Impaired Heart Rate Response to Graded Exercise
Prognostic Implications of Chronotropic Incompetence in the Framingham Heart Study
Background Previous reports have suggested that an attenuated exercise heart rate response may be associated with coronary heart disease risk and with mortality. These observations may parallel the association between reduced heart rate variability during normal activities and adverse outcome. This investigation was designed to look at the prognostic implications of exercise heart rate response in a population-based sample.
Methods and Results In this prospective cohort investigation, 1575 male participants (mean age, 43 years) in the Framingham Offspring Study who were free of coronary heart disease, who were not taking β-blockers, and who underwent submaximal treadmill exercise testing (Bruce protocol) were studied. Heart rate response was assessed in three ways: (1) failure to achieve 85% of the age-predicted maximum heart rate, which has been the traditional definition of chronotropic incompetence; (2) the actual increase in heart rate from rest to peak exercise; and (3) the ratio of heart rate to metabolic reserve used by stage 2 of exercise (“chronotropic response index”). Proportional hazards analyses were used to evaluate the associations of heart rate responses with all-cause mortality and with coronary heart disease incidence during 7.7 years of follow-up. Failure to achieve target heart rate occurred in 327 (21%) subjects. During follow-up there were 55 deaths (14 caused by coronary heart disease) and 95 cases of incident coronary heart disease. Failure to achieve target heart rate, a smaller increase in heart rate with exercise, and the chronotropic response index were predictive of total mortality and incident coronary heart disease (P<.01). Failure to achieve target heart rate remained predictive of incident coronary heart disease even after adjusting for age, ST-segment response, physical activity, and traditional coronary disease risk factors (adjusted hazard ratio, 1.75; 95% confidence interval, 1.11 to 2.74; P=.02). After adjusting for the same factors, the increase in exercise heart rate remained inversely predictive of total mortality (P=.04) and coronary heart disease incidence (P=.0003). The chronotropic response index also was predictive of total mortality (P=.05) and incident coronary heart disease (P=.001) after adjusting for age and other risk factors.
Conclusions An attenuated heart rate response to exercise, a manifestation of chronotropic incompetence, is predictive of increased mortality and coronary heart disease incidence.
Recently, there has been increasing interest in the role of autonomic nervous system regulation in heart disease. Heart rate (or RR cycle) variability, one marker of autonomic activity, has been found to be an important marker of risk both among survivors of myocardial infarction1 and among healthy adults in the Framingham Heart Study.2 Therefore, we were interested in studying an easily obtained measure that reflects autonomic regulation, namely, the heart rate response to a standard, graded exercise test.3 4 We hypothesized that just as reduced heart rate variability is associated with an adverse outcome,1 2 so would an attenuated heart rate response to exercise be predictive of risk.
In patients with known coronary heart disease an attenuated exercise heart rate response, sometimes referred to as chronotropic incompetence,5 has been associated with an adverse prognosis.5 6 7 8 Although a blunted heart rate response to exercise has been shown to be a predictor of subsequent coronary heart disease events in asymptomatic individuals, this finding was not clearly independent of concurrent ischemic ST-segment changes, age, physical fitness, and standard coronary heart disease risk factors.9
The purpose of this study was to determine the relation between exercise heart rate response and prognosis in asymptomatic subjects from the Framingham Heart Study. End points of interest included all-cause mortality and coronary heart disease events.
The Framingham Heart Study was initiated in 1948 as an epidemiological study of cardiovascular disease precursors. In the early 1970s, offspring and spouses of offspring of the original study cohort were enrolled in the Framingham Offspring Study. Study design and selection criteria for the original and offspring cohorts have been described in detail elsewhere.10 11 12 Members of the Offspring Study have undergone follow-up examinations 8, 12, and 16 years after the initial examination. At the time of the second Framingham Offspring Study examination, subjects underwent a physician-obtained medical history, physical examination, ECG, exercise treadmill test according to the Bruce protocol,13 and fasting blood tests including glucose and lipid profiles. Informed consent was obtained from all subjects.
To be eligible for this study, subjects had to undergo exercise treadmill testing. Exclusion criteria included prevalent coronary heart disease, technical problems with the treadmill, an inability to reach stage 2 in a standard Bruce protocol (to eliminate confounding effects of very early heart rate responses due to anxiety and also to reduce possible effects of occult angina presenting as very poor exercise tolerance), and use of β-blockers at the time of the treadmill test. Heart rate blunting calcium blockers like verapamil and diltiazem were not yet in widespread use. This analysis was restricted to men because there were only 20 deaths in the women studied.
