- Email: [email protected]
Effects of a prescribed supervised exercise program on mortality and cardiovascular morbidity in patients after a myocardial infarction
Effects of a Prescribed Supervised Exercise Program on Mortality and Cardiovascular Morbidity in Patients After a Myocardial Infarction The National Exercise and He@ Disease Project*
LAWRENCE W. SHAW for the Project Staff Washington,
D.C.
THE NATIONAL EXERCISE AND HEART DISEASE PROJECT: l
PARTICIPATING CENTERS:Universifyof A/a&ma Collaborating Center, Birmingham, Alabama: Albert Oberman, MD, Glenda Barnes, RN, Del Eggert, MM, Stephen N. Barton, MD; Case Western Reserve Collaborating Center, Cleveland. Ohio: Herman Hellerstein, MD. FACC, Jorge Insua, MD, Chaim Yoran, MD, Paul Fardy, PhD, Barry A. Franklin, PhD; Emory University Collaborating Center, Atlanta, Georgia: Charles Gilbert, MD. FACC, Daniel Lee Blessing, MS, Barbara Johnson, RN; The George Washington University Collaborating Center, Washington, D.C.: Patrick Gorman, MD, Margie LaVelle, RN, Marcia Everett, MS; Lankenau Hospital Collaborating Center, Philadelphia, Pennsylvania: Alan Barry, PhD, James Daly, MD’, John Satinsky, MD, William Marley, MD. COORDINATING CENTERS: The George Washington University, Washington, D.C. and State University of New York, at Buffalo, Buffalo, New York: Lawrence Shaw, AM, Patricia Cleary, MS, Jorge Rios, MD, Melvin Stern, MD, Donald Paup, PhD, Dan Bogarty, Patricia Kavanaugh, Sarah Schlesselman, MS, John LaRosa, MD and John Naughton, MD, FACC. SPONSORING AGENCY: National Institute for Handioapped Research, Department of Education, Washington, DC: Barbara E. Moriarity.
This project was supported in part by Grant 13P-5738-07 from the Rehabilitation Services Administration of the Department of Health, Education, and Welfare, Washington, D.C. Manuscript received December 2. 1980; revised manuscript received February 24, 1981, accepted March 3, 1981. + Deceased January 1980. Address for reprints: John Naughton, MD, School of Medicine, State University of New York at Buffalo, Bartley Avenue, Farber Hall, Room 154, Buffalo. New York 14214.
This study enrolled 651 men with myocardial infarction in flve participating centers in a randomized 3 year clinical trial of the effects of prescribed supervised exercise. The subjects, aged 30 to 64 years, were screened for eligibility 2 to 36 months after their qualifying myocardlal infarction. The men In the exercise group pursued intensive exercise In the laboratory for 6 weeks and then in a gymnasium for 34 months. The experience of the exercise group was more favorable than that of the control group In most of the comparisons made. The cumulative 3 year total mortallty rate was 7.3 percent for the control group and 4.6 percent for the exercise group; the 3 year rate for recurrent myocardlal infarction was 7.0 and 5.3 percent, respectively. Mortality rates In the two groups did not differ slgnlflcantly, but the data were consistent with an assumption of substantial benefit from exercise. Adjustment for small differences In basellne varlables by multivariate methods did not materially alter the estimate of effect of exercise. Certain subgroups showed a greater benefit from exercise.
It is widely believed that regular physical activity deters the development and progression of coronary heart disease. Although there are many descriptive reports suggesting that regular exercise is beneficial and safe for patients after a myocardial infarction,‘p2 an adequate scientific evaluation of the merit of exercise programs for such subjects has not been conducted. In 1971, Fox et a1.3 reported that more studies are urgently needed, “particularly concerning whether increased physical activity will contribute to cardiovascular and general health enhancement, increased total human performance and a vigorous creative society.” In 1972, the Rehabilitation Services Administration of the Department of Health, Education, and Welfare sponsored a controlled clinical trial of prescribed physical activity for patients after a myocardial infarction. A large trial was planned, but the funding agency was forced to limit its size. The study, called the National Exercise and Heart Disease Project, enrolled 651 participants. This report presents the findings of this study on cardiovascular morbidity and mortality, and deaths from all causes, during 3 years of observation per patient. Other reports will provide information on the effects of the exercise program on cardiovascular fitness and other variables of interest. Methods Study patients: This research project was a collaborative randomized clinical trial of the effects of individually prescribed, medically supervised regular physical activity on the rehabilitation of male survivors (aged 30 to 64 years) of a myocardial infarction. Details on the design and conduct of the study have been published.” Candidates for the study had to have a documented acute myocardial
July 1981
The American Journal of CARDIOLOGY
Volume 48
39
NATIONAL EXERCISE AN0 HEART DISEASE PROJECT-SHAW
infarction within 3 years (preferably 1 year) of admission to the study. All candidates entered a preliminary prerandomization program of low level exercise of 6 weeks’ duration that was designed to identify and exclude likely nonadherers to the planned exercise program without improving cardiovascular fitness. A study protocol, a manual of operations and standard data forms were prepared to guide the staffs of the participating clinics. The collaborating centers recruited patients,
performed evaluations and carried out the exercise training programs. A coordinating center was established to perform
TABLE I Selected Characteristics of the 651 Participants at Entry Into the Study Control Group (n = 328) Demographic findings Fhy;~ e)(yr) (SEM)
% Married (%)
;z;
(0.4)
_
Exercise Group (n = 323) ;5:; (0.4)
.
Working (more than 10 h/wk) Education (maximal) (%) High school or less College level Graduate training Income (% ) <$10.000
(%)
Smokers (some form daily) (%) Alcohol drinkers (at least once a week) 4::; weight (kg) (SEhn) Cardiovascular findings by history lnterter;rom QMI to entry (%) 7-12 mo 113 mo Previous MI in addition to QMI (%) Complication during QMI (%) Angina withln past 3 mo (%) History of hypertension (%) Father died of coronary heart disease
91.2 76.8
92.6 74.9
30.8 43.0 25.6
31.6 44.3 22.9
14.9 61.3 21.3 25.9 66.2
14.9 65.5 18.0 29.3 63.2
79.1 (0.6)
79.7 (0.6)
20.4 36.9 42.7 16.2 46.3 29.1 33.6 31.1
19.5 36.5 44.0 18.6 48.1 27.5 32.0 31.6
(%)
Cardiovascular findings by examination Mean resting systolic blood pressure (mm Rg) (SEM) Mean restina diastolic blood pressure (mm Hg) (SEM) Mean resting heart rate (beatslmin) (SEM) Lipid measurements Mean cholesterol (mg/dl) (SEM) Mean triglycerides (mg/dl) (SEM) Cardiovascular findings by treadmill test Reasons for stopping test (%) Symptoms Signs Reached age-predicted heart rate W3y; zeyaFity at entry (%) 7-9 Mets I10 Mets S-T segment depression at peak exercise (%) Peak systolic blood pressure, I140 Rg (%)
mm
122.8 (0.9)
124.6 (0.9)
84.9 (0.5) 61.3 (0.6)
85.5 (0.6) 62.8 (0.6)
222.8 (3.8) 172.4 (8.6)
223.3 (2.2) 186.4 (9.5)
69.8 16.5 13.7
67.5 16.7 15.8
30.5 49.4 20.1 14.9
30.7 48.3 21.0 13.0
19.9
18.5
One Met approximates the oxygen uptake at rest (that is, 3.5 ml/kg per min). Ml = myocardial infarction; QMI = qualifying myocardial infarction: SEM = standard error of the mean. l
40
July 1981
The American Journalof CARDIOLOGY Volume 48
centralized functions such as electrocardiographic interpretations and statistical analysis of the data. Uniformity of conduct of the study was promoted by a central coordinator for exercise programs and for psychological testing. Exercise testing and exercise prescription: Before admission, and semiannually during the study, candidates and participants were given standardized multistage electrocardiographically monitored treadmill exercise tests to determine their work capacity. The results of this test were used to formulate the prescription for exercising for the period from one test until the next. The prescription, expressed in terms of target heart rate, was 85 percent of the peak heart rate achieved on the test. Exercise program: The postrandomization exercise program consisted of two parts: 1. During the first 8 weeks, the participant attended an exercise laboratory 1 hour/day, 3 days/week. Under supervision, with continuous electrocardiographic monitoring, he exercised for a total of 24 minutes, by exercising for 4 minutes on each of six stationary devices, and resting for 2 minutes after use of each device. The work load on each device was set to yield the target heart rate early in each 4 minute exercise period. Exercise was halted if the participant exhibited adverse symptoms, signs or electrocardiographic changes. 2. Thereafter the exercise program was conducted in a gymnasium without electrocardiographic monitoring. It consisted of supervised physical activities designed to yield the prescribed target heart rate. The activities included 15 minutes of continuous jogging, cycling or swimming, followed by 25 minutes of games. The attained heart rate was periodically checked. The participant was urged to attend three sessions/week, but in some instances was allowed to exercise on his own. Follow-up examinations: All study group members were examined at 9 weeks and 6 months after randomization and semiannually thereafter. The examination included an interval history, physical examination and a treadmill test. The test findings were used to judge the effect of the physical activity intervention on work capacity, blood pressure, heart rate and electrocardiographic measures. Twelve lead and X, Y and Z orthogonal lead recordings (paper and tape) of electrocardiograms were obtained at each stage of the treadmill test. Special forms were prepared for the reporting of all deaths, recurrent myocardial infarctions and, at a later time, other periods of rehospitalization. Data analysis: During the conduct of the study, the accumulating results were not shown to the participating field staff or the sponsor. However, they were shown periodically to a data-monitoring committee, which determined whether any substantial change in the conduct of the study should be recommended to the investigators. An expert panel of three cardiologists independently reviewed clinical summaries, death certificates and autopsy reports, if made, and classified ail deaths by cause. They also reviewed all reported episodes of myocardial infarction and classified them into three categories: definite, suspected or no myocardial infarction. The observation period for physiologic and pyschologic variables was closed at the last semiannual visit before De-
cember 1,1978. Observations for morbidity and mortality were continued to May 31, 1979, thus providing a minimum of 31 months’ observation. Statistical methods: When the patients were categorized into classes (for example, smokers versus nonsmokers) and the benefit of exercise was studied within each class, a Z test was performed on the difference between the differences of
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
the observed proportions to assess whether the benefit was greater for one class than for the other. When multivariate methods were used (for example, logistic regression) several variables, including exercise status, were used in the model. The effect of the exercise program was assessed by calculating the probability of death for each of the 651 participants under two circumstances: (1) assuming that each participant was a control subject, and (2) assuming that each participant was in the exercise group. The sum of the first set of probabilities gives the expected number of deaths for the entire population if they were all control subjects; similarly, the sum of the second set gives the expected number of
CUMULATIVE MORTALITY RATES 8-
deaths if all persons were in the exercise group. The ratio of the two numbers (second over first) establishes the relative risk adjusted for the other factors used in the model. Life table methods were used to calculate survival curves and their complements, cumulative mortality rates, as a function of time.
Results Description
Years of Observation YEHDP
4.
of Study Group
There were 931 referrals of interested and presumptively eligible candidates.4 Nearly one third (280) of these men were eliminated before formal entry into the study, because they were found ineligible or had dropped out during the screening process. From September 1974 to October 1976, a total of 651 eligible candidates who were willing to accept the random assignment of “exercise” or “no exercise” were admitted to the study. In Table I, selected characteristics of these participants are depicted, as are certain aspects of their cardiovascular health. The two groups are very similar in all characteristics. The participants were predominantly white, married, college-educated men with an above average income. Most of them had had a qualifying myocardial infarction more than 6 months before admission and most had already returned to work by the time of entry into the study. Many had associated coronary risk factors: smoking (28 percent), hypertension (33 percent), ele-
*
One standard
error
FIGURE 1. Cumulative mortality rates.
vated serum lipids (36 percent), and death of father from coronary heart disease (31 percent). A large proportion (85 percent) had cardiovascular limitations in that they could not reach the age-predicted peak heart rate on the admission treadmill test: The test was stopped in 68 percent because of symptoms and in 17 percent because of signs. Adherence
During the study there was only a modest rate of loss from the periodic examinations, 6 percent in the control subjects and 7 percent in the exercise group. A loss was defined as a participant’s failure to report for the semiannual observation on two consecutive appointments.
TABLE II Mortality of Participants by Study Group and Estlmates of Effectiveness of the Exercise Program
Control Group (n = 328)
Exercise Group (n = 323)
Estimates of Effectiveness of Program (%)
Comparison Statistics Minimum
Events Counted All deaths Cardiovascular deaths Acute myocardial infarction Other definite* Subtotal Sudden deaths+ Total Indeterminate cause
Observed
Maximum
n
%
n
%
X2
P
(CL)
(El
24
7.3
15
4.6
1.62
0.22
-15
37
68
2.4
1 5
0.3
3.96
0.047
a7
98
1.9
2.42
0.13
-5
56
84
4.3
0.70
0.40
-33
29
66
t 14 6 20 4
4.3 6.1
8” 14 1
22
GJ)
Includes six deaths from arrhythmias, two from congestive cardiac failues, one from cardiogenic shock and two from cerebrovascular accidents. + Within 1 hour. CL = lower 95 percent confidence limit: C u = upper 95 percent confidence limit; E = observed effectiveness; p = probability. l
July 1981
The American Journal 01 CARDIOLOGY
Volume 48
41
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
the true effectiveness of exercise ranged widely-from a minimum of -15 percent to a maximum of 68 percent (95 percent confidence limits). Cardiovascular deaths: If the analysis is restricted to cardiovascular deaths, there is little change, because only five deaths were not allocated to cardiovascular causes. If sudden deaths (within 1 hour of onset of symptoms) are excluded, the number of cardiovascular deaths becomes 20 (4.3 percent in the control group and 1.9 percent in the exercise group; difference not significant [NS]). The estimate of effectiveness of treatment is 56 percent, but the confidence interval for this estimate extends from -5 percent to 84 percent. Only nine deaths were due to recurrent myocardial infarction; eight of these occurred in the control group and one in the exercise group (p = 0.05). The estimate of effectiveness was 87 percent, but again there was a wide range of confidence limits (from 22 to 98 percent).
