In response to the conversation on “Obesity, Evolution and Delayed Gratification” on the main page and Razib’s coverage of a fascinating new study on the relationship to the lactase gene and obesity, I thought now would be a good time to write about an important new study that helps define the boundaries of what normal and healthy weights are in humans.
This study, entitled Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies is a whopper of a meta-analysis. That is, a study that increases the power of other similar studies by combining their results so that, in this case, data from hundreds of thousands of patients can be aggregated. Meta-analyses have their flaws, and I criticize them frequently when poorly-done or poor-quality studies end up being averaged-in with the results of better-designed studies, but this one is large enough and thorough enough that its results should not be dismissed.
What this study describes is the mortality, and causes of mortality, one observes when one sorts people by body mass index. Body mass index also has it’s flaws but it is a useful, if imperfect method of describing one’s relative contribution of body fat to their total mass. It is calculated by taking and individual’s body weight in kilograms and dividing by the square of their height in meters. “Normal” is defined between 18.5-25, overweight is 25-30, and obese is greater than 30. These numbers do not describe all people well, and you may be an exception to these predictions. This usually occurs if you have a large amount of muscle mass relative to your height, so Arnold Schwarzenegger would be obese according to these scales. However, most people are not Arnold Schwarzenegger and the scale fits, it’s better not to let the perfect spoil the good. One must also remember that it would be unethical to design a study in which we prospectively made people overweight or obese, since we suspect that will cause poor health, so this is necessarily a correlative study of BMI and health. But this information combined with what we know about mechanisms of cardiovascular disease, diabetes, etc., makes a lot of sense, and I believe in the context of the literature we can make a safe assumption the effects we see are causal.
Overall what the study suggests is that the current 18.5-25 recommended BMI is probably about right, BMI of 25-30 marginally increases morbidity and mortality, and BMIs much greater than 30 significantly shorten one’s life. The reason I like this study is that they have aggregated such a huge data set, they demonstrate a clear dose-response curve between obesity and mortality, and they’ve done a better job than most in teasing out the relationship between health, weight, smoking and other co-morbidities at all BMIs.
Let’s take a look at some of the data.
Starting with their first figure, it becomes readily apparent what correlates with obesity in these groups.
The main contributor to early death one would expect would be the effect obesity has on vascular risk factors such as blood pressure, diabetes, cholesterol etc. These data, which are controlled for age, sex, and baseline smoking status (except where smoking is studied), show that as your BMI increases (the X axis), your risk factors increase. Worse, there is a counter-correlation with other possible causes of disease with decreased smoking and drinking at higher BMI.
The authors do many other analyses to show this is not likely due to other risk factors like drinking or smoking. Instead, when the relationships are teased apart, the risks were mainly additive.
When one then looks at all cause mortality of participants in this study, one gets a feel for what risk is taken as BMI increases. The top line is for smokers, the bottom for never smokers.
While being too skinny puts one at some increased risk, and too fat increases your risk commensurate with the level of obesity, smoking no matter what is bad bad bad. I read this as showing a healthy BMI to be between about 22 and 27, with risk increasing linearly below or over those levels.
And finally my favorite aspect of this study is the dose-response relationship between obesity and all cause mortality.
In these figures, a left shift of the curve = bad. It means in general you will die sooner if you are in this category. The BMI ranges of 20-25 and 25-30 aren’t too different and are the best curves to be on. But once you go over 30, and especially over 40, the risk of death increases dramatically, just like a dose-response. The risk is kicking in at ages over 50 for males and over 60 for females. Indeed, for morbid obesity the authors compared the risk as being about equivalent to being a life-long smoker, and reducing your life by about 10 years.
We’ve talked a lot about obesity here at denialism blog, and as it is with all situations in which there is a group of people that desperately wants the truth not to be true, there is some crankery surrounding the health effects obesity. But I think that based on this and other studies we can assert some things as facts.
1. Obesity increases your risk factors for vascular death and some cancers, and in a dose-response relationship increases your risk of death especially over a BMI of 30.
2. We have sufficient evidence for a causal nature of this relationship based on studies such as these, as well as our knowledge of physiology, the relationship between obesity and the risk factors such as diabetes and cholesterol, and the reversal of these risk factors after interventions such as bariatric surgery (both gastric banding and bypass although bypass also has additional effects due to changes in physiology after the surgery and before the weight is lost). When the weight goes on you get the risk factors and the mortality, when you correct the weight the risk factors disappear, and the mortality (aside from surgical mortality) reflects the new weight status. It would be unethical to do the gold standard test of making people gain weight and then increase their risk of death directly, so drop it.
3. “Normal” or “Ideal” weight is cultural and individual, but based on BMI what should be “normal” or “healthy” is between the range of 22 and 27. Currently > 25 is overweight, and we can argue that 25-27 is the “high” range of normal. But there is no point arguing over such small differences when it is clear that after this level obesity becomes increasingly dangerous.
4. It is easier to never gain weight than it is to lose weight. I argue for primary prevention rather than secondary prevention. People with BMI over 30 should try to lose weight. However at BMIs > 40 bariatric surgery should be a consideration, and earlier rather than late to avoid increased risk of surgical complications, and increased time with higher risk factors for vascular disease.
Prospective Studies Collaboration, . (2009). Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies The Lancet, 373 (9669), 1083-1096 DOI: 10.1016/S0140-6736(09)60318-4