What is Fatty Liver?

Fatty liver is characterised by the retention of triglycerides in liver cells (hepatocytes). Within hepatocytes, triglycerides are stored in the cytosol, in lipid droplets. Non-alcoholic fatty liver disease (NAFLD) is diagnosed when:

1. There is accumulation of triglycerides in  more than 5% of hepatocytes
2. This accumulation is not due to excessive alcohol consumption

Current estimates suggest that around 25% of the global population may exhibit NAFLD, correlating with the prevalence of the metabolic syndrome related to the clustering of cardio-metabolic risk factors arising from increased hepatic fat. The accumulation of fat in the liver is a major driver of cardio-metabolic disease, increasing risk for both cardiovascular disease and type-2 diabetes, while also increasing risk of cirrhosis and hepatocellular carcinoma (the most common type of primary liver cancer).

Risk Factors for NAFLD

In relation to risk factors for NAFLD, non-modifiable risk factors include:

1. Sex: Males have greater risk than females
2. Ethnicitiy: Asian and Hispanic ethnicity exhibit greater prevalence than Black or White populations
3. Genetics: Particular genetic variants increase risk

Related modifiable risk factors include:

1. Central adiposity and visceral fat
2. Sedentary behaviour
3. Diet

Although NAFLD tends to be referred to in the singular, the condition in fact refers to a spectrum of liver disorders from hepatic steatosis (non-alcoholic fatty liver) to NASH (non-alcoholic steatohepatitis) and liver cirrhosis. The scope of this Statement will be hepatic steatosis, i.e., 'fatty liver' or 'metabolic NAFLD', and the nutritional determinants of hepatic fat together with a summation of evidence for interventions targeting reductions in fatty liver.

Pathways of Liver Fat Accumulation

The net retention of triglyceride in liver cells is a reflection of a situation where there is a surplus of fatty acids entering the liver, relative to fatty acids leaving the liver. There are a number of different pathways of delivery of fatty acids to the liver, which differ depending on the fed vs. fasted state, and level of adiposity. Once delivered to the liver, non-esterified fatty acids (NEFA) form a fatty acid pool, from which there are two primary disposal pathways:

1. Esterification
2. Beta-oxidation

Beta-oxidation  of fatty acids provides energy (oxidizing fatty acids is colloquilly phrased as "burning" fat), and occurs in the mitochondria of cells. Esterification involves the production of new triglycerides in the liver from NEFA (recall that NEFA are 'non-esterified'). These triglycerides can either be stored in hepatocytes as lipid droplets or packaged into very-low density lipoproteins (VLDL) and exported into systemic circulation as VLDL-TGs. Measuring the fatty acid content of VLDL-TGs provides a means of determining the respective contribution of the different sources of hepatic fatty acids.

By Cruithne9 - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=45500333

Adipose tissue lipolysis is the breakdown of triglycerides stored in a fat cell into NEFA, or free fatty acids, and glycerol, which are then in turn released from the fat cell. In the fasted state, adipose tissue lipolysis constitutes the primary endogenous pathway that delivers fatty acids to the liver. In the fasted state, these NEFA that are released into circulation contribute to the formation of triglycerides that are carried in VLDLs, which are derived from the liver. In longer duration fasts, NEFA derived from adipose tissue lipolysis may account for nearly 100% of liver fatty acids used for the synthesis of VLDL-triglycerides. In general however, the contribution of systemic NEFA to liver fatty acids may be in the region of 45-75%.

However, humans spend the majority of any day in the post-prandial, fed state, and average daily eating durations may be up to 16 hours. Thus, fatty acid flux to the liver from dietary intake is a crucial determinant of hepatic fat content, and three main exogenous pathways are related to post-prandial metabolism:

1. Chylomicron-derived fatty acid spillover.
2. Chylomicron remnants.
3. De novo lipogenesis.

© 2019 Caroline C. Duwaerts, Jacquelyn J. Maher. Published by Elsevier Inc. on behalf of the AGA Institute. From: Duwaerts & Maher, 2019.

The chylomicron pathways are derived from dietary fat:

1. In the digestion of dietary fat, it (in the form of triglycerides) enters circulation from the intestines packaged inside chylomicrons.  Chylomicrons are large, triglyceride-rich lipoproteins.
2. The triglycerides (TG)  in chylomicrons are hydrolysed (broken down) into free fatty acids by a group of enzymes known as lipases. In particular  by lipoprotein lipase (LPL), which may act preferentially on chylomicron-TGs compared to VLDL-TGs (the triglycerides in VLDL particles). In other words, LPL preferentially acts on the exogenous pathway (i.e. from diet), rather than the endogenous pathway (i.e. liver TG synthesis).
3. When LPL acts to hydrolyse the triglcerides in chylomicrons, a proportion of those fatty acids are not taken up by adipose tissue, and so they "spillover" into the pool of systemic free fatty acids (or NEFA). This fatty acid pool contributes the greatest proportion of fatty acids to intra-hepatic triglycerides (IHTG), i.e. triglycerides within the liver.
4. The pathway of LPL-mediated metabolism of chylomicron-TGs also produces what are known as remnants. These are formed when the hydrolysis of chylomicron-TGs results in a smaller lipoprotein being leftover, i.e., a chylomicron-remnant.
5. These chylomicron-remnants are taken up by the liver, and the remaining TGs in the remnant particle may be repackaged into VLDL.

