Risk of sudden versus nonsudden cardiac death in patients with coronary artery disease

Final Reports from the AHRQ Sudden Death Patient Outcomes Research Team

Risk of sudden versus nonsudden cardiac death in patients with coronary artery disease Nathan Every, MD,a Alfred Hallstrom, PhD,a Kathryn M. McDonald, MM,b,c Lori Parsons,a David Thom, MD, PhD,c Douglas Weaver, MD,d and Mark A. Hlatky, MDb,c Seattle, Wash, Stanford, Calif, and Detroit, Mich

Background Patients at high risk of sudden cardiac death, yet at low risk of nonsudden death, might be ideal candidates for antiarrhythmic drugs or devices. Most previous studies of prognostic markers for sudden cardiac death have ignored the competitive risk of nonsudden cardiac death. The goal of the present study was to evaluate the ability of clinical factors to distinguish the risks of sudden and nonsudden cardiac death.

Methods We identified all deaths during a 3.3-year follow-up of 30,680 patients discharged alive after admission to the cardiac care unit of a Seattle hospital. Detailed chart reviews were conducted on 1093 subsequent out-of-hospital sudden deaths, 973 nonsudden cardiac deaths, and 442 randomly selected control patients. Results Patients who died in follow-up (suddenly or nonsuddenly) were significantly different for many clinical factors from control patients. In contrast, patients with sudden cardiac death were insignificantly different for most clinical characteristics from patients with nonsudden cardiac death. The mode of death was 20% to 30% less likely to be sudden in women, patients who had angioplasty or bypass surgery, and patients prescribed ␤-blockers. The mode of death was 20% to 30% more likely to be sudden in patients with heart failure, frequent ventricular ectopy, or a discharge diagnosis of acute myocardial infarction. A multivariable model had only modest predictive capacity for mode of death (c-index of 0.62).

Conclusion Standard clinical evaluation is much better at predicting overall risk of death than at predicting the mode of death as sudden or nonsudden. (Am Heart J 2002;144:390-6.)

Sudden cardiac death (SCD) remains a major cause of death in the United States, despite the considerable advances over the past 2 decades in the prevention and treatment of heart disease.1-3 Most SCDs occur in patients with clinically recognized heart disease, particularly previous myocardial infarction and congestive heart failure.4-7 The identification of patients at high risk of SCD might facilitate specific treatment aimed at cardiac arrhythmias, the proximate cause of SCD in most instances. Recent clinical trials have demonstrated the efficacy of amiodarone8 and the implantable cardioverter-defibrillator (ICD)9,10 in specific patient groups at high risk of SCD.

From the aDepartment of Biostatistics, Northwest Health Services Research and Development Center of Excellence, VA Puget Sound Healthcare System, University of Washington, Seattle, Wash, the Departments of bHealth Research and Policy and cMedicine, Stanford University, Stanford, Calif, and the dDivision of Cardiology, Henry Ford Healthcare System, Detroit, Mich. Supported by grant HS08362 from the Agency for Health Care Policy and Research, Rockville, Md, and the Department of Veterans Affairs Health Services Research and Development. Submitted May 22, 2000; accepted October 26, 2000. Reprint requests: Mark A. Hlatky, MD, Stanford University School of Medicine, HRP Redwood Building, Room 150, Stanford, CA 94305-5405. E-mail: [email protected] © 2002, Mosby, Inc. All rights reserved. 0002-8703/2002/$35.00 ⫹ 0 4/1/125495 doi:10.1067/mhj.2002.125495

Epidemiologic studies have established various clinical factors as markers of high risk of SCD, including prior myocardial infarction, reduced left ventricular ejection fraction (LVEF), and evidence of ventricular arrhythmias.11-14 Almost all previous investigations have measured the risk of SCD by comparison with surviving patients and have omitted patients who died nonsuddenly of cardiac causes. There is, however, evidence that patients at high risk of SCD are also at high risk of non-SCD. It is possible that clinically evident risk factors for SCD simply indicate high risk of cardiac death in general, with similar elevation in the risk of both SCD and non-SCD. It is not clear that any clinical risk factor indicates a disproportionate risk of SCD, such as a 2-fold increase in the risk of SCD without a higher risk of non-SCD. Patients at high risk of SCD and low risk of non-SCD should, in principle, be ideal candidates for specific therapies aimed at cardiac arrhythmias. An implantable cardioverter defibrillator (ICD), for example, might provide particular benefit to such patients, because the reduction in SCDs by the ICD would not be offset by an increase in non-SCDs (eg, due to heart failure). The goal of this study was to identify clinical risk factors that characterize patients specifically at high risk of SCD and low risk of non-SCD.

American Heart Journal Volume 144, Number 3

Methods This study was based on an analysis of 30,680 patients discharged alive with suspected acute myocardial infarction from 1 of 16 Seattle area hospitals between 1988 and 1993. Patients discharged with either noncardiac diagnoses or the diagnosis of nonspecific or atypical chest pain were excluded. Characteristics of the Myocardial Infarction Triage and Intervention (MITI) Registry, data-gathering procedures, and reliability have been previously described.15 For patients transferred between institutions during the index hospitalization, medical records were abstracted at both facilities so that each patient had a continuous care record. The study was approved by the University of Washington and State of Washington Human Subjects Review Committee. Trained abstractors collected data from patient records by use of standardized forms and a manual of operation. Demographic variables included age, sex, and race. Prehospital variables included type of transport to the hospital (911 call or other) and duration of cardiac symptoms before emergency department evaluation. Information from the medical history included noncardiac comorbidity, alcohol and tobacco use, prior myocardial infarction, heart failure, angina, hypertension, percutaneous coronary angioplasty, bypass surgery, and atrial or ventricular arrhythmia. Data on hospital course included location of the infarct on electrocardiography (ECG), vital signs on admission, details of arrhythmia during the hospitalization, new evidence of congestive heart failure, shock, infarct extension, recurrent chest pain, use of thrombolytic therapy, cardiac catheterization, coronary angioplasty, or bypass surgery. Results of any exercise treadmill tests, echocardiograms, LVEF, and coronary angiography were recorded. Data were also collected on the results of ambulatory ECG, signal-averaged ECGs, and electrophysiologic testing, but these tests were performed too infrequently (⬍1% of patients) to judge their value in risk assessment.

Design Deaths during follow-up were identified by linking the MITI Registry to the State of Washington Comprehensive Hospital Abstract Reporting System (CHARS), which includes death certificate data, and to the National Death Index (NDI). Washington State death certificates do not include an “immediate” cause of death. Instead, the State uses a combination of a computer program and a trained nosologist to translate the “immediate” and “underlying” causes of death into a single underlying cause of death that is assigned an International Classification of Diseases-9 code. As a result of this methodology, Washington State death certificate data do not classify a death as the result of cardiac arrest but rather as the result of the presumed underlying cause of the cardiac arrest. This methodology is consistent with World Health Organization policies and is practiced throughout the United States. To identify episodes of ventricular arrhythmia in follow-up, we linked the MITI Registry to paramedic databases of the City of Seattle and King County, which include data on condition of patient on arrival, initial cardiac rhythm, and the medical or surgical diagnosis. The paramedic databases include all patients who underwent attempted resuscitation but not those who were classified as “dead on arrival.” The data

Every et al 391

from ECG rhythm strips and prehospital reports were reviewed by a paramedic quality assurance committee that includes the medical director. Medical diagnoses and mechanism of illness were determined by the paramedics at the scene but were reviewed and edited by study personnel. For patients who survived to hospitalization, hospital records were reviewed, and paramedic prehospital reports may be edited to include information obtained during the hospitalization and/or autopsy. For purposes of this study, SCD patients were identified by 1 of 2 criteria. Paramedic-defined SCD cases were patients evaluated by Seattle or King County paramedics who were resuscitated from cardiac arrest or in either ventricular fibrillation (VF), ventricular tachycardia (VT), asystole, or idioventricular rhythm on paramedic arrival as well as a cardiac mechanism of illness. Death certificate SCD was defined16 as death occurring either at home or in a public place as a result of an underlying medical cause (excluding trauma, suicide, and cancer deaths). For the purpose of this analysis, SCD cases include both paramedic and death certificate SCD. Non-SCD was defined as death after hospital admission with a cardiac cause as the discharge diagnosis.

Statistical analysis To identify risk factors that uniquely identified patients at risk for SCD, we performed a series of Cox regression models restricted to patients who died of cardiac causes after hospital discharge and were classified as either sudden or non-SCD. The first model included data available on all patients, including demographics, history, and hospital course. To assess the incremental prognostic value of various tests, we first determined the prognostic value of selection to undergo a particular test and then assessed the value of the specific results of that test. Separate models were used to assess the prognostic value of LVEF, coronary angiography, and echocardiography. All models were repeated to analyze only SCD patients defined by the paramedic evaluation (ie, excluding the death certificate patients).

Results Outcomes A total of 30,680 patients were discharged alive after admission for suspected acute myocardial infarction. After a mean follow-up of 3.3 years, 4820 (16%) of these patients had died, and an additional 85 patients had been resuscitated from an out-of-hospital cardiac arrest. The cause of death was cardiac in 2686 patients, noncardiac in 1040 patients (444 as a result of neoplasms, 106 as a result of chronic obstructive lung disease, 91 as a result of diabetes, 65 as a result of pneumonia, 63 as a result of trauma, and 271 as a result of other specified causes), and unknown in 1094 patients. The cardiac deaths included 490 sudden outof-hospital deaths with an attempted resuscitation by the paramedics and 603 sudden out-of-hospital deaths without an attempted resuscitation. Non-SCDs included 975 patients who died of cardiac causes after hospital admission. An additional 618 cardiac deaths

American Heart Journal September 2002

392 Every et al

Table I. Baseline characteristics P

Age (mean ⫾ SD) Sex (female) (%) White (%) Married (%) Previous MI (%) Heart failure (%) Hypertension (%) Diabetes (%) Prior bypass surgery (%) Syncope (%) Current smoker (%) Alcohol abuse (%)

Control (n ⴝ 442)

SCD case (n ⴝ 1093)

Non-SCD case (n ⴝ 973)

SCD vs control

SCD vs non-SCD

66.0 ⫾ 13.3 37.8 91.0 61.9 28.1 18.1 57.4 15.4 11.8 2.9 25.4 9.0

70.7 ⫾ 12.2 35.7 89.3 55.4 37.0 37.2 54.0 25.5 16.7 3.9 26.2 10.8

73.2 ⫾ 10.9 45.4 93.6 57.0 38.5 38.5 56.1 29.4 18.4 4.3 20.7 6.0

⬍.001 – – .02 ⬍.001 ⬍.001 – ⬍.001 .02 – – –

⬍.001 ⬍.001 ⬍.001 – – – – .05 – – .004 ⬍.001

MI, Myocardial infarction.

were not classified as sudden or nonsudden because the death either occurred in a nursing home or the location and timing of death could not be established.

Univariate comparison We performed detailed chart reviews on a sample of the patients. We reviewed all 575 cases of out-of-hospital cardiac arrests with attempted resuscitation by the paramedics (85 successfully resuscitated, 203 unsuccessfully resuscitated from ventricular tachycardia/fibrillation, and 287 unsuccessfully resuscitated from asystole or electromechanical dissociation), and 518 cases randomly selected from the 603 cases of out-ofhospital sudden death without an attempted resuscitation. We performed the same chart review on all 975 patients identified as having non-SCD and for 442 randomly selected control patients. To evaluate factors that could differentiate SCD from non-SCD after discharge, we compared detailed data in 1093 SCD, 973 non-SCD, and 442 control patients. Patients who died of cardiac causes, either suddenly or nonsuddenly, were significantly different from randomly selected control patients on multiple demographic and clinical variables (Tables I-IV). In particular, patients who subsequently died were significantly older, more likely to have had a prior myocardial infarction or congestive heart failure, and more likely to have diabetes (Table I). Patients who died, either suddenly or nonsuddenly, were more likely to have had evidence of heart failure or pulmonary edema during the index hospitalization and were more likely to have had atrial arrhythmias (Table II) but, surprisingly, were not more likely to have had ventricular arrhythmias during the index admission (Table II). Patients who died had more extensive coronary disease and lower

LVEFs (Table III), were less likely to have been discharged on ␤-blockers or aspirin, and were more likely to have received angiotensin-converting enzyme inhibitors, digoxin, and diuretics (Table IV). In contrast to the numerous significant differences between all of the patients who died and randomly selected control patients, there were very few factors that distinguished patients with subsequent SCD from patients with subsequent non-SCD (Tables I-IV). SCD cases were significantly younger, less likely to be female or white, and more likely to smoke cigarettes and consume alcohol (Table I). The hospital course (Table II) of SCD cases was more likely to include monomorphic premature ventricular contractions, but other ventricular arrhythmias were not significantly different. Cardiac test results (Table III) and discharge medications (Table IV) of patients who died suddenly were in large part quite similar to those of patients who died nonsuddenly.

Multivariate comparisons To identify factors known at the time of index hospitalization that would predict and differentiate sudden from non-SCD, we performed a series of Cox models limited to the patients who died after discharge (Table V). In the first model, we analyzed factors known for most patients at the time of discharge. In this model, several variables were statistically significant, but none was a strong predictor of mode of death. Female sex (hazard ratio 0.78), being married (0.78), undergoing coronary angioplasty, undergoing bypass surgery, the prescription of ␤-blockers at discharge, and slower discharge heart rate each predicted a slightly lower likelihood of SCD versus non-SCD. History or symptoms of heart failure during the hospitalization (hazard

American Heart Journal Volume 144, Number 3

Every et al 393

Table II. Presentation and hospital course P

ECG (%) Segment elevation ST-segment depression Normal ECG Syncope (%) Clinical shock (%) CHF, pulmonary edema, or rales on admission (%) Documented heart block or bradycardia (%) CHF/PE during hospital stay (%) Shock after admission (%) Hospital arrhythmia (%) VF Sustained VT Nonsustained VT Frequent polymorphic PVCs Frequent monomorphic PVCs Atrial fibrillation flutter Last recorded pulse prior to discharge (mean ⫾ SD)

Control (n ⴝ 442)

SCD case (n ⴝ 1093)

Non-SCD case (n ⴝ 973)

SCD vs control

SCD vs non-SCD

22.1 29.3 13.6 3.9 0.2 25.8

27.3 36.3 2.1 4.2 0.6 48.9

20.8 38.7 2.4 3.9 0.9 49.6

.04 .01 ⬍.001 – – ⬍.001

⬍.001 – – – – –

6.3

7.9

7.7

–

–

27.2 2.3

54.9 4.4

54.7 3.0

⬍.001 .05

– –

0.9 1.6 4.1 2.3 13.4 14.5 76.0 ⫾ 14.2

1.1 1.9 4.6 4.3 12.3 22.7 78.5 ⫾ 14.4

0.7 1.5 5.1 4.8 8.8 22.9 77.2 ⫾ 13.4

– – – .06 – ⬍.001 .002

– – – – .01 – .03

CHF, Congestive heart failure; PE, pulmonary edema; PVC, premature ventricular contraction.

ratio 1.2), frequent premature ventricular contractions (hazard ratio 1.2), and a discharge diagnosis of acute myocardial infarction (hazard ratio 1.3) predicted a slightly higher likelihood of SCD after discharge. The predictive ability of the model containing all variables was only modest, however, with a c-index (equivalent to the area under a receiver operating curve) of 0.62, where 0.5 indicates no discrimination and 1.0 perfect discrimination. Tests such as coronary angiography and echocardiography were performed only on selected patients. To separate the prognostic information conveyed by selection for a test from the actual test results, we performed a series of models. In the first stage, we tested in the entire patient sample the prognostic significance of selection for coronary angiography, echocardiography, or measurement of LVEF. There was no association between the mode of death (sudden vs nonsudden) and selection for angiography (hazard ratio 0.9, 95% CI 0.73-1.13), echocardiography (hazard ratio 1.0, 95% CI 0.84-1.24), or measurement of LVEF (hazard ratio 1.05, 95% CI 0.85-1.29). Once the effect of selection for testing was included in the model, the actual test results carried little prognostic information. In the model that included LVEF data (available in 48.5% of patients), there was no association between low LVEF and mode of death (hazard ratio of LVEF ⬍30% 1.001, 95% CI 0.8-1.3). Further, in

a model that included an imputed LVEF for patients without an LVEF measurement, we found no association between lower LVEF and mode of death. In the model that included angiography data (available in 22.2% of patients), there was no association between the number of diseased vessels of ⬎70% and mode of death (hazard ratio 1.1, 95% CI 0.9-1.3). Finally, in the model that included echocardiographic data (available in 28.8% of patients), none of the echocardiographic variables collected were able to differentiate between SCD and non-SCD: moderate or severe left ventricular hypertrophy hazard ratio 0.93 (95% CI 0.7-1.2), left ventricular end diastolic dimension hazard 0.99 (95% CI 0.98-1.00), and moderate or severe mitral regurgitation hazard 1.2 (95% CI 0.9-1.6).

Discussion The major finding of this study is that patients who experience SCD have a clinical profile very similar to that of patients who die nonsuddenly, despite the striking differences between either group and the surviving patients. Most risk factors for SCD appear to be predictors of the overall chance of cardiac death rather than an indicator of the specific mode of death. Standard clinical evaluation does not appear to provide the information needed to identify patients who are

American Heart Journal September 2002

394 Every et al

Table III. Cardiac test results P

Coronary angiography stenosis of Left main (%) LAD system (%) LCX system (%) RCA system (%) Number of vessels ⱖ70% stenosis (mean ⫾ SD) LV ejection fraction (mean ⫾ SD) Echocardiogram data Left ventricular end diastolic diameter (mm) (mean ⫾ SD) Intraventricular septum dimension (diastolic) (mean ⫾ SD) Left ventricular hypertrophy (%) None Mild (1⫹) Moderate (2⫹) Severe (3⫹) Mitral regurgitation (%) Moderate or severe

Control (n ⴝ 442)

SCD case (n ⴝ 1093)

Non-SCD case (n ⴝ 973)

SCD vs control

SCD vs non-SCD

3.2 49.1 42.7 52.2 1.4 ⫾ 1.0

9.1 65.0 55.0 59.3 1.8 ⫾ 1.0

8.1 72.9 59.5 65.3 2.0 ⫾ 1.0

.03 .002 .02 – ⬍.001

– – – – –

51.7 ⫾ 15.1

40.8 ⫾ 15.7

42.5 ⫾ 15.8

⬍.001

–

52.4 ⫾ 10.0

57.3 ⫾ 11.4

55.2 ⫾ 10.7

⬍.001

.02

11.4 ⫾ 3.2

11.8 ⫾ 7.5

11.8 ⫾ 4.6

–

–

56.3 13.5 10.1 3.4

52.0 17.2 11.1 3.2

47.7 20.0 13.9 5.5

– – – –

– – – –

27.7

38.7

39.7

.03

–

LAD, Left anterior descending coronary artery; LCX, left circumflex artery; RCA, right coronary artery; LV, left ventricle.

Table IV. Discharge medications P

Calcium antagonists (%) ␤-Blockers (%) Antiarrhythmics (%) Diuretics (%) Digoxin (%) Aspirin (%) ACE inhibitors (%)

Control (n ⴝ 442)

SCD case (n ⴝ 1093)

Non-SCD case (n ⴝ 973)

SCD vs control

SCD vs non-SCD

44.8 22.6 11.3 32.8 26.2 51.1 23.3

39.9 14.4 12.3 59.0 41.7 40.9 35.4

48.5 17.8 13.1 61.6 41.8 40.7 36.3

– ⬍.001 – ⬍.001 ⬍.001 ⬍.001 ⬍.001

⬍.001 .03 – – – – –

ACE, Angiotensin-converting enzyme.

simultaneously at high risk of SCD and low risk of nonSCD. Defining the specific risk of SCD is of interest for several reasons. Deaths that result from cardiac arrhythmias are generally sudden and unexpected and might be prevented by treatment with drugs such as amiodarone or devices such as the ICD. In principle, the greatest benefit of these therapies will be among patients at high risk of SCD, but this benefit may not be realized if the patient also has a high risk of nonSCD. In such instances, the potential to extend survival meaningfully may be limited by the competing

risk of non-SCD, such as heart failure. The main potential for benefit of specific antiarrhythmia/arrhythmic therapy arises when the risk of SCD is substantially higher than that of non-SCD. Previous studies of clinical and laboratory risk markers of SCD have almost always contrasted cases of SCD with surviving control patients, not with cases of nonSCD. The few other studies that have compared cases of SCD and non-SCD have generally found few, if any, features that distinguish between these modes of death. In a combined analysis of the Albany and Framingham studies, no combination of risk factors distin-

American Heart Journal Volume 144, Number 3

guished SCD from non-SCD.17 The National Mortality Followback Survey found that current smokers had a 1.3 times greater chance of SCD than non-SCD but no other differences in modifiable risk factors.3 A more recent community-based study in Britain found the risk factors for SCD and non-SCD were generally similar, except for an association of heavy alcohol intake with sudden death.18 These studies both included subjects without preexisting heart disease and focused more on traditional cardiac risk factors than clinical measures of coronary artery disease severity. One study from the Coronary Artery Surgery Study (CASS) Registry19 found no clinical factors that differentiated the 557 SCDs from the 813 non-SCDs during a 5-year follow-up of medically treated patients with coronary artery disease. Similar findings were reported in a smaller cohort of patients from the Montreal Heart Institute.20 Our study confirms this previous work and extends it by studying more cases (with greater statistical power) and by enrolling more contemporary patients, many of whom were treated with reperfusion therapy. Newer tests may provide better tools for risk stratification than we were able to evaluate here. Heart rate variability,21,22 baroreflex sensitivity,22 and the signalaveraged ECG23 have recently shown promise for the identification of SCD risk. Our results suggest that it will be critical to evaluate these tests by their ability to distinguish cases of SCD from non-SCD. Most studies to date have only examined the ability of these tests to distinguish sudden death cases from survivors, a far less stringent standard.

Study limitations Although our findings suggest that traditional risk factor evaluation does not differentiate SCD from nonSCD in patients with known or presumed coronary disease, there are limitations to our analysis. Our data were obtained from retrospective chart review, and not all data were available for all patients. In particular, LVEF, cardiac catheterization, and echocardiographic data were available in only a subset of patients. Thus, our results illustrate the limited value of these tests as used in usual care. Whether these tests, if performed in all patients at risk, might differentiate SCD from non-SCD could not be answered with this study. However, adding a variable of whether a cardiac test such as cardiac catheterization or echocardiography was completed during the index hospitalization did not differentiate SCD from non-SCD. Our data on cardiac arrhythmia during hospital admission were obtained by reviewing telemetry strips and progress notes in the chart, which are not as sensitive as Holter monitor data. Our definition of SCD has been validated against paramedic reports; however, there is some misclassification in our death certificate– based system. Bias in our classification scheme would tend to classify

Every et al 395

Table V. Multivariable independent predictors of the hazard of sudden death versus the hazard of nonsudden death SCD vs non-SCD Factor Female Married Bypass surgery during admission Angioplasty during admission ␤-Blockers on discharge Heart failure during admission Frequent PVCs Discharge heart rate ⬎90 beats/min Discharge diagnosis of MI

Hazard ratio

95% CI

0.78 0.78 0.70 0.72 0.76 1.24 1.18 1.20 1.27

0.67–0.90 0.68–0.89 0.51–0.96 0.53–0.99 0.63–0.92 1.05–1.46 1.00–1.39 1.01–1.43 1.10–1.48

Factors included in the model but not statistically significant (with hazard ratio for sudden versus nonsudden death in parentheses) were age (1.00), white race (1.02), diabetes (0.91), history of MI (0.90), history of heart failure (1.05), prior coronary artery bypass graft (0.88), history of syncope (0.95), current smoking (1.04), alcohol abuse (1.13), hypotension on admission (1.38), ST elevation (1.13), atrial fibrillation or flutter during hospitalization (1.12), and ventricular fibrillation or tachycardia during hospitalization (1.31).

patients with SCD when the deaths were not arrhythmic. To address this limitation, we performed a subgroup analysis where SCD classification was limited to patients with witnessed cardiac arrest by paramedics. In this analysis, risk factors for SCD were not substantially different from the main analysis.

Conclusions Clinical risk assessment can identify patients at increased risk of cardiac death in general but have limited ability to predict the mode of death as sudden or nonsudden. Evaluation of newer tests for the prediction of SCD should compare SCD with non-SCD patients.

References 1. Zipes DP, Wellens HJJ. Sudden cardiac death. Circulation 1998; 98:2334-51. 2. Myerburg RJ, Kessler KM, Castellanos A. Sudden cardiac death: epidemiology, transient risk, and intervention assessment. Ann Intern Med 1993;119:1187-97. 3. Escobedo LG, Zack MM. Comparison of sudden and nonsudden coronary deaths in the United States. Circulation 1996;93:2033-6. 4. Schaffer WA, Cobb LA. Recurrent ventricular fibrillation and modes of death in survivors of out-of-hospital ventricular fibrillation. N Engl J Med 1975;293:259-62. 5. Weaver WD, Lorch GS, Alvarez HA, et al. Angiographic findings and prognostic indicators in patients resuscitated from sudden cardiac death. Circulation 1976;54:895-900. 6. Goldstein S, Landis JR, Leighton R, et al. Characteristics of the resuscitated out-of-hospital cardiac arrest victim with coronary heart disease. Circulation 1981;64:977-84. 7. Myerburg RJ, Kessler KM, Estes D, et al. Long-term survival after prehospital cardiac arrest: analysis of outcome during an 8 year study. Circulation 1984;70:538-46.

396 Every et al

8. Sim I, Gupta M, McDonald K, et al. A meta-analysis of randomized trials comparing coronary artery bypass grafting with percutaneous transluminal coronary angioplasty in multivessel coronary artery disease. Am J Cardiol 1995;76:1025-9. 9. Moss AJ, Hall WJ, Cannom DS, et al. Improved survival with an implanted defibrillator in patients with coronary disease at high risk for ventricular arrhythmia. N Engl J Med 1996;335:1933-40. 10. The Antiarrhythmics Versus Implantable Defibrillators (AVID) Investigators. A comparison of antiarrhythmic-drug therapy with implantable defibrillators in patients resuscitated from near-fatal ventricular arrhythmias. N Engl J Med 1997;337:1576-83. 11. Shaw LJ, Peterson ED, Kesler K, et al. A metaanalysis of predischarge risk stratification after acute myocardial infarction with stress electrocardiographic, myocardial perfusion, and ventricular function imaging. Am J Cardiol 1996;78:1327-37. 12. Peterson ED, Shaw LJ, Califf RM. Risk stratification after myocardial infarction. Ann Intern Med 1997;126:561-82. 13. Volpi A, DeVita C, Franzosi MG, et al. Determinants of 6-month mortality in survivors of myocardial infarction after thrombolysis: results of the GISSI-2 data base. Circulation 1993;88:416-29. 14. The Multicenter Postinfarction Research Group. Risk stratification and survival after myocardial infarction. N Engl J Med 1983;309: 331-6. 15. Weaver WD, Litwin PE, Martin JS. Use of direct angioplasty for treatment of patients with acute myocardial infarction in hospitals with and without on-site cardiac surgery. The Myocardial Infarction, Triage, and Intervention Project Investigators. Circulation 1993;88:2067-75.

American Heart Journal September 2002

16. Every NR, Parsons L, Hlatky MA, et al. Use and accuracy of state death certificates for classification of sudden cardiac death in high-risk populations. Am Heart J 1997;134:1129-32. 17. Doyle JT, Kannel WB, McNamara PM, et al. Factors related to suddenness of death from coronary disease: combined AlbanyFramingham studies. Am J Cardiol 1976;37:1073-8. 18. Wannamethee G, Shaper AG, Macfarlane PW, et al. Risk factors for sudden cardiac death in middle-aged British men. Circulation 1995;91:1749-56. 19. Holmes DR, Davis K, Gersh BJ, et al. Risk factor profiles of patients with sudden cardiac death and death from other cardiac causes: a report from the Coronary Artery Surgery Study (CASS). J Am Coll Cardiol 1989;13:524-30. 20. Puddu PE, Bourassa MG, Lesperance J, et al. Can the mode of death be predicted in patients with angiographically documented coronary artery disease? Clin Cardiol 1983;6:384-95. 21. Zuanetti G, Neilson JMM, Latini R, et al. Prognostic significance of heart rate variability in post-myocardial infarction patients in the fibrinolytic era: the GISSI-2 results. Circulation 1996;94:432-6. 22. LoRovere MT, Bigger JT, Marcus FI, et al. Baroreflex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction. Lancet 1998;351:478-84. 23. Hartikainen JEK, Malik M, Staunton A, et al. Distinction between arrhythmic and nonarrhythmic death after acute myocardial infarction based on heart rate variability, signal-averaged electrocardiogram, ventricular arrhythmias and left ventricular ejection fraction. J Am Coll Cardiol 1996;28:296-304.