The obesity paradox paradox

by John Q on June 19, 2013

I see lots of stories made up of handwringing over the “obesity paradox”, normally presented as saying that even though obesity is a risk factor for all kinds of diseases, obese people appear to have lower mortality than others. A typical finding is the one reported here

being overweight or slightly obese was linked to about a 6 percent lower risk of dying, compared to people considered “normal weight. Being severely obese, however, was still tied to an almost 30 percent higher risk of death.

People are tying themselves in knots over this, but it doesn’t seem to me that there is any paradox to be explained. The obvious reading of the data is that the Body Mass Index.[^1] ranges used for the various categories (20-25 Normal, 25-30 Overweight etc) were set a bit too low when they were originally estimated, or rather, guessed. From my quick look at the data, if you bumped the ranges up by a couple of points, the paradox would disappear. People at the bottom of the current normal range, who tend to have high mortality, would be classed as underweight, while those currently classed as slightly overweight would be reclassified as normal, and so on.

Am I missing something?

[^1]: This point is logically separate from the general problems of the BMI, regarding muscle mass and so on.



Marc 06.19.13 at 9:08 pm

The fact that the recommended ranges are in units of 5 lends credence to what you’re saying. If you were designing a diagnostic tool you’d look for a BMI point where mortality increased measurably, or where there was a change in slope of the mortality-BMI curve (e.g. BMI higher than this level is correlated with a 10% reduced lifespan over the study period, etc.)


adam.smith 06.19.13 at 9:11 pm

The other thing is reverse causality, as explained in the article:
“Some have argued that the pattern is a statistical one only because being thin, especially in old age, is often a sign or a result of serious illness – so the thinner people seem to have higher mortality.”

Attempting to control for that (which is hard) can make quite a difference, IIRC.
That said, the lower threshold for normal weight seems insanely low to me, so your theory sounds plausible to me. I’m skinny by any reasonable measure and would have to lose almost 10kg to get to a BMI below 20 (underweight is sometimes <20, sometimes<18.5). Obviously that's anecdotal and BMI isn't really supposed to be useful at the individual level, but still.


Omega Centauri 06.19.13 at 9:33 pm

One problem is that BMI uses an integral exponent on the height, because they thought doctors wouldn’t be able to handle a non-integer power. So the extrapolation with height is poor. I don’t remember the correct figure, but those with heights not near the center of the range will have skewed BMI values.


Donald Johnson 06.19.13 at 9:34 pm

The whole BMI thing is a little weird anyway–why should one take the square of the height and divide by the weight? If people were geometrically similar you’d cube the height–since we’re not, there’s no particular reason why one would use the square as opposed to some other exponent, except that’s what Quetelet did when he invented the idea–

Quetelet article at wikipedia (Someone pointed this out Quetelet’s role to me in some earlier thread where BMI came up.)

So maybe a 6 foot person with a BMI of 25 (for example) really isn’t in the same cohort (healthwise) as a 5 footer with the same BMI. It might be better just to show what weight range gives the lowest mortality at each height, rather than use some artificial statistic like BMI .


Anderson 06.19.13 at 9:37 pm

Yeah, BMI is crap. (And not because it says I’m overweight. I *am* overweight.)


John Quiggin 06.19.13 at 9:44 pm

To restate the point in footnote 1, the difficulties with BMI aren’t the problem here. The paradox arises in people who aren’t athletes, and whose height distribution is that of the general population, so that BMI is, on average, a pretty good proxy for body fat percentage. The problem is with the ranges, not the measure.


lupita 06.19.13 at 10:57 pm

The incidence of diabetes among Asians and Latin Americans is higher at comparable weights so the tables had to be recalculated in these countries. The original weight-height tables had just been adopted as gospel before doctors realized that certain races are more tolerant of fat and sugar than others. In some places, doctors recommend that people be even thinner than those tables suggest.


bob mcmanus 06.20.13 at 12:09 am

And James Gandolfini died today of a massive heart attack at 51.


Glen Tomkins 06.20.13 at 1:24 am

This is basically the answer.

The “ideal” body weight was originally set empirically, as the height-weight with the best survivability, least incidence of disease. The reasonable interpretation of there being such a trough in the relationship between height-weight and disease, is that high weight generally causes health problems, while low weight is caused by health problems.

In the decades since that actuarial determination of IBWs, the mean body weight of the population has been creeping up. People who are now at the old healthiest body weight have been left behind at what is now statistically aberrant. In enough cases to affect the mortality statistics, the reason for that aberrancy is that they have some disease that is keeping their weight down. The general population is a bit sicker because of that overall increase in weight, but people who haven’t shared in that increase have failed to share those “gains” so often because some illness is holding them back, that being even a bit thinner than the new mean is correlated with bad health outcomes.


Andrew 06.20.13 at 1:33 am

As I understand it the range was lowered sometime in the late 1990s, ‘healthy’ was up to 27 or 28, then it was dropped to 25. It made about 300 million overweight or obese overnight! For me the peculiar part of BMI is the rampant use of it as a measure of individual health. It is a population measure of mass, not an individual measure of health.


duckbilledplacelot 06.20.13 at 2:37 am

Straight from wikipedia:
“In 1985 the National Institutes of Health (NIH) consensus conference recommended that overweight BMI be set at a BMI of 27.8 for men and 27.3 for women. In 1988 a NIH report concluded that a BMI over 25 is overweight and a BMI over 30 is obese. In the 1990s the World Health Organization (WHO) decided that a BMI of 25 to 30 should be considered overweight and a BMI over 30 is obese, the standards the NIH set. This became the definitive guide for determining if someone is overweight.”

So the definitive guide for determining if someone is overweight came about from someone’s affinity for counting by fives. Meanwhile, HAES makes very little progress, and the idea of ‘getting healthier’ still includes surgeries with huge complication rates that remove major parts of people’s digestive systems…..


Yarrow 06.20.13 at 4:02 am

Here’s a graph from Nature: at younger ages (say below 50), viewing a BMI of 18-25 as normal and 25-30 as overweight makes some sense; even at 50, a BMI of 18 is no worse in terms of mortality than a BMI of 30. At age 60, a BMI of 21 or less is worse than a BMI of 30; and age 70, a BMI of 24 or less is worse than a BMI of 30. (And a BMI of 18 is about as bad as a BMI of 40.)

I think the “paradox” is mostly aesthetic — thin people are supposed to be healthier than fat people.

(Article here)


Glen Tomkins 06.20.13 at 4:32 am


At younger ages, being at a lower than average weight is more likely to be the result of a healthy lifestyle, so better health correlates with lower weight. At increasing age, having some disease or multiple diseases takes over as the most likely cause of weighing less than average, so the trough of maximum correlation to good health moves up the BMI scale.

Much of the confusion over the “paradox” results from people misunderstanding the point that the ideal body weight trough reflects not just the fact that weighing less causes good health outcomes, but that bad health causes lower weights. Because of that latter dynamic, the actual weight you should aim for to maximize health is a bit less than the “ideal” body weight they arrived at actuarially. The weight that correlates with best health outcomes is pulled towards higher BMIs by that reverse causal dynamic, that bad health tends to take the weight off. But the weight that tends to cause good health outcomes stays low as that weight of max correlation moves up the BMI scale.


Evan 06.20.13 at 4:56 am

The reason people struggle to understand this is that they’re viewing body size as a moral issue, not a physical measurement. Their religion says anyone with a BMI of 28 is a sinner, and it’s bewildering to them that God isn’t punishing sinners.


Glen Tomkins 06.20.13 at 5:11 am


The existence of people who are self-righteous about overweight does not diminish in the least the real health hazards of overweight. God doesn’t punish the overweight any more than He punishes smokers, but both nevertheless have bad health outcomes at higher rates than non-overweight non-smokers.


pjm 06.20.13 at 6:56 am

@14, Glen properly speaking we don’t really know if being overweight is the cause of health problems or an association. High blood pressure is often assumed to have a causal connection to a number of serious health issues but the connection between hbp and obesity is not so direct. The commonly observed pattern in low-carb treatment of obesity is that blood pressure improves very quickly, i.e., before much weight has come off, so it is apparently not the extra weight per se that is the cause of the hbp. But to the extent that most obesity of probably derives a particular pathogenesis this is somewhat a distinction without a difference.


reason 06.20.13 at 7:24 am

What Adam Smith @2 says. Some illnesses make people thin.


