Neutrophil Extracellular Traps Are Associated With Early Neurological Deterioration in Patients With Acute Ischemic Stroke Receiving Intravenous Thrombolysis: A Prospective Cohort Study

Abstract

Background

Neutrophil extracellular traps contribute to thrombolytic resistance and the no‐reflow phenomenon after acute ischemic stroke, thereby impairing microvascular reperfusion and leading to unfavorable clinical outcomes. This study investigated the association between prethrombolytic circulating neutrophil extracellular trap biomarkers, CitH3 (citrullinated histone 3) and MPO (myeloperoxidase)‐DNA complexes, and early neurological deterioration (END) in patients with acute ischemic stroke treated with intravenous thrombolysis alone.

Methods

We prospectively enrolled patients with acute ischemic stroke who received standard‐dose alteplase within 4.5 hours of symptom onset (January 2019–April 2023). Venous blood was drawn before intravenous thrombolysis to measure serum CitH3 and MPO‐DNA. The primary outcome was END, defined as an increase of ≥4 points in NIHSS score at 24 hours versus baseline. Logistic regression with restricted cubic splines assessed associations between neutrophil extracellular trap biomarkers and END.

Results

Among 287 patients (mean age 61.69±12.41 years, 70.4% male), 45 (15.7%) developed END. Serum CitH3 >17.67 ng/mL (odds ratio [OR], 3.81 [95% CI, 1.75–8.29]) and MPO‐DNA >79.72 ng/mL (OR, 5.58 [95% CI, 2.36–13.21]) were independently associated with END. The associations remained significant when treated as continuous variables. Restricted cubic splines confirmed a linear dose–response relationship between higher CitH3 levels and END risk (P for nonlinearity=0.475). Subgroup analyses revealed stronger associations for CitH3 in patients with coronary artery disease or stroke history (P for interaction=0.018) and D‐dimer <0.50 μg/mL (P for interaction=0.013). A significant interaction of smoking was found between MPO‐DNA and END (P for interaction=0.026).

Conclusions

Serum neutrophil extracellular traps may serve as promising biomarkers associated with END after intravenous thrombolysis in patients with acute ischemic stroke.

Nonstandard Abbreviations and Acronyms

AIS

acute ischemic stroke

CitH3

citrullinated histone 3

END

early neurological deterioration

IVT

intravenous thrombolysis

MPO‐DNA

myeloperoxidase‐DNA

NIHSS

National Institutes of Health Stroke Scale

OTT

onset‐to‐treatment time

Clinical Perspective.

What Is New?

• In this prospective real‐world cohort of patients with acute ischemic stroke treated exclusively with intravenous thrombolysis, higher prethrombolytic serum levels of neutrophil extracellular trap biomarkers, CitH3 (citrullinated histone 3), and MPO (myeloperoxidase)‐DNA, were independently associated with early neurological deterioration within 24 hours after treatment.

• This study provided quantitative evidence linking neutrophil extracellular traps to the risk of early neurological deterioration, revealing a linear dose–response relationship for CitH3 and a threshold‐like nonlinear pattern for MPO‐DNA.

What Are the Clinical Implications?

• Baseline neutrophil extracellular trap levels measurement may help identify patients at higher risk of early neurological deterioration and highlight circulating neutrophil extracellular traps as potential therapeutic targets to improve microvascular reperfusion and clinical outcomes following intravenous thrombolysis.

As the global burden of stroke continues to escalate, it has emerged as the third leading cause of mortality and the fourth foremost contributor to disability‐adjusted life years. ¹ For patients with acute ischemic stroke (AlS), accounting for 60% to 70% of all strokes, intravenous thrombolysis (IVT) remains a cornerstone for achieving reperfusion. ² However, despite its proven efficacy, early neurological deterioration (END) still occurs in a substantial proportion of patients, with an overall pooled incidence of ∼14% after reperfusion therapy as reported by a recent meta‐analysis. ³ Given that END significantly increases the risk of poor functional outcomes, identifying reliable biomarkers is critical for improving clinical management. ⁴

Cerebral ischemia induces oxidative stress, damage‐associated molecular patterns release, and platelet–neutrophil interactions, which rapidly initiate the process of NETosis. Over time, the progressive deposition of neutrophil extracellular traps (NETs) within evolving thrombi leads to the formation of chromatin networks composed of DNA, histones, and granular enzymes. These structures provide a scaffold for fibrin and platelets, amplifying thromboinflammation, aggravating endothelial injury, and increasing blood–brain barrier permeability. ⁵ As the thrombus matures, dense NETs structures reduce fibrinolytic permeability and promote resistance to tissue plasminogen activator. ⁶ Excessive NET accumulation impairs microvascular reperfusion, thereby contributing to the no‐reflow phenomenon, while directly disrupting blood–brain barrier integrity and promoting vasogenic edema and infarct expansion. ⁷ Taken together, these mechanisms exacerbate ischemic injury and ultimately limit the efficacy of thrombolytic therapy in AIS.

Clinically, histopathological studies of retrieved thrombi have revealed widespread NETs deposition, with NETs detected in nearly all thrombi and higher NETs content correlating with greater stroke severity. ⁸ , ⁹ , ¹⁰ Circulating NETs levels were significantly elevated in patients with AIS compared with healthy subjects and independently associated with adverse clinical outcomes. ¹¹ , ¹² In a cohort study, baseline plasma NETs levels were higher in patients with AIS without early improvement after tissue plasminogen activator than in those with National Institutes of Health Stroke Scale (NIHSS) score reduction at 24 hours. ¹³ Similarly, MRCLEAN‐MED (Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands: The Effect of Periprocedural Medication) trial ¹⁴ reported that increased baseline plasma citrullinated histone 3 (CitH3) levels were associated with severe NIHSS score at 24 hours after endovascular thrombectomy in patients treated with endovascular thrombectomy alone. However, evidence regarding postreperfusion neurological improvement remains limited, as most existing studies did simply descriptive comparisons and the sample size was small. Experimental and ex vivo studies have shown that NET inhibition (eg, DNase I) significantly reduces NET burden and improves outcomes in mouse stroke models or cerebral thrombi, whereas human data remain scarce. ¹³ , ¹⁵ Collectively, these findings suggest that NETs may serve as both potential biomarkers and therapeutic targets in AIS, especially among patients treated with IVT.

Based on prior mechanistic and clinical evidence, this study aims to further elucidate the associations of circulating NETs biomarkers, CitH3 and MPO (myeloperoxidase)‐DNA complexes, with a standardized END end point in a prospective, real‐world IVT‐only cohort with AIS.

METHODS

Transparency and Openness Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request. Individual‐level data cannot be shared publicly due to patient privacy and ethical restrictions. The analytical methods and statistical code used in this study are available upon request from qualified researchers for purposes of reproducing the results.

Study Design and Patient Enrollment

This cohort study was based on the Ischemic Cerebrovascular Disease Database of the First Affiliated Hospital of Zhengzhou University, as previously reported. ¹⁶ , ¹⁷ The current study prospectively enrolled consecutive patients with AIS receiving IVT from 2 branch hospitals of the First Affiliated Hospital of Zhengzhou University, the Heyi Main Hospital and the Zhengdong Branch Hospital, between January 2019 and April 2023. Both branch hospitals shared standardized protocols, staffing practices, electronic health record systems, and research infrastructure. AIS was diagnosed based on the World Health Organization criteria. ¹⁸

The inclusion criteria were (1) age ≥18 years old, and (2) patients with AIS who receive standard‐dose alteplase within 4.5 hours of clearly defined symptom onset.

The exclusion criteria included (1) failure to collect blood samples before IVT; (2) discharge within 24 hours; (3) endovascular treatment; (4) premature termination of IVT due to severe adverse effects; (5) incomplete or missing clinical data; (6) comorbidities of severe systemic illnesses, such as malignant brain tumor, autoimmune diseases, severe hepatic or renal dysfunction, etc; and (7) active severe inflammatory or infectious diseases.

This study was approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University (2024‐KY‐2285‐001), and written informed consents were obtained from all participants or their designated proxies.

Data Collection

Clinical and demographic data (age, sex, vascular risk factors, medical history, NIHSS score, etc) were collected by well‐trained neurologists using a standardized case report form at admission. Laboratory tests, imaging features, and treatment information were obtained before discharge from the hospital's electronic medical records. Onset‐to‐treatment (OTT) time was the interval between the onset of stroke symptoms and the initiation of IVT. Stroke cause was classified using the TOAST (Trial of ORG 10172 in Acute Stroke Treatment) criteria, as large‐artery atherosclerosis, cardioembolism, small‐vessel occlusion, stroke of other determined cause, and stroke of undetermined cause. ¹⁹ Tests included prethrombolysis white blood cell count, red blood cell count, platelet count, neutrophil count, and lymphocyte count, D‐dimer, glycated hemoglobin (HbA1c), total cholesterol, triglycerides, high‐density lipoprotein, low‐density lipoprotein, etc. All laboratory tests were completed in the central laboratory of the hospital.

Measurement of NET Biomarkers

Blood samples were collected before IVT and unified stored in the Biobank of The First Affiliated Hospital of Zhengzhou University and National Human Genetic Resources Sharing Service Platform (Grant No. 2005DKA21300). Among NET‐related markers identified in peripheral blood, CitH3 and MPO‐DNA are considered more specific for NETs. ¹⁵ Both were measured at the Henan Key Laboratory of Cerebrovascular Diseases in this study for the purpose of evaluating systemic NETs burden after AIS. All assays were conducted in strict accordance with the manufacturer's instructions. To ensure optimal sample stability, processing was performed promptly with serum separated via centrifugation at 4000× g for 4 minutes within 60 minutes of collection. To quantify NET biomarkers, we measured serum levels of CitH3 and MPO‐DNA using ELISA. Serum CitH3 levels were measured using the Citrullinated Histone H3 (Clone 11D3) ELISA kit (501 620, Cayman Chemical, Ann Arbor, MI, USA). Serum MPO‐DNA levels were quantified using the Myeloperoxidase‐DNA Complex ELISA research kit (JM‐1227H1, Jiangsu Jingmei Biological Technology Co., Ltd., China). The laboratory personnel were blinded to the clinical data.

Magnetic Resonance Imaging Protocol

All patients underwent 3.0 T magnetic resonance imaging scans (MAGNETOM Vida, Siemens Healthineers, Erlangen, Germany) immediately after thrombolysis. Sequences included T1WI, T2WI, diffusion‐weighted images, apparent diffusion coefficient and fluid‐attenuated inversion recovery. Diffusion‐weighted imaging was acquired using a readout‐segmented spin‐echo echo planar imaging sequence (b=0 and 1000 s/mm², 20 axial slices, 5 mm thickness, 1.0 mm gap, voxel size 0.8×0.8×5.0 mm).

Measurement of Infarct Core Volume

Acute infarcts were defined as hyperintense lesions on diffusion‐weighted images with corresponding hypointensity on apparent diffusion coefficient maps. Lesions were manually segmented in 3D Slicer (Version 5.8.1; https://www.slicer.org/) by 2 independent neurologists blinded to outcomes. Segmentation was conducted with special attention to exclude artifacts, tumor, or edema. For cases involving multiple vascular territories, lesions were segmented separately. Infarct core volume (mL) was automatically calculated using the Segment Statistics module of 3D Slicer. Twenty cases were randomly selected to test interrater reliability by using intraclass correlation coefficients, which yielded an intraclass correlation coefficient of 0.959 (95% CI, 0.900–0.984, P<0.001).

