Novel Model Predicts Type 2 Diabetes Mellitus Patients Complicated With Metabolic Syndrome Using Retrospective Dataset From First Affiliated Hospital of Shenzhen University, China.

Objective: Metabolic syndrome (MS) is the most important risk factor for Type 2 diabetes mellitus (T2DM) and cardiovascular disease. This study used a retrospective dataset from the First Affiliated Hospital of Shenzhen University and aimed to develop and validate a novel model nomogram based on clinical parameters to predict MS in patients with T2DM. Methods: A total of 2854 patients with T2DM between January 2014 and May 2022 were selected and divided into a training dataset (n = 2114) and a validation dataset (n = 740). This study used multivariate logistic regression analysis to develop a nomogram for predicting MS in patients with T2DM that included candidates selected in the LASSO regression model. The data were set standardized before LASSO regression. The area under the receiver operating characteristic curve (AUC-ROC) was used to assess discrimination in the prediction model. The calibration curve is used to evaluate the calibration of the calibration nomogram, and the clinical decision curve is used to determine the clinical utility of the calibration diagram. The validation dataset is used to evaluate the performance of predictive models. Results: A total of 2854 patients were eligible for this study. There were 1941 (68.01%) patients with MS. The training dataset included 20 potential risk factors of the patient's demographic, clinical, and laboratory indexes in the LASSO regression analysis. Gender, hypertension, BMI, WC, HbA1c, TG, LDL, and HDL were multivariate models. We obtained a model for estimating MS in patients with T2DM. The AUC-ROC of the training dataset in our model is 0.886, and the 95% CI is 0.871-0.901. Similar to the results obtained from the training dataset, the AUC-ROC of the validation dataset in our model is 0.859, and the 95% CI is 0.831-0.887, thus proving the robustness of the model. The prediction model is as follows: logit (MS) = -9.18209 + 0.14406 ∗ BMI (kg/m²) + 0.09218 ∗ WC (cm) + 1.05761 ∗ TG (mmol/L)-3.30013 ∗ HDL (mmol/L). The calibration plots of the predicted probabilities show excellent agreement with the observed MS rates. Decision curve analysis demonstrated that the new nomogram provided significant net benefits in clinical applications. Conclusion: The prediction model of this study covers four clinically easily obtained parameters: BMI, WC, TG, and HDL, and shows a high accuracy rate in the validation dataset. Our predictive model may provide an effective method for large-scale epidemiological studies of T2DM patients with MS and offer a practical tool for the early detection of MS in clinical work.