© Sebastian
A sea change is underway. Ten years ago, it was heresy to propose that a ketogenic (a.k.a low carb high fat) diet was in any way healthier than the low fat high carb diet supported by public health authorities. It was branded a "gimmick" diet. In some places, doctors who prescribed it to their patients risked having their medical licenses revoked.

The German physicist Max Planck is often misquoted as having said that "science advances one funeral at a time". Well, the man who gave birth to the low fat high carb dogma, Ancel Keys, died in 2004. His first generation of acolytes have now joined him in oblivion. The men (they were with very few exceptions all men) who created the current dietary guidelines back in the late 70's and early 80's are also gone, after having presided over a massive explosion in the number of people suffering from obesity and type 2 diabetes.

The newer generations of nutrition researchers do not appear to be as wedded to the old dogma. This is visible in the increasing number of studies being published on a ketogenic diet. Some of these are even appearing in the most prestigious and conservative nutrition journals.

One such study was recently published in Advances in Nutrition, a journal owned by the American Society for Nutrition. It was a systematic review looking at randomized trials of a ketogenic diet as a treatment for Alzheimer's disease, which is the most common cause of dementia. We're going to get to that study in a minute, but first, a little detour.

There is some evidence to support the notion that dementia can in part be caused by a high carbohydrate diet. An observational study was published back in 2012 in The Journal of Alzheimer's Disease in which 937 elderly people were followed for four years. The median age at the start of the study was 80 years, and at the beginning, all the participants were asked to fill in a diet questionnaire and were also evaluated for cognitive function. Four years later, 200 of the 937 participants had developed some level of cognitive impairment.

When the researchers correlated this with dietary carbohydrate intake, they found that the quartile with the highest intake had an 89% increased relative risk of developing cognitive impairment during the four years of follow-up, as compared to the group with the lowest intake. And that's after adjusting for known confounders like gender, BMI, co-morbidities, and APOE4 status (APOE4 is a gene variant that is strongly associated with increased risk of Alzheimer disease). The difference was statistically significant (p-value 0.004).

The quartile with the highest fat intake, on the other hand, had a 56% decreased relative risk of cognitive impairment as compared to the quartile with the lowest fat intake (p-value 0.03).

Interestingly, the differences between the quartiles in terms of carbohydrate and fat consumption weren't actually that big. The highest quartile in terms of carbohydrate consumption was getting more than 58% of calories from carbohydrates, while the lowest quartile was getting less than 47%. Not a huge difference. The same was true for fat intake. The quartile with the highest fat intake was getting more than 35% of calories from fat, while the quartile with the lowest fat intake was getting less than 27%. This would seem to suggest that even relatively modest differences in consumption of carbohydrates and fats can have big effects on cognitive function over time, and that an even bigger reduction in relative carbohydrate intake might have achieved an even bigger reduction in risk of cognitive impairment.

Of course, this was an observational study, and although the results are suggestive, it can't prove the existence of a cause and effect relationship between carbohydrate/fat intake and dementia. The results could have been caused by residual confounders that the researchers were not able to adjust for. For proof of a cause-effect relationship you need randomized controlled trials. Which is where the recent systematic review published in Advances in Nutrition comes in. As mentioned earlier, it was looking at the randomized trials that exist of a ketogenic diet as a treatment for Alzheimer's disease and mild cognitive impairment.

Ten trials were identified, with a total of only 456 participants, which really shows how under-researched this area is. And things get worse. Only three of the trials, with a total of only 47 participants, were actually testing a ketogenic diet (i.e. a diet in which carbohydrates are restricted to the point where the body significantly increases production of ketone bodies). The rest were testing supplements containing medium chain triglycerides (MCT's), which the body preferentially converts to ketones. From my perspective, these are two very different interventions. A ketogenic diet has many different effects on our metabolism, and I am inclined to believe that the beneficial effects come primarily from the reduction in carbohydrates and insulin, not from the increase in ketones.

Taking an MCT containing supplement is obviously not the same thing as following a ketogenic diet. The seven studies of ketogenic supplements were, with only one exception, either funded by companies that sell supplements, or they failed to disclose their funding (which means they were probably funded by companies that sell supplements). Most of these studies were never registered at, and of the ones that were, this was done after the trials were already underway, which is highly suspect behaviour, because it means the researchers could know whether the trials were going well or not before they let the world know about them. In other words, it's possible they were simultaneously running other trials that weren't going so well, and that were therefore never posted on, which could lead to massive publication bias.

The three small studies of a ketogenic diet compared it with the traditionally recommended low fat high carb diet. One of the three ran for twelve weeks, while the other two ran for six weeks, so these were short term interventions. In terms of outcomes, there were improvements in some of the cognitive functions tested, but not in others. Overall, the results really don't tell us anything useful, as you would expect from tiny trials run for short periods of time.

The seven studies of MCT supplements appeared to show some benefit in terms of cognitive function in Alzheimer patients, although the fact that these were mostly industry funded studies, that weren't pre-registered at, makes the results hard to trust. Strangely, the systematic review only reports whether there was a "benefit" or not, but not what the size of the benefit was, or whether it was statistically or clinically significant. This feels like a rather weird omission in a systematic review. So I decided to look up the two biggest trials, with 152 and 131 participants respectively. According to the systematic review, the first showed an "improvement" in ADAS-Cog (a test of cognitive function used in Alzheimer's disease) and MMSE (a test for dementia), while the second showed an "improvement" in ADAS-Cog .

When we look at the first of these trials, we find that the difference between the group getting MCT and the placebo group at 104 days (the longest follow-up) was less than one point on the 70 point ADAS-Cog scale. One point on a 70 point scale is not a noticeable difference. Additionally, the difference wasn't statistically significant. In other words, there was no clinically meaningful or statistically significant difference between the groups on ADAS-Cog. If we move on to MMSE, we find no difference whatsoever between the groups. Yet this study is reported as being "positive" in the systematic review. Odd.

When I moved on and looked at the second of these trials, I immediately realized that it was just a duplicate report of the same study, with a few new analyzes of the same data set. Researchers often do this, to maximize the number of publications they can get out of one data set (since career success in research is largely determined by number of publications). How the authors of the systematic review didn't realize this is beyond me.

So basically, one negative study was reported as two positive studies in the systematic review. And these were the two "big" studies, supposedly representing 62% of the participants in the systematic review. All the other studies were much smaller.

This weirdness really makes me wonder about the motives of the authors of the systematic review. No conflicts of interest were reported, and they reported receiving no specific funding to carry out the review. But seriously, they went through all this data in detail and didn't realize that they were looking at the same data set twice! And then, to top it off, Advances in Nutrition, the fourth highest ranked nutrition journal, went ahead and published it, no questions asked!

This really speaks to the poor state of nutrition research more than anything else, and to the low added value provided by the process of peer review. If peer review was the rigorous process that the general public thinks it is, this nonsense would have been noticed and called out, and the article wouldn't have been published.

What can we conclude?

Although I am a strong proponent of a ketogenic diet as an effective therapy for metabolic syndrome, obesity and type 2 diabetes, and therefore think it's likely that it also has beneficial effects in terms of preventing or delaying dementia (which is far more common in people suffering from these diseases), the evidence that exists today cannot prove that that is the case. Nor does the current evidence support the use of MCT supplements as a way to treat or prevent dementia.