Dietary pattern modifies the risk of MASLD through metabolomic signature

Background & Aims The EAT-Lancet Commission in 2019 advocated a plant-centric diet for health and environmental benefits, but its relation to metabolic dysfunction-associated steatotic liver disease (MASLD) is unclear. We aimed to discover the metabolite profile linked to the EAT-Lancet diet and its association with MASLD risk, considering genetic predisposition. Methods We analyzed data from 105,752 UK Biobank participants with detailed dietary and metabolomic information. We constructed an EAT-Lancet diet index and derived a corresponding metabolomic signature through elastic net regression. A weighted polygenic risk score for MASLD was computed from associated risk variants. The Cox proportional hazards model was employed to estimate hazard ratios (HRs) and 95% CIs for the risk of MASLD (defined as hospital admission or death). Results During a median follow-up period of 11.6 years, 1,138 cases of MASLD were documented. Participants in the highest group for the EAT-Lancet diet index had a multivariable HR of 0.79 (95% CI 0.66–0.95) for MASLD compared to the lowest group. The diet's impact was unaffected by genetic predisposition to MASLD (p = 0.42). Moreover, a robust correlation was found between the metabolomic signature and the EAT-Lancet diet index (Pearson r = 0.29; p <0.0001). Participants in the highest group for the metabolomic signature had a multivariable HR of 0.46 (95% CI 0.37–0.58) for MASLD, in comparison to those in the lowest group. Conclusions Higher intake of the EAT-Lancet diet and its associated metabolite signature are both linked to a reduced risk of MASLD, independently of traditional risk factors. Impact and implications: Our analysis leveraging the UK Biobank study showed higher adherence to the EAT-Lancet diet was associated with a reduced risk of metabolic dysfunction-associated steatotic liver disease (MASLD). We identified a unique metabolite signature comprising 81 metabolites associated with the EAT-Lancet diet, potentially underlying the diet's protective mechanism against MASLD. These findings suggest the EAT-Lancet diet may offer substantial protective benefits against MASLD.

EAT-Lancet diet index (Pearson correlation coefficient = 0.84, P <0.0001).Therefore, if participants completed dietary assessments multiple times, the study employed both the earliest dietary consumption date and the averaged dietary assessment EAT-Lancet diet index to maximize follow-up time.

Metabolomics measurement
A high-throughput nuclear magnetic resonance (NMR) metabolomics platform was used to analyze baseline plasma samples from approximately 280,000 randomly selected UK Biobank participants.Detailed protocols on sample collection and metabolomic quantification are presented elsewhere [2][3][4].This simultaneously quantified 251 metabolic biomarkers (170 directly measured and 81 ratios of these).
The biomarkers investigated in this study encompass multiple metabolic pathways.These include lipoprotein lipids in 14 subclasses, fatty acids and their compositions, as well as various low-molecular weight metabolites, such as amino acids, ketone bodies, and glycolysis metabolites quantified in molar concentration units.In the present study, we included a subset of 170 metabolic biomarkers (Table S2) that were directly measured for subsequent analyses.Metabolites with less than 10% missing were included, and missing data for each metabolite were imputed using half of the minimum measured value.Metabolites were highly correlated within each category (Fig. S2).

Polygenic risk score for MASLD
Based on the selected SNPs, the polygenic risk score (PRS) for MASLD was calculated as follows: PRS = β1 ×SNP1 + β2 × SNP2 + …+βn × SNPn, where SNPn is the risk allele number of each SNP.A higher PRS indicates a higher genetic predisposition to the disease.Participants were further divided into low (quintile 1), medium (quintiles 2-4), and high (quintile 5) PRS groups based on the quintiles of the PRS.

Statistical analysis
Models were adjusted for age, sex, BMI, smoking status, alcohol intake, education levels, Townsend deprivation index, total energy consumption, physical activity, individual history of the disease (cancer, cardiovascular disease, hypertension and diabetes), fasting duration, spectrometer, MASLD-PRS, first 10 principal components of ancestry, and genotype measurement batch.We also estimated the joint association of EAT-Lancet diet index and PRS with MASLD risk by creating a combined variable based on categories of PRS and EAT-Lancet diet index (12 categories).The reference group consisted of individuals with the highest risk combination, which involved having the lowest EAT-Lancet diet index and the highest PRS.In addition, we performed mediation analyses using the method proposed by Lange et al. [5] to evaluate the mediation of the association between EAT-Lancet diet and risk of MASLD by metabolite signature score or selected metabolites.a Values are hazard ratios (95% confidence interval) (all such values).b P for trend was calculated across quartiles using multivariable Cox regression models.
Model 1 was adjusted for age, sex, and BMI.Model 2 was additionally adjusted for total energy intake, smoking status, alcohol intake, educational level, Townsend deprivation index, physical activity, hypertension, diabetes, cancer, cardiovascular disease, fasting duration, spectrometer, first 10 principal components of ancestry, and genotype measurement batch.b Hazard ratios (95% confidence interval) (all such values).c P for trend was calculated across quartiles using multivariable Cox regression models.A two-tailed p <0.05 was considered statistically significant.Model 1 was adjusted for age, sex, and BMI.Model 2 was additionally adjusted for total energy intake, smoking status, alcohol intake, educational level, Townsend deprivation index, physical activity, hypertension, diabetes, cancer, cardiovascular disease, fasting duration, spectrometer, MASLD-PRS, first 10 principal components of ancestry, and genotype measurement batch.b Hazard ratios (95% confidence interval) (all such values).c P for trend was calculated across quartiles using multivariable Cox regression models.A two-tailed p <0.05 was considered statistically significant.Model 1 was adjusted for age, sex, and BMI.Model 2 was additionally adjusted for total energy intake, smoking status, alcohol intake, educational level, Townsend deprivation index, physical activity, hypertension, diabetes, cancer, cardiovascular disease, fasting duration, spectrometer, MASLD-PRS, first 10 principal components of ancestry, and genotype measurement batch.Model 2 + mutual adjustment was additionally included both EAT-Lancet diet index and the metabolic signature of EAT-Lancet diet simultaneously in the model2 to examine association independence.a The initial model was same as the model 2 in Table2.b Hazard ratios (95% confidence interval) (all such values).c P for trend was calculated across quartiles using multivariable Cox regression models.A two-tailed p <0.05 was considered statistically significant.Sensitivity analysis 1: further adjusted for baseline albumin, alanine aminotransferase, and gamma-glutamyltransferase.

Categories
Sensitivity analysis 2: further adjusted for waist circumference.Sensitivity analysis 3: excluded participants with less than twice dietary assessment, n = 64,373.Sensitivity analysis 4: excluded participants with less than 2 years of follow-up, n = 105,660.Sensitivity analysis 5: expanded the definition of MASLD, n = 105,670, case = 1,724.

Fig. S1 .
Fig. S1.Selection of study participants in the UK Biobank Cohort.

Table 2 .
A two-tailed p <0.05 was considered statistically significant.Joint association of the EAT-Lancet diet index and genetic susceptibility with risk of MASLD.Adjusted for age, sex, BMI, total energy intake, smoking status, alcohol intake, educational level, Townsend deprivation index, physical activity, hypertension, diabetes, cancer, cardiovascular disease, fasting duration, spectrometer, first 10 principal components of ancestry, and genotype measurement batch.Correlation between the EAT-Lancet diet and the corresponding metabolite signature score in repeated assessments.Adjusted HR (95% CI) for per 1-score EAT-Lancet diet index increment and risk of MASLD stratified by potential risk factors.All models were adjusted the same as the Model 2 in the table 2. A two-tailed p <0.05 was considered statistically significant.

Table S1 . Criteria for the EAT-Lancet diet index constructed to evaluate the EAT-Lancet diet and example of food items in the UK Biobank a Food components in the EAT- Lancet diet index b
Milk, dairy smoothie, flavored milk, yogurt, ice-cream, cheesecake, milk-based pudding, other milk-based pudding, low fat hard cheese, hard cheese, soft cheese, blue cheese, low fat cheese spread, cheese spread, cottage cheese, feta cheese, mozzarella cheese, goat's cheese,

Table S2 . Metabolic biomarkers and abbreviations included the UK Biobank
Other liver diseases including liver diseases as well as liver cancer mentioned in ICD-9 and ICD-10 coding. a

Table S7 . Association between repeated assessments of metabolic signature with MASLD risk a
a Obtained by using multivariable Cox regression model.

Table S9 . Sensitivity analyses for associations between EAT-Lancet diet index and risk of MASLD a
MASLD, metabolic dysfunction-associated steatotic liver disease.