Skip to main content
NCBI home page
Science Advances logoLink to Science Advances
. 2026 May 13;12(20):eaeg4939. doi: 10.1126/sciadv.aeg4939

The metabolomic signatures mediate associations between physical frailty and metabolic dysfunction–associated steatotic liver disease

Rongtao Jiang 1,2,*, Matthew Rosenblatt 3, Shile Qi 4, Vince D Calhoun 5, Qian Wang 1, Peng Wang 1, Jing Wu 6, Dustin Scheinost 2,3,7,8,9, Jing Sui 1,*
PMCID: PMC13170646  PMID: 42127175

Abstract

Prevention remains a key strategy to address the growing burden of metabolic dysfunction–associated steatotic liver disease (MASLD), highlighting the importance of exploring modifiable risk factors. Accumulating evidence suggests a close link between physical frailty and MASLD. However, how frailty interacts with metabolic syndrome to affect MASLD and the causality and direction of the association remain largely unknown. Leveraging data from 405,224 UK Biobank participants with a 13.65-year follow-up, we found that physical frailty was associated with an increased risk of clinically diagnosed MASLD and exacerbated the adverse effect of metabolic syndrome on MASLD incidence, implying that frail people may be more vulnerable to this disease because of metabolic syndrome. Mendelian randomization provided evidence for a potential causal effect of physical frailty on MASLD but not the reverse direction. Moreover, the metabolome-wide association analysis revealed widespread associations of plasma metabolites with both frailty and MASLD, suggesting a shared metabolomic foundation between them. Some metabolites, including fatty acids and triglyceride-rich lipoprotein biomarkers, partially explained the frailty-MASLD relationship, indicating a potential metabolomic mechanism. If confirmed in further studies, frailty screening may help identify high-risk individuals and inform early prevention for MASLD, especially for those with metabolic syndrome.

Physical frailty predicts MASLD risk and shares metabolomic signatures with the disease.

INTRODUCTION

Metabolic dysfunction–associated steatotic liver disease (MASLD), affecting 38% of the world’s population (1), often progresses asymptomatically until irreversible complications like cirrhosis and hepatocellular carcinoma occur (2–4). With no approved therapies, developing preventative interventions remains a key strategy to address its growing burden (5).

Physical frailty—defined as the concurrent presence of five components, including weight loss, exhaustion, weakness, physical inactivity, and slow walking speed—is a promising modifiable risk factor for MASLD (6, 7). Frailty is highly prevalent (8), prospectively associated with a 3.32-fold increased risk of developing MASLD (9), and predictive of disease progression (10). Nevertheless, several knowledge gaps remain regarding whether physical frailty can be integrated into MASLD prevention. First, a previous study explored the association of MASLD with general frailty, potentially overlooking differential contributions of individual components of frailty. Given that modifying all aspects of frailty is challenging, elucidating specific components can pinpoint high-priority targets and enable focused interventions (11). Second, we do not fully understand how physical frailty interacts with other preexisting risk factors, especially metabolic syndrome (12), to affect MASLD risk. Considering an increased vulnerability of frail people to stressors, frailty may exacerbate the adverse effect of metabolic syndrome on MASLD (13). However, existing research has exclusively focused on multiplicative interactions (9), leaving the synergistic interaction largely understudied, which holds more clinical implications for identifying high-risk individuals (14). Third, most existing evidence is from observational studies, which are limited by residual confounding and reverse causation. If the relationship is not causal, interventions targeting physical frailty would yield few benefits for MASLD (15). Mendelian randomization (MR) provides an alternative approach for estimating causal effects when clinical trials are challenging to do (16).

Moreover, the biological mechanisms linking physical frailty with MASLD remain poorly understood. The plasma metabolome carries dynamic biological signals reflecting homeostatic processes and health status (17), thereby providing an important perspective to assess the molecular mechanisms underlying MASLD. Disturbed plasma metabolites including amino acids and lipids are major factors contributing to the pathophysiology of MASLD (18–20). Likewise, metabolic signatures are essential to the maintenance of physiological homeostasis (21), and having an older metabolic age is associated with more severe frailty (22, 23). These findings shed light on certain metabolites as bridges between physical frailty and MASLD, yet studies regarding whether frailty-associated metabolic signatures may explain the increased risk of MASLD are scarce. A deep understanding of the metabolic mechanism holds promise in informing potential prevention strategies.

To fill these research gaps, this study leveraged data from the UK Biobank to investigate the prospective association of physical frailty and its components with MASLD incidence identified through hospital admission or death records. We also examined how physical frailty interacts synergistically with other factors to affect MASLD risk. Subsequently, we used bidirectional MR to distinguish causation from association (16). Capitalizing on high-throughput metabolome data, we conducted a metabolome-wide association analysis aiming to identify frailty-associated metabolites and dissect their role in linking frailty with MASLD risk. Last, we validated the above analyses using magnetic resonance imaging (MRI)–derived proton density fat fraction (PDFF) as a secondary outcome of MASLD (9), which provides precise estimation for fat concentration in the liver.

RESULTS

Study sample

We used phenotypic and hospitalization data from 405,224 MASLD-free participants from UK Biobank (mean age, 56.28 ± 8.08 years; 53.79% female; 95.27% white); of these, 14,502 were categorized as frail, 161,224 as prefrail, and 229,498 as nonfrail. Table S1 presents the baseline sample characteristics by frailty status. Compared with nonfrail individuals, participants with prefrailty and frailty were older, more likely to be female, and more deprived and had lower educational attainment and household income. They were also more likely to smoke, reported more sedentary behavior, and had a high prevalence of metabolic syndrome. Figure 1 shows a schematic representation of the study design and main analyses performed in this study.

Fig. 1. Overview of study workflow.

Fig. 1.

Open in a new tab

Leveraging phenotypic and hospitalization data from UK Biobank, this study investigated the prospective and causal association of physical frailty and its components with MASLD incidence, as well as the multiplicative and additive interactions with covariates. Then, capitalizing on high-throughput metabolomics data, we conducted a metabolome-wide association analysis to identify frailty-associated metabolites and dissect their role in mediating MASLD risk. Last, we reran the above analyses using MRI-derived PDFF as a secondary outcome and validated the association analysis on matched data using a propensity score matching procedure. Icons were made from www.svgrepo.com/.

Associations between physical frailty and MASLD incidence

Over a median follow-up of 13.65 years (interquartile range, 12.91 to 14.38 years), 4741 new-onset MASLD cases recorded in hospital admission or death registries were documented. Compared with nonfrail individuals, those with prefrailty [hazard ratio (HR) = 2.06] and frailty (HR = 5.00) showed a substantially increased risk of developing MASLD in the minimally adjusted model (Fig. 2A). Further adjustment for lifestyle or socioeconomic factors only slightly affected the associations, while adjusting metabolic syndrome attenuated the association of prefrailty and frailty with MASLD incidence by 28.3 and 41.0%, respectively (table S2). In the fully adjusted model, the HRs of MASLD risk were 1.61 in prefrailty {HR = 1.61, 95% confidence interval (CI) = [1.52 to 1.72], P = 5.47 × 10−51} and 2.67 in frailty (HR = 2.67, 95% CI = [2.40 to 2.97], P = 3.88 × 10−74, table S3). The association showed a monotonic increasing trend between frailty severity and MASLD risk (Fig. 2B). Moreover, strong evidence supported a nonlinear association between frailty severity and MASLD risks (P = 6.0 × 10−4), with steeper slopes at higher exposures (Fig. 2C).

Fig. 2. Prospective association between physical frailty and the risk of MASLD.

Fig. 2.

Open in a new tab

(A) Compared with nonfrail individuals, those with prefrailty and frailty showed a higher risk of developing MASLD. Among all covariates, metabolic syndrome attenuated the associations the most. (B) The association showed a monotonic increasing trend between frailty severity and MASLD. (C) The association between frailty severity and MASLD risks was nonlinear, with steeper slopes at higher exposures. (D) Each of the five components of frailty showed significant associations with MASLD risks, with slow walking speed demonstrating the strongest association. (E) Multiplicative interactive analyses supported a substantially modifying effect of sex and metabolic syndrome on the associations. (F) People having frailty and metabolic syndrome simultaneously conferred a higher risk of MASLD than the sum of the excess risk of the two factors, suggesting an additive interaction. SI, synergy index.

