«Latest  ‹Forward   News item: 3436  Back›  Oldest» 

More INTERPHONE wierdness as significant "protective effect" reported as "no association"
Finland Created: 11 Sep 2008
The latest Interphone paper by Lahkola et al from Finland [Lahkola 2008] incorrectly summarises its findings as "Our results do not provide support for an association between mobile phone use and risk of meningioma". In actual fact, their results suggest a very distinct and fully significant protective association between mobile phones and meningioma risk. Their figures suggest that being a mobile phone user can lower your lifetime risk of developing a meningioma by as much as 25%!

In reality of course this association is unlikely, but to conclude that there was no support for an association is misleadingly hiding behind the fact there are some very serious flaws in the study design that would prevent a positive association from being found. There are 65 odds ratios in the paper, of which 62 of them are below 1 - we would expect about 50% below and 50% above if there was truly no association and no significant design flaws.

In the paper text itself the authors do recognise the over-abundance of low ORs, and explain that this is likely due to "selection bias" which would lead to an "underestimation of risk". This is true, and is an issue covered repeatedly in great detail in Lloyd Morgan's excellent column. What they then failed to mention is the three ORs that were over 1 were all in the heaviest user category. As we covered before on the last paper that we found with this significantly protective effect (also, curiously, from the same author), a crude adjustment for the selection bias gives a very different picture for risk in the highest user group. Their 2007 paper[Lahkola 2007] would have found approximately a 40% increase in risk for the highest user group with ajustment for the nonexistent "protective" effect, and this paper would have found a 30% increase (of borderline significance). This is crude statistical maths and is not in any way a water-tight argument, but to have not covered (even speculatively) what controlling for the selection bias would have done to the heaviest user group is a very serious omission by the study authors. With 62 of the 65 ORs below 1 (showing a protective effect), and 27 of these being statistically significant, there are clearly very serious issues with the paper. There is a very real chance that this study does in fact support an association between long term mobile phone use and an increase in brain cancer risk once the study flaws are adjusted for, and this possibility should have been mentioned in the study abstract.

Louis Slesin has also covered this issue on Microwave News, and looks at the findings in context with the other papers in the Interphone project, including the continuing rift between project researchers about how to interpret the low ORs generally and the consistent increase for long term phone users.

References

1. N Lahkola A et al, (August 2008) Meningioma and mobile phone use--a collaborative case-control study in five North European countries, Int J Epidemiol. 2008 Aug 2.
http://www.ncbi.nlm.nih.gov/pubmed/18676984

2. P Lahkola A et al, (April 2007) Mobile phone use and risk of glioma in 5 North European countries, Int J Cancer. 2007 Apr 15;120(8):1769-75
http://www.ncbi.nlm.nih.gov/pubmed/17230523
Click here to view the source article.
Source: PowerWatch UK, 03 Sep 2008

«Latest  ‹Forward   News item: 3436  Back›  Oldest»