This retrospective observational cohort study included 655 patients diagnosed with ventricular aneurysm (VA) at the First Affiliated Hospital of Wenzhou Medical University from January 2004 to December 2024. An outline of the screening workflow is illustrated in Fig. 1. Ultrasound echocardiography data were retrieved from January 2004 to December 2024 (n = 8829). Participants without a diagnosis of ventricular aneurysm or those with non-persistent ventricular enlargement during follow-up were excluded (n = 7132), leaving 1697 participants with ventricular aneurysm. Further exclusions included: (1) patients without hospitalization records (n = 106); (2) participants for whom triglyceride (TG) and fasting blood glucose (FBG) data were unavailable (n = 420) (these missing records were from earlier years of the dataset); (3) participants aged less than 60 years (n = 165); (4) participants without diabetes (n = 218); (5) participants have a history of previous MI (n = 133). This resulted in 655 eligible patients with first-ever acute MI complicated by new-onset LVA for the final analysis.

Fig. 1.

Flowchart for subject screening and study design.

Patient clinical information was collected from prior hospital records and included demographics (age, sex, and body mass index [BMI] (BMI was dichotomized at 24 kg/m² according to the World Health Organization (WHO) Asia-Pacific criteria and the Chinese adult overweight guidelines, which define BMI $\ge$24 kg/m² as overweight [13])); clinical risk factors (blood pressure indices [systolic blood pressure (SBP), diastolic blood pressure (DBP)], lifestyle factors such as smoking and alcohol consumption, and medical history of diabetes mellitus, hypertension, and hyperlipidemia (the latter defined based on prior clinical diagnosis supplemented by baseline lipid profiles: total cholesterol (TC) $>$5.2 mmol/L, TG $>$1.7 mmol/L, low-density lipoprotein cholesterol (LDL-C) $>$3.4 mmol/L, or high-density lipoprotein cholesterol (HDL-C) 1.0 mmol/L [14])). and pharmacotherapy-related covariates included the use of lipid-lowering drugs, angiotensin receptor blockers (ARBs), angiotensin-converting enzyme inhibitors (ACEIs), beta-blockers, aldosterone antagonists, antiarrhythmic agents, and antidiabetic medications prior to admission, as well as the in-hospital use of beta-blockers, aldosterone antagonists, and antiarrhythmic drugs. In addition, the analysis included two percutaneous coronary intervention (PCI) variables: PCI status (No PCI, Emergency, Elective) and PCI type (Conservative, Stent, Balloon, or combinations).

Laboratory indices, including TC, TG, LDL-C, HDL-C, glycosylated hemoglobin (HbA1c), FBG, estimated glomerular filtration rate (eGFR), white blood cell (WBC) count, B-type natriuretic peptide (BNP), c reactive protein (CRP), monocytes count, lymphocyte count and neutrophil (NE) count , were obtained from routine clinical testing performed during hospitalization. All measurements adhered to standardized protocols and testing systems.

Echocardiographic data were collected from routine examinations performed during hospitalization. Given the long study span, two objectively verifiable data sources were available. For earlier cases, measurements were extracted directly from archived echocardiography reports. For more recent cases with preserved digital records, original DICOM images were retrieved and independently reviewed by two experienced echocardiographers following a standardized interpretation protocol. The echocardiographic data included left ventricular end-systolic dimension (LVESD), left ventricular end-diastolic dimension (LVEDD), and left ventricular ejection fraction (LVEF).

The study diagnosed a VA using the transthoracic echocardiography (TTE). The diagnostic criteria were: persistent outward bulging of the left ventricular wall during the cardiac cycle, characterized by thinning of the ventricular wall, absence of contraction, or paradoxical motion, and a clear demarcation from adjacent normal myocardium [15]. Diagnosis of diabetes mellitus followed the American Diabetes Association (ADA) standards, including a documented diagnosis of diabetes in the electronic medical record, treatment with insulin or oral antidiabetic agents, and laboratory findings of FBG $\ge$126 mg/dL or HbA1c $\ge$6.5% [16].

The TyG index quantifies insulin IR by combining fasting glucose with triglyceride levels. It was calculated as Ln (fasting TG [mg/dL] $\times$ FBG [mg/dL] / 2) [17]. The participants were classified into four groups (Q1, Q2, Q3, Q4) by the quartiles of the TyG index, with Q1 serving as the reference group. Using the fully adjusted model, receiver operating characteristic (ROC) curve analysis was performed to determine the optimal cutoff value for discrimination.

