Thromboinflammation and hemostatic activation drive early mortality risk in pulmonary embolism

Abstract

Background

Despite advances in diagnosis and treatment, pulmonary embolism (PE) remains a challenging condition with high mortality. Current prognostic models lack integration of the biological underpinnings driving adverse outcomes.

Objectives

This study profiles thromboinflammation and hemostatic biomarkers in patients with PE to evaluate their association with 30-day mortality.

Methods

In a prospective-specimen collection, retrospective-blinded evaluation study, plasma samples from 500 patients collected at time of diagnosis were analyzed. Biomarkers reflecting endothelial dysfunction, platelet activation, coagulation imbalance, impaired fibrinolysis, and inflammation were quantified. Hematological indices (neutrophil-to-lymphocyte ratio [NLR], platelet-to-lymphocyte ratio [PLR], and systemic immune–inflammation index [SII]) were calculated. Receiver operating characteristics analysis, multivariable logistic regression, and Chi-square automatic interaction detection (CHAID) recursive partitioning tree were used to assess predictors of mortality.

Results

Of 500 patients (median age, 64 years; 52% males), 48 (9.7%) died within 30 days. Levels of von Willebrand factor, P-selectin, tissue factor, tissue-type plasminogen activator, plasminogen activator inhibitor (PAI)-1, D-dimer, C-reactive protein, interleukin 6, tumor necrosis factor (TNF)-α, NLR, PLR, and SII were significantly higher, while FVII, FX, FXIIIA, and thrombin-activatable fibrinolysis inhibitor levels were significantly lower in 30-day nonsurvivors. Multivariable analysis revealed that elevated von Willebrand factor, E-selectin, tissue factor, tissue-type plasminogen activator, PAI-1, D-dimer, C-reactive protein, interleukin 6, TNF-α, and NLR, and decreased FVII, FX, FXIIIA, and TAFI levels, were independently associated with 30-day mortality. Chi-squared automatic interaction detection analysis identified elevated NLR, PAI-1, and TNF-α as key predictors of mortality.

Conclusion

Current PE scoring systems have a low positive predictive value for clinical deterioration and mortality. This study uniquely integrates endothelial, platelet, coagulation, fibrinolytic, inflammatory, and hematologic biomarkers to identify novel mortality–predictive thresholds, highlighting the value of thromboinflammation, and hemostatic activation biomarkers associated with the pathobiology of PE for enhancing risk stratification of patients with PE.

Essentials

• •Current pulmonary embolism (PE) scoring systems have a low predictive value for mortality.
• •In a large cohort of patients with PE, altered hemostatic and inflammatory biomarkers were identified.
• •Mortality-predictive quantifiable thresholds were established for all biomarkers.
• •Neutrophil-to-lymphocyte ratio, plasminogen activator inhibitor-1, and tumor necrosis factor α were key predictors of mortality.

1. Introduction

Pulmonary embolism (PE) remains a leading cause of cardiovascular-related mortality, with clinical manifestations ranging from asymptomatic incidental findings to sudden cardiac death [1]. In the United States alone, PE affects an estimated 370,000 individuals annually and contributes to ∼100,000 deaths [2]. Hemodynamic instability occurs in 25% of PE cases, often resulting in rapid clinical deterioration, underscoring the need for timely and precise risk stratification. Despite advancements in risk stratification, existing tools frequently fail to capture the evolving pathobiology of acute PE, particularly in the early stages, when timely intervention is most critical.

Robust clinical scoring algorithms such as the pulmonary embolism severity index and its simplified version have improved early decision making by incorporating demographic, clinical, and select biomarker data [[3], [4], [5]]. The European Society of Cardiology (ESC) guidelines and the Bova Outcomes Venous thromboembolism Assessment score with added lactate have further refined prognostication through incorporation of cardiac biomarkers, including troponin, natriuretic peptides, and lactate, as well as echocardiographic findings [6,7]. The Registro Informatizado de la Enfermedad TromboEmbólica (RIETE) registry additionally evaluated a multimarker approach combining simplified pulmonary embolism severity index, cardiac markers such as natriuretic peptide and troponin, and concomitant lower-limb deep vein thrombosis (DVT) [8]. Although these models are used to guide clinical decision making and resource allocation, their capacity to accurately predict all-cause mortality and clinical deterioration is constrained, as evidenced by a positive predictive value of only 2% [9,10]. These models largely rely on markers of end-organ injury, reflecting myocardial strain or metabolic stress, rather than the upstream biological processes driving PE progression. Current prediction models overlook evolving thromboinflammatory changes, which may be critical for accurately identifying patients at risk of early mortality.

Epidemiological factors such as surgery, trauma, obesity, elderliness, prolonged immobility, cancer, congenital or acquired thrombophilia, complement activation, and inflammation are known to increase the risk of developing thrombosis [[11], [12], [13], [14], [15], [16], [17], [18], [19]]. The concept of thromboinflammation, the integrated interplay between thrombosis and inflammation, has emerged as central to the pathogenesis and progression of PE [20]. The pathophysiology of PE is driven not only by thrombus formation but also by the inflammatory responses that sustain clot propagation and induce vascular injury. Recent studies highlight the critical roles of endothelial dysfunction, platelet activation, impaired fibrinolysis, and immune cell–mediated inflammation in facilitating thrombogenesis and disease progression [[21], [22], [23], [24], [25]].

This study hypothesizes that profiling thromboinflammation and hemostatic activation pathways can improve the prediction of early mortality in PE beyond current clinical models. The primary objective of this study was to discover novel, quantifiable biomarkers associated with thromboinflammation and hemostatic activation in acute PE as predictors of PE early (30-day) mortality. The secondary objective of this study was to compare biomarker levels with those of healthy individuals.

