Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome

Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome View ORCID Profile Ana Marco-Rico , View ORCID Profile Ihosvany Fernández-Bello , View ORCID Profile Jorge Mateo-Sotos , View ORCID Profile Pascual Marco-Vera doi: https://doi.org/10.1101/2025.03.11.25323564 Ana Marco-Rico 1 Hospital General Universitario Dr. Balmis-ISABIAL , Alicante, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ana Marco-Rico Ihosvany Fernández-Bello 1 Hospital General Universitario Dr. Balmis-ISABIAL , Alicante, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ihosvany Fernández-Bello For correspondence: ihosvanyf{at}yahoo.es Jorge Mateo-Sotos 2 Medical Analysis Expert Group, Universidad de Castilla-La Mancha , Cuenca, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jorge Mateo-Sotos Pascual Marco-Vera 1 Hospital General Universitario Dr. Balmis-ISABIAL , Alicante, Spain 3 Department of Clinical Medicine, Universidad Miguel Hernández , Alicante, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Pascual Marco-Vera Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Thrombotic Antiphospholipid Syndrome (TAPS) is an autoimmune disorder associated with a high risk of recurrent thromboembolic events. Despite advances in anticoagulation, predicting recurrence remains challenging, underscoring the need for more precise risk stratification to optimize personalized treatment. Traditional predictive models struggle to integrate the complexity of clinical and biochemical risk factors, creating an opportunity for Machine Learning (ML) to enhance prognostic accuracy. In this study, we evaluated the performance of the Extreme Gradient Boosting (XGBoost) model in predicting recurrent thrombotic events in TAPS, comparing it to Support Vector Machine, Decision Tree, Gaussian Naive Bayes, and K-Nearest Neighbors. Using demographic and clinical data, model performance was assessed through multiple metrics, including accuracy, recall, specificity, precision, Youden’s Index (DYI), F1 score, Matthews Correlation Coefficient (MCC), and the area under the receiver operating characteristic curve (AUC-ROC). XGBoost outperformed all other models, achieving an AUC-ROC of 0.91, an F1-score of 91.24, and an MCC of 80.98. Recall and accuracy exceeded 92.23% and 91.35%, respectively, demonstrating robust predictive capabilities. Key predictors identified included renal insufficiency, age, and lupus anticoagulants, reinforcing the clinical relevance of these factors in risk assessment. These findings highlight the potential of XGBoost to improve risk stratification and support clinical decision-making in TAPS. By identifying critical predictors, this approach may optimize anticoagulation strategies and enhance resource allocation. However, further validation in larger cohorts and prospective studies is necessary before clinical integration. Introduction Antiphospholipid Syndrome (APS) is an autoimmune disorder characterized by recurrent thrombotic events and obstetric morbidity caused by antiphospholipid antibodies (aPL), including lupus anticoagulant (LAC), anticardiolipin (aCL), and anti-β2 glycoprotein I antibodies (anti-β2GPI) 1 . APS is a heterogeneous condition with a wide spectrum of clinical manifestations, ranging from isolated thrombotic events to obstetric complications and, in severe cases, catastrophic presentations involving multi-organ failure 2 . The syndrome is classified into thrombotic APS (TAPS), characterized by venous, arterial, or microvascular thrombosis, and obstetric APS, involving complications such as recurrent fetal loss and placental insufficiency 1 . APS can occur as a primary condition or in the setting of autoimmune diseases like systemic lupus erythematosus (SLE) 3 . Thrombotic events in APS predominantly affect venous territories, with deep vein thrombosis of the lower limbs being the most common clinical presentation 4 . Pulmonary embolism, hepatic vein thrombosis, and ischemic strokes are also frequently observed 5 . The risk of recurrence is strongly influenced by the presence of aPL, being LAC the most pathogenic, followed by aCL and anti-β2GPI antibodies 6 – 8 . Patients with triple positivity for these antibodies have the highest risk of thrombosis 9 – 12 . Despite anticoagulant therapy, recurrence rates of thrombosis in patients with APS remain high. A study