Exploration of the Precision of Classifying Sudden Cardiac Death
Implications for the Interpretation of Clinical Trials
Background As cardiovascular clinical trials improve in sophistication and therapies target specific cardiac mechanisms of death, a more objective and precise system to identify specific cause of death is needed. Ideally, sudden cardiac death would describe patients dying of ventricular tachycardia and ventricular fibrillation. In this context, we explored the precision of current sudden death classification and implications for clinical trials.
Methods and Results Deaths were analyzed in 834 patients who received an automatic implantable cardioverter-defibrillator (ICD). Three arrhythmia experts used a standard prospective classification system to classify deaths into accepted categories: sudden cardiac, nonsudden cardiac, and noncardiac. New aspects to this study included analysis of autopsy results and ICD interrogation for arrhythmias at the time of death. All of the patients receiving the ICD previously had documented sustained ventricular tachycardia/fibrillation or cardiac arrest. Of the 109 subsequent deaths in the 834-patient database, 17 (16%) were classified as sudden cardiac. Compared with the nonsudden cardiac and noncardiac categories, sudden cardiac death was more often identified in outpatients (59% versus 10%) and witnessed less often (41% versus 86%; both P<.001). The autopsy information contradicted and changed the clinical perception of a “sudden cardiac death” in 7 cases (myocardial infarction [n=1], pulmonary embolism [n=2], cerebral infarction [n=1], ruptured thoracic [n=1], and abdominal aortic aneurysms [n=2]). Interpretable ICD interrogation was available in 53% of the deaths (47% unavailable: buried, programmed off, or other technical reasons). When evaluated, only 7 of 17 “sudden deaths” were associated with ICD discharges near the time of death.
Conclusions Even in a group of patients with an ICD, deaths classified as sudden cardiac frequently were not associated with ventricular tachycardia or ventricular fibrillation and were often noncardiac. It is possible to create a wide range of sudden cardiac death rates (more than fourfold) using the identical clinical database despite objective, prespecified criteria. Autopsy results frequently reveal noncardiac causes of clinical events simulating sudden cardiac death. ICD interrogation revealed that ICD discharges were often related to terminal arrhythmias incidental to the primary pathophysiological process leading to death.
Most pivotal clinical trials in cardiovascular disease involve large numbers of patients, allowing the evaluation of objective clinical end points, especially mortality.1 2 3 4 5 6 7 The methodology for clinical trials has improved in sophistication, and new therapies target specific mechanisms (for example, heart failure or ventricular tachycardia), creating the need for the objective and precise classification of the specific cardiac cause of death.8 9 10 11 12 13 14 15 16 17 18 This is especially important for the category of sudden cardiac death, which has assumed major importance in cardiovascular clinical trials. The practical consequence of classifying a death as sudden cardiac is the presumption that such deaths are due to ventricular tachyarrhythmias. If this assumption is accurate, such sudden cardiac deaths might benefit from antiarrhythmic drugs and/or implantable cardioverter-defibrillators (ICDs). However, if this presumption is inaccurate and deaths classified as sudden cardiac include a variety of arrhythmic and nonarrhythmic mechanisms, the lack of precision has serious implications. Currently used methods to classify death markedly differ, resulting in sudden cardiac death rates that vary widely, as depicted in Table 1⇓.2 3 4 5 6 7 For example, two trials reporting on similar patients with clinical heart failure have sudden death rates varying by threefold (22% versus 61%).4 7
The purpose of this investigation was to explore the precision with which current sudden death classifications identify patients dying of ventricular tachycardia and/or ventricular fibrillation and the implications of these observations to the appropriate interpretation of the results of cardiovascular clinical trials.
The death classification approach was prospectively agreed to by an independent committee comprised of three university cardiologists with expertise in arrhythmias, appointed by Telectronics Pacing Systems, Inc (Englewood, Colo). The committee independently classified deaths in patients enrolled in the three Guardian ICD trials from July 1990 through June 1994.
Death Classification System
The classification selected was a modified Cardiac Arrhythmia Suppression Trial (CAST) approach,2 3 18 with the following definitions: (1) sudden cardiac death, defined as death of cardiac origin that occurred unexpectedly within 1 hour of the onset of new symptoms or a death that was unwitnessed and unexpected, unless a specific noncardiac cause of death was confirmed, (2) nonsudden cardiac death, defined as a death of cardiac origin but not within the sudden cardiac death definition, and (3) noncardiac death, defined as a death that was not primarily due to cardiac causes.
