A multidimensional biomarker model of vitality and its associations with intrinsic capacity and frailty Authors: J Chew, JQ Chia, JP Lim, A Yeo, CN Tan, SP Chan, CL Lim, WS Lim Published: 2026 ABSTRACT AbstractBackgroundThe World Health Organisation has identified intrinsic capacity (IC) as a key determinant of healthy ageing. Within the IC framework, vitality reflects physiological reserve arising from neuromuscular, metabolic and inflammatory/stress-response systems. However, this multidomain physiological construct has not been empirically validated, and vitality is often operationalised using nutritional assessments alone.ObjectivesWe evaluated a biomarker-derived vitality construct and examined its associations with IC domains and frailty, compared to nutritional measures.MethodsCross-sectional analysis of 300 community-dwelling, functionally-independent older adults (mean (SD) age 67.4 (7.1) years; 69% female) was conducted using exploratory (EFA) and confirmatory factor analyses (CFA) to characterise the latent structure of biomarker-defined vitality. Factor scores were examined for associations with IC domains, composite IC and frailty using linear regression, compared with the Mini Nutritional Assessment–Short Form (MNA-SF).ResultsEFA supported a multidimensional structure with biomarkers clustering into neuromuscular, metabolic and inflammatory domains. CFA supported a correlated three-factor model for vitality; while a bifactor model showed superior global fit, general factor dominance was limited (ECV = 0.28; ωH = 0.26), with most variance at the domain level. Higher neuromuscular vitality was associated with greater composite IC (β = 1.39, 95% CI 0.25–2.54) and cognitive capacity (β = 1.74, 95% CI 0.52–2.95). A better metabolic vitality profile was associated with higher locomotor capacity (β = 1.74, 95% CI 0.58–2.91), while a more favourable inflammatory vitality profile was associated with higher locomotor capacity (β = 1.33, 95% CI 0.09–2.57) and reduced frailty (β = −0.11, 95% CI − 0.22 to −0.022). In contrast, MNA-SF scores were associated only with lower frailty (β = −0.09, 95% CI − 0.17 to −0.004), and not with IC.ConclusionsVitality is better represented as a multidimensional, physiological construct than by nutrition-based measures alone. Distinct biomarker-derived vitality domains were associated with IC and frailty, supporting its role in characterising physiological determinants of healthy ageing. Longitudinal research is required to establish its prognostic value.Keywords: Intrinsic capacity, Vitality, Frailty, Biological ageing, Biomarkers, Nutrition Background Intrinsic capacity (IC) has been proposed as a framework for healthy ageing that emphasises functional ability rather than disease [1]. In practice, IC has been operationalised within the World Health Organisation (WHO) Integrated Care for Older People (ICOPE) framework [2] primarily as a population-level screening approach to identify early functional decline and guide community-based interventions [3]. IC can also be viewed as an integrative phenotype that captures the expression of underlying physical and mental capacities across the life course [4], with declines preceding the emergence of overt disability and adverse health outcomes [5]. Within this framework, vitality is a foundational domain reflecting physiological reserve and the integrity of systems that underpin the expression of locomotor, cognitive, sensory and psychological capacities [6,7], with impairments preceding declines in IC domains and capturing inter-individual differences in underlying reserve [8]. Despite this central role, vitality remains variably defined and inconsistently measured [9]. The WHO working definition describes vitality as a physiological state arising from normal or accelerated biological ageing processes, resulting from interactions among multiple physiological systems, which is reflected in the level of energy and metabolism, neuromuscular function, and immune and stress response functions of the body [10]. However, in research and clinical practice, vitality is frequently operationalised using nutritional status, most commonly with the Mini Nutritional Assessment (MNA) [11], reflecting the pragmatic need to identify clinically actionable nutritional risk within care pathways. While nutrition is a crucial health determinant, it may not fully represent the broader concept of physiological reserve, particularly in healthier, community-dwelling populations in whom undernutrition is infrequent and declines in reserve may precede detectable nutritional impairment. This approach effectively characterises vitality as a downstream nutritional phenotype rather than the underlying physiological reserve it is intended to represent [12]. In line with this framing of vitality as physiological reserve across multiple biological systems rather than a single observable deficit, we aimed to develop and validate a biomarker-based operationalisation of vitality within the IC framework. Guided by the WHO working definition encompassing neuromuscular, metabolic and inflammatory domains, we aimed to characterise the latent structure of biomarkers across these domains and to examine their associations with other IC domains and