A recent NY Times article claims that people who are slightly overweight have “less risk of dying than people of normal weight.” The author goes on, as any responsible author should, to hedge her claim with quotes from various doctors suggesting that we shouldn’t “start gorging on fried Belgian waffles or triple cheeseburgers.”

Unfortunately, the damage is done. My Facebook feed is already inundated with this article, and it is only a matter of time before people simply start disregarding their doctor’s advice, waiving a NY Times article in the air as they enjoy their deep fried bacon-wrapped Twinkies generic snack cake.

To begin, let’s go back to the original report and look at what they actually found. Katherine Flegal, PhD, and colleagues decided to conduct a meta-analysis of 97 different studies examining BMI and mortality. As a baseline, they took the group of individuals typically classified as “normal” (BMI 18.5-<25). They then compared the mortality rates of different BMI groups using what’s called a Hazard Ratio. A Hazard Ratio is simply a comparison between two groups on an outcome (in this case, death). In other words, how likely are you to die?

What did they find?

1. Overweight group (BMI 25-30) were 6% less likely to die than the normal (18.5-25) group.
2. Slightly obese group (BMI 30-35) were 5% less likely to die than the normal group.
3. Very obese group (BMI over 35) were 29% more likely to die than the normal group.

 

The inclination for many people here is to measure one’s own BMI and feel relieved as long as your BMI is less than 35 (“34! I’m totally healthy!” *eats chocolate donut*).

Of course this is ludicrous. As the NY Times article begins to touch on, “BMI is an imperfect measure of the risk of mortality.” More accurately, BMI is a measure of relative bodyweight (to height). It is simply the ratio of your total weight to your total height. In other words, it is not a measure of body fat at all! BMI was created in the 1800s by Adolphe Quetelet (a Belgian statistician) so that he could quantify body size for some of his research. It was never designed to act as a replacement measure for obesity; at best, it acts as a first step in diagnosis (if this patient has a very abnormal BMI, we should examine further to ensure they are not underweight or obese).

 

In other words, BMI is not a measure of obesity, and the NY Times article fundamentally misses this key point. As it goes off on a tangent discussing the the protective effects of different types of fat, it forgets to address a more important issue we face in medicine: proper diagnosis. When we use BMI to declare someone obese in the absence of other measures, we undermine our ability to help truly obese patients. “Arnold Schwarzenegger has a BMI of 33, therefore I must be in perfect health too!”

In short, this excellent new research doesn’t point to the protective effects of fat as much as the need to stop relying on BMI to assess obesity.

 

Update (02 Jan, 2013): Response to comments on Facebook

A thoughtful reader responds:

Here is one paper and a version of a comment I just left on my own facebook page, finding that BMI has a good correlation with BF% ( in men, R^2 = 0.44 and women R^2 = 0.71 — in women that’s just crazy high; in men it’s still strong. R^2 is the square of the correlation, so that’s c correlation of .66. Maybe it’s because I’m a social scientist, but that’s what I call a strong correlation.) The paper finds that BMI is sensitive but not specific for diagnosis, i.e. that more people have adipose tissue in excess of (current, perhaps mistaken) recommendations than have a “high” BMI. If *that’s* the most common problem with BMI, I don’t see how that does anything to contradict the inference from the current meta-analysis that fat may be protective (it might complicate it — maybe the riskiest group to be in is “skinny fat,” having plenty of adipose tissue, but not so much with the muscle mass or bone density.) This is not to say that BMI is what people should be using to treat individual patients; I’m only talking about inferences that would be reasonable to draw from the meta-analysis. Obviously, it’s just one paper — I probably shouldn’t spend the time to do an entire review of the literature on BMI/adiposity correlation right now.

I agree completely that BMI can be useful as an initial diagnostic tool to determine whether or not further assessment is necessary. This is similar to total cholesterol levels or liver enzyme levels. On their own, they are insufficient as a diagnostic criteria, but can certainly be used as one of many tests to ascertain health. The concern I have with the NY Times article (not so much the JAMA article) is that the author uses sensational journalistic language (e.g., the title “Study Suggests Lower Death Risk for the Overweight”) that many lay-readers would take as gospel and simply assume that the next health craze should be to become overweight at any cost in order to increase longevity. This has happened countless times before: most fad diets come from these kinds of oversimplified descriptions of research (e.g., Atkins diet). It’s not that there is no truth, but rather that nuance is already lost when we try to explain research in non-scientific terms, so to go out of one’s way to sensationalize and grossly oversimplify the findings is, in my view, harmful.

 

Reader Rebecca Weinberger responds:

You don’t think BMI is an accurate measure of fat or risk (this we agree on) but you do feel comfortable saying it’s important for people to be scared of getting fat, lest they eat things you don’t want them to eat, “My Facebook feed is already inundated with this article, and it is only a matter of time before people simply start disregarding their doctor’s advice, waiving a NY Times article in the air as they enjoy their deep fried bacon-wrapped Twinkies generic snack cake.” That’s pretty fucked up. That leads me to believe that you’re not really interested in finding a more accurate measurement tool, but you are definitely interested in holding on to authority through scaring people.

Rebecca makes an extremely important and excellent point regarding our role and stance as healers in the community. A deeper look at the discussion that followed and the implications for the medical community can be found here.