Study reviewed: K.D. Hall, K.Y. Chen, J. Guo, Y.Y. Lam, R.L. Leibel, L.ES. Mayer, M.L. Reitman, M. Rosenbaum, S.R. Smith, B.T. Walsh, and E. Ravussin; Energy expenditure and body composition changes after an isocaloric ketogenic diet in overweight and obese men (2016)

Lawrence Judd

Ah, the carbohydrate-insulin model of obesity.

Like a jilted lover who won’t stop calling, an overbearing mother-in-law or your local MLM sales rep, you really want it to go away… but it lingers. Clinging to life like the child’s blanket that desperately needs to be binned, it lurks in corners of the Internet rarely frequented by those of an evidence-based persuasion for fear of dogma, ad hominem insults and “well, it works for me.”

(I promise, I’m done with the similes now.)

This hypothesis essentially claims that calories are NOT the most important factor in fat gain; rather, that a higher proportion of dietary carbohydrate will promote insulin secretion to the extent that there will be a net storage of fat. Conversely, when carbohydrates are lowered, the lower circulating insulin levels will allow for greater fat oxidation and a net reduction in body fat – independently of calorie intake. Lower carbohydrate diets have been proposed to have a 300-600 kcal per day “metabolic advantage” over higher carbohydrate intakes, due to increased energy expenditure.

This is a testable hypothesis – have study subjects eat a maintenance diet, with a normal amount of carbohydrates, and then switch to a diet with the same amount of calories but a very low amount of dietary carbohydrates. You could then see what happens to energy expenditure, weight and body composition, and the results from this could effectively put the insulin hypothesis to bed once and for all.

However, it’s been very difficult to test this in the past – a study like this would require:

  • Exceptional dietary adherence, which is very rare in most studies.
  • An accurate way (or ways) of quantifying energy expenditure. This is very time-consuming and effort-intensive, and as a result is very expensive.

That’s exactly what Kevin Hall’s research group at the NIDDK were able to do, and that’s what we’ll be looking at today. Before we get into dissecting the study itself, though, I want to give a brief low-down on the science behind how researchers can circumvent the above issues.

We’re about to turn the world of Personal Trainer education upside down.

We can’t say too much just yet – but when we do, we’ll tell our email list first. Pop your email address in the box to hear about what moves we’re making next.

Metabolic Wards And Chambers

Vital to the detailed study of the effects of different dietary protocols is the use of metabolic wards and chambers. Metabolic ward is the name given to a portion of a research facility in which subjects are confined for the duration of a study, in order for all of their meals to be provided to them and in order to completely restrict the consumption of non-experimental food. The food is usually prepared on site in a metabolic kitchen, which allows for the ultra-precise provision of various calorie and macronutrient intakes.

Did you know – The first recorded example of a “metabolic ward” was in 1747 by a gentleman called James Lind, who was investigating how best to treat sailors with scurvy?

A metabolic chamber is something a little different. A metabolic chamber is a completely sealed room into which a subject can be placed, and measures the amount of oxygen and carbon dioxide flowing in and out of the room. The energy expending processes in the body produce carbon dioxide as a waste product, which we then breathe out – so by measuring the amount of carbon dioxide exiting the chamber (and performing some calculations), we can effectively calculate just how much energy someone is expending over the course of a period of time.

Keto 1
An open-access metabolic chamber built in South Africa. Source:

Pretty useful, huh?

Doubly Labelled Water

Doubly Labelled Water (DLW) is another method of measuring energy expenditure, which takes advantage of a technique called isotope tracing. In doubly labelled water, some of the hydrogen and oxygen atoms are replaced with “heavy” oxygen and hydrogen atoms, or isotopes. These isotopes have a heavier nucleus than regular oxygen and hydrogen, and so can be separated out via a process called mass spectrometry.

The heavy oxygen can leave the body in two ways – through bodily fluids, and through exhaled carbon dioxide. However, the heavy hydrogen (also called deuterium) can only leave via bodily fluids. We can measure the amount of deuterium in urine, which we can then use to figure out how much heavy oxygen has left via urine. By comparing this to the amount of heavy oxygen that was present after the dose of DLW was consumed, we can figure out how much left the body via CO2 – this can then (similarly to in a metabolic chamber) be used to estimate energy expenditure.

The great thing about DLW is that it’s genuinely non-invasive, isn’t dangerous, and can be used over periods of 10-20 days without confining the subject to a metabolic ward.

Long story short – both of these provide highly accurate estimates of energy expenditure by measuring the amount of carbon dioxide excreted by the body.   

Now we’re all on the same page…

The Study

The following table gives a breakdown of the methodology and the outcomes that Hall et al assessed. Give it a read, and then I’ll give my comments on the methodology.

