Thursday, March 29, 2012

Compliance bias

Recently, I read analysis of a research study (I can't remember where right now), that was talking about compliance bias.  Some of you may know, I'm taking a grad class in Epidemiology right now, so much of my semester has been spend talking about different factors that cause studies to show a result that isn't valid. 

For example, if you surveyed the heights of all the men in the NBA and averaged them, you would get a number that would be the average height of an NBA player -- but would not be the average height of a man in the same age group in the United States.  NBA players are screened for height -- that's part of the reason they are NBA players.  So the results you get from your study is not applicable to an unscreened group of people like the general population. (Selection bias)

Compliance bias occurs because human behavior is very complex.  Some people are just more compliant than others.  The article I was reading cited a study done in the early 70's investigating the effects of cholesterol medications.

What that study found was that those who took either the actual medicine or the placebo with the greatest degree of compliance to the instructions, had the greatest reduction in risk of death during the study.

Those who took either the medication or the placebo with the least amount of compliance to the instructions (took it less regularly) experienced an increase risk of dying during the study than did the compliant group.

Although those that took the medication as directed had a slightly decreased risk over those that took the placebo as directed -- even the placebo group experienced better results than the non-compliant medication takers.  Know why?

Because medication compliance says something about who these people were.  If they were willing to follow directions to take a pill regularly is it likely they were following other instructions to make healthy choices, as well.

The reason I bring this up is I just read another study saying those who eat the most blueberries and apples have the lowest risk for diabetes.  Is this because there is some magical compound in these two fruits that stave off diabetes?  Maybe.  But it is more likely that those who eat the most apples and blueberries are also making many other small choices that bring them greater health.  Perhaps they drink less (or no) soda.  They may be more likely to walk or bike somewhere instead of driving their car.  They probably have health insurance (blueberries and apples are relatively pricey and those that can afford more of them probably have better paying jobs that come with benefits) and those with health insurance are likely to be healthier than those without.  There are, likely, many more factors involved in the way blueberries and apples are related to a reduced incidence of diabetes than just a magic compound in the fruit.

Once you understand compliance bias, it seems to make more sense that processing down apples and blueberries into a pill and then taking that pill (on top of all the other behaviors you already engage in) may not be enough to significantly alter your risk of diabetes (or other negative health outcomes).  Adding more apples and blueberries to an already overflowing diet, also may not be the best way to reduce your risks.

Making health changes that are common sense (eat less, move more) will allow you to tap into some of the "magic" that the apples and blueberry study was actually measuring.

And all of this should be a reminder to us:

There are no magic beans for weight management -- you have to do the work!

No comments:

Post a Comment