UPDATE: Click on the screenshot to read Flegal’s new essay about what happened to her. Thanks to several readers for sending me the link.
______________________
The Sunday edition of the Boston Globe has a free article (click on screenshot below) by Amy Crawford discussing the results of Katherine Flegal and her colleagues on the relationship between weight and health. Flegal was a CDC scientist when she did this work, though she’s now at Stanford. What she and her coworkers did was correlate the relative risk of mortality in five classes of people with different body mass indices (BMIs; which are calculated by one’s weight in kilomgrams divided by the square of your height in meters). The classes are these:
BMI < 18.5, classified as “underweight”
BMI 18.5 to less than 25, classified as “normal weight”
BMI 25-30, classified as “overweight”
BMI 30 to less than 35, classified as “obese”
and BMI over 35, classified (I think) as “morbidly obese”
Flegal et al. also looked at the relationship between weight and mortality in three age classes: 25-59 years, 60-69 years, and 70 years or older. They used mortality data from surveys by the National Center for Health Statistics over three periods of time: 1971-1975, 1976-1980, and 1988-1994, as well as the single year 2000. Since their first paper was published in 2005. there have been more studies and one meta-analysis of all the data (see second screenshot below). Not all the results jibe, but most of them do, at least in showing the most surprising outcome. Read on.
The data, published in the prestigious Journal of the American Medical Association in 2005 (see second screenshot below), showed, with appropriate controls, that while people who were obese or morbidly obese had, as expected, a higher death rate than those of normal weight, as did people who were underweight, people with BMIs of 25-30 (overweight but not obese) had the lowest death rates of all! In other words, you were better off being a tad plump than being “normal”. I had heard this before but hadn’t seen the data.
The Boston Globe article concentrates on the firestorm that the data of Flegal et al. produced, apparently because it went counter to the conventional wisdom that to be the healthiest, you should have the “normal” BMI. The study was apparently sound, and yet a professor of epidemiology and nutrition at Harvard’s School of Public Health took out after the study, calling it “deeply flawed”, and her results were attacked over the years. It was an early example of “cancel culture.”
You can read the Boston Globe article by clicking on the screenshot below:
While it’s true that a few similar studies gave different results, most of them confirmed that a little avoirdupois was good for you, which enraged many people. As the Globe notes (my emphasis):
In 2013, Flegal followed up her 2005 paper with a meta-analysis — a review of published papers — in JAMA. In this research she and her coauthors reviewed 97 studies covering nearly 3 million people. Once again, they found people in the overweight category were least at risk of premature death. And once again, the paper roiled the fields of nutrition and public health, attracting special censure from Willett and his Harvard colleagues.
Willett emailed the CDC director to complain, arguing that the meta-analysis had caused damage and confusion and undermined public confidence in science. “Kathy Flegal just doesn’t get it,’’ he grumbled to a reporter for The Atlantic (“I go by Katherine,’’ Flegal notes). In an interview with National Public Radio, Willett said, “This study is really a pile of rubbish and no one should waste their time reading it.’’
Flegal’s confidence in her data, methods, and findings was never shaken. Still, the attacks took their toll. “There were all kinds of little things—it was like, wow, we’re under siege here!’’ she says. “I lost control of the narrative completely.’’
Flegal retired from the CDC in 2016 and took a part-time role as a consulting professor at Stanford. With more time to reflect on her experience, she finally decided that it was important to share her story. “It’s about getting it off my chest,’’ she says. “But it’s also like, wait a minute, this is not really right! People should know about this.’’
Flegal has now written an essay about her experience of being demonized, but I can’t find it online. I hope a reader can find it for us, and I’ll put it here above the fold. Several people in the article, including Alice Dreger, defend Flegal against the attacks.
Here’s the original study of Flegal et al (click to read, and there’s a free pdf). I’ll show both a graph of the relative risks from the paper, separated by BMI group and age class, and then the numerical data.