Associations persisted in subgroups stratified by covariates (fig. S1). Sensitivity analyses, including censoring participants experiencing events within the first 10 years of follow-up, using a broad definition of MASLD, or excluding those who drink alcohol excessively, mirrored the results of the main analysis (fig. S2). The associations were also replicated in covariate-matched data generated using a propensity score matching procedure (fig. S3 and table S4) (24).

Each of the five components of frailty showed significant associations with MASLD incidence, even in the mutually adjusted models (Fig. 2D and table S5). Specifically, slow walking speed demonstrated the strongest association (HR = 1.56, 95% CI = [1.44, 1.70], P = 1.38 × 10−25), while physical inactivity demonstrated the least significant association (HR = 1.14, 95% CI = [1.05, 1.24], P = 2.56 × 10−3).

Interplay between physical frailty and metabolic syndrome

Multiplicative interactive analyses supported a significantly modifying effect of sex and metabolic syndrome (Bonferroni-corrected P < 0.05/10) on the association between physical frailty and MASLD incidence (Fig. 2E). Specifically, the positive associations were more pronounced in females than in males and in people without metabolic syndrome than in those with.

Figure 2F shows the joint associations of physical frailty and metabolic syndrome with MASLD. The risk of MASLD was the highest among people who were frail and had metabolic syndrome, which was 8.27 times higher (HR = 8.27, 95% CI = [7.28, 9.41], P = 1.80 × 10−228; table S6) relative to those who were nonfrail and without metabolic syndrome. Furthermore, there was notable additive interaction, indicating that having frailty and metabolic syndrome simultaneously conferred to 2.00-fold higher risk than the sum of the excess risk of the two factors {relative excess risk due to interaction (RERI) = 2.00, 95% CI = [0.92, 3.10]}. Similar findings were observed for prefrailty, where having prefrailty and metabolic syndrome simultaneously conferred to 0.66 times higher risk than the sum of individual risks (RERI = 0.66, 95% CI = [0.27, 1.003]). The attributable proportions of joint effects were 12.65% for prefrailty (AP = 12.65%, 95% CI = [5.34%, 19.41%]) and 24.22% for frailty {attributable portion due to interaction (AP) = 24.22%, 95% CI = [12.01%, 34.64%]}. In a post hoc analysis exploring individual components of metabolic syndrome, the additive interaction appeared to be primarily driven by type 2 diabetes (table S7). No additive interaction was observed for other covariates (table S8).

Causal association between physical frailty and MASLD

MR analyses provided causal evidence of a detrimental effect of physical frailty on MASLD risk (Fig. 3A), with a nonsignificant effect in the reverse direction (Fig. 3B). Specifically, using 27 frailty-associated single-nucleotide polymorphisms (SNPs) as proxies (table S9), the inverse-variance weighted (IVW) method under fixed effect indicated a 4.62 times higher risk of MASLD per one-point increment in physical frailty severity {odds ratio (OR) =4.62, 95% CI = [1.90, 11.21], P = 7.10 × 10−4}. Model-based sensitivity analyses, including weighted median (P = 1.41 × 10−2), simple median (P = 7.10 × 10−4), and IVW under a random effect (P = 1.30 × 10−4), yielded a similar pattern of effects but were nonsignificant for MR Egger (P = 0.47). Figure 3C shows the scatterplot of SNP effects on physical frailty and MASLD.

Fig. 3. Bidirectional causal associations between physical frailty and MASLD.

Fig. 3.

Open in a new tab

(A) MR analysis provided causal evidence for a significantly detrimental effect of frailty on MASLD. (B) The causal effect of MASLD on physical frailty was nonsignificant. (C) The scatterplot shows SNP effects on physical frailty and MASLD, and the forest plot of leave-one-SNP-out sensitivity analysis indicated that no single SNP drove the MR estimation.

The MR Egger intercept test detected no obvious directional pleiotropy (intercept = −0.0011, P = 0.97), and the modified Cochran’s Q test suggested no notable heterogeneity (Q = 28.68, P = 0.28). MR-PRESSO (MR pleiotropy residual sum and outlier) detected no outliers, and analyses leaving out each SNP revealed that no single SNP drove the estimation. All SNPs had F-statistics >30, suggesting sufficient instrument strength. We found no evidence of a potential causal effect of any of the frailty components on MASLD risk (table S10).

Associations of metabolites with physical frailty and incident MASLD

The metabolome-wide association analysis encompassed a maximum of 225,698 participants with complete metabolome data. Together, 221of 251 metabolites were significantly associated with physical frailty while controlling for covariates (Bonferroni-corrected P < 0.05/251 for 251 tests; Fig. 4A and table S11). Among them, 36 showed positive associations and 185 showed negative associations. The effect size ranged from β = −0.088 for total esterified cholesterol (95% CI = [−0.092, −0.084], P < 5 × 10−324) to β = 0.089 for the phospholipid–to–total lipid percentage in very small very-low-density lipoprotein (XS_VLDL_PL%, 95% CI = [0.085, 0.093], P < 5 × 10−324).

Fig. 4. Mediating effect of plasma metabolites on the association between physical frailty and MASLD.

Fig. 4.

Open in a new tab

(A) Of 251 metabolites, 221 showed Bonferroni-corrected significant associations with physical frailty while controlling for covariates (P < 0.05/251). (B) A total of 187 of 251 metabolites showed Bonferroni-corrected significant associations with the risk of MASLD after covariate adjustment. (C) The metabolome-wide association pattern of physical frailty was significantly similar to that of MASLD (r = 0.73, P = 1.60 × 10−43), and 162 metabolites were consistently associated with both physical frailty and MASLD risk. (D) On the basis of the 162 metabolites, a general metabolomic signature generated using the first principal component of PCA explained 36.74% of the total variance. (E) The general metabolomic signature mediated 7.68% of the association between physical frailty and MASLD. (F) Each of the top 20 MASLD-associated metabolites also significantly mediated the frailty-MASLD association, with the mediated variance varying between 2.91 and 6.86% (P < 0.05/20, bootstrapping test).

Over a median follow-up of 13.70 years (interquartile range, 12.94 to 14.92 years), 187 of 251 metabolites at the baseline showed significant associations with the risk of developing MASLD after covariate adjustment (P < 0.05/251; Fig. 4B and table S12). Among them, 65 showed potentially detrimental effects and 122 showed protective effects. The effect sizes ranged from HR = 0.68 for cholesteryl esters in very large high-density lipoprotein (HDL) (XL_HDL_CE, 95% CI = [0.64, 0.72], P = 3.00 × 10−36) to HR = 1.34 for the monounsaturated fatty acid–to–total fatty acid percentage (MUFA%, 95% CI = [1.28, 1.40], P = 2.99 × 10−40). Notably, among the top 20 metabolites exhibiting the numerically most significant associations with MASLD, 15 were ratio measures and only 5 were absolute measures.

Sensitivity analyses, including conducting a 10-year landmark analysis, removing metabolite outlier data, excluding heavy alcohol drinkers, applying a broad MASLD definition, and logarithmically transforming metabolites, yielded nearly identical findings to the main results, with the association patterns highly correlated between each other (r > 0.92; figs. S4 and S5).

Metabolic signatures partially explained the frailty-MASLD associations

The metabolome-wide association pattern of physical frailty was significantly similar to that of MASLD (r = 0.73, P = 1.60 × 10−43), and 162 metabolomic biomarkers were consistently associated with both frailty and MASLD (Fig. 4C). On the basis of the 162 metabolites, we derived a general metabolomic signature of physical frailty using the first principal component of principal components analysis (PCA), accounting for 36.74% of the total variance (Fig. 4D). The metabolomic signature demonstrated a significant linear association with future risk of MASLD after adjusting for covariates (Pnonlinear = 0.93, Poverall = 7.31 × 10−49). A one-SD increase in the metabolomic signature was associated with a 25% lower risk of developing MASLD (HR = 0.75, 95% CI = [0.72, 0.78], P = 1.83 × 10−45). Moreover, elastic net–penalized Cox regression identified a subset of ~30 metabolites that achieved a mean C-index of 0.724 (95% CI = [0.709, 0.738]) for predicting MASLD incidence in the testing data (fig. S6). Most of these metabolites overlapped with those identified in the metabolome-wide association analyses.

