The Social Affairs Unit

Print Version • Website Home • Weblog Home


Use the buttons below to change the style and font size of our site.
Screen version     Print version:   
June 07, 2005

British Medical Journal launches new global health scare

Posted by John Adams

Misleading, alarmist headlines linking childhood cancers with living close to power lines appeared in the media around the world last week. Prof. John Adams - Britain's leading academic expert on risk and the author of the seminal Risk - examines how such misleading headlines came to be published from London to Sidney, via China, the USA and Russia.

On 4th June the BMJ [British Medical Journal] published an article (by Draper, Vincent, Kroll and Swanson) entitled: "Childhood cancer in relation to distance from high voltage power lines in England and Wales: a case-control study".

It was accompanied by a short piece entitled: "Editor's Choice: Making sense of Things". The deputy editor observed:

The finding about power lines [is] puzzling. As the authors themselves say, there is no accepted biological mechanism to explain their results and 'the relation may be due to chance or confounding'.

Despite this editorial puzzlement, the BMJ held a press conference on 3rd June to publicize the article and ran a trailer for the media on its website. It had a bold headline, a scary photo, and a message that reserved any doubt for the last sentence:

Leukaemia: avoid living near high voltage lines... Children who, at birth, live near high voltage power lines in England and Wales are more at risk of getting leukaemia than children who don't. In a case-control study of more than 29,000 children with cancer and their matched controls, Draper and colleagues (p 1290) found that - compared with children who lived more than 600 m away from the power lines - children who lived within 200 m and between 200 m and 600 m away had a significantly increased risk of leukaemia (relative risk 1.69 and 1.23, respectively). The authors think the relation may be due to chance or confounding [emphasis on last sentence added].

BMJ 2005;330 (4 June), doi:10.1136/bmj.330.7503.0

The trailer was remarkably effective. In cyberspace, alarming headlines travel quickly. By 10 am on 3rd June, Google turned up fear-generating headlines on 116 websites of major media outlets all around the world, relating to an article nominally published on 4th June.

A few examples:

Study: Children near power lines may face higher cancer risk
CNN International

Children Living By Power Cables Have Cancer Risk, Study Shows
Bloomberg

Proximity to power lines feared to increase child leukaemia risk
Xinhua, China

Child Leukemia Again Linked to Power Lines
FOX News

Childhood leukaemia linked to power lines
Daily Times, Pakistan

Childhood cancer linked to overhead power lines
Zee News, India

Power line link to leukaemia
Irish Health, Ireland

Cancer higher near pylons
The Sun, UK

Babies at risk with high-voltage cancer link
Sydney Morning Herald

'Pylons caused my son's cancer'
BBC News, UK

Babies Living Close To Power Lines May Have Increased Leukemia ...
CBS2 Chicago, IL

Power Lines-Leukemia Debate Lingers
ABC News

Study links childhood leukaemia to power lines
Reuters.uk, UK

Fears over child leukaemia link to power lines
New Zealand Herald, New Zealand

Leukaemia among children near high voltage power line
EITB, Spain

High-voltage power lines increases the risk for childhood leukemia ...
Pravda, Russia

Pylons 'may be a leukaemia risk'
Journal of Turkish Weekly, Turkey

Most of the articles were less sensational than the headlines beneath which they appeared, and included many of the study's own caveats. The headline over the New Scientist article, for example, read: "Large study links power lines to childhood cancer". But, quoting the lead author, it observed:

The confusing message, which comes from the largest study to date - of over 29,000 children with cancer - is that since 'there is no biological mechanism to explain the higher risk', the results, 'although statistically significant, may be due to chance'... We don't think it is possible that a magnetic field of these low magnitudes [beyond 60 metres] could have a causative effect on childhood leukemia.

Nevertheless most of the articles reported that the study found "an association" between high-voltage power lines and childhood cancer. This conclusion, reported with scary, caveat-free headlines, leaves those readers predisposed to worry about their children's safety, with the idea that the connection between power lines and cancer is real, and mysterious and spooky. This fear is amplified by the news that it has been established by "the largest study to date", and the assertion that the association is "statistically significant".

