The association between liver fat content and plasma metabolite profiles in fasting and postprandial states: an integration of a cohort study and a randomized controlled trial.

This paper combines two linked studies:

**Study 1 (Cohort):** The researchers examined whether liver fat content (measured by MRI) was associated with specific blood metabolite profiles in the fasting state and after a high-fat meal challenge. They compared people with low liver fat (<5% liver fat content) versus those with high liver fat (≥5% liver fat content, i.e., non-alcoholic fatty liver disease or NAFLD).

**Study 2 (RCT):** In a separate group of people with high liver fat, the researchers tested whether a 12-week exercise intervention (aerobic exercise, 3 sessions per week at moderate-to-vigorous intensity) could reduce liver fat content and simultaneously shift blood metabolite profiles toward those seen in people with low liver fat.

**Outcome measures:**

**Primary:** Liver fat content measured by MRI-proton density fat fraction (MRI-PDFF)

**Secondary:** Fasting and postprandial plasma metabolite profiles (targeted metabolomics measuring 186 metabolites including amino acids, acylcarnitines, glycerophospholipids, sphingolipids, and hexoses)

**Additional:** Body composition (BMI, waist circumference), insulin resistance (HOMA-IR), and lipid panel (triglycerides, HDL, LDL)

**Study 1 (Cohort):**

120 participants recruited from a university hospital in China

Age range: 18–65 years

BMI range: 18.5–35 kg/m²

Exclusion criteria: diabetes, cardiovascular disease, liver disease other than NAFLD, alcohol consumption >20 g/day for women or >30 g/day for men, pregnancy, and contraindications to MRI

60 participants with low liver fat (<5%), 60 participants with high liver fat (≥5%)

Groups were matched for age, sex, and BMI

**Study 2 (RCT):**

48 participants with high liver fat (≥5%) from the same hospital

Age range: 18–60 years

BMI range: 22–32 kg/m²

24 participants randomized to exercise, 24 to control (usual lifestyle)

44 completed the study (22 exercise, 22 control)

2 dropouts in exercise group (1 due to knee injury, 1 due to time constraints), 2 dropouts in control group (1 moved away, 1 lost interest)

**Liver fat content:**

MRI-proton density fat fraction (MRI-PDFF) using a 3.0 Tesla scanner

This is a non-invasive, quantitative measure of liver fat percentage

A value of ≥5% defines NAFLD

**Plasma metabolites:**

Blood samples collected after an overnight fast (≥10 hours)

For the postprandial challenge: a high-fat meal (800 kcal, 50% fat, 35% carbohydrate, 15% protein) was consumed, and blood was drawn at 0, 1, 2, 3, and 4 hours after the meal

Targeted metabolomics using liquid chromatography-tandem mass spectrometry (LC-MS/MS)

186 metabolites measured: 14 amino acids, 40 acylcarnitines, 90 glycerophospholipids, 15 sphingolipids, and 1 hexose (glucose)

Metabolite concentrations reported in μmol/L

**Other measures:**

Body weight and height (BMI calculated as kg/m²)

Waist circumference (measured at the midpoint between the lowest rib and the iliac crest)

Fasting blood glucose and insulin (HOMA-IR calculated as [glucose (mmol/L) × insulin (mU/L)] / 22.5)

Fasting triglycerides, HDL cholesterol, LDL cholesterol, total cholesterol

Blood pressure (systolic and diastolic)

**Study design:** This is an integration of two study designs — a cross-sectional cohort comparison (Study 1) and a parallel-group randomized controlled trial (Study 2). The cohort study establishes associations between liver fat and metabolite profiles. The RCT tests whether an intervention that reduces liver fat also changes metabolite profiles in the same direction.

**Study 1 (Cohort):**

Participants were grouped based on their liver fat content (low vs. high)

Fasting blood samples and postprandial metabolite responses were compared between groups

Statistical analysis used multivariable linear regression adjusted for age, sex, and BMI

False discovery rate (FDR) correction was applied for multiple comparisons (q-value <0.05 considered significant)

**Study 2 (RCT):**

Randomization: Participants were randomly assigned to exercise or control using a computer-generated random number sequence. Allocation concealment was achieved using sealed opaque envelopes.

Blinding: This was an open-label trial — participants and exercise trainers knew group assignment. Outcome assessors (MRI technicians and lab analysts) were blinded to group allocation.

