Buckle your seat belts, ladies and gentlemen-- we're going on a long ride through the scientific literature on sugar and body fatness. Some of the evidence will be surprising and challenging for many of you, as it was for me, but ultimately it paints a coherent and actionable picture.

Introduction: What Exactly are Sugar and Starch?

In this post, I'll be using the word 'sugar' to refer to three things: 1) sucrose, or table sugar, 2) high-fructose corn syrup (HFCS), and 3) fruit and honey sugars. Sucrose is composed of one molecule of glucose and one molecule of fructose, linked together, and it is therefore 50:50 glucose:fructose. When you eat sucrose, this bond is rapidly broken, releasing free glucose and fructose, which are then absorbed. HFCS is a mixture of free glucose and fructose that comes in two common varieties: 42:55 and 53:42 glucose:fructose. The former is most often used in soft drinks, while the latter is most often used in baked goods. Since sucrose and HFCS are both refined sugars composed of roughly half glucose, half fructose, one would expect them to have similar effects on the body, and controlled experiments overall have confirmed this. Fruit sugar is a mixture of sucrose, free glucose and free fructose. Honey is composed of roughly half free fructose, and half free glucose, but the ratio depends on the type of flowers visited by the bees.

Glucose is the form of carbohydrate that predominates in the blood, and along with fatty acids, it's one of the two primary fuel sources for the body. Starch is composed of long chains of glucose that are rapidly broken up during digestion. Because of this, dietary starch is similar to dietary glucose in the way it affects the body after ingestion.

Primate and Human Evolutionary History with Sugar

The earliest primates living more than 60 million years ago likely subsisted on fruit, leaves and insects, as most primates do today (Richard Klein. The Human Career. 2009). Chimpanzees and bonobos, our closest living primate relatives (~6 million years of divergence), are mostly frugivorous, meaning they're fruit eaters that obtain much of their energy from fruit sugars. For what it's worth (perhaps not much), these primates are both extremely lean in the wild. It's likely that our last common ancestor with them relied heavily on fruit.

Humans probably began to diverge from our ancestral primate dietary patterns around 2.6 million years ago when we developed stone tools. Archaeologists think this is when pre-humans began hunting/scavenging animal foods more frequently and relying more on animal sources of nutrition (Richard Klein. The Human Career. 2009). Some time over the ensuing two million years or so, we began using fire regularly and eating more starch foods, also presumably at the expense of leaves, fruit and insects.

However, our ancestors never abandoned their fruit-eating ways, and modern hunter-gatherers continue to eat fruit and honey when available. In equatorial and many temperate regions, fruit and honey together are a major source of calories during much of the year, and thus the diet contains a significant amount of sugar. For example, the Hadza hunter-gatherers of Tanzania rely heavily on fruit and honey (1). !Kung hunter-gatherers of the Kalahari desert also eat a substantial amount of fruit and sometimes honey (Richard Lee. The !Kung San. 1985). Both groups are characteristically lean. These cultures presumably eat in a way that resembles the diet of our own human ancestors.

There are also many examples of non-industrial horticultural and agricultural groups that eat fruit regularly. For example, according to the work of Dr. Staffan Lindeberg, residents of the Melanesian island of Kitavans obtain 50 g of carbohydrate per day from fruit, most of which would presumably be sugar (2, 3). This is about half the amount of sugar Americans eat today (discussed below). Dr. Lindeberg's research showed that Kitavans are quite lean, and have an undetectable incidence of heart attacks, stroke and diabetes. Their fasting insulin levels are low by Western standards.

It is difficult to escape the conclusion that our ancestors have been consuming sugar, in the form of whole fruit, continuously for the last 60 million years.

Recent History of Sugar Consumption in the US and Correlation with the Obesity Epidemic

Sugar intake in the US has increased greatly over the last 200 years, mostly due to an increase in added refined sugars. In 1822, the average American consumed an estimated 6.3 pounds of added sugar per year, while in 2004, consumption had increased to 101 pounds per year, a 16-fold increase.

This sounds like it could seriously implicate sugar in the obesity epidemic, but a closer inspection of the numbers complicates matters.

The CDC's NHANES surveys documented the beginning of the US "obesity epidemic" between the 1978 and 1991 survey periods, as well as a continued increase in obesity rates throughout the 2000s. Between 1978 and 2004, the prevalence of obesity in US adults increased from 15 to 33 percent (NHANES). Over that same time period, added sugar intake increased from 86 to 101 pounds per year, a 17 percent increase. On this basis, it could potentially be a factor in the obesity epidemic. However, there are clearly other factors involved, since a 17 percent increase in sugar consumption cannot reasonably account for the full 120 percent increase in obesity prevalence over this time period.

