Is being overweight better for life expectancy?

If you believe a recent article in the NY Times  captioned "Our Absurd Fear of Fat" you'd conclude that being overweight but not obese is actually better for your life expectancy.  The article points to a recent piece of research published in the Journal of the American Medical Association which suggests that the relationship between mortality rates and BMI may not be quite as simple as suggested.  In fact, the research concludes that "Relative to normal weight, ... Grade 1 obesity overall was not associated with higher mortality, and overweight was associated with significantly lower all-cause mortality."

Let me play this back in normal English.  

BMI is a measure of weight proportionate to height.  It is considered a good way to determine whether a person is overweight.  Usually, BMI of 18.5 to 24.9 is considered normal.  People with BMI of 25 to 29.9 are considered overweight, while Grade 1 obesity is BMI of 20 to 34.9.   Usually, we are told that being overweight is bad and being Grade 1 obese is associated with substantially elevated risk of risk of a bunch of diseases including heart disease, diabetes, etc.

This research is suggesting that being overweight may slightly improve your chances of surviving longer and there is hardly any difference between Grade 1 obesity and normal weight.  

If true, this would be astonishing.  It is so counter intuitive that I decided to investigate whether there are other explanations.

The thing that bothers me about this analysis is that it suggests that the BMI has the same interpretation irrespective of what caused you to get to that weight.   

For instance, imagine someone who has an active lifestyle with moderate calorie intake vs someone that has ridiculously high activity with a very high calorie intake.  If their BMI was the same, would that mean the same life expectancy?  If people were machines, we'd probably conclude that the higher activity level leads to higher wear and tear which may in fact result in lower durability, or in human terms higher mortality than people with a more normal level of exercise and calorie intake.  Could it be that the elevated risk of being overweight is sometimes offset by the higher levels of wear and tear that one would need to to sustain to maintain that weight without reducing calorie intake?  I could not find any study that had specifically addressed this.

I also speculated that it could be because of factors such as smoking and illness.  Smokers are known to have lower weight for the same lifestyle.  Could higher incidence of smokers in the normal weight population be skewing the results?  Also, sick people generally lose weight, so could preexisting diseases be skewing the results?

I had more success with these latter speculations.  A few Google searches later, I found this fascinating piece of research that looks at several possible causes, particularly race, preexisting diseases, smoking and age.  

What they conclude is that once you adjust for preexisting conditions and smoking, you do see the risk of death increase as expected as you get beyond normal BMI.  However, a close look at the charts suggests that it depends on  the age of the participant  and that while the optimal BMI was somewhere between 23 and 26, the mortality risk was actually roughly the same for people who were thinner vs those who were slightly overweight.

At a high level, what this suggests that BMI does in fact affect risk of mortality but that we need to consider more than just the raw BMI in assessing risks.  We need to factor in how one got to the weight.  

A more practical point is that people who are overweight (not obese) probably don't need to obsess about weight and should instead focus on other risk factors that are probably significantly more predictive of their mortality risk than their weight.