Should we abandon the carbohydrate hypothesis of obesity?

I have read Good Calories Bad Calories. At just under a kilogram there are minor points within it with which I disagree. But, for the majority of the people who have read it, it is basically correct.

One of the most recent critical appraisals of the carbohydrate hypothesis of obesity was posted by Stephan over at Whole Health Source. Obviously, I disagree with Stephan's appraisal. That's fine, to disagree is perfectly OK. We'd get nowhere if we all sang from the same hymn sheet. This post is basically my take on the evidence used to destroy the carbohydrate hypothesis. It's depressing to have to do this but reassuring at the same time.

So why do I cling to this apparently incorrect and outdated hypothesis? Let's look at the points in approximate order as taken by Stephan.

A defect of fat metabolism?

Taubes ignored leptin to concentrate on insulin; this appears to be the main conclusion in this section. Stephan cites a neat paper by Leibel et al which demonstrated that in four healthy, never-obese humans the fall in metabolic rate induced by 10% weight loss could be reversed by physiological leptin replacement. That's cool if you want to be a young, fit, healthy, never-obese experimental volunteer desperate to live comfortably at 10% below your normal, slim weight.

If you are currently morbidly obese it may be of some comfort to know that leptin might be able to help you correct your hypometabolism should you manage to lose some weight.

If you are ex-morbidly obese and have managed to lose a few hundred pounds of fat you will still own a set of injured adipocytes. These injured adipocytes refuse to produce physiologically appropriate levels of leptin for their fat stores. Now THERE is a role for leptin. It might even reverse the persistent hyperinsulinaemia present even during starvation in the morbidly obese...

If you are currently morbidly obese and think leptin will help you lose weight, think again.

So do I think leptin is unimportant? Of course not. Does this invalidate the carbohydrate hypothesis? Shrug.

What about the morbidly obese ob/ob mouse, which cannot make leptin? There are a handful of human families on the whole of the earth with this problem. They need leptin and it will work for them.

The population of the USA is around 300 million. Of the adults in this population, as of 2008, 34.2% are overweight, 33.8% are obese and 5.7% are morbidly obese. They are not going to benefit from leptin supplementation to lose weight.

According to Stephan many, if not most, of these few million people will benefit, for reasons which are not entirely clear, from carbohydrate restriction. But it's not due to lowered insulin levels... Fascinating conclusion.

How can anyone be so sure that it is not from a reduction in insulin levels?

Here's why, watch very carefully:



You may think you have seen this clip before but no, although the child is the same the chocolate is different. This is 90% cocoa chocolate, none of your boring 74% sugary stuff...

If your baby is going to self feed chocolate you are going have to bath her. Bathing babies is dangerous. We also have one of these:



I always try not to throw out the baby with the bathwater. You can't be too careful...

The big problem with insulin, as any obesity researcher will tell you, is that it is a satiety hormone. I've said it before, all you have to do is have it injected in to your brain and you won't feel like eating a steak for the next few hours. Let's get a nice juicy quote from this paper by Velloso and Schwartz, hot off the press in 2011:

"A major and persisting source of confusion surrounding the hypothesis that insulin action in the brain reduces food intake and body weight while also lowering hepatic glucose production and increasing thermogenesis stems from evidence that following systemic insulin administration, the subsequent fall in glucose levels potently increases food intake while also increasing liver glucose production and reducing sympathetically driven thermogenesis. Thus, insulin-induced hypoglycemia potently overrides virtually all of insulin’s central effects, an observation that for many years has confounded research in this field."

Did you see the baby go? Here it is again:

"Thus, insulin-induced hypoglycemia potently overrides virtually all of insulin’s central effects".

That's it: Baby, bathwater, gone. How can anyone be so careless? Oh, did you miss it?

