Neurodegenerative disease and magnetic field exposure in UK electricity supply workers

Abstract

Background

Previous research has suggested a possible link between neurodegenerative disease and exposure to extremely low-frequency electric and magnetic fields.

Aims

To investigate whether risks of Alzheimer’s, motor neurone or Parkinson’s disease are related to occupational exposure to magnetic fields.

Methods

The mortality experienced by a cohort of 73051 employees of the former Central Electricity Generating Board of England and Wales was investigated for the period 1973–2010. All employees were hired in the period 1952–82, were employed for at least 6 months and had some employment after 1 January 1973. Detailed calculations had been performed by others to enable an assessment to be made of exposures to magnetic fields. Poisson regression was used to calculate relative risks (rate ratios) of developing any of the three diseases under investigation for categories of lifetime, distant (lagged) and recent (lugged) exposure.

Results

No statistically significant trends were shown for risks of any of these diseases to increase with estimates of lifetime, recent or distant exposure to magnetic fields.

Conclusions

There is no convincing evidence that UK electricity generation and transmission workers have suffered elevated risks from neurodegenerative diseases as a consequence of exposure to magnetic fields.

Introduction

There have been many studies into exposure to extremely low-frequency electric and magnetic fields (ELF-EMF) and risks of neurodegenerative disease (NDD) and two meta-analyses are available for Alzheimer’s disease (AD) [1,2]. Garcia et al. [1] reported an increased risk of AD from ELF-EMF exposure based on pooled risk estimates from nine case–control studies [odds ratio (OR) 2.03, 95% confidence interval (95% CI) 1.38–3.00] and five cohort studies [rate ratio or relative risk (RR) 1.62, 95% CI 1.16–2.27]. They also found considerable heterogeneity between study results and an indication of publication bias in the case–control studies. Various definitions of ‘exposure’ were used in the different studies and it was not possible to select a single ‘main result’ from each study in a standardized manner (these are limitations for all the meta-analyses under discussion). A more recent meta-analysis is also available [2]. Increased risks were shown for 36 risk estimates from 13 case–control studies (RR 1.29, 95% CI 1.11–1.50), and 15 risk estimates from six cohort studies (RR 1.39, 95% CI 1.10–1.75). In this meta-analysis, a study could contribute more than a single risk estimate.

Two meta-analyses are also available for motor neurone disease (MND) or its principal subtype amyotrophic lateral sclerosis (ALS). Zhou et al. [3] reported significant increases in risk of ALS due to ELF-EMF exposure based on nine case–control studies (RR 1.39, 95% CI 1.05–1.84), but no significant increase in eight cohort studies (RR 1.16, 95% CI 0.80–1.69). Similar findings are shown in the meta-analysis from Vergara et al. [2]. There was a significant increased risk for MND/ALS based on 27 risk estimates from 12 case–control studies (RR 1.38, 95% CI 1.13–1.68), but no significant increased risk from 23 risk estimates from nine cohort studies (RR 1.14, 95% CI 0.92–1.42).

If the above meta-analyses are taken at face value, it would not be possible to exclude a possible role for ELF-EMF exposure on risks of NDD, but convincing evidence requires cohort studies that show statistic- ally significant positive trends of disease risk with quantitative exposure estimates (there is no requirement for such dose–response effects to be monotonic). The aim of this article was to obtain important new information on the topic of occupational magnetic field exposure and risks of mortality from NDD by examining data from an ongoing, pseudonymized epidemiological study of UK electric utility workers. An earlier report considered follow-up to the end of 2004 [4]; a further 6 years of mortality data are now available.

Methods

The materials and methods have been summarized in an earlier companion article on brain tumour risks [5]. The analysis reported here is based on the same cohort of 73051 study subjects (62825 men and 10226 women) first employed in the period 1952–82 for whom a work history was available. The current article, however, only considers death certificate data as there are no national registers for NDD and follow-up is censored at age 95 years (rather than 85 years in the earlier analysis) [5] because an important percentage of death certificates mentioning NDD occur after the age of 85 years. The survey was established with the approval of the Central Ethical Committee of the British Medical Association, and the author is currently accredited by the Health and Social Care Information Centre as the ‘Approved Researcher’ of this cohort study.

Results

RRs for AD (any mention on the death certificate: 170 cases in total) are shown in Table 1 for four categories of estimated cumulative occupational exposure to magnetic fields relative to the corresponding rates in the lowest (baseline) category of exposure (Model 1). Corresponding RRs are also shown for a simultaneous analysis of distant (lagged) and recent (lugged) exposures (Model 2). RRs in the left-hand side of the table were adjusted for age and sex. RRs in the right-hand side of the table were additionally adjusted for calendar period and socio-economic status (three categories: managers, scientists and engineers; administrative and clerical workers; industrial and construction workers). To be concrete, the table summarizes four separate analyses. None of the individual point estimates of risk are significantly different from unity and there is no suggestion that risks increase with exposure. Findings were little different with or without adjustment for calendar period and socio-economic status.