Study Subject Number N=17
Subject Age 33 ± 1.8 years
Subject Starting Weight and Body Composition Weight: 87.4 ± 3.7 kg

BMI: 28.8 ± 0.8

Body Fat % (Measured by DEXA): 28.9% ± 1.1%


Other relevant subject details ·       Subjects were not allowed to participate if their weight wasn’t stable (i.e. had changed more than 5% over the past 6 months)

·       Reduced-obese individuals were not allowed to take part – the cut off point for this was if they were less than 92% of their lifetime maximum weight

·       If subjects typically ate a very low or very high carbohydrate diet (<30% or >65%) then they weren’t permitted to take part

Experimental Design ·       Total of 8 weeks in a metabolic ward

·       Total of 8 x 2-day visits to a metabolic chamber to accurately measure energy expenditure

·       2 doses of Doubly-Labelled Water – 1 at 2 weeks in, and the other at 2 weeks from the end of the study – to provide another measure for energy expenditure

·       Weeks 1-4, subjects consumed a baseline diet (2398 kcal, 91 g protein, 300 g carbohydrate, 93 g fat, 26 g fibre). This was adjusted up until day 15 to try and match energy expenditure, and then clamped.

·       Weeks 5-8, subjects consumed an isocaloric ketogenic diet (2394 kcal, 91 g protein, 31 g carbohydrate, 212 g fat, 12 g fibre)

Exercise Intervention ·       90 minutes of low intensity cycling was done per day
Outcomes Assessed ·        Overall physical activity measured via a hip-mounted accelerometer (this was measured both inside and outside of the metabolic chamber)

·       Energy expenditure whilst in the metabolic chamber (EEchamber)

·       Sleeping energy expenditure whilst in the metabolic chamber (SEE)

·       Energy expenditure via doubly labelled water (EEDLW)

·       Respiratory Quotient  (RQ)

·       Body weight

·       Body fat % via DEXA

·       Blood and urine measurements: C-peptide (a measure of insulin), ketones, nitrogen, urea, ammonia, creatinine, adrenaline, norepinephrine, glucose, triglycerides


Comments On The Methodology

This is an unbelievably well-controlled study. The researchers:

  • Looked to quantify energy expenditure as much as possible – non-invasively via a hip-mounted pedometer, via doubly labelled water and via regular visits to a metabolic chamber. This not only provided multiple measures but allowed researchers to pinpoint where extra energy expenditure might occur – whether it was due to increased activity or not.
  • Kept protein controlled (a really common flaw in a lot of high CHO vs. low CHO studies).
  • Ensured dietary adherence via the use of a metabolic ward.
  • Actually compared “high” and “low” carbohydrate intakes – the baseline diet contained ~50% of calories from carbohydrates, whereas the ketogenic diet had ~6% of calories from carbohydrates.
  • Kept protein at ~1g/kg – whilst many of you might be looking at this and thinking that it seems really low, don’t forget the habitual diet of these individuals. A sudden increase in protein might well have caused changes in body composition during the baseline diet, which would have artificially made the baseline diet look far better than the ketogenic diet, as well as potentially prolonging the amount of time taken for the body to reach ketosis due to increased gluconeogenesis.
  • Accounted for any metabolic adaptation effects by excluding reduced-obese individuals from the study.
  • Body composition was assessed via multiple DEXA scans… big tick.

Another little detail from the manuscript that I’d like to highlight – the authors fully disclosed how they dealt with statistically anomalous results, but also revealed why they were outliers (there were two very interesting results… more on them later). This is very useful for us as Personal Trainers, as it not only helps to put the published results into context but also helps to quantify the extent to which individual response can vary. After all – we deal with individuals, not average.


Remember, the insulin hypothesis states that a lower carbohydrate diet will:

  • Increase fat loss because of lowered insulin.
  • Have a 300-600 kcal per day “metabolic advantage”, which you would expect to lead to a greater rate of weight loss.

So… how did the hypothesis fare against the experimental results?

Body Composition and Energy Balance

The figure to the right shows how body weight, body fat % and the calculated energy balance changed between the baseline diet and ketogenic diet conditions (the ketogenic diet started at day 0).Keto 2

We can see that the rate of weight loss was greatly accelerated for the first week, but then slowed right down for the remaining 3 weeks. This weight drop was largely caused by loss of water weight, not body fat; if you look at graph B, you can see that the rate of body fat loss actually slowed with the introduction of the ketogenic diet – completely the opposite effect of that predicted by the insulin hypothesis.

