Development and validation of a new nomogram model for predicting acute ischemic stroke in elderly patients with non-valvular atrial fibrillation:A single-center cross-sectional study

Abstract

Background and aims: Non-Valvular atrial fibrillation (NVAF) patients face a 3-5 times greater risk of acute ischemic stroke (AIS) compared to those without NVAF. Currently employed models for predicting stroke risk in NVAF patients exhibit limitations. It becomes evident that risk profiles for thrombosis and embolism can differ based on race and geographical location. Hence, seeking a new predictive model tailored for the local region to assess the risk of AIS in NVAF patients could lead to ongoing improvements in the model's performance and enhanced predictive efficacy. This study aims to establish a novel clinical prediction model for AIS in elderly patients with NVAF by incorporating relevant biomarker indicators. Methods: A total of 313 individuals were selected from May 2020 to May 2023 for this investigation at the Third Affiliated Hospital of Anhui Medical University. They were patients diagnosed with NVAF. Their clinical data was amassed for retrospective analysis. Based on the presence of AIS, patients were categorized into two groups: the Stroke Cohort (143 cases, NVAF patients with concurrent AIS) and the Non-Stroke Cohort (158 cases, patients with isolated NVAF). Predictor screening was performed using the least absolute shrinkage and selection operation (LASSO) regression algorithm. The binary logistic regression equation was applied to fit the model, followed by internal validation using the bootstrap resampling method (1000 times). Receiver operating characteristic (ROC) curve, calibration degree curve plots and Clinical decision curve analysis (DCA) were generated, respectively. Finally, a Nomogram was constructed to present the prediction model. Result: The final results of this study revealed that neutrophil-to-lymphocyte ratio (NLR), red cell distribution width (RDW), lipoprotein(a) (Lp(a)), systolic pressure, history of stroke, hyperlipidemia were independent risk factors for AIS in elderly patients with NVAF (P < 0.05). On the other hand, high-density lipoprotein cholesterol (HDL-C) were independent protective factors (P < 0.05). By incorporating these seven indicators, a Nomogram prediction model for predicting AIS in elderly patients with NVAF was constructed. The results demonstrate that the area under the ROC curve (AUC) for the modeling dataset is 0.915, and the AUC for the validation dataset is 0.860. The DCA for the modeling set and validation set exhibited clinical net benefits ranging from 0 to 1. Internal validation demonstrated that the model exhibited favorable discriminative ability, calibration, and clinical benefit for AIS in NVAF patients. Comparative analysis between the nomogram predictive model and CHA2DS2-VASc score revealed that the AUC of the nomogram predictive model surpassed that of the CHA2DS2-VASc score (AUC of nomogram predictive model: 0.881, 95% CI: 0.8430-0.9193, sensitivity: 0.7552, specificity: 0.8797; AUC of CHA2DS2-VASc-60 score: 0.850, 95% CI: 0.8177-0.8965, sensitivity: 0.7832, specificity: 0.7841). The DCA plots for both models exhibited clinical net benefit rates spanning 0% to 100%, signifying high clinical utility for both models. Conclusions: NLR, RDW, Lp(a), SP, history of stroke, hyperlipidemia and HDL-C emerge as independent prognostic factors for acute ischemic stroke in elderly patients with non-valvular atrial fibrillation. The predictive utility of nomogram model may potentially surpass that of the CHA2DS2-VASc scoring system, particularly with regard to predictive specificity.