Published September 25, 2026
In this blog post, I’d like to discuss a particularly well-done randomized controlled clinical trial that shed new light on the best diet to lower liver fat and improve insulin sensitivity in people with fatty liver disease. As always, we’ll discuss it in the context of the entire literature in this field, and conclude with what we now know about the dietary treatment of fatty liver disease and related metabolic issues, such as insulin resistance and glucose intolerance.
Specifically, researchers studied 42 participants with metabolically unhealthy obesity. That means that participants had obesity and also both prediabetes and elevated liver fat content. They were also very insulin resistant, with HOMA-IR values around 6.8 at baseline. Remember that 1 is normal and even 2 is mildly insulin resistant, so 6.8 is what I would call very insulin resistant. And in line with their fatty livers, they also had elevated fasting triglycerides, averaging around 170, 180 mg/dL in all three groups. I have said before, anytime your fasting triglycerides are higher than 100 mg/dL or so, you should wonder whether you may carry too much fat in your liver. There are some other factors that can raise fasting triglycerides, but most commonly, these are a sign of your liver trying to get rid of excess fat.
As I have discussed in several prior blog posts, carrying excess ectopic fat, that is, in organs not designed for fat storage such as the liver, is a major root cause of insulin resistance. So people who store a lot of fat in their livers usually have insulin resistance, and often also glucose intolerance, such as prediabetes or type 2 diabetes. Exactly what we are seeing in these participants here.
A key piece of advice usually given to such patients is to lose weight, in order to reduce their ectopic and visceral fat. And generally speaking, we so far assumed that it wouldn’t matter too much how that weight is lost, that is, which specific diet someone follows to lose that weight.
Specifically, the researchers had participants follow one of three diets:
- A very-low-carbohydrate ketogenic diet.
- A Mediterranean diet.
- A very-low-fat plant-forward diet.
So these were very different diets in terms of what participants ate. On the keto diet, only 4% of calories came from carbs, 23% from protein, and 73% from fat. Most of the meals focused heavily on animal foods, including meat, fish, eggs, and dairy, but also non-starchy vegetables and nuts. On the very-low-fat plant-forward diet, 70% of calories came from carbs, 15% from protein, and 15% from fat, and it was heavily based on legumes, whole grains, vegetables, fruit, and berries. Some animal foods as well, but mostly plant-based. The Mediterranean diet was somewhere in the middle in terms of its content of carbs and fat, and also in terms of the percentage of animal vs. plant foods.
The researchers studied participants comprehensively at baseline, before they adopted these diets and lost weight, and again after they had lost exactly 10% of their baseline weight.
This is a massive strength of the study. Often, in diet trials, one group loses more weight than another, so you struggle to tell whether differences in health outcomes are due to differences in their diets or differences in how much weight they lost. Here, weight loss was experimentally controlled. So if we see changes in all three diets, that would most likely be explained by the weight loss that was, by design, the same in all diet groups. If we see changes only in one or two groups, we could attribute that to the specific diet composition, beyond the effects of weight loss.
So what happened?
Effects of Weight Loss Seen in All Diet Groups
Let’s start with what happened across the board, in all three diet groups. Weight loss was successful and similar in all groups. Not surprisingly, because, again, this was by design, body weight, fat mass, and fat-free mass decreased in all three groups without significant differences between them. Note that a significant portion of weight loss was in fat-free mass. That’s mostly a loss of muscle tissue, and I’ll get back to that finding later in this blog post.
When we look at insulin sensitivity, and now specifically in muscle tissue, all three diets were winners. Using the hyperinsulinemic-euglycemic clamp technique, the gold standard for measuring insulin sensitivity, the researchers found that weight loss increased muscle insulin sensitivity by about 50% in all three groups, with no meaningful difference between them.
So that’s already a very encouraging finding. No matter which diet people were on, losing weight substantially improved insulin sensitivity in their muscle tissue. This wasn’t measured here, but one mechanism may be that participants had insulin resistance in their muscle tissue at baseline due to intramuscular fat. Fat in muscle tissue is a type of ectopic fat, similar to liver fat, and because these participants were explicitly chosen to have elevated liver fat, it’s likely that they also had elevated ectopic fat in their muscle tissue. Losing weight likely reduced the amount of fat stored in the muscle, removing the root cause of that muscle insulin resistance and, as a result, improving insulin sensitivity in this important tissue.
Where the Diets Differed
However, when we zoom in on the liver, the story changes dramatically.
The researchers found that while all diets reduced liver fat, the very-low-carbohydrate ketogenic diet clearly outperformed the others. The keto group saw a 67% reduction in liver fat, compared to 45% in the Mediterranean and very-low-fat plant-forward diet groups. These two graphs above are simply two different ways to show these data. On the left side, we see that the average liver fat content was around 18-20% at baseline in all three groups. That is clearly very much elevated. Ideally, liver fat content should be 0%, or close to it, and when it’s 5% or higher, we call this fatty liver disease. The 10% weight loss then led to a reduction in liver fat content to somewhere around 6- 10%. Very impressive, but still, by the way, suboptimal. However, in the graph on the right side, we see that if we express the change in liver fat in percent from baseline, the reduction was around 45% in the Mediterranean and very low-fat plant-forward diet, but 67% in the keto diet.
As I’ve discussed in prior blog posts, liver fat is a key root cause of insulin resistance in the liver. So, unsurprisingly, liver insulin sensitivity improved in all groups as liver fat decreased with weight loss. But because the liver fat content decreased the most in the keto diet group, insulin sensitivity of the liver also increased more in the keto group. As we see in the graph below, on the right side, whereas hepatic insulin sensitivity improved by around 50-80% in the Mediterranean and low-fat plant-forward diets, liver insulin sensitivity almost tripled (+~200%) in the keto diet group.
This is a significant finding. I was actually somewhat surprised by how much more effective the keto diet was at improving insulin sensitivity in the liver, and I’ll share later in this blog post how this finding affects my recommendations for people with elevated liver fat, insulin resistance, and glucose intolerance. First, let’s take a look at the rest of the results.
Fasting glucose was reduced significantly in all three diet groups, which we’d expect with such substantial weight loss, but – again – that reduction was most pronounced in participants following the ketogenic diet. Very similarly, fasting insulin was reduced in all three diets, and again, that reduction was most pronounced in the keto diet group. HOMA-IR, a simple measure of insulin resistance based on fasting glucose and fasting insulin, also improved meaningfully in all three groups, but improved most in the keto diet group. And, as an aside, Fasting glucose, fasting insulin, and HOMA-IR are mostly affected by the insulin sensitivity in the liver, and – related to this – the basal glucose production by the liver in the fasting state. That is because all of these are fasting measures, and in the fasting state, the main insulin-sensitive tissue regulating blood glucose levels is the liver. And because the ketogenic diet had the most profound impact of these three intervention diets on liver fat, the greater improvements in fasting glucose and HOMA-IR are no surprise.
How about HbA1c? As you probably know, this measures the average glucose concentration in blood over the past 3-4 months. In the table below, we are seeing moderate reductions in HbA1c in all three groups, so we can say that the degree of improvement is the result of 10% weight loss. And again, we are seeing a slightly stronger reduction in the keto group on top of what we are seeing in the other two diet groups.
Fasting triglycerides, which here are relevant as an indirect measure of liver fat content, did decrease in all three diets, but again that reduction was most pronounced on the keto diet. Very much in line with the greater reduction in liver fat content seen in that group.
Now, a lot of things have improved, but most of these data are not perfect yet in any group. Liver fat was still elevated across the board, fasting triglycerides and HOMA-IR were still elevated, and as a group, participants still had a body mass index in the obese category. Still, some participants reversed their prediabetes; specifically, 50% of participants in the keto group no longer met clinical criteria for prediabetes, compared to 29% of participants in the Mediterranean diet group and only 7% of participants in the low-fat plant-forward diet group.
Why These Diets Improved Metabolic Health, and Why the Keto Diet Was Most Effective
Let’s talk a bit more about mechanisms: why did weight loss lead to these improvements in insulin sensitivity, glucose tolerance, and glycemic control, and why was the keto diet more effective than the other two?
First, we need to acknowledge that even though the researchers called the very-low-carb diet a ketogenic diet, it wasn’t really typical for how we clinically define ketogenic diets. That is because, to achieve ketosis, you ideally should limit both carbs and protein, so a diet with a protein content of 23% of total calories doesn’t usually qualify as keto. So I’d call this a very low-carb diet. In this particular trial, participants may well have achieved ketosis because they were on a calorie-reduced diet plan as well, so their absolute consumption of carbs and protein was probably sufficiently low.
I am mentioning this because the high protein content of the keto diet could well have played a role here in this study. Let’s take a look again at the macronutrient composition of the three diets (see table below). The keto diet was very low in carbs, at 4% of total energy, and remember that these diets were also calorie-reduced, so as a result, total digestible carb intake averaged only 20g per day. And we’ll talk in a minute about why that probably was important. But before we get to that, let’s also acknowledge that the protein content was about 50% higher in the keto diet than in the other two diets, at 23% of total calories compared to 15%.