published in 2002 found that patients undergoing high-intensity anticoagulation therapy with an international normalized ratio (INR) target of 3.5 still experienced a recurrence rate of 9.1% patient-years, highlighting the persistent challenge of thrombotic events 13 . Later, in 2017, another study reported that recurrence rates in APS patients treated with anticoagulants reached 23.7%, compared to 37.2% in those on antiplatelet therapy and 6.9% in those receiving combined therapy, with an estimated 20% recurrence occurring within 3.4, 7.3, and 16.3 years, respectively 14 . More recently, in 2018, a meta-analysis of APS patients treated with direct oral anticoagulants (DOACs) reported a recurrence rate of 16%, with a mean duration of 12.5 months until the next thrombotic event, underscoring the limitations of DOACs compared to traditional anticoagulants 15 . Given these challenges, it is urgently needed to find new methodologies to achieve the most accurate possible prediction of thrombosis recurrence in this patient population. However, traditional statistical methods for predicting thrombotic recurrence in APS are limited by their inability to capture the multifactorial and non-linear nature of the syndrome 16 . The heterogeneity of APS, including variations in clinical presentation, antibody profiles, and response to therapy, highlights the inadequacy of conventional models for stratifying risk. Machine learning (ML) algorithms have emerged as transformative tools in medicine, offering predictive capabilities that surpass traditional methods by integrating complex, multidimensional datasets, including clinical, demographic, and laboratory variables 17 . Unlike conventional approaches, ML models are very promising for identifying intricate patterns and interactions, making them particularly well-suited for risk prediction in APS 18 . By addressing the limitations of conventional analysis, this study aims to assess whether ML algorithms can provide a robust framework for personalized thrombosis risk stratification in individuals with TAPS undergoing anticoagulation therapy. Among these algorithms, Extreme Gradient Boosting (XGBoost) has proven highly effective in integrating clinical data, identifying complex relationships, and improving predictive accuracy compared to traditional methods. XGBoost has been successfully applied across various medical domains, including cardiovascular event prediction 19 , early cancer detection 20 , and risk stratification in autoimmune diseases 21 . Material and Methods Study Design This retrospective study includes 72 patients with TAPS who were undergoing anticoagulation treatment and were followed at the Thrombosis and Hemostasis Unit of Hospital General Universitario Dr. Balmis in Alicante, Spain. Inclusion and Exclusion Criteria Patients aged 18 years or older with a diagnosis of TAPS, confirmed by clinical thrombotic events and persistent positive aPL in at least two samples taken 12 weeks apart, and who were receiving antithrombotic treatment, were eligible for the study. Patients with incomplete clinical data were excluded. Additional exclusion criteria comprised the use of intensive immunosuppressive therapies that could interfere with the interpretation of thrombotic recurrence, documented poor treatment adherence, discontinuation of anticoagulant therapy during the follow-up period, end-stage renal disease requiring dialysis, and active cancer. Clinical and Laboratory Data Variables were selected based on their known relevance to thrombotic risk in patients with TAPS. Data was extracted from an anonymized institutional database and included clinical, laboratory, and treatment-related variables documented during routine follow-up. Demographic characteristics (age and sex), common cardiovascular risk factors (including hypertension, diabetes mellitus, dyslipidemia, renal insufficiency, obesity, and smoking status), and pre-existing autoimmune conditions (such as SLE, inflammatory bowel disease, and rheumatoid arthritis) were recorded. Thrombosis-related variables encompassed the location of the first thrombotic event (venous, arterial, or small vessel) and thrombotic recurrence (the variable we aim to predict) during follow-up. Laboratory parameters included the titers of LAC, aCL (IgG and IgM), anti-β2GPI (IgG and IgM), and D-dimer levels. LAC was confirmed by dilute Russelĺs viper venom time (DRVVT) and the use of silica as a reagent sensitive to LAC derived from APTT (Diagnostica Stago, Paris, France) following manufactureŕs protocols. APL were analyzed by chemiluminescence (Werfen, Boston, USA) and D-dimer levels were determined by immunoprecipitation (Diagnostica Stago, Paris, France), in agreement with manufactureŕs instructions. Antithrombotic treatments were categorized into vitamin K antagonists (e.g., acenocoumarin, warfarin), low-molecular-weight heparin, DOACs, and antiplatelet agents. For patients on vitamin K antagonists, the last value of the International Normalized Ratio (INR) was recorded and dichotomized as 1 if INR > 60% of the upper limit of the normal range, otherwise 0. Predictive Machine Learning Modeling and Data Analysis In this study, we propose the use of the XGBoost model for predicting recurrent thrombotic events and compare its performance with other ML algorithms. To evaluate the effectiveness of the proposed XGBoost model, we compared it with four ML algorithms: Support Vector Machine (SVM), Decision Tree (DT), Gaussian Naive Bayes (GNB), and K-Nearest Neighbors (KNN). Continuous variables were standardized using z-scores, and features with high collinearity (r > 0.8) were excluded to avoid redundancy. Model performance was assessed using 5-fold cross-validation with principal metrics including the area under the receiver operating characteristic curve (AUC-ROC), F1-score (the harmonic mean of precision and recall), and MCC (Matthews Correlation Coefficient), as they provide a comprehensive assessment of predictive performance. AUC-ROC evaluates the model’s ability to differentiate between classes across various thresholds, making it particularly suitable for imbalanced data. F1-score balances precision and recall, offering a reliable metric for minimizing false positives and false negatives, while MCC considers all elements of the confusion matrix, providing a robust evaluation, especially in cases of class imbalance. Recall, accuracy, specificity, precision and DYI (Youden’s Index) were used as secondary metrics, as they can be influenced by class imbalance and may introduce bias. Metrics such as recall and precision, for instance, focus on specific aspects of model performance and may not reflect overall quality, while accuracy and specificity may overestimate performance in datasets where one class dominates. On the other hand, DYI adds value as a secondary metric to identify the most predictive model by balancing the trade-offs between sensitivity and specificity. While it may not provide as comprehensive an assessment as AUC-ROC, F1-score, or MCC, it complements these metrics by offering an additional perspective on model performance. By prioritizing MCC, F1-score, and AUC-ROC, we ensured a more balanced and reliable evaluation of model efficiency. This combination enabled a thorough and objective model comparison. Hyperparameter tuning was performed using grid search optimization to identify the best model configurations. All analyses were conducted using Python (version 3.9), leveraging libraries such as Scikit-learn for model implementation, Pandas and NumPy for preprocessing, and Matplotlib for visualization. To evaluate the association between categorical variables (e.g., presence of venous location [VL], thrombophilia, renal insufficiency [RI]) and thrombotic recurrence, contingency tables were created, and the Chi-squared test of independence was used. For small sample sizes or when assumptions of the Chi-squared test were not met, Fisher’s exact test was applied. A p-value < 0.05 was considered statistically significant. All analyses were performed using Python, or SPSS, with appropriate statistical libraries. Results All data used for the analysis in this study are available in supplemental material file Table S1, which includes the complete clinical characteristics of the patients. A total of 72 patients met the inclusion criteria, with a mean age of 60.5 ± 14.7 years, of whom 40% were male. Hypertension was present in 54.2% of patients, diabetes mellitus in 19.4%, and renal insufficiency in 15.3%. Laboratory results showed that 30.6% of patients were positive for LAC, 4.2% for aCL antibodies, and 26.4% for anti-β2GPI. Triple positivity was not observed. A comparison of clinical and demographic variables in patients with and without