Prevalence and Prognostic Significance of Malnutrition Assessed by Two Scoring Systems in Hospitalised Elderly Patients with Chronic Heart Failure

Malnutrition is common among patients with chronic heart failure (CHF).^1–3 As a possible contributor to systemic inflammation, autonomic dysfunction and cachexia, malnutrition is associated with longer hospital stays and worsened prognosis.^4–6

Elderly patients (≥65 years), who often have multiple comorbidities, are frail, experience cognitive impairment, or come from vulnerable social backgrounds, are at a greater risk of malnutrition and may require more detailed assessment and care.^7,8 Although several screening tools for malnutrition exist, the absence of standardised diagnostic criteria makes it challenging to compare its impact on morbidity and mortality across different study populations.^9–12 The Global Leadership Initiative on Malnutrition (GLIM) criteria are considered the diagnostic gold standard, but their application in acutely hospitalised elderly patients with CHF is limited.¹³ Specifically, GLIM requires historical data on weight loss and dietary intake, and objective assessment of muscle mass (e.g. via dual-energy X-ray absorptiometry [DXA], bioelectrical impedance analysis [BIA] or CT), which are often unavailable or infeasible in acute settings.¹³

In this context, nutritional scores such as the Prognostic Nutritional Index (PNI) and Geriatric Nutritional Risk Index (GNRI) offer practical alternatives.^14,15 Both rely on laboratory measures routinely collected at admission.¹⁶ In contrast, scores such as the controlling nutritional status (CONUT) and the triglycerides, total cholesterol and body weight index (TCBI) require lipid measurements, which are not routinely obtained during acute hospitalisation.¹⁵

Several studies have reported the utility of these nutritional indices for risk stratification and prognostic prediction in patients with acute and chronic heart failure (HF).^12,17 However, their association with established prognostic factors, particularly inflammatory marker levels, remains unclear. Moreover, it is still unclear which nutritional index offers superior prognostic utility in elderly patients with CHF. In this study, we also focused on high-sensitivity C-reactive protein (hsCRP) and soluble urokinase plasminogen activator receptor (suPAR), both of which are associated with adverse outcomes and reflect different aspects of systemic inflammation.^18–20

The aims of this study were to evaluate the relationship between PNI and GNRI and two inflammatory markers (hsCRP and suPAR), and to compare their predictive value for all-cause mortality and adverse hospital outcomes in a cohort of elderly patients (≥65 years), hospitalised with CHF, for whom the application of GLIM criteria was not feasible.

Methods

Database Introduction

This single-centre retrospective cohort study was conducted at the Department of Emergency Medicine, Copenhagen University Hospital, Amager and Hvidovre, Denmark. It used prospectively collected data from adult patients (≥18 years) who were admitted to the emergency department (ED) between 10 March 2020 and 31 March 2022 with clinical signs of infection (i.e. chest pain, dyspnoea, cough, fever or headache). During this period, all acutely admitted medical patients underwent routine COVID-19 testing and were included in the cohort. The follow-up period extended for at least 100 days following the index admission.

Clinical information was retrieved from the electronic health record (EHR) system covering EDs in the Capital Region, and subsequently entered into REDCap (Research Electronic Data Capture) once patient enrolment had concluded. The EHR system includes records on diagnoses, prescriptions and mortality occurring outside the hospital setting.

The database and data handling procedures were approved by the Danish Data Protection Agency (P-2020-513) and the Danish Patient Safety Authority (31-1521-319). The study was observational and did not interfere with standard diagnostic or therapeutic procedures.

Study Population and Definition

All patients’ records were screened, and patient data were extracted from the EHR system using each patient’s unique civil registration number (Central Person Register). The patient selection included all those aged ≥18 years. A total of 10,027 individuals were admitted to Copenhagen University Hospital, Amager and Hvidovre, during the study period. After applying predefined exclusion criteria (Supplementary Figure 1), 7,220 patients (72.0%) admitted to the ED were included in the cohort. Exclusions were made for patients with no primary care contact during the study period, hospital contacts limited to outpatient visits, absence of blood sampling within 24 hours of admission, absence of a diagnosis of CHF, missing anthropometric data (height and/or weight), or lacking measurements of serum albumin and/or total lymphocyte count. Identification of patients with cardiopulmonary diagnoses was based on International Classification of Diseases, Ninth Revision (ICD-9) codes retrieved from the registry.

