Development and Internal Validation of a Risk-Prediction Nomogram for Calcific Aortic Valve Stenosis: A Single-Center Real-World Study

Background Calcific aortic valve stenosis (CAVS) is a common valvular disease, and early identification of high-risk populations can help optimize screening and follow-up strategies. However, there is still a relative lack of risk prediction models for CAVS based on routine clinical indicators that have undergone internal validation. This study aims to develop and validate a clinically applicable CAVS prediction model, assessing its discrimination, calibration, and clinical net benefit. Methods This was a single-center retrospective study including patients diagnosed with CAVS at our hospital between 2019 and 2024 and individuals undergoing health examinations during the same period. Participants were classified according to echocardiographic aortic valve area (AVA): AVA < 3 cm² as the CAVS group and AVA ≥ 3 cm² as the normal group. A total of 580 participants were enrolled (450 controls and 130 CAVS cases) and randomly split at a 7:3 ratio into a training set (n = 406; 315 controls and 91 CAVS cases) and a validation set (n = 174; 135 controls and 39 CAVS cases). In the training set, the least absolute shrinkage and selection operator (LASSO) was used for variable selection, followed by multivariable logistic regression to construct the prediction model and nomogram. Model calibration, discrimination, and clinical net benefit were evaluated in both datasets using calibration curves/Hosmer–Lemeshow (H–L) test, receiver operating characteristic (ROC) curve with area under the curve (AUC), and decision curve analysis (DCA), respectively. Results Based on LASSO selection and multivariate logistic regression, 11 independent predictors were identified: ApoB/ApoA1 ratio, age, SIRI(Systemic Inflammation Response Index), sex, fasting glucose, history of CKD, triglycerides, history of hypertension, CRP, history of diabetes, and uric acid. The model demonstrated good calibration: Hosmer–Lemeshow test χ²=6.43, P = 0.600 for the training set; χ²=6.86, P = 0.552 for the validation set, with calibration curves closely approaching the ideal line. The model exhibited high discrimination: AUC = 0.932 (95% CI 0.905–0.959) for the training set and AUC = 0.907 (95% CI 0.855–0.958) for the validation set. At a cutoff probability of 0.269, the sensitivity and specificity were 81.3% and 88.3% for the training set, and 82.1% and 82.2% for the validation set, respectively. DCA indicated that the nomogram provided higher net benefits across a wide range of threshold probabilities (0.01–0.94 for the training set; 0.01–0.89 and 0.96–0.99 for the validation set). Conclusion Using single-center retrospective data, we developed and internally validated a nomogram for predicting CAVS risk. This tool may facilitate individualized risk assessment and inform screening decisions for CAVS.