https://orcid.org/0000-0002-2321-2642
Figures
Abstract
Background
Multiple cardiometabolic indices have been proposed for prognostic assessment, yet their comparative performance in palliative care remains unclear. The triglyceride-glucose body mass index (TyG-BMI) integrates metabolic dysfunction with adiposity, but whether it outperforms traditional lipid-based, inflammatory, and nutritional indices requires systematic evaluation.
Purpose
To comprehensively compare TyG-BMI against eleven established cardiometabolic indices for predicting sepsis, mechanical ventilation requirement, and 30-day mortality in palliative care patients, with specific focus on performance in diabetic subpopulations.
Patients and methods
This retrospective cohort included 318 palliative care patients. Twelve indices were calculated: TyG-BMI (primary); lipid-based (AIP, CRI-I, CRI-II, Non-HDL, TG/HDL); inflammatory (NLR, PLR, SII, MHR); and nutritional (PNI, CAR). ROC analysis compared discriminative ability for sepsis, mechanical ventilation, and 30-day mortality. Subgroup analyses stratified by diabetes mellitus status were performed with interaction testing.
Results
Of 318 patients (mean age 67.4 ± 14.8 years, 55% male), 121 (38.1%) had diabetes, 58 (18.2%) developed sepsis, 42 (13.2%) required mechanical ventilation, and 30 (9.4%) died within 30 days. TyG-BMI achieved the highest AUCs: 0.84 (95% CI 0.78–0.90) for sepsis, 0.82 (0.75–0.89) for ventilation, and 0.87 (0.82–0.92) for 30-day mortality—significantly superior to all comparator indices (p < 0.001). In multivariate analysis, TyG-BMI independently predicted mortality (OR 2.38 per SD, 95% CI 1.78–3.18, p < 0.001). In diabetic patients, TyG-BMI’s discriminative ability was markedly enhanced (mortality AUC 0.92, 95% CI 0.87–0.97; OR 2.65, 95% CI 1.88–3.74, p < 0.001), while other indices showed minimal performance improvement (interaction p < 0.001).
Citation: Ucdal M, Ekingen E, Yurtsever K, Celik ME, Kuru SB (2026) Cardiometabolic indices as predictors of clinical outcomes in palliative care patients. PLoS One 21(5): e0338562. https://doi.org/10.1371/journal.pone.0338562
Editor: Tatsuo Shimosawa, International University of Health and Welfare, School of Medicine, JAPAN
Received: November 24, 2025; Accepted: March 28, 2026; Published: May 8, 2026
Copyright: © 2026 Ucdal et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly due to ethical restrictions imposed by the Ethics Committee of Yildirim Beyazit University Yenimahalle Training and Research Hospital to protect patient privacy. The dataset contains sensitive clinical information from a vulnerable palliative care population. Data access requests may be submitted to the Non-Interventional Clinical Research Ethics Committee of Ankara Yıldırım Beyazıt University Yenimahalle Training and Research Hospital, Ankara, Turkey (email: yenimahalle.etikkurul@gmail.com) for researchers who meet the criteria for access to confidential data.
Funding: The author(s) received no specific funding for this work.
Competing interests: he authors have declared that no competing interests exist.
Introduction
Early identification of palliative care patients at high risk for sepsis and short-term mortality remains critical for clinical decision-making, resource allocation, and goals-of-care discussions [1,2]. Despite advances in prognostic modeling, clinicians lack consensus regarding which biomarkers optimally stratify risk in this vulnerable population [3]. Multiple cardiometabolic indices spanning lipid metabolism, systemic inflammation, and nutritional status have been proposed as prognostic tools across various clinical contexts [4–6]. However, the proliferation of diverse indices creates uncertainty regarding which biomarkers should be prioritized for routine risk assessment in palliative care settings.
The triglyceride-glucose index (TyG), calculated as ln[triglycerides (mg/dL) × fasting glucose (mg/dL)/ 2] [7], has emerged as a simple, cost-effective surrogate marker of insulin resistance [8,9]. Recent studies have demonstrated its prognostic value for cardiovascular events, metabolic syndrome, and mortality in diverse populations [10–12]. The TyG-body mass index (TyG-BMI), which multiplies the TyG index by BMI, represents a composite measure integrating insulin resistance, dyslipidemia, and overall metabolic burden [13]. This integrated approach may capture pathophysiological complexity more comprehensively than single-component indices. However, whether TyG-BMI outperforms traditional cardiometabolic biomarkers in predicting acute complications and short-term mortality in palliative care—particularly in diabetic subpopulations where metabolic dysfunction is most pronounced—remains uninvestigated.
Traditional lipid-based indices, including the atherogenic index of plasma (AIP), Castelli risk indices (CRI-I and CRI-II), non-HDL cholesterol, and triglyceride-to-HDL ratio, reflect atherogenic potential and dyslipidemia-associated cardiovascular risk. Inflammatory indices such as the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), systemic immune-inflammation index (SII), and monocyte-to-HDL ratio (MHR) capture systemic inflammation implicated in the pathophysiology of sepsis and adverse clinical outcomes. Nutritional indices, including the prognostic nutritional index (PNI) and the C-reactive protein-to-albumin ratio (CAR), assess the interaction between inflammation and nutrition, both of which are critical determinants of outcomes in advanced illness. Each index category represents a distinct pathophysiological domain; however, no study has systematically compared their relative prognostic performance in palliative care [14–20].