Observational Studies in Humans

Observational diet studies attempt to measure what people are eating in their day-to-day lives, and correlate that with other measures such as weight or health. There have been a number of observational studies on sugar consumption and body fatness, and they have yielded surprising but relatively consistent results. I'll let the authors of a review paper summarize (4):

In all cases in which the association is significant, there is an inverse relation between sucrose or carbohydrate intake and [body mass index]...

We find no reason at present to associate high consumption of sugar with obesity.

What the authors are referring to is that four out of the six studies they reviewed found that people who eat the most sugar are the leanest, and the remaining two studies found no association. The authors found an inverse relationship between sugar intake and dietary fat intake, and speculate that the reason people who report higher sugar intake are leaner is because they eat less fat.

This conclusion must be viewed with caution, because the method that's typically used to assess food intake in these studies, the food frequency questionnaire, is not very good at measuring the intake of sugar-sweetened foods and other foods that people view as unhealthy (5). That's probably because people wishfully under-report them. According to a paper evaluating the food frequency questionnaire:

These findings for specific foods suggest that participants over-reported consumption of foods often considered desirable or healthy, such as fruit and vegetables, and underestimated foods considered less desirable.

The degree of under-reporting is probably not evenly distributed across the populations they're studying. This is a problem that often plagues observational studies but is typically ignored.

What about sugar-sweetened beverages (SSBs) such as soda? In contrast to total sugar intake, higher intake of SSBs is generally associated with higher body fatness (6, 7).

In recent years, researchers have gotten more sophisticated about measuring fat mass, and now they often look at body fat distribution rather than simply total fat mass. A recent study found that fructose intake, mostly from sweetened beverages and food, was not associated with total body fatness in US adolescents, but it was associated with higher visceral (belly) fat mass specifically (8). It was also associated with insulin resistance, markers of inflammation and higher blood pressure-- all components of the metabolic syndrome. As we will see, this finding is quite representative of the research on sugar as a whole.

Overall, the observational evidence suggests that sugar in the form of SSBs is associated with elevated body fatness, but total sugar intake is not. This is our first clue that this story may be more complicated than "sugar makes you fat".

Controlled Studies in Animals

There have been literally hundreds of animal studies on sugar and body weight/fat gain, most of which were conducted in rodents. Sugar, when compared to starch in the context of a pelleted diet, can promote body fat gain in rodents in some contexts (9, 10, 11), but in some cases sugar actually leads to similar or even less less fat deposition than starch (12, 13, 14, 15). Diets high in sugar and fat together tend to be the most fattening of all (16, 17). The most commonly used fattening rodent diet, Research Diets D12492 (and related diets), is 34 percent fat and 9 percent sugar by weight (60 and 7% by calories, 18). This recipe was created by trial and error, and it's very effective at rapidly producing obesity in rodents-- more effective than any high-sugar low-fat diet. We'll come back to why the composition of D12492 is an obesity "sweet spot" in a moment.

There's another way to fatten rodents using sugar-- add it to the drinking water. Offering rodents sweetened liquids consistently causes body fat accumulation (19).

How can we explain the discrepancy between studies that reported that sugar is fattening, and those that reported that it isn't, or even that it's slimming? One thing I've noticed is that the pelleted diets that are the lowest or highest in sugar (e.g. 0% or 64 -73%) tend to be the least fattening, while diets that are in between (9 to 54%) tend to be the most fattening. This is not what we would expect if sugar were inherently fattening, in which case more sugar should lead to more body fat accumulation. There's a parsimonious explanation for these observations, which is summarized in this graph showing the sugar taste preferences of rats (20). Rats were offered a choice between plain distilled water and sugar water, and this was repeated with solutions containing different concentrations of sucrose:

Rats like sweet fluids, and the more sugar you add, the more they drink relative to plain water... up to a point. Beyond 27 percent sucrose (w/v), their sugar water intake plummets, and they go back to preferring plain water by 37 percent. There's a simple reason for this: they don't like the taste. Rats and humans alike have a palatability "sweet spot" for sugar, and exceeding it makes food unpalatable (commercial sweetened beverages are all within the human sweet spot). Sugar tastes good, but imagine yourself getting 70 percent of your calories from plain sugar cubes. You would get sick of it almost immediately. Together, this suggests that sugar may cause body fat accumulation in rodents in large part because it tastes good and it's rewarding, rather than because of some inherent metabolic effect of sugar.

To further explore this point, let's take a look at studies that compared refined glucose and fructose feeding. Glucose, as we've discussed, is metabolically very similar to starch after consumption, because starch breaks down into glucose rapidly in the gut. Sugar (sucrose or HFCS) is roughly 50:50 glucose and fructose. So any inherent metabolic effects of sugar that are different from starch would be expected to come from the fructose portion. Directly comparing dietary glucose to fructose allows us to approximately control for the sweetness (flavor) of the diets while identifying any unique fattening effects of fructose. This is the limitation of studies that compared sugar to starch: palatability/reward differed between conditions, so there's no way to know for sure which effects are due to palatability/reward, and which are due to inherent metabolic properties of sugar.

The studies that compared glucose to fructose are surprisingly consistent with one another in rodents, dogs, and (discussed below) humans. When diets high in glucose are compared head-to-head with diets high in fructose (and sometimes sucrose), the fructose diet increases circulating insulin, causes insulin resistance in the liver, but total body fatness remains similar between the glucose and fructose groups in almost every case (22, 23, 24, 25, 26, 27, 28). This is also true if the sugars are administered side-by-side in drinking water (28a). There is no difference in the effects of glucose, fructose or sucrose on total body fatness in animal models, suggesting that sugar probably does not have an inherently fattening effect that is independent of its calorie content and flavor.

Now, let's get back to that extremely fattening pelleted diet, Research Diets D12492. A central reason why it's so fattening to rodents is probably because it combines 1) a preferred concentration of fat, 2) a preferred concentration of sugar, 3) a high energy density, 4) a soft and preferred texture, and 5) constant accessibility. Both fat and sugar are in the palatability/reward sweet spot, neither too high nor too low, and rodents are crazy about the taste of it. It's right under their noses all day, and they have nothing better to do than eat it*.

As a whole, the animal research suggests that sugar and fructose are probably not inherently fattening, but that they can be fattening when they are used to increase the palatability, reward value and energy density of foods and beverages.

Controlled Studies in Humans

Fortunately for us, there are many controlled human studies investigating the effects of sugar consumption on body fatness, and there is more than enough information to come to a reasonable conclusion. Let's start with the most extreme example I have: a 1971 paper titled "Physiological Effects of a Mainly Fruit Diet in Man" (29). Subjects ate nothing but fruit (82% of kcal) and nuts (18% of kcal) for six months. The diets were 52-65 percent carbohydrate (mostly from fruit sugar), 37-45 percent fat (nuts and avocado), and 5.6-8 percent protein. These fascinating quotes sum up their findings:

A considerable number of the normal subjects claimed that their physical condition improved while they were on the diet. Some were convinced that their stamina increased and that their ability to undertake strenuous physical tasks and to compete in sport improved.

From these figures it is clear that most subjects lost weight initially... After the initial loss, the weights of the subjects in group B leveled off at figures which corresponded more or less with the calculated theoretical normal weights...

A few of these who were mildly overweight, found this diet an excellent incentive to reduce, and consequently showed a greater decrease in weight than the others. An interesting aspect of the diet was the tendency for the weights to level off more or less at the 'theoretically ideal' weight for the subject. This may partly explain why some lost more weight than others.

These people obtained the majority of their calories from unrefined sugar for 6 months, yet their body weights approached the 'theoretical ideal', with lean subjects remaining lean and overweight subjects losing weight. This finding is difficult to reconcile with the idea that sugar is inherently fattening. It is also an excellent illustration of a 'simple' (reduced reward) whole food-based diet normalizing body fat mass in humans. I think this diet isn't nourishing enough to be sustainable in the long term, but it certainly emphasizes the benefit of whole, simple food on body fat mass, regardless of sugar content.

A second example of weight loss on a high sugar diet is the "bland liquid diet" study I've referenced on this blog several times. Volunteers were restricted to a bland liquid formula that was high in sugar, but were not asked to restrict calories. While lean subjects maintained a normal calorie intake and body weight, obese subjects experienced a greatly reduced appetite and rapidly lost weight, with one man losing 200 lbs over 255 days (30). This once again shows that weight loss is possible on a high-sugar diet if it's lower in palatability and reward value.

A third very instructive study was published in 2002 (31). Investigators recruited overweight volunteers with the metabolic syndrome (people who should be sugar sensitive if anyone is) and assigned them to one of three diets for 6 months:

1. Replacing 1/4 of daily fat intake with 'simple carbohydrates' (sugared foods)


2. Replacing 1/4 of daily fat intake with 'complex carbohydrates' (both refined and unrefined starch foods)


3. A control diet in which nothing was deliberately changed

None of the diets were calorie restricted. Over the course of 6 months, group #2 lost 9 pounds (4.25 kg), but there were no significant changes in group #1 or #3 (although group #1 did lose 0.6 lbs). This suggests that sugared and fatty foods are equally fattening in the context of a typical diet, since substituting one for the other had no effect on body weight. However, replacing fatty foods with starchy foods produced weight loss, suggesting that the most commonly eaten sugared and fatty foods are fattening relative to starch foods. This study was unique because it replaced one thing with another rather than simply restricting or adding foods.

There are several studies showing that high-sugar diets do not impair weight loss in the context of a low-calorie diet (32, 33), once again confirming the primacy of calorie intake in weight loss.

Now let's examine studies that compared glucose to fructose feeding in humans, which would identify any unique fattening and metabolic effects of fructose, and by inference, sugar. Perhaps the most interesting study was conducted by Dr. Peter J. Havel's group in 2009 (34). They had volunteers obtain 25 percent of their energy needs from beverages sweetened with either glucose or fructose, for 10 weeks, without controlling total calorie intake. They used precise measures of body composition and distribution before and after the intervention. Here's how total body fat mass changed in each group:

In perfect agreement with the animal studies in rodents and dogs, total body fatness increased the same amount in both groups provided with refined glucose- or fructose- sweetened beverages. Also reminiscent of the animal studies, the fructose group saw a rapid increase in visceral fat specifically, an increase in fasting insulin and insulin resistance, and an increase in blood pressure. Thus although the refined fructose strongly promoted elements of the metabolic syndrome, it did not have any special ability to increase total fat mass beyond what occurred simply by adding a sweetened beverage to the diet. It's worth noting that these volunteers were receiving very large quantities of refined fructose, which it would be difficult to obtain via a normal diet.

This result is consistent with a number of other fructose feeding studies in humans, reviewed in a brand new paper in the Annals of Internal Medicine (35). The investigators reviewed 41 human fructose feeding trials. They concluded that fructose is not fattening when substituted for other carbohydrate in the diet, but it can be fattening if consumed in addition to the typical diet. This fattening effect is equivalent to what one would expect based on the increase in calorie consumption in these trials.

OK, so high-sugar diets don't necessarily produce body fat accumulation in humans, and they can even allow body fat loss under some circumstances, but are there situations where sugar can cause fat accumulation? The answer is an emphatic yes.

The most convincing studies in this regard were aimed at reducing the consumption of SSBs such as soda. These generally show that replacing sugar with starch or non-caloric sweeteners leads to reduced calorie intake and weight loss (36). In one recent study, investigators recruited obese and overweight volunteers, and had them replace the SSBs that they were already consuming with either diet beverages or water (37). Over a 6 month period, both groups lost about 5 pounds, much of it from their abdominal region, suggesting that the SSBs were one of the reasons they carried excess fat to begin with.

The controlled human studies once again suggest that sugar and fructose are not inherently fattening, but that refined sugar can cause body fat accumulation when it's used to increase the palatability, reward value and energy density of food and beverages. However, excessive consumption of refined sugar can promote elements of the metabolic syndrome, and this is due specifically to its fructose content.

Conclusions

Here are the take-home points from this post:

1. Sugar, including fructose, is not inherently fattening relative to other calorie sources, and unrefined sugar is compatible with fat loss in the context of simple whole food diets.


2. Sugar can be fattening in certain contexts, specifically if it is added to foods and beverages to increase their palatability, reward value and energy density.


3. Sugar-sweetened beverages are probably one of the most fattening elements of the modern diet.


4. Fruit is not fattening, and it may actually be slimming.


5. In excess, refined sugar can cause body fat to redistribute from the subcutaneous depot (under the skin, where you want it) to the visceral depots and the liver (where you don't want it). It can also cause insulin resistance in the liver and increase blood pressure, all components of the 'metabolic syndrome'. This is caused specifically by the fructose portion of the sugar.

Here are the implications:

1. Avoiding sugar-sweetened foods, and particularly sugar-sweetened beverages (soda, punch, sweetened coffee, cocktails, maybe fruit juice as well?) can prevent and to some extent reverse fat gain and metabolic dysfunction.


2. I see no reason to believe that refined and unrefined sugars, used in the same context (e.g. muffins baked with white vs. brown sugar), would have different effects on body fatness. However, unrefined sugars may be less harmful to other aspects of health, because they contain other substances that may be protective. Mark Sisson discussed this idea in a recent post on honey (38).


3. Eating fruit does not contribute to fat gain in most people, but instead probably favors leanness. Fruit is a whole food with a low energy density and a moderate palatability and reward value.

* This is actually a bit of an oversimplification. Rats will gorge themselves on this food for about two weeks, gain a bunch of body fat, but then their food intake returns to normal (this is presumably the energy homeostasis system kicking in as leptin and insulin levels rise). Despite the normalized calorie intake, they continue to gain fat at a slower rate. Furthermore, they will gain fat even if, from the very beginning of the experiment, you only allow them to eat the same number of calories as the regular unrefined chow group (39). This shows that these diets are not simply fattening because they cause increased food intake-- it's more complicated than that. This could still be explained by the reward/palatability value of the diet, if we invoke the idea that reward and hedonic centers act on energy homeostasis centers in the brain, favoring the 'defense' of a higher fat mass (40). But there are other contributing factors as well, such as the inflammation and lipid accumulation that develop in the hypothalamus and lower leptin and insulin responsiveness.