Background/Purpose: The most common cause of death among patients with systemic lupus erythematosus (SLE) is cardiovascular disease (CVD). Previous studies have performed cardiovascular risk stratification; however, the association by different ethnicities/races including Hispanics has rarely been evaluated. For this reason, we aim to use machine learning (ML) to identify features associated to CVD risks in patients with SLE among various ethnic groups.

Methods: The data was extracted from NIH All of Us Program. A total of 2,684 individuals with a diagnosis of SLE were selected as our SLE cohort. Demographic parameters, comorbidities, physical measures, other conditions, and observations were included as features (a total of 300 features). Major adverse cardiovascular events (MACE) were the class, defined as a composite outcome of developing acute myocardial infarction, stroke, or cardiovascular death. Individuals (samples) were divided and compared for the following groups: Hispanics (3.8%), Non-Hispanic-White-American (58.2%), Non-Hispanic-African-American (38.0%). The Recursive Feature Elimination algorithm and several ML algorithms (SVC [Support Vector Machine Classifier], GaussianNB [Naive Bayes Classifier], LogisticRegression, LinearDiscriminantAnalysis, RandomForestClassifier, and KNeighborsClassifier) were implemented. We obtained the most precise ML models for each group, along with the corresponding important features associated with MACE. The important features were computed through the permutation_importace function and the “Important Indexes” (IIs) were obtained.

Results: Out of the SLE cohort (n=2,684), 1,381 (51.6%) had a MACE event. Among Hispanics with SLE, pulmonary hypertension (0.0049), and pulmonary embolism (0.0047), and cardiomegaly (0.0069) exhibit stronger associations with CVD. In Afro-American patients, diabetes (0.0645), hyperlipidemia (0.0104), and heart failure (0.0300) are more closely linked to CVD risk. Conversely, among whites, hypertension (0.0076), chronic kidney disease (0.0131), and gastrointestinal disorders (0.0154) are predominant factors associated with CVD risk.

Conclusion: In this multiethnic cohort of patients with SLE, cardiovascular disease risk factors in patients with systemic lupus erythematosus (SLE) exhibit variations across different ethnic groups. Distinct patterns emerge for Hispanics with SLE, with pulmonary hypertension and pulmonary embolism being more prevalent when compared to non-Hispanics, which in turn is associated with more severe SLE disease and increased CVD risk. Overall, the varying profiles of CVD risk factors among different ethnic groups with SLE highlight the need for personalized approaches to CVD prevention and management in this patient population, taking into account ethnic-specific risk factor profiles and disease manifestations.

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/machine-learning-methods-to-predict-cardiovascular-risk-in-hispanic-patients-with-systemic-lupus-erythematosus/