John Quiggin 06.20.13 at 7:41 am

To restate my position: it’s true by definition that being overweight is bad for you, otherwise it wouldn’t be overweight. The question is “what weight is overweight?”.


PlutoniumKun 06.20.13 at 8:01 am

I’m no expert on the topic, but I believe that one known issue with conventional measurements (not just BMI) of obesity is that they don’t actually measure fat. Hence a very well muscled person (take the average rugby player in the Oz vs Lions match) might be technically ‘obese’ by some measures. This again, may be a component in Johns suggestion that the definition of obese is too tight and we need to just expand the definition of ‘normal’ to make that paradox go away.

Unfortunately, I can’t find the link, but I recall reading a summary of some research which suggested that quite thin people may have higher mortality rates when they fall sick (as opposed to rates of sickness). The idea being that somewhat overweight people have more ‘reserves’ to get them through a health crisis. Although I think later research cast doubt on this.

And as a thoroughly unscientific anecdotal comment, I live in an area with several hostels providing help for drug addicts and alcoholics. Almost none of them are obese, many are fashion model skinny. Might be a confounding factor.


Walt 06.20.13 at 8:46 am

Glen, researchers — not being completely stupid — try to control for the fact that bad health lowers your weight. As far as they can control for it, being slightly overweight still leads to better health outcomes.


conallboyle 06.20.13 at 8:50 am

Obesity is an effect not a cause. High time this blog took the writings of Gary Taubes seriously (The Diet Delusion, 2008 (different title in the US)). Most of the ‘official’ dietary advice is not just wrong; it is perverse!


prasad 06.20.13 at 9:11 am

Nature has a great article on this topic, with much more detail than the Reuters story. As Walt says, researchers are well aware of the fact that the connection between ill-health and weight can go both ways – they’re not idiots. To give one important example, they’ve been trying to separate out the effect of smoking for decades now, since smoking increases mortality but decreases weight.

I think it’s correct to say (based on current knowledge) the BMI ranges need to be shifted up, but that won’t be enough. The really important nuance (to me) is that the correct ranges are highly age dependent – as you grow older it’s actually good to gain some weight. The suggestion is that growing ill calls upon stores of energy. And the bodies of thin people don’t allocate/can’t summon sufficient resources to fighting disease.


prasad 06.20.13 at 9:12 am

Screwed up the link, sorry. The Nature story is worth much more than my blather. Science reporting at its best:


ajay 06.20.13 at 10:01 am

I believe that one known issue with conventional measurements (not just BMI) of obesity is that they don’t actually measure fat. Hence a very well muscled person (take the average rugby player in the Oz vs Lions match) might be technically ‘obese’ by some measures

But most technically obese people aren’t prop forwards – at a population level, it’s a good measure, and the existence of the Lions pack doesn’t undermine it any more than the existence of a 95-year-old smoker in perfect health undermines the link between smoking and morbidity, or the existence of an unemployed Harvard graduate undermines the link between a Harvard degree and a good job.
It undermines its use on an individual basis, but you shouldn’t be using it for that anyway.

I think it’s correct to say (based on current knowledge) the BMI ranges need to be shifted up,

The really obvious intuitive reason for this is that people are taller now. For purely geometrical reasons, BMI will be too ready to call people overweight when they’re tall. So the population-level health correlation will break down for a population that’s generally tall.

The breakdown at older ages is interesting; hadn’t heard that before but it makes sense.


Ray 06.20.13 at 11:59 am

The question is “what weight is overweight?”

That may also be the wrong question, because it assumes there is a single ‘correct’ weight, when there is a right weight range for a 20 year old student, a 30 year old rugby player, a 40 year old runner… and BMI is not a useful guide for any of these people.


Steve LaBonne 06.20.13 at 12:50 pm

There is no evidence I know of that, IF your blood pressure, blood lipids, and glucose metabolism are in fine shape, you’re at any significant health risk merely from being a little overweight per the standard charts. Somebody who fits that profile should not be harassed to lose weight.


Katherine 06.20.13 at 1:09 pm

it’s true by definition that being overweight is bad for you, otherwise it wouldn’t be overweight.

That’s nonsensical. “Overweight” is only “overweight” because we have added the word “over” to “weight”.

And as someone else said above, this seems to be classic case of causation vs correlation.

And there is some research now suggesting that it’s not the fat as such that is even correlated with bad health, but the effects of dieting. Because as any fule kno, dieting to lose weight has a 95%+ failure rate over the medium to long term.

Time to do some serious rethinking. I’d recommend some research into the Health At Every Size movement.


Glen Tomkins 06.20.13 at 1:31 pm


Some of the data out there isn’t adjusted for known health conditions because the big data sets that record the height-weight of large numbers of people, don’t have any data on their medical history, or have inconsistent data on that history. You can do such a case-control study just by looking through already gathered records, and get a huge n from which to do your statistical analysis.

To adjust for known health conditions, you really have to do a much more expensive sort of study, where you go out and query a cohort to get both height-weight and medical history, then wait the years it takes until they start to die off in numbers large enough to draw conclusions. But even if you do that cohort study, you can only adjust for medical conditions the patient knows he or she has, and for conditions you suspect correlate with weight.

Be that as it may, people still pay attention to the case-control, actuarial sort of study, because when you do the more careful cohort study, you find the same trough, you find that there is an “ideal” body weight for a given height, where the mortality is the lowest, and that trough isn’t in a much different place from the actuarial trough.

The point that JQ is addressing is that the trough has moved, that the ideal height-weight is now at a higher weight than it was decades ago, and that some people have taken that fact to mean that somehow the entire idea that overweight is bad for your health has been thus refuted. JQ points out that the difference in the data isn’t very big, and certainly not big enough that people should tie themselves up in knots over the difference.

My point is that you can make stronger statements against the overweight nihilism that this new data has brought out in so many quarters. The difference in the data, the upward march of the trough, is secular, and it coincides with an upward march in average weight. It was completely to be expected, and does not at all tend to refute the idea that overweight is bad for you, that this IBW trough would march north as the average weight of the population marches north. People left behind by that march towards obesity were left behind for a reason. Sure, for some, the reason is better diet, but the average weight of the population is going north, so that’s not the predominant reason. The reason to be at a relatively low height-weight, in a population gaining weight, is going to be predominantly that some illness is causing the lag, is the reason those relatively thin people aren’t keeping up with the Joneses in increasing their weight.


Josh G. 06.20.13 at 1:34 pm

The existing BMI recommendations are really an aesthetic and moral judgment masquerading as medical advice.


MPAVictoria 06.20.13 at 1:56 pm

“The existing BMI recommendations are really an aesthetic and moral judgment masquerading as medical advice.’

This is probably correct. No one is likely to be in danger of dying just because they are 5 or 10 pounds over their “ideal” weight. Now being a hundred pounds over is probably not the healthiest.


Rich Puchalsky 06.20.13 at 2:17 pm

I think a good way of disentangling some of the social thinking about this, as opposed to any strictly medical issues, is imagining an index based on frequency of sexual activity. Clearly people who score high are oversexed, right? There must be all kinds of risk factors for STDs, not to mention time lost from other activities and so on. While people who score low are showing healthy restraint. And there’s no reason to think that people slightly above whatever societal average may be enjoying anything, or that enjoyment itself may have some relationship to health.


reason 06.20.13 at 2:22 pm

John Quiggin @18
Yes but arguing from morbidity statistics is a bad start – because some illnesses make you thinner before they kill you.


pjm 06.20.13 at 2:24 pm

@18. John that is a a tautology. There is some really sound evidence that it is not obesity which is the cause of disease but rather the physiological state the accompanies most adult weight gain is what causes the damage, so it ain’t necessarily so. Relatively skinny people can have the blood chemistry, inflammation even ateriovascular buildup of people gaining weight (and die from the same chronic diseases associated with obesity – which also include cancer and neurodegenerative disease – without being overweight themselves).

@21, conal, OT warning: I think Taubes is going to emerge as scientific hero in not too many years. His writing gives a very sobering picture of how little hard evidence (i.e. things that are not nutritional epidemiology) goes into medical conventional wisdom (at least about diet and nutrition). For those who have read not him, he points out that current medical paradigm for how obesity happens is one that appears to have been adopted mid-century largely for non-evidence related reasons and contradicted most of the dominant school of thought among biomedical researchers in Europe prior to WWII.

