The paradox of the BMI paradox

2016-10-19-17_52_02-physbe-talk-bs-pdf-adobe-reader

I had the honor to be invited to the PHYSBE research group in Gothenburg, Sweden. I got to talk about the paradox of the BMI paradox. In the announcement abstract I wrote:

“The paradox of the BMI paradox”
Many fields have their own so-called “paradox”, where a risk factor in certain
instances suddenly seems to be protective. A good example is the BMI paradox,
where high BMI in some studies seems to be protective of mortality. I will
argue that these paradoxes can be explained by a form of selection bias. But I
will also discuss that these paradoxes have provided researchers with much
more than just an erroneous conclusion on the causal link between BMI and
mortality.

I first address the problem of BMI as an exposure. Easy stuff. But then we come to index even bias, or collider stratification bias. and how selections do matter in a recurrence research paradox -like PFO & stroke- or a health status research like BMI- and can introduce confounding into the equation.

I see that the confounding might not be enough to explain all that is observed in observational research, so I continued looking for other reasons there are these strong feelings on these paradoxes. Do they exist, or don’t they?I found that the two sides tend to “talk in two worlds”. One side talks about causal research and asks what we can learn from the biological systems that might play a role, whereas others think with their clinical  POV and start to talk about RCTs and the need for weight control programs in patients. But there is huge difference in study design, RQ and interpretation of results between the studies that they cite and interpret. Perhaps part of the paradox can be explained by this misunderstanding.

But the cool thing about the paradox is that through complicated topics, new hypothesis , interesting findings and strong feelings about the existence of paradoxes, I think that the we can all agree: the field of obesity research has won in the end. and with winning i mean that the methods are now better described, better discussed and better applied. New hypothesis are being generated and confirmed or refuted. All in all, the field makes progress not despite, but because the paradox. A paradox that doesn’t even exist. How is that for a paradox?

All in all an interesting day, and i think i made some friends in Gothenburg. Perhaps we can do some cool science together!

Slides can be found here.

Advertisements

Opzet en interpretatie van mensgebonden onderzoek – slides online

Tomorrow I will teach at the graduate course ‘Design and analysis of clinical research’. My part is to introduce the concept of confounding which i demonstrate through the general  idea of ‘confusing of effects’.  Perhaps a bit ‘oldskool’, but it works as a nice introduction to the concept without a direct confrontation with DAGs etc, especially since it helps to think in ways to prevent / solve this problem in data analyses. What ‘arrow’ in the classic confounding triangle can be removed?

I also go into the concept of ceteris paribus, which is further explored through examples of IV analyses. These examples can be historical (Boylston and inoculation) or recent (mendelian randomisation on CRP and CVD disease).

Slides are present in the presentations section of this website.