[Nomogram prediction model for factors associated with vascular plaques in a physical examination population].

OBJECTIVES: Cardiovascular disease (CVD) poses a major threat to global health. Evaluating atherosclerosis in asymptomatic individuals can help identify those at high risk of CVD. This study aims to establish an individualized nomogram prediction model to estimate the risk of vascular plaque formation in asymptomatic individuals. METHODS: A total of 5 655 participants who underwent CVD screening at the Health Management Center of The Third Xiangya Hospital, Central South University, between January 2022 and June 2024 we retrospectively enrolled. Using simple random sampling, participants were divided into a training set (n=4 524) and a validation set (n=1 131) in an 8꞉2 ratio. Demographic and clinical data were collected and compared between groups. Multivariate logistic regression analysis was used to identify independent factors associated with vascular plaques and to construct a nomogram prediction model. The predictive performance and clinical utility of the model were evaluated using receiver operating characteristic (ROC) curves, the Hosmer-Lemeshow goodness-of-fit test, calibration plots, and decision curve analysis (DCA). RESULTS: The mean age of participants was 52 years old. There were 3 400 males (60.12%). The overall detection rate of vascular plaque in the screening population was 49.87% (2 820/5 655). No statistically significant differences were observed in clinical indicators between the training and validation sets (all P>0.05). Multivariate Logistic regression analysis identified age, systolic blood pressure, high-density lipoprotein (HDL), low-density lipoprotein (LDL), lipoprotein(a), male sex, smoking history, hypertension history, and diabetes history as independent risk factors for vascular plaque in asymptomatic individuals (all P<0.05). The area under the curve (AUC) of the nomogram model for predicting vascular plaque risk were 0.778 (95% CI 0.765 to 0.791, P<0.001) in the training set and 0.760 (95% CI 0.732 to 0.787, P<0.001) in the validation set. The Hosmer-Lemeshow goodness-of-fit test indicated good model calibration (training set: P=0.628; validation set: P=0.561). The calibration curve plotted using the Bootstrap method demonstrated good agreement between predicted probabilities and actual probabilities. DCA showed that the nomogram provided a clinical net benefit for predicting vascular plaque risk when the threshold probability ranged from 0.02 to 0.99. CONCLUSIONS: The nomogram prediction model for vascular plaque risk, constructed using readily available and cost-effective physical examination indicators, exhibited good predictive performance. This model can assist in the early identification and intervention of asymptomatic individuals at high risk for cardiovascular disease.