The Keto-Med Study Published in the American Journal of Clinical Nutrition

Our new Keto-Med Study, comparing two popular low-carbohydrate diets, Ketogenic and Mediterranean, to determine what constitutes an optimal diet in individuals with prediabetes and type 2 diabetes mellitus (T2DM), has been published in the American Journal of Clinical Nutrition. The randomized

Quicklinks

crossover trial compared the two diets with 3 key similarities (incorporating nonstarchy vegetables and avoiding added sugars and refined grains) and 3 key differences (incorporating compared with avoiding legumes, fruits, and whole, intact grains) for their effects on glucose control and cardiometabolic risk factors in individuals with prediabetes and T2DM.

We recruited 40 participants with Type 2 diabetes or prediabetes for our crossover study. Over a 12 week period half of the participants ate Ketogenic meals while the other half ate Mediterranean meals. For the first 4 weeks of each diet the participants were provided ready-to-eat Ketogenic or Mediterranean meals provided by the food delivery service Methodology after which they were responsible for choosing and preparing their own meals. At the end of the 12 weeks, the participants swapped diets for an additional 12 weeks. This crossover allowed participants to act as their own controls.

Both diets received high marks in controlling blood glucose levels and aiding weight loss, but the Ketogenic diet was lower in several nutrients, particularly fiber, and was more difficult for study participants to follow in the long run.

In the end, there was no additional overall health benefits to cutting out legumes, fruits, and whole grains to achieve an ultra-low-carb Ketogenic diet. For people with diabetes or prediabetes, the less restrictive Mediterranean diet from Methodology was similarly effective in controlling glucose and likely more maintainable long term. "Restricting added sugars and refined grains and emphasizing the inclusion of vegetables should be the focus," Gardner said. "There’s no reason to restrict heart-healthy, quality carbohydrate foods above and beyond."

Visit the Keto-Med Study page

Read the AJCN article

Read the Stanford press release

Visit Methodology

Keto-Med Press Release Notes

Controversy and confusion vs. agreement

While many people tend to consider the world of nutrition and diet advice to be plagued with controversy and confusion, Christopher Gardner, PhD, Professor of Medicine and Director of Nutrition Studies at the Stanford Prevention Research Center, finds strong consensus and general agreement on most topics.
He is currently the chair of the Nutrition Committee of the American Heart Association, and in 2019 was a member of a panel of scientists that updated nutrition therapy guidance for the American Diabetes Association.
Over the past 29 years at Stanford, he has conducted more than 20 human nutrition intervention trials involving more than 2,000 study participants, examining the health impacts of nutrients like omega-3 fats and antioxidants, foods like garlic and soy, or diet patterns like vegetarian, Low-Fat or Low-Carb.
The latest comparison of the Gardner research group was between a Ketogenic and a Mediterranean diet among 40 adults with either pre-diabetes or type 2 diabetes (T2D), with the results being published this week in the American Journal of Clinical Nutrition.

Ketogenic – Very Low Carbohydrate (and high fat; not high meat or high protein)

The "classic" ketogenic diet is a high-fat, low-carbohydrate diet determined 100 years ago to help control seizures in some people with epilepsy.
The popularity of the ketogenic diet in the past five years has increased dramatically in the general public, repurposed for potentially addressing weight loss and blood sugar control.
The central tenet of this diet pattern is the extreme restriction of carbohydrates.
The goal is generally to lower intake to no more than 20-50 grams of carbohydrate per day, with the qualifier that this be “net carbs”, which suggests that fiber intake, a type of carbohydrate, need not be restricted.
For context, typical carbohydrate intake across most of the planet, including a wide range of cultural differences and agricultural practices, is usually in the 45%-65% range; let’s call it 50% for a reasonable round number.
For an adult consuming a typical range of 2,000-3,000 calories/day, that would be 250-375 grams of carbohydrate/day.
So, getting to 20-50 grams/day would mean eliminating 80% or more of usual carbohydrate intake for most people – that is substantial, and challenging.
The main sources of dietary carbohydrates are broad – vegetables, fruits, legumes, grains and dairy (the lactose in the dairy).
In order to limit carbs in accordance with guidelines for the ketogenic diet, this means eliminating all grains, all legumes (beans, pulses, chickpeas, lentils), most fruits (berries are an exception, and are often included in keto), and all starchy vegetables (sometimes referred to as the “below ground” list that include potatoes, carrots, yams). Some full-fat dairy is allowed (e.g., cream).
What remains acceptable in the carbohydrate-rich food groups is “above ground” vegetables, for example: leafy greens, bell peppers, and the brassica family (broccoli, cauliflower, Brussels sprouts, cabbage, kale).
Gardner says that many people trying to follow a Ketogenic diet assume that Low-Carb means High-Protein, which may sound intuitive because animal sources of food (beef, pork, poultry, fish) do not have carbohydrates, their calories come entirely from protein and fat.
But in fact, a Well Formulated Ketogenic Diet, as perhaps best described in a book on Low-Carb diets by Jeff Volek and Steve Phinney, is described as being primarily high in fat (70% or more of daily calories), and only moderate in protein (~20% of calories).
Being that high in fat in general, a ketogenic diet also tends to be high in saturated fat.
Given that the US Dietary Guidelines for Americans and the recommendations of the World Health Organization, the American Heart Association and national/international health organizations promote the health benefits of fruits, legumes and grains, and recommend limiting saturated fat intake, the ketogenic diet strikes most health professionals as inconsistent with current guidance.
Gardner looks for areas of agreement, rather than disagreement.