Exercise intervention: 12 weeks of supervised aerobic exercise, 3 sessions per week, 45–60 minutes per session at 60–75% of heart rate reserve (moderate-to-vigorous intensity). Exercise modes included treadmill walking/jogging, cycling, and elliptical training.

Control group: Continued usual lifestyle, no exercise prescription

Duration: 12 weeks of intervention, with pre- and post-intervention measurements

Statistical analysis: Between-group differences were analyzed using ANCOVA with baseline values as covariates. Within-group changes were analyzed using paired t-tests. FDR correction was applied.

**What this design can prove:**

Study 1 can establish associations between liver fat and metabolite profiles but cannot prove causality (cross-sectional design)

Study 2 can demonstrate that an exercise intervention reduces liver fat and changes metabolite profiles, but cannot prove that the metabolite changes are directly caused by liver fat reduction (the exercise may have independent effects on metabolism)

**What this design cannot prove:**

Study 1 cannot determine whether high liver fat causes the metabolite changes or vice versa

Study 2 cannot separate the effects of exercise on liver fat from the effects of exercise on other tissues (muscle, adipose tissue) that also influence metabolite profiles

The open-label design in Study 2 means participants knew their group assignment, which could influence lifestyle behaviors outside the intervention

**Major methodological weaknesses:**

Small sample size in the RCT (n=44 completers) limits statistical power for detecting subtle metabolite changes

No blinding of participants or exercise trainers (open-label design)

No control for dietary intake during the intervention period (participants were asked to maintain usual diet, but no dietary monitoring was reported)

Single-center study in China limits generalizability to other populations

The high-fat meal challenge was standardized in calories and macronutrient composition but not in food type (e.g., saturated vs. unsaturated fat content)

**Study 1 (Cohort): Fasting metabolite differences between high and low liver fat groups**

**Amino acids:** Participants with high liver fat had significantly higher fasting levels of:

- Valine: 23% higher (mean 245 vs. 199 μmol/L, p<0.001, q<0.001)

- Leucine: 18% higher (mean 152 vs. 129 μmol/L, p<0.001, q<0.001)

- Isoleucine: 21% higher (mean 78 vs. 64 μmol/L, p<0.001, q<0.001)

- Phenylalanine: 12% higher (mean 68 vs. 61 μmol/L, p=0.002, q=0.008)

- Tyrosine: 19% higher (mean 62 vs. 52 μmol/L, p<0.001, q<0.001)

- Alanine: 15% higher (mean 385 vs. 335 μmol/L, p=0.003, q=0.010)

- Glutamate: 22% higher (mean 54 vs. 44 μmol/L, p<0.001, q<0.001)

**Acylcarnitines:** Several medium- and long-chain acylcarnitines were elevated in the high liver fat group:

- C12:0 (dodecanoylcarnitine): 34% higher (p=0.001, q=0.006)

- C14:0 (tetradecanoylcarnitine): 28% higher (p=0.002, q=0.009)

- C16:0 (hexadecanoylcarnitine): 19% higher (p=0.004, q=0.015)

- C18:0 (octadecanoylcarnitine): 17% higher (p=0.008, q=0.028)

**Glycerophospholipids:** Several phosphatidylcholines were lower in the high liver fat group:

- PC aa C36:4: 15% lower (p=0.001, q=0.006)

- PC aa C38:6: 12% lower (p=0.003, q=0.011)

- PC ae C40:6: 14% lower (p=0.002, q=0.009)

**Sphingolipids:** Sphingomyelin SM C26:0 was 22% higher in the high liver fat group (p=0.001, q=0.005)

**Hexose (glucose):** Fasting glucose was 8% higher in the high liver fat group (mean 5.4 vs. 5.0 mmol/L, p=0.002, q=0.008)

**Study 1 (Cohort): Postprandial metabolite differences**

After the high-fat meal, the high liver fat group showed:

- Greater increase in triglycerides: peak at 3 hours, 42% higher than low liver fat group (mean 2.8 vs. 2.0 mmol/L, p<0.001)

- Slower clearance of triglycerides: at 4 hours, triglycerides remained 35% above baseline in the high liver fat group vs. 18% above baseline in the low liver fat group (p=0.002)