The primary difference between the chylomicron spillover and chylomicron remnant pathways is the speed of incorporation of TGs into VLDL. Fatty acids delivered from the spillover pathway and systemic free fatty acid pool occurs more rapidly, and appear as VLDL-TGs at a faster rate than fatty acids derived from remnants are incorporated into VLDL. However, this difference should not be interpreted as the spillover pathway contributing to greater fatty acid uptake by the liver. Over a 24 hour period, the contribution of NEFA derived from chylomicron-remnants to VLDL-TGs has been shown to be greater than the contribution of chylomicron spillover NEFA.

The final exogenous pathway is de novo lipogenesis (DNL). This is a process where fatty acids are synthesised in the liver from non-fat precursor sources, primarily from excess dietary carbohydrate (proteins contribute very little to DNL), in particular free sugars. The contribution of DNL to hepatic NEFA (fatty acids in the liver) in metabolically healthy individuals is relatively small (<5%). However, the presence of fatty liver substantially modifies the rate of DNL, which may be up ~22-24% of fatty acids in VLDL-TGs in individuals with NAFLD. Insulin resistance strongly modifies post-prandial DNL, which increases in individuals with elevated insulin levels.

It is clear that the metabolic health status of the individual also mediates the respective origins of hepatic fat, as evident with DNL. In addition, the contribution of  lipolysis in the visceral abdominal area (splanchnic lipolysis) to liver fatty acids is in the region of 5-10% in lean individuals, but may be up to ~30% in individuals with high visceral abdominal adiposity. The reason for this increased delivery of NEFA in individuals with visceral adiposity appears to be the direct release of these fatty acids into the portal vein, and thus direct uptake by the liver. Therefore, increased visceral adipose depots results in a greater proportion of NEFA derived from lipolysis in the visceral abdominal area, as opposed to being derived from the systemic NEFA pool, which ordinarily provides the greatest delivery of fatty acids to the liver. It should be noted that chylomicron-spillover is a physiologically normal characteristic and is observed to a greater degree in lean individuals and in women, whom exhibit a greater capacity for post-prandial triglyceride clearance, factors which may be mediated by sex hormones and adipose tissue distribution.

In sum, liver fat originates from:

1. The endogenous systemic fatty acids derived from adipose tissue and splanchnic lipolysis
2. The exogenous dietary fatty acids derived from chylomicron spillover or chylomicron remnants
3. The de novo synthesis of fatty acids from non-fat precursors, in particular carbohydrate

The respective contribution of fatty acids to the triglcerides contained within VLDL particles (VLDL-TG) have been shown to be in the region of:

• 75-84% from the systemic NEFA pool
• 12-39% from dietary fatty acids,
• 5-22% from DNL

The ranges listed above reflect the wide variability that is due to the metabolic health of the individual, in addition to dietary composition. Increased intra-hepatic triglycerides (fat in the liver cells) and consequent hepatic and adipose tissue insulin resistance results in increased concentrations of circulating non-esterified fatty acids (NEFA) in the postprandial period, upregulating triglyceride synthesis and impairing postprandial TG clearance. In addition to increased TG synthesis from increased circulating NEFA, de novo lipogenesis of liver TGs may be increased from carbohydrate overfeeding, and exacerbated by the presence of fatty liver. Thus, both dietary fat and carbohydrate influence the accumulation of liver fat.

Cardio-Metabolic Consequences of Liver Fat

The accumulation of liver fat relates to multiple components of the metabolic syndrome. Metabolic syndrome is generally defined as at least 2 of the following risk factors:

• High fasting glucose levels
• High triglycerides (known as 'hypertriglyceridemia')
• High blood pressure
• Low HDL-cholesterol levels

This clustering of risk factors is strongly associated with increased risk of both cardiovascular disease (CVD) and type-2 diabetes (T2DM), in addition to advanced liver diseases. More particularly, each of the components of the metabolic syndrome strongly correlate with liver fat, to the extent that NAFLD has been termed the “hepatic manifestation of the metabolic syndrome”. The presence of NAFLD also poses a significant increase in risk for the development of non-alcoholic steatohepatitis and cirrhosis, however for the purposes of this Statement we will not expand on these outcomes.

