NOTE: Orac is on semi-vacation this week, trying very hard to recharge his Tarial cells. Actually, although he is at home, he is spending much of his time in his Sanctum Sanctorum (i.e., his home office) working on an R01 for the February submission cycle. Given that the week between Christmas and New Years Day tends to be pretty boring, both from a blogging and blog traffic standpoint, he’s scaling back the new, original stuff and mixing in some “best of” reruns, as well as some more recent stuff that appeared in a different form elsewhere, modified a bit to be more appropriate to this blog. Why? Because he likes them enough that he wants to make sure that as many people as possible see them.
Fear not, though. If something sufficiently interesting happens, Orac will welcome a break from the drudgery of grant writing to apply some appropriate not-so-Respectful Insolence to it. (Yes, I’m referring to Deepak Chopra, who, I’m told, returned over the weekend after Christmas, surpassing even his own usual level of vitriol towards “skeptics.”)
It’s been about a year and a half since I’ve written about this topic; so I thought I’d better update the disclaimer that I wrote at the beginning:
Before I start into the meat of this post, I feel the need to emphasize, as strongly as I can, four things:
- I do not receive any funding from the telecommunications industry in general, or wireless phone companies in particular. None at all. In other words, I’m not in the pocket of “big mobile” any more than I am in the pocket of big pharma.
- I don’t own any stock in telecommunications companies, other than as parts of mutual funds in which my retirement funds are invested that purchase shares in many, many different companies, some of which may or may not be telecommunications companies.
- None of my friends or family work for cell phone companies.
- I don’t have a dog in this hunt. I really don’t.
There. That’s better. Hopefully that will, as it did last time, serve as a shield against the “shill” argument, which is among the frequent accusations I hear whenever I venture into this particular topic area. So, as I did back in 2008, I just thought I’d clear that up right away in order (hopefully) to preempt any similar comments after this post. Unfortunately, as I have known for a long time, I’m sure someone will probably show his or her lack of reading comprehension and post one of those very criticisms of me. It’s almost inevitable, either here or elsewhere. Posting such disclaimers never seems to work against the “pharma shill” gambit when I write about vaccines or dubious cancer cures. Even so, even after nearly ten years involved in skepticism and promoting science-based medicine, hope still springs eternal.
There are two reasons that I think the issue of mobile phones and cancer needs an update on our blog: First, it has been a year and a half since I last wrote about it. At that time I castigated Dr. Ronald B. Herberman, who at that time was director of the University of Pittsburgh Cancer Institute for what I viewed as fear mongering over cell phones and cancer based on at best flimsy evidence. Second, there have been two fairly high profile studies looking at whether there is a link between mobile phone use and cancer. One of these my colleague and friend Steve Novella has already discussed, but there was another one that he didn’t see because it didn’t get quite as much publicity, possibly because the corresponding author is based in Korea. I will take this opportunity to discuss them both.
Note: Given that Orac is on vacation (sort of) this week (in reality I’m spending a lot of time in my home office writing a grant), posting this week will be a mix of the old and new. I’ll be reposting some Classic Insolence reruns interspersed with some new stuff as the mood strikes me. Full regular posting will resume on January 4. In any case, this particular post you may have seen in a different form in a different place, but if you haven’t, it’s new to you! I like it enough that I had to make sure that it appears here too.
Over the last several years, as cell phones have become not only ubiquitous but have morphed into in essence pocket computers with Internet connections, there has developed a cottage industry of cell phone “shields” that allegedly protect people from horrible microwaves emitted by cell phones that supposedly cause cancer. These shields, of course, do nothing of the sort; objective tests of many of them show that they don’t even do a good job of blocking electromagnetic radiation emitted by cell phones. However, the data looking at the question of whether cell phones can cause some form of cancer has been mixed at best and consistent with no detectable association or possibly a very tiny association that just rises above the background noise, but even that is arguable.
One reason we have a lot of doubts over whether cell phone radiation can actually cause cancer goes back to a longtime focus of this blog and one reason why I try to emphasize science-based medicine rather than evidence-based medicine, namely scientific plausibility. From a biological standpoint, a strong link between cell phone use and brain cancer (or any other cancer) is not very plausible at all; in fact, it’s highly implausible. Cell phones do not emit ionizing radiation; they emit electromagnetic radiation in the microwave spectrum whose energy is far too low to cause the DNA damage that leads to mutations that lead to cancer. True, it is possible that perhaps heating effects might contribute somehow to cancer, but most cell phones, at least ones manufactured in the last decade or so, are low power radio transmitters. It is also possible that there is an as yet undiscovered biological mechanism by which low power radio waves can cause cancer, perhaps epigenetic or other, but the evidence there is very weak to nonexistent as well. Worse, as has been pointed out many times, epidemiological evidence for people who have used cell phones heavily for more than 10 years is sparse.
