ECG-based predictors of sudden cardiac death in chagas' disease

Abstract

With nearly six million infected subjects, Chagas' disease is becoming an alarming public health problem, especially in Latin America where it is endemic. This disease is caused by a parasite infecting heart tissue, which can degenerate into serious rhythm disturbances and high risk of sudden cardiac death (SCD). This study aims at stratifying the SCD risk in patients with Chagas' heart disease (CHD). A database composed by 22 Holter ECG recordings from CHD patients with 11 alive and 11 SCD cases was studied. Classical heart rate turbulence (HRT) and heart rate variability (HRV) parameters in time domain were extracted from the signals divided in two 12 h periods (day and night). These parameters were used as input for two multivariate linear models - logistic regression (LR) and linear Fisher discriminant (LDA). When computed separately, HRT and HRV indices cannot properly discriminate alive from SCD patients with CHD. Their discrimination capability increases when HRT is combined with standard HRV indices and they are computed in night recordings, where vagal tonus is increased. Indeed, both resulting models included three parameters from the night period: turbulence slope, standard deviation of all NN intervals and the proportion of successive normal RR intervals with more than 50 ms. The best model (LDA) provided 82.4% accuracy, 87.5% sensitivity and 77.8% specificity.