First, a plot of the relative risk data, showing that at all age classes it’s higher for people who are underweight, obese, or morbidly obese, set at 1.0 for BMI 18.5 to <25 (“normal”) and slightly lower for BMI 25-30, the “overweight”. The figures in the table below the graph (see below) are more informative.
Figure 1. Relative Risks of Mortality by BMI Category, Survey, and Age
And the table of relative risks, “relative” being to people of normal BMI. Notice that the error bars for the overweight often overlap 1, but are consistently lower than 1, suggesting a mortality advantage for being overweight. Note that the relative risk is set at 1 for the “normal class”. Note as well that from BMI 25-30, relative risks are lower, and substantially lower for never-smokers from 25-59 years old. If you’re between 25-59 and obese, the data also show a reduced relative risk, which is weird, but the rest of the data is as you’d expect.
The authors barely mention the benefits of being “overweight”, mentioning it only twice, and very briefly, and not dwelling on it at all. Did they sense the furor it would cause (my emphasis):
Descriptive data for the 3 survey cohorts are shown in Table 1. The numbers of deaths in the 3 cohorts were 3923, 2133, and 2793, for a total of 8849 deaths. Estimated relative risks are shown in Figure 1 by BMI category, age group, and survey, and relative risks from the combined data set and their SEs are shown in Table 2. Obesity (BMI ≥30) was associated with increased risk, particularly at the younger ages; the relative risks were lower in the oldest group. The relative risk in the overweight category (BMI 25 to <30) was low, often below 1. Relative risks in the underweight category usually exceeded unity (1.00). Relative risks were generally modest, in the range of 1 to 2 in most cases.
They found pretty much the same thing in the data for 2000, with BMIs between 25 and 30 associated with lower relative risks.In general, the authors say that they found less increase in relative risk among the underweight and obese than they expected.
Figure 2. Estimated Numbers of Excess Deaths in 2000 in the United States Relative to the Healthy Reference BMI Category of 18.5 to <25, Shown by Survey and BMI Category
Now why, if these data be true, does it seem to be good to have a little excess poundage? Nobody knows, but perhaps “normal” BMI has been improperly defined. It may be the median BMI (I doubt it), but most probably the BMI that doctors think is the healthiest. But it doesn’t seem to be, at least if you consider mortality as the ultimate arbiter of unhealthiness. If I were to hazard a guess, I’d say that in our ancestors, being a bit overweight gave you a selective advantage, as you’d be better able to weather times of famine, and so the evolutionary mortality “setpoint” was set at having a bit more fat than doctors like to see today. But this is just a theory which is mine, and I have no idea whether it’s true.
Below is the meta-analysis by Flegal et al. mentioned above, also published in JAMA, and I haven’t yet read it. I’ll leave that as an exercise for readers (it’s free if you click on the screenshot), who will surely want to discuss this long-known but surprising result. Remember, we don’t know how much confidence we should put in it, as all scientific “facts” are tentative, and later analyses may give different results:
h/t: Tim
Her account of what happened, maybe this: https://www.sciencedirect.com/science/article/pii/S0033062021000670?via%3Dihub
Thanks Maurice for the link.
Wow!! She does an admirable job at summarizing the many attacks, but sheesh, it must be exhausting!
This reminds me of the attacks on Alfred Wegener for his theory on continental drift.
BMI is a ridiculously simplistic measurement. Most athletic people are classified as overweight because muscle weighs more than fat. Since I was a teenage, sometimes I have been physically fit and sometimes I have been fat but I have always been BMI “overweight.”
Curtis, I promise you, one pound of muscle weighs the same as one pound of fat.
Muscle is denser, by quite a bit, than fat. Thus, it is true as Curtis says, that many athletic people (including essentially all body builders) would be classified as “overweight” and even “obese” by the BMI measure, which was an ad hoc measurement improvised by a mathematician, never originally based on or reflecting any verified medically relevant information.