Moreover, it’s not possible to study the issue by randomized studies, because it is impractical to the point of being virtually impossible to randomize people into groups that do and do not use cell phones given how essential cell phones have become to most people in industrialized nations and then to follow them for the 20 years or so that it would take to identify a link. That leaves retrospective data, with all the perils and pitfalls inherent in retrospective studies. One potential approach to such a study is to formulate a simple hypothesis. If mobile phone use causes cancer, then after the widespread introduction of cell phones into a population there should be a detectable significant increase in the incidence of tumors that could potentially be due to cell phone use, particularly 10 to 20 years later. That is the hypothesis that the investigators who published recent Danish study (Deltour et al, Time Trends in Brain Tumor Incidence Rates in Denmark, Finland, Norway, and Sweden, 1974-2003) decided to examine. It is a good population to examine this question in because the populations of these countries is homogeneous, cell phone use became widespread earlier than it did in the U.S., and these countries have nationalized health systems that allow centralized collection of cancer data in national cancer registries.
The basic design of the study was as follows. Investigators examined the cancer registries of these nations for the first incidence of brain tumors in patients aged 20-79, noting this background:
Previous investigations in Denmark, Finland, Norway, and Sweden found that the incidence of glioma was relatively stable from 1983 to 1998 ( 7 ) and that the incidence of meningioma increased from 1968 to 1997, more so for women than for men (8). Time trends in brain tumor incidence after 1998 are likely to be relevant for evaluating possible associations with respect to radio frequency exposure from mobile phones after 5 – 10 years of exposure. We investigated time trends in brain tumor incidence rates in these four Nordic countries to evaluate whether trends in the incidence of brain tumor changed in Denmark, Finland, Norway, and Sweden from 1998 to 2003.
This information is important to know, because if brain tumor incidence were rising before the widespread use of cell phones, then to find a correlation that might indicate causation, there would have to be a more rapid increase in brain cancer, starting with an appropriate lag time after the use of cell phones became so prevalent, likely at least five to ten years. This is very much like the evidence for an epidemiological link between smoking tobacco and lung cancer, except that for the tobacco-lung cancer link there was a 20-30 year lag between the introduction of inexpensive, mass-marketed cigarettes and the increase in lung cancer incidence. In any case, the resultant population examined in this study was, in essence, the entire populations of these four countries, where 59,984 brain tumor cases that were diagnosed from 1974 to 2003 among 16 million adults aged 20 -79 years. They say that a picture is worth a thousand words; so here is Figure 1 from the paper:
Panels A and B present incidence rates for gliomas and meningiomas, respectively, in men; panels C and D present incidence rates for gliomas and meningiomas, respectively, in women. Circles indicate rates for those aged 20 – 39 years, squares indicate rates for those aged 40-59 years, triangles indicate rates for those aged 60-79 years, and a solid line indicates the regression curve. As described in the paper:
During this time, the incidence rate of cancers known as gliomas increased gradually by 0.5% per year among men and by 0.2% per year among women.
For cancers known as meningioma, the incidence rate increased by 0.8% among men and, after the early 1990′s, by 3.8% among women.
This more rapid change for women was driven, the researchers say, by the 60-79 year age group.
In other words, there was a slow rate of increase in these tumors that did not change in the 1998-2003 cohort. The reasons for this slow rate of increase are unclear, but because it began before the widespread insinuation of mobile phones into the population it is almost certainly not due to mobile phone radiation. One possibility that could have contributed to this is the increasing use and sophistication of imaging technology like CT and MRI, which, as the authors pointed out in the introduction, can lead to an increased apparent incidence without any changes in etiological factors through the detection of asymptomatic meningioma, for example. (Sound familiar?) In any case, there was no detectable evidence of an uptick in the incidence of these brain tumors after 1998. The lack of evidence for a change in the rate of increase of these tumors is consistent with three conclusions. Either there is no link between cell phone use and these brain tumors; the “lag time” for such an effect is greater than 5-10 years; or the effect is too small to be detected in an overall population level. One area where I will disagree with our fearless leader (somewhat) is that I don’t consider this study to be weak evidence. It is, in fact, strong evidence that, if a link between cell phone use and brain cancer exists, it is almost certainly weak and small. Add to that the biological implausibility of a link, given our current knowledge about cancer, and I remain less than impressed with the claims that cell phones cause brain cancer. I would be happy to change my mind if new evidence, either in the form of a biological mechanism being discovered that could explain how long term exposure to low energy radio signals could cause cancer or epidemiological evidence showing a clear association between cell phone use and cancer (preferably both), came to light.
