Obesity Definitions and Hospital Stay in Older Adults with Heart Failure: A Post Hoc Analysis of a Prospective Cohort

Heart failure (HF) is a complex and progressive clinical syndrome characterised by high morbidity, mortality and frequent hospitalisations, posing a significant global public health burden.^1–3 As of 2019, HF affected approximately 64 million individuals worldwide, with age-standardised prevalence rates reaching 711.9 per 100,000 and a total of 5.05 million years lived with disability attributed to the condition.⁴ Despite advances in pharmacological and device-based therapies, the global burden of HF has increased substantially over the past decades, particularly in ageing populations and regions undergoing rapid epidemiological transitions.⁵ Notably, recent data from the Global Burden of Disease 2021 analysis indicate a rise in age-standardised HF prevalence from 641.1 in 1990 to 676.7 per 100,000 in 2021, with a potential inflection point observed after 2019.⁵ Socio-demographic disparities persist, with high-income regions showing a decline in HF burden, while South Asia and sub-Saharan Africa have experienced sharp increases.^4,5 The predominant aetiologies contributing to the HF burden globally include ischaemic heart disease, hypertensive heart disease and cardiomyopathies, accounting collectively for more than 60% of attributable cases.⁴ Given these trends, HF remains a critical challenge for healthcare systems worldwide, particularly in older adults, who often present with multimorbidity and functional decline that further exacerbate clinical outcomes.⁶

Obesity is a well-established risk factor for the development and progression of HF, exerting both direct and indirect effects on cardiac structure, function and systemic haemodynamics.⁷ Excess adiposity promotes maladaptive remodelling, increases ventricular stiffness and contributes to the development of HF with preserved ejection fraction (HFpEF), particularly in older adults and women.^7,8 In parallel, obesity is associated with a higher burden of HF symptoms, reduced quality of life and diminished exercise tolerance.^9,10 Among hospitalised patients with HF, obesity prevalence varies from 23% to 33%, depending on the HF phenotype, with HFpEF patients disproportionately affected. Despite this, the relationship between obesity and clinical outcomes in HF remains complex and often paradoxical. While observational data suggest that obesity may confer a survival benefit – commonly referred to as the ‘obesity paradox’ – this effect is largely attributed to confounding by cachexia and frailty.^9,10 Recent data challenge the validity of the obesity paradox, showing that its apparent protective effect diminishes when accounting for biomarkers such as natriuretic peptides or using measures of central adiposity.¹¹ Waist-to-height ratio, for instance, is positively associated with worse outcomes in HF patients, suggesting that BMI-based assessments may obscure true risk.¹² Importantly, recent data highlight that obesity is associated with longer hospital length of stay (LOS), particularly in the context of frailty and prior HF hospitalisations.¹³ Nevertheless, the magnitude of this effect remains poorly characterised due to inconsistencies in obesity definitions, variability in comorbidity adjustment and heterogeneous study populations.^9,13 Therefore, improved phenotyping of obesity in HF is essential to refine prognostic stratification and personalise treatment strategies.¹⁴

In recognition of the limitations of BMI as a sole diagnostic tool, the 2024 Lancet Diabetes & Endocrinology Commission proposed a novel framework distinguishing clinical obesity – a chronic, systemic disease driven by excess adiposity and associated organ dysfunction – from preclinical obesity, where risk is elevated but end-organ damage is not yet apparent.^15,16 This paradigm shift underscores the need for individualised assessment based on function, health impact and metabolic profile, rather than anthropometric thresholds alone.¹⁶

The aims of this study were to compare the predictive value of three distinct definitions of obesity – BMI-based, waist circumference-based and clinical obesity as defined by the 2024 Lancet Diabetes & Endocrinology Commission – for hospital LOS among older adults (≥60 years) with acute decompensated HF. The study also aimed to identify predictors of LOS within these three obesity definitions.

Methods

Participants

This analysis was conducted as a post hoc exploratory evaluation within a prospectively enrolled cohort originally designed to investigate cognitive function in patients with HF. This study enrolled 250 patients with a confirmed diagnosis of HF who were hospitalised for acute decompensation at the Institute of Heart Diseases, Department of Cardiology, University Clinical Hospital, Wrocław Medical University, Poland. Eligibility criteria included age ≥60 years, a documented diagnosis of HF according to the current guidelines of the European Society of Cardiology, a disease duration of at least 6 months, and New York Heart Association (NYHA) functional class II to IV at the time of admission. Patients were required to be cognitively intact, with a Mini-Mental State Examination (MMSE) score of ≥24.¹⁷ Exclusion criteria encompassed NYHA class I, an MMSE score <24, current or treated depressive disorder, active malignancy, active systemic infection (e.g. pneumonia) or lack of consent to participate.