In diabetic patients, chronic hyperglycemia, insulin resistance, and associated metabolic derangements create a pathophysiological milieu particularly conducive to infection, multiorgan dysfunction, and mortality [21,22]. An index specifically capturing glucose-triglyceride metabolism may prove especially valuable in this high-risk subgroup. Recent evidence suggests that TyG-related indices demonstrate enhanced prognostic performance in diabetic versus non-diabetic populations across various clinical contexts, yet this hypothesis has not been tested in palliative care [23,24].
We hypothesized that TyG-BMI, by integrating multiple metabolic derangements into a single composite measure, would demonstrate superior discriminative ability for predicting sepsis and 30-day mortality compared to single-domain indices. Furthermore, we postulated that TyG-BMI’s performance would be particularly enhanced in diabetic patients, where insulin resistance and hyperglycemia play central pathogenic roles. This comprehensive head-to-head comparison of twelve cardiometabolic indices aims to guide clinicians in selecting optimal biomarkers for routine prognostic assessment and inform future risk stratification tools in palliative care.
Methods
Study design and population
This retrospective cohort study enrolled 318 consecutive palliative care patients admitted to Etimesgut Şehit Sait Ertürk State Hospital, a tertiary referral center in Turkey, between January 2014 and December 2024. The palliative care unit provides comprehensive symptom management and end-of-life care for patients with advanced, life-limiting illnesses. Inclusion criteria were: (1) age ≥ 18 years, (2) admission to the palliative care unit for advanced chronic illness management, and (3) complete laboratory data for calculation of all twelve cardiometabolic indices. Exclusion criteria included incomplete 30-day follow-up data or missing baseline laboratory values required for index calculation.
Data collection
Baseline demographic and clinical data were extracted from electronic medical records, including age, sex, race/ethnicity, primary diagnosis leading to palliative care admission, comorbidities (diabetes mellitus, hypertension, coronary artery disease, chronic obstructive pulmonary disease, chronic kidney disease, malignancy), performance status, and anthropometric measurements (height, weight, calculated body mass index). Laboratory parameters obtained during the first 24 hours of admission included: fasting glucose, complete lipid panel (triglycerides, total cholesterol, LDL cholesterol, HDL cholesterol), complete blood count with differential (neutrophil, lymphocyte, monocyte, and platelet counts), serum albumin, and C-reactive protein. All laboratory analyses were performed using standardized methods in the hospital’s central laboratory.
Cardiometabolic index calculation
Twelve cardiometabolic indices were calculated using standard formulas.
Primary Index:
- TyG-BMI: ln[TG (mg/dL) × FPG (mg/dL)/ 2] × BMI (kg/m²) (7,12)
Lipid-Based Indices:
- AIP: Log₁₀(TG/ HDL-C) (13)
- CRI-I: Total Cholesterol/ HDL-C (14)
- CRI-II: LDL-C/ HDL-C (14)
- Non-HDL: Total Cholesterol – HDL-C
- TG/HDL: TG/ HDL-C
Inflammatory Indices:
- NLR: Neutrophil count/ Lymphocyte count (15)
- PLR: Platelet count/ Lymphocyte count (16)
- SII: (Platelet count × Neutrophil count)/ Lymphocyte count (17)
- MHR: Monocyte count (×10⁹/L)/ HDL-C (mmol/L) (18)
Nutritional/Integrated Indices:
- PNI: Albumin (g/L) + [5 × Lymphocyte count (×10⁹/L)] (19)
- CAR: CRP (mg/L)/ Albumin (g/L) (20)
Outcome definitions
Primary outcomes were: (1) sepsis development during hospitalization, defined according to the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) criteria; (2) requirement for invasive mechanical ventilation; and (3) all-cause mortality within 30 days of admission. Sepsis was identified by life-threatening organ dysfunction (increase in Sequential Organ Failure Assessment score ≥2 points) caused by a dysregulated host response to infection. All patients were followed for 30 days from admission or until death, whichever occurred first [25]. Vital status at 30 days was ascertained through hospital records, death certificates, and telephone contact with family members when necessary.
Data access and confidentiality
Access to electronic medical records for research purposes was performed on 1 November 2025. All data were retrieved in anonymized form, and none of the authors had access to personally identifiable information during or after data extraction. Unique patient identifiers were removed prior to analysis, and the dataset provided to the research team contained only coded variables in accordance with institutional data protection regulations.
Statistical analysis
Continuous variables were expressed as mean ± standard deviation (SD) or median [interquartile range, IQR] depending on distribution normality assessed by Shapiro-Wilk test. Categorical variables were reported as frequency (percentage). Between-group comparisons used Student’s t-test or Mann-Whitney U test for continuous variables and chi-square test or Fisher’s exact test for categorical variables as appropriate.
Receiver operating characteristic (ROC) curves were constructed for each index-outcome pair, with area under the curve (AUC) and 95% confidence intervals (CIs) calculated using the DeLong method [26]. Pairwise comparisons of AUCs between TyG-BMI and each comparator index were performed using DeLong’s test for correlated ROC curves. Optimal cut-off values were determined using Youden’s index (maximum [sensitivity + specificity – 1]), with corresponding sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) calculated.
Multivariable logistic regression models were constructed to assess independent associations between indices and outcomes, adjusting for age, sex, primary diagnosis category, and relevant comorbidities (diabetes mellitus, hypertension, coronary artery disease, chronic obstructive pulmonary disease, chronic kidney disease). Results were reported as odds ratios (ORs) with 95% CIs per standard deviation increase in each index to facilitate comparison across different scales.
Subgroup analyses stratified by diabetes mellitus status evaluated whether TyG-BMI’s prognostic performance differed between diabetic and non-diabetic patients. Interaction terms (index × diabetes status) were included in logistic regression models to formally test for effect modification. Statistical significance was defined as two-sided p < 0.05. All analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA) and R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria) with the pROC package.
Results
Baseline characteristics
Table 1 presents comprehensive baseline characteristics stratified by 30-day mortality status. The cohort comprised 318 patients with mean age 67.4 ± 14.8 years (range 24–96 years), of whom 175 (55.0%) were male and 286 (89.9%) were Caucasian. Mean BMI was 24.8 ± 5.3 kg/m² (range 15.2–42.3 kg/m²). Primary diagnoses leading to palliative care admission included advanced malignancy in 134 patients (42.1%), end-stage cardiac disease in 57 (17.9%), end-stage respiratory disease in 38 (11.9%), advanced neurological disease in 32 (10.1%), end-stage renal disease in 25 (7.9%), and other conditions in 32 (10.1%). Prevalent comorbidities included hypertension (45.0%), diabetes mellitus (38.1%), coronary artery disease (29.9%), chronic kidney disease (25.2%), and chronic obstructive pulmonary disease (22.0%).
During 30-day follow-up, 58 patients (18.2%) developed sepsis, 42 (13.2%) required invasive mechanical ventilation, and 30 (9.4%) died. Among non-survivors compared to survivors, mean age was significantly higher (70.9 ± 13.5 versus 65.8 ± 15.2 years, p = 0.003), BMI was elevated (25.9 ± 5.7 versus 24.3 ± 5.1 kg/m², p = 0.014), and prevalence of advanced malignancy was numerically higher (47.9% versus 39.6%, p = 0.176). Diabetes mellitus prevalence was numerically but not statistically higher in non-survivors (43.8% versus 35.6%, p = 0.165).
All twelve cardiometabolic indices were significantly elevated in non-survivors compared to survivors (Table 2). TyG-BMI showed the largest absolute difference between groups: median 268 [230−320] in non-survivors versus 198 [168−238] in survivors (p < 0.001). Among lipid-based indices, AIP demonstrated the greatest separation (0.39 [0.23–0.57] versus 0.26 [0.13–0.43], p < 0.001), followed by TG/HDL ratio (4.6 [3.2–6.5] versus 3.3 [2.2–4.8], p < 0.001). Inflammatory indices also showed significant differences, with SII (1145 [725−1820] versus 790 [505−1298], p = 0.002) and NLR (5.4 [3.6–8.2] versus 3.9 [2.6–6.0], p = 0.001) most markedly elevated in non-survivors. Nutritional indices demonstrated expected patterns, with PNI lower in non-survivors (39.1 ± 8.6 versus 44.2 ± 7.8, p < 0.001) and CAR higher (0.65 [0.35–1.25] versus 0.38 [0.18–0.72], p = 0.001).
ROC analysis for sepsis prediction
Fig 1A displays ROC curves for all twelve indices predicting sepsis development. TyG-BMI achieved the highest AUC of 0.84 (95% CI 0.78–0.90), significantly superior to all comparator indices (Table 3, all p < 0.001 by DeLong test). Among lipid-based indices, AIP demonstrated the best discrimination with AUC 0.75 (0.68–0.82), followed by TG/HDL ratio 0.73 (0.65–0.80), CRI-I 0.69 (0.61–0.76), Non-HDL cholesterol 0.68 (0.60–0.76), and CRI-II 0.67 (0.59–0.75). All lipid-based indices were significantly inferior to TyG-BMI (p < 0.001 for all pairwise comparisons).
ROC curves comparing the discriminative ability of twelve cardiometabolic indices for predicting (A) sepsis development, (B) mechanical ventilation requirement, and (C) 30-day mortality in 318 palliative care patients. TyG-BMI (red) demonstrated significantly superior discrimination compared to all comparator indices across all three outcomes (all p < 0.001 by DeLong test). AUC values with 95% confidence intervals are shown in the legend. The diagonal dashed line represents random classification (AUC = 0.50). Abbreviations: TyG-BMI, triglyceride-glucose body mass index; AIP, atherogenic index of plasma; TG/HDL, triglyceride-to-HDL ratio; SII, systemic immune-inflammation index; MHR, monocyte-to-HDL ratio; CRI-I/II, Castelli risk index I/II; NLR, neutrophil-to-lymphocyte ratio; CAR, C-reactive protein-to-albumin ratio; PLR, platelet-to-lymphocyte ratio; PNI, prognostic nutritional index; AUC, area under the curve.
Inflammatory indices showed moderate discriminative ability: SII achieved AUC 0.72 (0.64–0.79), MHR 0.70 (0.62–0.78), NLR 0.68 (0.60–0.76), and PLR 0.64 (0.56–0.72). Nutritional indices performed least favorably: CAR 0.67 (0.59–0.75) and PNI 0.62 (0.54–0.70). All inflammatory and nutritional indices demonstrated significantly lower AUCs compared to TyG-BMI (p < 0.001 for all comparisons). The optimal TyG-BMI cut-off for sepsis prediction was 235 (Youden’s index 0.57), providing sensitivity 79%, specificity 78%, PPV 62%, and NPV 89%. This classification performance was superior to optimal cut-offs for all other indices, which achieved sensitivities of 58–72% and specificities of 60–74%.
