Development and validation of a new prediction model for calcific aortic valve stenosis

Calcific aortic valve stenosis (CAVS) is a common valvular heart disease, but there are limited reports on the construction of prediction models for CAVS. This study aimed to investigate the risk factors for CAVS and construct a predictive model for CAVS based on its common clinical features.

Patients with CAVS who underwent surgical treatment in our hospital from 2016 to 2020 and those who underwent physical examination during the same period were retrospectively studied and placed in the CAVS group and normal group based on the area of aortic valve orifice less than or more than 3 cm². A total of 548 patients were included in this study, including 106 CAVS patients and 442 normal patients. Subjects were randomly divided into training and validation sets at a 7:3 ratio. The features were dimensionally reduced using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm in the training set, and the optimal clinical features were selected. The independent predictors of patients with CAVS were determined by univariate and multivariate logistic regression, and nomogram was constructed. The calibration curve, receiver operating characteristic (ROC) curve, and decision curve analysis (DCA) were used to evaluate the model in both the training set and the validation set.

In this study, 11 independent predictors were distinguished by multivariate logistic regression analysis: history of hypertension, history of carotid atherosclerosis, age, diastolic blood pressure, C-reactive protein, direct bilirubin, alkaline phosphatase, low-density lipoprotein (LDL), lipoprotein(a) [Lp(a)], uric acid, and cystatin C. A nomogram was constructed using the above indicators. The model was well-calibrated and showed good discrimination and accuracy [the area under the curve (AUC) =0.981] in the training set, with a sensitivity of 91.89% and a specificity of 95.48%. More importantly, the nomogram displayed a good performance in the validation set (AUC =0.955, 95% CI: 0.925-0.985), with a sensitivity of 93.75% and a specificity of 84.09%. Additionally, DCA revealed that the nomogram had high clinical practicability.

This study successfully established a risk prediction model for CAVS based on 11 conveniently accessible clinical indicators, which might easily be used for individualized risk assessment of CAVS.