
CC BY-NC-ND image from Arthaey via Flickr
Last week, I took a look at the WHO’s announcement classifying radiofrequency electromagnetic fields, such as those emitted by mobile phones, as being possibly carcinogenic to humans, and dubbed it a bad call. Their decision was announced via a press release, which stated that an article would go live “in a few days” on The Lancet Oncology website that would give a concise report of their findings. This has yet to surface at the time of this writing, which is incredibly disappointing – engaging a sensationalist and fear-loving media with an international health policy statement without disclosing the specific evidence-based thought process behind it is poor planning at best, and at worst, implies that the thought process behind it may be suspect.
This leaves us with only the press release to work with, which notes that the decision was based on an extensive literature review. That said, the release contains only a single piece of hard data from one paper: a statistic from the INTERPHONE study which stated that there was a “40% increased risk for gliomas in the highest category of heavy users (reported average: 30 minutes per day over a 10‐year period)”. That seems pretty impressive and dramatic, as I’m sure the author of the press release intended it to be.
It’s also wrong.
It turns out that the correct way to portray the finding the release is referencing is actually a bit confusing, and somewhat unclear from the INTERPHONE data. It’s also interesting to note that most media and blogs reported INTERPHONE to show no correlation overall between mobile phone use and brain cancer. How, then, are we to reconcile that with the above statistic? It turns out INTERPHONE did show an overall correlation of tumors with phone use – a negative one. They found a reduced risk for gliomas in people who had ever used a mobile phone overall, including a reduced risk for gliomas in the second highest category of heavy phone users. Yet, we don’t see anyone making the claim that moderate mobile phone use is somehow protective against brain cancer.
There are several reasons for these confusing results. Firstly, there is no plausible way that radiofrequency electromagentic fields could cause cancer in the first place, so one would an expect a study examining this question to give unclear results at best. Also, INTERPHONE has several sources of error and bias that the authors of the study discuss in detail, and as such caution that these “limit the strength of the conclusions we can draw from these analyses and prevent a causal interpretation”. As such, INTERPHONE provides an excellent lesson in evidence-based medicine, as well as the truth behind the Mark Twain-popularized truism, “There are three kinds of lies: lies, damned lies, and statistics.”
What is the INTERPHONE study? INTERPHONE is the largest case control study every done looking at mobile phone use and brain tumors. It was conducted from 2000 through 2004 in 13 countries: Australia, Canada, Denmark, Finland, France, Germany, Israel, Italy, Japan, New Zealand, Norway, Sweden, and the United Kingdom. The study focused on two specific types of brain tumors – gliomas and meningiomas. Patients diagnosed with either of these tumors during the study period were interviewed in detail about their prior mobile phone use habits, as well as other possible exposures to radiation. This is a case-control study, meaning that a control group was created by “matching” at least one cancer-free person to each cancer patient based on having otherwise similar demographics. Overall, 2425 persons with meningiomas, 2765 persons with gilomas, and 7658 controls were included.
What does it show? The survey data from the study participants was analyzed to see if there was any correlation between having either a meningioma or glioma and different parameters of mobile phone use – such as number of years using a mobile phone, or total amount of calls made during one’s lifetime – as well as the anatomic location of the tumor. A statistic called an Odds Ratio (OR) was calculated for each comparison, which is standard for a case control study. An OR compares the odds of one thing divided by the odds of another; in this case, the odds of having a brain tumor if one has ever used a mobile phone divided by the odds of having a brain tumor if one has never used a mobile phone.
An OR of 1.0 means that there is no difference between the two groups, whereas an OR above 1.0 would represent an increased risk for a tumor, and an OR of less than 1.0 shows a decreased risk. However, it is important to note that the OR gives a ratio of the odds that a thing will happen, and does not give a direct percentage chance of increased or decreased risk. To determine that, one would need to use a different statistical technique than is used in case control studies, such as the Relative Risk.
In INTERPHONE, the overall OR’s for a meningioma and glioma being associated with using a mobile phone for greater than one year were 0.79 and 0.81 respectively. This means that the odds of having a brain tumor if one has ever used a mobile phone for more than a year were actually less than the odds of having a brain tumor if one had not used a mobile phone. The authors then broke down the patients into ten subgroups based on number of hours spent on the phone, and found that the vast majority of OR’s were less than 1.0, again showing a decreased risk. In fact, the ninth (second highest) category for cumulative call time showed the lowest OR’s in this breakdown – 0.71 for gliomas and 0.76 for meningiomas. The tenth, or highest cumulative call time, category is the one that showed showed elevated OR’s of 1.15 for meningiomas and 1.4 for gilomas.
This latter OR of 1.40 is where the press release is getting the 40% increase statistic, or so it seems. If that is the case, this is an incorrect reporting of the Odds Ratio, as 1.4 does not mean a 40% increase. This is a common error that is made when reporting on OR’s, because they are often confused with the other statistic mentioned above – the Relative Risk. If a Relative Risk was calculated to be 1.4, then it would be correct to say that there would be a 40% increased risk, because Relative Risk directly compares two probabilities, not two sets of odds.
So what is the the direct percentage increased risk of gilomas in that group, or any of the groups for that matter? Simply put, the data needed to calculate that is not in the INTERPHONE paper.