Mediation analyses provided evidence of a significantly mediating effect of the general metabolomic signature on the association between physical frailty and MASLD risk while controlling for covariates. The proportion of mediated variance was 7.68% (95% CI = [6.31%, 9.73%], bootstrapping test, P < 2.0 × 10–4; Fig. 4E). The mediation analysis also revealed a partial but significant mediating effect of each of the top 20 MASLD-associated metabolites on the association between frailty and MASLD after adjusting for confounders and multiple comparisons (P < 0.05/20, Bonferroni correction, the mediated effect varying between 2.91 and 6.86%; Fig. 4F). The linoleic acid–to–total fatty acid percentage (LA%, proportion mediated 6.86%, 95% CI = [5.86%, 8.36%]) showed the largest mediating effects, followed by cholesteryl esters in large (L_HDL_CE, 6.17%, 95% CI = [4.63%, 8.08%]) and very large HDL (XL_HDL_CE, 6.72%, 95% CI = [5.40%, 8.48%]), cholesterol in large and very large HDL (L_HDL_C, 5.84%, 95% CI = [4.33%, 7.19%]; XL_HDL_C, 6.17%, 95% CI = [4.83%, 7.53%]), and polyunsaturated fatty acid–to–MUFA ratio (PUFA/MUFA, 5.80%, 95% CI = [4.83%, 7.17%]). Detailed results can be found in table S13.

Physical frailty, metabolites, and MRI-derived PDFF

A total of 40,834 participants with MRI-derived PDFF and 23,459 of them also having complete metabolome biomarkers were used for the validation analyses. The association between physical frailty and MASLD risk defined as PDFF > 5% was slightly attenuated compared with that from the main analysis. Specifically, the risk was 1.33 times higher for prefrailty (OR = 1.33, 95% CI = [1.26, 1.40], P = 1.02 × 10−27) and 2.05 times higher for frailty (OR = 2.05, 95% CI = [1.68, 2.51], P = 2.84 × 10−12; Fig. 5A and table S14) after controlling for covariates. The metabolome-wide association analysis identified 200 metabolites demonstrating significant associations with PDFF (P < 0.05/251), with an effect size ranging from OR = 0.49 to OR = 1.59 (Fig. 5B and table S15). These metabolites demonstrated a generally stronger association with PDFF than with International Classification of Diseases (ICD)–defined MASLD, but the general association pattern was highly similar to each other (r = 0.81, P = 1.77 × 10−60; Fig. 5C). Overall, 153 metabolites showed consistently significant associations with both ICD-defined MASLD and PDFF.

Fig. 5. Analyses using the PDFF-defined MASLD as an outcome yielded nearly unchanged results as those from the main analyses.

Fig. 5.

Open in a new tab

(A) Compared with nonfrailty, the risk was 1.33 times higher for prefrailty and 2.05 times higher for frailty. (B) The metabolome-wide association analysis identified 200 metabolites showing significant associations with PDFF-defined MASLD (P < 0.05/251). (C) The metabolome-wide association pattern of PDFF was highly similar to that of ICD-defined MASLD (r = 0.81, P = 1.77 × 10−60), and 153 metabolites showed consistently significant associations with both measures. (D) Each of the top 20 MASLD-associated metabolites significantly mediated the frailty-PDFF association, with the mediated effect varying between 2.48 and 14.00% (P < 0.05/20, bootstrapping test).

The top 20 MASLD-associated metabolites mediated more variance on the association between frailty and PDFF than between frailty and ICD-defined MASLD (the mediated effect varying between 2.48 and 14.00%; Fig. 5D and table S16). Of these metabolites, L_HDL_CE showed the greatest proportion of mediated variance (14.0%), followed by L_HDL_C (13.7%) and XL_HDL_CE (13.6%).

DISCUSSION

In this population-based cohort study, we found that physical frailty was prospectively and causally associated with an increased risk of clinically diagnosed MASLD. Additive interaction analysis indicated that physical frailty amplified the adverse effect of metabolic syndrome on MASLD risk, implying that frail people may be more vulnerable to MASLD because of metabolic syndrome. Moreover, plasma metabolites, including fatty acids and triglyceride- and cholesteryl ester–rich lipoprotein biomarkers, partially explained the frailty-MASLD relationship, implying a potential metabolomic mechanism.

Accumulating evidence has linked physical frailty and MASLD. However, most studies have focused on using small samples or relied on a cross-sectional design. Only two recent studies have investigated the longitudinal association using data from UK Biobank (9, 25). Specifically, Yang et al. (25) observed a 1.50- and 1.98-fold increased risk of MASLD in prefrail and frail participants, respectively, regardless of genetic predisposition. Zhong et al. (9) demonstrated a similar magnitude of association using two distinct frailty measurements and extended the associations to cirrhosis, liver cancer, and liver-related mortality. Together, they provided important insights into using frailty as a risk factor for multiple chronic liver diseases. Despite a prospective design and adequate adjustment for many confounders, these studies do not allow causal inferences because of biases stemming from reverse causation and residual confounding in observational studies (26). Physical frailty may be associated with MASLD for noncausal reasons, representing a prodromal syndrome preceding the disease occurrence. In this context, improving physical frailty cannot be leveraged as a direct modifier of MASLD risk (27). Our current study corroborated existing results and broadened them by providing causal evidence. Specifically, MR analysis supported a detrimental causal effect of physical frailty on MASLD risk, while MASLD does not appear to have such an effect on frailty. This finding accords with a recent MR study (28) showing a causal relationship between frailty and MASLD. However, this study adopted the deficit-accumulation index to quantify frailty severity (29), which is less modifiable than the frailty phenotype and has limited implications for designing interventions (30). Nevertheless, considering the relatively small number of MASLD cases in the available genome-wide association study (GWAS), statistical power may have been limited. Therefore, larger and updated MR analyses are needed to validate the causal estimates.

Of the five frailty components, slow walking speed and weakness were the most strongly associated with the risk of MASLD, while physical inactivity showed the least associations. This finding provides valuable guidance for devising cost-effective strategies in resource-intensive trials to maximize the benefits of frailty interventions. Identifying high-priority factors and delivering focused intervention to modify specific frailty components may maximize frailty’s effectiveness as a preventative treatment target (11, 31). The MR analyses did not support a causal effect of any individual frailty component on MASLD. Instead, the association may reflect the cumulative physiological decline captured by the overall frailty phenotype, consistent with the multidimensional nature of frailty. Nevertheless, the null findings may also partly reflect limited statistical power or weaker genetic instruments for specific frailty traits.

Metabolic health is a key determinant of MASLD (32–35), but not all individuals with metabolic diseases develop MASLD. The joint association of physical frailty and metabolic health in influencing MASLD risk has not been investigated. Our interaction analysis revealed notable synergistic effects of physical frailty in amplifying the association between metabolic syndrome and MASLD incidence. Specifically, the RERI analysis found that the combined presence of frailty and metabolic syndrome doubled the risk of MASLD compared to the sum of individual risks. Such an interaction may explain the separation of risk trajectories between people with and without metabolic syndrome (11, 25). It also defines a group of high-risk individuals who are extremely vulnerable to developing MASLD and consequently should be given more attention in clinical practice. Accordingly, individuals experiencing either of the two conditions should take actionable measures early to prevent the occurrence of the other. Those people are also likely to benefit the most from intensified interventions. Furthermore, our findings lend support to the hypothesis that physical frailty might function as fuel to exacerbate the adverse effect of metabolic factors (36, 37). Metabolic syndrome and frailty share common risk factors and pathophysiology (38) and act synergistically in a vicious cycle (21). This may partly explain the synergistic effect of these two factors in influencing MASLD, yet whether coexisting metabolic syndrome and physical frailty represent a distinct entity merits further inquiry.

Current research on MASLD prevention has mostly focused on lifestyle and environmental factors (20, 39–41). Our findings have important implications for clinicians in hepatology to integrate frailty assessments into primary care and routine monitoring. However, survey data from a multicenter study of clinicians revealed a relatively low assessment rate of frailty in daily practice, which was primarily limited to geriatricians (42). Although the prevalence of frailty increases with age, middle-aged people with frailty showed a comparable risk of developing MASLD with their old counterparts, thereby justifying increased attention to individuals in midlife.

Moreover, we observed that physical frailty had a greater impact on MASLD than previously established lifestyle factors, such as smoking and sedentariness (7, 42), yet had a comparable effect to that of metabolic syndrome. Compared with indicators of metabolic syndrome, physical frailty is more amenable through nutritional supplementation and enhanced exercise in community settings (7, 21, 42). In this regard, the burden of MASLD attributed to frailty might be alleviated by delaying its onset, although relevant evidence is scarce. Future studies should clarify the extent to which interventions targeting physical frailty could translate into clinical benefits in reducing MASLD risk.