Noncompliance was defined as failure to attend the exercise sessions. By the end of 2 years, 23 percent of the exercise group were not attending the organized exercise program and did not claim to be exercising elsewhere. At the same time, 31 percent of the control subjects alleged that they were exercising regularly. Mortality
During the 3 years of observation per patient, there were 24 deaths (7.3 percent) in the control group of 328 subjects and 15 deaths (4.6 percent) in the exercise group of 323 subjects (Table II). This difference was not statistically significant (p = 0.05). Cumulative mortality rates for progressive time periods are shown in Figure 1. The curves do not differ significantly, at any time or collectively. One can calculate whether the observed results, while not proving a clear benefit, could have by chance underestimated a true benefit of much greater magnitude. To answer this question confidence limits for
Role
the observed benefit are derived. The observed benefit is expressed as follows: Effectiveness
=
X
of risk factors
and cardiovascular
status:
Whether the treatment is appreciably better for some than for other subgroups is evaluated in Table III using coronary risk factors, and in Table IV using factors more directly indicating the cardiovascular status of the subject. (Note that the proportions for the subgroups in the control group give some indication of risk, whereas the estimates of effectiveness give an indication of whether exercise was more beneficial for one group than another.) Table III shows that in the control group, except in the age subdivision, each subgroup with a risk factor had a greater incidence of death. However, there was no
100 = 37 percent,
(that is, exercise appears to have reduced the incidence of death by 37 percent). However, this estimate of effectiveness is rather unstable because both of the component parts are subject to statistical variation. There are no exact methods for determining confidence limits, but two approximate methods were used5,6 and they gave similar results. These methods indicate that TABLE Ill
Relation of Mortallty to Coronary Risk Factors at Entry Control Group
Exercise Group
Deaths Subgroup All subjects Age 150 years Yes No Smoking Yes No Elevated lipids’ Yes No Missing History of hypertension Yes No Father died of coronary heart disease Yes No Presence of above factors Three Two or more One or none
Deaths
Subgroup (n)
n
%
328
24
213 115 a5 243
Estimate of Effectiveness of Program (%)
Subgroup fn)
n
%
7.3
323
15
4.6
37
15 9
7.0 7.8
198 125
11 4
5.6 3.2
21 59
168
9.4 6.6
94 227
2 13
2.1 5.7
77 13
117 189 17
5 9 1
4.3 4.8 5.9
51 4
104 202 22
1: 5
a.7 5.0 22.7
110 218
12 12
10.9 5.5
102 221
8’
6.9 3.6
33:
102 226
9 15
6.8 6.6
102 221
7”
7.8 3.2
::
67 109 110
9 3’
10.3 6.4 2.7
:: 119
5 8 1
5.5 8.3 0.8
-“3; 69
Cholesterol 2250 mg/dl, or triglyceride 2200 mg/dl. In each segment of this table the category shown first represents those presumed to have a higher probability of death; similarly, in the last segment those with the highest probability of death are shown first. l
42
July 1981
The American Journal of CARDIOLOGY
Volume 48
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
TABLE IV Relallon of Mortality to Cardiovascular Status at Entry
ControlGroup
Exercise Group
Subgroup (n)
n
%
Subgroup (n)
n
%
Estimate of Effectiveness of Program (%)
328
24
7.3
323
15
4.6
37
150 178
4 20
2.7 11.2
131 192
2 13
1.5 6.8
::
100 228
15 9
15.0 3.9
2::
11 4
11.1 1.8
26 55
1;: 70
1; 4
12.3 5.9 5.7
64 186 72
: 4
7.8 3.2 5.6
Z
2::
12.5 6.2 0.0
2::
6
1: 0
1:; 39
:
73
:
7.8 23.1 3.6 5.5
ZB 152 75
Deaths Subgroup All subjects Previous MI or complicated QMI
l
No Yes Work capacity <7 Mets Yes No S-T segment depression (> 1 mm at last stage of exercise test) Yes No Missing Systolic blood pressure (~140mmHgatlast stage of exercise test) Yes No Missing Presence of these factors Two Three or four One or none Incomplete information l
Deaths
! 2
4
I :
13.6 5:::
-8 69+
3.5 15.4 1.3 6.7
:: 63
Qualifying myocardial infarction. + The difference between 6.2 and 1.9 is statistically significant (0.01 < p <0.05).
consistent pattern with respect to the benefit of exercise, and no clear-cut benefit from exercise emerged for any subgroup. The greatest benefit was enjoyed by those smoking at entry to the study, but the difference in death rate of smokers and nonsmokers, (9.4 versus 2.1 percent) was not statistically significant. (Moreover, although a 77 percent benefit appeared much better than the 13 percent benefit for the nonsmokers, the difference between these estimates was not significant.) The last panel of Table III examines the joint influence of the risk factors. There was a trend in the death rates in the control group, but no apparent trend in the estimates of benefit from exercise. In Table IV, the importance of each cardiovascular factor selected is fairly clear in the proportions for the control group, but in general the benefit of exercise was no better or worse in the various subgroups. However, the results in one subgroup suggest a “significant” finding: The subjects who had satisfactory systolic blood pressure responses during the last stage of the exercise
test (that is, more than 140 mm Hg) enjoyed a 69 percent benefit from exercise in reduction of mortality. Finally, the last panel of Table IV presents the joint influence of these selected factors and demonstrates a clear trend in incidence of death in the control subjects. The analysis of effect of exercise suggests that men with more evidence of disease benefited less from the exercise program. However, the trend, from 33 to 63 percent, was not statistically significant. Multivariate analysis: Multivariate regression procedures that adjust for any imbalance in the study groups were used in the 490 subjects whose electrocardiographic S-T segment data during exercise were available. The techniques yielded little change from the crude data, as indicated in Table V. The factors used in the logistic model and the Cox model are shown in Table VI. Moreover, when age (below or above age 50 years) and smoking status at entry were added to each model, the estimate of effectiveness did not change, very likely because the two study groups were well balanced on
TABLE V Comparison of Crude Death Rates With Two Types of Adjusted Rates
Sample size Crude death rates (3 yr) Rates using logistic model7 (3 yr) Rates using Cox models (3 yr)
Control Group
Exercise Group
Estimate of Effectiveness
249 8.0% 8.0% 8.2%
241 4.1% 4.1% 4.1%
48% 46% 48%
July lQ81
The American Journal of CARDIOLOGY
Volume 48
43
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
TABLE VI Factors Used in Logistic Model and the Cox Model
Myocardial infarction history* Work capacity (Mets) S-T segment depression > 1 mm at last stage of exercise test History of hypertension Heart size (vol/body surface area) Treatment group l
More than one myocardial infarction or complicated
Favorable
Unfavorable
No >_7 No No
Yes 16 Yes Yes lMedian Control
qualifying myocardial infarction
these factors. Finally, inclusion of peak systolic blood pressure during the admission treadmill test produced no substantial change in the analysis of the effect of exercise on mortality.
The total number of rehospitalizations for all reasons other than myocardial infarction was examined
Morbidity
If the need for coronary arterial surgery is viewed as a measure of failure of therapy, no differential between
for any indication of a difference due to exercise. None was found; the rehospitalization rates were virtually identical in the two study groups.
In addition to the 34 patients who died of cardiovascular causes (Table II), 26 patients had a nonfatal myocardial infarction; thus there were 60 cases of serious cardiovascular morbidity to analyze (Table VII, subtotal line). The case rate was 9.5 percent for control subjects and 9.0 percent for the exercise group; the estimate of benefit from exercise was only 5 percent.
the two groups occurred. Coronary arterial bypass surgery was performed in 16 control subjects and 17 in the exercise group. Although there was no apparent benefit from exercise in relation to morbidity, no harm was evident (that is, the event rates in the exercise group were not higher than in the control group).
The domain of review was broadened to include all instances of cardiovascular illness, by adding five
Discussion
suspected myocardial infarctions and 50 hospitalizations for other cardiovascular illnesses (Table VII). This yielded a total of 115 participants experiencing some kind of cardiovascular morbidity. The event rate was 17.7 percent for the control subjects and 17.6 percent for the exercisers, indicating no apparent benefit from exercise. The upper confidence limit of 31 percent indicates that the true benefit from exercise, in preventing aggregate cardiovascular morbidity, does not exceed a 31 percent reduction. The results for all myocardial infarctions, fatal and nonfatal, are shown in an added line of Table VII. The estimate of benefit from exercise was only 25 percent, with confidence limits of -34 and 66 percent.
Exercise and cardiac mortality: The results of this study suggest that a program of prescribed supervised physical activity for patients after myocardial infarction may be beneficial in reducing subsequent cardiac mortality, but the evidence is not convincing. Depending on whether one focuses on the narrow domain of deaths due to recurrent myocardial infarction or looks more broadly at all cardiovascular mortality, a very different estimate of benefit from the exercise program is observed. The apparent reduction in deaths from myocardial infarction was 87 percent, whereas it was 29 percent for total cardiovascular mortality. The latter degree of success would certainly be a worthy benefit,
TABLE VII Morbidity Experience of Study Participants Control Group (n = 328) Events Counted
Exercise Group (n = 323)
Estimates of Effectiveness (%)
Comparison Statistics P
Minimum (CL)
Observed (E)
Maximum (Cu)
0.70
0.40
-33
29
66
0.005
0.95
-57
5
42
n
%
n
%
20
6.1
14
4.3
;:
9.5
:9”
9.0
::
17.7
2; 57
17.6
0.00
1.00
-44
0
31
All recurrent myocardial infarctions
23
7.0
17
5.3
0.59
0.46
-34
25
66
Total hospitalizations for reasons other than myocardial infarction
90
27.4
92
28.5
0.04
0.86
-34
Cardiovascular deaths Nonfatal infarctions Subtotal Suspected infarctions Other events Total
Abbreviations
44
July 1991
2
as in Table II.
The American
X2
Journal of CARDIOLOGY
Volume
49
-4
21
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
but because the study group was small, the difference in death rates, 6.1 versus 4.3 percent, was not statistically significant. These results may be compared with those reported by Wilhelmsen et a1.g In a randomized clinical trial of the value of exercise, they observed 4 year cardiac mortality rates of 21.0 and 14.6 percent in their control and exercise groups, respectively. This yielded an effectiveness rate of 31 percent, which is remarkably close to our estimate of 29 percent despite very wide confidence limits pertaining to each estimate. Results of the two studies (either separately or collectively) were not statistically significant. Exercise and cardiovascular morbidity: There is no suggestion of benefit from the exercise program in relation to cardiovascular morbidity; the event rates in the two study groups were virtually identical (18 percent of the subjects in each group had some cardiovascular episode in 3 years of follow-up study). This may contrast somewhat with the favorable findings of Wilhelmsen et a1.,g who reported a 22 percent reduction in cardiovascular morbidity. The results reported here can also be compared with the data of Rechnitzer’O from the randomized clinical trial conducted in Ontario, Canada. In that study a high and a low intensity exercise program were compared. The preliminary report showed no benefit; indeed, patients in the high intensity group had a greater rate of reinfarction than did those in the low intensity group. Methodologic considerations: The value of having a randomized control group is apparent in this report. The total mortality rate in the exercise group was far below that predicted (only 4.6 percent in 3 years versus an “expectation” of 12 to 15 percent). Without an appropriate control group, it might have been speculated (as some investigators have donell), that a very substantial benefit from exercise was demonstrated. But the data on the control group show that the participants in this study were a rather favorable subset of subjects in the postmyocardial infarction state; their mortality rate was half that “expected” in advance of the trial. (A report from the Coronary Drug Project12 showed that it is not difficult to identify a subgroup of patients after a myocardial infarction who may be expected to have a relatively low death rate.) The “null hypothesis”is important but not the only hypothesis to consider. One could ask: How good could
exercise really be, and yet the observed findings occur? The upper confidence limit indicates that the true benefit of exercise could be to reduce the cardiac mortality rate by 66 percent, but in this study, with a small sample of 651 patients, the observed reduction was only 29 percent. At the same time, it should be stated that
the lower confidence limit of -33 percent means that exercise could be hazardous, but this study failed to show a risk. This study, because of its sample size limitations, was not designed to give a definitive answer to the question of mortality (or morbidity) in respect to the effect of an exercise program. (It was originally estimated that a large study of 4,200 participants was desirable.) Because the exercise group was seen much more often than the control group in this nonblind clinical trial, there may have been some excess ascertainment of the milder forms of morbidity in the exercise group. Did the participants
follow the prescribed
therapy?
It appears that adherence to the therapy plan was good in the exercise group and only fair in the control group. Some lagging in enthusiasm and attendance was noted after the 1st year of the exercise program and some “exercising on their own” was done by some control subjects. This response, of course, dilutes the estimates of true benefit of exercise but it does not interfere with assessment of benefit of the “program” (that is, intensive long-term exercise). Some persons have postulated that the preliminary “prerandomization exercise program” damaged the value of the postrandomization research program, in that enthusiasm for exercise was engendered and some control subjects rejected their assignment and continued to exercise. If this occurred, the study yielded a very useful “public health” finding, namely, that an expensive long-term vigorous exercise program does not offer significantly greater benefit than a short-term introduction to mild exercise. Did those who faithfully exercised have a better outcome than the control subjects who did not exercise?
It does not seem appropriate to answer this question because of the likelihood of selection bias; that is, sicker people may withdraw from the exercise program and healthier control subjects may pursue exercise, making the adherers “noncomparable.” Finally, whereas the analyses by subgroups often showed expected patterns in respect to the natural history of chronic heart disease, the suggestion of greater benefit from exercise for some subgroups may easily be due to chance-especially because so many comparisons were examined. Implications of study: The case for exercise in persons with known myocardial infarction is neither proved nor disproved. Enough information was gathered from this trial and sufficient success in obtaining continued participation was achieved to conclude that a full scale trial of the merits of exercise for patients with cardiac disease is feasible. It is of interest that if the 37 percent benefit from exercise reported had been observed in a comparable study of as few as 1,400 patients, the result would have been judged “statistically significant.”
References 1. Heiierstein HK, Buriando A, Hirsch EZ, et al. Active physical reconditioning of coronary patients. Circulation 1965;31,32:Suppl li:ll-110-l. 2. Naughton J, Lategoia MT, Shanbour K. A physical rehabilitation program for cardiac patients. A progress report. Am J Med Sci
1966;252:545-53. 3. Fox SM iii, Naughton JP, Haskeii WL. Physical activity and the prevention of coronary heart disease. Ann Clin Res 1971;3: 404-32. 4. Naughton J. The national exercise and heart disease project: de-
July 1981
The American Journal of CARDIOLOGY
Voiume 48
45
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
5.
6. 7.
8.
46
velopment recruitment, and implementation. In: Wenger NK, ed. Exercise and the Heart. Philadelphia, PA: Cardiovasc Clin, 1978;205-22. Noether GE. Two confidence intervals for the ratio of two probabilities and some measures of effectiveness. J Am Stat Assoc 1957;52:36-45. Brass I. A confidence interval for a percentage increase. Biometrics 1954;10:245-50. Cornfleld J. Joint dependence of risk of coronary heart disease on serum cholesterol and systolic blood pressure: a discriminant function analysis. Fed Proc 1962;21:58-61. Cox DR. Regression models and life tables (with discussion). J R Stat Sot (B) 1972;34: 187-220.
July 1961
The American Journal of CARDIOLOGY
Volume 48
9. Wilhelmsen L, Sanne H, Elmfeldt D, Grfmby G, Tlbblln G, Wedel H. A controlled trial of physical training after myocardial infarction. Prev Med 1975;4:491-508. 10. Rechnltzer PA. The effects of training: reinfarction and death-an interim report. Med Sci Sports 1979;11:382. 11. Kavanagh T, Shephard RJ, Chlsholm AW, Qureshi S, Kennedy J. Prognostic indexes for patients with ischemic heart disease enrolled in an exercise-centered rehabilitation program. Am J Cardiol 1979;44:1230-40. 12. Coronary Drug Project Research Group. Factors influencing long-term prognosis after recovery from myocardial infarction. J Chron Dis 1974;27:267-85.