Follow-up data in this study were collected from medical records of both inpatients and outpatients, with the follow-up period calculated from the date of VA diagnosis until discharge or death, whichever occurred first. The main outcome assessed in this study was the occurrence of in-hospital malignant arrhythmias, which specifically included sudden cardiac death and episodes of ventricular tachycardia (VT) or ventricular fibrillation (VF). Sudden cardiac death was defined as sudden cardiac arrest due to cardiac causes in previously stable individuals, excluding non-cardiac causes such as trauma, poisoning, or cerebrovascular events. Ventricular tachycardia or ventricular fibrillation was defined as episodes requiring treatment with an implantable cardioverter-defibrillator (ICD) or untreated episodes of ventricular tachycardia lasting longer than 30 seconds detected by an ICD. The diagnosis was based on electrocardiogram (ECG) or ICD recordings [18].

In this study, continuous data were summarized as mean $\pm$ standard deviation (SD) or median (interquartile range, IQR) depending on normality. Differences were evaluated using one-way analysis of variance (ANOVA) for normally distributed variables and the Kruskal-Wallis test for non-normal data. Categorical variables, reported as frequencies (percentages), were compared via the chi-square test or Fisher’s exact test.

To comprehensively evaluate the association between the TyG index and in-hospital malignant arrhythmias, this metric was analyzed in three complementary ways: (1) as a continuous variable for Cox regression and restricted cubic spline (RCS) analyses to characterize the overall dose-response relationship; (2) as quartiles to illustrate risk gradients across TyG levels using Kaplan-Meier curves; (3) as a binary variable, using an optimal cutoff determined by ROC analysis (derived from the multivariate model) to assess its potential clinical applicability. Cox regression models were performed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Three models were constructed: Model 1 was unadjusted; Model 2 adjusted for age, gender, BMI, smoking, and alcohol consumption; Model 3 additionally incorporated blood parameters, medication use, and echocardiographic variables. Kaplan-Meier curves with log-rank tests were used to compare event incidence across TyG quartiles and across ROC-derived high- and low-TyG groups. RCS analysis was applied to assess potential nonlinear associations between continuous TyG levels and clinical outcomes.

Inflammatory markers were considered as potential biomarkers mediating the relationship between the TyG index and in-hospital malignant arrhythmias [19]. Exploratory pathway analyses using structural equation modelling (SEM) were conducted to examine whether inflammatory markers could statistically account for the link between the TyG index and in-hospital malignant arrhythmias. Exploratory single-pathway models were first constructed for NE and WBC to estimate the proportion of the association they explained. A multiple-pathway model was then fitted to evaluate their independent and combined contributions, considering potential overlap between these inflammatory measures. The statistical significance of these pathways was assessed using 5000 bootstrap resamples to obtain robust estimates. Additionally, the Systemic Inflammation Response Index (SIRI) was also evaluated for its potential to statistically explain this association. Subgroup analyses were conducted across key demographic, clinical, metabolic, lifestyle, and treatment-related categories, with interaction effects evaluated using the likelihood ratio test. Variance inflation factor (VIF) analysis revealed that values for all covariates were less than 5, suggesting the absence of substantial multicollinearity (Supplementary Table 1). Statistical computations were carried out via R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria), defining statistical significance as a two-tailed p-value 0.05.

Table 1 presents the baseline characteristics of the 655 participants according to TyG index quartiles. The cohort had a mean age of 75 years, with 57.56% being male. The average TyG index was 8.82 (8.42–9.33). Compared with the lowest quartile group, participants with the highest quartile group tended to be younger, female, obese, and non-smokers, and had a higher prevalence of hyperlipidemia and hypertension, along with greater use of beta-blockers during hospitalization. The biochemical parameters showed significant differences across the groups. Significantly increased DBP, HbA1c, LDL-C, CRP, WBC, monocyte count and NE were observed in the highest quartile. ROC curve analysis identified 9.01 as the optimal TyG index threshold for predicting in-hospital malignant arrhythmias (Supplementary Fig. 1). By using this cutoff value to stratify the TyG index, Supplementary Table 2 presents the baseline characteristics of the patients. The baseline characteristics of both the high TyG group and the low TyG group were largely consistent with the results of the quartile stratification.

