Results
Population
The baseline characteristics of the study population are listed in Table 1. Table 2 shows the significant univariate associations of different measures of HR and baseline variables of interest. As shown all measures of HR are higher in women, smokers, diabetics, and subjects with a lower level of physical activity. In multiple regression analyses, resting HR was related to smoking and high-sensitive C-reactive protein, while 24 h HR was associated with sex, smoking, high-sensitive C-reactive protein, NT-proBNP, and SBP, and night-time HR was associated with sex, smoking, high-sensitive C-reactive protein, and low physical activity level.
Follow-up and Endpoints
The median follow-up was 76 months (inter-quartile range: 74–78). During the follow-up period, 80 participants died (22 from CV death, 26 from cancer, 15 of unknown causes, and 17 from other causes), and 54 had a CV event (AMI = 20, CV death = 22, revascularization = 26). Fourteen subjects had more than one event.
Event rates for quintiles of the three different measures of HR are illustrated in Figure 2. As shown there seems to be a threshold of risk at about 70 beats/min for resting HR and 24 h HR, but for night-time HR, it seems to be at about 60 beats/min.
(Enlarge Image)
Figure 2.
Events (all-cause mortality) per 1000 patient-years for each quintile of different measures of heart rate. Numbers at the bottom represent the median value of the heart rate of the corresponding quintile above.
Figure 2 Events (all-cause mortality) per 1000 patient-years for each quintile of different measures of heart rate. Numbers at the bottom represent the median value of the heart rate of the corresponding quintile above.
All-cause Mortality
All three measures of HR were significantly associated with all-cause mortality in sex- and age-adjusted Cox proportional hazard models ( Table 3 ). In multivariable analyses, adjusted for sex, age, and conventional risk factors (total cholesterol, diabetes, current smoking, and hypertension), the associations remained significant. Further adjustment for prognostic significant biomarkers (high-sensitive C-reactive protein and NT-proBNP) and use of β-blockers did not change the associations. Nor did additional adjustment for other relevant medication (ACE-inhibitors/angiotensin II receptor blockers, diuretics, and calcium channel blockers) ( Table 3 ). Using forward selection including all three HR measures, only night-time HR stayed in the model suggesting that this has the strongest correlation to all-cause mortality, hazard ratio = 1.07 (1.02–1.33), P = 0.02. When only resting HR and 24 h HR were tested in a forward selection model, only 24 h HR remained in the model. Figure 3 illustrates the event-free survival estimates of the subjects in quintiles of different measures of HR.
(Enlarge Image)
Figure 3.
Event-free survival estimates for the quintiles of the three different measures of heart rate.
Cardiovascular Events
All three measures of HR were associated with CV events both in univariate and in age and gender adjusted analysis (Table 3). However, when adjusting for CV risk factors, the association with resting HR and 24 h HR was no longer present. Yet, when further adjusting for biomarkers and β-blocker usage, the association with 24 h HR and resting HR reappeared (Table 3, Model 4). After additional adjustment for other relevant medication (ACE-inhibitors/angiotensin II receptor blockers, diuretics, and calcium channel blockers), all three measures of HR remained significantly associated with CV events (Table 3, Model 5). In a forward selection model with inclusion of all three measures of HR, conventional risk factors, biomarkers, and β-blockers usage, only night-time HR, among the three measures of HR, remained in the model, hazard ratio = 1.17 (1.05–1.30), P = 0.005. When only resting HR and 24 h HR were tested in a forward selection model, none remained in the model.
The median value of fluctuation from the mean night-time HR to maximum night-time HR was 32 (range: 11–86). These fluctuations were not associated with all-cause mortality but were inversely associated with CV events in a multivariate Cox-model: hazard ratio = 0.84, 95% confidence interval (CI): 0.72–0.98, P = 0.029 after adjustment for all conventional risk factors, high-sensitive C-reactive protein, NT-proBNP, and use of β-blockers.
Sensitivity Analysis
The night-time HR measured from 2:00 to 2:15 a.m. (65.9 ± 10.4) was closely correlated with night-time HR measured from 1 to 1:15 a.m. (67.6 ± 10.5), and from 3 to 3:15 a.m. (66.3 ± 10.5), respectively, with correlation coefficients of 0.90 and 0.92; P < 0.0001 for both associations. Multivariable adjusted hazard ratio and 95% CI for the association with CV events for night-time HR at 1 a.m. was 1.16 (1.01–1.35), P = 0.036, and at 3 a.m. 1.25 (0.98–1.30), P = 0.09, respectively. Very similar results were found for all-cause mortality.
Reanalysis of the data after exclusion of β-blocker users did not change the results (data not shown).
Collinearity Between HR Variables
Collinearity was found with Pearson correlation coefficients of 0.83 (between 24 h HR and night-time HR), 0.62 (between 24 h HR and resting HR), and 0.50 (between night-time HR and resting HR), all P < 0.0001.
Stratified by Risk Profile
To investigate whether the key findings were consistent in subjects with low-risk profile (0–1 risk factor) as well as in subjects with high-risk profile (≥2 risk factors), age and gender-adjusted Cox proportional hazard analyses were performed stratified by risk groups with all-cause mortality as endpoint. We found night-time HR to be predictive in both subjects with 0–1 risk factor (hazard ratio = 1.22, 95% CI: 1.04–1.43, P = 0.02) and in subjects with ≥2 risk factors (hazard ratio = 1.29, 95% CI: 1.14–1.46, P < 0.0001). Twenty-four-h HR was borderline significant in subjects with 0–1 risk factor (hazard ratio = 1.24, 95% CI: 1.01–1.53, P = 0.045) and significant in subjects with ≥2 risk factors (hazard ratio = 1.26, 95% CI: 1.08–1.48, P = 0.004). Resting HR was still a significant predictor in the subjects with 0–1 risk factor (hazard ratio = 1.03, 95% CI: 1.01–1.06, P = 0.01) but not in subjects with ≥2 risk factors (hazard ratio = 1.02, 95% CI: 0.99–1.04, P = 0.118).