Very low energy ketogenic diet vs. Mediterranean diet for MASLD: Superior steatosis reduction in a randomised pilot study

Abstract

Background & Aims

Weight loss is the cornerstone of treatment for metabolic dysfunction-associated steatotic liver disease (MASLD). This pilot study compared the efficacy and safety of a ketogenic very low energy diet (VLED) vs. a Mediterranean diet (MD) in improving hepatic steatosis and liver histology in individuals with overweight or obesity and MASLD.

Methods

We conducted a pilot randomised controlled trial in adults with histologically confirmed MASLD and a BMI of 27-35 kg/m². Participants were assigned to either a 12-week VLED (3,151 kJ/day) or MD program (8,950 kJ/day) and monitored for 24 weeks. The VLED group received low-dose semaglutide (0.5 mg/weekly) from week 13 for weight maintenance. The primary outcome was change in hepatic steatosis by MRI liver-fat-fraction at 12 weeks. Secondary outcomes included total body weight loss (TBWL) and change in liver histology over 24 weeks.

Results

The VLED group (n = 14) achieved significantly greater reduction in MRI-liver-fat-fraction (-77% relative reduction [IQR 51–88]) compared to the MD group (n = 11, -14% [IQR 0–30], p <0.01). The VLED also produced greater TBWL at week 12 (-13% [IQR -17 to -9] vs. -4% [-4.4 to -0.2], p <0.01). At 24 weeks, the VLED/semaglutide group maintained a -14% TBWL (IQR -17 to -10) from baseline vs. -3% TBWL (IQR -4 to 0) in the MD group. Liver histology improved in both groups, with greater improvements in the VLED group (NAFLD activity score reduction VLED: -2 [IQR -3.5 to -2] vs. MD: -1 [IQR -1 to -1], p <0.01).

Conclusions

The ketogenic VLED resulted in significantly greater reduction in hepatic steatosis and weight loss compared to a MD. The very low energy diet is widely available, easily accessible and should be more commonly considered for patients with MASLD and overweight or obesity.

Impact and implications

This pilot randomised controlled trial provides the first direct comparison between a ketogenic very low energy diet (VLED) and a Mediterranean diet (MD) for the treatment of metabolic dysfunction-associated steatotic liver disease (MASLD), demonstrating greater reductions in hepatic steatosis (77% vs. 14%) and body weight (13% vs. 4%) with the VLED. The growing burden of people who are overweight and obese with early-stage MASLD means that effective dietary weight loss interventions with proven hepatic and metabolic benefits are urgently required. The accessibility, effectiveness, and feasibility of the VLED in clinical settings suggest that it may represent a valuable therapeutic option for selected patients with MASLD and overweight or obesity.

Clinical trial registration

www.anzctr.org.au trial ID: ACTRN12623000756628.

Graphical abstract

Highlights

• •Ketogenic VLED achieved 77% hepatic steatosis reduction vs. 14% with a MD.
• •69% of VLED participants normalised liver fat within 12 weeks vs. 0% with MD.
• •VLED produced greater weight loss (13% vs. 4%) and histological improvements.
• •Low-dose semaglutide maintained VLED benefits from the end of the diet through to week 24.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most common liver disease worldwide, characterised by excessive hepatic fat accumulation and strong associations with obesity and the metabolic syndrome.¹ MASLD carries a significant disease burden, ranging from hepatic complications (cirrhosis, decompensated liver disease and hepatocellular carcinoma) to extrahepatic events (including type 2 diabetes mellitus [T2DM] and major adverse cardiovascular events).^2,3 Weight loss remains the cornerstone of MASLD management. However, treatment practices are heterogeneous, and compensatory hormonal changes make diet-induced weight loss difficult to achieve and maintain.^4,5 Guidelines recommend at least 5% weight loss to improve hepatic steatosis, and 10% or greater for hepatic fibrosis regression.⁶ Weight loss is primarily achieved through an energy deficit facilitated by dietary energy restriction. The Mediterranean diet (MD) is often recommended for MASLD as it can improve liver fat content and cardiometabolic health, even without significant weight loss,⁷ with some arguing that the MD may be easier to maintain in the long term.⁶

The medical management of obesity utilises a ketogenic very low energy diet (VLED), which induces rapid weight loss, is associated with remission of T2DM, and is often used short-term before bariatric surgery to reduce hepatic steatosis.[8], [9], [10] The carbohydrate restriction on the VLED leads to fatty acid oxidation in the liver, resulting in the production of ketones which reduce appetite through the suppression of appetite-regulating hormones, thus helping patients to maintain the strict energy restriction.¹¹ Expert obesity guidelines recommend the VLED to achieve target weight loss, before transitioning to a portion-controlled weight maintenance diet to help prevent weight regain.¹² Due to hormonal and metabolic changes after weight loss which increase hunger, many obesity experts argue that pharmacotherapy is required to prevent weight regain during the maintenance phase.^4,13,14 This approach has been shown to improve weight maintenance post-VLED vs. control and is a practice followed within local weight management services.[15], [16], [17] Semaglutide is a glucagon-like peptide-1 receptor agonist (GLP-1RA) and has been shown to induce weight loss, improve markers of hepatic steatosis and inflammation, and, more recently, hepatic fibrosis.[18], [19], [20] GLP-1RAs have recently become the preferred medication in weight management; however, they are expensive at high doses, and use of low-dose semaglutide presents a pragmatic approach to weight maintenance in clinical practice.

In this study, we aimed to determine the effectiveness of a 12-week ketogenic VLED in reducing hepatic steatosis in overweight and obese individuals with MASLD, compared to a MD. Secondary aims included evaluating the transition to a weight maintenance phase (with low-dose semaglutide commenced in the VLED group and persistence of MD principles assessed in the MD group), and changes in body weight, body composition, cardiometabolic profile, and liver histology at 24 weeks.

Patients and methods

Study design

We conducted a single-centre, unblinded, randomised controlled trial, comparing a 12-week ketogenic VLED vs. prescribed MD in overweight and obese patients with MASLD. Participants were randomised 1:1, with stratification for T2DM and sex, using computerised randomisation managed by REDCap electronic data capture tools.^21,22 The study design is outlined in Fig. 1. This study was approved by the Human Research Ethics Committee of St Vincent’s Hospital Melbourne (HREC 277/18, Project ID Number 50048), and registered in the Australian Registry of Clinical trials (www.anzctr.org.au ACTRN12623000756628).

Fig. 1.

Study design and intervention timeline.

Participants were randomised to either a VLED or MD for a 12-week weight loss phase. During the subsequent 12-week weight maintenance phase (weeks 12–24), the VLED group transitioned to an energy-controlled diet with weekly subcutaneous semaglutide 0.5 mg, while the MD group continued with an energy-controlled diet alone. MD, Mediterranean diet; VLED, very low energy diet.

The primary outcome was change in hepatic steatosis on MRI liver-fat-fraction (MRI-LFF) at the end of the 12-week diet. Secondary outcomes included change in body weight and metabolic profile as well as paired liver biopsies at baseline and week 24. Participants and participant-facing trial staff were not blinded, but intervention allocation was concealed from outcome assessors.

Participants

Patients were screened from liver clinics at St Vincent’s Hospital Melbourne between April 2021 and July 2023 to recruit adults aged 18-70 years with a BMI of 27-35 kg/m² and MASLD, defined by elevated alanine aminotransferase (ALT) and/or gamma-glutamyltransferase (GGT), ≥10% hepatic steatosis on MRI and Metavir fibrosis score of ≥F1 on liver histology. Key exclusions: excessive alcohol (>20 g/day men, >10 g/day women), concomitant liver disease, current GLP-1RA or insulin use, recent weight loss (>5% total body weight [TBW]) or previous bariatric surgery. Participants continued their usual medications throughout the study period.

Intervention

All participants had dietitian consultations at baseline and week 6, with telephone support at weeks 2, 4 and 9. Both groups were advised to perform 150-300 min of weekly exercise per EASL guidelines.⁶

Ketogenic VLED: Two meal replacement products (Optifast VLCD shake (Nestlé) provided) plus one very low-carbohydrate, low-fat meal daily (3,151 kJ total, 40% protein, 27% fat, 31% carbohydrates) were advised for the VLED group. A booklet containing nine recipes and resource guides outlining suitable foods and beverages was provided.

Prescribed MD: Modified MD providing up to 8,950 kJ daily (17% protein, 44% fat [50% of total fat from monounsaturated fats], 30% carbohydrate) was advised, adjusted from a previously published ad libitum MD to allow participants to remain eucaloric.²³ A two-week rotating meal plan with recipes was provided alongside essential components of the MD, such as extra virgin olive oil, tinned fish, baked beans and almonds.

Maintenance phase (Weeks 13-24): The dietitian provided individualized energy requirement counselling from week 13 onwards, but no meal plans were provided. The VLED group commenced semaglutide (Ozempic, Novo Nordisk) at week 13 (0.25 mg weekly via subcutaneous injection for 4 weeks, then 0.5 mg weekly for 8 weeks). The MD group were encouraged to maintain MD dietary habits.

Adherence

Dietary adherence was assessed via 3-day food diaries (patient record on paper) at baseline, week 12 and 24. Nutritional analysis was conducted using FoodWorks version 10 (Xyris, Brisbane, QLD). The VLED group were assessed for presence of ketosis (serum 3-hydroxybutyrate measured at week 2 and 6). Urinary ketone testing strips were provided to participants to self-monitor compliance. MD adherence was assessed using the PREDIMED MEDAS (Mediterranean Diet Adherence Screener) score, with >7 indicating adequate compliance.²⁴ At week 24, PREDIMED MEDAS score was calculated from food diaries. Medication adherence was monitored through self-reporting, pharmacy records and returned medication devices.

Outcomes

Hepatic steatosis was measured using MRI liver fat fraction (MRI-LFF) on a 3T magnet (Siemens Skyra and Vida) following the Siemens LiverLab protocol (software version XA50). Liver stiffness measurement (LSM) and controlled attenuation parameter (CAP) obtained using Fibroscan (Echosens, France). Liver biopsies obtained under ultrasound guidance by radiologists and liver histology was reviewed by a pathologist using the NASH Clinical Research Network (CRN) scoring.²⁵ Body composition was assessed via whole body dual-energy x-ray absorptiometry scan (GE Lunar iDXA Pro, enCORE software version 18) and physical activity was measured using an activPAL (PAL technologies, Scotland).

Sample size

Based on previous MD studies in patients with MASLD, which found an MRI liver fat reduction of 12-25%, we anticipated a 15% mean reduction in liver fat on the MD with a SD of 8%.^7,26 This was a pilot study and 20 participants, with 10 in each arm would provide an 80% power with 0.05 alpha to detect a difference of 20%. We aimed to recruit 24 participants to allow for 20% drop out.

Statistical analysis

Data normality was assessed using the Shapiro-Wilk test. Normally distributed data are presented as mean ± SD, non-parametric data as median (IQR). Between-group comparisons used Fisher's exact test or chi-squared test, t-test, or Wilcoxon rank sum as appropriate. Signed rank test was used for within-group differences, and Spearman correlation to assess the relationship between two continuous variables. Per-protocol analysis was performed, including participants with available data at each time point; no imputation was performed for missing data. Intention-to-treat analysis is provided in the supplementary material (Appendix E). Subgroup and adjusted analyses were not performed due to the small sample size of this pilot study. All statistical analysis was performed using Stata version 18 (StataCorp. 2023; Texas USA).

Results

Study population

Twenty-five participants were randomised, with 24 completing the 12-week intervention and 23 completing the full 24-week study (Fig. 2). The cohort was predominantly middle-aged and Caucasian (76%, n = 19), with a mean BMI of 31.7 ± 3.1 kg/m². Baseline median MRI-LFF was 21% (IQR 14.95–28.7), and only one participant had advanced liver fibrosis on histology (F3-4) (Table 1).

Fig. 2.

CONSORT flow diagram of participant enrolment, allocation and follow-up.

Flow of participants through the trial from screening to final analysis.MD, Mediterranean diet; MRI-LFF, MRI liver-fat-fraction; VLED, very low energy diet.

Table 1.

Baseline characteristics of study participants randomised to VLED or MD interventions.

Characteristics	Cohort N = 25	VLED (n = 14)	MD (n = 11)	p value
Male sex, n (%)	14 (56%)	9 (64%)	5 (45%)	0.35
Age, y (IQR)	49 (42, 57)	45 (35, 59)	50 (44, 57)	0.53
Type 2 diabetes mellitus, n (%)	2 (8%)	1 (7%)	1 (7%)	0.86
Hypertension, n (%)	7 (28%)	4 (29%)	3 (27%)	0.94
Dyslipidaemia	6 (24%)	3 (21%)	3 (27%)	0.73
Ethnicity, n (%)				0.12
Oceanian Caucasian	15 (60%)	7 (50%)	8 (73%)
South + Central Asian	5 (20%)	5 (36%)	0 (0%)
South + East European	4 (16%)	2 (14%)	2 (18%)
Pacific Islander	1 (4%)	0 (0%)	1 (9%)
Smoking, n (%)	4 (16%)	4 (29%)	0 (0%)	0.05
Daily energy intake (kJ)	7,454 (5,960–9,653)	8,944 (6,326–11,607)	7,308 (6,116–10,534)	0.54
Daily step count (total steps)	6,075 (4,909–7,839)	5,629 (4,856–7,534)	6,348 (5,235–8,144)	0.62
Sedentary time (min/day)	537 (439–641)	529 (489–621)	546 (401–716)	0.79
MRI liver-fat-fraction, % (IQR)	21.2 (14.95–28.7)	19.4 (16.1–29.6)	21.4 (13.8–28)	0.84
Liver NAS (IQR)	4 (4–5)	4 (4–5)	4 (4–5)	0.34
Liver fibrosis score, n (%)				0.53
Grade 1-2	23 (96%)	12 (92%)	11 (100%)
Grade 3-4	1 (4%)	1 (8%)	0
Liver stiffness measurement, kPa (IQR)	5.6 (4.9–6.4)	5.7 (4.6–7.3)	5.6 (4.9–6)	0.64
CAP, dB/m (IQR)	322 (295–333)	327 (295–341)	317 (292–325)	0.32
ALT, IU/L (IQR)	51 (37–67)	46.5 (37–58)	52 (30–99)	0.66
GGT, IU/L (IQR)	55 (37–114)	49.5 (30–86)	67 (40–140)	0.11
Body weight, kg (SD)	91.9 (±13.6)	90.4 (±13.1)	93.9 (±14.5)	0.44
BMI, kg/m2 (SD)	31.7 (±3.1)	31.2 (±3.2)	32.3 (±3.1)	0.34
Waist circumference, cm (SD)
Women	108.6 (±7.7)	107.7 (±10.0)	109.4 (±6.2)	0.86
Men	108.8 (±8.9)	105.6 (±8.3)	114.6 (±7.3)	0.053
HOMA-IR score (IQR)	4.1 (2.8–6.4)	5.1 (3.2–6.6)	3.8 (2.3–6.3)	0.34

Data presented as median (IQR) for non-parametric continuous variables, mean (±SD) for parametric continuous variables, and number (percentage) for categorical variables. p values were calculated using Fisher's exact test or chi-square test for categorical variables, Wilcoxon rank sum test for non-normally distributed continuous variables, and t-test for normally distributed continuous variables. p <0.05 considered statistically significant. ALT, alanine aminotransferase; CAP, controlled attenuation parameter; GGT, gamma-glutamyltransferase; HOMA-IR, homeostatic model assessment for insulin resistance; MD, Mediterranean diet; NAS, NAFLD activity score; VLED, very low energy diet.

Primary outcome: liver fat fraction (0-12 weeks)

The VLED resulted in a significant improvement in MRI-LFF compared to the MD at 12 weeks (Table 2, Fig. 3). The VLED achieved an absolute reduction in MRI-LFF of -18% (IQR -24 to -10), representing a relative reduction of -77% (IQR -88 to -55) from baseline (p <0.01), with 69% (n = 9/13) achieving normalisation (≤5%) of hepatic steatosis. Comparatively, the MD group had a non-significant absolute reduction of -5% (IQR -8 to 0) from baseline (p = 0.15) (Table 2). Similar results were observed using intention-to-treat analysis of the entire cohort (supplementary material).

Table 2.

Comparison of clinical outcomes between VLED and MD interventions at 12 and 24 weeks.

	Week 12			Week 24
	VLED	MD	p value∧	VLED	MD	p value∧
Relative reduction in MRI liver-fat-fraction (MRI-LFF)(%)	-77 (-88 to -55)∗	-14 (-30 to 0)	<0.01	-64 (-83 to -57)∗	-4 (-22 to 2)	<0.01
Absolute reduction in LFF (% steatosis)	-18 (-24 to -10)∗	5 (-8 to 0)	<0.01	-16 (-25 to -9)∗	-0.6 (-5 to 0.6)	<0.01
Normalisation of hepatic steatosis (≤5%)	69% (n = 9)	10% (n = 1)	<0.01	46% (n = 6)	0% (n = 0)	<0.01
Weight (kg)	-12 (±5.8)∗	-3.5 (±3.9)∗	<0.01	-13 (±4.4)∗	-2 (±5.4)	<0.01
Total body weight loss (%)	-13 (±6)∗	-4 (±4)∗	<0.01	-15 (±5)∗	-2 (±5)	<0.01
Clinically significant weight loss (>10% TBWL)	71% (n = 10)	10% (n = 1)	<0.01	79% (n = 11)	10% (n = 1)	<0.01
BMI (kg/m2)	-4 (±1.8)∗	-1.2 (±1.3)∗	<0.01	-4.5 (±1.7)∗	-0.6 (±1.8)	<0.01
LSM (kPa)	-1.1 (-2 to 0)∗	-0.25 (-1.3 to 0.25)	0.32	-0.9 (-1.6 to 0.1)	-0.1 (-1.5 to 0.5)	0.56
CAP (dB/m)	-82.5 (-123 to -41.5)∗	-34.5 (-48 to 4)	0.049	-85 (-98 to -40)∗	-16 (-37 to 11)	<0.01
ALT (IU/L)	-26 (-3 to 39)∗	-13.5 (-45 to -5)∗	0.88	-24 (-43 to -6)∗	-18 (-30 to -5)∗	0.61
GGT (IU/L)	-25 (-32 to -14)∗	-13 (-19 to -1)	0.21	-23 (-31 to -4)∗	-13 (-18 to -8)	0.38
Albumin (g/L)	1 (0 to 1)∗	1 (-1 to 3)	0.59	0 (-3 to 3)	-2 (-4 to 1)	0.34
HOMA-IR score	-1.8 (-2.9 to -1.3)∗	-1.2 (-4.4 to -0.6)∗	0.74	-1.5 (-2.8 to 0.1)∗	-0.1 (-0.2 to 1.2)	0.36
HbA1C (%)	-0.4 (-0.6 to -0.1)∗	-0.3 (-0.3 to -0.1)∗	0.35	-0.5 (-0.6 to -0.3)∗	-0.3 (-0.3 to 0.1)	0.055
LDL-cholesterol (mmol/L)	0 (-0.2 to 0.35)	0.3 (-0.05 to 0.7)	0.35	0 (-0.3 to 0.4)	0.1 (-0.4 to 0.7)	0.74
HDL-cholesterol (mmol/L)	0.05 (-0.02 to 0.2)	-0.01 (-0.05 to 0.2)	0.86	0.19 (0.05 to 0.3)∗	0.03 (-0.01 to 0.14)	0.44
Triglycerides (mmol/L)	-0.3 (-0.6 to -0.1)∗	-0.8 (-1.2 to 0.1)	0.78	-0.4 (-0.6 to 0.1)	-0.2 (-1.2 to 0.1)	0.82
Diastolic BP (mmHg)	-9 (-11 to -2)∗	-1 (-9 to 4)	0.14	-6.5 (-11 to -5)∗	-5 (-11 to -1)	0.3
Systolic BP (mmHg)	-8 (-17 to 6)	7 (-10 to 19)	0.12	-14 (-18 to -6)∗	-3 (-11 to 4)	0.09

Data presented as median (IQR) for non-parametric variables, mean (±SD) for parametric variables, or % (n) for categorical variables and represents change from baseline at each time point. Per protocol analysis performed. p values calculated for between-group comparisons were performed using Wilcoxon rank-sum test for continuous variables and chi-square test for categorical variables. Within-group changes from baseline were assessed using Wilcoxon signed-rank test. p <0.05 considered statistically significant. ∗p <0.05 in-group change from baseline for continuous variables. ^∧p value VLED vs. MD. ALT, alanine aminotransferase; BP, blood pressure; CAP, controlled attenuation parameter; GGT, gamma-glutamyltransferase; HbA1c, glycated haemoglobin; HDL, high-density lipoprotein; HOMA-IR, homeostatic model assessment for insulin resistance; LDL, low-density lipoprotein; LFF, liver-fat-fraction; LSM, liver stiffness measurement; MD, Mediterranean diet; TBWL, total body weight loss; VLED, very low energy diet.

Fig. 3.

Comparison of liver-related biomarkers and metabolic parameters between VLED and MD over 24 weeks.

Time course of (A) MRI liver-fat-fraction, (B) total body weight (C) ALT, (D) gamma- GGT, (E) LSM, and (F) CAP score in participants randomised to VLED or MD interventions. Data presented as median with interquartile range. p values were calculated using Wilcoxon rank sum test. p <0.05 considered statistically significant. ∗p <0.05 for between-group comparison (VLED vs. MD) at that timepoint. ALT, alanine aminotransferase; CAP, controlled attenuation parameter; GGT, gamma-glutamyltransferase; LSM, liver stiffness measurement; MD, Mediterranean diet; VLED, very low energy diet.

Secondary outcomes (0 -12 weeks)

Weight loss

Greater weight loss was achieved with the VLED compared to the MD at week 12 (-13% vs. -4% TBWL; p <0.01), with both groups achieving statistically significant weight loss compared to baseline (p <0.01 for VLED and p = 0.01 for MD) (Table 2, Fig. 3). Clinically significant weight loss (>10% TBWL) was achieved in 71% (n = 10/14) of VLED participants vs. 10% (n = 1/10) in the MD group (p <0.01). There was a strong positive correlation between TBWL and relative liver fat reduction (Spearman's rho = 0.68, R² = 0.46, p <0.01).

Liver and metabolic parameters

The VLED was associated with significant improvement in GGT, LSM and CAP scores (GGT p <0.01, LSM p = 0.03; CAP p <0.01 from baseline; Table 2, Fig. 3). Both groups showed improvements in ALT (VLED p <0.01; MD p <0.01), with no between-group difference (p = 0.88).

Both diets significantly improved markers of insulin sensitivity at week 12 (HOMA-IR [homeostatic model assessment for insulin resistance]: VLED p <0.01, MD p = 0.04; HbA1c [glycated hemoglobin]: VLED: p <0.01, MD p = 0.02 from baseline; Table 2), with no significant between-group differences. The VLED also led to a significant reduction in triglycerides (p = 0.03) and diastolic blood pressure (p = 0.02) at week 12 vs. baseline (Table 2).

Dietary adherence

The VLED group showed good adherence to the diet, with 13/14 participants able to demonstrate ketosis during the diet intervention period (12 via serum ketones, 1 via urinary ketones). All MD participants demonstrated good adherence to the prescribed diet (week 12 MD score >7). Analysis of participant food diaries confirmed VLED participants achieved significant energy restriction (4,791 kJ/day [IQR 3,745–5,377] at week 12 vs. 8,944 kJ/day [IQR 6,326–11,607] at baseline, p <0.01), while MD participants remained eucaloric (7,440 kJ/day [IQR 6,038–10,788] at week 12 vs. 7,308 kJ/day [IQR 6,116–10,534] at baseline, p = 0.51), with macronutrient composition changes in line with dietary prescriptions (supplementary material).

Maintenance phase (12-24 weeks)

The VLED/semaglutide group maintained a stable MRI-LFF between week 12 and 24 (-0.1% [IQR -1.8 to 3.7], p = 0.75), while the MD group showed a small but significant increase in MRI-LFF (+3.3% [IQR 1.7–6.5], p = 0.03) (Fig. 3). Both groups remained weight stable during this maintenance period (VLED/semaglutide: -0.8 kg [SD 4.2], p = 0.33; vs. MD: +1.9 kg [SD 2.9], p = 0.09). No change was seen in Fibroscan LSM, CAP, or liver enzymes between week 12 and 24 in either group (Table 2).

Compared to baseline, at week 24 the VLED/semaglutide group maintained a significant -15% TBWL (SD 5.5, p <0.01), with 79% (n = 11/14) achieving clinically significant (>10% TBWL) weight loss. The VLED/semaglutide group also maintained a significant relative reduction in LFF of -64% (IQR -83 to -57, p <0.01). The MD group showed no significant changes from baseline in weight (-2% TBWL [SD 5.3], p = 0.41) or LFF (-4% [IQR -22 to 2], p = 0.41).

The VLED/semaglutide group increased their dietary energy intake by 1,208 kJ/day (IQR -224 to 3,680, p = 0.047) between week 12 and 24 during the maintenance phase, while the MD group maintained consistent intake (-256 kJ/day [IQR -1,022 to 505], p = 0.68). There was a trend towards a lower energy intake compared to baseline in the VLED/semaglutide group (-517 kJ/day [IQR -3,119 to 710], p = 0.13) at week 24, but this was not statistically significant, and no change was seen in the MD group (191 kJ/day [IQR 7–1,209], p = 0.31). MD scores remained >7 at week 24 in 7/8 MD participants. Both groups’ macronutrient intake at week 24 returned towards their baseline levels (supplementary material).

Liver histology: improvement at week 24 vs. baseline

Paired liver biopsies obtained at baseline and week 24 (n = 12 VLED, n = 9 MD) showed significant improvements in NAFLD activity score in both groups, with greater benefits seen in the VLED/semaglutide arm (Fig. 4). The VLED/semaglutide group demonstrated significant improvements in hepatic steatosis (-1.5 stage [IQR -2 to -1], p <0.01) and hepatocellular ballooning (-1 stage [IQR -1 to 0], p = 0.02). The MD also resulted in an improvement in steatosis (-1 stage [IQR -1 to 0], p = 0.01), but to a lesser degree than the VLED (p = 0.02) (Fig. 4). No change in histological fibrosis stage was noted within either group (VLED: F score change 0 [IQR -1 to 0], p = 0.18; vs. MD: 0 [IQR 0–0], p = 0.16), though participants achieving >10% TBWL regardless of intervention were more likely to show fibrosis improvement than those who did not (40% [n = 4/10] vs. 0% [n = 0/11], p = 0.02).

Fig. 4.

Histological assessment of liver biopsy samples showing changes in NAS components and fibrosis between baseline and 24 weeks.

Comparison of (A) total NAS, (B) steatosis score, (C) ballooning score, (D) inflammation score, and (E) fibrosis score in participants receiving MD or VLED interventions. Data shown at baseline and week 24 for each treatment group. Individual data points are overlaid on bar graphs showing median and interquartile range. p values were calculated using Wilcoxon rank-sum test for between group comparisons and Wilcoxon signed-rank test for within-group changes. p <0.05 considered statistically significant. ∗p <0.05 for within-group change from baseline to week 24; ∗∗p <0.05 for between-group comparison (VLED vs. MD) at week 24. MD, Mediterranean diet; NAS, NAFLD activity score; VLED, very low energy diet.

Body composition

At week 12, the VLED resulted in reductions from baseline in fat mass (p <0.01), including both visceral (p <0.01) and subcutaneous adipose tissue (p <0.01), and a reduction in lean mass (p <0.01), which was noted in both the appendicular and trunk regions (Table 3). Lean mass loss accounted for 24% of the TBWL achieved at week 12. The MD group showed modest fat mass reduction from baseline (p = 0.03), specifically visceral adipose tissue (p = 0.01), with no change in lean mass (p = 0.09). Across all participants weight loss was predominantly fat-driven, with TBWL strongly correlated with fat mass reduction (Spearman's rho = 0.94, R² = 0.88, p <0.01) but also associated with lean mass loss (rho = 0.66, R² = 0.44, p <0.01).

Table 3.

Changes in body composition from baseline at 12 and 24 weeks in participants receiving VLED or MD interventions.

	Week 12			Week 24
	VLED	MD	p value∧	VLED	MD	p value∧
Body fat (%)	-4.2 (-7.8 to -3.1)∗	-2.0 (-2.4 to -0.3)∗	0.01	-7.4 (-9.8 to -3.2)∗	-0.9 (-2.3 to 0.0)	<0.01
Fat mass (kg)	-8.4 (-11.7 to -5.8)∗	-2.9 (-3.8 to -1.7)∗	0.01	-11.1 (-13.0 to -5.7)∗	-2.0 (-3.5 to -0.1)∗	<0.01
Lean mass (kg)	-2.7 (-3.7 to -2.0)∗	-0.2 (-1.5 to -0.2)	<0.01	-2.6 (-3.6 to -1.4)∗	-0.4 (-0.7 to 0.7)	0.01
Appendicular	-1.7 (-2.4 to -1.1)∗	-0.1 (-0.6 to 0.1)	<0.01	-1.8 (-2.3 to -0.7)∗	-0.1 (-0.4 to 0.1)	<0.01
Trunk	-0.8 (-1.5 to -0.3)∗	-0.2 (-0.4 to 0.1)	0.15	-0.8 (-1.0 to -0.5)	0.1 (-0.5 to 0.4)	0.16
VAT mass (g)	-661.7 (-1,066.0 to -513.0)∗	-393.8 (-626.1 to -116.9)∗	0.15	-691.5 (-873.6 to -547.2)∗	-325.5 (-430.6 to -71.0)∗	0.01
SAT mass (g)	-637.2 (-864.8 to -491.3)∗	-97.9 (-136.0 to -28.9)	0.01	-809.6 (-967.2 to -651.2)∗	-175.9 (-344.1 to 6.9)	<0.01

Data presented as median (IQR) and represents change from baseline at each time point. Per protocol analysis performed. p values were calculated using Wilcoxon rank-sum test for between group comparisons and Wilcoxon signed-rank test for within group changes. p <0.05 considered statistically significant. ∗p <0.05 for within-group change from baseline. ^∧p values for between-group comparisons (VLED vs. MD) at each time point. MD, Mediterranean diet; SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue; VLED, very low energy diet.

During the maintenance phase, lean mass remained unchanged in both groups from week 12. The MD group showed a trend toward fat mass gain (p = 0.051), while the VLED/semaglutide group maintained stable fat mass (p = 0.09). No changes were measured in physical activity in either group at any time point during the trial (supplementary material).

Adverse events

Five adverse events were recorded during the study, including three liver-biopsy-related complications. The biopsy-related adverse events (3 of 25 participants) were predominantly minor (pain, minor bleeding) and resolved without sequelae. Mild gastrointestinal side effects were reported in both diet groups, predominantly constipation in the VLED group and bloating in the MD group (supplementary material). No additional side effects were reported with the use of semaglutide.

Discussion

This pilot randomised controlled trial demonstrates that a 12-week ketogenic VLED is significantly more effective than a prescribed MD in reducing hepatic steatosis and achieving clinically significant weight loss in overweight/obese individuals with MASLD. The VLED achieved a 77% relative reduction in MRI-LFF compared to 14% with the MD (p <0.01), with 69% of VLED participants normalising hepatic steatosis (<5% MRI-LFF) vs. none in the MD group. Weight loss was also significantly greater with the VLED (13% vs. 4% TBWL, p <0.01), with 71% achieving clinically significant weight loss (>10% TBWL) compared to only 10% of the MD group. These benefits were maintained through to week 24 with low-dose semaglutide, and were accompanied by greater histological improvements including reduced hepatocellular ballooning.