- Higher postprandial levels of branched-chain amino acids (valine, leucine, isoleucine): area under the curve (AUC) 18–25% higher (all p<0.01)

- Higher postprandial acylcarnitine levels: AUC for C12:0 was 31% higher (p=0.001)

- Lower postprandial phosphatidylcholine levels: AUC for PC aa C36:4 was 13% lower (p=0.003)

**Study 2 (RCT): Effects of 12-week exercise intervention**

**Liver fat content:**

- Exercise group: decreased from mean 12.4% to 8.1% (absolute reduction of 4.3 percentage points, relative reduction of 35%, p<0.001)

- Control group: no significant change (mean 11.8% to 11.5%, p=0.42)

- Between-group difference: −3.9 percentage points (95% CI: −5.2 to −2.6, p<0.001)

**Fasting metabolites (exercise group vs. control):**

- Valine: decreased by 18% in exercise group vs. 2% increase in control (between-group difference: −20%, p=0.002)

- Leucine: decreased by 15% in exercise vs. 1% increase in control (between-group difference: −16%, p=0.004)

- Isoleucine: decreased by 17% in exercise vs. 3% increase in control (between-group difference: −20%, p=0.003)

- Glutamate: decreased by 14% in exercise vs. 2% increase in control (between-group difference: −16%, p=0.005)

- C12:0 acylcarnitine: decreased by 22% in exercise vs. 4% increase in control (between-group difference: −26%, p=0.001)

- PC aa C36:4: increased by 11% in exercise vs. 3% decrease in control (between-group difference: +14%, p=0.002)

- Fasting glucose: decreased by 6% in exercise vs. 1% increase in control (between-group difference: −7%, p=0.008)

**Postprandial metabolites (exercise group vs. control):**

- Postprandial triglyceride AUC: decreased by 28% in exercise group vs. 5% increase in control (between-group difference: −33%, p<0.001)

- Postprandial branched-chain amino acid AUC: decreased by 15–20% in exercise vs. no change in control (all p<0.01)

- Postprandial acylcarnitine AUC: decreased by 18–25% in exercise vs. no change in control (all p<0.01)

**Body composition and insulin resistance:**

- BMI: decreased by 1.8 kg/m² in exercise vs. 0.2 kg/m² increase in control (between-group difference: −2.0 kg/m², p<0.001)

- Waist circumference: decreased by 4.2 cm in exercise vs. 0.5 cm increase in control (between-group difference: −4.7 cm, p<0.001)

- HOMA-IR: decreased by 32% in exercise vs. 5% increase in control (between-group difference: −37%, p<0.001)

**Liver fat reduction:** A 12-week exercise program (3 sessions per week, 45–60 minutes at moderate-to-vigorous intensity) reduced liver fat by an average of 4.3 percentage points — from 12.4% to 8.1%. This is a 35% relative reduction. To put this in perspective, a 4.3 percentage point reduction in liver fat is roughly equivalent to moving from "moderate NAFLD" to "mild NAFLD" on the clinical spectrum. For someone with a starting liver fat of 10%, this would bring them to 6.5% — still above the 5% threshold for NAFLD, but substantially closer to normal.

**Metabolite changes:** The exercise intervention shifted metabolite profiles toward those seen in people with low liver fat. The branched-chain amino acids (valine, leucine, isoleucine) decreased by 15–20% — roughly the same magnitude as the difference between the high and low liver fat groups at baseline. The acylcarnitine C12:0 decreased by 22%, and the phosphatidylcholine PC aa C36:4 increased by 11%. These changes suggest improved mitochondrial fatty acid oxidation and reduced incomplete fat burning.

**Postprandial triglyceride response:** The 28% reduction in postprandial triglyceride AUC after exercise means that after a high-fat meal, blood triglycerides peaked lower and cleared faster. This is clinically meaningful because postprandial hypertriglyceridemia is an independent risk factor for cardiovascular disease.

**Insulin resistance:** The 32% reduction in HOMA-IR (a measure of insulin resistance) is substantial. For comparison, metformin (a common diabetes drug) typically reduces HOMA-IR by 20–30% in people with prediabetes. This suggests the exercise intervention had a powerful effect on metabolic health beyond just liver fat reduction.

**Acknowledged by authors:**

Small sample size in the RCT (n=44 completers) — may