thrombosis recurrence is summarized in Table 1 . View this table: View inline View popup Download powerpoint Table 1. Comparison of clinical and demographic variables in patients with and without thrombosis recurrence. A comprehensive comparison of model performance metrics is presented in Table 2 . Among the ML algorithms evaluated, XGBoost demonstrated superior performance during the testing phase. The relative contributions of the study variables are illustrated in Figure 1 . The feature importance analysis of XGB identified renal insufficiency, LAC, VL, and age as the most significant predictors of thrombotic recurrence. XGBoost achieved an AUC-ROC of 0.9125, an F1-score of 91.24, and an MCC of 80.98, with recall and accuracy exceeding 92.23% and 91.35%, respectively, highlighting its robust predictive capability compared to other models Figures 2 - 4 . View this table: View inline View popup Download powerpoint Table 2. Performance metrics achieved during the training and testing phases of the machine learning models. Download figure Open in new tab Figure 1. Variable and weights assigned to each variable, indicating their relative contribution to the model. Download figure Open in new tab Figure 2. Radar Chart of machine learning algorithm performance during the training phase. XGB: Extreme Gradient Boosting. KNN: K-nearest neighbors. SVM: Support Vector Machine. DT: Decision Tree. GNB: Gaussian Naive Bayes Download figure Open in new tab Figure 3. Radar Chart of machine learning algorithm performance during the testing phase. XGB: Extreme Gradient Boosting. KNN: K-nearest neighbors. SVM: Support Vector Machine. DT: Decision Tree. GNB: Gaussian Naive Bayes Download figure Open in new tab Figure 4. A multi-model ROC curve comparing the AUC-ROC of all five algorithms. XGB: Extreme Gradient Boosting. KNN: K-nearest neighbors. SVM: Support Vector Machine. DT: Decision Tree. GNB: Gaussian Naive Bayes Discussion The findings of this study underscore the potential of ML, particularly XGBoost, in predicting recurrent thrombotic events in patients with TAPS. XGBoost outperformed traditional predictive models by effectively integrating clinical, demographic, and laboratory variables to generate accurate risk assessments. Its ability to handle heterogeneous data, detect complex interactions among multiple variables, and mitigate overfitting through regularization contributes to its superior predictive capabilities also observed in previous studies 22 – 24 . The identification of renal insufficiency, age, and LAC as key predictors aligns with existing clinical knowledge, reinforcing the understanding that renal dysfunction exacerbates endothelial damage and significantly contributes to thrombotic risk 25 . Age, a well-established risk factor, is associated with physiological changes such as arterial stiffening, endothelial dysfunction, and increased coagulability, all of which elevate thrombotic risk 26 . Additionally, aging is linked to enhanced activation of the coagulation cascade and elevated levels of pro-inflammatory biomarkers that promote thrombosis 27 . Older patients frequently present with comorbidities, including hypertension, diabetes, and dyslipidemia, which further contribute to vascular dysfunction and increase the likelihood of recurrent thrombotic events 28 . The presence of LAC has been identified as the most potent thrombotic risk factor in APS, significantly increasing the risk of arterial and venous thrombosis 12 . The presence of LAC in patients with TAPS has been associated with a significant increase in resistance to activated protein C (APC), a key mechanism in coagulation regulation 29 . This resistance decreases APC’s ability to inactivate factors Va and VIIIa, promoting a prothrombotic state due to higher levels of thrombin generation after coagulation activation. Excess thrombin subsequently activates thrombin-activatable fibrinolysis inhibitor, an enzyme that stabilizes the clot by removing plasminogen-binding sites on fibrin, thereby reducing plasminogen conversion into plasmin and blocking fibrinolysis 30 . In addition, enhanced thrombin generation also induces the production of plasminogen activator inhibitor-1 (PAI-1) 31 , a potent fibrinolysis suppressor that prevents the activation of plasminogen by tissue plasminogen activator (t-PA) 32 . Elevated PAI-1 levels further contribute to fibrin accumulation and persistent clot formation. Together, these mechanisms generate a prothrombotic state in TAPS patients, explaining the role of LAC as a key risk factor for arterial and venous thrombosis in this population. In addition to the presence of LAC, the presence of aCL and anti-β2GPI antibodies are contributing variables to our predictive model, which aligns with previous findings indicating that triple positivity has been strongly associated with a higher likelihood of recurrent thrombotic events in TAPS patients 33 . A significant observation in this study is that the INR did not exhibit predictive value for thrombotic recurrence. This may be attributed to the presence of confounding factors not captured by the model, and/or that INR alone is not a reliable predictor of thrombotic events 34 – 36 . While INR is a standard tool for monitoring anticoagulation with vitamin K antagonists, recent studies suggest that it does not always accurately reflect the procoagulant state in APS patients, which may explain its limited utility in our predictive model 37 , 38 . Limitations Although the sample size in this study may be considered a limitation, recent evidence suggests that ML models can achieve reliable performance even in small cohorts when appropriate validation techniques, such as cross-validation, are applied 39 . To further mitigate the impact of the limited sample size, this study employed data synthesis techniques, a strategy proven to enhance the performance of ML models in prior studies 40 . To improve predictive performance, future research should consider incorporating additional aPL, such as anti-phosphatidylserine/prothrombin and anti-annexin-V, which have demonstrated strong associations with thrombotic events in recent studies 41 , 42 . Additionally, platelet lipidomic may offer valuable insights into thrombotic risk stratification 43 , 44 . In addition, altogether using a larger, multicenter cohorts design could further improve the accuracy and applicability of the predictive models. From a clinical perspective, the adoption of ML tools like XGBoost has the potential to revolutionize patient care by enabling personalized risk stratification and optimizing anticoagulation strategies. However, for successful real-world implementation, user-friendly interfaces and integration into electronic health record systems will be essential to facilitate clinical decision-making. Conclusions This study demonstrates the potential of XGBoost in TAPS management, achieving an AUC-ROC of 0.91, an F1-score of 91.24, and an MCC of 80.98, with recall and accuracy exceeding 92.23% and 91.35%, respectively, highlighting its robust predictive capability compared to other models. Key predictors such as renal insufficiency, age, and LAC highlight the importance of integrating clinical and laboratory data for precise risk stratification. Integrating multi-omics data, such as platelet lipidomic, and additional aPL, like anti-phosphatidylserine/prothrombin and anti-annexin-V, could boost predictive accuracy by uncovering molecular interactions and pathways driving thrombotic risk in platelets, the main target of aPL antibodies. These approaches provide a comprehensive view of the biological processes underlying TAPS, enabling the identification of novel biomarkers and therapeutic targets, ultimately improving patient outcomes and resource allocation in TAPS care. Data Availability All data produced in the present study are available upon reasonable request to the authors Author Contributions Conceptualization, A.M.R., I.F.B. and P.M.V.; Methodology, A.M.R., I.F.B. and P.M.V.; Software, J.M.S.; Formal analysis, J.M.S.; Data curation, A.M.R., I.F.B., J.M.S and P.M.V.; Writing – original draft preparation, I.F.B.; Writing – review and editing, all authors; Visualization, J.M.S.; Supervision, I.F.B and P.M.V.; All authors have read and agreed to the published version of the manuscript. Availability of Data and Materials The data supporting the findings of this study are provided as supplemental material in File 1. Consent for Publication Written informed consent was obtained from patients or legal guardians before inclusion. Declaration of Conflicting Interests The authors declared that they had no interests that could be perceived as creating a conflict or bias. Ethics Approval and Consent to Participate The study was approved