Blood–Brain Barrier Permeability Dynamics and Mediation of Triglyceride–Glucose Index on Acute Ischemic Stroke Outcomes

Abstract

Background

The evolution and severity of blood–brain barrier (BBB) permeability in the core and penumbra over time in acute ischemic stroke remain poorly understood. Recent studies have found that elevated triglyceride–glucose (TyG) index is associated with higher stroke recurrence, functional deterioration, and death. We investigated BBB permeability dynamics in patients with acute ischemic stroke and the impact of the BBB disruption–mediated TyG index on outcomes.

Methods

This double‐centric retrospective study included patients with acute anterior large‐artery occlusion from January 2019 to December 2023. BBB permeability was measured as flow extraction product (FED) values in the largest core and penumbra slices using computed tomography perfusion. The TyG index was calculated from triglyceride and glucose levels. Poor outcome was defined as a 90‐day modified Rankin Scale score of 3 to 6. Linear correlation evaluated relationships between FED values, time, and TyG index. Logistic regression analyzed associations between FED, TyG index, and outcomes. Mediation analysis assessed BBB disruption’s role in the insulin resistance–outcome relationship.

Results

This study included 274 patients (median age, 67 years [interquartile range, 57–74]; 179 men [66%]). Penumbra FED values increased with time (r=0.143; P=0.022), and core FED values were higher than penumbra (P<0.01). A higher TyG index was associated with poor outcomes (odds ratio, 6.6 [95% CI, 1.0–41.7]; P=0.04) and higher core FED values (odds ratio, 4.3 [95% CI, 1.4–12.9]; P=0.01). The highest TyG tertile had 9.96 and 17.51 times higher odds of increased core FED values (P=0.004) and poor outcomes (P=0.044). Mediation analysis indicated higher core FED values fully mediated the insulin resistance–outcome relationship (P=0.041).

Conclusions

FED values were higher in the core and increased over time, contributing to worse outcomes. An elevated TyG index mediates worse outcomes through increased BBB permeability, highlighting BBB stabilization as a potential therapeutic target.

Nonstandard Abbreviations and Acronyms

AIS

acute ischemic stroke

BBB

blood–brain barrier

CTP

computed tomography perfusion

FED

flow extraction product

IR

insulin resistance

NIHSS

National Institutes of Health Stroke Scale

TyG

triglyceride–glucose

Clinical Perspective.

What Is New?

• This study delineates the temporal evolution of blood–brain barrier permeability in patients with acute ischemic stroke, establishing that insulin resistance exacerbates this disruption specifically within the ischemic core.

• It further identifies blood–brain barrier permeability as a critical mediator linking insulin resistance to adverse clinical outcomes, thereby revealing a promising therapeutic target for stroke management.

What Are the Clinical Implications?

• An elevated triglyceride–glucose index may predict adverse outcomes in patients with acute ischemic stroke by increasing blood–brain barrier permeability, underscoring the importance of early identification of insulin resistance and the need for targeted therapies to stabilize the blood–brain barrier, potentially improving the prognosis in these patients.

Acute large‐vessel occlusion of head and neck severely affect quality of life. One of the critical pathological events associated with prognosis of acute ischemic stroke (AIS) is the disruption of the blood–brain barrier (BBB), a highly selective barrier maintaining the brain’s microenvironment. ¹ Ischemia causes BBB dysfunction, ² exacerbated by peripheral immune cell infiltration, ³ leading to increased parenchymal injury, hemorrhage, ⁴ and edema. ⁴ , ⁵ Therefore, accurately determining the evolution and severity of BBB permeability are crucial for prognostic evaluation in patients with AIS.

Research on BBB disruption in patients with AIS primarily focuses on pathological mechanisms and imaging evaluations. Numerous studies use animal models and cell culture experiments to elucidate the physiological and molecular bases of BBB disruption. ² , ⁵ , ⁶ , ⁷ , ⁸ Additionally, imaging techniques like magnetic resonance imaging are widely used to assess evolution of BBB permeability ¹ , ⁹ , ¹⁰ , ¹¹ , ¹² and demonstrate correlations between BBB damage, cerebral edema, hemorrhagic transformation, and poor prognosis. Despite significant advances, several issues remain unresolved. First, differences between animal models and human disease limit clinical applicability. Second, while magnetic resonance imaging is widely used to provide quantitative data on BBB disruption, they still face the limitations of being time consuming and inconvenient in the clinical environment. However, BBB disruption in patients with AIS based on computed tomography perfusion (CTP) needs further study.