frailty. Establishing a biomarker-based operational definition of vitality clarifies this construct beyond nutrition-based proxies and supports a clearer model of physiological reserve capacity that will guide the operationalisation of vitality in future research and clinical practice in older adults. Methods Study population Participants were recruited from the Longitudinal Assessment of Biomarkers for Characterisation of Early Sarcopenia and Predicting Frailty and Functional Decline in Community-Dwelling Asian Older Adults (GeriLABS-2) study, which is a prospective cohort of community-dwelling adults aged 50–99 years old. The present analyses included 300 participants who met eligibility criteria, including residence in the community, independence in activities of daily living and instrumental activities of daily living, intact cognition and the ability to walk at least 4.5 metres independently. Individuals with dementia, cognitive impairment or residence in sheltered or nursing homes were excluded. Vitality biomarkers Vitality was operationalised using a panel of physiological and circulating biomarkers aligned with the WHO working definition of vitality capacity [10]. This definition was developed through an expert consensus process to clarify the attributes of vitality capacity and distinguish it from the expressed domains of IC. In this definition, vitality is a physiological state arising from the interaction of multiple systems and reflected in energy and metabolism, neuromuscular function, and immune and stress response processes. The expert group further identified candidate markers within each of the domains. Accordingly, biomarkers were selected to represent these consensus-derived physiological domains contributing to vitality capacity. Neuromuscular function Neuromuscular biomarkers included appendicular lean mass index (ALMI), handgrip strength, and forced expiratory volume in 1 second (FEV1). ALMI was derived from dual-energy X-ray absorptiometry (DXA, Horizon W, Hologic Inc.) as the sum of appendicular lean mass normalised for height squared (kg/m2). Maximum handgrip strength was measured using a handheld Jamar dynamometer according to a standardised protocol [13], and FEV1 was obtained using a portable spirometer (EasyOne Air, NDD Medical Technologies) following standardised procedures. Together, ALMI, handgrip strength and FEV1 captures complementary aspects of neuromuscular physiological reserve. ALMI and handgrip strength are core components of sarcopenia, a key antecedent of frailty and functional decline, and reflects the muscle reserves that underlie expressed capacities, including locomotion. Handgrip strength has also been associated with better IC across domains [14]. FEV1 extends the construct to respiratory reserve, a biomarker reflecting integrated cardiopulmonary, ventilatory and muscular capacity rather than airway pathology alone [15]. FEV1 is a well-established correlate of functional limitation and disability, with higher levels of lung functional reserve associated with better mobility and functional performance [16]. Energy and metabolism Metabolic biomarkers included insulin-like growth factor-1 (IGF-1), visceral adipose tissue index (VATI; kg/m2, derived from DXA-measured visceral fat mass normalised to height squared), leptin, and insulin resistance using the homeostasis model assessment (HOMA-IR). Circulating IGF-1 and leptin concentrations were measured from fasting blood samples, and HOMA-IR was calculated from fasting glucose and insulin concentrations. These biomarkers were selected to represent metabolic reserve, reflecting the capacity to maintain energy homeostasis through anabolic signaling, adipose-derived endocrine signals and insulin-mediated metabolic pathways. IGF-1 captures anabolic endocrine signaling through the insulin/IGF-1 axis, with age-related decline in IGF-1 contributing to anabolic resistance in skeletal muscle [17]. Lower circulating levels of IGF-1 has also been independently associated with frailty and low muscle mass in community-dwelling older adults [18]. VATI and leptin were included to capture adiposity-related contributions to metabolic reserve. VATI captures visceral adiposity, a metabolically active fat depot that is more closely linked to dysmetabolism than overall adiposity [19]. Given the role of visceral adipose tissue in insulin resistance, adipokine dysregulation and inflammatory cytokine production, leptin was included as an adipose-derived endocrine signal involved in energy homeostasis, regulation of appetite and glucose metabolism [20]. HOMA-IR captures insulin resistance and impaired glucose–insulin homeostasis, which have been associated with frailty [21] and cognitive decline in older adults [22]. These biomarkers therefore capture complementary dimensions of metabolic reserve rather than conventional cardiometabolic risk alone, linking altered energy homeostasis to vulnerability to reduced IC and frailty. Immune and stress response Pro-inflammatory and stress response biomarkers included C-reactive protein (CRP), interleukin-6 (IL-6), tumour necrosis factor receptor 1 (TNFR1), monocyte chemoattractant protein-1 (MCP-1) and growth differentiation factor-15 (GDF-15). These markers were selected to capture