by the local ethical committee. Funding No funding was needed. Supplemental material Supplemental material for this article is available online in the file Table S1. Acknowledgements The authors thank Bayer Hispania, S.L. for its contribution to the creation of the Cátedra de Inteligencia Artificial , an academic initiative focused on Artificial Intelligence-driven data analysis, which carried out the data analysis for this study. Footnotes Emails Ana Marco Rico: annetamarco1{at}gmail.com , Jorge Mateo Sotos: jorge.mateo{at}uclm.es , Pascual Marco Vera: pascualmarco1851{at}gmail.com Abbreviations APS Antiphospholipid Syndrome aPL Antiphospholipid antibodies LAC Lupus anticoagulant aCL Anticardiolipin anti-β2GPI Anti-β2 glycoprotein I antibodies TAPS Thrombotic Antiphospholipid Syndrome SLE Systemic Lupus Erythematosus INR International Normalized Ratio DOACs Direct Oral Anticoagulants ML Machine Learning XGBoost Extreme Gradient Boosting SVM Support Vector Machine DT Decision Tree GNB Gaussian Naive Bayes KNN K-Nearest Neighbors DRVVT Dilute Russell’s viper venom time APTT Activated Partial Thromboplastin Time DYI Youden’s Index MCC Matthews Correlation Coefficient AUC-ROC Area Under the Receiver Operating Characteristic Curve VL Venous Location RI Renal Insufficiency APC Activated Protein C PAI-1 Plasminogen Activator Inhibitor-1 t PA Tissue Plasminogen Activator References 1. ↵ Miyakis S , Lockshin MD , Atsumi T , et al. International consensus statement on an update of the classification criteria for definite antiphospholipid syndrome (APS) . Journal of thrombosis and haemostasis . 2006 ; 4 ( 2 ): 295 – 306 . OpenUrl 2. ↵ Cervera R , Piette JC , Font J , et al. Antiphospholipid syndrome: clinical and immunologic manifestations and patterns of disease expression in a cohort of 1,000 patients . Arthritis & Rheumatism: Official Journal of the American College of Rheumatology . 2002 ; 46 ( 4 ): 1019 – 1027 . OpenUrl 3. ↵ Brooker SM , Fokas JA , Larson DN , Grebenciucova E . The Spectrum of Movement Disorders Associated with Systemic Lupus Erythematosus . Current neurology and neuroscience reports . 2025 ; 25 ( 1 ): 1 – 11 . OpenUrl 4. ↵ Kopydłowska AL , Wojtasińska E , Kurosz J , Zawilska K , Gil L . Arterial and Venous Thrombosis as a Symptom of Antiphospholipid Syndrome–A Noticeable Analysis from the Past . Medical Research Archives . 2021 ; 9 ( 3 ) 5. ↵ Waisayarat J , Plumworasawat S , Vilaiyuk S , Sirachainan N . Intracardiac thrombus in a patient with catastrophic antiphospholipid syndrome: an autopsy case report and review of the literature . Vascular health and risk management . 2019 : 253 – 258 . 6. ↵ Horbach DA , Oort EV , Donders RC , Derksen RH , de Groot PG . Lupus anticoagulant is the strongest risk factor for both venous and arterial thrombosis in patients with systemic lupus erythematosus . Thrombosis and haemostasis . 1996 ; 76 ( 12 ): 0916 – 0924 . OpenUrl 7. Forastiero R , Martinuzzo M , Pombo G , et al. A prospective study of antibodies to β2-glycoprotein I and prothrombin, and risk of thrombosis . Journal of Thrombosis and Haemostasis . 2005 ; 3 ( 6 ): 1231 – 1238 . OpenUrl 8. ↵ Pengo V , Tripodi A , Reber G , et al. Update of the guidelines for lupus anticoagulant detection . Journal of Thrombosis and Haemostasis . 2009 ; 7 ( 10 ): 1737 – 1740 . OpenUrl 9. ↵ Devreese K , Hoylaerts MF . Challenges in the diagnosis of the antiphospholipid syndrome . Clinical chemistry . 2010 ; 56 ( 6 ): 930 – 940 . OpenUrl Abstract / FREE Full Text 10. Pengo V , Ruffatti A , Tonello M , et al. Antibodies to Domain 4/5 (Dm4/5) of β2-Glycoprotein 1 (β2GP1) in different antiphospholipid (aPL) antibody profiles . Thrombosis research . 2015 ; 136 ( 1 ): 161 – 163 . OpenUrl CrossRef PubMed 11. Arachchillage DR , Laffan M . What is the appropriate anticoagulation strategy for thrombotic antiphospholipid syndrome? British journal of haematology . 2020 ; 189 ( 2 ): 216 – 227 . OpenUrl PubMed 12. ↵ Yin D , de Groot PG , Ninivaggi M , Devreese KM , de Laat B . Clinical relevance of isolated lupus anticoagulant positivity in patients with thrombotic antiphospholipid syndrome . Thrombosis and haemostasis . 2021 ; 121 ( 09 ): 1220 – 1227 . OpenUrl PubMed 13. ↵ Ruiz-Irastorza G , Khamashta MA , Hunt BJ , Escudero A , Cuadrado MJ , Hughes GR . Bleeding and recurrent thrombosis in definite antiphospholipid syndrome: analysis of a series of 66 patients treated with oral anticoagulation to a target international normalized ratio of 3.5 . Archives of Internal Medicine . 