This matters because dietary protein has been shown to reduce fat accumulation in the liver. We see this in randomized controlled trials where protein replaces carbs or fat, and we even see it in a trial in which protein is added to a high-calorie, high-sugar, high-fat diet. The evidence in this area is not perfect, because every dietary intervention study that aims to investigate the impact of a macronutrient, such as protein, always changes at least one other variable when manipulating dietary protein content (carb intake, fat intake, or calorie intake), any of which could by itself affect liver fat content. Studies showing that replacement of either carbs or fat with protein tends to lower liver fat may seem reassuring that indeed protein is the active component here; however, it is as well possible that protein is simply neutral and that both carbs and fat increase liver fat. However, a particularly insightful study was done by Bortolli and colleagues in 2009. These investigators compared a control diet that was eucaloric (designed to keep body weight stable) with one that contained the same amounts of carbs and protein, but substantially more fat, thereby making it both richer in fat and richer in calories. This diet, as predicted, induced an increase in liver fat content. They then included a third group that also included additional fat plus additional protein, thereby making the calorie excess even more extreme. One would have assumed that this extremely hypercaloric diet would induce an even greater increase in liver fat, but the addition of more protein instead lowered liver fat accumulation, providing the strongest evidence to date that protein actively reduces fat accumulation in the liver.
I am therefore comfortable with the conclusion that even though the evidence is suboptimal in some ways, it seems clear that higher protein intake tends to lower liver fat content, and that therefore the much higher protein intake on the keto diet was one factor contributing to the greater reductions in liver fat and improvements in liver insulin sensitivity.
Now, on top of that, it seems very likely that the lower carb intake was beneficial here as well. Note that we do not know this for sure; it is possible that either of the other two intervention diets could have lowered liver fat to a similar degree as the keto diet if these had been matched for protein content. However, it does seem possible, even likely, that the lower carb intake on the keto diet also played a role here.
To explain why a low-carb intake probably offers benefits in this particular patient population, let me give you a bit of background information. What I’d like to do is briefly review some of the most important causes of insulin resistance.
Anytime we gain weight, the body needs to store excess calories as body fat. And the only safe storage location for body fat is in the subcutaneous fat tissue, right under our skin. This can continue for many years, depending on how much fat a person can store there. At some point, that subcutaneous fat tissue is no longer able to store fat. We call that point the personal fat threshold, and if that person now continues to consume more calories than they burn, another location needs to be found somewhere else in the body where the excess fat can be stored. In that situation, we often find that the concentrations of fatty acids in blood are elevated. We call these free fatty acids, or FFA, because they are not bound to a lipoprotein particle. And these free fatty acids then find their way into tissues that are not intended for primary fat storage, such as the visceral adipose tissue depots surrounding the inner organs and also ectopic fat depots such as the liver and the muscle tissue. Because these tissues are not made for fat storage, the accumulation of fatty acids there tends to cause inflammation and also the buildup of certain lipids, all of which together cause insulin resistance. One thing that happens specifically in the liver and the muscle is that the free fatty acids are taken up, and used to make triglycerides, the main storage form for fat in the human body. Now you have more triglycerides in the liver and muscle cells, which makes them insulin resistant.
OK, now the body tries to compensate for this insulin resistance by producing more insulin. If it cannot do that because the beta-cells are maxed out, then this is the time when glucose intolerance develops, and blood glucose levels start to rise. That’s not good, but usually, this doesn’t happen immediately. In most people, the body can compensate for insulin resistance, at least initially, and produces more insulin. So the result is that in the fasting state and particularly after every meal containing carbs, a lot of insulin will be circulating in the blood. We call this hyperinsulinemia.
Now, having enough insulin in the blood is good in some ways because it helps keep blood glucose levels in the normal range. However, it has some negative side effects. One is that insulin stimulates a process in the liver called de novo lipogenesis. De novo lipogenesis basically means making fats from scratch. This means that when insulin levels in the blood are high, the liver is told to gather raw materials and convert them into fatty acids. The liver then converts these fatty acids into triglycerides. Again, this doesn’t immediately have to be a metabolic catastrophe, because the liver could just assemble a type of lipoprotein called very low-density lipoprotein, or VLDL, and transport the triglycerides out of the liver to fat tissue. But, hey, remember, in our scenario, the subcutaneous fat tissue is already full, so now the fat may come right back in the form of these free fatty acids. So ultimately, this combination of having crossed the personal fat threshold plus chronically high insulin levels due to insulin resistance is the perfect recipe that will lead to a buildup of fatty acids and triglycerides in the liver, and also in muscle, making the insulin resistance worse and worse.
But there is more: if now in this situation, the person is also consuming a lot of foods that can serve as substrate for de novo lipogenesis in the liver, the situation gets even worse. What I am talking about here in particular is alcohol, fructose, and glucose from high-carb foods with a high glycemic index. Let’s say the above person who has crossed their personal fat threshold AND who is insulin resistant and has hyperinsulinemia now drinks a few beers, the alcohol from these beers will quickly be absorbed, and transported right to the liver. And now the liver needs to deal with all of that incoming alcohol, and it does that by converting the alcohol to … fatty acids. The same goes for fructose. Now, not all fructose we consume even reaches the liver. We have data suggesting that small amounts of fructose consumed as part of whole foods, such as whole fruit and berries, can be utilized or converted by the epithelial cells in the gastrointestinal tract. And data suggest that the negative metabolic effects of dietary fructose, such as elevated liver fat, hepatic insulin resistance, and elevated fasting triglycerides, manifest when fructose is metabolized by the liver. I therefore don’t see whole fruit or berries as a problem here.
