Questioning the Reliability of the " Mobile Phone Use and the Risk of Acoustic Neuroma" Study
I've been receiving emails questioning the significance of the reported results, an increase in the incidence of acoustic neuromas associated with analog cell phone use.
It's important to understand that these researchers utilized the case-control study method which is the least reliable compared to a prospective cohort study, or the even better randomized clinical trial (RCT), also prospective. The case-control study is retrospective, relying on investigating people who have already contracted a particular disease (acoustic neuroma), and then looking at a group of people who are similar except for not being exposed to the supposed cause (cell phone use) for comparison.
In the interest of clarity, I'll quote from a book on evaluating clinical tests, because this author has such a facility for explaining statistics. It's "Studying a Study and Testing a Test," by Richard K. Riegelman:
Case-control studies have the distinct advantage of being useful for studying rare conditions or diseases. If a condition is rare, case-control studies can detect differences between groups using far fewer individuals than other study designs require. Much less time is often needed to perform a case-control study because the disease has already developed. This method also allow investigators to simultaneously explore multiple characteristics or exposures that are potentially associated with a disease. One could examine, for instance, the many variables that are possibly associated with colon cancer, including prior diet, surgery, ulcerative colitis, polyps, alcohol, cigarettes, family history, and many other variables.
The major objection to case-control studies is that they are prone to various errors and biases that will be explained in the following chapters. Despite these problems, case-control studies are very important for establishing associations. They are capable of showing that a potential "cause" and a disease or other outcome occur together more often than expected by chance alone.
First, we can classify the Swedish study as really a historic, or retrospective cohort study since we are looking at one supposed cause: analog cell phone use in one cohort as opposed to the control group which didn't. Second, notice that Riegelman uses "cause" in quotes. Association doesn't necessarily mean causality. There might be some other confounding factors causing disease that are not apparent in this group. However, in this study, the incidence of disease was associated more often with the same side of cell phone use. This lends creedence to cell phone use as the cause. Third, we are dealing with a potential reporting bias. A person with a diagnosis is more likely to make an association with a supposed cause as opposed to a person who never had a reason to consider any problem with their behavior. Observational studies suffer from this bias.
Finally, how do make an extrapolation from such a study? What do you recommend and with what certainty? Predicting rare events is the biggest challenge of clinical medicine. It makes the greatest demands on the clinical tools we have. What's the feasibility of doing a prospective study involving tens of thousands of participants for TEN YEARS to observe an effect? (If I'm not mistaken, symptomatic acoustic neuromas have a prevalance of 1:100,000?)
We all want certainty, but we have to appreciate the effort it might entail to feel comfortable about living our ever mobile, digital lives.
8:44:41 PM
|