The baby is the peripheral effect of insulin on lipolysis, which is discarded without mention. Because hypoglycaemia in your brain (a central effect) makes you hungry, the fact that hypoglycaemia can steamroller insulin's central effects appears to have allowed the discard of insulin's peripheral adipocyte effects. Insulin's inhibition of lipolysis, in a normal human being, occurs at concentrations which do not even budge muscle glucose uptake. Infuse it directly in to the arterial supply to the fore arm and the systemic hypoglycaemic effect is lost. All you get at low infusion rates is inhibited lipolysis. This is the baby in the bathwater. Up the rate a bit and potassium uptake is increased. Bugger glucose uptake, this needs far more insulin that inhibition of lipolysis or promotion of potassium translocation. To summarise, if abnormally high insulin levels are needed to deal with unwanted hyperglycaemia then lipolysis will be inhibited until such a time as fat cells become so distended they refuse to listen to this excessive insulin, ie when they have become insulin resistant.

There are certain other spectacularly obvious problems with this accidental baby loss. Once we have all accepted that insulin is a satiety hormone it becomes perfectly obvious that people on low carbohydrate diets, with their chronically reduced insulin levels, should be hungry. After all, I have seen it suggested by Stephan that insulin might assist weight loss. I'm still trying to get my head around that one, while eating low carb and trying to remember what it felt like to be hungry. Trouble is it's all so many years ago... Of course, as Stephan points out, hypoglycaemia is a potent appetite stimulant. Again LC eaters, with their chronically low blood glucose levels, should be ravenous. I'm also trying to get my head round that one too. Reality occasionally gets in the way of great theories. Sigh.

I'm not quite sure where to put in the neuronal insulin receptor knock-out mouse. It's fat, so the conclusion appears to be that brain insulin receptors are important to satiety. I'm sure they are. However, these fat mice are also hyperinsulinaemic and will be lipolytically challenged. I love these particular KO-mice... They do not have me mainlining insulin as a weight loss drug. I love the impaired spermatogenesis and ovarian follicular maturation too. I still would not decry leptin here but these hyperleptinaemic mice don't do reproduction terribly well. Dare I use the I-word when talking about fertility?

Insulin inhibits lipolysis. Don't forget that when we come to talk about the Pima.

Before we move on let's look at the satiating effects of foods. Stephan's refs 4, 5, 6 and 7 suggest no macronutrient matters much and ref 8 shows protein is more satiating than carbohydrate. But reference number 9 is the absolute beauty.

Satiety is proportional to the insulin response to protein. Wow! Must be the anorexic effect of insulin.

But there are problems, wouldn't you guess. I don't have the insulin/glucose data following ingestion of any of the proteins mentioned in the abstract but let's look at the effect of casein, which I do have data for. The principle is identical.



Casein raises blood insulin level from 39pmol to over 100pmol and it's still at 90pmol by the three hour mark when sampling stopped.

Amen, RIP the insulin hypothesis.

Protein=insulin=satiety=anorexia=lipolysis.

But just a minute.... There is no sugar in casein, any more than there is sugar in beef. If we took these same seven volunteers and, without feeding them, injected them with enough exogenous insulin to raise blood level to 100pmol and peg it there while simultaneously locking the canteen door, they would die in hypoglycaemic seizures somewhere around the 10 minute mark. We could even throw in a little amylin (which obese people happen overproduce, odd that) to stop them being hungry as they die.

But elevating insulin to lethal levels using glucose-free casein, beef, whey, eggs etc all produce acute, severe, unremitting, paradoxical normoglycaemia.

I can't blame Stephan for not mentioning glucagon as it doesn't help destroy the carbohydrate hypothesis. Explaining the physiology seems to be my problem.

In healthy people eating neat protein there is a rise in glucagon which slightly under compensates for, as far as blood glucose is concerned, the rise in insulin. There is normally a slight fall in blood glucose.

Does glucagon increase lipolysis? It certainly does in pharmacological doses, as any physiology text will explain. In real life people seem rather unwilling to publish the data. You would have thought that 30 seconds on pubmed would have shelled out the effect of isolated protein on lipolysis but there you go, the insulin hypothesis, while defunct, discourages dabbling...

In this study they fed children consistent meals of mixed formula for a couple of weeks, then they switched them to a split meal protocol with most of the carbohydrate in the morning meal and all of the protein in the afternoon meal, fat being held constant for both meals.

The morning high carbohydrate meal suppressed FFAs as you would expect because insulin inhibits lipolysis. The afternoon meal of reduced carbohydrate, high protein content spiked insulin all right, but also increased FFAs. As dietary fat was held constant those FFAs almost certainly came from lipolysis. The group didn't measure glucagon but normoglycaeimia in the presence of insulin smells of glucagon to me.

Whenever someone does a hatchet job on the carbohydrate hypothesis using the insulinogenic index of beef without mentioning glucagon I am left wondering why they were carrying an axe in the first place. I find this thought very uncomfortable.

As a complete aside, people may enjoy this snippet on glucagon receptor deficient mice. You can eliminate the diabetic phenotype induced by massive streptozotocin overdose so long as glucagon cannot act. Interesting stuff but off topic really. But you cannot have death by lipolysis under hypoinsulinaemia without the lipolytic action of glucagon...

We next have two excellent studies correctly showing that resting energy expenditure is higher in both Pima Indians and schizophrenics in direct proportion to their hyperinsulinaemia. Oddly enough they don't simply melt away to size zero supermodels under the anorexic effect of insulin because their post prandial thermogenesis is depressed to almost exactly the same amount as REE is increased. Neat huh? Did you realise when you read the citation?

There comes a point at which fat cells become sufficiently insulin resistant that they cannot hang on to the their fat content. You can still put fat in there with minimal insulin and minimal insulin sensitivity. Once adipocytes are sufficiently insulin resistant and they are leaking sufficient free fatty acids to match input, obviously weight gain stops. The inappropriate spilling of FFAs causes palmitate deriviatives to be produced which worsen insulin resistance, whole body, and obesity flips in to diabetes.

I am in complete agreement with Stephan here. What I object to is citing a situation where insulin is failing to progress obesity, when it is doing its best to, as evidence it did not cause it in the first place. Insulin is trying and FAILING to make the adipocytes fatter. The more impossible the task, the more insulin is produced.

So we have a muscle cell, for example, which is wondering what the hell is going on as it sits in a sea of glucose and free fatty acids which is physiologically completely inappropriate. As we have been told by Stephan:

"Let me explain what the primary role of insulin is. It is to coordinate the metabolic shift between burning primarily fat, to burning primarily carbohydrate. Any time insulin suppresses fat oxidation, it increases carbohydrate oxidation by an equivalent amount. That is what it is designed to do."

In morbidly obese people, as they flip in to diabetes, this is EXACTLY what insulin is NOT doing. If you make fat cells more insulin sensitive (or generate some new, insulin-sensitive adipocytes), say with with PPAR alpha agonists, you will correct the elevated FFAs as insulin starts working on fat cells again and diabetes will abate slightly until the ability to store fat under the influence of chronic hyperinsulinaemia is once again lost, but at a higher fat mass.

I dunno, maybe PPAR gamma agonists simply increase food reward??????????????

Why is resting energy expenditure high in the obese? The body hates hyperglycaemia and wants to burn glucose whenever it's high. FFAs are a supply led system. Failure of adipocytes to respond to insulin increases supply. You then have excess glucose from the diet and excess FFAs from leaky adipocytes. You have to do something with the calories.

So does this destroy the "insulin locks fat away and decreases the metabolic rate of hyperinsulinaemic people" hypothesis? This has particular relevance to the multiple observations of utterly impoverished communities were adult obesity co exists with infant malnutrition. I would stress that this does not reflect the situation in the Pima community as studied in the 1990s, where childhood obesity has certainly been an issue and calorie malnutrition is not. Everyone has enough junk food to eat. Everyone can be obese, everyone can have enough calories to run a high REE, keep total caloric output down by decreased post prandial thermogenesis and still manage to gain a few grams of adipose tissue a day until diabetes sets in.

