Sage Journals: Discover world-class research

Abstract

Purpose

This study aimed to evaluate the influence of the triglyceride-glucose index (TyG index) on clinical outcomes and to develop nomogram models for predicting early neurological deterioration (END) and long-term prognosis in acute ischemic stroke (AIS) patients following intravenous thrombolytic (IVT) therapy.

Methods

We conducted a multi-center retrospective cohort study involving 333 AIS patients treated with IVT. The short-term and long-term outcomes were defined as the occurrence of END and 90-day prognosis. Multivariate logistic regression was used to develop nomogram models for forecasting these clinical outcomes.

Results

Patients in the high-TyG group exhibited significantly higher risks of END (P = 0.0010) and poor 90-day outcomes (P = 0.0012). Independent risk factors for END included a lower baseline NIHSS score, delayed door-to-needle time (DNT), reduced ASPECTS score, elevated TyG index, higher potassium (K+) levels, and incomplete Willis artery. Additionally, a higher initial NIHSS, increased TyG levels, presence of END, and a history of hypertension were predictors of poor prognosis. Based on the identified risk factors, two nomogram models yielded AUC values of 0.746 and 0.849 for predicting END and poor prognosis, respectively. NIHSS scores, TyG index, and admission glucose levels (Glu) emerged as prognostic indicators across all patients, while higher mean platelet volume (MPV) and history of stroke were identified as novel risk factors for poor prognosis in NO-END group.

Conclusion

A higher TyG index correlates with poor clinical outcomes in AIS patients post-IVT. The nomograms combining the TyG index with various factors enhanced risk prediction for END and poor prognosis.

Keywords

acute ischemic stroke (AIS)triglyceride-glucose (TyG) index early neurologic deterioration (END)prognosis outcome nomogram models

Backgrounds

Alteplase-based intravenous thrombolysis constitutes the benchmark for treating acute ischemic strokes.¹ The WAKE-UP and EXTEND trials, together with the implementation of emergency ‘green channel’ systems, have significantly expanded the accessibility of intravenous thrombolytic therapy (IVT), resulting in substantial enhancements in functional outcomes for patients with acute ischemic stroke (AIS).^2,3 Early neurological deterioration (END) is a prevalent complication, occurring in about 5% to 40% of AIS patients following IVT. This complication exacerbates neurological deficits, extends hospital stays, and is linked to poor long-term outcomes.^4–6 Therefore, there is an urgent need for valid markers and efficient tools that can forecast the onset of END and subsequent poor long-term prognosis.

The Triglyceride-Glucose (TyG) index serves as a convenient indicator of chronic insulin resistance (IR), being derived from readily available fasting measures of glucose (Glu) and triglycerides (TG).^7,8 Recent research has highlighted the TyG index as a significant risk factor for the incidence and progression of ischemic stroke.^9–12 Some studies have also suggested an association between an elevated TyG index and the development of END^13,14 and a poor long-term prognosis.^15–18

Although the TyG index has been extensively studied in AIS, its prognostic value in IVT-treated patients remains undetermined. This multicenter cohort study systematically evaluated the longitudinal relationship between TyG index and the risk of END, as well as poor prognosis in AIS patients receiving IVT. Additionally, we developed two comprehensive nomogram prediction models incorporating the TyG index and multi-dimensional variables to precisely estimate the occurrence of END and poor 90-day prognosis.

Materials and Methods

Study Patients and Treatment

A retrospective, multicenter investigation was conducted among all individuals who were admitted to the stroke units at Shanghai Tongji Hospital affiliated to Tongji University and The Fourth Affiliated Hospital affiliated to Guangzhou Medical University from January 2021 to June 2024. The investigation obtained ethical approval from the Ethics Review Committee of Tongji Hospital in Shanghai (No. K-2020 021) and the Ethical Review Board of the Fourth Affiliated Hospital of Guangzhou Medical University (No. 2024-H-022). In alignment with national regulations and institutional guidelines, no written informed consent was solicited from participants. The inclusion parameters for the study were: (1) individuals diagnosed with AIS and administered IVT with tissue plasminogen activator; (2) subjects 18 years of age or older; (3) patients having blood glucose and TG measurements within 24 h post-admission; (4) patients who underwent a pre-treatment computed tomography (CT) scan; (5) patients who received cerebral computed tomography angiography (CTA) or magnetic resonance angiography (MRA), as well as a head magnetic resonance imaging (MRI) during their hospital stay; and (6) patients with comprehensive hospitalization and follow-up data available. The exclusion criteria comprised: (1) patients not receiving IVT treatment due to time constraints or medical contraindications; (2) patients undergoing concurrent mechanical thrombectomy; (3) patients with preexisting disability or prior stroke-related functional impairment (Baseline mRS ≥ 2); (4) patients with images of insufficient quality for accurate evaluation; and (5) patients with incomplete data or those who were unavailable for follow-up. Ultimately, 333 cases satisfied the inclusion criteria. A flowchart depicting the patient selection process was provided in Figure 1.

Figure 1.

Flow Chart of Patient Selection. AIS = Acute Ischemic Stroke, IVT = Intravenous Thrombolysis, Glu = Glucose, TG = Triglycerides, CTA = Computed Tomography Angiography, MRA = Magnetic Resonance Angiography

Multi-Dimensional Dataset

We meticulously compiled a comprehensive dataset on patient characteristics. This included demographic information (including age and sex); previous medical history (eg, stroke, diabetes); clinical indictors such as National Institute of Health stroke scale (NIHSS),¹⁹ systolic and diastolic blood pressures; imaging biomarkers like the Alberta Stroke Program Early CT (ASPETS) Score to evaluate the size of infarction; and relevant laboratory markers (blood glucose levels, TG, total cholesterol, etc.).

