Critical Analysis of Wilson et al’s. 2017 Keto Study

Written by
: Adam Tzur
Acknowledgements: Brandon Roberts, Greg Nuckols, and Anthony Roberts
Article length: ~3000 words
Last updated: 17.04.2017


  • Wilson et al. have published a study where 25 trained lifters were divided into two groups: a keto group and a western diet group. Both groups did 9 weeks of strength training and results for hypertrophy, LBM, fat mass, etc. were collected.
  • The results were, at face value, quite extraordinary for the keto group.
  • There are also several issues with the study methodology and lacking reporting of conflicts of interest.
  • Scroll down to the results to see tables and figures

If you have any comments regarding this article, please send an e-mail to

Sign up for the newsletter and stay updated when new articles, videos, and e-books are published on SCI-FIT!


The study: facts and information

Title: "The Effects of Ketogenic Dieting on Body Composition, Strength, Power, and Hormonal Profiles in Resistance Training Males"

Here is a link to the study


“We hypothesized that the KD would decrease body fat to a greater extent than a WD group, while maintaining skeletal muscle hypertrophy, strength, and power.”



There is little data on the subjects. The authors state that 30 subjects joined, but 5 dropped out due to various reasons. Participants needed to have a 1.5x bodyweight squat if they wanted to be included in the study.

“(...) subjects had an average squat of 1.56 ± 0.14 times their body weight”


They used DXA for LBM and ultrasound for body composition measurements. I go into more detail on this further down in the article.

  • DXA: “The coefficient of variation (12) between body composition assessments was 1.5%.”
  • Ultrasound; “The coefficient of variation between muscle thickness assessments was 3.4%.”

Resistance training program

“All subjects were instructed to not perform any additional resistance training or endurance training throughout the duration of the study. For all criterion lifts, subjects were given a target repetition range based on a percentage of their 1RM”


“Following the two-week diet adaptation period, all participants performed a 7-week high volume resistance training protocol followed by a 2 week taper”


“Carbohydrates were reintroduced to the subjects in the KD treatment from week 10 to week 11”.

Statistical analyses

  • Data are presented as mean±SD.
  • Confidence intervals not shown.
  • 2-way ANOVA with repeated measures (time and group interactions).
  • Post-hoc with Tukey's adjustment
  • Data on LBM and fat mass were only presented in graphs, not as absolute numerical values


Body weight

Note that the keto group gained 5 kg of total body mass in one week (10->11). They consumed the same amount of calories as they did in previous weeks.

Fat mass

Note that the keto group had an average gain of 2 kg of fat mass in one week (10->11). They consumed the same amount of calories as they did in previous weeks.

Lean Body Mass (LBM)

LBM was not reported in absolute numbers, but as percentages. I had to look at the graphs and try to estimate absolute LBM values. Standard deviation is not reported, only displayed. The graph below is an estimation based on my reading of the graph! After I read the baseline LBM from the graph, I used the reported percent changes to estimate absolute LBM changes.

Muscle thickness

Bench press strength

Keto group added ~9 lbs to their bench press from week 1 to 10.

Keto group added ~14 lbs to their bench press from week 10 to 11.

WD added ~14.5 lbs to their bench press from week 1 to 10.

WD added ~2 lbs to their bench press from week 10 to 11.

Are the gains realistic?
Comparing results to previous research

Deficit gaining and glycogen loading

The data presented in this study suggests that both keto and WD groups gained muscle mass and strength in a caloric deficit. I’ve previously reviewed the deficit gaining literature, and concluded that it is possible to gain LBM and strength in a deficit. However, the keto group in this study lost 4 kg of fat mass while gaining ~1.5 kg LBM and increasing their muscle thickness. However, when your glycogen stores deplete (as they do with a very low carb diet), you will “lose” LBM.

From my article on detraining:

When muscle glycogen stores shrink during detraining, we “lose” muscle mass, or more accurately, the water contents of our muscles decreases (Costill et al., 1985). On the flipside, you could even “trick” body estimates by going on a high carb diet to glycogen load your muscles (Nygren et al., 2001; Rouillier et al., 2015; Bone et al., 2016). Eating low-carb or going on a cut can lower your glycogen and water stores. This affects “wet" muscle mass (Hulmi et al., 2016)

This is why we have to take the atrophy and hypertrophy studies with a grain of salt. When beginners start exercising, their muscle glycogen stores will grow quickly and they will retain more water, as shown below:

What 16 weeks of strength training does for total body water in beginners.
Figure by Ribero et al., 2014

It is very unlikely that trained lifters will have “glycogen/water gains” since these are some of the first adaptations the body makes.

