|» Printer friendly version
Brain tumours and EMFs
Health index » Overview | Childhood leukaemia | Brain tumours | Electromagnetic hypersensitivity | Other health effects
Please note that this page is now available in Russian.
There has been repeated coverage in the last few years that mobile phones will increase your chances of getting both malignant and benign forms of brain tumours. There has been vehement criticism of this coverage, claiming that there is no scientific evidence to support such an association, and that reporting a link is inappropriate at best, irresponsible scaremongering at worst.
In actual fact, there is now a large amount of epidemiological literature assessing the risk of mobile phone usage and brain cancer. In reality we would not expect studies to have found a link at this point in time, as the latency period (time between exposure to cause and diagnosis of cancer) of most brain tumours is between 15 and 25 years, and mobile phones have only been in widespread usage for about 10 years.
Despite this, a number of papers are now showing signs of a significantly increased risk of brain cancer incidence from long term usage (over 10 years) have now been published:
Lennart Hardell and Swedish Research
Oncologist Lennart Hardell and his colleagues has been researching this association for about 10 years and has published numerous papers covering their findings. Starting with preliminary work at the turn of the century where they found a statistically significant 1.4-fold increase in risk for brain tumours (not sub-categorised) on the same side of the head as the mobile phone was used[Hardell 1999, Hardell 2000].
In 2002, looking at 1617 patients histopathologically diagnosed with brain tumours, they found that use of analogue mobile phones was associated with a significant 30% increase in risk for brain tumours (overall). This increased to 80% when only looking at patients who had used their phone for over 10 years, and further to 150% (2.5 times as likely to develop a brain tumour) when side of the head was taken into account. For acoustic neuromas, the increase in risk was 250% [Hardell 2002]. This was followed in 2003 by a separate analysis that found that Astrocytomas also had a significantly increased risk of 80%[Hardell Feb 2003], and a subsequent paper finding a 3.5-fold risk of Acoustic Neuroma from mobile phone use (CI 1.77-6.76)[Hardell Mar 2003]
By 2005 most papers generally sub-categorised brain tumour risk from mobile phone usage into acoustic neuromas, meningiomas and gliomas. Hardell produced a paper in 2005 analysing the increase in risk for acoustic neuromas and meningiomas. With a very good response rate (85%) he found that, for a mobile phone usage of greater than 10 years, the odds ratio for meningiomas was 2.1 (CI 1.1-4.3) and for acoustic neuromas this was split further into digital mobile phones (2.0 - CI 1.05-3.8) and analogue mobile phones (4.2 - CI 1.8-10)[Hardell 2005]. By this stage his results were consistently pointing towards a possible doubling in risk, and this for mobile phone users who had not used their phone for as long as the typical latency period for the tumours! Hardell also published one of the only papers to date looking at risk of non-hodgkin's lymphoma, finding a 6-fold increase for over 10 years of use. He did highlight however that there were very few cases and that the results should be interpreted with a fair degree of caution[Hardell Sept 2005].
In February 2006 Hardell published a paper using more recent diagnoses (patients diagnosed between 2000 and 2003), and found that the increase in risk was steadily strengthening in magnitude and statistical significance as the length of phone usage was increasing. For malignant tumours he found that the OR for analogue phone use was 2.6 (CI 1.5-4.3), for digital phone use was 1.9 (CI 1.3-2.7), and for cordless phone use was 2.1 (CI 1.4-3.0). Looking at patients who had used their phones for 10 years or more this increased to 3.5 (CI 2.0-6.4), 3.6 (CI 1.7-7.5) and 2.9 (CI 1.6-5.2) respectively[Hardell Feb 2006]. This work was followed up in October 2006 looking specifically at acoustic neuromas (a benign form of brain tumour), astrocytomas and non-hodgkin's lymphomas. He found acoustic neuromas and the higher grades of astrocytoma (Grades III and IV) to have significant increases from all forms of mobile and cordless phone usages (around 50% increase in risk in each case), which increased further for those who had used their phone greater than 10 years. However, he found no increase for lower grade astrocytomas (Grade I and II), and no increase for non-hodgkin's lymphomas in contrast to his paper from the previous year[Hardell Oct 2006].
