Cellphones and cancer

(graphic from Ask Santosh)

There's an article in yesterday's New York Times that starts like this:

WARNING: Holding a cellphone against your ear may be hazardous to your health. So may stuffing it in a pocket against your body.

Sounds scary, doesn't it? Wasn't all this cellphones-cause-cancer thing settled? The author, Randall Stross, is a professor of business, and that's what he thought, too, until:

The cellphone instructions-cum-warnings were brought to my attention by Devra Davis, an epidemiologist who has worked for the University of Pittsburgh and has published a book about cellphone radiation, “Disconnect.” I had assumed that radiation specialists had long ago established that worries about low-energy radiation were unfounded. Her book, however, surveys the scientific investigations and concludes that the question is not yet settled.

OK, so someone has just written a new book and contacted Stross to get the word out. Well, there's nothing wrong with marketing yourself, but I have to wonder how a professor of business is suited to judge a scientific question like this.

But he certainly seems to be worried. Well, who wouldn't be? However,... if cellphones cause brain cancer, wouldn't you expect that to be obvious by now? After all, we went from a time when no one used a cellphone, since they hadn't been invented, to now, where they're ubiquitous. I certainly don't want to dismiss this out of hand, but as a practical matter, I don't see how this question could still be unsettled, one way or the other.

After a little bit of "he said, she said," Stross gets to this:

The largest study of cellphone use and brain cancer has been the Interphone International Case-Control Study, in which researchers in 13 developed countries (but not the United States) participated. It interviewed brain cancer patients, 30 to 59 years old, from 2000 to 2004, then cobbled together a control group of people who had not regularly used a cellphone.

The study concluded that using a cellphone seemed to decrease the risk of brain tumors, which the authors acknowledged was “implausible” and a product of the study’s methodological shortcomings.

Er, what? The study showed that using a cellphone seemed to decrease the risk of brain tumors? But let's dismiss that, because it's "implausible" and likely a product of the study's "methodological shortcomings." Instead, let's leap to another part of the study that gives us the results we'd rather see:

The authors included some disturbing data in an appendix available only online. These showed that subjects who used a cellphone 10 or more years doubled the risk of developing brain gliomas, a type of tumor.

OK, so I'm certainly no medical researcher, but I'll bite. I went to that Appendix 2 to see what the fuss was all about. This is the first line:

We observed an overall decrease in risk of glioma and of meningioma with any regular use of a mobile phone...

Um, excuse me? An overall decrease in risk of gliomas? But again, that can't be right, can it? So let's throw that out. The researchers decide to simply ignore people who don't use a cellphone, since that gives a result they don't want, and just compare different amounts of cellphone use.

Analyses of the INTERPHONE non-response questionnaire suggest the presence of participation bias: less participation of non-users of mobile phones than users. In addition, controls were less likely to participate than cases.

OK, that seems reasonable. Cellphone users have been sufficiently scared that they're more likely to respond to this questionnaire than people who don't use cellphones. On the other hand, who would be the most likely to respond to this kind of questionnaire? That would be people who use a cellphone and have a brain tumor, right?

This exposure category includes some highly implausible reported values of mobile phone use (e.g., 12+ reported hours of use per day), which were more common in glioma cases than in controls.

So people who use cellphones and have a brain tumor report "highly implausible" reported values of cellphone use. Of course, all of these people have likely heard the scare stories, and no doubt most are already convinced that their cancer was caused by cellphones. So they simply remember their cellphone use as being greater than it really was, in some cases so high as to be completely unbelievable.

Think about this. Anyone who has used a cellphone and who also has a brain tumor is likely to exaggerate their past cellphone use. It's not that they're deliberately lying, of course not. Most simply remember it differently, because they've all heard about a cellphone-brain cancer connection, and most are probably convinced that it's true (since "I don't know" isn't a very satisfactory answer for anyone).

Here's the researchers' conclusion:

Conclusion
Analyses excluding never regular users of mobile phones may have reduced downward bias in ORs [odds ratios] for menigioma and glioma due to selective non-participation of people who were never regular users. There is evidence, however, of persisting bias in the results of these analyses and it is possible that the exclusion of never regular users has produced upward bias in the ORs, particularly for glioma. Thus biases and error prevent a causal interpretation of these results.