In accordance with the European Society of Cardiology guidelines for the diagnosis and treatment of acute and chronic HF, patients with CHF were identified via manual chart review by experienced cardiologists.²¹ In ambiguous cases, classification was achieved through consensus among the study investigators. In this CHF population, a clinically distinct subgroup of patients hospitalised with acute decompensated heart failure (ADHF), or worsening CHF, was identified based on documentation of signs and symptoms consistent with acute decompensation. This subgroup was defined retrospectively and included patients with pulmonary congestion or peripheral oedema, oxygen requirement >1 l/min, and/or evidence of left ventricular enlargement or dysfunction documented by chest X-ray and/or echocardiography, as well as those who received IV furosemide within 24 hours of presentation. The final cohort consisted of 597 patients (6.0%) admitted to the ED with CHF (including ADHF) with available data to calculate both PNI and the GNRI (Supplementary Figure 1).

Malnutrition Screening Tools

Patients were assessed for malnutrition using the two indices: PNI and GNRI. The PNI was calculated as: serum albumin (g/l) + 5 × total lymphocyte count (10⁹/l).²² A PNI >38 indicated normal nutritional status, while scores of 35–38 and <35 represented moderate and severe malnutrition, respectively. The PNI does not define a ‘mild malnutrition’ category. The GNRI was calculated as: 1.489 × serum albumin (g/l) + 41.7 × (actual body weight/ideal body weight in kilograms).²³ Ideal body weight was defined as 22 × height² (in metres).²⁴ GNRI scores >98 indicated normal nutritional status, while scores of 92–98, 82–91 and <82 corresponded to mild, moderate, and severe malnutrition, respectively. Meeting either PNI or GNRI criteria, patients were stratified as malnourished (moderate or severe category) or well-nourished (Table 1).

BMI is calculated as weight (kg) divided by height² (in metres). According to the WHO BMI classification, patients were categorised into four groups: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²) or obese (≥30.0 kg/m²).²⁵ Notably, a BMI of exactly 30 kg/m² is classified in the obesity category. Additionally, to evaluate the prognostic impact of inflammatory markers, patients were stratified based on baseline levels of suPAR and hsCRP.

Data Extraction and Processing

Laboratory Measurements

Blood samples were obtained in the first 2 hours after admission at the ED and analysed at the Department of Clinical Biochemistry, Copenhagen University Hospital, Hvidovre. White blood cell counts, albumin, haemoglobin, creatinine, estimated glomerular filtration rate (eGFR), carbamide, hsCRP and alanine aminotransferase were measured using a COBAS 8,000 analyser (Roche Diagnostics). Cell counts (leukocytes, eosinophils, total lymphocyte count and neutrophils) were measured using flow cytometry on a Sysmex XN 9000 (Sysmex Corporation).

The suPAR was measured using the suPARnostic Quick Triage point-of-care test (ViroGates) according to the manufacturer’s instructions and quantified using an aLF reader (QIAGEN). Blood for this test (4ml) was drawn and collected in an EDTA (ethylenediaminetetraacetic acid) tube on arrival at the ED (in the first 2 hours after admission) and centrifuged for 3 minutes. Recent evidence supports using specific cut-offs for risk stratification: suPAR <4 ng/ml identifies low-risk patients, 4–6 ng/ml denotes medium risk, and >6 ng/ml indicates high risk.^20,26–28 In addition, evidence supports using specific cut-offs for hsCRP in cardiovascular risk stratification: hsCRP <1 mg/l identifies low-risk individuals, 1–3 mg/l denotes intermediate risk, and >3–10 mg/l indicates high risk. Levels >10 mg/l indicate potential acute inflammation and, if persistently elevated, a non-cardiovascular cause should be considered.^18,19,29

Clinical Data at Admission

At admission, all patients underwent a standardised clinical assessment including vital signs: systolic and diastolic blood pressure, heart rate, respiratory rate, peripheral oxygen saturation (via pulse oximetry), oxygen supplementation and body temperature. The National Early Warning Score (NEWS) was calculated to evaluate clinical acuity.³⁰ Standard 2D and Doppler transthoracic echocardiography was performed using Vivid E7 or E9 systems (GE Healthcare) to assess cardiac morphology and estimate left ventricular ejection fraction (LVEF). Demographic data, baseline comorbidities (e.g. prior stroke, chronic obstructive pulmonary disease, diabetes), and physical examination findings (including height and weight) were collected. BMI and nutritional indices were recorded at the time of admission.

