Machine learning integration of multimodal data identifies key features of circulating NT-proBNP in people without cardiovascular diseases

Abstract

Abstract N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP) is important for diagnosing and predicting heart failure or many other diseases. However, few studies have comprehensively assessed the factors correlated with NT-proBNP levels in people with cardiovascular health. We used data from the 1999–2004 National Health and Nutrition Examination Survey (NHANES). Machine learning was employed to assess 66 factors that associated with NT-proBNP levels, including demographic, anthropometric, lifestyle, biochemical, blood, metabolic, and disease characteristics. The predictive power of the model was assessed using five-fold cross-validation. The optimal features predicting NT-proBNP levels were identified using univariate and step-forward multivariate models. Weighted least squares regression (WLS) was applied for supplementary analysis. Finally, the relationship between the corresponding features and NT-proBNP was validated using weighted and adjusted generalized additive models (GAM). We included 12, 526 participants without cardiovascular diseases. In the univariate model, age exhibited the highest association with NT-proBNP levels (the coefficient of determination (R2) = 36.91%). The multivariate models revealed that age, gender, red blood cell count, race/ethnicity, systolic blood pressure, and total protein level were the top six predictors of NT-proBNP. GAM demonstrated a noteworthy non-linear association between NT-proBNP and age, red blood cell count, systolic blood pressure, and total protein. Our study contributes to explaining the biological mechanisms of NT-proBNP and will facilitate the design of relevant cohort studies. We underscore the significance of assessing various population subgroups when employing NT-proBNP as a biomarker, and the need for developing innovative clinical algorithms to establish personalized levels.

Keywords

• Machine learning
• Circulating NT-proBNP
• National health and nutrition examination survey (NHANES)
• People without cardiovascular diseases
• Public health
• Personalized medicine