Data Collection

Patients were recruited between September 2022 and June 2023. All participants were hospitalised due to acute decompensated HF and were enrolled once clinical stabilisation was achieved, prior to discharge. Anthropometric measurements (body weight, height and waist circumference) were performed at that time. Data collection was performed by trained healthcare professionals using a standardised protocol. Sociodemographic and clinical data were obtained through structured interviews and medical chart review during hospitalisation. Laboratory and diagnostic parameters were extracted directly from electronic health records to ensure the completeness and reliability of the dataset.

Collected data included demographic variables such as age, sex, marital status, years of education, place of residence (urban or rural) and occupational status. Clinical parameters comprised NYHA functional class, duration of hospitalisation and number of hospitalisations in the previous year. Laboratory biomarkers included haemoglobin concentration (g/dl), high-sensitivity C-reactive protein (mg/l), serum albumin (g/dl), N-terminal pro-brain natriuretic peptide (NT-proBNP, pg/ml) and estimated glomerular filtration rate (ml/min/1.73 m²). Comorbidities were assessed through medical records and self-report and included: type 2 diabetes, chronic obstructive pulmonary disease or asthma, chronic kidney disease, prior MI, previous stroke or cerebrovascular disease, connective tissue disease, malignancies, liver disease and peptic ulcer disease. Tobacco exposure was measured as pack-years. Nutritional status was evaluated using the Mini Nutritional Assessment tool.¹⁸ Frailty was assessed according to the Fried phenotype, with frailty defined as the presence of three or more of the five components: unintentional weight loss, exhaustion, weakness, slowness and low physical activity.¹⁹

The study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines to ensure transparency and methodological rigour.

Research Instruments

Obesity in this study was defined and classified according to three approaches. First, obesity was assessed based on the WHO criteria, with a BMI of ≥30 kg/m² considered as classical obesity.²⁰ Second, central obesity was identified using sex-specific waist circumference thresholds recommended by the Polish Society of Hypertension and collaborating societies: ≥102 cm for men and ≥88 cm for women.²¹ Third, and most importantly, we adopted the recently proposed definition of clinical obesity published by The Lancet Diabetes & Endocrinology Commission.¹⁶ According to this framework, clinical obesity is diagnosed when two components are present: an anthropometric criterion indicating excess adiposity and at least one obesity-related disease. This definition was applied in a post hoc analysis.

The anthropometric criterion was met if the patient had either: a BMI ≥30 kg/m² and waist circumference ≥102 cm (men) or ≥88 cm (women), or a BMI between 25–29.9 kg/m² and waist circumference ≥110 cm (men) or ≥95 cm (women).

The clinical criterion required the presence of at least one obesity-related disease, such as type 2 diabetes, hypertension, cardiovascular disease (e.g. HF with reduced/preserved ejection fraction [HFrEF/HFpEF]), non-alcoholic fatty liver disease or an obesity-related cancer. Given that all patients in our cohort were diagnosed with HF and had hypertension, the clinical component of the definition was inherently fulfilled in all individuals with excess adiposity. As a result, patients who met the anthropometric criteria outlined above were classified as having clinical obesity. The Lancet clinical obesity framework was applied as originally proposed; however, given the presence of HF in all participants, the clinical component could not be independently evaluated.

Ethical Considerations

The research protocol was reviewed and approved by the Bioethics Committee of Wrocław Medical University (approval no. KB-651/2022) and was carried out in full compliance with the principles outlined in the Declaration of Helsinki. Prior to enrolment, all participants provided written informed consent. To protect patient confidentiality, all personal data were anonymised during data handling and analysis.

Statistical Methods

Categorical variables were compared between groups using the χ² test (with Yates’ correction for 2 × 2 tables) or Fisher’s exact test when expected cell counts were low. Continuous variables were compared between two groups using the Mann–Whitney U-test due to non-normal distributions. To assess the association between potential predictors and the number of days of hospitalisation, we used univariable and multivariable negative binomial regression models with a log link function, appropriate for count data with overdispersion. Model estimates are reported as incidence rate ratios (IRRs) with 95% CI and are interpreted as the relative (%) difference in LOS. Variables with a statistically significant or near-significant association (p<0.2) in univariable models were included in multivariable analysis. A p-value <0.05 was considered statistically significant. All analyses were performed using R software (version 4.5.0).