Ironically in medical school, doctors are taught both the mechanism for the conventional modern view and the pre-1950 related view (caloric imbalance vs adipose disregulation via insulin) and doctors don’t even realize they are distinct (and unrelated) mechanisms. And only one mechanism is actually based in biochemistry (and it’s not the conventional, properly speaking I don’t even think a biochemical mechanism that corresponds to calorie imbalance even has been identified – probably doesn’t exist) On top of that, the animal models pretty much all support the older view not the current one. None of this contradicts that obesity is at least a marker for chronic disease but it does affect what you do about it and how you protect your health.


Marc 06.20.13 at 2:28 pm

Epidemiology is a tricky business. Our tools are pretty inexact – we can detect huge risks (smoking is correlated with lung cancer, extreme obesity with diabetes.) There is an online strain of denialism that shades into claiming that there is no relationship between weight and health at all that is misguided. The difference between a BMI of 25 and 35 for someone my height is ~40 kilos, and a weight difference like that has real consequences that are very well documented.

But more subtle effects are difficult to capture. I think it’s pretty likely that the difference between 25 and 26 BMI (3-4 kilos) won’t matter much. But we see fierce discussions about whether tiny levels of second-hand cigarette smoke carry health risks, or very low levels of drinking during pregnancy. It’s entirely possible that there is a statistical optimal weight for a given height / age and a few extra kilos do (slightly) reduce longevity. It’s also entirely possible that we can’t precisely measure the fulcrum point where that ideal place is.

But just because we can’t quantify small effects doesn’t imply they are absent.


Murray Reiss 06.20.13 at 2:38 pm

No statistician, barely numerate, I’m still back at the “6 percent lower risk of dying.” Ever? What does this mean?


Donald Johnson 06.20.13 at 3:01 pm

“The paradox arises in people who aren’t athletes, and whose height distribution is that of the general population, so that BMI is, on average, a pretty good proxy for body fat percentage. ”

Yes, I know you aren’t writing about BMI and I read the phrase where you could make the paradox disappear by slightly raising the recommended BMI numbers, but maybe it would be still better to stop using the BMI. Are there studies which show that non-athletic people with BMI’s of a given value all have similar percentages of bodyfat, even if they are of different height? If it’s supposed to be used as a proxy for bodyfat, then maybe someone should fit the data on bodyfat to height and find the most accurate formula. We all have calculators, so an exponent of 2.389 or whatever shouldn’t be a big problem here.


Donald Johnson 06.20.13 at 3:20 pm

Or better yet, just use some more direct method of determining bodyfat percentage if that’s the issue and then do studies to determine how that correlates with mortality. The idea that some simple formula works unless you’re an athlete seems a little odd, because it implies some sharp distinction between groups. Some people are just naturally more muscular than others, and people work out or don’t work out to varying degrees, or some work out and watch their diet and some work out and don’t.

Anyway, I’ll shut up now, except to complain about one other tangential point–if you follow the fitness and/or nutrition fields by reading the popular press (which is what I do), it’s all chaos and confusion if you get past the basics (exercise and eat a balanced diet). I could go on about that, but I just promised I’d shut up.


Bruce Wilder 06.20.13 at 3:25 pm

an aesthetic and moral judgment masquerading as medical advice

I’m wondering why sound medical advice would not, in the main, coincide with sound aesthetic and moral judgment.

I suppose the point may be analogous to the distinction between an architect’s interest in sculptural geometry and opinion about sound structural engineering. We expect that sound medical advice will derive from a technical analysis of underlying mechanisms.

pgm’s observations seem on target in this regard. BMI, derived from actuarial rather than epidemiological data, and connected to the know-almost-nothing model of caloric balance, does seem more than a bit thin, technically. I see JQ’s point, which is that leaping onto a contradiction or paradox bandwagon, when the effect is only an artifact of a tiny mis-calibration of ranges, seems unjustified.

Where’s the disease? remains a valid question, even if one gives up rationalization and excuses.


Limericky Dicky 06.20.13 at 3:28 pm

The best of all Body Mass Indices
Contemplates fat, but not thin, disease.
It floats like a freighter
In oceans of data
But sinks when it’s out of, not in, the seas.


Random Lurker 06.20.13 at 4:06 pm

“being overweight or slightly obese was linked to about a 6 percent lower risk of dying”

I can explain this paradox easily:
When I was 14, I was bit by a radioactive marble dog, and therefore I’m now immortal but, being a marble-man, I’m also slightly overweight.
Most other people, on the other hand, tend to have a risk of dying of 100%.
As I was in the panel in the study with other 16 overweight guys, our cohort had a 94% average risk of dying, as opposed to other cohorts who usually had a 100% average. The obese cohort had a 130% average risk of dying, which surprised a little the researchers, but is in line with the common sense theory that being overweight increases your risk of dying.


Josh G. 06.20.13 at 4:38 pm

Marc @ 34: ” But we see fierce discussions about whether tiny levels of second-hand cigarette smoke carry health risks, or very low levels of drinking during pregnancy.

And this needs to stop. It’s driven by thinly veiled sexism: the widespread societal belief that the bodies of pregnant women belong to society.


ajay 06.20.13 at 4:50 pm

No statistician, barely numerate, I’m still back at the “6 percent lower risk of dying.” Ever? What does this mean?

Per year. If 5 out of every 100 smokers die in any one year, and 3 out of every hundred non-smokers die in any one year, then not smoking gives you a 40% lower risk of dying.


Evan 06.20.13 at 4:55 pm

Glen @15, I’m not arguing that obesity has no health effects. I’m explaining the “paradox” John asked about. The people who decide these things could easily adjust BMI scales so that the purportedly “healthy” weight range actually lined up with statistical mortality, but I predict they will resist doing so as long as they can, because to most people weight is an issue of moral purity, not statistics.


ajay 06.20.13 at 5:01 pm

40: the drinking one; not the second-hand smoke one.


Bob Duckles 06.20.13 at 5:10 pm

Over the last two years, I’ve come down from a high of 248 pounds (BMI = 33.6-obese-) to 190 pounds to 190 pounds (BMI=25.6 – barely overweight – ) today. When I dropped below 200, my blood glucose got better and my blood pressure got better in daily checks. It’s an n=1, but suggest that in my case getting below 200 was a pretty good idea.


lemmy caution 06.20.13 at 5:17 pm


James Wimberley 06.20.13 at 6:01 pm

Omega Centauri (#3) is on to something: “they thought doctors wouldn’t be able to handle a non-integer power. ” The practical definition of numeracy has now changed with the universal availability of spreadsheets. It would be unrealistic to expect doctors or any other group of professionals to master all the mathematical operators offered by Excel, but it would be reasonable to expect them to understand a formula like “SUM((height_variable)^2.5)”.
Similarly there’s no reason not to adopt progressive tax rates as continuous spreadsheet functions, without bands.


pjm 06.20.13 at 6:05 pm

@44. Bob, congratulations. But you give a good example of the trickiness of this. Was “getting” below 200 a good idea, or “being” below 200 a good idea? (a flow vs stock distinction, as economists would say).
While I think there are arguments for why a high weight contributes to unhealthy blood pressure (which there is some reason to question) the relation between blood sugar and a given weight is not apparent to me. I.e., dieting usually means calorie restriction and carbohydrate restriction inextricably tied together. The connections between blood sugar and carbs are pretty clear, the relation between blood sugar and lower calories per se, not so much.

Does BMI, or any obesity measure, measure weight “gain” or rather the the condition of being heavy? This is very likely a distinction that has import for understanding the disease process. As for the paradox, that could be explained by the possibility that weight gain (which could, for example, be indicative of chronically elevated insulin levels) and not the weight level is really the determining factor. E.g., so if you have been gaining weight slowly for 10 years and now have a slightly elevated BMI, assuming it is the disease processes kicked off by weight gain that is the issue, your increased BMI is in contradiction to the fact that you may be no more measurably unhealthy than the population at large. Stock v. flow.


