This cross-sectional study included community-dwelling older adult women from São Paulo, Brazil, between July 2022 and September 2023. The protocol was approved by the Ethics Committee of the Hospital das Clínicas, School of Medicine, University of São Paulo (HC-FMUSP, CAAE: 31958520.4.0000.0068) and conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. All participants provided written informed consent.

Eligible individuals were women aged ≥65 years. Participants were recruited via geriatrics outpatient clinics (HC-FMUSP), prior study registries, and direct community referrals. Exclusion criteria included a diagnosis of dementia, physical disability, use of insulin, glucocorticoids, or medications affecting muscle function or bone mass; chronic obstructive pulmonary disease (COPD) or chronic kidney disease; uncontrolled chronic non-communicable diseases (including diabetes, hypertension, dyslipidemia, thyroid disorders, or liver disease); restrictive diets; or recent use (within the prior 12 months) of bisphosphonates, denosumab, teriparatide, or romosozumab. All those who did not meet the exclusion criteria were assessed for frailty.

Frailty was assessed according to the Fried frailty phenotype (2001), which evaluates unintentional weight loss (≥4.55 kg or >5% body weight in the preceding year), self-reported exhaustion, weakness (handgrip strength), slow walking speed (4.6 m), and low physical activity. Participants were categorized as robust (no criteria), pre-frail (1–2 criteria), or frail (≥3 criteria). Only pre-frail and frail community-dwelling older women were included.

A standardized face-to-face questionnaire collected demographic and clinical data, including age, comorbidities, medication use, smoking status, alcohol intake, supplementation, and physical activity. Anthropometric measurements (height, weight, and body mass index—BMI) were obtained according to standard procedures to confirm self-reported information. Obesity was defined as BMI ≥30 kg/m² (WHO, 1997).

Fasting blood samples (8-h fast) were collected between 7:30 and 9:30 a.m. Measurements included glycated hemoglobin, insulin, very low-density lipoprotein (VLDL), LDL-cholesterol, HDL-cholesterol, total cholesterol, triglycerides, ferritin, and erythrocyte sedimentation rate, analyzed using automated standardized methods in an accredited hospital laboratory.

Body composition was evaluated by whole-body dual-energy X-ray absorptiometry (DXA) using a Lunar iDXA scanner (GE Healthcare). Total fat mass, appendicular lean mass (ALM), and total lean mass were assessed. ALM was defined as the sum of lean soft tissue in both upper and lower limbs (Siervo et al., 2015) and used to classify low ALM (LALM) according to fat-adjusted criteria previously applied in the São Paulo Aging and Health (SPAH) Study (Lopes et al., 2011).

Low appendicular lean mass (LALM) was defined using São Paulo Ageing and Health Study (SPAH) criteria from prior Brazilian studies (Domiciano et al., 2013; Figueiredo et al., 2014). Expected ALM was calculated via the linear regression model proposed by Newman et al. (2003), which incorporates height (m²) and total fat mass (kg). The residual (measured ALM minus expected ALM) reflects deviation from predicted muscle mass: positive values denote higher-than-expected ALM, whereas negative values indicate lower ALM.

For women, the model equation is: ALM (kg) = −14.51 + 17.27 × height (m) + 0.20 × fat mass (kg). The 20th percentile of the residual distribution (−1.45) was used as the cutoff for LALM in Brazilian older women (Domiciano et al., 2013).

Lower-extremity function was assessed using the five-times chair stand test, in which participants are timed while rising from and sitting back down on a chair five consecutive times without using their arms (Pavasini et al., 2016).

Handgrip strength was measured with a Jamar® dynamometer (Lafayette Instrument Company, United States). Participants performed three trials on each hand, alternating sides with one-minute intervals. The highest value obtained from the dominant hand was used for analysis (Roberts et al., 2011).

Chair stand and grip strength were selected as direct, frailty-sensitive measures aligned with Fried criteria (Suzuki et al., 2023; Kawamura et al., 2025). Bohannon (2019) reiterates grip strength as a biomarker for identifying at-risk older adults, while Johansson et al. (2023) demonstrate their independent mortality associations, supporting combined rather than interchangeable use.

