Hormone-defined metabolic phenotypes in institutionalized older adults: a cluster analysis

INTRODUCTION

Ageing is associated with a progressive decline in the capacity to maintain metabolic and functional integrity, placing older adults at higher risk of disorders such as insulin resistance, type 2 diabetes, dyslipidaemia, and cardiovascular disease [1]. These changes arise from a complex interplay of intrinsic cellular aging, chronic low-grade inflammation, hormonal alterations, and shifts in body composition. Hormonal regulation is central to metabolic control, integrating peripheral signals with central mechanisms that coordinate energy intake, nutrient utilization, and storage. Key roles in this process are played by hormones such as leptin, ghrelin, and insulin.
Leptin, secreted primarily by adipose tissue, regulates appetite and energy expenditure through hypothalamic pathways [2], while ghrelin and insulin interact with leptin to fine-tune glucose and lipid metabolism and overall energy balance, linking peripheral nutrient status with central regulatory networks [3]. Dysregulation of these hormonal networks, arising from genetic, environmental, and lifestyle factors, may produce heterogeneous metabolic phenotypes that are not always detectable through conventional anthropometric measures such as body mass index (BMI). Identification of such phenotypes is critical for understanding inter-individual variability in metabolic health and for detecting subtle dysfunctions that may predispose older adults to metabolic complications.
The present study focuses on a psychogeriatric popu­lation of institutionalized older adults. Previous assessments in this cohort have reported low physical activity levels and a predominance of overweight and obesity, accompanied by caloric surplus and insufficient protein intake. Notably, in the same cohort, only approximately 12% of participants presented with a HOMA-IR index ≥ 2.5, a commonly used index of insulin resistance, despite the high prevalence of overweight and obesity [4]. This apparent dissociation between adiposity and insulin resistance suggests substantial metabolic heterogeneity and emphasizes that standard anthropometric or nutritional indicators may not fully capture underlying metabolic function or risk in this population.
To our knowledge, no studies have systematically characterized metabolic phenotypes in older adults residing in institutionalized settings. To address this gap, we applied a data-driven, cluster-based phenotyping approach, integrating hormonal, metabolic, nutritional, and anthropometric markers into coherent physiological profiles. This method allows the identification of patterns that may not be evident when examining indivi­dual parameters in isolation. Accordingly, the aim of this study was to identify distinct metabolic phenotypes in institutionalized psychogeriatric older adults using k-means cluster analysis and to evaluate their associations with circulating hormones, metabolic markers, and anthropometric characteristics [5].

MATERIAL AND METHODS

Study population and assessment

The study included 95 residents of the Nursing and Treatment Facility in Cibórz, an institution integrated within a psychiatric hospital providing regular medical and psychiatric care. Due to inclusion criteria, participants were ≥ 55 years old, had resided in the facility for at least three months, and were free from acute infections, malignancy, and severe psychiatric disorders. Before enrolment, all participants underwent a standard medical evaluation that included a clinical examination performed by an internist, anthropometric measurements, blood pressure assessment, and routine laboratory testing. Written informed consent was obtained from all individuals, and the study protocol was approved by the Bioethics Committee of the Medical College at the University of Zielona Góra, Poland (approval no. 16/2024), in accordance with the Declaration of Helsinki (annex Ref. KB-889/18).
Anthropometric measurements, including body weight, height, waist, and hip circumferences, were performed to calculate BMI and waist-to-hip ratio, following previously established protocols [4].
Blood samples were collected in the morning between 08:00 and 08:30 following an overnight fast of at least 8 hours. Serum was separated and stored at −70°C until analysis. Serum concentrations of insulin, leptin, and ghrelin were measured using human-specific ELISA kits (Qayee Biotechnology Co., Shanghai, China). All samples and standards were analyzed in duplicate according to the manufacturers’ protocols, and calibration curves were prepared using serial dilutions of the provided standards. Intra-assay coefficients of variation were below 15% for all assays.
Serum high-density lipoprotein (HDL) cholesterol, triglycerides (TG), and glucose were measured using colorimetric assay kits (Pointe Scientific, Canton, MI, USA) according to the manufacturer’s instructions. Non- esterified fatty acids (NEFA) were quantified using the Fujifilm Wako NEFA-HR(2) colorimetric assay kit validated for human samples. Serum albumin concentrations were measured using standard automated laboratory assays. Optical densities were recorded for all colorimetric assays using a Synergy H1 microplate reader (BioTek, Winooski, VT, USA).
All procedures were carried out under standardized laboratory conditions to minimize variability and ensure reliability.