Marc 06.20.13 at 6:28 pm

Gaining weight certainly impacts your ability to exercise and the stress on your joints and heart from activity. How could it not?


stacks of books 06.20.13 at 6:42 pm

well according to reinhart-rogoff if BMI exceeds 90% of GDP then…oh wait…nevermind.


Wonks Anonymous 06.20.13 at 7:49 pm

The Willet vs Flegal spat (discussed in the article prasat linked) definitely makes me think something is rotten in public health science.


Andrew Edwards 06.20.13 at 7:55 pm

Seems like “mortality” is a really strange way of thinking about “health” for those under, say, 40.

Cause of death for those under 40 is virtually never related to weight, right?


leederick 06.20.13 at 9:05 pm

The BMI is set too low argument doesn’t get anywhere. The paradox is that obesity is a risk factor for all kinds of diseases, but obese people appear to have lower mortality rates. If we relabel obese as normal you still have the paradox that normal people appear to have lower mortality rates, even though normal weight is a risk factor.

The paradox is a discrepancy between high risk of specific disease, but low overall mortality, and our inability to reconcile the two findings. Nothing to do with arbitrary bands.


adam.smith 06.20.13 at 9:16 pm

reading the Nature article (thanks to prasad for the link and from me, too, warmly recommended), yes, leederick is right.
Changing the bands doesn’t make the paradox go away. The plausible theory there is that people with somewhat higher wait – especially older people – are more likely to survive certain diseases. E.g. a thin persons has lower risk for diabetes, but if s/he has diabetes it has a much higher mortality risk for her/him.
But also, the Nature article seems to side with Flegal somewhat, but it’s also clear that not everyone believes that the paradox even exists once you control for smoking and illness.


ezra abrams 06.20.13 at 11:31 pm

go here
and look at figure 2, bar chart of BMI (x axis) and incidence of diabets (y axis)
nuf said

Grandpa, is it true that you used to offer people donuts and stuff like that for free at work – people actually brought donuts in ? I mean, didn’t they know htat donuts are as addictive and toxic as cigarettes ?
Well son, attitudes are awfully hard to change. It seems incredible now, but cigarettes used to be “healthy” and it was considerd a good deed to bring in free donuts for your coworkers.


pjm 06.21.13 at 12:45 am

@54, 51. “something wrong in public health” Yup. Btw, the title of the article is almost certainly an allusion to Taubes. There is a quote from Taubes to the effect that every association touted by Willett and his Harvard team in a series of well publicized studies over the last 10 years has failed to hold up under scrutiny. Every single one, batting a big goose egg.

The big scientific issue is not about that obesity in general is unhealthy, almost all sides in the controversy agree about that to some extent. The big issues what are its causes and its relations to disease and both of those wrapped in the contentious issue of the role of dietary fat. And to some extent, there is a insurgency building in the medical community, especially among bariatric physicians but also endocrinologists that are looking askance at the received wisdom.


pjm 06.21.13 at 12:58 am

Also, I don’t know whether renormalizing makes the paradox go away (it would depend on how a number of statistical tests perform after the changing the model). To my mind, the significant problem is that even with a model of the relation between disease and BMI, there are very different interpretations about what the data shows and the science underlying some widespread interpretations of this are just not that solid.


Glen Tomkins 06.21.13 at 3:50 pm

This isn’t really complicated.

Obesity causes the metabolic syndrome. The metabolic syndrome causes coronary artery disease, and other atherosclerotic diseases. Coronary artery disease is far and away the biggest killer in the US and the rest of the developed world.

But coronary artery disease, while the single biggest killer, is responsible for far less than half of the mortality out there. Other diseases, each individually responsible for <1% of deaths, together cause most deaths. We need to speak even less concretely of "disease processes" causing deaths, because quite a large percentage of deaths have no firm attribution of cause. Many of these diseases cause weight loss, or at least weight to lag behind the population mean, months to years, to even decades, before death.

The result is a trough, a sweet spot, in the graph of BMI against mortality, the minimum of a rough parabola.

Why this state of affairs and its consequences should be thought of as a paradox, is beyond me. What is obvious, just from this discussion, is that a big part of the problem is essentialism, the idea that if a theory is true, it has to be the exclusive true theory of everything. The fact that the graph of BMI against mortality is parabolic, and not linear, in which case higher weight would always mean higher mortality, is taken as at least a difficulty for, if not a refutation of, the theory that obesity causes mortality. People who go along with the obvious truism that obesity kills are accused of maintaining that only obesity kills, which is a ridiculous straw man.

I think that many people have some misplaced idea of where medical advice such as the advice to try to get to IBW comes from. The IBW for a given height was not arrived at by any sort of process of analyzing the operations of the human body from the subatomic level on up to the individual to determine what height-weight would be the optimum for maintaining health. Medicine is a descriptive science. The IBW was arrived at actuarially, as that sweet spot on the graph of mortality against BMI.

Now, if we actually did understand human physiology well enough to be able to arrive at an IBW from the ground up, from first principles rather than from observation and actuarially, we could individualize advice to patients about their individual ideal weight. But we don't. What we know is that there is a welter and gemisch of normal and aberrant physiology working on them that we have no hope of disentangling. You can and should look for known diseases known to cause weight loss or weight lag if the individual patient presents with such. But you often won't find a culprit even if you hold out for some extreme cut-off as your weight loss of concern, and the culprit process for lesser degrees of weight loss or lag is almost certainly not well-observed or described enough to guide any advice or treatment.

What you're left with when faced with a 30 year-old obese patient who is not diabetic, hypertensive or hypercholesterolemic, is the fact that obesity puts him or her at greater risk for developing one or more of these three within the next three decades, and those diseases in turn greatly increase the risk of the greatest killer of your patients, atherosclerosis. No doubt, no doubt at all, that most of these 30 year-olds you hector about their weight, will not get any of these diseases no matter what their weight, so they will derive no benefit whatever even if they (miracle of miracles) actually follow your advice and get down to IBW. But that low rate of return is absolutely standard in medicine. Screening colonoscopy will, maybe, at the most optimistic estimate, prevent 2% of deaths. The other 98% of the patients you manage to convince to get the thing, will derive exactly zero benefit, and at the personal cost to them of enduring the thing. We accept such low rates of return because people tend to be so attached to living that they will accept low payoff means to that end.

Until and unless we can disentangle the gemisch, however theoretically complicated the underlying physiology, the practical question of what to tell 30 year-olds who are above their IBW is really quite simple and utterly non-paradoxical — "Lose weight or accept a higher risk of death." We don't know enough to be more targeted in our advice.

Arguing anything else but that question –what we tell the 30 year old who's overweight — may indeed often be very interesting and cutting edge, much more interesting than boring old medical practice. I don't begrudge such discussions. But they should be kept clearly separate from the boring old practice of medicine.


Glen Tomkins 06.21.13 at 5:19 pm


If you see the actuarial sweet spot as the result of competing forces, the mortality caused by obesity pulling against the marker of higher mortality that low weight constitutes, you will see that even the old IBW is too far in the direction of tolerating high weights.

If we could identify all the disease processes that cause low weight, we could adjust the mortality statistics to arrive at a true ideal body weight for avoiding a cardiac death. But we can’t, and the result is that the IBW is further right on the graph of mortality against BMI than it would be if we did have the ability to sort out all the wasting illnesses. Even the old IBW was too lax.

We certainly shouldn’t move the recommended weights even further right in response to the statistical sweet spot moving right. The mortality trough moved right because it works off the population mean weight, which also moved right (on the graph of BMI against time) during this period. There is no reason to think that this shift means that obesity has become less dangerous than fifty years ago. Obesity, which pulls the sweet spot left with the same force, simply gets to equilibrium further to the right in 2013 because it starts from higher mean BMIs in 2013.


soru 06.21.13 at 6:26 pm

Consider a spherical human, with a propensity to develop the sole cause of death straightforwardly proportional to radius. If that process of mortality is slow, and involves shrinkage, then the measured relationship between radius and mortality will be more or less arbitrary, depending on the exact parameters chosen.

Without in any way changing the causual relationship, and so appropriate control strategy.

Surely you can compensate for this straightforardly by measuring the weight say 5 years prior to death?


My iron lung 06.21.13 at 8:30 pm

A 6 foot 150 lbs adult has a BMI of 20.3. Which is supposedly “normal.” A 6 foot, 150 lbs guy would look like a skeleton.

BMI is bunk.


Walt 06.21.13 at 8:37 pm

Glen, there is essentially no evidence for your position on John’s suggestion. Sure, there could be lots of undiagnosed illnesses that reduce weight, so many that they are producing a large skew in the stats, but now that we’re in the realm of speculation there could be undiagnosed illnesses that increase weight. You need to ask yourself why you are so committed to a view that has so little evidence in favor of it.


LizardBreath 06.21.13 at 9:11 pm

A 6 foot, 150 lbs guy would look like a skeleton.

I’m married to one, and he really doesn’t look that bad. (6’2″, ranging between 150 and 165, depending on stress level. And when I met him, he was 135, and still didn’t look unhealthy.) Most people would look starved at that height/weight, but frame size is incredibly variable. I would bet that if he ever managed to get up near ‘overweight’ by the standard BMI range (over 195lbs, at 6’2″), he’d be seriously at risk of metabolic syndrome, because that’s way over where his body naturally settles.

Any individual’s specifics are peripheral to the conversation, of course, but I think it’s worth remembering that wherever you set the BMI range, it may correspond to the height/weight ranges where the population as a whole tends to have the lowest risk of mortality, but it’s going to be way off for lots of individuals.


pjm 06.21.13 at 9:35 pm

@58. Glen I think your point about disaggregation is valid but if I can nitpick I don’t think is quite so simple. Why?
1) First of all, you probably have the causation between obesity and metabolic syndrome exactly reversed (though even then it is not simple because central obesity – aka intra-organal fat – is metabolically active and reinforces M.S. though probably not the sole causative agent). And people who gain fatty tissue outside the abdomen are not nearly as likely to have m.s. as people who do.

Obesity in general is not the cause but probably a symptom (though perhaps with some feedback). Fructose’s effects on the liver and endocrine system is the most likely cause of met.syn..Elevated insulin is a constituent part of metabolic syndrome and it’s also been known for a hundred years that insulin is the master hormone in the accumulation of body fat. So to say obesity is the cause is questionable or a simplification.

Also there is ample clinical experience (with about every physician who treats patients with low carb diets) that the indicators of met-syn go away with long before patients become lean (though they do eventually loose the weight too). Blood pressure, blood sugar, triglycerides, LDL/HDL ratios all improve rapidly. You could argue weight loss (though not weight level) is what makes these patients better but they get obviously better while still technically obese.

2). Metabolic syndrome is not just associated with heart disease with but with a number of other diseases as well, for some of which it may also be a causative agent. M.s. is a cause of diabetes II, another major cause of mortality of morbidity, and it well could be for others, including cancer (via chronic increased inflammation), mult. scler (via same)., alzheimer (same), and so on. This should not be surprising, chronic elevated insulin almost certainly wreaks havoc in a multi-system, multi-organ sort of way. I.e., the more health problems associated with met-syn the flatter the parabola has to be.

3) So what the medical advice should be is that no matter what your weight, you should be trying to combat the met-sym (i.e.reduce or eliminate carbs, especially fructose) though in most cases there will be a reduction of adipose tissue and weight loss as well as a result of this protocol.

The likely reason any of the above is not more widely accepted is because it contradicts the medical establishment’s dominant paradigm (and public health recommendations dating from the McGovern committee) against saturated fats in the diet (which frankly never had very strong evidence backing it – mostly somewhat ambiguous epidemiology from the 60’s) Contradictory evidence has been piling up in the opposite direction for some time now, in fact, the centrality of metabolic syndrome in thinking about chronic disease is part of the tide turning in medical opinion away from the old viewpoint.


Katherine 06.21.13 at 10:22 pm

Obesity causes the metabolic syndrome.

Citation needed.


Glen Tomkins 06.21.13 at 11:38 pm


The anti-fructose theorizing is very interesting, but what do you tell the 30 year old who’s overweight? Go on a milkshake diet, as long as the chocolate syrup you use doesn’t have any fructose?


Glen Tomkins 06.22.13 at 12:22 am


As far as I’m concerned, “metabolic syndrome” is simply the observed tendency of hypertension, DM-II, hypercholesterolemia and obesity to flock together. Obesity seems to be reasonably characterized as causing the other three, because you can make them better, often even “cure” them, by reversing the obesity. Some take the concept further and I’m aware that there is a body of work out there seeking to understand the mechanics of the observed association, but I don’t keep up with that. If they find something clinically useful, I’ll find out about it soon enough.

But until then, all I know is that I’ve only had one hypertensive patient the past twenty years who was at IBW, and that circumstance was so unusual that I worked him up for secondary causes of hypertension and found the only pheochromocytoma of my career. And speaking of wasting illnesses, he put on quite a bit of weight after his pheo was removed, got closer to American standard weight. Even the few hypertensives who aren’t that way because of obesity seem to contribute to this “paradox”.

Whatever the theoretical underpinning of the associations among these four conditions, that association is undeniable in clinical practice. I’ll stop calling it the “metabolic syndrome” if you convince me that there are sound theoretical reasons to avoid the term, but convincing me that thin people get hypertension and DM II would involve convincing me that I’ve hallucinated the past 30 years.


Glen Tomkins 06.22.13 at 1:25 am


Just look at the graphs. This thing is easier to understand graphically. Prasad cites this article in Nature at his comment #23 that is a big part of this nonsense about some “paradox”. The curves in the first graph in that article show a trough in the graph of mortality against BMI.

It’s not me claiming that lower weight than the trough weight is associated with higher mortality, it’s the very existence of that left-handed limb of the curves, where they arc up into higher mortality as the weight decreases. That is this supposed “paradox”, the supposedly heretical idea that mortality, at least on that part of the curve, increases as you lose weight. These people actually seem to believe that the idea that overweight causes mortality is refuted by the fact that the opposite association holds true in that left-hand limb of the curves.

As for illness causing low weight, any illness at all tends to do that as it becomes more severe. By end stage, they all cause weight loss by one mechanism or the other.

And while end-stage of diagnosable illnesses generally are recognized as such, and could have been adjusted for in some of this data, there is still a whole lot of room for wasting illnesses that we haven’t diagnosed in individual patients, and even more room for wasting illnesses we haven’t even categorized and classified at all, to exert huge population effects. The reason there is such a rush to attribute all manner of non-specific ailments to the latest “it” diagnosis (Lyme disease, chronic fatigue syndrome, lupus — I’m old enough to remember when it was hypoglycemia!), is because there is a vast amount of undiagnosable illness out there. These people aren’t crazy, they aren’t imagining things, they just have stuff that we can’t categorize, that isn’t any recognized disease. As a primary care physician, I spend most of my time dealing with this stuff. This stuff exceeds the actual diagnosable and treatable illness I deal with, by a lot. I assume that people die of this stuff, because they very often don’t die of any disease we have characterized, and they get “cardiorespiratory arrest” written in as cause of death on their death certificates. We might characterize some of these cases on autopsy, but no one gets autopsied anymore except on crime shows.

But if you really want to defy common sense, go right ahead. Attribute the higher mortality with decreasing weight in the left hand limb of the curves to some idea that these people’s mortality is caused by weight loss, and not the other way around. They were dumb enough to listen to doctors hectoring them to lose weight, and then they got some illness and didn’t have enough fat stores to weather the crisis! That’s the ticket, that’s sound common sense, and I’m just some crank for believing that bad health causes low weight.

What is especially ironic about this talk of the “paradox” of higher weight associated with better health, is that if you go back 50 years, we had the same paradox in reverse. The Nature article speaks of the insurance industry as the source of the current supposed conventional wisdom that obesity causes bad health outcomes. But what it doesn’t go into, is the fact that that idea from the 1960s was the Sunday Supplement (okay, for the benefit of any children out there who don’t remember Sunday Supplements, they were what Nature has become today, uplifting scientoid infotainment) “paradox” challenging what everyone at the time had known for ages, that fat was healthy and thin people were that way because of consumption or stomach cancer or poverty or some other condition that boded ill for their survivability.


Katherine 06.22.13 at 10:15 am

As far as I’m concerned, “metabolic syndrome” is simply the observed tendency of hypertension, DM-II, hypercholesterolemia and obesity to flock together. Obesity seems to be reasonably characterized as causing the other three, because you can make them better, often even “cure” them, by reversing the obesity

So, no citation then.