Three areas of agreement, three areas of differences

He cites a 2020 analysis of dietary trends in the US for the past 20 years, that indicates approximately 50-55% of daily calories in the US coming from carbohydrates, with ~80% of those coming from low quality carbohydrates, such as added sugar and refined grains (i.e., soda, candies, desserts, snacks, and white bread in bagels, pastries, pizza crust, etc.).
In the context of the type of carbohydrate, Gardner suggests that both ketogenic proponents and other nutrition health professionals are in strong agreement about restricting intakes of low quality carbs.
Many recent studies of the ketogenic diet have set up the trial design as a comparison of Low-Carb (Keto) vs. various versions of Low-Fat (e.g., Dietary Guidelines for Americans, American Heart Association).
A challenge in those studies is that sugar sweetened beverages and refined carbohydrates are “Low-Fat”, but are not considered healthy Low-Fat options.
To address this, what Gardner and colleagues addressed in the current study was a comparison of two Low-Carb diets that differed in “how low in carb”.
The comparison diet in the study was the Mediterranean diet, which includes olive oil, fatty fish, nuts, seeds, and is in general a relatively higher fat, lower carb diet.
The two study diets were designed to be the same in three priorities of agreement – 1) high in vegetable intake (particularly above ground vegetables), and eliminate or strongly limit 2) added sugars and 3) refined grains.
The primary differences in the two dietary approaches involved legumes (beans and more), fruits, and grains. For the Mediterranean Diet, participants were encouraged to include all of these (with an emphasis in the grain category on whole intact grains, like whole oats, whole barley, whole wheat kernels), while for the Ketogenic Diet, participants were instructed to avoid all of these (with the exception of occasional intake of berries).
Therefore, by design, the study was set up to address the following question: “Beyond the consensus agreement to consume a high level of vegetable intake, and eliminate or strongly restrict added sugars and refined grains, is there an additional health benefit to be experienced from also eliminating legumes (beans), fruits and whole grains?”
They hypothesized that most of the benefit would come from the first three areas of agreement, and that beyond that, the further elimination of legumes, fruits and whole grains would be more problematic than beneficial.