At baseline examination, which occurred between 1979 and 1983, all subjects had height and weight measured. Obesity was assessed using body mass index (weight in kilograms divided by height in meters squared, kg/m2). Systolic blood pressure was obtained by a physician using a mercury column sphygmomanometer: two readings were averaged. Baseline hypertension was defined as a systolic blood pressure of 140 mm Hg or more, a diastolic blood pressure of 90 mm Hg or more, or use of antihypertensive medication.14 Diabetes was defined as use of insulin or oral hypoglycemic agents or a fasting blood glucose of at least 140 mg/dL (7.77 mmol/L) at the index examination. Usual levels of physical activity were assessed by a questionnaire in which subjects were asked how many hours per day were spent engaging in sleep, sedentary activity, and slight, moderate, and heavy physical activity. Based on the answers to these questions a physical activity index was calculated.15
Subjects were followed for a mean of 7.7 years for all-cause mortality and incident coronary heart disease events. The coronary heart disease end points included incident angina pectoris, coronary insufficiency (unstable angina with documented ischemic ST-segment changes), myocardial infarction, sudden and nonsudden coronary heart disease deaths, and coronary revascularization (coronary artery bypass grafting or percutaneous transluminal coronary angioplasty). All available medical records, ECGs, and laboratory data were reviewed by a committee of three physicians to assign cardiovascular disease diagnoses. This committee had no knowledge of the subjects’ exercise responses. The criteria for cardiovascular disease diagnoses have been described in detail elsewhere.16
All subjects underwent standard treadmill exercise testing on the same day as the baseline examination according to the Bruce protocol.13 Exercise was stopped when subjects achieved a target heart rate (in beats per minute) defined as 85% of the age- and sex-predicted maximum heart rate,17 which in men were age 16 to 20 years, maximum heart rate 179; age 21 to 24 years, 177; age 25 to 29 years, 175; age 30 to 34 years, 173; age 35 to 39 years, 172; age 40 to 44 years, 170; age 45 to 49 years, 168; age 50 to 54 years, 166; age 55 to 59 years, 164; age 60 to 64 years, 162; and age 65 to 69, 158. Other reasons for stopping exercise included participant request, limiting chest discomfort, dyspnea, fatigue, leg discomfort, hypotension, an excessive increase in systolic blood pressure (peak systolic pressure ≥250 mm Hg), ≥2 mm ST-segment depression, or significant ventricular ectopy (including frequent ventricular beats and nonsustained ventricular tachycardia). Heart rate and blood pressure were measured at rest, during each stage of exercise, at peak exercise and during recovery. An ST-segment response was considered ischemic if there was at least 1 mm (0.1 mV) of additional (versus the resting ECG) horizontal or downsloping ST-segment depression measured 80 msec after the J-point. Exercise capacity in metabolic equivalents (METs) was estimated based on a previously published nomogram relating to the Bruce protocol18 : stage 2, 7.1 METs; stage 3, 9.9 METs; stage 4, 13.5 METs; and stage 5, 16.5 METs. Subjects were not credited with achieving any given stage of exercise unless they reached the point when blood pressure was measured, namely, 1.5 to 2.0 minutes into that stage.
Assessment of heart rate response. The exercise heart rate response was assessed in three ways: (1) ability or failure to achieve the target heart rate, (2) actual increase in heart rate from rest to peak exercise (in beats per minute), and (3) the ratio of heart rate to metabolic reserve used by stage 2 of exercise. All subjects by definition had to achieve stage 2; thus using measurements at this stage of exercise enables inclusion of all subjects.19 This last measure of chronotropic competence is based on the notion that a valid measure of heart rate response must take into account age, resting heart rate, and exercise capacity19 ; it is not merely a marker of treadmill time. Percent metabolic reserve (MR) used by any stage of exercise can be defined as
In an analogous fashion, the percent heart rate reserve (HRR) used by any stage of exercise is
In a healthy subject, the ratio of heart rate to metabolic reserve used by any stage of exercise (heretofore referred to as the chronotropic response index) is roughly 1, a reflection of the association between heart rate response and metabolic work during exercise.19 A low ratio implies chronotropic incompetence, which can be distinguished from effects of age, resting heart rate, and physical fitness on the heart rate response to exercise.19 This ratio should be accurate despite the submaximal nature of the exercise test because of the linear relationship between heart rate and metabolic work during exercise.19
To confirm that the chronotropic response index is relatively independent of age, resting heart rate, and physical fitness, Pearson correlation coefficients (along with standard errors of the estimate) were calculated relating chronotropic response index and heart rate increase with exercise to age, resting heart rate, and exercise capacity in METs.