The metabolome-wide association analyses revealed that nearly 88% of the included plasma metabolites showed significant associations with frailty severity. This finding is consistent with the multifactorial nature of physical frailty (43, 44). Likewise, baseline levels of these metabolites also showed widespread associations with MASLD incidence (45). The association patterns of MASLD and frailty were highly similar, indicating a shared metabolomic foundation. Mediation analysis provided further evidence regarding the potential molecular mediators linking physical frailty to MASLD by showing the mediating role of the identified metabolomic signature (40). Notably, ratio measures demonstrated stronger associations with MASLD than absolute measures, possibly because the ratio measures were more stable. It is also possible that the relative balance and interplay of certain metabolites are more important in maintaining liver health than their absolute concentrations (46). The proportion mediated by the metabolomic signature was relatively small, yet it reflects effects beyond adjustment for demographic, anthropometric, socioeconomic, and metabolic covariates and highlights biologically coherent lipid pathways relevant to MASLD. The remaining association likely involves additional mechanisms, such as inflammation and hormonal or muscle-related metabolic dysregulation. Nevertheless, mediation estimates should be interpreted cautiously, as metabolites were measured only at the baseline and repeated assessments are needed to establish temporal ordering and provide more precise estimates of indirect effects.

Among all metabolites, fatty acids and triglyceride-, phospholipid-, and cholesteryl ester–rich lipoprotein biomarkers showed relatively large mediating effects. Triglycerides are the primary form of lipid accumulation in the liver. An imbalance between intrahepatic triglyceride production and secretion can induce hepatic steatosis (18), and the excessively accumulated lipids were hepatotoxic and could propagate an inflammation response and promote oxidative stress, which further contributed to the development and progression of MASLD (19). Certain triglyceride-rich lipoprotein characteristics have a causal relationship with MASLD (45), and evidence from clinical trials demonstrated that triglycerides lowering drugs offered promise for MASLD prevention and treatment (47). The detrimental effect of MUFA% and the protective effect of LA%, PUFA%, and PUFA/MUFA on MASLD also agree with existing studies (2, 45, 48). Specifically, a high MUFA level was detrimental for atherogenesis because of the lipotoxic and adipogenic effects, and its impact on MASLD may occur through cardiovascular factors (49). The protective effect of PUFA in MASLD may originate from its role in lowering inflammation, combating oxidative stress, and enhancing insulin sensitivity (2, 50).

Potential limitations deserve attention. First, four of the five frailty components were self-reported, which can introduce recall bias. Nonetheless, prior studies have shown that self-reported frailty and its test-based alternatives had similar characteristics (51) and were comparable in predicting incident health outcomes (52). Furthermore, given the dynamic nature of physical frailty, a single baseline assessment may not fully capture changes in frailty status over time (13). Repeated frailty assessments will be necessary to determine whether frailty trajectories better predict MASLD risk. Second, because physical frailty and circulating metabolites were measured concurrently at the baseline, the temporal relationship between these variables cannot be established. Thus, the mediation analysis should be interpreted as statistical rather than causal mediation. The identified metabolites may reflect concurrent manifestations of poor health rather than metabolic changes directly caused by frailty. Third, using hospital admission and death records to ascertain MASLD primarily captures individuals who come to medical attention and may therefore represent clinically recognized or relatively advanced MASLD, while milder or subclinical disease in the community may not be fully captured (9). However, our sensitivity analyses using MRI-derived PDFF as a secondary outcome to capture undiagnosed mild cases of MASLD yielded similar results. Therefore, our findings should be interpreted primarily in the context of severe, clinically recognized MASLD. Fourth, the metabolomic coverage of the nuclear magnetic resonance (NMR) platform is lipid focused, mostly covering large molecules and representing only a fraction of the metabolites in human plasma. More diverse metabolomic analyses based on mass spectrometry will be needed to complement the current findings. Fifth, the study population was predominantly of white European ancestry, which may limit the generalizability of our findings to other ethnic groups. Given that the risk of MASLD and related metabolic profiles may differ across populations, further studies in more diverse cohorts are warranted. Last, despite adequate adjustment of multiple confounds, unmeasured residual confounding may inflate the magnitude of the observed associations.

Overall, this study demonstrated that physical frailty and its individual components were associated with a higher risk of clinically recognized MASLD. Moreover, we found causal evidence for physical frailty on MASLD but not for any of the frailty components or the reverse direction. We also provided insights into the molecular basis linking frailty to MASLD by showing that plasma metabolomic signatures may partially explain the association. Incorporating frailty assessment into MASLD screening may help identify high-risk individuals—especially for those with metabolic syndrome, whose risk of developing MASLD might be amplified by frailty status.

MATERIALS AND METHODS

Study population

The UK Biobank is a prospective cohort study comprising more than 500,000 participants aged 37 to 73 years. The study protocol has been described in detail elsewhere (53). Baseline data collection, including touchscreen questionnaires, physical measurements, and biological samples, occurred between 2006 and 2010. UK Biobank had approval from the North West Multicenter Research Ethics Committee (no. 11/NW/0382), and all participants provided written informed consent. Of the 502,180 participants, we excluded those who had missing data or responded “prefer not to answer” or “do not know” for physical frailty or any covariates and had a history of MASLD, other liver-related diseases, or alcohol/drug use disorders at/before the baseline. Figure S7 shows a flowchart illustrating the criteria for participant selection.

Physical frailty assessment

We adopted the Fried frailty phenotype (6) to assess physical frailty, which was based on five indicators, including weight loss, weakness, physical inactivity, exhaustion, and slow walking speed. As per previous studies (54, 55), we modified definitions of some criteria to adapt the data for use in UK Biobank (table S17). Briefly, each of the five indicators was dichotomized into yes or no, and the number of criteria met was summed together, resulting in a severity score ranging from 0 to 5. For consistency with existing studies (54–56), participants were classified as frail, prefrail, or nonfrail if they fulfilled three to five, one to two, or no criteria, respectively.

MASLD assessment

Severe MASLD was defined as hospitalization or death resulting from MASLD or metabolic dysfunction–associated steatohepatitis. The date and cause of hospital admissions were ascertained through record linkage to Health Episode Statistics (England and Wales) and the Scottish Morbidity Records (Scotland). Death register was identified through linkage with the Hospital Episode Statistics for England, Scottish Morbidity Records for Scotland, or Patient Episode Database for Wales. MASLD was defined as ICD-10 codes K76.0 [fatty (change of) liver, not elsewhere classified] and K75.8 (other specified inflammatory liver diseases) (57). Accordingly, our analyses primarily pertain to severe MASLD requiring clinical recognition or hospitalization. Table S18 details the ICD-10 codes used in this study. The follow-up period was calculated from the date of recruitment registration to the first diagnosis of MASLD, death, loss to follow-up, or the censoring date (31 October 2022), whichever came first.

Given that using hospitalization and death register data to identify MASLD incidence may miss some relatively mild cases, we used MRI-derived liver PDFF as a secondary outcome (9). PDFF is an accurate and precise approach for quantifying intrahepatic liver fat content (58). Liver scans were performed using a Siemens 1.5T scanner as part of the abdominal imaging protocol in 2014 [MRI-derived PDFF was measured at the imaging visit, on average ~8 to 9 years after baseline assessment (59)]. Fat-referenced PDFF was calculated from nine regions of interest in the liver, placed while avoiding any inhomogeneities, major vessels, or bile ducts. In accordance with previous studies (9, 60), MASLD diagnosis was identified as PDFF > 5% (fig. S8).

Plasma metabolomics profiling

Metabolomic analysis was performed on baseline EDTA plasma samples from a randomly selected subset of participants using a high-throughput NMR spectroscopy platform provided by Nightingale Health Ltd. (61). Potential systemic and technical variations were mitigated by implementing accredited quality control, and further details can be found at https://biobank.ndph.ox.ac.uk/ukb/ukb/docs/nmrm_companion_doc.pdf. A total of 251 metabolomic biomarkers (170 measures in absolute levels and 81 ratio measures) were identified, which spanned multiple metabolic pathways, including fatty acids, fatty acid compositions, lipoprotein lipids in 14 subclasses, and low-molecular-weight metabolites such as amino acids, ketone bodies, and glycolysis metabolites (62).

