Think of this article as a Rorschach test. I’ll describe some data from a recent experiment; you decide for yourself what meaning to extract from it. This, as far as I can tell, is the only safe way to write about low-carb, high-fat (a.k.a. ketogenic) diets for endurance athletes, without getting bogged down in endless debates about your motives, word choice, and sanity. So here goes nothing!
The study in question, newly published in Medicine & Science in Sports & Exercise, comes from a group at the Sports Performance Research Institute New Zealand at Auckland University of Technology, led by doctoral student David Shaw. It’s not a perfect study, but it has some notable strengths compared to a lot of diet-related research: a randomized trial with strict dietary control, and a 31-day period to adjust to the unfamiliar diet. And it tackles a very simple but athletically important question: do you get faster on a ketogenic diet, which is designed to teach your body to rely almost exclusively on fats and ketones (a supplementary fuel that your body produces in the near-absence of carbohydrate)?
Some details: the study started with 10 trained male endurance athletes, all sub-3:30 marathoners running more than 30 miles per week, none with previous experience on a ketogenic diet. Two of them failed to stick to the study’s conditions, so the final analysis includes 8 subjects. All of the subjects, in randomized order, completed two 31-day protocols: one consuming their normal diets, the other on a ketogenic diet, with a series of performance and physiological tests before and after each 31-day block.
The normal diets for these particular people averaged 42.9 percent carbohydrate, 38.5 percent fat, and 18.6 percent protein. The prescribed targets for the ketogenic diet were less than 50 grams per day of carbohydrate, 15 to 20 percent of calories from protein, and 75 to 80 percent from fat. The subjects were given free coconut oil, olive oil, LCHF cereal, and discounted fruits and vegetables, and their dietary reports were carefully monitored. They ended up averaging 34 grams of carbohydrate per day (4.1 percent of their calories), with 77.7 percent of their calories from fat and 18.2 percent from protein. Regular blood and urine tests confirmed that the subjects were indeed consistently in ketosis.
There were two basic assessments. One was a progressively accelerating treadmill test, which allowed the researchers to measure VO2max at exhaustion, and also to assess efficiency at a range of different speeds as the treadmill accelerated. The other was a plain old run to exhaustion at pace equivalent to 70 percent of VO2max, which the subjects could maintain on average for about four hours (so, in other words, significantly slower than marathon pace, which is typically somewhere around 80 percent of VO2max). During the run to exhaustion, they either received a carbohydrate-based sports drink (during the normal diet trial) or an artificially sweetened drink with the same number of calories from coconut oil (during the ketogenic trial).
In the efficiency test, there was no difference between the diets at the lower speeds corresponding to below about 60 percent of VO2max. Once the pace picked up to above 70 percent of VO2max, however, the runners on the ketogenic diet became significantly less efficient: they needed more oxygen and more energy to sustain a given pace. Their VO2max itself—that is, the maximum amount of oxygen they could use per minute—stayed the same on both diets, but the speed they could run at while consuming that oxygen was lower on the ketogenic diet.
Interestingly, that echoes what Kieran Clarke, the co-developer of the ketone ester drink sold by HVMN, told me last year: “As soon as you’re up to 75 percent of your maximum workload,” she said, “I wouldn’t even go near a ketone.”
That makes the run to exhaustion extra-interesting, because it was at 70 percent of VO2max, right around the threshold where efficiency seems to start suffering. So here’s the Rorschach part of the article. On the left, you’ve got the before-and-after results for each of the eight subjects on their habitual diet (plus the average results with standard deviations shown); on the right, the same thing for the 31-day ketogenic diet.
So what’s the verdict? The statistical analysis tells us that time to exhaustion was similar in both conditions: the average times before and after the ketogenic diet were 239 and 219 minutes, respectively, with a p value of 0.36. Three of the subjects lasted longer; five gave up sooner. The variation is much higher than after the habitual diet: some seem to have thrived, others tanked.
The big question is how much we can or should read into those results. Would the three people who got better have shown similar results in another test a week later? A month later? How about the five people who got worse? Or is it just random scatter, since time-to-exhaustion tests are very sensitive to small perturbations in how you’re feeling?
The researchers offer one speculative answer to this question. You can divide the subjects into two groups based on their respiratory exchange ratio (RER) at the end of the VO2max test. The RER is the ratio between exhaled carbon dioxide and inhaled oxygen, and (with a few caveats) it tells you what mix of fat and carbohydrate you’re burning: a value of 0.7 corresponds to pure fat, 1.0 corresponds to pure carbohydrate, and greater than 1.0 suggests you’re going so hard that you can’t supply oxygen quickly enough and are adding a significant amount of anaerobic energy.
After 31 days of ketogenic diet, all the subjects had ramped up their fat-burning abilities (which is good), but had also lost some of their carbohydrate burning abilities (which is not good, particularly at near-maximal intensities where you’re consuming energy very rapidly). In their post-keto VO2max test, 4 of the subjects had a final RER above 1.0, suggesting that they could still access carbohydrate and anaerobic energy at a reasonable rate, while the other 4 had a final RER below 1.0. Those with the lower RER were the ones who subsequently struggled in the time-to-exhaustion test: their average time decreased significantly by 237 to 174 minutes, and they had higher lactate levels at the end of the test. In contrast, those with the higher RER had lower lactate levels and no significant change in time to exhaustion (increasing from 241 to 265 minutes).
Again, this is a highly speculative suggestion. The interpretation of RER at high intensities is problematic, and combing through post-hoc sub-groups of 4 people is a good way of finding patterns that don’t really exist. But it’s an idea to file away for future investigation: could a simple measurement of RER at VO2max give you a quick and reliable way of predicting who’s likely to perform well or poorly once they’ve adjusted to a ketogenic diet?
One final historical note: it’s interesting to see how many echoes there are here from Stephen Phinney’s 1983 paper on four weeks of ketogenic diet for cyclists, which has near-scriptural status in the ketogenic community. In Phinney’s study, fat-burning was ramped up but high-intensity power was throttled: he noted “a severe restriction on the ability of subjects to do anaerobic work.” In a time-to-exhaustion test, there was no significant change on average (147 to 151 minutes), but huge individual variations: one of the five subjects improved from 148 to 232 minutes, another decreased from 140 to 89 minutes. Shaw’s new results seem surprisingly similar.
To sum up the key points from the inkblot: at speeds faster than 70 percent of VO2max (i.e. well below marathon pace), efficiency was significantly impaired on a ketogenic diet. At speeds slower than 60 percent of VO2max, efficiency was unchanged. Right at 70 percent VO2max, time to exhaustion was unchanged on average—but the individual results in the graph above suggest the possibility of a more nuanced picture. You can decide which dot you think you’d be.
My new book, Endure: Mind, Body, and the Curiously Elastic Limits of Human Performance, with a foreword by Malcolm Gladwell, is now available. For more, join me on Twitter and Facebook, and sign up for the Sweat Science email newsletter.