The imaging assessment was conducted by a team of four neuroradiologists, each boasting at least 5 years of collective expertise. The ASPECTS represents a standardized ordinal scale that gauges the extent of early ischemic alterations on CT images, ranging from 0 (signifying complete infarction) to 10 (absence of infarction).²⁰ The Willis Artery status was determined as complete, partially complete, or incomplete via CTA or MRA.²¹ Additionally, the degree of white matter hyperintensities was quantitatively measured using the Fazekas scale, with scoring ranging from 0 to 6 points.²²

SPAN-100 Index and THRIVE Score

The SPAN-100 index, a composite score derived from the sum of age and the NIHSS score, serves as a quantitative indicator of stroke severity, with higher values signifying more pronounced stroke manifestations. A threshold of ≥100 signifies a positive SPAN-100 index, indicating a greater level of stroke severity.²³ The Health Risks in Vascular Events (THRIVE) scoring system, on the other hand, encapsulates a comprehensive assessment of AIS patients by incorporating five key factors: NIHSS score, age, hypertension, diabetes, and atrial fibrillation.²⁴ This scoring system is designed in such a way that higher scores correspond to a more adverse prognosis.

TyG Index

The TyG index was calculated as: Ln [TG (mg/dL) × Glu (mg/dL)÷2]. The fasting venous blood samples were drawn to measure TG and Glu within the first 24 h on admission.

Outcomes

The short-term outcome of this examination was the emergence of END, evidenced by a nominal NIHSS score escalation of at least two points during the initial 24-h period. The long-term outcome was the subject’s functional status at the 90-day mark (90d-mRS) with a unfavorable outcome denoted by a mRS score within the range of 3 to 6 according to established criteria.^1–3

Statistical Analysis

Participants were categorized into groups based on the presence or absence of END and poor 90-day prognosis. The NO-END group underwent a detailed subgroup analysis to identify prognostic risk factors. Categorical data were presented as frequencies and percentages, and differences across groups were evaluated via the chi-squared test. For continuous variables, medians along with interquartile ranges (IQRs) were reported, and the Mann-Whitney U test was employed for their analysis. Subgroup analysis, stratified by TyG index levels, employed the chi-square test to examine the discrepancies in the incidence of END and the likelihood of poor 90 day-prognosis between the Low-TyG and High-TyG groups. The Forward Stepwise Method of logistic regression and restricted cubic splines (RCS) was utilized to evaluate the association between the included factors and the risk of END as well as the likelihood of a poor prognosis. Nomogram models were constructed based on the outcomes of the multivariate logistic regressions. Receiver Operating Characteristic (ROC) analysis was conducted to assess the diagnostic efficacy, while Decision Curve Analysis (DCA) was employed to evaluate the clinical utility. Calibration curves were used to assess the accuracy of the models, and the Delong test was applied to compare the ROC curve performances. All statistical analyses were performed using R software version 4.0.3, with all P-values presented as two-tailed, and a P < 0.05 considered statistically significant.

Results

Baseline Characteristics in Patients within END and 90 Day-Prognosis Subgroups

Table1 illustrated the baseline features of the study population. Our statistical analysis identified various factors predictive of END, such as neutrophil counts, serum potassium levels (K+), NIHSS scores at admission, door-to-needle time (DNT), TyG index, TG, systolic blood pressure (SBP), baseline ASPECTS, atrial fibrillation status, and the integrity of Willis Artery. Additionally, certain variables were found to be indicative of 90-day prognosis, including age, NIHSS scores at admission, TyG index, admission glucose levels (Glu), baseline ASPECTS, and Fazekas score.

Table 1.