Statistical analysis

All statistical analyses were performed using STATISTICA 13 (Tibco Software Inc., Palo Alto, CA, USA) for Windows (Microsoft Corporation, Redmond, WA, USA). The distribution of continuous variables was assessed using the Shapiro-Wilk test to evaluate normality and skewness. Variables exhibiting significant right-skewness, including leptin, ghrelin, insulin, non- esterified fatty acids (NEFA), and HDL cholesterol, were logarithmically transformed to improve normality. TG were not log-transformed, as their distribution did not show significant skewness and met the assumptions of parametric analysis. Glucose, albumin, and BMI were also not subjected to logarithmic transformation. All variables included in the clustering procedure were subsequently standardized using z-scores to ensure equal contribution to the analysis.
k-means++ cluster analysis was then applied to identify homogeneous groups of participants based on the standardized variables. Following cluster formation, one-way analysis of variance (ANOVA) with Welch correction for unequal variances was used to assess differences between clusters. A two-sided p-value < 0.05 was considered statistically significant.

RESULTS

Characteristics of the study population

Baseline demographic, anthropometric, and clinical characteristics of the cohort are summarized in Table 1. Overall, the group was characterized by overweight, with a mean BMI within the range classified as overweight [5]. Mean arterial pressure values were within the normal range, indicating generally adequate blood pressure control at the group level [6].
The mean waist circumference exceeded commonly accepted thresholds for increased cardiometabolic risk, and the mean waist-to-hip ratio approached or exceeded values typically associated with visceral fat accumulation [7].

Raw biochemical, metabolic, and anthropometric parameters

Descriptive statistics for raw (non-transformed) biochemical, metabolic, and anthropometric variables included in the clustering procedure are presented in Table 2. These data illustrate the absolute levels and variability of hormones, metabolic markers, and anthropometric measures in the entire cohort prior to transformation and standardization. Raw values were not used directly in the k-means clustering but are provided to facilitate clinical interpretation.

Identification of metabolic phenotypes

To identify metabolic phenotypes, k-means cluster analysis was performed using standardized variables, with log transformation applied where appropriate to reduce skewness. Hormonal, metabolic, nutritional, and anthropometric parameters were included in the clustering procedure.
Two metabolic phenotypes were identified, comprising 69 (73%) and 26 (27%) participants, respectively. Cluster centers (means ± SD of standardized values) and between-cluster differences assessed by one-way ANOVA with Welch correction for unequal variances are summarized in Table 3.
Participants classified as phenotype 2 exhibited significantly higher standardized levels of leptin, ghrelin, and insulin compared with phenotype 1 (all p < 0.001). HDL cholesterol was also modestly but significantly higher in phenotype 2 (p = 0.01). In contrast, no statistically significant differences were observed between clusters for NEFA, TG, glucose, albumin, or BMI.

Visualization of cluster separation

To aid interpretation of the clustering results, a two-dimensional visualization based on leptin and ghrelin was constructed. These hormones were selected for visualization because they showed the strongest between-cluster differences and are central regulators of appetite and energy balance.
Figure 1 illustrates the separation of the two metabolic phenotypes using standardized log-transformed leptin and ghrelin values. The clusters show clear separation along both hormonal axes, reflecting distinct hormonal profiles underlying the identified phenotypes.

Supplementary material

Additional exploratory analyses are presented in Supplementary Table 1. These include an alternative clustering solution assuming three clusters (k = 3; unstable solution). This supplementary table provides additional context and support for the main findings but was excluded from the primary analysis due to considerations of stability and interpretability.

DISCUSSION

To our knowledge, no previous studies have systematically characterized metabolic phenotypes in institutionalized psychogeriatric populations with a high prevalence of overweight and obesity using an integrative, hormone-centered clustering approach. Although age-related obesity is typically associated with metabolic deterioration, institutionalized older adults often present a paradoxical profile characterized by excess adiposity, low physical activity, and relatively preserved classical metabolic markers. This heterogeneity suggests that traditional anthropometric and biochemical indices may be insufficient to capture early metabolic vulnerability in this population.

