brain that presents as dementia, usually in people over the age of 65. The diagnosis includes symptoms of dementia (e.g. loss of memory and mental and cognitive functions) and excludes other causes such as Parkinson disease, head trauma, alcoholism, and stroke (vascular dementia). Alzheimer disease is biologically distinct from other dementias, but its characteristic form of brain degeneration is seen only at autopsy. Alzheimer disease is the commonest form of dementia: e.g. of the 77 cases of dementia studied by Feychting et al. (Feychting et al., 1998), 71.4% were diagnosed as Alzheimer disease, 15.6% as vascular dementia, 10.4% as other forms, and 2.6% as unspecified dementia. Two hypotheses have been advanced with regard to the effect of EMF on dementia. Sobel and Davanipour (Sobel & Davanipour, 1996) proposed that EMF contributes to the neurodegenerative processes behind Alzheimer disease, while Feychting et al. (Feychting et al., 1998) hypothesized that the brain damage caused by EMF could increase the risks for all forms of dementia. EMF as a risk factor for Alzheimer disease or all dementias has been the subject of six studies (Table 4.37).

Sobel et al. (Sobel et al., 1995) used three case-control sets to study 'sporadic' Alzheimer disease, i.e., patients who had no first-degree relatives with the disease. The cases were diagnosed at hospitals and university clinics in Finland (53 in 1982-85 and 198 in 1977-78) and southern California (136 in 1982-83). For the older Finnish series, the diagnostic criteria were different than those used in California and for the other Finnish series (those of the National Institute of Neurologic and Communicative Disorders and Storke-Alzheimer's Disease and Related Disorders Association; NINCDS-ADRDA), and the agreement between the clinical diagnosis and autopsy results in a sample of cases was only 82%, in comparison with 90-98% in the other series. In the first Finnish series and the California series cases with familial disease were eliminated by questioning next-of-kin about Alzheimer disease among relatives. The controls were patients with sporadic vascular dementia in the first Finnish series, hospital patients with not neurological disease in the second series, and neighborhood volunteers who were neurologically normal in the California series. In formation on the primary lifetime occupation was obtained by interviewing a proxy respondent for demented subjects and by direct interview for healthy controls. High, medium, or low exposure to ELF magnetic fields was attributed to each job by an industrial hygienist. Since the commonest occupation with medium-to-high exposure in this population were seamstresses and tailors, magnetic fields were measured around a sample of industrial and home sewing machines; the average spot measurement was 1.93 µT. [To judge the quality of these exposure assessment, the occupation judged to involve medium-to-high exposure was compared in Table 2.4 with measured TWAs; 87% are above the 540th percentile for men, excluding occupations for which these were no measurements.]

A significant association was found in women with exposure to medium-to-high magnetic fields (OR = 3.9; 95% CI, 1.7-8.9) and for the men and women combined (2.9; 1.6-5.4). Despite the different composition of the three case-control series, elevated risks were found consistently, except among men in the series in which people with vascular dementia were used as controls (0.7; 0.1-8.9) with one exposed case). [The limitations of the study are use of different control groups in the three series, particularly patients with vascular dementia who may in fact have had Alzheimer disease; obtaining job histories by questionnaires; lack of validation of exposure of the study population; dependence on proxy respondents for job histories for cases but not for some of the control; and failure to take into account potential confounding by heredity. These limitations make interpretation of the results problematic.]

Sobel et al. (Sobel et al., 1996) used a similar study design for a new case-control series drawn from an Alzheimer disease clinic in California. The cases were dementia or probably Alzheimer disease diagnoses in 86 men and 240 women at the clinic [period unspecified] by the criteria of the NINCDS-ADRDA. The controls were 76 male and 76 female patients at the clinic in whom other dementias (excluding vascular dementia and mixed diagnoses) were diagnosed. In a departure from the method of Sobel et al. (Sobel et al., 1995), subjects under the age of 65 were excluded from the analysis, but those reporting a family history of Alzheimer disease were included because the reports could not be verified. Information on primary occupation and education was available from clinical records. Work with sewing machines was again prevalent among cases. High, medium, or low exposure to ELF magnetic fields were attributed to each of these jobs following the classification used in the previous study. The high and medium exposure categories were combined into a single category for the analyses. Odds ratios were adjusted for education (an apparent risk factor), gender, and age of onset (possible confounders) by logistic regression. An association with magnetic fields was noted, which was significant in men, but not in women: the odds ratios were, respectively, 4.9 (1.3-7.9) and 3.4 (0.76-16) for men and women. [The odds ratios in this study were higher for men than for women, contrary to what had been observed in the previous study. The limitations of this study were that the cases included 24 patients with unclear diagnoses; the controls were not matched by age or gender to the cases and were from the same clinic, which specialized in Alzheimer disease; job histories were obtained by questionnaire; and the exposure of the study population was not validated. The different designs used in this study and in the three other studies of Sobel et al. lead to a diverse collection of relative risks and potential biases that make interpretation of these results very difficult.]