Description of baseline and exercise characteristics. Subjects were grouped according to ability to achieve target heart rate and group-specific values of baseline and exercise characteristics were determined. In separate analyses, subjects were split according to tertiles of the chronotropic response index, with the lowest tertile representing the greatest degree of chronotropic incompetence. The tertile partition values were 0.15 to 0.95 for the lowest tertile, 0.95 to 1.13 for the middle tertile, and 1.13 to 1.77 for the highest tertile. Subjects were not divided into quartiles because this would have left too few events per quartile for meaningful descriptions. ANOVA and χ2 tests were used to compare means and proportions among groups.
Outcome analyses. The end points of this study were all-cause mortality and incident coronary heart disease; none of the subjects had coronary heart disease at baseline. For descriptive purposes, Kaplan-Meier cumulative incidence plots of all-cause mortality and incident coronary heart disease were constructed according to ability to achieve target heart rate and according to tertiles of the chronotropic response index.
The Cox proportional hazards model20 was used to quantify the associations between exercise heart rate responses and these end points. For analyses of mortality, Cox models were adjusted for age, ST-segment response, valvular disease, pulmonary disease, body mass index, smoking status, hypertension, hypertension treatment, diabetes, physical activity index, and the ratio of total to HDL cholesterol. Because valvular disease and pulmonary disease were uncommon, they were combined for modeling purposes. Models of coronary heart disease incidence were adjusted for age, ST-segment response, body mass index, smoking status, hypertension, hypertension treatment, diabetes, and the ratio of total to HDL cholesterol.
In supplementary analyses, the sample was restricted to subjects who remained free of events during the 2 years of follow-up to reduce any potential bias caused by subclinical conditions at the time of exercise testing.
To investigate the possibility of a J-shaped curve (in which subjects with an excessive heart rate response in early exercise might be at increased risk) further analyses were restricted to those with a chronotropic response index of at least one. Subjects with a chronotropic response index of ≥1.3 were considered to have a high chronotropic response. Their risks for all-cause mortality and incident coronary heart disease were compared with subjects whose chronotropic response indexes were between 1.0 and 1.3 using Cox regression analyses.
Proportional hazards analyses were performed using PROC PHREG of the SAS statistical packages21 on a Sun Sparcstation 2.
The baseline characteristics of eligible subjects are summarized in Tables 1⇓ and 2⇓. None of the subjects had atrial fibrillation or congestive heart failure at baseline and none was taking digitalis. Of the 1575 men, 327 (21%) failed to achieve 85% of the age-predicted maximum heart rate. Compared with subjects who achieved their target heart rate, those who failed to do so were older, had higher mean values for body mass index, resting blood pressures, and total to HDL cholesterol ratio, and were more likely to smoke cigarettes, have clinical hypertension, and pulmonary disease (Table 1⇓). The most commonly used antihypertensive drugs included thiazide diuretics (n=96), α-methyldopa (n=27), loop diuretics (n=3), spironolactone (n=3), and miscellaneous other agents including α-blockers and direct vasodilators (n=54).
When subjects were divided according to tertiles of chronotropic response index, there were no marked differences in age, blood pressure, body mass index, resting heart rate, and the ratio of total to HDL cholesterol; smoking was somewhat more prevalent among those in the lowest tertile of chronotropic response index (Table 2⇑). The physical activity index was unrelated to chronotropic response (Tables 1⇑ and 2⇑).
The exercise characteristics of subjects are summarized in Tables 3⇓ and 4⇓. Reasons for stopping exercise included dyspnea (n=83), leg discomfort (n=75), marked ischemic ST-segment response (n=24), excessive increase in systolic blood pressure (n=8), participant request (n=7), complex ventricular ectopy (n=4). One subject stopped because of chest discomfort. As expected, the chronotropic response index was about 1.0 in subjects who reached their target heart rate but was lower (0.86±0.22, P<.0001) in those who did not achieve the target heart rate. Subjects who failed to achieve target heart rate were more likely to exhibit an ischemic ST-segment response (18% versus 13%, P=.054); they also had a lower exercise capacity (Table 3⇓). When divided according to tertiles of chronotropic response index, there were no marked differences in abnormal ST-segment response; exercise capacity was slightly higher among subjects in the highest tertile of chronotropic response index (Table 4⇓).