ageing-related immune and stress-response dysregulation, spanning systemic inflammation, inflammatory pathway activation, innate immune recruitment and cellular stress signalling. CRP and IL-6 reflect chronic low-grade inflammation, a hallmark of inflammageing that has been associated with mortality, frailty and functional decline [23,24]. TNFR1 captures tumour necrosis factor pathway activation, linking inflammatory signaling to catabolism, endothelial dysfunction and ageing-related vulnerability, including low IC [25]. MCP-1 reflects chemokine-mediated monocyte recruitment and innate immune activation, while GDF-15 captures cellular stress-response biology related to mitochondrial stress, inflammation and senescence-associated pathways [26]. Collectively, this panel operationalises the immune and stress-response component of vitality capacity, linking inflammageing and stress-response dysregulation to reduced IC, frailty and adverse ageing-related outcomes. Blood sample collection, processing and quantification For biomarker quantification, fasting venous blood samples were collected in plain serum tubes and processed according to standardised protocols. Following collection, samples were allowed to clot at room temperature for 60 min. Clotted samples were then centrifuged at 3,000 × g for 10 min at 24 °C. The resulting serum was aliquoted and stored at −80 °C until analysis. Serum IGF-1 was quantified using a magnetic bead-based multiplex immunoassay (MILLIPLEX® MAP Human IGF-1, II Magnetic Bead Panel, Merck Millipore) on a Luminex platform (R&D Systems, Inc.). Leptin, TNFR1, MCP-1 and GDF-15 were measured using Human Luminex Discovery Assays (R&D Systems, Inc.), and IL-6 was measured using the SIMOA IL-6 Advantage PLUS Assay (Quanterix). Myostatin was measured using a Quantikine ELISA (R&D Systems, Inc.). CRP and fasting glucose were measured using standard automated clinical chemistry methods, employing immunoturbidimetric and hexokinase assays, respectively. Serum insulin was quantified using a chemiluminescent immunoassay (Alinity i Insulin assay, Abbott). HOMA-IR was calculated from fasting glucose and insulin concentrations. Biomarker distributions were inspected prior to analysis and variables showing skewness were log-transformed, and standardised with z-scores prior to latent variable modelling. Nutritional assessment Nutritional status was assessed using the Mini Nutritional Assessment Short Form (MNA-SF) [27], a validated multidimensional nutrition screening tool incorporating appetite, recent weight loss, mobility, psychological stress or acute illness, neuropsychological problems, and body mass index. MNA-SF was included to enable comparison between a conventional nutrition-based measure and biomarker-derived vitality constructs. Exploratory factor analysis (EFA) EFA was performed on standardised vitality biomarkers to examine their underlying covariance structure. Sampling adequacy was assessed using the Kaiser–Meyer–Olkin (KMO) measure, and sphericity was evaluated using Bartlett’s test. Principal component analysis (PCA) was used for component extraction [28]. The number of components retained was guided by eigenvalues greater than one, inspection of scree plots, and parallel analysis comparing observed eigenvalues with those derived from randomly generated data [29]. Components were rotated using varimax rotation. Component loadings were examined to characterise biomarker contributions to each component. Biomarkers with loadings ≥0.50 were considered to contribute meaningfully to a component, while cross-loadings were evaluated for interpretability and stability [30]. Biomarkers with weak loadings were not retained in subsequent confirmatory factor analyses. Confirmatory factor analysis (CFA) Confirmatory factor analysis (CFA) was conducted to evaluate candidate measurement models of vitality informed by the EFA and the conceptual framework of vitality capacity. Alternative model specifications were tested. These included a unidimensional first-order model in which all biomarkers loaded onto a single latent factor. First-order multidimensional models were also examined, with biomarkers loading onto domain-specific latent factors representing neuromuscular, metabolic and inflammatory domains, specified as either uncorrelated or correlated. Second-order models were subsequently evaluated, in which correlations among domain-specific factors were accounted for by a higher-order vitality factor. Finally, a bifactor model was specified, comprising a general vitality factor alongside orthogonal domain-specific factors, to evaluate the relative contributions of shared and domain-specific variance. All CFA models were fitted to standardised biomarker data using maximum likelihood estimation [31]. Model fit was evaluated using the χ2 statistic, comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA), and standardised root mean square residual (SRMR) [32]. For the bifactor model, the omega total (ω) and omega hierarchical (ωH) indices were examined to quantify the reliability of the general factor, alongside the explained common variance (ECV), to assess the extent to which the variance was attributable to the general factor relative to domain-specific factors [33]. Standardised