Feychting et al. (Feychting et al., 1998) examined exposure to magnetic fields in the population of a study of dementia in Swedish twins (Gatz et al., 1997) in which twins in the Swedish registry were screened for dementia by a complete clinical work-up including neurological and neuropsychological assessments and neuro-imaging (81% response rate); the diagnoses followed the criteria of the NINCDS-ADRDA but were not confirmed by autopsy. All 77 cases in this study were dementia, and 71% (55) were Alzheimer disease. When both twins were demented, one of them was selected randomly to be a case in the study. Controls were twins found to be free of dementia at clinical examination. To obtain enough older controls, 14% of the control population was recruited from a twin registry population different than that from which the cases were derived. Two independent control groups of 228 and 238 subjects, respectively, were formed so that sibling twins could be assigned to different groups, eliminating double representation of the same genes among the controls. Complete histories of jobs and exposure in the workplace were obtained by structured interviews with controls and people who knew the case patients. For the job held the longest and the job held last, TWA magnetic field magnitudes were assigned from a JEM based on full-shift TWA magnetic field magnitudes were assigned from a JEM based on full-shift measurements (Floderus et al., 1994). Risk estimates were adjusted for age at onset, education, and birth date.

A significant association was found between all dementias and exposure to TWA magnetic fields > 0.2 µT in the job held last (OR, 1.6; 95% CI, 0.6-4.0 for TWA 0.12-0.19 µT and 3.3; 1.3-8.6 for > 0.20 µT, compared with < 0.12 µT). The association was smaller and not significant when the primary occupation was considered. The association between Alzheimer disease and exposure to TWA magnetic fields > 0.2 µT in the job held last was also increased but not significantly so (2.4; 0.8-6.9). When the disease was diagnosed in the patient before the age of 75, significant relative risks were found between exposure to TWA magnetic fields > 0.2 µT and all forms of dementia combined (5.8; 1.4-24) and with Alzheimer disease alone (5.0; 1.1-21). The risks reportedly increased further when fewer than 10 years had elapsed between the last job and disease onset. The relative risks for a small number of non-Alzheimer dementia cases are also elevated. [The limitations are the small number of cases, particularly of Alzheimer disease; possible selection bias due to twins who refused to be examined; potential information biases in the job histories, which were obtained for cases from proxy respondents; lack of autopsy confirmation of the diagnosis of Alzheimer disease; and failure to evaluate potential confounding by heredity.]

Johansen and Olsen conducted a cohort mortality study of neurological disorders among electrical utility workers, using a JEM based on expert judgments from magnetic field measurements (Johansen, 1998). The study design is similar to that of their cancer cohort study, reviewed in section 4.2.1.1 (Johansen & Olsen, 1998). Only four deaths due to senile dementia were found, and the SMRs were not elevated at any level of exposure to magnetic fields. [The major limitations of this study are the use of death certificates to assess dementia, which is often not the immediate cause of death; failure to take into account potential confounding by heredity; and the very small number of cases.]

Savitz et al. also studied mortality from neurodegenerative disease and exposure to magnetic fields in the cohort of men working at five US electric utilities (described in section 4.2.1.1) (Savitz et al., 1998a). The cohort was followed from mortality in 1950-88. These were 24 cases of Alzheimer disease, 45 of Parkinson disease, and 28 of amyotrophic lateral sclerosis (ALS) listed on death certificates as either the underlying or secondary cause of death. The magnitudes of cumulative exposures to TWA magnetic fields were calculated for five durations (total career, ³ 2, 2-10, 10-20, and ³ 20 years) from job histories and a JEM based on full-shift measurements taken for the previous study. Mortality rate ratios relative to the cohort were computed of the five exposure categories by Poisson regression with adjustment for age, decade of death, race, social class, retirement status, and exposure to solvents.

For Alzheimer disease, a non-significant association with career exposure was noted, with relative risks of 1.4 (0.4-5.3) for 0.54-1.13 µT-years, 1.8 (0.5-6.7) for 1.13-2.06 µT-years, and 2.0 (0.6-7.0) for 2.06-4.75 µT-years. The risk was highest for the highest exposure category (2.7; 0.8-8.9) for exposures received ³ 20 years previously. No increase in risk or exposure-response relationship was apparent for Parkinson disease. [The major limitations of this study are the use of death certificate to assess outcome, particularly since Alzheimer disease is difficult to diagnose and is often underreported on death certificates; the failure to take into account potential confounding by heredity; and the very small number of cases.]