LAWRENCE W. SHAW for the Project Staff Washington,
D.C.
THE NATIONAL EXERCISE AND HEART DISEASE PROJECT: l
PARTICIPATING CENTERS:Universifyof A/a&ma Collaborating Center, Birmingham, Alabama: Albert Oberman, MD, Glenda Barnes, RN, Del Eggert, MM, Stephen N. Barton, MD; Case Western Reserve Collaborating Center, Cleveland. Ohio: Herman Hellerstein, MD. FACC, Jorge Insua, MD, Chaim Yoran, MD, Paul Fardy, PhD, Barry A. Franklin, PhD; Emory University Collaborating Center, Atlanta, Georgia: Charles Gilbert, MD. FACC, Daniel Lee Blessing, MS, Barbara Johnson, RN; The George Washington University Collaborating Center, Washington, D.C.: Patrick Gorman, MD, Margie LaVelle, RN, Marcia Everett, MS; Lankenau Hospital Collaborating Center, Philadelphia, Pennsylvania: Alan Barry, PhD, James Daly, MD’, John Satinsky, MD, William Marley, MD. COORDINATING CENTERS: The George Washington University, Washington, D.C. and State University of New York, at Buffalo, Buffalo, New York: Lawrence Shaw, AM, Patricia Cleary, MS, Jorge Rios, MD, Melvin Stern, MD, Donald Paup, PhD, Dan Bogarty, Patricia Kavanaugh, Sarah Schlesselman, MS, John LaRosa, MD and John Naughton, MD, FACC. SPONSORING AGENCY: National Institute for Handioapped Research, Department of Education, Washington, DC: Barbara E. Moriarity.
This project was supported in part by Grant 13P-5738-07 from the Rehabilitation Services Administration of the Department of Health, Education, and Welfare, Washington, D.C. Manuscript received December 2. 1980; revised manuscript received February 24, 1981, accepted March 3, 1981. + Deceased January 1980. Address for reprints: John Naughton, MD, School of Medicine, State University of New York at Buffalo, Bartley Avenue, Farber Hall, Room 154, Buffalo. New York 14214.
This study enrolled 651 men with myocardial infarction in flve participating centers in a randomized 3 year clinical trial of the effects of prescribed supervised exercise. The subjects, aged 30 to 64 years, were screened for eligibility 2 to 36 months after their qualifying myocardlal infarction. The men In the exercise group pursued intensive exercise In the laboratory for 6 weeks and then in a gymnasium for 34 months. The experience of the exercise group was more favorable than that of the control group In most of the comparisons made. The cumulative 3 year total mortallty rate was 7.3 percent for the control group and 4.6 percent for the exercise group; the 3 year rate for recurrent myocardlal infarction was 7.0 and 5.3 percent, respectively. Mortality rates In the two groups did not differ slgnlflcantly, but the data were consistent with an assumption of substantial benefit from exercise. Adjustment for small differences In basellne varlables by multivariate methods did not materially alter the estimate of effect of exercise. Certain subgroups showed a greater benefit from exercise.
It is widely believed that regular physical activity deters the development and progression of coronary heart disease. Although there are many descriptive reports suggesting that regular exercise is beneficial and safe for patients after a myocardial infarction,‘p2 an adequate scientific evaluation of the merit of exercise programs for such subjects has not been conducted. In 1971, Fox et a1.3 reported that more studies are urgently needed, “particularly concerning whether increased physical activity will contribute to cardiovascular and general health enhancement, increased total human performance and a vigorous creative society.” In 1972, the Rehabilitation Services Administration of the Department of Health, Education, and Welfare sponsored a controlled clinical trial of prescribed physical activity for patients after a myocardial infarction. A large trial was planned, but the funding agency was forced to limit its size. The study, called the National Exercise and Heart Disease Project, enrolled 651 participants. This report presents the findings of this study on cardiovascular morbidity and mortality, and deaths from all causes, during 3 years of observation per patient. Other reports will provide information on the effects of the exercise program on cardiovascular fitness and other variables of interest. Methods Study patients: This research project was a collaborative randomized clinical trial of the effects of individually prescribed, medically supervised regular physical activity on the rehabilitation of male survivors (aged 30 to 64 years) of a myocardial infarction. Details on the design and conduct of the study have been published.” Candidates for the study had to have a documented acute myocardial
July 1981
The American Journal of CARDIOLOGY
Volume 48
39
NATIONAL EXERCISE AN0 HEART DISEASE PROJECT-SHAW
infarction within 3 years (preferably 1 year) of admission to the study. All candidates entered a preliminary prerandomization program of low level exercise of 6 weeks’ duration that was designed to identify and exclude likely nonadherers to the planned exercise program without improving cardiovascular fitness. A study protocol, a manual of operations and standard data forms were prepared to guide the staffs of the participating clinics. The collaborating centers recruited patients,
performed evaluations and carried out the exercise training programs. A coordinating center was established to perform
TABLE I Selected Characteristics of the 651 Participants at Entry Into the Study Control Group (n = 328) Demographic findings Fhy;~ e)(yr) (SEM)
% Married (%)
;z;
(0.4)
_
Exercise Group (n = 323) ;5:; (0.4)
.
Working (more than 10 h/wk) Education (maximal) (%) High school or less College level Graduate training Income (% ) <$10.000
(%)
Smokers (some form daily) (%) Alcohol drinkers (at least once a week) 4::; weight (kg) (SEhn) Cardiovascular findings by history lnterter;rom QMI to entry (%) 7-12 mo 113 mo Previous MI in addition to QMI (%) Complication during QMI (%) Angina withln past 3 mo (%) History of hypertension (%) Father died of coronary heart disease
91.2 76.8
92.6 74.9
30.8 43.0 25.6
31.6 44.3 22.9
14.9 61.3 21.3 25.9 66.2
14.9 65.5 18.0 29.3 63.2
79.1 (0.6)
79.7 (0.6)
20.4 36.9 42.7 16.2 46.3 29.1 33.6 31.1
19.5 36.5 44.0 18.6 48.1 27.5 32.0 31.6
(%)
Cardiovascular findings by examination Mean resting systolic blood pressure (mm Rg) (SEM) Mean restina diastolic blood pressure (mm Hg) (SEM) Mean resting heart rate (beatslmin) (SEM) Lipid measurements Mean cholesterol (mg/dl) (SEM) Mean triglycerides (mg/dl) (SEM) Cardiovascular findings by treadmill test Reasons for stopping test (%) Symptoms Signs Reached age-predicted heart rate W3y; zeyaFity at entry (%) 7-9 Mets I10 Mets S-T segment depression at peak exercise (%) Peak systolic blood pressure, I140 Rg (%)
mm
122.8 (0.9)
124.6 (0.9)
84.9 (0.5) 61.3 (0.6)
85.5 (0.6) 62.8 (0.6)
222.8 (3.8) 172.4 (8.6)
223.3 (2.2) 186.4 (9.5)
69.8 16.5 13.7
67.5 16.7 15.8
30.5 49.4 20.1 14.9
30.7 48.3 21.0 13.0
19.9
18.5
One Met approximates the oxygen uptake at rest (that is, 3.5 ml/kg per min). Ml = myocardial infarction; QMI = qualifying myocardial infarction: SEM = standard error of the mean. l
40
July 1981
The American Journalof CARDIOLOGY Volume 48