Apparently one pound of humor isn’t equal to one pound of laughs around here….
The humor gods need their pound of flesh.
It depends on whether it’s vitreous or aqueous.
Robert, the BMI categories (at which BMIs you are underweight, normal weigh, overweight, obese, etc) where set based on medical information.
See World Health Organization: Obesity: Preventing and managing the global epidemic. Geneva, 1998
The categories were based on some information, but it is not very convincing, because it’s such a blunt instrument. But the very measure itself was not created based on any demonstrated medical research at the time. It just became popular because it’s easy.
But neither muscle nor fat contribute significantly to height, which is the other factor in the BMI calculation.
Okay…
Joking? Or did you really misunderstand what Curtis meant?
I read it as a joke. But the original point is the one I was going to make as well, about the inaccuracy of BMI. It’s misleading. A more accurate measure is the ratio of fat to muscle.
No doubt you are right, my bad. My jokometer is woefully out of wack these days.
One of the huge problems with BMI is it does not distinguish between lean tissue and fat. Any normal person who engages in moderate strength training (or works with their hands) will put on enough lean body mass to tip into the overweight category. Also, on the other side, you have smokers with COPD and emphysema and diabetics with low BMI’s as a result of serious health problems, so you are sampling people with low BMI’s because they are really sick, and people with high BMI’s who are really not very fat, and any conclusions derived concerning “obesity” should be taken with a grain of salt. BMI is easy to measure, body fat is harder and more complex (not everyone puts it on in the same place so skin fold tests, etc., aren’t necessarily super accurate and lean body mass scales are expensive), but you really need body fat to make any meaningful interpolations.
Very good points.
Yes. A good summary here: https://www.everydayhealth.com/diet-nutrition/bmi/bmi-flaws-history-other-ways-measure-body-weight/
The authors conclude: “BMI has a lot going for it. It’s a simple, inexpensive, and noninvasive way to measure body fat. It can be calculated quickly and has been widely used, which is why it’s an easy way for statisticians to define obesity in general populations.
Health practitioners should be aware of BMI’s downsides and should recognize the limitations. Even if doctors measure BMI at routine appointments, it shouldn’t be the only factor that’s used to diagnose someone as overweight or obese”.
BMI is a lousy measure, waist to hip ratio (no, not the one about sexual attractiveness) is somewhat better, but also far from perfect. There are many other measures, but they are not as easily measured. Somehow I’m not surprised that slight overweight (as determined by BMI) is associated with better outcomes .
Note, the Flak Flegel got was unwarranted, and way beyond what a scientific discussion should be. The way she was dissed was really unconscionable.
The intensity of the debate reflects the high stakes for public health. If obesity increases morbidity and mortality, then researchers like Harvard’s Walter Willett had good reasons to be very annoyed with Flegal’s work.
Yes, I’m no fan of the BMI. I also think underweight people are completely passed by when it comes to being ill. As a long time anorexic of the past, I can tell you it’s no party but of course our society favours slim bodies, especially with women (who are often starving), and conversely sees those who are overweight as having a character flaw (I think in a way to dislike them and still feel good that they are disliking them for an accepted reason). Thank goodness fitness programs like Apple’s Apple One show women with actual muscles and actual athletic bodies – remember the crazy thin bodies of exercise instructors in the 80s? How could they walk with those stick legs let alone exercise?
I watched the Taylor Swift doc on Netflix (it was good! She seems like a very neat and talented young lady!) and she talked about how during her thinner times, she was always feeling on the verge of passing out after shows, and thought that that was a normal thing. Once she began eating more, she realized that it was not, in fact, normal to feel that way all of the time.
Swift is a good example of another factor not being considered. She started her career as a teenager. I think it’s normal in teen years to be thin. When I was in high school I weighed about 130 lbs at 5’8″, no matter how much I ate. I’m quite heavier than that now decades later, but don’t consider myself obese.