Cytokines (IL-6, TNF-α), adipokines (adiponectin, resistin), and myokines [growth differentiation factor-15 (GDF-15), brain-derived neurotrophic factor (BDNF), and myostatin] were quantified using the Luminex® xMAP multiplex platform with commercial Milliplex MAP kits (HAGE1MAG-20K, HADCYMAG-61K, HMYOMAG-56K; Millipore Corp., Billerica, MA, United States). All assays were run simultaneously by a blinded technician following manufacturer instructions.

Serum IGF-1 and IGFBP-3 were measured via chemiluminescence in an accredited clinical laboratory. Intra-assay coefficients of variation were: GDF-15 (1.51%), IL-6 (1.41%), TNF-α (1.17%), adiponectin (0.98%), resistin (2.84%), BDNF (0.69%), and myostatin (0.63%).

Sample size estimation was based on skeletal muscle index (SMI, kg/m²) differences between older adults with obesity versus sarcopenia (Fukuda et al., 2018), yielding a required sample of 44 participants to achieve 95% power at α = 0.05 using G*Power (version 3.1.2).

Normality assumptions were verified prior to analyses. Generalized Estimating Equations (GEE) with independent working correlation structure was adopted, providing robust standard errors appropriate for unbalanced cross-sectional data. Both unadjusted (P¹) and age-adjusted (P²) P-values are reported for the group main effect. Bonferroni post hoc testes, estimated marginal means (EMMEANS), provide pairwise unadjusted (Bonferroni post hoc P¹) and age-adjusted (Bonferroni post hoc P²) P-values for multiple group comparisons, using a two-tailed significance threshold of 0.05.

Categorical variables were evaluated with χ² or Fisher’s exact tests. Pearson correlation coefficients were computed separately for each dependent variable—chair stand time (seconds) and handgrip strength (kg)—against cytokines, adipokines, myokines, and health-related markers as independent variables. Multiple linear regressions were then performed separately for each dependent variable (chair stand time, handgrip strength and body composition parameters), adjusting for age. Full model specifications, including β coefficients, unstandardized B with 95% CIs, and P-values were also reported.

Data are presented as mean ± SD, median (interquartile range), or n (%). Statistical analyses were conducted using IBM SPSS Statistics version 26.0.

The markedly lower IGFBP-3 levels observed in the obesity plus LALM group align with findings by Gomez et al. (2004), who reported reduced IGFBP-3 in women with higher BMI and fat mass. IGFBP-3 plays a dual metabolic and anti-inflammatory role—preserving muscle homeostasis, modulating IGF-1 signalling, and exerting protective effects against obesity-related inflammation (Palau et al., 2012; Mohanraj et al., 2013). Individuals presenting both obesity and low muscle mass exhibit two overlapping metabolic stressors: elevated adiposity (Kim, 2013) and reduced muscle anabolic capacity (Ferrari et al., 2021). This combined phenotype may therefore deplete IGFBP-3 as part of a broader catabolic–inflammatory response, consistent with its association with multiple age-related diseases such as cognitive impairment, neurodegeneration, and osteoporosis (Yang et al., 2005; Wennberg et al., 2018).

Furthermore, experimental evidence suggests that IGFBP-3 may mitigate obesity-induced insulin resistance (Mohanraj et al., 2013), reinforcing the plausibility of the more insulin-resistant profile observed in the obesity and obesity plus LALM groups in our study.

Resistin levels were significantly higher in individuals with obesity, including in the sensitivity analysis restricted to frail participants, consistent with prior studies linking resistin to chronic inflammation, metabolic dysfunction, frailty, sarcopenia, and all-cause mortality in older adults (Khan et al., 2014; Parkkila et al., 2021; Kwak et al., 2024; Kim et al., 2025). Kim et al. (2025) recently identified resistin as a promising biomarker of frailty, reporting markedly elevated levels in frail individuals; notably, each 1-SD increase in resistin was associated with a 67% higher risk of frailty. Our findings extend this concept by showing that, even within a frailty-enriched sample, the metabolic signature of obesity is associated with higher resistin concentrations. This supports the notion that inflammatory adipokines may help discriminate frailty phenotypes with distinct metabolic risks, even when clinical frailty is already established.