Hardell's published a meta-analysis in September 2007 of the existing literature to date (2 cohort studies and 16 case-control studies). His findings were a 140% increase in risk for benign acoustic neuromas (CI 1.1-5.3) and a 100% increase in risk for malignant gliomas (CI 1.2-3.4), with further increased risks when looking at ipsilateral exposure[Hardell Sept 2007]. His summary so far on all the work (including work from othor researchers) is that "Results from present studies on use of mobile phones for > or =10 years give a consistent pattern of increased risk for acoustic neuroma and glioma. The risk is highest for ipsilateral exposure.
Flaws, Recall Bias and Selection Bias
As he is one of the only researchers consistently finding an effect, he has received a lot of criticism for the expected quality (or lack of) his work. The primary criticism that has been directed at his work is that of recall bias (failure of the mobile phone users to correctly remember the amount of usage and the side of the head). However (and primarily for the Interphone study), this has been addressed by Vrijheid in a study that found that heavy users generally overestimated their use and light users generally underestimated their use[Vrijheid 2006]. As the high risk category is the heavy users, an overestimation of use would imply the risk is actually attributed to lower usage, which will lead to an underestimation of risk. Vrijheid also found a small element of inaccuracy of users recalling which side of the head their phone used which may contribute to an slight overestimation of risk in the highest user category[Vrijheid 2008]. The other Interphone study to look at recall bias found that users were more accurate at recalling phone usage in terms of number of calls made than total phone usage, but without comment as to how this was likely to effect the reported risk[Samkange-Zeeb 2004]. One of the Interphone papers looked at selection bias, and found that approximately 10% less cases responded than controls (70% and 80% response rates respectively)[Lahkola 2005]. If this difference is real, it would lead to an overestimation in risk, but it is only based on one of the Interphone studies and Lloyd Morgan's pooled analysis found only 60% of controls responded on average (See the section on flaws of the Interphone Project below) - if so, this would instead lead to an underestimation in risk.
The Interphone Project
The Interphone project, initiated in 1999, is a multi-national initiative aiming to fully examine the association between long term mobile phone usage and increases in risk for all forms of brain cancer. The results of the project were expected to be published in 2005, but the findings were conflicting and the researchers involved in the project were divided into three opinions on how the results should be summarised: a) that there was a possible increase in brain tumour risk that is real and warrants further investigation, b) that the increase is an artifact due to bias in the study, and c) that with the data currently available it is not possible to tell whether a) or b) are correct, with both being a feasible possibility. Louis Slesin has covered the various updates in the Interphone project in great detail on his Microwave News website. The results of the Interphone project have still (as of August 2008) not been published.
Unfortunately, here are a number of flaws with the design of the studies in the Interphone project. One of the major flaws is a lack of inclusion in most of the studies of any form of recognition of digital cordless phone users. With the handset similar to a phone and the base unit exposing users all the time whilst they are in the building, this becomes a very large confounder for using "length of phone usage" as a metric of exposure level.
Lloyd Morgan, an American researcher and electronic engineer (and currently a member of the board of directors at the Central Brain Tumor Registry of the United States), has been writing about the Interphone papers as each one is published, and has explained the major flaws with the Interphone protocol in great detail. The summary of the 6 primary flaws are listed below:
Flaw 1: Selection Bias
The first flaw is called selection bias. It is likely the result of the low percentage of controls that participated in the studies (weighted average of 59%). Think about being randomly selected for a cellphone study. You are told you will be asked to answer a long questionnaire. If you use a cellphone you are more likely to agree to participate than if you do not use a cellphone. If this happens it is called selection bias. Selection bias will result in an underestimation of risk.
Flaw 2: Inclusion of Tumors Outside the Cellphone's Radiation Plume
The second flaw is the inclusion of all brain tumors without regard to their location. Because the cellphone's radiation plume only penetrates a short distance into the head, nearly all of this radiation is absorbed by the temporal lobe, the acoustic nerve, or the parotid gland. Even when cellphone exposure of one side of the head is considered on the side where the cellphone was held, a substantial portion of half the brain is unexposed (the opposite side is completely unexposed). Studies that include brain tumors outside of the cellphone radiation plume contribute to an underestimation of the risk of brain tumors.
Flaw 3: Latency Time and Definition of Regular User
The third flaw is the definition of "regular" cellphone use in relation to a reasonable latency time. "Regular" cellphone use is defined as use of a cellphone on average once per week for at least 6 months. Exposure within 1 year of the diagnosis date is not considered. The result of this definition, combined with the incredibly fast rate of new cellphone users, is to overweight "regular" users with an incredibly large group of short-term users - far too short a time to expect a tumor to be diagnosed.
