Dietary index for gut microbiota, plasma metabolome, and risks of metabolic dysfunction-associated steatotic liver disease and other chronic liver diseases

Abstract

Background

The dietary index for gut microbiota (DI-GM) is a newly proposed metric for assessing diet quality linked to gut microbiota. However, prospective evidence is scarce on the associations between DI-GM and adverse liver outcomes.

Methods

The DI-GM was calculated by averaging the intakes of 12 foods and nutrients. Elastic net regression was performed to identify metabolites associated with DI-GM and metabolic signature reflecting higher adherence to DI-GM was constructed. Cox proportional hazards regression and mediation analyses were employed to explore the potential associations and mechanisms.

Results

This prospective cohort study included 168,456 participants from the UK Biobank. Compared to participants with DI-GM scores of 0–3, those scoring ≥ 6 presented 22% lower risk of MASLD (HR = 0.78, 95% CI = 0.68–0.90). Metabolic signature for DI-GM and dietary index beneficial to gut microbiota (BDI-GM) were also inversely correlated with MASLD. Similar inverse correlations between DI-GM and BDI-GM and the risks of other chronic liver diseases were identified. Furthermore, phenotypic age, body mass index, metabolic score, inflammatory score, and metabolic signature significantly mediated the relationship between DI-GM and MASLD. No significant interactions were observed between DI-GM and polygenic risk score of hepatic steatosis, and the associations between DI-GM and adverse liver outcomes persisted regardless of genetic risk.

Conclusions

Higher adherence to DI-GM significantly correlates with reduced risks of MASLD and other chronic liver diseases, independent of genetic susceptibility. And the apparent mediating effects of five indices highlight the role of aging, obesity, metabolic disorders, inflammation, and metabolomic alterations in the association between DI-GM and MASLD. Further research is warranted to evaluate the utility of metabolic signatures in metabolic profile monitoring and risk stratification.

Impact and implications

This large-scale cohort study first demonstrates that higher adherence to a gut microbiota-beneficial diet (DI-GM) is associated with a lower risk of MASLD and other chronic liver diseases, independent of genetic susceptibility. The estimated population attributable fractions, while derived from observational data and requiring cautious interpretation, suggest that a substantial portion of liver disease cases in the study population might be linked to suboptimal DI-GM adherence. These findings underscore the importance of integrating gut microbiome health into public health strategies for liver disease prevention, offering a practical approach to reduce disease burden at both individual and population levels. The DI-GM-associated metabolic signature represents a candidate objective biomarker meriting evaluation in future studies for its potential in early risk assessment. Mediation analyses further reveal that a diet promoting healthy gut microbiota may reduce MASLD risk by maintaining gut microbiota homeostasis, decelerating biological aging, ameliorating obesity, attenuating metabolic disorders, alleviating inflammation, and altering metabolome. Collectively, this study generates important hypotheses and provides a rationale for future interventional research to determine whether promoting DI-GM-aligned diets can effectively reduce liver disease risk at the population level.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12986-026-01145-w.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly termed non-alcoholic fatty liver disease (NAFLD), is characterized by hepatic steatosis – detected via imaging or histology – alongside the presence of at least one cardiometabolic risk factor and the exclusion of other causes of hepatic fat accumulation [1]. With a global prevalence of approximately 30%, MASLD is the most common cause of chronic liver disease, and the burden of MASLD increases worldwide in parallel with the escalating rates of obesity and metabolic syndrome [2]. The disease progression can lead to cirrhosis, hepatocellular carcinoma (HCC), and liver-related mortality, posing substantial public health and economic challenges [3]. However, effective strategies for MASLD prevention and management remain elusive, underscoring the urgent need to identify actionable and non-invasive approaches to mitigate MASLD risk [4].

Emerging evidence indicates the interplay between gut microbiota and pathogenesis of MASLD, suggesting the potential of taking gut microbiota as a target for MASLD treatment and prevention [4, 5]. Additionally, studies have established a strong correlation between dietary patterns and gut microbiota composition [6]. Specifically, Kase et al. developed the novel dietary index for gut microbiota (DI-GM) through a comprehensive review of longitudinal studies investigating diet-gut microbiota association in adults [7]. These findings indicate that DI-GM positively correlates with urinary enterodiol and enterolactone levels, proposing DI-GM as a standardized tool for evaluating diet quality in relation to maintaining a healthy gut microbiota [7]. However, the longitudinal association between DI-GM and MASLD and the biological mechanisms underlying the association remain poorly understood.

Previous studies have linked aging, obesity, metabolic disorders, inflammation, and metabolomic alterations to MASLD [8–12], and gut microbiota has been associated with these factors [12–16]. These interconnections suggest that a diet promoting healthy gut microbiota may reduce MASLD risk by maintaining gut microbiota homeostasis, decelerating biological aging, ameliorating obesity, attenuating metabolic disorders, alleviating inflammation, and altering the metabolome. Therefore, phenotypic age (Phenoage), body mass index (BMI), metabolic score (MS), inflammatory score (IS), and metabolic signature were employed to elucidate the potential mechanisms underlying the correlation between DI-GM and MASLD.

This study aims to investigate the associations between DI-GM and the incidence of MASLD and other chronic liver diseases, including cirrhosis, liver cancer, and liver-related mortality. Given that metabolic signature based on metabolomics provides objective metrics for assessing the internal metabolic response to dietary shifts, we also evaluated the correlations between metabolic signature for DI-GM and the risks of adverse liver outcomes. To uncover biological mechanisms, we explored the mediating roles of Phenoage, BMI, MS, IS, and metabolic signature in the associations between DI-GM and adverse liver outcomes. Furthermore, we examined whether those associations are modified by genetic risk of hepatic steatosis, as determined by the polygenic risk score of hepatic steatosis (HS-PRS). We also estimated the public health implications utilizing the population attributable fraction (PAF) and examined the relationship between DI-GM and magnetic resonance imaging (MRI)-derived liver proton density fat fraction (PDFF).

Methods

Study population

The UK Biobank (UKB) is a large-scale prospective cohort study comprising over 500,000 individuals from 22 assessment centers. After providing informed consent at baseline, participants completed self-administered, touch-screen questionnaires detailing sociodemographic information and lifestyle factors. Trained nurses conducted anthropometric measurements, including weight, height, and waist circumference. Additionally, non-fasting blood samples were collected for future analyses across a wide range of biomarkers.

Participants were included in this analysis if they completed at least one online 24-hour dietary recall questionnaire and had available data to calculate DI-GM. Individuals were excluded if they : (1) had a diagnosis of cancer or liver disease at baseline, including chronic liver disease, viral hepatitis, cirrhosis, and liver cancer; (2) reported implausible energy expenditure (< 500 kcal or > 5000 kcal); (3) lacked necessary covariate data.

Assessment of dietary index for gut microbiota

Utilizing a validated web-based 24-hour dietary assessment tool, the average dietary intake was calculated for participants who completed multiple questionnaires between 2009 and 2012. Total energy and nutrient intakes were determined based on the fifth edition of McCance and Widdowson’s “The composition of food”. The NOVA classification system was employed to differentiate processed from unprocessed foods. Following the methodology established by Kase et al. [7], 12 foods and nutrients were selected as DI-GM components. Comprehensive information on the components, field IDs, and scoring criteria for the DI-GM is available in Table S1. The overall DI-GM ranged from 0 to 12, with both beneficial and unfavorable components ranging from 0 to 8. DI-GM scores were categorized into four groups: 0–3, 4, 5, and ≥ 6.

Ascertainment of outcomes

The primary outcome was MASLD and the secondary outcomes were other chronic liver diseases, including cirrhosis, liver cancer, and liver-related mortality. Based on the 10th edition of International Classification of Diseases (ICD-10) and the Expert Panel Consensus Statement, study outcomes were identified using linkage with National Health Service hospital admission data and death register records. Details concerning ICD-10 codes used to define the outcomes are provided in Table S2. Hospital admissions and mortality data were available until 31 December 2021, and follow-up was censored on this date. The time-to-event was calculated from the date of the last dietary assessment to the date of liver disease diagnosis, death, loss to follow-up, or censorship, whichever occurred first. MRI-PDFF was employed as a supplementary approach for MASLD ascertainment (supplementary materials 1). MASLD was defined as an MRI liver PDFF > 5% [17].

Metabolomic profiling and metabolic signature for DI-GM

Metabolomic profiling of 249 plasma metabolites was conducted in a subsample of over 110,000 individuals employing a high-throughput 1 H-NMR platform (Nightingale Health; biomarker quantification version 2020) [18, 19]. Details about sample collection and metabolomic quantification are available elsewhere [20]. To derive metabolic signatures for DI-GM, we leveraged metabolomic profiles from the UK Biobank, utilizing baseline data as the training set and the first repeated assessment data as the testing set. Elastic net regression models were applied to identify metabolites associated with DI-GM, and the metabolic signatures were constructed as the weighted sum of these selected metabolites (Supplementary Materials 1).