Insulin resistance (IR) is linked to adverse cardiovascular ¹³ , ¹⁴ and metabolic outcomes. ¹⁵ , ¹⁶ Therefore, identifying patients with IR aids early risk stratification and management. The triglyceride–glucose (TyG) index, which combines fasting triglyceride and glucose levels, has been proposed as a marker of IR. Currently, research on the TyG index primarily focuses on type 2 diabetes, ¹⁷ obesity, ¹⁸ and cardiovascular diseases. ¹⁹ Recent studies have found that an elevated TyG index is associated with higher stroke recurrence, ¹⁹ , ²⁰ functional deterioration, ²¹ , ²² and death. ²³ However, whether the underlying mechanisms involved in the association of IR with stroke outcomes have not been fully understood. IR exacerbated vascular inflammation and endothelial dysfunction, ²⁴ , ²⁵ , ²⁶ which were critical contributors to BBB disruption in AIS. Previous studies have shown that metabolic dysregulation, including hypertriglyceridemia ²⁷ and hyperglycemia, ²⁸ aggravates BBB permeability by triggering oxidative stress and inflammatory pathways. ²⁹ , ³⁰ Specifically, IR leads to the generation of reactive oxygen species and the activation of NADPH oxidase, which contribute to oxidative stress and endothelial damage. ³⁰ , ³¹ Inflammatory cytokines such as tumor necrosis factor‐α, interleukin‐1β, and interleukin‐6 are also elevated in the insulin‐resistant state, activating nuclear factor‐κB and Janus kinase signaling pathways that further compromise endothelial tight junction proteins such as occludin, claudin‐5, and zonula occludens‐1, crucial for BBB integrity. ²⁹ , ³² These mechanisms are known to increase BBB permeability, facilitating the entry of harmful substances into the brain and contributing to worsened stroke outcomes. These findings indicated that the relationship between IR and stroke outcomes may be mediated by increased BBB permeability. However, the TyG index’s relationship with BBB disruption and outcomes in patients with AIS is not well studied.

Therefore, we hypothesized that BBB permeability varied in core and penumbra with time onset, and the more severe degree of BBB permeability was commonly seen in core. BBB disruption could mediate the relation between IR and functional outcomes. Therefore, the purpose of the current study included (1) to investigate the dynamic changes of BBB permeability in patients following AIS with onset time, (2) to investigate the severity of BBB injury in different core and penumbra ischemic area, and (3) to explore the meditation effect of BBB permeability with TyG index on functional outcomes.

METHODS

Study Design

The data that support the findings of this study are available from the corresponding author upon reasonable request. This double‐centric retrospective study reviewed patients with AIS admitted between January 2019 and December 2023. Inclusion criteria were as follows: (1) aged >18 years; (2) stroke‐related computed tomography (CT) imaging on admission, including noncontrast CT, CT angiography (CTA), and CTP; (3) onset to CTP time ≤24 hours; (4) unilateral internal carotid artery or middle cerebral artery occlusion; and (5) received reperfusion therapy. Exclusion criteria were the following: (1) concurrent posterior circulation large‐vessel occlusion; (2) bilateral chronic large‐vessel occlusion; (3) incomplete clinical or imaging data; (4) poor‐quality CT perfusion images; and (5) other significant medical conditions. A total of 274 patients met these criteria. This retrospective study was approved by the Institutional Review Board of our center, with informed consent waived due to its retrospective nature.

Data Collection

Demographic and clinical variables documented included age, sex, vascular risk factors, time from symptom onset to CTP, the Alberta Stroke Program Early CT Score, vascular occlusion location, core and penumbra volume, and treatment type. Patients were categorized by onset time: T≤3 hours, T 3 to ＜6 hours, and T 6 to ≤24 hours. Laboratory tests included fasting plasma glucose, triglycerides, total cholesterol, and so on. The TyG index, calculated as Ln [triglycerides (mg/dL)×fasting plasma glucose (mg/dL)/2], assessed IR status, with a TyG index ≥8.4 indicating IR in Asian patients. ³³ Additional data included the National Institute of Health Stroke Scale (NIHSS) scores on admission and 90‐day modified Rankin Scale score. Patients were categorized into 2 groups on the basis of the median NIHSS score: 1 group with NIHSS ≤10 and the other with NIHSS >10. The 90‐day outcomes were assessed via follow‐up or telephone, with a modified Rankin Scale score of 0 to 2 considered a good outcome.

Imaging Protocol

Participants underwent multimodal CT imaging using multidetector CT scanners (Revolution CT, GE Healthcare, Milwaukee, WI), including noncontrast CT, multiphase CTA, and CTP before endovascular treatment. Noncontrast CT parameters were 120 kV, 380 mA, and 5‐mm slice thickness. CTP was performed 5 minutes after multiphase CTA with 140 mm z axis coverage; CTP parameters were 80 kV, 260 mA, and 5‐mm slice thickness. A total of 20 consecutive spiral acquisitions were conducted. Forty frames of the total brain were scanned, and the total scanning duration was 40 seconds. Participants remained on the table for 5 minutes before multiphase CTA. Multiphase CTA scans included arterial (arch to vertex), venous, and late venous (skull base to vertex) phases. Parameters were 100 kV; 445 mA; pitch factor, 0.992; rotation time, 0.28 milliseconds; field of view, 200 mm; and image matrix, 512×512. Sixty milliliters of intravenous contrast (370 mg/mL; iopamidol, Bracco Imaging, Milan, Italy) was followed by a 35‐mL saline flush at 5 mL/s. Dynamic monitoring started 8 seconds after injection, with bolus tracking in the descending aorta at 120 Hounsfield units, with subsequent phases acquired at 10‐ and 18‐second delays.

