Debatable issues in automated ECG reporting.

Although automated ECG analysis has been available for many years, there are some aspects which require to be re-assessed with respect to their value while newer techniques which are worthy of review are beginning to find their way into routine use. At the annual International Society of Computerized Electrocardiology conference held in April 2017, four areas in particular were debated. These were a) automated 12 lead resting ECG analysis; b) real time out of hospital ECG monitoring; c) ECG imaging; and d) single channel ECG rhythm interpretation. One speaker presented the positive aspects of each technique and another outlined the more negative aspects. Debate ensued. There were many positives set out for each technique but equally, more negative features were not in short supply, particularly for out of hospital ECG monitoring.

Decision to implant a cardioverter defibrillator after myocardial infarction: the role of ejection fraction v. other risk factor markers.

BACKGROUND: The Multicenter Automatic Defibrillator Implantation Trial showed that in post-myocardial infarction patients with a left ventricular ejection fraction (EF) 0.30, an implantable cardioverter defibrillator (ICD) resulted in a 31% relative reduction in the risk of death when compared with a conventional therapy group. Whether further refinement in risk estimation could be achieved with additional clinical testing to qualify patients for primary prevention with ICDs remains problematic. METHODS: The authors analyzed Cardiac Arrhythmia Suppression Trial registry data to estimate sensitivity and specificity of EF, ventricular premature frequency, and nonsustained ventricular tachycardia for predicting death. They combined the results with similar data from the literature and used summarizing receiver operating characteristic (meta-ROC) curves to estimate overall operational values for sensitivity and specificity for each clinical test. They estimated aggregate values for prior probability to project risks when tests were used singly and in combination. RESULTS: The authors used arrhythmia markers and heart rate variability to further stratify low-EF patients (prior risk = 20.3%); proportionately, 20.4% were predicted at high risk (>30%) and 40.5% at low risk (<10%). When heart rate variability is normal, those at high risk reduced proportionately to 9.2%, and those at low risk increased to 51.6%. CONCLUSIONS: The combined use of noninvasive markers for arrhythmia substrate and altered autonomic tone can improve risk stratification in low EF without optimal beta-block therapy, whereas for those with optimal beta-block therapy, markers for arrhythmia substrate alone work. Ancillary use of electrophysiologic stimulation can improve results.

Characteristics of heart beat intervals and prediction of death.

OBJECTIVE: To assess the value for improving risk stratification of measures, unadjusted and adjusted for heart rate, of heart rate variability (HRV) and heart rate turbulence (HRT) based on 2- to 24-h ambulatory electrocardiographic recordings; and to relate this to the decision to use an implantable cardiac defibrillator (ICD) and the attendant consequences on effectiveness and cost-effectiveness. BACKGROUND: Risk stratification for high risk or low risk of lethal ventricular arrhythmic events, and hence for a decision about defibrillator implant, most commonly utilizes the left ventricular ejection fraction (LVEF). Electrocardiographic (ECG) approaches include 24-h ambulatory ECG recordings, with counts of ventricular premature contractions (VPCs), measures of heart rate variability (HRV), and heart rate turbulence (HRT). HRT has two components: turbulence onset (TO) and turbulence slope (TS). METHODS AND RESULTS: We evaluated the qualifying ambulatory ECG recordings from 744 patients in the active treatment arms of the Cardiac Arrhythmia Suppression Trial (CAST). Beat characteristics, VPC counts, normal-to-normal beat intervals, and time-domain measures of HRV and HRT were calculated. Tachograms were rescaled to a heart rate of 75 and the resulting "normalized" measures evaluated as risk predictors for death, compared to unnormalized measures. Measures based on 2-h ECGs were also evaluated as risk predictors. The most powerful univariate predictor of survival was the normalized turbulence slope. The best multivariate prediction model had six components: history of angina, hypertension, diabetes, and absence of post-myocardial infarction revascularization, the log of LVEF, normalized TS, HR, and an interaction term of HR and normalized TS. Gains in effectiveness from use of this model cost between $0 and $4000 per year of life saved. CONCLUSIONS: Turbulence slope substantially exceeded other ECG-based measures in improving prediction of subsequent death in models which included LVEF, and other clinical parameters. Use of this model would improve the effectiveness and cost-effectiveness of the ICD.

Structural relationships between measures based on heart beat intervals: potential for improved risk assessment.