We compared the VLED to MD, which is recommended in international guidelines.⁶ While MD did not achieve statistically significant reductions in hepatic steatosis by MRI-LFF, it did demonstrate improvements in ALT, HOMA-IR, and HbA1c. Participants in the MD group remained largely eucaloric, which limited significant weight change. Histological improvement in steatosis in the MD despite modest weight loss was consistent with previous studies and suggests benefits beyond weight reduction, likely related to the anti-inflammatory effects of the MD(7).

The data highlight the importance of clinically significant weight loss in managing patients with overweight/obesity and MASLD, with the VLED an effective strategy to achieve this. The magnitude of weight loss and hepatic fat reduction after the 12-week VLED exceeded that reported in previous lifestyle intervention studies and was comparable to outcomes observed in phase II trials of high-dose GLP-1RAs.²⁷ In MASH (metabolic dysfunction-associated steatohepatitis) with F2–3 fibrosis, semaglutide 0.4 mg daily resulted in 13% weight loss at 72 weeks, while in MASH cirrhosis, semaglutide 2.4 mg weekly for 48 weeks produced an 8.8% weight loss and a 31% relative reduction in hepatic steatosis on MRI-proton density fat fraction.^19,28 Pemvidutide (GLP-1RA/glucagon dual receptor agonist) induced a 69% relative reduction in MRI liver fat after 12 weeks in patients with MASLD and BMI ≥28 kg/m².²⁹ Comparatively, the VLED offers a viable non-pharmacological alternative as a safe, rapid, and effective initial weight loss strategy in MASLD.

The histological findings at 24 weeks are noteworthy and demonstrate benefits of the VLED-semaglutide approach extending beyond steatosis reduction to improvements in hepatocellular ballooning, a marker of liver cell injury, likely reflective of the substantial weight loss achieved with the VLED. The maintenance phase with semaglutide may have consolidated these benefits by preventing weight regain, potentially complemented by indirect hepatic effects of GLP-1RAs, including reduced de novo lipogenesis and anti-inflammatory properties, that appear to be at least partially independent of weight loss.^30,31 Regardless, the net clinical effect of the combined intervention on histology represents a pragmatically relevant outcome. Participants achieving >10% TBWL were more likely to have fibrosis improvement, regardless of intervention, reinforcing the importance of substantial weight loss in MASLD. This aligns with current guidelines recommending ≥10% TBWL and supports a dose-response relationship between weight loss and histological improvement. The importance of sustained weight loss is further demonstrated by the phase III ESSENCE study, in which high-dose semaglutide (2.4 mg weekly) resulted in 11% weight loss and fibrosis regression in 37% of participants, compared with 22% in the placebo group after 72 weeks.²⁰ Our study did not detect a change in fibrosis; however, it was not powered to do so, and the follow-up interval was not optimised to capture this outcome.

Lean mass loss occurred with the VLED but not with the MD; however, this occurred to an expected degree, with 25-30% loss of lean mass typical for dietary weight loss interventions.^32,33 This is thought to be at least partly reflective of the reduction in axial load required to be carried, necessitating less lean mass.³² While lean mass loss can be a concern in patients with chronic liver disease, this was a young cohort without cirrhosis, with no functional impact on physical activity.³⁴

Mediterranean-style diets are optimal for MASLD; however, they are difficult to translate into clinical practice, as replication of the diets that have shown benefit in research studies requires comprehensive education from a dietitian.^7,23,35,36 In contrast, the VLED can be utilised by any trained healthcare provider, including general practitioners, nurses and allied health practitioners.^12,37 VLED programs using meal replacement products are accessible and can be readily administered in an outpatient setting. The VLED guidelines are simple to follow and evidence-based, utilising ketosis for appetite suppression, assisting adherence.¹¹ The greater metabolic benefits observed with the VLED vs. MD may reflect the greater weight loss achieved, making it a better option in patients for whom >5% weight loss is desirable.

The maintenance of weight loss and hepatic steatosis reduction observed in the VLED/semaglutide group through week 24 addresses a critical issue in managing patients with MASLD and overweight or obesity after weight loss – weight regain. While weight change during the maintenance phase did not differ significantly between groups, the VLED group had achieved substantially greater weight loss (>10% TBWL), and therefore faced greater physiological drive towards weight regain due to metabolic adaptation.^4,13 Despite this, and despite increased energy intake and macronutrient changes observed in the VLED group during the maintenance phase, low-dose semaglutide effectively maintained the benefits achieved with the VLED, providing a practical weight maintenance strategy. While GLP-1RAs alone are effective for weight loss, side-effect tolerability and high cost can be prohibitive to more widespread use.³⁸ Low-dose semaglutide has been utilised in the medical management of obesity when higher doses are not required to achieve weight loss goals, or not tolerated due to side effects or from a cost perspective.^39,40 This pragmatic use of the VLED to induce clinically significant weight loss followed by low-dose semaglutide to prevent weight regain presents a more affordable alternative to high-dose GLP-1RA monotherapy with lower medication costs and improved tolerability.

