Sunday, June 05, 2005

Correlation does not equal causation-back to the basics

A recent post by California Medicine Man is worth reading in part because it discusses a text book example of over hyping a statistical association and engaging in hypertrophied, sweeping and unjustified policy recommendations. There are at least two issues here: 1.the basic statistical concepts 2. recommendations for policy changes based on inadequate data and the fallacy that there is a government answer to every purported problem.Everyone should know and could not attend statistics 101 and be even occasionally conscious and not come away knowing that correlation does not equal causation. How many times does a student hear that causation is not proven by observational studies and case control studies?. How many times is the term "hypothesis generating" applied to those types of data?
We are inundated with medical headline news often from observational or case control studies about X risk factor increasing or decreasing the risk of disease Y. Here are a few recent ones'
1."tree pollen peaks are associated with increased nonviolent suicide in women"
2.Thiazide use linked to cholecytitis risk.(here the reported relative risk was 1.16). Should the ALLHAt advocates rush to retract their recommendations for use of diuretics as first line hypertension treatment?
3. High Folate intake is associated with more rapid cognitive decline ( high is is 400 micrograms which is what is in Foltx and similar popular pills used to hopefully decrease CAD risk). Even the authors expressed surprise with the findings.
4. Milk consumption in midlife increases Parkinson's Risk. This is from the Honolulu Heart study.The risk was 2.3 fold higher in men who drank more than 16 ounces of milk per day.
5.Recent studies have also shown that early oophorectomy and history of exposure to pesticides increased Parkinson risk. On the other hand, diet high in Vitamin E is said to decrease risk. A decreased risk has also been reported with ibuprofin use. (In my brief effort to round up some of these factoid articles, I considered adding a feature entitled "the Journal of Silly Articles" along the lines of the Ministry of Silly Walks. Please note that I do not mean to imply the above mentioned articles would be considered for that mythical journal,though some of them may have been funded by "government grants".)
There are so many studies showing weak correlations, multiple studies showing contradictory findings and so many of these studies getting headlines on TV and in the newspapers and the medical news publications that the credibility level of medical news in general has to suffer. For most of the studies an appropriate comment to a patient who asks about it might be " This is very preliminary, it is the type of research that may suggest relationships but cannot prove cause and effect and we will just have to wait and see if the questions raised by this study will be verified or validated as more is learned about the issue."

No comments: