I wrote the book Always Hungry? to present the Fat Cell Model (also called the Carbohydrate-Insulin Model) of body weight control, as an alternative to the Calories In, Calorie Out approach to obesity treatment. According to this unconventional way of thinking, weight gain occurs because fat cells are stimulated by insulin and other anabolic signals to take in and store excessive calories. When this happens, the concentration of calories in the blood becomes depleted, leaving too few for the rest of the body. Perceiving this problem, the brain responds by increasing hunger and lowering metabolic rate — akin to a state of starvation — antagonizing long-term weight loss. In this sense, the conventional low-calorie diet is symptomatic treatment that makes the fundamental problem worse, by further restricting the available fuel supply in the blood stream.
One major driver of increased insulin secretion is the highly processed carbohydrate that inundated the American diet during the low-fat craze beginning in the 1970s (corresponding to the start of the obesity epidemic). Non-dietary factors, including stress, sleep deprivation, physical inactivity and endocrine-disrupting environmental contaminants, may also affect fat cells through hormonal or neurologic inputs.
We’ve known for nearly a century that body weight is under primary biological control over the long-term. Unfortunately, that understanding has failed to alter a clinical treatment paradigm that remains focused on behavioral methods to alter calorie balance, despite extraordinarily poor results in practice. It’s time for new thinking.
Below, I respond to the arguments (in italics) of Stephan Guyenet published on his Whole Health Source blog on January 9, 2016 (accessed January 20, 2016). A more complete discussion of these issues can be found in my book, my academic reviews [Ludwig , Ludwig] and in the works of Gary Taubes and others.
1. Overeating does make you fat. Randomized controlled trials have shown that eating excess calories causes fat gain, whether the extra calories come from fat or carbohydrate, and regardless of their impact on insulin levels (4, 5). If you eat too many calories, regardless of why you overate, you will gain fat (although some people are intrinsically more resistant to overeating-induced fat gain than others). That’s why overeating remains a key concept for understanding body fatness.
Not over the long-term. In the classic force-feeding studies, volunteers can be made to gain weight in the short-term by massive overfeeding. However, the body responds dynamically: the volunteers lose all interest in food and their metabolic rate tends to increase in the body’s attempt to burn off the excess calories. Indeed, volunteers in overfeeding studies characteristically report feeling intensely uncomfortable. Once the protocol ends, weight typically decreases back to baseline [Leibel, Roberts, Norgan, Sims] — a phenomenon that has also been well documented with experimental animals. Consistent with the Carbohydrate-Insulin Model, the metabolic responses protecting against long-term weight gain with overfeeding may not fully engage when excess calories are provided primarily as high glycemic load carbohydrate (candy) rather than protein and fat (peanuts) [Claesson].
Moreover, Guyenet ignores many experimental models of obesity notably lacking hyperphagia (above normal calorie intake), or in which overeating occurs secondary to increasing adiposity, such as defects in melanocortin signaling [Asai]. With regard to diet-induced obesity, rodents fed high versus low glycemic index (GI) starch showed the following sequence of events: first, hyperinsulinemia; second, anabolic changes in adipose tissue; third, greater adiposity; fourth, lower energy expenditure; and only then, fifth, increased energy intake. With food restriction to prevent excessive weight gain, the high GI-fed animals still develop excessive adiposity (and associated increased CVD risk factors) — findings that defy the conventional Calories In, Calories Out model [Kabir, Kabir, Lerer-Metzger, Pawlak, Pawlak].
Even with the prototypic models of hyperphagia such as hypothalamic damage and leptin deficiency, metabolic defects precede overeating, and restriction of food intake to normal does not prevent development of excess adiposity [Friedman, Dubuc, Bray].
2. Hunger is only one of the reasons we eat. We don’t generally eat dessert because we’re still hungry at the end of a meal. We don’t drink alcohol or put cream and sugar in our coffee because we’re hungry. Much of the eating we do in the affluent world has little to do with hunger — a phenomenon researchers call “non-homeostatic eating”.
Again, not over the long-term. Food intake can vary greatly from day to day based on numerous environmental and psychological influences. Many of us overeat on Thanksgiving. But we consistently compensate for under- or overeating over the long term, by eating more after a fast and less after a feast. Even in someone gaining weight at the rate of 2 pounds per year (enough to cause obesity by middle age), integrated calorie balance remains remarkably stable, to within >99%.
Although long-term studies in humans to distinguish hedonic from homeostatic eating have not been done and would be methodologically daunting, the dominance of the latter is self-evident. A starving person will eat virtually anything edible, whereas someone who has overeaten will find even the most palatable foods less appealing.
Animal studies demonstrate that the metabolic effects of food dominate eating behavior over the long term. Rodents appear to like the taste of sugar water and, with unlimited access, will overconsume it and become fat (not unlike humans). But if the sugar water is spiked with an intensively bitter chemical, the rodents overcome their instinctual dislike of that aversive taste and still become fat [Sclafani]. In humans, a high GI meal was shown to activate the nucleus accumbens, a critical brain center that mediates reward and cravings, evidently through metabolic effect independent of palatability [Lennerz].
3. Blood levels of fat and glucose tend to be normal or elevated in people with obesity and high insulin, not lower (6, 7, 8). That’s because they’re insulin resistant, meaning that insulin isn’t doing its job of constraining blood glucose and fat levels as effectively. Since people with obesity/overweight don’t have lower circulating energy levels than lean people, this cannot explain why they eat more. Obesity is not a condition of “internal starvation”.
Static analysis after obesity has developed, rather than during its dynamic stages, are misleading. The physiological mechanisms of many diseases can only be observed during development, before the body reaches a new (albeit pathological) steady state. Consider for example Addison’s disease, in which the adrenal glands lose the ability to produce aldosterone, a critical hormone that helps the kidneys retain sodium. As this condition develops, sodium in the urine is definitively elevated. But as the concentration of sodium in the blood continues to fall, less sodium is delivered to the kidneys and urinary sodium decreases. A static analysis of urine sodium late in the disease, without considering the full picture, could be falsely interpreted as normal (or possibly even low).
Indeed, many secondary pathophysiological changes develop with obesity that would obscure etiology, including hypothalamic inflammation, central leptin resistance, and peripheral insulin resistance (although fortunately, these are potentially reversible through diet). For this reason, cross-sectional analysis of calorie concentration in the blood is uninformative in the same way that measuring serum insulin levels in a state of insulin resistance can’t tell us anything about insulin action in target organs. Moreover, cross-sectional studies don’t tell us whether the individuals under study are in the process of gaining or losing weight — a critical distinction.
Although the natural history of obesity has not been well studied in humans, animal research provides key insights into the relevant mechanisms. Immediately following experimental hypothalamic damage in rodents, insulin secretion and autonomic nervous system stimulation of fat cells increase, redirecting calories to fat tissue at the expense of lean body mass before overeating has developed [Friedman, Bray, Penicaud]. These findings are fully consistent with the high GI diet model of obesity in rodents.
The total concentration of calories in the blood — from glucose, free fatty acids, and ketones — is tightly controlled because of the body’s continuous need for fuel (ranging between 4 and 6 kcal/L according to one study [Walsh]). An acute decrease in the circulating concentration or oxidation of these fuels provokes intense hunger and food intake [Friedman, Thompson]. Conversely, pharmacological manipulations that increase metabolic fuel availability, such as fatty acid synthase inhibition or b3 agonist administration, lower food intake [Cha, White]. Although disregarded by the Calorie In, Calorie Out model, fuel availability can be acutely affected by what we eat, independently of how much we eat.
In a crossover feeding study, 12 children were given low, moderate, or high glycemic load meals with similar calorie content at breakfast and lunch. As expected, blood glucose and insulin levels were initially higher after the high glycemic load meal, compared to the other two meals. However, 3 to 5 hours later, the concentration of glucose and free fatty acids in the blood were lowest after that meal. Also at that time, stress hormones surged and overeating ensued after the high glycemic load meal — evidence for the clinical relevance of these metabolic events [Ludwig]. In another study, the combined total calories in the blood was lower in the late postprandial period after a high carbohydrate meal compared to moderate or low carbohydrate meals, coincident with a fall in energy expenditure [Walsh].
Indeed, low blood glucose after consumption of high GI carbohydrate is so common as to be considered “normal” [Lev-Ran, Brun]. Even when glucose is given intravenously to prevent hypoglycemia, raising insulin level has been reported to increase hunger, heighten the palatability of sugar, and cause overeating (presumably due to suppression of free fatty acids) [Rodin].
Together, these findings provide a mechanism for understanding how the hormonal response to ingestion of high glycemic load carbohydrates could limit access to metabolic fuels and cause weight gain.
4. Fat cells do not have an increased affinity for fat in people with obesity and high insulin. In fact, people with obesity and elevated insulin release fat from their fat tissue at a higher rate than lean people with lower insulin (higher total lipolysis rate; 9). Again, this may relate to the fact that they’re insulin resistant.
Again, cross-sectional analyses can be misleading. The relevant question is not whether someone is obese or lean, but whether that person is in the process of gaining (or losing) weight. Such fundamental information cannot be discerned from static analyses. Many animal models of obesity clearly show increased insulin sensitivity in fat and insulin resistance in muscle during the critical early stages of development, such as with hypothalamic damage, a high GI diet (both considered above), or peripheral insulin administration [Cusin].
5. Body fatness is regulated by the brain, not by fat tissue or the pancreas. There is a vast research literature showing that the brain regulates food intake, energy expenditure, and fat tissue metabolism to regulate the size of body fat stores (10). There is no known mechanism intrinsic to fat tissue or the insulin-secreting pancreas that does this. Genetic differences that impact body fatness tend to be located in genes that affect brain function, not fat tissue or insulin signaling (11, 12).
The brain of course mediates hunger and metabolic rate, but that fact tells us nothing about peripheral and external influences. In any model of metabolic homeostasis, the brain must receive inputs from the body, and these plausibly include the concentration of circulating metabolic fuels, peripheral hormones, and the rich afferent autonomic innervation that transmits information from the gastrointestinal track and adipose tissue to the brain [Yamada, Coppack].
Contrary to claim, straightforward mechanisms relate increased insulin secretion to obesity, as illustrated by excessive weight gain with insulin-producing tumors, genetic variants of the insulin promoter, direct peripheral insulin administration, and other models [Le Stunff, Cusin, Sigal].
In experimental animals, genetic manipulation of numerous biochemical pathways outside the brain results in obesity, including muscle-specific insulin receptor ablation [Kim], adipose-specific overexpression of 11-beta hydroxysteroid dehydrogenese type 1 (an enzyme involved in glucocorticoid metabolism) [Masuzaki], and liver-specific overexpression of sterol regulatory element-binding protein-1c (a transcription factor regulating de novo lipogenesis) [Knebel]. Indeed, the development of obesity following experimental insulin administration provides compelling evidence that the peripheral anabolic effects of this hormone dominate any central catabolic effects [Cusin].
Serum insulin levels provide no meaningful measure of insulin action. Circulating insulin levels may increase for two reasons, each with opposite physiological significance. Primary hyperinsulinemia — resulting from diet, endogenous causes, or direct insulin administration — produces increased insulin action and weight gain. Secondary hyperinsulinemia is a compensatory process arising from insulin resistance that results in decreased insulin action and serves to protect against ongoing weight gain. Most observational analyses fail to make this distinction and are therefore uninformative.
A simple model of primary hyperinsulinemia is insulin administration. People with type 1 diabetes who receive inadequate insulin treatment invariably lose weight, no matter how much they consume. Conversely, insulin initiation in type 2 diabetes or overtreatment in type 1 diabetes predictably causes weight gain. A component of this effect is metabolic, not simply differences in glucose loss in urine [Carlson].
In animals, insulin administration causes weight gain. Even when food intake is restricted to prevent weight gain, the animals still become excessively fat, showing that insulin redirects metabolic fuels to adipose tissue at the expense of lean body mass [Torbay]. Moreover, a transgenic experiment in mice found that reduced primary hyperinsulinemia increased white adipose tissue expression of uncoupling protein, increased energy expenditure, and protected the animals from diet-induced obesity. The investigators concluded that “circulating hyperinsulinemia drives diet-induced obesity and its complications” [Mehran].
7. If high insulin were a major contributor to obesity, weight loss would be a positive feedback process. In other words, the more weight you lost, the easier it would become to lose further weight. This is because weight loss itself reduces insulin levels, both between and after meals (15, 16). Yet what we observe is the opposite: weight loss becomes more difficult the more you lose, despite declining insulin levels (a negative feedback process).
As above, this reasoning conflates insulin levels and insulin action. With weight loss on a conventional diet, fat cells become increasing sensitive to insulin action, but if the diet remains high in glycemic load, the resulting hypersecretion of insulin will exert too great an anabolic effect on fat cells. Consequently, an increasing proportion of the (reduced) calorie supply becomes directed to adipose, pushing the body into starvation physiology — increased hunger, decreased metabolic rate, rising stress hormone production — even though total calorie stores in fat may remain above normal. This mechanism provides an explanation for the difficulties most people have adhering to a calorie restriction, even while body fat stores remain high.
8. Foods that lead to higher blood levels of glucose and insulin do not result in greater subsequent hunger. The most comprehensive study examined 38 common foods and found no relationship between glycemic index and subsequent hunger, and an inverse relationship between insulin levels and hunger (i.e., foods that caused greater insulin release tended to be more filling; 17).
The metabolic problems following a high glycemic load meal occur several hours after eating. The review cited by Guyenet from 20 years ago examined glycemic and insulinemic responses 2 hours after the meal. As extensively reviewed [Ludwig], the metabolic problems with high glycemic load foods emerge in the late postprandial period (after about 3 to 5 hours), when availability of metabolic fuels is reduced [Ludwig, Walsh, Ludwig, Roberts]. Analyses that fail to distinguish between early effects (when blood glucose surges) and later effects (when metabolic fuels decrease) will be uninformative.
9. Diets that reduce blood glucose and insulin swings (low-glycemic) are not an effective tool for weight control. This has been shown repeatedly in RCTs lasting longer than two months (18, 19, 20, 21, 22, 23), including an 18-month study by Ludwig’s group that found a low-glycemic-load diet to provide the same weight and fat loss, and the same participant satisfaction, as a standard low-fat diet (24). This is despite the fact that these studies often don’t control for confounding dietary factors like fiber content, calorie density, protein, and/or palatability (i.e., the “low-glycemic” diet is often a whole-food-based diet).
Most long-term behavioral diet studies suffer from severe non-compliance, limiting inferences. The vast majority of long-term behavioral studies of macronutrients and body weight fail to produce meaningful differentiation between treatment groups due to their weak interventions and the inherent challenges of long-term behavior change. Therefore, it’s not surprising that these studies don’t show impressive differences in weight between groups. We certainly need to develop better behavioral methods, but to understand the intrinsic efficacy of diet requires higher quality research, such as feeding studies. The best and largest of these studies quite clearly demonstrate superiority of low carbohydrate and low GI diets, including the Direct trial and Diogenes [Shai, Larsen]. Furthermore, Guyenet disregards a major finding in our 18-month clinical trial of special relevance to the Carbohydrate-Insulin Model.
In 1999, my coauthors and I iterated a central component of the Carbohydrate-Insulin Model as follows: “[V]arious factors that augment insulin secretion or action (intrinsic to islet cells, at peripheral sites of action, or dietary) might promote obesity. Individuals who, perhaps for genetic reasons, have an exuberant insulin response to glucose may be especially sensitive to dietary GI” [Ludwig].
Results from translational research, observational research, and feeding studies provide strong support for this hypothesis. In these studies, insulin secretion:
- predicted most of the variation in weight gain among rodents fed a high GI diet (R2 = 0.84, P < 0.0001) but none of the variation among animals fed a low GI diet (R2 = 0.003, P = 0.94) [Pawlak].
- was associated with 6-y weight gain (R2 = 0.26, P < 0.0001) and change in waist circumference (R2 = 0.30, P < 0.0001) in the Quebec Family Study among those eating a low fat/high glycemic load diet, but not among those eating a high fat/low glycemic load diet (P > 0.05) [Chaput].
- influenced response to diet in our 18-month clinical trial such that individuals with high insulin secretion showed significantly greater weight loss on a low glycemic load diet (13 lb) than a low fat diet (3 lb, P = .004; P for interaction = 0.02) [Ebbeling]
- predicted adverse changes in body composition and metabolism following weight loss [Hron]
10. Billions of people globally eat high-glycemic diets and remain lean. Many traditional diets are very high in starch and low in fat. If foods that promote large blood sugar and insulin spikes were the primary factor in obesity, shouldn’t these people be obese?
So-called “ecological” studies comparing different populations comprise the lowest quality epidemiological data. The white rice consumed by peasants in China may keep them from starvation — but that says nothing about effects in other populations. Now that those individuals are moving to the cities, taking their high carbohydrate diets, but leaving behind the high levels of physical activity, rates of obesity and diabetes are skyrocketing. Which isn’t to say all low fat diets are inherently unhealthy. But as Americans cut back fat beginning in the 1970s, they also increased high glycemic load grains, potatoes, and added sugars — not whole fruits, non-starchy vegetables, or legumes.
11. There is no evidence that our appetites increase, and our energy level drops, because our fat cells are hoovering up fat from the bloodstream. You would think, with how often this is repeated, that there would be some kind of evidence that this process is actually happening in common obesity. Yet despite having read a number of works by Taubes and Ludwig, I haven’t found anything more concrete than speculation and analogies. The concrete evidence I have encountered (#3 and 4 above) is at odds with the claim.
The existing paradigm doesn’t work and it’s time for new thinking! The Calories In, Calories Out model states that weight control is simply a matter of eating a bit less, and moving a bit more. Although our modern food environment may offer many temptations, it’s ultimately a question of will power. However, this way of thinking disregards a century of research showing that body weight is controlled more by biology than will power over the long term. Indeed, the obesity epidemic has progressed despite an incessant focus on calorie balance by the government, professional nutrition establishment, and, recently, the food industry (witness the 100 Calorie Pack). To add insult to injury, the conventional model blames people with obesity for failure to control their calorie balance. But if conscious control of calorie balance were so crucial, how did people manage to avoid massive swings in weight before the notion of the calorie was invented a century ago?
Contrary to claim, the Carbohydrate-Insulin Model is founded on solid science, as detailed above and summarized here:
- The body fights back against calorie restriction, with increased hunger and lower metabolic rate — biological adaptations that make weight loss maintenance progressively more difficult over time.
- Metabolic dysfunction involving fat cells has been definitively shown to precede overeating, at least in some models of obesity.
- Primary hypersinsulinemia promotes calorie storage in adipose tissue, lowers the concentration of calories in the blood stream, triggers overeating, and causes long-term weight gain. In a state of hyperinsulinemia, restriction of food intake may diminish the rate of weight gain, but does not prevent excess adiposity.
- Highly processed carbohydrates cause more insulin secretion, calorie for calorie, than any other food, and are consistently associated with the most weight gain in the best cohort studies [Mozaffarian].
- The highest quality weight loss diet trials — with measures to assure differentiation between treatment groups — show clear advantages of a reduced glycemic load diet compared to a low fat diet.
- Reducing glycemic load appears to attenuate the biological adaptations antagonizing weight loss, including decreased energy expenditure [Pereira, Ebbeling].
Thus, the Carbohydrate-Insulin Model argues that calorie restriction causes a state of deprivation in the body, resulting in the hallmarks of starvation: rising hunger and falling energy expenditure. This pushback from the body can be mitigated by diet quality — especially by reducing consumption of highly processed carbohydrates — resulting in long-term weight loss with less difficulty. The Carbohydrate-Insulin Model also provides a more plausible explanation than lack of willpower for the poor results of conventional diets (as most people with obesity have a strong desire to achieve lasting weight loss).
Of course, much more research will be needed to reach a full understanding of the etiologies of obesity and individual variability in response to treatment. Until then, we must keep our minds open to new, potentially more effective approaches to the public health crisis of obesity-related disease.