Baseline Characteristics for END and 90-Day Prognosis Groups

	No-END (n = 149)	END (n = 184)	P value	Good (n = 223)	Poor (n = 110)	P value
Demographics
Age (Median, IQR)	66 (58, 75)	65 (55, 76)	0.560	64 (55, 72)	68 (60, 78)	0.004*
Male (Percentile: %)	107 (32.1%)	131 (39.3%)	0.901	61 (18.3%)	34 (10.2%)	0.499
Medical history (percentile: %)
Coronary atherosclerosis	15 (4.5%)	18 (5.4%)	0.931	18 (5.4%)	11 (3.3%)	0.969
Atrial fibrillation	19 (5.7%)	11 (3.3%)	0.032*	15 (5%)	12 (3.6%)	0.395
Hypertension	93 (27.9%)	112 (33.6%)	0.773	122 (36.6%)	83 (24.9%)	<0.001*
Diabetes	28(9.4%)	36 (12.1%)	0.836	44 (13.2%)	28 (8.4%)	0.233
Stroke	30 (9%)	28 (8.4%)	0.239	33 (9.9%)	25 (7.5%)	0.073
Clinical variables (Median, IQR)
NIHSS on admission	9 (5, 15)	4 (3, 9.25)	<0.001*	5 (3, 9)	12 (8, 17)	<0.001*
DNT (min)	52 (40, 73)	60 (46.75, 85.75)	0.005*	58 (43, 80)	60 (45, 84)	0.689
Systolic BP (mm Hg)	150 (135, 166)	156 (140, 170)	0.012*	154 (140, 168)	155.5 (142, 169.75)	0.271
Diastolic BP (mm Hg)	86 (78, 98)	90 (80, 100)	0.139	88 (80, 100)	88 (80, 96)	0.849
Imaging variables (Median, IQR/Percentile: %)
ASPECTS	9 (8, 10)	8 (7, 10)	0.009*	8 (7, 10)	9 (8, 10)	0.055*
Willi’s artery			0.005*			0.083
Complete	24 (7.3%)	13 (3.9%)		29 (8.8%)	8 (2.4%)
Partial complete	59 (17.8%)	60 (18.1%)		85 (25.7%)	34 (10.3%)
Incomplete	66 (19.9%)	109 (32.9%)		109 (32.9%)	66 (22.1%)
Fazekas (Median, IQR)						<0.001*
	2 (1, 3)	2 (1, 3)	0.995	2 (1, 3)	3 (2, 3)
Laboratory values (Median, IQR)
Fibrinogen (FIB: g/L)	2.89 (2.42, 3.45)	2.93 (2.45, 3.52)	0.768	2.90 (2.45, 3.45)	3.09 (2.40, 3.77)	0.269
PT (sec)	11.2 (10.6, 12)	11.1 (10.6, 12)	0.632	11.1 (10.55, 11.8)	11.25 (10.63, 12.10)	0.260
APTT (sec)	26.6 (24.1, 29.6)	25.5 (23.5, 27.9)	0.038*	26.2 (23.8, 28.6)	25.6 (23.5, 27.9)	0.166
TT (sec)	17.7 (16.6, 18.5)	17.5 (16.6, 18.62)	0.699	17.5 (16.6, 18.6)	17.65 (16.4, 18.5)	0.807
INR	0.96 (0.91, 1.01)	0.96 (0.92, 1.00)	0.910	0.95 (0.91, 1)	0.97 (0.92, 1.03)	0.164
MPV (fL)	9.80 (9.15, 10.75)	9.80 (9.15, 10.30)	0.393	9.70 (9.12, 10.40)	9.90 (9.17, 0.62)	0.129
Neutrophils (%)	69.5 (57.4, 75.7)	70.25 (62, 79.25)	0.039*	69.4 (59.85, 76.15)	71.9 (60.4, 79.9)	0.138
Lymphocytes (%)	21.9 (16.1, 32.7)	20.55 (14.85, 28.52)	0.088	21.5 (16.3, 30)	19.5 (13.77, 29.47)	0.248
PDW (fL)	12.15 (10.3, 15.9)	10.7 (9.65, 12.5)	<0.001*	10.9 (9.7, 13.1)	11.55 (10.3, 15.32)	0.033*
MPV (fL)	9.8 (9.18, 10.7)	9.80 (9.15, 10.30)	0.754	9.7 (9.2, 10.4)	9.9 (9.12, 10.67)	0.122
NLR	3.12 (1.79, 4.79)	3.40 (2.16, 5.28)	0.104	3.30 (2.01, 4.59)	3.68 (2.06, 5.79)	0.211
K+ (mmol/L)	3.84 (3.64, 4.12)	3.98 (3.69, 4.25)	0.018*	3.92 (3.69, 4.2)	3.87 (3.59, 4.21)	0.459
Na+ (mmol/L)	138.5 (136.9, 140.2)	139 (136.97,140.93)	0.263	138.8 (137, 140.5)	138.85(136.2141.78)	0.556
Cl- (mmol/L)	102.9 (101, 105.4)	102.55(100.07,104.7)	0.144	102.7(100.75,104.8)	102.75(100.12,105.9)	0.653
Glu (mmol/L)	5.39 (4.58, 6.99)	5.84 (4.97, 7.92)	0.048*	5.29 (4.62, 6.68)	6.79 (5.22, 8.96)	<0.001*
HbA1c (mmol/L)	6.2 (5.8, 6.8)	6.05 (5.8, 7.22)	0.363	6.1 (5.75, 7.2)	6.3 (5.9, 6.95)	0.125
TG (mmol/L)	1.23 (0.9, 1.57)	1.34 (0.98, 1.87)	0.018	1.28 (0.93, 1.76)	1.33 (0.96, 1.77)	0.960
CHOL (mmol/L)	4.96 (4.29, 5.71)	5.12 (4.34, 5.73)	0.322	5.01 (4.29, 5.72)	5.16 (4.44, 5.76)	0.339
LDL (mmol/L)	2.9 (2.35, 3.31)	2.94 (2.48, 3.39)	0.397	2.84 (2.37, 3.38)	2.96 (2.66, 3.36)	0.143
HDL (mmol/L)	1.12 (0.97, 1.33)	1.11 (0.95, 1.24)	0.301	1.1 (0.95, 1.28)	1.13 (0.99, 1.29)	0.177
UA (μmol/L)	353 (286, 425)	348 (280.25, 420.5)	0.619	359 (286, 432)	347.5 (275, 408.02)	0.201
TyG index	7.05 (6.64, 7.46)	7.21 (6.86, 7.64)	0.002*	7.09(6.71, 7.51)	7.29 (6.86, 7.70)	0.014*
Classic scoring tools (Median, IQR/Percentile: %)
SPAN-100	75 (67, 86)	71 (61, 83)	0.017*	70 (61, 80)	81 (70, 92)	<0.001*
SPAN-100 negative	139 (42.2%)	172 (52.3%)	0.660	2 (1, 3)	3 (2, 4.25)	<0.001*
THRIVE	2 (2, 4)	2 (1, 3)	0.013*	2 (1, 3)	3 (2, 4.25)	<0.001*
END (percentile: %)	/	/	/	37(12.4%)	71(23.8%)	0.018*

IQR-interquartile range; NIHSS-National Institute of Health stroke scale; ASPECTS-Alberta Stroke Program Early CT Score; PT-Prothrombin Time; APTT-Activated Partial Thromboplastin Time; TT-Thrombin time; INR-International Normalized Ratio; PDW-Platelet Distribution Width; MPV-Mean Platelet Volume; NLR-Neutrophil-to-lymphocyte Ratio; K+-Potassium; Na+-Sodium; Cl-Chloride; Glu-Glucose on admission; HbA1c-Glycated hemoglobin; TG-Triglycerides; CHOL-Total cholesterol; LDL-Low-density lipoprotein; HDL-High-density lipoprotein; UA-Uric acid; TyG-Triglyceride-glucose index; END-Early neurological deterioration. THRIVE-The Health Risks in Vascular Events. For continuous variables, the Mann Whitney U test was used. For categorical variables, chi-square test was used. *Statistical significance: P < 0.05.

Subgroups Analysis Based on TyG Level for END and Poor 90 Day-Prognosis

In the absence of consensus cutoff values for the TyG index in stroke populations, patients were stratified into two groups based on the median TyG index value: the Low-TyG group (≤7.12) and the High-TyG group (>7.12). Importantly, individuals in the High-TyG group were identified to be at a significantly higher risk of developing into END (P = 0.0010, see Figure 2A) or poor 90-day prognosis (P = 0.0012, see Figure 2B). Figure 2C depicted the grouping flow of END occurrence and 90-day prognosis in patients from different TyG subgroups, intuitively showing that AIS patients post IVT in the High-TyG group were more likely to develop END and exhibit unfavorable prognosis. Approximately 39.67% patients with END experienced a poor 90-day outcome, compared to 24.83% of those from NO-END group.

Figure 2.

Comparisons of the Occurrence of END (A) and Poor 90-Day Prognosis (B) Between the Low-TyG and High TyG Groups. Sankey Diagram Illustrating the Distribution and Interconnections Among TyG Index Groups, END Categories, and 90-Day Prognosis Categories (C). TyG = Triglyceride-Glucose Index; END = Early Neurological Deterioration; NO-END = Without Early Neurological Deterioration; Good = Good 90-Day Prognosis; Poor = Poor 90-Day Prognosis

Early Neurological Deterioration

Predictive Model for END Prediction (Nomogram _END Model)

Due to the significant multicollinearity observed between triglycerides (TG) and the TyG index (Pearson r = 0.704, P < 0.001), TG was excluded from the subsequent multivariate logistic regression analysis. The analysis revealed several independent predictors of END, as outlined below: the NIHSS score at admission (OR 0.914 [95% CI 0.879-0.950], P < 0.001), DNT (OR 1.008 [95% CI 1.001-1.015], P = 0.035), ASPECTS score (OR 0.812 [95% CI 0.682-0.966], P = 0.019), Willis Artery integrity (OR 1.658 [95% CI 1.145-2.399], P = 0.007), K+ (OR 1.714 [95% CI 1.053-2.791], P = 0.03), and TyG index (OR 1.850 [95% CI 1.233-2.755], P = 0.003). The results of this analysis were presented in Figure 3A, and these factors were then integrated into the Nomogram _END Model (Figure 3B).

Figure 3.

Multivariable Logistic Regression Analyses With Respective Odds Diagrams Assessing Independent Factors of END (A). Nomogram _END Model for Predicting the Probability of END at 24 h (B). Calibration Plots of the Nomogram (C). The AUCs of Willis, DNT, TyG, K+, NIHSS Score, ASPECTS and Nomogram Model to Predict END for Patients With AIS Following IVT (D). Decision Curve Analysis Evaluating the Independent Predictors and the Nomogram Model (E). DNT-Door-To-Needle Time; TyG-Triglyceride-Glucose Index; K+-Serum Potassium Level; NIHSS-National Institute of Health Stroke Scale; ASPECTS-The Alberta Stroke Program Early CT Score; IVT-Intravenous Thrombolytic Therapy

To determine whether the Nomogram _END model could improve the differentiation between the NO-END and END subgroups, the AUC for the Nomogram _END and individual predictors was evaluated. Nomogram _END demonstrated superior performance, achieving an AUC of 0.746 (95% CI 0.693-0.800), with a sensitivity of 0.735 (95% CI 0.664-0.798) and a specificity of 0.674 (95% CI 0.597-0.748) (Figure 3C). The DeLong’s test revealed a statistically significant difference in the AUC between the Nomogram _END Model and the individual predictors (Table 2). The calibration plot (Figure 3D) confirmed the accuracy of the Nomogram _END Model, reporting a C-Statistic of 0.746 (95% CI 0.692-0.801). Additionally, the DCA curve suggested that the END _Model provided greater benefits than the individual predictors across specific threshold probabilities (Figure 3E).

Table 2.

Results of DeLong’s Test for ROC Curves Between the Nomogram _END Model and Independent Predictors for END

	z statistic	P value
NIHSS vs Nomogram _END	2.134	0.0328
TyG vs Nomogram _END	4.360	P < 0.0001
DNT vs Nomogram _END	4.514	P < 0.0001
ASPECTS vs Nomogram _END	3.923	0.0001
Willis artery vs Nomogram _END	4.587	P < 0.0001

NIHSS-National Institute of Health stroke scale; TyG-Triglyceride-glucose index; DNT-Door-to-needle time; ASPECTS-Alberta Stroke Program Early CT Score.

Potential Nonlinear Analysis for Prognostic Factors of END

The analysis employed restricted cubic spline models (RCS) to elucidate the possible nonlinear correlations between the occurrence of END and pertinent indicators. Our findings indicated a nonlinear declination in the risk of END for patients, particularly when NIHSS surpassed 5, as demonstrated in Figure 4A (P < 0.001). A significant nonlinear reduction in the likelihood of END was also observed when the ASPECTS was greater than 6 (P = 0.047), as depicted in Figure 4B. Furthermore, there was a linear increment in the risk of developing into END as the TyG index,K+ level and DNT increased.

Figure 4.

Cubic Regression Model Illustrating the Relationship Between the National Institute of Health Stroke Scale (NIHSS) Score at Admission and the Door-To-Needle Time (DNT) in Predicting the Occurrence of END

90-Day Prognosis

Predictive Model for 90 Day-Prognosis Prediction (Nomogram _Prognosis Model)

Significant multicollinearity was observed between the TyG index and its component variables - triglyceride (TG: Pearson r = 0.704, P < 0.001) and glucose (Glu: Pearson r = 0.549, P < 0.001). Therefore, both TG and Glu were excluded from the subsequent multivariate logistic regression analysis. The regression revealed several independent predictors of a poor prognosis, including the NIHSS score at the time of admission (OR 1.267 [95% CI 1.190-1.349], P < 0.001), TyG index (OR 1.752 [95% CI 1.093-2.808], P = 0.02), END (OR 0.156 [95% CI 0.075-0.325], P < 0.001), and a history of hypertension (OR 1.267 [95% CI 1.190-1.349], P < 0.001). The findings were presented in Figure 5A, and these significant factors were integrated into the Nomogram _Prognosis Model (Figure 5B).

Figure 5.

Multivariable Logistic Regression Analyses, Accompanied by Odds Ratio Plots, Examining the Independent Predictors of Poor 90-Day Prognosis (A). Nomogram _Prognosis Model Designed to Estimate the Probability of a Poor 90-Day Outcome (B). Calibration Curve of the Nomogram, Validating its Predictive Accuracy (C). The AUCs for the NIHSS, Hypertension, END, TyG and the Nomogram Model in Predicting Poor Outcomes for Patients With Acute Ischemic Stroke (D). Decision Curve Analysis is also Presented to Evaluate the Clinical Utility of the Independent Predictors and the Nomogram Model (E). NIHSS = National Institute of Health Stroke Scale; END = Early Neurological Deterioration; TyG = Triglyceride-Glucose Index

The Nomogram _Prognosis Model exhibited excellent performance, highlighted by an AUC of 0.849 (95% CI 0.814-0.896), a sensitivity of 0.750 (95% CI 0.657-0.828), and a specificity of 0.737 (95% CI 0.674-0.794), as depicted in Figure 5C. The DeLong’s test indicated a statistically significant difference between the AUC of the Nomogram _Prognosis Model and that of the individual predictors, as shown in Table 3. Furthermore, the calibration curve (Figure 5D) and the DCA curve (Figure 5E) supported the robust diagnostic efficacy of the Nomogram _Prognosis Model, suggesting potential clinical utility. The C-Statistic of the calibration curve achieved 0.812 (95% CI 0.764-0.860), displaying its accuracy.

