Mobile phones 'affect the brain'The paper's from Nora Volkow and colleagues from NIDA in the USA. Volkow's best known for her work on addiction.
47 people got 18FDG Positron Emission Tomography. This method measures brain glucose use as a proxy for how hard cells are working. They say that this makes it better than other kinds of PET which merely measure regional blood flow. I bet they really wanted to do this study with fMRI, because PET scans cost loads, but of course you can't take a cellphone into an MRI scanner.There were two conditions: a control in which they had a phone stuck to each ear but they were both off, and an active condition in which the right-ear phone was switched on and receiving a call - but muted so they couldn't hear anything. Each subject was scanned twice, once under each condition, so that's 94 scans.
What happened? In the Results section they say that (my emphasis):
SPM comparisons on the absolute metabolic measures showed significant increases (35.7 vs 33.3 µmol/100 g per minute for the on vs off conditions, respectively; mean difference, 2.4 [95% CI, 0.67-4.2]; P=.004) in a region that included the right orbitofrontal cortex and the lower part of the right superior temporal gyrus. No areas showed decreases.In other words a highly signficiant finding of increased glucose uptake in the areas of the brain closest to the cell phone. Whoa, that's big. However, it seems that this result was not corrected for multiple comparisons, because in the table of results they give the corrected p value for the activated cluster as p=0.05 - bang on exactly low enough to be considered significant, but no lower.
Their method for correcting for multiple comparisons was also quite unusual and I'm not quite sure what to make of it. It's on the right-hand column of Page 810. Maybe commentators will be able to offer some insight.
There's a few other things to note here. They show a nice big colorful This Is Your Brain On Phone image but it's a "representative" image of one brain, rather than an averaged image from all subjects. This is really not good practice. It's acceptable - but only because there's no alternative - for data which can't be averaged, like microscope pics.
With between-group comparisons of neuroimaging data, the averages are computed as part of the statistical analysis, and should be shown. With single-subject data we're left having to trust the authors to have really picked a "representative" image as opposed to "the best image".
Second, this has nothing to do with cancer. Brain activation happens all the time and very rarely does it have cancerous consequences. In fact this is so unrelated to cancer that I shouldn't even be mentioning cancer in this post. However, I feel the need to because the BBC (and most other outlets) did. Thus we saw curious paragraphs like this (direct quote):Since the boom in mobile phone use, there has been considerable interest in the effect on the body. The largest study on 420,000 mobile phone users in Denmark, has not shown a link between phone use and cancer. This small study on 47 people...Why mention cancer, if the only thing you say about it is that there's no link? Presumably because of the following chain of associations: cell phones use radiation...radiation causes cancer...cell phones and cancer!
I have no idea if cell phones cause cancer. Just from basic biology though, if they were going to cause any cancer, it'd probably be skin cancer rather than brain cancer, since a) they're closest to the skin, not the brain and b) brain cancer is incredibly rare because the brain contains no rapidly dividing cells, whereas skin cancer is common because skin is made of exactly that.
So even if if this increased brain glucose metabolism somehow was related to cancer of the brain, this would be the least of our worries, because if cell phones somehow caused brain cancer, they'd almost certainly cause many times more cases of skin cancer and the brain cancer would be a footnote.
But the point is, this study has nothing to do with cancer so forget I said that. If you have trouble forgetting, just hold your mobile phone over your temporal lobes until your hippocampus is overloaded and you suffer memory loss.
Link: Also blogged here and here.
Volkow, N., Tomasi, D., Wang, G., Vaska, P., Fowler, J., Telang, F., Alexoff, D., Logan, J., & Wong, C. (2011). Effects of Cell Phone Radiofrequency Signal Exposure on Brain Glucose Metabolism JAMA: The Journal of the American Medical Association, 305 (8), 808-813 DOI: 10.1001/jama.2011.186
17 comments:
Seems plausible, if you consider that an electric field might plausibly influence local field potential of the extracellular space around cells, which plausibly influences firing patterns, which, of course, affects cell function.
I live 3 km (as the crow flies) from NATO's main Southern Eastern observation post. The radiation the post pumps out could drive a train.
It's so hugely powerful it's impossible to use any kind of wireless device at a distance of more then 2 meters.
Cellphones hardly work here, electronic devices spontaneously do weird things.
Now with the ME issues, the post is running at full. If i stick a nail in an amplifier i capture radio, quite clearly.
I'm not the only one living here, obviously. About a 100.000 people live with a 20 km radius.
If ever there was a connection between EM radiation and cancer this should be the place.
But, nope. Cancers of any kind are not clustered around here.
Otoh, it does make your head spin at times. Lucid dreams are common amongst most people i talk to.
Neurosceptic: Nice detailed write-up. Thanks for the link to my post.
Cheers,
Arunn
Kevin: It's not entirely implausible, but it strikes me as unlikely. For one thing, if mobile phones affected brain function, why don't we see cellphone induced seizures in people with epilepsy? Just about everything else which stimulates the brain, from TMS to flashing lights, can trigger seizures.
That's not a knockdown argument because it might affect the brain in some subtle long-term fashion but still.
