Phenotypic Identification and Risk Stratification of Venous Thromboembolism in Heart Failure: A Clustering Analysis Across Ejection Fraction Categories

Phenotypic Identification and Risk Stratification of Venous Thromboembolism in Heart Failure: A Clustering Analysis Across Ejection Fraction Categories | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Phenotypic Identification and Risk Stratification of Venous Thromboembolism in Heart Failure: A Clustering Analysis Across Ejection Fraction Categories hong Ran, zhiling luo, LuLu Zhang, XiaoLi Dong, RongHui Tang, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9064579/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Background Heart failure (HF) patients exhibit significant heterogeneity in venous thrombo-embolism (VTE) risk, which conventional scoring systems fail to adequately capture, and create a clinical dilemma in accurately predicting VTE risk in HF populations. This study aimed to derive data-driven phenotypes of HF across the full spectrum of left-ventricular ejection fraction (LVEF) and quantify their utility for VTE risk characterization and prediction. Methods In this retrospective cohort study, we enrolled 276 consecutive in-patients with HF, standardized clinical, echocardiographic, laboratory and biomarker data were extracted. After z-score normalization, K-means clustering was used to identify natural patient groupings; the optimal cluster number was selected by the elbow method and silhouette analysis. Cluster separation was visualized with principal-component analysis. Results Two pathophysiologically distinct phenotypes emerged: Cluster 1 (high VTE-risk, n = 127) displayed a pro-thrombotic signature hallmarked by heightened cardiac stress, volume overload and systemic inflammation; Cluster 2 (low VTE-risk, n = 149) was driven primarily by metabolic and anthropometric factors and maintained a relatively stable haemodynamic profile. VTE rates differed significantly between clusters: 63.0% (n = 87/127) vs 44.2% (n = 61/149), p < 0.01. Membership in Cluster 1 remained an independent predictor of incident VTE after adjustment for established risk markers, with the strongest discriminative effect observed in heart failure with mildly reduced ejection fraction (HFmrEF) and heart failure with preserved ejection fraction (HFpEF) sub-groups. Conclusion Unsupervised clustering uncovers two VTE-risk phenotypes that transcend conventional LVEF-based classification. Integration of these data-driven phenotypes into routine risk algorithms could enable personalized, phenotype-guided thromboprophylaxis for patients with HF. Clustering Analysis heart failure༛Venous thromboembolism phenotypic identification risk characterization Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Heart failure (HF) is a common clinical syndrome characterized by reduced heart pumping function and the prevalence is increasing annually. Venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), HF is associated with increased risk of long-term VTE. VTE not only increases morbidity and mortality but also leads to hospital readmission in 3.9% of patients with HF owing to thromboembolic events, significantly increasing the burden on healthcare resources [1–3] . Previous studies have mainly focused on risk factor analysis and prevention strategies, identifying advanced age, cardiac insufficiency, prolonged bed rest, diuretic use, and inadequate anticoagulant therapy as risk factors for VTE in patients with HF [4] . Other studies have constructed models to predict VTE based on traditional risk factors (e.g., the Wells and Caprini scores), However, such models are based on the general population or hospitalized patients and do not adequately consider the pathophysiological characteristics of patients with HF. The risk of VTE in patients with HF is still not well defined, appears to vary widely in clinical studies and cannot be accurately predicted [5] . Recent clustering-based studies have redefined HF beyond LVEF categories, demonstrating that data-driven phenogroups stratify outcomes more accurately than traditional classification [6] . These phenogroups also underscore the marked heterogeneity of VTE risk within HF, yet a validated VTE-risk phenotype is still lacking. This evidence gap complicates routine thromboprophylaxis decisions. In this study, we integrate unsupervised clustering with prospective VTE-risk stratification across the full LVEF spectrum. we derive data-driven HF phenotypes, quantify their VTE discriminative power, and outline how a phenotype-aware risk algorithm could personalise thromboprophylaxis for patients with HF. Methods Data sources This retrospective cohort study analyzed anonymized data from 375 consecutive patients diagnosed with HF at Yunnan Fuwai Cardiovascular Hospital between August 1, 2021, and August 1, 2024. After excluding 87 patients with incomplete clinical data and clinical treatment; 12 patients who lost contact in the later period, the final analysis included 276 HF patients. Data preprocessing steps including handling of missing values, outliers, and standardization. Variables with > 15% missing data were excluded and multiple imputation (m = 5) was performed using predictive mean matching with the 'mice' R package, all predictor variables underwent preprocessing and continuous variables were Z-score normalized. The study was approved by the Ethics Committee of Fuwai Yunnan Hospital (No. 2025-122-01), which waived the requirement for informed consent as the data were anonymized before analysis. Study design and participants Figure 1 shows the study workflow. After excluding 87 patients with incomplete clinical data and clinical treatment; 12 patients who lost contact in the later period, the final analysis included 276 HF patients. The inclusion criteria: (1) patients diagnosed with HF who were hospitalized for the management of symptomatic heart failure, The diagnosis was rigorously established using a multi-parameter framework adapted from contemporary heart failure guidelines [7] : Heart Failure with Reduced Ejection Fraction (HFrEF) was defined as LVEF ≤ 40%, Heart Failure with mildly Reduced Ejection Fraction (HFmrEF) was defined as LVEF was 41% -49%,Heart Failure with Preserved Ejection Fraction (HFpEF) was defined as LVEF ≥ 50%; (2) VTE events (deep vein thrombosis or pulmonary embolism) were confirmed by imaging. The exclusion criteria were: (1) previous lower-limb venous thrombosis or recurrent pulmonary embolism; (2) treated patients with concomitant VTE; (3) history of trauma, surgery, tumor, or acute infarction in the previous year; and (4) previous anticoagulation patients for atrial fibrillation or any reason; (5) incomplete clinical data or loss to follow-up. Selection of predictor variables: Clinical, demographic, echocardiographic, and laboratory data were collected at baseline. A total of 23 candidate variables included sex, age, height, weight, body mass index, left ventricular ejection fraction (LVEF), left ventricular end-diastolic velocity (LV EDV), total cholesterol level, high-density lipoprotein level, fibrinogen (Fib) level, prothrombin time (PT), D-dimer level, hemoglobin level, total protein (TP) level, neutrophil count (NEUT), lymphocyte count (LYMP), alkaline phosphatase level, heart rate, brain natriuretic peptide (BNP-pro) level, and New York Heart Association (NYHA) classification. These candidate predictors were selected based on their pathophysiological relevance and established associations with VTE pathogenesis in CHF. Clustering Analysis Variables were standardized (z-scores). K-means clustering was performed, and the optimal number of clusters was determined using the elbow method (within-cluster sum of squares) and silhouette width. Principal component analysis (PCA) visualized cluster separation. The tidyverse package was used to Compare the clinical characteristics of different clustering groups, The ggplot2 package was used to analysis VTE risk stratified by HF sub-type. Statistical Analysis For parameters with continuous variables, normal distribution was expressed as mean ± standard deviation, whereas those with skewed distributions were expressed as median [M] and interquartile range [Q25–Q75]. and compared using t-tests or Mann–Whitney U tests. Categorical variables were presented as n (%) and compared by chi-square tests. To evaluate the association between the identified clinical phenotypes and the risk of VTE, we performed univariable and multivariable logistic regression analyses. The occurrence of VTE (stata = 1) was set as the dependent variable. with stepwise adjustment for demographics (age, sex, height, weight, BMI) and cardiac biomarkers (BNP, LVEF). Predictive performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC). Analyses were performed using R (version 4.4.2) with a two-sided p < 0.05 considered significant. Results Characteristics of the study cohort: A total of 276 patients with HF were included in this study, and their baseline characteristics are shown in Table 1. The majority of patients in the cohort were male (64.5%), and their heart function was generally poor. More than half (58.7%) of the patients had HFrEF, and half (50.4%) of the patients had heart function in NYHA class III-IV. It is worth noting that the overall incidence of VTE is very high, reaching 53.6%. After standardizing key continuous variables such as age and BNP, it was found that the study cohort was close to the overall average level in these dimensions. Through unsupervised clustering analysis, Both the Elbow Method and the Silhouette Method consistently suggest that k = 2 is the optimal number of clusters for this dataset (Figure S1). the heterogeneous population was further divided into two clinical phenotypes with significant differences. Cluster Identification and Characteristics The heatmap highlights two distinct pathophysiological profiles: Cluster 1 shows notably elevated levels of D-dimer, Fib, PT, BNP-pro, LVEDV, and NEUT, suggesting a phenotype associated with hypercoagulability, cardiac stress, volume-overloaded and systemic inflammation. In contrast, Cluster 2 is characterized by higher values in anthropometric and metabolic indicators such as Weight, BMI, Age, Height, TC, and HDL, alongside lower