Despite taking a serious battering over the last few years*, the carbohydrate-insulin hypothesis of obesity remains popular in the public sphere, and to a lesser extent, among researchers. Public advocates such as Gary Taubes, and researchers such as David Ludwig and Robert Lustig, have vehemently argued that obesity has little to do with calories. Rather, it's a condition in which the body is primed for hunger and fat storage due to a hormonal disorder-- and the culpable hormone is insulin.
Insulin is an attractive mechanism because of its ability to affect blood levels of sugar and fat, and fat flux in/out of fat cells. According to one popular iteration of the carbohydrate-insulin hypothesis, high levels of insulin reduce blood sugar and fat in the bloodstream, and also trap fat inside fat cells, leading to hunger and fat gain. Since we know that carbohydrates are particularly good at elevating insulin levels, and obese people tend to have high levels of insulin, it all seems to add up pretty well. Adding to the attractiveness of the hypothesis, under free-living conditions, overweight people tend to lose more weight when they follow a moderate low-carbohydrate diet than when they follow a moderate low-fat diet.
These facts led to the development and meteoric rise of the modern carbohydrate-insulin hypothesis of obesity, complete with ardent claims that carbohydrate and insulin are the only relevant causes of obesity. I can't explain this better than Kevin Hall and colleagues did in the introduction of their paper (1):
Weight loss diets often recommend targeted restriction of either carbohydrates or fat. While low-fat diets were popular in the latter part of the 20th century, carbohydrate restriction has regained popularity in recent years, with proponents claiming that the resulting decreased insulin secretion causes elevated release of free fatty acids from adipose tissue, increased fat oxidation and energy expenditure, and greater body fat loss than restriction of dietary fat (Ludwig and Friedman, 2014, Taubes, 2007, Taubes, 2011, Westman et al., 2007). One influential author concluded that “any diet that succeeds does so because the dieter restricts fattening carbohydrates …Those who lose fat on a diet do so because of what they are not eating—the fattening carbohydrates” (Taubes, 2011). In other words, body fat loss requires reduction of insulinogenic carbohydrates. This extraordinary claim was based on the observation that even diets targeting fat reduction typically also reduce refined carbohydrates. Since the primary regulator of adipose tissue fat storage is insulin, and a reduction in refined carbohydrates reduces insulin, carbohydrate reduction alone may have been responsible for the loss of body fat—even with a low-fat diet.This hypothesis may be a bit extreme, but it's a good hypothesis in the sense that it makes testable predictions. For example, it predicts that specifically dropping carbohydrate from the diet should cause more fat loss than specifically dropping the same number of fat calories from the diet, if everything else is held constant. Surprisingly, no study has ever directly tested this! Many studies have compared low-carbohydrate to low-fat weight loss diets under equal-calorie conditions, but all of them changed the intake of multiple macronutrients from baseline values, rather than manipulating only single macronutrients.
A new study by Hall's group directly tests this prediction, making it one of the most pertinent tests of the carbohydrate-insulin hypothesis of obesity ever conducted. Onward!
Hall's team started with a group of 19 male and female obese volunteers without overt disease. Volunteers were randomly assigned to one of two groups:
- Controlled baseline diet for 5 days, followed by 6 days of 30% calorie restriction. Calorie intake was reduced by reducing carbohydrate intake ONLY.
- Controlled baseline diet for 5 days, followed by 6 days of 30% calorie restriction. Calorie intake was reduced by reducing fat intake ONLY.
All diet phases were strictly controlled under metabolic ward conditions. The baseline phase was used to determine each volunteer's calorie needs. After completing one diet phase (e.g., low-carbohydrate), each volunteer went home for 2-4 weeks, then completed the other diet phase (e.g., low-fat). This is called a crossover study design, and it's extremely robust because each person is compared to himself, which reduces measurement variability. In other words, this study compares Bob eating a low-carbohydrate diet to Bob eating a low-fat diet.
It's worth noting that sugar intake was the same between the baseline and low-fat periods. Any observed effects cannot be attributed to reduced sugar intake.
At several points during the experiment, Hall's team took a wide variety of metabolic measurements, including metabolic rate and fat oxidation. To do this, they measured the breath exchange of carbon dioxide and oxygen that indicate how many calories a person is burning, and whether those calories are coming from carbohydrate or fat. By subtracting the amount of fat a person is burning from the amount of fat they're eating, researchers can calculate how much fat a person is losing (or gaining)**. They also measured body weight, and directly measured body composition using DXA, a gold-standard method.
They also measured blood insulin levels, urine C-peptide levels, and several other hormones. C-peptide is a protein that's co-released with insulin but has a longer half-life in circulation than insulin. It ends up in the urine, and it's commonly used as a marker of overall insulin secretion over the course of the day.
I want to note that this study's methods were downright obsessive. The overall study design and diets were extremely tightly controlled, and the researchers took a large number of measurements using gold-standard methods. Even though the study was short, it must have required a huge effort and cost a lot of money. The study's rigor is reflected in the fact that it was published in Cell Metabolism, a high-quality journal.
As expected, the low-carbohydrate diet led to a 22 percent reduction in overall insulin secretion, as measured by urine C-peptide. In contrast, the low-fat diet had no impact on insulin levels.
Also as expected, the low-carbohydrate diet caused greater weight loss than the low-fat diet. The low-carbohydrate diet also led to a higher rate of fat burning. Yet this did not quite make up for the fact that the low-carbohydrate diet contained more fat than the low-fat diet.
Metabolic measures indicated that the low-carbohydrate diet caused a 245 gram (0.5 lbs) loss of body fat over the 6-day diet period, while the low-fat diet led to a 463 gram (1 lb) loss of body fat over the same period. Total energy expenditure decreased significantly in the low-carbohydrate group, but not in the low-fat group (-98 vs. -50 kcal/d).
DXA analysis showed that both groups lost fat, but the changes were too small and the variability too large to detect a significant difference between groups.
First, let's make sense of the results. Body weight loss was higher on the low-carbohydrate diet, yet body fat loss was higher on the low-fat diet. This might seem difficult to imagine, but in fact it's not as weird as it might seem, because low-carbohydrate diets cause a rapid loss of water weight from liver and muscle tissue. This is one of the reasons why they're so popular-- they deliver big, immediate weight loss results.
Yet obesity is a condition of excess body fat, not excess water weight. What matters is losing fat. In this study, the low-fat diet caused twice as much fat loss as the low-carbohydrate diet over the 6-day period.
The study's title is provocative: "Calorie for Calorie, Dietary Fat Restriction Results in More Body Fat Loss than Carbohydrate Restriction in People with Obesity". And that is indeed the case-- over a 6-day period. Yet surprisingly, using their evidence-based mathematical model of human metabolism, Hall's team predicts that the low-fat diet would continue to outpace the low-carbohydrate diet out to 6 months. I don't pretend to understand the mathematical underpinnings of the model, but I will note that previous experience shows that it's pretty darn good at predicting the metabolic and body weight outcomes of human diet studies.
How do we reconcile this with previous research suggesting that over longer periods of time, diet composition stops being very important for weight loss and only calories matter? And how is the difference in fat loss they observed possible, since calorie intake was the same on both diets? I believe part of the explanation may be relatively straightforward: macronutrient extremes result in metabolic inefficiency (just to be clear, this is a hypothesis that I believe explains the evidence, but it remains to be tested directly). By default, the body uses a liberal mix of carbohydrate and fat, and a bit of protein to fuel its daily activities. When you restrict one of those to an extreme, you take the body outside its normal operating parameters, and metabolism becomes less efficient. In other words, your body requires more energy to perform the same tasks.
In the current study, calories were subtracted exclusively from dietary carbohydrate or fat, but because of the composition of the baseline diets, the low-fat diet ended up with a much more extreme composition than the low-carbohydrate diet. Here are pie charts showing the composition of each diet, including the baseline diet:
As you can see, the low-fat diet was a spartan 8 percent fat! That is extremely low-- among the lowest of any diet study I've ever seen. Meanwhile, the low-carbohydrate diet was a comfortable 29 percent carbohydrate (although that represents a 60% reduction in absolute carbohydrate intake vs. baseline). This was necessitated by the study design, but it did lead to a more extreme diet composition in the low-fat group.
My hypothesis about metabolic efficiency is supported by the fact that the low-fat diet didn't cause a reduction in energy expenditure, while the low-carbohydrate diet did. In other words, the low-fat diet had a small "metabolic advantage", demonstrating that it was more inefficient. Yet this metabolic advantage isn't big enough to account for the difference in fat loss, leaving most of it unexplained. Assuming the fat mass data are correct, this suggests that the low-carbohydrate group may have led to a greater loss of calories from sources other than body fat (e.g., glycogen and/or lean tissue), which they allude to in the paper.
The "metabolic advantage" is usually associated with low-carbohydrate/high-protein diets. For example, in a study by David Ludwig's group, a spartan 10 percent carbohydrate, higher-protein diet led to a higher energy expenditure following weight loss than a moderate low-fat diet (2). Yet more moderate low-carbohydrate diets don't seem to yield a metabolic advantage (3). Again, an extreme diet composition seems to be the key, not carbohydrate restriction per se.
It's interesting to note that diet advocates in both the very-low-carbohydrate and very-low-fat camps claim a metabolic advantage and superior fat loss. If my hypothesis is correct, they're both right!
Hall's group actually alludes to this hypothesis in their paper. According to their metabolic model, if the low-carbohydrate diet had been even lower in carbohydrate (with a corresponding increase in fat intake to maintain calories), it would have matched the very-low-fat diet in the fat loss department. Their data suggest that both macronutrient extremes are a bit more effective for fat loss than being in the middle, even when calories are held constant.
Another explanation for the possible metabolic advantage of typical low-carbohydrate diets is the higher protein intake, which can lead to both short-term and long-term increases in energy expenditure, particularly during and after weight loss. That wasn't a factor in Hall's study, but it could have been a factor in others.
These results directly refute the carbohydrate-insulin hypothesis of obesity (at least one if its most popular incarnations). The low-carbohydrate diet caused a 22 percent reduction in insulin secretion, but only half the fat loss of the low-fat diet. Under the conditions of this study, the relationship of insulin levels to fat loss was the opposite of what the carbohydrate-insulin hypothesis predicts. Let me pass the mic to Hall and colleagues:
We can definitively reject the claim that carbohydrate restriction is required for body fat loss.
Now, let's talk about what this study is, and what it isn't:
- It is an investigation of the mechanisms of short-term weight loss, specifically the mechanistic importance of calories vs. diet composition.
- It is a well-designed test of the carbohydrate-insulin hypothesis of obesity.
- It isn't a test of which diet works best under real-world conditions, or how different diets affect hunger, food motivation, or food intake.
We already have many other studies that test the real-world effectiveness of low-fat and low-carbohydrate diets. These studies show that neither diet is very effective as a stand-alone fat loss measure, but moderate low-carbohydrate diets tend to be more effective than moderate low-fat diets, perhaps due to their greater ability to control appetite. We still need more data on the real-world effectiveness of very-low-fat diets, which may be more effective than the moderate low-fat diets typically used as control groups in low-carbohydrate diet studies.
This is a very sophisticated study that takes us several steps toward understanding the metabolic and energy balance effects of different diet compositions. The results are surprising and provocative.
The results suggest that over a 6-day period, specifically dropping fat from the diet leads to greater body fat loss than specifically dropping carbohydrate from the diet, when calories are held constant. Yet, the difference in fat loss may depend on the fact that the composition of the low-fat diet was much more extreme than that of the low-carbohydrate diet. If both diets were equally extreme, Hall's model predicts that fat loss would have been similar. This hypothesis needs to be taken ex silico to be confirmed, hopefully in a longer-duration diet trial.
Since dropping carbohydrates from the diet lowered insulin but slowed fat loss when compared to dropping dietary fat, this study falsifies one of the most popular incarnations of the carbohydrate-insulin hypothesis of obesity. I hope regular readers of my blog aren't too shocked.
* Which I will modestly take some credit for.
** Their method also takes de novo lipogenesis into account.