However, consuming sugar-sweetened beverages, such as soda, lemonade, or energy drinks, as well as beverages with natural sugars, such as fruit juices, could allow some of that fructose to reach the liver, and again, the liver would need to convert the fructose mostly to fatty acids. That’s particularly the case because such sugary beverages also tend to increase overall calorie intake, and that combination of excess calories that need to be stored plus substrate in the form of fructose or alcohol can lead to a rapid increase in liver fat content. The same would happen if someone consumed a lot of easily digestible carbs in the context of a high-calorie meal, as would be typical if we had cake, cookies, donuts, ice cream, or chips, but even white flour bread or white rice. That combination of a lot of calories and a lot of glucose coming into the body triggers a massive insulin spike, particularly if someone is insulin resistant, and some of that glucose would again be converted by the liver to fatty acids and eventually triglycerides.
If you look at the figure above, you’ll notice that a vicious cycle develops over time. Once we have developed some degree of insulin resistance and we keep eating easily digestible carbs, we are constantly in a state of hyperinsulinemia. And that hyperinsulinemia then triggers an increase in de novo lipogenesis, and this is particularly bad if we are in a state of caloric excess, because the body needs to convert incoming alcohol or carbs to fat for storage. That increases liver fat, which worsens insulin resistance, and so on.
Imagine a typical person who has been eating a Western diet high in ultra-processed foods, fast foods, soda, alcoholic beverages, candy, sweets, chips, and ice cream for many decades. They have gained a few pounds every year, and have now reached the point where they have crossed their personal fat threshold; they are very insulin resistant with massive hyperinsulinemia, and a lot of fat accumulated in visceral fat depots, the liver, and the muscle. And now their pancreatic beta-cells are starting to fail (meaning insulin secretion could still be very high, just not high enough for their degree of insulin resistance), and they have developed moderate glucose intolerance, diagnosed clinically as prediabetes.
That person could be one of the participants in our study at baseline.
So now what happened to all of this as they lost weight? Well, the data are very clear: the average person lost about 18 to 19 pounds, 8 to 9 kg, of fat mass, so this is a lot less fat to store in the body. It’s possible that some of these participants moved below their own personal fat threshold, meaning that now they can store all or most of their body fat again in their subcutaneous fat tissue. So this mechanism discussed earlier, where fatty acids are spilling out of an overfull subcutaneous fat tissue, looking for a new home somewhere else, is much less of a problem now that we’ve lost some of that excess weight. And as a result, we would assume that the amount of fat that needs to be stored in visceral fat, muscle, and the liver would have been very substantially reduced. And indeed, we see substantially reduced liver fat in all three of these groups. I would also add that participants were probably discouraged from drinking alcoholic beverages during the study, so this may have helped lower liver fat as well in all three diet groups. And note specifically that even on the high-carb plant-forward diet, the weight loss helped lower liver fat and improve insulin sensitivity quite a lot. So keep this in mind: simply being in a calorie deficit to some degree broke that vicious cycle.
But we can still recognize why the keto diet was most effective at lowering liver fat specifically. That is probably partly because being on keto largely removes substances the liver needs to convert to fatty acids, including fructose and glucose. And keto meals also do not raise blood insulin levels a lot, so the insulin levels would be expected to be much, much lower on the ketogenic diet. And if that were true, we would expect less stimulation of de novo lipogenesis in the liver, meaning fewer fatty acids would be made from scratch and fewer triglycerides would accumulate in the liver. That would partly explain the lower liver fat. Do we have evidence that this is what happened in our study?
Well, the authors did collect blood regularly over a 24-hour period as participants were consuming first standardized Western meals at baseline, and then – after 10% weight loss, meals typical of their assigned diets.
The figure below shows the data for the 24-hour insulin levels. The left panel shows the blood insulin concentrations in the keto group. The grey circles show baseline data before participants began the keto diet. Insulin levels were very high throughout the day, reflecting insulin resistance at this time point and a rather unhealthy Western diet high in easily digestible carbs. The red circles show the insulin levels on the keto diet after 10% weight loss. Much lower insulin levels throughout the day reflect the much improved insulin sensitivity, and also the lower glycemic and insulinemic load of the ketogenic diet.
We see that circulating insulin levels were also reduced after the 10% weight loss in the other two diet groups, but because these participants were still a bit more insulin resistant and their diets had a much higher glycemic and insulinemic load, the reduction in 24-hour insulin levels was much less pronounced when compared to the keto diet.