Image Analysis

F‐STROKE software (Naoxi Intelligent Technology, Shanghai, China) analyzed noncontrast CT and CTP data, measuring Alberta Stroke Program Early CT Score, cerebral blood volume, cerebral blood flow, mean transit time, and time to peak. The middle cerebral artery was the arterial input, and the superior sagittal sinus was the venous output. Time to maximum >6 seconds delineated the penumbra, while relative cerebral blood flow <30% of in the contralateral hemisphere defined the core. Patients were categorized by Alberta Stroke Program Early CT Score into <7 and ≥7 groups.

CT Neuro–Workflow on syngo.via software (Siemens Healthcare, Erlangen, Germany) delineated regions of interest in the largest core and penumbra slice, with mirror regions of interest in the contralateral hemisphere (Figure S1). Flow extraction product (FED) values were calculated using a deconvolution model, relying on arterial time–density curves. ³⁴ , ³⁵ Data S1 provides relevant information about the FED values. This model calculates the residual impulse response function, linking arterial and tissue time–density curves.

$$c\left(t\right)=\left(a\otimes \text{impulse response function}\right)\left(t\right).$$

Two neuroradiologists (W.F.Q. and Z.C.X.; mean experience, 8 years) independently performed postprocessing, blinded to study data.

Statistical Analysis

Continuous variables following a normal distribution were expressed as means±SDs and evaluated using t tests or ANOVA. Nonnormally distributed variables were described using medians and interquartile ranges (IQRs), assessed using the Mann–Whitney U or Kruskal‐Wallis test, with Bonferroni correction for multiple comparisons. Categorical variables were represented by frequencies and percentages, using the χ ² or Fisher’s exact test. Interobserver agreement was assessed using the intraclass correlation coefficient in the appendix. The receiver operating characteristic curve for core FED values was plotted calculating sensitivity and specificity in predicting outcomes.

Univariate binary logistic regression was conducted, and variables with P<0.05 were included in a multivariate model to assess the association between the TyG index and either FED values or outcomes. Outcomes were expressed as odds ratios (ORs) with 95% CI. Spearman’s rank correlation analyzed the relationship between the TyG index and FED value of the core and penumbra. Analyses were performed using SPSS version 25.0 (IBM, Armonk, NY).

Mediation analysis, using SPSSAU 24.0, assessed whether core FED values mediated the relationship between IR and patient outcome. A statistically significant difference was defined as P<0.05.

RESULTS

Participant Characteristics

A total of 414 patients with acute anterior circulation large‐vessel occlusion who underwent CTP were reviewed (Figure 1). Exclusions included posterior circulation stroke (n=46), bilateral chronic occlusion (n=3), incomplete clinical information or imaging data (n=7), poor image quality (n=11), and serious medical histories (n=73). Finally, 274 eligible patients (aged 67 [IQR, 57–74] years; 66% men) were analyzed. Demographic characteristics, stratified by time from symptom onset to CTP, are summarized in Tables 1 and 2. Patients were categorized by time from onset to CTP: T≤3 hours (n=64), T 3 to <6 hours (n=115), and T 6 to ≤24 hours (n=95). To evaluate the degree of reperfusion following endovascular therapy, the modified Thrombolysis in Cerebral Infarction score was used. In this study, endovascular treatment encompassed mechanical thrombectomy alone, bridging therapy (intravenous thrombolysis followed by thrombectomy), and angioplasty (including balloon dilation or stenting). Among the 90 patients who achieved modified Thrombolysis in Cerebral Infarction 2b‐3 reperfusion, 46 received mechanical thrombectomy alone, 34 received bridging therapy, and 10 underwent angioplasty. Reperfusion therapy methods differed significantly among the groups (P<0.001).

Figure 1. Flowchart of patient selection.

CT indicates computed tomography.

Table 1.

Patient Characteristics of Study Cohort Stratified by Time From Symptom Onset to CT Perfusion

