Machine learning-based investigation of the relationship between immune status and left ventricular hypertrophy in patients with end-stage kidney disease.

Abstract

Left ventricular hypertrophy (LVH) is the most frequent cardiac complication among end-stage kidney disease (ESKD) patients, which has been identified as predictive of adverse outcomes. Emerging evidence has suggested that immune system is implicated in the development of cardiac hypertrophy in multiple diseases. We applied machine learning models to exploring the relation between immune status and LVH in ESKD patients. A cohort of 506 eligible patients undergoing immune status assessment and standard echocardiography simultaneously in our center were retrospectively analyzed. The association between immune parameters and the occurrence of LVH were evaluated through univariate and multivariate logistic analysis. To develop a predictive model, we utilized four distinct modeling approaches: support vector machine (SVM), logistic regression (LR), multi-layer perceptron (MLP), and random forest (RF). In comparison to the non-LVH group, ESKD patients with LVH exhibited significantly impaired immune function, as indicated by lower cell counts of CD3+ T cells, CD4+ T cells, CD8+ T cells, and B cells. Additionally, multivariable Cox regression analysis revealed that a decrease in CD3+ T cell count was an independent risk factor for LVH, while a decrease in NK cell count was associated with the severity of LVH. The RF model demonstrated superior performance, with an average area under the curve (AUC) of 0.942. Our findings indicate a strong association between immune parameters and LVH in ESKD patients. Moreover, the RF model exhibits excellent predictive ability in identifying ESKD patients at risk of developing LVH. Based on these results, immunomodulation may represent a promising approach for preventing and treating this disease.