centralized functions such as electrocardiographic interpretations and statistical analysis of the data. Uniformity of conduct of the study was promoted by a central coordinator for exercise programs and for psychological testing. Exercise testing and exercise prescription: Before admission, and semiannually during the study, candidates and participants were given standardized multistage electrocardiographically monitored treadmill exercise tests to determine their work capacity. The results of this test were used to formulate the prescription for exercising for the period from one test until the next. The prescription, expressed in terms of target heart rate, was 85 percent of the peak heart rate achieved on the test. Exercise program: The postrandomization exercise program consisted of two parts: 1. During the first 8 weeks, the participant attended an exercise laboratory 1 hour/day, 3 days/week. Under supervision, with continuous electrocardiographic monitoring, he exercised for a total of 24 minutes, by exercising for 4 minutes on each of six stationary devices, and resting for 2 minutes after use of each device. The work load on each device was set to yield the target heart rate early in each 4 minute exercise period. Exercise was halted if the participant exhibited adverse symptoms, signs or electrocardiographic changes. 2. Thereafter the exercise program was conducted in a gymnasium without electrocardiographic monitoring. It consisted of supervised physical activities designed to yield the prescribed target heart rate. The activities included 15 minutes of continuous jogging, cycling or swimming, followed by 25 minutes of games. The attained heart rate was periodically checked. The participant was urged to attend three sessions/week, but in some instances was allowed to exercise on his own. Follow-up examinations: All study group members were examined at 9 weeks and 6 months after randomization and semiannually thereafter. The examination included an interval history, physical examination and a treadmill test. The test findings were used to judge the effect of the physical activity intervention on work capacity, blood pressure, heart rate and electrocardiographic measures. Twelve lead and X, Y and Z orthogonal lead recordings (paper and tape) of electrocardiograms were obtained at each stage of the treadmill test. Special forms were prepared for the reporting of all deaths, recurrent myocardial infarctions and, at a later time, other periods of rehospitalization. Data analysis: During the conduct of the study, the accumulating results were not shown to the participating field staff or the sponsor. However, they were shown periodically to a data-monitoring committee, which determined whether any substantial change in the conduct of the study should be recommended to the investigators. An expert panel of three cardiologists independently reviewed clinical summaries, death certificates and autopsy reports, if made, and classified ail deaths by cause. They also reviewed all reported episodes of myocardial infarction and classified them into three categories: definite, suspected or no myocardial infarction. The observation period for physiologic and pyschologic variables was closed at the last semiannual visit before De-
cember 1,1978. Observations for morbidity and mortality were continued to May 31, 1979, thus providing a minimum of 31 months’ observation. Statistical methods: When the patients were categorized into classes (for example, smokers versus nonsmokers) and the benefit of exercise was studied within each class, a Z test was performed on the difference between the differences of
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
the observed proportions to assess whether the benefit was greater for one class than for the other. When multivariate methods were used (for example, logistic regression) several variables, including exercise status, were used in the model. The effect of the exercise program was assessed by calculating the probability of death for each of the 651 participants under two circumstances: (1) assuming that each participant was a control subject, and (2) assuming that each participant was in the exercise group. The sum of the first set of probabilities gives the expected number of deaths for the entire population if they were all control subjects; similarly, the sum of the second set gives the expected number of
CUMULATIVE MORTALITY RATES 8-
deaths if all persons were in the exercise group. The ratio of the two numbers (second over first) establishes the relative risk adjusted for the other factors used in the model. Life table methods were used to calculate survival curves and their complements, cumulative mortality rates, as a function of time.
Results Description
Years of Observation YEHDP
4.
of Study Group
There were 931 referrals of interested and presumptively eligible candidates.4 Nearly one third (280) of these men were eliminated before formal entry into the study, because they were found ineligible or had dropped out during the screening process. From September 1974 to October 1976, a total of 651 eligible candidates who were willing to accept the random assignment of “exercise” or “no exercise” were admitted to the study. In Table I, selected characteristics of these participants are depicted, as are certain aspects of their cardiovascular health. The two groups are very similar in all characteristics. The participants were predominantly white, married, college-educated men with an above average income. Most of them had had a qualifying myocardial infarction more than 6 months before admission and most had already returned to work by the time of entry into the study. Many had associated coronary risk factors: smoking (28 percent), hypertension (33 percent), ele-
*
One standard
error
FIGURE 1. Cumulative mortality rates.
vated serum lipids (36 percent), and death of father from coronary heart disease (31 percent). A large proportion (85 percent) had cardiovascular limitations in that they could not reach the age-predicted peak heart rate on the admission treadmill test: The test was stopped in 68 percent because of symptoms and in 17 percent because of signs. Adherence
During the study there was only a modest rate of loss from the periodic examinations, 6 percent in the control subjects and 7 percent in the exercise group. A loss was defined as a participant’s failure to report for the semiannual observation on two consecutive appointments.
TABLE II Mortality of Participants by Study Group and Estlmates of Effectiveness of the Exercise Program
Control Group (n = 328)
Exercise Group (n = 323)
Estimates of Effectiveness of Program (%)
Comparison Statistics Minimum
Events Counted All deaths Cardiovascular deaths Acute myocardial infarction Other definite* Subtotal Sudden deaths+ Total Indeterminate cause
Observed
Maximum
n
%
n
%
X2
P
(CL)
(El
24
7.3
15
4.6
1.62
0.22
-15
37
68
2.4
1 5
0.3
3.96
0.047
a7
98
1.9
2.42
0.13
-5
56
84
4.3
0.70
0.40
-33
29
66
t 14 6 20 4
4.3 6.1
8” 14 1
22
GJ)
Includes six deaths from arrhythmias, two from congestive cardiac failues, one from cardiogenic shock and two from cerebrovascular accidents. + Within 1 hour. CL = lower 95 percent confidence limit: C u = upper 95 percent confidence limit; E = observed effectiveness; p = probability. l
July 1981
The American Journal 01 CARDIOLOGY
Volume 48
41
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
the true effectiveness of exercise ranged widely-from a minimum of -15 percent to a maximum of 68 percent (95 percent confidence limits). Cardiovascular deaths: If the analysis is restricted to cardiovascular deaths, there is little change, because only five deaths were not allocated to cardiovascular causes. If sudden deaths (within 1 hour of onset of symptoms) are excluded, the number of cardiovascular deaths becomes 20 (4.3 percent in the control group and 1.9 percent in the exercise group; difference not significant [NS]). The estimate of effectiveness of treatment is 56 percent, but the confidence interval for this estimate extends from -5 percent to 84 percent. Only nine deaths were due to recurrent myocardial infarction; eight of these occurred in the control group and one in the exercise group (p = 0.05). The estimate of effectiveness was 87 percent, but again there was a wide range of confidence limits (from 22 to 98 percent).