Table 3.

Results of DeLong’s Test for ROC Curves Between the Nomogram _Prognosis Model and Independent Predictors for 90-Day Prognosis

	z statistic	P value
NIHSS vs Nomogram _Prognosis	2.739	0.0062
TyG vs Nomogram _Prognosis	7.157	P < 0.0001
END vs Nomogram _Prognosis	8.662	P < 0.0001
Hypertension vs Nomogram _Prognosis	8.109	P < 0.0001

NIHSS-National Institute of Health stroke scale; TyG-Triglyceride-glucose index; END-Early neurological deterioration.

Potential Nonlinear Analysis for Predictive Factors of 90-Day Prognosis

We further performed RCS analyses to assess the associations of NIHSS score and TyG index with poor 90-day outcomes in AIS patients. The analysis revealed a significant non-linear relationship between NIHSS score and poor outcome risk (P = 0.002; Figure 6A), with a sharp increase in risk observed when NIHSS scores exceeded 6. In contrast, the TyG index exhibited a linear positive association with poor outcome risk (P = 0.630), indicating that higher TyG values corresponded to progressively greater risks.

Figure 6.

(A)Figure Depicting a Cubic Regression Analysis Showing the Relationship Between the National Institute of Health Stroke Scale (NIHSS) Score at Admission and the 90-Day Prognosis Outcomes. (B) The AUCs for SPAN-100 Binary, SPAN-100, THRIVE and Nomogram _Prognosis Model, Evaluating Their Performance in Identifying Poor Prognostic Risks for Patients With Acute Ischemic Stroke

Comparisons with the Prior Classic Scoring Tools

To rigorously evaluate the prognostic performance of our newly developed Nomogram _Prognosis model, we established reference 90-day poor outcome models based on the established SPAN-100 and THRIVE scoring systems within our AIS study population. The distribution of patients’ SPAN-100 and THRIVE scores was presented in Table 1. We developed three distinct models: one utilizing a binary variable of the SPAN-100 (SPAN-100 Binary), another employing the continuous variable version of the SPAN-100 (SPAN-100), and the third using the THRIVE score (THRIVE). In predicting the 90-day poor prognosis, the SPAN-100 Binary yielded an AUC of 0.563 (95% CI, 0.529-0.596); the SPAN-100 achieved an AUC of 0.698 (95% CI, 0.637-0.760); THRIVE followed closely with an AUC of 0.747 (95% CI, 0.692-0.802); and the Nomogram _Prognosis model demonstrated the highest prognostic accuracy with an AUC of 0.849 (95% CI, 0.807-0.892) (Figure 6B). The Delong test confirmed that there were statistically significant differences in the performance of the various ROC curves (P < 0.05), thus displaying the added value of our Nomogram _Prognosis model in prognostic evaluation.

90d-Prognostic Subgroup Analysis in NO-END Group

Our study further investigated prognostic factors for poor 90-day outcomes in patients who did not experience END. Univariate analysis revealed significant associations (all P < 0.05) for mean platelet volume (MPV), NIHSS scores, TyG index, glucose levels (Glu), and stroke history (Table 4). In the overall cohort of AIS patients post IVT, NIHSS scores, TyG index, and baseline Glu levels consistently predicted poor outcomes. Notably, in the NO-END subgroup, elevated MPV and prior stroke history were newly identified as independent risk factors for unfavorable prognosis.

Table 4.

Results of Univariable Regression Analysis for Predictors of Poor Prognosis in Patients From NO-END Group

	Odds Ratio (95% CI)	P value
Demographics
Age	0.978 (0.947-1.010)	0.184
Male	0.653 (0.294-1.448)	0.294
Medical history
Coronary atherosclerosis	0.456 (0.150-1.381)	0.165
Atrial fibrillation	0.685 (0.240-1.955)	0.480
Hypertension	2.202 (0.950-5.103)	0.066
Diabetes	0.519 (0.221-1.222)	0.133
Stroke	3.216 (1.364-7.585)	0.008*
Clinical variables
NIHSS on admission	0.862 (0.807-0.920)	<0.001*
DNT (min)	1.011 (0.997-1.025)	0.130
Systolic BP (mm Hg)	1.004 (0.987-1.021)	0.652
Diastolic BP (mm Hg)	1.027 (0.998-1.057)	0.070
Imaging variables
ASPECTS	0.918 (0.712-1.182)	0.507
Willi’s artery
Complete	Reference
Partial complete	1.244 (0.421-3.675)	0.693
Incomplete	0.808 (0.270-2.418)	0.703
Fazekas	0.869 (0.635-1.190)	0.382
Laboratory values
Fibrinogen (FIB: g/L)	0.987 (0.939-1.037)	0.602
PT (sec)	0.876 (0.666-1.154)	0.347
APTT (sec)	1.006 (0.978-1.035)	0.687
TT (sec)	0.990 (0.898-1.091)	0.842
INR	0.087 (0.002-3.808)	0.205
MPV (fL)	9.90 (9.17, 0.62)	0.129
Neutrophils (%)	1.006 (0.978-1.034)	0.684
Lymphocytes (%)	0.999 (0.967-1.033)	0.975
PDW (fL)	0.996 (0.970-1.023)	0.774
MPV (fL)	0.709 (0.503-0.999)	0.050*
NLR	1.000 (0.908-1.101)	0.998
K+ (mmol/L)	1.222 (0.690-2.165)	0.491
Na+ (mmol/L)	0.946 (0.844-1.061)	0.342
Cl- (mmol/L)	0.971 (0.900-1.046)	0.435
Glu (mmol/L)	6.79 (5.22, 8.96)	<0.001*
HbA1c (mmol/L)	1.012 (0.939-1.092)	0.748
TG (mmol/L)	0.702 (0.437-1.127)	0.143
CHOL (mmol/L)	0.896 (0.645-1.244)	0.512
LDL (mmol/L)	0.726 (0.457-1.152)	0.174
HDL (mmol/L)	1.314 (0.417-4.139)	0.641
UA (μmol/L)	0.999 (0.996-1.002)	0.549
TyG	0.352 (0.181-0.688)	0.002*

NIHSS-National Institute of Health stroke scale; ASPECTS-Alberta Stroke Program Early CT Score; PT-Prothrombin Time; APTT-Activated Partial Thromboplastin Time; TT-Thrombin time; INR-International Normalized Ratio; PDW-Platelet Distribution Width; MPV-Mean Platelet Volume; NLR-Neutrophil-to-lymphocyte Ratio; K+-Potassium; Na+-Sodium; Cl-Chloride; Glu-Glucose on admission; HbA1c-Glycated hemoglobin; TG-Triglycerides; CHOL-Total cholesterol; LDL-Low-density lipoprotein; HDL-High-density lipoprotein; UA-Uric acid; TyG-Triglyceride-glucose index. *Statistical significance: P < 0.05.