While there are initial costs associated with this approach (meal replacement products and semaglutide), they should be weighed against the significant long-term burden of progressive liver disease. If this strategy can prevent progression to advanced fibrosis, the long-term cost-effectiveness might be favourable despite higher upfront costs. A formal health economic evaluation, which is beyond this pilot study's scope, will be crucial to guide implementation and health policy.

This study has several important strengths that differentiate it from previous work. This is the first randomised trial to directly compare the ketogenic VLED to the MD for the management of MASLD, a clinically relevant comparison with both approaches recommended for metabolic health but without any direct head-to-head comparisons to guide treatment selection. Unlike previous small uncontrolled VLED studies in MASLD, which used heterogeneous diagnostic criteria, we employed paired liver biopsies with standardised histological scoring (NASH-CRN), demonstrating improvements not only in steatosis but also in hepatocellular ballooning – a key marker of liver injury.[41], [42], [43] Another key strength is the comprehensive assessment of liver outcomes, including both non-invasive (MRI-LFF, Fibroscan) and invasive (histology) measures. Both groups received matched levels of dietitian support throughout, minimising contact bias that can confound dietary comparisons. Finally, the sequential strategy of VLED induction followed by low-dose GLP-1RA maintenance represents a novel, pragmatic approach to sustained weight management in MASLD that has not been previously evaluated.

Several limitations should be acknowledged. The relatively low prevalence of T2DM and advanced fibrosis in our cohort may limit generalizability to populations with more severe metabolic dysfunction or liver disease; however, the significant improvements in early-stage MASLD seen may be beneficial in preventing disease progression. The 24-week duration may not fully capture the long-term sustainability of the observed benefits, and the study occurred during COVID outbreaks with periods of prolonged community lockdown in Melbourne, which may have impacted our results.⁴⁴ This was a pilot study with a small sample size, so while it was designed with sufficient power to detect the primary outcome, study power is limited for some secondary analyses. The small sample size precluded subgroup analysis based on factors such as sex and smoking status; however, the magnitude of the primary outcome difference (77% vs. 14% relative reduction in MRI-LFF) substantially exceeds what could be plausibly attributed to baseline differences between groups.

Our findings have important clinical implications. The superior efficacy of the VLED approach, combined with its good tolerability, minimal adverse events, wide applicability and low cost, supports the consideration of VLED as a first-line intervention for patients with overweight/obesity and MASLD. The ability to achieve and maintain substantial weight loss with a dual diet and medication approach addresses a critical unmet need in MASLD management. Future studies with a larger cohort, including more advanced liver disease and a longer period of follow-up are warranted.

A ketogenic 12-week VLED is more effective than a prescribed MD in reducing hepatic steatosis and achieving clinically meaningful weight loss in patients with obesity and MASLD, with greater improvements in liver enzymes, liver histology and cardiometabolic parameters. The VLED followed by low-dose semaglutide offers an effective strategy to address both the initial weight loss and subsequent weight maintenance challenges in MASLD and demonstrates the benefits of adapting the medical weight loss approach to the management of overweight and obese patients with MASLD.

Abbreviations

ALT, alanine aminotransferase; CAP, controlled attenuation parameter; CRN, Clinical Research Network; GGT, gamma-glutamyl transferase; GLP-1RA, glucagon-like peptide-1 receptor agonist; LSM, liver stiffness measurement; MASLD, metabolic dysfunction-associated steatotic liver disease; MD, Mediterranean diet; MRI-LFF, MRI liver-fat-fraction; T2DM, type 2 diabetes mellitus; TBWL, total body weight loss; TBW, total body weight; VLED, very low energy diet.

Authors’ contributions

Ann Farrell: Conceptualisation, Methodology, Formal analysis, Investigation, Data curation, Writing - Original Draft, Visualisation, Project administration, Funding acquisition. Tonya Paris: Investigation, Data curation, Writing - Review & Editing, Project administration. Evelyn B Parr: Methodology, Investigation, Resources, Formal analysis, Data curation, Writing - Review & Editing. Elena George: Methodology, Resources, Writing - Review & Editing. Jessica Howell: Methodology, Data curation, Writing - Review & Editing, Visualisation. Catherine Croagh: Methodology, Writing - Review & Editing, Visualisation. Tom Sutherland: Investigation, Resources. Mark Page: Investigation, Data curation. Penny McKelvie: Investigation, Resources, Data curation. Alexander Thompson: Conceptualisation, Methodology, Writing - Review & Editing, Visualisation, Funding acquisition. Marno Ryan: Conceptualisation, Methodology, Writing - Review & Editing, Visualisation, Funding acquisition, Project administration, Supervision.

Data availability

The datasets generated and analysed during the current study are not publicly available due to participant privacy, but are available from the corresponding author upon reasonable request subject to ethical approval and data use agreements. Analytical methodologies and research materials are described in detail in the methods section and supplementary materials.

Writing assistance

During the preparation of this work the author(s) used Grammarly in order to improve language and readability. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

Financial support

This research was supported by a research grant from St Vincent’s Hospital Melbourne Research Endowment Fund (Grant #91940). AF was supported by an Australian Government Research Training Program (RTP) Scholarship. In-kind contributions of food products were accepted from Cobram estates, Australian Almond Board, Safcol and Heinz.

Conflicts of interest

The authors declare the following financial and non-financial competing interests: AF, TP, EBP, EG, JH, CC, TS, MP, PM, AT: no competing interests. MR: has received speaking honorarium from Novo Nordisk.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article: