Two articles in the press caught my eye last month. One reported that eating avocados twice a week lowered your risk of heart disease by 20% and the other that drinking 2-3 cups of coffee daily lowered your risk of death and heart disease by 10-15%. Since I need my two cups of coffee to get going in the morning but have never eaten avocados, I had to dig a bit deeper.
Both these studies, it turned out, were observational studies.
The gold standard for testing the value of a new treatment is the controlled trial. You take two groups of people who are similar in all respects and randomly give one half of the group treatment A and the other half treatment B. You then compare the results and if the results are substantially different, you can assume this is because A or B is better.
When the treatment is something people cannot do by themselves, this type of trial is straight-forward to design and conduct. When you are looking at diet, exercise, smoking or other habits, things get a lot harder. To compare the effects of eating avocados twice a week or not, you would have to control the subjects’ diets 24/7, clearly impractical. For this type of comparison, people are recruited and asked to do or not do things, but the researcher cannot control their actual behavior, or their usual behaviors are ascertained by questioning them.
Many large groups have been studied over the years in observational trials, and much useful information has been learned by following their health outcomes, but findings from such studies are almost never proof that the behaviors caused the outcomes.
The classic example of this misunderstanding about observational trials is the effect of post-menopausal estrogen use. For decades, almost every doctor believed that taking estrogens after menopause prevented heart disease. Why? Because women who took estrogen had much less heart disease. Only after the Women’s Health Initiative trial seemed to refute this did doctors pause to think that women who took estrogen were different in many other ways from women who did not. They smoked less, exercised more, saw doctors more often and were generally more health conscious. It seemed they were healthier to begin with.
More recently, the “fact” that light to moderate alcohol consumption benefits heart health has also been questioned. Researchers looked at over 371,000 people in the United Kingdom Biobank. They found, as expected, that light to moderate drinkers had the lowest heart disease risk. They also found that this group tended to have healthier lifestyles than abstainers: they smoked less, ate more vegetables and were more physically active. Taking the lifestyle factors into account eliminated any beneficial effect that could be attributed to their drinking habits.
Bottom line: observational studies may suggest harms or benefits but rarely if ever can they prove such effects. If you enjoy avocados or your morning coffee, go on consuming them, but I would not depend on either to keep you forever young.
Prescription for Bankruptcy. Buy the book on Amazon
Spot on. This post hoc ergo propter hoc reasoning is so common--and so unwarranted.ReplyDelete