Heart Rate Measures and Age, Resting Heart Rate, and Exercise Capacity
The increase in heart rate with exercise was weakly correlated with age (r=−.41, SEE=0.023), resting heart rate (−.62, SEE=0.020), and exercise capacity in METs (r=.52, SEE=0.022). In contrast, the chronotropic response index was uncorrelated with age (r=.005, SEE=0.025), resting heart rate (r=−.01, SEE=0.025), and it was only weakly correlated with exercise capacity in METs (r=.21, SEE=0.024). The physical activity index was not correlated with the chronotropic response index (r=−.003). The chronotropic response index represents an exercise heart rate response variable that is uncorrelated or minimally correlated with age or level of physical fitness.
Prognostic Implications of Exercise Heart Rate Response
During a mean of 7.7 years of follow-up there were 55 deaths: 14 from coronary heart disease, 27 from cancer, 2 from other cardiovascular causes, and 12 from miscellaneous causes. There were 95 incident coronary heart disease events including 45 myocardial infarctions, 44 cases of angina pectoris or coronary insufficiency, and 6 sudden cardiac deaths. No cases of coronary revascularization occurred as incident coronary heart disease events.
An inability to achieve target heart rate and a lower chronotropic response index were related to a higher incidence of all-cause mortality (Figs 1⇓ and 2⇓) and coronary heart disease events (Figs 3⇓ and 4⇓).
Of the 327 men who failed to reach the target heart rate, 21 (6%) died; however, of the 1248 who achieved their target heart rate only 34 (3%) died. There were 23 deaths (4%) among those in the lowest tertile of chronotropic response index, 16 (3%) deaths in the middle tertile, and 16 (3%) in the highest tertile. In unadjusted analyses, failure to achieve target heart rate was associated with an increased risk of death, but after adjusting for several covariates it was no longer predictive of death risk. The increase in heart rate was inversely related to mortality risk after adjusting for the same covariates (P=.04). Similarly, an impaired early heart rate response through stage 2 of exercise, as assessed by the chronotropic response index was predictive of mortality (P<.05). See Table 5⇓.
Coronary Heart Disease Events
Of the 327 men who failed to reach their target heart rate, 44 (14%) experienced a coronary event (25 had myocardial infarctions, 15 developed angina pectoris, and 4 had sudden cardiac death as the initial presentation of coronary heart disease); of the 1248 men who reached their target heart rate, only 51 (4%) experienced a coronary event. There were 44 (8%) coronary heart disease events among those in the lowest tertile of chronotropic response index, 35 (7%) in the middle tertile, and 16 (3%) in the highest tertile. Failure to achieve target heart rate was associated with an increased risk of incident coronary heart disease when studied alone. After adjusting for age, ST-segment response, body mass index, smoking, hypertension, hypertension treatment, diabetes, physical activity index, and the ratio of total to HDL cholesterol, failure to achieve target heart rate remained predictive of coronary heart disease events (adjusted hazard ratio, 1.75; 95% confidence interval, 1.11 to 2.74; P=.02). The increase in heart rate was inversely predictive of incident coronary heart disease after adjusting for the same covariates (P=.0003). Similarly, the chronotropic response index was inversely predictive of incident coronary heart disease (P=.001). See Table 5⇑.
In analyses restricted to subjects who had at least 2 years of event-free follow-up, similar associations of heart rate response with late incident coronary heart disease events were noted, with no material difference from the prior results (Table 6⇓ is provided for reviewers). None of the heart rate variables was independently predictive of late all-cause mortality.
Impact of an Excessive Heart Rate Response in Early Exercise
It might be postulated that subjects with high chronotropic responses (chronotropic response index ≥1.3) are at increased risk for an adverse outcome. There were 891 subjects with a chronotropic index of at least 1; during follow-up there were 26 deaths and 44 incident coronary heart disease events. Compared with subjects with a chronotropic index between 1.0 and 1.3, those with an index exceeding 1.3 had a similar all-cause mortality rate (hazard ratio, 1.34; 95% confidence interval, 0.54 to 3.35; P>.5) and comparable risk of coronary heart disease events (hazard ratio, 0.88; 95% confidence interval, 0.39 to 1.99; P>.7).
