Weight and Mortality II: what do thin people die of?

Now that we know that being a bit overweight is good and being too thin is roughly as bad as being merely obese (except for the elderly, for whom it is worse), I was wondering: what are all those thin people dying of? And how about the "regular" weight ones? It seems like for the big killers the obese are at higher (cardiovasclar disease, diabetes) or equal risk (cancer) compared to others.

According to the CDC, the 5 biggest killers are (along with deaths and % of deaths):

Diseases of the heart 725,192 30.3%
All cancers 549,838 23.0%
Stroke 167,366 7.0%
Chronic obstructive pulmonary disease 124,181 5.2%
Diabetes 68,399 2.9%

Well, COPD is almost entirely a disease of smokers. The study corrected for smoking, so we'll assume that has no effect. Whether obesity is an independent risk factor for heart disease and stroke is unclear, but 2 factors not accounted for in the study, high fat diet and lack of physical activity clearly are risk factors and also lead to obesity. Diabetes is also clearly worse in the obese (at least type 2 which makes up the majority of the cases).

So I'm still not clear exactly what thin (and regular weight people) are dying of.

One explanation is that there are very few underweight people. In the survey data sets, only 2-3% of the sample qualify as underweight while 16-23% are obese and 34-36% overweight, so the increased deaths in the underweight don't make a big impact in overall mortality numbers.

The paradox of increased deaths among the normal weight (versus the overweight) remains. One possibility is that the data are wrong (after all the differences didn't make statistical significance). Another is that there is a group of individuals suffering from chronic diseases who are both think and likely to die. These could include those with type 1 diabetes, arthritis, lupus, etc. Follow up in the study was long enough it is unlikely that undiagnosed cancer or HIV plays a huge role, but maybe cancer and HIV treatments are good enough now that the two lower weight groups contains a subset of people who will eventually die of those illnesses but not for many years. Finally, other causes of deaths such as accidents may be higher in the relatively thin.

Note that increased rates of smoking in the thin should have been accounted for by the multivariate analysis.

The article suggest that part of the explanation is improved prevention and treatment of heart disease and hypertension (which predisposes to strokes) but that can't be the whole story or they wouldn't be such important causes of death.

I remain at a loss as to what causes of death are so increased among the "regular" weight that they outweight presumed increased rates of more common diseases in the overweight and obese.

UPDATE: more recent statistics from the CDC suggest accidents have overtaken diabetes for the #5 spot.