Prognostic significance of nonfatal myocardial reinfarction

253

JACC Vol. 15, No.2

February 1990:253-8

COOPERATIVE STUDIES

Prognostic Significance of Nonfatal Myocardial Reinfarction JESAIA BENHORIN, MD, ARTHUR J. MOSS, MD, FACC, DAVID OAKES,

PHD,*

AND THE MULTICENTER DILTIAZEM POSTINFARCTION TRIAL RESEARCH GROUP

Rochester, New York

In most risk stratification and intervention postinfarction trials, cardiac mortality is used as the major outcome end point either alone or in combination with nonfatal reinfarction. However, the independent risk carried by nonfatal reinfarction for subsequent cardiac death has not been quantified. The prognostic significance of nonfatal reinfarction was determined from the multicenter diltiazem trial data base of 1,234 patients treated with placebo foUowed up for 1 to 4 years after acute myocardial infarction. One hundred sixteen patients had at least one nonfatal reinfarction, 14 (12%) of whom subsequently experienced cardiac death. Of the remaining 1,118 patients without nonfatal reinfarction, 110 (9.8%) experienced cardiac death. Compared with event-free patients, patients with nonfatal reinfarction were more likely (p < 0.05) to be women, to have had an infarction before their index event and to have had prior cardiac-related symptoms.

In most risk stratification and intervention postinfarction trials, cardiac mortality is used as a major outcome end point, but it is often combined with nonfatal reinfarction to form a single composite end point, the first recurrent cardiac event (1-9). Use of this combined end point is based on the assumption that similar pathophysiologic mechanisms (namely, ischemia) operate in both, and that nonfatal reinfarction increases the subsequent risk for cardiac mortality, mainly by increasing the degree of mechanical dysfunction. The relation between nonfatal reinfarction and subsequent cardiac death has not been evaluated in detail or quantified, although an understanding of it is essential for secondary

From the Heart Research Follow-Up Program and the *Division of Biostatistics, University of Rochester School of Medicine and Dentistry, Rochester, New York. This study was supported in part by a consortium grant from Godecke Aktiengesellschaft, West Germany; Laboratorios de Dr Esteve, SA, Spain; Marion Laboratories, Inc., USA; Nordic Laboratories, Inc., Canada; Lars Synthelabo, France; Tanabe Seiyaku Co., Ltd., Japan and Warner-Lambert International, USA. Manuscript received December 29, 1988; revised manuscript received April 9, 1989, accepted September 19, 1989. Address for reprints: Jesaia Benhorin, MD, Box 653, University of Rochester Medical Center, Rochester, New York 14642. ©199O by the American College of Cardiology

Cox survivorship analyses, using pertinent baseline clinical variables along with nonfatal reinfarction as a timedependent predictor variable, revealed that nonfatal reinfarction carried a significant and independent risk for subsequent cardiac mortality (hazard ratio 3.0, p =0.002), which was greater than that carried by other significant predictor variables (New York Heart Association functional class, pulmonary congestion on chest radiograph, blood urea nitrogen level, predischarge Holter-recorded ventricular premature complexes and radionuclide ejection fraction). The cardiac mortality risk associated with nonfatal reinfarction was further increased in patients whose index event was their first infarction (hazard ratio 5.4, p = 0.0006). Thus, nonfatal reinfarction carries a strong, significant and independent risk for subsequent cardiac death in patients surviving an acute myocardial infarction. (J Am Coil CardioI1990;15:253-8)

prevention strategies. The purpose of this study is to examine the clinical determinants of cardiac death and nonfatal reinfarction and to determine the independent contribution of nonfatal reinfarction to the risk of subsequent cardiac death.

Methods Study patients. A detailed description of the design, population, data acquisition, data management and followup of the Multicenter Diltiazem Postinfarction Trial has been previously reported (9). The population for the present study included only the placebo-assigned patients of the total trial population. There were 38 participating hospitals from 23 centers, 19 in the United StateS and 4 in Canada. The placebo group included 1,234 patients, aged 25 to 75 years, who had a documented acute myocardial infarction between February 1983 and June 1986. The diagnosis of acute myocardial infarction required serum enzyme confirmation in patients with symptoms or electrocardiographic changes, or both, suggestive of an acute coronary event. None of these patients received thrombolytic therapy. Patients had regular 0735-1097/90/$3.50

254

BENHORIN ET AL. NONFATAL REINFARCTION

follow-up until the common termination date of June 30, 1987. Follow-up duration ranged from 12 to 52 months (average 25 months per patient). Data acquisition. Data acquisition was accomplished as previously reported (9). The clinical variables collected on each patient included demographic data, prior cardiac history, coronary care unit course, the qualifying electrocardiogram (ECG) and evaluation by radionuclide ejection fraction measurement and 24 h Holter ECG monitoring, which were performed 7.9 ± 3.9 and 6.9 ± 3.0 days (mean ± SD) after infarction, respectively. Routine clinical variables were missing in <0.5% of the patients, radionuclide ejection fraction measurement was not performed in 12% and Holter monitoring was not performed in 13% of the patients. Definitions. Aftrst recurrent cardiac event was defined as a nonfatal reinfarction (nonfatal event) or cardiac death (fatal event), whichever occurred first. Patients who experienced cardiac death after surviving hospitalization for a nonfatal reinfarction were categorized in the nonfatal event group. Variables. The following variables were preselected and dichotomized as the principal, clinically meaningful covariates for the analyses reported here: age (~60 years), male gender, prior myocardial infarction (self-reported), New York Heart Association functional classification class II to IV I month before the index infarction, any grade of pulmonary congestion on chest radiograph, index infarction type (Q wave anterolateral, Q wave inferoposterior and non-Q wave) according to the Manhattan code (10), blood urea nitrogen (>35 mg/dl), radionuclide left ventricular ejection fraction «40%), Holter-recorded ventricular premature complexes (~IO/h) and occurrence of the first nonfatal reinfarction during follow-up. Nonfatal reinfarction was categorized by a three member subcommittee that reviewed all information obtained by the study coordinators regarding patients rehospitalized with acute coronary events. The criteria for reinfarction were the same as those used for defining the index infarction. Patients who had survived their reinfarction hospitalization were. categorized as having had a nonfatal reinfarction. Therapy and end points. Concurrent therapy was not controlled and was left to the discretion of the patient's personal physician. At the time of randomization to placebo therapy 53% of the patients were taking a beta-adrenergic blocking agent, 62% were taking nitrates and 9% were taking an antiarrhythmic agent. The primary end point of the analysis was cardiac death. A four member mortality subcommittee reviewed each death case using information obtained from relatives, witnesses, death certificates, attending physicians, hospital records and autopsy reports. Each death case was categorized as cardiac or noncardiac. Statistical analysis. Baseline characteristics of patients experiencing fatal and nonfatal first recurrent cardiac events were compared using a chi-square test. Separate comparisons were made comparing patients with no event with

lACC Vol. 15, No.2

February 1990:253-8

patients with a nonfatal event, and comparing patients with a fatal event with those with a nonfatal event. A two-sided p value of <0.05 was considered statistically significant. In these analyses, nonfatal reinfarction served as an end point. Cumulative reinfarction and cardiac mortality rates over time in different subsets were described by Kaplan-Meier curves (11), and differences between curves were assessed by the log-rank test (12). A stepwise logistic regression technique (13) was used to determine the clinical variables that best predicted the fatality of a given follow-up cardiac event. This analysis included only patients with an event, using the fatality or nonfatality of the event as an end point. To determine the independent risk carried by nonfatal reinfarctionfor subsequent cardiac death over time, we used the Cox proportional hazards survival model (14). In these analyses, nonfatal reinfarction served as a time dependent predictor variable. The independent risk for subsequent cardiac death that is carried by each predictor variable is reported in terms of hazard ratio-the ratio of the risk of experiencing a cardiac death per unit of time for patients with a certain predictor variable to the risk of patients without that predictor variable (while adjusting for all other variables in the model). The mortality experience of patients with nonfatal reinfarction was compared with that of a control group. This subset analysis included only patients without a history of a pre-enrollment infarction. The selection procedure of the control group was designed after that of Prentice (15). The control group consisted of five matched control patients for each patient with nonfatal reinfarction. Matching criteria included event-free survival time to reinfarction, left ventricular ejection fraction dichotomized at 40% and the presence or absence of pulmonary congestion on chest radiograph. After the matching procedure, control patients were randomly selected, five at a time, without replacement. To avoid retrospective bias, the matching technique allowed selection of patients with "late" nonfatal reinfarction as control patients for patients with "early" nonfatal reinfarction. Consequently, a patient who experienced a nonfatal reinfarction 6 months after his first infarction could be randomly selected as a control case for a patient who experienced a nonfatal reinfarction 2 months after his first infarction. Such a control patient was kept in the control group until reinfarction occurred, at which time he was censored from the control group and moved to the reinfarction group. Six independent sets of 5: 1 matched control groups were randomly selected. Comparisons of the cardiac mortality over time between the patients with nonfatal reinfarction and each of the six matched control groups were similar. A representative control group was selected for graphic cardiac mortality comparison with the nonfatal reinfarction group using the method of Kaplan and Meier (11). The reported analyses utilized the Multicenter Diltiazem