Discussions

The triglyceride-glucose (TyG) index has emerged as an innovative biomarker for insulin resistance (IR) assessment. Growing evidence supports its prognostic value in AIS, demonstrating consistent associations with stroke recurrence, neurological deterioration, poor functional outcomes, and all-cause mortality.^13–18 A prospective observational study¹⁴ established the TyG index as an independent risk factor for both cervicocerebral atherosclerosis progression and post-ischemic major adverse cardiovascular events. Additional studies have shown that elevated TyG index values correlate with increased in-hospital mortality in severe stroke cases¹⁵ and higher rates of early stroke recurrence.¹⁶

Our dataset revealed that patients with a high TyG index are at greater risk of experiencing END and exhibit a higher risk of poor 90-day prognosis for AIS post IVT therapy. Baixiang Zhang and colleagues demonstrated that increasing TyG index was associated with a higher risk of END and a lower probability of early neurological improvement (ENI) in patients with AIS treated with intravenous thrombolysis.¹³ The study by Emma M. S. Toh et al further highlighted that TyG index was significantly associated with 90-day mortality (OR: 2.12, 95% CI: 1.39-3.23, P = 0.001), poor functional outcome (OR: 1.41, 95% CI: 1.05-1.90, P = 0.022), and negatively associated with ENI (OR: 0.68, 95% CI: 0.52-0.89, P = 0.004).¹⁴ Our research expanded upon previous findings by exploring the associations between the TyG index and END, as well as the 90-day outcomes. We furtherly enhanced the analysis by incorporating multi-dimensional imaging variables and laboratory metrics, and by creating predictive nomogram models within a well-defined patient cohort. In addition, the study conducted by Jia Wang and colleagues focused on a cohort of patients who were not administered intravenous thrombolysis, yet their findings aligned closely with our own.¹⁵ These findings indicated that TyG can be served as a potent predictive indicator for AIS patients across various treatment modalities.

This study revealed an intriguing pattern: while diabetes history prevalence showed no significant differences between the END cohort and 90-day prognosis cohort, substantial differences emerged in TyG index values. We propose two explanations for this finding: First, real-time biochemical markers like the TyG index provide more accurate reflections of patients’ current metabolic status than historical medical records. Second, insulin resistance (IR), characterized by reduced sensitivity and responsiveness to insulin, has been established as a pivotal indicator of type 2 diabetes mellitus (T2DM). IR often occurring several years before the onset of the disease.²⁵ These results highlight the TyG index’s superior prognostic value over traditional diabetes history.

The TyG index serves as a reliable and accessible surrogate marker for IR in routine clinical practice, the potential mechanisms by which the IR influences the onset of END and 90-day prognosis are multifaceted: Firstly, IR may elevate the risk of END and poor prognosis by accelerating atherosclerosis, interfering with platelet function, and exacerbating modifiable risk factors for AIS, such as hypertension, diabetes, and dyslipidemia.²⁶ Secondly, inflammation plays a pivotal role in the pathogenesis of brain injury associated with AIS. Research indicated that IR can stimulate the recruitment of monocytes and activate proinflammatory macrophages, leading to the production of monocyte chemoattractant protein. This interaction between inflammation and IR creates a vicious cycle that can adversely affect the occurrence of END and the prognosis of AIS.²⁷

Our study also identified several other risk factors that are closely associated with the development of END, including NIHSS score, ASPECTS score, DNT, potassium levels, and the integrity of the Willis artery. A comprehensive review by Benjamin D. has underscored the potential significance of potassium intake in both the etiology of stroke and its management, as indicated by a convergence of evidence from animal research, epidemiological studies, and clinical trials.²⁸ The ASPECTS score functions as a semi-quantitative index of infarct size and emerges as a pivotal prognostic marker for AIS.²⁹ In consonance with this, Vynckier et al have reported that a lower NIHSS score at hospital admission (aOR 0.81, P = 0.006), the presence of capsular warning syndrome (aOR 7.00, P < 0.001), ventral pontine infarction (aOR 3.49, P = 0.008), and imaging-confirmed hypoperfusion lesions (aOR 2.13, P = 0.026) are predictive of END in individuals with lacunar stroke.³⁰ Nonetheless, the aforementioned study merely stratified risk factors without developing a predictive model. Drawing on this, we conducted a comprehensive analysis of imaging, clinical, and laboratory parameters to construct an objective predictive model for END with an AUC of 0.746, which held practical utility in clinical settings.

Building on our END prediction, we further forecasted the 90-day prognosis. Additional predictors of the 90-day outcome include the baseline NIHSS score, blood glucose levels, the occurrence of END, and hypertension—factors that aligned with a broad spectrum of scientific literature.^31,32 We have developed another Nomogram _Prognosis model with the objective of accurately predicting the long-term functional outcomes for AIS patients treated with IVT. The nomogram model created by Yinglei Li and co-workers included analogous predictors (such as NIHSS, DNT, and baseline blood glucose) for 90-day prognosis and demonstrated a robust diagnostic performance, with an AUC of 0.822 (95% confidence interval: 0.769 to 0.875).³³ Ke-Jia Zhang and colleagues have devised a personalized nomogram model for forecasting poor functional outcomes, incorporating BNP, NIHSS scores, and baseline glucose levels, which yielded an AUC of 0.764.³⁴ Our study distinguished itself from earlier researches by integrating a more extensive range of variables, particularly the TyG index and neuroimaging data, leading to the predictive nomogram models showing superior performance with AUC of 0.849. Moreover, our results revealed that the Nomogram _Prognosis surpassed the performance of the SPAN-100 and THRIVE scores, which were established and validated as dependable prognostic indicators for AIS patients.