What she’s facing now is the expectation that she continue to look like a teenager, when that is not realistic anymore. She’s 32.
Gaining weight disproportionately as you age (which Taylor Swift has not done) is not normal biology, though it is a statistical reality in the US and in many other rich countries and some poorer ones.
Japan has an obesity rate of about 4%. The US rate is about 10 times higher. Japanese immigrants to the US who adopt the nowadays typical US diet (SAD = standard American diet) and lifestyle gain weight disproportionately as do US non-immigrants and get the same chronic diseases. So it’s not genetics that explains Japan’s 4% obesity rate. It’s mostly that the Japanese eat less and eat differently (a more traditional diet).
How the rest of the world sees Americans (even if the actors cast as Americans in this instance are of the Canadian persuasion):
I suspect there are regional differences in the U.S. on this. I recall on every occasion when visiting the east or west coast, we’d notice how thin and fit people seemed (well, there were a lot of them, always running and looking toned and beautiful). But then as we’d drive back to the midwest we would ascend the gradient of obesity.
Anecdotally, I went the reverse direction. I was a bit chubby in my teens and early twenties and dropped it all off by 30 and now have a pleasantly slimmer plateau where I know what I can eat to maintain without effort.
Agreed. Why should one weigh significantly more at 55 than they did at say, 25? Unless one was severely underweight and/or deconditioned when they were younger, the only real explanation for large weight gains in middle age and beyond is greatly increased fat storage. Gaining many kilos of healthy muscle tissue in one’s later years is extremely difficult, and therefore almost never the reason for the weight gain.
So yes, no reason for older people to be walking about with these big guts and butts and blaming this on a “natural process”.
It’s worth pointing out that Taylor Swift is also above average height – apparently she’s 5″10″. So I’d guess she tends to have a naturally fairly slim frame that goes along with that.
Oh yes. You have no energy. I used to sleep all the time. It got hard to keep up with school work. I’d get bruises easily that wouldn’t heal very fast and I’d get about 13 colds a year.
”The 20 Minute Workout’, Canadian TV, was an example back in the early ’80s.
So the Canadian Ski Marathon, not a race, but 170 km over Sat/Sun, I used to call my 20 Hour Workout. That was considered slow. I don’t think I’d ever have been considered skinny. No TV there.
But in Norway TV they make a big fuss about really old farts, 80+, who finish the
(real) Birkebeiner, 55km over top of two little mountain ranges, Rena (of some historic fame for the King’s escape from Hitler) to Lillehammer (1994 Olympic stadium these days). Bigger fuss almost than the winners get.
But back on topic, getting away from the unrealistic woman weight expectations has been a very good thing, though carried too far in the opposite direction at times. Media should just shut up about specific aims and try to use only advisers who know what they are talking about.
Another factor: Osteoporosis is a major cause of death of older women, due to the number who die after hip fractures. A little extra weight stresses the bones and encourages one to retain bone mass. Some people have speculated that this may be the reason that women are so prone to overweight after menopause; that it is nature’s way of protecting the bones.
BMI is a poor indicator of longevity except the extreme cases. Especially, as others indicate, not distinguishing between fat and muscle.
As I recall, there is plenty of difference between male and female as regards to average percentage of body fat. This came a few decades ago to me when studying literature related to endurance sports. In my case that is mainly nordic (AKA cross country) ski racing, which is still large in my rather elderly interests. I’m wondering if anybody here has well considered opinions concerning how accurate any methods are for determining that percentage.
In any case, and despite the fact that many people are not able to extract info very well from a graph, it seems pretty obvious that guessing longevity would be far better if one had graphs giving likely mortality as a function of TWO variables, (BMI , % BODY FAT).
A few colleagues and I were playing with the navy version of the body fat % calculator the other day. It had you take your height, weight, gender, neck measurement, hip measurement & waist measurement and give a percentage response. To get the number to drop from normal to athletic, you have to increase the neck measurement, however, you could decrease the weight to zero and it had no effect.