2. Methods

2.1. Study design and patient population

This study implemented a prospective-specimen-collection, retrospective-blinded-evaluation (PRoBE) design to investigate thromboinflammation and hemostatic activation pathways in patients with acute symptomatic PE. The PRoBE design is a rigorous methodological framework designed for evaluating the clinical accuracy of biomarkers, emphasizing prospective sample collection, blinded outcome assessment, and predefined analysis to minimize bias and ensure validity [26]. Patients with PE were stratified according to the ESC criteria [6]. Additionally, 30-day all-cause mortality rates, defined as primary outcomes, were collected independently of the biomarker status. This single-center study evaluated 500 patients with acute PE enrolled in the Pulmonary Embolism Response Team (PERT) program of the Loyola University Health System (LUHS) between 2018 and 2023. The LUHS PERT functions at both the Loyola University Medical Center and Gottlieb Memorial Hospital.

PE was diagnosed using imaging modalities, including echocardiography, computed tomography, pulmonary angiography, or ventilation–perfusion scanning. Specifics of the LUHS PERT program have been published [27]. Inclusion criteria required that participants were adults (≥18 years) with a radiographically confirmed diagnosis of acute PE, with no additional exclusion criteria. The study protocol was approved by the institutional review board (IRB 209457032217) of Loyola University Chicago. The majority of patients were followed up for 6 months to 1 year at the outpatient clinic for ongoing evaluation and assessment of long-term outcomes.

2.2. Sample size justification and power calculation

The sample size was determined based on the anticipated effective size of biomarker associations in predicting 30-day mortality in patients with acute PE. Power analysis targeted detection of an odds ratio (OR) of >2.0 at 80% power and a 2-sided significance level of 0.05. Based on an estimated mortality rate of 8% to 10%, the minimum sample size required for adequately powered multivariable analysis was 470 patients. Power calculations were performed using G∗Power 3.1.

2.3. Sample collection and processing

Patients admitted to the emergency department due to PE, eligble for the inclusion in the study, were identified by the PERT clinicians. Single blood samples per patient (0.3 mL of 3.2% sodium citrate to 2.7 mL whole blood) collected within 24 to 72 hours of PE diagnosis were immediately transported to the Hemostasis and Thrombosis Research Laboratories at Loyola University Chicago. Upon arrival, samples were centrifuged at 3000 rpm to isolate platelet-poor plasma, which was immediately aliquoted and stored at −70 °C until analysis. Healthy individuals consisted of 25 males and 25 females with no known medical conditions and not taking any medications, sourced from George King Biomedical. All analyses were conducted within 1 year of sample collection.

2.4. Data collection and patient information

Demographic characteristics and comorbid conditions were retrospectively extracted from electronic health records, including gender, age, weight, body mass index, and established risk factors for PE, such as active cancer, recent surgery, prolonged immobility, and pregnancy. Comorbidities included a history of DVT or PE, leg varicosities, chronic heart failure, chronic lung disease, diabetes, hypertension, myocardial infarction, cerebral ischemia, peripheral artery disease, statin use, smoking, and recent major bleeding episodes.

Clinical signs and symptoms, such as systolic blood pressure, heart rate, oxygen saturation, syncope, chest pain, and dyspnea, were systematically recorded. Laboratory parameters included complete blood counts (neutrophil, lymphocyte, and platelet counts), cardiac biomarkers (B-type natriuretic peptide and troponin), and echocardiographic assessment of right ventricular diameter to left ventricular diameter ratio. Patients were stratified according to the ESC guidelines for PE severity.

Mortality data, including both in-hospital and postdischarge mortality, were collected from medical records and outpatient follow-up visits. Both outcomes were confirmed through electronic data entry and supplemented with direct follow-up telephone calls for patients without documented outcomes.

2.5. Quantification of biomarker levels

Biomarkers were quantified using sandwich ELISAs. To ensure quality control, pooled normal human plasma was used as an internal standard in each assay. Components of the thromboinflammatory response evaluated included markers of endothelial injury (von Willebrand factor [VWF], E-selectin, and tissue factor [TF]), platelet activation (P-selectin), coagulation (factor [F]VII, FIX, FX, and FXIIIA), fibrinolytic function (tissue-type plasminogen activator [tPA], plasminogen activator inhibitor [PAI]-1, thrombin-activatable fibrinolysis inhibitor [TAFI], and D-dimer), and systemic inflammation (C-reactive protein [CRP], interleukin [IL]-6, and tumor necrosis factor [TNF]-α). Each of these biomarkers plays a distinct role in thromboinflammation and hemostatic activation and were chosen for this study to identify physiological pathways that may contribute to the complex pathophysiology of PE (Table 1) [[28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46]].

Table 1.

Overview of thromboinflammation and hemostatic activation markers and their clinical significance in PE.

Pathophysiological Process	Biomarkers	Biological role	Prognostic implication
Endothelial activation	VWF, E-selectin, and TF	Initiates prothrombotic shift; enhances platelet adhesion [[28], [29], [30], [31]]	Reflects vascular injury; elevated levels of VWF associated with increased risk of VTE and related complications [32]
Impaired fibrinolysis	PAI-1, D-dimer, TAFI, and tPA	Inhibits clot breakdown; promotes thrombus persistence [[33], [34], [35], [36]]	Elevated PAI-1 and D-dimer indicate clot burden and impaired resolution, and PE severity [22]
Platelet activation	P-selectin and FXIIIA	Promotes clot stabilization and propagation [29,37,38]	Associated with increased thrombus formation and PE severity [39]
Systemic inflammation	CRP, IL-6, and TNF-α	Signals acute-phase response and immune activation [[40], [41], [42], [43]]	Elevated levels of CRP were linked with VTE; IL-6 levels varied with PE severity [21,44]
Hematologic indices of immune dysregulation	NLR, PLR, and SII	Reflect systemic inflammation, platelet activation, and immune imbalance	Readily available clinical data; higher NLR and SII were associated with increased risk of mortality and major bleeding in VTE [45,46]

CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; PLR, platelet-to-lymphocyte ratio; SII, systemic immune–inflammation index; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

Assays for VWF, FVII, FIX, FX, FXIIIA, tPA, PAI-1, TAFI, D-dimer, and CRP were sourced from Aniara Diagnostics, HYPHEN BioMed; kits for IL-6, TNF-α, and P- and E-selectin were obtained from R&D Systems; and the TF kit was supplied by BioMedica Diagnostics. All assays were conducted according to the manufacturers’ protocols. Biomarkers were measured as antigen levels, which are not influenced by anticoagulants, ensuring that the values accurately reflected thromboinflammatory activity in patients with acute PE.

2.6. Hematologic indices

Hematology parameters from complete blood cell counts were collected from patients’ medical records. Calculations of these parameters included the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and the systemic immune-inflammation index (SII), calculated as (platelet count × neutrophil count)/lymphocyte count.

2.7. Statistical analysis

All statistical analyses were conducted using GraphPad Prism software, version 10.0 (GraphPad Software); SPSS statistics, version 31.0 (IBM); and SPSS modeler, version 18.1 (IBM). Biomarker assay evaluations remained blinded of demographic data and outcome until the final statistical analysis. Importantly, mortality status did not influence laboratory analyses, which were conducted independently. Biomarkers were analyzed as continuous, categorical, and cutoff-based variables to comprehensively evaluate their prognostic utility. Based on data distribution using the Kolmogrov–Smirnov test, median values and IQRs (25-75) were used. The nonparametric 2-tailed Wilcoxon Mann–Whitney U-test was used to compare biomarker levels in the following groups: patients with PE vs healthy individuals; stratification based on the PE severity; and 30-day survivors vs nonsurvivors, with significance determined at P < .05. Fold changes relative to healthy individuals were calculated to quantify biomarker elevation and prioritize them based on the magnitude of change. Spearman correlation analysis assessed multicollinearity among biomarkers, with significant correlations (r) determined at P < .05.

To enhance clinical interpretability, receiver operating characteristic (ROC) curve analyses were performed to determine the optimal cutoff values for each biomarker in predicting 30-day mortality. Significant markers were identified at P < .05, with the area under the curve (AUC) and 95% CIs reported. These cutoffs were then used to categorize biomarkers for further analysis, with results presented as numbers, percentages, and frequencies. Univariate analysis was conducted using chi-squared tests to evaluate associations between biomarkers and 30-day mortality outcomes, with ORs and 95% CIs calculated at P < .05.

Multivariable logistic regression models were constructed to assess the independent predictive value of each biomarker for 30-day mortality, adjusting for PE severity based on ESC guidelines. A backward conditional selection method was applied to identify the most relevant predictors. Each biomarker was analyzed individually in separate models alongside PE severity, with results presented as adjusted ORs (aORs), 95% CIs, and P values.

Chi-square automatic interaction detection (CHAID) recursive partitioning was used to predict mortality and identify the most robust biomarkers [47]. This nonparametric decision tree method determines optimal data splits based on chi-squares tests of association between predictors and the outcome. Splits were evaluated using chi-squared tests with Bonferroni-adjusted P values. Tree growth continued until no further significant splits were detected or minimum node size criteria were met. Model performance was assessed using classification accuracy and risk estimates.

3. Results

3.1. Demographics and clinical characteristics

The study cohort had a near-equal distribution of males (52.1%) and females (47.9%), a median age of 64 years, and a median body mass index of 29.7 kg/m² (Table 2). The most common presenting symptoms were dyspnea (88.6%), chest pain (30.0%), and syncope (12.0%). According to ESC classification, 10.4% (n = 51) of patients were estimated to be at low risk, 80.9% (n = 397) as intermediate risk, and 8.8% (n = 43) as high risk. During the first 30 days, 48 patients (9.7%) died, while 73 patients (14.8%) died at 90 days.

Table 2.

Baseline clinical characteristics and parameters in the LUHS PERT cohort.

Variables	Counts	LUHS PERT cohort
Patients	500	—
Demographics
Age (y)	493	64 (53-73)
Male sex	493	257 (52.1)
Weight (kg)	492	87 (71-10)
Body mass index (kg/m2)	492	30 (25-36)
Clinical presentation
Systolic blood pressure (mm Hg)	493	117 (103-130)
Heart rate (beats/min)	493	97 (85-112)
Oxygen saturation (%)	493	92 (89-95)
Respiratory rate (breaths/min)	493	21 (18-26)
Dyspnea	493	437 (88.6)
Syncope	493	59 (12.0)
Chest pain	493	148 (30.0)
Predictors of PE severity
RV/LV ratio	339	0.93 (0.80-1.17)
Troponin	493	0.05 (0.02-0.27)
Brain natriuretic peptide	449	129 (47-3612)
PESI score	493	111 (82-148)
Risk stratification
Low-risk PE	492	51 (10.4)
Intermediate-risk PE	492	397 (80.9)
High-risk PE	492	44 (8.8)
Comorbidities
Active cancer	493	117 (23.7)
Chronic heart failure	493	63 (12.8)
Chronic lung disease	493	60 (12.2)
Chronic kidney disease	493	53 (10.8)
Mortality
30-day mortality	493	48 (9.7)
Laboratory parameters
Hemoglobin (g/dL)	493	12.3 (10.1-13.9)
Neutrophils (×10/μL3)	493	7.1 (4.7-9.8)
Lymphocytes (×10/μL3)	493	1.4 (0.8-2.0)
Platelets (×10/μL3)	493	228.0 (166.5-302.5)

Values are n (%) or median (IQR).

LUHS, Loyola University Health System; PERT, pulmonary embolism response team; PESI, pulmonary embolism severity index; RV/LV ratio, right ventricular/left ventricular ratio.

3.2. Biomarker levels in PE vs healthy individuals

Upon admission to hospital patients with PE, compared with healthy individuals, exhibited significantly elevated levels of biomarkers (Supplementary Table 1). Among endothelial markers, VWF was notably elevated (P < .0001), whereas E-selectin showed a modest but nonsignificant increase. Platelet activation markers, particularly P-selectin, were significantly (P < .0001) high in patients with PE. Coagulation factors revealed distinct alterations in patients with PE. TF (P < .0001) demonstrated the most pronounced increase, while FIX (P = .0138) and FX (P = .0400) showed only minimal elevation. In contrast, FVII (P = .0412) and FXIIIA (P < .0001) were decreased in patients with PE. Fibrinolytic markers, including tPA, PAI-1, and D-dimer were significantly elevated (P < .0001) in patients with PE, while TAFI showed no measurable change. Systemic inflammatory markers were markedly upregulated, especially CRP (P < .0001), followed by IL-6 (P < .0001) and TNF-α (P < .0001). Hematological indices showed most pronounced increase in NLR (P < .0001), followed by PLR (P = .0003) and then SII (P < .0001).

3.3. Severity stratification and biomarker levels

With increasing PE severity, biomarkers exhibited divergent patterns, with some levels increasing and others decreasing in association with disease severity (Figure 1, Supplementary Table 2).

Figure 1.

Comparison of thromboinflammation and hemostatic activation biomarkers in patients with PE based on the PE severity. CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; NS, not significant; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

3.3.1. Endothelial dysfunction

VWF showed a consistent increase from low-risk to high-risk PE, with significant increase from low-risk to intermediate-risk (P < .001) and high-risk PE (P = .007). The change from intermediate- to high-risk PE was minimal. E-selectin showed a nonsignificant decrease from low-risk PE to intermediate-risk, and high-risk, also from intermediate-risk to high-risk PE.

3.3.2. Platelet activation

P-selectin increased across the risk groups. The increase in P-selectin from low-risk patients to high-risk patients (P = .005) and from intermediate-risk patients to high-risk patients (P < .001), were statistically significant. A modest increase in P-selectin was observed from low-risk to intermediate-risk PE.

3.3.3. Coagulation

TF levels increased with increase in PE severity from low-risk PE to intermediate-risk PE, and high-risk PE; however, the increase from low-risk to intermediate-risk PE (P = .013) was the only significant increase. Factor VII levels significantly decreased from low-risk patients to intermediate-risk (P = .045) and high-risk (P = .011) patients, respectively. The decrease from intermediate risk to high-risk was not significant. Factor IX showed increased levels from low-risk to intermediate-risk, while the level in the high-risk patients decreased compared with both low-risk and intermediate-risk PE. Factor X demonstrated a consistent and significant decrease in levels with increased PE severity, from low-risk to intermediate-risk (P = .018) and high-risk (P < .001) PE, respectively, also from intermediate-risk to high-risk PE (P = .003). Factor XIIIA followed the same trend as FX, with a consistent and nonsignificant decrease in the levels from low-risk to intermediate-risk and high-risk PE.

3.3.4. Fibrinolysis

tPA significantly increased with the severity of PE, from low-risk to intermediate-risk, and high-risk PE. TAFI showed a modest increase from low-risk PE to intermediate-risk PE. PAI-1 increased across PE severity, with a modest rise from low-risk to intermediate-risk PE. A significant increase was observed in high-risk PE compared with both low-risk (P = .001) and intermediate-risk (P = .004) PE. However, TAFI levels significantly decreased in high-risk PE compared with both low-risk (P = .020) and intermediate-risk (P = .009) patients with PE. D-dimer elevated significantly with increasing severity from low-risk to intermediate-risk (P = .003), and high-risk patients with PE (P < .001), and from intermediate-risk to high-risk (P < .001) patients with PE.

3.3.5. Systemic inflammation

CRP showed a variation with a modest increase and decrease in the levels across the PE severity. Increase in IL-6 was consistent with the severity of PE, which significantly increased from low-risk PE to intermediate-risk PE (P = .016), and high-risk PE (P = .001), and from intermediate-risk PE to high-risk PE (P = .038). TNF-α also showed a variation across the risk categories with a significant increase from low-risk PE to intermediate-risk PE (P = .0015). The level of TNF-α increased in high-risk PE compared with low-risk PE, while it decreased when compared with intermediate-risk PE.

3.3.6. Hematologic indices

NLR consistently increased from low-risk PE to intermediate-risk PE and high-risk PE. Only the increase from low-risk to intermediate-risk PE (P = .036) was significant. PLR increased from low-risk PE to intermediate-risk PE but decreased in high-risk PE. However, the decrease from intermediate-risk subgroup to high-risk PE (P = .011) was the only significant change. SII also increased from low-risk PE to intermediate-risk PE and then decreased in high-risk PE; however, changes were not significant.

3.4. Biomarkers and 30-day mortality

Biomarker analysis revealed distinct patterns associated with 30-day mortality (Table 3). ROC curve analysis demonstrated that both elevated and reduced levels of specific biomarkers were predictive of mortality (Figure 2). In univariate analysis, elevated levels of several biomarkers were significantly associated with 30-day mortality (Table 4). After adjusting for ESC-defined severity in multivariable logistic regression, most biomarkers remained independently associated with early mortality (Table 4, Figure 3).

Table 3.

Comparison of thromboinflammation and hemostatic activation biomarkers in patients with PE by 30-day mortality status.