The final graph shows the energy balance as calculated using both body composition changes (i.e. using changes in fat/muscle mass to approximate energy balance via metabolizable energies) and doubly labelled water. Whilst these might look different, the error bars show that they’re not significantly different from each other – the calorie deficit was pretty much identical in both conditions. This played out in the weight loss, too: during the last 2 weeks of the baseline diet, the subjects lost a little under 1 kg of weight on average (~0.5 kg per week). During the ketogenic phase, the subjects lost on average 2.2 kg of bodyweight (give or take 0.3 kg). Divide 2.2 by 4 weeks, and you get – you guessed it – ~0.5 kg per week.

The subjects also lost half a kilo of body fat over the entire 4-week ketogenic dieting period – exactly the same amount as lost during the last 2 weeks of the baseline diet phase.

One outlier that was mentioned lost over 2 kg of body fat within the 2 weeks leading up to the introduction of the ketogenic diet, despite only losing 0.5 kg of weight.  He was excluded from the statistical analysis. Putting on 1.5 kg of muscle in 2 weeks whilst losing weight seems a little incredible, but don’t forget that these were completely untrained, overweight individuals – he may have been a hyper-responder to exercise.

It’s not looking good for the insulin hypothesis so far – rate of weight loss stayed the same, and fat loss actually slowed during the ketogenic phase.

Energy Expenditure

The table below shows all the data that was collected regarding energy intake and energy expenditure.

There are a lot of acronyms here, so let’s translate:

  • The first part of the table gives you the breakdown of the actual energy and macronutrient intakes for each dieting period.
  • EEchamber is the daily energy expenditure whilst the subjects were in the metabolic chamber. This was very slightly lower than calculated energy intake during both periods, which means the subjects should actually have gained a very small amount of weight. However, the bodyweight changes and doubly labeled water tells us that the subjects were, in fact, in a calorie deficit. We can see that there was roughly a ~60 kcal per day increase in the ketogenic phase – hardly the 300-600 kcal advantage that has previously been predicted.
  • SEE is the sleeping energy expenditure – we can see that this went up by ~90 kcal per day in the ketogenic diet phase.
  • RQ stands for respiratory quotient, which is a measure of which substrates the body is predominantly using for fuel. When pure carbohydrate is being oxidized, RQ = 1, and when pure fat is being oxidized, RQ = 0.7. RQ dropped significantly towards 0.7 during the ketogenic diet, showing that more fat was being oxidized.
  • The next 4 rows in the table give the various energy expenditures when adjusted for both bodyweight and for body composition of the various subjects. This is important as smaller or greater individual changes in body composition and weight can influence the average energy expenditure, which they clearly did – the difference in energy expenditure goes up to ~90-100 kcal per day.
  • EEexercise is the rate of calorie expenditure during the 90 minutes of low intensity cycling in both conditions, which wasn’t significantly different between conditions. Over the 90 minutes, this amounted to an ~8 kcal difference.
  • SPA refers to Spontaneous Physical Activity (effectively, NEAT). We can see that the calorie expenditure per minute dropped ever so slightly in the ketogenic condition, but again – not enough to be statistically significant.
  • AFT is the Awake and Fed Thermogenesis – basically, the combined thermic effect of the food eaten, plus the energy expenditure incurred by being awake. This wasn’t different between the baseline and ketogenic conditions.
  • EEsedentary is the energy expended whilst not moving or exercising – this significantly increased during the ketogenic condition.
  • EEDLW (energy expenditure via doubly labelled water) showed a ~150 kcal increase between the baseline and ketogenic diets. This increase above and beyond the difference calculated using the chamber values is likely a result of increased physical activity outside of the chamber, which can be confirmed by looking at the difference between the PAE (physical activity expenditure) values for each.

Keto 3

Does this tell the full story, though? Not quite – what’s also interesting is if we look at the time course of the EEchamber and SEE over the 4 weeks, as opposed simply to averages.

The graph to the right shows that change over time. During the first week of the ketogenic dKeto 4iet, overall energy expenditure went up by ~100 kcal as shown by the averages, but SEE went up by ~200 kcal. Both of those then declined steadily over the remaining 3 weeks. This shows that any metabolic “advantage” is virtually eliminated after a month of a ketogenic diet.
The authors explain that this is likely a result of the following series of occurrences:

  • Increased ketone production (which was observed in the blood samples), which causes an increase in energy expenditure – the biochemical pathways required to synthesise ketones are a little more “energetically expensive” than the ones required for carbohydrate metabolism.
  • Increased gluconeogenesis and triglyceride synthesis during the first period as the brain still attempts to run on glucose. After this, the brain switches predominantly to ketone use, and the energy expenditure increase from gluconeogenesis drops off.
  • The researchers observed decreases in leptin, insulin, catecholamines and thyroid hormone – over time, this will all lead to a drop in energy expenditure; all of these hormones help to influence total body energy expenditure, as their levels are somewhat symbolic of the “energy status” of the body. This might help explain why the increase in energy expenditure drops off as the ketogenic period goes on.

