Risk of Sudden Death with Heart Failure
Abstract & Commentary
Synopsis: Spontaneous ventricular arrhythmias on an AECG are nonspecific predictors of total mortality and sudden death and their detection provides limited added prognostic significance over other clinical variables in patients with advanced heart failure.
Source: Teerlink JR, et al. Circulation 2000:101;40-46.
The prospective randomized milrinone survival evaluation (PROMISE) was designed to evaluate the safety and efficacy of oral milrinone in patients with class III and class IV heart failure. PROMISE enrolled 1088 patients with class III or class IV symptoms for more than or equal to three months and a left ejection fraction less than or equal to 35%. Twenty-four-hour ambulatory electrocardiographic (AECG) recordings were obtained within seven days before randomization. During the study, the modes of death were reviewed in a prospectively defined, blinded fashion by an independent mortality committee that classified deaths as sudden or nonsudden. Data from the AECGs were then correlated with mortality to test the hypothesis that ambulatory ventricular ectopy predicts outcome in patients with severe heart failure. Selected AECG variables were then examined in univariate and multivariate Cox proportional hazard analyses. Two multivariate logistic models were developed that used the clinical variables with and without the most powerful AECG variables to assess the incremental value of the latter. Finally, sensitivity and specificity analyses were performed on these models, and receiver operating characteristic (ROC) curves were generated.
The study group was 78% male and coronary artery disease was the etiology of heart failure in 54%. Five percent of the patients were on amiodarone and 17% were on other antiarrhythmic drugs but antiarrhythmic drug use did not influence any of the analyses. Sixty percent of the group had more than 30 PVCs per hour, 85% had couplets, and 61% had nonsustained ventricular tachycardia. Twenty-nine percent of the group had more than five episodes of nonsustained VT and in 10% a single nonsustained VT episode lasted more than 10 beats.
During the course of the study, 290 patients died and 139 of these deaths were classified as sudden. Frequent PVCs, the presence of nonsustained VT, and longer runs of non-sustained VT identified groups with increased all-cause mortality using a univariate analysis. Frequent PVCs, the presence of couplets, and the presence and frequency of nonsustained VT were also univariate predictors of sudden death. However, multivariate general linear proportional hazard models demonstrated that selected clinical variables (age, New York Heart Association class, presence of coronary disease as etiology, ejection fraction, and systolic blood pressure) were the best independent predictors of overall mortality. Ejection fraction was the most powerful clinical predictor of sudden death. An interaction between milrinone therapy was noted for NYHA class in that class IV but not class III patients had greater excess mortality with milrinone. When ambulatory ECG model variables were added to the analysis, the number of nonsustained VT episodes was a significant independent predictor of sudden death, nonsudden death, and overall mortality. However, further analysis revealed that ventricular arrhythmias were not specific predictors of sudden death. At all sensitivity levels greater than 50% for predicting sudden death, the false-positive rate for the best of the ambulatory ECG variables was greater than 80%. ROC curves of a multivariate logistic regression model with and without inclusion of the ambulatory ECG variables showed that both models were associated with poor sensitivity and specificity and that the ROC curves of the two models were essentially superimposable.
Teerlink and colleagues conclude that spontaneous ventricular arrhythmias on an AECG are nonspecific predictors of total mortality and sudden death and that their detection provides limited added prognostic significance over other clinical variables in patients with advanced heart failure.
Comment by John P. DiMarco, MD, PhD
Over the years, numerous studies have shown that detection of frequent or "complex" ventricular arrhythmias in patients with advanced heart disease is a univariate predictor of mortality. The use of the AECG for this purpose seemed intuitive since most sudden deaths are due to ventricular fibrillation or ventricular tachycardia. It was also initially felt that suppression of the ventricular ectopy could be guided by responses during serial monitoring during drug therapy. When the Cardiac Arrhythmia Suppression Trial (CAST) tested this hypothesis, however, it was found that, at least with the antiarrhythmic drugs used in that trial, suppression of spontaneous ventricular ectopy was associated with an increased, rather than a decreased, mortality. Other studies have shown that spontaneous ventricular ectopy is a finding that often manifests significant day-to-day variability in both frequency and complexity in the same individual. Thus, since a high proportion of patients with advanced heart failure will have complex ectopy on a single monitor, it seems likely that most, if not all, will have frequent or complex ectopy at some point if multiple monitors are obtained. Since CAST, the use of AECG in asymptomatic patients has declined in frequency.
Two recent studies, however, have used the presence of nonsustained VT during monitoring as a starting point for further studies. The Multicenter Automatic Defibrillator Implantation Trial (MADIT) and the Multicenter Unsustained Tachycardia Trial (MUSTT) both took patients with prior myocardial infarctions (MI) and nonsustained VT and then subjected the patients to electrophysiologic studies for further risk stratification. Both studies concluded that defibrillator implantation was associated with improved outcome. However, in MUSTT, the electrophysiologic study was only a relatively weak predictor of outcome since a high mortality was also noted in the group without inducible VT. This paper and the data from MUSTT suggest that just symptoms of heart failure and a depressed ejection fraction are sufficient to define a high-risk group. Since the proportion of patients who have abnormal ambulatory monitor results is much greater than the frequency of sudden death, the management algorithms based on ambulatory ECG variables will always be somewhat insensitive and highly nonspecific. Unfortunately, based on the data from MUSTT, it appears that the same problem persists even if electrophysiologic study results are included.
This difficulty in defining a high-risk subgroup would not be a problem if simple and highly effective interventions had been identified. This would apply to the use of ACE inhibitors or beta-blockers but certainly does not apply to defibrillator implantation because of the latter’s expense.
The most powerful predictor of sudden death in heart failure patients is:
a. PVCs on AECG.
b. salvos of PVCs on AECG.
c. nonsustained VT on AECG.
d. ejection fraction.