Also, you’re aware I assume of the 98% medium to long term failure rate in “reversing” the obesity. And worse, that the act of “reversing” the obesity is known to potentially increase the obesity. Doesn’t sound terribly much like reversing, and more like the cure being worse than the disease.


Glen Tomkins 06.22.13 at 2:39 pm


There is no doubt that we have no effective medical intervention against obesity. But that has absolutely no bearing on the question of whether or not obesity is a medical problem. There are many diseases for which we have no cures. That doesn’t mean we should pretend that they aren’t diseases.

But while there are no medical interventions for obesity, people can lose weight by changing their diet. For diabetes, a 50 lb. weight loss often means they can come off insulin and their other diabetic meds. No medical intervention can do that, remove the need for insulin, essentially cure type-2 diabetes, but the patient can do that by losing weight. The fact that only 2% succeed hardly means that we should stop urging everyone to do this, because we can’t tell beforehand who will pull it off, and the payoff is a cure that is beyond any medical intervention, well, any medical intervention aside from our pointing out the health benefits of losing weight so that the patient will do the rest.

If only 2% succeed, we need to review what kind of advice we give people about dieting, because the success rate, historical and even contemporary, at arriving at a diet that doesn’t make you obese is a lot better than 2%. Sure, some of the historical “success” of traditional diets at keeping the weight off had to have been that these were often diets of scarcity. No doubt a large part of the old conventional wisdom that thin means bad health arose from the effects of a food scarcity that we do not want to replicate. But it is also possible that industrial societies such as ours have gone a bit too far in specialization and division of labor in respect to food preparation. We tell people who are obese to go on some sort of reduction diet, to throw “bad” things out of their diet (what’s “bad” can vary from pet theory to pet theory, but there’s always some food cast in the role of villain), when the problem is that their diet is already too reduced, is so impoverished by the decline of cooking that it has too few choices. They’re already on a reduction diet, and that’s why they’re obese, there isn’t a full range in their diet for their appetites to choose from. Narrowing the range further is likely to be the worst thing to do.

Yep, no citation.

If you ever get bacterial meningitis, are you going to ask for a literature citation from the ER doc before you let him give you antibiotics? There is no such literature, beyond a few very small case series, supporting the efficacy or safety of that intervention. Medical journals mostly refuse to even consider publishing a case series these days, though the literature used to be full of them, because they really are incredibly weak. They amount to no more than a write up of received medical wisdom. You’re going to let doctors give you a potentially deadly cocktail of antibiotics, based on what, the bald assertion of received medical wisdom?

It gets worse. Why did you let them stick that needle in your back? Where’s the literature citation, where’s the evidence for the existence of such a thing as “bacterial meningitis”, for the idea that it’s likely to be bacteria growing in your spinal fluid that’s giving you high fever and the worst headache of your life? There is no quantitative evidence for our nosology, for these concepts we impose on reality dividing the universe of disease off from the rest of the universe of bad things, and within the universe of disease, into distinct diseases. That’s all based on gestalt and pattern-recognition, and the resulting consensus among a bunch of old white guys (mostly long since dead) about where the boundary lines lie. Of course these divisions are often wrong. How could they fail to be? But they are redrawn over time by the same process of pattern-recognition attaining consensus, that is just as uninformed by any of the quantitative methods that constitute the medical literature as it exists today. No one can argue that medicine shouldn’t be evidence based, but medicine relies, it has to rely, on more sorts of evidence than are dreamt of in the philosophy of the “Evidence-Based Medicine crowd.

Which is all a long way around the barn to saying that I don’t need no stinking literature to know that obesity causes DM II and HTN. A non-obese person with either of those two is such a rarity that it makes me consider that the diabetic is really type-1, and that it must be secondary HTN. People who lose 20 lbs predictably, almost universally, need to have their diabetic meds reduced, and if they gain 20 lbs, need to have their meds increased. Ask anybody who treats diabetics and hypertensives. That’s stronger evidence than you’re going to find anywhere in the literature.

I will freely admit that it was a mistake on my part to use “metabolic syndrome” as a shorthand way of referring to that observed connection of obesity and HTN, DM II and HLP. That phrase implies more than that bare observational linkage, and there is indeed a literature that reports on attempts to find a pathophysiology behind the connection, and “metabolic syndrome” is connected to one pet theory in that arena. I am not familiar with that literature, and don’t feel any need to become familiar with it, because what’s important for my practice is the bare observed linkage, and theories about why the linkage exists are just pet theories at this point. The medical literature was once nothing but bloviation about pet theories, and case series were actually an improvement on that. That stuff is as relevant to my practice as Tiger Rag, which I don’t read either. Lab bench researchers need to jaw at each other over such things, but it’s not for me.


pjm 06.22.13 at 2:49 pm

@69, Katherine it is a good point and it part and parcel of the insulin disregulation camp’s arguments against the reigning wisdom. I.e., most medical dietary advice (except, it should be noted, from 50% of the bariatric medical community which has gone increasingly low carb despite official hostility to the contrary) is all about calorie restriction (though many popular diet advocating physcians are “kitchen sinkers” – effectively low carb and low glycemic index – bascially a c.y.a. position). This has been compounded by the food industry’s replacing on a widespread basis sugar for fat in foods and marketing them as “healthy”, resulting in a substitution of foods that naturally regulate appetite for those that do the opposite.

One of the reasons dieting is unsuccessful and often results in the yo-yo is that in a diet that achieves lower calories by restricting fat and letting the proportion of carbs increase is that the result is often the loss of lean muscle mass. High fat, low calorie diets, on the other hand, result in the minimal loss of protein.

@66 Glen,you tell the 30 year old to restrict carbs not just sugar. On average they will lose weight and have better health indicators. As you probably know, anything with table sugar (sucrose) in it is a significant source of fructose. If your implication is why can’t you just eat sugarless ice cream
all day long and not be unhealthy, the answer is you won’t. Typically, fat in the diet satisfies appetite and carbohydrates (fructose or not) tend to do the opposite (and in fact are argued to be addictive – I think that is a plausible claim though I don’t know much about the research). Also not enough protein in sugarless ice cream, so I wouldn’t recommend it as a staple.

The insulin disregulation school is all about the idea that in the default case, appetite and weight are pretty tightly regulated by the body (which is an idea that explains some of the findings in diet research which caloric imbalance paradigm doesn’t handle so well) and the main source of this disruption is the effect of dietary carbohydrates on insulin levels. (And in case you are wondering, fat in the diet has exactly zero effect on insulin levels, less than even protein).

And if you are wondering why insulin is so central in this science, afaik, 1) it is literally impossible for fat cells to take up fatty acids without it 2) very hard for fat cells to release fatty acids to be metabolised in the presence of high insulin levels. In the animal models, it has been repeatedly demonstrated that in animals not taking in enough calories to maintain their lean body mass (i.e. negative caloric balance) will still accumulate fat if their insulin levels are high (generally the experimental setup is that they mostly eating carbs). I.e. in such models the body
is literally cannibilizing itself to pursue fat storage (probably relevant to the issue of yo-yo dieting).

I think the anthropological evidence is strongly suggestive that humans are just not adapted to getting so much of their calories from carbohydrates (obviously with some genetic variation) but even among individuals who do okay with carbs in their diet, fructose’s effective role in metabolic syndrome is to reduce the ability of the body to process carbs without insulin resistance, i.e. chronically elevated insulin.


Walt 06.22.13 at 3:00 pm

Glen, you can’t just look at the graphs, because the graphs are consistent with many stories. That’s why we use statistics (and even statistics can only get us so far). Parsimony and common sense are weak criteria, since the temptation to regard your preconceived notions as both parsimonious and common-sensical is hard to resist.

For example, if we’re going to posit the existence of large numbers of undiagnosed wasting diseases, then it’s perfectly possible that fat people take longer to waste away from them than do thin people.

If you’re routinely seeing patients that cure themselves of obesity, then you are seeing a very atypical population of patients.