Outcomes

The primary outcome was 1-year all-cause mortality. Secondary outcomes were in-hospital mortality, post-discharge mortality, hospital length of stay and all-cause rehospitalisations in the first year following admission. Survival time was defined as the period from the date of admission to the date of death.

Statistical Analysis

All analyses were calculated using the software R, version 4.3.2 (R Foundation for Statistical Computing). Based on the distribution and variance of data, continuous variables are expressed as mean ± SD or median (range with minimum and maximum values in the dataset) and were compared using Student’s t-test or Mann–Whitney U-test. Categorical variables are represented by composition ratio and were compared using the chi-squared test or Fisher’s exact test.

Kaplan–Meier curves are used to illustrate cumulative survival across subgroups, and differences were assessed using the log-rank test. Associations between nutritional status and mortality were examined using Cox proportional hazards models, with results presented as HR and corresponding 95% CI. To adjust for potential confounding, the final model included clinically relevant covariates selected a priori, including age, sex, BMI, vital signs and laboratory parameters obtained at admission. The laboratory parameters were chosen due to their significant influence on albumin and lymphocyte concentrations.³¹

To avoid confusion of direct-effect estimates with total-effect estimates (also known as ‘the table 2 fallacy’), these covariates were included for adjustment purposes only and not interpreted individually.

To test whether the prognostic association between nutritional indices and mortality varied by inflammation level, multiplicative interaction terms between nutritional scores (GNRI, PNI) and inflammatory biomarkers (hsCRP, suPAR) were included in Cox models. Interaction was evaluated by testing the statistical significance of the interaction term in the Cox models and, when relevant, by examining HRs in stratified subgroups. Interaction analysis between the GNRI/PNI and BMI groups was inconclusive due to data sparsity and model instability, and therefore these analyses were not included in the final results.

To assess the prognostic value of each nutritional index and their individual components for mortality, we estimated the area under the receiver operating characteristic (ROC) curve (AUC) for each variable. Comparison between AUCs was conducted using DeLong’s test to evaluate whether nutritional scores outperformed conventional biomarkers. Two-tailed p<0.05 was considered statistically significant. In addition, we constructed logistic regression models combining nutritional indices with inflammatory markers (e.g. GNRI + suPAR), and compared their AUCs with corresponding single-variable models using DeLong’s test to evaluate incremental prognostic value.

Results

Clinical Baseline Characteristics

The baseline characteristics of patients stratified by nutritional status using both the PNI and GNRI are listed in Table 1. Of the 597 patients included, 307 (51.4%) were classified as malnourished and 290 (48.6%) as well-nourished.

The median age was 76.0 years (range, 65.0–99.0 years), and 38.9% of the cohort were female. Malnourished patients had significantly lower BMI and serum albumin levels than the well-nourished group. These patients also had higher levels of inflammatory and catabolic markers including hsCRP, suPAR, neutrophils and carbamide, as well as lower haemoglobin, eosinophils and eGFR. At the time of admission, malnourished patients had significantly lower systolic and diastolic blood pressure and a higher prevalence of confirmed infections compared with the well-nourished group. No significant difference between the two groups was observed in the presence of ADHF, supplementary oxygen requirement, body temperature or LVEF. Regarding outcomes, malnourished patients had higher 1-year mortality (38.8% versus 21.7%; p<0.001), in-hospital mortality (9.8% versus 4.8%; p=0.031) and post-discharge mortality in the first year after admission (29.0% versus 18.6%; p=0.003), as well as longer hospital stays (5.0 days versus 3.0 days; p=0.004). The number of rehospitalisations in the first year after admission was also more frequent in the malnourished group (2.0 versus 1.0; p=0.025).

Prognosis

Figure 1A shows Kaplan–Meier survival curves stratified by PNI group (normal, moderate, severe). One-year survival probability declined progressively with increasing severity of malnutrition, and patients in the PNI-severe group had the lowest cumulative survival (log-rank p<0.0001). A similar pattern was observed in Figure 1B, which shows survival curves stratified by GNRI category; here, too, patients with moderate and especially severe GNRI-defined malnutrition had significantly poorer 1-year survival compared with the normal group (log-rank p<0.0001).

