Clinical Prognosis And Predictive Model Study Of Surgical Treatment For Patients With Chronic Valvular Heart Disease And Atrial Fibrillation

Objectives:To evaluate the efficacy of Cox-Maze IV procedure（CMP-IV）in patients with chronic valve disease combined with atrial fibrillation,to identify potential risk factors for atrial fibrillation recurrence using machine learning algorithms,to construct a predictive model for atrial fibrillation recurrence after CMP-IV,and to enhance individualized atrial fibrillation treatment plans.Methods:1.1026 eligible patients with chronic valvular heart disease and atrial fibrillation from the Second Xiangya Hospital of Central South University and New Bridge Hospital Affiliated to Army Medical University were enrolled from January 2012 to December 2019,with 555patients in the CMP-IV group and 471 patients in the non-CMP-IV group.The Kaplan-Meier method was used to analyze the sinus rhythm maintenance rate in each group.10 machine learning models were constructed,and the performance of the models was evaluated using five-fold cross-validation and model evaluation metrics,including accuracy,precision,recall,F1 score,and the area under the ROC curve.The four models with the best performance were selected for further analysis,including feature importance evaluation and SHAP analysis,to identify the risk factors for atrial fibrillation recurrence and build an atrial fibrillation recurrence risk prediction model.2.779 eligible patients with chronic valvular heart disease and atrial fibrillation from January 2011 to June 2019（a retrospective cohort study）and 360 eligible patients from June 2019 to December 2021（a prospective cohort study）were enrolled.The main endpoint events in the retrospective and prospective cohorts were analyzed,and the effectiveness of the HAS-BLED score and the CHA₂DS₂-VASc score in predicting the prognosis of patients with valvular heart disease and atrial fibrillation was evaluated.Logistic regression and Cox regression methods were used to build a prediction model for the main endpoint events of atrial fibrillation recurrence,death,and stroke.3.ROC analysis of the data was performed using SPSS to calculate the Jorden index for each point on the ROC curve and to analyze the best classification points for preoperative left atrial and atrial fibrillation history risk factors that significantly affect atrial fibrillation recurrence.Results:1.The 5-year postoperative sinus rhythm maintenance rate of patients in the two-center study was 82.13%（95%CI:78.51%,85.93%）.among the 10 machine learning models,the XGBoost,Light GBM,Cat Boost and Random Forest models in the CMP-IV group performed the best,with an area under the ROC curve of 0.768（95%CI:0.742,0.786）,0.766（95%CI:0.744,0.792）,0.762（95%CI:0.723,0.801）and0.732（95%CI:0.701,0.763）,and had higher accuracy,precision,recall and F1 values among the 10 models.Feature importance and SHAP analysis showed that the length of AF history,preoperative left ventricular ejection fraction,postoperative rhythm,preoperative left atrial internal diameter,preoperative neutrophil lymphocyte ratio（NLR）,preoperative heart rate and preoperative white blood cell count were important factors for AF recurrence,and a machine learning-based risk prediction model for AF recurrence was successfully constructed.2.The effect of HAS-BLED and CHA₂DS₂-VASc scores on the risk analysis of the occurrence of the primary study endpoint events in the two cohorts was not significant3.Combining the retrospective cohort study and the prospective cohort study showed that concomitant CMP-IV during valve surgery has good safety..Within the medium-to long-term follow-up period,the CMP-IV group had significantly better results in terms of sinus rhythm occurrence rate,stroke rate,and survival rate compared to the non-CMP-IV procedure group.Successful construction of predictive models for the main endpoint events of atrial fibrillation recurrence,death risk,and stroke risk.4.The best classification point for the preoperative left atrium and atrial fibrillation history risk factors affecting atrial fibrillation recurrence predicted a Youden index of 0.272 for 5-year atrial fibrillation recurrence,corresponding to a left atrial size value of 54.5 mm.The Youden index for predicting 5-year atrial fibrillation recurrence was 0.484,corresponding to a value of 4.45 years.Conclusions:The CMP-IV treatment for atrial fibrillation has good sinus rhythm maintenance,concomitant CMP-IV during valve surgery has good safety,the HAS-BLED and CHA₂DS₂-VASc scores did not have significant effects on the risk analysis of the main study endpoint events in this study.This study successfully identified various factors for atrial fibrillation recurrence after CMP-IV using machine learning algorithms,and successfully constructed four machine learning-based atrial fibrillation recurrence risk prediction models and three predictive models for the main endpoint events based on regression analysis.This could help in clinical decision-making,optimizing individualized surgical management of atrial fibrillation,and postoperative risk assessment.