I don't care that much about cell phones, but I do care about multiple comparisons correction, so I took a look at that paragraph. I think it makes slightly more intuitive sense if you flip it around. Without correction, your alpha criterion is 0.05. For each cluster, multiply that alpha by Cv/Sv (i.e., the proportion of the search volume taken up by that cluster). The idea (presumably) being that if you have three clusters of different sizes, you can correct the alpha criterion (or their p values) in proportion to their sizes, and preserve the Bonferroni inequality.
I'm not sure how valid this is, but it makes a little more sense each time I read it. However, I've never really had a good intuitive understanding of how these cluster level statistics are calculated (pointers welcome).
The other oddity about that paragraph was the idea that they were only going to correct in some regions. I presume this is meant to convey that they didn't look elsewhere, but that's an odd way of putting it.
Focusing your investigation on certain areas of the brain is a common practice in neuroimaging, so I don't fault them for that. The more you can narrow down your search the better.
I agree with DanK on the multiple comparisons issue. At first glance it looks alright. Lots of people have used Bonferroni-based corrections to neuroimaging data before. I just haven't ever seen the cluster threshold put into the equation like that.
The cluster threshold is being used to make it easier to get significance, since you are dividing the search volume by that value. My only thought is that this assumes all the voxels are independent samples of data, which they are not. This is especially true in PET imaging which has a lower spatial resolution. They are being conservative by using the Bonferroni correction, but maybe being too liberal in backing it off with the cluster threshold.
Kevin: Electric fields definitely do influence cell firing patterns, though the studies I'm aware of have used stronger and lower-frequency fields. There was a Nature paper a few years ago where applied fields were used to amplify slow-wave sleep oscillations in humans, leading to improved declarative memory (Marshall et al. 2006). And transcranial direct current stimulation has been used for several decades for treating a variety of disorders. It is surprising that cell phone EMFs would affect the brain, however, because they're so small and high-frequency.
Neuroskeptic: I don't think seizures are relevant since cell phone EMFs are so small. Under normal conditions, tDCS uses fields too low in magnitude for inducing seizures, but still modulates neuronal function, either long-term (by affecting plastic changes) or short-term (oscillating fields can affect synchrony). And cell phones should theoretically have an even smaller effect than normal experimental tDCS, especially since radio frequencies should be too high-frequency to affect syncrony.
Would have been interesting to have half the subjects hold a dead salmon next to their ear...
The most arcane correction for multiple comparisons I've ever seen. Highly unconvincing - I can't think of a reason why they wouldn't use conventional methods for such a bombshell of a paper, other than that traditional methods didn't "work".
One thing that's certain is that cell phones don't produce ionising radiation, which is the cancer big bad guy. The cell phone photons have enough energy to shake a molecule but not to rip bond asunder.
So, if cell phones causes cancer, it must be by some other much weaker mechanism, something like sustained neuron activation which might cause a bit of cell damage by fatigue, or some downstream hormonal effect, or something, who knows. Studies to date put an limit on it as a pretty weak effect, either negative or positive, but lost in the noise, probably on par with something like excessive chess analysis or too much facebooking...
Yeah I noticed the image as well. I mean it's a 7% difference from control! There is no way that image is representative. Then again, it probably makes for much better press than the composite image.
Anonymous 2:26 - It does seem arcane. They used SPM2 to do their analysis. SPM2 is ancient (there's since been 5 and 8) but it still incorporated standard Gaussian random field theory based cluster correction.
I don't know if it would allow them to "correct over a limited area" as they wanted, though.
The potentially most convincing argument is the purported relationship between field strength and change in activity, but:
I didn't see whether they actually checked the assumptions that go into linear regression, which is important given how small the effect was.
I think the most suspect thing about the study was that...they just used cell phones. That's fine, but if you're positing a relationship between the field intensity and the brain activity, you should further investigate that by varying the intensity/frequency, whatever with a device producing emf in a similar range as that of cell phones.
If they could show that changes in frequency and intensity of the field produced by such a device reliably changed activity, then I would be more excited. Just using a cell phone reeks of High Splash Science.
Another piece that seems to be missing is that they didn't test whether turning on the _left_ phone caused increased activity in the _left_ side of the brain. That would be strong evidence for an effect caused strictly by physical proximity. Otherwise, IMO they haven't ruled out some sort of cognitive effect: if phones vibrate or get slightly warmer when in use, subjects may have been (perhaps without being aware of it) trying to "listen."
Maybe that isn't likely, but the check needs to be done.
Sat phones are quite a bit stronger then cell phones. Query if they have more of an impact?
Also, one wonders if some of the effect of the cell phone being on as distinct from reactions to a call is anticipatory, like a Pavlov reaction. Your ear expects to leap into action when the cell phone is near and may be "on alert."
Sat phones are quite a bit stronger then cell phones. Query if they have more of an impact?
Also, one wonders if some of the effect of the cell phone being on as distinct from reactions to a call is anticipatory, like a Pavlov reaction. Your ear expects to leap into action when the cell phone is near and may be "on alert."
The most plausible explanation for this result -assuming it is not a statistical artifact - is simply the heat from the cell phone. The cell phone was active for 50 minutes prior to the measurement. During this time the phone will heat up considerably. This heat can be felt, and might well cause a slight change in metabolism. For a more complete analysis on this study and the subject of EMF & health, see my web site at www.emfandhealth.com
Post a Comment