Cox regression analysis for 1-year mortality adjusted for relevant confounding factors is shown in Figure 1C and D. Severe PNI-based malnutrition was independently associated with increased risk of 1-year mortality (HR 1.43; 95% CI [1.02–1.99]; p=0.038), while moderate malnutrition did not reach statistical significance. For GNRI, both moderate (HR 1.58; 95% CI [0.99–2.53]; p=0.058) and severe malnutrition (HR 1.78; 95% CI [1.02–3.12]; p=0.043) were associated with increased risk of 1-year mortality, although only the severe GNRI category was statistically significant.

Figure 2A and Table 2 detail the ROC curves for predicting 1-year all-cause mortality, comparing nutritional indices, their individual components, selected biomarkers, and combined models. Of the single markers, suPAR had the highest AUC (0.693), followed by GNRI (0.663) and PNI (0.631). Lymphocytes (AUC 0.554) showed the poorest performance and were significantly outperformed by both PNI and GNRI (p<0.001). Adding hsCRP to PNI or GNRI yielded only marginal improvements (AUC 0.643 and 0.670, respectively; both p>0.28 versus base models). In contrast, adding suPAR improved GNRI performance (AUC 0.723, p=0.077) and showed a trend toward improved PNI (AUC 0.694, p=0.076), although neither reached statistical significance.

Figure 2B and Table 3 detail the corresponding ROC curves for post-discharge mortality. GNRI outperformed PNI (AUC 0.666 versus 0.586, p=0.002). Adding hsCRP to either index did not meaningfully enhance predictive accuracy. Adding suPAR to PNI (AUC 0.636, p=0.197) or GNRI (AUC 0.679, p=0.697) was associated with numerically higher AUCs but these differences were not statistically significant.

For the well-nourished patients, obesity (BMI ≥30 kg/m²) was associated with lower 1-year all-cause mortality compared with patients with normal BMI (Table 4). This association was not observed in malnourished patients, in whom obesity conferred no survival benefit. Supplementary Figure 2 shows the Kaplan–Meier curves for the total cohort stratified by BMI category: <18.5, 18.5–24.9, 25–29.9 and ≥30 kg/m². While the survival trajectories of the normal and overweight groups are largely overlapping, patients with BMI ≥30 kg/m² consistently show higher survival, and those with BMI <18.5 kg/m² the lowest.

Prevalence

Table 5 lists the prevalence of malnutrition and associated 1-year mortality, first stratified by BMI category, using the two nutritional risk indices: PNI and GNRI. For each BMI category, the first row under each index indicates the number and proportion of patients classified as moderately to severely malnourished. The second row shows 1-year mortality rates for the nutritional strata, allowing direct comparison of mortality risk across levels of malnutrition severity within each BMI category.

Across both PNI and GNRI, the prevalence of moderate–severe malnutrition declined progressively with increasing BMI: from 66.7% (PNI) and 97.9% (GNRI) in underweight patients to 34.0% and 3.1%, respectively, in obese individuals. Similarly, 1-year mortality for patients with moderate–severe malnutrition also decreased with increasing BMI. According to PNI, mortality declined from 39.6% (BMI <18.5 kg/m²) to 11.9% (BMI ≥30 kg/m²), and according to GNRI from 45.8% (BMI <18.5 kg/m²) to 1.3% (BMI ≥30 kg/m²).

The mortality impact of malnutrition was most pronounced in the BMI 18.5–24.9 kg/m² group, for whom GNRI-defined moderate–severe malnutrition was associated with a 1-year mortality of 24.2% versus 2.9% in the well-nourished group (p=0.008).

Furthermore, GNRI-defined malnutrition had consistent and statistically significant interactions with both suPAR and hsCRP levels. Severe GNRI-defined malnutrition is associated with increased 1-year mortality in patients with low inflammation (suPAR <4 ng/ml: HR 5.79, p=0.001; hsCRP <1 mg/l: HR 2.44, p=0.001), but this association is significantly attenuated in those with high inflammation (suPAR >6 ng/ml: HR 0.27, p=0.032; hsCRP >10 mg/l: HR 0.09, p=0.021). Similarly, severe PNI-defined malnutrition was associated with increased mortality at low inflammation levels (suPAR <4 ng/ml: HR 3.08, p=0.047; hsCRP <1 mg/l: HR 14.56, p=0.020), but this association was not observed in patients with higher inflammation (suPAR >6 ng/ml: HR 0.41, p=0.148; hsCRP >10 mg/l: HR 0.11, p=0.058).

