Statistics from Altmetric.com
While Savitz's point of view expressed in the editorial1 that epidemiological methodology faces its limits when the risk is small, exposure assessment is poor, and biological insight is lacking, must be reinforced, it is not so clear whether or not this view is applicable to the field of exposure to extremely low frequency electromagnetic fields (ELF EMF). Unfortunately some of the studies that could contribute to an evaluation of the relation between ELF EMF exposure and cancer have serious deficits. This is apparently also the case for the paper by Sorahan and colleagues.2 First it has to be stressed that there is no such diagnostic entity as “brain tumour”. Brain tumours comprise a heterogeneous group of both malign and benign neoplasms generating from different tissues, with different growth rates and other essentially different features (for an overview see Black3). The authors do not even mention the number of cases of different tumour types, let alone discuss why they feel that all these completely different entities could be affected by a single cause.
Another crucial point is latency. The only essential criterion of causation in the assessment of epidemiological evidence is “temporal relation”. It is crucial that provisions are made to allow for biologically reasonable latencies. Instead the authors report on estimates based on the most recent (!) five years of exposure, thus choosing an exposure metric that has nothing to do with the vast majority of brain tumours that have latencies of at least five (but many 20 or more) years (for example, Strojan et al4). Most of the brain tumours will have been already initiated before the point in time the exposure was accumulated to give the indicator the authors have chosen. At least the last 10 years prior to diagnosis of the tumour have to be truncated in computation of the exposure metric and all cases occurring earlier than 10 years after onset of exposure have to be omitted.
To choose Tesla-years as the exposure variable is also questionable because we do not know whether or not risk is cumulative. A more sophisticated exploitation of information on exposure could be expected from the authors. For example, time spent under peak exposures (e.g. exceeding 10% of the exposure limit) would be a meaningful surrogate. Tesla-years introduces an equivalence that has never shown to be meaningful: that exposure duration and intensity are commutative (that is, 10 years exposure to 1 μT is equivalent to one year exposure to 10 μT).
Overall the study in its presented form cannot be considered to contribute to the assessment of a relation between ELF EMF exposure and brain tumours.
Professor Kundi implies that, in our analyses of brain tumour risks and magnetic field exposure, we only considered exposures occurring in the most recent five years. We did not. Analyses of total cumulative exposures to magnetic fields in relation to mortality risks from primary brain tumours were reported in table 3, and analyses of the potential role of recent exposures were reported in table 4.1 Confirmation of diagnosis had also been sought from cancer registration particulars. These analyses were planned in advance as tests of the main hypotheses of interest. These hypotheses had been derived from a review of the current literature, and for neither analysis was there any suggestion of magnetic field exposure being implicated in mortality risks for brain tumours. The ICD codes we used to define the health outcome and the use of micro-Tesla years as the unit of magnetic field exposure enabled our study findings to be compared to other reports. Their use appears, at least to us, to be eminently sensible. We remain open to the possibility that other exposure metrics may come to be appreciated as more biologically relevant but we doubt whether the proposal of Prof. Kundi (time spent exceeding an arbitrary percentage of a contemporaneous exposure limit) will gain favour.
We hope our study makes a useful contribution to the practice of occupational health and that employees in the UK electricity supply and transmission industry treat the findings as good news.