February 1990:253-8

0.20

r-Prior Infarction (n 0.15

~

« a:

r-

i=: 0.10 (,,)

-

Nonfatal Reinfarction No

Status

p = 0.0002

Survived Cardiac death Noncardiac death

:•

•

I

r


Table 1. Mortality Events in 1,234 Patients With and Without Nonfatal Reinfarction

I

37 events)

.

r"

;


~ 258,

... ------~ ,..----,jI ,._ ,.-"

,.~

a: ~

255

BENHORIN ET AL. NONFATAL REINFARCTION

lACC Vol. 15, No.2

(n

= 1,118)

(n

Yes = 116) 97 14 5

970 110 38

,I

ijj

ex:

r ,r I

0,05

r

0r----,----,------.----,--,-------,

o

200

400

600

800

1000

1200

DAYS FROM INDEX INFARCTION

Figure 1. Cumulative first nonfatal reinfarction rate in 1,233 patients with and without prior (pre-enrollment) infarction. Postinfarction Trial analytic data base released November 1, 1987.

Results Nonfatal reinfarction: frequency of occurrence and characteristics. One hundred sixteen patients (9.4%) experienced a nonfatal reinfarction as their first recurrent cardiac event during an average follow-up period of 25 months. The clinical presentation of patients with nonfatal reinfarction included typical central chest pressure in 96 patients (82.8%); EeG changes during reinfarction included new Q waves alone in 29 cases (25%) and new ST or T wave changes in 79 (68.1%). Reinfarction was procedure-related in only seven cases (6%). A first nonfatal reinfarction was followed by a second nonfatal reinfarction in 13 patients (11.2%) and by two nonfatal reinfarctions in 1 patient. The cumulative occurrence of nonfatal reinfarction over time in patients with and without history of an infarction before the index event is presented in Figure 1. Nonfatal reinfarction occurred more frequently and earlier among patients with a history of a pre-enrollment infarction than among those without (p = 0.0002). Mortality. Mortality (due to cardiac and noncardiac causes) in patients with and without nonfatal reinfarction is summarized in Table 1. A first nonfatal reinfarction was followed by subsequent cardiac death in 14 (12.1%) of 116 patients and by noncardiac death in five patients (4.3%). Among the 1,118 patients who did not experience nonfatal reinfarction, 110 (9.8%) had a cardiac death and 38 (3.4%) died from a noncardiac cause. Univariate associations between clinical characteristics and recurrent cardiac events. The clinical characteristics of patients without postinfarction cardiac events, with nonfatal

cardiac events and with fatal cardiac events are presented in Table 2. Patients with nonfatal reinfarction were more likely (p < 0.05) to be women and to have had a history of an infarction before the index event, prior cardiac symptoms and an index infarction that was not of the inferioposterior Q-wave type than were patients without any events. Markers of cardiac dysfunction (ejection fraction <40%, pulmonary congestion, blood urea nitrogen >35 mg/dO after the index infarction were more frequently observed (p < 0.05) among patients who had a cardiac death than among those who had a nonfatal reinfarction. Clinical predictors of first cardiac event fatality. To identify clinical variables that can predict the fatal outcome of a given recurrent cardiac event, we used a multiple logistic regression technique (13). Only patients with an event (cardiac death or nonfatal reinfarction) were entered into the model. The model included 191 patients who subsequently experienced a cardiac event during follow-up; of these 191 patients, 91 had cardiac death and 100 had nonfatal reinfarction as their first recurrent cardiac event. The same preselected 11 clinical variables as in Table 2 were entered as Table 2. Baseline Characteristics of 1,008 Patients Without a Cardiac Event, 116 Patients With Nonfatal Reinfarction and 110 Patients With a Fatal First Recurrent Cardiac Event First Recurrent Cardiac Event No

Nonfatal

Characteristic

(n = 1,008) (%)

(n = 116) (%)

Age (::=060 years) Male gender Prior MI NYHA class II to IV Acute anterolateral Q wave Acute inferoposterior Q wave Acute non-Q wave MI LVEF «40%) Pulmonary congestion VPCs (::=olO/h) BUN (>35 mgldl)

49 80 17 16 31 42 27 26 5 15 6

53 71 32

24 38 29 33 34 6 22

8

(n

Fatal = 110) (%)

65 81 42 33 38 34 28 61 17 32 23

BUN = blood urea nitrogen; LVEF = left ventricular ejection fraction; MI = myocardial infarction; NYHA class = New York Heart Association functional classification I month before entry; VPCs = ventricular premature complexes recorded on Holter monitor.

256

JACC Vol. 15, No.2 February 1990:253-8

BENHORIN ET AL. NONFATAL REINFARCTION

Table 3. Hazard Ratios for Postinfarction Cardiac Death by Baseline Clinical Characteristics and the Interim Occurrence of Nonfatal Reinfarction in All Patients (modell) and in Patients Without Prior Infarction (model 2) Modell (n = 934)

0.20

r--------

Model 2 (n = 747)

Predictor Variable

HR

95%CI

HR

95%CI

Nonfatal reinfarction NYHA class II-IV Pulmonary congestion LVEF «40%) VPCs (~IO/h) BUN «35 mg/dl)

3.0 1.9 1.3 2.4 2.3 2.3

(1.5,6.1) (1.2, 3.2) (0.8,2.2) (1.5, 3.9) (1.4, 3.7) (1.3,4.1)

5.4 1.6 2.0 2.0 2.4 1.8

(2.1, 14.4) (0.7,3.3) (1.1, 3.9) (1.1,3.6) (1.3,4.3) (0.8,4.2)

CI = confidence intervals; HR = hazard ratio; other abbreviations as in Table 2.

I

I

0.15 Non-fatal Reinfaretion (n = 67; 8 events) ...

~

:::i

I

~ 0

u

-J

I

I I

It ~

i I

I

r-~

0.10

I I I

S

I

Q

r---- ..

It

I

~

I

....__ J

p = 0.013

I I

.'

0.05

I

I

I I

~ I

1

predictor variables using first cardiac event fatality as the outcome end point. This procedure selected three variables as significant (p < 0.05) and independent clinical predictors of the fatality of a given first cardiac event: reduced «40%) left ventricular ejection fraction (chi square 15.7), increased (>35 mg%) blood urea nitrogen level (chi square 5.7) and an index Qwave inferior-posterior infarction (chi square 6.6). Nonfatal reinfarction as a time-dependent predictor of subsequent cardiac death. To determine the independent contribution of nonfatal reinfarction to the risk of subsequent cardiac death while adjusting for baseline clinical variables and the variability in time of occurrence of nonfatal reinfarction during follow-up, a time-dependent Cox proportional-hazards survival model technique was employed (14). Nonfatal reinfarction was entered into the different models as a time-dependent predictor variable along with the other pertinent baseline clinical variables. The results of two models are presented in Table 3. Nonfatal reinfarction was a significant and independent predictor of cardiac death in both models yet it carried more risk in patients without a history of a pre-enrollment infarction (hazard ratio 5.4) than in the total population (hazard ratio 3.0). An actuarial comparison (Kaplan-Meier) of the cardiac mortality rate in patients after a second nonfatal myocardial infarction and in a 5: 1matched control group is presented in Figure 2. The cardiac mortality rate was significantly greater (p = 0.013) over time in the reinfarction group than in the control population.

Discussion The principal findings of this study indicate that nonfatal reinfarction carries a significant and independent risk for subsequent cardiac mortality in postinfarction patients. The relative contribution of nonfatal reinfarction to the risk of cardiac death is higher in patients with a first myocardial infarction than in the total postinfarction population.