As noted above, increased delivery of NEFA to the liver results in increased triglyceride production and an upregulation of VLDL synthesis. The triglycerides are exported from the liver as VLDL-TG. And the resultant hypertriglyceridemia is central to the CVD risk related to fatty liver, through driving what is known as the 'atherogenic lipoprotein phenotype' (ALP). The ALP is characterised by the combination of:

• Elevated triglycerides
• Low levels of HDL (high-density lipoproteins)
• Predominance of abnormally small, dense LDL (low-density lipoproteins)

Hypertriglyceridemia results in an overburdening of HDL with excess triglycerides. The HDL return to the liver and are catabolised, thus leading to the observed low HDL levels in the ALP. In addition, impaired clearance of circulating triglycerides results in LDL remodelling into small, dense LDL (sdLDL). These processes are inherently linked to the impaired glucose tolerance and insulin resistance induced by fatty liver. In the post-prandial period liver production of VLDL is generally inhibited by insulin. However, the fatty liver state impairs insulin action, while VLDL clearance continues uninhibited, resulting in hypertriglyceridemia. This milieu of metabolic complications of fatty liver may result in vascular complications, increasing risk of coronary artery atherosclerosis, ischemic heart disease and increased carotid intima media thickness. Patients with NAFLD exhibit significantly greater risk for both fatal and non-fatal CVD events compared to non-NAFLD controls.

Fatty Liver & Type-2 Diabetes

Insulin resistance in both liver and fat tissue, in addition to the spillover of fat from the liver to the pancreas, may all contribute to increased risk of T2DM. NAFLD and T2DM may co-exist, which is associated with worse metabolic profiles than either condition alone. A recent meta-analysis indicated that the association between NAFLD and T2DM incidence was as high as 87% in certain studies, with a pooled prevalence of 59.6%. Fatty liver influences a number of diabetes risk factors. Insulin inhibition of hepatic (liver) glucose production is impaired by fatty liver, thus increasing fasting glucose levels and resulting in elevated insulin levels. The excess contribution of splanchnic lipolysis to circulating levels of free fatty acids may play a role in the pathogenesis of insulin resistance, as elevated NEFA induces both hepatic and adipose tissue insulin resistance while potentially reducing beta-cell function. Excess accumulation of fat in the liver spills over into other visceral organs and tissues, in particular the pancreas. Increasing fat depots in the cells of the pancreas inhibits expression of insulin genes, defective processing of proinsulin precursors to insulin, and progressive decline in beta-cell function, which predates diagnosis of T2DM. Thus, the hyperglycaemic, hyperinsulinemic complications of fatty liver, and the bi-directional relationship with hypertriglyceridemia and circulating NEFA, creates a complex interaction where the presence of NAFLD may drive progression to T2DM, and vice versa.

Currently, there are no available pharmaceutical interventions specifically for NAFLD, although a number of diabetic drugs may exert benefits. Dietary influences on liver fat (in addition to wider lifestyle interventions, e.g., exercise) are therefore of central importance to the management and treatment of fatty liver.

Diet & Fatty Liver: Observational Evidence

A number of observations related to dietary intake indicate that both fat composition and carbohydrate type may be important determinants of liver fat. Cross-sectional data (here & here) has shown an association between icreased liver fat and:

• High calorie diets
• High total fat intake
• High saturated fat intake
• High total carbohydrate and added sugar/glycaemic index diets
• Low omega-3 intakes

Further cross-sectional research has indicated that patients with NAFLD have low levels of polyunsaturated fatty acids (PUFA) in liver cells, compared to non-NAFLD controls. Food source may have important implications for the nutrient-based findings. Circulating levels of C15:0 and C17:0, both biomarkers of dairy fat consumption, have been associated cross-sectionally with lower liver fat content. Conversely, in the Multi-Ethnic Cohort, red and processed meat were both associated with increased risk of NAFLD, while fibre intake was inversely associated. In an Israeli cohort study, high intake of sugar-sweetened beverages and red meat intake were associated with significantly increased risk of NAFLD, while low omega-3 fish intake was non-significantly associated.

Observational studies have the ability to adjust for total energy intake in determining associations between an exposure and outcome, however, it is known that total energy intake is strongly associated with NAFLD. There may be effects of specific foods and nutrients however, which will be further elucidated below. Given the relationship between both total energy and specific nutrients, the following sections will address the evidence from intervention studies from the perspective of:

1. Hypercaloric (overfeeding) diets
2. Eucaloric (energy balance) diets
3. Hypocaloric (energy deficit) diets
4. Specific nutrients of interest

Hypercaloric Diets: Overfeeding Studies

Both dietary sugar and fat intake have been the subject of a number of overfeeding studies, either in isolation or comparing the effects of one against the other. With regard to the role of sugars, there are two potential pathways through which sugars may increase risk:

1. Through increased DNL and triglyceride synthesis (with downstream effects on lipoprotein remodelling). This may be considered the direct pathway
2. Through increased body fat, which strongly correlates with accumulation of visceral and liver fat. This may be considered the indirect pathway.

In relation to the indirect pathway, increasing caloric intake from sugar under ad libitum conditions (eating as much or as little as desired) is associated with increased weight gain. This is attributable to the lack of apporpirate compensation for the additional energy intake. The implication is that the adverse effects of free sugars may primarily be through increased adiposity. Intervention studies, however, have used different exposures; with some studies using monosaccharide forms of fructose or glucose (which are not present in the food supply, as the commonly used added sugars are a mix of fructose and glucose), while other studies have used added sugars specifically. In one free-living intervention by Maersk et al., participants were randomly assigned to one of four groups where they consumed 1 litre of that beverage each day for six months:

1. Coca Cola - 1 litre per day (100g sucrose)
2. Milk - 1 litre per day (47g milk sugars)
3. Diet Coke - 1 litre (0g sugars, 0 kcal)
4. Water - 1 litre (0g sugars, 0 kcal)

The group consuming one litre (1L) of Coca-Cola per day exhibited significant increases in liver fat, from 3.7% to 5%. However, this was not a like-for-like comparison in terms of the dose of sugar between the calorie-containing beverages, given that milk had less than half the sugar content of the Coca-Cola.

A further hypercaloric intervention over 3 weeks of overfeeding added sugars by 1,000 kcal/d extra in participants with NAFLD demonstrated a significant 27% increase (from 9.2% to 11.7%) in liver fat. This was proportional to the increase in DNL from sugar overfeeding. This study lacked a control or comparative arm, but it is consistent with the prior study and suggests an influence of added (free) sugars in the habitual food supply on increasing liver fat, when provided as an energy surplus.

A number of studies have compared the effects of overfeeding from simple sugars against overfeeding from dietary fat in conditions of overfeeding, which will be discussed further below. However, it is important to discuss the metabolic health of the host as an important determinant of responses to dietary sugar intake. Umpleby et al. compared high and low added sugar diets, corresponding to the upper and lower 2.5th percentile of intake in the UK population, i.e.:

• High: The 2.5% of the population with the highest intake (around 26% of calories from added sugars)
• Low: The 2.5% of the population with the loweest intake (around 6% of calories from added sugars)

Participants included men with NAFLD and low-liver fat controls, and diets were matched for total energy intake. Fatty liver modified the effects of the diets: participants with NAFLD exhibited greater total VLDL-TGs in response to both high and low sugar diets. Further, while the high sugar diet resulted in an upregulation of larger, less atherogenic VLDL1, in participants with NAFLD the high sugar diet resulted in a greater upregulation of more atherogenic VLDL2. Splanchnic-derived fatty acids contributed more to VLDL-TG in participants with NAFLD. Thus, it can be concluded that the state of fatty liver alters the metabolic response to dietary added sugars.

Glucose vs. Fructose

Before examining the overfeeding studies comparing different macronutrients, however, it is important to discuss the effects of fructose vs. glucose. Fructose has attracted much attention - and hysteria - as uniquely responsible for increased adiposity, on the mechanistic basis that fructose metabolism bypasses the rate-limiting step in glucose metabolism in the liver, resulting in a greater availability of fructose for DNL and triglyceride synthesis. Indeed, compared to glucose, fructose may increase DNL to a greater degree than glucose. However, in the context of overfeeding it does not appear that the difference in circulating triglycerides between fructose and glucose is significantly different. Elevated triglycerides in response to fructose overfeeding exhibits a dose-response, with >50g/d increasing postprandial triglycerides, while 100g/d increases fasting circulating triglycerides . However, these levels reflect fructose as a monosaccharide (indiviudual sugar unit; mono- = “one”; sacchar- = “sugar”), which is not present in the food supply (fructose is typically present when combined with glucose in the form of sucrose or high-fructose corn syrup). Additionally, fructose overfeeding studies have averaged 3-4 g/kg body weight per day or an average of ~187g monosaccharide fructose. While overfeeding fructose to these levels has been shown to significantly increase liver fat content, studies comparing fructose to glucose have found no significant difference between either sugar monosaccharide type in the increase in liver fat.

For example, Johnston et al. compared 25% energy each from monosaccharide fructose vs. glucose, in both hypercaloric and isocaloric phases, in healthy participants. During overfeeding (hypercaloric conditions), both fructose and glucose significantly increased intra-hepatic triglycerides (IHTG) by 1.7% and 2.1%, respectively. But these effects which were not evident during the isocaloric feeding phase, despite the dose of each sugar remaining the same. More particularly, the outcome was influenced more by energy status and increased body weight, rather than monosaccharide type. Overfeeding fructose vs. glucose by 3 g/kg/d for seven days resulted in a significant increase in liver fat content in both groups, effects which were not observed at 1.5g/d fructose. A further study overfeeding fructose vs. glucose by 600 kcal/d found no significant effect of either diet, or difference between diets, on liver fat over 10 weeks.

Sugar Overfeeding: Conclusions

A number of conclusions may be drawn from the available evidence for sugar overfeeding:

1. There does not appear to be any material difference comparing monosaccharide fructose vs. glucose overfeeding on liver fat content or fasting insulin sensitivity.
2. The monosaccharide overfeeding studies may lack ecological validity for both habitual population levels of added sugar consumption and the type of sugar utilised in the food supply.
3. While monosaccharide fructose does increase circulating TGs in a dose-dependent manner, it appears the increase in TGs from overfeeding either monosaccharide or added sugars is similar, i.e., overfeeding simple sugars per say may increase circulating TGs, and this is supported by food-based interventions.
4. Increased DNL in response to sugar overfeeding may explain the increase in liver fat, however, the sugar overfeeding studies have all resulted in significant increases in bodyweight, rendering any particular effect of sugars difficult to disentangle from the effects of energy excess and increased adiposity per se.

