Machine Learning-Assisted Analysis of Sublingual Microcirculatory Dysfunction for Early Cardiovascular Risk Evaluation and Cardiovascular-Kidney-Metabolic Syndrome Stage in Patients With Type 2 Diabetes Mellitus.

Abstract

To examine whether sublingual microcirculation can be used as an effective and noninvasive method for assessing cardiovascular, kidney, and metabolic risks in patients with type 2 diabetes mellitus (T2DM). This cross-sectional observational study enrolled 186 patients with T2DM. All patients were evaluated using the Framingham General Cardiovascular Risk Score (FGCRS) and cardiovascular-kidney-metabolic (CKM) syndrome stage. Side-stream dark-field microscopy was used for sublingual microcirculation, including total and perfused vessel density (TVD and PVD). Multiple machine-learning prediction models have been developed for CKM risk and stage assessment in T2DM patients. Receiver operating characteristic (ROC) curves were generated to determine cutoff points. Compared to patients with T2DM, diabetic patients with subclinical atherosclerosis (SA) had a greater CV risk, as measured by the FGCRS, accompanied by markedly decreased microcirculation perfusion. Microcirculatory parameters (TVD and PVD), including carotid intima-media thickness (IMT), brachial-ankle pulse wave velocity (ba-PWV), and FGCRS, were closely associated with SA incidence. Microcirculatory parameters, Index (DMSA screen), and cut-off points were used to screen for SA in patients with T2DM. Furthermore, a new set of four factors identified through machine learning showed optimal sensitivity and specificity for detecting CKM risk in patients with T2DM. Decreased microcirculatory perfusion served as a useful early marker for CKM syndrome risk stratification in patients with T2DM without SA. Sublingual microcirculatory dysfunction is closely correlated with the risk of SA and CKM risk in T2DM patients. Sublingual microcirculation could be a novel tool for assessing the CKM syndrome stage in patients with T2DM.