Cases of discordance among the members were reviewed and the final classification was made either by unanimity or majority vote. Every effort was made to obtain confirmatory medical records documenting an irreversible “downward spiral” (for example, congestive heart failure) before death when attempting to assess the relative contribution of incidental arrhythmias occurring at the end stage of such a process. The results of autopsy information were used in the final classification. The data log telemetered from the ICD was not included in the death classification system. However, a separate analysis of the ICD log information examined the potential impact of interrogation on the final classification.
Description of ICD Devices
Patients with life-threatening ventricular arrhythmia events received one of three investigational ICD devices. The ICD devices are designed to automatically monitor, detect, and treat ventricular tachycardia, ventricular fibrillation, and bradycardia. The three models included 75 model 4204 devices, which provide shock and bradycardia therapy; 516 model 4210 devices, which, in addition to defibrillation therapy, incorporate antitachycardia pacing for hemodynamically tolerated ventricular tachyarrhythmias; and 243 model 4211 implants, which offer programming options for two antitachycardia pacing zones, separate low-energy cardioversion and high-energy defibrillation modes, and biphasic or monophasic shock waveforms. All devices use the same detection scheme, which is based on X out of Y intervals exceeding a programmed rate criteria. Information regarding the detection of tachyarrhythmia episodes and the delivered therapy is stored for later retrieval. Each detection as well as the results of all therapies used and the outcome of each episode is detailed. Sensed intervals and intracardiac electrograms are also stored at selected points during the event.
Descriptive results are summarized by mean±SD. Some comparisons of death classification results were made using Fisher’s exact test to examine the null hypothesis of equality of proportion between classification groups. Test results were interpreted at a two-sided α=0.05 significance level.19
A total of 834 patients with life-threatening ventricular arrhythmias (695 men, 139 women; mean age, 61±13 years) were enrolled in the Guardian ICD trials from July 1990 through June 1994; 109 patients died, forming the basis of this report. The mean duration of follow-up for all patients was 590±389 days (range, 1 to 1525 days). Patients still enrolled as of the June 1994 cutoff date were followed for a mean of 684±379 days (range, 10 to 1525 days), whereas the patients who died were enrolled for a mean of 284±270 days (range, 1 to 1081 days).
Death Classification Results
The classification of the 109 deaths is presented in Table 2⇓. A total of 17 deaths were classified as sudden cardiac (17 of 834, 2% of the study population, or 16% of the total deaths). An important consideration in the mortality classification was the location of the patient at the time of death and the presence or absence of witnesses to the death. As presented in Table 3⇓, 90% of the deaths classified as nonsudden cardiac occurred in-hospital, whereas only 41% of the deaths classified as sudden cardiac occurred in-hospital (P<.001). Only 41% of the deaths classified as sudden cardiac were witnessed, whereas 86% of nonsudden cardiac deaths were witnessed (P<.001). The majority of deaths classified as noncardiac occurred in-hospital (78%) and were witnessed (83%) compared with deaths classified as sudden cardiac (P<.01).
Despite the prespecified system of death classification and identical patient data sets, there was initial disagreement in 36% (39 of 109) of the cases (Table 4⇓). Disagreements were usually clarified by additional information obtained and subsequent discussion. The most frequent disagreements occurred between the nonsudden cardiac and noncardiac categories. Initial disagreements and the final classification are as follows: (1) There were 13 deaths that at least one committee member initially classified as sudden cardiac death, which, after review, 11 were classified by consensus as nonsudden cardiac (n=11) or noncardiac (n=2). (2) There were 17 deaths that at least one committee member initially classified as nonsudden cardiac, of which 3 were changed to sudden cardiac and 14 classified as noncardiac. (3) There were 4 deaths initially classified as noncardiac by at least one committee member; classification changed in all cases to nonsudden cardiac. Examples of cases with clinical presentations creating difficulties in classification include the following.
Case 1. This was a 44-year-old woman with an ICD implanted for 369 days. The patient had a 1-week history of progressive deterioration of congestive heart failure after a 6-month history of worsening congestive heart failure, progressing from New York Heart Association functional class II to class IV. During the 24 hours preceding death, progressive hypotension developed despite pressor drugs. Recurrent sustained ventricular tachycardia occurred in the hours before death, which was detected and treated by the ICD. After discussion of the primary mechanism leading to death, one member of the committee changed his vote from sudden cardiac death to the final consensus classification, nonsudden cardiac death.
Case 2. This was a 68-year-old man with ventricular tachycardia and ventricular fibrillation with an ICD implanted for 825 days. The patient was hospitalized for progressive congestive heart failure with severe dyspnea at rest and treated with dobutamine and nitroglycerin. In contrast to the previous case, this patient had symptomatic improvement and was able to ambulate in the days before death. The subsequent death was witnessed, unexpected, and occurred ≤1 hour from the onset of symptoms; it was classified as a sudden cardiac death.
Case 3. This was a 67-year-old man with an ICD implanted for sustained ventricular tachycardia. He awoke one morning complaining of severe shortness of breath and was observed by his wife to collapse immediately. He was brought to the local emergency room, where cardiopulmonary resuscitation was unsuccessful, and the patient was pronounced dead within 30 minutes. Although the temporal sequence was consistent with a witnessed sudden cardiac death, an autopsy documented massive pulmonary embolism. Based on the autopsy results, this death was classified as noncardiac.
In these case examples, ICD detection of ventricular tachycardia near the time of death can be incidental (case 1). The perfect clinical circumstances simulating sudden cardiac death can be misleading and may not equate to an arrhythmic death (case 3). In many cases, there were competing cardiac causes of death, with an arbitrary decision as to which was predominant (case 2).
Device Interrogation: Relationship to Clinical Death Classification
As a separate analysis, the impact of ICD device interrogation on the death classification was assessed. Of the 109 patients, device information was unavailable in 51 of 109 (47%) (1) because the device was buried with the patient (n=20); (2) the device was programmed to “off” before death (n=18); telemetry was incomplete (n=5); (3) no printouts were available (n=4); and (4) in the remaining cases (n=4) the device was explanted with therapy programmed to “on,” resulting in noise overwriting existing data. Thus, exclusive reliance on device information as the basis for a classification system of deaths would have resulted in 47% of the deaths remaining unclassified.
A comparison of information derived from the interrogation of the ICD relative to the clinical classification of death is presented in Table 5⇓. ICD discharges were detected within 6 hours of death in 20 patients but detected in only 7 of 17 (41%) patients whose death was classified as sudden cardiac. In the 10 unwitnessed sudden deaths, 3 had ICD discharges, 3 had no ICD detections, and 4 had no data. Four of 7 witnessed sudden cardiac deaths had ICD discharges in the hours preceding death. Two witnessed sudden cardiac deaths with no ICD discharges were explained by a confirmed lead dislodgment and a monitored asystolic death; in the remaining witnessed sudden deaths there was no ICD information. Deaths classified as nonsudden cardiac and noncardiac had ventricular tachycardia and/or ventricular fibrillation detections with shock and/or antitachycardia pacing within 6 hours of death in 9 of 51 and 4 of 40, respectively, as reported in Table 5⇓.
Two examples of contrasting clinical presentations, when ICD interrogations correctly document and treat ventricular tachycardia within 1 hour of death, are presented in Figs 1⇓ and 2⇓. One patient had a clinical course of weeks of inexorable progression of congestive heart failure and therefore was classified as nonsudden cardiac (Fig 1⇓) despite multiple arrhythmic episodes detected on the data log in contrast to a death classified as sudden cardiac (Fig 2⇓), which was an unwitnessed and unexpected death associated with ventricular tachycardia detection. Although ICD detection and appropriate discharges are documented in both cases, in only one is this information relevant to the primary mechanism leading to death.
Relevance of Autopsy Information for Death Classification
Autopsy information was available in 29 of 109 deaths. Seven deaths that would have been classified as sudden cardiac based solely on temporal criteria were clarified by the results of autopsy, which included 2 cases of ruptured abdominal aortic aneurysm, 1 case of massive cerebral infarction, 1 case of ruptured thoracic aortic aneurysm, 2 cases of massive pulmonary embolism (see case 3), and 1 acute anterior myocardial infarction. Based on the autopsy information, these 7 deaths were subsequently classified as noncardiac (n=6) and nonsudden cardiac (n=1). Had autopsy information not been available or considered, the reported rate for sudden cardiac death would have increased to 24 of 834 (2.9%). A wide spectrum of sudden death rates can be created from this database, depending on the definitions and the types of information used (Table 6⇓).
In this study of patients who had received an ICD for sustained ventricular tachycardia/fibrillation or cardiac arrest, many subsequent deaths that clinically simulated sudden cardiac death were in fact due to noncardiac causes with no ICD evidence of ventricular arrhythmia. This investigation focuses on the limitations and imprecision inherent in the current approach to classifying sudden cardiac death. The important conclusion from this investigation is that standard definitions of sudden death fail to precisely identify patients dying of ventricular tachycardia or ventricular fibrillation. This important conclusion is based on new observations from a substantial autopsy and defibrillator database.
Sudden death is usually classified by considering the temporal sequence from the onset of symptoms to death. For example, in CAST, sudden cardiac death or arrhythmic death included witnessed death within 1 hour of the onset of symptoms.2 3 Most multicenter clinical trials such as the CAST receive autopsy information on deaths in only 1% to 2% of cases. Ongoing trials such as the Canadian Implantable Defibrillator Trial (CIDS) and the Multicenter Unsustained Tachycardia Trial (MUSTT) distinguish instantaneous death as within 1 to 5 minutes.20 21 This latter approach presumably improves specificity for primary arrhythmic causes but limits sensitivity, since not all arrhythmic deaths occur instantaneously. Other defibrillator and/or antiarrhythmic drug trials presently under way have selected a classification strategy closer to the approach used in CAST, such as the Antiarrhythmics Versus Implantable Defibrillators (AVID) trial.22
The use of temporal criteria obviously cannot apply to the unwitnessed category, which usually is in the outpatient setting, where relevant clinical information is frequently lacking. Based on results presented in this investigation, noncardiac causes of death can frequently simulate the clinical circumstances of instantaneous sudden death, whether witnessed or not witnessed and unexpected.
An ideal death classification system would limit the identification of sudden death to those cases in which antiarrhythmic therapy could be useful. There has been limited attention to the accuracy of current death classification approaches, although it has been the focus of a recent policy conference, and the variability in reporting sudden death rates has long been recognized.23 24 Sudden death classifications usually focus on a time-based definition and whether a death was witnessed.12 15 The events classification committee of the Cardiac Arrhythmia Pilot Study (CAPS) acknowledged that a time-based definition was unlikely to accurately identify deaths potentially preventable by an ICD or an antiarrhythmic drug.18
The present study has identified unwitnessed outpatient deaths as the category most likely to inappropriately ascribe a death as sudden cardiac. It is intuitive that in most clinical trials the majority of deaths are unwitnessed (in the field) and the time from symptoms to death not known, resulting in a large potential error rate, resulting in the classification of these deaths as sudden cardiac by default. To highlight the potential for misclassification in this category, only 50% of the deaths classified as unwitnessed sudden cardiac in the present report were temporally associated with ventricular tachycardia and/or ventricular fibrillation documented by ICD interrogation.
Another important observation from this study is that unexpected cardiovascular medical catastrophes, identified by autopsy, closely simulate the circumstances of unexpected (sudden) death. Such deaths (especially those with a history of life-threatening ventricular arrhythmia) would be classified as sudden cardiac if strict temporal criteria alone were used. The additional autopsy data established that ventricular tachycardia and/or ventricular fibrillation was not the predominant mechanism responsible for a number of sudden, unexpected deaths.
Is ICD Interrogation the Solution for Determining the Cause of Death?
The present investigation does not question the established clinical utility of device interrogation for assessing shock delivery and appropriateness of shocks in live patients. In contrast, this study identifies significant limitations when considering ICD interrogation in place of a clinically based system for classifying death. In the present investigation, despite the best efforts of clinical investigators, ICD information was not obtained in 47% (51 of 109) of the patients, with most of the devices either programmed “off” or buried with the patient. Moreover, device detections frequently occurred in the final hours of life, which, although correctly identifying and treating ventricular tachycardia and ventricular fibrillation, were incidental to an inexorable disease process (for example, end-stage heart failure). The subset of deaths in which interrogation appears most informative is the category of unwitnessed, unexpected sudden death. One half of the deaths fitting this profile with available ICD information had no ventricular tachycardia or ventricular fibrillation detections in the hours preceding death, eliminating a primary arrhythmic mechanism from consideration. A corollary conclusion is that the selection of ICD discharges as the primary end point in a clinical trial or its use as a surrogate for sudden death presents substantial interpretive problems.
