Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly known as non-alcoholic fatty liver disease, affects approximately one-quarter of the global population1,2. This condition has emerged as a significant global health concern due to its increasing prevalence, its association with severe metabolic comorbidities, and its potential progression to cirrhosis and hepatocellular carcinoma3. Obesity, hyperglycemia, hyperlipidemia, and insulin resistance are well-established independent risk factors for MASLD4,5,6. Despite the close relationship between obesity, type 2 diabetes, and MASLD, 20% of individuals with one of these conditions do not exhibit the other two7. Therefore, identifying novel risk factors for MASLD will aid in refining population stratification and facilitate the development of effective prevention and treatment programs8.

Recent studies have identified MASLD as closely associated with patients’ mental states. Although most MASLD patients recognize that poor lifestyle choices, such as high-calorie diets and lack of physical activity, are key risk factors, adherence to improved lifestyle practices remains poor9. Approximately 40% of MASLD patients undergoing liver transplantation experience disease recurrence within five years post-transplantation10. Research indicates that MASLD patients display significantly higher scores on the Standard Assessment of Personality-Abbreviated Scale, uncontrolled eating, and cognitive restraint measures compared to healthy individuals11. Additionally, the prevalence of MASLD is higher among individuals using antipsychotics (8.70%) compared to those who do not (5.45%)11. However, there is a paucity of longitudinal studies examining the association between mental health and MASLD12.

Recent studies have suggested potential mechanisms linking mental status to MASLD progression, highlighting that the brain is physiologically involved in disease progression. For instance, the brain-liver-gut axis maintains homeostasis of the digestive system’s immune function via peripheral regulatory T cells13. Additionally, the brain-brown fat-liver axis regulates inflammation and mental health through the IL-6 signaling pathway14.

Metabolomics has been widely applied to interpret metabolite characteristics in metabolic diseases, especially providing valuable insights into metabolic-related diseases, like MASLD15,16,17,18.

In this study, we explored the complex relationship between mental health and the development of MASLD by utilizing a large cohort from the UK Biobank. Our findings reveal that common mental health disorders, including anxiety, depression, mania, and trauma, significantly contribute to the risk of MASLD. These conditions were found to influence the incidence of MASLD through metabolic dysregulation, with metabolic pathways, particularly those involving nitrogen metabolism and lipid partitioning, mediating this association. As far as we are aware, these results provide molecular insights into the psychosomatic pathways linking mental health to liver disease and emphasize the need for interdisciplinary approaches that address both mental and metabolic health for effective prevention and intervention strategies.

Materials and methods

Study design and population

The UK Biobank is a prospective cohort study that recruited >500,000 individuals aged 40–69 years between 2006 and 2010. Participants completed questionnaires and biospecimen collection. UK Biobank study was approved by the North–West Multi-center Research Ethics Committee and conducted according to the Declaration of Helsinki. Our analytic procedures were conducted under the UK Biobank application number “96962”. Exclusion criteria included: (1) confirmed liver disease, (2) absence of metabolomics data, or (3) incomplete mental health assessment. Written informed consent was obtained from all participants. Figure 1.

Fig. 1: The flowchart of the UK Biobank participants included and excluded in this study.
Fig. 1: The flowchart of the UK Biobank participants included and excluded in this study.The alternative text for this image may have been generated using AI.
Full size image

This figure illustrates the process of participant selection from the UK Biobank dataset (n = 502,480). After applying exclusion criteria, the final groups included 55,012 participants in the Trauma group, 15,318 participants in the Mania group, 30,798 participants in the Depression group, and 17,596 participants in the Anxiety group.

Mental health assessment

Mental health was evaluated for anxiety, depression, mania, and trauma using UK Biobank survey instruments, developed under expert supervision. Based on questionnaire responses, we calculated composite scores for each domain (Supplementary Data 1).

Identification of MASLD

MASLD cases were identified using ICD-10 codes K76.0 and K75.8. For patients with multiple admissions, the first diagnosis and admission date were used.

Covariates

Potential confounders were selected based on previous studies, including age, sex, ethnicity, education level, household income, BMI, diabetes, smoking, alcohol use, sleep duration, and physical activity.