Discussion

Main Findings

This study evaluates the association between nutritional status assessed using the PNI and the GNRI, and 1-year all-cause mortality in hospitalised elderly patients with CHF. First, we found that malnutrition was frequently observed in patients with CHF. The prevalence of malnutrition was 51–61%, with the higher estimate including patients classified as mildly malnourished according to the GNRI. Importantly, the prevalence of moderate–severe malnutrition was inversely related to BMI, ranging from 66.7% (PNI) and 97.9% (GNRI) in underweight patients to 34.0% and 3.1% in individuals with obesity. Second, according to the Kaplan–Meier curves, survival probability declined progressively with increasing malnutrition severity across both indices. Severe malnutrition by either PNI or GNRI was independently associated with higher 1-year all-cause mortality, consistent with prior findings on objective nutritional indices in this population.¹⁵ Both Cox models demonstrated the obesity paradox, with BMI ≥30 kg/m² associated with lower mortality (PNI HR 0.31; GNRI HR 0.45) compared with underweight. In subgroup analysis, obesity (BMI≥30 kg/m²), compared with normal BMI, was independently associated with lower 1-year all-cause mortality in well-nourished patients, supporting the obesity paradox. This survival advantage was not observed in malnourished patients. Third, ROC analysis further supported the prognostic value of these indices, with GNRI outperforming PNI for post-discharge mortality prediction, highlighting its utility in risk stratification beyond hospitalisation. And last, mortality rose progressively with increasing suPAR and hsCRP levels in malnourished patients, suggesting that systemic inflammation modifies risk in this population. Overall, GNRI appears more sensitive to inflammatory status than PNI in predicting outcome, suggesting that GNRI and inflammatory biomarkers may interact biologically in risk stratification. Although formal tests for interaction did not reach statistical significance for PNI (e.g. hsCRP interaction p=0.058), the findings are suggestive of effect modification by inflammation.

These findings address our aim and reinforce the prognostic relevance of malnutrition, assessed by PNI and GNRI, given that patients classified as malnourished had worse short- and long-term outcomes, including significantly higher 1-year and in-hospital mortality, more frequent rehospitalisations, and longer hospital stays. The prognostic value of PNI and GNRI is not uniform but varies according to BMI and systemic inflammation in elderly patients with CHF. To the best of our knowledge, this is the first study to examine how these nutritional indices relate to established prognostic factors (including biomarkers of inflammation), and to directly compare their ability to predict mortality and adverse hospital outcomes in this population.

In our previous study involving a separate cohort of patients with acute heart failure, we similarly observed that low PNI (malnutrition) was associated with increased mortality, further supporting the importance of nutritional status in refining risk stratification.¹⁷

Nutritional Particularities in Elderly Patients with Chronic Heart Failure

Elderly patients (≥65 years) characteristically present with multiple comorbidities, age-related physiological changes, atypical disease presentations (e.g. delirium as a first sign of acute coronary syndrome), and a high prevalence of geriatric syndromes such as frailty, falls, polypharmacy and cognitive impairment.²¹ Some of these characteristics, as observed in our study, underscore the importance of performing a comprehensive geriatric assessment. The risk of malnutrition increases with age, frailty and the burden of comorbidity, with a prevalence of approximately 40% in the community environment and exceeding 70% in cases of ADHF.^12,32,33 Malnutrition is associated with increased risk of mortality, disability and institutionalisation.³⁴ Therefore, routine malnutrition screening may be considered in elderly patients with CHF, irrespective of care setting, to guide the design of personalised care plans.³⁴

Biochemical Markers

Serum visceral proteins such as albumin (half-life 14–21 days) and prealbumin (half-life 2–3 days) have been used as indicators of nutritional status in patients with HF and hold prognostic value. However, their interpretation is limited by their sensitivity to systemic inflammation and congestion. Albumin, a negative acute-phase reactant, has reduced hepatic synthesis in inflammatory states and should therefore be interpreted alongside C-reactive protein (CRP). To account for this, our multivariate analysis included either GNRI and PNI (nutritional indices that incorporate albumin alongside other variables), thereby enabling us to capture the combined effects of nutritional status and inflammation. Although prealbumin is considered a more reliable marker of protein malnutrition because it is less influenced by hydration status compared with albumin, it is not routinely measured in hospitalised patients. Like albumin, it acts as a negative acute-phase reactant and should therefore be interpreted together with CRP. Owing to its shorter half-life, prealbumin is particularly useful for monitoring short-term changes in nutritional status.¹