Implications of the Present Trial
As part of the government requirement for an Investigational Device Exemption, the Food and Drug Administration requires the sponsor of a new ICD to classify cause-specific mortality based on an objective assessment of each death. In the present investigation, the final sudden cardiac death rate was determined to be 2% (17 of 834) over the 1.6-year mean follow-up, which generally is comparable to other databases.25 However, a wide disparity in sudden death rates can be reported from this clinical database. For instance, if the definition of sudden cardiac death was restricted to those that were witnessed and occurred within 1 hour of observation, as suggested by some investigators,15 the reported sudden death rate would be 0.8% (7 of 834). Alternatively, had the classification of sudden cardiac death been applied to every death initially so classified by at least one committee member, the incidence of sudden cardiac death would have increased to 3.6% (30 of 834), a fourfold spectrum created from the identical clinical information.
Autopsy data, if uniformly obtained in all deaths, could improve the precision of a clinical classification system. Unfortunately, it required a great effort to obtain autopsy information in even one fourth of the deaths; thus, this effort may represent a “best-case scenario” for a clinical trial. Autopsy data are especially useful in identifying noncardiac causes of unwitnessed, unexpected sudden deaths.
ICD databases such as the one in this report focus on sudden cardiac death as the primary end point, but such trials lack a parallel control group. Based on this report, there are pitfalls when attempting a comparison of the sudden death rates in separate uncontrolled ICD databases because of the variations in death classification methods as well as the concern for baseline demographic comparability. The large potential for misclassification and the competing cardiac and noncardiac mechanisms involved in many deaths support the wisdom of considering total mortality as the only interpretable end point of cardiovascular clinical trials.25
- Received June 26, 1995.
- Revision received September 11, 1995.
- Accepted September 18, 1995.
- Copyright © 1996 by American Heart Association
Cohn JN, Archibald DG, Ziesche S, Franciosa JA, Harston WE, Tristani FE, Dunkman WB, Jacobs W, Francis GS, Flohr KH. Effect of vasodilator therapy on mortality in chronic congestive heart failure: results of a Veterans Administration cooperative study. N Engl J Med. 1986;314:1547-1552.
Ruberman W, Weinblatt E, Goldberg JD, Frank CS, Chaudhary BS, Shapiro S. Ventricular premature complexes and sudden death after myocardial infarction. Circulation. 1981;64:297-305.
Moss AJ, Davis HT, DeCamilla J, Bayer LW. Ventricular ectopic beats and their relation to sudden and nonsudden cardiac death after myocardial infarction. Circulation. 1979;60:998-1003.
Hinkle LE Jr, Thaler HT. Clinical classification of cardiac deaths. Circulation. 1982;65:457-464.
WHO Scientific Group. The pathological diagnosis of acute ischemic heart disease. WHO Tech Rep Ser. 1970;441:5-27.
Joint International Society and Federation of Cardiology/World Health Organization Task Force on Standardized of Clinical Nomenclature. Report: nomenclature and criteria for diagnosis of ischemic heart disease. Circulation. 1979;59:607-609.
Marcus FL, Cobb LA, Edwards JE, Kuller L, Moss AJ, Bigger T, Fleiss JL, Rolnitzky L, Serokman R, and the Multicenter Post Infarction Research Group. Mechanism of death and prevalence of myocardial ischemic symptoms in the terminal event after myocardial infarction. Am J Cardiol. 1988;6:8-15.
Greene HL, Richardson DW, Barker AH, Roden DM, Capone RJ, Echt DS, Friedman LM, Gillespie MJ, Hallstrom AP, Verter J, and the CAPS Investigators. Classification of deaths after myocardial infarction as arrhythmic or nonarrhythmic (the Cardiac Arrhythmia Pilot Study). Am J Cardiol. 1989;63:1-6.
Conover WG. Practical Nonparametric Statistics. 2nd ed. New York, NY: John Wiley & Sons, Inc; 1980.
Antiarrhythmics Versus Implantable Defibrillators (AVID) Research Protocol. Bethesda, Md: National Heart, Lung, and Blood Institute, HC-92-10; 1993:18.
Kim SG, Fogoros RN, Furman S, Connolly SJ, Kuck KH, Moss AJ. Standardized reporting of implantable cardioverter defibrillator patient outcome: the report of a North American Society of Pacing and Electrophysiology Policy Conference, February 9-10, 1993. PACE Pacing Clin Electrophysiol. 1993;16:1358-1362.