Covariates

This study included a total of 10 covariates collected at the baseline. These covariates were chosen on the basis of the existing literature investigating the association between exposures and the risk of MASLD (57, 60). Demographic factors included age [categorized as 37 to 45, 45 to 55, 55 to 65, and 65 to 73 years (55)], sex, and race (white and ethnic minorities). Socioeconomic factors included educational level (above college degree, below college degree, and no formal education), household income [low (<£52,000), middle (£52,000 to £100,000), and high (>£100,000)]; participants reporting “unknown” were designated as a separate group to retain more participants], and material deprivation (quantified using the Townsend deprivation index and divided into quartiles). Lifestyle factors included smoking status (never and ever), alcohol intake status (never and ever), and sedentary status [determined from television watching time and categorized as low (≤2 hours/day), intermediate (2 to 4 hours/day), and high (≥4 hours/day)]. Metabolic syndrome was determined as the presence of at least three of the following indicators: central obesity, diabetes, hypertension, low HDL, and high triglycerides (57). UK Biobank field identifications for all variables used in this study are described in table S19.

Statistical analysis

Association analyses of physical frailty and its components with MASLD incidence

Complete case analysis was performed in this study. Sample characteristics were presented as the means with SD for quantitative variables and as frequencies and percentages for categorical variables. Cox proportional hazard models were used to estimate the HRs and 95% CIs for the association between physical status and MASLD incidence with follow-up time as the time-dependent variable. The analysis used nonfrail individuals as the reference group and adjusted for the 10 covariates described above. The proportional hazard assumption was checked using Schoenfeld residuals, and no violations were detected.

To gauge the contribution of baseline covariates in explaining the frailty-MASLD associations, we calculated the percentage of excess risk mediated (PERM) (63) for socioeconomic variables, lifestyle factors, and metabolic syndrome in separate models. The minimally adjusted model controlled for demographic variables. For each risk factor group, we estimated the PERM as follows: PERM = [HR(demographic variables adjusted) − HR(demographic variables and specific risk factor adjusted)]/[HR(demographic variables adjusted) − 1].

To explore whether the MASLD risk increased alongside the number of frailty indicators, a further model was established by treating frailty scores as a categorical variable. We also fitted exposure-response curves using restricted cubic splines in Cox proportional hazard models to assess potential nonlinearity with the R package “rms.” We investigated the association between five frailty components and the risk of MASLD individually (by establishing a separate model for each component of frailty) and mutually (by including the mutual adjustment of all five components) (54, 56, 64). All analyses were performed using a 5-year landmark analysis, censoring participants experiencing events within the first 5 years of follow-up to account for potential reverse causality (9).

Multiplicative and additive interactions

We performed both multiplicative and additive interaction analyses to assess the modifying effect of covariates on the association between frailty and MASLD incidence. The multiplicative interaction was assessed by incorporating a product term of each of the 10 covariates with frailty into the model. The additive interaction was measured by RERI, AP, and synergy index (SI) (65) using the R package “interactionR,” as previously described (11, 14). The 95% CIs were estimated by drawing 5000 bootstrap samples (11), and if an additive interaction exists, the 95% CI for RERI and AP will not contain zero.

MR analysis

We performed bidirectional two-sample MR analyses to investigate the bidirectional causal association between physical frailty and MASLD using the “TwoSampleMR” package in R (66). Summary statistics for physical frailty were obtained from a recent GWAS study involving 386,565 UK Biobank participants of European descent (67). Summary statistics for MASLD were downloaded from the latest FinnGen consortium (https://r11.finngen.fi/pheno/NAFLD), which included 3006 cases and 450,727 controls. We selected SNPs at a significance threshold of P < 5 × 10−8 and clumped them for independence at r2 > 0.001 with a window of 10,000 kb on the basis of European ancestry reference data from the 1000 Genomes Project. SNPs associated with MASLD (P < 0.01) were excluded from the analyses to prevent horizontal pleiotropy. In the MR analysis, the fixed-effects IVW method was used as the primary method, and the weighted median, simple median, and MR-Egger were implemented as sensitivity analyses.

We performed several additional sensitivity analyses to assess the robustness of the MR estimates, including a modified Cochran’s Q test to evaluate heterogeneity and the MR-Egger regression intercept test to detect horizontal pleiotropy (68). Moreover, we performed the MR-PRESSO test to detect outliers and leave-one-SNP-out analyses to assess whether the overall effect was driven by any single SNP. We also performed MR analysis for each of the five components of physical frailty. Detailed information about the GWAS data is provided in table S10.

Metabolome-wide association analyses

We investigated how physical frailty relates to the 251 metabolomic biomarkers using linear-mixed effect models (69, 70). Within the same analytical framework, physical frailty was fitted as a fixed effect, and the same set of covariates mentioned above was included as additional predictors of no interest (13). The UK Biobank assessment center was modeled as a random effect, and each of the 251 metabolites was fitted as the dependent variable. We extracted the standardized β coefficients and 95% CIs. The prospective association between each of the 251 metabolites and MASLD risk was investigated using Cox proportional hazards models while simultaneously accounting for variables. We used a conservative Bonferroni correction to establish top hits (P < 0.05/251).

To explore the predictive value of metabolites for MASLD incidence, we applied elastic net–penalized Cox regression including all 251 metabolites. The dataset was randomly divided into training (70%) and testing (30%) sets, with model tuning performed using nested 10-fold cross-validation and the penalty parameter selected by the one-standard-error rule (71). The procedure was repeated 100 times to minimize variability resulting from random partitioning, and predictive performance was assessed using the concordance index (C-index).

Mediation analyses

We used the “mediation” package in R to examine the extent to which the metabolomic markers can explain the association between physical frailty and the risk of MASLD (72). For each of the metabolomic biomarkers showing significant associations with both frailty and MASLD risk, we established a standard three-variable path model. In line with previous studies (73, 74), the frailty-metabolite association was examined using multiple linear regression, while the frailty-MASLD and metabolite-MASLD association was examined using survival regression. The significance of mediating effects was determined using 5000 bootstrap iterations.

Given the high intercorrelations between metabolomic biomarkers, we performed a PCA on metabolites identified from the association analysis. The PCA aims to reduce the dimensionality of the data while retaining the most important information related to the metabolic biomarkers. We extracted the first principal component as a general representation of the metabolomic signature of physical frailty (75). Then, we ran the mediation analysis to assess the role of this metabolomic signature in mediating the association between physical frailty and MASLD (20).

We replicated the above association and mediation analyses using PDFF >5% as the definition of MASLD, and associations of physical frailty and metabolites with PDFF were examined using logistic regression rather than Cox proportional models (9). All analyses were conducted in R 4.3.3 and adjusted for the same set of covariates and multiple comparison using Bonferroni correction.

Acknowledgments

Funding:

This work is supported in part by The Scientific and Technological Innovation 2030–The Major Project of the Brain Science and Brain-Inspired Intelligence Technology, 2021ZD0200500 (to J.S.); National Natural Science Foundation of China, 62373062 (to J.S.); the Startup Funds for Talents at Beijing Normal University (to R.J.); and The Fundamental Research Funds for the Central Universities (to J.S.).

Author contributions:

Conceptualization: R.J. and J.W. Methodology: R.J. Investigation: R.J. and J.S. Visualization: R.J., S.Q., and J.W. Validation: R.J., Q.W., and P.W. Software: R.J. and S.Q. Formal analysis: R.J. Resources: R.J. and J.S. Data curation: R.J. and J.S. Supervision: J.S. Writing—original draft: R.J., J.W., and D.S. Writing—review and editing: M.R., S.Q., V.D.C., P.W., J.W., D.S., and J.S. Project administration: R.J. and J.S. Funding acquisition: R.J. and J.S. V.D.C., M.R., and D.S.’s roles were limited to scientific discussion and manuscript contribution. They did not provide data, funding, or other resources for this study.

Competing interests:

The authors declare that they have no competing interests.