Assessment of END

The assessments of NIHSS score ²⁰ were systematically conducted at admission, during IVT, and once per hour within 24 hours after IVT by qualified neurologists. We defined END as an increase of ≥4 points in NIHSS score at 24 hours from baseline NIHSS, which was commonly used in large‐scale studies. ²¹

Statistical Analysis

The distribution of continuous variables was assessed using the Shapiro–Wilk test in combination with graphical inspection. Continuous variables with an approximately normal distribution were presented as mean±SD, and Student's t test was employed for comparisons. Variables with a nonnormal distribution were presented as median with interquartile range (M [Q25–Q75]), and the Mann–Whitney U test was used for comparisons. Categorical variables were summarized as number (percentage) and compared using the chi‐square test or Fisher's exact test, as appropriate.

Multivariable logistic regression analyses were performed to evaluate the associations of serum CitH3 and MPO‐DNA with END. Results were represented as odds ratios (ORs) with 95% CIs. Candidate covariates were prespecified a priori based on prior literature and clinical relevance. Model specification was guided by information criteria to achieve a parsimonious adjustment set. Because infarct core volume exhibited a highly right‐skewed distribution, a log‐transformed value (ln [infarct core volume+1]) was used in all adjusted models.

In the primary analyses, serum CitH3 and MPO‐DNA levels were analyzed as categorical variables using optimal cutoff values determined by the Youden Index derived from receiver operating characteristic (ROC) curves. To further characterize associations on a continuous scale, both biomarkers were additionally modeled as standardized continuous variables (per 1‐SD increment).

Restricted cubic spline functions within logistic regression models were used to explore the potential nonlinear concentration response relationships between NETs biomarkers and END. Following Harrell's recommended quantile‐based knot placement, spline specifications with 3 to 5 knots were compared on the common complete‐case data set using Akaike information criterion; models within ΔAkaike information criterion ≤2 were considered equivalent, and the more parsimonious specification was preferred. The reference point for spline curves was set at the median biomarker concentration. Nonlinearity was evaluated using likelihood ratio tests comparing models with linear terms to those including spline terms.

Prespecified subgroup analyses were conducted to assess the consistency of associations across clinically relevant strata, including age, sex, body mass index, smoking status, hypertension, diabetes, hyperlipidemia, coronary artery disease (CAD) or stroke history, atrial fibrillation, NIHSS score on admission, HbA1c, D‐dimer, and OTT. CitH3 and MPO‐DNA were standardized (per SD) before constructing interaction terms. To preserve statistical power given the limited number of END events, all subgroup analyses were performed using a simplified adjustment set consisting of age, sex, NIHSS score on admission, HbA1c, and D‐dimer. The stratifying variable was not included as a covariate within its corresponding stratum. Effect modification was evaluated by adding multiplicative interaction terms between biomarkers and subgroup variables to the logistic regression models. Age was dichotomized as <60 and ≥60 years. Body mass index categories were defined according to the World Health Organization classification, as follows: <25 kg/m² and ≥25 kg/m². ²² Stroke severity on admission was categorized as minor (NIHSS score ≤3) and nonminor (NIHSS score >3), according to the AHA/ASA guideline. ²³ HbA1c categories followed the American Diabetes Association criteria (<5.7%, 5.7%–6.49%, and ≥6.50%). ²⁴ D‐dimer was categorized into <0.50 and ≥0.50 μg/mL. ²⁵ OTT was dichotomized at the cohort median (162 minutes) to ensure balanced strata.

Several sensitivity analyses were conducted to evaluate the robustness of the primary findings. First, to address potential selection bias due to missing NETs biomarker measurements, baseline characteristics of included and excluded patients were compared. In addition, inverse probability weighting was applied, in which the probability of inclusion in the complete‐case analysis was modeled using baseline covariates only. Second, additional multivariable logistic regression models were constructed with further adjustment for OTT, while treating NETs biomarkers as categorical and continuous variables, respectively. Third, we prespecified 2 nested logistic regression models for the END in the cohort. Fully adjusted model included all variables from the simplified model and additionally incorporated cause classification, OTT, and ln (infarct core volume+1). The two models were prespecified and compared using Akaike and Bayesian information criteria and likelihood ratio tests to confirm the stability of the selected model. Their performance was further evaluated by discrimination (area under the ROC with 95% CIs via DeLong's method) and calibration (Hosmer–Lemeshow test). Fourth, we repeated all subgroup analyses using the fully adjusted model, and we additionally conducted an cause‐based subgroup analysis. In strata with sparse data, estimates were marked as not estimable.

All statistical analyses were performed by the SPSS 26.0 software (SPSS, IBM, West Grove, PA, USA) and R4.5.2 software (R Foundation for Statistical Computing, Vienna, Austria). P<0.05 (2 sides) was considered statistically significant.

RESULTS

Baseline Characteristics

From January 2019 to April 2023, a total of 510 patients with AIS receiving alteplase treatment were registered in the database, of whom 331 had prethrombolysis serum samples collected. Following the exclusion of 44 patients, a total of 287 patients were ultimately included in the final analysis. Detailed information regarding the excluded patients can be found in the flow chart (Figure 1). Normality assessment for continuous variables is provided in Table S1. The average age of enrolled patients was 61.69±12.41 years, with 70.4% (N=202) being male. The median NIHSS score at admission was 5 (3–9) points. Regarding stroke cause, small‐vessel occlusion was the most common subtype, accounting for 39.0% of all patients, followed by large‐artery atherosclerosis (31.4%), cardioembolism (12.5%), and stroke of other determined cause/stroke of undetermined cause (17.1%). The distribution of infarct core volume was markedly right skewed, with a median of 1.44 (0.50–6.30) mL for the overall cohort. The serum concentrations of CitH3 and MPO‐DNA were 16.26 (12.72, 20.07) ng/mL and 60.03±22.61 ng/mL, respectively. The median of serum MPO‐DNA levels was 59.99 ng/mL. The median OTT was 162 (120–210) minutes (Table 1).

Figure 1. Flow chart of the study.

AIS indicates acute ischemic stroke; END, early neurological deterioration; IVT, intravenous thrombolysis; and rt‐PA, recombinant tissue plasminogen activator.

Table 1.

Baseline Characteristics of Patients With AIS With and Without END

Characteristics	All patients (N=287)	END		P value
		No (n=242)	Yes (n=45)
Age, y	61.69±12.41	60.87±11.88	66.11±14.28	0.024
Male sex, n (%)	202 (70.4)	169 (69.8)	33 (73.3)	0.637
BMI, kg/m2	24.85±3.34	24.74±3.36	25.43±3.18	0.207
Smoking, n (%)	103 (35.9)	84 (34.7)	19 (42.2)	0.335
Alcohol consumption, n (%)	80 (27.9)	70 (28.9)	10 (22.2)	0.357
Risk factors, n (%)
Hypertension	177 (61.7)	146 (60.3)	31 (68.9)	0.278
Diabetes	67 (23.3)	57 (23.6)	10 (22.2)	0.846
Hyperlipidemia	49 (17.1)	42 (17.4)	7 (15.6)	0.768
Coronary artery disease	41 (14.3)	35 (14.5)	6 (13.3)	0.842
Atrial fibrillation	22 (7.7)	16 (6.6)	6 (13.3)	0.129
Stroke history	65 (22.6)	57 (23.6)	8 (17.8)	0.395
Medical history, n (%)
Lipid‐lowering therapy	41 (14.3)	37 (15.3)	4 (8.9)	0.260
Antiplatelet therapy	39 (13.6)	35 (14.5)	4 (8.9)	0.316
Clinical assessment
SBP, mm Hg	147.15±20.11	146.90±20.61	148.44±17.29	0.638
DBP, mm Hg	85.94±13.53	85.80±13.47	86.67±13.97	0.694
National Institutes of Health Stroke Scale score on admission	5.00 (3.00–9.00)	5.00 (3.00–8.25)	5.00 (4.00–12.50)	0.046
Onset‐to‐treatment time, min	162 (120–210)	160 (120–213)	165 (120–210)	0.595
Laboratory data
Glycated hemoglobin, %	6.00 (5.60–7.20)	6.00 (5.60–6.81)	6.27 (5.80–8.60)	0.014
White blood cell count (×109/L)	7.19 (5.96–8.67)	7.08 (5.97–8.60)	7.80 (5.75–9.29)	0.389
Red blood cell count (×1012/L)	4.51 (4.18–4.87)	4.51 (4.17–4.89)	4.45 (4.26–4.73)	0.720
Platelet (×109/L)	211 (176–254)	213.50 (176–258.25)	199 (162–236.50)	0.139
Neutrophil (×109/L)	4.74 (3.75–6.56)	4.66 (3.80–6.35)	5.40 (3.61–7.13)	0.180
Lymphocyte (×109/L)	1.56 (1.14–2.06)	1.57 (1.20–2.10)	1.46 (0.92–1.88)	0.100
Total cholesterol, mmol/L	4.22±0.92	4.24±0.93	4.17±0.83	0.638
Triglycerides, mmol/L	1.25 (0.95–1.68)	1.27 (0.96–1.73)	1.07 (0.79–1.55)	0.088
High‐density lipoprotein, mmol/L	1.05 (0.90–1.24)	1.06 (0.92–1.21)	1.05 (0.90–1.26)	0.937
LDL, mmol/L	2.71±0.83	2.72±0.84	2.67±0.77	0.705
D‐dimer, μg/mL	0.13 (0.06–0.25)	0.12 (0.06–0.22)	0.18 (0.10–0.74)	0.001
Infarct core volume, mL	1.44 (0.50–6.30)	1.25 (0.48–5.42)	3.38 (0.93–20.06)	0.001
Cause classification				0.039
Large‐artery atherosclerosis	90 (31.4)	71 (29.3)	19 (42.2)
Cardioembolism	36 (12.5)	28 (11.6)	8 (17.8)
Small‐vessel occlusion	112 (39.0)	103 (42.6)	9 (20.0)
Stroke of other determined or undetermined cause	49 (17.1)	40 (16.5)	9 (20.0)
Citrullinated histone 3, ng/mL	16.26 (12.72–20.07)	15.71 (11.71–18.63)	20.96 (16.65–25.82)	<0.001
MPO‐DNA, ng/mL	60.03±22.61	56.13±20.69	80.96±21.16	<0.001

Normally distributed variables (eg, age, BMI, SBP, DBP, LDL and MPO‐DNA) are presented as mean±SD; nonnormally distributed continuous variables are presented as median with interquartile range, ie, M (Q25–Q75). AIS indicates acute ischemic stroke; BMI, body mass index; DBP, diastolic blood pressure; END, early neurological deterioration; LDL, low‐density lipoprotein; MPO‐DNA, myeloperoxidase‐DNA; and SBP, systolic blood pressure.

Serum CitH3/MPO‐DNA Levels and END

A total of 45 patients (15.7%) experienced END within 24 hours of IVT. Table 1 showed the baseline characteristics comparison of patients with and without END. Patients with END were much older, were more likely to have higher admission NIHSS scores, and had elevated levels of HbA1c and D‐dimer than those without (all P values <0.05). The END group also showed markedly greater infarct core volumes (3.38 versus 1.25 mL, P=0.001). The distribution of stroke cause differed significantly between the 2 groups (P=0.039) Notably, patients with END showed markedly higher concentrations of serum CitH3 (20.96 [16.65, 25.82] versus 15.71 [11.71, 18.63] ng/mL, P<0.001) and MPO‐DNA (80.96±21.16 versus 56.13±20.69 ng/mL, P<0.001), compared with those without (Table 1). The optimal cutoff values of serum CitH3 and MPO‐DNA levels for END were 17.67 and 79.72 ng/mL with Youden Indexes of 0.419 and 0.416 in ROC analyses, respectively.

In the primary analyses (Table 2), higher serum NETs biomarker levels were significantly associated with an increased risk of END. Patients with elevated CitH3 levels had a markedly higher odds of developing END compared with those below the optimal cutoff (OR, 3.81 [95% CI, 1.75–8.29]; P=0.001). Similarly, elevated MPO‐DNA concentrations were independently associated with END (OR, 5.58 [95% CI, 2.36–13.21]; P<0.001) after multivariable adjustment. When analyzed as continuous variables (Table 3), both biomarkers demonstrated consistent associations. Each 1‐SD increase in CitH3 (5.50 ng/mL) was associated with 2.64‐fold higher odds of END (95% CI, 1.69–4.13; P<0.001), and each 1‐SD increase in MPO‐DNA (22.61 ng/mL) corresponded to 3.53‐fold higher odds (95% CI, 2.05–6.09; P<0.001).

Table 2.

Multivariable Logistic Regression Analyses of the Associations Between Serum CitH3/MPO‐DNA Levels (as Binary Variables) and END in Patients With AIS Receiving IVT

Variables	Model 1		Model 2
	OR (95% CI)	P value	OR (95% CI)	P value
Age, y	1.03 (1.00–1.06)	0.103	1.04 (1.01–1.07)	0.023
Male sex	2.06 (0.85–4.98)	0.109	2.03 (0.81–5.10)	0.131
NIHSS score on admission	1.05 (0.98–1.12)	0.179	0.98 (0.91–1.06)	0.614
HbA1c, %	1.27 (1.03–1.57)	0.024	1.25 (1.01–1.55)	0.040
D‐dimer, μg/mL	2.03 (0.90–4.60)	0.089	2.14 (0.96–4.78)	0.064
ln (infarct core volume+1), mL	1.14 (0.86–1.52)	0.352	1.18 (0.89–1.57)	0.253
Cause classification		0.604		0.658
LAA	Ref.		Ref.
Cardioembolism	1.15 (0.40–3.34)		0.88 (0.30–2.59)
SVO	0.57 (0.22–1.47)		0.54 (0.21–1.42)
ODC or UDC	0.93 (0.35–2.48)		0.86 (0.32–2.34)
CitH3 >17.67 ng/mL	3.81 (1.75–8.29)	0.001	–	–
MPO‐DNA <79.72 ng/mL	–	–	5.58 (2.36–13.21)	<0.001

Model 1 indicated multivariable logistic regression analyses including age, sex, NIHSS score on admission, HbA1c, D‐dimer, ln (infarct core volume+1), cause classification and CitH3 >17.67 ng/mL. Model 2 indicated multivariable logistic regression analyses including age, sex, NIHSS score on admission, HbA1c, D‐dimer, ln (infarct core volume+1), cause classification, and MPO‐DNA >79.72 ng/mL. AIS indicates acute ischemic stroke; CitH3, citrullinated histone 3; END, early neurological deterioration; HbA1c, glycated hemoglobin; IVT, intravenous thrombolysis; LAA, large‐artery atherosclerosis; MPO‐DNA, myeloperoxidase‐DNA; NIHSS, National Institutes of Health Stroke Scale; ODC, stroke of other determined cause; OR, odds ratio; SVO, small‐vessel occlusion; and UDC, stroke of undetermined cause.

Table 3.

Multivariable Logistic Regression Analyses of the Associations Between Serum CitH3/MPO‐DNA Levels (as Continuous Variables) and END in Patients With AIS Receiving IVT

Variables	Model 1		Model 2
	OR (95% CI)	P value	OR (95% CI)	P value
Age, y	1.02 (0.99–1.05)	0.217	1.02 (0.99–1.06)	0.181
Male sex	2.24 (0.88–5.69)	0.090	1.88 (0.74–4.78)	0.183
NIHSS score on admission	1.02 (0.94–1.09)	0.691	0.96 (0.89–1.04)	0.287
HbA1c, %	1.20 (0.96–1.50)	0.114	1.13 (0.90–1.42)	0.305
D‐dimer, μg/mL	1.97 (0.89–4.37)	0.097	2.53 (1.13–5.64)	0.204
ln (infarct core volume+1), mL	1.14 (0.85–1.53)	0.385	1.11 (0.83–1.48)	0.494
Cause classification		0.682		0.650
LAA	Ref.		Ref.
Cardioembolism	1.16 (0.39–3.44)		1.05 (0.36–3.09)
SVO	0.61 (0.23–1.63)		0.56 (0.21–1.50)
ODC or UDC	1.02 (0.37–2.81)		0.98 (0.35–2.72)
Serum CitH3 level per SD (5.50 ng/mL) increment	2.64 (1.69–4.13)	<0.001	–	–
Serum MPO‐DNA level per SD (22.61 ng/mL) increment	–	–	3.53 (2.05–6.09)	<0.001

Model 1 indicated multivariable logistic regression analyses including age, sex, NIHSS score on admission, HbA1c, D‐dimer, ln (infarct core volume+1), cause classification and CitH3 as a continuous variable. Model 2 indicated multivariable logistic regression analyses including age, sex, NIHSS score on admission, HbA1c, D‐dimer, ln (infarct core volume+1), cause classification and MPO‐DNA as a continuous variable. AIS indicates acute ischemic stroke; CitH3, citrullinated histone 3; END, early neurological deterioration; HbA1c, glycated hemoglobin; IVT, intravenous thrombolysis; LAA, large‐artery atherosclerosis; MPO‐DNA, myeloperoxidase‐DNA; NIHSS, National Institutes of Health Stroke Scale; ODC, stroke of other determined cause; OR, odds ratio; SVO, small‐vessel occlusion; and UDC, stroke of undetermined cause.

Figure 2A shows the crude restricted cubic spline association between serum CitH3 levels and the risk of END. Under this framework (Tables S2 through S4), a 3‐knot spline was selected for CitH3. In the crude spline model, higher CitH3 levels were significantly associated with an increased risk of END (P for overall <0.001), with no evidence of a nonlinear association (P for nonlinearity =0.475), indicating an approximately linear dose–response pattern across the observed range.

Figure 2. Associations between serum NET biomarker levels and END in patients with AIS receiving IVT.

A, The restricted cubic splines based on logistic regression models depicted the concentration‐response relationships between CitH3 levels and END, with 3 knots at 10th, 50th, and 90th percentiles. B, The restricted cubic splines based on logistic regression models depicted the concentration‐response relationships between MPO‐DNA levels and END, with 4 knots at the 5th, 35th, 65th, 95th percentiles. The references CitH3 and MPO‐DNA levels were set at the median values of 16.26 and 59.99 ng/mL, respectively. The results were adjusted for age, sex, NIHSS score on admission, HbA1c, D‐dimer, ln (infarct core volume+1), and cause classification. Solid lines and the gray areas represented the adjusted ORs and 95% CIs of END. Red lines are the reference OR of 1. AIS indicates acute ischemic stroke; CitH3, citrullinated histone 3; END, early neurological deterioration; HbA1c, glycated hemoglobin; IVT, intravenous thrombolysis; MPO‐DNA, myeloperoxidase‐DNA; NET, neutrophil extracellular trap; NIHSS, National Institute of Health Stroke Scale; and OR, odds ratio.

Figure 2B shows the corresponding crude restricted cubic spline association between serum MPO‐DNA levels and END risk. Using the same Akaike information criterion‐based selection strategy (Tables S2 through S4), a 4‐knot spline was selected for MPO‐DNA. In contrast to CitH3, MPO‐DNA exhibited a significant nonlinear association with END (P for overall <0.001; P for nonlinearity=0.010, Figure 2B), suggesting that the relationship between MPO‐DNA and END risk was not adequately captured by a simple linear term.

Subgroup Analyses

In subgroup analyses (Table 4), the positive associations of serum CitH3 and MPO‐DNA levels with END were largely consistent across most clinical and laboratory strata. Notably, significant interactions were observed for CitH3 by CAD or stroke history (P for interaction=0.018) and D‐dimer levels (P for interaction=0.013). The association between CitH3 and END was stronger in patients with a history of CAD or stroke (OR, 9.11 [95% CI, 2.63–31.55]) than in those without (OR, 1.94 [95% CI, 1.19–3.19]), and in those with lower D‐dimer concentrations (OR, 3.48 [95% CI, 1.95–6.22]) compared with higher D‐dimer (OR, 1.85 [95% CI, 0.79–4.31]). For MPO‐DNA, a significant interaction was identified with smoking status (P for interaction =0.026). The association with END was markedly stronger among smokers (OR, 17.99 [95% CI, 3.81–85.03]) than nonsmokers (OR, 2.24 [95% CI, 1.23–4.08]).

Table 4.

Subgroup Analyses of the Associations Between Serum CitH3/MPO‐DNA Levels and END in Patients With AIS Receiving IVT

Subgroups	No. of END/patients	Each SD increment in CitH3 for END		Each SD increment in MPO‐DNA for END
		OR (95% CI)	P for interaction	OR (95% CI)	P for interaction
Age, y			0.987		0.275
<60	17/127	2.87 (1.41–5.81)†		3.53 (1.60–7.80)†
≥60	28/160	2.70 (1.53–4.79)‡		4.53 (2.08–9.89)‡
Sex			0.289		0.981
Female	12/85	4.45 (1.48–13.38)†		2.95 (1.07–8.13)*
Male	33/202	2.38 (1.42–4.00)†		3.72 (1.93–7.17)‡
Body mass index, kg/m2			0.852		0.449
<25.00	22/158	2.36 (1.19–4.69)*		4.14 (1.63–10.50)†
≥25.00	23/129	2.99 (1.56–5.71)‡		3.80 (1.78–8.13)‡
Smoking			0.455		0.026
No	26/184	2.51 (1.46–4.31)‡		2.24 (1.23–4.08)†
Yes	19/103	3.32 (1.44–7.63)†		17.99 (3.81–85.03)‡
Hypertension			0.261		0.950
No	14/110	2.10 (0.99–4.45)		4.49 (1.51–13.36)†
Yes	31/177	3.76 (2.03–6.95)‡		3.71 (1.90–7.22)‡
Diabetes			0.461		0.584
No	35/220	2.32 (1.39–3.86)†		2.93 (1.54–5.58)†
Yes	10/67	5.48 (1.73–17.37)†		8.14 (2.12–31.22)†
Hyperlipidemia			0.617		0.932
No	38/238	2.92 (1.74–4.88)‡		3.61 (2.02–6.43)‡
Yes	7/49	1.51 (0.38–5.96)		2.22 (0.33–15.02)
Coronary artery disease or stroke history			0.018		0.271
No	31/189	1.94 (1.19–3.19)†		3.32 (1.74–6.33)‡
Yes	14/98	9.11 (2.63–31.55)‡		5.68 (1.79–18.00)†
Atrial fibrillation			0.354		0.486
No	39/265	2.55 (1.60–4.08)‡		3.42 (1.95–6.00)‡
Yes	6/22	4.99 (0.68–36.65)		10.39 (0.61–178.04)
NIHSS score on admission			0.910		0.914
≤3	8/92	3.32 (1.08–10.21)*		4.02 (1.05–15.35)*
>3	37/195	2.70 (1.65–4.43)‡		3.07 (1.8–0‐5.25)‡
HbA1c, %			0.107		0.606
<5.70	8/85	2.56 (0.90–7.26)		5.52 (1.18–25.94)*
5.70–6.49	15/101	2.77 (1.27–6.04)*		2.54 (1.08–5.97)*
≥6.50	22/101	4.77 (2.16–10.50)‡		5.67 (2.34–13.69)‡
D‐dimer, μg/mL			0.013		0.078
<0.50	31/254	3.48 (1.95–6.22)‡		3.62 (1.90–6.89)‡
≥0.50	14/33	1.85 (0.79–4.31)		2.72 (0.89–8.29)
Onset‐to‐treatment time, min			0.604		0.054
<162	23/143	2.94 (1.55–5.58)‡		2.83 (1.45–5.55)†
≥162	22/144	3.02 (1.56–5.82)‡		6.14 (2.34–16.15)‡

ORs were adjusted for age, sex, NIHSS score on admission, HbA1c, and D‐dimer. Subgroup analyses were performed one subgroup at a time. When stratifying by a variable, that variable was not included in the covariate set for that stratum. AIS indicates acute ischemic stroke; BMI, body mass index; CitH3, citrullinated histone 3; END, early neurological deterioration; HbA1c, glycated hemoglobin; IVT, intravenous thrombolysis; MPO‐DNA, myeloperoxidase‐DNA; NIHSS, National Institute of Health Stroke Scale; OR, odds ratio.

^* P<0.05.

^† P<0.01.

^‡ P<0.001.

Sensitivity Analyses

Baseline characteristics were generally comparable between included and excluded patients, with no statistically significant differences observed in age, sex, NIHSS score on admission, HbA1c, D‐dimer, or stroke cause distribution (all P>0.05) (Table S5). Weight diagnostics indicated well‐behaved weight distributions after truncation at the 1st and 99th percentiles, with mean weights close to 1 and minimal variability, suggesting adequate control of potential selection bias (Table S6).

After further adjustment for OTT, the associations of both CitH3 >17.67 ng/mL and MPO‐DNA >79.72 ng/mL with END remained significant (Table S7). In these extended models, elevated CitH3 and MPO‐DNA levels were associated with 3.80‐fold (95% CI, 1.75–8.26; P=0.001) and 5.71‐fold (95% CI, 2.40–13.63; P<0.001) higher odds of END, respectively. These results remained consistent when NETs were analyzed as continuous variables (Table S8). Each SD increase in CitH3 and MPO‐DNA was associated with 2.63‐fold (95% CI, 1.68–4.12) and 3.59‐fold (95% CI, 2.07–6.23) higher odds of END, respectively (both P<0.001).

As Table S9 shown, in this nested logistic regression comparison, adding OTT, cause classification, and ln (infarct core volume+1) to simplified adjusted model did not significantly improve model fit (likelihood ratio tests: chi‐square test=5.36, df=5, P=0.374). Although the area under the ROC increased slightly from 0.749 (95% CI, 0.668–0.830) to 0.774 (95% CI, 0.700–0.849), the improvement was modest and not statistically significant. Both models demonstrated good calibration (Hosmer–Lemeshow P>0.4). Overall, the simplified adjusted model provided a more parsimonious fit, whereas the fully adjusted model offered minimal gain in discrimination at the cost of increased complexity.

In an extended subgroup framework (Table S10), the associations between serum CitH3 and MPO‐DNA levels and END remained consistent with those in the main analyses. Additional adjustment for OTT, ln (infarct core volume+1), and stroke cause did not materially alter the direction or magnitude of these relationships, indicating their robustness across diverse clinical contexts.

All sensitivity analyses produced results consistent with the primary models, confirming the robustness of the observed associations.

DISCUSSION

Our findings revealed that elevated prethrombolytic serum NET concentrations demonstrated a significant independent association with END risk in patients with AIS undergoing IVT. Patients with serum CitH3 levels exceeding 17.67 ng/mL were associated with a 3.8‐fold higher risk of END compared with those with ≤17.67 ng/mL. Similarly, patients presenting with serum MPO‐DNA >79.72 ng/mL showed a 5.6‐fold higher risk of END relative to those below this threshold. Both CitH3 and MPO‐DNA levels, when analyzed as continuous variables, showed a significant positive correlation with END risk. Notably, the relationship between CitH3 and END was linear and dose dependent, suggesting a progressive increase in END risk with higher CitH3 concentrations. Significant interactions were observed for CitH3 with D‐dimer and with prior CAD or stroke and for MPO‐DNA with smoking. Specifically, the association between higher NET levels and END was stronger in patients with D‐dimer <0.50 μg/mL, a history of CAD or stroke, and in smokers. These findings provided clinical evidence linking circulating NETs burden with early worsening after reperfusion therapy in a real‐world IVT population.

NETs have recently emerged as a novel marker and a potential therapeutic target to enhance thrombolysis efficacy in patients with AIS, motivating recent clinical investigations. In a small human cohort (N=60), ¹³ baseline plasma NETs levels were higher in AIS patients without early improvement after tissue plasminogen activator. The MRCLEAN‐MED trial ¹⁴ demonstrated that increased baseline plasma CitH3 levels were associated with severe NIHSS scores at 24 hours after endovascular thrombectomy in patients treated with endovascular thrombectomy alone (n=94, r=0.28, P=0.03). Our study advances the field by focusing on a real‐world IVT‐only cohort with a larger sample (N=287) and by targeting a standardized primary end point of END. Rather than simply descriptive comparisons, our study provides quantitative estimates on the associations using multivariable logistic regression (both ROC‐derived cutoffs and per‐SD effects), characterized dose–response with restricted cubic spline, and examined prespecified interactions.

Notably, CitH3–END associations appeared stronger in patients with prior CAD or stroke or in those with lower D‐dimer (<0.50 μg/mL) levels. These findings imply a feedback loop where cardio‐/cerebrovascular incidents, elevated serum NETs concentrations, and deteriorating clinical outcomes mutually reinforce each other. D‐dimer also appeared to modify the effect of NETs on END. At lower D‐dimer, when fibrin turnover is relatively limited, NET‐driven microvascular obstruction and impaired fibrinolysis may play a larger role. At higher D‐dimer, broader coagulation and inflammation activity likely predominates, attenuating the relative impact of NETs. We further observed that the association between higher MPO‐DNA and END was stronger in smokers, consistent with the concept that tobacco exposure amplified neutrophil activation, oxidative stress, and platelet–endothelium crosstalk, thereby facilitating NETosis after thrombolysis. ²⁶ Clinically, this heterogeneity suggests that patients who smoke could be particularly susceptible to NET‐mediated early deterioration after IVT and might derive greater benefit from NET‐targeted adjuncts. However, these interaction findings are exploratory and warrant external validation. Our study findings provide real‐world clinical evidence that bridges preclinical discoveries and these ongoing translational trials, highlighting the potential of NETs as both personalized therapeutic targets and prognostic biomarkers in the context of reperfusion therapy.

In AIS, NETs contribute to cerebral injury through multiple thromboinflammatory mechanisms. After ischemic onset, neutrophils execute the NETosis program via distinct intracellular cascades. NETs provide a scaffold for platelet adhesion and fibrin deposition, activate the coagulation cascade, and aggravate neuroinflammation and blood–brain barrier disruption, thereby amplifying ischemic damage. ¹¹ , ²⁷ Within thrombi, dense DNA–protein networks interlaced with fibrin and von Willebrand factor limit tissue plasminogen activator penetration and promote thrombolytic resistance. NET‐derived histones and DNA further stabilize fibrin structure and enhance plasminogen activator inhibitor‐1 release, thereby impairing fibrinolysis and microvascular reperfusion after intravenous thrombolysis. ⁵ In our study, we evaluated 2 specific NET biomarkers with distinct biological implications. The relationship between serum MPO‐DNA and END displayed a threshold‐like, nonlinear pattern, in contrast to the linear dose–response observed for CitH3. CitH3 reflects histone citrullination, an essential PAD4 (peptidylarginine deiminase 4)‐mediated step in chromatin decondensation during NETosis, and serves as a relatively specific structural marker of NET formation. ²⁸ Unlike CitH3, MPO‐DNA is a circulating complex formed after NET release, which undergoes DNase I/DNase IL3 (interleukin‐3)‐mediated degradation and may bind plasma proteins or be phagocytosed, leading to reduced stability and assay signal variability. ²⁹ When NETs burden is moderate, increases in MPO‐DNA may proportionally represent ongoing NETosis. But at higher concentrations, compensatory clearance or assay saturation could attenuate this relationship, producing a plateau effect. Although a statistical nonlinearity was observed for MPO‐DNA, the biological significance is uncertain. Rather than focusing on the curve pattern, the main implication is that higher levels of both NETs markers uniformly indicate greater END risk.

Another mechanistic explanation for our findings is the no‐reflow phenomenon, defined as failure of microvascular reperfusion despite successful macrovascular recanalization. ³⁰ In an experimental ischemic stroke model study, CitH3‐positive neutrophils and extracellular DNA–histone projections within capillary lumens, which constitutes morphological evidence of intravascular NETosis. They proposed that such intravascular NETs may induce secondary microthrombosis, delaying reperfusion and contributing to the no‐reflow phenomenon. ⁷ No‐reflow occurs in roughly 30% of patients with AIS and is associated with poor outcomes. ³¹ Clinically, a cohort study shows neutrophil‐to‐lymphocyte ratio and D‐dimer are associated with no‐reflow risk. ³² Experimental data support a neutrophil/NETs‐mediated pathway: insufficient DNase I/DNase IL3‐dependent clearance permits intravascular NET accumulation and microvascular occlusion, ²⁹ and 2‐photon imaging demonstrates neutrophil plugging of distal capillaries after recanalization, with an antineutrophil monoclonal antibody improving reperfusion. ³³ Although we did not directly quantify no‐reflow in the present study, our findings are consistent with this proposed mechanism. The convergence of clinical and experimental evidence supports a NETs‐no‐reflow pathway to early deterioration. There is a need to integrate standardized no‐reflow assessment with circulating NET markers in future work. ³⁴

To enhance statistical power and balance selection bias, this observational study enrolled consecutive patients from 2 geographically distinct branches of the same tertiary hospital. Nevertheless, several important limitations warrant consideration. First, the relatively small sample size may have introduced unavoidable selection bias, particularly in subgroup analyses where wide CIs reflected reduced statistical power and unstable effect estimates. Second, the patients in current study exhibited predominantly mild‐to‐moderate neurological deficits (median NIHSS score 5 [3–9]). This inherent selection bias may limit the external validity of our findings, particularly in relation to patients presenting with severe stroke. Third, only a limited number of confounders were controlled in this study and the results from this study may be subject to the residual confounding. Fourth, owing to the observational design, our analyses demonstrated associations rather than predictive or causal relationships. Subgroup analyses should be regarded as exploratory. The limited number of events meant these findings remained hypothesis‐generating and required external validation. Finally, NETs biomarkers were measured only once at baseline, which limited our ability to capture their dynamic changes during ischemia and reperfusion. Including both IVT and non‐IVT treated populations with AIS within 4.5 hours of onset would better clarify whether the association is specific to thrombolysis or generalizable to all patients with AIS. However, in real‐world settings, it is challenging to obtain multiple blood samples within such a short time window, and the number of non‐IVT patients was very limited. These limitations underscore the need for future studies with larger cohorts, more comprehensive confounding adjustment, and rigorous stratified analyses.

CONCLUSIONS

In this real‐world cohort, higher prethrombolytic circulating NET burden was associated with END risk after IVT in patients with AIS. A dose–response pattern that was linear for CitH3 and threshold‐like for MPO‐DNA was found. These results provide clinical evidence supporting NETs as potential biomarkers and mechanistic contributors to early neurological worsening following reperfusion therapy. These findings represent associations rather than predictive relationships and should be interpreted accordingly. Future multicenter studies with larger sample size and serial sampling are warranted to validate our findings, elucidate the temporal kinetics of NETs, and to explore whether targeting NETs can improve outcomes in acute ischemic stroke.

Sources of Funding

This work was supported by the National Natural Science Foundation of China (Nos. 82171324 and 82471349).

Disclosures

None.

Supporting information

STROBE Statement—Checklist

Tables S1–S10