Let us discuss the murderous cell phones stalking our fair land

Cancer and cell phones - I meant to blog about this for some time, since it has long trended among the most read articles at the Times website.

To be charitable, the article did make me go and look up the literature, so that's not a bad thing. In short, however, the Times treatment is irresponsible and fear-mongering.

First, let me remark that the Times article mentions by name a refereed study of cellphones in humans only in the 14th paragraph. And it neglects to mention the multiple studies which have shown no connection.

Now, let's consider the INTERPHONE study referred to in the Times piece (it's one of these with the fake acronyms). It showed no connection between cell phone use and cancers, when all brain cancers are taken together. Now, it's reasonable for them to analyze different cancers separately, since they are of different severity and prevalence. It's not questionable in itself that they looked at an effect on gliomas. However, this effect was not significant. Only when they looked at cell phone use for 10 years or longer did they find an association with gliomas.

Several caveats - screaming sumo-size caveats - were not mentioned in the Times piece. (A science reporter presumably would have read the article.)

1. As far as I can figure out from reading the article, it's a secondary analysis. There was no a priori hypothesis that cell phone use for 10 years would be associated with glioma. Post hoc analyses are suspect - as you know - since data mining is biased. How many associations were fished through and discarded before this positive one was found? There is always a probability of a false positive, so if there were twenty post-hoc associations (properly consigned to an on-line appendix, pace the Times's conspiratorial mutterings), the chance of one positive finding is 5% - just by probability.

2. The association itself is not statistically significant! This is mentioned nowhere in the Times article, but the authors of the study themselves make haste to note this up front, in the abstract - which makes them responsible. I would not call this a hook to hang anyone's hat on.

3. The INTERPHONE study is a case-control study. A big question in any study of this kind is how we are to judge the accuracy of the cases' self reporting. People with cancer are understandably eager to find a cause, and might recall cell phone use out of proportion to the controls. Such recall bias is hard to control for.

4. Even if recall bias is controlled for, the correlation between recollected number of cell phone calls and the actual number of cell calls is not perfect. Heavier cell phone users tend to overestimate the number of calls they have made. In addition, the correlation between subject recall and their actual exposure to electromagnetic frequency is not airtight either.

5. Let's say the effect is real (which I very much doubt by reason of the sumo caveats just mentioned). (This would contradict another case-control study done on the very relationship between gliomas and cell phone use, in 2005, which was negative.) How high should this putative danger even rank on our public-health agenda?  Gliomas are rare.

To quote a 2004 study in the journal Cancer (first thing I could Google): "The incidence rate of central nervous system (CNS) tumors in 2000 was 6.7 per 100,000 persons as reported from the Surveillance, Epidemiology, and End Results (SEER) registry and gliomas account for approximately 51% of all CNS tumors.". Let's say then 3.5 per 100,000 people. or 10,000 cases a year, more or less. Horrible cases to be sure. If cell phone use increased this number to 20,000 cases a year that would be a tragedy, but a tragedy comparable to the million deaths caused yearly by malaria?

6. In toto: bullshit.

Update: but see some second thoughts here.