Data, code, and materials availability:

The data used in this study were obtained from the UK Biobank resource under application numbers 784650 and 42009. Because of UK Biobank access restrictions, individual-level data cannot be directly shared or linked. Researchers can access the same data by applying to the UK Biobank (www.ukbiobank.ac.uk/) and requesting access to the relevant data fields via the Access Management System (www.ukbiobank.ac.uk/enable-your-research/apply-for-access). Summary statistics for physical frailty are publicly available on Figshare: https://figshare.com/s/6683396c68807fe4e729. Summary statistics for MASLD were downloaded from the latest FinnGen consortium (https://r11.finngen.fi/pheno/NAFLD). Scripts used to perform the analyses are available at Zenodo (https://doi.org/10.5281/zenodo.19121927). All data and code needed to evaluate and reproduce the results in the paper are present in the paper and/or the Supplementary Materials. This study did not generate new materials.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8

Tables S1 to S19

sciadv.aeg4939_sm.pdf (3.1MB, pdf)

REFERENCES

  • 1.Wong V. W., Ekstedt M., Wong G. L., Hagstrom H., Changing epidemiology, global trends and implications for outcomes of NAFLD. J. Hepatol. 79, 842–852 (2023). [DOI] [PubMed] [Google Scholar]
  • 2.Guo C., Liu Z., Fan H., Wang H., Zhang X., Zhao S., Li Y., Han X., Wang T., Chen X., Zhang T., Machine-learning-based plasma metabolomic profiles for predicting long-term complications of cirrhosis. Hepatology 81, 168–180 (2025). [DOI] [PubMed] [Google Scholar]
  • 3.Karlsen T. H., Sheron N., Zelber-Sagi S., Carrieri P., Dusheiko G., Bugianesi E., Pryke R., Hutchinson S. J., Sangro B., Martin N. K., Cecchini M., Dirac M. A., Belloni A., Serra-Burriel M., Ponsioen C. Y., Sheena B., Lerouge A., Devaux M., Scott N., Hellard M., Verkade H. J., Sturm E., Marchesini G., Yki-Jarvinen H., Byrne C. D., Targher G., Tur-Sinai A., Barrett D., Ninburg M., Reic T., Taylor A., Rhodes T., Treloar C., Petersen C., Schramm C., Flisiak R., Simonova M. Y., Pares A., Johnson P., Cucchetti A., Graupera I., Lionis C., Pose E., Fabrellas N., Ma A. T., Mendive J. M., Mazzaferro V., Rutter H., Cortez-Pinto H., Kelly D., Burton R., Lazarus J. V., Gines P., Buti M., Newsome P. N., Burra P., Manns M. P., The EASL-Lancet Liver Commission: Protecting the next generation of Europeans against liver disease complications and premature mortality. Lancet 399, 61–116 (2022). [DOI] [PubMed] [Google Scholar]
  • 4.Powell E. E., Wong V. W.-S., Rinella M., Non-alcoholic fatty liver disease. Lancet 397, 2212–2224 (2021). [DOI] [PubMed] [Google Scholar]
  • 5.Riazi K., Azhari H., Charette J. H., Underwood F. E., King J. A., Afshar E. E., Swain M. G., Congly S. E., Kaplan G. G., Shaheen A. A., The prevalence and incidence of NAFLD worldwide: A systematic review and meta-analysis. Lancet Gastroenterol. Hepatol. 7, 851–861 (2022). [DOI] [PubMed] [Google Scholar]
  • 6.Fried L. P., Tangen C. M., Walston J., Newman A. B., Hirsch C., Gottdiener J., Seeman T., Tracy R., Kop W. J., Burke G., McBurnie M. A., Cardiovascular Health Study Collaborative Research Group , Frailty in older adults: Evidence for a phenotype. J. Gerontol. A Biol. Sci. Med. Sci. 56, M146–M156 (2001). [DOI] [PubMed] [Google Scholar]
  • 7.Kim D. H., Rockwood K., Frailty in older adults. N. Engl. J. Med. 391, 538–548 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Khanmohammadi S., Masrour M., Fallahtafti P., Habibzadeh A., Schuermans A., Kuchay M. S., The relationship between nonalcoholic fatty liver disease and frailty: A systematic review and meta-analysis. Diabetes Metab. Syndr. Clin. Res. Rev. 19, 103187 (2025). [DOI] [PubMed] [Google Scholar]
  • 9.Zhong Q., Zhou R., Huang Y. N., Huang R. D., Li F. R., Chen H. W., Wei Y. F., Liu K., Cao B. F., Liao K. Y., Xu Z. Y., Wang S. A., Wu X. B., Frailty and risk of metabolic dysfunction-associated steatotic liver disease and other chronic liver diseases. J. Hepatol. 82, 427–437 (2025). [DOI] [PubMed] [Google Scholar]
  • 10.Wang S., Whitlock R., Xu C., Taneja S., Singh S., Abraldes J. G., Burak K. W., Bailey R. J., Lai J. C., Tandon P., Frailty is associated with increased risk of cirrhosis disease progression and death. Hepatology 75, 600–609 (2022). [DOI] [PubMed] [Google Scholar]
  • 11.Kim M. S., Shim I., Fahed A. C., Do R., Park W. Y., Natarajan P., Khera A. V., Won H. H., Association of genetic risk, lifestyle, and their interaction with obesity and obesity-related morbidities. Cell Metab. 36, 1494–1503.e3 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.European Association for the Study of the Liver (EASL), European Association for the Study of Diabetes (EASD), European Association for the Study of Obesity (EASO) , EASL-EASD-EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). J. Hepatol. 81, 492–542 (2024). [DOI] [PubMed] [Google Scholar]
  • 13.Jiang R., Noble S., Sui J., Yoo K., Rosenblatt M., Horien C., Qi S., Liang Q., Sun H., Calhoun V. D., Scheinost D., Associations of physical frailty with health outcomes and brain structure in 483 033 middle-aged and older adults: A population-based study from the UK Biobank. Lancet Digit Health 5, e350–e359 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Petermann-Rocha F., Zhou Z., Mathers J. C., Celis-Morales C., Raubenheimer D., Sattar N., Pell J. P., Forrest E., Ho F. K., Diet modifies the association between alcohol consumption and severe alcohol-related liver disease incidence. Nat. Commun. 15, 6880 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Choi K. W., Chen C. Y., Stein M. B., Klimentidis Y. C., Wang M. J., Koenen K. C., Smoller J. W., Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium , Assessment of bidirectional relationships between physical activity and depression among adults: A 2-sample Mendelian randomization study. JAMA Psychiatry 76, 399–408 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Davies N. M., Holmes M. V., Davey Smith G., Reading Mendelian randomisation studies: A guide, glossary, and checklist for clinicians. BMJ 362, k601 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Buergel T., Steinfeldt J., Ruyoga G., Pietzner M., Bizzarri D., Vojinovic D., Belzen J. U. Z., Loock L., Kittner P., Christmann L., Hollmann N., Strangalies H., Braunger J. M., Wild B., Chiesa S. T., Spranger J., Klostermann F., van den Akker E. B., Trompet S., Mooijaart S. P., Sattar N., Jukema J. W., Lavrijssen B., Kavousi M., Ghanbari M., Ikram M. A., Slagboom E., Kivimaki M., Langenberg C., Deanfield J., Eils R., Landmesser U., Metabolomic profiles predict individual multidisease outcomes. Nat. Med. 28, 2309–2320 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Masoodi M., Gastaldelli A., Hyotylainen T., Arretxe E., Alonso C., Gaggini M., Brosnan J., Anstee Q. M., Millet O., Ortiz P., Mato J. M., Dufour J. F., Oresic M., Metabolomics and lipidomics in NAFLD: Biomarkers and non-invasive diagnostic tests. Nat. Rev. Gastroenterol. Hepatol. 18, 835–856 (2021). [DOI] [PubMed] [Google Scholar]
  • 19.Tilg H., Adolph T. E., Dudek M., Knolle P., Non-alcoholic fatty liver disease: The interplay between metabolism, microbes and immunity. Nat. Metab. 3, 1596–1607 (2021). [DOI] [PubMed] [Google Scholar]
  • 20.Ran S., Zhang J., Tian F., Qian Z. M., Wei S., Wang Y., Chen G., Zhang J., Arnold L. D., McMillin S. E., Lin H., Association of metabolic signatures of air pollution with MASLD: Observational and Mendelian randomization study. J. Hepatol. 82, 560–570 (2025). [DOI] [PubMed] [Google Scholar]