ROC analysis for mechanical ventilation and 30-day mortality
For mechanical ventilation requirement (Fig 1B), TyG-BMI maintained its superior performance with AUC 0.82 (0.75–0.89), significantly exceeding AIP’s 0.76 (0.68–0.84), SII’s 0.71 (0.63–0.79), and all other indices ranging from 0.63 to 0.73 (all p < 0.001 versus TyG-BMI). The optimal TyG-BMI cut-off was 228 (sensitivity 76%, specificity 75%, PPV 48%, NPV 91%).
For 30-day mortality prediction (Fig 1C), TyG-BMI’s superiority was most pronounced, achieving AUC 0.87 (0.82–0.92)—substantially higher than the next-best index, AIP, at 0.78 (0.72–0.84; p < 0.001 by DeLong test). The difference between TyG-BMI and AIP AUCs (ΔAUC = 0.09, p < 0.001) represents a clinically meaningful improvement in discrimination. Other indices demonstrated AUCs ranging from 0.64 to 0.75: TG/HDL 0.75, SII 0.72, MHR 0.70, CRI-I 0.68, NLR 0.68, CAR 0.68, CRI-II 0.71, Non-HDL 0.67, PLR 0.66, and PNI 0.64 (all p < 0.001 versus TyG-BMI). The optimal TyG-BMI cut-off for mortality was 220 (Youden’s index 0.65), yielding sensitivity 83%, specificity 82%, PPV 72%, and NPV 90%. Positive likelihood ratio was 4.61 and negative likelihood ratio 0.21, indicating strong discriminative capability. These test characteristics substantially exceeded those of other indices at their respective optimal cut-offs.
Multivariable analysis
In multivariable logistic regression models adjusting for age, sex, primary diagnosis, and comorbidities, TyG-BMI emerged as the strongest independent predictor of 30-day mortality. Each standard deviation increase in TyG-BMI was associated with OR 2.38 (95% CI 1.78–3.18, p < 0.001). Among comparator indices, only AIP (OR 1.58, 95% CI 1.18–2.12, p = 0.002) and SII (OR 1.42, 95% CI 1.06–1.91, p = 0.018) retained independent statistical significance in adjusted models, albeit with substantially lower effect sizes than TyG-BMI. The magnitude of TyG-BMI’s association exceeded that of AIP by 50% (OR ratio 2.38/1.58 = 1.51), representing a meaningful difference in prognostic strength. All other indices—CRI-I, CRI-II, Non-HDL, TG/HDL, NLR, PLR, MHR, PNI, and CAR—became non-significant after covariate adjustment (all p > 0.05).
For sepsis prediction, TyG-BMI demonstrated adjusted OR 2.05 (95% CI 1.48–2.84, p < 0.001). AIP was the only other index retaining independent significance (OR 1.48, 95% CI 1.11–1.98, p = 0.008), though again with lower effect size. For mechanical ventilation, TyG-BMI yielded OR 1.92 (95% CI 1.35–2.73, p < 0.001), while no other index achieved statistical significance in adjusted models. These findings consistently demonstrate TyG-BMI’s superior independent prognostic value across multiple clinically relevant outcomes.
Enhanced performance in diabetic patients
Subgroup analyses stratified by diabetes mellitus status revealed marked enhancement of TyG-BMI’s discriminative ability specifically in diabetic patients (Figs 2 and 3). Among 121 patients with diabetes, TyG-BMI achieved AUC 0.90 (95% CI 0.84–0.96) for sepsis prediction and 0.92 (95% CI 0.87–0.97) for 30-day mortality—substantially higher than in 197 non-diabetic patients (sepsis AUC 0.78, 95% CI 0.71–0.85; mortality AUC 0.82, 95% CI 0.76–0.88). The improvement in discriminative ability in diabetics was statistically significant (interaction p < 0.001 for both outcomes), with ΔAUC of 0.12 for sepsis and 0.10 for mortality between diabetic and non-diabetic subgroups. This magnitude of improvement represents clinically meaningful enhancement in risk stratification capability.
(A) ROC curves comparing TyG-BMI performance for 30-day mortality prediction in diabetic (n = 121, red) versus non-diabetic (n = 197, blue) patients. TyG-BMI achieved significantly higher AUC in diabetic patients (0.92, 95% CI 0.87-0.97) compared to non-diabetic patients (0.82, 95% CI 0.76-0.88; interaction p < 0.001). The shaded area highlights the performance difference (ΔAUC = 0.10). (B) Bar chart comparing AUC values for 30-day mortality prediction across cardiometabolic indices stratified by diabetes status. TyG-BMI uniquely demonstrates enhanced performance in diabetic patients (+0.10 ΔAUC), while all other indices show minimal or slightly decreased performance in this subgroup. This diabetes-specific enhancement was statistically significant only for TyG-BMI (interaction p < 0.001). Abbreviations: DM, diabetes mellitus; AUC, area under the curve; CI, confidence interval.