Biomarkers	30-d mortality
	Dead		Alive		P
	n	Median (IQR)	n	Median (IQR)
Endothelial dysregulation markers
VWF (%)	47	144.3 (140.0-153.7)	447	138.4 (128.1-147.2)	.003
E-selectin (ng/mL)	44	38.1 (20.3-67.7)	426	30.9 (19.1-49.0)	NS
Platelet activation markers
P-selectin (ng/mL)	47	117.6 (71.6-219.3)	447	89.8 (52.6-149.6)	.018
Coagulation factors
TF (pg/mL)	22	189.3 (124.9-343.3)	358	115.4 (72.1-165.7)	<.001
FVII (%)	47	87.5 (55.1-97.9)	441	97.3 (78.8-115.1)	<.001
FIX (%)	47	124.7 (75.6-147.2)	441	121.1 (91.3-142.5)	NS
FX (%)	47	86.0 (63.6-104.1)	441	95.7 (83.4-110.8)	.002
FXIIIA (%)	48	57.5 (43.8-83.2)	452	76.0 (56.2-95.4)	.001
Fibrinolytic deficit markers
tPA (ng/mL)	48	14.9 (7.4-20.9)	451	11.2 (7.2-17.1)	.043
PAI-1 (ng/mL)	48	64.9 (40.6-85.5)	452	48.6 (33.9-65.3)	.001
TAFI (%)	46	94.5 (84.0-111.4)	440	102.0 (92.8-112.0)	.031
D-dimer (ng/mL)	48	7174.1 (3167.8-14,283.6)	452	5392.0 (2814.8-9692.8)	.048
Systemic inflammatory markers
CRP (μg/mL)	48	14.6 (13.3-18.2)	452	13.9 (12.1-16.0)	.004
IL-6 (pg/mL)	41	79.9 (28.0-287.1)	413	25.5 (9.8-59.8)	<.001
TNF-α (pg/mL)	44	2.2 (1.0-3.6)	426	1.2 (0.8-1.8)	.001
Hematologic indices
NLR	48	12.6 (6.1-23.2)	445	4.7 (2.7-8.4)	<.001
PLR	48	284.8 (122.4-476.7)	445	173.1 (104.6-263.0)	.001
SII	48	2583.0 (1277.0-2820.0)	445	1120.0 (567.3-2043.0)	<.001

CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; NS, not significant; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

Figure 2.

Receiver operating characteristics curve analysis of thromboinflammation and hemostatic activation biomarkers with 30-day mortality. CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; NS, not significant; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; PLR, platelet-to-lymphocyte ratio; SII, systemic immune–inflammation index; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

Table 4.

Univariable chi-squared and multivariable regression analysis of thromboinflammation and hemostatic activation biomarkers with 30-day mortality.

Biomarkers	Chi-squared analysis			Regression analysis
	n (%)	OR (95% CI)	P	aOR (95% CI)	P
Endothelial dysregulation markers
VWF > 140%	35 (74.5)	3.37 (1.71-6.67)	<.001	3.13 (1.57-6.24)	.001
E-selectin > 48 ng/mL	20 (45.5)	2.33 (1.24-4.38)	.007	2.24 (1.17-4.03)	.015
Platelet activation markers
P-selectin > 110 ng/mL	26 (55.3)	1.95 (1.06-3.57)	.029	—
Coagulation factors
TF > 130 pg/mL	16 (72.7)	3.92 (1.50-10.25)	.003	5.07 (1.79-14.35)	.002
FVII < 65%	19 (40.4)	5.60 (2.91-10.80)	<.001	5.01 (2.54-9.88)	<.001
FIX > 135%	20 (42.6)	1.51 (0.82-2.78)	NS	—
FX < 65%	13 (27.7)	5.54 (2.63-11.68)	<.001	4.61 (2.12-10.10)	<.001
FXIIIA < 58%	25 (52.1)	3.08 (1.68-5.64)	<.001	3.38 (1.81-6.31)	<.001
Fibrinolytic deficit markers
tPA > 15 ng/mL	24 (50.0)	2.24 (1.23-4.09)	.007	2.09 (1.13-3.89)	.019
PAI-1 > 63 ng/mL	27 (56.3)	3.44 (1.88-6.32)	<.001	2.98 (1.60-5.56)	<.001
TAFI < 95%	24 (52.2)	2.52 (1.36-4.65)	.003	2.14 (1.14-4.03)	.019
D-dimer > 13,688 ng/mL	16 (33.3)	4.56 (2.32-8.96)	<.001	3.61 (1.71-7.60)	<.001
Systemic inflammatory markers
CRP > 12 μg/mL	46 (95.8)	7.37 (1.76-30.87)	.001	6.82 (1.62-28.70)	.009
IL-6 > 48 pg/mL	29 (70.7)	5.37 (2.65-10.87)	<.001	5.13 (2.52-10.50)	<.001
TNF-α > 2 pg/mL	23 (47.9)	3.68 (2.00-6.79)	<.001	3.76 (1.99-7.09)	<.001
Hematologic indices
NLR > 7	36 (75.0)	6.40 (3.23-12.67)	<.001	6.31 (3.15-12.70)	<.001
PLR > 220	30 (62.5)	3.15 (1.70-5.83)	<.001	—
SII > 1600	33 (68.8)	4.12 (2.17-7.81)	<.001	—

aOR, adjusted odds ratio; CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; NS, not significant; OR, odds ratio; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; PLR, platelet-to-lymphocyte ratio; SII, systemic immune–inflammation index; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

Figure 3.

Multivariable regression analysis for thromboinflammation and hemostatic activation biomarkers in predicting 30-day mortality. Note: The model is adjusted for the ESC severity. aOR, adjusted odds ratio; CRP, C-reactive protein; F, factor; IL, interleukin; NLR, neutrophil-to-lymphocyte ratio; OR, odds ratio; PAI, plasminogen activator inhibitor; PE, pulmonary embolism; TAFI, thrombin-activatable fibrinolysis inhibitor; TF, tissue factor; TNF, tumor necrosis factor; tPA, tissue-type plasminogen activator; VWF, von Willebrand factor.

3.4.1. Endothelial dysfunction

Markers of endothelial injury were significantly elevated in nonsurvivors. Median VWF was markedly higher among those who died; ROC analysis demonstrated that VWF of >140% was associated with mortality (AUC, 0.632; 95% CI, 0.544-0.720; P = .003). E-selectin was numerically elevated (>48 pg/mL) but did not reach statistical significance (AUC, 0.563; 95% CI, 0.466-0.659; P = NS). Univariate analysis confirmed that elevated levels of VWF and E-selectin were significantly associated with 30-day mortality. In regression analysis, both VWF (aOR, 3.13; 95% CI, 1.57-6.24; P = .001) and E-selectin (aOR, 2.24; 95% CI, 1.17-4.30; P = .015) remained independent predictors of 30-day mortality.

3.4.2. Platelet activation

P-selectin was elevated in nonsurvivors compared with the survivors. ROC analysis identified P-selectin of >110 ng/mL as a significant predictor of 30-day mortality (AUC, 0.606; 95% CI, 0.522-0.691; P = .014). P-selectin showed an association and predictive value with 30-day mortality by univariate analysis; however, this association did not remain significant in the regression analysis.

3.4.3. Coagulation

While TF demonstrated the highest discriminative ability among coagulation-related biomarkers, several other coagulation factors were decreased in patients who died. Among the biomarkers analyzed, ROC curve analysis showed that TF of >130 pg/mL (AUC, 0.790; 95% CI, 0.705-0.874; P < .001), while FVII of <65% (AUC, 0.790; 95% CI, 0.589-0.754; P < .001), FX of <65% (AUC, 0.636; 95% CI, 0.545-0.727; P = .004), and FXIIIA of <58% (AUC, 0.642; 95% CI, 0.558-0.725; P = .001) significantly predicted 30-day mortality. In contrast, FIX (>135%) levels were marginally higher in nonsurvivors (AUC, 0.507; 95% CI, 0.414-0.600; P = NS), although the difference was not statistically significant. In both, univariate and regression analysis, increased levels of TF, and decrease in FVII, FX, and FXIIIA were significantly associated with 30-day mortality. Notably, TF of >130 pg/mL (aOR, 5.07; 95% CI, 1.79-14.35; P = .002), FVII of <65% (aOR, 5.01; 95% CI, 2.54-9.88; P < .001), FX of <65% (aOR, 4.61; 95% CI, 2.12-10.10; P < .001), and FXIIIA of <58% (aOR, 3.38; 95% CI, 1.81-6.31; P < .001) demonstrated the strongest association with 30-day mortality in regression analysis.

3.4.4. Fibrinolysis

Markers of fibrinolysis including tPA, PAI-1, and D-dimer levels were significantly elevated in nonsurvivors, while TAFI decreased. ROC curve analysis revealed that elevated levels of tPA of >15 ng/mL (AUC, 0.591; 95% CI, 0.502-0.680; P = .046), PAI-1 of >63 ng/mL (AUC, 0.643; 95% CI, 0.551-0.735; P = .002), and D-dimer of >13,688 ng/mL (AUC, 0.588; 95% CI, 0.494-0.681; P = .007) as significant predictors of 30-day mortality. In contrast, decreased levels of TAFI of <95% (AUC, 0.599; 95% CI, 0.501-0.696; P = .047) were also associated with an increased mortality risk. Elevated levels of tPA, PAI-1, and D-dimer, as well as a decrease in TAFI, were significantly associated with 30-day mortality in both univariate and multivariate regression analyses. Among these, D-dimer (aOR, 3.61; 95% CI, 1.71-7.60; P < .001) showed the strongest association with 30-day mortality in regression, followed by PAI-1 (aOR, 2.98; 95% CI, 1.60-5.56; P < .001), TAFI (aOR, 2.14; 95% CI, 1.14-4.03; P = .019), and tPA (aOR, 2.09; 95% CI, 1.13-3.89; P = .019).

3.4.5. Systemic inflammation

CRP, IL-6, and TNF-α levels were significantly elevated in nonsurvivors. ROC analysis showed that CRP of >12 μg/mL (AUC, 0.626; 95% CI, 0.548-0.704; P = .001), IL-6 of >48 pg/mL (AUC, 0.727; 95% CI, 0.644-0.810; P < .001), and TNF-α of >2.0 pg/mL (AUC, 0.649; 95% CI, 0.541-0.756; P = .007) were strongly associated with early mortality. Both the univariate and logistic regression analyses identified CRP, TNF-α, and IL-6 as an independent predictor of 30-day mortality, with CRP showing the highest association (aOR, 6.82; 95% CI, 1.62-28.70; P = .009), followed by IL-6 (aOR, 5.13; 95% CI, 2.52-10.50; P < .001) and TNF-α (aOR, 3.76; 95% CI, 1.99-7.09; P < .001) in the multivariate model.

3.4.6. Hematologic indices

Indices including NLR, PLR, and SII were significantly higher in patients who died. ROC analysis showed NLR of >7 as a robust predictor (AUC, 0.758; 95% CI, 0.683-0.834; P = .000) of 30-day mortality, followed by SII of >1600 (AUC, 0.706; 95% CI, 0.624-0.789; P = .000) and PLR of >220 (AUC, 0.642; 95% CI, 0.547-0.736; P = .003). In univariate analysis, all hematological indices were associated with mortality, with NLR showing the highest association, followed by SII and PLR. To assess and address potential multicollinearity among predictors, Spearman correlation analysis was performed. Given the strong correlations between NLR and PLR, as well as NLR and SII, only NLR was included in the final model to avoid collinearity. In multivariate analysis NLR of >7 emerged as an independent predictor of 30-day mortality (aOR, 6.31; 95% CI, 3.15-12.70; P < .001).

CHAID recursive partitioning tree identified NLR of >7 as the primary discriminator of 30-day mortality, representing the most influential split in the decision tree (Figure 4). Downstream splits incorporated biomarkers reflecting fibrinolytic dysfunction and cytokine activation. Based on the model-generated predictor importance ranking, variables contributed to mortality prediction in the following hierarchical order: NLR > 7 (P < .0005; $\chi$² = 25.699); PAI-1 > 63 ng/mL (P = .028; $\chi$² = 4.813); and TNF-α > 2 pg/mL (P = .035; $\chi$² = 4.437). Overall, cohort mortality at the root node was 5.8% (22/380), but it was 41.2% (7/17) at the terminal node for those who met all 3 of these biomarker values (node 10). All other terminal nodes revealed a mortality distribution that was similar to the root node.

Figure 4.

Chi-squared automatic interaction detection recursive partitioning tree for thromboinflammation and hemostatic activation biomarkers in predicting 30-day mortality. NLR, neutrophil-to-lymphocyte ratio; PAI, plasminogen activator inhibitor; TNF, tumor necrosis factor.

4. Discussion

This study provides an in-depth characterization of thromboinflammation and hemostatic activation biomarkers in patients with PE, offering insight into the biological mechanisms underpinning disease severity and clinical outcomes. To our knowledge, such a large biomarker profiling study conducted in an integrated manner simultaneously evaluating endothelial dysfunction, platelet activation, coagulation, fibrinolysis, and inflammatory pathways has not been reported to date. Although there are several reports on inflammatory biomarkers and their diagnostic potential, these studies did not target a large PE cohort but were related to postthrombotic syndrome and venous thrombosis [38,39]. Furthermore, this study also addressed a relatively large cohort of 500 patients, in contrast to previous studies with a limited number of patients. In addition, this study evaluated cellular indices alongside other circulating biomarkers, providing a more comprehensive assessment of thromboinflammatory processes in PE.

In our study, severity stratification of PE revealed significant associations between biomarker levels and clinical risk categories. These findings underscore the prognostic significance of the studied biomarkers and suggest that incorporation of thromboinflammatory and hemostatic activation biomarkers into stratification scores could refine these clinical tools. The further novelty of our investigation was that relevance to mortality was addressed. Moreover, our study, for the first time, established quantifiable thresholds of biomarkers in predicting early mortality in patients with acute PE.

Endothelial dysfunction markers showed marked variations with PE severity, suggesting that endothelial damage intensifies with increasing severity of PE. In particular, VWF levels increased proportionally with PE severity, as well as in nonsurvivors. Published studies reported an association of VWF with VTE [32] and provoked and unprovoked DVT [48]. Our results are also consistent with previous studies that reported no significant change in E-selectin levels between 30-day nonsurvivors and survivors [49]. However, our study is the first to identify specific quantification thresholds of >140% VWF and >48 ng/mL E-selectin as predictors of 30-day mortality. While previous studies have acknowledged the role of endothelial injury in thrombosis, the specific quantifiable thresholds correlating with 30-day mortality in PE presented in this study adds a new dimension to our understanding of how endothelial damage contributes to PE progression.

P-selectin, a key indicator of platelet activation, was significantly elevated in patients with more severe PE and among those who experienced 30-day mortality. This increase reflects sustained platelet–endothelial interactions and a prothrombotic state. The observed increase in P-selectin may be due to the heightened platelet activation and degranulation in response to disease progression. However, despite its association with disease severity and mortality, P-selectin was not found to be a reliable predictive of clinical outcomes. These findings are consistent with previous studies in DVT, as well as in both uncomplicated and complicated PE [39].

Coagulation markers showed a complex profile: upstream activators such as TF were elevated, while downstream components, including FVII, FX, and FXIIIA, were reduced with increase in severity. Changes were only significant with TF, FVII, and FX. This suggests a possible consumption or dysregulation of the coagulation cascade, reflecting the overall activation of the coagulation cascade and a hypercoagulable state in the patients that increases with increasing severity of PE. A prospective study reported the decrease of FXIII in patients with PE compared with patients diagnosed with other serious conditions of acute coronary syndrome and aortic dissection, consistent with our findings [50]. Contrary to our results, one study reported elevated levels of FVII, FIX, and FX and their association with venous thrombosis compared with a control group [51]. A systematic review and meta-analysis also reported an association between elevated FIX and increased risk of VTE [52]. In our study, the trend of change in the levels remained with mortality, where TF was significantly increased, while FVII, FX, and FXIIIA were significantly decreased, in patients who experienced 30-day mortality with thresholds predictive of 30-day mortality identified as TF of >130 pg/mL, FVII of <65%, FX of <65%, and FXIIIA of <58%. Since the levels of these factors were measured by immunologic methods, which are prone to epitopic modification by interactions with other proteins, these changes need further validation by functional assays and potential interference of binding proteins.

The markers of the fibrinolytic process—tPA, PAI-1, and TAFI—showed various degrees of elevation across PE severity. This is consistent with the response of D-dimer (a current marker in clinical scores), reinforcing the concept of increased fibrin formation, clot degradation, and impaired thrombus resolution in PE. Furthermore, tPA, PAI-1, TAFI, and D-dimer levels were significantly altered in 30-day nonsurvivors. These findings are consistent with our earlier report, which demonstrated that elevated PAI-1 levels and reduced TAFI levels were associated not only with increased PE severity but also with 30-day mortality [22]. Similarly, published studies reported an association of elevated PAI-1 and TAFI with increased risk of venous thrombosis [53] and elevated D-dimer linked to disease severity but without prediction of long-term outcomes [54]. Advancing the field, our study identified clinically relevant threshold values associated with 30-day mortality as follows: tPA > 15 ng/mL, PAI-1 > 63 ng/mL, TAFI < 95%, and D-dimer > 13,688 ng/mL. While D-dimer and PAI-1 are well established and traditionally used in PE diagnosis, our findings suggest they may have an additional unrecognized prognostic value.