2002 ; 162 ( 10 ): 1164 – 1169 . OpenUrl CrossRef PubMed Web of Science 14. ↵ Jackson WG , Oromendia C , Unlu O , Erkan D , DeSancho MT . Recurrent thrombosis in patients with antiphospholipid antibodies and arterial thrombosis on antithrombotic therapy . Blood advances . 2017 ; 1 ( 25 ): 2320 – 2324 . OpenUrl Abstract / FREE Full Text 15. ↵ Dufrost V , Risse J , Reshetnyak T , et al. Increased risk of thrombosis in antiphospholipid syndrome patients treated with direct oral anticoagulants. Results from an international patient-level data meta-analysis . Autoimmunity reviews . 2018 ; 17 ( 10 ): 1011 – 1021 . OpenUrl PubMed 16. ↵ Zhao Y , Huang C , Qi W , et al. Validation of three prediction models for thrombosis recurrence in antiphospholipid syndrome patients based on a prospective cohort . RMD open . 2023 ; 9 ( 3 ): e003084 . OpenUrl Abstract / FREE Full Text 17. ↵ Topol EJ . High-performance medicine: the convergence of human and artificial intelligence . Nature medicine . 2019 ; 25 ( 1 ): 44 – 56 . OpenUrl CrossRef PubMed 18. ↵ Chen JH , Asch SM . Machine learning and prediction in medicine—beyond the peak of inflated expectations . The New England journal of medicine . 2017 ; 376 ( 26 ): 2507 . OpenUrl CrossRef PubMed 19. ↵ Dong T , Oronti IB , Sinha S , et al. Enhancing Cardiovascular Risk Prediction: Development of an Advanced Xgboost Model with Hospital-Level Random Effects . Bioengineering . 2024 ; 11 ( 10 ): 1039 . OpenUrl PubMed 20. ↵ Abuelmakarem HS , Majdy A , Maher G , Khaled H , Emad M , Asem Shaker E . Precancer Detection Based on Mutations in Codons 248 and 249 Using Decision Tree (DT) and XGBoost Deep Learning Model . International Journal of Industry and Sustainable Development . 2025 ; 6 ( 1 ): 67 – 77 . OpenUrl 21. ↵ Chen Y , Huang S , Chen T , et al. Machine learning for prediction and risk stratification of lupus nephritis renal flare . American Journal of Nephrology . 2021 ; 52 ( 2 ): 152 – 160 . OpenUrl CrossRef PubMed 22. ↵ Wardhani KDK , Akbar M . Diabetes Risk Prediction Using Extreme Gradient Boosting (XGBoost) . Jurnal Online Informatika . 2022 ; 7 ( 2 ): 244 – 250 . OpenUrl 23. Yang JC . The prediction and analysis of heart disease using XGBoost algorithm . Applied and Computational Engineering . 2024 ; 41 : 61 – 68 . OpenUrl 24. ↵ Rani R , Gill KS , Upadhyay D , Devliyal S . XGBoost-Driven Insights: Enhancing Chronic Kidney Disease Detection . IEEE ; 2024 : 1131 – 1134 . 25. ↵ Karandyszowska N , Alagündüz H , Widaeus J , et al. POS1512 THROMBOCYTOPENIA AND IMPAIRED KIDNEY FUNCTION ARE MAJOR RISK FACTORS FOR RECURRENT THROMBOSIS IN PATIENTS WITH ANTIPHOSPHOLIPID SYNDROME . BMJ Publishing Group Ltd ; 2023 . 26. ↵ Hooglugt A , Klatt O , Huveneers S . Vascular stiffening and endothelial dysfunction in atherosclerosis . Current opinion in lipidology . 2022 ; 33 ( 6 ): 353 – 363 . OpenUrl CrossRef PubMed 27. ↵ Dugué P-A , Hodge AM , Ulvik A , et al. Association of markers of inflammation, the kynurenine pathway and B vitamins with age and mortality, and a signature of inflammaging . The Journals of Gerontology: Series A . 2022 ; 77 ( 4 ): 826 – 836 . OpenUrl 28. ↵ Li A , Yan J , Zhao Y , et al. Vascular aging: assessment and intervention . Clinical Interventions in Aging . 2023 : 1373 – 1395 . 29. ↵ Pengo V . Interaction between antiphospholipid antibodies and protein C anticoagulant pathway: a narrative review . Thieme Medical Publishers, Inc.; 2022 : 971 – 977 . 30. ↵ Sillen M , Declerck PJ . Thrombin activatable fibrinolysis inhibitor (TAFI): an updated narrative review . International journal of molecular sciences . 2021 ; 22 ( 7 ): 3670 . OpenUrl PubMed 31. ↵ Huebner BR , Moore EE , Moore HB , et al. Thrombin stimulates increased plasminogen activator inhibitor-1 release from liver compared to lung endothelium . Journal of Surgical Research . 2018 ; 225 : 1 – 5 . OpenUrl PubMed 32. ↵ Milenković J , Miljković E , Milenković K , Bojanić N . PLASMINOGEN ACTIVATOR INHIBITOR 1 (PAI-1) AS A POTENTIAL DIAGNOSTIC AND THERAPEUTIC TARGET . Acta Medica Medianae . 2018 ; 57 (2) 33. ↵ Laurent C , Ricard L , Nguyen Y , et al. Triple positive profile in antiphospholipid syndrome: prognosis, relapse and management from a retrospective multicentre study . RMD open . 