As noted above, many of the overfeeding studies have compared monosaccharide fructose vs. glucose, or examined the effects of overfeeding added sugars without a comparative macronutrient group. However, a number of studies have directly investigated the effects of overfeeding sugars compared to fats on liver fat and related pathways.

Overfeeding Fat vs. Sugar

Sobrecases et al. compared the effects of overfeeding by:

• 35% with fructose
• 30% with saturated fat
• 65% with both fructose + saturated fat.

Fructose overfeeding increased IHTG by 16%, and VLDL-TGs by 58%, while saturated fat overfeeding increased IHTG by 86%. The combination of fructose and saturated fat overfeeding  led to a 133% increase in IHTG. In this study, the effects of fructose on circulating VLDL-TGs did not parallel the changes in IHTG, which goes to confirming previous fructose overfeeding studies insofar as the primary effect of overfeeding is on DNL (as reflected in VLDL-TGs). While in contrast, overfeeding saturated fat leads to direct increases in intracellular fat in the liver.

In an intervention by Luukkonen et al., the impact of three overfeeding conditions on liver fat were examined:

1. Overfeeding by 1,000 kcal/d from saturated fat
2. Overfeeding by 1,000 kcal/d from unsaturated fat
3. Overfeeding by 1,000 kcal/d from simple sugars

The study found that intra-hepatic triglycerides (IHTG) increased by:

• 55% in response to saturated fat overfeeding
• 33% in response to simple sugars overfeeding
• 15% in response to unsaturated fat overfeeding

While DNL increased significantly only in response to simple sugars, the contribution of lipolysis (the breakdown of stored fats into free fatty acids) as a pathway contributing to IHTG was increased significantly only in the saturated fat group. Metabolic outcomes were also differentially affected by overfeeding, with insulin resistance increased by 23% in the saturated fat group, an effect which may have been mediated by a 49% increase in plasma ceramides. Ceramides are compounds synthesised in the body from saturated fatty acids that increase insulin resistance and interfere with blood glucose regulation. There is mechanistic support for how ceramides lead to insulin resistance: circulating ceramides lead to inhibition of the  GLUT4 glucose transporter responsible for shuttling glucose into the cell, thereby reducing glucose uptake. The increase in plasma ceramides were not observed in either the unsaturated fat or simple sugar groups.

The relatively minor increase in liver fat from unsaturated fat (which was predominantly monounsaturated fat) in the Luukkonen study (an absolute difference of 4.8% at baseline to 5.5% at the end) indicates that the degree of unsaturation of fatty acids may be relevant for the post-prandial pathways influencing liver fat. Rosqvist et al. compared overfeeding of saturated fat to omega-6 polyunsaturated fat (PUFA) in an intervention designed to cause a 3% gain in bodyweight. After seven weeks, while both groups had gained the same amount of absolute weight (1.6kg), liver fat increased by drastically different amounts, specifically:

• Liver fat increased by 58% on the saturated fat diet
• Liver fat increased by only 5% on the omega-6 PUFA diet

Interestingly, overfeeding omega-6 PUFA caused greater lean mass increases, and the ratio of lean mass to fat mass gained was 1:1 in the PUFA group, compared to 1:4 in the saturated fat group. In the saturated fat group total body fat and visceral fat both increased in addition to liver fat. Thus, the increase in liver fat was independent of the gain in bodyweight, which was similar between groups.

Meal Frequency in Overfeeding

Finally, the potential role of meal frequency may be relevant. Koopman et al. conducted a six week overfeeding study comparing the effects of overfeeding by 40% with high-fat/high-sugar (HFHS), or high-sugar alone (HS). They also compared the effect of the overfeeding coming as a result of either greater meal size or increased meal frequency. And so there were four conditions to compare:

1. 40% calorie surplus via increased high-fat + high-sugar - Increased meal size (liquid meal replacement alongside 3 main meals)
2. 40% calorie surplus via increased high-fat + high-sugar - Increased meal frequency (liquid meal replacement as a snack between the three main meals)
3. 40% calorie surplus via increased high-sugar alone - Increased meal size (liquid meal replacement alongside 3 main meals)
4. 40% calorie surplus via increased high-sugar alone - Increased meal frequency (liquid meal replacement as a snack between the three main meals)

Interestingly, while there was no effect of either HFHS or HS alone when 40% energy surplus was consumed as 3 large meals, liver fat (IHTG) increased significantly with increased meal frequency by 45% in the HFHS group and 110% in the HS group. The fact that the effects were observed in both high-frequency diets  indicates that independent of the macronutrient composition of overfeeding, the increase in liver and abdominal fat resulted from higher frequency alone, and that the same level of carbohydrate and/or fat had little to no effect when consumed as 3 meals per day. Further, the fact that the HFHS meal replacement contained primarily unsaturated fats may explain why there was less of an effect of the HFHS frequency diet vs. the HS frequency diet, given that overfeeding studies have generally demonstrated a greater effect of saturated fat compared to simple sugars in increasing IHTG.