Findings for MND (86 cases) are shown in Table 2. There are no significant trends between disease risks and exposure, but the point estimate of risk for the second exposure category is significantly elevated both for lifetime exposure and for exposures received >10 years ago. Point estimates of risk for lifetime and distant exposures were greater than unity for all four higher exposure categories. Findings were little different with or without adjustment for calendar period and socio-economic status.

Findings for Parkinson’s disease (PD) (278 cases) are shown in Table 3. There are no statistically significant positive trends of disease risk with exposure, but point estimates of risk are significantly raised for the third exposure category, both for lifetime exposures and for exposures received >10 years ago. Point estimates of risk for lifetime and distant exposures were greater than or equal to unity for all four higher exposure categories. Findings were little different with or without adjustment for calendar period and socio-economic status.

The analyses summarized in Tables 1–3 were then repeated for the sub-cohort of those 48768 employees first employed in power stations; these analyses were carried out because the exposure assessments for power station workers are more detailed than for other groups of workers. RRs for AD (125 cases) are shown in Table 4. None of the individual point estimates of risk are significantly different from unity and there is no suggestion that risks increase with increasing exposure. Findings were little different with or without adjustment for calendar period and socio-economic status.

Findings for MND (68 cases) are shown in Table 5. There are no significant trends between disease risks and exposure, but the point estimate of risk for the second exposure category is significantly elevated for lifetime exposure and approaches statistical significance for exposures received >10 years ago. Point estimates of risk for lifetime and distant exposures were greater than unity for all four higher exposure categories. Findings were little different with or without adjustment for calendar period and socio-economic status.

Findings for PD (205 cases) are shown in Table 6. There are no statistically significant positive trends of disease risk with exposure, but point estimates of risk are significantly raised for the third exposure category, both for lifetime exposures and for exposures received >10 years ago. Point estimates of risk for lifetime and distant exposures were greater than unity for three of the four higher exposure categories. Findings were little different with or without adjustment for calendar period and socio-economic status.

Discussion

This large cohort of UK electricity generation and transmission workers found no significant positive trends between estimated exposures to magnetic fields and the risks of three NDDs (AD, MND and PD). This was the case for lifetime occupational exposures, distant (lagged) and recent (lugged) exposures. The findings are consistent with the hypotheses that both distant and recent magnetic field exposures are not causally related to MND or PD and that distant magnetic field exposures are not causally related to AD. The study is uninformative about recent magnetic field exposures and AD because most death certificates mentioning AD occur many years after retirement age (range 60–94 years). These summaries are not dependent on the selection of covariates in the analysis or on the selection of sub-cohorts for analysis (all employees or power station workers only).

The analysis has many strengths including the large size of the cohort, long period of follow-up, large number of NDD cases available for analysis and detailed exposure assessments that used the physics of exposure to magnetic fields as a starting point [6]. However, there are limitations to be attached to the work. It was necessary to assume that for those workers hired before 1973, job and place of work in the 1950s and 1960s were the same as those pursued in the early 1970s, and it was also assumed that working patterns (time spent by different groups of workers in different parts of power stations) are the same in different power stations. These assumptions will have introduced errors into the exposure assessments. We remain confident, however, that the exposure assessments have value particularly if we accept the relative rankings of the five exposure categories and do not attach overwhelming importance to the absolute values. It must be the case, however, that the current exposure estimates fall short of an ideal survey that would include measured individual exposures over time. In addition, exposures were considered in relation to the date of death rather than the date of diagnosis, and there will have been diagnoses of NDD that do not appear on the death certificates. It is not possible on the basis of death certificate data to separate ALS from other types of MND.

This analysis was designed to carry out a minimum of multiple testing; there was one set of cut-off points for each of the three exposure metrics, and the principal test was a single test for trend across all exposure categories; there is, of course, no requirement for trends to be monotonic (risks in each exposure category greater than or equal to the risk in the preceding exposure category). One could only be confident of finding a monotonic trend in an infinitely large study with all relevant variables known with complete accuracy. These analyses do not consider the possible role of threshold effects (no effects at lower exposures) or saturation effects (same effects at moderate and higher exposures) and it is possible that, in the course of time, physiological considerations might lead to very different exposure metrics being investigated. There were some positive findings for MND and for PD, including significant findings for MND at exposures of 2.5–4.9 µT year and for PD at exposures of 5.0–9.9 µT year; confident interpretation of these isolated findings is not possible, but emphasis on these findings would involve making unattractive post hoc arguments. It would also be possible to combine the exposure categories to achieve either more positive or more negative results, but the resulting P values would be meaningless.

In conclusion, the current UK study does not provide convincing evidence that magnetic field exposures are a risk factor for NDD.

Funding

Energy Networks Association.

Conflicts of interest

The University salary of T.S. was defrayed, in part, by the research award from the Sponsor.

Acknowledgements

I thank the Office for National Statistics and the Health and Social Care Information Centre for follow-up details, Dr David Renew and other members of the EMF Research Section, National Grid Company plc, for the detailed exposure assessments and the pensions and human resources departments of the participating companies for solving data queries.

References