Throughout the median follow-up period of 8 days (6.00, 14.00), 67 (10.23%) patients showed in-hospital malignant arrhythmias. Cox regression analysis (Table 2) showed that higher TyG index levels were significantly linked to an elevated risk of in-hospital malignant arrhythmias (HR = 1.57, 95% CI: 1.06–2.32). In the quartile analysis, patients in Q3 (HR = 2.87, 95% CI: 1.22–6.75) and Q4 (HR = 2.99, 95% CI: 1.26–7.13) had markedly higher risks of malignant arrhythmias compared with those in Q1.

The Kaplan-Meier (K-M) survival curves (Fig. 2A) showed that the incidence of malignant arrhythmias was markedly elevated in the Q4 group relative to Q1. Building on this gradient across TyG quartiles, Kaplan-Meier curves based on the ROC-derived exploratory cutoff (Supplementary Fig. 2) further demonstrated that patients in the high-TyG group experienced an elevated risk of in-hospital malignant arrhythmias compared to those with lower TyG levels. This exploratory cutoff serves as a preliminary reference for identifying high-risk participants, and requires further external validation before clinical application. In addition, the RCS model revealed a clear linear increase in the risk of in-hospital malignant arrhythmias with rising TyG levels (Fig. 2B).

Fig. 2.

The relationship between TyG and in-hospital malignant arrhythmias. (A) Kaplan-Meier survival curves for in-hospital malignant arrhythmias according to TyG quartiles. (B) Restricted cubic spline regression analysis illustrating the association between the TyG index and in-hospital malignant arrhythmias.

The exploratory pathway analysis demonstrated a significant relationship between the TyG index and the risk of in-hospital malignant arrhythmias, with an overall association of $\beta$ = 1.11 (95% CI: 0.52–1.71). In the single-pathway models, both neutrophil count and white blood cell count statistically explained a significant proportion of this relationship. The pathway involving NE was $\beta$ = 0.26 (95% CI: 0.08–0.43), accounting for 30.2% of the overall association, whereas WBC showed a similar statistical pathway of $\beta$ = 0.26 (95% CI: 0.09–0.43), also explaining 30.2% of the overall association (Fig. 3). These findings indicate that inflammation, reflected by NE and WBC, partially accounts for the association between TyG and malignant arrhythmias.

Fig. 3.

Exploratory pathway analysis of the association between the TyG index and in-hospital malignant arrhythmias involving NE and WBC.

Considering the biological correlation and potential overlap between NE and WBC, a multiple-pathway model incorporating both markers was further constructed. The independent pathway associations of NE and WBC were diminished and no longer significant, while the combined pathway association remained significant (Supplementary Table 3). This pattern suggests partial redundancy between the two markers and supports inflammation as a shared underlying pathway.

To further characterize inflammatory mechanisms, we conducted additional analyses focusing on the composite inflammatory index SIRI. We found that SIRI exhibited a statistically significant pathway association ($\beta$ = 0.09, 95% CI: 0.00–0.20), accounting for 8.5% of the overall association, further supporting a potential role of systemic inflammation in this link (Supplementary Fig. 3).

Given the complex medication regimen of these patients, subgroup analyses were performed to verify the consistency of TyG’s predictive value across different therapeutic backgrounds. Subgroup analyses revealed that the relationship linking the TyG index and in-hospital malignant arrhythmias was consistent across all examined strata, including gender, BMI, smoke, alcohol consumption, hyperlipidemia, hypertension, ACEI/ARB therapy, lipid-lowering drugs, Beta-blocker use, antidiabetic medication, aldosterone antagonist use, and PCI status, with no significant interactions observed (Fig. 4; all p for interaction $>$ 0.05).

Fig. 4.

Subgroup analyses of TyG index and in-hospital malignant arrhythmia in elderly diabetic patients with ventricular aneurysm.

Despite its strengths, this study has several limitations. First, the retrospective design and the exclusion of cases with missing earlier records may have introduced a degree of selection bias. Second, the relatively short observation window precludes the evaluation of long-term prognostic implications. Third, the lack of temporal separation between the assessment of the TyG index and inflammatory markers restricts our exploratory pathway analysis to being hypothesis-generating, precluding any direct causal inferences. In addition, the cohort was limited to elderly patients with LVA and diabetes mellitus, which may constrain the generalizability of the results. Therefore, external validation in larger and more diverse populations is needed. Future prospective studies with longer follow-up periods will be essential to validate these findings and further delineate the temporal associations underlying these mechanisms.