Conclusions from Overfeeding Studies

In sum, in the context of overfeeding and weight gain, the type of fatty acids and the degree of saturation appear to be important mediators of the effect of dietary fat on liver fat. In relation to liver fat specifically, a hierarchy of effect appears present during overfeeding with the increases in liver fat being greatest in order of:

1. Saturated fat
2. Simple sugars
3. Unsaturated fat (with potentially even a more protective effect of polyunsaturated fat vs. monounsaturated fat).

Conversely, simple sugars increase DNL and VLDL-TGs to a greater degree than dietary fat, although there is a disconnect between the magnitude of increase in DNL and accumulation of IHTGs. However, as previously stated, total energy is an important moderating factor with any analysis of specific nutrients, and thus studies conducted in energy balance may further elucidate more specific nutrient effects.

Eucaloric Diets: Energy Balance Studies

A number of eucaloric intervention studies have been conducted comparing carbohydrates vs. fats, and comparing different fat subtypes, for effects on liver fat. In the HEPFAT trial, the effects of a diet containing 15% omega-6 PUFA was compared to that of 20% saturated fat (with both diets containing the same total fat of ~38%), over 10 weeks, with diets aimed at maintaining weight. While weight increased non-significantly by 0.4kg and 0.8kg in the PUFA and saturated fat groups, respectively, the omega-6 PUFA diet resulted in a 9% decrease in liver fat content, while the saturated fat diet resulted in a 7% increase. Increases in circulating serum linoleic acid (an omega-6 PUFA) levels were shown to correspond to the reduction in liver fat in the PUFA group, while increased liver fat positively correlated with greater circulating saturated fatty acids. While the study was conducted as a free-living intervention, and therefore it is possible that the diets may not have been fully complied with, the study demonstrated that in the absence of weight loss, higher omega-6 PUFA resulted in decreased liver fat while higher saturated fat increased liver fat content.

In the context of energy balance, Parry et al. compared two diets (protein intake was matched between diets at 15%):

1. High simple sugar diet: 65% carbohydrate (of which 20% simple sugars) and 20% fat
2. High saturated fat diet: 45% total fat (20% saturated fat) and 40% carbohydrate

Diets were consumed for four weeks in free-living conditions, before a seven week washout period, followed by 'crossing over' to the other diet. While the study was aimed to be eucaloric, the saturated fat diet resulted in a 1.5kg gain in bodyweight compared to maintenance achieved in the sugar diet (0.2kg increase in bodyweight). Liver fat (IHTG) increased by 39% in the saturated fat group, whilst there was no change in liver fat in the added sugar diet. Linear regression analysis of the relationship between weight gain and liver fat demonstrated that the increase in bodyweight explained only 17% of the change in IHTG. In other words, the majority of the increase in liver fat was attributable to the diet. There was no significant difference between the diets in post-prandial DNL, while post-prandial glucose and insulin excursions were greater and more prolonged over the entire post-prandial period, after the saturated fat diet.

A couple of studies have investigated the effects of monounsaturated fats (MUFA) and fibre in isocaloric interventions. Over a 12-week study in participants with NAFLD and prediabetes, Errazuriz et al. compared:

• A diet containing 28% MUFA (with half of this derived from extra-virgin olive oil)
• A diet containing 20g fibre per 1,000 kcal of intake
• A control group consuming their habitual diet

Weight remained stable throughout the intervention in all diet groups. In the MUFA-diet, there was a statistically significant reduction in measured liver fat from 9.7% to 8%. This magnitude of difference represented an 18% reduction in liver fat compared to baseline in the MUFA-diet. There was no significant difference in the Fibre-diet group, which increased insignificantly by 2% from baseline. Liver fat non-significantly increased by 13% in the control group. There was no significant difference in any of the glucose tolerance or insulin parameters in any diet group. However, it is also worth mentioning here again that carbohydrate content was not matched across diets, and average intake during the study was 188g/d in the MUFA-diet, compared to 256g/d on the Fibre-diet, and 241g/d on the control, and the lower carbohydrate content may have modified the effect of the MUFA-diet.

In an eight week study in participants with T2DM, Bozzetto et al. also compared 28% MUFA diet to a high-carb/high-fibre/low-GI diet containing 28g of fibre per 1,000kcal, at eucaloric energy intakes. Liver fat decreased by 29% in the MUFA diet group, compared to a 4% decrease in the high-fibre group. These studies demonstrate that in the context of energy balance, MUFA-enriched diets decrease liver fat while fibre intakes in a range of 20-28g/1,000kcal have little effect.