In addition to the above, the investigators further broke the data down to see if there was a relationship between tumor location in the brain with amount of phone use. Again, they found reduced OR’s overall and in most categories, although OR’s above 1.0 were seen in the highest use category both for tumors being on the same side of the head as phone use and in the temporal lobe of the brain.
What are the strengths? INTERPHONE managed to include large numbers of patients with meningiomas and gliomas. They focused on younger patients in the age range of 30-59 years old, as these were most likely to have used mobile phones over the prior 5-10 years. They also focused on urban areas, which also have the longest and greatest amount of mobile phone usage.
What are the weaknesses? The authors discussed a large number of possible sources of bias in the study itself. Many patients diagnosed with brain tumors and asked to participate in the study refused for a number of reasons; only 64% of glioma patients they identified agreed to participate, for instance. Being unable to include one third of glioma patients in the study could easily have impacted any associations seen. Not every established confounding factor was controlled for – for example, higher socio-economic status has been implicated by other studies as being a associated with an increased risk of brain cancer, but INTERPHONE did not record participant economic status in a rigorous way. Also, the timing and nature of the interviews of participants was variable.
It was further noted that the glioma group had a greater amount of potential sources of bias, especially for those in the higher use category for mobile phones. That group had a higher proportion of someone responding to the survey on behalf of the patient instead of the patients themselves, which could have led to greater inaccuracies. Also, that group had a much greater amount of missing data from the interviews, for which a “best guess” was made by the investigators and plugged into the calculations. Also at play is the normal human condition of recall bias and subjective memory, especially in the context of being asked detailed questions about things that happened several years prior. It is also worth noting that depending on its location and the clinical circumstances, a brain tumor and the treatment thereof might interfere with memory and the interview process.
What conclusions can we draw from it? Given the amount of potential error and bias involved, there are very few definite conclusions to be had. None of the OR’s were very low or very high, and their distribution is clearly confusing. Certainly there’s no reason to suspect that mobile phones somehow reduce the risk of gliomas or other brain cancer, just as there’s no plausible reason that the radiation they emit should cause brain cancer in the first place.
The one obvious conclusion is that there was no strong correlation between mobile phone use and brain cancer seen in the study. Even with the methodological issues involved, a strong correlation between phones and cancer should have been seen if it was there, and it was not. The authors themselves said it best:
An exhaustive analysis of this large data set involved estimation of hundreds of ORs; rather than focus on the most extreme values, the interpretation should rest on the overall balance of evidence. The null hypothesis of no association would be expected to produce an approximately symmetric pattern of negative and positive log ORs. A skewed distribution could be due to a bias or to a true effect. Our results include not only a disproportionately high number of ORs <1, but also a small number of elevated ORs. This could be taken to indicate an underlying lack of association with mobile phone use, systematic bias from one or more sources, a few random but essentially meaningless increased ORs, or a small effect detectable only in a subset of the data.
Overall, no increase in risk of either glioma or meningioma was observed in association with use of mobile phones. There were suggestions of an increased risk of glioma, and much less so meningioma, at the highest exposure levels, for ipsilateral exposures and, for glioma, for tumours in the temporal lobe. However, biases and errors limit the strength of the conclusions we can draw from these analyses and prevent a causal interpretation. (emphasis mine)
This brings us back to the WHO’s decision, which in theory is based on all the medical evidence, inclusive of INTERPHONE. The majority of the studies done, including one other large cohort study involving over 420,000 patients in Denmark, show no association of brain cancer with mobile phone use. A minority of studies have seemed to show an association, but their results are not strong, and the methods they used questionable. Overall, neither INTERPHONE itself nor the sum total of studies done to date makes a case for even a possible association of mobile phone use and brain cancer of any type from an evidence-based medicine perspective. There is no scientifically plausible way that radiofrequency electromagnetic fields can cause brain cancer that we know of, let alone protect against it, from a science-based medicine viewpoint.
I have previously stated that the WHO does not do science. Yes, they count scientists among their number, and yes, they are involved in scientific research – the WHO’s IARC was a big part of the INTERPHONE study, for example. That noted, their primary mission, i.e. the thing that they do, is to shape international health policy and procedures. Said policies should be based on rational appraisal of scientific evidence, and that is why I am incredibly disappointed in last week’s announcement and the use of an incorrect, wrongly reported “shock statistic” in their press release. It may be that they have considered unpublished data in their review, that the public does not yet have access to. If so, however, this would constitute an example of the loathsome practice of “science by press conference”, in order to appeal in a dramatic way to the public and the media.
As such, the WHO’s classification of radiofrequency electromagnetic fields as being a Grade 2B possible human carcinogen is not a mere scientific statement, but one of public health policy with far reaching implications. These include for individuals, whom the WHO is counseling to use hands-free alternatives to standard phone use, as well as economic ones for the mobile phone industry and the stock market. All of this would be fine if there was credible evidence of even the possibility of a correlation, but there is not, at least in the public domain. If there is credible evidence, then it is both confusing and concerning that it was not published prior to the WHO engaging the media with a press release. It will be interesting to see what the WHO’s thought process and evidence base was for this decision once their article in The Lancet Oncology comes to press.
Reference: The INTERPHONE Study Group. Int. J. Epidemiol. (2010) 39 (3): 675-694. doi:10.1093/ije/dyq079 (2010).