Mediation analyses

Mediation analyses were employed to explore whether the relationships between DI-GM and the risks of adverse liver outcomes could be explained by metabolomic alterations, aging, obesity, inflammation, and metabolic disorders. Firstly, we constructed five indices, including metabolic signature, Phenoage, BMI, IS, and MS (supplementary materials 1). We then estimated (a) the total effect of DI-GM on each outcome, (b) the effect of every mediator on the outcome, and (c) the effect of DI-GM on each mediator, allowing decomposition of the total effect into direct and indirect components. The mediation analyses were conducted under the assumptions of correct model specification and absence of unmeasured confounding for the mediator-outcome associations.

Polygenic risk score of hepatic steatosis

Genetic variants that correlate with steatotic liver disease were retrieved to establish HS-PRS (supplementary materials 1). The overall HS-PRS varied from 0 to 4 and was converted into tertiles. Subsequently, we estimated the relative excess risk (RERI) and attributable proportion (AP) for additive interactions between DI-GM and HS-PRS.

Statistical analysis

Multivariable Cox proportional hazards models estimated hazards ratios (HRs) and 95% confidence intervals (CIs) for the associations of DI-GM and metabolic signature with the risk of MASLD and other chronic liver diseases. The effects of DI-GM and metabolic signature on adverse liver outcomes were explored through three models, with potential confounders adjusted (supplementary materials 1). DI-GM was analyzed categorically (0–3, 4, 5, 6) and continuously (per 1-unit increment) based on previous studies [21, 22]. Proportional hazards were checked using Schoenfeld residuals. We also assessed the associations between each DI-GM component and adverse liver outcomes. To evaluate dose-response relationships, restricted cubic spline (RCS) regression models were fitted within Cox hazard regressions. Furthermore, multivariable logistic regression analyzed the association between DI-GM and MRI-PDFF. To explore modification by genetic MASLD risk, analyses were stratified by HS-PRS tertiles and the interactions between DI-GM and HS-PRS were tested. Details of PAF, an epidemiological measure widely utilized to assess the public health impact of exposures in populations, were provided in supplementary materials 1.

To ensure the robustness of our findings, we conducted extensive sensitivity analyses by : (1) excluding incident cases within the first two years of follow-up to mitigate reverse causality; (2) removing participants who completed only one questionnaire and re-running the models; (3) applying multiple imputation for missing covariate data followed by completed-cases analysis; (4) constructing a weighted DI-GM score and exploring the associations between the weighted DI-GM score and adverse liver outcomes aiming at enhancing estimation efficiency and accuracy; (5) applying a Fine and Gray competing risks regression model with all-cause mortality treated as a competing risk. Subgroup analyses were also conducted (supplementary materials 1).

All statistical analyses were performed utilizing R software (version 4.3.3, R Project for Statistical Computing), with two-tailed P-values and significance set at P < 0.05.

Results

Characteristics of the participants

From the UKB cohort of 502,412 participants, 291,463 were excluded for missing data required for calculating DI-GM and 10,956 for diagnosis of cancer or prevalent liver disease at baseline. Of the remaining 199,993 participants, 883 were excluded for misreported total energy intake and 30,654 for incomplete covariate information (Fig. 1). This left 168,456 participants (33.53%) for the main analysis, including 91,751 (54.47%) women and 76,705 (45.53%) men, with a mean age of 55.90 (SD: 7.95) years. Over a median follow-up of 12.4 years (IQR: 11.8–13.2), we recorded 1,519 MASLD cases, 663 cirrhosis, 257 liver cancer, and 233 liver-related mortality. Compared to those with DI-GM scores of 0–3, participants scoring ≥ 6 were more likely to be non-smokers, less deprived, non-obese, and physically active. Participants scoring ≥ 6 also exhibited more favorable cardiometabolic profiles than those scoring 0–3, including higher high-density lipoprotein cholesterol, lower triglycerides, and lower prevalence of diabetes and hypertension at baseline (Table S3).

Fig. 1.

Flowchart of participants screening for this study. DI-GM, dietary index for gut microbiota; MASLD, metabolic dysfunction-associated steatotic liver disease

Associations between DI-GM and adverse liver outcomes

An apparent inverse association between DI-GM and the risk of MASLD was observed in Table 1. After completely adjusting for potential confounders, each one-point increase in DI-GM was associated with a reduced MASLD risk (HR = 0.94, 95% CI = 0.91–0.97). Compared to participants with DI-GM scores 0–3, those scoring ≥ 6 exhibited 22% lower risk of MASLD (HR = 0.78, 95% CI = 0.68–0.90). A linear dose-response relationship between DI-GM and MASLD was confirmed by RCS (P-nonlinearity = 0.766, Fig. S1). The inverse association between DI-GM and the risk of MASLD was maintained in a subsample of participants who underwent abdominal MRI (Table S8). Additionally, the dietary index beneficial to gut microbiota (BDI-GM) was associated with reduced risk of MASLD (HR = 0.91, 95% CI = 0.87–0.94).

Table 1.

Associations between DI-GM and the risks of MASLD and other chronic liver diseases

	Events	Model1a		Model2b		Model3c
		HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
MASLD
DI-GM group
0–3	405	1.00(Ref)		1.00(Ref)		1.00(Ref)
4	351	0.83 (0.72–0.96)	0.013	0.91 (0.79–1.05)	0.204	0.91 (0.79–1.05)	0.188
5	340	0.74 (0.64–0.86)	<0.001	0.86 (0.74–0.99)	0.038	0.85 (0.74–0.98)	0.029
>=6	423	0.59 (0.52–0.68)	<0.001	0.79 (0.69–0.91)	0.001	0.78 (0.68–0.90)	0.001
Continuous (per 1 unit)	1519	0.88 (0.86–0.91)	<0.001	0.95 (0.92–0.98)	0.001	0.94 (0.91–0.97)	<0.001
Trend test			<0.001		0.001		<0.001
Beneficial to gut microbiota	1519	0.85 (0.82–0.88)	<0.001	0.92 (0.89–0.95)	<0.001	0.91 (0.87–0.94)	<0.001
Unfavorable to gut microbiota	1519	1.00 (0.96–1.06)	0.852	1.03 (0.98–1.08)	0.294	1.04 (0.98–1.09)	0.187
Liver cirrhosis
DI-GM group
0–3	184	1.00(Ref)		1.00(Ref)		1.00(Ref)
4	162	0.83 (0.67–1.03)	0.090	0.89 (0.72–1.10)	0.268	0.89 (0.72–1.10)	0.271
5	152	0.71 (0.57–0.88)	0.002	0.79 (0.63–0.98)	0.030	0.79 (0.64–0.98)	0.032
>=6	165	0.49 (0.40–0.61)	<0.001	0.61 (0.49–0.76)	<0.001	0.61 (0.49–0.76)	<0.001
Continuous (per 1 unit)	663	0.86 (0.82–0.90)	<0.001	0.91 (0.87–0.95)	<0.001	0.91 (0.87–0.95)	<0.001
Trend test			<0.001		<0.001		<0.001
Beneficial to gut microbiota	663	0.80 (0.76–0.85)	<0.001	0.85 (0.81–0.90)	<0.001	0.85 (0.80–0.90)	<0.001
Unfavorable to gut microbiota	663	1.06 (0.98–1.14)	0.139	1.07 (1.00–1.16)	0.066	1.07 (0.99–1.16)	0.084
Liver cancer
DI-GM group
0–3	65	1.00(Ref)		1.00(Ref)		1.00(Ref)
4	59	0.82 (0.58–1.17)	0.273	0.85 (0.60–1.22)	0.384	0.85 (0.60–1.22)	0.381
5	56	0.67 (0.47–0.96)	0.030	0.72 (0.50–1.03)	0.073	0.72 (0.50–1.03)	0.072
>=6	78	0.59 (0.42–0.82)	0.002	0.68 (0.49–0.95)	0.025	0.68 (0.48–0.95)	0.024
Continuous (per 1 unit)	257	0.89 (0.82–0.96)	0.002	0.92 (0.85–0.99)	0.035	0.92 (0.85–0.99)	0.035
Trend test			0.001		0.018		0.018
Beneficial to gut microbiota	257	0.87 (0.80–0.95)	0.002	0.91 (0.83–0.99)	0.031	0.90 (0.83–0.99)	0.028
Unfavorable to gut microbiota	257	0.96 (0.85–1.09)	0.538	0.98 (0.87–1.11)	0.763	0.98 (0.86–1.11)	0.751
Liver-related mortality
DI-GM group
0–3	63	1.00(Ref)		1.00(Ref)		1.00(Ref)
4	56	0.83 (0.58–1.19)	0.317	0.88 (0.61–1.26)	0.480	0.88 (0.62–1.27)	0.499
5	53	0.71 (0.49–1.02)	0.062	0.78 (0.54–1.13)	0.186	0.79 (0.55–1.14)	0.213
>=6	61	0.52 (0.36–0.74)	<0.001	0.63 (0.44–0.91)	0.013	0.65 (0.45–0.93)	0.019
Continuous (per 1 unit)	233	0.86 (0.79–0.93)	<0.001	0.90 (0.83–0.98)	0.011	0.90 (0.83–0.98)	0.017
Trend test			<0.001		0.010		0.016
Beneficial to gut microbiota	233	0.80 (0.73–0.88)	<0.001	0.84 (0.77–0.92)	<0.001	0.85 (0.77–0.93)	0.028
Unfavorable to gut microbiota	233	1.05 (0.92–1.19)	0.467	1.08 (0.95–1.23)	0.266	1.06 (0.93–1.21)	0.387