It may be compelling to interpret these studies as demonstrating greater efficacy for high-fat diets in reducing liver fat, but this is not necessarily the case. Isocaloric diets comparing low-carb/high-fat diets vs. low-fat/high-carb diets consistently show greater reductions in liver fat on the low-fat/high-carb diets. Westerbacka et al. compared two isocaloric diets in participants with fatty liver, with participants consuming each diet for two weeks each. The diets contained:

1. Low-fat/high-carb diet: 16% fat, 61% carbohydrate, 19% protein
2. High-fat/low-carb diet: 56% fat, 31% carbohydrate, 13% protein

All foods were provided to participants and weight remained stable throughout the study. During the low-fat diet phase, liver fat decreased by 20%. While during the high-fat diet phase, liver fat increased by 35%.  The difference in total fat was marked by the difference in saturated fat content, which constituted 5% and 28% of the low-fat and high-fat diets, respectively.

Similarly, van Herpen et al. compared low-fat and high-fat diets over 3-weeks in healthy overweight men. The diets contained:

1. Low-fat diet: 20% fat, 65% cabohydrate, 15% protein
2. High-fat diet: 55% fat, 30% cabohydrate, 15% protein

The low-fat diet resulted in a 13% decrease in liver fat, compared to a 17% increase in the high-fat group.

Conclusions from Energy Balance (Eucaloric) Studies

Results of eucaloric intercentions allow one to conclude the following:

• The lack of effect of high dietary sugars on post-prandial DNL during conditions of relative energy balance suggests that this effect of sugars is primarily mediated by energy surplus.
• Meanwhile the increase in liver fat from saturated fat is evident in both energy balance and energy surplus, with the magnitude of effect increasing with greater calorie intake.
• In relation to low-fat/high-carb vs. low-carb/high-fat diets, it is important to distinguish effects: lowering carbohydrate intake may be effective at reducing liver fat, therefore the increases in liver fat demonstrated on these isocaloric low-carb/high-fat diets appears to be attributable to the high fat content of the diets.
• In the context of energy balance, it appears that the total fat content of the diet (with the level of saturated fat comprising that being a critical factor) is the primary mediator of increases or decreases in hepatic fat content. This is particularly the case where fibre intake does not appear to have any appreciable effect on liver fat reductions.

Hypocaloric Diets: Energy Deficit Studies

Energy restriction may have profound effects on reducing liver fat content. Reductions in adiposity may result in decreases in liver fat of up to 30-45% after energy-restricted, very-low calorie diets (and a threshold of 7% weight loss from baseline). Thus, while weight-neutral interventions as outlined above are important, the potential additive effect of reducing adiposity and dietary modifications could have significantly greater clinical impact on a condition with substantial comorbidity.

In a small trial of 8 participants with T2DM, Peterson et al. utilised a low calorie (1,200 kcal/d) very low-fat diet (3% fat, 43% protein and 50% carbohydrate) intervention which resulted in an 81% reduction in IHTG content, bringing liver fat to near-normal levels (i.e. reducing liver fat down to 2.2% from a basline level of 12% liver fat content).

In a comparative intervention between low-carb and low-fat diets, Haufe et al. compared two diets with a 30% calorie deficit, that were either:

• A diet of less than 90g of total carbohydrate per day
• A diet of less than 20% total fat per day

Both diet groups contained subgroups of participants with either low IHTG (~3%) or high IHTG (~15%) levels at baseline. The mean reduction in IHTG was similar in the low-carb and low-fat groups. However in participants with high baseline IHTG, the relative reduction in liver fat content was greater on the low-carb diet.

Browning et al. compared a low-carb diet group consuming <20g/d carbohydrate to an energy-restricted mixed macronutrient diet. In the two week study they had women and men consuming 1,200kcal and 1,500kcal per day, respectively. Irrespective of diet group, participants lost an average of 4.3% body weight and experienced a 43%  reduction in liver fat content. However, despite the similar weight reduction, the carbohydrate-restricted diet resulted in a significantly greater absolute reduction (5%  vs. 12% in the low-calorie and low-carb groups, respectively) in liver fat.

Conversely, Kirk et al. compared two energy-restricted diets (1,000 kcal/d deficit) over an 11-week period, with a 2-day in-patient metabolic ward study period prior to the remaining free-living period. The diets being compared were:

1. Low-carb diet: 75% fat, 10% carbohydrate, 15% protein
2. Low-fat diet: 20% fat, 65% carbohydrate, 15% protein

Both diets resulted in the same weight loss in the 48 hour in-patient metabolic study, and over the 11-weeks of the free-living intervention. However, IHTG decreases to a significantly greater degree during the 48-hr in-patient metabolic study, while over the 11-week period there was no statistically significant difference between diets (trend toward greater reduction in the low-fat group) and a mean relative reduction in IHTG of 50% from baseline in both groups.

These studies suggest that under hypocaloric conditions, short-term carbohydrate restriction may facilitate a more immediate reduction in IHTG levels, while over the longer-term reductions in IHTG may reflect the magnitude of energy restriction.

However, some more recent evidence does suggest that macronutrient manipulations may have additive effects to energy-restriction on visceral and hepatic fat. The CENTRAL intervention trial compared a low-carb (70g/d), unsaturated fat-enriched Mediterranean diet to a low-fat diet (<30% energy) over 18-months, with the first 6-months confined to dietary intervention only, before further randomisation to additional physical activity or continued diet only. The LC-Med diet resulted in a 1.56% greater absolute reduction in liver fat compared to the LF diet over 18-months, with the magnitude of effect being significantly greater with the addition of physical activity. The reductions in IHTG in both diet groups was independent of the weight loss observed in the study.

