Heart Rate Recovery after Submaximal Exercise Testing as a Predictor of Mortality in a Cardiovascularly Healthy Cohort

We previously reported on the prognostic importance of heart rate recovery after symptom-limited exercise in an intermediate-risk group of patients who were referred for exercise testing with thallium single-photon emission computed tomography in a tertiary-care center (1). It is unknown, however, whether these findings can be generalized to healthy adults undergoing submaximal exercise. Our goal was to examine heart rate recovery as a predictor of long-term mortality in a population-based cohort of adults without evidence of cardiovascular disease who underwent submaximal exercise testing. All-cause mortality was the outcome of interest because of its relevance and its objective and unbiased nature (2). Methods The study cohort was derived from the Lipid Research Clinics Prevalence Study, the selection of which has been described in detail elsewhere (3-6). Briefly, that study was designed to determine the prevalence of lipid abnormalities among North Americans. Of the 68 317 population-based persons recruited from 10 primary care centers, those with lipid abnormalities (approximately 10% of the population) and an additional randomly selected 15% were invited to undergo more extensive investigation (3). Of the 13 852 who appeared for this second visit, 8681 underwent exercise testing. Reasons for exclusion from exercise testing included resting systolic blood pressure less than 90 or greater than 200 mm Hg, resting diastolic blood pressure greater than 120 mm Hg, inability to complete a treadmill test on the basis of physician examination, or participant request (5, 6). To assure selection of a cardiovascularly healthy asymptomatic cohort similar to that in previous population-based studies (7), persons were excluded if they were younger than 30 years of age (n=131 [2%]); were taking -blockers, digoxin, antiarrhythmic agents, or nitrates (n=356 [4%]); had a history of cardiovascular disease (defined as previous heart surgery or history of angina, myocardial infarction, claudication, stroke, previous vascular surgery, congenital disease, or arrhythmia) (n=2413 [28%]); or were unable to reach stage 2 of exercise (n=1620 [19%]). Before testing, participants underwent a detailed medical and family history that elicited information about medical conditions, alcohol use, smoking, education level, socioeconomic status, exercise habits, and medications. Lipid profiles were obtained in all patients. The treadmill exercise protocol used was the Bruce or modified Bruce protocol (5, 6). Participants exercised until they attained at least 85% to 90% of their age- and fitness-predicted maximum heart rate for 1 minute or until fatigue or medical contraindications to continued exercise were observed. Data on symptoms, heart rate, and blood pressure were collected and electrocardiography was done before exercise, at the end of each stage, immediately after exercise, and 2 minutes into recovery. Estimated workload, expressed in metabolic equivalents, was based on total treadmill time. Chronotropic response during exercise was assessed by using the chronotropic indexthe ratio of heart rate to metabolic reserve usedduring stage 2 of exercise (8, 9). We used a cutoff value (0.86) that maximized the log-rank chi-square statistic (10). Immediately after exercise, participants were helped to chairs, and recovery data were obtained. Heart rate recovery was defined as the change from peak heart rate to that measured after 2 minutes of recovery (that is, heart rate recovery=heart ratepeak heart rate2-min recovery). An abnormal value for heart rate recovery was determined by finding the maximum value for the log-rank chi-square test statistic for all possible cutoff points between the 10th and 90th percentiles (10); this turned out to be 42 beats/min or less. This cutoff value differed from and could not be validly compared to that of our previous study (1) because 2-minute values (rather than 1-minute values) were used and because exercise was submaximal, not symptom-limited. We also considered indexing the 2-minute heart rate recovery by the maximum heart rate, but this resulted in a substantially lower maximum log-rank chi-square value (44 compared with 77). Mean follow-up time was 12 years. The primary end point was all-cause mortality, which was determined through annual telephone interviews with study participants or their families or employers. Cause of death was assessed by review of death certificates and interviews of physicians or next of kin (6). If the patient could not be contacted, registries were searched for mortality information. For assessment of vital status, follow-up was 100% complete. All analyses were performed by using version 6.12 of the SAS statistical package (SAS Institute, Inc., Cary, North Carolina). For descriptive purposes, patients were divided into two groups based on heart rate recovery. Continuous variables are presented as the mean