The latency time for brain tumors is between 15 and 25 years. For the Interphone studies, using weighted averages for cases or controls, we see that 6.3% of cases and 6.4% of controls have used a cellphone for 10 years or more, and 18% of cases and 21% of controls have used a cellphone for 5 years or more (Weighted average of 10 Interphone brain tumor studies - 3 Interphone studies of 5 countries which the 10 studies are excluded). For a reasonable latency time, it would be unlikely to find any risk of tumors, given the percentage of cases and controls. Yet some of the Interphone studies are already finding a risk. Because such a large percentage of "regular" users have used a cellphone for an unreasonably short latency time the reported results for < 10 years as well as for > 10 years (6.3% of cases) are an underestimation of risk.
Flaw 4: Children and Young Adult Are Not Included in Interphone Studies
The Interphone Protocol states that cases be between 30 and 59 years of age While a few studies have included cases as young as 20, the non-inclusion of < 20 year olds is likely to result in an underestimation of risk. Research has shown that children's brains (due to skull formation through the childhood years) absorb a greater proportion of the radiation emitted by a mobile phone[Ghandi 1996, Ghandi 2002, Christ 2005, de Salles 2006, Wiart 2008]. They are also likely to be at a greater risk due to their higher rate of cell division (than adults). It is generally accepted that teenagers and young adults are the primary users of mobile phones.
Flaw 5: Cellphone's Radiated Power
It is reasonable to expect that risk of a tumor from a cellphone, after a reasonable latency time, would be the cellphone's power multiplied by cumulative time of use. In the early days of cellphone use all cellphones used analog technology. These always radiated a fixed amount of power (~2 Watts). Analog cellphones use has been totally displaced by digital cellphones. Digital cellphones have a feature called Automatic Power Control or APC. At the beginning of a call the cellphone radiates maximum power (~2 Watts) but quickly reduces the power so the radiated power is sufficient to have a reliable link to the cell tower (AKA masks or base stations). The result is that cellphones radiate far less power in urban areas compared to rural areas. This is because cell phone towers are much closer in urban areas compared to rural areas so the cellphone radiates less power in urban areas and more power in rural areas. When rural and urban cellphones are not reported separately the result is an underestimation of risk.
Flaw 6: Number of Cases Included in a Study
The weighted average time in these 10 studies for a case to be eligible for inclusion in the study was only 2.6 years. When one considers 4 of the 5 previous flaws, it becomes obvious that such a short period of time for eligibility will result in too few cases to resolve these flaws. For example, if tumors were limited only to the exposed region of the brain then there would be far fewer cases; if a reasonably long latency time was included, again there would be far fewer cases; if children had been included there would have been more cases; and, if rural users were to be compared to the far larger number of urban users a much larger number of cases would need to be eligible to participate in the Interphone Study.
In this year's (2008) BEMS (Bioelectromagnetics Society) meeting, Lloyd presented a thorough talk outlining all of these flaws, their implications, and how this affected the statistical data represented in the papers. His indication that the flaws in the Interphone protocol would significantly underestimate the overall risk would closely match our original theory as speculated in January 2007 when one of Lahkola's studies[Lahkola 2007] was published, based on crude statistical analysis of suspicious looking figures. Lahkola published a meta-analysis of Scandinavian papers in August 2008 that similar found these strongly significant protective effects, with an OR 0.76 (CI 0.65-0.89)[Lahkola 2008]!
It wasn't just Lahkola's work though, a number of the Interphone studies have found statistically significant protective effects from mobile phone usage[Schuz Mar 2006, Schuz Dec 2006, Christensen 2005, Lonn 2005, Klaeboe 2007], and despite these flaws some have still found statistically significant increase in risk for the heavier group of mobile phone users[Hepworth 2006, Lonn Nov 2004, Schoemaker 2005, Takebayashi 2008, Hours 2007].
A number of the papers have not shown this unlikely looking protective effect and have still not shown an increase[Christensen 2004, Takebayashi 2006], but it is important to remember that with around 5% of the cases having used their phone for 10 years or more we would not expect to see a risk anyway (as the latency period for these tumours tends to be between 15 and 25 years). The fact that any papers are showing a risk is very concerning.