Noncompliance was defined as failure to attend the exercise sessions. By the end of 2 years, 23 percent of the exercise group were not attending the organized exercise program and did not claim to be exercising elsewhere. At the same time, 31 percent of the control subjects alleged that they were exercising regularly. Mortality
During the 3 years of observation per patient, there were 24 deaths (7.3 percent) in the control group of 328 subjects and 15 deaths (4.6 percent) in the exercise group of 323 subjects (Table II). This difference was not statistically significant (p = 0.05). Cumulative mortality rates for progressive time periods are shown in Figure 1. The curves do not differ significantly, at any time or collectively. One can calculate whether the observed results, while not proving a clear benefit, could have by chance underestimated a true benefit of much greater magnitude. To answer this question confidence limits for
Role
the observed benefit are derived. The observed benefit is expressed as follows: Effectiveness
=
X
of risk factors
and cardiovascular
status:
Whether the treatment is appreciably better for some than for other subgroups is evaluated in Table III using coronary risk factors, and in Table IV using factors more directly indicating the cardiovascular status of the subject. (Note that the proportions for the subgroups in the control group give some indication of risk, whereas the estimates of effectiveness give an indication of whether exercise was more beneficial for one group than another.) Table III shows that in the control group, except in the age subdivision, each subgroup with a risk factor had a greater incidence of death. However, there was no
100 = 37 percent,
(that is, exercise appears to have reduced the incidence of death by 37 percent). However, this estimate of effectiveness is rather unstable because both of the component parts are subject to statistical variation. There are no exact methods for determining confidence limits, but two approximate methods were used5,6 and they gave similar results. These methods indicate that TABLE Ill
Relation of Mortallty to Coronary Risk Factors at Entry Control Group
Exercise Group
Deaths Subgroup All subjects Age 150 years Yes No Smoking Yes No Elevated lipids’ Yes No Missing History of hypertension Yes No Father died of coronary heart disease Yes No Presence of above factors Three Two or more One or none
Deaths
Subgroup (n)
n
%
328
24
213 115 a5 243
Estimate of Effectiveness of Program (%)
Subgroup fn)
n
%
7.3
323
15
4.6
37
15 9
7.0 7.8
198 125
11 4
5.6 3.2
21 59
168
9.4 6.6
94 227
2 13
2.1 5.7
77 13
117 189 17
5 9 1
4.3 4.8 5.9
51 4
104 202 22
1: 5
a.7 5.0 22.7
110 218
12 12
10.9 5.5
102 221
8’
6.9 3.6
33:
102 226
9 15
6.8 6.6
102 221
7”
7.8 3.2
::
67 109 110
9 3’
10.3 6.4 2.7
:: 119
5 8 1
5.5 8.3 0.8
-“3; 69
Cholesterol 2250 mg/dl, or triglyceride 2200 mg/dl. In each segment of this table the category shown first represents those presumed to have a higher probability of death; similarly, in the last segment those with the highest probability of death are shown first. l
42
July 1981
The American Journal of CARDIOLOGY
Volume 48
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
TABLE IV Relallon of Mortality to Cardiovascular Status at Entry
ControlGroup
Exercise Group
Subgroup (n)
n
%
Subgroup (n)
n
%
Estimate of Effectiveness of Program (%)
328
24
7.3
323
15
4.6
37
150 178
4 20
2.7 11.2
131 192
2 13
1.5 6.8
::
100 228
15 9
15.0 3.9
2::
11 4
11.1 1.8
26 55
1;: 70
1; 4
12.3 5.9 5.7
64 186 72
: 4
7.8 3.2 5.6
Z
2::
12.5 6.2 0.0
2::
6
1: 0
1:; 39
:
73
:
7.8 23.1 3.6 5.5
ZB 152 75
Deaths Subgroup All subjects Previous MI or complicated QMI
l
No Yes Work capacity <7 Mets Yes No S-T segment depression (> 1 mm at last stage of exercise test) Yes No Missing Systolic blood pressure (~140mmHgatlast stage of exercise test) Yes No Missing Presence of these factors Two Three or four One or none Incomplete information l
Deaths
! 2
4
I :
13.6 5:::
-8 69+
3.5 15.4 1.3 6.7
:: 63
Qualifying myocardial infarction. + The difference between 6.2 and 1.9 is statistically significant (0.01 < p <0.05).
consistent pattern with respect to the benefit of exercise, and no clear-cut benefit from exercise emerged for any subgroup. The greatest benefit was enjoyed by those smoking at entry to the study, but the difference in death rate of smokers and nonsmokers, (9.4 versus 2.1 percent) was not statistically significant. (Moreover, although a 77 percent benefit appeared much better than the 13 percent benefit for the nonsmokers, the difference between these estimates was not significant.) The last panel of Table III examines the joint influence of the risk factors. There was a trend in the death rates in the control group, but no apparent trend in the estimates of benefit from exercise. In Table IV, the importance of each cardiovascular factor selected is fairly clear in the proportions for the control group, but in general the benefit of exercise was no better or worse in the various subgroups. However, the results in one subgroup suggest a “significant” finding: The subjects who had satisfactory systolic blood pressure responses during the last stage of the exercise
test (that is, more than 140 mm Hg) enjoyed a 69 percent benefit from exercise in reduction of mortality. Finally, the last panel of Table IV presents the joint influence of these selected factors and demonstrates a clear trend in incidence of death in the control subjects. The analysis of effect of exercise suggests that men with more evidence of disease benefited less from the exercise program. However, the trend, from 33 to 63 percent, was not statistically significant. Multivariate analysis: Multivariate regression procedures that adjust for any imbalance in the study groups were used in the 490 subjects whose electrocardiographic S-T segment data during exercise were available. The techniques yielded little change from the crude data, as indicated in Table V. The factors used in the logistic model and the Cox model are shown in Table VI. Moreover, when age (below or above age 50 years) and smoking status at entry were added to each model, the estimate of effectiveness did not change, very likely because the two study groups were well balanced on
TABLE V Comparison of Crude Death Rates With Two Types of Adjusted Rates
Sample size Crude death rates (3 yr) Rates using logistic model7 (3 yr) Rates using Cox models (3 yr)
Control Group
Exercise Group
Estimate of Effectiveness
249 8.0% 8.0% 8.2%
241 4.1% 4.1% 4.1%
48% 46% 48%
July lQ81
The American Journal of CARDIOLOGY
Volume 48
43
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
TABLE VI Factors Used in Logistic Model and the Cox Model
Myocardial infarction history* Work capacity (Mets) S-T segment depression > 1 mm at last stage of exercise test History of hypertension Heart size (vol/body surface area) Treatment group l
More than one myocardial infarction or complicated
Favorable
Unfavorable
No >_7 No No
Yes 16 Yes Yes lMedian Control
qualifying myocardial infarction
these factors. Finally, inclusion of peak systolic blood pressure during the admission treadmill test produced no substantial change in the analysis of the effect of exercise on mortality.
The total number of rehospitalizations for all reasons other than myocardial infarction was examined
Morbidity
If the need for coronary arterial surgery is viewed as a measure of failure of therapy, no differential between
for any indication of a difference due to exercise. None was found; the rehospitalization rates were virtually identical in the two study groups.
In addition to the 34 patients who died of cardiovascular causes (Table II), 26 patients had a nonfatal myocardial infarction; thus there were 60 cases of serious cardiovascular morbidity to analyze (Table VII, subtotal line). The case rate was 9.5 percent for control subjects and 9.0 percent for the exercise group; the estimate of benefit from exercise was only 5 percent.
the two groups occurred. Coronary arterial bypass surgery was performed in 16 control subjects and 17 in the exercise group. Although there was no apparent benefit from exercise in relation to morbidity, no harm was evident (that is, the event rates in the exercise group were not higher than in the control group).
The domain of review was broadened to include all instances of cardiovascular illness, by adding five
Discussion
suspected myocardial infarctions and 50 hospitalizations for other cardiovascular illnesses (Table VII). This yielded a total of 115 participants experiencing some kind of cardiovascular morbidity. The event rate was 17.7 percent for the control subjects and 17.6 percent for the exercisers, indicating no apparent benefit from exercise. The upper confidence limit of 31 percent indicates that the true benefit from exercise, in preventing aggregate cardiovascular morbidity, does not exceed a 31 percent reduction. The results for all myocardial infarctions, fatal and nonfatal, are shown in an added line of Table VII. The estimate of benefit from exercise was only 25 percent, with confidence limits of -34 and 66 percent.