Results

Baseline Characteristics Stratified by Obesity Definition

When comparing the three definitions of obesity, 92 patients (36.8%) were classified as having obesity according to BMI (≥30 kg/m²), 162 (64.8%) according to waist circumference criteria and 125 (50.0%) according to the 2024 Lancet Commission framework. Substantial overlap between definitions was observed: 91 patients met all three criteria, 92 fulfilled both BMI- and Lancet-based definitions, 91 fulfilled both BMI- and waist-based definitions, and 124 fulfilled both Lancet- and waist-based definitions. Notably, 33 patients (13.2%) were classified as having obesity according to the Lancet Commission definition, despite having BMI <30 kg/m², and 71 patients (28.4%) were classified as having obesity according to waist circumference criteria despite having BMI <30 kg/m².

The Lancet Commission Definition

It was significantly more common for patients without clinical obesity to be men than those with clinical obesity. Patients without clinical obesity had significantly longer hospital stays. NT-proBNP concentrations were also significantly higher in patients without clinical obesity. A history of MI was more frequently reported in patients without clinical obesity, whereas the prevalence of diabetes was significantly higher among those meeting the criteria for clinical obesity. Baseline characteristics by clinical obesity status are presented in Table 1.

Classical Obesity by BMI

Patients without classical obesity (BMI <30 kg/m²) had significantly longer hospital stays. NT-proBNP concentrations were also significantly higher in the patients without obesity. The proportions of patients with HF with mildly reduced ejection fraction and HFrEF were both higher among those without classical obesity. In contrast, diabetes was more prevalent among patients with obesity classified based solely on BMI. Full characteristics are shown in Table 2.

Central Obesity Defined by Waist Circumference

Central obesity was defined as a waist circumference ≥102 cm in men or ≥88 cm in women. Male sex was more common among patients without central obesity. Hospital stays and NT-proBNP concentrations were significantly higher in the patients without central obesity. A history of MI was also more frequently reported in patients without central obesity. Detailed comparisons are provided in Table 3.

Exploratory Analysis of Continuous Obesity Measures

Generalised additive models (GAM) were used to examine the associations between continuous anthropometric measures (BMI, waist circumference in women and waist circumference in men) and LOS. For BMI, a U-shaped relationship was observed, with LOS decreasing until approximately 32 kg/m², followed by a gradual increase. For waist circumference in women, a similar U-shaped pattern was found, with the shortest LOS at around 95 cm. In men, the association between waist circumference and LOS was approximately linear, with LOS decreasing as waist circumference increased and no clear inflection point identified. Detailed GAM plots with smoothing curves are presented in Supplementary Figures 1–3.

Association Between Obesity (Defined by Various Criteria) and Length of Stay

Univariable negative binomial regression models were used to assess the association between three obesity definitions and LOS. None of the obesity types – clinical obesity (the Lancet definition), classical obesity (BMI ≥30 kg/m²), or central obesity (based on waist circumference) – showed a statistically significant relationship with the number of hospital days (all p>0.05). Results are summarised in Supplementary Table 1. We performed three separate multivariable negative binomial regression models – one for each obesity definition – adjusted for age and sex. In all models, obesity remained non-significantly associated with LOS (all p>0.05), suggesting no independent effect of obesity – regardless of definition – after accounting for these covariates. Detailed results are presented in Supplementary Table 2.

Fully Adjusted Models

In the multivariable model among patients with clinical obesity, only NT-proBNP remained an independent predictor of LOS. Specifically, each 100 pg/ml increase in NT-proBNP was associated with a 0.4% increase in LOS (IRR 1.004; p=0.004).

For patients with classical obesity (BMI ≥30 kg/m²), three variables were independently associated with prolonged LOS in multivariable analysis: NYHA class IV (IRR 2.355; p=0.004), NT-proBNP per 100 pg/ml (IRR 1.004; p=0.004), and a smoking history of 41–50 pack-years (IRR 1.946; p=0.022).

In the central obesity group, two variables remained independently associated with LOS in the multivariable model: haemoglobin and NT-proBNP. Each 1 g/dl increase in haemoglobin was associated with a 7.7% decrease in LOS (IRR 0.923; p=0.029), while each 100 pg/ml increase in NT-proBNP corresponded to a 0.4% increase in LOS (IRR 1.004; p<0.001). Detailed multivariable regression outputs are presented in Table 4, with full model specifications available in Supplementary Tables 3–5.