LVEF. This suggests a phenotype leaning toward metabolic and anthropometric-driven, with relatively preserved stable status (Fig. 2). Visualization of the two identified clusters (Cluster 1 and Cluster 2) projected onto the first two principal components (Dim1: 17% variance explained; Dim2: 14.7% variance explained). Clear separation between clusters is observed in the reduced dimensional space, supporting the validity of the clustering solution (Fig. 3A). Stacked bar chart showing the proportion of each heart failure subtype (HFrEF, HFmrEF, and HFpEF) within the two clusters. Cluster1 contains a higher fraction of HFrEF patients (60.9% vs 56.5% (Fig. 3B). Bar plot comparing the overall VTE rate between clusters. Cluster 1 exhibits a markedly higher incidence than Cluster 2 (63.0% vs 44.2%, p < 0.01), confirming the high VTE-risk designation assigned in Table 1 (Fig. 3C). VTE rates stratified by HF subtype and cluster. across all three subtypes the highest column corresponds to Cluster 1, the absolute peak is seen in HFrEF patients within Cluster 1 (≈ 95%). The clustering algorithm therefore captures clinically relevant heterogeneity in VTE risk that is independent of, yet interactive with, traditional HF classification (Fig. 3D). The radar chart visually reinforces the two phenotypically distinct patient subgroups identified in the analysis: Cluster 1 (solid line) shows prominent peaks in BNP, LVEDV, PT, and LYMP, consistent with a profile of cardiac stress, volume overload, coagulation activity, and inflammatory involvement. Cluster 2 (dashed line) exhibits higher values in anthropometric and metabolic variables such as Weight, BMI, TC, and HDL, along with lower LVEF, with comparatively lower coagulation and inflammatory markers, supporting a phenotype influenced by metabolic factors and body composition. Variables such as Height, Age, AST, ALP, TP, and HGB show smaller differences between clusters, indicating these are less distinguishing features in this clustering solution (FigureS2). The parallel coordinate plot exhibits Cluster 1 consistently higher values in BNP-pro, LVEDV, PT, Fib, D-dimer, and NEUT, forming a profile characterized by cardiac stress, volume overload, hypercoagulability, and inflammation. Cluster 2 elevated values in anthropometric and metabolic measures such as Weight, BMI, TC, and HDL, with lower LVEF, supporting a metabolic and body composition-driven phenotype (Figure S3). VTE Incidence and Risk Stratification by Cluster and HF subtype: The overall incidence of VTE is 53.6%, Cluster 1 significantly higher than in Cluster 2 (63.0% vs 44.2%, p < 0.01), consistent with the high VTE-risk phenotype defined in Table 1. Within HFrEF: Cluster 1 93.5% vs Cluster 2 71.4%; OR = 1.00 (95% CI 0.28–3.60), indicating a large but statistically non-significant excess. Within HFmrEF: Cluster 1 43.8% vs Cluster 2 33.3%; OR = 0.32 (95% CI 0.08–1.22), a trend towards higher risk without reaching statistical significance. Within HFpEF: Cluster 1 exhibited the highest absolute incidence, with a statistically meaningful elevation compared with Cluster 2(Fig. 4). Association Between Clinical Phenotypes and VTE To evaluate the association between the identified clinical phenotypes and the risk of VTE, we performed univariable and multivariable logistic regression analyses. in the Model 1 (Unadjusted, Cluster only): Cluster 1 was associated with significantly increased odds of VTE (OR = 2.14, 95% CI 1.41–3.34, p = 0.001). in the Model 2, after controlling for demographic characteristics (age, sex, height, weight, and BMI) and sequential adjustment for potential confounders, this association remained robust (OR = 2.02,95% CI 1.28–3.29, p = 0.003). Most importantly, even after further adjustment for key indicators of heart failure severity-BNP and LVEF, Phenotype cluster membership remained an independent predictor of VTE (OR = 1.88, 95% CI 1.14–3.18, p = 0.012). This demonstrates that the VTE risk captured by the phenotypic classification extends beyond what is explained by conventional measures of cardiac dysfunction alone (Table 2). Receiver Operating Characteristic (ROC) Curves Comparing VTE Prediction Models, Model A (BNP only): AUC = 0.585; Model B (BNP + D-dimer): AUC = 0.697; improved discrimination compared to BNP alone; Model C (BNP + D-dimer + Cluster assignment): AUC = 0.722, revealed the highest discriminative performance among the three models (Figure S4). In our multivariable logistic regression Forest plot, several clinical and laboratory markers were significantly associated with the high VTE-risk phenotype (Cluster 1) after adjusting for age and sex. Higher BMI (OR = 1.18, 95% CI: 1.05–1.32), LVEDV (OR = 1.02 per ml, 95% CI: 1.01–1.03), BNP (OR = 1.45 per 1000 pg/mL, 95% CI: 1.22–1.72), PT (OR = 1.25 per second, 95% CI: 1.10–1.42), Fib (OR = 2.15 per g/L, 95% CI: 1.64–2.82), D-dimer (OR = 1.89 per mg/L, 95% CI: 1.35–2.64), and NEUT percentage (OR = 1.12 per %, 95% CI: 1.07–1.18) were independently associated with an increased likelihood of being classified as high VTE -risk. Conversely, higher LVEF (OR = 0.96 per %, 95% CI: 0.94–0.98), LYMP percentage (OR = 0.92 per %, 95% CI: 0.89–0.95), and hemoglobin level (OR = 0.97 per g/L, 95% CI: 0.95–0.99) were protective factors associated with the low VTE-risk phenotype (Fig. 5). Clinical outcomes 53.6% happened VTE events (n = 148/276), with significant differences observed by cluster: 63.0% (n = 87/127) vs 44.2% (n = 61/149, P < 0.01; clusters 1and clusters 2, respectively. Discussion Our study demonstrates that unsupervised machine learning successfully identify two clinically distinct phenotypes of heart failure patients with markedly different VTE risk profiles. The high VTE-risk phenotype (Cluster 1) was characterized by significantly worse cardiac function (lower LVEF, higher LVEDV, elevated BNP), hypercoagulability (prolonged PT, elevated fibrinogen and D-dimer), systemic inflammation (higher neutrophil percentage, lower lymphocyte percentage), and malnutrition status (lower hemoglobin and total protein). In contrast, the low VTE-risk phenotype (Cluster 2) exhibited a phenotype leaning toward metabolic and anthropometric-driven, with relatively preserved stable status. Overall VTE incidence was significantly higher in Cluster 1 than in Cluster 2. VTE-risk stratified analyses by heart failure subtype further elucidated this relationship: While VTE risk was universally high in the HFrEF subtype regardless of phenotype, Phenotype 1 identified patients at significantly elevated risk within the HFmrEF and HFpEF categories, subgroups in which ejection fraction-based risk assessment is inherently less discriminative, this exceeded the explanatory range of traditional indicators of heart failure severity. Accumulating evidence indicates that HF is closely associated with VTE, and the risk of VTE in HF patients is significantly higher than in the general population. Important gaps remain in our understanding of HF phenotypic heterogeneity, particularly among patients with HFpEF or HFmrEF [8–10] , who exhibit considerable diversity in pathophysiological mechanisms and clinical presentations [11] . In our study, phenotype1 characterized by simultaneous elevation of inflammatory, prothrombotic, and neurohormonal biomarkers, likely represents a state of maximal pathophysiological derangement and decompensation, This triad likely operates through a mutually reinforcing cycle: sustained systemic inflammation promotes endothelial activation and dysfunction, leading to increased expression of tissue factor and adhesion molecules, which in turn primes the coagulation cascade and promotes platelet aggregation [12–13] . Concurrently, venous stasis resulting from impaired cardiac output and elevated filling pressures further exacerbates prothrombotic tendencies by permitting localized accumulation of activated coagulation factors and inflammatory cells [14–15] . Additionally, hypoalbuminemia exacerbates hypercoagulability by altering anticoagulant protein synthesis [16] . This integrative pathophysiology aligns with the contemporary concept of the “inflammation–thrombosis nexus” in HF. Importantly, our findings extend prior observations by demonstrating that this “thromboinflammatory” phenotype is not merely a surrogate of HF severity but constitutes a distinct risk stratum that transcends traditional ejection fraction-based classification, particularly identifying high-risk individuals among HFmrEF and HFpEF patients in whom conventional risk assessment is less informative [17] . Compared with existing risk-assessment models (e.g., the Caprini and Wells scores), our phenotypic model incorporates HF-specific indicators together with systemic markers of inflammation and coagulation, thereby better reflecting the multisystem involvement in HF. Importantly, our findings assessment by highlighting the synergistic role of cardiac dysfunction (BNP, LVEF) and inflammatory-coagulation crosstalk (Fib, D-dimer, NEUT/LYMP ratio)—a dimension not fully captured by existing scores [18] . The proposed phenotypic model may provide a more tailored risk-stratification tool for VTE and other adverse events in HF, especially for identifying asymptomatic high-risk patients with underlying hypercoagulable and pro-inflammatory states [19] . Although Phenotype 1 was most prevalent in HFrEF patients, its presence across all ejection fraction categories and its consistent association with elevated VTE risk regardless of LVEF underscore the primacy of phenotypic profiling over systolic function alone in thrombotic risk assessment. This may explain why some HFpEF patients experience VTE events despite relatively preserved systolic function—a observation consistent with earlier studies reporting that both HFrEF and HFpEF increase the long-term risk of VTE approximately fivefold [20] . The identification of Phenotype 2 patients, who exhibit favorable biomarker profiles and lower VTE risk, suggests opportunities for risk-adapted thromboprophylaxis strategies. While current guidelines often recommend uniform anticoagulation based primarily on ejection fraction criteria, our data support a more personalized approach in which treatment intensity could be aligned with phenotypic risk. Several limitations merit consideration. First, the single-centre, retrospective design, modest sample size (n = 276) and the small number of patients in some subgroups (especially HFmrEF) may limit generalisability; validation in large, multicentre prospective cohorts is required. Second, the high observed VTE incidence (53.6%) suggests potential selection bias. Third, key pathophysiological variables (renal function, multi-organ injury, immobilisation duration) were unavailable, and their inclusion—together with novel biomarkers or imaging metrics—might further refine phenotypic discrimination. Finally, prospective confirmation of phenotype-guided risk stratification is essential before clinical implementation. Future VTE risk models for HF patients should consider incorporating metrics of cardiac function together with biomarkers of inflammation and coagulation to develop an HF-specific risk score. The prominent hypercoagulable and inflammatory profile in the high VTE-risk phenotype suggests that anticoagulant and potentially anti-inflammatory therapies might be particularly beneficial in this subgroup, warranting further interventional research. Moreover, risk stratification based on clustering and regression analysis could facilitate precision, phenotype-guided management, enabling tailored monitoring and treatment strategies for distinct phenotypic subgroups. For example, patients with Phenotype 1 might benefit from more aggressive anti-inflammatory strategies in addition to anticoagulation, whereas Phenotype 2 patients could derive greater benefit from metabolic interventions. Conclusion Our study identifies two clinically distinct phenotypes of heart failure patients with differential VTE risks that cut across traditional LVEF-based classifications. The incorporation of these phenotypes into VTE-risk score enables clinicians to better stratify patients and optimize treatment plans, Implementation of phenotype-based risk assessment could enable more personalized thromboprophylaxis in heart failure patients, potentially improving outcomes while minimizing bleeding risks in lower VTE-risk subgroups. Abbreviations HF, heart failure; HFrEF, heart failure with reduced ejection fraction; HFmrEF, heart failure with mildly reduced ejection fraction; HFpEF, heart failure with preserved ejection fraction; VTE, venous thromboembolism; ROC, receiver operating characteristic; LVEF, left ventricular ejection fraction; LVEDV, left ventricular end diastolic velocity; NYHA, New York Heart Association classification; BNP, brain natriuretic peptide; PT, prothrombin time; FIB, fibrinogen; NEUT, neutrophil granulocyte; LYMP, lymphocyte; TP, total protein. Declarations Acknowledgements Not applicable. Funding The study was supported by the Yunnan Provincial Clinical Medicine Research Special Program (Project Number: 202405AJ310003); Yunnan Basic Research Program (Project Number: 202401AY070001-019) Data availability statement The datasets used and/or analyzed in the current study are available upon request from the corresponding author. Ethical approval and consent to participate This study was approved by the Ethics Committee of Fuwai Yunnan Hospital, Chinese Academy of Medical Sciences (approval number: 2025-122-01) Consent for publication The Ethics Committee of Fuwai Yunnan Hospital (No. 2025-122-01) waived the requirement for additional consent as all data were de-identified and analyzed anonymously. Conflict of interests The author(s) of this work have nothing to disclose. Author contributions: Hong Ran, Writing- original draft; Yan Shen, review and editing; ZhiLing Luo, Supervision; XiaoLi Dong, Methodology; LuLu Zhang, Project administration; RongHui Tang, Formal analysis; Ting Jiang, YaNan Xie, Investigation References Rashedi 1BB, Pfeferman S et al. MB,. Venous Thromboembolism: Global Burden of Disease Estimates Are Missing a Common Cardiovascular Condition. J Am Coll Cardiol . 2025;86(22):2135–2138. doi: 10.1016/j.jacc.2025.08.078. PMID: 41298028. Xu T, Huang Y, Liu Z, et al. Heart Failure Is Associated with Increased Risk of Long-Term Venous Thromboembolism. Korean Circ J. 2021;51(9):766. 10.4070/kcj.2021.0213 . Yuan JL, Xiao WK, Zhang CQ, et al. Incidence and characteristic of deep venous thrombosis in hospitalized chronic heart failure patients. Heart Vessels. 2024;39(7):597–604. 10.1007/s00380-024-02377-7 . Bechlioulis A, Lakkas L, Rammos A, et al. Venous Thromboembolism in Patients with Heart Failure. CPD. 2022;28(7):512–20. 10.2174/1381612827666210830102419 . Ng TM, Tsai F, Khatri N, et al. Venous thromboembolism in hospitalized patients with heart failure: incidence, prognosis, and prevention. Circ Heart Fail. 2010;3(1):165–73. 10.1161/CIRCHEARTFAILURE.109.892349 . Locatelli G, Iovino P, Pasta A, et al. Cluster analysis of heart failure patients based on their psychological and physical symptoms and predictive analysis of cluster membership. J Adv Nurs. 2024;80(4):1380–92. 10.1111/jan.15890 . Heidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA guideline for the management of heart failure: A report of the American college of cardiology/american heart association joint committee on clinical practice guidelines. Circulation. 2022;145(18). 10.1161/CIR.0000000000001063 . Gouda P, Alemayehu W, Rathwell S, et al. Clinical phenotypes of heart failure across the spectrum of ejection fraction: A cluster analysis. Curr Probl Cardiol. 2022;47(11):101337. 10.1016/j.cpcardiol.2022.101337 . Meijs C, Brugts JJ, Lund LH, et al. Identifying distinct clinical clusters in heart failure with mildly reduced ejection fraction. Int J Cardiol. 2023;386:83–90. 10.1016/j.ijcard.2023.05.024 . Watson C, Saaid H, Vedula V, et al. Venous Thromboembolism: Review of Clinical Challenges, Biology, Assessment, Treatment, and Modeling. Ann Biomed Eng. 2023. 10.1007/s10439-023-03390-z . Wiercioch W, Nieuwlaat R, Akl EA, et al. Methodology for the American Society of Hematology VTE guidelines: current best practice, innovations, and experiences. Blood Adv. 2020;4(10):2351–65. 10.1182/bloodadvances.2020001768 . Chaudhury P, Alvarez P, Michael M, et al. Incidence and Prognostic Implications of Readmissions Caused by Thrombotic Events After a Heart Failure Hospitalization. JAHA. 2022;11(10):e025342. 10.1161/JAHA.122.025342 . Ilieva R, Kalaydzhiev P, Slavchev B, et al. Clinical phenotypes of severe atrial cardiomyopathy and their outcome: A cluster analysis. IJC Heart Vasculature. 2025;58:101679. 10.1016/j.ijcha.2025.101679 . Karwath A, Bunting KV, Gill SK, et al. Redefining β-blocker response in heart failure patients with sinus rhythm and atrial fibrillation: A machine learning cluster analysis. Lancet. 2021;398(10309):1427–35. 10.1016/S0140-6736(21)01638-X . Karaban K, Słupik D, Reda A, et al. Coagulation disorders and thrombotic complications in heart failure with preserved ejection fraction. Curr Probl Cardiol. 2024;49(1):102127. 10.1016/j.cpcardiol.2023.102127 . Goldhaber SZ. Venous thromboembolism in heart failure patients: pathophysiology, predictability, prevention. J Am Coll Cardiol. 2020;75:159–62. Farmakis D, Chrysohoou C, Giamouzis G, et al. The management of atrial fibrillation in heart failure: an expert panel consensus. Heart Fail Rev. 2021;26(6):1345–58. 10.1007/s10741-020-09978-0 . Mebazaa A, Spiro TE, Büller HR, et al. Predicting the risk of venous thromboembolism in patients hospitalized with heart failure. Circulation. 2014;130(5):410–8. 10.1161/CIRCULATIONAHA.113.003126 . Huang SL, Xin HY, Wang XY, et al. Recent Advances on the Molecular Mechanism and Clinical Trials of Venous Thromboembolism. J Inflamm Res. 2023;16:6167–78. 10.2147/JIR.S439205 . Published 2023 Dec 14. Piccini TIJP, Mahaffey KW, et al. A Cluster Analysis of the Japanese Multicenter Outpatient Registry of Patients With Atrial Fibrillation. Am J Cadiol. 2019;124(6):871–8. https://doi.org/10.1016/j.amjcard.2019.05.071 . Tables Table 1: Baseline Characteristics and Clinical Profiles of Patients Stratified by Clustering Phenotypes Characteristic Total (n = 276) Cluster 1 (high-risk phenotype) Cluster 2 (low-risk phenotype) p value demographic Age (years) 64.8 ± 14.2 67.1 ± 12.5 62.9 ± 15.3 0.08 Sex Male n (%) 178 (64.5 %) 92 (66.7 %) 86 (62.3 %) 0.45 Anthropometric indicators Height(cm) 164.5 ± 8.7 163.1 ± 8.9 165.3 ± 8.4 0.015 Weight(kg) 63.5 ± 14.2 60.1 ± 13.5 65.6 ± 14.3 <0.001 BMI (kg /m²) 24.6 ± 4.1 25.2 ± 3.9 24.0 ± 4.2 0.02 Cardiac function LVEF (%) 39.6 ± 15.8 36.2 ± 14.1 42.7 ± 16.5 <0.01 LVEDV (ml) 221.6 ± 78.3 258.4 ± 82.1 199.1 ± 65.8 <0.001 BNP-pro (pg/mL) 1 865 (523–5 840 2 410 (710–7 930) 1 290 (420–3 560) <0.001 Inflammatory and coagulation markers PT (S) 12.8 ± 3.1 13.5 ± 3.6 12.4 ± 2.6 <0.001 Fib (g/L) 3.28 ± 1.12 3.85 ± 1.25 2.92 ± 0.86 <0.001 D-dimer (mg/L) 0.38 (0.11–1.57) 0.59 (0.18–2.83) 0.21 (0.08–0.74) <0.01 NEUT (%) 65.8 ± 12.4 72.5 ± 10.1 61.6 ± 11.9 <0.001 LYMP (%) 24.5 ± 10.2 19.8 ± 8.7 27.4 ± 10.1 <0.001 Metabolic and nutritional indicators HB (g/L) 138 ± 24 128 ± 21 144 ± 23 <0.001 TP (g/L) 66.5 ± 8.9 63.8 ± 9.2 68.2 ± 8.2 <0.001 TC (mmol/L) 4.23 ± 1.12 3.85 ± 1.05 4.46 ± 1.10 <0.001 HDL (mmol/L) 1.12 ± 0.45 0.95 ± 0.38 1.22 ± 0.46 <0.001 liver function index ALP (U/L) 94.6 ± 45.2 102.3 ± 50.1 89.8 ± 40.8 0.007 AST (U/L) 28.5 ± 22.1 32.1 ± 25.4 26.3 ± 19.3 0.012 HF category HFrEF 162 (58.7 %) 78 (56.5 %) 84 (60.9 %) <0.01 HFmrEF 37 (13.4 %) 25 (18.1 %) 12 (8.7 %) <0.01 HFpEF 77 (27.9 %) 35 (25.4 %) 42 (30.4 %) <0.01 VTE events 148 (53.6 %) 87 (63.0 %) 61 (44.2 %) <0.01 Table 2： Logistic regression for VTE occurrence (cross-sectional) Model OR (95% CI) p-value Model 1 (Cluster only) 2.14 (1.41–3.34) 0.001 Model 2 + Demographics | 2.02 (1.28–3.29) 0.003 Model 3 + BNP + LVEF 1.88 (1.14–3.18) 0.012 Additional Declarations No competing interests reported. Supplementary Files FigureS1S4.