Cell Phones and Cancer: Correlation is Not Causation
When you're looking at the results of data analysis, it's always good to keep in mind what different analyses do and don't "prove." This is especially true when you're reading about health-related studies in the popular press.
For whatever reason, the media frequently imply that a study has revealed some cause-and-effect relationship, even when the study's authors detail precisely the limitations of their research.
I read an article last week that said the World Health Organization (WHO) recently announced that radiation from cell phones can possibly cause cancer. The article has a lot of merit and serves as a good health reminder for all cell phone users, but I think it also brings up a few basic statistical limitations that are important to remember.
According to the article, A team of 31 scientists from 14 countries, including the United States, made the decision after reviewing peer-reviewed studies on cell phone safety.
- I'm not sure that WHO scientists actually combined the results of all the peer-reviewed studies into a statistical meta-analysis, but it’s always good to remember that the “file drawer problem” can prevent scientists from reviewing studies that show no significant results. Often, studies with no statistical significance aren't published or publicized as much as those that show statistical significance. It's also difficult to know for sure how many studies have been conducted, but never had results formally written or shared. Did WHO scientists consult the results of every peer-reviewed cell phone study when they made their announcement? What about the non-peer-reviewed studies?
- “Cherry picking” studies to include in your meta-analysis that align with your hypothesis or agenda can also be damaging to the integrity of your results. Ignoring studies that don’t align with your hypothesis can be just as detrimental. Did WHO scientists unintentionally "cherry pick" studies to review?
The team found enough evidence to categorize personal exposure as possibly carcinogenic to humans.
- It was a good move for WHO to use the world “possibly.” I think this helps to address some of the concerns I mentioned above. It’s sometimes overused, but “correlation does not imply causation” is a good reminder when you’re dealing with statistics. Correlation between two variables does not mean that one variable causes the other, especially if correlation statistics are the only statistics you are using in your data analysis.
The article makes a good point to recognize that more long-term studies need to be done, especially since brain cancer is known for its slow development cycle and widespread cell phone use is relatively new. Even if it's possible that some of the statistics aren't so sound, it's usually a good thing to publicize possible health concerns resulting from something many of us use daily.
Do you ever question the statistics used in health studies?