So yes, circulating insulin levels were much, much lower on the keto diet indeed. Did this actually reduce de novo lipogenesis in the liver, as we would predict? Fortunately for us, the researchers measured this as well. In the figure below, can clearly see that de novo lipogenesis dropped by about 80% from baseline in the participants on the ketogenic diet, and by about 40% in participants on the Mediterranean diet, whereas de novo lipogenesis did not change meaningfully on the low-fat plant-forward diet.
In fact, the degree to which de novo lipogenesis was reduced in the liver was strongly correlated with the reduction in 24-hour insulin levels. In the figure below, the red dots represent participants on the ketogenic diet, the blue dots represent participants on the Mediterranean diet, and the yellow dots represent participants on the low-fat plant-forward diet. So the more 24-hour insulin levels dropped, and – I would add – the fewer carbs were consumed that could serve as a substrate for de novo lipogenesis, the more de novo lipogenesis was reduced, and the more liver fat declined.
But there is more. If you think about the one thing that is characteristic of a ketogenic diet, it is that carb intake is so low that the body and specifically the liver dramatically increases its production of ketones. That is so that tissues all around the body, such the brain, that usually rely on glucose for energy, now have an alternative fuel source, namely these ketones.
OK, and what is the substrate for ketones in the liver? It’s fatty acids.
And in fact, the flow of free fatty acids from the fat tissue to the liver is actually increased in people on ketogenic diets. But now the liver does not use these fatty acids to make triglycerides, but instead uses them to make ketones. So, in a way, forcing the body into ketosis gives the liver an option to do something else with the arriving fatty acids. That may be another mechanism by which the ketogenic diet effectively lowered liver fat content.
So one thing we can say here is that all three diets are effective at addressing some of the key problems in these participants. Specifically, reducing the amount of fat that needs to be stored in the body overall, which substantially lowered liver fat and probably muscle and visceral fat as well. However, the ketogenic diet was particularly effective at disrupting the vicious cycle of elevated liver fat causing insulin resistance causing hyperinsulinemia causing increased de novo lipogenesis causing more liver fat buildup.
Limitations of This Study
This was a very nicely done study, but there is one thing I think could and should have been done better. These researchers assessed beta-cell function before and after the 10% weight loss. Beta-cell function, as measured by the amount of insulin secreted at specific given blood glucose levels, did not meaningfully improve in either group. They also assessed glucose tolerance, as measured at the 2-hour time point in an oral glucose tolerance test, and found that it improved equally in all three groups.
The problem is that a very low-carb diet, such as the keto diet tested here, is well known to temporarily and reversibly reduce insulin secretion, particularly the so-called first-phase insulin response. The first-phase insulin response consists of pre-formed insulin stored inside pancreatic beta-cells that can be secreted within seconds of eating carbs. Now, if you were in charge of managing your own pancreas, and for several months, you had observed that there are no meaningful increases in blood glucose because you are not consuming any carbs, what would you do with the insulin waiting around in the beta-cells? You’d say, ah, that insulin doesn’t seem to be needed, let’s reduce that a bit. And so, as a result, the first-phase insulin response, meaning beta-cell function, and as a result also glucose tolerance in an oral glucose tolerance test, are reduced in people eating very low-carb diets. That may sound like it’s a bad thing (and it’s sometimes sold as a bad thing by opponents of low-carb diets, such as low-fat plant-based diet advocates), but it is not. It is simply a very reasonable adjustment the body makes to use its resources efficiently. The body does not maintain perfect insulin secretion and glucose tolerance on keto because it doesn’t need to.
While on keto, someone would have to gradually increase their carb consumption to get back to their normal state, because why would the body maintain normal beta-cell function and normal glucose tolerance if no carbs are ever consumed? So, in short, I suspect that if people had been prepared for a few days with some carbs prior to the follow-up assessment, their beta-cell function and glucose tolerance would have improved more than seen in the study. We cannot know this for sure, but the data, as reported in the study, given this limitation, are not informative without understanding this important caveat.
Impact of the Study Diets on Risk Factors For Atherosclerotic Cardiovascular Disease
The investigators also assessed the impact of the three study diets on other major risk factors for atherosclerotic cardiovascular disease (ASCVD). LDL-cholesterol and apo B tended to decrease on all three study diets, but that reduction was only significant on the Mediterranean diet. However, the changes in LDL-cholesterol and apo B were not significantly different between the three diets.
