Comment My take on the study (Score 1) 282
First, please be aware that there is apparently a link between "texting" (sending an SMS) an being shot dead in your house:
http://sfgate.com/cgi-bin/article.cgi?f=/c/a/2006/ 04/02/BAGEUI24SP1.DTL
The former association is probably tenuous at best, however the latter article actually seems fairly well done. The Sweeds are able to do some great large-scale studies, as they have a fairly homogenous population with amazing levels of data collection.
I read the article, and this study was fairly well designed.
The important conclusions of the study was that long term, high quantity use of phones (both cellular, and cordless) was associated with a several fold increase in brain cancer risk (about 3-6x the risk in those with over 2000 cumulative hours of use) and that this result was seen most clearly after 10+ years after the exposure, consistent with the possible long-term / slow developing nature of cancer.
To put this in perspective, the lifetime risk for brain cancer is about 0.5%. If the results of this study were true, and you were a cell phone addict, perhaps you could increase your risk 10-fold, meaning your chance of getting a tumor had risen to 5%? The risk of dying of other things is much higher (heart disease and stroke: 40%, other cancers 20%, etc.) so it may not be worth worrying about.
Could the results be bogus?
This study is a retrospective case-control design, which means that they took people diagnosed with tumors, found healthy control subjects (matched on age, sex and geographic location), and then asked both groups about their cell phone use over the prior 20 years. It's fairly easy with these sorts of designs to have confounds. One of the most common and easiest to understand confound are 'response biases.'
In this study, this could work in several ways: Perhaps people with brain tumor are pissed off, have heard about possible connections with cellphone use, and are (perhaps subconsciously) looking for something to blame. Thus, they tend to overstate their cellphone use. The controls, who have no tumors, have no particular agenda and report their cell use more accurately. Another form this could take would be if the tumor folks with high cell use respond to the surveys accurately, whereas the controls with high cell use just throw them in the trash.
The authors have a few of points arguing against this thesis : first, the observed relationships follow a dose-response-latency model, which they suggest may not be likely to be how unequal response bias would look. Second, they had nearly identical and high response rates (90%) for both groups. Third, they found the relationships even when looking at cordless phones (which have generally not been demonized in the popular press).
One thing that did jump out at me were the demographic issues -- the authors mention that both gender and class (wealth, or socioeconomic status) are related to cell use (e.g. rich men use the most). They claim to have adjusted statistically for these issues, but the method was not described in this paper (it may be in their prior papers). However, in a retrospective case/control study like this, one should always question "is there anything unusual about group X". People with brain tumors? Wealthy males with very high cellphone use who were early adopters? Perhaps there is a core group of subjects that is driving the statistical relationship? E.g. high users are computer nerds who don't get enough sunlight, eat crappy diets and are vitamin deficient, and spend their lives in server farms inhaling chemicals from their computers? The authors did not report having measured any of these obvious confounds, which does leave a pretty big possible explanation: that the association between high cell use and cancer is true, but that the correlation is not causal.
http://sfgate.com/cgi-bin/article.cgi?f=/c/a/2006
The former association is probably tenuous at best, however the latter article actually seems fairly well done. The Sweeds are able to do some great large-scale studies, as they have a fairly homogenous population with amazing levels of data collection.
I read the article, and this study was fairly well designed.
The important conclusions of the study was that long term, high quantity use of phones (both cellular, and cordless) was associated with a several fold increase in brain cancer risk (about 3-6x the risk in those with over 2000 cumulative hours of use) and that this result was seen most clearly after 10+ years after the exposure, consistent with the possible long-term / slow developing nature of cancer.
To put this in perspective, the lifetime risk for brain cancer is about 0.5%. If the results of this study were true, and you were a cell phone addict, perhaps you could increase your risk 10-fold, meaning your chance of getting a tumor had risen to 5%? The risk of dying of other things is much higher (heart disease and stroke: 40%, other cancers 20%, etc.) so it may not be worth worrying about.
Could the results be bogus?
This study is a retrospective case-control design, which means that they took people diagnosed with tumors, found healthy control subjects (matched on age, sex and geographic location), and then asked both groups about their cell phone use over the prior 20 years. It's fairly easy with these sorts of designs to have confounds. One of the most common and easiest to understand confound are 'response biases.'
In this study, this could work in several ways: Perhaps people with brain tumor are pissed off, have heard about possible connections with cellphone use, and are (perhaps subconsciously) looking for something to blame. Thus, they tend to overstate their cellphone use. The controls, who have no tumors, have no particular agenda and report their cell use more accurately. Another form this could take would be if the tumor folks with high cell use respond to the surveys accurately, whereas the controls with high cell use just throw them in the trash.
The authors have a few of points arguing against this thesis : first, the observed relationships follow a dose-response-latency model, which they suggest may not be likely to be how unequal response bias would look. Second, they had nearly identical and high response rates (90%) for both groups. Third, they found the relationships even when looking at cordless phones (which have generally not been demonized in the popular press).
One thing that did jump out at me were the demographic issues -- the authors mention that both gender and class (wealth, or socioeconomic status) are related to cell use (e.g. rich men use the most). They claim to have adjusted statistically for these issues, but the method was not described in this paper (it may be in their prior papers). However, in a retrospective case/control study like this, one should always question "is there anything unusual about group X". People with brain tumors? Wealthy males with very high cellphone use who were early adopters? Perhaps there is a core group of subjects that is driving the statistical relationship? E.g. high users are computer nerds who don't get enough sunlight, eat crappy diets and are vitamin deficient, and spend their lives in server farms inhaling chemicals from their computers? The authors did not report having measured any of these obvious confounds, which does leave a pretty big possible explanation: that the association between high cell use and cancer is true, but that the correlation is not causal.