Savitz et al. considered mortality rates from neurodegenerative disease among working-age men in 25 states in the USA, using death certificates from National Center for Health Statistics (Savitz et al., 1998b). Over the period 1985-91, 1,768,411 death certificates met the inclusion criteria. The underlying cause of death was Alzheimer disease for 256 subjects, Parkinson disease for 168, and ALS for 114. For each case, three controls who had diet of other causes were selected by frequency matching on cases' age and year of death. A subject was considered to have been exposed if the job on his death certificate was one of 14 electrical occupations. [As shown in Table 2.4, 71% of these electrical occupations involve exposure to TWA magnetic fields above the 50th percentile for male workers.] Internally standardized mortality odds ratios for the electrical occupations were calculated by logistic regression with adjustment for age, year of death, social class, and race.

The relative risk for Alzheimer disease for men in all electrical occupations was weak but significantly elevated (1.2; 1.0-1.4), mainly due to an excess of deaths in men over 65 (1.2; .1-1.4). No association between electrical occupations and Parkinson disease was observed (1.1; 0.9-1.2). [As in the two previous studies, the use of death certificates for diagnosis of Alzheimer disease is a limitation as is failure to take into account confounding by heredity and the absence of validation of exposure in the study population; see also section 4.2.1.1 for comments on limitations of this study.]

(b) Amyotrophic lateral sclerosis and other motor neuron diseases.

Deapen and Henderson (Deapen & Henderson, 1986) conducted a case-control study of ALS with an a priori hypothesis that the cause is electrical shocks. Their 678 cases had voluntarily registered with the Amyotrophic Lateral Sclerosis Society in 1977-79 and had filled out a mailed questionnaire. The names of the 518 controls of the same gender and age were solicited from the patients [no criteria given]. All 518 case-control pairs completed questionnaires about job histories, genetic factors, electric shocks resulting in unconsciousness, and other exposure. Nineteen cases and five controls reported electrical occupations. Odds ratios were determined by a matched analysis for all risk factors, including a group of 19 electrically related occupation-industry combinations. [As indicated in Table 2.4, 71% of the electrical occupations in this paper involve exposure to TWA magnetic fields above the 50th percentile for male workers.] The relative risks were significant for both electrical occupations (3.8; 1.4-13) and shocks (2.8; 1.0-9.9), but the number of exposed cases was small. [Limitations of this study are that exposure to EMF was assessed from job titles based on responses to the questionnaire, failure to report the criteria for control selection, and the potential recall bias inherent in using occupational histories and reports of electric shock.]

Davanipour et al. (Davanipour et al., 1997) conducted a case-control study involving 28 patients at an ALS clinic. The cases were diagnosed ALS with no history of polio or ALS apparent in the family. The controls were one blood relative and one non-blood relative (except spouses), if they existed and were willing to participate, matched on age and gender when possible. In total, 28 controls were recruited, but eight cases had no control. Occupational histories were obtained by interviewing the subjects. For each job, exposure to magnetic fields was judged to be high, medium, or low by an industrial hygienist with experience in measuring EMF but blinded to the disease status of the person. [As shown in table 2.4, 75% of the occupations judged to involved medium-to-high exposure were those in which the TWA magnetic fields were above the 50th percentile for male workers.] Scores for cumulative exposure to magnetic fields were calculated for each subject's working life. In the highest exposure category, the odds ratios adjusted for gender were 2.5 (0.80-6.6) for all subjects and 7.5 (1.4-38) for subjects with ³ 20 years' work experience. Logistic regression analysis revealed a statistically significant exposure-response relationship for workers with ³ 20 years' experience (p < 0.02) but not for all subjects. [The study is limited by the small sample size and potential control selection bias.]

Johansen and Olsen (Johansen & Olsen, 1998) conducted a cohort study of mortality among electrical utility workers, using a JEM based on expert judgment from magnetic field measurements. Details of the study design are described in section 4.2.1.2. There were 14 deaths from ALS. The risks for this condition appeared to increase across exposure categories (SMR = 2.8 for TWA magnetic fields > 1.0 µT) but are too imprecise to be significant. The authors attributed the association between ALS and exposure to magnetic fields to confounding by non-lethal electric shocks. Although they had no data on exposure to shocks to test their hypothesis of confounding, they noted that deaths from electrical accidents correlate with exposure to magnetic fields (SMR = 0 for background vs. SMR = 31 for fields > 1.0 µT; p < 0.05). [The very small number of ALS cases and the absence of confidence intervals make these results difficult to interpret; diagnosis of ALS from death certificates has a 5-10% error rate; and family histories of ALS were not assessed.]

