Before I go on, a brief word about BMI. Roughly speaking, BMI is a way to compare body weights between people while correcting for height. If we simply compare the body weights of two people on paper, we might think a lean tall person is fatter than an obese short person simply because he weighs more. BMI corrects for this, allowing us to compare weight for height, and it does so reasonably well. The main problem with BMI is that it doesn't differentiate between lean and fat tissue, such that a bodybuilder might appear overweight by BMI. Despite that limitation, BMI is a useful measure of body fatness in scientific studies that compare averages. Its main advantage is that it's easy to measure: all you need are height and weight.
The paper is a goldmine, and I don't intend to discuss it all today, but there's one thing in particular that I do want to discuss because I think it's under-appreciated: the relationship between body fatness and blood lipids (e.g., cholesterol). The cardiovascular research world has a mystery on its hands: why is the blood cholesterol of modern affluent nations so much higher than that of most non-industrial cultures? Saturated fat and dietary cholesterol have been proposed as an explanation, but even if you're generous with the evidence, these cannot explain the full effect.
The current paper shows us the relationship between BMI and blood lipids. This includes two lipid measures: 1) HDL cholesterol, which is considered protective or at least a useful marker of lower heart attack risk, and 2) non-HDL cholesterol, which is considered harmful, and mostly reflects LDL cholesterol concentration. In summary, HDL = good; non-HDL = bad.
On to the data. Have a look at the black squares, which represent non-HDL cholesterol in men, with BMI on the horizontal axis:
Going from a BMI of 19 to a BMI of 30 (very lean to obese) is associated with an increase of approximately 40 mg/dL (1 mmol/L) in non-HDL cholesterol, and a similar association is present in women (open circles). That's a 27 percent increase-- highly significant from a cardiovascular risk perspective.
What's less surprising is the decrease in HDL cholesterol with obesity. This has been described many times and is common knowledge. But when you put the two together into a ratio (non-HDL to HDL), which has more ability to predict heart attacks than either alone, the result is even more striking. From a BMI of 19 to 30 in men (black diamonds), the non-HDL to HDL ratio goes from 2.6 to 4.3, an 65 percent increase! This is illustrated by the black diamonds in the figure above. Again, the finding is similar in women.
Obesity is a known risk factor for cardiovascular disease-- we don't need this paper to tell us that. However, what I think is most interesting is that excess body fatness, and/or higher food intake itself, may account for part of the difference in circulating cholesterol between industrial and non-industrial cultures*. Body fatness may be a factor that explains our more dangerous lipid profile and higher prevalence of heart attacks relative to our ancestors.
* Including ApoB-containing lipoproteins like LDL.