Clinical Outcomes And Predictors Of Patients With Hypertrophic Cardiomyopathy

Part ? Machine learning-based sudden cardiac death prediction of Chinese hypertrophic cardiomyopathy patientsBackgroud and aim:It is important to correctly identify people at high risk of sudden cardiac death(SCD)in patients with hypertrophic cardiomyopathy(HCM).At present,the C statistics of the prediction models proposed by the American College of Cardiology Foundation and American Heart Association(ACCF/AHA)in 2019 and the European Society of Cardiology(ESC)in 2014 in the Chinese cohort of HCM patients were 0.647 and 0.605,respectively.We aimed to explore whether machine learning approaches based on large amounts of baseline data can be more effective in identifying SCD risk in patients with HCM.Methods:We analyzed from January 2002 to December 2013 in 1282 patients with HCM admitted to Fuwai hospital.Clinical baseline data was collected.Logistic regression model,decision tree model,support vector machine model,the random forest model and K nearest neighbor model were constructed by using the method of machine learning,and were compared with the traditional HCM-risk-SCD model.Results:These patients were randomly assigned by the machine to a training set(70%)and a validation set(30%).The training set included 897 patients with HCM[52±13 years old,69.5%males].The validation set included 385 patients[52±12 years old,67.3%males].There were no significant differences in baseline characteristics between the two datasets(P>0.05).At a mean follow-up of 6.3±3.2 years,48 patients(3.7%)with HCM developed SCD events during follow-up.During the observed 5-year follow-up period,29 patients(2.3%)developed SCD,including 20 patients(20/897,2.2%)in the training set and 9 patients(9/376,2.3%)in the validation set.In the evaluation of six machine learning models,it was found that the predicted area under the curve(AUC)of the random forest model was the highest in the validation cohort of 1,2,3,4 and 5 years,respectively,with the values of 0.719,0.821,0.840,0.826,and 0.871,and the 5-year mean AUC was 0.815.The AUCs of HCM-risk-SCD model at 1,2,3,4 and 5 years of follow-up were 0.629,0.610,0.612,0.630 and 0.572,respectively,and the average AUC was 0.611.The prediction effect of HCM-risk-SCD was inferior to that of random forest model.The optimal cut-off point value of the random forest model for SCD risk prediction in 5 years was 0.272,and its sensitivity,specificity and accuracy were 0.778,0.848 and 0.764,respectively.In the random forest model,the importance feature in the top 15 predictors of SCD was increased NT-proBNP,increased FT4,higher serum creatinine,decreased left ventricular ejection fraction,elevated alkaline phosphatase,older age,higher total cholesterol,use of diuretics,decreased heart rate and higher maximum wall thickness,the maximum left ventricular outflow tract pressure decreased,reduced triglycerides,elevated lactate dehydrogenase,left atrial diameter expanding and atrial fibrillation.According to the predicted risk of SCD calculated by the random forest model,patients were divided into four groups,Q1,<0.1;Q2,0.1-0.14;Q3,0.14-0.22;Q4 is ?0.22.The survival curves of the four groups were significantly different(log-rank P<0.0001).Because the survival curves of the first and second groups were similar,we combined the two groups into one group and reclassified patients as low-risk,medium-risk,and high-risk.Survival differences were also found in all three groups(log-rank P<0.0001).In the random forest model,38 patients with medium-high risk SCD predicted in the HCM-risk-SCD model(model score ?40)were reclassified as low risk patients in the random forest model(HtoL group),and 64 patients with low risk(model score<4%)were reclassified as the high risk patients in the random forest model(LtoH group).No statistically significant difference was found in survival between the two groups during the 5-year follow-up(log-rank P=0.15),but there was a decreasing trend of long-term survival in the LtoH group compared with the HtoL group.Conclusions:We found that several machine learning approaches were more accurate in predicting SCD risk in patients with HCM than the HCM-risk-SCD model,and the random forest model had the best prediction effect.Using random forest algorithm is helpful to accurately identify HCM population suitable for ICD implantation and reduce the medical burden on patients and society.Part ? Association of QTc interval and V4-S wave with appropriate ICD therapy in patients with hypertrophic cardiomyopathyBackground and aim:The occurrence of ventricular arrhythmias in patients with hypertrophic cardiomyopathy(HCM)may lead to sudden cardiac death(SCD).Current predictive models for SCD include multiple baseline variables but do not use electrophysiological indicators.We aimed to assess the predictive value of electrocardiogram(ECG)indicators for appropriate therapy in HCM patients receiving implantable cardioverter-defibrillator(ICD)placement.Methods:This study was a single-center,retrospective,and observational study.A total of 164 HCM patients receiving ICD implantation were enrolled consecutively.Clinical data of patients was collected,and ECG parameters were measured and recorded.QT interval correction(QTc)with Bazett 's formula was used,and V4-S wave anomaly is defined as long or deep S wave in V4 lead,that is,the duration of V4 lead S wave is prolonged or the amplitude is deepened(duration time>50 ms and/or voltage amplitude>0.6 mV).The end point in our study was at least once ICD appropriate therapy triggered by ventricular tachyarrhythmia(VT)or ventricular fibrillation(VF),including anti-tachyarrhythmia pacing(ATP)and electrical shock.Results:A total of 149 patients with HCM(mean age 53±14 years,male 69.8%)were studied in the further