  • 21.Mishra M., Wu J., Kane A. E., Howlett S. E., The intersection of frailty and metabolism. Cell Metab. 36, 893–911 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Zhang S., Wang Z., Wang Y., Zhu Y., Zhou Q., Jian X., Zhao G., Qiu J., Xia K., Tang B., Mutz J., Li J., Li B., A metabolomic profile of biological aging in 250,341 individuals from the UK Biobank. Nat. Commun. 15, 8081 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Mutz J., Iniesta R., Lewis C. M., Metabolomic age (MileAge) predicts health and life span: A comparison of multiple machine learning algorithms. Sci. Adv. 10, eadp3743 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Caliendo M., Kopeinig S., Some practical guidance for the implementation of propensity score matching. J. Econ. Surv. 22, 31–72 (2008). [Google Scholar]
  • 25.Yang H., Ou F., Chang Q., Jiang J., Liu Y., Ji C., Chen L., Xia Y., Zhao Y., Physical frailty, genetic predisposition, and the risks of severe non-alcoholic fatty liver disease and cirrhosis: A cohort study. J. Cachexia. Sarcopenia Muscle 15, 1491–1500 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Ebrahim S., Davey Smith G., Mendelian randomization: Can genetic epidemiology help redress the failures of observational epidemiology? Hum. Genet. 123, 15–33 (2008). [DOI] [PubMed] [Google Scholar]
  • 27.Choi K. W., Stein M. B., Nishimi K. M., Ge T., Coleman J. R., Chen C.-Y., Ratanatharathorn A., Zheutlin A. B., Dunn E. C., 23andMe Research Team, Breen G., Koenen K. C., Smoller J. W., An exposure-wide and Mendelian randomization approach to identifying modifiable factors for the prevention of depression. Am. J. Psychiatry 177, 944–954 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.He C., Zheng S., Liu X., Reassessing the causal association between frailty and chronic liver disease: Evidence from genetic analyses. J. Hepatol. 82, e124–e125 (2025). [DOI] [PubMed] [Google Scholar]
  • 29.Rockwood K., Mitnitski A., Frailty in relation to the accumulation of deficits. J. Gerontol. A Biol. Sci. Med. Sci. 62, 722–727 (2007). [DOI] [PubMed] [Google Scholar]
  • 30.Hoogendijk E. O., Afilalo J., Ensrud K. E., Kowal P., Onder G., Fried L. P., Frailty: Implications for clinical practice and public health. Lancet 394, 1365–1375 (2019). [DOI] [PubMed] [Google Scholar]
  • 31.Dent E., Martin F. C., Bergman H., Woo J., Romero-Ortuno R., Walston J. D., Management of frailty: Opportunities, challenges, and future directions. Lancet 394, 1376–1386 (2019). [DOI] [PubMed] [Google Scholar]
  • 32.Targher G., Byrne C. D., Tilg H., MASLD: A systemic metabolic disorder with cardiovascular and malignant complications. Gut 73, 691–702 (2024). [DOI] [PubMed] [Google Scholar]
  • 33.Byrne C. D., Targher G., NAFLD: A multisystem disease. J. Hepatol. 62, S47–S64 (2015). [DOI] [PubMed] [Google Scholar]
  • 34.Eslam M., Fan J. G., Mendez-Sanchez N., Non-alcoholic fatty liver disease in non-obese individuals: The impact of metabolic health. Lancet Gastroenterol. Hepatol. 5, 713–715 (2020). [DOI] [PubMed] [Google Scholar]
  • 35.Eslam M., El-Serag H. B., Francque S., Sarin S. K., Wei L., Bugianesi E., George J., Metabolic (dysfunction)-associated fatty liver disease in individuals of normal weight. Nat. Rev. Gastroenterol. Hepatol. 19, 638–651 (2022). [DOI] [PubMed] [Google Scholar]
  • 36.Solfrizzi V., Scafato E., Custodero C., Loparco F., Ciavarella A., Panza F., Seripa D., Imbimbo B. P., Lozupone M., Napoli N., Piazzolla G., Galluzzo L., Gandin C., Baldereschi M., Di Carlo A., Inzitari D., Pilotto A., Sabba C., Italian Longitudinal Study on Aging Working Group , Liver fibrosis score, physical frailty, and the risk of dementia in older adults: The Italian Longitudinal Study on Aging. Alzheimers Dement. 6, e12065 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Flores Ruano T., Hoogendijk E. O., Romero Rizos L., Ariza Zafra G., León Ortiz M., Luengo Márquez C., Martín Senbastiá E., Navarro López J. L., Fernández Sánchez M., García Molina R., Resting metabolic rate in relation to incident disability and mobility decline among older adults: The modifying role of frailty. Aging Clin. Exp. Res. 35, 591–598 (2023). [DOI] [PubMed] [Google Scholar]
  • 38.Hoogendijk E. O., Huisman M., van Ballegooijen A. J., The role of frailty in explaining the association between the metabolic syndrome and mortality in older adults. Exp. Gerontol. 91, 5–8 (2017). [DOI] [PubMed] [Google Scholar]
  • 39.He P., Zhang Y., Ye Z., Li H., Liu M., Zhou C., Yang S., Gan X., Zhang Y., Qin X., A healthy lifestyle, Life’s Essential 8 scores and new-onset severe NAFLD: A prospective analysis in UK Biobank. Metabolism 146, 155643 (2023). [DOI] [PubMed] [Google Scholar]
  • 40.Wu H., Wei J., Wang S., Chen L., Zhang J., Wang N., Tan X., Dietary pattern modifies the risk of MASLD through metabolomic signature. JHEP Rep. 6, 101133 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Ge X., Wang X., Yan Y., Zhang L., Yu C., Lu J., Xu X., Gao J., Liu M., Jiang T., Ke B., Song C., Behavioural activity pattern, genetic factors, and the risk of nonalcoholic fatty liver disease: A prospective study in the UK Biobank. Liver Int. 43, 1287–1297 (2023). [DOI] [PubMed] [Google Scholar]
  • 42.Ryu S., Chang Y., Jung H. S., Yun K. E., Kwon M. J., Choi Y., Kim C. W., Cho J., Suh B. S., Cho Y. K., Chung E. C., Shin H., Kim Y. S., Relationship of sitting time and physical activity with non-alcoholic fatty liver disease. J. Hepatol. 63, 1229–1237 (2015). [DOI] [PubMed] [Google Scholar]
  • 43.Howlett S. E., Rutenberg A. D., Rockwood K., The degree of frailty as a translational measure of health in aging. Nat. Aging 1, 651–665 (2021). [DOI] [PubMed] [Google Scholar]
  • 44.Mak J. K. L., Kananen L., Qin C., Kuja-Halkola R., Tang B., Lin J., Wang Y., Jaaskelainen T., Koskinen S., Lu Y., Magnusson P. K. E., Hagg S., Jylhava J., Unraveling the metabolic underpinnings of frailty using multicohort observational and Mendelian randomization analyses. Aging Cell 22, e13868 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Gagnon E., Manikpurage H. D., Mitchell P. L., Girard A., Gobeil E., Bourgault J., Begin F., Marette A., Theriault S., Arsenault B. J., Large-scale metabolomic profiling and incident non-alcoholic fatty liver disease. iScience 26, 107127 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Hu W., Wang W., Liao H., Bulloch G., Zhang X., Shang X., Huang Y., Hu Y., Yu H., Yang X., He M., Zhu Z., Metabolic profiling reveals circulating biomarkers associated with incident and prevalent Parkinson’s disease. NPJ Parkinsons Dis. 10, 130 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Das Pradhan A., Glynn R. J., Fruchart J. C., MacFadyen J. G., Zaharris E. S., Everett B. M., Campbell S. E., Oshima R., Amarenco P., Blom D. J., Brinton E. A., Eckel R. H., Elam M. B., Felicio J. S., Ginsberg H. N., Goudev A., Ishibashi S., Joseph J., Kodama T., Koenig W., Leiter L. A., Lorenzatti A. J., Mankovsky B., Marx N., Nordestgaard B. G., Pall D., Ray K. K., Santos R. D., Soran H., Susekov A., Tendera M., Yokote K., Paynter N. P., Buring J. E., Libby P., Ridker P. M., PROMINENT Investigators , Triglyceride lowering with pemafibrate to reduce cardiovascular risk. N. Engl. J. Med. 387, 1923–1934 (2022). [DOI] [PubMed] [Google Scholar]