The subgroup analysis revealed that mean platelet volume (MPV) served as a prognostic indicator for 90-day outcomes in patients who have not experienced END. An elevated MPV can enhance platelet adhesion and aggregation, promoting atherosclerosis and increasing the likelihood of thrombotic events following plaque rupture, indicative of a more advanced pre-thrombotic condition.^35–37 Concurrently, higher MPV levels are linked to reduced bleeding times, which may diminish the therapeutic effectiveness of antiplatelet and thrombolytic treatments.^38,39 Therefore, for patients without END, it is crucial to closely monitor MPV levels and to identify any early signs of adverse prognostic developments.

Our study has several limitations. Firstly, the scope of our population sample may be limited, and our findings would benefit from broader validation. Secondly, the study did not comprehensively document the factors such as the pre-existing use of lipid-lowering medications or the presence of complications. The omission of these potential confounders from our statistical analysis may have contributed to bias in our results. Thirdly, the retrospective design of our study inherently possessed certain methodological shortcomings.

Conclusion

The TyG index serves as a surrogate marker for IR and is prognostically linked to END and poor 90-day outcomes in patients with AIS treated with IVT. By integrating the TyG index with a panel of other complex variables within nomograms, the accuracy of predicting END and poor long-term prognosis in AIS patients post thrombolysis has been significantly enhanced. These predictive models were developed to offer evidence-based decision support, particularly for AIS patients presenting near the terminal time window for IVT or those with borderline contraindications.

Footnotes

ORCID iDs

Lai Wei

Zhihua Wu

Xiang Zhou

Zhifeng Liu

Xiaoyan Wu

Kangwei Zhang

Peijun Wang

Haoyi Ye

Ethical Considerations

The research obtained ethical approval from the Ethics Review Committee of Tongji Hospital in Shanghai (No. K-2020 021) and the Ethical Review Board of the Fourth Affiliated Hospital of Guangzhou Medical University (No. 2024-H-022).

Author Contributions

Lai Wei, study concept and design, analysis and interpretation of data, drafting the manuscript. Zhihua Wu, acquisition and interpretation of data, manuscript revision. Xiang Zhou, acquisition and interpretation of data, manuscript revision. Zhifeng Liu, acquisition and interpretation of data, manuscript revision. Xiaoyan Wu, acquisition and interpretation of data, manuscript revision. Kangwei Zhang, methodology, analysis and interpretation of data, manuscript revision, critical review of manuscript for intellectual content. Peijun Wang, analysis and interpretation of data, manuscript revision, funding acquisition, critical review of manuscript for intellectual content. Haoyi Ye, study concept and design, interpretation of data, funding acquisition, critical review of manuscript for intellectual content. All authors read and approved the final manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Major project of the National Natural Scientific Foundation of China (Grant No. 82227807), the Research Project of Shanghai Municipal Health Commission (Grant No. 2022JC017), the National Key R&D Program of the Ministry of Science and Technology (Grant No. 2022YFC2009904), the Guangzhou Municipal Science and Technology Project (Grant No. 2025A03J4017) and the Research Project of Guangdong Health Information Net Association (Grant No. HX-202407-0002).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.*

References

Campbell

BCV

Ringleb

Parsons

Churilov

, et al. Extending thrombolysis to 4·5-9 h and wake-up stroke using perfusion imaging: a systematic review and meta-analysis of individual patient data. Lancet. 2020;395(10241):1906.

Thomalla

Simonsen

Boutitie

, et al. MRI-guided thrombolysis for stroke with unknown time of onset. N Engl J Med. 2018;379(7):611-622.

Campbell

BCV

Parsons

, et al. Thrombolysis guided by perfusion imaging up to 9 hours after onset of stroke. N Engl J Med. 2021;384(13):1278.

Bonkhoff

Hong

Bretzner

, et al. Association of stroke lesion pattern and white matter hyperintensity burden with stroke severity and outcome. Neurology. 2022;99(13):e1364-e1379.

Wei

Pan

Deng

, et al. Predicting long-term outcomes for acute ischemic stroke using multi-model MRI radiomics and clinical variables. Front Med. 2024;11:1328073.

AYX

Austin

Rashid

, et al. Sex differences in intensity of care and outcomes after acute ischemic stroke across the age continuum. Neurology. 2023;100(2):e163-e171.

Hoshino

Mizuno

Ishizuka

, et al. Triglyceride-glucose index as a prognostic marker after ischemic stroke or transient ischemic attack: a prospective observational study. Cardiovasc Diabetol. 2022;21(1):264.

Kim

Ahn

Kang

Nam

Kim

Park

. Relationship between the triglyceride glucose index and coronary artery calcification in Korean adults. Cardiovasc Diabetol. 2017;16:108.

Jiang

, et al. Correlation of TyG-BMI and TyG-WC with severity and short-term outcome in new-onset acute ischemic stroke. Front Endocrinol. 2024;15:1327903.

10.

Guo

Liu

, et al. Association between triglyceride-glucose index and 1-year recurrent stroke after acute ischemic stroke: results from the Xi’an stroke registry study of China. Cerebrovasc Dis. 2024;53(4):391-402.

11.

Zhang

Liu

Ruan

, et al. Triglyceride-glucose index linked to hospital mortality in critically ill stroke: an observational multicentre study on eICU database. Front Med. 2020;7:591036.

12.

Wang

Tang

Wang

, et al. Association of triglyceride-glucose index with early neurological deterioration events in patients with acute ischemic stroke. Diabetol Metab Syndr. 2023;15(1):112.

13.

Toh

EMS

Lim

AYL

Ming

, et al. Association of triglyceride-glucose index with clinical outcomes in patients with acute ischemic stroke receiving intravenous thrombolysis. Sci Rep. 2022;12(1):1596.

14.

Zhang

Lei

Ambler

, et al. Association between triglyceride-glucose index and early neurological outcomes after thrombolysis in patients with acute ischemic stroke. JCM. 2023;12(10):3471.

15.

Wang

Tang

Tian

Xie

. Non-insulin-based insulin resistance indices predict early neurological deterioration in elderly and middle-aged acute ischemic stroke patients in Northeast China. Sci Rep. 2024;14(1):16138.

16.

Nam

Kwon

Lee

. High triglyceride-glucose index is associated with early recurrent ischemic lesion in acute ischemic stroke. Sci Rep. 2021;11(1):15335.