Conclusions and Practical Recommendations

There we have it – the two main claims of the carbohydrate-insulin hypothesis, refuted by an incredibly well-controlled study.

However, I hate leaving you without any form of practical recommendations. I’ll say this now – this section is a little more “speculative” than the rest of the article. Quite a lot of this is simply my opinion; feel free to take it or leave it as you wish.

I’m going to give you these as a FAQ section, for 2 reasons. 1 being that I’m bored of bullet points (as I’m sure you are too), and the second being that it gives you ready-made answers that you can use in Internet debates if that’s a pastime of yours.

Is a low-carbohydrate diet necessary to lose fat?

No, not at all. In this study, the subjects actually lost the same amount of fat in a shorter time period whilst eating a much higher quantity of carbohydrates. Not only that, but fat loss was demonstrably slower during the first 2 weeks of the ketogenic diet.

Now, that’s not to say that this would happen if carbohydrates were lowered but fat intake wasn’t raised to the same degree, however, the data here simply show that a low-carbohydrate diet isn’t any better than a higher carbohydrate one.


Is a low-carbohydrate diet better for preserving muscle mass?

At this point, I’m going to rely less on the weight and body composition data presented here. This is largely due to the paucity of protein in both diets relative to what most of you reading would be consuming in order to preserve muscle mass while dieting, which would likely fall at around the 25% of total calories mark compared with 16% in this study.

However, what we do know is that insulin secretion is protein-sparing. A lower carbohydrate intake, as demonstrated in this study, will significantly reduce insulin levels. Is this likely to lead to greater muscle loss than a higher carbohydrate intake, even when protein intake is around the 2 g/kg mark or higher? I’m not sure. The only time I would consider this a valid concern is if you’re preparing for a bodybuilding show – I would speculate that in most other circumstances, the differences would be small.

Another thing to consider is how a lower carbohydrate intake affects your training volume. Anecdotally, some people report that their training isn’t affected by a ketogenic diet, whereas some others find that their ability to accumulate volume is massively reduced to a reduced work capacity. If this is the case, then arguably a ketogenic diet will be inferior for preserving muscle mass for that individual.


What about Cyclic Ketogenic Diets?

Cyclic ketogenic diets (CKD) are generally set up with periods of a combined calorie deficit with ketosis, followed by low fat, high carbohydrate “refeeding” periods. Typically, the period of ketosis is 4-5 days long followed by a 2-3 day refeed.

The issue with a dietary set up with this is that we can now see that the first week of a ketogenic diet is when the least fat loss actually occurs. The most weight loss occurs, sure – however, the reality is that most people are more bothered about body composition than weight.

Also, from a practical perspective – with a CKD approach, it’s going to be incredibly difficult to track progress. Weight will be yoyoing, your appearance will be yoyoing due to the rapid changes in water and glycogen… it’s probably not a particularly sensible approach unless you’re not that bothered and just like eating like that!

One way to potentially avoid the issue of slower body fat loss during that first week is to not adopt a ketogenic approach, and slash both fat and carbohydrates (similar to Lyle McDonald’s UD 2.0 approach). However, this is an even more extreme approach than a standard CKD, and not one I would recommend in 99.9% of circumstances due to the absolutely meticulous planning required and the levels of adherence needed.

Hat tip here has to go to Aadam Ali of Physiqonomics, who has written a book about his experiences with various forms of ketogenic diets. You can read the first chapter for free on his website, and then you should definitely spare the $3 to buy the 9,000-word book because 1) it’s great, I read the whole thing and loved it and 2) Aadam, like you, also needs to eat.

So… how should I determine my carbohydrate intake?

The above points have all been made with two relatively extreme ends of the carbohydrate intake spectrum, as ketogenic diets require exceptionally low carbohydrate intakes. With more moderate reductions in carbohydrate intake, the differences are likely to be a lot less noticeable. The body of evidence comparing various carbohydrate intakes would suggest that socioeconomic factors aside, the best determinants of your carbohydrate intake – or rather, your carbohydrate:fat ratio –  are likely to be:

  • Personal preference
  • Activity levels and type
  • Protein intake (which is dependent on both personal preference and diet phase)


We’re about to turn the world of Personal Trainer education upside down.

We can’t say too much just yet – but when we do, we’ll tell our email list first. Pop your email address in the box to hear about what moves we’re making next.