Resting heart rate (HR) is an easy to measure, readily available, inexpensive physiological parameter measurable in all clinical conditions. Decades of research have evaluated the prognostic value of resting HR in different clinical conditions along with its role as a cardiovascular and metabolic risk factor. Its relevance has been confirmed both among healthy subjects and in patients with known cardiovascular diseases such as hypertension, metabolic syndrome, and diabetes.1 The deleterious effect of increased HR has been also recognized as a factor known to promote the development and progression of coronary artery disease.2

In patients with acute chest pain and clinical suspicion of an acute coronary syndrome (ACS), HR at admission is the very first clinical sign recorded at presentation, often before any other clinical or laboratory parameter such as blood pressure, ECG, or cardiac biomarkers.

Due to its easy, fast, and reliable assessment, its obvious clinical usefulness and the continuous nature of this variable, HR has been recognized as one of a few factors able to stratify prognosis of ACS patients at admission and is thus included in several clinical risk prediction scores such as PURSUIT and GRACE.3,4 The importance of these scores in clinical practice has been stressed by current European guidelines suggesting that, beside clinical assessment, use of the GRACE risk score is able to predict short- and long-term outcomes in ACS patients.5

While GRACE adopted an incremental model to account for the detrimental prognostic impact of HR in ACS with a progressive increase in the risk of in-hospital and 6-month mortality for HR only above 60 beats per minute, assuming that patients with HR <50 bpm have the same risk as those with a HR of 60 bpm,4,6 some evidence suggested a nonlinear, relationship between HR and event rates, with high event rates evident also for low HRs.

In particular, numerous retrospective studies describing the relationship between adverse events and admission HR reported the pattern known as the “J-shaped curve”, with higher rates at both extremities of HR distribution (ie, very low and high HR).7–11 This finding has also been confirmed by a large post hoc analysis of the CRUSADE quality improvement initiative including data from 135 164 patients.12

Nonetheless, the association between bradycardia and an increased likelihood of short-term events remains elusive and potentially biased. This is because, even in large ACS cohorts, a definite association could not be ultimately confirmed due to the relatively small sample size of patients presenting with low HR at admission,7,9,11 the lack of relevant data on associated atrioventricular conduction disorders,11 and the presence of critical baseline differences in the absence of adequate corrections adopted to balance the effect of clinical characteristics with renewed prognostic impact among different HR groups.

Thus, the aim of our study was to evaluate the association between bradycardia at admission and clinical outcomes in myocardial infarction (MI) in a large cohort of contemporary patients enrolled in the multicenter, nationwide Acute Myocardial Infarction in Switzerland (AMIS) Plus registry.

Out of 59 354 patients enrolled in the registry in the predefined study period, 51 001 (85%), with a median age of 66 years [interquartile range, 56-76 years], fulfilled study inclusion criteria. Figure 1 shows patients’ flow. Baseline clinical characteristics stratified according to HR at admission are reported in table 1. Table 1 of the supplementary data reports the acute treatments offered at admission.

Figure 1.

Patient flowchart for study inclusion. AV, atrioventricular; HR, heart rate.

(0.21MB).

HR groups differed in several clinically relevant baseline characteristics with a known impact on short-term adverse event rates, including rates of resuscitated OHCA, cardiogenic shock, and presentation with STEMI vs NSTEMI.

Overall, 2525 in-hospital deaths were observed, leading to a crude in-hospital mortality of 5.0%. Crude estimates of in-hospital mortality were significantly different among HR groups, replicating the so-called “J shape curve”, with a bimodal distribution of events showing higher rates at very low and high admission HR.

In patients with HR <50 bpm, 66 (5.7%) deaths were observed (reference), while 642 (2.8%; OR, 0.47; 95%CI, 0.37-0.62; P <.001) were observed in patients with HR 50 to 75 bpm, 1053 (5.0%) in patients with HR, 76 to 100 bpm (OR, 0.87; 95%CI, 0.67-1.13; P=.290), and 764 (13.7%) in patients with HR> 100 bpm (OR, 2.62; 95%CI, 2.04-3.43; P <.001).