factor loadings and inter-factor correlations were examined to assess the interpretability of latent constructs. Model selection was informed by global fit indices in conjunction with factor reliability metrics (ω, ωH, and ECV), the magnitude and pattern of factor loadings, and the interpretability of general and domain-specific factors, rather than relying on improvements in comparative fit indices alone. Derivation of vitality factor scores Factor scores were derived from the final CFA model to obtain individual-level estimates of vitality and its constituent domains. Scores were estimated using the regression method [34]. These factor scores were used in subsequent analyses examining associations with frailty and IC domains. Frailty and intrinsic capacity (IC) measures IC domains other than vitality were assessed in accordance with the WHO IC framework. Locomotor capacity was assessed using the Short Physical Performance Battery (SPPB) [35]. Psychological capacity was assessed using the 15-item Geriatric Depression Scale (GDS-15) [36]. Cognitive capacity was assessed using a modified version of the Chinese Mini-Mental State Examination (mCMMSE) [37], with a maximum score of 28. Sensory capacity was assessed using hearing (whispered voice test) and vision (Snellen chart). Following a previously described approach [38], each IC domain was rescaled to a 0–100 score, with 100 representing the best performance and 0 the worst performance on the original measure. Sensory capacity was calculated as the average of vision and hearing scores. A composite IC score was derived as the arithmetic mean of the five domain scores. Frailty was defined using the Fried criteria [39]. Frailty components included: (1) “weight loss,” defined as body mass index (BMI) <18.5 kg/m2; (2) “exhaustion,” assessed by self-reported fatigue based on two items from the Center for Epidemiologic Studies Depression scale [40]; (3) “slowness,” defined as gait speed <1.0 m/s according to Asian Working Group for Sarcopenia criteria [41]; and (4) “low physical activity,” defined as a Frenchay Activities Index score ≤29 [42]. Each frailty component was operationalised as a binary indicator (1 = criterion present, 0 = absent), and the total frailty score (range 0–4) was derived as the sum of these components. Handgrip strength was not included in the frailty definition to avoid overlap with neuromuscular vitality indicators. Statistical analysis Descriptive statistics were used to summarise participant characteristics and study variables. Continuous variables were presented as means and standard deviations or medians and interquartile ranges, as appropriate, and categorical variables as counts and percentages. Associations between vitality factor scores and other IC domains and composite scores were examined using linear regression models. Frailty was analysed as a continuous variable, and associations between vitality factor scores and frailty were also evaluated using linear regression. Both unadjusted and adjusted models were fitted. Adjusted models included age, sex and comorbidity burden as covariates. As a sensitivity analysis, the frailty phenotype was additionally analysed as a count outcome using Poisson regression to assess the robustness of findings to model specification. To address potential overlap between the BMI component of the frailty phenotype and the nutritional assessment captured within the MNA-SF, an additional analysis was performed in which the BMI item was excluded from the MNA-SF. Statistical significance was defined as a two-sided p value <0.05. All analyses were performed using R (version 4.5.2). Results Participant characteristics A total of 300 community-dwelling older adults were included in this study. The mean age was 67.4 years (SD 7.1) and 68.7% were women participants. Participants were predominantly Chinese (95.0%) and had a mean of 11.9 years (SD 3.8) of formal education. Hypertension (32.0%), hyperlipidaemia (50.0%) and diabetes mellitus (14.0%) were the most common comorbidities. Overall, the cohort was functionally robust at baseline, with a mean Modified Barthel Index (MBI) score of 98.1 (SD 3.5) and a mean Lawton Instrumental Activities of Daily Living (IADL) score of 22.5 (SD 1.1), indicating high levels of independence in both basic and instrumental daily activities. Baseline IC domain scores were generally high across domains, with mean scaled scores ranging from 88.3 (SD 15.3) for sensory capacity to 94.6 (SD 11.4) for locomotor capacity. Baseline participant characteristics and IC domain scores are summarised in Table 1. Table 1Baseline characteristics, intrinsic capacity scores, and frailty status of study participants (N = 300).Table 1VariableValueDemographics, mean (SD) or n (%)Age, years67.4 (7.1)Female sex206 (68.7)Ethnicity– Chinese285 (95.0)– Malay1 (0.3)– Indian10 (3.3)– Other4 (1.3)Education, years11.9 (3.8)Smoker19 (6.3)Comorbidities, n (%)Hypertension96 (32.0)Hyperlipidaemia150 (50.0)Diabetes mellitus42 (14.0)Ischemic heart disease9 (3.0)Atrial fibrillation5 (1.7)Stroke or transient ischemic attack8 (2.7)Chronic kidney disease4 (1.3)Lung disease12 (4.0)Arthritis20 (6.7)Autoimmune disease6 (2.0)Functional and clinical measures, mean (SD)Activities of daily living (Modified Barthel Index, range 0−100)98.1 (3.5)Instrumental activities of daily living (Lawton Index, range 0−28)22.5 (1.1)Body mass index (BMI), kg/m²23.4 (3.1)Mini Nutritional Assessment-Short Form (MNA-SF) score, range 0−1413.2 (1.1)Geriatric Depression Scale (GDS-15), range 0−151.5 (1.6)mCMMSE score, range 0−2826.1 (1.7)Intrinsic capacity domain scores (0–100), mean (SD)*Cognitive capacity93.1 (5.9)Psychological capacity90.3 (12.4)Locomotor capacity94.6 (11.4)Sensory capacity (vision and hearing)88.3 (15.3)Composite intrinsic capacity score (excluding vitality)91.6 (6.9)Note: *Intrinsic capacity (IC) domain scores were scaled from 0 to 100, with higher scores indicating better capacity. The composite IC score was calculated as the mean of all intrinsic capacity domains excluding vitality.Abbreviations: mCMMSE, modified version of the Chinese Mini-Mental State Examination. Baseline characteristics, intrinsic capacity scores, and frailty status of study participants (N = 300). Note: *Intrinsic capacity (IC) domain scores were scaled from 0 to 100, with higher scores indicating better capacity. The composite IC score was calculated as the mean of all intrinsic capacity domains excluding vitality. Abbreviations: mCMMSE, modified version of the Chinese Mini-Mental State Examination. Biomarker characteristics Baseline distributions of biomarkers included in the exploratory factor analysis are summarised in Supplementary Material, Table S1. Biomarkers spanned neuromuscular, metabolic and inflammatory domains. Exploratory factor analysis (EFA) Exploratory factor analysis was performed on 12 standardised biomarkers. Sampling adequacy was supported by a Kaiser–Meyer–Olkin measure of 0.78, and Bartlett’s test of sphericity was significant (χ2 = 324.9, p < 0.001). Inspection of eigenvalues and scree plots suggested a multidimensional structure, with five components exceeding an eigenvalue of one. Parallel analysis supported retention of three components, which together accounted for 51.7% of the total variance. Detailed eigenvalues, scree plots and parallel analysis results used to guide factor retention are provided in Supplementary Material, Tables S2, S3; Figure S1. In the varimax-rotated solution, three components were retained. Using a loading threshold of ≥0.50, the first component was characterised by ALMI, HGS and FEV1. The second component was characterised by VATI, HOMA-IR and leptin. The third component was characterised by TNFR1, GDF-15 and IL-6. Loadings clustered into neuromuscular, metabolic and inflammatory or stress response domains (Table 2). Table 2Exploratory factor analysis of biomarker-defined vitality domains.Table 2BiomarkerNeuromuscularMetabolicInflammatory/Stress responseAppendicular Lean Mass Index (ALMI)0.863−0.178−0.076Maximum Handgrip Strength (HGS)0.8610.0670.104Forced Expiratory Volume in 1 s (FEV1)0.7770.1820.283Visceral Adipose Tissue Index (VATI)−0.0120.8030.225HOMA-IR−0.0400.7810.183Leptin0.3130.646−0.148Tumour Necrosis Factor Receptor 1 (TNFR1)−0.1640.2470.714Growth Differentiation Factor-15 (GDF-15)−0.1390.0650.716Interleukin-6 (IL-6)0.0990.0330.557Insulin-like Growth Factor-1 (IGF-1)0.246−0.3310.459Monocyte Chemoattractant Protein-1 (MCP-1)0.1140.0590.233Myostatin−0.301−0.0920.241Note: Loadings ≥0.50 are bolded as meaningful contributions (per Methods). Exploratory factor analysis of biomarker-defined vitality domains. Note: Loadings ≥0.50 are bolded as meaningful contributions (per Methods). Confirmatory Factor Analysis (CFA) CFA was conducted to evaluate structural representations of biomarker-defined vitality, as shown in Fig. 1. A unidimensional model specifying all biomarkers to load onto a single latent factor demonstrated poor fit (χ2 (27) = 224.9, CFI = 0.53, TLI = 0.38, RMSEA = 0.19), indicating that a single common factor did not adequately capture the covariance structure. A correlated three-factor model corresponding to neuromuscular, metabolic and inflammatory domains showed improved fit (χ2 (24) = 103.2, CFI = 0.88, TLI = 0.81, RMSEA = 0.11), with moderate to strong domain-specific loadings (Neuromuscular: 0.73–0.88; Metabolism: 0.40–0.90; Inflammation: 0.33–0.79; all p < 0.001). A second-order model specifying a higher-order vitality factor demonstrated similar fit (χ2 (24) = 104.2, CFI = 0.87, TLI = 0.81, RMSEA = 0.11). However, estimation did not achieve convergence, indicating limited empirical support for a stable higher-order vitality factor in this model. Model fit indices are presented in Table 3. Candidate confirmatory factor analysis (CFA) models for biomarker-derived vitality. Note: A: unidimensional CFA model; B: correlated first-order CFA model; C: second-order CFA model; D: bifactor CFA model. Table 3Fit indices for candidate CFA models of biomarker-defined vitality.Table 3Modelχ2 (df)CFITLIRMSEASRMRAICBICOne-factor (unidimensional)356.53 (27)0.480.310.200.156984.047084.04Three-factor (orthogonal)127.54 (27)0.840.790.110.116755.056855.05Three-factor (correlated)103.16 (24)0.880.810.110.096736.676847.78Second-order104.21 (24)0.870.810.110.096737.726848.83Bifactor35.87 (18)0.970.940.060.066681.386814.72Abbreviations: χ2 (df) = chi-square test statistic (degrees of freedom); CFI = Comparative Fit Index; TLI = Tucker–Lewis