Unfortunately, the second study does not qualify as either form of evidence. (How’s that for a segue?) In fact, from my perspective, it is one of the best examples of how meta-analyses can be tortured to find tenuous correlations where none probably exist. The study, which appeared in the November 20 issue of the Journal of Clinical Oncology, apparently slipped by my notice when it first appeared as an online publication on October 13. In any case, the study (Myung et al, Mobile Phone Use and Risk of Tumors: A Meta-Analysis) is, as the title says, a meta-analysis, meaning it’s a formal way of combining multiple studies that may or may not have statistically significant results on their own, that may or may not have found a correlation, and trying to see if the weight of the evidence suggests a correlation between mobile phone use and cancer. In this case, 465 articles were winnowed down to 22 articles using specific selection criteria, which were:
We included epidemiologic studies that met all of the following criteria: case-control study (to date, no randomized controlled trials and only one retrospective cohort study published in four different articles have been reported; therefore, we included only case-control studies in this study); investigated the associations between the use of mobile phones, cellular phones, or cordless phones and malignant or benign tumors; reported outcome measures with adjusted odds ratios and 95% CIs, crude odds ratios and 95% CIs, or values in cells of a 2(1)2 table (from which odds ratios could be calculated). If data were duplicated or shared in more than one study, the first published or more comprehensive study was included in the analysis.
I’ll cut to the chase right now. This meta-analysis does not–I repeat, does not–show any correlation between cell phone use and cancer, at least not in the overall results. The authors even say so:
As shown in Figure 2, the overall use of mobile phones (use v never or rarely use) was not significantly associated with the risk of tumors in a random-effects model meta-analysis of all 23 case-control studies (odds ratio 0.98; 95% CI, 0.89 to 1.07).
Let me repeat that again. The overall meta-analysis did not find any significant association between mobile phone use and tumors.
Of course, whenever an investigator does a meta-analysis and finds a result like this, he can never rest there. He has to slice and dice the data to try to find a group for which there is a correlation. There’s nothing wrong with that in and of itself. In prospective studies, post hoc subgroup analyses that were not planned in the original protocol right from the beginning are generally frowned upon because they have a high tendency to find associations that are usually spurious. I tend to look at it as being a lot like making multiple comparisons but not controlling for them. By doing a study to look at one population and a set of outcome measures but then, upon finding a negative result, going back and doing subgroup analyses, one is basically doing multiple comparisons without correcting for multiple comparisons. Anyone who’s been a regular reader of this blog should know that if one doesn’t correct for multiple comparisons, the more comparisons the higher the likelihood of finding one or more false positives.
Even so, it’s not necessarily scientifically dubious to do subgroup analyses if one looks at it as a hypothesis-generating exercise, rather than any actual conclusions. Correlations that are found may or may not be “real,” but strong correlations may be worth further investigation. What bothers me about this study is not so much that it did subgroup analysis on the populations studied, but rather how it did a sort of dubious “subgroup analysis” on the actual studies themselves:
However, a significant positive association (ie, harmful effect) was observed in eight studies 7,12,14-16,18,23 and one study by another group10) using blinding (odds ratio1.17; 95% CI, 1.02 to 1.36), whereas a significant negative association (ie, protective effect) was observed in 15 studies (nine INTERPHONE-related studies17,20-22,24-28 and six studies by other groups8,9,11,13,19) not using blinding (odds ratio 0.85; 95% CI, 0.80 to 0.91). No publication bias was observed in the selected studies (Begg’s funnel plot was symmetric; Egger’s test, P for bias .21; Fig 3)
Subgroup meta-analyses by methodologic quality of study revealed a significant positive association in the high-quality studies (odds ratio 1.09; 95% CI, 1.01 to 1.18), whereas a negative association was observed in the low-quality studies. In subgroup meta-analyses by malignancy of tumor, no significant association was observed for malignant tumors. However, a significant negative association was observed for benign tumors. Neither the use of analog phones nor the use of digital phones was associated with the risk of tumors.
First, note that these odds ratios are barely statistically significant, ranging from 1.02 to 1.36 and 1.01 to 1.18; in other words, the 95% confidence interval barely misses overlapping with 1.0. More importantly, I was very puzzled by the way that they chose to differentiate “high” quality studies from “low” quality studies. Basically, although it is only one of the eight criteria used in the Newcastle-Ottawa Scale (NOS) for Assessing the Quality of Case-Control Studies to evaluate study quality, for some reason the Myung et al decided to focus primarily on whether the studies were blinded or not, specifically whether the status of patient cases and controls was blinded at interview (blinded or not blinded/no description). This struck me as most curious. So studies that were blinded showed a positive association between cell phone use and brain tumors upon meta-analysis of just their results. The authors also report that their subgroup analysis demonstrated that studies of higher methodological quality (greater than or equal 7 points) also showed a positive association between cell phone use and cancer for just these studies.
