To paraphrase Oscar Wilde, I never travel without my copy of the Journal of Epidemiology and Community Health. One should always have something sensational to read on the train.
 
This week’s shocker (1) was a paper published online claiming that use of a mobile phone while pregnant, and after giving birth, leads to a greater risk of behavioural problems when the child is seven.
 
It got widespread coverage in the papers: The Independent, Metro, the Press and Journal, the Daily Mail, the Daily Express, the Herald, the Scotsman, the Sun, the Daily Star and the Daily Telegraph all had a nibble at it. But – and this really is a cause for celebration – they didn’t all give it an easy ride.
 
The study is large, but the results implausible. What possible biological mechanism could apply to babies in the womb, and also once they are out of it? The odds ratios for behavioural problems are small, and not very consistent. But the numbers are large, and the study properly constructed. What are journalists to do?
 
In the case of one I know, it was to e-mail the authors at the University of Southern California well in advance seeking answers to several questions, such as the one about biological mechanisms I posed above. The response was unconvincing.
 
Others chose not to write the study up at all, or managed to make it dull enough not to get in the paper. No mention I could find in The Times, for example, or The Guardian.  
 
The most plausible explanation for the finding - unless it is chance - is that mothers who use mobile phones more tend to neglect their babies. The authors did attempt to correct for “inattention” and doing so weakened the link. But the measures they had of inattention were very poor, with quite a lot of missing data, and a more complete correction for this factor might have eliminated the link altogether.
 
The Science Media Centre played a big role in getting this story downplayed. It  assembled a series of critical comments that were used, to a greater or lesser extent, by many of the journalists. The full list of comments is below, and it’s clear from these that several distinguished statisticians and epidemiologists were unconvinced.
 
Professor Patricia McKinney, Emeritus Professor of Paediatric Epidemiology at the University of Leeds, said: "The conclusions from this large study, associating behavioural problems in very young children with mobile phone use, over-interpret the results.  There is no scientific basis for investigating exposure of the growing baby when pregnant mothers use a mobile phone.  The risks linked to prenatal exposure are therefore questionable."
 
Professor David Coggon, Professor of Occupational and Environmental Medicine at the University of Southampton, said: "This study appears to have been well conducted, but the pattern of results suggests that the observed increase in behavioural problems may have been caused by factors other than mobile phone use.  When the authors took into account considerations such as quality of parenting, estimates of risk were reduced, but confounding influences of this sort are unlikely to have been fully eliminated."  
 
Professor David Spiegelhalter, Professor of Biostatistics at the University of Cambridge, said: “I am sceptical of these results, even though they will get a lot of publicity.  One finding is that very young children who use mobile phones show more behavioural disorders: this may well be the case, but is it plausible that the first causes the second?  The authors suggest that precautionary measures may be warranted because they have ‘virtually no cost’, but they ignore the cost of giving intrusive health advice based on inadequate science.
 
"The authors say that 'early exposure to cell phones could carry a risk which, if real, would be of public health concern'.  Well, I might just as well say 'Paul's psychic abilities, if real, would revolutionise our thinking about molluscs.'"
 
It’s just possible, I suppose, that Dr Leeka Kheifets and her co-authors have discovered something worth reporting - though even they say it is premature to interpret these results as causal. But it seems unlikely. I think it’s time the editors of the Journal of Epidemiology and Community Health remembered that statistical significance is not the gold standard of good science. It should not be mistaken for evidence of an association that actually matters.
 
References
1 Cell phone use and behavioural problems in young children, Hozefa A Divan, Leeka Kheifets, Carsten Obel and Jorn Olsen, J Epidemiol Community Health (2010) doi 10.1136/jech.2010.115402