  • 48.Liu Z., Huang H., Xie J., Xu Y., Xu C., Circulating fatty acids and risk of hepatocellular carcinoma and chronic liver disease mortality in the UK Biobank. Nat. Commun. 15, 3707 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Zhuang P., Liu X., Li Y., Wu Y., Li H., Wan X., Zhang L., Xu C., Jiao J., Zhang Y., Circulating fatty acids, genetic risk, and incident coronary artery disease: A prospective, longitudinal cohort study. Sci. Adv. 9, eadf9037 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Jia X., Fan J., Wu X., Cao X., Ma L., Abdelrahman Z., Zhao F., Zhu H., Bizzarri D., Akker E., Slagboom P. E., Deelen J., Zhou D., Liu Z., A novel metabolomic aging clock predicting health outcomes and its genetic and modifiable factors. Adv. Sci. 11, e2406670 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Theou O., O’Connell M. D., King-Kallimanis B. L., O’Halloran A. M., Rockwood K., Kenny R. A., Measuring frailty using self-report and test-based health measures. Age Ageing 44, 471–477 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Papachristou E., Wannamethee S. G., Lennon L. T., Papacosta O., Whincup P. H., Iliffe S., Ramsay S. E., Ability of self-reported frailty components to predict incident disability, falls, and all-cause mortality: Results from a population-based study of older British men. J. Am. Med. Dir. Assoc. 18, 152–157 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Sudlow C., Gallacher J., Allen N., Beral V., Burton P., Danesh J., Downey P., Elliott P., Green J., Landray M., UK biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 12, e1001779 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Jiang R., Noble S., Rosenblatt M., Dai W., Ye J., Liu S., Qi S., Calhoun V. D., Sui J., Scheinost D., The brain structure, inflammatory, and genetic mechanisms mediate the association between physical frailty and depression. Nat. Commun. 15, 4411 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Hanlon P., Nicholl B. I., Jani B. D., Lee D., McQueenie R., Mair F. S., Frailty and pre-frailty in middle-aged and older adults and its association with multimorbidity and mortality: A prospective analysis of 493 737 UK Biobank participants. Lancet Public Health 3, e323–e332 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Petermann-Rocha F., Lyall D. M., Gray S. R., Esteban-Cornejo I., Quinn T. J., Ho F. K., Pell J. P., Celis-Morales C., Associations between physical frailty and dementia incidence: A prospective study from UK Biobank. Lancet Health Longev. 1, E58–E68 (2020). [DOI] [PubMed] [Google Scholar]
  • 57.Petermann-Rocha F., Gray S. R., Forrest E., Welsh P., Sattar N., Celis-Morales C., Ho F. K., Pell J. P., Associations of muscle mass and grip strength with severe NAFLD: A prospective study of 333,295 UK Biobank participants. J. Hepatol. 76, 1021–1029 (2022). [DOI] [PubMed] [Google Scholar]
  • 58.Starekova J., Hernando D., Pickhardt P. J., Reeder S. B., Quantification of liver fat content with CT and MRI: State of the art. Radiology 301, 250–262 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Wilman H. R., Kelly M., Garratt S., Matthews P. M., Milanesi M., Herlihy A., Gyngell M., Neubauer S., Bell J. D., Banerjee R., Thomas E. L., Characterisation of liver fat in the UK Biobank cohort. PLOS ONE 12, e0172921 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Lv Y., Rong S., Deng Y., Bao W., Xia Y., Chen L., Plant-based diets, genetic predisposition and risk of non-alcoholic fatty liver disease. BMC Med. 21, 351 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Soininen P., Kangas A. J., Wurtz P., Suna T., Ala-Korpela M., Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics. Circ. Cardiovasc. Genet. 8, 192–206 (2015). [DOI] [PubMed] [Google Scholar]
  • 62.Julkunen H., Cichonska A., Tiainen M., Koskela H., Nybo K., Makela V., Nokso-Koivisto J., Kristiansson K., Perola M., Salomaa V., Jousilahti P., Lundqvist A., Kangas A. J., Soininen P., Barrett J. C., Wurtz P., Atlas of plasma NMR biomarkers for health and disease in 118,461 individuals from the UK Biobank. Nat. Commun. 14, 604 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Lin D. Y., Fleming T. R., De Gruttola V., Estimating the proportion of treatment effect explained by a surrogate marker. Stat. Med. 16, 1515–1527 (1997). [DOI] [PubMed] [Google Scholar]
  • 64.Zheng Z., Lv Y., Rong S., Sun T., Chen L., Physical frailty, genetic predisposition, and incident Parkinson disease. JAMA Neurol. 80, 455–461 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Li R., Chambless L., Test for additive interaction in proportional hazards models. Ann. Epidemiol. 17, 227–236 (2007). [DOI] [PubMed] [Google Scholar]
  • 66.Hemani G., Zheng J., Elsworth B., Wade K. H., Haberland V., Baird D., Laurin C., Burgess S., Bowden J., Langdon R., Tan V. Y., Yarmolinsky J., Shihab H. A., Timpson N. J., Evans D. M., Relton C., Martin R. M., Davey Smith G., Gaunt T. R., Haycock P. C., The MR-Base platform supports systematic causal inference across the human phenome. eLife 7, e34408 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Ye Y., Noche R. B., Szejko N., Both C. P., Acosta J. N., Leasure A. C., Brown S. C., Sheth K. N., Gill T. M., Zhao H., Falcone G. J., A genome-wide association study of frailty identifies significant genetic correlation with neuropsychiatric, cardiovascular, and inflammation pathways. Geroscience 45, 2511–2523 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Burgess S., Thompson S. G., Interpreting findings from Mendelian randomization using the MR-Egger method. Eur. J. Epidemiol. 32, 377–389 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Firth J., Firth J. A., Stubbs B., Vancampfort D., Schuch F. B., Hallgren M., Veronese N., Yung A. R., Sarris J., Association between muscular strength and cognition in people with major depression or bipolar disorder and healthy controls. JAMA Psychiatry 75, 740–746 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Jiang R., Westwater M. L., Noble S., Rosenblatt M., Dai W., Qi S., Sui J., Calhoun V. D., Scheinost D., Associations between grip strength, brain structure, and mental health in> 40,000 participants from the UK Biobank. BMC Med. 20, 286 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 71.Sui J., Jiang R., Bustillo J., Calhoun V., Neuroimaging-based individualized prediction of cognition and behavior for mental disorders and health: Methods and promises. Biol. Psychiatry 88, 818–828 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 72.Tingley D., Yamamoto T., Hirose K., Keele L., Imai K., Mediation: R package for causal mediation analysis. J. Stat. Softw. 59, 1–38 (2014).26917999 [Google Scholar]
  • 73.Deng Y. T., Li Y. Z., Huang S. Y., Ou Y. N., Zhang W., Chen S. D., Zhang Y. R., Yang L., Dong Q., Feng J. F., Suckling J., Smith A. D., Cheng W., Yu J. T., Association of life course adiposity with risk of incident dementia: A prospective cohort study of 322,336 participants. Mol. Psychiatry 27, 3385–3395 (2022). [DOI] [PubMed] [Google Scholar]
  • 74.Jiang R., Geha P., Rosenblatt M., Wang Y., Fu Z., Foster M., Dai W., Calhoun V. D., Sui J., Spann M. N., Scheinost D., The inflammatory and genetic mechanisms underlying the cumulative effect of co-occurring pain conditions on depression. Sci. Adv. 11, eadt1083 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.Xu H., Xu H., Wu J., Wang L., Guo B., Li W., Zhang J., Xiao X., Zhao X., Ambient air pollution exposure, plasma metabolomic markers, and risk of type 2 diabetes: A prospective cohort study. J. Hazard. Mater. 463, 132844 (2023). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figs. S1 to S8

Tables S1 to S19

sciadv.aeg4939_sm.pdf (3.1MB, pdf)

Data Availability Statement

The data used in this study were obtained from the UK Biobank resource under application numbers 784650 and 42009. Because of UK Biobank access restrictions, individual-level data cannot be directly shared or linked. Researchers can access the same data by applying to the UK Biobank (www.ukbiobank.ac.uk/) and requesting access to the relevant data fields via the Access Management System (www.ukbiobank.ac.uk/enable-your-research/apply-for-access). Summary statistics for physical frailty are publicly available on Figshare: https://figshare.com/s/6683396c68807fe4e729. Summary statistics for MASLD were downloaded from the latest FinnGen consortium (https://r11.finngen.fi/pheno/NAFLD). Scripts used to perform the analyses are available at Zenodo (https://doi.org/10.5281/zenodo.19121927). All data and code needed to evaluate and reproduce the results in the paper are present in the paper and/or the Supplementary Materials. This study did not generate new materials.

Articles from Science Advances are provided here courtesy of American Association for the Advancement of Science

RESOURCES