Savitz et al. included ALS in their cohort study of mortality from neurodegenerative disease and exposure to magnetic fields (see above) (Savitz et al., 1998a). No association was seen between total career exposure and mention of ALS as a cause of death (28 cases), but a non-significant increase was seen for exposure in the highest category for ³ 20 years (RR = 2.4; 95% CI, 0.7-8.0). [Limitations of this study are the modest number of ALS cases, diagnosis from death certificates, and the absence of data on electric shocks or the family's disease history.]

Savitz et al. examined ALS in their study of mortality from neurodegenerative diseases in US working-age men (see above) (Savitz et al., 1998b). There were 114 deaths from ALS. The mortality odds ratio for this disease category was slightly elevated for electrical occupations (1.3; 1.1-1.6). [The diagnosis of ALS from death certificates in this study was based on ICD9, which groups ALS with other motor neuron diseases. Other limitations of this study include the fact that only one occupation was taken from death certificates and the absence of data on important confounders, such as familial neurodegenerative disease or exposure to electric shocks.]

4.5.1.3 Suicide and depression

Savitz et al. studied depression and exposure to EMF in a cohort of US Army veterans who had been examined in the Vietnam Experience Study (Savitz et al., 1994). In this study, a sample of male veterans whose service had started in 1965-71 and lasted for a single term underwent medical and psychological examinations (52% participation rate). The population of the study of EMF was 4044 men who were employed at the time of the examination. They were asked in an interview about the job held currently, the job held the longest and its duration, demographic information, and data on potential confounders. At the time of the examination, the 183 exposed subjects had a job belonging to a group of 15 electrical occupations (Table 2.4). Depression was assessed on the basis of the Diagnostic Interview Schedule and the Minnesota Multiphasic Personality Inventory. For this study, the results of the former were used to test for three diagnoses of depression and eight symptoms; and the latter gave eight indicators of depression and five symptoms. The risk ratios for these outcomes were adjusted for race, marital status, education, alcohol use, and duration of employment.

Subjects who had held their most recent job for less than 10 years were generally more likely to be depressed. Some non-significant associations were found between electric work and markers of depression (see Table 4.39). Neither electricians nor electrical workers had an increased frequency in thoughts of death (OR = 0.8; 95% CI, 0.3-1.8). [Some of the elevated risks could be due to chance, given the many outcomes analyzed. Other problems are the very small number of exposed cases, exposure assessment based on job title, and the difficulty in generalizing results for Army veterans, especially with the low participation rate.]

Baris et al .(Baris et al., 1996a) examined suicide in a case-cohort study of the Hydro Québec workers who had been studied by Thériault et al. Deaths in the cohort were tracked over 18 years through company records for current employees, pension records for retirees, and Québec death certificates for former employees. Forty-nine suicides were found on the death certificates. The reference group was a 1% sample of the entire cohort (n = 217). Exposures were assessed from JEMs for electric fields, magnetic fields, and PEMF as described above (Armstrong et al., 1994; Thériault et al., 1994). The TWA and the GM over the shift were used as exposure metrics for EMF, while the TWA was the only metric for PEMF. For each employee, both cumulative and current exposures were calculated from the company's job records and JEMs for these five metrics. The company's records from periods before employment of the subject and annual medical examinations was examined for information on alcohol consumption and changes in marital status. The cohort's sick-leave records were scanned for any diagnoses of mental illness, which was a strong risk factor for suicide in this population (RR = 15; 95% CI, 6.8-34).

The results showed no significant association with cumulative or current TWA exposure to magnetic fields (OR = 1.6; 0.69-3.8 for TWA ³ 0.37 µT). Of the 10 types of exposure analyzed, only cumulative GM electric fields in the intermediate category were associated with a significant suicide rate (3.1; 1.9-8.2), but the risks declined with greater exposure, making a causal relationship less likely. In addition, the association with intermediate electric fields decreased somewhat after adjustment for previous mental disorders (2.8; 0.93-8.1). According to the authors, mental illness could be either an intermediate variable, if depression were caused by EMF, or a confounder, if psychological disorders predispose workers toward jobs with high or low exposure to EMF. The hypothesis that mental illness is an intermediate variable receives some support from the finding of a decrease in suicide risk after adjustment for mental disorders. Support for the hypothesis that it is a confounder is provided by the finding that 70.2% of the 20 cohort members with absences for mental illness had had geometric mean exposure to electric fields above the 50th percentile. The authors noted that their study did not have the complex design needed to distinguish a confounding from an intermediate variable.[For former employees of Hydro Québec, minor errors may have been introduced owing to lack of information on suicides occurring outside the Province. Data on confounders of suicide, such as alcohol use, are difficult to obtain accurately from company records. Moreover, the analysis of 10 exposure measures may have generated associations by chance.].

Johansen and Olsen (Johansen & Olsen, 1998) also examined suicides and exposure to magnetic fields in the cohort study described in section 4.2.1.2. A slight nonsignificant increase in the relative risk for suicide (OR = 1.4) was reported for exposure to TWA magnetic fields estimated to be > 1.0 µT.

4.5.1.4 Cardiovascular disease

For a cohort of Hydro QuÈbec workers included by Th ériault et al. (see section 4.2.1.1) (Thériault et al., 1994), Baris et al. (Baris et al., 1996a) reported a significant decrease in the SMR for death reported on death certificates as due to 'circulatory diseases' among workers exposed to > 1.6 µG when compared with workers with lower exposure (based on 217 deaths). [The grouping 'circulatory disease' might be too broad, and the 1.6 µG cut-point might not include groups with sufficiently high exposure.]

Savitz et al. (Savitz et al., 1998c) reported the results of an analysis of mortality from specific types of cardiovascular disease in their study of five utilities (Savitz & Loomis, 1995) (described in section 4.2.1.1) in relation to exposure to magnetic fields. The investigation was motivated by clinical findings that suggested a decrease in heart rate variability in people exposed to 20 µT 60 Hz magnetic fields. Various categories of causes of death from heart disease, based on the International Classification of Diseases, were combined into four groupings: arrhythmia-related ( 212 deaths), acute myocardial infarct (4238 deaths), atherosclerosis (142 deaths), and chronic coronary heart disease (2210 deaths). These causes accounted for 88% of all cardiovascular deaths identified in the cohort. The groupings were chosen a priori because the underlying pathological lesions of the first two potentially involve disruption of autonomic nervous system control of cardiac function, whereas loss of autonomic nervous system control is not believed to be critical for the second pair. The exposure of a random sample of workers in 28 occupational categories was measured over 2842 full shifts with an AMEX 3D meter. In all of the analyses, adjustment was made for age, calendar year, race, social class, and work status (active versus inactive).

As predicted, no association was found between total exposure to magnetic fields and the risk for atherosclerosis or chronic coronary heart disease (see Table 4.40). The relative risks for myocardial infarct were consistently and statistically significantly increased in each category of total exposure: 1.1 (1.0-1.3) for cumulative exposure of 0.4² 1.2 µT-years, 1.2 (1.1-1.3) for 1.2² 2.0 µT-years, 1.4 (1.2-1.5) for 2.0 ² 4.3 µT-years, and 1.6 (1.5-1.8) for ³ 4.3 µT-years. Significant increases were also seen for each category of exposure more than 5, 10, and 20 years previously, while significant decreases were seen for people in the highest exposure categories during the previous five years. In addition, the increased relative risks showed a statistically significant dose-response trend. Arrhythmia-related diseases were related similarly but less consistently to cumulative exposure. Although no direct information on known risk factors was available, the authors argued that it is unlikely that smoking and other lifestyle factors were important confounders, since no association was seen between exposure to magnetic fields and death from atherosclerosis or coronary heart disease. [The analysis of death from cardiovascular disease was an addition to the existing cohort study of cancer, and death certificates were used for the diagnosis of cardiovascular disease. See section 4.1.1.1 for comments on the limitations of the study. It is unclear why significant decreases in risk were observed for myocardial infarct and arrhythmia-related diseases after exposure during the five-year period prior to death, while significant increases were consistently found for exposure at the other times considered.]

4.5.1.5 Summary

Adverse reproductive effects from mother's exposure to EMF

The relationship between spontaneous abortion and exposure to ELF EMF was considered in two studies. In the first (Schnorr et al., 1991), no association with VDT use was observed; however, 'exposed' and 'unexposed' subjects had similar levels of exposure to ELF fields but different levels of exposure to VLF fields. In the second study, a significant association was seen with exposure to high magnetic and VLF fields; however, the response rate was very poor, particularly among controls.

The association with low birthweight was assessed in two studies (Bracken et al., 1995a; Grajewski et al., 1997), which found no association with either use of VDTs or exposure to magnetic fields.

Intrauterine growth retardation was considered by Bracken et al., who found no association with either use of VDTs or exposure to magnetic fields.

One study addressed the association between preterm birth and use of (Grajewski et al., 1997). No association was found.

Adverse reproductive effects from the father's exposure to EMF

The effect of paternal exposures to EMF on the frequency of congenital anomalies was considered in one study (Tornqvist, 1998). No increase was seen among the offspring of exposed male workers.

Alzheimer disease

Feychting et al. (Feychting et al., 1998) also considered a possible association with all dementia, including Alzheimer disease, and found a significant association with exposure to TWA magnetic fields > 0.2 µT. As mentioned above, this study has limitations and the results are therefore difficult to interpret.

Amyotrophic lateral sclerosis

Suicide and depression

One study addressed the risk for depression among electrical workers in a cohort of Viet Nam veterans. A nonsignificantly increased risk was observed for only one of the 24 measures of depression used. Given the small number of cases and the many outcomes analyzed in the study, the finding may be due to chance.

The risk for suicide was considered in two studies. No significant association with exposure to EMF was seen by Johansen and Olsen (Johansen, 1998), while an increased risk was seen with only one of 10 measures of exposure by Baris et al. (Baris et al., 1996a).

Cardiovascular disease

Savitz et al. (Savitz et al., 1998c) examined data from their study of five utilities, motivated a priori by a biological hypothesis based on clinical data and the results of human studies in vivo, which predicted increased numbers of deaths due to arrhythmia and acute myocardial infarct but no increase in risk for arteriosclerosis and chronic coronary heart disease. Significant, exposure-dependent associations were reported. Although no direct information on potential confounder was available, the authors provided indirect justification that smoking and life-style factors are unlikely to strongly confound the observed association.

Evaluation

There is inadequate evidence that maternal occupational exposure to ELF EMF causes adverse birth outcomes.

[This conclusion was supported by 22 Working Group members; there were 2 votes for 'lack' of evidence, 1 abstention, and 4 absent.]

There is inadequate evidence that paternal occupational exposure to ELF EMF causes reproductive effects.

[This conclusion was supported by 20 Working Group members; there were 3 votes for 'lack' of evidence, 2 abstentions, and 4 absent.]

There is inadequate evidence that occupational exposure to ELF EMF causes Alzheimer disease.

[This conclusion was supported by 23 Working Group members; there was 1 vote for 'lack' of evidence, 1 abstention, and 4 absent.]

There is inadequate evidence that occupational exposure to ELF EMF causes amyotrophic lateral sclerosis.

[This conclusion was supported by 24 Working Group members; there was 1 abstention and 4 absent.]

There is inadequate evidence that occupational exposure to ELF EMF causes suicide or depression.

[This conclusion was supported by 17 Working Group members; there were 6 votes for 'lack' of evidence, 2 abstentions, and 4 absent.]

There is inadequate evidence that occupational exposure to ELF EMF causes cardiovascular disease.

[This conclusion was supported by 24 Working Group members; there was 1 abstention and 4 absent.]

4.5.2 Environmental exposure

4.5.2.1 Pregnancy outcome

Seasonal patterns of occurrence of slow fetal development were observed among users of electric waterbeds and blankets, suggesting that use of these appliances at the time of conception might have had an adverse effect. The proportion of above-median gestational periods was significantly greater for conceptions in September-June than for those in July-August (56% vs. 38%, p < 0.0005) and was shown to follow a marked seasonal pattern among users, being high for babies conceived in winter and low in summer; the proportion was approximately constant at about 58% in the period October-April and decreased regularly down to around 40% during May-September [data presented only in graphical form]. No seasonal pattern was observed among non-users of electric blankets and water-beds. The prevalence of birth weight < 2500 g was similar in users and non-users (4.5 and 4.1%, respectively). Owing to a tendency not to announce the births of babies with congenital malformations, these were analyzed only among the 528 sibling births. One of 335 births among non-users (0.3%) and five of 193 among users (2.6%) had congenital defects. The frequency of abortions during the year preceding conception of a live infant was significantly higher (p < 0.05) among electric blanket users (7.8%; 24 cases)-but not among water-bed users (6.3%; 28 cases)-than among those who had used neither electric blankets nor water-beds (4.2%; 50 cases). A significant excess of abortions in September-January was noted among users of electric blankets and water-beds. [The ascertainment of pregnancy outcome may have been biased by the fact that the study population was defined on the basis of publication of birth announcements (only 42% of all births at the hospital) and that the sample was further restricted to those who could be located by telephone, who were presumably of higher socioeconomic status than the general population. Further, although efforts were made to exclude induced abortions by studying fetal loss in the year before conception of a live infant, some induced abortions were probably still included; the frequency of these is likely to have varied with socioeconomic status, as does the frequency of use of electric blankets and water-beds. It is also unclear whether the ascertainment of spontaneous fetal loss is complete from birth records. Finally, only recognized fetal losses followed by a live birth are included.]