analysis according to the inclusion and exclusion criteria.Appropriate ICD therapy occurred in 47 patients(31.5%)during a median follow-up of 2.9(1.7-5.6)years.Concerning ECG parameters,the proportion of long or deep S wave in V4 in patients receiving ICD therapy(63.8%)was significantly higher than those without ICD therapy(42.2%)(P=0.021).Moreover,QTc interval was significantly longer in patients receiving ICD therapy(464±56ms)than in those without ICD therapy(436±36ms)(P=0.003).After adjusting for confounding factors,Cox regression analysis showed that long or deep V4-S wave[hazard ratio(HR)1.955,95%confidence interval(CI)1.017-3.759,P=0.045]and prolonged QTc interval(HR 1.014,95%CI 1.008-1.021,P<0.001)were independent risk factors for appropriate ICD therapy in HCM patients.Receiver operating curve(ROC)showed that the optimal cut-off point value for QTc interval to predict the appropriate ICD therapy was 464ms,and the area under the curve(AUC)was 0.658(95%CI 0.544-0.762,P=0.002).The AUC of the ROC curve for long or deep V4-S wave anomalies to predict appropriate ICD therapy was 0.608(95%CI 0.511-0.706,P=0.034).Kaplan-Meier survival curves showed that patients with QTc?464ms had a significantly higher risk of developing appropriate ICD therapy during long-term follow-up than patients with QTc<464ms(Log-rank P<0.0001).Similarly,patients with long or deep S-wave in lead V4 on ECG had a significantly higher risk of appropriate ICD therapy than those without long or deep S-wave(Log-rank P=0.009).Subgroup analysis showed that HCM patients with QTc>464ms+long or deep V4-S wave had the highest risk of developing appropriate ICD therapy(log-rank P<0.0001).Meanwhile,we also verified that the HCM-risk-SCD score was an independent risk factor in our cohort(HR 1.110,95%CI 1.003-1.229,P=0.043).After we added QTc interval and long or deep V4-S wave into the HCM-risk-SCD model,the prediction effect of the new model was significantly improved,and the net reclassification index(NRI)was 0.302.Conclusions:Prolonged QTc interval and long or deep V4-S wave are independent predictors of appropriate ICD therapy in HCM patients with ICDs.Patients with QTc?464ms and long or deep S-wave were at higher risks of appropriate ICD therapy,and patients with both had the highest risk.After adding these two features into the HCM-risk-SCD model,which has been widely used in clinical practice,the prediction effect of the new model is significantly improved.Part ? Echocardiographic predictors of all-cause mortality in patients with hypertrophic cardiomyopathy following pacemaker implantationBackground and aim:Patients with Hypertrophic cardiomyopathy(HCM)are often associated with Left atrial enlargement.And Left atrial diameter(LAD)is often associated with prognosis in patients with HCM.The purpose of this study was to investigate the relationship between LAD and Left ventricular end-diastolic diameter(LVEDD)measured by echocardiography and the long-term risk of all-cause mortality in adult HCM patients after pacemaker implantation.Methods:This study was a single-center,retrospective,observational study.We analyzed 94 adult HCM patients who received pacemaker implantation for symptomatic bradycardia in Fuwai hospital from November 2002 to June 2013 and did not receive Implantable cardioverter-defibrillator(ICD)or Cardiac resynchronization therapy(CRT)during the follow-up period.The endpoint event was all-cause death.Results:Seventy-six patients(57.6± 15.2 years old,51.3%female)had LVEDD records,in whom 74 patients(58.1 ± 14.9 years old,52.7%female)had LAD records.General data showed that the most common comorbidity of HCM patients was atrial fibrillation,accounting for 41.9%of the total population.The second was hypertension,accounting for 28.4%.According to the receiver operating characteristic(ROC)curve,the optimal cut-off values of LAD and LVEDD for predicting all-cause mortality in HCM patients were 44 mm and 43 mm,respectively.Compared with patients with LAD<44mm,patients with LAD?44mm had a significantly higher proportion of atrial fibrillation(57.1%vs.32.6%,P=0.038),a lower critical proportion of Left ventricular outflow tract obstruction(35.7%vs.56.5%,P=0.082),and a significantly larger LVEDD(48.3±6.5mm vs.43.7±5.9mm,P=0.003).Patients with HCM(LAD?44mm)had significantly lower left ventricular ejection fraction(59.9± 10.5%vs.65.0±7.0%,P=0.027)and more angiotensin-converting enzyme inhibitors/angiotensin receptor antagonists(46.4%vs.19.6%,P=0.014).Patients with LVEDD>43mm had a larger LAD(43.8±8.4mm vs.38.3±5.4mm,P=0.003)and a lower left ventricular ejection fraction(61.2±8.9mm vs.66.1±7.4mm,P=0.016).In Kaplan-Meier survival analysis,LAD?44mm and LVEDD?43mm were significantly associated with all-cause mortality(Log-rank P<0.05).During a mean follow-up of 7.1±3.4 years,13 patients with LAD?44mm died(including one who received a heart transplant)and 5 patients with LAD<44mm died;Seventeen patients with LVEDD?43mm died(one of whom received a heart transplant),and one patient with LVEDD<43mm died.In the Kaplan-Meier survival analysis,the cumulative survival risk of patients with LAD?44mm and LAD<44mm was significantly different(Log-rank ?2=8.836;P=0.003).Similarly,patients with LVEDD?43mm had a higher risk of all-cause death compared to patients with LVEDD<43mm(log-rank test ?2=5.285;P=0.022).Multivariate Cox regression analysis showed that LAD?44mm(hazard ratio 3.58,95%confidence interval 1.055-12.148,P=0.041)was an independent predictor of all-cause mortality,while LVEDD?43mm was not significantly associated with all-cause mortality(hazard ratio 4.141,95%confidence interval 0.472-36.352,P=0.200)after adjusting for confounding factors.Conclusions:In HCM patients with pacemaker implantation,LAD?44 mm was an independent predictor of all-cause mortality.