I

r'

:

200

400

600

800

1000

1200

DAYS FROM REINFARCTION

Figure 2. Cumulative cardiac mortality rate in 67 patients after a

second nonfatal myocardial infarction and in a 5: I matched control group (n = 335). See text for details.

Limitations. An important question is what influence concurrent therapy such as beta-blocking agents or revascularization procedures during follow-up might have had on the reported findings. Medications other than calcium channel blockers were prescribed by each patient's personal physician. Information concerning medication used in this population was available only at specific follow-up visits during the study period. The interpretation of the effect of medication usage such as beta blockers on reinfarction and mortality rates is too complex to be incorporated as a baseline covariate in the Cox model. Data concerning beta-blocker therapy discontinuation before or after the occurrence of nonfatal reinfarction were incomplete. However, we know that beta-blockers at randomization were utilized by approximately one-half of the study population, and the percentage of patients receiving beta-blockers was similar in the 79 patients with nonfatal reinfarction and in the 5: 1 matched control patients. Moreover, repeated Cox analyses including beta-blocker therapy at randomization as one of the baseline covariates gave similar results with regard to the independent contribution of nonfatal reinfarction to the risk of subsequent cardiac death. Thus, we believe that the use of beta-blockers in this population did not exert a bias on the mortality risk contributed by nonfatal reinfarction. Revascularization procedures were uncommon: 155 patients (12.6%) had coronary artery bypass graft surgery, and 19 patients (1.5%) had coronary angioplasty during the first year of follow-up. Of the 116 patients with nonfatal reinfarc-

lACC Vol. 15, No.2 February 1990:253-8

tion, 25 had coronary bypass surgery during the first year of follow-up. Of these 25 patients, only 6 had surgery before their reinfarction whereas 19 patients (76%) had surgery after their reinfarction. Reinfarction rate in the total followup period was higher in patients who had coronary bypass surgery during the first year (16.1%) than in those who did not (8.4%). In view of these findings, revascularization procedures were not included in the Cox analyses. Cardiac mortality and nonfatal reinfarction rates. The nonfatal reinfarction rate in our study was 7.7% in the first year after the index infarction. This is in accordance with previous studies (5-7,16) that reported rates of 3% to 21% during the first year. According to the cumulative reinfarction rate curves (Fig. 1), two main slopes could be identified: the steeper occurred 0 to 6 months after the index infarction and the less steep >6 months after the index event. The relative changes in the slope of the curve are similar for patients with and without prior infarction, yet reinfarction occurred earlier in patients with than in those without prior infarction. These time-dependent changes in nonfatal reinfarction rates are similar to mortality rate changes over time observed in several postinfarction populations (1,5-9). It is the combination of these two rates over time that describes coronary disease activity and severity. Univariate predictors of nonfatal reinfarction. The association of prior infarction and prior cardiac symptoms with the occurrence of nonfatal reinfarction has been reported previously (1,3,6) and is related to the fact that both variables are good clinical measures of the activity of the disease. The association of a non-Q wave index infarction (7,17,18) and of a Q wave inferoposterior index infarction (19) with the subsequent occurre~ce of nonfatal reinfarction has been reported previously as well, but the same association for an anterolateral Q-wave infarction has not. These univariate associations, however, must be interpreted with caution because they do not take into account the fact that nonfatal reinfarction occurrence rate is indirectly influenced by postinfarction mortality, which is mainly dependent on the varying degrees of mechanical dysfunction that are observed after these different index infarction types. Independent predictors of the fatality of a given cardiac event. Different measures of left ventricular dysfunction were associated with a fatal cardiac event, and two of them (reduced ejection fraction and elevated blood urea nitrogen level) were shown to independently predict fatality in a multivariate model. This is in agreement with the strong association of cardiac mortality with left ventricular dysfunction (2-4,6-9,20), which has been shown to operate in concert with ischemia during the terminal event (21) in postinfarction patients. The finding of an independent association between an index infarction of the Q-wave inferoposterior type and the fatality ofgiven events probably relates to the more extensive additional myocardial damage that can occur during a subsequent event in these patients.

