Lloyd Morgan's Column
[Return to column index]
Eye Cancer in Germany from Cellphone Use
The recently published study by Andreas Stang and team is an amazing turn about from his previous paper. The previous paper reported a strong risk of eye cancer (uveal melanoma) from "radiofrequency-transmitting devices" and from "probable/certain exposure to mobile phones" (OR = 3.0, 95% CI = 1.4-6.3 and OR = 4.2, 95% CI = 1.2-14.5, respectively).1 The current paper is a repudiation of the previous paper. "Risk of uveal melanoma was not associated with regular mobile phone use (OR = 0.7, 95% CI = 0.5 to 1.0 vs population control subjects);"2
Yet we see an immediate problem in this sentence. This sentence should have read, "Statistically significant protection from uveal melanoma was associated with regular mobile phone use (OR = 0.7, 95% CI = 0.5 to 1.0 vs population control subjects)," because this is what was found, a statistically significant protective effect!
How could this be? First the authors criticize their previous study because it had "incomplete exposure assessment." Let's examine the implicitly complete exposure assessment to determine if it is even a reasonable exposure assessment.
The authors have adopted, without explanation, the definition of "Regular" cellphone use used found in the industry-funded Interphone Study, where "regular" use of a cellphone is defined as "use for at least once a week for 6 months or more." "Regular" cellphone use in all Interphone brain tumor studies published to date was also found to protect the cellphone user from a brain tumor. What is going on here? Why have these authors adopted the already discredited Interphone Study's regular cellphone use?
As discussed in a previous Powerwatch column, either such use does protect the user, or the studies have major design flaws. There are no other possible conclusions.
It is useful to consider such a definition if it were applied to smoking and lung cancer. If a "regular" smoker smoked at least once a week for 6 months or more would we expect to find a risk of lung cancer? Adding insult to incredulity, they also reported risk for "sporadic" users (defined as use of a cellphone at least once but less than a "regular" user). It would appear, by adding "sporadic" cellphone users to their published results, that they are trying to out-do the Interphone Study's incredulous findings. Sporadic use found a use of a cellphone had a non-significant protective effect, OR=0.9 (CI: 0.7 to 1.3).
What would we think of a study of smoking and lung cancer where the smoker had smoked at least one time, but less than once a week for less than 6 months? The authors report "sporadic use" in 2 tables (with a total of 15 results). What is the point? Was it just to filler?
Let's examine the duration of cellphone use in this study. The latency time between exposure and diagnosis of a cancer is multiple decades. "Regular" phone use for any duration of was a mere 30% of the eye cancers cases. What about use for longer periods of time? Sixty-four (14%) of the eye cancers cases had used a cellphone for >5 years of use with the exact same protective find for all "regular" users. Fourteen (3%) of eye cancer cases had used a cellphone for >10 years and a non-significant protective effect was found.
1 - p=0.0035 (equivalent to 99.7% confidence) and p=0.021 (equivalent to 97.9% confidence) respectively.
2 - p=0.05 (equivalent to 95% confidence).
The goal of an implicitly improved exposure assessment to resolve the previous study's "incomplete exposure assessment" morphed into an incompetent exposure assessment bordering on farce.
It is not only important to read a paper carefully, it is also important to read it with an eye towards what is not being reported. The on-line Supplemental Methods states "participants were asked questions about the use of cordless phones and radio sets." Results for radio sets are reported but results for cordless phone use are not reported. In an email to Dr. Stang I asked why cordless phone use was not reported. His reply was stunning, "Until now, we did not do these analyses." Was the published analysis incomplete, incompetent, or both?
The prevalence of cordless phone use during the time of this study was likely much higher than cellphone use. The German Interphone study on the risk of brain tumors (glioma and meningioma) from cellphone use reported, "In the 1990s, cordless phones became very popular in Germany and are now replacing the common fixedline phones."
Cordless phones in Germany use DECT technology, which is based on GSM cellphone technology. Exposure from cordless phones is very similar to exposure from cellphones. Ignoring cordless phone use (treating users as unexposed) result in an underestimation of risk. This suggests that a portion of the protective results can be explained by the deliberate ignoring of cordless phone use.
There are additional problems in this recent study which I will not address per se but will describe two additional links to the Interphone study that raise critical questions.
First, "For the detailed assessment of mobile phone use, we used the questionnaire of the Interphone study." This may have been convenient, but because the Interphone brain tumor studies have shown cellphone use provides consistent protection from a brain tumor, it may not have been wise.
Second, "We used the same categories of the cumulative exposure measures as the German part of the Interphone study." This seems bizarre. Why would it be assumed that cumulative exposure would be similar between these 2 studies? Even stranger, the German Interphone study used the controls for cumulative exposure measures (including cut-points) yet it was a study of the risk of brain tumors. If the purpose of the study is to determine the risk of brain tumors from cellphone use, why would controls that by definition do not have brain tumors be used to determine cumulative exposure measures? In the same email I asked Dr. Stang why had they used the procedure of the German Interphone study noting, in the "German Interphone study 61% glioma controls and 69% meningioma controls had never used a cellphone. And in your study the number of controls who had never used a cellphone was a remarkably different 17% to 24% depending on the control group." Dr. Stang replied, "From a statistical point of view, the best cutpoints are based on the distribution in the study base (=control group). this is how wie [sic] originally did it. However, the reviewers asked us to use the same cutoffs as in the German part of the Interphone study."
Reviewers are anonymous. Were they involved with the Interphone study? Were they associated with the cellphone industry? There is no way to know. Were the original results different from the re-analyzed results? An email asking for an answer to this latter question went unanswered.
Stang A et al
, (January 2009) Mobile phone use and risk of uveal melanoma: results of the risk factors for uveal melanoma case-control study
, J Natl Cancer Inst. 2009 Jan 21;101(2):120-3. Epub 2009 Jan 13 [View Author's abstract conclusions
3. Andreas Stang, Personal communication, 21 January 2009
[Return to column index]