If there is enough obesogenic food for all, a mother will be hypermetabolic at rest and her kids will be fat. The insulin hypothesis predicts that if there is a restricted supply of hyperinsulinaemic food the mother will remain fat due to her hyperinsulinaemia, while the child will remain emaciated while ever she maintains some degree of insulin sensitivity.

Let's do reductio ad absurdum: Mother and daughter have 8000kcal of hyperinsulinaemia generating food available. Mother eats 4500kcal, becomes as fat as her adipocytes will allow her to, then she leaks FFAs from her adipocytes to become diabetic. Daughter eats 3500kcal and does the same. Both are hypermetabolic at rest, the mother more so as she is not growing. Both become obese.

Now lets say mother and daughter have 2000kcal between them. Mother eats 1100kcal, moves as little as she can, drops her metabolic rate, is hungry all the time but stays fat. Daughter eats 900kcal and is malnourished, becomes emaciated.

This is an aspect of the insulin hypothesis which has not been tested for obvious reasons. It will be correct, in my opinion. I am unaware of any evidence base for this.

As I understand the reward hypothesis, the mother and daughter eat a high reward diet, hit their dopamine system, desensitise it by over rewarding and this ups the hypothalamic fat set-point. Mother increases her calorie intake to maintain her set point level of fatness and eats her starving daughter's food to stay there. Fascinating.

The dietary practices of the Pima under severe calorie restriction are a complete unknown to me but I have serious problems with the reductio ad absurdum example I've just discussed. I've never met a mother who appears to behave that way, but maybe I've never met anyone with adipose depots far enough below their bodyfat set-point to behave this way...... Even folks on WeightWatchers seem mostly human.

So looking at modern Pima Indians or schizophrenics fed to satiety in no way tests the insulin hypothesis of restricted metabolism under conditions where insulin remains elevated and people are hungry. In fact Stephan's neat leptin reference suggests if we went in and injected the hungry, obese mother with leptin twice a day we would reverse her hypothyroid state, fire up a few uncoupling proteins and stop her being hungry. We might even drop her chronically elevated insulin levels. She would lose a ton of weight, give all of her food to her daughter and die of a starvation related illness herself. Injectable altruism...

Looking at multiple studies where adipocyte insulin resistance has occurred under ad libitum conditions certainly demonstrates how metabolism breaks under free access to insulogenic calories. I can't see how it refutes the role of insulin in obesity. It utterly destroys a straw man, but you have to actually do some thinking to understand what is going on.

It's all genetic:

Twenty monogenetic obesity syndromes! All in leptin signalling! Unfortunately a) That hasn't given us 20 solutions to help the 200 million overweight and obese people in the USA. and b) In the last 30 years fat people must have instigated a covert "fatties only" breeding program. Think of orgies where skinny people get castrated by fatties as part of BDSM games. We all know it's happening and there is a government cover up. From about the 1970s onwards.

Ultimately life is genetic and if there wasn't variation in response to insult there would be limited ability to select for surviving that insult.

If you want to REALLY look at what a blind alley the genetics of obesity are leading you up just try ref 35 from Stephan. Same genes in Mexican Pima and USA Pima. Only the USA Pima are fed on "D12451" and look like ob/ob mice. Mexican Pima eat Mexican food and blend in to the population. I looks like my BDSM hypothesis on generation of the obesity epidemic might be incorrect. Ah well, back to the drawing board.

Let's look at the natives:

Starch based diets are not associated with obesity. They do not cause hyperinsulinaemia, post prandial or fasting. They do not cause insulin resistance. You can, with significant effort, become obese on starch but only if you force yourself to do so.

I agree with this, in unacculturated people.