17.

Liu

Zuo

. Association between insulin resistance markers and poor prognosis in patients with acute ischemic stroke after intravenous thrombolysis. Neurol. 2024;29(4):218-224.

18.

Sun

Shen

, et al. Influence of TyG index on large vascular occlusive stroke following endovascular treatment. CNS Neurosci Ther. 2024;30(12):e70143.

19.

Thompson

Luo

Gardiner

Burke

Nickles

Reeves

. Quantifying selection bias in National Institute of Health stroke scale data documented in an acute stroke registry. Circ Cardiovasc Qual Outcomes. 2016;9(3):286-293.

20.

Puetz

Dzialowski

Hill

Demchuk

. The Alberta Stroke Program Early CT Score in clinical practice: what have we learned? Int J Stroke. 2009;4(5):354-364.

21.

Hartkamp

van Der

Grond J

van Everdingen

, et al. Circle of Willis collateral flow investigated by magnetic resonance angiography. Stroke. 1999;30(12):2671-2678.

22.

Jacob

Verhoeven

, et al. Baseline and longitudinal MRI markers associated with 16-year mortality in patients with cerebral small vessel disease. Neurology. 2024;103(6):e209701.

23.

Ovbiagele

Reeves

Nasiri

Johnston

Bath

Saposnik

VISTA-Acute Collaboration Steering Committee . A simple risk index and thrombolytic treatment response in acute ischemic stroke. JAMA Neurol. 2014;71(7):848-854.

24.

Flint

Faigeles

Cullen

, et al. THRIVE score predicts ischemic stroke outcomes and thrombolytic hemorrhage risk in VISTA. Stroke. 2013;44(12):3365-3369.

25.

Tao

Wang

Hua

. Triglyceride-glucose index as a marker in cardiovascular diseases: landscape and limitations. Cardiovasc Diabetol. 2022;21(1):68. Published 2022 May 6.

26.

Chen

Wang

Pan

. Causal associations of insulin resistance with coronary artery disease and ischemic stroke: a Mendelian randomization analysis. BMJ Open Diabetes Res Care. 2020;8(1):e001217.

27.

Jayaraj

Azimullah

Beiram

Jalal

Rosenberg

. Neuroinflammation: friend and foe for ischemic stroke. J Neuroinflammation. 2019;16:142.

28.

Hunt

Cappuccio

. Potassium intake and stroke risk: a review of the evidence and practical considerations for achieving a minimum target. Stroke. 2014;45(5):1519-1522.

29.

Barber

Demchuk

Zhang

Buchan

. Validity and reliability of a quantitative computed tomography score in predicting outcome of hyperacute stroke before thrombolytic therapy. ASPECTS study group. Alberta stroke programme early CT score. Lancet. 2000;355(9216):1670-1674.

30.

Vynckier

Maamari

Grunder

, et al. Early neurologic deterioration in lacunar stroke: clinical and imaging predictors and association with long-term outcome. Neurology. 2021;97(14):e1437-e1446.

31.

IST-3 Collaborative Group . Association between brain imaging signs, early and late outcomes, and response to intravenous alteplase after acute ischaemic stroke in the third international stroke trial (IST-3): secondary analysis of a randomised controlled trial. Lancet Neurol. 2015;14(5):485-496.

32.

Dai

, et al. Predictors of early neurologic deterioration in acute pontine infarction. Stroke. 2020;51(2):637-640.

33.

Qie

. Developing a nomogram model for 3-month prognosis in patients who had an acute ischaemic stroke after intravenous thrombolysis: a multifactor logistic regression model approach. BMJ Open. 2024;14(7):e079428.

34.

Zhang

Jin

Zhang

Guo

Yang

. N-terminal pro-brain natriuretic peptide is associated with hemorrhagic transformation and poor outcomes in patients with stroke treated with intravenous thrombolysis. Front Mol Neurosci. 2021;14:758915.

35.

Gawaz

Langer

May

. Platelets in inflammation and atherogenesis. J Clin Investig. 2005;115:3378-3384.

36.

Kamath

Blann

Lip

. Platelet activation: assessment and quantification. Eur Heart J. 2001;22:1561-1571.

37.

Chu

Becker

Berger

, et al. Mean platelet volume as a predictor of cardiovascular risk: a systematic review and meta-analysis. J Thromb Haemostasis. 2010;8:148-156.

38.

Maden

Kacmaz

Selcuk

, et al. Relationship of admission hematological indexes with myocardial reperfusion abnormalities in acute ST segment elevation myocardial infarction patients treated with primary percutaneous coronary interventions. Can J Cardiol. 2009;25:e164-e168.

39.

Dębiec

Pogoda-Wesołowska

Piasecki

Stępień

Staszewski

. Mean platelet volume as a potential marker of large vessel occlusion and predictor of outcome in acute ischemic stroke patients treated with reperfusion therapy. Life. 2021;11(6):469.

Nomogram Models Integrating TyG Index for Predicting Early Neurological Deterioration and 90-Day Outcomes in AIS Patients Undergoing IVT

Abstract

Purpose

Methods

Results

Conclusion

Keywords

Backgrounds

Materials and Methods

Study Patients and Treatment

Multi-Dimensional Dataset

SPAN-100 Index and THRIVE Score

TyG Index

Outcomes

Statistical Analysis

Results

Baseline Characteristics in Patients within END and 90 Day-Prognosis Subgroups

Subgroups Analysis Based on TyG Level for END and Poor 90 Day-Prognosis

Early Neurological Deterioration

Predictive Model for END Prediction (Nomogram END Model)

Potential Nonlinear Analysis for Prognostic Factors of END

90-Day Prognosis

Predictive Model for 90 Day-Prognosis Prediction (Nomogram Prognosis Model)

Potential Nonlinear Analysis for Predictive Factors of 90-Day Prognosis

Comparisons with the Prior Classic Scoring Tools

90d-Prognostic Subgroup Analysis in NO-END Group

Discussions

Conclusion

Footnotes

ORCID iDs

Ethical Considerations

Author Contributions

Funding

Declaration of Conflicting Interests

Data Availability Statement

References

Predictive Model for END Prediction (Nomogram _END Model)

Predictive Model for 90 Day-Prognosis Prediction (Nomogram _Prognosis Model)