Index; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square Residual; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion. Fit indices for candidate CFA models of biomarker-defined vitality. Abbreviations: χ2 (df) = chi-square test statistic (degrees of freedom); CFI = Comparative Fit Index; TLI = Tucker–Lewis Index; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square Residual; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion. The bifactor model demonstrated the best global fit (χ2 (18) = 35.9, CFI = 0.97, TLI = 0.94, RMSEA = 0.06). However, reliability indices indicated limited dominance of the general factor (ECV = 0.28; ωH = 0.26), with domain-specific factors accounting for the majority of common variance (Supplementary Material, Table S8). Standardised loadings on the general factor were variable and generally modest in magnitude, whereas loadings on domain-specific factors were larger overall across neuromuscular, metabolic and inflammatory domains. As the bifactor structure did not support a substantively interpretable general vitality factor, the correlated three-factor model was retained as the primary representation of vitality. Factor loadings for all CFA models are presented in Supplementary Material, Tables S4 to S7. Associations of vitality with intrinsic capacity and frailty Using subdomain factor scores, higher neuromuscular vitality was associated with higher composite IC and cognitive capacity in adjusted models. Each standard deviation increase in neuromuscular vitality was associated with higher composite IC score (β = 1.39, 95% CI 0.25–2.54) and higher cognitive capacity scores (β = 1.74, 95% CI 0.52–2.95). Associations with locomotor capacity and frailty observed in unadjusted analyses were attenuated after adjustment and were no longer statistically significant. In adjusted models, higher inflammatory vitality (reflecting a more favourable inflammatory profile) was associated with higher locomotor capacity (β = 1.33, 95% CI 0.09–2.57) and lower frailty scores (β = −0.11, 95% CI − 0.22 to −0.002). Higher metabolic vitality was also associated with higher locomotor capacity (β = 1.74, 95% CI 0.58–2.91), but was not associated with frailty or other IC domains in adjusted models. In contrast to the domain-specific associations observed for vitality measures, MNA-SF scores were not associated with composite IC or individual IC domains but were associated with lower frailty scores (β = −0.085, 95% CI −0.17 to −0.004), but the associations were modest. Full results are presented in Fig. 2 and Supplementary Tables S9. Associations of vitality domains and MNA-SF with intrinsic capacity and frailty. Note: Results represent adjusted β coefficients with 95% confidence intervals; models adjusted for age, sex and vascular risk factors. Abbreviations: MNA-SF, Mini-Nutritional Assessment-Short Form; IC, Intrinsic Capacity. Sensitivity analyses using Poisson regression showed a similar pattern of associations to the primary linear models, with MNA-SF remaining significantly associated with frailty, while the vitality domains were not (Supplementary Table S10). Findings were also unchanged after exclusion of the BMI component from the MNA-SF (Supplementary Table S11). Discussion In this community-dwelling cohort, a biomarker-informed physiological construct of vitality showed a pattern of associations with IC and frailty that differed from those observed for nutritional status assessed using the MNA-SF. While MNA-SF was associated with frailty, it was not associated with composite IC or individual IC domains. In contrast, vitality subdomains demonstrated associations with IC, with neuromuscular vitality relating to composite IC and cognitive capacity, and metabolic and inflammatory vitality associated with locomotor capacity, although these associations were not consistent across domains and were attenuated after adjustment for some outcomes. Taken together, these findings suggest that biomarker-derived vitality measures may capture dimensions of physiological capacity that are not reflected by nutrition-based screening tools in relatively healthy, functionally independent older adults. The MNA was developed as a multidimensional screening tool integrating appetite, weight loss, dietary intake, mobility, acute illness and psychological stress, and has been extensively validated for predicting adverse outcomes including frailty, functional decline, hospitalisation and mortality in older adults [43]. As such, it is designed to identify established nutritional risk that co-occur with functional compromise, rather than to characterise the upstream biological conditions through which ageing influences capacity. In this relatively healthy cohort, MNA-SF scores were high with limited variability, reflecting the low prevalence of overt undernutrition. This ceiling effect is consistent with the context of its intended role as a screening tool for clinically expressed nutritional risk and may limit its ability to detect associations in healthy, functionally independent populations. In contrast, biomarker-derived vitality measures capture underlying physiological systems linked to core ageing processes, including inflammatory, metabolic and neuromuscular regulation, that shape how IC