Rationale for measured parameters

Leptin, ghrelin, and insulin were selected as central regulators of energy homeostasis and appetite control. These hormones provide complementary insight into endocrine regulation of energy balance [8] that may precede alterations in conventional metabolic markers. TG and HDL cholesterol were included as established indicators of lipid metabolism and cardiometabolic risk, while NEFA were assessed as markers of adipose tissue lipolysis and metabolic flexibility [9, 10]. Serum albumin was incorporated to provide contextual information on metabolic stability, recognizing that it is not an appropriate standalone marker of nutritional status but can aid interpretation of metabolic homeostasis in older adults [11].

Characterization of metabolic phenotypes

Using a data-driven, cluster-based approach integrating hormonal, metabolic, anthropometric, and nutritional markers, we identified two distinct metabolic phenotypes within this relatively homogeneous institutionalized cohort.
Cluster 1 comprised the majority of participants and displayed relatively stable metabolic characteristics despite elevated BMI and central adiposity. This phenotype aligns with the concept of metabolically healthy obesity (MHO), describing individuals with excess adiposity who lack overt metabolic disturbances despite unfavourable body fat distribution [12].
Cluster 2 exhibited a distinct endocrine profile, with higher standardized levels of leptin, ghrelin, and insulin compared with Cluster 1. Notably, these differences were not accompanied by marked alterations in classical metabolic parameters such as glucose, TG, NEFA, albumin, or BMI, highlighting that the primary distinction between clusters was hormonal rather than anthropometric or metabolic.
Higher leptin concentrations in Cluster 2 may reflect relative leptin resistance, commonly associated with increased adiposity and impaired central energy signalling. Under physiological conditions, leptin acts as a key satiety signal to the hypothalamus, suppressing appetite and promoting energy expenditure [13]. In the context of increased adiposity, reduced central responsiveness to leptin may impair appetite regulation and energy homeostasis, resulting in persistently higher circulating leptin concentrations despite limited effectiveness of this signal. Such a pattern has been linked to functional leptin resistance, potentially driven by overactivation of inhibitory feedback pathways within leptin signalling cascades [14, 15].
Higher ghrelin levels in Cluster 2 indicate altered endocrine regulation of energy balance. Although ghrelin normally stimulates appetite and promotes energy storage by enhancing lipogenesis [16], in individuals with greater adiposity its higher concentration likely reflects dysregulation of orexigenic–anorexigenic signalling rather than a compensatory adaptation to energy deficit. In this context, elevated ghrelin may exacerbate early endocrine instability by failing to be adequately counterbalanced by leptin, contributing to less stable regulation of energy balance [17].
Insulin concentrations were also higher in Cluster 2, while glucose levels remained largely similar between clusters. This pattern may reflect early alterations in insulin regulation, suggesting a tendency toward insulin resistance even in the absence of changes in circulating glucose [18]. The skewed distribution of insulin values, with several high outliers, should be considered when interpreting these results.
HDL cholesterol was slightly higher in Cluster 2 compared with Cluster 1, but absolute HDL levels across the cohort remained below recommended thresholds. Therefore, this difference should not be interpreted as metabolically protective. Rather, the relatively higher HDL observed in the less favorable endocrine phenotype may reflect a transient or compensatory response in lipid handling associated with positive energy balance, as previously reported during experimental hyperalimentation [19]. NEFA, TG, and albumin were similar between clusters, further supporting that the primary distinctions are hormonal rather than conventional metabolic or nutritional markers.
Collectively, our findings suggest that Cluster 2 represents a metabolically less stable phenotype, characterized by early endocrine alterations that occur prior to changes in classical metabolic markers. This underscores the potential value of hormone-centered phenotyping in predicting future metabolic risk among older adults.

Study limitations

While this study is observational and does not allow for causal inferences, the reported endocrine differences, particularly in leptin and ghrelin, provide valuable insights that warrant further investigation. The relatively small sample size and the focus on an institutionalized psychogeriatric cohort may limit generalizability, highlighting the need for prospective and mechanistic studies to confirm whether these endocrine phenotypes predict future metabolic changes or respond differently to targeted interventions.

CONCLUSIONS

Cluster-based phenotyping using hormonal, meta­bolic, and anthropometric parameters revealed subgroups of institutionalized older adults with distinct endocrine profiles, even when classical metabolic markers appeared relatively normal. These findings emphasize the potential utility of early hormonal assessment to identify at-risk individuals and inform precision interventions aimed at promoting healthy aging.

Disclosure

The authors report no conflict of interest.

References