Variable	All patients (N=274)	T ≤3 h (n=64)	T 3 to <6 h (n=115)	T 6 to ≤24 h (n=95)	P value
Age, y, median (IQR)	67 (57–74)	68 (57–72)	68 (59–76)	65 (57–70)	0.05
Male sex, n (%)	179 (65.3)	41 (64.1)	79 (68.6)	59 (62.1)	0.64
Vascular risk factors, n (%)
Current smoking	129 (47)	23 (36)	62 (54)	44 (46)	0.07
Current drinking	78 (28)	7 (11)	45 (39)	26 (27)	0.24
Hypertension	200 (73)	48 (75)	86 (75)	66 (69)	0.76
Hyperlipidemia	114 (42)	24 (38)	55 (48)	35 (37)	0.21
Diabetes	129 (47)	29 (45)	64 (56)	36 (38)	0.08
Coronary artery disease	46 (17)	6 (9)	25 (22)	15 (16)	0.85
Atrial fibrillation	58 (21)	13 (20)	29 (25)	16 (17)	0.38
Clinical assessment
Occlusion location, n (%)					0.56
ICA	68 (25)	16 (25)	33 (29)	19 (20)
ICA and MCA	36 (13)	10 (16)	12 (10)	14 (15)
MCA	170 (62)	38 (59)	70 (61)	62 (65)
Admission NIHSS scores, median (IQR)	10 (4–15)	11 (5–16)	11 (6–15)	8 (3–14)	0.06
Early CT changes, n (%)					<0.001
ASPECTS 0–7	100 (37)	30 (47)	50 (43)	20 (21)
ASPECTS >7	174 (64)	34 (53)	65 (57)	75 (79)
90‐day mRS	3 (1–4)	2 (1–4)	3 (1–4)	3 (1–4)	0.78
Functional outcomes, n (%)					0.53
Good (mRS score ≤2)	130 (47)	34 (53)	51 (44)	45 (47)
Poor (mRS score >2)	144 (53)	30 (47)	64 (56)	50 (53)
Reperfusion therapy, n (%)					<0.001
Intravenous thrombolysis	97 (35)	22 (34)	54 (47)	21 (22)
Bridging therapy	35 (13)	9 (14)	18 (16)	8 (8)
Mechanical thrombectomy alone	48 (18)	14 (22)	13 (11)	21 (22)
Recanalization, mTICI scores, 2b‐3, n (%)	90 (33)	21 (33)	35 (30)	34 (36)	0.29
Biological characteristics, median (IQR)
C‐reactive protein, mg/L	5.0 (1.5–15.1)	8.1 (4.7–16.1)	4.9 (1.2–12.2)	4.6 (1.4–27.9)	0.30
Hemoglobin A1c	5.9 (5.6–6.4)	5.8 (5.5–5.9)	5.8 (5.6–6.4)	6 (5.6–7.2)	0.34
Total cholesterol, mmol/L	4.3 (3.6–4.9)	4.0 (3.1–4.3)	4.5 (3.6–5.1)	4.2 (3.5–4.8)	0.33
HDL‐C, mmol/L	1.1 (0.9–1.3)	1.1 (0.9–1.4)	1.1 (0.9–1.4)	1.2 (1.0–1.3)	0.90
LDL‐C, mmol/L	2.8 (2.1–3.4)	2.5 (1.9–2.9)	2.8 (2.3–3.5)	2.8 (2.1–3.3)	0.24
VLDL‐C, mmol/L	0.23 (0.13–0.48)	0.15 (0.12–0.56)	0.21 (0.12–0.46)	0.33 (0.23–0.48)	0.58
Homocysteine, μmol/L	15.4 (11.3–19.8)	13.6 (10.8–18.9)	14.8 (12.1–21.6)	16.1 (11.5–18.9)	0.56
TyG index, mean±SD	8.8±0.5	8.6±0.6	8.9±0.5	8.8±0.5	0.22

The mTICI 2b‐3 group includes patients treated with thrombectomy alone, bridging therapy, and angioplasty. ASPECTS indicates Alberta Stroke Program Early CT Score; CT, computed tomography; HDL‐C, high‐density lipoprotein cholesterol; ICA, internal carotid artery; IQR, interquartile range; LDL‐C, low‐density lipoprotein cholesterol; MCA, middle cerebral artery; mRS, modified Rankin Scale; mTICI, modified Thrombolysis in Cerebral Infarction; NIHSS, National Institutes of Health Stroke Scale; TyG, triglyceride–glucose; and VLDL‐C, very‐low‐density lipoprotein cholesterol.

Table 2.

Radiologic Characteristics of Participants With AIS Stratified by Time From Symptom Onset to CT Perfusion

Variable	All patients (N=274)	T ≤3 h (n=64)	T 3 to <6 h (n=115)	T 6 to ≤24 h (n=95)	P value
FEDP, mL·100 g−1·min−1, median (IQR)	1.27 (0.76–1.98)	1.27 (0.55–1.54)	1.21 (0.78–1.85)	1.37 (0.76–2.59)	0.06
FEDpm, mL·100 g−1·min−1, median (IQR)	0.41 (0.18–0.74)	0.24 (0.12–0.52)	0.45 (0.20–0.77)	0.45 (0.20–0.84)	0.006
FEDpr, mL·100 g−1·min−1, median (IQR)	3.44 (1.71–6.78)	4.68 (2.15–6.92)	3.12 (1.57–6.96)	3.24 (1.64–6.38)	0.36
FEDc, mL·100 g−1·min−1, median (IQR)	2.24 (1.28–3.91)	1.89 (1.16–2.38)	2.58 (1.38–5.66)	2.23 (1.42–3.85)	0.01
FEDcm, mL·100 g−1·min−1, median (IQR)	0.38 (0.21–0.67)	0.34 (0.14–0.48)	0.43 (0.19–0.87)	0.35 (0.24–0.59)	0.23
FEDcr, mL·100 g−1·min−1, median (IQR)	6.21 (3.42–8.96)	5.14 (3.17–9.12)	6.73 (3.61–9.63)	5.61 (3.90–8.60)	0.46
Core volume, mL, median (IQR)	8.3 (0–39.6)	5.9 (0–50.1)	12.2 (0–55.7)	7.8 (0–24.1)	0.16
Penumbra volume, mL, median (IQR)	133.1 (62.7–209.8)	130.9 (60.4–191.6)	149.1 (69.7–246.5)	123.3 (56.4–184.9)	0.10
Ratio, median (IQR)	5.2 (2.2–14.9)	4.8 (2.4–16.1)	5 (2.1–11.3)	6.4 (2.5–15.9)	0.72
Collateral vessel density	0.009 (0.003–0.020)	0.008 (0.002–0.018)	0.009 (0.003–0.019)	0.012 (0.005–0.024)	0.14
Grouping of high and low FED values of the core, n (%)					0.02
Low FED values	109 (59)	34 (76)	41 (50)	34 (59)
High FED values	76 (41)	11 (24)	41 (50)	24 (41)
Grouping of high and low FED values of the penumbra, n (%)					0.53
Low FED values	132 (51)	32 (52)	60 (55)	40 (47)
High FED values	125 (49)	29 (48)	50 (45)	46 (54)

AIS indicates acute ischemic stroke; CT, computed tomography; FED, flow extraction product; FED_P, FED value of the penumbra; FED_pm, FED value of the mirror penumbra; FED_pr, relative FED value of the penumbra; FED_c, FED value of the core; FED_cm, FED value of the mirror core; FED_cr, relative FED value of the core; and IQR, interquartile range.

Changes of FED Values of the Core and Penumbra With Onset Time

FED values of the penumbra positively correlated with the longer onset time (r=0.143; P=0.022) and larger volume of penumbra (r=0.134; P=0.03). Conversely, no evidence indicated that the FED values of the core were related to the onset time (r=0.094; P=0.205) or the core volume (r=0.142; P=0.051) (Figure 2). In groups within 3 hours, FED values of the penumbra increased over time (r=0.372; P=0.003) (Figure 3A). However, no correlation was found between FED values and time in the T 3 to <6 hours (r=−0.088; P=0.36) and T 6 to ≤24 hours groups (r=−0.054; P=0.62). We found no evidence of a linear correlation between FED values of the core and any of the 3 time intervals (r=0.181; P=0.23; r=0.045; P=0.69; r=−0.150; P=0.26).

Figure 2. Heatmap of the correlations between the FED values in ischemic regions and other imaging and clinical variables.

An increase in time and the penumbra volume correlates with higher FED_P values. There is no linear correlation between FED_c and time, nor with the core volume. CRP indicates C‐reactive protein; FED, flow extraction product; FED_P, FED value of the penumbra; FED_pm, FED value of the mirror penumbra; FED_pr, relative FED value of the penumbra; FED_c, FED value of the core; FED_cm, FED value of the mirror core; FED_cr, relative FED value of the core; HbA_1c, hemoglobin A_1c; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; TyG, triglyceride–glucose; TC, total cholesterol; and VLDL‐C, very‐low‐density lipoprotein cholesterol.

Figure 3. Temporal evolution of the FED values in the ischemic core and penumbra.

A, Scatter plot illustrates the correlation between the time from symptom onset to CT perfusion (T≤3 hours) and FED values of the core and penumbra, respectively. In groups within 3 hours, FED values of the penumbra were positively correlated with time from symptom onset to CT perfusion, while FED values of the core were not correlated with the time. B, Line chart of FED values in the penumbra and core across 3 temporal groups. In the T 3 to <6 hours group, FED values of the core are significantly higher compared with the T ≤3 hours group. No significant differences in FED values of the penumbra are found across time intervals. Across all time groups, FED values are consistently higher in the core compared with the penumbra. CT indicates computed tomography; and FED, flow extraction product.

In the T 3 to <6 hours group, FED values of core were significantly higher than in the T ≤3 hours group (2.58 [IQR, 1.38–5.66] versus 1.89 [IQR, 1.16–2.38] mL·100 g⁻¹·min⁻¹; P=0.007). For T 6 to ≤24 hours, FED values of the core did not significantly differ from the T ≤3 hours (2.23 [IQR, 1.42–3.85] versus 1.89 [IQR, 1.16–2.38] mL·100 g⁻¹·min⁻¹; P=0.31) or T 3 to <6 hours groups (2.23 [IQR, 1.42–3.85] versus 2.58 [IQR, 1.38–5.66] mL·100 g⁻¹·min⁻¹; P=0.50) (Figure 3B and Figure 4). No significant differences in FED values of the penumbra were found across time intervals. Consistently, FED values of the core were higher than those of the penumbra across all time groups (P<0.01). Patients with poor outcomes had higher FED values of the core and penumbra than those with favorable outcomes (Table 2 and Table S1; Figure 5; Figure S2).

Figure 4. A 3‐dimensional ribbon chart demonstrates the variations in FED values within the core, classified by the time from symptom onset to CT perfusion scanning, along with changes in the volume of the core.

In the T 3 to <6 hours group, FED values of the core are significantly higher compared with those in the group with T ≤3 hours. For T 6 to ≤24 hours group, FED values do not show significant differences when compared with both the T ≤3 hours group and the T 3 to <6 hours group. CT indicates computed tomography; and FED, flow extraction product.

Figure 5. Comparative case illustration of computed tomography perfusion maps in patients with acute stroke with and without IR.

A, Cerebral blood volume, (B) tissue map, and (C) FED map in a 66‐year‐old woman with acute left middle cerebral artery occlusion with no IR (TyG index, 8.1). The 90‐day mRS for this patient was 2. The patient’s FED values are 3.28 mL·100 g⁻¹·min⁻¹ in the core and 0.82 mL·100 g⁻¹·min⁻¹ in the penumbra. D through F, images acquired with the same protocols in a 75‐year‐old woman with acute right middle cerebral artery occlusion with IR (TyG index, 8.86). The mRS score for this patient was 3. The patient’s FED values are 6.96 mL·100 g⁻¹·min⁻¹ in the core and 3.45 mL·100 g⁻¹·min⁻¹ in the penumbra. The FED map in the patient with IR (F) shows higher FED values at both core and penumbra as compared with the FED map in the patient with no IR (C). FED indicates flow extraction product; IR, insulin resistance; mRS, modified Rankin Scale; and TyG, triglyceride–glucose.

Association Between TyG Index and FED Values of the Core and Penumbra

A positive correlation was found between the TyG index and FED values of the core (r=0.352; P=0.003), but no correlation was found in the penumbra (Figure S3). Receiver operating characteristic analysis identified 2.535 as the optimal cutoff for FED values to predict adverse outcomes, with an area under the curve of 0.68, sensitivity of 53%, and specificity of 81% (Figure S4).

Univariable logistic regression analysis showed that higher TyG index values were associated with elevated FED values of the core (OR, 5.3 [95% CI, 1.8–15.8]; P=0.002). Multivariable logistic regression, adjusting for core volume and atrial fibrillation, confirmed the TyG index as an independent risk factor for increased core FED values (OR, 4.3 [95% CI, 1.4–12.9]; P=0.01) (Table 3). A significant trend was noted across tertiles of TyG index levels concerning the risk of augmented BBB permeability. Controlling for age, sex, atrial fibrillation, and core volume, the highest TyG tertile had significantly higher FED values of the core risk than the lowest (OR, 9.96 [95% CI, 2.09–47.58]; P=0.004) (Table S2).

Table 3.

Logistic Regression Models for Predicting High FED Values of the Core^*

Variable	Univariable	P value	Multivariable	P value
	OR (95% CI)		OR (95% CI)
TyG index†	5.3 (1.8–15.8)	0.002	4.3 (1.4–12.9)	0.01
Core volume	1.007 (1.002–1.011)	0.003	1.0 (0.99–1.01)	0.18
Atrial fibrillation	2.1 (1.1–4.1)	0.03	0.4 (0.1–1.4)	0.17

In univariable and multivariable analysis, only variables significantly associated were shown. FED indicates flow extraction product; OR, odds ratio; TyG, triglyceride‐glucose.

^*High FED values are >2.535 mL·100 g⁻¹·min⁻¹, which is the optimal cutoff value obtained from the receiver operating characteristic analysis to predict adverse outcomes.

^† TyG index is treated as a continuous variable and is dimensionless (unitless).

Association Between TyG Index, FED Values of the Core and Penumbra, and Clinical Outcomes

In the univariable analysis, the TyG index (OR, 4.7 [95% CI, 2.0–11.3]; P<0.001), FED values of the core (OR, 1.5 [95% CI, 1.2–1.8]; P<0.001), FED values of the penumbra (OR, 1.4 [95% CI, 1.2–1.8]; P=0.001), and NIHSS scores (OR, 1.2 [95% CI, 1.1–1.3]; P<0.001) were independent predictors for poor outcomes (all P<0.05). To assess potential nonlinear effects, we further modeled NIHSS scores as a continuous variable using restricted cubic splines (Figure S5). The spline analysis showed no evidence of nonlinearity (P for nonlinearity=0.241), supporting the use of NIHSS scores as a linear predictor of poor outcome. In the multivariable analysis, a higher TyG index increased the risk of poor outcomes (OR, 6.6 [95% CI, 1.0–41.7]; P=0.04) (Table 4). Patients with AIS without IR had a significantly higher rate of favorable outcomes (modified Rankin Scale score, 0–2) (73.9% vs 39.2%; P=0.007) (Figure 6A).

Table 4.

Logistic Regression Models for Predicting Poor Clinical Outcomes

Variable	Univariable		Multivariable
	OR (95% CI)	P value	OR (95% CI)	P value
TyG index*	4.7 (2.0–11.3)	<0.001	6.6 (1.0–41.7)	0.04
FED values of the core	1.5 (1.2–1.8)	<0.001	1.4 (0.9–2.4)	0.15
FED values of the penumbra	1.4 (1.2–1.8)	0.001	4.5 (1.1–18.7)	0.04
NIHSS scores	1.2 (1.1–1.3)	<0.001	1.6 (1.2–2.0)	<0.001

In univariable analysis, only variables significantly associated were shown. FED indicates flow extraction product; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; TyG, triglyceride–glucose.

^*TyG index is treated as a continuous variable and is dimensionless (unitless).

Figure 6. The IR, assessed by the TyG index, predicts 90‐day functional outcomes in acute ischemic stroke through mediation by FED values of the core.