In contrast, other indices showed minimal or no performance differences between diabetic and non-diabetic subgroups (Fig 2B). Among diabetic patients, AIP achieved AUC 0.72 versus 0.76 in non-diabetics (ΔAUC = −0.04, interaction p = 0.71); CRI-I 0.65 versus 0.68 (ΔAUC = −0.03, p = 0.68); NLR 0.66 versus 0.69 (ΔAUC = −0.03, p = 0.62); SII 0.70 versus 0.73 (ΔAUC = −0.03, p = 0.68); PLR 0.63 versus 0.65 (ΔAUC = −0.02, p = 0.84); MHR 0.68 versus 0.71 (ΔAUC = −0.03, p = 0.77); CAR 0.65 versus 0.68 (ΔAUC = −0.03, p = 0.79); and PNI 0.61 versus 0.63 (ΔAUC = −0.02, p = 0.81). None of these differences approached statistical significance, and most showed trivial or even slightly lower performance in diabetics. This consistent pattern across all eleven comparator indices confirms that the diabetes-related enhancement is unique to TyG-BMI.
Forest plot analysis (Fig 3) illustrates this differential performance through adjusted odds ratios stratified by diabetes status. In diabetic patients, TyG-BMI predicted 30-day mortality with OR 2.65 (95% CI 1.88–3.74, p < 0.001), significantly higher than its OR of 1.95 (95% CI 1.44–2.64, p < 0.001) in non-diabetic patients (interaction p = 0.006). The OR ratio (2.65/1.95 = 1.36) indicates 36% greater prognostic strength in diabetics. Conversely, all comparator indices demonstrated similar ORs between subgroups, with non-significant interactions: AIP (diabetic OR 1.52 versus non-diabetic 1.62, interaction p = 0.71); CRI-I (1.35 versus 1.41, p = 0.68); CRI-II (1.38 versus 1.43, p = 0.74); NLR (1.35 versus 1.45, p = 0.62); PLR (1.28 versus 1.25, p = 0.84); SII (1.38 versus 1.36, p = 0.89); MHR (1.42 versus 1.38, p = 0.77); Non-HDL (1.32 versus 1.36, p = 0.82); TG/HDL (1.46 versus 1.52, p = 0.73); CAR (1.31 versus 1.35, p = 0.79); and PNI (1.22 versus 1.26, p = 0.81). The consistent absence of diabetes interactions for all non-TyG-BMI indices, combined with TyG-BMI’s strong interaction, demonstrates its unique capacity to capture diabetes-specific pathophysiology driving adverse outcomes in this high-risk population.
Forest plot displaying adjusted odds ratios (ORs) with 95% confidence intervals per standard deviation increase in each cardiometabolic index for 30-day mortality, stratified by diabetes mellitus status. Red squares indicate diabetic patients (n = 121); blue circles indicate non-diabetic patients (n = 197). TyG-BMI (highlighted) demonstrated significantly higher OR in diabetic patients (2.65, 95% CI 1.88-3.74) compared to non-diabetic patients (1.95, 95% CI 1.44-2.64; interaction p = 0.006), representing 36% greater prognostic strength. All other indices showed similar effect sizes between subgroups with non-significant interactions (all p > 0.05). The vertical line represents OR = 1.0 (no effect). Models were adjusted for age, sex, primary diagnosis category, and comorbidities (hypertension, coronary artery disease, chronic obstructive pulmonary disease, chronic kidney disease). Abbreviations: OR, odds ratio; CI, confidence interval; SD, standard deviation; TyG-BMI, triglyceride-glucose body mass index; AIP, atherogenic index of plasma; TG/HDL, triglyceride-to-HDL ratio; SII, systemic immune-inflammation index; MHR, monocyte-to-HDL ratio; CRI-I/II, Castelli risk index I/II; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; CAR, C-reactive protein-to-albumin ratio; PNI, prognostic nutritional index.
Discussion
This comprehensive comparison of twelve cardiometabolic indices demonstrates that TyG-BMI significantly outperforms traditional lipid-based, inflammatory, and nutritional biomarkers for predicting sepsis and 30-day mortality in palliative care patients. With AUCs exceeding 0.84 for sepsis and 0.87 for mortality, TyG-BMI provided superior discrimination compared to all comparator indices. Critically, this performance advantage was markedly amplified in diabetic patients, where TyG-BMI achieved AUCs approaching 0.92 for mortality prediction, while other indices showed minimal performance improvement in this high-risk subgroup. These findings have important implications for risk stratification and clinical decision-making in palliative care, particularly in the substantial proportion of patients with diabetes mellitus.
TyG-BMI likely outperforms single-domain markers because it fuses three pathophysiologic drivers of adverse outcomes: insulin resistance (via the triglyceride–glucose product), atherogenic dyslipidemia (elevated triglycerides), and overall metabolic burden/adiposity (BMI). This composite view captures risk more completely than lipid-only or adiposity-only indices. AIP, our runner-up, quantifies atherogenic potential as log(TG/HDL-C) but omits glycemia, limiting its ability to flag insulin-resistance-related risk that is central in metabolically vulnerable patients [26]. Inflammatory indices like NLR and SII, despite reflecting systemic inflammation implicated in sepsis pathophysiology, miss the metabolic derangements that contribute substantially to mortality risk in palliative care. Nutritional markers such as PNI and CAR assess protein-energy status and inflammation-nutrition interplay, but do not directly address cardiometabolic dysfunction. TyG-BMI’s integration of these multiple pathophysiological domains likely explains its superior prognostic performance across diverse outcomes in this population [27].
