Map of life expectancy at birth from Global Education Project.

Tuesday, May 31, 2011

Hang up and read

I feel compelled to comment on the massive worldwide freakout over the WHO's International Agency for Research on Cancer declaring that cell phone use might be a cause of brain cancer. I have not made any deep study of this issue but I can step up and give you some clarity and perspective that generally seems to be lacking.

As you ought to know, I am the last person in the world who would ever shill for the mobile phone industry. (I do own 20 shares of AT&T stock, which doesn't exactly make me their punk, I don't think.) I have been outspoken here more than once about the very substantial risk of talking on the phone while driving, which for some reason doesn't seem to bother people; and the profound idiocy of talking on the phone all the time about trivial nonsense when you could be actually existing in the world. I make phone calls when I have a good reason, and save the idle chit chat for when I'm just hanging out with people who are actually sharing my region of the space-time continuum.

But let's get real here. What the IARC has said is that very heavy, long term cell phone use might be associated with a slightly elevated risk of a rare brain cancer called glioma. The press release refers to as yet unpublished results from the so-called Interphone study, which might suggest to readers that the study found out something really scary.

Okay, here's what the Interphone study found. It's a case control study, which means that the proper way to look at the conclusion is that it's telling you the risk of having used a cell phone that is associated with being diagnosed with glioma. If that sounds kind of backwards, that's because it is. So no matter what, it can't prove causation: maybe cell phone users tend to have something else going on that's associated with cancer. At most it can be suggestive.

But it's not even all that suggestive. In fact, the odds of ever having regularly used a cell phone are lower for people with glioma than for people without it, as are the odds for people with another kind of cancer called meningioma. Both these results are statistically significant. But they don't want to say that means cell phone use at moderate levels protects you against cancer, which would be the obvious conclusion. Instead they say this possibly reflects "participation bias or other methodological limitations." So they make this observation, but they don't believe it, without specifying any plausible reason, just guessing there might be one.

But, they also found that there was an elevated association between glioma and the highest reported decile of recalled cumulative call time over ten years, and odds ratio of 1.4, although it was barely statistically significant. As I've told you a million times, you're likely to see a supposedly significant finding in a sub-group comparison that is actually spurious, because you're making multiple comparisons and you could just stumble across a coincidence. And they say, "there are implausible values of reported use in this group." So maybe people with brain cancer have problems with memory or understanding the questions? But they aren't willing to suggest that this result possibly reflects any methodological limitations.

Finally, although for technical reasons the case control study can't directly output the actual population risk, but only the odds ratio for the study population. But I can tell you that there are about 10,000 gliomas diagnosed in the U.S. each year, which means the risk you will be diagnosed this year is 3.3 in 100,000. The odds ratio is almost identical to the risk ratio for such a rare event so the actual elevated risk from implausibly heavy cell phone use over ten years is 1.3 cases in 100,000 people per year. If I have done all that correctly. Okay, it's higher than that because we probably shouldn't include children, maybe twice as high if we're limiting this to adults.

Feel better now? Can you think of anything you maybe should be more worried about? That maybe we aren't talking about?

No comments: