Not only did the VLC diet lead to a higher total energy expenditure than the LF and LGI diets, the most remarkable finding is that it led to a higher resting energy expenditure. Basically, people on the VLC diet woke up in the morning burning more energy than people on the LGI diet, and people on the LGI diet woke up burning more than people on the LF diet. The VLC dieters burned 326 more calories than the LF dieters, and 200 more than the LGI dieters.
It's always tempting to view each new study in isolation, without considering the numerous studies that came before it, but in this case it's essential to see this study through a skeptical lens that places it into the proper scientific context. Previous studies have suggested that:
- The carbohydrate:fat ratio of the diet has little or no detectable impact on energy expenditure in people who are not trying to lose weight (2, 3).
- The carbohydrate:fat ratio of the diet has little or no detectable impact on energy expenditure in people who are being experimentally overfed, and if anything carbohydrate increases energy expenditure more than fat (4, 5).
- The carbohydrate:fat ratio of the diet has little or no detectable impact on energy expenditure during weight loss (6, 7, 8), and does not influence the rate of fat loss when calories are precisely controlled.
With that, let's see what could have accounted for the differences observed in Dr. Ludwig's study.
Was it Insulin?
I know it will be tempting to some people to attribute these results to changes in circulating insulin*. The hypothesis here is that lowering insulin increases energy expenditure. Let's see how plausible that is.
The first thing to point out is that the paper didn't actually report levels of fasting or meal-stimulated insulin, so we have no idea whether they differed, or if so, by how much. Most studies have shown that fasting insulin tends to be lower on diets that are very low in carbohydrate, but again this was not reported in the current study so all we can do is speculate. One would expect that the higher carbohydrate content of the low-fat diet in particular would lead to higher insulin after meals. However, the VLC diet was the highest in protein (30% vs. 20%), which also stimulates insulin release to a surprisingly large degree-- often more than an equivalent serving of carbohydrate (9). The point here is that we simply don't know how much insulin differed between the three groups.
The second, more important point is that lowering insulin levels does not appear to increase energy expenditure. The drug diazoxide, in addition to its effects in the brain and elsewhere, reduces both fasting and post-meal insulin secretion substantially (35-50%). In some but not all studies, diazoxide accelerates fat loss, and this may be related to its effects in a part of the brain called the hypothalamus
(10, 11, 12). When you give obese people diazoxide in conjunction with a low-calorie diet, it has no impact on energy expenditure, regardless of whether you look at the studies where it accelerated fat loss, or the ones where it didn't (13, 14). This suggests that when it works, it's because it reduces food intake.
Just to add more support to this idea, in free-living people, elevated insulin is associated with higher, not lower resting energy expenditure (15, 16). Insulin actually increases energy expenditure following a meal, at least in part via the brain, and so it is said to contribute to "diet-induced thermogenesis" (17). This is part of the reason why carbohydrate consumption is often associated with a higher post-meal energy expenditure than fat consumption (with protein being the highest, 18). It's worth noting that the difference in diet-induced thermogenesis between carbohydrate and fat is quite small.
Furthermore, it has been shown multiple times that on average, people with higher fasting insulin levels, and more insulin resistance, gain less fat over time than people with normal insulin levels (19). A recent study suggested that when you control for baseline fat mass however, this relationship disappears-- in other words, circulating insulin is not related to future fat gain (20). This is reminiscent of what we see in animal models, where a) increasing circulating insulin by making the liver insulin resistant does not cause fat gain (21), and b) preventing the increase in circulating insulin that normally occurs when they're given a fattening diet has no impact on the rate of fat gain (22). This shows that elevated circulating insulin is neither necessary nor sufficient to cause fat gain in animals, and suggests that the same is probably true in humans.
The overall point here is that insulin is not a compelling explanation for the effect they observed in this paper-- we need to look elsewhere.
Diet-induced Thermogenesis due to Increased Protein Intake
As mentioned above, protein increases energy expenditure following a meal (called 'diet-induced thermogenesis'), and differences in protein content between diets can result in differences in energy expenditure of up to 100 calories over the course of a day (23). The VLC diet was 30 percent protein, vs. 20 percent for the LGI and LF diets. This could account for a portion of the difference in total energy expenditure, although it can't account for the difference in resting energy expenditure because that's measured prior to eating.
The body contains and uses large quantities of protein, fat and carbohydrate. At macronutrient extremes, the body has to synthesize the missing macronutrient, and this is an energetically costly process. In the case of the VLC diet, participants' bodies would have had to make carbohydrate from amino acids and glycerol-- a process called gluconeogenesis. This uses energy, possibly accounting for a portion of the increased energy expenditure in that arm.
It would have been interesting to see a very low fat (10% or less) diet as a comparison group. It's possible that energy expenditure would have increased as participants approached the other end of macronutrient extremes.
How About Leptin and Insulin in the Brain?
One of the things I'd really like people to take away from my writing and talks is that where energy intake, energy expenditure, and body fatness are concerned, the brain is key. A particularly important system for fat mass regulation is the negative feedback loop between fat tissue and the brain that is completed by the hormone leptin. Leptin is by far the most important hormone for the regulation of body fatness in mammals. There are many things that influence food intake and energy expenditure, and some of these don't have a lot to do with leptin or physiology in general-- e.g. aunt Sally baked you a pie, or you're having drinks with friends, or you buy a dog and have to walk it twice a day. But leptin is the key hormone a circuit that does its best to regulate the balance between energy in, and energy out, and so it's always important to consider in cases like this.
As far as I know, the brain is the only organ that's capable of regulating energy expenditure, and it does so by controlling thyroid signaling, the HPA axis, the sympathetic/parasympathetic nervous systems, and physical movement. It has already been demonstrated that the metabolic response to fat loss in humans-- including the characteristic decline in energy expenditure per unit lean mass-- depends on the reduction in leptin that accompanies fat loss (24). The citation I just made was also in Dr. Ludwig's discussion section, and this is no coincidence! Although any explanation for Dr. Ludwig's finding is speculative at this point (duly acknowledged in the paper), leptin is a compelling possibility.
Leptin declined the most on the VLC diet, followed b the LGI diet, followed by LF. This doesn't tell us much about what's happening in the brain however, because we don't know how each diet affected leptin sensitivity. If changes in leptin signaling can account for the findings, then we would have to speculate that leptin sensitivity was increased in the VLC and LGI diets relative to the LF diet.
How the diets could have led to different leptin sensitivity in the brain is unknown. One intriguing possibility is that the decline in triglycerides improved leptin sensitivity. Dr. Bill Banks's research suggests that high circulating triglycerides impair leptin sensitivity in rodents (25). Dr. Banks is now at the University of Washington and I've had the pleasure of interacting with him a few times about this.
Another possibility is that the VLC diet led to improved insulin sensitivity in the brain. Insulin acts in the brain in a manner similar to leptin-- it suppresses food intake and increases energy expenditure, opposing fat gain (although to a much lesser degree than leptin, and the effect has historically been fickle). The VLC diet did appear to improve insulin sensitivity in the liver more than the other diets, so it's possible that it had such an effect in the brain as well.
It's always tempting to put the cart before the horse, and come to conclusions before we really know what's going on, but in this case all we can do is speculate. In my opinion, protein-induced thermogenesis, metabolic inefficiency, and alterations in the brain are the most likely explanations for the findings in this study, but we'll have to wait for future studies to see if this is true, and if so, why.
* Such as Gary Taubes's op-ed article in the NYT today. It proclaimed triumphantly that "a new study suggests that carbs, not calories, lead to weight gain". Back in reality-land, the study had nothing to do with what causes weight gain-- it was about fat loss maintenance in people who are already overweight (which are not metabolically analogous to pre-obese people as Taubes claims). There are numerous studies that compare the ability of fat and carbohydrate to actually cause fat gain, but these are inconvenient because they invariably show that 1) calorie intake is the dominant factor, and 2) fat is equally or more fattening than carbohydrate at a given calorie intake-- for example the study that concluded: "Excess dietary fat leads to greater fat accumulation than does excess dietary carbohydrate..." (26). In Dr. Ludwig's paper and the accompanying editorial in JAMA, they made no claims about carbohydrate being more fattening than fat, and they didn't even broach the topic, because Dr. Ludwig's study is not relevant to the question. I would love to see Dr. Ludwig set the record straight on this, but unfortunately most researchers don't fight back when their results are abused in the popular media.