SD. Differences between groups were compared by using the Student t-test, Wilcoxon rank-sum test, and chi-square test, as appropriate. Heart rate recovery was related to all-cause mortality by using univariable and multivariable Cox regression analyses. Stratified analyses were performed on prespecified subgroups according to age, sex, exercise patterns, smoking, cholesterol level, resting blood pressure and heart rate, chronotropic response to exercise, and medication use. Logarithmic and quadratic transformations and potential interactions were assessed for improvement of fit; the strata option of PROC PHREG was used to allow for accurate estimation of main effects relative risks. The Cox proportional-hazards assumption was confirmed by inspection of log (log [survival]) curves. Attributable risk was calculated as P (RR 1)/[P (RR 1) + 1], where P is prevalence and RR is relative risk. The estimated actuarial survival rate for the population was estimated on the basis of data from 1985 U.S. life tables (11). According to the policy at our institution, this research was considered exempt from institutional review board approval because it used existing, publically available data. In addition, the data were recorded in such a manner that participants could not be identified directly or through identifiers linked to them. Results Baseline and Exercise Characteristics A total of 5234 adults met all inclusion criteria. The median value for heart rate recovery was 49 beats/min (25th and 75th percentiles, 39 and 59 beats/min). An abnormal heart rate recovery of 42 beats/min or less was seen in 1715 participants (33%). Baseline and exercise characteristics of the study participants according to heart rate recovery are summarized in Table 1. Table 1. Baseline and Exercise Characteristics according to Heart Rate Recovery after Exercise Heart Rate Recovery and Mortality During 12 years of follow-up, 325 patients (6.2%) died. According to U.S. life tables, the actuarial predicted death rate was 7%. Abnormal heart rate recovery was strongly predictive of death (10% compared with 4% among participants with normal heart rate recovery; relative risk, 2.58 [95% CI, 2.06 to 3.20]) (P<0.001). The sensitivity, specificity, and positive and negative predictive values of an abnormal heart rate recovery for prediction of death over 12 years were 54%, 69%, 10%, and 96%, respectively. Of the participants who died, 116 (36%) were thought to have died of cardiovascular causes. An abnormal heart rate recovery was even more strongly predictive of cardiovascular death than of death in general (4% compared with 1%; relative risk, 3.06 [CI, 2.10 to 4.44]) (P<0.001). Results of analyses stratified by age, sex, chronotropic response to exercise, regular exercise, smoking, resting hemodynamics, cholesterol level, and use of vasodilator medication are shown in Table 2. Abnormal heart rate recovery was predictive of death in all subgroups except participants taking vasodilators. Table 2. Association of Abnormal Heart Rate Recovery with Mortality according to Prespecified Subgroups Multivariable Cox Regression Analyses After adjustment for age, sex, body mass index, ethnicity, resting systolic blood pressure, use of vasodilators, exercise habits, physical fitness, smoking, diabetes, lipid profiles, ST-segment response, resting heart rate, chronotropic index, and educational and socioeconomic status and after consideration of interaction terms, a slower decrease in heart rate remained predictive of death (adjusted relative risk, 1.55 [CI, 1.22 to 1.98]) (P<0.001 by chi-square test). The adjusted attributable risk for death related to abnormal heart rate recovery was 15% (CI, 7% to 24%). Abnormal heart rate recovery was also predictive of cardiovascular death after adjustment for potential confounders and interactions (adjusted relative risk, 1.95 [CI, 1.11 to 3.42]) (P=0.02). Discussion Heart rate recovery after submaximal exercise was a powerful predictor of mortality in a population-based cohort of adults without clinically evident cardiovascular disease, even after we adjusted for multiple, potentially confounding factors. These findings confirm the results of our previous study (1) and expand them by demonstrating the prognostic importance of heart rate recovery in healthy persons undergoing submaximal, as opposed to symptom-limited, exercise testing. Heart rate recovery may therefore be a clinically relevant predictor of risk among patients undergoing screening exercise testing. Furthermore, because an abnormal heart rate recovery accounted for 15% of deaths, this measure may be useful for insurance underwriting assessments. It has been suggested that the link between heart rate recovery and mortality may be related to vagal tone and physical fitness (12). We also noted an association between fitness levels and recovery heart rate; participants with abnormal heart rate recovery were less likely to exercise regularly and to participate in strenuous exercise (Table 1). It should be noted that heart rates d