Final Report Summary - CARDIODEF (Multiscale investigation of drug - implantable cardioverter defibrillator interactions for antiarrhythmic therapy)
Sudden Cardiac Death (SCD) represents approximately 20 per cent of all deaths and its incidence is similar in all western populations. Most SCDs are caused by abnormal heart rhythms called arrhythmias, being the most common sequence the degeneration of ventricular tachycardia (VT) into ventricular fibrillation (VF). Moreover, about 70% of SCDs are related to ventricular fibrillation induced by ischemia or coronary artery disease. The only effective therapy to avert SCD is an electrical cardioverter shock via an automated external
defibrillator (AED) or with an implantable-cardioverter defibrillator (ICD) defibrillator, that has become a first-line treatment and preventive therapy for patients at risk for VT/VF. Besides, VT is routinely treated in the clinical practice with anti-arrhythmic drugs that can lead to direct and indirect interactions with ischemia, with arrhythmia mechanisms and in turn with the ICD shock.
The goal of this project was to understand the interplay between ischemia, antiarrythmic drugs and implantable cardioverter defibrillators (ICD). A combination of signal processing techniques applied to the electrocardiographic signals and computational modelling and simulation of the human ventricles was used.
The scientific objectives were:
First Aim: To develop computational models of human ventricular electrophysiology for the simulation of the effects of antiarrhythmic class I and III drugs in ischemic and non ischemic conditions.
Second Aim: To investigate the mechanisms of class I and III drug therapy in ischemic and non-ischemic patients, and the identification of biomarkers predicting proarrhythmic effects
Third Aim: To investigate the mechanisms underlying drug/ICD interaction in human ventricular models under ischemia and non ischemia and evaluation with ECG from patients
Computational studies:
Firstly, we analyzed the applicability of different human computational models, which have been built with data from healthy cells, to studies of ischemia. This study allowed us to identify the ischemic computational models that are in agreement with experimental ranges at single cell level and tissue level. With the selection of the most suitable model, we were able to conduct simulations to understand the effect of drugs during ischemia or analyze biomarkers measured at the ECG level.
The importance of variability during ischemia observed in experiments is an important issue that we addressed in this project by studying the mechanisms underlying inter-subject variability in the response of the human ventricles to acute myocardial ischemia. This study also showed the variability of the most important mechanisms during ischemia that may lead to arrhythmia at single cell level.
Regarding class I antiarrhythmic drugs, which interfere with the sodium (Na) channel, we have made use of computer modeling to understand the role of the Na inactivation gates on the stability and evolution dynamics of rotors. Ischemia affects electric activation patterns and rotor dynamics during VF. The main mechanisms are action potential duration (APD) shortening, effective refractory period (ERP) prolongation and conduction velocity (CV) decrease. ERP prolongation and CV decrease are mainly affected by hyperkalemia through Na channel availability. We hypothesize that restitution of Na channel availability per se (without changes in refractoriness) is a key modulator of rotor dynamics during acute ischaemia. Results showed that Na current restitution determines rotor stability through the modulation of CV restitution, independently to refractoriness. This collaboration with the cardiologist Peter Taggart on Na channel availability also involves clinical data.
Finally, we analyzed the interplay between arrhythmia, ischemia and drugs by investigating arrhythmic mechanisms of class III anti-arrhythmic drugs during regional ischemia in whole human ventricle simulations. Simulation results showed that, in agreement with experimental recordings, figure-of-eight re-entry is established around the ischemic region following ectopic excitation during the vulnerability window in control and also for tested drug doses. Increasing class III anti-arrhythmic drug dose reduces the probability of re-entry in a biphasic manner. For low doses, prolongation of repolarization in normal tissue decreases the likelihood of reentry by 24%. Increasing the dose, however, increases the likelihood and complexity of re-entry. This is due to establishment of transmural re-entry at the ischemic border zone, facilitated by a modest increase in transmural action potential duration heterogeneity (up to 20 ms).
In conclusion, simulations using a human regionally-ischemic ventricular model showed that (1) anti-arrhythmic effects of class III drugs are based on prolonged repolarization particularly in the normal tissue; (2) pro-arrhythmic mechanisms become predominant at increased doses, as a result of increased transmural repolarization gradients in the ischemic border zone. Our results support the need for a dose-dependent evaluation of the safety and efficacy of these agents in patients with coronary heart disease.
Identification of biomarkers predicting proarrhythmic effects from the ECG:
Increased dispersion of restitution associated with an increased arrhythmic risk was quantified from the body surface ECG (DRest). We have evaluated this novel ECG-based biomarker to predict arrhythmic risk following the administration of an antiarrhythmic drug (sotalol). Results supports the potential of DRest for arrhythmic risk stratification by showing an improved performance compared to the classical QTc interval, in the available data following sotalol administration. In healthy subjects, no differences were found in biomarkers before and after sotalol administration, in agreement with computer simulations of sotalol effects on a human tissue electrophysiological model.
During the past few years, interest has increased in the use of the ICD in genetic cardiac diseases associated with sudden cardiac death, such as hypertrophic cardiomyopathy (HCM) characterized by thickening of a portion of the myocardium. HCM affects 1/500 individuals being the main cause of sudden cardiac death in children and young adults, and the identification of high-risk patients for primary prevention in this disease remains a challenge. Implanting ICDs in patients with HCM is an important clinical consideration since many individuals could achieve normal or near-normal lifespans with this protection. Even having abnormal ECG, reliable methods for risk stratification are lacking as well as ECG-based biomarkers characterizing these patients.
We evaluated ECG-based features from activation and repolarization phases and investigate their relationship with HCM and arrhythmic risk. Results showed higher values for HCM in QRS-duration, QT interval (QTc), T peak to T end interval corrected (Tpe ) and DRest quantifying dispersion of restitution. Also, morphological features such as ST level, T and QRS amplitudes and energies and the first four Karhunen-Loeve transform (KLT) coefficients were analysed. Results showed statistically significant differences between HCM patients and controls (p-value< 0.02) in the QTc interval, ST level, energies and the first and third KLT coefficients. DRest showed higher values in HCM patients (0.08 versus 0.05 in controls) in agreement with in vivo studies suggesting a correlation between increased dispersion and arrhythmic risk.
Regarding the activation phase accounted for the QRS complex, machine learning techniques were applied in order to identify arrhythmic risk in HCM patients. A “blind approach” considering the QRS complex as a feature and a “prior knowledge approach” deriving biomarkers related to the disease were used, obtaining 99% and 97% accuracy in the classification based on support vector machines. The “blind approach” suggested the S wave of the QRS as the one more affected by HCM, and the “prior knowledge approach” showed that the QRS was more upright in the healthy than in HCM subjects.
These two works related to HCM received two awards:
- "Electrographic abnormalities in hypertrophic cardiomyopathy” awarded the Best Poster Presentation Prize. Computing in Cardiology 2014
- Honorary mention in the Best Poster Award competition at the British Heart Foundation (BHF) Center of Research Excellence (CRE) 2014 symposium
All this mechanisms spans from ionic level to whole organ, and body-surface electrocardiogram. We have integrated all these levels, being able to derive simulated ECGs. As a first study we study the effect of an old myocardial infarct scar with a slowly conducting border zone on the ECG, obtaining results in agreement with clinical data.
All these studies and results have been presented in conferences, workshop an invited talks, and have been submitted to journals or are in preparation for submission. Most of these projects, especially the one related to HCM and the one related to Na channel availability, have promoted the collaboration between the clinical and the computational research sides.
defibrillator (AED) or with an implantable-cardioverter defibrillator (ICD) defibrillator, that has become a first-line treatment and preventive therapy for patients at risk for VT/VF. Besides, VT is routinely treated in the clinical practice with anti-arrhythmic drugs that can lead to direct and indirect interactions with ischemia, with arrhythmia mechanisms and in turn with the ICD shock.
The goal of this project was to understand the interplay between ischemia, antiarrythmic drugs and implantable cardioverter defibrillators (ICD). A combination of signal processing techniques applied to the electrocardiographic signals and computational modelling and simulation of the human ventricles was used.
The scientific objectives were:
First Aim: To develop computational models of human ventricular electrophysiology for the simulation of the effects of antiarrhythmic class I and III drugs in ischemic and non ischemic conditions.
Second Aim: To investigate the mechanisms of class I and III drug therapy in ischemic and non-ischemic patients, and the identification of biomarkers predicting proarrhythmic effects
Third Aim: To investigate the mechanisms underlying drug/ICD interaction in human ventricular models under ischemia and non ischemia and evaluation with ECG from patients
Computational studies:
Firstly, we analyzed the applicability of different human computational models, which have been built with data from healthy cells, to studies of ischemia. This study allowed us to identify the ischemic computational models that are in agreement with experimental ranges at single cell level and tissue level. With the selection of the most suitable model, we were able to conduct simulations to understand the effect of drugs during ischemia or analyze biomarkers measured at the ECG level.
The importance of variability during ischemia observed in experiments is an important issue that we addressed in this project by studying the mechanisms underlying inter-subject variability in the response of the human ventricles to acute myocardial ischemia. This study also showed the variability of the most important mechanisms during ischemia that may lead to arrhythmia at single cell level.
Regarding class I antiarrhythmic drugs, which interfere with the sodium (Na) channel, we have made use of computer modeling to understand the role of the Na inactivation gates on the stability and evolution dynamics of rotors. Ischemia affects electric activation patterns and rotor dynamics during VF. The main mechanisms are action potential duration (APD) shortening, effective refractory period (ERP) prolongation and conduction velocity (CV) decrease. ERP prolongation and CV decrease are mainly affected by hyperkalemia through Na channel availability. We hypothesize that restitution of Na channel availability per se (without changes in refractoriness) is a key modulator of rotor dynamics during acute ischaemia. Results showed that Na current restitution determines rotor stability through the modulation of CV restitution, independently to refractoriness. This collaboration with the cardiologist Peter Taggart on Na channel availability also involves clinical data.
Finally, we analyzed the interplay between arrhythmia, ischemia and drugs by investigating arrhythmic mechanisms of class III anti-arrhythmic drugs during regional ischemia in whole human ventricle simulations. Simulation results showed that, in agreement with experimental recordings, figure-of-eight re-entry is established around the ischemic region following ectopic excitation during the vulnerability window in control and also for tested drug doses. Increasing class III anti-arrhythmic drug dose reduces the probability of re-entry in a biphasic manner. For low doses, prolongation of repolarization in normal tissue decreases the likelihood of reentry by 24%. Increasing the dose, however, increases the likelihood and complexity of re-entry. This is due to establishment of transmural re-entry at the ischemic border zone, facilitated by a modest increase in transmural action potential duration heterogeneity (up to 20 ms).
In conclusion, simulations using a human regionally-ischemic ventricular model showed that (1) anti-arrhythmic effects of class III drugs are based on prolonged repolarization particularly in the normal tissue; (2) pro-arrhythmic mechanisms become predominant at increased doses, as a result of increased transmural repolarization gradients in the ischemic border zone. Our results support the need for a dose-dependent evaluation of the safety and efficacy of these agents in patients with coronary heart disease.
Identification of biomarkers predicting proarrhythmic effects from the ECG:
Increased dispersion of restitution associated with an increased arrhythmic risk was quantified from the body surface ECG (DRest). We have evaluated this novel ECG-based biomarker to predict arrhythmic risk following the administration of an antiarrhythmic drug (sotalol). Results supports the potential of DRest for arrhythmic risk stratification by showing an improved performance compared to the classical QTc interval, in the available data following sotalol administration. In healthy subjects, no differences were found in biomarkers before and after sotalol administration, in agreement with computer simulations of sotalol effects on a human tissue electrophysiological model.
During the past few years, interest has increased in the use of the ICD in genetic cardiac diseases associated with sudden cardiac death, such as hypertrophic cardiomyopathy (HCM) characterized by thickening of a portion of the myocardium. HCM affects 1/500 individuals being the main cause of sudden cardiac death in children and young adults, and the identification of high-risk patients for primary prevention in this disease remains a challenge. Implanting ICDs in patients with HCM is an important clinical consideration since many individuals could achieve normal or near-normal lifespans with this protection. Even having abnormal ECG, reliable methods for risk stratification are lacking as well as ECG-based biomarkers characterizing these patients.
We evaluated ECG-based features from activation and repolarization phases and investigate their relationship with HCM and arrhythmic risk. Results showed higher values for HCM in QRS-duration, QT interval (QTc), T peak to T end interval corrected (Tpe ) and DRest quantifying dispersion of restitution. Also, morphological features such as ST level, T and QRS amplitudes and energies and the first four Karhunen-Loeve transform (KLT) coefficients were analysed. Results showed statistically significant differences between HCM patients and controls (p-value< 0.02) in the QTc interval, ST level, energies and the first and third KLT coefficients. DRest showed higher values in HCM patients (0.08 versus 0.05 in controls) in agreement with in vivo studies suggesting a correlation between increased dispersion and arrhythmic risk.
Regarding the activation phase accounted for the QRS complex, machine learning techniques were applied in order to identify arrhythmic risk in HCM patients. A “blind approach” considering the QRS complex as a feature and a “prior knowledge approach” deriving biomarkers related to the disease were used, obtaining 99% and 97% accuracy in the classification based on support vector machines. The “blind approach” suggested the S wave of the QRS as the one more affected by HCM, and the “prior knowledge approach” showed that the QRS was more upright in the healthy than in HCM subjects.
These two works related to HCM received two awards:
- "Electrographic abnormalities in hypertrophic cardiomyopathy” awarded the Best Poster Presentation Prize. Computing in Cardiology 2014
- Honorary mention in the Best Poster Award competition at the British Heart Foundation (BHF) Center of Research Excellence (CRE) 2014 symposium
All this mechanisms spans from ionic level to whole organ, and body-surface electrocardiogram. We have integrated all these levels, being able to derive simulated ECGs. As a first study we study the effect of an old myocardial infarct scar with a slowly conducting border zone on the ECG, obtaining results in agreement with clinical data.
All these studies and results have been presented in conferences, workshop an invited talks, and have been submitted to journals or are in preparation for submission. Most of these projects, especially the one related to HCM and the one related to Na channel availability, have promoted the collaboration between the clinical and the computational research sides.