BENHORIN ET AL. NONFATAL REINFARCTION

257

Nonfatal reinfarction as a predictor of cardiac mortality. To our knowledge, this is the first study that included the occurrence of nonfatal reinfarction as a time-dependent predictor variable for subsequent cardiac mortality. The time-dependent nonfatal reinfarction analysis includes all cardiac deaths among the group under study. In principle, the analysis can be characterized as follows. For each time (from randomization), a separate two-way classification is formed for all patients still at risk at that time. The "row" classification corresponds to whether or not a patient had suffered a nonfatal reinfarction before that time. The "column" classification corresponds to cardiac death or survival at that time. Association in this table implies that the occurrence of reinfarction changes the risk of cardiac death. Such a table can be constructed at each day during follow-up and a cardiac death hazard ratio calculated for nonfatal reinfarction. The Cox survival analysis combines these hazard ratios over time and has the additional capability of adjusting these ratios for differences in baseline characteristics. Thus, the reported hazard ratios for nonfatal reinfarction reflect the subsequent risk of experiencing cardiac death for patients with a nonfatal reinfarction as compared with the risk for patients without a nonfatal reinfarction. The independent contribution of nonfatal reinfarction to the risk of cardiac death was found to be highly significant and was further increased in patients without prior infarction. This can be accounted for by the relatively greater contribution of reduced ejection fraction to the mortality risk in patients with prior infarction who had significantly lower mean ejection fraction measurements than those of patients without prior infarction (40.7 ± 14.4 versus 48.1 ± 13.5, respectively, p < 0.0001). Identification of patients with subsequent reinfarction. Our findings indicate the need to identify those patients who are at increased risk of developing reinfarction. Most previous studies (6,7,19,22) have failed to predict nonfatal reinfarction from baseline clinical variables. However, Dwyer et al. (16) found that predischarge angina and prior infarction predicted nonfatal reinfarction. Postinfarction exercise testing variables (7,8,16,23-25) and coronary angiographic findings (7,19,23) have not been good predictors of nonfatal reinfarction in most previous studies. Nuclear perfusion imaging techniques were more useful (23,24), but have not been prospectively employed in large scale studies. In a few previous medical intervention postinfarction trials (5,6), medical therapy reduced both overall nonfatal reinfarction and mortality rates over long-term follow-up. In the BetaBlocker Heart Attack Trial (BRAT) study (5), propranololtreated postinfarction patients had a lower reinfarction rate than did placebo-assigned patients over a follow-up period of 36 months (4.4% versus 5.3%, respectively). This difference was not statistically significant; however, the cardiac mortality rate was significantly reduced by 27% in propranololtreated patients. In the Norwegian timolol study (6), postin-

258

BENHORIN ET AL. NONFATAL REINFARCTION

farction timolol therapy reduced the reinfarction rate by 28.3% at 33 months of follow-up when compared with placebo (14.4% versus 20.1%, respectively, p = 0.0006), whereas the cardiac mortality rate was significantly reduced (p < 0.001) in timolol-treated patients as well. The reinfarction rate was higher in patients with than in those without prior infarction in both treatment groups. These similar trends of reduction in reinfarction and mortality rates in the aforementioned intervention trials suggest a common mechanism for the two outcome events and might indicate that reinfarction prevention played a role in the mortality reduction reported in these studies. Implications. Nonfatal reinfarction carries a significant and independent risk for subsequent cardiac death. Further studies are needed to identify postinfarction patients who are at increased risk for subsequent nonfatal reinfarction. There is some evidence that reinfarction prevention can be achieved by early interventions (medical or other) in some subsets of postinfarction patients. The results of this study suggest that prevention of reinfarction might contribute to a significant mortality reduction in postinfarction patients, particularly after a first infarction. We thank Shirley Eberly, Eric Carleen, Mark Andrews and Michael McDermott for assistance in the data analysis and Anita Oberer for secretarial proficiency.

References I. Vedin A, Wilhelmsson C, Elmfeldt D, Save-Soderbergh 1, Tibblin G,

2. 3. 4. 5. 6. 7.