A two year intervention which compared a 'Paleo' diet (40% fat, 30% carbohydrate and 30% protein) to a low-fat diet based on the Nordic Nutrition Recommendations (25-30% fat, 55-60% carbohydrate and 15% protein), while liver fat was reduced to a greater degree at 6-months in the Paleo diet, at two years there was no significant difference between diets and both interventions had resulted in a ~50% relative reduction in hepatic fat. This is relatively comparable with other areas of research, including T2DM and general weight loss, comparing low-fat to low-carb diets, which often demonstrate greater short-term effects for low-carb diets (i.e., ~6-months) but similar effects over 12-24-months.

There has been some suggestions that ketogenic diets may be more effective than non-ketogenic low-carb diets for inducing reductions in liver fat content. Luukkonen et al. conducted a 6-day intervention using a ketogenic diet in 10 participants, with the diet constituted as:

• 1,440kcal/d
• 6% carbohydrate (<25g/d)
• 64% fat
• 28% protein

All meals were prepared and provided to participants. Bodyweight was reduced by 3% over 6 days, while the diet resulted in a relative decrease in liver fat of 31%. The study utilised stable isotope tracers to investigate pathways of fat oxidation, which demonstrated that the reductions in liver fat reflected hydrolysis of IHTG and shuttling of fatty acids into ketone production. While this provides a mechanistic support for the effects of hypocaloric ketogenic diet in partitioning fatty acids towards ketogenesis instead of IHTG synthesis, the study lacked a control group and it is unclear how this effect may compare to non-ketogenic diets. Ketogenic diets may also reduce liver volume due to rapid reductions in liver glycogen levels. To date, many of the ketogenic diet interventions are small studies over the very short term (less than one week). However, the recent mechanistic support for the effects  of ketogenesis on inducing hepatic fat reductions suggest some promise for the use of short-term ketogenic diets.

Conclusions from Energy Deficit (Hypocaloric) Studies

In sum, although there is evidence of specific metabolic effects of ketogenesis which may induce short-term reductions in liver fat, these effects require confirmation in larger comparative trials. For non-ketogenic low-carb diets, low-fat diets, and very-low-calorie diets, the evidence to date is relatively equivocal, indicating that reductions in hepatic fat in hypocaloric interventions are primarily a function of an energy deficit. While there have been differences observed in certain interventions, these differences are often only apparent over the shorter-term, or are relatively minor differences in magnitude of effect. Specific macronutrient manipulations may have some utility in the context of hypocaloric studies, however this remains to be fully elucidated; at this juncture it is evident that negative energy balance, whether short-term and more severe or moderate and longer-term, reduce liver fat content.

Nutrients of Interest: Omega-3 & Vitamin E

There is interest in the potential effects of long-chain omega-3 fatty acids, which have been shown to reduce postprandial concentrations of triglycerides, an effect which may relate to faster clearance of triglycerides from chylomicrons. Green et al. conducted an open-label intervention which investigated both the in vivo effects of omega-3 fatty acid supplementation on lipid and glucose metabolism and an in vitro cellular study to investigate the mechanistic effects of omega-3 fatty acids in liver cells. Supplementation of 4 g/d of combined EPA + DHA resulted in a significant 18% reduction in IHTG after 8-weeks, compared to placebo. Interestingly, omega-3 supplementation suppressed fasting and post-prandial DNL by 30%. Using stable isotope tracers, the study demonstrated that omega-3 supplementation resulted in a significant 20% decrease in chylomicron-TGs, which is relevant for the origins of hepatic fat given both chylomicron remnants and spillover pathways deliver fatty acids to the liver. The in vitro aspect of the study demonstrated that EPA+DHA significantly decreased intracellular TG content. EPA+DHA suppressed DNL from glucose, reducing the proportion of fatty acids derived from glucose. Of particular note in this study is that the in vitro results corroborated the in vivo effects of reduced TG and DNL in liver cell lines. Ultimately, the significant reduction in intra-hepatic triglycerides, post-prandial TGs, and hepatic DNL, suggest numerous pathways through which omega-3 supplementation may protect against liver fat accumulation, and further research may confirm the therapeutic effect of omega-3's in the treatment of NAFLD.

Vitamin E also may have promise as a supplemental intervention in NAFLD, potentially due to its action as the primary fat-soluble antioxidant. However, these interventions have largely been conducted in participants with non-alcoholic steatohepatitis (NASH), a severe form of NAFLD characterised by inflammation and ballooning of the liver, in addition to accumulated liver fat. Interventions to date utilising doses of 800IU/d vitamin E have found opposition effects: the TONIC study found no difference between vitamin E or metformin compared to placebo, while the PIVENS trial demonstrated that vitamin E was more effective than placebo in reducing inflammation and liver ballooning over 96-weeks. However, the effect of vitamin E in reducing IHTG remains to be determined.