"Biases and error prevent a causal interpretation of these results." Can it get any clearer than that? But that's not the conclusion of this New York Times article. No, according to a quote from an author pushing her own book, this is "an epidemic in slow motion."

Would this article have been better if it had been written by someone with a scientific background? A professor of business would hardly seem to have the necessary expertise. Heck, I certainly don't, and I don't mean to imply otherwise. But just checking the links provided in the article itself gives me serious reservations about the whole thing. Why didn't Stross see that?

Of course, a newspaper article scaring people about cellphones and cancer is always going to be a lot more popular - with readers and with the publishing company - than one that's less sensational. And the entire impetus for the article came from the woman pushing her new book. Without that, there wouldn't have been an article.

I don't know anything about this, myself. I'm definitely no expert. But as we've clearly seen in the vaccine-autism scare, some people will never be convinced about something like this, no matter how much research shows that their concerns are unfounded. That's just human nature. If you get a brain tumor - or if your kid has autism - you want to find an explanation (reasonably enough) and, if possible, someone to blame. And it is very hard to accept that, often enough, we simply don't know why these bad things happen to good people.

Also, it's undeniable that corporations will fight long and hard against anything that could affect their bottom line. Tobacco companies did everything they could to deny that their product was unhealthy, or at least to muddy the waters for awhile. And oil companies are doing the same thing today about global warming. But that doesn't mean that all corporate products are killing us, either. What we need to do in all of these cases - tobacco, global warming, vaccines, and cellphones - is to look for the scientific consensus.

The scientific consensus can be wrong, but it's far more likely to be right than the opinion of some corporate PR flack, lobbyist, politician, concerned parent, or even the minority scientific view. And if it is wrong, scientists are likely to be the first to discover that, and the consensus will change. For us laymen, believing anything other than the scientific consensus is just believing what we want to believe, whether it's that smoking is harmless or that cellphone use is dangerous.

There's nothing wrong with being careful. If you want to limit cellphone use - or hold the phone away from your head, as this article recommends - go ahead. But I'm quite disappointed in the New York Times for publishing an article like this, one which seems to be poorly researched and deliberately sensational. Next time, why not hire an author with a scientific background?

PS. While writing this, I stumbled upon a blog post from Ask Santosh on the same topic, though focusing on a study from Denmark:

A study published today by the Journal of the National Cancer Institute should put chronic cell phone users at ease. Researchers in Denmark studied 420,000 cell phone users whose cases were followed for up to 21 years. The study focused on both men and women who began using cellular telephone service between 1982 and 1995, and followed them through 2002 for signs of cancer. The bottom line:

"We found no evidence for an association between tumor risk and cellular telephone use among either short-term or long-term users."

Researchers took advantage of the fact that Denmark maintains a national cancer registry, and correlated that to cell phone subscriber records. Statistically, they expected to find 15,000 incidences of cancer (brain tumors, acoustic neuromas, leukemias, etc.) in the group, but the number turned out to be less -- only 14,249 indicents of cancer were noted in the group.

Just to clarify, I don't mean to say they expected to find 15,000 cancers related to cell phones... I'm saying that based on medical stats they knew that in ANY group of 400,000 people, they could expect to find that many cases of cancer. If there was a definitive link between cell phones and cancer, you'd expect the numbers for this group, comprised ONLY of cell phone users, to be much higher.

But contrary to the widely-held suspicion that cell phone usage causes cancer, this study indicates a lower rate of cancer among long-term cell phone users. The study also found no link between the side of the head on which brain cancers occurred and the side on which the cellular phone was used. If you want more details on the study, you can see the facts and figures for yourself.

Note that this study did not have to rely on people remembering past cellphone use, since the researchers correlated Denmark's national cancer registry with cellphone subscriber records. That would seem to solve the bias problem mentioned above, don't you think?

Santosh points out other studies which have come to the same conclusion. It's an excellent post. Check it out for yourself.