This is a result that often confuses people. Why was there a significant change within one diet group only (the Mediterranean diet group) but not the other two, yet the overall test comparing whether these changes in the three groups differed from each other showed no statistically significant difference? First of all, we need to be aware that these were what we call secondary endpoints. The trial was not designed or powered to assess the impact of the diets on these risk factors. It is therefore possible that the trial was underpowered to detect differential effects in these risk factors. Second, in a trial like this, the most important statistical test we should be paying attention to is the one that compared changes between the diet groups, not a test done within a diet group. That is because simply participating in a trial like this, or something else that happens over time, could cause a change within a diet group. So my conclusion here is that the diets did not differentially affect LDL-cholesterol and apo B, but we cannot have a lot of conviction in this conclusion given the probably inadequate power to detect differential changes in these endpoints. We can still use these results to hypothesize that a Mediterranean diet may be preferable for lowering LDL-cholesterol and apo B, and that is very consistent with existing evidence.
With regard to blood pressure, we see exactly the same picture: no differential change between diet groups, but an obvious trend for greater lowering on the Mediterranean diet. So my suspicion is that if a similar study was done that was adequately powered for these endpoints, we may see superiority for the Mediterranean diet compared to the other diets, or at least compared to the keto diet.
How This Trial Compares to Prior Randomized Controlled Trials
The finding of reduced liver fat and improved hepatic insulin sensitivity on a high-protein ketogenic diet was confirmed by a small single-arm study by Luukkonen and colleagues in 10 participants with overweight or obesity who completed six days on a low-calorie, low-carb (6% of calories), but protein-enriched (28%) ketogenic diet. Weight was reduced by 3% during this time, whereas liver fat content was reduced by 31% and hepatic insulin resistance by 58%. An impressive result. However, this study suffered from the fact that no control group was included, and it remains unclear again to what degree reductions in liver fat were caused by the high-protein or low-carb content of the diet.
This was also the problem in a trial by Browning and colleagues. These researchers randomized 18 men and women with non-alcoholic fatty liver disease to two weeks of a hypocaloric diet or an ad libitum low-carb diet, with less than 20g of carbs per day. Again, protein intake was much higher in the low-carb diet, at 33% vs. 16% of total calorie intake. Both groups lost similar amounts of weight over 2 weeks, because participants spontaneously reduced their calorie intake on the low-carb diet to roughly match the controlled calorie intake in the low-calorie diet group. Liver fat decreased by 28% in the low-calorie group and by 55% in the low-carb group, again confirming that even under conditions of similar weight loss, a low-carb high-protein dietary approach is preferable for the reduction of liver fat.
Several studies compared the impact of low-carb vs. high-carb diets on liver fat and related metabolic endpoints.
For example, Dalby-Hansen and colleagues randomized 185 men and women with type 2 diabetes to a low-carb diet with less than 20% of calories from carbs or a high carb diets with 50-60% of calories from carbs. Both diets were similar in the protein content. Participants were studies at baseline and again at 3 and 6 months. Overall, weight loss was greater on the low-carb diet, and this was associated with a greater reduction in HOMA-IR and a trend towards greater reduction in liver fat content. Sadly, because energy intake and weight loss were not standardized, it remained unclear from this very nice trial whether the differences in liver fat and insulin sensitivity were mostly due to the differential weight loss or the differences in carb intake.
The same was seen in a small trial by Cunha and colleagues. Again, men and women with obesity were randomized to a very-low-calorie ketogenic diet (600-800 kcal/day) or a control diet low in calories (1,400 – 1,800 kcal/day). The intervention duration was two months. During this time, participants on the ketogenic diet lost 9.6% of their baseline weight and 39% of their liver fat, compared to a weight loss of 1.9% and a reduction in liver fat of 3% in the low-calorie control group. The main finding of this trial is that over two months, a very low-calorie ketogenic diet is more effective at helping people lose weight than the simple advice to cut calories, which is not overly surprising, but it remained unclear to what extent the much greater reduction in liver fat was due to the greater weight loss only or also due to the much lower carb intake.
Lastly, in the CARBFUNC study, Horn Sommersten and colleagues enrolled participants with central obesity, and randomized them to two high-carb diet or a low-carb diet (11-15% of total calories as carbs). The intervention duration was 12 months. One notable feature of this trial was that protein and fiber content were fairly well matched accross intervention groups. All three groups lost similar amounts of weight, which was associated with similar reductions in visceral fat and liver fat. This study would suggest that the carbohydrate content of the diet is less important than the protein content, and that the more substantial reductions in liver fat on low-carb, high-protein diets is more due to their high protein content. However, this study suffered from several important limitations. Most importantly, 193 participants were randomized , but only 78 completed the study and were included in the final analysis. Such a high dropout rate raises concerns, because (a) investigators have to use statistical methods to estimate missing data points for the gold-standard intention-to-treat analysis (this basically means that all participants that were randomized need to be included in the final statistical analysis); and (b) the final result may be biased by the dropout of those participants who did not do well on either diet (who had side effects or who did not lose weight, for example). I would therefore give this trial a low weighing when assessing the overall evidence.