Discussion

This prospective cohort study examined whether three obesity definitions – BMI, waist circumference and clinical obesity per the 2024 Lancet Commission – predict hospital LOS in older adults (≥60 years) hospitalised for acute HF. None showed independent prognostic value, underscoring the limitations of both anthropometric and functional classifications in this context. LOS is influenced by clinical, organisational and logistical factors and therefore represents an imperfect surrogate of clinical severity in HF.

Previous studies have reported inconsistent associations between obesity and clinical outcomes in patients with HF. Some have described an ‘obesity paradox,’ whereby higher BMI may confer a protective effect or, at minimum, not worsen prognosis in acute HF.^22–26 However, recent analyses – such as a post hoc evaluation of the PARADIGM-HF trial – have cast doubt on the obesity paradox, demonstrating that its effect is attenuated or eliminated after adjusting for key prognostic markers such as NT-proBNP and using alternative indices such as waist-to-height ratio.^11,27 Recent studies in elderly patients with HF further indicate that nutritional status substantially modifies the relationship between obesity and clinical outcomes, including mortality, highlighting the importance of nutritional assessment alongside anthropometric measures.^28,29 Others have underscored the adverse impact of obesity on cardiovascular outcomes and healthcare usage.^30,31 Our findings challenge both narratives by showing that none of the evaluated obesity definitions – including the clinically oriented model proposed by the Lancet Commission – were independently associated with LOS in older adults with HF. Even this comprehensive, phenotype-based framework failed to provide incremental prognostic value in our cohort.

However, distinct clinical risk patterns were observed across classifications. Among patients with clinical obesity, NT-proBNP emerged as the sole independent predictor of prolonged hospitalisation. As a robust biomarker of HF severity, NT-proBNP reinforces the primacy of clinical and biochemical indicators over anthropometric measures in prognostic assessment.^32,33 These findings suggest that even phenotype-driven obesity frameworks may be insufficient to capture short-term risk in acutely decompensated HF.

In patients classified as patients with obesity by BMI, advanced NYHA class (IV), elevated NT-proBNP, and heavy smoking history (41–50 pack-years) were independently associated with prolonged LOS. These findings underscore that clinical severity and cumulative lifestyle burden may outweigh static anthropometric definitions in determining hospitalisation risk. The prognostic relevance of obesity may therefore be context-dependent, with traditional metrics offering limited insight in the setting of acute HF.³⁴ In the central obesity group defined by waist circumference, lower haemoglobin levels and higher NT-proBNP concentrations were independently associated with prolonged LOS. Although waist circumference is a well-established marker of cardiometabolic risk, our findings suggest limited utility in acute HF, potentially due to fluid overload and dynamic metabolic alterations not captured by static anthropometric measures.^12,35

Taken together, these results highlight the limited prognostic value of both anthropometric and functional obesity classifications in older adults with acute HF. The interaction between obesity phenotypes, nutritional status and frailty is likely to be clinically relevant in HF. Geriatric domains such as malnutrition and frailty primarily influence longer-term outcomes, including readmission and mortality, rather than in-hospital LOS, which is strongly affected by organisational and logistical factors.^36,37 Across all definitions, clinical indicators of HF severity – particularly NT-proBNP and NYHA class – consistently outperformed obesity metrics in predicting LOS. An important interpretative consideration concerns the application of the Lancet clinical obesity framework in an HF-only cohort. Although not all patients were classified as having clinical obesity, HF was present in all participants and represents one of the obesity-related clinical manifestations explicitly listed in the Lancet framework. Consequently, the clinical component of the definition did not contribute to discrimination between patients and the classification of clinical obesity in this cohort was largely driven by anthropometric criteria. The present findings should therefore be interpreted as an evaluation of the performance of the Lancet definition in a population with established HF, rather than as a formal test of the full clinical obesity framework across heterogeneous disease states.

Prior studies have identified several clinical factors influencing LOS in HF. In a large Japanese cohort, advanced age, low systolic blood pressure, elevated brain natriuretic peptide and worsening HF were associated with prolonged hospitalisation.³⁸ Similarly, data from Ethiopia linked pleural effusion, hepatomegaly and third heart sound with increased LOS.³⁹ Reynolds et al. further demonstrated that LOS >5 days predicted higher short- and long-term mortality, reinforcing LOS as a proxy for disease complexity.⁴⁰ These findings align with our results, where markers of clinical severity, rather than obesity definitions, shaped LOS.