A population-based case-control study of congenital malformations among children born in the 57 counties in upstate New York and registered in the New York State Congenital Malformation Registry was carried out by Dlugosz et al. (Dlugosz et al., 1992). The cases were 224 neural tube defects and oral cleft defects (121 cleft palates and 197 cleft lips) diagnosed among children in 1983-84 and 224 neural tube defects diagnosed in 1983-86. Three controls were randomly selected from birth registration records for each case and individually matched by maternal race, age, home county, month of mother's last menstrual period, and child's gender. Maternal exposure to heated water-beds or electrical blankets one month before to three months after conception was assessed from responses to a questionnaire mailed or administered by telephone. Information on maternal education, prenatal vitamin intake, parity, smoking, high fever during pregnancy, and occupation was also requested. The response rates were 83% for the cases and 77% among controls. No association was found, with or without adjustment for other risk factors, for various measures of exposure to EMF. For use of electric blankets at any time during pregnancy, the adjusted odds ratios were 0.69 (95% CI, 0.25-1.9) for cleft palate, 0.70 (0.36-1.4) for cleft lip, and 0.77 (0.40-1.5) for neural tube defects. For use of waterbeds at any time during pregnancy, the corresponding odds ratios were 0.66 (0.23-1.9), 0.63 (0.33-1.2), and 1.1 (0.59-1.9). No association was seen when different levels of exposures were considered, such as duration of use and heat-control setting. No seasonal pattern of time of conception was observed. [No information was available on patterns of use, such as continual use throughout the night, and other sources of exposure were not collected.]

The association between loss in early pregnancy and domestic exposure to ELF EMF was investigated in a case-control study in Finland (Juutilainen et al., 1993). The study was nested within a cohort study of work and fertility in 443 healthy volunteer women who were trying to become pregnant over the period 1984-86. Each woman in the cohort was followed-up for six months or until she became pregnant. Pregnancy and loss in early pregnancy (defined as pregnancies that ended before they became clinically apparent) were detected by measuring serum human chorionic gonadotropin on the first day of menstruation. All 107 women in this cohort whose first pregnancy resulted in an early loss were sought for inclusion in the study, with a random sample of 122 women with normal pregnancies from the cohort. After exclusion of cases and controls for whom information on exposure to EMF could not be obtained (refusals, address not found or incomplete), 89 cases and 102 controls remained for analysis. Magnetic field intensity was measured in the residences of the women at the time they participated in the study of work and fertility. Measurements were made at the front door of each residence and, for 48% of case and 57% of control residences, in the living-room, the kitchen, and the parents' bedroom, with a self-constructed magnetic field meter with a small ferrite-core measuring coil similar to that used by Juutilainen and Saali (Juutilainen & Saali, 1986b). Occupational exposure to EMF was assessed on the basis of job classification and some measurements. Odds ratios were calculated by the Mantel-Haenszel procedure, with adjustment for mother's smoking, age, and type of dwelling; separate analyses were carried out for front-door measurements, average magnetic field in the home, and maximum field. When the analyses were based on exposure in three categories, elevated odds ratios were observed for the highest group versus the lowest: front door, > 0.126 µT vs. < 0.63 µT (OR = 5.1; 95% CI, 1.0-26); average field strength, ³ 0.252 µT vs. < 0.63 µT A/m (4.6; 0.9-25); maximum field, ³ 0.63 µT vs. < 0.126 µT (2.7; 0.6-12), but intermediate groups did not have a higher risk than baseline. For dichotomous exposure classification, only average magnetic field strengths was associated with loss in early pregnancy (5.4; 1.1-28 for ³ 0.252 vs. < 0.252 µT) The authors concluded that their results suggest a causal association, possibly with a non-linear dose-response relationship. [Measurements inside houses were available for only a small proportion of subjects.]

Bracken et al. (Bracken et al., 1995a) carried out a prospective study of pregnancy outcome in 3591 pregnant women living in the New Haven, Connecticut, area and receiving prenatal care from 11 private obstetrical practices and two health maintenance organizations in the period 1988-91. The outcomes considered were intrauterine growth retardation (being in the lowest tenth percentile of birth weight for each week of gestational age), low birth-weight prevalence, birth weight, and gestational age. Information on pregnancy outcome was obtained from the records of the health facilities. Women were asked in a questionnaire interview to provide information on their frequency and duration of use of electric blankets and water-beds and also on the type and temperature setting. Electric-blanket field strength was estimated in temperature-controlled chambers on the basis of data about several blankets at selected chamber temperatures and blanket settings. Other information collected during the interview included age, race, marital status, education, religion, medical history, contraceptive use, smoking, job title, drug use, alcohol and caffeine consumption, and exercise. The participation rate was 82.6%: thus, information was available on 2967 women.

A nested study design was used to monitor exposure at various stages of pregnancy by both direct and indirect methods. Women who reported use of electrically heated beds and a random sample of non-users were placed in an intensively monitored group for subsequent monitoring at approximately 20 (446 women), 28 (431 women), or 36 (427 women) weeks of gestation. The entire monitoring protocol was completed by 307, 281, and 261 women, respectively (compliance was higher among electric bed users than among non-users: 70.8 vs. 62.7%). The monitoring program included seven-day personal exposure assessment with AMEX-2 personal monitors; 24-h measurements of EMF in the center of a room [not otherwise specified] with EMDEX meters; duty cycle measurements of water-beds (i.e. proportion of time the bed's heating elements were switched on); and wire coding of residences (up to four for women with more than one residence). The wire coding was evaluated by the approach of Wertheimer and Leeper; however, as the thickness of the wires could not be assessed because of insulation, electric utility company circuit maps were consulted to determine actual wire gauges.