Anthropometry and Body Composition

Anthropometric methods (including weight, height, BMI, skin folds, arm muscle circumference [AMC] and waist circumference) are widely used in HF nutritional assessment due to their low cost and accessibility.¹ In HF, subclinical volume overload and oedema can confound these measures, producing falsely elevated BMI and masking true weight loss, with rapid weight changes often driven by fluid shifts rather than nutritional alterations. Skin folds (typically measured at bicipital, tricipital, subscapular and suprailiac sites using the Durnin–Womersley equation) and AMC provide more accurate assessments of body fat and fat-free mass because they are less influenced by fluid status. Waist circumference, correlated with visceral fat, helps predict metabolic complications, cardiovascular disease and mortality, even in those with normal BMI.³⁵ Findings by Butt et al. question the obesity paradox in HF with reduced ejection fraction, showing that BMI overestimates the protective effect of obesity, while waist-to-height ratio better identifies the adverse impact of adiposity on outcomes such as HF hospitalisation.³⁵ These challenges are reflected in the nutritional indices used in this study: the GNRI (1.489 × serum albumin (g/l) + 41.7 × actual/ideal body weight) incorporates weight and may be biased by volume overload, whereas the PNI (serum albumin (g/l) + 5× total lymphocyte count (10⁹/l)) avoids this matter by excluding weight from its calculation. Nonetheless, albumin concentration can also be diluted by volume overload and/or congestion, which may compromise the accuracy of both indices in HF.³⁶

Diagnosis and Management of Malnutrition

Although no consensus exists on the best method to assess malnutrition in CHF, the PNI and GNRI are recommended as multidimensional tools for evaluating nutritional status in this population.^12,37 Therefore, any of the available tools, such as the Nutritional Risk Screening 2002 (NRS-2002), Subjective Global Assessment, Mini Nutritional Assessment (MNA), Malnutrition Universal Screening Tool or Short Nutritional Assessment Questionnaire may be used, provided that their application is appropriate to the specific clinical setting (e.g. hospitalised, outpatient, or elderly patients).^38–42

In CHF populations, the Mini Nutritional Assessment-Short Form (MNA-SF) has shown predictive value for muscle wasting and mortality, as demonstrated in 130 ambulatory patients from the SICA-HF programme.⁴³ Sarcopenia screening tools such as the SARC-F (a 5-item questionnaire) have high specificity but limited sensitivity, which may be improved by adding calf circumference (SARC-CalF) or using the Mini Sarcopenia Risk Assessment (MSRA), which offers better sensitivity.^44–46 For critically ill patients, the American Society for Parenteral and Enteral Nutrition (ASPEN) recommends the Nutrition Risk in the Critically Ill (NUTRIC) score alongside NRS-2002, unlike the European Society for Clinical Nutrition and Metabolism (ESPEN).^47,48 The GLIM criteria have also been proposed for diagnosing malnutrition and may help predict refeeding syndrome.^49,50 However, none of these tools has been specifically validated in patients with CHF for risk screening or diagnosis of malnutrition.

Other European studies, such as the one by Kwaśny et al., have used the NRS-2002 to assess malnutrition in hospitalised patients with HF and reported similar prognostic relevance as in our study, including sex-specific effects.⁵¹ This underscores the heterogeneity in screening practices across Europe and the world and highlights the need for comparative validation of different tools in diverse settings. Although the GLIM criteria are endorsed for diagnostic purposes, their application is often limited in acute care settings due to the need for prior weight trajectories, dietary intake data and objective body composition measurements: parameters that are frequently unavailable or unreliable in hospitalised elderly patients with CHF.¹³ Therefore, diagnosing malnutrition in CHF remains challenging due to limited evidence on the optimal diagnostic approach and the reduced reliability of anthropometric measures and traditional biochemical markers in the context of volume overload associated with CHF.