A, Distribution of functional outcomes based on the TyG index. The figure presents the distribution of functional outcomes at 90 days, as defined by the mRS. Differences in the number of patients with good functional outcomes (green) between the groups are indicated with the dashed connecting lines. Patients with AIS without insulin resistance (TyG index <8.4) had a significantly higher rate of favorable outcomes (mRS scores 0–2). B, The Sankey diagram illustrates the correspondence between TyG index (surrogate marker for IR), high–low FED values of the core, and clinical outcomes. Compared with patients with TyG index <8.4, patients with insulin resistance (TyG index ≥8.4) tended to have higher FED values, and patients with higher FED values tended to have poor clinical outcomes. Mediation analysis of the impact of IR on unfavorable functional outcomes at 90‐day mRS as mediated by the FED values of the core (C) and penumbra (D). In this model, a represents the path coefficient from X to M, b represents the path coefficient from M to Y, and c′ represents the direct path coefficient from X to Y after accounting for the mediator M. Associations are given as standardized regression coefficients β. The β value on the solid line indicates the total effect, while the product of the 2 β values on the dashed line represents the indirect effect. The impact of insulin resistance on prognosis is fully mediated by FED values of the core, whereas the FED values of the penumbra demonstrate no mediating effect. FED indicates flow extraction product; IR, insulin resistance; mRS, modified Rankin Scale; and TyG, triglyceride–glucose.

A significant trend was noted across tertiles of TyG index levels concerning the risk of unfavorable clinical outcomes. Controlling for FED values in the core and penumbra, as well as classifications based on the NIHSS scores, the highest TyG index tertile had significantly greater unfavorable outcomes risk than the lowest (OR, 17.51 [95% CI, 1.09–282.38]; P=0.044) (Table S3). We further conducted subgroup analyses stratified by treatment modality to assess the consistency of this association. Due to limited sample sizes, bridging therapy and direct thrombectomy cases were combined into a total endovascular thrombectomy group. Although statistical significance was observed only in the overall reperfusion group, the direction of the association remained consistent across subgroups, with most adjusted ORs exceeding 1, suggesting a persistent trend toward worse outcomes among patients with IR regardless of treatment type (Tables S4 and S5).

Mediation Analysis Between IR, FED Values of the Core, and Poor Outcomes at 90 Days

A Sankey diagram (Figure 6B) illustrated the relationships between the TyG index, FED values of the core, and clinical outcomes. Compared with patients with a TyG index <8.4, patients with IR (TyG index ≥8.4) had higher core FED values, and higher core FED values were associated with worse clinical outcomes.

Mediation analysis assessed whether the FED values of the core (Figure 6C) and penumbra (Figure 6D) mediated the association between IR status and clinical adverse outcomes in patients with AIS. IR significantly influenced adverse outcomes (β=0.241; P=0.046) and FED values of the core (β=0.248; P=0.04). However, the effect of IR on adverse outcomes was not significant (β=0.197; P=0.109), suggesting that FED values of the core fully mediated the association. The indirect effect was significant (indirect effect, 0.047 [95% bootstrap CI, 0.001–0.093]; P=0.041), not including 0, confirming the mediation, which accounted for 100% of the total effect (Table S6). In contrast, for the penumbra, the mediation analysis showed that the indirect effect was not significant, with the 95% bootstrap CI including 0 (indirect effect, 0.017 [95% bootstrap CI, −0.006 to 0.040]; P=0.153), indicating that the FED values of the penumbra did not mediate the association between IR status and adverse outcomes (Table S7).

DISCUSSION

This retrospective study demonstrated BBB permeability evolution with onset time in patients with AIS, its relationship with the TyG index, and their impact on functional prognosis. The main findings of this study were as follows: (1) FED values in the penumbra increased over time, especially within the first 3 hours; (2) BBB disruption was more severe in the core than in the penumbra, with core FED values peaking between 3 and 6 hours after onset, indicating the core’s higher vulnerability; and (3) elevated TyG index levels were associated with higher core FED values and worse 90‐day outcomes. Compared with the lowest tertile of the TyG index, the highest TyG tertile had 9.96 times higher odds of increased FED values and 17.51 times higher odds of poor outcomes. Mediation analysis suggested that higher core FED values significantly mediated the relationship between IR and poor outcomes, consistent with a full mediation pattern.

Our study found that FED values of the penumbra increased significantly with time, especially within the hyperacute phase (T ≤3 hours), exhibiting a time‐dependent pattern. In contrast, the FED values of the core showed an initial rise during the 3‐ to 6‐hour window, then declined gradually from 6 to 24 hours after onset, suggesting a biphasic pattern of BBB disruption with early damage and later repair. In the acute phase, ischemic events may cause endothelial cell injury and inflammation, increasing BBB permeability. ¹⁰ , ¹² As time progresses, BBB integrity may partially recover due to neoangiogenesis and blood flow redistribution. ³⁶ These findings align with reports from Abo‐Ramadan et al ³⁷ and Morgan et al ⁸ on early BBB permeability increases during AIS. However, the late‐phase decline in BBB permeability observed in our study contrasts with reports of sustained disruption. ⁸ These discrepancies could be attributed to variations in study design, imaging techniques, and patient characteristics. Moreover, factors such as individual genetic predispositions, comorbidities, and treatment strategies (eg, reperfusion therapy) may further influence the temporal dynamics of BBB permeability.