Decreased left ventricular ejection fraction is the most commonly used risk factor for identification of patients at high-risk for lethal ventricular arrhythmic events. Twenty-four-hour electrocardiographic (ECG) approaches to risk stratification include: counts of ventricular premature contractions (VPCs), measures of heart rate variability (HRV), and heart rate turbulence (HRT) which has two components, turbulence onset and turbulence slope (TS). Refinement of these ECG risk stratifiers could enhance their clinical utility. We explored the structural relationships between heart rate (HR) and HRV and HRT measures. Our goal was to separate out the component of these measures due to the underlying average heart rate (HR), thus potentially reducing the variability of the measures and increasing their power to stratify risk. We proposed re-scaling tachograms of heart-beat intervals so that the re-scaled tachogram has a HR of 75 (or equivalently an average interval of 800 ms) and calculating HRV and HRT from the rescaled time series. We also explored the relationship between the number of VPCs and HRT. We showed that TS is structurally related to the number of VPCs (and hence to the length of the ECG recording). We proposed an adjusted TS that is independent of the number of VPCs. We also addressed the ability of shorter ECG recording to estimate HRV and HRT measures. We evaluated standard and rescaled HRV and HRT measures using qualifying ambulatory ECG recordings from 744 patients in the Cardiac Arrhythmia Suppression Trial. We found that measures based on the rescaled tachogram had reduced variance (20% to 40%). Correlations between measures were also substantially reduced. We also found substantial circadian effects on some, but not all HRV indices, not explained by the circadian pattern in HR and possibly pointing to additional measures for risk prediction. In conclusion, we found that adjusting for HR and the number of VPCs in heart-beat related ambulatory ECG measures has the potential to significantly improve the power of these measures to risk stratify cardiac patients.

Lack of usefulness of electron beam computed tomography for detecting coronary allograft vasculopathy.

Fifty-five patients with cardiac allografts were studied by electron beam computed tomography for coronary calcification (EBCT CC) and coronary arteriography, and from the latter, a coronary index was calculated using the size, degree of obstruction, and linear extent of disease of each vessel. There was a significant correlation between EBCT CC score and coronary index, but receiver-operating characteristic (ROC) analysis demonstrated unsatisfactory performance of EBCT CC, and 6 patients had no coronary calcification despite having very abnormal coronary indexes. There are pathologic differences between coronary allograft vasculopathy and atherosclerosis, and correspondingly, EBCT CC has limited usefulness in the cardiac transplant population.

Risk stratification in diabetic patients with a previous myocardial infarction.

We used Kaplan-Meier 2-year survival analysis on CAST registry patients to estimate prognostic power of VPC frequency (> or =10/hr), presence of nonsustained ventricular tachycardia (NSVT), left ventricular ejection fraction, and presence of diabetes. We also used meta-analysis of reports in the literature to estimate prognostic power of signal-averaged electrocardiogram (SAECG) and electrophysiological tests (EPS) as well as VPCs, NSVT, and LVEF. Combined results from CAST analysis and literature meta-analysis yielded sensitivity and specificity for VPCs, NSVT, SAECG, LVEF, Diabetes, and EPS. The overall 2 year event rate for life-threatening arrhythmias or death was 7.88% for 51,144 cases in the combined CAST and literature data. After segmenting the population 21.3% were diabetic with a predicted 2 yr event rate of 13.5% and 78.7% were nondiabetic event rate of 6.4%. We defined low risk as <10% and high risk as > or =30%. Otherwise predicted event rate was classified as "unstratified." When all possible combination of noninvasive tests were applied, a prominent difference in the proportions of cases at risk between the diabetics and nondiabetics was revealed. When the unstratified cases were subsequently tested with EPS, the difference between the two groups was even more marked.

Risk stratification applied to CAST registry data: combining 9 predictors. Cardiac Arrhythmia Suppression Trial.

Over 200,000 people in the United States die of sudden cardiac death (SCD) every year. Although many of these deaths occur in asymptomatic individuals, the vast majority of deaths occur in people who are under care for existing coronary heart disease. Implantable cardioverter/defibrillators (ICDs) have been shown in several randomized trials to be effective in prolonging lives of those at high risk for sudden cardiac death, but the criteria used in these trials and the ACC/AHA consensus guidelines would cover only a minority of patients. Developing methods to assign risk to individual patients without prior SCD events could promote the use of this life-saving therapy in those with especially high risk. Given sufficient physiologically relevant measurements from electrocardiogram analysis, clinical assessment, and demographic status, multivariate statistical methods for predicting survival can be used to combine many predictors of risk and calculate the risk for an individual patient. A survival analysis using Cox regression on data from the Cardiac Arrhythmia Suppression Trial (CAST) illustrates this concept. Patient age, sex, ejection fraction, smoking history, and prior myocardial infarction history, along with the frequency of premature beats and the presence of runs of ventricular tachycardia on Holter monitoring and the time from the index myocardial infarction to the baseline Holter and to recruitment into CAST were combined in a multivariate predictor derived from the Cox regression; this predictor significantly outperforms the individual predictors. A proposed test based on this predictor would identify as positive 7% of the CAST registry, with an average risk of death among the positives of 47%; 20% of those dead at 2 years would be positive. With improved component measurements, this approach has the potential for significantly improving risk stratification for the prevention of SCD.