None of the measures of exposure was significantly associated with any of the adverse outcomes considered. There was no suggestion of an effect of use or heat setting of electrically heated beds at the time of conception or during the first four or the last three months of pregnancy on birth weight, low birth-weight rate (< 2500 g), or intrauterine growth retardation. The results were similar with use of wire codes and of directly measured exposures in the monitored sub-sample.

Belanger et al. (Belanger et al., 1998) reported on the 135 spontaneous abortions that occurred in the same cohort between 7 and 25 weeks. No significant association was found between the incidence of spontaneous abortion and any measure of exposure to EMF. For blanket use at the time of conception, the odds ratio, adjusted for ethnicity, gestational age at interview, maternal age, and caffeine consumption, was 1.7 (95% CI, 0.96-3.2); no association, however, was seen for daily rather than less frequent use. Logistic regression analyses for duration of exposure (hours of use per day) and intensity of exposure showed a slight increase in the risk for spontaneous abortion with increasing duration and intensity of use. The small numbers of exposed cases, however, precluded a conclusive interpretation of these results. The odds ratio for water-bed use was 0.59 (95% CI, 0.33-1.1); no association was found with frequency, duration, or intensity of use. Although strong associations were observed with several other risk factors such as intrauterine growth retardation and marital status, ethnicity, education, reproductive history, and smoking, these did not have a confounding effect. [Assertainment of loss in early pregnancy was not complete, as only 14% of the cohort was interviewed before 10 weeks of gestation. Spontaneous abortion was assessed by interview.]

Li et al. (Li et al., 1995) studied electric blanket use during pregnancy in relation to the risk for congenital urinary tract anomalies in a case-control study in Washington State, USA. The cases were identified from the Washington State Birth Defects Registry over the period 1990-91, and 118 cases with no chromosomal abnormalities and 369 controls, consisting of births randomly selected from all single live infants delivered in five hospitals of King County in the study period, were included. Questionnaire interviews were completed with 62.6% of the case mothers and 67.6% of control mothers. Information was collected on history of reproduction, contraceptive use, smoking, alcohol consumption, and other substances [not specified]. The hypothesis was that any effect on fetal development of exposure to EMF during pregnancy might be more pronounced in susceptible subpopulations, such as women with a history of subfertility.

No association was observed between the incidence of congenital urinary tract anomalies and use of water-beds or blankets during the entire pregnancy or the first trimester. Elevated risks associated with electric blanket use were found among subfertile women, although this finding was based on few exposed cases (OR = 4.4; 95% CI, 0.9-22; 5 cases during the entire pregnancy; OR = 10; 95% CI, 1.2-85; 3 cases during the first trimester). The risk also appeared to increase with increasing duration of use of electric blankets. No association was found with use of water-beds or VDTs during the entire pregnancy or the first trimester. Recall bias might have influenced the results owing to media attention, but the association was observed only in subfertile women. No information on other sources of exposure to EMF was available. The authors hypothesized that hormonal disturbances caused by EMF (estrogen and progestogen production is affected by melatonin) could explain the findings. [The definition of subfertility as no pregnancy after unprotected intercourse for a year is inadequate, and applied to 23% of the study population. The finding in 'subfertile' women is hypothesis generating.]

4.5.2.2 Neurodegenerative diseases and neurobiological disorders

4.5.2.3 Summary

Pregnancy outcome

Pregnancy loss was investigated in two cohort studies. Early pregnancy loss was studied in one study in Finland, in which an increased risk was observed in the highest but not the intermediate category of exposure to magnetic fields. Pregnancy loss toward the end of the first trimester or later was studied in the USA; an elevated risk was found among users of electric blankets but not among users of water-beds.

In the same study in the USA, intrauterine growth retardation was assessed in a cohort of pregnant women. No association with residential exposure to EMF was found on the basis of a variety of measures, including use of electric blankets or water-beds and wire coding.

Depression

There is inadequate evidence that environmental exposure to ELF EMF has adverse effects on pregnancy outcome or is associated with depression.

[This conclusion was support by 23 Working Group members; there was 1 vote for 'no evidence', 1 abstention, and 4 absent.]

Table 4.36 Adverse pregnancy outcomes and maternal exposure to EMF

Table 4.37 Alzheimer disease and dementia in association with exposure to EMF

Table 4.38 Amyotrophic lateral sclerosis in association with exposure to EMF

Table 4.39 Suicide and depression in association with exposure to EMF

Table 4.40 Mortality from cardiovascular disease in relation to exposure to magnetic fields (Savitz et al., 1998c)