–
One of the reasons I think the healthy at every size movement was on to something was the focus on being able to do the physical activities you want to do in a way that you feel strong and healthy, rather than clothes size, weight, BMI or other measurements.
I always thought you could never be too rich or too thin. At least the first one still holds.
Randall, as to “You can never be too rich”- well, it depends on how you amass wealth. See, for instance, Allen Weisselberg: he supposedly tried to become more rich by cheating the IRS. That’s risky. I think he’s going to prison unless he helps prosecutors to land bigger fish.
Well, as we often read here (and elsewhere), correlation is not causation. Separating non-smokers makes sense, but there may be other illnesses lurking under the radar in the lowest BMI (and normal?) categories. Anorexia has been mentioned above. Cancers?
PCC’s hypothesis is akin to the “Thrifty Gene” hypothesis for the prevalence of Type 2 diabetes in certain populations.
https://en.wikipedia.org/wiki/Thrifty_gene_hypothesis
That was the usual explanation with which one justified not mentioning the higher risk of slim people in the past. But what’s normally done in such studies is to discount the deaths that occurred in the first years after the body mass measurement to rule out undiagnosed severe illness. Flegal’s results weren’t at all unusual. Only normally inconvenient parts of the results that don’t fit an established agenda don’t actually get mentioned. It’s like that in lots of subjects.
Maybe being a bit overweight gives you that little bit of an edge during a severe bout of illness, because you got more reserves?
Ya beat me to it.
Meanwhile, I’m gonna sing along with Weird Al.
BMI is a rotten measure of individuals. Adolphe Quetelet was interested in comparing the relative weights of populations and realised that he needed a correction for height. Obviously a power of one in the denominator was wrong, since it assumed that taller people were the same width as shorter ones. A strict geometric ratio with a power of three was equally wrong, since it assumed that taller people were scaled in exact proportion to their height. He settled on a power of two as a quick and dirty correction. Recent studies have suggested that a power of 2.3 to 2.7 is more appropriate. The blind use of a power of two is biased against taller people, and as people grow taller many more are condemned as overweight.
With regard to weight, this is the paradigm I’ve lived by for the last 10 years or so: “Weight is the dependent variable.” The independent variables are as follows:
– Your health metrics
– How you feel physically
– What you would like your body to be able to do, and
– Aesthetics/how you look naked
How we weight (excuse the pun) each variable is up to each individual, although I would argue that the health metric variable should come first.
So for example, if Person A is into distance running, Person B loves to bodybuild, and Person C is only seeking to do the minimal physical activity necessary to maintain health, then their “correct weight” may be widely different, even if they are similar height, age and gender.
In my case, I was advised by my doctor to lose about 5 kilos due to high blood pressure, even though I felt fine, and was satisfied with how my body performed and how I looked. So therefore the paradigm kicked out a weight for me of about 85 kilos, which did the job of reducing my blood pressure, and it’s a weight I’m now used to, even though it still puts me in the BMI slightly overweight category.
I could easily see my paradigm resulting in some people needing to GAIN weight.
In 2013 the AMA designated obesity a disease. Not everybody agrees. Dr. Pieter Cohen of Harvard Medical School says that’s absurd, since it would mean that the 35% of adult Americans categorized as obese are in a diseased state or sick. Theodore Dalrymple, who once worked as a prison psychiatrist in Great Britain, has a similar attitude toward viewing alcoholism as a disease, calling it “a moral confusion typical of our age.” His view is that alcoholism is a vice, not a fate: “‘An admirable evasion of whoremaster man,’ as Edmund says in King Lear, ‘to lay his goatish disposition to the charge of a star!’ – or his neurones.” Some also call cigarette smoking a disease but nobody got cancelled for ‘tobacco shaming’ back when I was quitting (or now). Bill Maher got cancelled recently for arguing that the lack of sympathy we give cigarette smokers helps them to quit.
