A k-nearest neighbors machine learning model predicted 12-month changes in NT-proBNP in patients with pulmonary arterial hypertension, achieving a Pearson correlation of 0.63 and an R² of 0.40.
Observational (n=181)
No
Can machine learning algorithms using baseline clinical data predict 12-month changes in NT-proBNP in patients with pulmonary arterial hypertension?
Machine learning models, particularly the k-nearest neighbors algorithm, can predict 12-month changes in NT-proBNP levels in PAH patients, potentially enabling earlier therapeutic adjustments.
Effect estimate: r 0.63
Abstract Background Pulmonary arterial hypertension (PAH) is a progressive disease leading to right heart failure and reduced survival. NT-proBNP is a key biomarker in PAH, correlated with disease decompensation, hospitalization, and clinical outcomes. Despite its central role in risk stratification and therapeutic decision-making, longitudinal prediction of NT-proBNP changes has not been attempted. Purpose To evaluate machine learning (ML) algorithms for predicting 12-month changes in NT-proBNP serum levels based on baseline clinical data, aiming to support early intervention and preventive treatment strategies. Methods This retrospective pilot study analyzed 181 patients with PAH, defined according to current diagnostic guidelines. Ninety-two baseline variables were incorporated, including demographic, hemodynamic, echocardiographic, functional, and laboratory data. ML models tested included Lasso and ridge regression, k-nearest neighbors (k-NN), decision trees, random forest, and gradient boosting. Model training used an 80/20 train–test split and five-fold cross-validation. Predictive performance was assessed using root mean squared error (RMSE), mean absolute error (MAE), median absolute error (Median AE), coefficient of determination (R²), and Pearson correlation coefficient (r). Results The k-NN model demonstrated the best predictive performance, achieving an r of 0.63, an R² of 0.40, and the lowest RMSE (0.52). Other models showed lower correlations and higher error metrics. Conclusions Machine learning approaches can predict clinically meaningful 12-month changes in NT-proBNP with reasonable accuracy. The k-NN algorithm showed the best performance, suggesting potential for personalized monitoring and earlier therapeutic adjustment. Predictive modeling may enhance risk stratification and support preventive, data-driven care in PAH. External validation in larger multicenter cohorts is warranted to confirm these findings and guide clinical integration.
Kramer et al. (Mon,) conducted a observational in Pulmonary arterial hypertension (PAH) (n=181). Machine learning algorithms (k-NN) vs. Other machine learning models was evaluated on 12-month changes in NT-proBNP serum levels (r 0.63). A k-nearest neighbors machine learning model predicted 12-month changes in NT-proBNP in patients with pulmonary arterial hypertension, achieving a Pearson correlation of 0.63 and an R² of 0.40.