is maintained with ageing. For instance, inflammatory markers such as IL-6 and TNF signaling capture chronic low grade inflammation that has been linked to declines in physical performance and gait speed over time, plausibly connecting inflammatory vitality to locomotor capacity and to later frailty risk [26]. In addition, GDF-15 reflects cellular stress and mitochondrial function across tissues, and has been associated with reduced physical function, frailty and mortality in older adults, supporting its role as a complementary indicator of systemic stress within the inflammatory vitality domain [44]. Metabolic markers indexing insulin resistance and adiposity-related regulation capture energy and anabolic-catabolic balance, which are closely tied to muscle function and mobility, and are also associated with cognitive performance in older adults, consistent with metabolic vitality relating to both locomotor and cognitive domains of IC [45]. Neuromuscular measures such as grip strength, muscle mass and ventilatory function reflect integrated system-level physiological performance, summarising the net effect of musculoskeletal integrity, neuromuscular control and cardiopulmonary reserve, and have been strongly associated with subsequent disability and mortality [46]. Within this framing, vitality subdomains are not interchangeable with the MNA, instead, they represent biologically grounded axes of physiological states that could shape how IC is expressed, while frailty remains a downstream clinical phenotype that becomes more apparent as reserve is eroded [47]. This interpretation is supported by prior research in community-dwelling older adults that directly compared MNA, handgrip strength and plasma biomarker-based vitality. In this analysis, only a biomarker-derived vitality index predicted longitudinal locomotor decline, suggesting that biomarker-informed vitality may capture earlier, subclinical physiological dysregulation not reflected by nutrition- or strength-based measures [48]. Within this conceptual frame, our CFA results indicate that vitality is better represented as a set of related physiological domains than as a single general factor. This correlated structure supports the notion of vitality capacity as arising from interactions among neuromuscular, metabolic and immune or stress-response systems, rather than from a unidimensional construct [49]. Circulating biomarkers reflect integrated, system-level physiological states shaped by multiple biological processes. In line with our results, prior research on multisystem physiological dysregulation shows that domain-specific components typically account for a substantial, and often greater, share of biomarker variability than any single general factor [50]. Together, these findings suggest that shared biological influences are expressed through interacting, but partially distinct domain-level processes. This is in keeping with evidence that biological ageing can diverge across physiological systems within individuals, in which different systems follow partly independent trajectories with distinct functional consequences [51]. Therefore, while a global vitality score may be useful for descriptive or comparative purposes, domain-level representations may better preserve biologically relevant heterogeneity across individuals and clinical contexts, offering greater interpretive precision. A further consideration is that neuromuscular domain measures such as handgrip strength and spirometry differ in biological level from the circulating metabolic and inflammatory biomarkers included in the vitality construct. While blood-based markers primarily index underlying physiological and molecular processes, grip strength and pulmonary function represent phenotypic measures of system performance, reflecting the integrated functioning of multiple biological subsystems. Nonetheless, this combination of physiological and performance-based indicators is still consistent with the proposed WHO vitality framework, which spans biological processes and their functional expression rather than restricting measurement to a single level of organisation. However, this layering implies that vitality integrates signals operating at different points along the biology-to-function continuum, which may explain our observed domain-specific multidimensional structure rather than a single dominant latent dimension. Overall, these findings have implications for how vitality measures may be used alongside existing tools in clinical practice and research, particularly among relatively healthy older adults. Biomarker-informed vitality measures offer a means of profiling domain-specific physiological states that vary across individuals even in the absence of overt clinical impairment. In this context, vitality measures may help explain inter-individual differences in IC by identifying the underlying system-level biological variation that contributes to differences in functional reserve across IC domains. At the same time, the multidimensional structure of vitality highlights challenges for routine clinical implementation, as it does not readily reduce to a single actionable metric. Rather than serving as a standalone clinical tool, biomarker-based vitality measures