Exercise and cardiac mortality: The results of this study suggest that a program of prescribed supervised physical activity for patients after myocardial infarction may be beneficial in reducing subsequent cardiac mortality, but the evidence is not convincing. Depending on whether one focuses on the narrow domain of deaths due to recurrent myocardial infarction or looks more broadly at all cardiovascular mortality, a very different estimate of benefit from the exercise program is observed. The apparent reduction in deaths from myocardial infarction was 87 percent, whereas it was 29 percent for total cardiovascular mortality. The latter degree of success would certainly be a worthy benefit,
TABLE VII Morbidity Experience of Study Participants Control Group (n = 328) Events Counted
Exercise Group (n = 323)
Estimates of Effectiveness (%)
Comparison Statistics P
Minimum (CL)
Observed (E)
Maximum (Cu)
0.70
0.40
-33
29
66
0.005
0.95
-57
5
42
n
%
n
%
20
6.1
14
4.3
;:
9.5
:9”
9.0
::
17.7
2; 57
17.6
0.00
1.00
-44
0
31
All recurrent myocardial infarctions
23
7.0
17
5.3
0.59
0.46
-34
25
66
Total hospitalizations for reasons other than myocardial infarction
90
27.4
92
28.5
0.04
0.86
-34
Cardiovascular deaths Nonfatal infarctions Subtotal Suspected infarctions Other events Total
Abbreviations
44
July 1991
2
as in Table II.
The American
X2
Journal of CARDIOLOGY
Volume
49
-4
21
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
but because the study group was small, the difference in death rates, 6.1 versus 4.3 percent, was not statistically significant. These results may be compared with those reported by Wilhelmsen et a1.g In a randomized clinical trial of the value of exercise, they observed 4 year cardiac mortality rates of 21.0 and 14.6 percent in their control and exercise groups, respectively. This yielded an effectiveness rate of 31 percent, which is remarkably close to our estimate of 29 percent despite very wide confidence limits pertaining to each estimate. Results of the two studies (either separately or collectively) were not statistically significant. Exercise and cardiovascular morbidity: There is no suggestion of benefit from the exercise program in relation to cardiovascular morbidity; the event rates in the two study groups were virtually identical (18 percent of the subjects in each group had some cardiovascular episode in 3 years of follow-up study). This may contrast somewhat with the favorable findings of Wilhelmsen et a1.,g who reported a 22 percent reduction in cardiovascular morbidity. The results reported here can also be compared with the data of Rechnitzer’O from the randomized clinical trial conducted in Ontario, Canada. In that study a high and a low intensity exercise program were compared. The preliminary report showed no benefit; indeed, patients in the high intensity group had a greater rate of reinfarction than did those in the low intensity group. Methodologic considerations: The value of having a randomized control group is apparent in this report. The total mortality rate in the exercise group was far below that predicted (only 4.6 percent in 3 years versus an “expectation” of 12 to 15 percent). Without an appropriate control group, it might have been speculated (as some investigators have donell), that a very substantial benefit from exercise was demonstrated. But the data on the control group show that the participants in this study were a rather favorable subset of subjects in the postmyocardial infarction state; their mortality rate was half that “expected” in advance of the trial. (A report from the Coronary Drug Project12 showed that it is not difficult to identify a subgroup of patients after a myocardial infarction who may be expected to have a relatively low death rate.) The “null hypothesis”is important but not the only hypothesis to consider. One could ask: How good could
exercise really be, and yet the observed findings occur? The upper confidence limit indicates that the true benefit of exercise could be to reduce the cardiac mortality rate by 66 percent, but in this study, with a small sample of 651 patients, the observed reduction was only 29 percent. At the same time, it should be stated that
the lower confidence limit of -33 percent means that exercise could be hazardous, but this study failed to show a risk. This study, because of its sample size limitations, was not designed to give a definitive answer to the question of mortality (or morbidity) in respect to the effect of an exercise program. (It was originally estimated that a large study of 4,200 participants was desirable.) Because the exercise group was seen much more often than the control group in this nonblind clinical trial, there may have been some excess ascertainment of the milder forms of morbidity in the exercise group. Did the participants
follow the prescribed
therapy?
It appears that adherence to the therapy plan was good in the exercise group and only fair in the control group. Some lagging in enthusiasm and attendance was noted after the 1st year of the exercise program and some “exercising on their own” was done by some control subjects. This response, of course, dilutes the estimates of true benefit of exercise but it does not interfere with assessment of benefit of the “program” (that is, intensive long-term exercise). Some persons have postulated that the preliminary “prerandomization exercise program” damaged the value of the postrandomization research program, in that enthusiasm for exercise was engendered and some control subjects rejected their assignment and continued to exercise. If this occurred, the study yielded a very useful “public health” finding, namely, that an expensive long-term vigorous exercise program does not offer significantly greater benefit than a short-term introduction to mild exercise. Did those who faithfully exercised have a better outcome than the control subjects who did not exercise?
It does not seem appropriate to answer this question because of the likelihood of selection bias; that is, sicker people may withdraw from the exercise program and healthier control subjects may pursue exercise, making the adherers “noncomparable.” Finally, whereas the analyses by subgroups often showed expected patterns in respect to the natural history of chronic heart disease, the suggestion of greater benefit from exercise for some subgroups may easily be due to chance-especially because so many comparisons were examined. Implications of study: The case for exercise in persons with known myocardial infarction is neither proved nor disproved. Enough information was gathered from this trial and sufficient success in obtaining continued participation was achieved to conclude that a full scale trial of the merits of exercise for patients with cardiac disease is feasible. It is of interest that if the 37 percent benefit from exercise reported had been observed in a comparable study of as few as 1,400 patients, the result would have been judged “statistically significant.”
References 1. Heiierstein HK, Buriando A, Hirsch EZ, et al. Active physical reconditioning of coronary patients. Circulation 1965;31,32:Suppl li:ll-110-l. 2. Naughton J, Lategoia MT, Shanbour K. A physical rehabilitation program for cardiac patients. A progress report. Am J Med Sci
1966;252:545-53. 3. Fox SM iii, Naughton JP, Haskeii WL. Physical activity and the prevention of coronary heart disease. Ann Clin Res 1971;3: 404-32. 4. Naughton J. The national exercise and heart disease project: de-
July 1981
The American Journal of CARDIOLOGY
Voiume 48
45
NATIONAL EXERCISE AND HEART DISEASE PROJECT-SHAW
5.
6. 7.
8.
46
velopment recruitment, and implementation. In: Wenger NK, ed. Exercise and the Heart. Philadelphia, PA: Cardiovasc Clin, 1978;205-22. Noether GE. Two confidence intervals for the ratio of two probabilities and some measures of effectiveness. J Am Stat Assoc 1957;52:36-45. Brass I. A confidence interval for a percentage increase. Biometrics 1954;10:245-50. Cornfleld J. Joint dependence of risk of coronary heart disease on serum cholesterol and systolic blood pressure: a discriminant function analysis. Fed Proc 1962;21:58-61. Cox DR. Regression models and life tables (with discussion). J R Stat Sot (B) 1972;34: 187-220.
July 1961
The American Journal of CARDIOLOGY
Volume 48
9. Wilhelmsen L, Sanne H, Elmfeldt D, Grfmby G, Tlbblln G, Wedel H. A controlled trial of physical training after myocardial infarction. Prev Med 1975;4:491-508. 10. Rechnltzer PA. The effects of training: reinfarction and death-an interim report. Med Sci Sports 1979;11:382. 11. Kavanagh T, Shephard RJ, Chlsholm AW, Qureshi S, Kennedy J. Prognostic indexes for patients with ischemic heart disease enrolled in an exercise-centered rehabilitation program. Am J Cardiol 1979;44:1230-40. 12. Coronary Drug Project Research Group. Factors influencing long-term prognosis after recovery from myocardial infarction. J Chron Dis 1974;27:267-85.