Wilhelmsen L. Deaths and non-fatal reinfarctions during two years' follow-up after myocardial infarction: a follow-up study of 440 men and women discharged alive from hospital. Acta Med Scand 1975;198:353-64. Davis HT, DeCamilla 1, Bayer LW, Moss AJ. Survivorship patterns in the posthospital phase of myocardial infarction. Circulation 1979;60: 12528. Luria MH, Knoke m, Wachs IS, Luria MA. Survival after recovery from acute myocardial infarction, two and five year prognostic indices. Am 1 Med 1979;67:7-14. The Multicenter Postinfarction Research Group. Risk stratification and survival after myocardial infarction. N Engll Med 1983;309:331-6. Goldstein S, The BHAT Research Group. Propranolol therapy in patients with acute myocardial infarction: the Beta Blocker Heart Attack Trial. Circulation 1983;67:53-6. The Norwegian Multicenter Study Group. Timolol-induced reduction in mortality and reinfarction in patients surviving acute myocardial infarction. N Eng) J Med 1981;304:801-7. Norris RM, Barnaby PF, Brandt PWT, et al. Prognosis after recovery from first acute myocardial infarction: determinants of reinfarction and sudden death. Am J CardioI1984;53:408-I3.

JACC Vol. 15, No.2 February 1990:253-8

8. Fioretti P, Brower RW, Simoons ML, et at. Prediction of mortality during the first year after acute myocardial infarction from clinical variables and stress test at hospital discharge. Am J CardioI1985;55:1313-8. 9. The Multicenter Diltiazem Post-infarction Trial Research Group. The effect of diltiazem on mortality and reinfarction after myocardial infarction. N Engl J Med 1988;319:385-92. 10. Greenberg H, Gillespie J, Dwyer EM Jr, the Multicenter Post-infarction Research Group. A new electrocardiographic classification for patients in post-myocardial infarction clinical trials. Am 1 Cardioll987;59:1057-63. 11. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-81. 12. Peto R, Pike MC. Conservation of the approximatics (O-E)2IE in the log rank test for survival data or tumor incidence data. Biometrics 1973;29: 579-84. 13. Dixon WJ, Engelman L. BMDP Biomedical Computer Programs. Berkeley: University of California Press, 1981:330-44. 14. Cox DR. Regression models and life tables (with discussion). J RStat Soc B 1972;34: 187-220. 15. Prentice RL. On the design of synthetic case-control studies. Biometrics 1986;42:301-10. 16. Dwyer EM, McMaster P, Greenberg H, The Multicenter Postinfarction Research Group. Nonfatal cardiac events and recurrent infarction in the year after acute myocardial infarction. 1 Am Coli CardioI1984;4:695-702. 17. Krone Rl, Friedman E, Thanavaro S, Miller JP, Kleiger RE, Oliver GC. Long term prognosis after first Q-wave (transmural) or non-Q-wave (nontransmural) myocardial infarction: analysis of 593 cases. Am J Cardiol 1983;52:234-9. 18. Marmor A, Geltman EM, Schechtman K, Sobel BE, Roberts R. Recurrent myocardial infarction: clinical predictors and prognostic implications. Circulation 1982;66:415-21. 19. Castaner A, Betriu A, Roig E, et al. Clinical course and risk stratification of myocardial infarct survivors with three-vessel disease. Am Heart J 1986;112: 1201-9. 20. Greenberg H, McMaster P, Dwyer EM lr, The Multicenter Postinfarction Research Group. Left ventricular dysfunction after acute myocardial infarction: results of a prospective multicenter study. 1 Am Coli Cardiol 1984;4:867-74. 21. Marcus FI, Cobb LA, Edwards JE, et al. Mortality after acute myocardial infarction in the Multicenter Post-infarction Program. Am J Cardiol 1988;61:8-15. 22. Vedin A, Wilhelmsen L, Wedel H, et al. Prediction of cardiovascular deaths and non-fatal reinfarctions after myocardial infarction. Acta Med Scand 1977;210:309-16. 23. Gibson RS, Watson DD, Craddock GB, et al. Prediction of cardiac events after an uncomplicated myocardial infarction: a prospective study comparing predischarge exercise thallium-201 scintigraphy and coronary angiography. Circulation 1983;68:321-36. 24. Rozanski A, Berman D. The efficacy of cardiovascular nuclear medicine exercise studies. Sem Nucl Med 1987;17:104-20. 25. Corbett JR, Dehmer OJ, Lewis SE, et al. The prognostic value of submaximal exercise testing with radionuclide ventriculography before hospital discharge in patients with recent myocardial infarction. Circulation 1981 ;64:535-44.