The classification of obesity as a disease remains a subject of on-going debate, carrying significant implications for clinical care, public health policy and weight-related stigma.^41–45

While our findings offer important insights, their generalisability may be constrained by the single-centre design and focus on older adults. Nonetheless, the absence of prognostic relevance across all obesity definitions – including the novel clinical obesity construct – raises critical questions about how obesity should be defined and operationalised in acute decompensated HF. In our cohort, the three obesity definitions were not fully interchangeable. While most patients meeting the BMI-based criterion also fulfilled the Lancet Commission and/or waist circumference–based definitions, 13% met the Lancet criteria and 28% met the waist-based criteria despite having BMI <30 kg/m². This underscores that non-BMI-based definitions identify additional individuals with obesity who might be overlooked by traditional BMI thresholds, potentially capturing clinically relevant phenotypes not detected otherwise.

Moreover, in our study, anthropometric measurements were obtained after clinical stabilisation and prior to hospital discharge, rather than at admission. This timing was chosen to minimise the confounding influence of acute volume overload on BMI and waist circumference, a limitation explicitly acknowledged in contemporary guidance. Because anthropometric assessments were performed after clinical stabilisation rather than at admission, the cohort represents a clinically stable subset of hospitalised HF patients. This approach reduced the risk of measurement bias related to volume overload and minimises the likelihood of falsely amplifying an ‘obesity paradox,’ which can occur when anthropometric indices are distorted by oedema. Both the 2025 European Society of Cardiology consensus on HF and obesity and the 2025 American College of Cardiology Scientific Statement on obesity management in HF recommend that anthropometric assessment be performed in a clinically stable state, as fluid retention in the acute setting may lead to substantial misclassification of obesity status.^10,14 While this methodological choice enhances internal validity for in-hospital analyses, it may yield anthropometric values that differ from those obtained during long-term outpatient follow-up, which should be considered when extrapolating these results.

Future studies in larger, multicentre cohorts should investigate whether dynamic body composition assessments – such as bioelectrical impedance analysis or dual-energy X-ray absorptiometry – can improve risk stratification in this population. Such efforts may support a broader reappraisal of obesity frameworks in acute HF, shifting the emphasis from static anthropometric thresholds toward dynamic pathophysiology and clinical severity.

Study Limitations

This study has several limitations. Although the cohort was prospectively enrolled and obesity classifications were based on predefined anthropometric criteria, the application of the clinical obesity construct from the 2024 Lancet Commission framework was performed post hoc and was not part of the original protocol. Moreover, because all participants had HF, one of the obesity-related clinical manifestations included in The Lancet framework was ubiquitous in the cohort, which limited the ability to independently evaluate the clinical component of the definition.

As the parent study was not designed to assess determinants of hospital LOS, several key factors influencing LOS such as guideline-directed medical therapy during hospitalisation, aetiological triggers of decompensation, intensive care unit transfer, in-hospital complications and nosocomial infections were not systematically collected. This may have resulted in residual confounding. Anthropometric measurements were obtained after clinical stabilisation rather than at admission, potentially limiting comparability with studies using admission-based values. Furthermore, because enrolment occurred only after stabilisation, patients with early in-hospital deterioration or death were not captured, introducing possible selection bias. Long-term outcomes, including readmissions and mortality, were unavailable. Some laboratory variables (e.g. HbA_1c) and detailed pharmacotherapy data were incomplete, limiting subgroup analyses. Finally, the single-centre design may reduce generalisability and social or logistical factors affecting LOS were not systematically assessed.

Conclusion

In this prospective cohort study of older adults (≥60 years) hospitalised for acute HF, none of the evaluated definitions of obesity – BMI-based, waist circumference-based or clinical obesity defined by the 2024 Lancet Diabetes & Endocrinology Commission – was independently associated with LOS in either univariable or multivariable models. These findings suggest that, in the context of acute hospitalisation for HF, the association between obesity definitions and hospital LOS appears to be limited. However, the set of independent predictors of LOS differed across obesity categories, highlighting potential differences in clinical risk profiles depending on the operational definition of obesity. Our results call for a re-evaluation of the role of obesity classifications as risk stratification tools in acutely decompensated HF. Further research in larger, more diverse populations is warranted to validate these findings and explore their generalisability.

Click here to view Supplementary Material.