GDF-15 has emerged as a robust biomarker of aging, systemic stress, mitochondrial dysfunction, and muscle catabolism. Elevated GDF-15 has been linked to frailty (Cardoso et al., 2018; Arauna et al., 2020; Suganuma et al., 2024), physical impairment (Tavenier et al., 2021), and pain-related disability (Tarabeih et al., 2019), and our findings confirm higher circulating GDF-15 in participants with obesity plus LALM. The literature supports a strong association between GDF-15 and unfavourable body composition phenotypes, including sarcopenia, obesity, osteosarcopenic obesity, and their combinations (Lee S. H. et al., 2022).

Mechanistically, GDF-15 promotes anorexigenic signalling, muscle wasting pathways, and reduced energetic capacity—potentially accelerating functional decline, especially in individuals with coexisting muscle and fat abnormalities. The observed positive correlation between GDF-15 and slower five-times chair stand performance strengthens the hypothesis that GDF-15 may serve as a mediator linking metabolic stress to impaired physical function in frailty. Similar associations were reported by Newman et al. (2016) in non-frail older adults, suggesting that this mechanism persists—or may even be amplified—in frail populations.

Both the obesity and obesity plus LALM groups required more time to complete the five-times chair stand test, indicating reduced lower-extremity functional capacity, although the difference reached statistical significance only in the comparison between the obesity and LALM groups. Poor physical performance has long been associated with elevated circulating cytokines in older adults, including slower chair stand time, lower SPPB scores (Kwak et al., 2024), and reduced gait speed (Wennberg et al., 2018). Consistent with these findings, we observed in this pre-frail and frail cohort significant positive associations between GDF-15 and chair stand duration (both Pearson r = 0.285, P = 0.006 and β = 0.309, P = 0.013), supporting the notion that GDF-15 may contribute to inflammation-related declines in muscle function, as previously described by Newman et al. (2016) in non-frail older adults.

From a clinical perspective, both obesity (15.4 s) and obesity plus LALM (13.3 s) groups exceeded the reference threshold of 12.6 s proposed by Bohannon (2006), reinforcing the functional impairment associated with these phenotypes. These values surpassed established cutoffs linked to increased fall risk (≥12 s) (Tiedemann et al., 2008) and, in the obesity group, approached the threshold associated with recurrent falls (>15 s) (Buatois et al., 2010). Taken together, these findings suggest that adiposity-related inflammatory stress is associated with greater lower-limb functional decline even within a population that already has frailty.

LALM adoption followed Domiciano et al. (2013), demonstrating that traditional ALM/height² cutoffs did not accurately discriminate sarcopenia in overweight/obese community-dwelling Brazilian older adult women. International consensus (EWGSOP2, AWGS, FNIH) endorses population-specific thresholds due to ethnic/anthropometric variability. Therefore, it was plausible to adopt regression models adjusted ALM for fat mass/height for enhanced LALM accuracy over rigid cutoffs.

Frailty is not a homogeneous inflammatory state, as inflammaging exhibits individualized expression modulated by body composition phenotypes. In our study, obesity plus LALM was associated with suppressed IGFBP-3, elevated GDF-15, and a trend towards slower chair-stand performance—potentially representing a high-risk phenotype among heterogeneous frailty presentations, warranting personalized phenotyping approaches.

This underscores the need to integrate body composition profiling—particularly DXA-derived parameters—with biomarker assessment when evaluating frailty risk, prognosis, and potential therapeutic targets. In clinical practice, these findings highlight that pre-frail and frail older women with concomitant obesity and low muscle mass may benefit from interventions targeting both muscle anabolism and adiposity-related inflammation.

Robust methodological controls minimized Type I errors (false positives) across multiple biomarker comparisons, including sample size exceeding requirements for 95% power (n = 44); GEE analyses (vs. GLM) accounting for data correlation; low intra-assay CVs (<3%, range 0.63%–2.84%) via Luminex® assays; and frail-only sensitivity analyses confirmed key resistin findings. All P-values are transparently reported with null findings labelled exploratory, enhancing interpretative caution. This approach acknowledges statistical uncertainties in exploratory biomarker research on inflammaging across obesity-frailty phenotypes in an understudied cohort.

Limitations include female-only Brazilian sample limits generalizability; cross-sectional design precludes causality; formal sarcopenia subanalysis (ALM < 15 kg or ALM/h² < 5.5 kg/m²) identified only 3 cases, precluding group comparisons; lack of visceral/regional fat distribution assessment. Longitudinal studies with comprehensive fat/muscle phenotyping are needed to establish causality.