I think it might even have applied to people in the USA of 1900.

I disagree with this if applied to the current industrialised world, especially anyone who has become obese. Why should this be?

Obesity was present at a low level in the USA of the 1900s. Rumour, without hard data that I can locate, suggests it affected less than 1% of the population. It increased slowly until the 1970s by which time it was present in around 15% of the population. From 1970 to 2000 it doubled to 30%.

There's a graph on wiki here.

We know form Stephan's neat graph that this gradual rise in obesity between 1900 and 1970 was associated with a fall in carbohydrate consumption and that the rise in obesity after 1970 was associated with a rise in simple sugar consumption.

If you were to include a separate line to show sucrose (plus, as it became available, HFCS) it would rather neatly parallel the obesity curve. Obviously no one wishing to discredit Gary Taubes would do this but, if you are interested in hyperinsulinaemia as a cause rather than a consequence of obesity, I would suggest that you might be rather interested in this line. Once you are insulin resistant carbohydrates become spontaneously fattening. No ritual needed, it happens very much against your will.

The body uses fructose to replenish liver glycogen. There, I said it. The occasional bit of fruit will not make you obese. You only convert fructose in to a fatty liver through denovo lipogenesis when intake is in excess of what humans are remotely able to make use of. Elite athletes consume rather a lot of fructose. It helps them win races. Try breaking your leg by falling off your pushbike and then still keep up the cola consumption needed to keep you in the yellow jersey... You may just develop a fatty liver. OK, you will.

With a sucrose content in the diet averaging 64 lb/year very few people would start on that journey to hepatic denovo lipogenesis in the USA of 1900. At 120 lb/y over a third of the population will go that route. Once you have accepted that dietary fat causes obesity you are then going to eat the replacement carbohydrate which will dial up your fasting insulin and hunger. Official fat phobia kicked in during the 1970s...

Oops, I forgot that insulin is a satiety hormone and facilitates weight loss and a low insulin level will make you hungry. That good old low carbohydrate paradox.

There is a rather stupid saying that "you are what you eat". It is slightly better phrased as "you are what you do with what you eat". You could go so far as to say "You are what your food does to you". I won't go in to epigenetics except to say that you can think about the phrase "You are what the food eaten by your mother and granny did to you". Certainly to your X chromosome(s) and your mitochondria.

In 1900 very few people had grannies who consumed even 64 lb/year of sucrose. More likely less than 30 lb/year. I remember dipping white bread toast spread with margarine and marmalade in to tea sweetened with three heaped spoonfuls of sugar at my granny's house in Bargeddie on the outskirts of Glasgow. And being amazed at how she could actually bring herself to inject her own leg every day with insulin. This was back in the 1960s, I'd have been about 5 years old.

If you are overweight and try going on a starch based spontaneously hypocaloric diet you may as well sign up for Barndard's disastrous diabetes diet. If you are far enough in to metabolic syndrome to find the label "diabetic" has been applied to yourself, going to a high carbohydrate diet will ruin you blood glucose control as soon as you stop losing weight. No one can lose weight for ever.

I was going to say that no one is going back to Kitava from modern Texas but this clearly depends on how permanent the damage done to you metabolism is. The more damaged you are, the more carbohydrate restriction is likely to benefit you long term.





As I read through this post there are two things which come to mind. First is that elevated insulin is core to weight gain. Second is that we have to be very careful about exactly what, under which circumstances, elevates insulin. Discarding insulin as a factor in obesity because there are circumstances in which starch does nor invariably elevate insulin is a serious case of throwing the baby out with the bath water. There are circumstances in which carbohydrate does not elevate insulin. Most of us don't live there, we can tell by our waist lines.


This post has been a long time coming. I've not particularly enjoyed writing it. But I have an insulocentric bias about obesity and its host of associated medical problems. Insulin provides a framework which, so far, paints a consistent picture of the way life works. It has served me pretty well.

Time to hit "post"

Peter