^aModel 1 was adjusted for age and sex

^bModel 2 was adjusted for Model 1+ TDI, ethnicity, smoking status, drinking status, physical activity, obesity, diabetes, hypertension, low HDL-C, and high TG

^cModel 3 was adjusted for Model 2+ energy

DI-GM, dietary index for gut microbiota; MASLD, metabolic dysfunction-associated steatotic liver disease; HR, hazard ratios; CI, confidence interval; TDI, townsend deprivation index; HDL-C, high-density lipoprotein cholesterol; TG, total triglycerides

Similar inverse associations were detected between DI-GM and BDI-GM and other chronic liver diseases (Table 1). Specifically, participants with DI-GM scores ≥ 6 had significantly lower risks of cirrhosis (HR = 0.61, 95% CI = 0.49–0.76), liver cancer (HR = 0.68, 95% CI = 0.48–0.95), and liver-related mortality (HR = 0.65, 95% CI = 0.45–0.93) than those scoring 0–3. The BDI-GM was associated with reduced risks of cirrhosis (HR = 0.85, 95% CI = 0.80–0.90), liver cancer (HR = 0.90, 95% CI = 0.83–0.99), and liver-related mortality (HR = 0.85, 95% CI = 0.77–0.93). RCS indicated non-linear relationships between BDI-GM and cirrhosis (P-nonlinearity = 0.036, Fig. S1).

Identification of DI-GM-related metabolic signature and associations between metabolic signature and liver outcomes

Metabolites such as fatty acids and lipoprotein lipids were identified to significantly correlate with DI-GM at baseline (Fig. S2A) and the findings from first repeat assessment data showed a high degree of concurrence (Fig. S2B). Strong associations among metabolites were observed in Fig. S3. Subsequently, 89 metabolites that were significantly associated with DI-GM (baseline data: r = 0.25, P < 2.2 × 10^− 16; repeat assessment: r = 0.27, P < 2.2 × 10^− 16; Fig. 2B and C) were selected from the elastic net regression model to establish the total metabolic signature. And relative lipoprotein lipid concentrations, lipoprotein subclasses, and fatty acids predominated in the metabolic signatures (Fig. 2). In the metabolic signature for DI-GM, polyunsaturated fatty acids to total fatty acids percentage made notable contributions to the positive coefficient, while omega-6 fatty acids to total fatty acids percentage played a significant role in influencing the reverse coefficient of the metabolic signature (Fig. 3).

Fig. 2.

The metabolic signature for DI-GM: flow chart of study design and analytical approach. (A) The training and testing procedures of a metabolic signature for DI-GM. (B) Correlation between DI-GM and the metabolic signature using baseline data (training set). (C) Correlation between DI-GM and the metabolic signature using first repeat data (testing set). DI-GM, dietary index for gut microbiota; Phenoge, phenotypic age; BMI, body mass index; MS, metabolic signature; IS, inflammatory score; MASLD, metabolic dysfunction-associated steatotic liver disease

Fig. 3.

Correlations between 249 metabolites and DI-GM, BDI-GM, UDI-GM, and incident adverse liver outcomes. Presented from top to bottom are the coefficients (weights) of metabolites in the metabolic signature and associations with DI-GM and adverse liver outcomes risks. Coefficients for associations with DI-GM indicate the standard deviation (SD) changes in metabolites per score increment of DI-GM. Coefficients for the adverse liver outcomes risks ln(hazard ratio) of adverse liver outcomes risks per SD increment in metabolites. Asterisks denote the significance level of associations (*P < 0.05 and **Bonferroni-corrected P < 0.05). CE, cholesteryl ester; FC, free cholesterol; L, large; M, medium; S, small; L, large; VLDL, very LDL; XL, very large; XS, very small; XXL, especially large

Associations of metabolic signature with the risks of MASLD and other chronic liver diseases using UK Biobank baseline data are presented in Table S4. Each one-point increase in the metabolic signature was associated with reduced risk of MASLD (HR = 0.57, 95% CI = 0.46–0.70). The inverse correlation remained statistically significant even after additionally adjusting for DI-GM.

Mediation analysis

Table S5 and Fig. 4 illustrate the mediating roles of Phenoage, BMI, MS, IS, and metabolic signature in the relationships between DI-GM and adverse liver outcomes. Notably, the remarkable mediating effect of metabolic signature (proportion of mediation: 36.00%, 95% CI = 19.40–79.00%), BMI (proportion of mediation:23.5%, 95% CI = 16.10–39.00%), MS (proportion of mediation:14.90%, 95% CI = 9.88–35.00%), Phenoage (proportion of mediation: 7.89%, 95% CI = 4.30–22.00%), and IS (proportion of mediation:7.22%, 95% CI = 3.78–21.00%) in the association between DI-GM and MASLD was observed. And BMI significantly mediated the correlations between DI-GM and other chronic liver diseases (proportion of mediation:8.10%, 95% CI = 4.22–16.00% for cirrhosis; 13.00%, 95% CI = 4.48–291.00% for liver cancer; 13.20%, 95% CI = 5.51–31.00% for liver-related mortality).

Fig. 4.

The proportion mediated by mediators on the associations of DI-GM with adverse liver outcomes. Colors indicate the magnitude of the mediated proportion, with red representing a higher proportion. BMI, body mass index; MASLD, metabolic dysfunction-associated steatotic liver disease

Associations between DI-GM and MASLD and other chronic liver diseases stratified by HS-PRS Tertiles

The elevated risks of adverse liver outcomes were perceived in participants with DI-GM scores of 0–3 and higher HS-PRS relative to those with scores of ≥ 6 and lower HS-PRS (Table S6). However, no significant interactions were observed between DI-GM and HS-PRS (Table S7).

Associations of each DI-GM component with adverse liver outcomes

In the fully adjusted model, higher consumption of coffee (HR = 0.80, 95% CI = 0.72–0.89), fermented dairy (HR = 0.85, 95% CI = 0.77–0.94), fiber (HR = 0.84, 95% CI = 0.75–0.93), soybean (HR = 0.41, 95% CI = 0.19–0.85), and whole grains (HR = 0.85, 95% CI = 0.77–0.95) were associated with a lower risk of MASLD (Table S9). Conversely, higher intake of refined grains (HR = 1.11, 95% CI = 1.01–1.23) was associated with elevated risk of MASLD (Table S9). Inverse correlations between intake of fermented dairy and whole grains and the risks of other chronic liver diseases were also identified (Table S10-12).

Sensitivity analyses

Table S13-17 demonstrate that our findings remained robust across most of sensitivity analyses. Subgroup analyses suggested that the associations of DI-GM with cirrhosis and liver-related mortality were stronger in individuals who did not engage in regular physical activity (both P-interaction < 0.05, Fig. S4). However, no significant interactions were detected for other factors, including age, sex, smoking status, alcohol consumption, and diabetes (all P-interaction > 0.05, Fig. S4).

Population attributable fraction (PAF) for adverse liver outcomes

Table S18 presents the percentage of liver disease cases that could have been prevented in UKB participants when adhering to a higher DI-GM dietary pattern. If we assume the associations to be causal, 10.75% of incident MASLD cases and 26.41% of incident cirrhosis cases in the study population were attributed to lower adherence to the highest DI-GM.

Discussion

Based on a large-scale prospective cohort study, we first demonstrated that adhering to higher DI-GM dietary pattern and diet beneficial to gastrointestinal microbiota significantly reduced the risk of MASLD. Participants with DI-GM scores ≥ 6 had a notably lower MASLD risk than those scoring 0–3. This inverse association was consistently observed in a subsample of participants who underwent abdominal MRI and persisted regardless of genetic risk. Our findings remained consistent across multiple sensitivity analyses, confirming the stability and reliability of the results. Similar inverse relationships were also identified for the risks of other chronic liver diseases. The RCS analyses confirmed significant dose-response relationships between DI-GM and MASLD and cirrhosis. Furthermore, mediation analyses revealed that Phenoage, BMI, MS, IS, and metabolic signature significantly mediated the associations between DI-GM and MASLD. These findings highlight the potential value of DI-GM as a broad-spectrum preventive target for adverse liver outcomes, beyond its protective role against MASLD.

Emerging evidence has emphasized the role of DI-GM in promoting human health. For example, an inverse correlation between DI-GM and the prevalence of MASLD was identified in a cross-sectional study [23]. Aligning with this, our study not only confirmed the longitudinal association between DI-GM and MASLD but also extended this relationship to other chronic liver diseases. Furthermore, DI-GM demonstrated an inverse linear association with stroke prevalence [22]. Individuals with higher DI-GM exhibited a lower risk of diabetes, a common metabolic disorder with substantial global morbidity and mortality worldwide [24]. Therefore, these studies, alongside our findings, collectively offer robust evidence supporting the role of adhering to higher DI-GM in attenuating the progression of various diseases and enhancing overall health.

The interplay between components of DI-GM and adverse liver outcomes has been highlighted in previous studies. Specifically, higher intake of legumes was linked to a reduced risk of MASLD [25]. A 24-week randomized controlled clinical trial confirmed that yogurt consumption improved hepatic steatosis and liver enzyme profiles in MASLD patients [26]. And each 5-g/d increment in dietary fiber was associated with 7% reduced MASLD risk [27]. Our findings coincide with these studies, suggesting that higher intake of soybean (a type of legume), fermented dairy, and dietary fiber correlated with decreased MASLD risk. Both previous research and our study support an inverse relationship between dietary fiber consumption and incident cirrhosis [27]. Furthermore, individuals with higher tea, yoghurt, or whole grains consumption presented reduced risk of HCC [28, 29]. Consistent with these findings, our study also identified significant inverse correlations between fermented dairy, green tea, and whole grains and liver cancer. Additionally, daily consumption of one or more cups of coffee per day was associated with significantly reduced risk of liver-related mortality [30].

The protective association between higher DI-GM and reduced risk of adverse liver outcomes is mediated through multiple interconnected biological pathways. Firstly, DI-GM may influence MASLD progression by modulating gut microbiota, although this inference is indirect given the limited coverage of microbially derived metabolites in our platform. For instance, a diet rich in fermented dairy – beneficial components of DI-GM – could enhance microbial diversity and alter compositions, thereby regulating MASLD development [31, 32]. Secondly, the mediating effect of metabolic signature underscores the role of metabolomic alterations in the relationship between DI-GM and MASLD. Our exploratory analyses identified 89 circulating metabolites significantly correlated with DI-GM, spanning key biological classes including lipoprotein subclasses, fatty acids, amino acids, and glycolysis-related metabolites (e.g., omega-3/6 fatty acids and lactate). These metabolites were integrated into a metabolic signature that significantly mediated the association between DI-GM and liver diseases. This metabolic signature primarily reflects perturbations in lipid and lipoprotein metabolism, consistent with prior metabolomic studies linking dietary patterns to MASLD pathophysiology. For example, high consumption of yogurt – a unique component of DI-GM – reduces plasma lipids such as total cholesterol and triacylglycerol, alleviating hepatic steatosis and injury [32]. Our exploratory analyses confirm these parameters are inversely correlated with MASLD risk and positively associated with DI-GM, highlighting the importance of metabolic regulation. Thirdly, the mediating role of IS indicates that adherence to higher DI-GM may mitigate liver disease by reducing inflammation. Whole grains abundant in dietary fiber, a key element of DI-GM, promote SCFA-producing bacteria, which alleviate inflammation [33–35]. Ameliorated inflammation improves gut dysbiosis, attenuating MASLD progression [36]. Ultimately, our mediation analyses suggest that adherence to higher DI-GM may attenuate liver disease by ameliorating obesity, aging, and metabolic disorders. Existing evidence supports these mechanistic links. For example, fermented foods consumption alleviates obesity and metabolic disorders by reducing body weight gain and fat tissue, enhancing glucose homeostasis and insulin resistance, thereby attenuating hepatic dysfunction [37, 38]. And coffee, a key element of DI-GM, enhances PPAR-α-mediated fatty acid oxidation and boosts antioxidants, decelerating aging-induced ferroptotic stress that exacerbates MASLD [8, 39]. Therefore, our findings highlight the roles of metabolomic changes, inflammation, obesity, metabolic disorders, and aging in the pathogenesis and progression of MASLD, with indirect support for the involvement of gut microbiota, offering integrated insights into the mechanisms linking DI-GM to liver health.

Given the role of gut microbiota as non-invasive biomarkers for liver disease, the inverse associations between DI-GM and adverse liver outcomes highlight DI-GM as a promising tool to evaluate microbiota-targeted diets and underscore the potential of nutritional therapy to prevent liver diseases. Our metabolic signature, derived from 89 DI-GM-correlated metabolites spanning pathways like fatty acid and amino acid metabolism, significantly mediated the DI-GM–liver disease association, providing novel insights into underlying metabolic alterations. While these findings are promising, several limitations of the metabolic signature are detailed in the limitations section. Notably, the inverse association between DI-GM and MASLD was consistent across all HS-PRS tertiles, indicating a gut-friendly diet may counteract genetically predisposition to liver disease. Furthermore, the inverse correlations were more pronounced in physically inactive individuals than in others, highlighting microbiota-targeted dietary interventions as a promising strategy for high-risk populations when sustained exercise is challenging. Aligning with prior studies, our findings also emphasize the protective role of fermented dairy and whole grains – key elements of DI-GM – in liver health, whereas refined grains are associated with elevated risks [26, 28, 33]. PAF analysis showed over one-fifth of chronic liver disease cases were attributable to lower DI-GM adherence. Thus, promoting diets rich in beneficial DI-GM components (e.g., fermented foods, whole grains) while reducing unfavorable ones (e.g., refined grains) is a promising prevention strategy, warranting testing in future trials. The etiology-specific predictive value and optimal thresholds of DI-GM for different liver diseases also require validation in targeted studies.

Several strengths of this study warrant attention. We first demonstrated the longitudinal associations between DI-GM and the risks of MASLD and other chronic liver diseases. To ensure the robustness and reliability of our findings, we conducted extensive subgroup and sensitivity analyses and the associations remained consistent. Additionally, the MRI-derived liver PDFF enables the identification of undiagnosed MASLD cases, further validating our primary findings. To explore the potential mechanisms, we examined the mediating roles of Phenoage, BMI, MS, and IS, revealing significant mediating effects in the relationships between DI-GM and adverse liver outcomes. Stratification by HS-PRS tertiles also allows us to assess the potential modification of associations by genetic risk factors.

However, several limitations should be noted. First, as an observational study, causal inferences cannot be established, and residual confounding may remain despite comprehensive adjustment. Second, dietary intake based on 24-hour recalls is susceptible to measurement bias, although the metabolic signature was constructed to improve objectivity. Third, MASLD ascertainment using ICD-10 codes, with limited MRI-PDFF validation, may lead to non-differential misclassification and severity bias, even though our case definition aligns with current MASLD nomenclature and prior research. Fourth, mediation analyses depend on untestable model assumptions and should be interpreted cautiously. Fifth, the NMR-based platform assessed only a subset of metabolites, and our metabolic signature was derived from these specific metabolomic assays, with modest correlation to DI-GM, limited coverage of microbial metabolites, and no external validation, which restricts its utility as a biomarker and limits mechanistic inferences on gut microbiota. Sixth, the UK Biobank cohort is predominantly composed of middle-aged and older adults of European ancestry, with inherent demographic and dietary homogeneity, which may constrain the broader generalizability of our findings. Seventh, defining follow-up from the last dietary assessment may introduce left truncation and immortal time bias, even though consistent associations were identified. Ultimately, some DI-GM components were not available in the dietary recall data (e.g., chickpeas, cranberries), potentially limiting the validity and comparability of the dietary index.

Conclusions

Adherence to a higher DI-GM dietary pattern and a diet beneficial to gut microbiota was associated with reduced risks of metabolic-associated steatotic liver disease and other chronic liver diseases, regardless of genetic risk. Metabolic signature for DI-GM also inversely correlated with MASLD. Mediation analyses revealed that the Phenoage, BMI, MS, IS, and metabolic signature significantly mediate the relationship between DI-GM and incident MASLD. These findings suggest that dietary interventions targeting gut microbiota could be a promising avenue for reducing the risk of adverse liver outcomes and warrant further investigation in interventional studies. The metabolites and signature identified may serve as candidates for future development of risk assessment tools.

Abbreviations

MASLD

Metabolic dysfunction-associated steatotic liver disease

DI-GM

Dietary index for gut microbiota

Phenoage

Phenotypic age

BMI

Body mass index

MS

Metabolic score

IS

Inflammatory score

HS-PRS

Polygenic risk score of hepatic steatosis

PAF

Population attributable fraction

MRI-PDFF

Magnetic resonance imaging-derived liver proton density fat fraction

BDI-GM

Dietary index beneficial to gut microbiota

UKB

UK Biobank

HR

Hazard ratio

Author contributions

HX and XC had full access to all of the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis.Conception and design: HX, WL, XC, JL, CY, YD. Acquisition of data, or analysis and interpretation of data: MZ, SW, YL, YH, XC. Drafting the manuscript: MZ, YH. Revising the manuscript critically for important intellectual content: QY, SW, YD, XC, CY, JL, YH. Statistical analysis: MZ, SW, YL, PZ. Obtained funding: HX. Administrative, technical, or material support: AC, WL, YD, XC, HX.Supervision: HX, WL.

Funding

Prof Xue is supported by National Natural Science Foundation of China (Grant number 82304130) and Basic research project of Guangzhou Science and Technology Bureau (Grant number SL2023A04J01145). The UK Biobank was approved by the North-West Multi-Centre Research Ethics Committee (reference 11/NW/0382).

Data availability

The UK Biobank data that support the findings of this study can be accessed by researchers on application (https://www.ukbiobank.ac.uk/).

Declarations

Competing interests

The authors declare no competing interests.