Harold 06.22.13 at 3:10 pm

Restoring nerve function may be key in fighting obesity (Science Daily)
Restricting the size of the stomach has some role in the effectiveness of gastric bypass, but it’s not the full story,” said Kirsteen Browning, assistant professor of neural and behavioral sciences. “It is not fully understood why the surgery works.” The researchers published their findings in the Journal of Physiology.
Complications from diseases such as diabetes can resolve before weight is lost, and sometimes before the person even leaves the hospital after gastric bypass surgery.
“This suggests an altering of the neural signals from the gut to the brain and back,” Browning said.
These nerve cells send signals to tell the body’s digestive system how to respond properly and regulate normal functions of digestion. In obese people, the nerve cells are less excitable, meaning they respond less tonormal stimulation. For example, there are neurons that help tell a person that their stomach is full, called satiation.
“These signals tell you to stop eating,” said study co-author Andrew Hajnal, professor of neural and behavioral sciences. “Obviously these signals are strong enough to be overcome by all of us and we can eat more even after we are told we are full. However, as obesity develops, it appears these signals are less strong and easier to overcome.”
Penn State Hershey researchers used a high-fat diet in rats to replicate long-term exposure to a Western diet. They then observed the effects of gastric bypass on the rats and have shown for the first time that the effects of diet on nerve cells seem to be restored to normal function after the surgery. This would help in restoring satiation signals so that they can be recognized more easily.
“We know gastric bypass improves the health of nerve cells and reverses the effects on the signals,” Browning said. “Even if the nerve cells have been affected over a long term, gastric bypass still improves toward normal function.”
The goal of this line of research is to find new treatments that will not require gastric bypass surgery, which is still considered an invasive procedure.
“Once we understand what gastric bypass is doing, we hope we can mimic that with other treatments,” Browning said. “Restricting the size of the stomach may not be the major player of the surgery; restoring normal neural function is also an important aspect.”


Walt 06.22.13 at 3:57 pm

Let me guess, Glen. You “know” that peptics ulcers are caused by stress, just like old white guys told you once upon a time.


Glen Tomkins 06.22.13 at 5:04 pm


I don’t treat dieting as any sort of medical intervention, therefore I don’t give advice on what to avoid and what to put in. I don’t think the science is anywhere close to mature on that subject. If it were, we would be beyond Walt’s skepticism over his mistaken impression that my patients routinely lose weight. You’re welcome to your pet theory on the subject, but it isn’t ready for prime time, and it won’t be until it can be at least notionally “bottled”, reproducibly make people lose weight. I assume that hasn’t happened yet, because if it had, my patients and the rest of the world would already be all over the cure, and within months they would all weigh IBW and I could dismiss 90% of them (the other 10% have occult wasting diseases) and finally actually retire.

What I do tell my patients is that their best bet is to learn how to cook. Reach a point where you don’t want processed dreck, reach the point where you can put a dozen choices made from scratch on the table for every meal, and your appetite will take care of the rest. I base that theory on the crude idea that that’s how people used to eat (when scarcity didn’t intervene) and people didn’t used to be so often so obese, so it’s got validity just from that. Weak as that evidence is, it’s the best evidence I see out there. Lab bench research about fructose is irreplaceable at generating hypotheses, but it isn’t even in the same ballpark as evidence about telling people what to eat.

I work in a free clinic, and most of my patients are non-citizens, currently unemployed, or both. They’re ahead of the game, because they tend to be nearer to some or other traditional diet, they have some memory of how their grandparents ate, and therefore some memory of what right tastes like. They also tend to have the free time to learn how to replicate the old way of eating. I tell them that’s their job now that they’ve been laid off, learning how to cook. No one’s going to give them a paycheck for that work, all they get out of it is coming off insulin and avoiding dialysis.


Glen Tomkins 06.22.13 at 6:16 pm


No, I don’t routinely see patients losing any amount of weight (permanently), and rarely see a stable 50 lb loss. But, however small the n, the results of a 50lb loss on diabetic control are dramatic and consistent. We treat bacterial meningitis with antibiotics based on an initial cases series with an n of 10. We tell people that smoking causes lung cancer based on case control studies. And we’re right to do these things, and tell diabetics to lose weight, because the results are stark enough that even weak study design and a small n leave no room for doubt.

Okay, if graphs don’t work for you, I’m sure you can still translate this simple take on the data into whatever form impresses you. But thinking of this graphically is not only natural and intuitive for most of us, you can characterize the positions of people in this fracas very naturally in terms of the curves.

If the graph of mortality against BMI were flat, a straight horizontal line, then the reasonable interpretation would be that weight has no effect on mortality. Sure, you could hypothesize that, yes, weight does effect mortality, but that for every weight, competing disease process were perfectly balanced so that they exactly canceled out each other. That seems a stretch.

If the graph were a straight, non-horizontal, line rising with some positive slope away from the origin, then we would say that this shows that increasing weight is at least correlated with increasing mortality, and you would probably infer that the relationship is causal, from high weight to high mortality. The folks pushing the idea that there is some sort of “paradox” introduced by the fact that some part of the curve shows the opposite, mortality rising as weight decreases, seem to think that the relationship has to be linear with a positive slope, or we must abandon the theory that increased weight causes increased mortality. They characterize the conventional wisdom as just that, the belief that there is this linear relationship whereby rising weight always means rising mortality.

Go back a century, and we have a conventional wisdom that was pretty much the opposite of this caricature of the current conventional wisdom. Decreasing weight was felt to correlate with increasing mortality. Fat was good and healthy, thin was a marker of disease. You have a line with a negative slope.

But none of these is the actual relationship of mortality and BMI. The actual relationship isn’t a horizontal line, or a line of positive or negative slope. It’s a curve, not a true parabola, but paraboloid. That shape of the curve really is central to this whole discussion. The whole idea of an ideal body weight is that the minimum of that paraboloid is the sweet spot between the old idea that low weight is bad, and the new idea that high weight is bad. The IBWs have built into them the concept that of course there is no linear relationship between BMI and mortality, of course there’s an inflection point beyond which decreasing weight becomes bad for your health. There’s no paradox.

To complete the characterization of different positions in graphic terms. What’s new lately is that the minimum, the sweet spot where BMI produces the lowest mortality, occurs further right, at a higher BMI, than it did back when the IBWs were set. Some say that therefore we should move the IBWs right, to a higher BMI, to accommodate the newer evidence. I disagree, on the grounds that the paraboloid, 50 years ago and today, is the result of the summation of two roughly linear relations, that of some diseases causing low weight, and high weight causing some diseases. Sum the positive slope line with the negative slop line, and you get the actually observed paraboloid.

JQs take seems to be that, new curve or old curve, the BMIs we’re talking about are in that nearly horizontal part of the curve near the minimum. Go a couple of point of BMI either way in that region, and the slope is so small that the resulting mortality change is not that big, not enough to worry about. What I have to say against that idea, is that it’s only that positive slope line we’re worried about when we give people advice about their weight. We don’t know where that line is, precisely because the horizontal part of the curve is exactly where the two linear relationships most interfere with each other, most muddy each other’s water. The relationship of concern, BMI causing increasing mortality, could have a much steeper slope in this horizontal area.

Which brings us to what seems to cause you distress over my supposedly cavalier appeal to occult wasting diseases. If you could adjust the data for all diseases which cause weight loss, you could get to the real curve of the relationship between BMI and mortality, so people who are eager to overthrow the imagined conventional wisdom are eager to believe that you can indeed so adjust the data. But that flies in the face of clinical experience.

You probably couldn’t get a study of a cardiac intervention published in the literature if you used cardiac death (death from some cardiac cause) as an endpoint, but failed to also look at “all-cause mortality”. It’s the standard to look at all cause mortality because it is generally recognized, not just a wild theory of Wild Man of North Borneo Tomkins, that even in patients being followed in a study by the Mayo Clinic itself, we don’t do a very good job of sorting out cause of death. Dead from not-dead, that’s easy, we can make that distinction readily. But dead because of the disease process of concern from dead because all other, often occult, causes — there we are terrible, even the Mayo Clinic. Even the Mayo Clinic has to report all cause mortality if it’s going to report cardiac mortality because we take for granted that even they will be unable to attribute cause of death reliably, and we don’t want to accept that a new intervention reduces cardiac deaths, when it might actually cause such, but in ways that we cannot reliably attribute to what the intervention did.