There have been a number of other issues on an individual study basis, such as the highly publicised Danish cohort study towards the end of 2006[Schuz Mar 2006, Schuz Dec 2006] which was hyped as being a "definitive study", containing some 400,000 people in the dataset. Sadly, as we covered in great detail in December 2006, the classification of these subjects was extremely misleading: Firstly, only contract users were considered as those were the only users that it was possible to identify. All "pay as you go" users were classified as "non-mobile phone users" and effectively moved into the control group - this would lead to an underestimation of risk. Originally the authors had 720,000 records, of which 100,000 were removed (quite validly) due to duplication. Out of the remaining 620,000, a further 200,000 were removed because they "couldn't identify the users" as the contracts were corporate and not linked to any specific individual. By the authors' own admission these were likely to be the heaviest phone users in the dataset, so another 33% of the heaviest users they were looking for were moved in to the "non-user" control group, which would also lead to an underestimate the risk. Strangely, the study's findings then highlighted a significant protective effect (7 of the 18 data points had a statistically significant reduction in risk), which disappeared in the highest usage category. One reason for this may be the flaws detailed by Lloyd Morgan above for the Interphone project papers, that would expect a significantly reduced OR in cases when compared to controls. If these are controlled for in the statistical calculations, we end up finding an increase in risk of 30% for the heaviest users, reaching borderline significance!
There have been a number of other epidemiological papers published over the last 10 years looking at mobile phones and brain cancer, but most of these have failed to find an effect. This is unsurprising, as the studies were published a number of years ago and the cases involved had used their phones for less than 5 years - far too short a period to find any risk of brain tumours[Lonn Jan 2004, Cook 2003, Muscat 2002, Johansen 2001, Inskip 2001, Muscat 2000, Auvinen 2002]. Despite this, two of the studies found increases for brain cancer that were of borderline significance[Muscat 2000, Auvinen 2002]. More recently Kan performed a meta-analysis on 9 case-control studies, finding no increase of risk overall but a statistically significant increase in risk (OR 1.25) for those who have used their phone for more than 10 years (CI 1.01-1.54)[Kan 2008].
Hardell L et al
, (May 2000) Case-control study on radiology work, medical x-ray investigations, and use of cellular telephones as risk factors for brain tumors
, MedGenMed. 2000 May 4;2(2):E2 [View Author's abstract conclusions
Hardell L et al
, (2005) Case-control study on cellular and cordless telephones and the risk for acoustic neuroma or meningioma in patients diagnosed 2000-2003
, Neuroepidemiology. 2005;25(3):120-8 [View Author's abstract conclusions
Hardell L et al
, (February 2006) Case-control study of the association between the use of cellular and cordless telephones and malignant brain tumors diagnosed during 2000-2003
, Environ Res. 2006 Feb;100(2):232-41 [View Author's abstract conclusions
Ghandi O, Kang G
, (1996) Effect of the head size on SAR for mobile telephones at 835 and 1900MHz
, Bioelectromagnetics Society 23rd Annual Meeting. St. Paul, Minnesota, USA, June 10-14, 2001, p. 52 [View Author's abstract conclusions
Ghandi O, Kang G
, (May 2002) Some present problems and a proposed experimental phantom for SAR compliance testing of cellular telephones at
835 and 1900 MHz
, Phys. Med. Biol. 47 1501 18 [View Author's abstract conclusions
Lahkola A et al
, (August 2008) Meningioma and mobile phone use--a collaborative case-control study in five North European countries
, Int J Epidemiol. 2008 Aug 2. [Epub ahead of print]Click here to read [View Author's abstract conclusions
Schoemaker MJ et al
, (October 2005) Mobile phone use and risk of acoustic neuroma: results of the Interphone case-control study in five North European countries
, Br J Cancer. 2005 Oct 3;93(7):842-8 [View Author's abstract conclusions
Takebayashi T et al
, (February 2008) Mobile phone use, exposure to radiofrequency electromagnetic field, and brain tumour: a case-control study
, Br J Cancer. 2008 Feb 12;98(3):652-9 [View Author's abstract conclusions
Hours M et al
, (October 2007) Cell Phones and Risk of brain and acoustic nerve tumours: the French INTERPHONE case-control study
, Rev Epidemiol Sante Publique. 2007 Oct;55(5):321-32 [View Author's abstract conclusions
This page has links to content that requires a .pdf reader such as Adobe Acrobat Reader