Metabolomics

Nuclear magnetic resonance

Plasma metabolomics profiling was performed by Nightingale Health Ltd. on behalf of UK Biobank using a high-throughput proton (^1H) nuclear magnetic resonance (NMR) spectroscopy platform (Bruker Avance III HD 600 MHz). A total of 251 metabolic measures, including absolute concentrations of small molecules (e.g., glucose, creatinine, amino acids) and derived traits such as lipoprotein particle sizes and fatty acid ratios, were quantified from plasma samples collected in 2019–2020. Data generation, including sample preparation, spectral acquisition, and quality control, followed standardized protocols established by Nightingale Health and UK Biobank. Details of the NMR metabolomics platform are available on the UK Biobank website (https://biobank.ndph.ox.ac.uk/ukb/label.cgi?id=220 and have been described in previous publications19,20,21.

Data processing

Metabolites were log-transformed and standardized to reduce skewness and improve comparability.

Statistical analysis

Baseline characteristics were summarized as medians (IQR) for continuous variables and frequencies (%) for categorical variables.

Survival analysis

Kaplan–Meier curves and log-rank tests were used to estimate cumulative MASLD risk across high vs. low mental health scores. Restricted cubic splines (three knots) were applied to explore non-linear associations.

Metabolite selection

Elastic net regularization (glmnet package) was applied to identify metabolites associated with mental health scores.

Cox regression

Multivariable Cox proportional hazards models were used to examine the relationship between mental health, metabolic signatures, and MASLD, adjusting for covariates. Proportional hazards assumptions were tested using Schoenfeld residuals.

Mediation analysis

The mediation package with 1000 bootstrap iterations quantified the proportion of the mental health–MASLD association mediated by metabolic signatures.

Robustness checks

Sensitivity analyses were performed with alternative covariate sets and cross-validation of models. False discovery rate (Benjamini–Hochberg) and Bonferroni corrections were applied to control for multiple testing.

All analyses were conducted in R. Data visualization was performed using ggplot2. A two-sided p-value < 0.05 was considered statistically significant.

Results

Baseline features

After collation, the study included 55,012 participants for the trauma score, 15,318 for the mania score, 30,798 for the depression score, and 17,596 for the anxiety score. The median follow-up times were as follows: 12.1 years (IQR 11.5–13.1 years) for the trauma group, 11.7 years (IQR 10.9–12.6 years) for the mania group, 11.9 years (IQR 11.0–12.7 years) for the depression group, and 11.6 years (IQR 10.8–12.7 years) for the anxiety group. White participants comprised the majority (>85%) in all four groups, and more than half of the participants in each group had attained higher education Table 1. (Supplementary table 1).

Table 1 The characteristics of the included participants
Full size table

3.2 Trauma, mania, depression, and anxiety increase the risk of developing MASLD

First, we stratified participants into high and low groups according to the median values of anxiety, depression, mania, and trauma scores. After adjusting for baseline covariates, the results revealed a consistent trend across all four dimensions: individuals in the high-score groups exhibited a significantly higher cumulative risk of MASLD compared to those in the low-score groups (log-rank P < 0.001). There were no significant nonlinear relationships between anxiety scores, depression scores, mania scores, trauma scores, and MASLD in the restricted cubic spline analyses (anxiety P for non-linear = 0.971, depression P for non-linear = 0.609, mania P for non-linear = 0.969, trauma P for non-linear = 0.174). However, a J-shaped association was observed between the four mental health scores and the temporal risk of MASLD, indicating that the risk of MASLD increased with higher mental health scores (anxiety P for overall trend <0.001, depression P for overall trend <0.001, mania P for overall trend <0.001, trauma P for overall trend <0.001) Fig. 2.

Fig. 2: Anxiety, depression, mania, and trauma are associated with an increased risk of MASLD.
Fig. 2: Anxiety, depression, mania, and trauma are associated with an increased risk of MASLD.The alternative text for this image may have been generated using AI.
Full size image

a–d Cumulative risks of developing MASLD based on mental health status (anxiety, P = 0.0009; depression, P = 0.0006; mania, P = 0.0002; traumma, P = 0.0007). e–h Dose-response relationships between mental health scores (anxiety, P = 0.0004; depression, P = 0.002; mania, P = 0.002; traumma, P = 0.005) and MASLD risk ratios. Pink lines depict the risk ratios, while shaded areas represent the 95% confidence intervals. Models were adjusted for age, sex, education level, household income, BMI, smoking status, alcohol consumption, physical activity, diabetes, and sleep time. Comparisons of cumulative incidence curves were performed using the two-sided Log-rank test. The P-values were adjusted for multiple comparisons using the Bonferroni method (2a–2h). The solid line represents the hazard ratios (HRs), and the shaded area represents the 95% confidence interval, derived from Cox proportional hazards models using restricted cubic splines. The P-values for the overall and nonlinear association are shown. All models were adjusted for multiple comparisons using the Bonferroni method.

3.3 Mental health scores and metabolic signatures associated with MASLD risk

The elastic net model, which integrates ridge regression and lasso regression, was employed to identify metabolites associated with anxiety, depression, mania, and trauma. Specifically, the analysis revealed 62 metabolites linked to anxiety, 76 to depression, 47 to mania, and 102 to trauma. Cox proportional hazards regression analyses revealed that 9 metabolites were associated with the risk of MASLD in patients with anxiety (P < 0.05), 10 metabolites were associated with the risk of MASLD in patients with depression, 6 metabolites were associated with the risk of MASLD in patients with mania, and 15 metabolites were associated with the risk of MASLD in patients with trauma. Metabolic signatures for anxiety, depression, mania, and trauma were established independently using Cox proportional hazards modeling. Creatinine exhibited a strong negative correlation with anxiety, depression, and trauma scores. Triglycerides-VLDL(very low-density lipoprotein)%, glutamine, and free cholesterol-VLDL% demonstrated the strongest positive correlations with the metabolic models of anxiety, depression, mania, and trauma, respectively. Notably, depression, mania, and trauma showed the most significant positive correlations within their respective models Fig. 3(Supplementary Data 2).

Fig. 3: Association between metabolite profiles and mental health outcomes, as well as MASLD risk.
Fig. 3: Association between metabolite profiles and mental health outcomes, as well as MASLD risk.The alternative text for this image may have been generated using AI.
Full size image

Utilizing Cox proportional hazards modeling to develop metabolic. The metabolites’ coefficients (weights) in the signature, associations with mental health, and subsequent MASLD risk. a Anxiety, this analysis included n = 17,596 participants with complete anxiety assessment data. b Trauma, this analysis included n = 55,012 participants with complete trauma assessment data. c Depression, this analysis included n = 30,798 participants with complete depression assessment data. d Mania, this analysis included n = 15,318 participants with complete mania assessment data. Forest plot shows the hazard ratios (HRs) and their 95% confidence intervals for the association between the mental health and incident MASLD. The vertical line of no effect (HR = 1) is indicated. The HR and their 95% CI for each metabolite are provided in the supplementary material of Fig. 3.

3.4 Mental health scores and metabolic signatures associated with MASLD

In Model 2, the prevalence of MASLD increased with higher scores of mental health, anxiety (HR = 1.13, 95% CI = 1.08–1.18, P < 0.05), depression (HR = 1.06, 95% CI = 1.03–1.10, P < 0.05), mania (HR = 1.15, 95% CI = 1.06–1.24, P < 0.05), trauma (HR = 1.22, 95% CI = 1.11–1.34, P < 0.05). Sensitivity analyses demonstrated the same results. Anxiety (HR = 1.21, 95% CI = 1.14–1.29, P < 0.05), depression (HR = 1.09, 95% CI = 1.06–1.15, P < 0.05), mania (HR = 1.18, 95% CI = 1.11–1.25, P < 0.05), trauma (HR = 1.22, 95% CI = 1.11–1.34, P < 0.05) related associated metabolic signatures were associated with increased HR for MASLD Fig. 4(Supplementary Data 3). Sensitivity analyses were consistent with the results of the main analyses.

Fig. 4: Association of mental health and metabolic signatures with the risk of MASLD.
Fig. 4: Association of mental health and metabolic signatures with the risk of MASLD.The alternative text for this image may have been generated using AI.
Full size image

a Anxiety, this analysis included n = 17,596 participants with complete anxiety assessment data. b Depression, this analysis included n = 30,798 participants with complete depression assessment data. c Trauma, this analysis included n = 55,012 participants with complete trauma assessment data. d Mania, this analysis included n = 15,318 participants with complete mania assessment data. Forest plot shows the hazard ratios (HRs) and their 95% confidence intervals for the association between the metabolic signature of anxiety and incident MASLD. The vertical line of no effect (HR = 1) is indicated. Cox proportional hazards models were employed to investigate the association between mental health and metabolic signatures and the risk of developing MASLD. Model 1 was adjusted for age and gender. Model 2 further adjusted for age, sex, education level, household income, BMI, smoking status, alcohol consumption, physical activity, diabetes, and sleep time. Model 3 was a multivariate model that included adjustments for both mental health and metabolic signatures, allowing for an exploration of their independent effects on MASLD. A p-value threshold of <0.05 was used to determine statistical significance. The HR and their 95% CI for each model are provided in the supplementary material of Fig. 4.

We further explored the independent roles of mental health scores and metabolic signatures on incident MASLD. As shown in the results of Model 3, after adjusting for the metabolic signatures, the risk of MASLD increased with higher anxiety, depression, mania, and trauma scores. Additionally, higher metabolic signature scores were associated with an increased risk of MASLD. Mental health and metabolic signatures remained risk factors for MASLD in all three sensitivity analyses (Supplementary Data 4).

3.5 Mediation analyses

Metabolic signatures significantly mediated the relationship between anxiety, depression, mania, trauma, and the risk of MASLD. Specifically, the analysis revealed that the indirect effects of these mental health conditions on MASLD were accounted for by changes in metabolic pathways. The mediation weights were 12.7%, 15.7%, 11.5%, and 11.2% (P < 0.05), respectively, indicating the proportion of the mental health-MASLD association attributable to metabolic alterations. Figure 5.

Fig. 5: Mediation analysis diagram illustrating the role of metabolic signatures in the relationship between mental health disorders and MASLD risk.
Fig. 5: Mediation analysis diagram illustrating the role of metabolic signatures in the relationship between mental health disorders and MASLD risk.The alternative text for this image may have been generated using AI.
Full size image

The model tests whether metabolic alterations mediate the association between mental health exposures (anxiety, depression, mania, trauma) and MASLD. Percentages indicate the proportion of the total effect mediated by metabolic pathways. All indirect effects were statistically significant (P < 0.05).

Discussion

Based on a comprehensive analysis of the relationship between mental health and MASLD in a large-scale prospective UK study, we identified a distinct metabolic signature linked to anxiety, depression, mania, and trauma. Our findings indicate that higher mental health scores are significantly correlated with an increased risk of MASLD. Even after adjusting for potential covariates such as diabetes, obesity, and alcohol consumption, and conducting three sensitivity analyses, the metabolic signatures associated with these mental health conditions consistently showed an elevated risk of MASLD. Our findings suggest that mental health conditions, such as anxiety, depression, mania, and trauma, are associated with an increased risk of MASLD, potentially through metabolic dysregulation acting as a mediator. The mediation analysis demonstrates that changes in metabolic pathways, including lipid and nitrogen metabolism, account for a significant portion of the mental health-MASLD association. These findings provide a mechanistic insight into how mental health conditions may indirectly influence MASLD development.

MASLD affects approximately one-quarter of the global population, with around 10% progressing to MASH22. However, its pathogenesis remains not fully understood. The traditional second-hit theory regarding the progression from MASLD to MASH is increasingly challenged by emerging research, which highlights synergistic mechanisms involving genetic predisposition, gut microbiota, and insulin resistance in the development of MASH23. While obesity is widely recognized as a significant risk factor for MASLD, population-based cohort studies have shown that about 20% of MASLD patients have a BMI below 23.9 kg/m²24. Therefore, it is crucial to accurately identify additional predisposing factors for MASLD25.

According to the World Health Organization, mental health is among the four major global health burdens26,27. Factors such as discrimination, disease, environmental changes, wealth disparity, and family issues can directly or indirectly impair mental health28. During the COVID-19 pandemic, billions worldwide experienced anxiety, depression, mania, and trauma29. Adverse outcomes of impaired mental health include self-harm, alcoholism, and drug addiction. However, despite significant attention on mental health itself, few studies have explored histopathological changes in other organs following mental health impairment, particularly in the liver, a central organ for substance metabolism and transport30.

Recent studies underscore the pivotal role of the gut-brain-liver axis in MASLD, involving the hypothalamus, gut, and liver31. Interactions between the central nervous system, gastrointestinal tract, and hepatic system are crucial for regulating glucose, lipid, and protein metabolism, and inflammatory responses32,33. Following the COVID-19 pandemic, a European cohort study noted a significant rise in anxiety, depression, and trauma scores. Individuals with anxiety and depression showed an increased proportion of Aspergillus flora and a decreased presence of Clostridium perfringens in their guts34. Altered levels of anaerobic and spore-forming enteric bacteria were observed among those experiencing trauma35. The gut microbiota plays a crucial role in modulating inflammatory responses within the body36. Blocking integrin α4β7 could mitigate inflammation and fibrosis in liver tissues affected by metabolic dysbiosis37. Turicibacter sanguinis has been linked to central nervous system disorders, including schizophrenia, anxiety disorders, and depression38. Recent studies have demonstrated a significantly higher abundance of spore-forming enteric bacteria in patients with MASLD39. Increased abundance of Turicibacter sanguinis was accompanied by increased abundance of pro-inflammatory Escherichia coli and Shigella taxa in the intestinal tract, along with a corresponding decrease in anti-inflammatory rectal E. coli40. Moreover, this change was associated with elevated levels of inflammatory mediators and MASH events41.

The liver serves as a central hub for the metabolism of proteins, lipids, and carbohydrates. Its function is intricately regulated by neural signals, hormones, and metabolites42. Additionally, metabolites produced by intestinal microorganisms can enter the liver through the portal venous system, thereby influencing hepatic function43,44. Metabolic reprogramming, encompassing hepatocyte redox state, hormonal circadian rhythms, gut microbiota, and their metabolites, modulates hepatocyte metabolism, stress responses, and intercellular interactions, ultimately contributing to pathological changes in MASLD45. Mental health also plays a significant role in liver metabolism. A prospective study identified a potential association between liver disease and alterations in brain white and gray matter volumes46. Specifically, patients with metabolic liver disease exhibited trends of decreased overall brain volume, as well as reduced white matter integrity and synaptic density46. Mechanistically, alterations in gut microbiota composition and impaired amino acid metabolism, both of which are consequences of compromised mental health, contribute to the development of MASLD47. Our study identified anxiety, depression, mania, and trauma as novel risk factors for incident MASLD. After adjusting for common confounders such as alcohol consumption, BMI, and diabetes, and conducting three sensitivity analyses, we found that a higher mental health score was associated with an increased risk ratio for MASLD. Metabolomics elucidates the potential mechanisms underlying the impact of mental health scores on MASLD through the characterization of metabolic pathways. Specifically, glutamine levels were found to be negatively correlated with mental health scores. Gómez found that MASH patients exhibited strong glutaminase 1 activity via the LPS/TLR4 axis, resulting in increased glutamine catabolism48. The catabolites from this process promote the progression of MASLD to MASH, which aligns with our findings. Additionally, studies have shown that an elevated Omega-6 to Omega-3 ratio is commonly observed in mental health issues49. In patients with MASLD, increased levels of oxidative stress and inflammation have been reported. Multiple studies have demonstrated a positive correlation between Omega-6 levels, oxidative stress, and inflammation.

However, this study has several limitations. First, the majority of participants were White, which limits the generalizability of the findings. Validation studies in more diverse populations from other countries and regions would enhance the applicability of the results. Second, investigating the impact of time-series changes in metabolic characteristics on MASLD events would provide more relevant insights for real-world scenarios. Finally, mental health assessments were conducted via questionnaires rather than through evaluations by specialized or community doctors, which may introduce potential biases and limitations.

Conclusion

In conclusion, our prospective study of a large cohort has identified distinct metabolic signatures associated with anxiety, depression, mania, and trauma. These findings add to the body of evidence supporting the role of metabolic signatures in the development of MASLD. While current non-invasive imaging modalities, such as Vibration-controlled transient elastography and Magnetic resonance elastography, are widely used and validated for detecting hepatic steatosis and fibrosis, our findings suggest that metabolic alterations associated with adverse mental health conditions may offer additional biological insight into MASLD pathogenesis. Although further validation is required, integrating metabolomic markers with existing screening approaches in high-risk psychiatric populations could potentially enhance early risk stratification and inform targeted prevention strategies.