could be best viewed as instruments for characterising heterogeneity in ageing trajectories [52], while underscoring the need for further work to translate such approaches into clinically practical applications. This study has several strengths, including the use of a multidimensional vitality construct informed by established expert consensus and is examined in a relatively healthy, functionally independent older population, where early heterogeneity in ageing processes is often difficult to detect. However, several limitations should also be acknowledged. First, vitality was assessed using static measures at a single time point, limiting inference about dynamic processes such as physiological adaptation or recovery from stressors; accordingly, the construct should not be interpreted as a direct measure of physiological resilience [53]. Second, the biomarkers and functional measures included may be influenced by comorbid conditions and underlying disease processes. Although analyses adjusted for comorbidity and cardiovascular risk factors to mitigate these effects, residual confounding cannot be excluded. Third, while the selected biomarkers capture key physiological domains relevant to vitality, they do not exhaustively represent all biological processes involved in ageing. Fourth, our findings should be interpreted in light of the use of MNA-SF as the nutritional measure. Although MNA-SF is a validated, feasible and widely used screening tool for malnutrition risk in older adults, its design reflects its role as a clinically oriented screening instrument, distinct from the full MNA or more comprehensive nutritional assessments. In this relatively well-functioning cohort, the ceiling effect in MNA-SF scores may also have limited its ability to discriminate subtle differences in nutritional risk. Fifth, frailty was assessed using a modified frailty phenotype rather than the original Fried phenotype. Handgrip strength was excluded because it was already included in the biomarker-derived vitality construct; retaining it in both the predictor and outcome would have introduced construct overlap and potentially inflated associations. Hence, frailty-related findings should be interpreted as associations with a modified phenotype-based measure, with its limitations in comparability to studies using the original criteria. Finally, the study population comprised relatively healthy older adults, and the extent to which these findings generalise to populations with greater multimorbidity or advanced functional impairment warrants further investigation. Conclusions In summary, this study supports vitality as a multidimensional, biomarker-derived construct reflecting interacting neuromuscular, metabolic and immune or stress-response domains, rather than a single underlying biological factor. Variation across these domains was associated with the expression of IC, particularly in relatively healthy older adults where overt functional deficits are uncommon. In this setting, established tools such as the MNA remain essential for the current clinical operationalisation of the vitality care pathway, particularly for identifying clinically apparent nutritional risk and guiding intervention, but may be less sensitive to earlier, subclinical physiological differences in more robust older persons, although this should be interpreted in the context of the present study using the MNA-SF. Biomarker-based vitality measures may provide a complementary lens by characterising upstream physiological states that shape functional expression and heterogeneity in ageing within the IC framework, and may help inform future refinement of vitality assessment beyond current approaches. Longitudinal studies are needed to establish their prognostic value for early risk stratification and monitoring age-related decline. CRediT authorship contribution statement Conceptualisation and methodology: JC, WSL, SPC; data curation: AY, CNT; formal analysis: JC, SPC; supervision: WSL, CLL; writing – original draft: JC; writing – review & editing: JC, WSL, CLL, JPL, JQC. All authors read and approved the final manuscript. Ethics statement The ethics committee of the National Healthcare Group Domain Specific Review Board (DSRB) approved the GeriLABS study (NHG DSRB Reference: 2017/00850). All participants gave their written informed consent. All authors of this manuscript comply with the guidelines of ethical authorship and publishing in the Journal of Nutrition, Health and Aging. Declaration of Generative AI and AI-assisted technologies in the writing process No AI tools were used in preparing this manuscript. Funding This research was supported by the Lee Foundation Grant 2019. Data statement The data that support the findings of this study are available from the corresponding author upon reasonable request, subject to institutional approval and data sharing agreements. Declaration of competing interest The authors have no conflict of interest to declare. Acknowledgments The authors thank the study participants who contributed to this study. Footnotes Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jnha.2026.100894. Supplementary data The following is Supplementary data to this article: References Associated Data Supplementary Materials