Keto studies

[Update: we have now written an article series on the ketogenic diet]

Generally speaking, keto isn’t known for it’s ability to build mass quickly.

From Tinsley and Willoughby, 2016:

“Jabekk et al. (2010) reported no change in fat-free mass in subjects following a resistance training program in combination with a ketogenic diet, and Wood et al. (2012) reported a decrease.”

From Meirelles and Gomes, 2016:

“Twenty-one volunteers engaged in an eight-week progressive RT program three times per week were assigned to a CRD [< 30 g carbohydrate; n = 12] or a CONV [30% energy deficit n=9].“


“In conclusion, hypoenergetic diets combined with RT led to significant increases in muscle strength and were capable of maintaining muscle thicknesses in the upper and lower limbs of overweight and obese participants, regardless of the carbohydrate content of the diets.“

From Paoli et al., 2012:

“No significant differences were detected between VLCKD and WD in all strength tests. Significant differences were found in body weight and body composition: after VLCKD there was a decrease in body weight (from 69.6 ± 7.3 Kg to 68.0 ± 7.5 Kg) and fat mass (from 5.3 ± 1.3 Kg to 3.4 ± 0.8 Kg p < 0.001) with a non-significant increase in muscle mass.”

It is already hard to increase LBM in trained lifters, so when this study shows a LBM gain of 4.5 kg in the keto group after 7+2 weeks of RT, I’m a bit skeptical. Even the 1.5 kg LBM increase at week 10 is impressive. So are the gains for WD!

Remember, the keto group:

  1. was in a caloric deficit
  2. was on a ketogenic diet
  3. had trained lifters

All these factors restrict how much gains you can make.

Conflicts of interest

[Update: read our article on keto COIs]

I searched the study paper for “coi”, “interest”, “conflict”, “fund”, and “funding”. I found no mention of funding or conflicts or interests.

This alone does not mean that we should automatically discard the results of the study, as I’ve written about in this article! But, we should be aware of these COIs.

Shout-out to Anthony Roberts for letting me know about D'Agostino and Volek!

Problems and errors in the study

Lack of anthropometric data

At the start of studies, it is customary to place a table with data about height, weight, age, and so on. This is not done in this study. They only tell you the age of the participants in the results section, late in the paper. And their weight is only shown in the results table. Height is not reported.

Lack of absolute LBM and fat mass values

LBM/fat mass is not reported in absolute numbers, but as percent changes. I had to look at the graphs and try to estimate absolute LBM/FM values. This could have been put in Table 3 with the rest of the data.

How long was the resistance training period?

  • Claim #1: “both groups participated in an 8 week supervised, periodized resistance training.”
  • Claim #2: “all participants performed a 7-week high volume resistance training protocol followed by a 2 week taper

Was it 8 or 7+2 weeks? According to the volume loads in Figure 2 in the study, it was 7 weeks of high volume and 2 weeks of tapering.

Baseline strength, inclusion criteria, and standard deviations

“25 subjects had an average squat of 1.56 ± 0.14 times their body weight''

The numbers here are a little strange. The standard deviation is small. Furthermore, the SD goes below the minimum required level for squat strength. Per the inclusion criteria, it is at least 1.5x BW. Given a normal distribution, several subjects would fall below 1,5x. Thanks to Jorn Trommelen for pointing this out. Thanks to Greg Nuckols for running the statistical test to check the probabilities.

"After a visual inspection of boxplots in order to identify any outliers, a normality test (i.e. Shapiro Wilk) confirmed the normality of the data"

Mean age and training experience

The participants were this old, in years: “KD = 23.5 ± 4.5 vs. WD = 21.3 ± 3.7”

The participants also had an average of 5.5 years resistance training experience.

This suggests that, on average, the subjects lifted weights since they were ~18 years. This is kind of impressive, given everything that happens during these transition years. It’s not impossible, but it is impressive that the researchers originally found so many people at this age with this type of experience.