So does it mean anything that lumping the studies together that are of “high quality” produces a positive result where the low quality studies don’t? Normally, I’d think that it might. After all, the studies showing the positive result are considered to be methodologically rigorous, although it should be noted that several of them individually didn’t find a significant association. In this case, I doubt it means much of anything, and here’s why. Seven of the eight “high quality” studies were all by a single group of researchers, Dr. Lennart Hardell’s group in Sweden. Whenever one group of researchers keeps finding a result that no other group seems able to replicate or that otherwise disagrees with what everyone else is finding, that’s a huge red flag for me. Remove those studies, and even the wisp of a hint of a shadow of the association between cell phone use and cancer found in this study disappears. I’d have a lot more confidence in this seeming association in “high quality” studies if the association didn’t depend upon a single researcher and if this researcher was not also known for being an expert witness in lawsuits against mobile phone companies. Don’t get me wrong; these studies could be correct, but replication is one of the checks on research in science-based medicine. If other groups can’t replicate Dr. Hardell’s work, I wonder why. Is it something about the methodology? Is it something specific about Swedes? Is it something about the population? Are there confounders unique to his work that aren’t operative elsewhere? Until I see other researchers independent of Dr. Hardell and using a variety of different analyses find the same results, I don’t have a lot of confidence in them. Even the authors acknowledge that a weakness of their study is that “we did not explore potential confounding factors in the studies by Hardell et al that reported positive results not found by other study groups.”
There was another passage in this paper that I also found most curious:
We feel the need to mention the funding sources for each research group because it is possible that these may have influenced the respective study designs and results. According to the acknowledgments that appeared in the publications, the Hardell et al group was supported by grants from the Swedish Work Environment Fund, Orebro Cancer Fund, Orebro University Hospital Cancer Fund, and so on. Most of the INTERPHONE-related studies were mainly supported by the Quality of Life and Management of Living Resources program of the European Union and the International Union Against Cancer; the International Union Against Cancer received funds for those studies from the Mobile Manufacturers Forum and the Global System for Mobile Communication Association.
I don’t know about you, but I’ve never seen a passage like this in a research paper or meta-analysis before. It stands out like the proverbial sore thumb, and I can only think that it was placed there to try to cast aspersions, representing Dr. Hardell as the exemplar of Truth, Justice, and Science while painting the studies that failed to find an association between cell phone use and cancer as hopelessly biased, the product of Big Cellular. At least that’s how it came across to me. It struck me as gratuitous.
So where do we stand now? My interpretation of the evidence thus far is that we can say with some confidence that there is no short term risk of brain cancer from cell phone use. However, after more than ten years the evidence is less clear but trends towards either no detectable risk or a very small risk that barely rises above the noise. It’s possible that there may well be a risk; it’s possible that there is hitherto undiscovered biology that provides a mechanism by which non-ionizing radiation like the radio waves from cell phones could over time induce cancer beyond ten years. Because more and more people are using cell phones over longer and longer periods of time, it’s worth studying this issue. If there is an increased risk, it may be possible to mitigate that risk by using headsets or through the design of phones that use less energy. Mobile phone technology is a relatively new technology, though, and has only been widely available since the mid-1990s. In the U.S., it didn’t become truly ubiquitous until the early part of this decade. Consequently, there hasn’t been enough lag time for us to be truly confident of studies showing little or no risk. On the other hand, it is reassuring that early studies are pretty resoundingly negative and that there really is no good biological mechanism that we have been able to find by which cell phone radiation could cause cancer. In the next 5-10 years, more studies will be done, and, over that time, I expect evidence will answer the question one way or the other.
In the meantime, it’s useful to keep things in perspective. Each and every year, there are approximately 40,000 to 45,000 fatalities due to auto collisions. Getting in a car and driving to work every day is among the most dangerous things the average person does. Yet we accept this risk because automobiles are such an incredibly useful tool in modern life; indeed, they are indispensable to most people. Cell phones are clearly in that category as well; so even if cell phones are ultimately found to increase the risk of brain cancer by, for example, 10-20% after 10-20 years, it would most likely be a risk that most people would end up living with in order to be able to use these incredibly useful devices. My personal reading of the data is that there probably isn’t a significant risk of brain tumors due to the use of cell phones, but I am not sufficiently certain to make any blanket statements. I, like everyone else, will have to wait for the evidence to settle things one way or the other.
Myung, S., Ju, W., McDonnell, D., Lee, Y., Kazinets, G., Cheng, C., & Moskowitz, J. (2009). Mobile Phone Use and Risk of Tumors: A Meta-Analysis Journal of Clinical Oncology, 27 (33), 5565-5572 DOI: 10.1200/JCO.2008.21.6366
Deltour, I., Johansen, C., Auvinen, A., Feychting, M., Klaeboe, L., & Schuz, J. (2009). Time Trends in Brain Tumor Incidence Rates in Denmark, Finland, Norway, and Sweden, 1974-2003 JNCI Journal of the National Cancer Institute DOI: 10.1093/jnci/djp415