Several other studies conducted in this area suffered from major limitations, including a very small sample size and minimal differences between study diets, small studies studying people with very low liver fat in whom differential changes between diets are difficult to detect, and a study in which a low-carb diet was included that wasn’t really low in carbs (33% of total calories).
Conclusions
OK, so what are we learning from this study, and the wider field overall?
First, if you carry any excess visceral or ectopic fat, it’s a good idea to lose weight. In fact, that should be your number one priority.
Now, this study was only five months long, and while every improvement has value, we should mostly be interested in weight loss and metabolic improvements that are sustainable. So ideally, you pick a way of eating that is highly satiating per calorie, nutrient dense, and rich in fiber and protein, and that you can happily maintain long-term, ideally forever, so that once you lose visceral and ectopic fat, you will be able to keep it off. If that’s keto or low-carb for you, but if these are not ways of eating you can or want to sustain long-term, it’s fine to adopt another way of high-satiety eating, even though the keto diet was more effective here. Because more effective in the short term doesn’t help you if you cannot stand low-carb and then you fall off the wagon and regain all the weight you lost, plus some more.
I have a strong additional suggestion that would make weight loss even more effective: if you embark on a weight loss journey, engage in regular resistance training, at least twice a week. That will help you better hold on to your muscle mass, or even build muscle, while losing visceral and ectopic fat. Of course, if you don’t lose muscle mass, your rate of body weight loss may be slower, but please don’t let that discourage you. Just indiscriminately losing body weight should not be the goal. Hanging on to muscle mass, or even building more, while losing excess visceral and ectopic fat is going to be way better for insulin sensitivity, better for glucose tolerance, better for long-term weight loss maintenance, better for bone mineral density, and better for healthy, functional aging. Meaning, better for getting up from a chair or down a flight of stairs when you are 80 or 90. So, while this point doesn’t really have anything to do with this study, I feel it’s a very important aspect that everyone should know about before starting a weight-loss journey.
OK, now the second point we have clearly learned from this study is that a higher protein and lower glycemic and insulinemic diet is better for breaking the vicious cycle of chronic hyperinsulinemia, elevated de novo lipogenesis in the liver, higher liver fat content, and insulin resistance. That could be a ketogenic diet, as in this study, and actually being in ketosis may offer some unique benefits because then the liver can convert fatty acids into ketones. However, I would assume that a less strict low-carb high-protein diet would also be a good option.
Now, what if you don’t want to follow a low-carb high-protein diet long-term, or maybe it’s just not feasible for you? In that case, I would adopt a ketogenic or low-carb diet just for the first few weeks to break that vicious cycle, and then gradually transition to a diet with more healthy, whole food sources of carbs that you can maintain long-term.
If you really cannot do low-carb at all, that’s also fine; it may just take a bit longer to get rid of accumulated liver fat. However, even if you do eat carbs, you really need to make sure to minimize added sugar, sugar-sweetened beverages, and high-glycemic starchy foods, i.e., those types of foods that tend to cause us to overconsume calories and that raise blood glucose and insulin levels the most. Aside from that, simply make sure you eliminate foods that cause you to overeat, such as ultra-processed and fast foods, treats, or things like chips or crackers. And design your diet such that it is highly satiating per calorie.
As part of this, I’d definitely suggest considering a higher protein intake than in the Mediterranean and low-fat plant-forward diets tested in this study. No need to go crazy here and have protein-enriched everything, but a slightly higher protein content could certainly help make the diet more satiating per calorie, help with maintaining or building muscle mass, lower the risk of blood sugar spikes, and also have metabolic benefits such as reduced liver fat content. One simple rule of thumb is to try to get at least 20g of protein per meal if you are women, or 30g of protein per meal if you are a man.
And, of course, whichever diet you choose, minimize or completely avoid alcoholic beverages. What this all comes down to is avoiding a state of caloric excess and ideally gradually losing weight, and removing all foods that trigger massive glucose and insulin spikes and that also contain substrates for de novo lipogenesis in the liver. And that means – most importantly – cutting out all beverages containing alcohol or sugar.
There are a few additional dietary factors that do have an impact on liver fat content. This includes dietary fatty acid composition and our gut microbiome, along with determinants such as dietary pre- and probiotic intake. We’ll cover these soon in a separate video and blog post.