In this group of 1575 men derived from a population-based sample, an impaired heart rate response on standard exercise testing was related to the risk of an adverse outcome after adjusting for age, ST-segment response, physical activity, and traditional coronary disease risk factors including diabetes, smoking, hypertension, antihypertensive therapy, and the ratio of total to HDL cholesterol. An impaired heart rate response was associated with higher total mortality and with increased risk of coronary heart disease. In addition, a blunted heart rate response was associated with late incident coronary heart disease events (events occurring at least 2 years after the treadmill test).
The heart rate response to exercise is related to a complex interplay among many factors including age, sex, physical conditioning, sympathetic drive, baroreceptor reflexes, and venous return, as has been reviewed elsewhere.7 Several studies have demonstrated an association between failure to achieve a predicted heart rate and coronary artery disease prognosis.5 6 7 8 9 Only one of those was population-based9 ; in that study, failure to attain 90% of the age-predicted maximum heart rate was predictive of coronary heart disease risk. That study, however, did not rigorously adjust for ST-segment changes and traditional coronary risk factors in multivariable models. Another study of employed men found sustained slow heart rates may be predictive of a higher risk of cardiac death.22 In that study there were no noted exclusions for β-blocker use or preexisting coronary heart disease.
The mechanism by which an impaired exercise heart rate response may be associated with increased risk of coronary heart disease is unclear. An attenuation of sympathetic drive has been demonstrated for patients with established heart failure23 ; perhaps a similar attenuation may occur in very early subclinical manifestations of cardiovascular disease. Another possible explanation is that an impaired exercise heart rate response may well represent an early manifestation of cardiac ischemia.5
The association of an impaired exercise heart rate response and increased mortality and coronary risk may be analogous to the association of reduced heart rate variability (or RR cycle variability) with adverse outcome.1 2 A recent report from the Framingham Heart Study2 found that time-domain and frequency-domain variables derived from a 2-hour Holter monitor recording predicted mortality risk in conjunction with standard cardiovascular risk factors. In contrast to these measures of heart rate variability, the exercise heart rate response may represent an easily and routinely obtained measure that is associated with autonomic regulation.
The increased risk associated with an impaired heart rate response may reflect decreased physical fitness, which has been shown to be predictive of cardiac risk.24 Since the exercise tests obtained in this study were submaximal, our ability to accurately assess the impact of physical fitness is limited. To separate effects of physical fitness from heart rate response, analyses were performed looking only at heart rate response during early exercise (through stage 2 of a Bruce protocol). In models of mortality and coronary heart disease risk, the chronotropic response index was predictive of risk. Of note, the chronotropic response index was only weakly (r=.21) associated with exercise capacity (and therefore was not merely a marker of treadmill duration or physical fitness) and was unassociated with age or resting heart rate. Among subjects in the highest tertile of chronotropic response index exercise capacity was slightly higher (13.5 versus 12.7), an association that was statistically significant given the large sample size but not particularly marked. In contrast, failure to achieve target heart rate was strongly associated with age (P<.0001), resting heart rate (P=.02), and physical fitness (P<.001) (Tables 1⇑ and 3⇑). Increase in heart rate with exercise was also correlated with age, resting heart rate, and physical fitness. Thus, of the three measures of heart rate response we considered, the chronotropic response index comes closest to a pure measure of chronotropic competence. Also, baseline physical activity, in the form of a physical activity index,15 was also considered and was found to be unassociated with chronotropic response. It did not influence the association between chronotropic incompetence and adverse outcome.
Inability to achieve target heart rate was associated with a higher likelihood of an ischemic ST-segment response, a lower exercise capacity, and an adverse coronary heart disease risk factor profile. As such, these true confounders of the association of heart rate response to exercise with risk for death and coronary heart disease needed to be considered. Stratification was impractical with several confounders,25 so multivariable proportional hazards models were adopted. We assumed that the primary variables have certain mathematical relations to the risks of death and of developing coronary disease, and also that those relations are not modified by confounders. Although the data were too sparse to check these assumptions fully, we found that the results from unadjusted and adjusted models were consistent with our expectations: the effect estimated for each primary variable was attenuated somewhat upon adjustment for confounders.