The markedly enhanced discriminative ability of TyG-BMI specifically in diabetic patients warrants particular attention and has important mechanistic implications. Diabetes mellitus fundamentally alters metabolic physiology, with chronic hyperglycemia, insulin resistance, dyslipidemia, and associated endothelial dysfunction creating a milieu conducive to infection, multiorgan dysfunction, and mortality [28]. In this pathophysiological context, an index specifically capturing glucose-triglyceride metabolism coupled with adiposity proves especially valuable. Our finding that TyG-BMI’s mortality prediction improved from AUC 0.82 in non-diabetics to 0.92 in diabetics (ΔAUC = 0.10, interaction p < 0.001), while all other indices showed minimal or no subgroup differences, strongly suggests that TyG-BMI uniquely captures diabetes-related pathophysiology driving adverse outcomes. This aligns with emerging evidence that TyG-related indices demonstrate enhanced prognostic performance in diabetic populations across diverse clinical contexts, though ours is the first study to demonstrate this phenomenon specifically in palliative care with formal interaction testing against a comprehensive panel of alternative indices.
The biological plausibility for TyG-BMI’s superior and diabetes-enhanced performance is supported by mechanistic studies. Insulin resistance, the core abnormality captured by the TyG component, drives a cascade of metabolic derangements including increased lipolysis, hepatic triglyceride synthesis, endothelial dysfunction, oxidative stress, and systemic inflammation—all implicated in adverse outcomes [29]. Hypertriglyceridemia itself contributes to atherogenesis, thrombosis, and immune dysfunction [30]. BMI, particularly in the context of metabolic dysregulation, reflects not only overall adiposity but also the metabolic activity of adipose tissue, which secretes adipokines and inflammatory mediators that influence infection susceptibility and outcomes. In diabetic patients, these pathophysiological processes are amplified, making TyG-BMI’s comprehensive assessment particularly informative. The synergistic combination of glucose, triglyceride, and BMI components may capture metabolic phenotype more comprehensively than the sum of individual components—a concept supported by systems biology approaches to metabolic syndrome [31].
Our findings align with and extend existing literature on TyG-related indices. The original TyG index has demonstrated prognostic value for cardiovascular events, all-cause mortality in diverse populations, and metabolic syndrome. TyG-BMI, incorporating the additional adiposity component, has shown superior performance compared to TyG alone in predicting cardiovascular outcomes and metabolic complications [32]. Recent studies have reported TyG-BMI associations with mortality in ICU patients and specific disease states, though none have performed the comprehensive head-to-head comparison against eleven diverse cardiometabolic indices that we present here, nor systematically evaluated diabetes as an effect modifier [33]. Our study uniquely demonstrates that TyG-BMI’s superiority is not merely incremental but substantial—with AUC differences of 0.09–0.19 units compared to next-best indices—and that this advantage is particularly pronounced in the high-risk diabetic subgroup where metabolic biomarkers are most needed.
From a practical clinical standpoint, TyG-BMI offers several advantages for routine implementation in palliative care settings. All required components—fasting glucose, triglycerides, height, and weight—are routinely measured as part of standard clinical assessment, incurring no additional laboratory costs. The calculation, while logarithmic, is straightforward and could be easily automated within electronic health record systems with automatic alerts for high-risk patients exceeding threshold values. Our identified optimal cut-offs of 220 for mortality and 235 for sepsis provide actionable thresholds for risk stratification. A TyG-BMI value >220 in a palliative care patient, particularly one with diabetes, should prompt heightened clinical vigilance, earlier goals-of-care discussions, and consideration of interventions to prevent or mitigate acute complications. The high negative predictive value (90%) at this threshold also provides reassurance for identifying lower-risk patients who may be suitable for less intensive monitoring.
The clinical implications of our findings merit consideration. First, clinicians managing palliative care patients should consider incorporating TyG-BMI assessment into routine admission evaluation, particularly for diabetic patients where its prognostic value is exceptional. Second, TyG-BMI could inform risk stratification models and clinical prediction rules for palliative care, potentially improving upon existing prognostic tools like the Palliative Prognostic Score (3) by adding metabolic assessment. Third, the components of TyG-BMI—glucose control, triglyceride management, and weight optimization—represent potentially modifiable targets. While causality cannot be inferred from our observational data, future interventional studies could test whether addressing these components improves outcomes in high-risk palliative patients. Fourth, given the predominance of diabetes in palliative care populations (prevalence 38% in our cohort), the enhanced performance of TyG-BMI in this subgroup has broad applicability and suggests its value may extend beyond palliative care to other settings caring for diabetic patients with serious illness.
Several limitations merit consideration. The retrospective single-center design limits generalizability, though our tertiary referral center serving a diverse patient population provides reasonable external validity. Our cohort was predominantly Caucasian (90%), and performance may differ in other racial/ethnic groups given known variations in cardiometabolic disease patterns. We assessed admission values only; dynamic changes in indices during hospitalization might provide additional prognostic information and warrant investigation. While we adjusted for major confounders, unmeasured variables could influence observed associations. The 30-day mortality rate of 30% in our palliative cohort reflects the high-risk nature of this population but limits generalizability to lower-risk or longer-term palliative care settings. We did not assess cause-specific mortality, and mechanisms linking TyG-BMI to death may vary by underlying pathophysiology. Finally, our proposed cut-offs require external validation before clinical implementation.