One last remark: in their paper, Petersen and colleagues conclude that “the results demonstrate that a very-low-carbohydrate diet has greater cardiometabolic benefits […] than matched 10% weight loss by a Mediterranean or a very-low-far diet in people with metabolically unhealthy obesity”. I feel that conclusion is not fully supported by their results because, as discussed above, there seemed to be a non-significant trend towards greater reductions in LDL-cholesterol, apo B, and blood pressure on the Mediterranean diet. Sure, these differential changes were not statistically significant, but we must note that these were secondary endpoints, the study was not powered to detect differential changes in these, and that’s potentially why no differences were found. Let’s assume for a minute that the authors had done this study in 100 or 200 participants and not 42. It seems at least possible that the Mediterranean diet may have been superior to the ketogenic diet in terms of it’s ability to lower LDL-cholesterol, apo B, and blood pressure. Given that these are major risk factors for cardiovascular disease, it seems premature to conclude that the ketogenic diet had greater cardiometabolic benefits than the other two diets. My conclusion would be that for people with abdominal obesity, fatty liver disease, insulin resistance, and glucose intolerance, a ketogenic diet is more effective at lowering liver fat and improving related metabolic parameters. I don’t feel that this study allows us to draw conclusions on the much wider term ‘cardiometabolic benefits’ that must include other risk factors, such as apo B and blood pressure, which this study could not adequately address.
I hope this makes sense to you. If you have questions, or would like to share your experiences getting rid of liver fat or putting prediabetes or type 2 diabetes in remission, please share in the comment section below.
If you enjoyed this blog post, and would like to support my work, please consider becoming a Patron. I don’t do sponsored content or affiliate marketing for any organization or company related to the food, supplement, or pharmaceutical industries. I want to make unbiased, purely evidence-based content for you guys, so if that’s something you value, I’d appreciate it if you could make a contribution, small or large, everything is welcome. Here is a link to the Nourished by Science Patreon page if you want to support my work on a regular basis, or click here if you want to make a one-time tip. And thank you for your support.
References
- Petersen et al.; Effect of diet macronutrient content on the cardiometabolic response to weight loss: a randomized clinical trial. Cell Metabolism 2026; 38: 1-12.
- Daley et al.; The ketogenic diet: clinical applications, evidence-based indications, and implementation. StatPearls 2025.
- Xu et al.; High-protein diet more effectively reduces hepatic fat than low-protein diet despite lower autophagy and FGF21 levels. Liver International 2020; 40: 2982-97.
- Surowska et al.; Effects of dietary protein and fat content on intrahepatocellular and intramyocellular lipids during a 6-day hypercaloric, high-sucrose diet: a randomized controlled trial in normal weight healthy subjects. Nutrients 2019; 11: 209.
- Bortolli et al.; High protein intake reduces intrahepatocellular lipid deposition in humans. American Journal of Clinical Nutrition 2009; 90: 1002-10.
- Jang et al.; The small intestine converts dietary fructose into glucose and organic acids. Cell Metabolism 2018; 27: 351-61.
- Andres-Hernando et al.; Deletion of Fructokinase in the liver or in the intestine reveals differential effects on sugar-induced metabolic dysfunction. Cell Metabolism 2020; 32: 117-27.
- Luukkonen et al.; Effect of a ketogenic diet on hepatic steatosis and hepatic mitochondrial metabolism in nonalcoholic fatty liver disease. Proceedings of the National Academy of Sciences 2020; 117: 7347-54.
- Browning et al.; Short-term weight loss and hepatic triglyceride reduction: evidence of a metabolic advantage with dietary carbohydrate restriction. American Journal of Clinical Nutrition 2011; 93: 1048-52.
- Dalby Hansen et al.; Effect of calorie-unrestricted low-carbohydrate, high-fat diet versus high-carbohydrate, low-fat diet on type 2 diabetes and nonalcoholic fatty liver disease. Annals of Internal Medicine 2023; 176: 10-21.
- Cunha et al.; Efficacy of a 2-month very low-calorie ketogenic diet (VLCKD) compared to a standard low-calorie diet in reducing visceral and liver fat accumulation in patients with obesity. Frontiers in Endocrinology 2020; 11: 607.
- Horn Sommersten et al.; Diets differing in carbohydrate cellularity and amount similarly reduced visceral fat in people with obesity – a randomized controlled trial (CARBFUNC). Clinical Nutrition 2022; 41: 2345-55.
- Ramon-Kraul et al.; A low-glycemic-load versus low-fat diet in the treatment of fatty liver in obese children. Childhood Obesity 2013; 9: 252-60.
- Martens et al.; The potential of a high protein-low-carbohydrate diet to preserve intrahepatic triglyceride content in healthy humans. PLOS One 2014; 9: e109617.
- Utzschneider et al.; Beneficial effect of a weight-stable, low-fat/low-saturated fat/low glycemic index diet to reduce liver fat in older subjects. British Journal of Nutrition 2013; 109: 1096-1104.
- Friden et al.; Effects of an anti-lipogenic low-carbohydrate high polyunsaturated fat diet or a healthy Nordic diet versus usual care on liver fat and cardiometabolic disorders in type 2 diabetes or prediabetes: a randomized controlled trial (NAFLDiet). Nature Communications 2025; 16: 11130.