Notably, our findings supported the hypothesis that BBB disruption may originate from the core and subsequently extend to the penumbra, ³⁷ , ³⁸ indicating spatial and temporal heterogeneity during AIS. Thus, future studies should focus on capturing the dynamic changes of BBB permeability across different stages and regions of ischemia using advanced imaging modalities and molecular assessments. This approach may clarify the mechanisms of BBB disruption and offer a foundation for targeted therapies in acute stroke management.

Our findings indicated that increased FED values of the core significantly impacted clinical outcomes and mediated the relationship between IR, quantified by the TyG index, and prognosis. This supports previous studies showing that BBB disruption worsens stroke outcomes. ⁴ , ⁶ , ³⁹ A compromised BBB allows neurotoxic proteins to enter the brain, triggering inflammation, edema, neuronal injury, and nerve fiber disruption, leading to poor outcomes. Furthermore, our results were consistent with research indicating that metabolic dysfunctions, including hypertriglyceridemia ⁴⁰ and IR, ⁴¹ increase BBB permeability. Our study demonstrated that BBB disruption mediated the impact of IR on clinical prognosis, which was rarely studied, suggesting that increased TyG index exacerbated BBB permeability and then increased the risk of poor outcomes. These findings reveal that BBB disruption is not merely a passive pathological event in AIS but a critical mediator linking IR to clinical outcomes. These findings provide a foundation for developing therapeutic strategies targeting BBB stabilization or repair, which may improve long‐term outcomes in patients with both AIS and IR.

Stabilizing the BBB is a key therapeutic target in AIS. Phase‐specific strategies, such as anti‐inflammatory and antioxidant agents (eg, interleukin‐1ra, tumor necrosis factor‐α, NADPH oxidase inhibitors) in the acute phase ² , ³ , ²⁹ and reparative mediators (eg, transforming growth factor‐β, interleukin‐10, angiopoietin‐1) in the subacute phase, ⁴² help preserve endothelial integrity. Matrix metalloproteinase‐9 inhibition further reduces tight junction damage and hemorrhagic risk. ⁴² In later phases, cell‐ or gene‐based approaches (eg, mesenchymal stem cells, sirtuin 6) ⁶ and metabolic interventions (eg, intranasal insulin, mammalian target of rapamycin targeting) ⁴¹ offer added benefit. Timing is crucial, with most agents being most effective in the early subacute phase. ⁴³ These findings support the development of tailored, stage‐specific BBB‐targeted treatments to improve prognosis in patients with AIS, particularly those with metabolic dysfunction.

However, our study has limitations. First, the relatively small sample size from 2 centers may have introduced selection bias and limits the generalizability of the findings. Larger, multicenter studies are needed to confirm and extend our results. Second, the retrospective design and single–time point data collection limited our ability to evaluate the temporal dynamics of BBB permeability and IR. Longitudinal, multiple–time point studies are needed to clarify the evolving relationship and explore potential causal mechanisms. Third, although longer acquisition times (>60 seconds) are known to reduce measurement errors and improve result stability, they may also lead to motion artifacts, incomplete imaging, and increased radiation exposure. Our protocol used deconvolution‐based analysis with a 40‐second acquisition time, which is supported by prior validation studies. Fourth, patients with hemorrhagic incidents were excluded, and BBB permeability was assessed using only CTP, without other imaging methods or biomarkers. Future studies incorporate multimodal approaches, including magnetic resonance imaging and serum biomarkers, to further enhance the accuracy and robustness of BBB permeability assessments. Fifth, the inclusion of patients with varying reperfusion treatments (eg, intravenous thrombolysis, bridging therapy, or mechanical thrombectomy alone) introduced clinical heterogeneity. Finally, the mediation analysis suggesting a full mediating effect of BBB permeability on the relationship between IR and functional outcomes should be interpreted with caution. This result was derived from a relatively small, retrospective data set without mechanistic validation. The wide CIs further indicate limited precision. Moreover, underlying molecular pathways (eg, involving inflammation or metabolic regulation) linking the TyG index to BBB permeability were not explored. Thus, these findings should be viewed as preliminary and hypothesis generating, warranting validation in larger, prospective studies that explore underlying molecular mechanisms.

In conclusion, this study highlighted the dynamic changes in BBB permeability following AIS, revealing increased FED values in the penumbra over time and greater disruption in the core. Additionally, FED values of the core mediated the relationship between IR and functional outcomes, emphasizing the need for targeted therapies to improve patient prognosis.

Sources of Funding

This work was supported by the National Natural Science Foundation of China (No. 82171916), Tianjin Health Science and technology project (specific projects of key disciplines) (No. TJWJ2022XK019), Tianjin Key Medical Discipline (Specialty) Construction Project (No. TJYXZDXK‐041A), and the Youth Innovation Talent Training Program of Tianjin First Central Hospital.

Disclosures

None.

Supporting information

Data S1

Tables S1–S7

Figures S1–S5

References 44–48