Nutritional alterations in CHF encompass a range of clinical conditions, including disease-related malnutrition (DRM) without inflammation, DRM with acute or chronic inflammation, cachexia, sarcopenia and sarcopenic obesity.^52,53 Accurate diagnosis of the specific type of malnutrition is essential for appropriate management, given that patients with inflammation-driven malnutrition will require more than nutritional support: namely, a multidisciplinary approach incorporating rehabilitation and optimal treatment of the underlying disease.

It is important to assess symptoms or medications that impair intake (such as anorexia, dysphagia or early satiety) to tailor nutrition and address deficiencies. In-hospital nutritional interventions for hospitalised patients with CHF reduce mortality and readmission, supporting routine screening for malnutrition in this population.⁵⁴

There are no CHF-specific nutritional formulas; choices should be guided by malnutrition type, clinical status and comorbidities. Some nutrients, such as omega-3 fatty acids and coenzyme Q10, may offer benefits, but stronger evidence is needed.^21,55–58 Cardiac rehabilitation is crucial, particularly for patients with sarcopenia, because it improves symptoms and reduces hospitalisations.^21,55

The Obesity Paradox in Heart Failure

A potential explanation for the obesity paradox in HF is that BMI does not differentiate between fat and lean mass. As highlighted in the 2025 American College of Cardiology Scientific Statement, higher BMI may reflect greater lean mass, which is associated with improved outcomes, whereas sarcopenia and low muscle mass, regardless of weight, are linked to poorer functional status and increased mortality.⁵⁹ Thus, the observed protective effect of obesity may partly reflect preserved muscle mass rather than excess adiposity per se.^59–61 This highlights the importance of how the weight loss occurs, given that the preservation of lean mass appears critical. These findings emphasise the need to assess and address nutritional status in CHF patients, given that unintentional or catabolic weight loss may contribute to worse outcomes despite higher body weight.

Limitations

In addition to the limitations mentioned in the discussion, there are several other important considerations. First, this was a single-centre, retrospective observational study, inherently susceptible to bias, unmeasured confounding, and limited generalisability beyond our healthcare setting. As an observational design, it enables identification of associations but does not establish causality. The limited sample size may have led to insufficient statistical power, affecting the robustness and generalisability of the results. Second, we did not account for important clinical variables such as frailty, immobility and concurrent medications, which may have influenced outcomes. We did not collect data on pharmacological treatments, such as glucagon-like peptide-1 receptor agonists or sodium–glucose co-transporter 2 inhibitors, which may have been more frequently prescribed to patients with diabetes and could have influenced mortality risk. Also, we did not collect data on physical performance measures (e.g. handgrip strength or gait speed) or functional status. These are known prognostic markers in elderly patients and may influence both nutritional status and mortality risk, representing potential unmeasured confounders. Given the use of BMI and weight-based indices without direct measures of body composition (e.g. muscle mass, sarcopenia assessment), our classification of nutritional status may be biased, especially in the context of fluid retention commonly seen in CHF. In addition, we did not capture data on dietary intake, nutritional interventions (e.g. enteral or parenteral nutrition), or referrals to dietitians during hospitalisation. These factors could have influenced both the trajectory of nutritional status and clinical outcomes. Moreover, the study did not assess longitudinal changes in nutritional status, which may have prognostic significance. Socioeconomic determinants of health, which are known to affect nutritional status, were also not collected.³⁷

Third, basic transthoracic echocardiography was often performed at the bedside during admission, primarily to estimate LVEF, limiting precision. Consequently, we lacked comprehensive data on ejection fraction and HF aetiology, warranting caution in interpreting related findings. Fourth, we did not collect data on active malignancy, which could have influenced both PNI and GNRI values and outcomes. Fifth, we relied on BMI values obtained during hospitalisation without follow-up measurements, which may reflect fluid retention and thus overestimate actual body weight in some patients. Finally, while the study focused on all-cause mortality, it did not explore cause-specific mortality or readmission outcomes, which are also clinically relevant.

Conclusion

In elderly patients (≥65 years) hospitalised with CHF, malnutrition, as assessed with PNI and GNRI, was common and associated with increased 1-year mortality. Prognostic value varied by BMI and inflammatory status, as measured by hsCRP and suPAR, with GNRI showing greater sensitivity to inflammation and better post-discharge prediction. Given that the GLIM criteria are often impractical in acute settings, our findings support the use of simple, routinely available indices for risk stratification in this setting.

Click here to view Supplementary Material.