Leave the Mayo Clinic, leave studies where study physicians or nurses pore over the medical records to sort out cause of death, and our ability to get the attribution right plummets to truly abysmal levels. Any filter or second-guessing you try to apply to death certificate data is jus going to introduce massive biases, so you’re often better off just using mortality and not trying to adjust data that is basically unadjustable.

So, yes, there is a huge place for all manner of undiagnosed and undiagnosable illness to cause weight loss for imponderable periods prior to death.

Feel free to go ahead and reverse the traditional assignment of the direction of the causal arrow, and say that it’s the low weight that causes death. But, as you yourself tell us, it’s not like there are hordes of people out there losing weight — quite the contrary. The left limb of the curve, where lower BMI means higher mortality, has migrated right, this higher mortality now occurs where the horizontal part of the curve used to be. Your theory can’t account for weights that used to be healthy, now putting people at risk of wasting away in the face of a disease crisis that more obese people would weather.


Michael Sullivan 06.22.13 at 6:28 pm

“But most technically obese people aren’t prop forwards.”

That’s true, but the use of BMI as an individual indicator is much worse than that. You don’t have to be a professional athlete for there to be a huge difference between what what BMI would suggest about your level of fitness and health vs. the reality. Those are just the most extreme examples.

Pretty much anyone who has done a healthy amount of strength training, or who regularly does physical work, will show up as overweight or even obese even if they are in great condition. Particularly for people who are taller than average, as you note.

When I was strength training hard a few years ago, I bulked up to where my bodyfat % was around 20%, while not losing any weight from being pretty fat. I was hardly a pro athlete, just a guy who hit the gym hard 2-3 times a week, and otherwise worked a relatively sedentary job.

By body composition I was on the chubby side of normal and that’s about how I looked. According to BMI, I was “morbidly obese” at 6’3″ 27o lbs.


pjm 06.22.13 at 6:30 pm

Glen, you are doing important work and you are right that the theory is not mature but to some
extent that is self-inflicted within the research community. My two cents, is that I think it is a significant trend the bariatric physicians have gotten to a point where half of those that use dietary intervention do carb restriction having progressed from the point where Atkins was widely believe to be a dangerous quack.

Indeed, what has shaped as something of a movement has included lots of physicians who have tried carb restriction on themselves, found it easy and effective, had improved blood lipid profiles, are usually surprised by these results and then either decide to look into the science and/or
treat their patients with it (in part out of frustration with other dietary interventions).

In 10 years my guess is it will be the carb restricton will be the best practice. From what I read, it is much more common for endocrinologists to recommend carb restriction as part of a diabetes protocol that it was a decade ago. As 2012, the Swedish National Board of Health is recommending a “modest” low carb diet for diabetics. Five years I was told the only way to the only known lifestyle intervention to mitigate NALD was rigorous exercise. But now, it turns out carb restriction helps with that as well.

I do think insulin disregulation (or something similar) is the right theory both about the cause of obesity and how it impacts health. But even if that’s true, it doesn’t mean your patients will be all over the cure. The entire food industry is based on the availability of cheap carbohydrates (mostly from grains). Besides being highly available and cheap, if there is addictive factor going on it might not be so easy. On the other hand, one of the reasons the bariatric community has moved towards carb restriction is because compliance is reported by patients to be much less of an issue. In any case, the best to you and your patients and hopefully we will all know more about the science as soon as we can.


Michael Sullivan 06.22.13 at 6:32 pm

“We tell people that smoking causes lung cancer based on case control studies. And we’re right to do these things, and tell diabetics to lose weight, because the results are stark enough that even weak study design and a small n leave no room for doubt.”

I have a problem with that Glen.

At least with the recommendation to lose weight. And that is that dieting, for the large number of people that will not permanently lose weight, often has the perverse long term effect of *increasing* weight and fat by playing havoc with our internal satiety signals and often inducing malnutrition.

A healthy eating and exercise program, that may occasional result in weight loss, but will almost always result in better health is appropriate. But going specifically for weight loss has a tendency to lead to eating behaviors that are worse, rather than better.


Harold 06.22.13 at 9:25 pm

It appears that the nerve system that regulates appetite is inactivated by overeating and not restored after dieting, making it virtually impossible to keep off weight. Also, dieting is associated with osteoporosis.

They need to find a way to restore the feedback mechanism.


Hector_St_Clare 06.23.13 at 12:29 am

Some people are also thinner than they would be otherwise because of smoking, which would also increase mortality for thin people.

That being said, it’s probably not a good idea to simply raise the ‘healthy’ range without more detailed recommendations. Different ethnic groups have different ‘healthy’ weight ranges. South Asians, for example, have elevated diabetes risk at a BMI of 23, iirc (similar things are true for Native Americans and Latinos).


ChrisB 06.24.13 at 12:28 am

The real paradox is that prosperous western populations (I’ve graphed Australia, here – show a rigid and unvarying increase in life expectancy of about two months a year since about 1890. If obesity has a strong connection with life expectancy, then there’s nowhere where any variation in obesity has impacted on the strongest measure of health; perhaps the two graphs are exactly parallel.
If we hadn’t been getting more obese, is the theory that we would have been rising by three, four, or six months a year?


Kyle C 06.24.13 at 2:19 am

Fantastic comments! Best thing I’ve read, um, all week.


Saul 06.24.13 at 6:49 am

There has been a great call for citations needed on the idea that obesity actually causes risk rather than being an epi-phenomenon. The problem is that obesity is almost entirely resistant to behavioral intervention on a systematic basis. People can lose lots of weight but they often regain it and trying to single out a group who have lost weight by choice and compare them properly is pretty tough – those who do lose weight by changing their behavior are inherently unusual.

So. Despite the distaste for such things evident in this thread, I give you some articles on bariatric surgery and ultra-low calorie diets. (Quick Medline, rather than deep knowledge. I’m a neuroscientist.) Not because those treatments are necessarily a good idea but because they work, at least for a while, on a good percentage of people. That allows the non obesity-related risk factors to be properly controlled for.

Bottom line looks like blood glucose, blood pressure and lipids predictably drop. Studies are not long enough to demonstrate survival. Bariatric surgery should have mortality data pretty soon, by the looks of things.

LAP-BAND® for lower BMI: 2 year results from the multicenter pivotal study. Obesity (Silver Spring). 2013 May 2 Epub ahead of print

Effect of weight loss on sympatho-vagal balance in subjects with grade-3 obesity: restrictive surgery versus hypocaloric diet. Acta Diabetol. 2013 Jan 26. [Epub ahead of print]

Weight loss independently predicts urinary albumin excretion normalization in morbidly obese type 2 diabetic patients undergoing bariatric surgery. Surg Endosc. 2013 Jun;27(6):2046-51

Abstracts for them all in Medline.

They are all pay-walled for the full article.

I would add that the idea that losing weight reverses your diabetes, lowers your blood pressure and reduces the rate of decline of your osteoarthritis is pretty widely accepted in the entire medical world. You are not going to find a direct study challenging that but these three all estimate the magnitude of those changes.

REALLY should go back to my ACTUAL work at this point…


Katherine 06.24.13 at 10:24 am

The fact that only 2% succeed hardly means that we should stop urging everyone to do this, because we can’t tell beforehand who will pull it off

Are you insane? At best, the advice given will help 2% of people and leave everyone else at baseline neutral (except that is for the complete waste of time and effort of dieting, losing weight and then gaining it all back again). From there, the benefits are downwards only, where the negative effects of dieting and regaining go up until you get to the worst case scenario of 2% helped and 98% of people actively harmed.

So, absolute tip-top best case scenario is Mostly Rubbish, with increasing harm all the way down to the worst case scenario of Mostly Harm. Even the best case scenario suggests, does it not, a pretty dramatic failure to understand the mechanisms at work? Honestly, it’s like medieval doctors insisting that bleeding is a great idea, even if they don’t understand how it works, because it appears to help 2% of people.

And presumably, since you say that it’s still good advice to give since you don’t know which people will be in the 2%, once someone has dieted and their death-fat hasn’t been “reversed”, you stop giving that advice, since it’s clearly useless to them? And it’s put in their notes not to offer that advice again, so that other doctors won’t give the same useless advice over and over again? That would seem like a basic for empirical good practice, no?

And yet, fat people will get hectored again and again to lose weight, which flies in the face of whatever evidence there is.


pjm 06.26.13 at 6:08 pm

Comments on this entry are closed.