Development and validation of a prediction model of left ventricular systolic dysfunction in type 2 diabetes mellitus.

Background: Left ventricular longitudinal myocardial systolic dysfunction (LVSD) represents a critical risk factor for diabetes-related cardiovascular events. This study aimed to develop a well-calibrated and convenient risk prediction model to investigate early predictive risk of LVSD in type 2 diabetes mellitus (T2DM) patients with preserved left ventricular ejection fraction (LVEF), and to evaluate its performance.

Methods: A total of 310 patients with T2DM from June 2020 to October 2021 at the Second Affiliated Hospital of Nanchang University were prospectively enrolled and randomly assigned to a training set (n=217) and a validation set (n=93) at a 7:3 ratio. Basic characteristics, laboratory tests, echocardiographic parameters, two-dimensional global longitudinal strain (GLS) parameters, and medication use were collected. LVSD in patients with T2DM with preserved LVEF was defined as an absolute value of GLS <18%. The least absolute shrinkage and selection operator (LASSO) regression was applied to optimize the screening variables, followed by multivariate logistic regression to identify independent risk factors for predicting LVSD, and a nomogram was established. The receiver operating characteristic (ROC) curves, area under the curve (AUC) values, calibration plot, and decision curve analysis (DCA) were used to verify and evaluate the nomogram's discrimination, calibration, and clinical validity.

Results: A total of 8 independent risk predictors of LVSD in T2DM were extracted and incorporated into the nomogram, as evaluated using LASSO regression analysis and multivariate logistic regression analysis, including body mass index (BMI), T2DM duration, blood urea nitrogen (BUN), left ventricular (LV) mass index, E/e', diabetic retinopathy, diabetic peripheral neuropathy, and diabetic nephropathy. The nomogram indicated excellent prediction properties with AUC values of 0.922 and 0.918 for the training set and validation set, respectively. Further, the predictive nomogram demonstrated outstanding consistency between the predicted probability and the actual probability in terms of the calibration plots. DCA showed also that the predicted nomogram was clinically beneficial.

Conclusions: This study identified independent risk factors for LVSD in patients with T2DM and developed a predictive nomogram. It allows for clinical decision-making to timely intervene or delay the occurrence of LVSD.