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 28 Apr, 2026 Reviewers agreed at journal 23 Apr, 2026 Reviewers invited by journal 21 Apr, 2026 Editor invited by journal 01 Apr, 2026 Editor assigned by journal 10 Mar, 2026 Submission checks completed at journal 10 Mar, 2026 First submitted to journal 08 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9064579","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":631215216,"identity":"c3736cda-f94d-43c9-8a59-a5a1ce1c9b1b","order_by":0,"name":"hong Ran","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"hong","middleName":"","lastName":"Ran","suffix":""},{"id":631215217,"identity":"1f222ea7-e97c-4276-bdfe-72f75a30b8b9","order_by":1,"name":"zhiling luo","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"zhiling","middleName":"","lastName":"luo","suffix":""},{"id":631215218,"identity":"2a99c843-85bf-49ca-a27e-757a4996fe22","order_by":2,"name":"LuLu Zhang","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"LuLu","middleName":"","lastName":"Zhang","suffix":""},{"id":631215220,"identity":"b27f63b9-71de-4dfc-b53e-b62de1679e86","order_by":3,"name":"XiaoLi Dong","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"XiaoLi","middleName":"","lastName":"Dong","suffix":""},{"id":631215224,"identity":"34fcba1b-a83d-4296-a9e2-226b252bbb1d","order_by":4,"name":"RongHui Tang","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"RongHui","middleName":"","lastName":"Tang","suffix":""},{"id":631215226,"identity":"bf30bec5-631b-4f7e-a874-dfed63c0d7d4","order_by":5,"name":"Ting Jiang","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ting","middleName":"","lastName":"Jiang","suffix":""},{"id":631215230,"identity":"e2c97190-8403-47ef-a507-b7e26ecdafeb","order_by":6,"name":"YaNan Xie","email":"","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"YaNan","middleName":"","lastName":"Xie","suffix":""},{"id":631215231,"identity":"7216ae15-ea9d-4c3e-bb41-424fd5e09381","order_by":7,"name":"Yan Shen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYDACCQglB6WZiddiTLqWxBlEa5Gf3XzsMU/NnfSZ006nSTBUWCc2sJ89gFeLwZ1j6cY8x57lzpbO3SbBcCY9sYEnLwG/FokcM2ketsO580BaGNsOJzZI8Bjgd9iM/G/SPP8Op8uBtfwjQgvDjRw2ad62wwnSYC0NRGgxuJFmJjm377DhzNm5my0SgB5r48kh5LDkZxJvvh2Wl7idu/HGhxpr2X72MwQcBgRMPDBWAhCzEVQPBIw/iFE1CkbBKBgFIxcAAHMWQgBXhuqeAAAAAElFTkSuQmCC","orcid":"","institution":"Chinese Academy of Medical Sciences, Affiliated Cardiovascular Hospital of Kunming Medical University","correspondingAuthor":true,"prefix":"","firstName":"Yan","middleName":"","lastName":"Shen","suffix":""}],"badges":[],"createdAt":"2026-03-08 13:39:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9064579/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9064579/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108492792,"identity":"f0d18361-3ab4-4047-8706-e27114f67c7c","added_by":"auto","created_at":"2026-05-05 09:58:39","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":142669,"visible":true,"origin":"","legend":"\u003cp\u003eThe study workflow\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/30d694ccee07b55ff6f701ee.jpg"},{"id":108493300,"identity":"1bad3b66-0caa-4533-8599-55c4095a008b","added_by":"auto","created_at":"2026-05-05 09:59:52","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":289420,"visible":true,"origin":"","legend":"\u003cp\u003ePrincipal Component Analysis and Clinical Characterization of Identified Clusters\u003c/p\u003e\n\u003cp\u003eThe two phenotypically distinct clusters derived from Principal component analysis show not only separation in feature space but also clinically relevant differences in heart failure subtype distribution and Venous thromboembolism risk.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/752db4416ad20a67c57b2db8.png"},{"id":108804285,"identity":"2eb4635a-19c1-4a6a-b216-a33da87b3aad","added_by":"auto","created_at":"2026-05-08 15:18:55","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":125171,"visible":true,"origin":"","legend":"\u003cp\u003eHeatmap of Standardized Clinical and Laboratory Characteristics Across Clusters\u003c/p\u003e\n\u003cp\u003eThe heatmap highlights two distinct pathophysiological profiles: Cluster 1: Inflammatory, hypercoagulable, and volume-overloaded phenotype; Cluster 2: Metabolic, anthropometric-driven phenotype with relatively preserved coagulation status.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/3e5c855a5725c9b8f91c2a40.png"},{"id":108385278,"identity":"3cb71cdd-eb50-49c0-986b-0e0b471535a3","added_by":"auto","created_at":"2026-05-04 06:02:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":152113,"visible":true,"origin":"","legend":"\u003cp\u003eVenous Thromboembolism Incidence and Risk Stratification by Cluster and Heart Failure Type\u003c/p\u003e\n\u003cp\u003eCluster assignment yields clinically relevant Venous thromboembolism stratification across all HF sub-types.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/9528bc1a456dc7352a381401.png"},{"id":108385280,"identity":"a07971d0-9590-438a-8648-0ee6b651b272","added_by":"auto","created_at":"2026-05-04 06:02:53","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":143489,"visible":true,"origin":"","legend":"\u003cp\u003eForest plot of multivariable logistic regression analysis for predicting high Venous thromboembolism risk phenotype in heart failure patients.\u003c/p\u003e\n\u003cp\u003eThese findings further highlight the integrative role of cardiac dysfunction, hypercoagulability, inflammation, and nutritional status in defining high-risk heart failure phenotypes.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/5f8d2552ccc00727d19683f0.png"},{"id":109206625,"identity":"1a5fd5a5-e45a-4080-a2a9-95e9be30aba6","added_by":"auto","created_at":"2026-05-13 15:14:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":936250,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/d220b023-317f-40b1-8b05-95c85932d588.pdf"},{"id":108385276,"identity":"34716f17-62cf-441a-8d0f-b85562370938","added_by":"auto","created_at":"2026-05-04 06:02:53","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":319279,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1S4.docx","url":"https://assets-eu.researchsquare.com/files/rs-9064579/v1/a1e15bfd2379803d0b87de4e.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Phenotypic Identification and Risk Stratification of Venous Thromboembolism in Heart Failure: A Clustering Analysis Across Ejection Fraction Categories","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHeart failure (HF) is a common clinical syndrome characterized by reduced heart pumping function and the prevalence is increasing annually. Venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), HF is associated with increased risk of long-term VTE. VTE not only increases morbidity and mortality but also leads to hospital readmission in 3.9% of patients with HF owing to thromboembolic events, significantly increasing the burden on healthcare resources \u003csup\u003e[1\u0026ndash;3]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003ePrevious studies have mainly focused on risk factor analysis and prevention strategies, identifying advanced age, cardiac insufficiency, prolonged bed rest, diuretic use, and inadequate anticoagulant therapy as risk factors for VTE in patients with HF \u003csup\u003e[4]\u003c/sup\u003e. Other studies have constructed models to predict VTE based on traditional risk factors (e.g., the Wells and Caprini scores), However, such models are based on the general population or hospitalized patients and do not adequately consider the pathophysiological characteristics of patients with HF. The risk of VTE in patients with HF is still not well defined, appears to vary widely in clinical studies and cannot be accurately predicted \u003csup\u003e[5]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eRecent clustering-based studies have redefined HF beyond LVEF categories, demonstrating that data-driven phenogroups stratify outcomes more accurately than traditional classification \u003csup\u003e[6]\u003c/sup\u003e. These phenogroups also underscore the marked heterogeneity of VTE risk within HF, yet a validated VTE-risk phenotype is still lacking. This evidence gap complicates routine thromboprophylaxis decisions.\u003c/p\u003e \u003cp\u003eIn this study, we integrate unsupervised clustering with prospective VTE-risk stratification across the full LVEF spectrum. we derive data-driven HF phenotypes, quantify their VTE discriminative power, and outline how a phenotype-aware risk algorithm could personalise thromboprophylaxis for patients with HF.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eData sources\u003c/h2\u003e \u003cp\u003eThis retrospective cohort study analyzed anonymized data from 375 consecutive patients diagnosed with HF at Yunnan Fuwai Cardiovascular Hospital between August 1, 2021, and August 1, 2024. After excluding 87 patients with incomplete clinical data and clinical treatment; 12 patients who lost contact in the later period, the final analysis included 276 HF patients. Data preprocessing steps including handling of missing values, outliers, and standardization. Variables with \u0026gt;\u0026thinsp;15% missing data were excluded and multiple imputation (m\u0026thinsp;=\u0026thinsp;5) was performed using predictive mean matching with the 'mice' R package, all predictor variables underwent preprocessing and continuous variables were Z-score normalized. The study was approved by the Ethics Committee of Fuwai Yunnan Hospital (No. 2025-122-01), which waived the requirement for informed consent as the data were anonymized before analysis.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStudy design and participants\u003c/h3\u003e\n\u003cp\u003eFigure 1 shows the study workflow. After excluding 87 patients with incomplete clinical data and clinical treatment; 12 patients who lost contact in the later period, the final analysis included 276 HF patients. The inclusion criteria: (1) patients diagnosed with HF who were hospitalized for the management of symptomatic heart failure, The diagnosis was rigorously established using a multi-parameter framework adapted from contemporary heart failure guidelines \u003csup\u003e[7]\u003c/sup\u003e: Heart Failure with Reduced Ejection Fraction (HFrEF) was defined as LVEF\u0026thinsp;\u0026le;\u0026thinsp;40%, Heart Failure with mildly Reduced Ejection Fraction (HFmrEF) was defined as LVEF was 41% -49%,Heart Failure with Preserved Ejection Fraction (HFpEF) was defined as LVEF\u0026thinsp;\u0026ge;\u0026thinsp;50%; (2) VTE events (deep vein thrombosis or pulmonary embolism) were confirmed by imaging. The exclusion criteria were: (1) previous lower-limb venous thrombosis or recurrent pulmonary embolism; (2) treated patients with concomitant VTE; (3) history of trauma, surgery, tumor, or acute infarction in the previous year; and (4) previous anticoagulation patients for atrial fibrillation or any reason; (5) incomplete clinical data or loss to follow-up.\u003c/p\u003e\n\u003ch3\u003eSelection of predictor variables:\u003c/h3\u003e\n\u003cp\u003eClinical, demographic, echocardiographic, and laboratory data were collected at baseline. A total of 23 candidate variables included sex, age, height, weight, body mass index, left ventricular ejection fraction (LVEF), left ventricular end-diastolic velocity (LV EDV), total cholesterol level, high-density lipoprotein level, fibrinogen (Fib) level, prothrombin time (PT), D-dimer level, hemoglobin level, total protein (TP) level, neutrophil count (NEUT), lymphocyte count (LYMP), alkaline phosphatase level, heart rate, brain natriuretic peptide (BNP-pro) level, and New York Heart Association (NYHA) classification. These candidate predictors were selected based on their pathophysiological relevance and established associations with VTE pathogenesis in CHF.\u003c/p\u003e\n\u003ch3\u003eClustering Analysis\u003c/h3\u003e\n\u003cp\u003eVariables were standardized (z-scores). K-means clustering was performed, and the optimal number of clusters was determined using the elbow method (within-cluster sum of squares) and silhouette width. Principal component analysis (PCA) visualized cluster separation. The \u003cem\u003etidyverse\u003c/em\u003e package was used to Compare the clinical characteristics of different clustering groups, The \u003cem\u003eggplot2\u003c/em\u003e package was used to analysis VTE risk stratified by HF sub-type.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eFor parameters with continuous variables, normal distribution was expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation, whereas those with skewed distributions were expressed as median [M] and interquartile range [Q25\u0026ndash;Q75]. and compared using t-tests or Mann\u0026ndash;Whitney U tests. Categorical variables were presented as n (%) and compared by chi-square tests. To evaluate the association between the identified clinical phenotypes and the risk of VTE, we performed univariable and multivariable logistic regression analyses. The occurrence of VTE (stata\u0026thinsp;=\u0026thinsp;1) was set as the dependent variable. with stepwise adjustment for demographics (age, sex, height, weight, BMI) and cardiac biomarkers (BNP, LVEF). Predictive performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC). Analyses were performed using R (version 4.4.2) with a two-sided p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 considered significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eCharacteristics of the study cohort:\u003c/h2\u003e \u003cp\u003eA total of 276 patients with HF were included in this study, and their baseline characteristics are shown in Table\u0026nbsp;1. The majority of patients in the cohort were male (64.5%), and their heart function was generally poor. More than half (58.7%) of the patients had HFrEF, and half (50.4%) of the patients had heart function in NYHA class III-IV. It is worth noting that the overall incidence of VTE is very high, reaching 53.6%. After standardizing key continuous variables such as age and BNP, it was found that the study cohort was close to the overall average level in these dimensions. Through unsupervised clustering analysis, Both the Elbow Method and the Silhouette Method consistently suggest that k\u0026thinsp;=\u0026thinsp;2 is the optimal number of clusters for this dataset (Figure S1). the heterogeneous population was further divided into two clinical phenotypes with significant differences.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCluster Identification and Characteristics\u003c/h3\u003e\n\u003cp\u003eThe heatmap highlights two distinct pathophysiological profiles: Cluster 1 shows notably elevated levels of D-dimer, Fib, PT, BNP-pro, LVEDV, and NEUT, suggesting a phenotype associated with hypercoagulability, cardiac stress, volume-overloaded and systemic inflammation. In contrast, Cluster 2 is characterized by higher values in anthropometric and metabolic indicators such as Weight, BMI, Age, Height, TC, and HDL, alongside lower LVEF. This suggests a phenotype leaning toward metabolic and anthropometric-driven, with relatively preserved stable status (Fig.\u0026nbsp;2).\u003c/p\u003e \u003cp\u003eVisualization of the two identified clusters (Cluster 1 and Cluster 2) projected onto the first two principal components (Dim1: 17% variance explained; Dim2: 14.7% variance explained). Clear separation between clusters is observed in the reduced dimensional space, supporting the validity of the clustering solution (Fig.\u0026nbsp;3A). Stacked bar chart showing the proportion of each heart failure subtype (HFrEF, HFmrEF, and HFpEF) within the two clusters. Cluster1 contains a higher fraction of HFrEF patients (60.9% vs 56.5% (Fig.\u0026nbsp;3B). Bar plot comparing the overall VTE rate between clusters. Cluster 1 exhibits a markedly higher incidence than Cluster 2 (63.0% vs 44.2%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), confirming the high VTE-risk designation assigned in Table\u0026nbsp;1 (Fig.\u0026nbsp;3C). VTE rates stratified by HF subtype and cluster. across all three subtypes the highest column corresponds to Cluster 1, the absolute peak is seen in HFrEF patients within Cluster 1 (\u0026asymp;\u0026thinsp;95%). The clustering algorithm therefore captures clinically relevant heterogeneity in VTE risk that is independent of, yet interactive with, traditional HF classification (Fig.\u0026nbsp;3D).\u003c/p\u003e \u003cp\u003eThe radar chart visually reinforces the two phenotypically distinct patient subgroups identified in the analysis: Cluster 1 (solid line) shows prominent peaks in BNP, LVEDV, PT, and LYMP, consistent with a profile of cardiac stress, volume overload, coagulation activity, and inflammatory involvement. Cluster 2 (dashed line) exhibits higher values in anthropometric and metabolic variables such as Weight, BMI, TC, and HDL, along with lower LVEF, with comparatively lower coagulation and inflammatory markers, supporting a phenotype influenced by metabolic factors and body composition. Variables such as Height, Age, AST, ALP, TP, and HGB show smaller differences between clusters, indicating these are less distinguishing features in this clustering solution (FigureS2).\u003c/p\u003e \u003cp\u003eThe parallel coordinate plot exhibits Cluster 1 consistently higher values in BNP-pro, LVEDV, PT, Fib, D-dimer, and NEUT, forming a profile characterized by cardiac stress, volume overload, hypercoagulability, and inflammation. Cluster 2 elevated values in anthropometric and metabolic measures such as Weight, BMI, TC, and HDL, with lower LVEF, supporting a metabolic and body composition-driven phenotype (Figure S3).\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eVTE Incidence and Risk Stratification by Cluster and HF subtype:\u003c/h2\u003e \u003cp\u003eThe overall incidence of VTE is 53.6%, Cluster 1 significantly higher than in Cluster 2 (63.0% vs 44.2%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), consistent with the high VTE-risk phenotype defined in Table\u0026nbsp;1. Within HFrEF: Cluster 1 93.5% vs Cluster 2 71.4%; OR\u0026thinsp;=\u0026thinsp;1.00 (95% CI 0.28\u0026ndash;3.60), indicating a large but statistically non-significant excess. Within HFmrEF: Cluster 1 43.8% vs Cluster 2 33.3%; OR\u0026thinsp;=\u0026thinsp;0.32 (95% CI 0.08\u0026ndash;1.22), a trend towards higher risk without reaching statistical significance. Within HFpEF: Cluster 1 exhibited the highest absolute incidence, with a statistically meaningful elevation compared with Cluster 2(Fig.\u0026nbsp;4).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eAssociation Between Clinical Phenotypes and VTE\u003c/h2\u003e \u003cp\u003eTo evaluate the association between the identified clinical phenotypes and the risk of VTE, we performed univariable and multivariable logistic regression analyses. in the Model 1 (Unadjusted, Cluster only): Cluster 1 was associated with significantly increased odds of VTE (OR\u0026thinsp;=\u0026thinsp;2.14, 95% CI 1.41\u0026ndash;3.34, p\u0026thinsp;=\u0026thinsp;0.001). in the Model 2, after controlling for demographic characteristics (age, sex, height, weight, and BMI) and sequential adjustment for potential confounders, this association remained robust (OR\u0026thinsp;=\u0026thinsp;2.02,95% CI 1.28\u0026ndash;3.29, p\u0026thinsp;=\u0026thinsp;0.003). Most importantly, even after further adjustment for key indicators of heart failure severity-BNP and LVEF, Phenotype cluster membership remained an independent predictor of VTE (OR\u0026thinsp;=\u0026thinsp;1.88, 95% CI 1.14\u0026ndash;3.18, p\u0026thinsp;=\u0026thinsp;0.012). This demonstrates that the VTE risk captured by the phenotypic classification extends beyond what is explained by conventional measures of cardiac dysfunction alone (Table\u0026nbsp;2).\u003c/p\u003e \u003cp\u003eReceiver Operating Characteristic (ROC) Curves Comparing VTE Prediction Models, Model A (BNP only): AUC\u0026thinsp;=\u0026thinsp;0.585; Model B (BNP\u0026thinsp;+\u0026thinsp;D-dimer): AUC\u0026thinsp;=\u0026thinsp;0.697; improved discrimination compared to BNP alone; Model C (BNP\u0026thinsp;+\u0026thinsp;D-dimer\u0026thinsp;+\u0026thinsp;Cluster assignment): AUC\u0026thinsp;=\u0026thinsp;0.722, revealed the highest discriminative performance among the three models (Figure S4).\u003c/p\u003e \u003cp\u003eIn our multivariable logistic regression Forest plot, several clinical and laboratory markers were significantly associated with the high VTE-risk phenotype (Cluster 1) after adjusting for age and sex. Higher BMI (OR\u0026thinsp;=\u0026thinsp;1.18, 95% CI: 1.05\u0026ndash;1.32), LVEDV (OR\u0026thinsp;=\u0026thinsp;1.02 per ml, 95% CI: 1.01\u0026ndash;1.03), BNP (OR\u0026thinsp;=\u0026thinsp;1.45 per 1000 pg/mL, 95% CI: 1.22\u0026ndash;1.72), PT (OR\u0026thinsp;=\u0026thinsp;1.25 per second, 95% CI: 1.10\u0026ndash;1.42), Fib (OR\u0026thinsp;=\u0026thinsp;2.15 per g/L, 95% CI: 1.64\u0026ndash;2.82), D-dimer (OR\u0026thinsp;=\u0026thinsp;1.89 per mg/L, 95% CI: 1.35\u0026ndash;2.64), and NEUT percentage (OR\u0026thinsp;=\u0026thinsp;1.12 per %, 95% CI: 1.07\u0026ndash;1.18) were independently associated with an increased likelihood of being classified as high VTE -risk. Conversely, higher LVEF (OR\u0026thinsp;=\u0026thinsp;0.96 per %, 95% CI: 0.94\u0026ndash;0.98), LYMP percentage (OR\u0026thinsp;=\u0026thinsp;0.92 per %, 95% CI: 0.89\u0026ndash;0.95), and hemoglobin level (OR\u0026thinsp;=\u0026thinsp;0.97 per g/L, 95% CI: 0.95\u0026ndash;0.99) were protective factors associated with the low VTE-risk phenotype (Fig.\u0026nbsp;5).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eClinical outcomes\u003c/h2\u003e \u003cp\u003e53.6% happened VTE events (n\u0026thinsp;=\u0026thinsp;148/276), with significant differences observed by cluster: 63.0% (n\u0026thinsp;=\u0026thinsp;87/127) vs 44.2% (n\u0026thinsp;=\u0026thinsp;61/149, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01; clusters 1and clusters 2, respectively.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study demonstrates that unsupervised machine learning successfully identify two clinically distinct phenotypes of heart failure patients with markedly different VTE risk profiles. The high VTE-risk phenotype (Cluster 1) was characterized by significantly worse cardiac function (lower LVEF, higher LVEDV, elevated BNP), hypercoagulability (prolonged PT, elevated fibrinogen and D-dimer), systemic inflammation (higher neutrophil percentage, lower lymphocyte percentage), and malnutrition status (lower hemoglobin and total protein). In contrast, the low VTE-risk phenotype (Cluster 2) exhibited a phenotype leaning toward metabolic and anthropometric-driven, with relatively preserved stable status. Overall VTE incidence was significantly higher in Cluster 1 than in Cluster 2. VTE-risk stratified analyses by heart failure subtype further elucidated this relationship: While VTE risk was universally high in the HFrEF subtype regardless of phenotype, Phenotype 1 identified patients at significantly elevated risk within the HFmrEF and HFpEF categories, subgroups in which ejection fraction-based risk assessment is inherently less discriminative, this exceeded the explanatory range of traditional indicators of heart failure severity.\u003c/p\u003e \u003cp\u003eAccumulating evidence indicates that HF is closely associated with VTE, and the risk of VTE in HF patients is significantly higher than in the general population. Important gaps remain in our understanding of HF phenotypic heterogeneity, particularly among patients with HFpEF or HFmrEF \u003csup\u003e[8\u0026ndash;10]\u003c/sup\u003e, who exhibit considerable diversity in pathophysiological mechanisms and clinical presentations \u003csup\u003e[11]\u003c/sup\u003e. In our study, phenotype1 characterized by simultaneous elevation of inflammatory, prothrombotic, and neurohormonal biomarkers, likely represents a state of maximal pathophysiological derangement and decompensation, This triad likely operates through a mutually reinforcing cycle: sustained systemic inflammation promotes endothelial activation and dysfunction, leading to increased expression of tissue factor and adhesion molecules, which in turn primes the coagulation cascade and promotes platelet aggregation \u003csup\u003e[12\u0026ndash;13]\u003c/sup\u003e. Concurrently, venous stasis resulting from impaired cardiac output and elevated filling pressures further exacerbates prothrombotic tendencies by permitting localized accumulation of activated coagulation factors and inflammatory cells \u003csup\u003e[14\u0026ndash;15]\u003c/sup\u003e. Additionally, hypoalbuminemia exacerbates hypercoagulability by altering anticoagulant protein synthesis \u003csup\u003e[16]\u003c/sup\u003e. This integrative pathophysiology aligns with the contemporary concept of the \u0026ldquo;inflammation\u0026ndash;thrombosis nexus\u0026rdquo; in HF. Importantly, our findings extend prior observations by demonstrating that this \u0026ldquo;thromboinflammatory\u0026rdquo; phenotype is not merely a surrogate of HF severity but constitutes a distinct risk stratum that transcends traditional ejection fraction-based classification, particularly identifying high-risk individuals among HFmrEF and HFpEF patients in whom conventional risk assessment is less informative \u003csup\u003e[17]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eCompared with existing risk-assessment models (e.g., the Caprini and Wells scores), our phenotypic model incorporates HF-specific indicators together with systemic markers of inflammation and coagulation, thereby better reflecting the multisystem involvement in HF. Importantly, our findings assessment by highlighting the synergistic role of cardiac dysfunction (BNP, LVEF) and inflammatory-coagulation crosstalk (Fib, D-dimer, NEUT/LYMP ratio)\u0026mdash;a dimension not fully captured by existing scores \u003csup\u003e[18]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe proposed phenotypic model may provide a more tailored risk-stratification tool for VTE and other adverse events in HF, especially for identifying asymptomatic high-risk patients with underlying hypercoagulable and pro-inflammatory states \u003csup\u003e[19]\u003c/sup\u003e. Although Phenotype 1 was most prevalent in HFrEF patients, its presence across all ejection fraction categories and its consistent association with elevated VTE risk regardless of LVEF underscore the primacy of phenotypic profiling over systolic function alone in thrombotic risk assessment. This may explain why some HFpEF patients experience VTE events despite relatively preserved systolic function\u0026mdash;a observation consistent with earlier studies reporting that both HFrEF and HFpEF increase the long-term risk of VTE approximately fivefold \u003csup\u003e[20]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe identification of Phenotype 2 patients, who exhibit favorable biomarker profiles and lower VTE risk, suggests opportunities for risk-adapted thromboprophylaxis strategies. While current guidelines often recommend uniform anticoagulation based primarily on ejection fraction criteria, our data support a more personalized approach in which treatment intensity could be aligned with phenotypic risk.\u003c/p\u003e \u003cp\u003eSeveral limitations merit consideration. First, the single-centre, retrospective design, modest sample size (n\u0026thinsp;=\u0026thinsp;276) and the small number of patients in some subgroups (especially HFmrEF) may limit generalisability; validation in large, multicentre prospective cohorts is required. Second, the high observed VTE incidence (53.6%) suggests potential selection bias. Third, key pathophysiological variables (renal function, multi-organ injury, immobilisation duration) were unavailable, and their inclusion\u0026mdash;together with novel biomarkers or imaging metrics\u0026mdash;might further refine phenotypic discrimination. Finally, prospective confirmation of phenotype-guided risk stratification is essential before clinical implementation.\u003c/p\u003e \u003cp\u003eFuture VTE risk models for HF patients should consider incorporating metrics of cardiac function together with biomarkers of inflammation and coagulation to develop an HF-specific risk score. The prominent hypercoagulable and inflammatory profile in the high VTE-risk phenotype suggests that anticoagulant and potentially anti-inflammatory therapies might be particularly beneficial in this subgroup, warranting further interventional research. Moreover, risk stratification based on clustering and regression analysis could facilitate precision, phenotype-guided management, enabling tailored monitoring and treatment strategies for distinct phenotypic subgroups. For example, patients with Phenotype 1 might benefit from more aggressive anti-inflammatory strategies in addition to anticoagulation, whereas Phenotype 2 patients could derive greater benefit from metabolic interventions.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur study identifies two clinically distinct phenotypes of heart failure patients with differential VTE risks that cut across traditional LVEF-based classifications. The incorporation of these phenotypes into VTE-risk score enables clinicians to better stratify patients and optimize treatment plans, Implementation of phenotype-based risk assessment could enable more personalized thromboprophylaxis in heart failure patients, potentially improving outcomes while minimizing bleeding risks in lower VTE-risk subgroups.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eHF, heart failure; HFrEF, heart failure with reduced ejection fraction; HFmrEF, heart failure with mildly reduced ejection fraction; HFpEF, heart failure with preserved ejection fraction; VTE, venous thromboembolism; ROC, receiver operating characteristic; LVEF, left ventricular ejection fraction; LVEDV, left ventricular end diastolic velocity; NYHA, New York Heart Association classification; BNP, brain natriuretic peptide; PT, prothrombin time; FIB, fibrinogen; NEUT, neutrophil granulocyte; LYMP, lymphocyte; TP, total protein.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was supported by the Yunnan Provincial Clinical Medicine Research Special Program (Project Number: 202405AJ310003); Yunnan Basic Research Program (Project Number: 202401AY070001-019)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed in the current study are available upon request from the corresponding author.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved\u0026nbsp;by the Ethics Committee of Fuwai Yunnan Hospital, Chinese Academy of Medical Sciences (approval number: 2025-122-01)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Ethics Committee of Fuwai Yunnan Hospital (No. 2025-122-01) waived the requirement for additional consent as all data were de-identified and analyzed anonymously.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author(s) of this work have nothing to disclose.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003econtributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHong Ran, Writing- original draft; Yan Shen, review and editing; ZhiLing Luo, Supervision;\u003c/p\u003e\n\u003cp\u003eXiaoLi Dong, Methodology; LuLu Zhang, Project administration; RongHui Tang, Formal analysis; Ting Jiang, YaNan Xie, Investigation\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRashedi 1BB, Pfeferman S et al. MB,. Venous Thromboembolism: Global Burden of Disease Estimates Are Missing a Common Cardiovascular Condition. \u003cem\u003eJ Am Coll Cardiol\u003c/em\u003e. 2025;86(22):2135\u0026ndash;2138. doi: 10.1016/j.jacc.2025.08.078. PMID: 41298028.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu T, Huang Y, Liu Z, et al. Heart Failure Is Associated with Increased Risk of Long-Term Venous Thromboembolism. Korean Circ J. 2021;51(9):766. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4070/kcj.2021.0213\u003c/span\u003e\u003cspan address=\"10.4070/kcj.2021.0213\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan JL, Xiao WK, Zhang CQ, et al. Incidence and characteristic of deep venous thrombosis in hospitalized chronic heart failure patients. Heart Vessels. 2024;39(7):597\u0026ndash;604. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00380-024-02377-7\u003c/span\u003e\u003cspan address=\"10.1007/s00380-024-02377-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBechlioulis A, Lakkas L, Rammos A, et al. Venous Thromboembolism in Patients with Heart Failure. CPD. 2022;28(7):512\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2174/1381612827666210830102419\u003c/span\u003e\u003cspan address=\"10.2174/1381612827666210830102419\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg TM, Tsai F, Khatri N, et al. Venous thromboembolism in hospitalized patients with heart failure: incidence, prognosis, and prevention. Circ Heart Fail. 2010;3(1):165\u0026ndash;73. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIRCHEARTFAILURE.109.892349\u003c/span\u003e\u003cspan address=\"10.1161/CIRCHEARTFAILURE.109.892349\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLocatelli G, Iovino P, Pasta A, et al. Cluster analysis of heart failure patients based on their psychological and physical symptoms and predictive analysis of cluster membership. J Adv Nurs. 2024;80(4):1380\u0026ndash;92. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/jan.15890\u003c/span\u003e\u003cspan address=\"10.1111/jan.15890\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA guideline for the management of heart failure: A report of the American college of cardiology/american heart association joint committee on clinical practice guidelines. Circulation. 2022;145(18). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIR.0000000000001063\u003c/span\u003e\u003cspan address=\"10.1161/CIR.0000000000001063\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGouda P, Alemayehu W, Rathwell S, et al. Clinical phenotypes of heart failure across the spectrum of ejection fraction: A cluster analysis. Curr Probl Cardiol. 2022;47(11):101337. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cpcardiol.2022.101337\u003c/span\u003e\u003cspan address=\"10.1016/j.cpcardiol.2022.101337\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeijs C, Brugts JJ, Lund LH, et al. Identifying distinct clinical clusters in heart failure with mildly reduced ejection fraction. Int J Cardiol. 2023;386:83\u0026ndash;90. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijcard.2023.05.024\u003c/span\u003e\u003cspan address=\"10.1016/j.ijcard.2023.05.024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWatson C, Saaid H, Vedula V, et al. Venous Thromboembolism: Review of Clinical Challenges, Biology, Assessment, Treatment, and Modeling. Ann Biomed Eng. 2023. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10439-023-03390-z\u003c/span\u003e\u003cspan address=\"10.1007/s10439-023-03390-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiercioch W, Nieuwlaat R, Akl EA, et al. Methodology for the American Society of Hematology VTE guidelines: current best practice, innovations, and experiences. Blood Adv. 2020;4(10):2351\u0026ndash;65. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/bloodadvances.2020001768\u003c/span\u003e\u003cspan address=\"10.1182/bloodadvances.2020001768\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChaudhury P, Alvarez P, Michael M, et al. Incidence and Prognostic Implications of Readmissions Caused by Thrombotic Events After a Heart Failure Hospitalization. JAHA. 2022;11(10):e025342. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/JAHA.122.025342\u003c/span\u003e\u003cspan address=\"10.1161/JAHA.122.025342\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIlieva R, Kalaydzhiev P, Slavchev B, et al. Clinical phenotypes of severe atrial cardiomyopathy and their outcome: A cluster analysis. IJC Heart Vasculature. 2025;58:101679. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijcha.2025.101679\u003c/span\u003e\u003cspan address=\"10.1016/j.ijcha.2025.101679\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarwath A, Bunting KV, Gill SK, et al. Redefining β-blocker response in heart failure patients with sinus rhythm and atrial fibrillation: A machine learning cluster analysis. Lancet. 2021;398(10309):1427\u0026ndash;35. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(21)01638-X\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(21)01638-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaraban K, Słupik D, Reda A, et al. Coagulation disorders and thrombotic complications in heart failure with preserved ejection fraction. Curr Probl Cardiol. 2024;49(1):102127. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cpcardiol.2023.102127\u003c/span\u003e\u003cspan address=\"10.1016/j.cpcardiol.2023.102127\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoldhaber SZ. Venous thromboembolism in heart failure patients: pathophysiology, predictability, prevention. J Am Coll Cardiol. 2020;75:159\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFarmakis D, Chrysohoou C, Giamouzis G, et al. The management of atrial fibrillation in heart failure: an expert panel consensus. Heart Fail Rev. 2021;26(6):1345\u0026ndash;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10741-020-09978-0\u003c/span\u003e\u003cspan address=\"10.1007/s10741-020-09978-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMebazaa A, Spiro TE, B\u0026uuml;ller HR, et al. Predicting the risk of venous thromboembolism in patients hospitalized with heart failure. Circulation. 2014;130(5):410\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIRCULATIONAHA.113.003126\u003c/span\u003e\u003cspan address=\"10.1161/CIRCULATIONAHA.113.003126\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang SL, Xin HY, Wang XY, et al. Recent Advances on the Molecular Mechanism and Clinical Trials of Venous Thromboembolism. J Inflamm Res. 2023;16:6167\u0026ndash;78. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/JIR.S439205\u003c/span\u003e\u003cspan address=\"10.2147/JIR.S439205\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Published 2023 Dec 14.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiccini TIJP, Mahaffey KW, et al. A Cluster Analysis of the Japanese Multicenter Outpatient Registry of Patients With Atrial Fibrillation. Am J Cadiol. 2019;124(6):871\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.amjcard.2019.05.071\u003c/span\u003e\u003cspan address=\"10.1016/j.amjcard.2019.05.071\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1: Baseline Characteristics and Clinical Profiles of Patients Stratified by Clustering Phenotypes\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"right\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCharacteristic\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal (n = 276)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCluster 1 (high-risk phenotype)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCluster 2 (low-risk phenotype)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003edemographic\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e64.8 \u0026plusmn; 14.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e67.1 \u0026plusmn; 12.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e62.9 \u0026plusmn; 15.3 \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eSex Male n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e178 (64.5 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e92 (66.7 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e86 (62.3 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAnthropometric indicators\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eHeight(cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e164.5 \u0026plusmn; 8.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e163.1 \u0026plusmn; 8.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e165.3 \u0026plusmn; 8.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e0.015\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eWeight(kg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e63.5 \u0026plusmn; 14.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e60.1 \u0026plusmn; 13.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e65.6 \u0026plusmn; 14.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eBMI (kg /m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e24.6 \u0026plusmn; 4.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e25.2 \u0026plusmn; 3.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e24.0 \u0026plusmn; 4.2 \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCardiac function\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eLVEF (%) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e39.6 \u0026plusmn; 15.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e36.2 \u0026plusmn; 14.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e42.7 \u0026plusmn; 16.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eLVEDV (ml)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e221.6 \u0026plusmn; 78.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e258.4 \u0026plusmn; 82.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e199.1 \u0026plusmn; 65.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eBNP-pro (pg/mL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e1 865 (523\u0026ndash;5 840\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e2 410 (710\u0026ndash;7 930) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e1 290 (420\u0026ndash;3 560)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 380px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInflammatory and coagulation markers\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003ePT (S)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e12.8 \u0026plusmn; 3.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e13.5 \u0026plusmn; 3.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e12.4 \u0026plusmn; 2.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eFib (g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e3.28 \u0026plusmn; 1.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e3.85 \u0026plusmn; 1.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e2.92 \u0026plusmn; 0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eD-dimer (mg/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.38 (0.11\u0026ndash;1.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e0.59 (0.18\u0026ndash;2.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e0.21 (0.08\u0026ndash;0.74)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eNEUT\u0026nbsp;(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e65.8 \u0026plusmn; 12.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e72.5 \u0026plusmn; 10.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e61.6 \u0026plusmn; 11.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eLYMP\u0026nbsp;(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e24.5 \u0026plusmn; 10.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e19.8 \u0026plusmn; 8.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e27.4 \u0026plusmn; 10.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 380px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMetabolic and nutritional indicators\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eHB (g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e138 \u0026plusmn; 24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e128 \u0026plusmn; 21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e144 \u0026plusmn; 23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eTP (g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e66.5 \u0026plusmn; 8.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e63.8 \u0026plusmn; 9.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e68.2 \u0026plusmn; 8.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eTC (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e4.23 \u0026plusmn; 1.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e3.85 \u0026plusmn; 1.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e4.46 \u0026plusmn; 1.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eHDL (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e1.12 \u0026plusmn; 0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e0.95 \u0026plusmn; 0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e1.22 \u0026plusmn; 0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eliver function index\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eALP (U/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e94.6 \u0026plusmn; 45.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e102.3 \u0026plusmn; 50.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e89.8 \u0026plusmn; 40.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 204px;\"\u003e\n \u003cp\u003eAST (U/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e28.5 \u0026plusmn; 22.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 241px;\"\u003e\n \u003cp\u003e32.1 \u0026plusmn; 25.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 251px;\"\u003e\n \u003cp\u003e26.3 \u0026plusmn; 19.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003e0.012\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHF category\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eHFrEF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e162 (58.7 %) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e78 (56.5 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e84 (60.9 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eHFmrEF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e37 (13.4 %) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e25 (18.1 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e12 (8.7 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eHFpEF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e77 (27.9 %) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e35 (25.4 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e42 (30.4 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVTE events \u0026nbsp; \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 176px;\"\u003e\n \u003cp\u003e148 (53.6 %)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 241px;\"\u003e\n \u003cp\u003e87 (63.0 %) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e61 (44.2 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 121px;\"\u003e\n \u003cp\u003e\u0026lt;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003c/p\u003e\u003cp\u003eTable 2： Logistic regression for VTE occurrence (cross-sectional)\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 198px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 217px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOR (95% CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 198px;\"\u003e\n \u003cp\u003eModel 1 (Cluster only)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 217px;\"\u003e\n \u003cp\u003e2.14 (1.41\u0026ndash;3.34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e0.001 \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 198px;\"\u003e\n \u003cp\u003eModel 2 + Demographics |\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 217px;\"\u003e\n \u003cp\u003e2.02 (1.28\u0026ndash;3.29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 198px;\"\u003e\n \u003cp\u003eModel 3 + BNP + LVEF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 217px;\"\u003e\n \u003cp\u003e1.88 (1.14\u0026ndash;3.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e0.012 \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-cardiovascular-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcar","sideBox":"Learn more about [BMC Cardiovascular Disorders](http://bmccardiovascdisord.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcar/default.aspx","title":"BMC Cardiovascular Disorders","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Clustering Analysis, heart failure༛Venous thromboembolism, phenotypic identification, risk characterization","lastPublishedDoi":"10.21203/rs.3.rs-9064579/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9064579/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eHeart failure (HF) patients exhibit significant heterogeneity in venous thrombo-embolism (VTE) risk, which conventional scoring systems fail to adequately capture, and create a clinical dilemma in accurately predicting VTE risk in HF populations. This study aimed to derive data-driven phenotypes of HF across the full spectrum of left-ventricular ejection fraction (LVEF) and quantify their utility for VTE risk characterization and prediction.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eIn this retrospective cohort study, we enrolled 276 consecutive in-patients with HF, standardized clinical, echocardiographic, laboratory and biomarker data were extracted. After z-score normalization, K-means clustering was used to identify natural patient groupings; the optimal cluster number was selected by the elbow method and silhouette analysis. Cluster separation was visualized with principal-component analysis.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eTwo pathophysiologically distinct phenotypes emerged: Cluster 1 (high VTE-risk, n\u0026thinsp;=\u0026thinsp;127) displayed a pro-thrombotic signature hallmarked by heightened cardiac stress, volume overload and systemic inflammation; Cluster 2 (low VTE-risk, n\u0026thinsp;=\u0026thinsp;149) was driven primarily by metabolic and anthropometric factors and maintained a relatively stable haemodynamic profile. VTE rates differed significantly between clusters: 63.0% (n\u0026thinsp;=\u0026thinsp;87/127) vs 44.2% (n\u0026thinsp;=\u0026thinsp;61/149), \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01. Membership in Cluster 1 remained an independent predictor of incident VTE after adjustment for established risk markers, with the strongest discriminative effect observed in heart failure with mildly reduced ejection fraction (HFmrEF) and heart failure with preserved ejection fraction (HFpEF) sub-groups.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eUnsupervised clustering uncovers two VTE-risk phenotypes that transcend conventional LVEF-based classification. Integration of these data-driven phenotypes into routine risk algorithms could enable personalized, phenotype-guided thromboprophylaxis for patients with HF.\u003c/p\u003e","manuscriptTitle":"Phenotypic Identification and Risk Stratification of Venous Thromboembolism in Heart Failure: A Clustering Analysis Across Ejection Fraction Categories","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 06:02:44","doi":"10.21203/rs.3.rs-9064579/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-28T15:56:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"207180316137890409702364100543881967553","date":"2026-04-23T12:53:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-21T09:16:11+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-01T14:11:21+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-10T05:41:25+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-10T05:40:27+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cardiovascular Disorders","date":"2026-03-08T13:24:19+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cardiovascular-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcar","sideBox":"Learn more about [BMC Cardiovascular Disorders](http://bmccardiovascdisord.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcar/default.aspx","title":"BMC Cardiovascular Disorders","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"82d9595f-16b7-4dd1-93eb-1428fb2ed293","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-04T06:02:45+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 06:02:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9064579","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9064579","identity":"rs-9064579","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}