But let's look at the numbers. The 29,081 cases of childhood cancer included in the study occurred over a period of 33 years. Of the distance bands in the table below, only the first, the 0-49 metre band, falls within the range over which the authors believe there might be a magnetic effect worth investigating. Within this range, over a period of 33 years they found 5 cases of leukaemia and 3 "controls" (children without leukaemia), giving a 5/3 = 1.67 relative risk ratio. But the largest relative-risk reported in the table is for "CNS/brain tumours" within 0-49 metres of a power line - but the effect is in the opposite direction; there are more controls than cases. The power lines appear to have a protective effect!

If we plot the relative risk ratios from the table, we see in the graph below a) that beyond 200 metres the relative risk ratios for leukaemia increase with distance, b) that up to 200 metres the "protective effect" for CNS/brain tumours diminishes with distance, and at greater distances the graph shows no trend, and c) that the graph for "other diagnoses" manifests no trend, and that over all distance bands below 600 metres there are 20 more controls than cases a slight protective effect?

Of course, with ratios based on such small numbers, it takes a very sharp-eyed statistician to see any "trends" or "effects" at all. But the same reasoning that, when applied to the distance band with the strongest magnetic effects, finds a "significant" association between power lines and leukaemia, finds the opposite association with respect to CNS/brain tumours. This finding conspicuously failed to generate any headlines about the protective effect of power lines.

A further problem is the study's failure to deal adequately with significant potential "confounders" i.e. other variables, not included in the analysis, that might provide an alternative explanation for the result. The authors use something called the Carstairs deprivation index to satisfy themselves that socio-economic status was not a possible confounder. But the index relates to census wards which, in rural areas, are probably at least 50 square kilometres in size. Thus the index they use has a level of resolution far too low to identify any possible effect associated with a corridor only 100 metres wide (or even 1200 metres wide). Given that close proximity to power lines is known to depress property values a recent report estimates by 30% - it is highly probable that those living close to them will be less affluent than others farther away living within the same ward.

Table 1 Distance of address at birth from nearest National Grid line for cases and controls in each diagnostic group, and estimated relative risk (RR).

LeukaemiaCNS/brain tumoursOther diagnoses
Distance to line (metres)CasesControlsRRCasesControlsRRCasesControlsRR
0-49531.67370.44761.17
50-9919111.79460.6915160.91
100-19940251.6426320.8237450.81
200-29944391.1638281.3566760.87
300-39961541.1535301.1979651.21
400-49978651.2340420.9680970.82
500-59975561.3654411.3386851.01
≥600 (reference group)937894471.00640564191.001240612 3861.00
Total97009700 66056605 1277612 776 
CNS = central nervous system.

In the same 4th June issue, the BMJ also published an editorial by Heather Dickinson entitled: "The causes of childhood leukaemia". It was clear and sensible and generated no headlines. It observed that:

factors are more likely to be causal if they are biologically plausible, are seen consistently in different populations, and have a dose-response relation to the risk of leukaemia.
It noted, along with the deputy editor, that the Draper study did not have a biologically plausible explanation for its findings, and that it included no measures of dose in the form of either estimates or measures of the magnetic fields to which the cases and controls were exposed. It also noted that:
its matching of controls to cases on the basis of administrative areas may have yielded controls who were not completely representative of the distance of children's homes from power lines.
This is a gentle way of saying that the socio-economic-status measure used to match the cases and the controls precluded the possibility of establishing that the two populations in the study had matching socio-economic characteristics. And it concluded that even if the association reported in the Draper study were statistically significant "it could account for only a tiny proportion of cases".

The Dickinson editorial did not generate headlines because they would have had to say something like:
New research might have found a possible, inexplicable, cause of a tiny proportion of cases of childhood cancer most of which are curable.

That would not sell many newspapers, or attract many new readers to the British Medical Journal.

[Currents of Death? by Adam Burgess is an excellent review of the history of this debate.]

John Adams is emeritus professor of geography at University College London.


Comments Notice
This comments facility is the property of the Social Affairs Unit.
We reserve the right to edit, amend or remove comments for legal reasons, policy reasons or any other reasons we judge fit.

By posting comments here you accept and acknowledge the Social Affairs Unit's absolute and unfettered right to edit your comments as set out above.
Comments
Post a comment








Anti-spambot Turing code







Creative Commons License
Except where otherwise noted, this site is licensed under a Creative Commons License.

The Social Affairs Unit's weblog Privacy Statement