In the following sentence about DXA, Wilson et al., link to a study:

“The coefficient of variation (12) between body composition assessments was 1.5%.”

Citation (12) is entitled: “Effect of a Geriatric Consultation Team on Functional Status of Elderly Hospitalized Patients”.

I haven’t been able to figure out how it connects to DXA CV. Looks like they just linked a random study.

Weird data reporting

  • No CI
  • Graphs are poorly labeled
  • No absolute values reported for LBM and fat mass

25 or 26 participants?

A poster-presentation from the study was published in 2014. In the presentation, it is stated that 26 participants were part of the study. However, the 2017 publication states that 25 participants were in the final analysis.

Thanks to Philip Bloem and Jorn Trommelen for pointing this out!

Bias in abstract

The abstract is formulated in such a way that it selectively interprets the data and favors keto.

Can we trust the results?

Accuracy of DXA and ultrasound

I’m not going to go into a whole discussion of the accuracy of these body comp measurement tools. Sufficient to say, all measurement tools (even MRI) are affected by carb intake. Muscle glycogen loading and depletion affect the results (Nygren et al., 2001; Rouillier et al., 2015; Bone et al., 2016).

If you want to learn more about glycogen and muscle mass, read:
The Science of Detraining: How long you can take a break from the gym before you lose muscle mass, strength, and endurance

You should also check out James Krieger's articles on DXA, BIA, etc. Here's one of his articles from weightology: Cheat Your Body Fat Test

Large LBM and muscle thickness gains

As discussed under “Are the gains realistic?”, lifters with an average of 5.5 years of lifting experience generally don’t make quick gains in muscle or strength. Especially if they are on a low-carb diet (low glycogen) and are eating at a deficit. Hence, we are likely looking at changes in muscle hydration.

Reputation and COIs

Wilson et al., have a reputation among some experts for publishing results that are extraordinary and unusual. And they don’t disclose COI.

Check out this letter to the editor by Stuart M. Phillips, Alan A. Aragon, Shawn M. Arent, Ben Esgro, D. Lee Hamilton, Eric R. Helms, Menno Henselmans, Jeremy P. Loenneke, Kevser Ermin, Layne Norton, Michael Ormsbee, Brad J. Schoenfeld, Matthew Vukovich, Colin Wilborn, and Darryn Willoughby.

Should we reject the paper?

First, I want to state that we should reject or accept a paper based on its quality. This means we must read beyond the abstract and conflicts of interest. We must dig into the meat and bones of the paper. The more errors and strange data, the more likely it is that something is off.

My conclusion is that we should reject the paper, for the following reasons:

  • There are too many errors and problems with the study, the data, the reporting, interpretation…
  • The researchers do not follow the conventions of the literature
  • Study design favors the keto group from week 10-11
  • The abstract only reports some results and it is biased towards the keto group
  • Potential conflicts of interest which are not disclosed
  • Authors' reputation and history

Is there anything we can conclude from this study?

If we ignore the results from weeks 10 to 11, the study suggests that ketogenic diets might be worse for LBM and strength.


It should be said that I’ve been very nit-picky in my analysis of this study. Very few studies are so good that they would be able to survive such scrutiny. In fact, it is said that we should be somewhat lenient when criticizing studies, because researchers are often constrained by financial resources and other limitations.

Furthermore, it can be argued that the carb refeed was necessary so that both groups had equal amounts of glycogen when the testing was done. This would give us fairer results, than comparing glycogen depleted subjects to subjects with normal glycogen levels.

'However, there are some issues.

  1. A one-week carb refeed might lead to glycogen supercompensation and excessive water retention. DXA is highly sensitive to changes in water weight, and it is generally not recommended to make large dietary changes right before a measurement. To fix this, the study would have needed either a 4C measurement analysis that could account for water weight, or the carb refeed would have had to occur over a longer period of time (i.e. 2-3 weeks). Since they didn't use 4C to check water weight, I don't think they should have done a refeed. This is because LBM and muscle thickness measurements are strongly confounded.
  2. Is the refeed truly justified in terms of external validity / ecological validity? From the abstract, it is clear that the authors are making claims about external validity (the usefulness of keto for lifters). If they just did carb refeed to check how DXA and ultrasound measurements are affected, then it wouldn't be much of an issue. There is, however, a problem when they interpret the findings the way they do.