Study Design Specifics

Given the current dual epidemics of obesity and diabetes, one area of health promotion and disease prevention/management that is particularly relevant to the ketogenic diet is blood sugar.
According to the CDC-P (Centers for Disease Control and Prevention), more than 37 million Americans have diabetes, and 90-95% of that is Type 2 Diabetes (T2D), or ~1/10 Americans.
The hallmark of diabetes is problems managing blood glucose levels, particular high levels (hyperglycemia).
One way to keep down blood levels of glucose is to limit intake of high-sugar foods.
A standard clinical measure of blood glucose control is called glycosylated hemoglobin (or glycated hemoglobin, aka HbA1c).
The duration of the trial phases was 12 weeks, because HbA1c reflects glucose control over this time frame.
The study population selected for this nutrition trial was adults with either T2D, or prediabetes, and the primary outcome being examined in the study was HbA1c.
A cross-over design was used so that all participants would try both diets, in random order, and therefore each participant would be their own control (compared to themselves on the two diet phases).
In addition to the primary study outcome of HbA1c, the investigators examined other relevant clinical measures as secondary measures: fasting lipid profile (LDL- and HDL-cholesterol, triglycerides), fasting blood glucose and insulin, a liver enzyme (ALT – alanine amino transferase), and body weight.
Exploratory measures in the study included using continuous glucose monitors, assessing adherence to the study diets as prescribed, assessing satisfaction with the study diet across multiple domains, and a characterization of what the study participants were eating on their own, 12 weeks after they had tried both diets and the official intervention was completed (i.e., longer term impact on habitual dietary intake).
Altogether this was 12 weeks on one of the diet types, 12 weeks on the other (in random order) and 12 weeks of follow-up after the first two phases were completed) = 36 weeks.
For both the Mediterranean and Ketogenic phases of the study, the participants had foods delivered to them by a delivery service (Methodology) for the first four of the 12 week diet phases.
For the latter 8 weeks of each 12 week phase, participants were instructed by health educators how to follow the diets on their own, purchasing and preparing their own foods, or ordering foods consumed outside the home that would be consistent with study guidelines.
Extensive dietary assessment was conducted to address diet "adherence" and to determine overall nutrient intakes while on the two diets.

Results – “Process evaluation” (likely too much detail, but included for transparency)

40 participants were enrolled, 21 with prediabetes and 19 with T2D
33 of them completed both phases of the crossover study and all data collections (Note: The study was ongoing when the pandemic hit. 4 participants dropped due to concerns with the pandemic, 1 dropped for personal reasons. 2 participants that completed both phases were unable to contribute blood samples at the end of the second phase [pandemic-related], but provided all other data [n=35 total]).
Due to the pandemic, some of the participants were on one of the two diets for up to 24 weeks rather than just 12 as they waited for our research clinics to open back up to accepting participants for blood sampling (some extended by 2 weeks, some by 4, 6, 8, 10 or 12).
In general, the qualitative aspects of the two diets were similar for both diet phases (acceptability of food delivery menus, cost, enjoyment, ease, etc.) – good for internal validity.
A completely separate paper on "Adherence" was published in 2021 – the main findings from this paper were:
- On a scale of 0-10 (10 being greatest adherence) the average scores were similar on both diets for the food-delivery phase (~7.5), and similar for the self-provided phase (~5.5)
- It was a bit surprising that the average scores weren’t higher on the food-delivery phase, given that the food delivered was intended to provide perfect adherence. Disappointing…but these are humans not mice. And, we appreciate that they were honest!
- For the self-provided phase, a subtle difference between the two diets was that the Keto group had a much greater range, such that ~4 participants were in the 9-10 range, and ~6 were in the 0-2 range. In contrast, although the average adherence score for Mediterranean was similar to Ketogenic, for Mediterranean the range was much more narrow - none of the participants were in either the 9-10 or the 0-2 range.
Very interestingly, 12 weeks after the study was finished, we characterized just how "Mediterranean" vs. "Ketogenic" the participants were choosing to eat, and we found that they scored an average of ~4 on the Mediterranean diet score, and an average of <2.5 on the Ketogenic diet score – They were eating more Mediterranean than Ketogenic after experiencing both for 12 weeks (or more) each → suggests Mediterranean is more sustainable.
BTW – "Adherence" for Ketogenic was simpler than Mediterranean. Most of the adherence for Keto was simply avoiding total carbohydrate. For Mediterranean, adherence meant eating legumes, grains, fruits, vegetables, no added sugar, no refined grains…

Main Results

For the primary outcome, HbA1c, level improved for both diet phases relative to pre-study, but were not significantly different between diet phases.
Most of the secondary outcomes improved in general, but were not different between the two phases: fasting insulin and glucose, HDL-cholesterol, the liver enzyme ALT, and body weight (weight loss was border line significantly greater for ketogenic, 8% weight loss Keto vs. 7% weight loss for Mediterranean).
The two statistically significant differences between the two diet phases were:
- LDL-cholesterol went up for Keto, and down for Mediterranean (lower is better)
- Triglyceride went down for both groups (lower is better) but dropped significantly more for Keto than Med.
Overall for the primary and secondary clinical measures, it was a tie – main outcome similar, most secondary outcomes similar, each diet phase did better than the other in one of the secondary outcomes.

