Wednesday, March 21, 2012

From Yahoo: "Can an aspirin a day keep cancer away?"

     If, like me, you are a professor looking for examples of bad science to share with your students, one fertile hunting ground is going to be Yahoo! "news."  It is usually dangerous when science ends up being discussed in the popular press, because there is no real way for any scientist who gets quoted to verify the interpretations of the writer ... who is not, generally, an expert in the field.
     This danger gets magnified when the "news" source is something like Yahoo!, which seems to rely ever less on reporters and more on bloggers.
     Whatever the credentials of the writer, however, Yahoo! is a common dumping ground for dumb comments by "professionals."  Take this article, for instance, in which two people who self-identifty as doctors (Borges and Turner) are reporting on three articles recently published in medical journals.
     There are two bits of relevant information in the Yahoo! story.  The first is that people on a daily aspirin regimen have a lower rate of occurence of cancer than do people not taking aspirin regularly.  The second bit is that, when they do develop cancer, those on the aspirin regimen have a reduced risk of death.
     If you are new to such epidemiological findings, what we are talking about here are correlational studies.  This means that researchers have collected data through some combination of reviewing medical charts and surveying patients, and at least one variable has been found to be statistically related to the incidence of some disease.  Any condition found to be so correlated is referred to as a "risk factor," or as something "linked to" a disease.
     What this does not mean is that the new "risk factor" is a cause of, or cure for, the disease in question.  Determining scientific causality requires a controlled experiment (in medicine, usually a randomized clinical trial).  These are very rare in medicine, because we need ethical approval to either 1) deliberately increase the risk of disease for some patients, or 2) to deny some patients a treatment we believe will improve their condition.  Because it is very difificult to get permission to inflict deliberate harm, we are left with statistical relationships between variables, without any way to tell how or why they are related.
     This is basic scientific methodology ... the sort we teach undergraduate students in their first class.  No one with a bachelor's degree in a scientific field should ever confuse a statistical relationship as being causal.  And if that's true for those holding the lowest of university degrees, it is infinitely more true for people who have doctoral degrees.
     So what do we get in this Yahoo! article, written by two doctors?  We get the following quote, without criticism ...
"No one knows exactly what the mechanism is of how aspirin helps prevent cancer," said Asad Umar, chief of the Gastrointestinal and Other Cancers Research Group of the National Cancer Institute.
     Aspirin prevents cancer?!? From the head of a research group AT THE NATIONAL CANCER INSTITUTE!?!
     Congratulations, Dr. Umar - and to you, Drs. Borges and Turner, who authored this trash - the Fist of Causality serves up to each of you a well deserved punch in the neck. 


 

In the beginning ...

     Hello, and welcome to The Fist of Causality!  This is my first blog, so please bear with me as I get used to the process.  By training, I am an experimental psychologist.  In practice, I am a biostatistician.  That is, I make my living largely by teaching research methods and statistics to college students ... and as a statistical/methodological "gun for hire."

     This blog is born out of my increasing frustration at the inability of people to properly interpret the data they are presented.  To be clear, I am not frustrated by different people looking at the same data and infering different theoretical implications.  Rather, this blog is fueled by people who ought to know better doing things they oughtn't do, with the scientific method and with statistics.

     I don't know how often I will post.  But when I do, rest assured that it will be as a result of the continuing abuse of cumulative human knowledge.