Inflammatory markers, specifically IL-6, were consistently increased with the severity of PE, indicating a direct correlation between a heightened systemic inflammatory response and disease progression. In contrast, CRP (a known marker of PE disease status) and TNF-α exhibited a variable pattern across severity. These results are consistent with our previous study, which reported varying degrees of change in IL-6 and TNF-α across PE severity [21]. A published study measured basal high-sensitivity CRP levels and found a significant increase in patients with spontaneous VTE compared with those in healthy individuals [44]. Notably, all 3 markers were significantly upregulated in 30-day nonsurvivors, underscoring their prognostic relevance. Importantly, we established optimal thresholds for predicting 30-day mortality as follows: CRP > 12 μg/mL, IL-6 > 48 pg/mL, and TNF-α > 2 pg/mL. These cutoff values reflect PE-related vascular dysfunction [21].

Hematological indices showed varying levels in relation to PE severity, suggesting a complex interplay between the immune response and inflammation. Elevated indices observed in 30-day nonsurvivors further support the interplay among inflammation, thrombosis, and vascular dysfunction in acute PE. Our results are aligned with previous studies: one study reported significant variations in NLR and PLR across PE severity, with notably higher levels in patients who died at 30 days, and it also reported only NLR as a predictor of all-cause mortality in PE [55]; another study also reported only NLR as a predictor of 30-day all-cause mortality in acute patients with PE [56]. Our novel finding of an optimal threshold of NLR > 7 as a predictor of 30-day mortality reinforces its potential as an early prognostic marker. This threshold aligns with our earlier studies, which not only proposed NLR > 7 as a significant predictor of 30-day mortality in patients with PE but also demonstrated that the levels varied with PE severity [45].

Statistical analysis by multiple programs demonstrated consistent data outcomes. However, the importance of the CHAID analysis identified NLR as the highest predictor of early mortality, followed by PAI-1 and TNF-α, adding to the uniqueness of our study. This ranking reflects the biological mechanism underlying PE severity: hematologic inflammatory imbalance (NLR) is the primary driver of early mortality, impaired fibrinolysis (PAI-1) contributes additional risk, and systemic cytokine activation (TNF-α) further amplifies disease severity. These findings are consistent with prior studies, highlighting the prognostic significance of inflammatory and fibrinolytic markers in PE. The dominance of NLR suggests that hematologic inflammatory imbalance may be an early therapeutic target, while PAI-1 and TNF-α provide additional stratification for high-risk patients. Overall, the tree structure and predictor hierarchy highlight that dysregulation of inflammation and fibrinolysis exert the strongest influence on early mortality, with cytokine elevations enhancing risk in susceptible subgroups.

Aside from the promising implications of our findings, several limitations of this study should be noted. (1) The relatively small sample size reduced the statistical power and limited the generalizability across different PE risk categories. (2) Specimen collection that was not timed precisely may have introduced variability in biomarker expression, particularly given the dynamic nature of thromboinflammatory responses in PE. (3) Ongoing clinical treatments, such as anticoagulation or anti-inflammatory therapies, may have confounded the biomarker levels, complicating the isolation of disease-specific signals. (4) Biomarker profiling was limited in scope, capturing only selected thromboinflammation and hemostatic activation markers, while emerging targets related to NETosis, oxidative stress, IL profiles, cytokine indices, complement, and lymphocyte subsets were underrepresented. (5) The absence of longitudinal sampling restricted assessment of temporal changes, which is essential for understanding disease progression, and therapeutic response.

The clinical implications of this study are significant, demonstrating that thromboinflammatory and hemostatic biomarkers can enhance risk assessment beyond traditional PE scoring systems. The identification of specific biomarker patterns with elevated VWF, TF, tPA, PAI-1, D-dimer, CRP, IL-6, TNF-α, and NLR, along with reduced FVII, FX, FXIIIA, and TAFI, provides more precise recognition of patients at risk for early mortality, while defined thresholds and biomarker cutoffs offer clinically relevant decision support to guide monitoring intensity, anticoagulation strategies, and overall therapeutic management. Overall, integrating these biomarkers into PE evaluation may strengthen risk stratification and support more personalized, biology-informed care.

The findings of the current study highlight the complexity of PE and underscore the need for broader, more integrative investigative approaches. Future studies should expand the scope of biomarker analysis beyond conventional markers to include immune and inflammatory mediators (cytokines and complement components) and innate immune-related processes (NETosis and pyroptosis), which play a central role in immunothrombosis and thrombus propagation. Investigating oxidative stress biomarkers that capture lipid, protein, and DNA oxidation along with overall antioxidant capacity further clarify the contribution of redox imbalance to thrombogenesis and clot persistence. Integrating microRNA profiling with circulating biomarkers may also enhance understanding of gene regulatory mechanisms underlying thrombosis and inflammation. Advances in molecular and cellular biology, including high-dimensional panel profiling, biosensor-based detection platforms, and cellular flow dynamics technologies, offer new opportunities to capture real-time thromboinflammatory and hemostatic processes with unprecedented resolution, thereby deepening the mechanistic insight into the pathogenesis of PE.

5. Conclusion

Taken together, these findings support a thromboinflammatory model of PE. Our results highlight the complex, multifaceted nature of PE, underscoring an interplay between endothelial injury, platelet activation, coagulation, fibrinolysis, systemic inflammation, and hematological indices as key biological mechanisms driving adverse clinical outcomes. The biomarkers in our study demonstrated an association with progression and prognosis of PE. This paradigm highlights the need to move beyond purely anatomical (troponin) or hemodynamic assessment (right ventricular diameter to left ventricular diameter ratio) toward a biology-informed approach to risk stratification.

While traditional clinical scores remain essential, incorporating quantifiable thromboinflammation and hemostatic activation biomarkers, such as VWF, E-selectin, TF, FVII, FX, FXIIIA, tPA, D-dimer, TAFI, CRP, and IL-6, and in particular NLR, TNF-α, and PAI-1, could enhance the precision of risk assessment. Integrating these markers into clinical practice could provide a more nuanced approach to PE severity diagnosis, improved prognostic accuracy, and facilitation of personalized therapeutic strategies.