Association between triglyceride-glucose-body mass index and adverse prognosis in elderly patients with severe heart failure and type 2 diabetes: a retrospective study based on the MIMIC-IV database

Abstract

Objective: The triglyceride-glucose (TyG) index is a validated marker of insulin resistance (IR) and predictor of cardiovascular outcomes. However, the prognostic utility of integrating TyG with body mass index (BMI) as the TyG-BMI index in elderly patients with severe heart failure (HF) and type 2 diabetes mellitus (T2DM) remains unestablished. We aimed to evaluate associations between TyG-BMI and all-cause mortality at multiple time points in this high-risk cohort.

Methods: This retrospective cohort study analyzed 4,523 elderly patients (aged >65 years) with severe HF and T2DM from the MIMIC-IV database. Participants were stratified into TyG-BMI quartiles (Q1-Q4) at ICU admission. Primary outcomes were 60-, 90-, 180-, and 365-day all-cause mortality. Associations were assessed using Kaplan-Meier analysis, Cox proportional hazards models, and restricted cubic splines (RCS).

Results: The cohort (mean age 72.79 ± 7.84 years; 41.5% male) demonstrated graded mortality reductions with increasing TyG-BMI quartiles. Compared to Q4, Q1 (lowest TyG-BMI) had significantly higher mortality at 90 days (58.70% vs. 48.45%; p = 0.008) and 365 days (80.54% vs. 73.91%; p < 0.001), with similar 60-day trends (58.79% vs. 39.34%; p = 0.059). Adjusted Cox models confirmed progressively lower mortality risk in higher quartiles (365-day HR for Q4 vs. Q1: 0.74, 95% CI: 0.68-0.93). Subgroup analyses demonstrated a consistent inverse TyG-BMI-mortality association across all strata (age, cardiac function, comorbidities), with pronounced risk reduction in HFrEF (LVEF ≤40%; all-timepoint HR >1, p<0.05) and patients without prior myocardial infarction (365-day aHR 0.69 vs. 0.81 with infarction). RCS analysis identified nonlinear thresholds (TyG-BMI = 148.73 for 60-day; 163.38 for 365-day mortality), below which each unit increase conferred greater protective effects.

Conclusion: Lower TyG-BMI independently predicted increased short-, intermediate-, and long-term mortality in elderly patients with severe HF and T2DM. This composite index-integrating metabolic (TyG) and nutritional (BMI) dimensions-provides practical risk stratification, particularly within identified threshold ranges.

Keywords: All-cause mortality; Heart failure; Insulin resistance; Triglyceride-glucose body mass index.

Fig. 1

Workflow of the present study

Fig. 2

(A) 60-day survival, (B) 90-day survival, (C) 180-day survival, (D) 365-day survival. Curves depict cumulative survival probabilities in elderly ICU patients with severe heart failure and type 2 diabetes (n=4,523), stratified by TyG-BMI quartiles at admission: Q1 (TyG-BMI <109.75, n=1,131), Q2 (109.75–127.80, n=1,131), Q3 (127.80–151.79, n=1,131), Q4 (>151.79, n=1,130). Log-rank tests confirm significantly reduced survival in Q1 (lowest TyG-BMI) versus Q4 (highest TyG-BMI) at 60-day (p<0.05), 90-day (p<0.001), 180-day (p<0.001) and 365-day (p<0.001), consistent with time-specific mortality differences reported in Table 1

Fig. 3

Threshold effect of TyG-BMI on mortality risk revealed by restricted cubic spline models. (A) RCS curves for 60-day all-cause mortality across different clinical subgroups. Hazard ratios (HRs) and 90% confidence intervals (CIs) are shown relative to TyG-BMI reference values. (B) RCS curves for 90-day all-cause mortality across subgroups. (C) RCS curves for 180-day all-cause mortality across subgroups. (D) RCS curves for 365-day all-cause mortality across subgroups

Fig. 4

Forest plots of stratified analyses of TyG-BMI index and 60-day all-cause mortality. Forest plot displaying hazard ratios (HR) and 95% confidence intervals (CI) from univariate Cox proportional hazards models. Variables include Age (stratified at 77.91 years), Gender, APSIII (cutoff: 53), SAPSII (cutoff: 44), Smoking status, Cerebrovascular disease, CK-MB (cutoff: 25 U/L), Troponin (cutoff: 0.2 ng/mL), BNP (stratified: ≤ 100, 100–500, >500 pg/mL), and LVEF (≤ 40%, 40–50%, ≥ 50%). Reference line at HR = 1

Fig. 5

Forest plots of stratified analyses of TyG-BMI index and 90-day all-cause mortality. Unadjusted hazard ratios (95% CI) for 90-day all-cause mortality across clinical subgroups. Variables analyzed: Age (>77.91/ ≤ 77.91 years), Gender, APSIII (>53/ ≤ 53), SAPSII (>44/ ≤ 44), Smoking, Cerebrovascular disease, CK-MB (>25/ ≤ 25 U/L), Troponin (>0.2/ ≤ 0.2 ng/mL), BNP (three-tiered), and LVEF (three categories). Dashed vertical line indicates null effect (HR = 1). P-values reflect significance of subgroup differences

Fig. 6

Forest plots of stratified analyses of TyG-BMI index and 180-day all-cause mortality. Univariate Cox regression results for 180-day mortality risk. Subgroups: Age (dichotomized at 77.91 years), Gender, APSIII (cutoff: 53), SAPSII (cutoff: 44), Smoking history, Cerebrovascular disease status, CK-MB levels (25 U/L threshold), Troponin levels (0.2 ng/mL threshold), BNP categories, and LVEF strata. Error bars represent 95% CIs; filled diamonds mark point estimates. All models are unadjusted for covariates

Fig. 7

Forest plots of stratified analyses of TyG-BMI index and 365-day all-cause mortality. Long-term (365-day) all-cause mortality risk stratification by clinical parameters. Univariate hazard ratios shown for: Age groups, Gender, disease severity scores (APSIII/SAPSII), Smoking, Cerebrovascular disease, cardiac biomarkers (CK-MB, Troponin, BNP), and LVEF categories. Note the progressive risk elevation with higher BNP levels (HR = 3.09 for >500 pg/mL) and persistent mortality risk in reduced LVEF (≤ 40%). Scale extended to HR = 3.5 to accommodate effect sizes