{
  "abstract": "Objectives Patients with systemic lupus erythematosus (SLE) face a dramatically increased risk of cardiovascular and thromboembolic events yet predicting who will develop these complications remains challenging. Leveraging machine learning, we aimed to uncover the most influential drivers of six major adverse outcomes, providing actionable insights for personalized care.Methods We conducted a retrospective prediction study patients with systemic lupus erythematosus (SLE) followed at the Internal Medicine and Rheumatology Department, Dr. Ion Cantacuzino Clinical Hospital (January 2015–June 2019), who subsequently developed cardiovascular comorbidities. Random Forest models were used to identify baseline predictors of six cardiovascular outcomes: deep vein thrombosis, stroke, acute myocardial infarction, heart failure, ischemic coronary disease, and pulmonary embolism (see flowchart 1). Baseline variables were grouped into five predefined domains—demographics, cardiovascular risk factors, lupus-specific features, cardiovascular therapies, and SLE therapies—to enhance interpretability and reduce overfitting. Separate domain-specific models were trained to identify key predictors, and the top variable from each domain was combined into a final outcome-specific model. Analyses were conducted in R version 4.4.3 using the RandomForest package.Results Our analysis comprised 140 patients, and their baseline demographic and clinical profiles are detailed in table 1. Across outcomes, a consistent pattern emerged (figure 1): traditional cardiovascular risk factors - dyslipidemia, diabetes, metabolic syndrome, smoking - and markers of lupus severity - nephritis, anemia, hypocomplementemia - were dominant drivers. Age and disease duration contributed meaningfully, while therapy variables influenced risk only in certain outcomes—highlighting that treatment choices often reflect underlying patient risk rather than causing events directly. Focused models built on the most influential predictors explained much of the outcome variability, highlighting a complex yet clinically actionable risk profile shared across the six major adverse events.Abstract PO:03:092 Figures and TablesConclusions Our findings signal the paradigm: rather than treatment alone, cardiovascular risk management and close monitoring of lupus severity are the strongest levers to prevent adverse outcomes. This work provides a data-driven roadmap for personalized interventions, highlighting which patients truly need aggressive preventive strategies. Incorporating these insights into routine care may potentially reduce the burden of cardiovascular and thromboembolic events in SLE.",
  "authors": [
    {
      "affiliations": [
        "Dr I Cantacuzino Clinical Hospital – Department of Internal Medicine and Rheumatology, Bucharest, Romania"
      ],
      "name": "Alexandru Garaiman"
    },
    {
      "affiliations": [
        "Dr I Cantacuzino Clinical Hospital – Department of Internal Medicine and Rheumatology, Bucharest, Romania",
        "Carol Davila University of Medicine and Pharmacy – Faculty of Medicine, Bucharest, Romania"
      ],
      "name": "Cristina Alexandru"
    },
    {
      "affiliations": [
        "Dr I Cantacuzino Clinical Hospital – Department of Internal Medicine and Rheumatology, Bucharest, Romania",
        "Carol Davila University of Medicine and Pharmacy – Faculty of Medicine, Bucharest, Romania"
      ],
      "name": "Ion Ancuta"
    },
    {
      "affiliations": [
        "Dr I Cantacuzino Clinical Hospital – Department of Internal Medicine and Rheumatology, Bucharest, Romania",
        "Carol Davila University of Medicine and Pharmacy – Faculty of Medicine, Bucharest, Romania"
      ],
      "name": "Mihai Bojinca"
    },
    {
      "affiliations": [
        "Dr I Cantacuzino Clinical Hospital – Department of Internal Medicine and Rheumatology, Bucharest, Romania",
        "Carol Davila University of Medicine and Pharmacy – Faculty of Medicine, Bucharest, Romania"
      ],
      "name": "Anca Bobirca"
    }
  ],
  "title": "PO:03:092 Learning from data: cardiovascular risk in systemic lupus erythematosus under the machine learning lens",
  "uid": "1889728d-8aed-5601-8a6c-abffa64f114c"
}