Intended for healthcare professionals


Childhood cancer in relation to distance from high voltage power lines in England and Wales: a case-control study

BMJ 2005; 330 doi: (Published 02 June 2005) Cite this as: BMJ 2005;330:1290

Authors' reply

We thank everyone who has commented on our paper; we respond here
only where we feel we can add anything to what we have already said in the

Various commentators have criticised us for publishing alarming
results that we are unable to explain. We should have preferred to delay
publication until we could offer a definitive explanation for our results.
However, once the analysis was complete, it would have been unethical not
to publish results of potential public health significance. Moreover, a
partial version of these results had been leaked and it became clear that
the only satisfactory way to respond to these leaks was to publish the
complete results.

We address first the responses concerning problems of methodology and
interpretation of the results, and then those that suggest possible
explanations of the results.

We do not agree with the statement by Hepworth et al that “the
findings are inconsistent with [the UKCCS1] study” – the only other UK
study with which comparison can be made. We consider that our results are
entirely consistent with that study: their relative risk (for acute
lymphoblastic leukaemia) of 1.42 for 0-400 m seems to agree rather well
with ours of 1.69 for 0-200 m and 1.23 for 200-600 m. This conclusion is
not weakened by the fact that the UKCCS estimate was not statistically
significant; this lack of statistical significance could be a consequence
of the smaller sample size in that study.

Hepworth et al and Kheifets et al raise questions concerning the
controls. As we stated in our paper, it seems possible that the elevated
relative risk for leukaemia depends, at least partly, on an
unrepresentative set of controls, since the addresses of the leukaemia
controls tend to be further from power lines than those of the controls
for the other diagnostic groups. We are puzzled by the suggestion by
Hepworth et al that findings can be considered to be less robust if the
estimates are noticeably different when the matching is broken, though, as
Kheifets et al point out, in the present analysis the estimates in fact
remain essentially unaffected. This is, however, quite separate from the
question of whether the complete set of controls should have been used.
Kheifets et al show that different estimates are then obtained. Although
these authors do not say so, these estimates would provide little evidence
for a relation between distance and leukaemia risk. There are two reasons
for regarding these latter estimates as unsatisfactory. First, they do not
take account of the original matching factors, particularly year of birth
and birth registration district. In fact, adjusting for birth year has
little effect on the estimates whichever set of controls is used. One
cannot, however, allow for a possible effect of birth registration
district in the unmatched analysis. Secondly, and in our view more
importantly, it is invalid to re-analyse the data using alternative
controls if this is done simply because the first set gives unexpected
results. (The situation is different if the original analysis is subject
to bias. It is extremely unlikely that there is any important source of
bias here.)

Hepworth et al suggest that adjustment for confounding factors might
explain our results. Neither our (admittedly less than adequate) measure
of socio-economic status (reported in the paper) nor population density
(not reported) explains the findings. We considered the question of
population mixing but it is not clear that an appropriate measure is
available for the whole of England and Wales over a period of 34 years. We
agree there might be other confounding factors that could explain our
results were we able to identify them.

Whitlock raises the question of bias arising from possible
differences in the likelihood of omitting cases near and far from lines.
We think this is unlikely, but such an effect would presumably apply also
to the controls and to the other diagnostic groups.

Coghill, Hepworth et al and O’Carroll refer to our calculation that
the association with distance that we reported implies that five cases of
childhood leukaemia a year in England and Wales would be attributable to
high voltage power lines if the association is causal. None of these
writers repeats our distinction between (chance) association and
causality. Coghill makes suggestions about the numbers of cases
attributable to 132kV lines. We do not agree with all his reasoning, but
in any event he goes beyond our data. O’Carroll’s calculations are based
on the assumption that the results at 0-200 metres are due to magnetic
fields while those at 200-600 metres are due to chance. We do not think it
justifiable to make an arbitrary division of our results into two bands.

Some of our correspondents over-interpreted, perhaps misinterpreted,
the findings. Burgess draws attention to the finding of a decreased
relative risk for CNS/brain tumours near the lines. But this decrease,
unlike the increase for leukaemia, is not part of a statistically
significant trend, nor does it correspond to any prior hypothesis. We
agree with Gaylord’s suggestion that the pattern of results for CNS/brain
and other tumours appears to be due to chance; this particular argument
cannot be applied to leukaemia though we have emphasised that, for other
reasons, we regard it as possible that the results are in fact due to
chance. Phillips appears to place too much emphasis on small, probably
chance, increases in relative risks at greater distances. His graph
appears to contain some inaccuracies and compares our leukaemia results
with the UKCCS ‘all malignancies’. As explained above, we do however agree
with him that our results relating to distance and leukaemia risk are
consistent with those of the UKCCS.

We, and our respondents, have considered a number of alternative
explanations for our results. Henshaw and Preece refer to Henshaw’s corona
ions hypothesis. Coulton questions its plausibility; we tested for it
without taking any view on its plausibility. We described our test as
“oversimplified”. Preece, who devised the method, points out the
simplification that all addresses in the north-east quadrant from the line
are considered “exposed”, i.e. that the wind that transports these ions is
assumed to be from the south-west, whereas one ought to consider actual
wind directions. Additionally, all addresses within 600 m are considered
equally exposed, without taking account of the actual distance or the
different propensity of different lines to produce ions, and the method
considers the closest point of the line only. We are analysing our data
using a better test, agreed with both Preece and Henshaw, avoiding these

Coghill, Juli, Lincoln and Preece raise questions about the
measurement of fields and about other sources of EMFs. We made no
assumption about a direct equivalence between field and distance. We shall
analyse calculated fields in a further paper; these fields, which are
still being checked, take into account the line characteristics mentioned
by Juli. We are investigating the possibility of analysing proximity to,
and calculated fields from, lower voltage distribution systems, but not
those from domestic appliances. Electric fields, suggested by Coghill,
appear no more likely to explain risks at 600 m than magnetic fields.

We shall investigate as many as possible of the various suggested
explanations put forward by Koenigbuescher, Netter, Poston, Coghill,
Henshaw, Preece and McDevitt, though in many cases the relevant data
will not be available.

1 UK Childhood Cancer Study Investigators. Childhood cancer and
residential proximity to power lines. Br J Cancer 2000; 83:1573-80.

Competing interests:
GJD, TJV and MEK: no conflict of interest. JS is employed by National Grid Transco and worked on this project with their permission

Competing interests: No competing interests

04 July 2005
Gerald Draper
Honorary Senior Research Fellow
Tim Vincent, Mary E. Kroll, and John Swanson
Childhood Cancer Research Group, University of Oxford, 57 Woodstock Rd, Oxford, OX2 6HJ