2023 ; 9 ( 1 ): e002534 . OpenUrl Abstract / FREE Full Text 34. ↵ Isert M , Miesbach W , Schüttfort G , et al. Monitoring anticoagulant therapy with vitamin K antagonists in patients with antiphospholipid syndrome . Annals of hematology . 2015 ; 94 : 1291 – 1299 . OpenUrl PubMed 35. Pokorney SD , Holmes DN , Thomas L , et al. Association between warfarin control metrics and atrial fibrillation outcomes in the outcomes registry for better informed treatment of atrial fibrillation . JAMA cardiology . 2019 ; 4 ( 8 ): 756 – 764 . OpenUrl PubMed 36. ↵ Afzal A , Gage BF , Luo S , Schoen MW , Sanfilippo KM . Elevated International Normalized Ratio (INR) Did Not Prevent Against Thrombosis in Patients with Chronic Liver Disease (CLD): A Population Based Study . Blood . 2020 ; 136 : 4 – 5 . OpenUrl CrossRef PubMed 37. ↵ Cohen H , Efthymiou M , Devreese KM . Monitoring of anticoagulation in thrombotic antiphospholipid syndrome . Journal of Thrombosis and Haemostasis . 2021 ; 19 ( 4 ): 892 – 908 . OpenUrl 38. ↵ Marco-Rico A , Marco-Vera P. Thrombotic antiphospholipid syndrome and direct oral anticoagulants: unmet needs and review of the literature . Thieme Medical Publishers, Inc .; 2023 : 736 – 743 . 39. ↵ Larracy R , Phinyomark A , Scheme E . Machine learning model validation for early stage studies with small sample sizes . IEEE ; 2021 : 2314 – 2319 . 40. ↵ Yang W , Nam W . Data synthesis method preserving correlation of features . Pattern Recognition . 2022 ; 122 : 108241 . 41. ↵ Egri N , Bentow C , Rubio L , et al. Anti-phosphatidylserine/prothrombin antibodies at two points: correlation with lupus anticoagulant and thrombotic risk . Frontiers in immunology . 2021 ; 12 : 754469 . OpenUrl PubMed 42. ↵ Roselli D , Bonifacio MA , Barbuti G , Rossiello MR , Ranieri P , Mariggiò MA . Anti-phosphatidylserine, anti-prothrombin, and anti-annexin V autoantibodies in antiphospholipid syndrome: a real-life study . Diagnostics . 2023 ; 13 ( 15 ): 2507 . OpenUrl PubMed 43. ↵ Tohidi-Esfahani I , Mittal P , Isenberg D , Cohen H , Efthymiou M . Platelets and Thrombotic Antiphospholipid Syndrome . Journal of Clinical Medicine . 2024 ; 13 ( 3 ): 741 . OpenUrl PubMed 44. ↵ Kulkarni PP , Shim YJ , Hubben AK , et al. Circulating Platelets in Thrombotic Antiphospholipid Syndrome Display High Procoagulant Activity and Surface Complement Deposition: Correlation with Thrombotic Risk . Blood . 2023 ; 142 : 5407 . OpenUrl View the discussion thread. Back to top Previous Next Posted March 13, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome Ana Marco-Rico , Ihosvany Fernández-Bello , Jorge Mateo-Sotos , Pascual Marco-Vera medRxiv 2025.03.11.25323564; doi: https://doi.org/10.1101/2025.03.11.25323564 Share This Article: Copy Citation Tools Machine Learning for Predicting Thrombotic Recurrence in Antiphospholipid Syndrome Ana Marco-Rico , Ihosvany Fernández-Bello , Jorge Mateo-Sotos , Pascual Marco-Vera medRxiv 2025.03.11.25323564; doi: https://doi.org/10.1101/2025.03.11.25323564 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Hematology Subject Areas All Articles Addiction Medicine (567) Allergy and Immunology (863) Anesthesia (296) Cardiovascular Medicine (4409) Dentistry and Oral Medicine (443) Dermatology (380) Emergency Medicine (606) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1505) Epidemiology (15202) Forensic Medicine (30) Gastroenterology (1119) Genetic and Genomic Medicine (6568) Geriatric Medicine (666) Health Economics (994) Health Informatics (4509) Health Policy (1365) Health Systems and Quality Improvement (1608) Hematology (537) HIV/AIDS (1262) Infectious Diseases (except HIV/AIDS) (15902) Intensive Care and Critical Care Medicine (1103) Medical Education (619) Medical Ethics (144) Nephrology (665) Neurology (6572) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1139) Occupational and Environmental Health (954) Oncology (3319) Ophthalmology (967) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (662) Pediatrics (1686) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5420) Public and Global Health (9203) Radiology and Imaging (2190) Rehabilitation Medicine and Physical Therapy (1367) Respiratory Medicine (1191) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (288) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fe6b7510b174807',t:'MTc3OTIzMTYwOA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();