There are other potential limitations of this study. The study sample was all male and overwhelmingly white, so the results may not apply to females and nonwhites. Despite the increased risks associated with attenuated exercise heart rate responses, most of the deaths and coronary heart disease events occurred in individuals who did achieve their target heart rate and who were in the top two tertiles of chronotropic response index. Exercise capacity in METs was estimated, not directly measured using gas exchange techniques. We were unable to document the accuracy of the estimated exercise capacity or metabolic reserve, which was used for calculating chronotropic index because metabolic testing was not used at Framingham during the testing period. Thus, the chronotropic index is a calculated, rather than a directly measured value. The use of the Bruce protocol, with its incremental stages, may result in an overestimation of exercise capacity.26 The major stopping point for our exercise test was achievement of a target heart rate, based on 85% of the age-predicted maximum, which has a fairly high standard deviation7 and which, therefore, may not be ideally applicable for individual subjects. The use of a submaximal test, instead of a symptom-limited test, may be a potential limitation, since many individuals could have achieved higher heart rates than they did. By truncating heart rate increase with exercise in this way, the differences were diminished between heart rate increases in those who failed, compared with those who reached their target heart rate. This, if anything, would attenuate the association between heart rate response and outcome; there would be a bias toward no effect.
Despite these limitations, heart rate response to exercise was associated with an increased risk for death and coronary heart disease even after considering ST-segment response to exercise, baseline physical activity, and traditional coronary disease risk factors. Furthermore, we have demonstrated that the chronotropic response index, a measure of exercise heart rate response that is relatively independent of age, resting heart rate, and physical fitness, is a strong predictor of outcome.
Reprint requests to Dr Lauer, Department of Cardiology, Cleveland Clinic Foundation, Desk F-15, 9500 Euclid Ave, Cleveland, OH 44195.
- Received September 27, 1995.
- Revision received January 16, 1996.
- Accepted January 22, 1996.
- Copyright © 1996 by American Heart Association
Bigger JT, Fleiss JL, Rolnitzky LM, Kleiger RE, Rottman JN. Frequency domain measures of heart period variability and mortality after myocardial infarction. Circulation. 1992;85:164-171.
Tsuji H, Venditti FJ, Manders ES, Evans JC, Larson MG, Feldman CL, Levy D. Reduced heart rate variability and mortality risk in an elderly cohort: the Framingham Heart Study. Circulation. 1994;90:878-883.
Gianrossi R, Detrano R, Mulvihill D, Lehmann K, Dubach P, Colombo A, McArthur D, Froelicher VF. Exercise-induced ST depression in the diagnosis of coronary artery disease: a meta analysis. Circulation. 1989;80:87-98.
McNeer JF, Margolis JR, Lee KL, Kisslo JA, Peter RH, Kong Y, Behar VS, Wallace AG, McCants CB, Rosati RA. The role of the exercise test in the evaluation of patients for ischemic heart disease. Circulation. 1978;57:64-70.
Ellestad MH, Wan MK. Predictive implications of stress testing: follow-up of 2700 subjects after maximum treadmill stress testing. Circulation. 1975;51:363-369.
Dawber TR, Meadors GR, Moore FE. Epidemiologic approaches to heart disease: the Framingham Heart Study. Am J Publ Health. 1951;41:279-286.
Dawber TR, Kannel WB, Lyell LP. An approach to longitudinal studies in a community: the Framingham Study. Ann N Y Acad Sci. 1963;107:539-556.
Kannel WB, Feinlieb M, McNamara PM, Garrison RJ, Castelli WP. An investigation of coronary heart disease in families: the Framingham Offspring Study. Am J Epidemiol. 1979;110:281-290.
Shurtleff D. Some characteristics related to the incidence of cardiovascular disease and death: the Framingham Study 18-year follow-up. In: Kannel WB, Gordon T, eds. The Framingham Study: An Epidemiologic Investigation of Cardiovascular Disease. Section 30. Washington, DC: Government Printing Office, 1974. DHEW publication No. (NIH) 74-599.
Sheffield LT. Graded exercise tests for ischemic heart disease. In: Exercise Testing and Training of Apparently Healthy Individuals: A Handbook for Physicians. Dallas, Tex: American Heart Association; 1975:35-38.
Cox DR. Regression models and life tables (with discussion). J R Stat Soc B. 1972;34:187-220.
SAS Institute Inc. SAS/STAT User’s Guide. Version 6, 4th ed. Cary, NC: SAS Institute Inc; 1989.
Paffenbarger RS Jr, Wing AL, Hyde RT. Physical activity as an index of heart attack risk in college alumni. Am J Epidemiol. 1978;108:161-175.
Rothman KJ. Modern Epidemiology. Boston, Mass: Little, Brown, & Co; 1986:306-307.