Even if obesity is not a disease, and all of the research that implicates obesity as a major cause of many diseases turns out to be false, and society does a 180 on beauty standards and considers obesity the pinnacle of sexiness, then I still have this question for obese people: are you actually comfortable with all of that weight on your body, and how does this affect your ability to move and do things?
I have several obese members of my extended family, and when we visit them, we are stuck doing sedentary things, like eating and watching tv. Even walking around a mall is considered taxing. Their teenage children are starting to become this way as well…we can’t do things like play basketball or go on a rope course because they get too tired. There’s a real problem when your 45 year old has-been uncle can outrun you on the basketball court!
I’m sorry, but I like to move around and still enjoy physically strenuous activities. I couldn’t see doing any of that if I was obese.
Assuming that the study is correct — that being slightly overweight correlates with longevity – then one has to ask why that might be. It could be b/c there is a co-association with a higher income, and hence access to a healthy diet (albeit with a bit of imbalance on in portion sizes and physical activity), education, and access to good to health care. Those who “let themselves go” too far in the overweight category could more likely to be missing one or more of these factors. I’m just spitballing here, but one has to wonder.
I am disappointed that nobody has picked up my point in comment #10. It is simple BMI is a faulty measure that tends to rate taller people as overweight, better nutrition means that people grow taller, and some of them then automatically fall into the “overweight” category when they are actually healthier.
The faults of BMI were pointed out by several other people.
They were nor specific: I was.
What’s for certain is that some topics seem to be able to boil the blood, in many different ways. When someone argues for a slimmer ideal, it’s guaranteed that the #fatphobia brigade throws their weight into the ring. Maybe everyone needs to learn that this “petition culture” is just how it is, from now on.
I’ve gone through thick and thin periods – right now I’m the most overweight (but not obese) I’ve been in my life. The times when I was at my lowest weights were times when I would get sick the most.
Under the premise that the BMI “overweight” range is ill-defined in terms of mortality, has anyone analyzed the Flegal, et al data to see if there is a better definition for that range? Is there an increased mortality at some point in that range, or is that not seen until BMIs that are labeled as “obese:”?
I’m 5’9 and 114 pounds or less for years now so I’m somewhat underweight – it is no picnic. Girls don’t like it for a start. (I don’t have an eating disorder).
D.A.
NYC
I was a “TOFI” – Thin Outside Fat Inside” as shown in a DEXA scan
(Dual-energy X-ray absorptiometry is a means of measuring bone mineral
density using spectral imaging – cost around AUD$100, done in a few minutes).
It’s low dose, but still some ionising radiation.
Main concern was excess visceral fat around my organs.
You get what looks like a CT scan of your body
showing muscle, fat and bone plus an estimate of your resting metabolic rate.
I tried the 5-2 diet, lost 10Kg and lost some of that visceral fat
which noticeably reduced the occurrence of my
Atrial Fibrillation (AF) symptoms.
Apparently that excess fat can cause inflammation of organs, eg heart muscle.
Now, aged 77, for other reasons, I am trying to put weight back on!
I believe that there is a major cognitive problem over issues like this. Yes, good quality epidemiological analysis will confirm tendencies and averages for a particular population but there is no assurance that all individuals will benefit from the findings. So, smoking cigarettes is generally bad for people, but some people can smoke and encounter no ill effects. The trick is identifying which people will suffer, and those that wont. The confusion between population based data and individual variation is too easily brushed off.
So some (probably many) people will suffer if they are too fat, and some (probably comparatively few) people will suffer if they are too thin. But there will be a lot of people who will not suffer, whatever their weight. Should they be concerned about entreaties from the health experts? There is a tendency for ‘expert advice’ to become hectoring if not challenged by reality, but it is also too easy to dismiss if you can think of one counterintuitive example…
Classic article explaining why BMI is a useless measure for individuals:
https://www.maa.org/external_archive/devlin/devlin_05_09.html