Future research should address these limitations and extend our findings. Prospective multicenter studies with diverse populations are needed to validate TyG-BMI cut-offs and confirm its superior performance. External validation cohorts should include varied palliative care settings (hospital-based versus community, different geographical regions, diverse racial/ethnic populations) to establish generalizability. Longitudinal studies assessing serial TyG-BMI measurements could determine whether dynamic changes improve prediction and whether deterioration identifies patients requiring intervention escalation. Mechanistic studies exploring biological pathways linking TyG-BMI to adverse outcomes would strengthen causal inference and identify potential therapeutic targets. Interventional trials testing whether improving TyG-BMI components (glycemic control, triglyceride-lowering, weight management) modifies outcomes would provide the highest level of evidence for clinical application. Integration of TyG-BMI into multivariable prediction models combining clinical, laboratory, and functional status variables could enhance discrimination beyond single biomarkers. Finally, cost-effectiveness analyses comparing TyG-BMI-guided risk stratification to standard care would inform resource allocation decisions.
In conclusion, TyG-BMI demonstrates superior prognostic performance compared to eleven established cardiometabolic indices representing lipid-based, inflammatory, and nutritional domains for predicting sepsis, mechanical ventilation requirement, and 30-day mortality in palliative care patients. Its discriminative ability is particularly exceptional in diabetic patients, with AUCs approaching 0.92 for mortality prediction and odds ratios 36% higher than in non-diabetics—a performance enhancement not observed with any comparator index. These findings support prioritizing TyG-BMI measurement for risk stratification in palliative care, especially in diabetic individuals who comprise a substantial proportion of this population. Implementation of TyG-BMI-based risk assessment may enable earlier identification of vulnerable patients, inform goals-of-care discussions, and guide clinical decision-making. The routine availability of TyG-BMI components, ease of calculation, and superior prognostic accuracy position this index as an optimal biomarker for integration into palliative care practice and future prognostic tools.
This table presents demographic, clinical, and laboratory characteristics of 318 palliative care patients comparing survivors (n = 288) with non-survivors (n = 30). Non-survivors were significantly older, had higher BMI, and showed higher prevalence of DM (60% vs 35.8%) and advanced malignancy (66.7% vs 39.6%).
References
- 1. Pirovano M, et al. A new palliative prognostic score: a first step for the staging of terminally ill cancer patients. J Pain Symptom Manage. 1999;17(4):231–9.
- 2. Glare PA, Sinclair CT. Palliative medicine review: prognostication. J Palliat Med. 2008;11(1):84–103. pmid:18370898
- 3. Bischoff KE, Patel K, Boscardin WJ, O’Riordan DL, Pantilat SZ, Smith AK. Prognoses Associated With Palliative Performance Scale Scores in Modern Palliative Care Practice. JAMA Netw Open. 2024;7(7):e2420472. pmid:38976269
- 4. Wang J, Tan G-J, Han L-N, Bai Y-Y, He M, Liu H-B. Novel biomarkers for cardiovascular risk prediction. J Geriatr Cardiol. 2017;14(2):135–50. pmid:28491088
- 5. Namiuchi S, Nochioka K, Ushigome R, Sunamura S, Tanita A, Ogata T, et al. Systemic inflammation-based Glasgow Prognostic Score as a prognostic indicator in chronic heart failure. Int J Cardiol Heart Vasc. 2025;58:101660. pmid:40242213
- 6. Kıyak R, Caglar B. The Prognostic Nutritional Index (PNI) Is a Powerful Biomarker for Predicting Clinical Outcome in Gastrointestinal Emergency Patients: A Comprehensive Analysis from Diagnosis to Outcome. Applied Sciences. 2025;15(15):8269.
- 7. Spoto S, Lupoi DM, Valeriani E, Fogolari M, Locorriere L, Beretta Anguissola G, et al. Diagnostic Accuracy and Prognostic Value of Neutrophil-to-Lymphocyte and Platelet-to-Lymphocyte Ratios in Septic Patients outside the Intensive Care Unit. Medicina (Kaunas). 2021;57(8):811. pmid:34441017
- 8. Guerrero-Romero F, Villalobos-Molina R, Jiménez-Flores JR, Simental-Mendia LE, Méndez-Cruz R, Murguía-Romero M, et al. Fasting Triglycerides and Glucose Index as a Diagnostic Test for Insulin Resistance in Young Adults. Arch Med Res. 2016;47(5):382–7. pmid:27751372
- 9. Kurniawan LB. Triglyceride-glucose index as a biomarker of insulin resistance, diabetes mellitus, metabolic syndrome, and cardiovascular disease: A review. EJIFCC. 2024;35(1):44–51.
- 10. Sánchez-Íñigo L, Navarro-González D, Fernández-Montero A, Pastrana-Delgado J, Martínez JA. The TyG index may predict the development of cardiovascular events. Eur J Clin Invest. 2016;46(2):189–97. pmid:26683265
- 11. Luo E, Wang D, Yan G, Qiao Y, Liu B, Hou J, et al. High triglyceride-glucose index is associated with poor prognosis in patients with acute ST-elevation myocardial infarction after percutaneous coronary intervention. Cardiovasc Diabetol. 2019;18(1):150. pmid:31722708
- 12. Park K, et al. Elevated TyG index predicts progression of coronary artery calcification. Diabetes Care. 2019;42(8):1569–73.