Qualitative findings

When assessing nutrient levels, the Ketogenic diet was found to provide several nutrients in lower amounts compared to the Mediterranean diet, fiber in particular, but also Thiamin, Vit B6, C, D, E and phosphorous. In contrast Keto was higher than Mediterranean in one nutrient, vit 12. (Notably, these nutrient differences were not linked to any specific changes in clinical outcomes…the study duration was not designed to assess this…12 weeks was not long enough to test for any of these).
Sustainability – From our diet assessment 12 weeks after the study, the participants were eating a diet that was more Mediterranean than Keto, despite scoring similarly in both the food delivery period and the self-provided period. This suggests the Mediterranean diet is more likely to be maintained (or, probably more accurately, that the Ketogenic diet is less likely to be maintained).

Additional findings of interest (for the true geeks)

Although LDL-cholesterol is widely accepted to be an adverse blood factor (lower is better), there is an interesting argument that the type of blood particle that carries the LDL-cholesterol can differ in terms of its "risk". For people whose triglyceride drops (as happened in the keto group), the rise in LDL-cholesterol may not increase CVD risk as much as someone whose LDL-cholesterol went up the same amount but whose triglyceride didn't change → suggests the elevated LDL-C finding for Keto may not be as bad as it appears.
We also have a lot of data that were collected using Continuous Glucose Monitoring. Some of these suggest a modest advantage to controlling blood sugar for the Ketogenic diet. This would not be surprising since the Ketogenic diet provides such low amounts of sugar/carbohydrate. These data are presented as exploratory in the supplemental materials. The reviewers of our manuscript suggested to the editors that we refrain from presenting these in the main paper because they were originally designated as "exploratory" outcomes (and the amount of data collected was substantial and somewhat complex).
Medications: Individuals with T2DM were instructed to stop sulfonylurea medications before starting the WFKD and to reduce the dose by 50% before starting the Med-Plus diet. Out of seven participants taking sulfonylureas at baseline, three used the same dosage for both diets, and four used different dosages between diets. When excluding these four participants (sensitivity analyses), we observed a significant difference for both HbA1c and HDL-C between diets, favoring the Ketogenic phase. In other words, the differential changes in medication use limited the opportunity to observe a benefit for the Keto group. Importantly, the changes in medication dose were made by a study physician regardless of glucose levels at the start of each diet phase and regardless of the glucose response to each diet phase.

Future directions

We collected stool samples from participants. We intend to use these to examine potential differences in the impact of the two diet phases on the microbiome – currently underway in the Sonnenburg lab.
We also collected blood samples to evaluate impact on immune profiling and inflammation. Those outcomes are likely related to potential changes in the microbiome. Those findings will be published in a separate research paper in the future.

Overall conclusions

Study participants experienced substantial improvements in many clinical markers in both diet phases.
Main outcome – HbA1c – improved for both, not significantly different in primary analysis.
Secondary outcomes:
- An adverse effect of the Keto diet on LDL-cholesterol (raised for Keto, lowered for Med).
- Triglycerides better than baseline for both diets (lowered), but significantly better/lower for Keto vs. Med
- Therefore, simplistically, a tie for clinical outcomes.
Given the overall clinical similarities while on both diet phases, there is not a strong rationale for avoiding the types of healthy carbohydrate-rich whole foods (fruits, legumes, whole grains) currently promoted by national and international guidelines but eschewed by/for the Ketogenic diet.
Given that sustainability (diet maintenance) is an important aspect of the potential usefulness of a diet pattern, the observation that the ketogenic diet was less sustained in general once the study concluded (12 weeks later) suggests an advantage to a Mediterranean style diet, low in added sugars and refined grains.
An additional factor in favor of the Mediterranean-style diet would be the more favorable nutrient profile relative to the highly restrictive Ketogenic diet.
Focusing on the widely agreed upon recommendations that vegetable intake be promoted and added sugars and refined grains be eliminated or restricted, a relatively lower carb, higher fat Mediterranean diet is a better choice for metabolic and overall health among adults with prediabetes or T2D – avoidance of fruits, legumes, whole intact grains (WFKD), not supported.

If you have questions, please email nutrition@stanford.edu.

Find a doctor

Clinics & Services

Research Resources

Professional Training

Education Resources

The Keto-Med Study Published in the American Journal of Clinical Nutrition

Quicklinks

Keto-Med Press Release Notes

Stanford Prevention Research Center