This is interesting in a "of course" kind of way.<p>If you look at a bunch of variables, including BMI, and then you remove the impact of well known negative health impacts of obesity (prediabetes/T2D), then you see that high BMI doesn't not correlate well with mortality. IMO, what this indicates to me is that (1) BMI is not a good indicator of obesity in this study. There are many healthy people with low body fat and high BMI, I am one of them. All you need to do is be tall and lift weights occasionally. If you remove the obesity related negative health signals, you also remove obesity(2) Being fat in and of itself is not the issue, the issue is prediabetes/T2D which is extremely reliably caused by obesity, and the treatment for prediabetes/T2D is weight loss.
I can't read the study directly (and it's not my field, which makes reading it directly challenging), but this is the abstract conclusion:<p>`
Higher fasting insulin and higher c-reactive protein confound the association between BMI and the risk of all-cause mortality. The increase in mortality that has been attributed to higher BMI is more likely due to hyperinsulinemia and inflammation rather than obesity.`
Key takeaway:<p>"Higher fasting insulin and higher c-reactive protein confound the association between BMI and the risk of all-cause mortality. The increase in mortality that has been attributed to higher BMI is more likely due to hyperinsulinemia and inflammation rather than obesity."
The researchers tried to subtract out the mortality risk of elevated C-reactive protein (aka CRP, inflammation market) and higher fasting insulin.<p>However, both elevated CRP and higher fasting insulin are correlated with obesity, so controlling for these variables seems misleading.
This isn't surprising. BMI increases bring reduced insulation sensitivity, which in turn is what causes the issues. The fix is to diet and exercise. At the end of the day we don't care about BMI, we care about insulin sensitivity - but it is easier to measure BMI than blood insulin. BMI isn't a perfect indicator but it's a pretty good one.
This seems like a textbook example of the model 11 "bad control" from <<a href="http://causality.cs.ucla.edu/blog/index.php/category/back-door-criterion/" rel="nofollow">http://causality.cs.ucla.edu/blog/index.php/category/back-do...</a>>. X is BMI, Y is mortality, and Z is fasting insulin and blood sugar. Basically, if X causes Z and Z causes Y, then controlling for Z hides the relationship between X and Y.
PDF here: <a href="https://www.nature.com/articles/s41366-022-01211-2.pdf" rel="nofollow">https://www.nature.com/articles/s41366-022-01211-2.pdf</a>
"The increase in mortality that has been attributed to higher BMI is more likely due to hyperinsulinemia and inflammation rather than obesity."<p>That's encouraging because I've been able to fix the hyperinsulinemia and inflammation, but not all of the excess BMI. Maybe fit & fat aren't mutually exclusive.
And... one big factor in inflammation is our modern lack of sun exposure (infrared).<p><a href="https://www.youtube.com/watch?v=wadKIiGsDTw" rel="nofollow">https://www.youtube.com/watch?v=wadKIiGsDTw</a><p>Taking someone sick with high inflammation and putting them in a hospital bed is terrible for their recovery.
As a side note, BMI is a terrible measure of obesity for outliers. I have several first hand experiences where the BMI indicates obese or even morbidly obese, while the DEXA scan would show fat in single digits.
TLDR this is only about obese people, don't bother reading if you have healthy BMI, heck median BMI of participants was 27, way beyond healthy since overweight starts at 25 and I would say even that is very generous<p>~20.3 here
<i>Using the primary cohort, time to mortality was regressed on all three exposures of interest: BMI, fasting insulin, and CRP. A number of parametrizations of these three exposures were explored. The models with the best fit used linear and quadratic terms for BMI, a linear term for fasting insulin, and the natural logarithm of CRP.</i><p>and we chose not to justify or explain those parameterizations in any way, because what we probably actually did was diddle SAS until the model did what it was supposed to.<p>yawn.