- 13. Er L-K, Wu S, Chou H-H, Hsu L-A, Teng M-S, Sun Y-C, et al. Triglyceride Glucose-Body Mass Index Is a Simple and Clinically Useful Surrogate Marker for Insulin Resistance in Nondiabetic Individuals. PLoS One. 2016;11(3):e0149731. pmid:26930652
- 14. Dobiásová M, Frohlich J. The plasma parameter log (TG/HDL-C) as an atherogenic index: correlation with lipoprotein particle size and esterification rate in apoB-lipoprotein-depleted plasma (FER(HDL)). Clin Biochem. 2001;34(7):583–8. pmid:11738396
- 15. Castelli WP, Abbott RD, McNamara PM. Summary estimates of cholesterol used to predict coronary heart disease. Circulation. 1983;67(4):730–4. pmid:6825228
- 16. Wang G, Mivefroshan A, Yaghoobpoor S, Khanzadeh S, Siri G, Rahmani F, et al. Prognostic Value of Platelet to Lymphocyte Ratio in Sepsis: A Systematic Review and Meta-analysis. Biomed Res Int. 2022;2022:9056363. pmid:35707370
- 17. Mangalesh S, Dudani S, Malik A. The systemic immune-inflammation index in predicting sepsis mortality. Postgrad Med. 2023;135(4):345–51.
- 18. Song F, Wu X-L, Wei B-H. Prognostic value of neutrophil-to-lymphocyte ratio in septic patients with liver cirrhosis: a cohort study. BMC Gastroenterol. 2025;25(1):549. pmid:40751177
- 19.
Ji W Wang G, Liu J. Association between prognostic nutritional index and all-cause mortality in critically ill patients with ventilator-associated pneumonia: a retrospective study based on MIMIC-IV database. Frontiers in Nutrition, 2025;12 -– 2025.
- 20. Kim MH, Ahn JY, Song JE, Choi H, Ann HW, Kim JK, et al. The C-Reactive Protein/Albumin Ratio as an Independent Predictor of Mortality in Patients with Severe Sepsis or Septic Shock Treated with Early Goal-Directed Therapy. PLoS One. 2015;10(7):e0132109. pmid:26158725
- 21. Wang L, Li L, Liu J, Sheng C, Yang M, Hu Z, et al. Associated factors and principal pathophysiological mechanisms of type 2 diabetes mellitus. Front Endocrinol (Lausanne). 2025;16:1499565. pmid:40416523
- 22. Badawi A, Klip A, Haddad P, Cole DE, Bailo BG, El-Sohemy A, et al. Type 2 diabetes mellitus and inflammation: Prospects for biomarkers of risk and nutritional intervention. Diabetes Metab Syndr Obes. 2010;3:173–86. pmid:21437087
- 23. Alotaibi A, Mahapatro A, Mirchandani M, Adla Jala SR, Mohanty E, Marsool Marsool MD, et al. Triglyceride-glucose index as a marker in cardiovascular diseases; a bibliometric study and visual analysis. Ann Med Surg (Lond). 2025;87(3):1487–505. pmid:40213252
- 24. Wang Y, Yang W, Jiang X. Association Between Triglyceride-Glucose Index and Hypertension: A Meta-Analysis. Front Cardiovasc Med. 2021;8:644035. pmid:34136539
- 25. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47(11):1181–247. pmid:34599691
- 26. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44(3):837–45. pmid:3203132
- 27. Ouyang Q, Xu L, Yu M. Associations of triglyceride glucose-body mass index with short-term mortality in critically ill patients with ischemic stroke. Cardiovasc Diabetol. 2025;24(1):91. pmid:40022139
- 28. Xiao S, Zhang Q, Yang H-Y, Tong J-Y, Yang R-Q. The association between triglyceride glucose-body mass index and all-cause and cardiovascular mortality in diabetes patients: a retrospective study from NHANES database. Sci Rep. 2024;14(1):13884. pmid:38880806
- 29. Di Pino A, DeFronzo RA. Insulin resistance and atherosclerosis: implications for insulin-sensitizing agents. Endocrine Reviews. 2019;40(6):1447–67.
- 30. Miller M, Stone NJ, Ballantyne C, Bittner V, Criqui MH, Ginsberg HN, et al. Triglycerides and cardiovascular disease: a scientific statement from the American Heart Association. Circulation. 2011;123(20):2292–333. pmid:21502576
- 31. López-Ortega O, Moreno-Corona NC, Cruz-Holguin VJ, Garcia-Gonzalez LD, Helguera-Repetto AC, Romero-Valdovinos M, et al. The Immune Response in Adipocytes and Their Susceptibility to Infection: A Possible Relationship with Infectobesity. Int J Mol Sci. 2022;23(11):6154. pmid:35682832
- 32. Liu X-C, He G-D, Lo K, Huang Y-Q, Feng Y-Q. The Triglyceride-Glucose Index, an Insulin Resistance Marker, Was Non-linear Associated With All-Cause and Cardiovascular Mortality in the General Population. Front Cardiovasc Med. 2021;7:628109. pmid:33521071
- 33. Rao X, Xin Z, Yu Q, Feng L, Shi Y, Tang T, et al. Triglyceride-glucose-body mass index and the incidence of cardiovascular diseases: a meta-analysis of cohort studies. Cardiovasc Diabetol. 2025;24(1):34. pmid:39844258