Overall, the secondary composite endpoint occurred in 3157 patients, with significant differences throughout the HR distribution (P <.001 among groups). Rates of the individual components of the composite endpoint showed significant differences among HR groups. Mortality and MACCE rates along with associated OR stratified by HR group are reported in figure 2. table 2 of the supplementary data reports additional details regarding in-hospital outcomes.

Figure 2.

Crude estimates of mortality (upper panel) and MACCE (lower panel) rates (bars) and associated odds ratios (whisker plots) among groups stratified by HR at admission. HR, heart rate; MACCE, major adverse cardiovascular and cerebrovascular events.

(0.15MB).

Clinical correlates of in-hospital mortality

After multivariable logistic regression analysis, admission HR 50 to 75 was recognized as independently associated with lower in-hospital mortality compared with HR <50bpm (OR, 0.67; 95%CI, 0.47-0.99 P=.038) while HR> 100bpm was associated with a further increase in mortality risk (OR, 2.13; 95%CI, 1.48-3.15 P <.001). Additional independent predictors of increased in-hospital mortality were age, female sex, impaired renal function, OHCA, cardiogenic shock at admission, STEMI presentation and chronic therapy with either beta-blockers or calcium channel blockers. Table 2 reports the results of the primary multivariable regression analysis.

Table 2.

Multivariable logistic regression analysis: independent clinical parameters associated with in-hospital mortality.

Variable	OR	95%CI	P
Heart rate group
<50	—	—
50-75	0.67	0.47-0.99	.038
76-100	1.18	0.83-1.73	.37
> 100	2.13	1.48-3.15	<.001
Age (per year increase)	1.08	1.07-1.09	<.001
Female sex	1.28	1.13-1.46	<.001
Creatinine <83μmol/L	0.41	0.35-0.47	<.001
Out-of-hospital cardiac arrest	4.56	3.76-5.51	<.001
Cardiogenic shock at admission	7.48	6.25-8.96	<.001
ACS subtype
STEMI	—	—
NSTEMI	0.71	0.62-0.80	<.001
Chronic therapy with CCB	1.18	1.02-1.36	.026
Chronic therapy with beta-blockers	1.19	1.06-1.35	.004

95%CI, 95% confidence interval; ACS, acute coronary syndrome; CCB, calcium channel blockers; NSTEMI, non–ST-elevation myocardial infarction; OR, odds ratio; STEMI, ST-elevation myocardial infarction.

Nonetheless, multiple sensitivity multivariable logistic regression analyses performed after exclusion of patients on regular therapy with either beta-blockers or calcium channel blockers or adopting a random effect for study center did not confirm the protective effect of HR 50 to 75 bpm, which was borderline significant after exclusion of patients with atrial fibrillation for in-hospital mortality (OR, 0.66; 95%CI, 0.45-1.00; P=.047), while the role of all remaining clinical predictors identified on the primary multivariable analysis was confirmed, including HR> 100 bpm (tables 3-7 of the supplementary data).

The restricted cubic spline analysis, adjusted for age, sex, creatinine <83μmol/L, OHCA, cardiogenic shock at admission, ACS type, and chronic therapy with beta-blockers or calcium channel blockers, revealed a nonlinear, complex relationship between HR and in-hospital mortality (figure 3). The P-value for the smoothed HR term was <.001.

Figure 3.

Restricted cubic spline analysis to model the relationship between heart rate and log-odds of in-hospital mortality, adjusted for age, sex, creatinine <83μmol/L, resuscitated out-of-hospital cardiac arrest, cardiogenic shock at admission, acute coronary syndrome subtype, and chronic therapy with beta-blockers or calcium channel blockers. Bar histogram shows the distribution of the heart rate values.

(0.21MB).

Propensity score matched analyses

Three sequential propensity score matched analyses were performed pairing patients with HR <50 bpm and those with HR 50 to 75 bpm, HR 76 to 100 bpm and HR> 100 bpm, identifying 1159, 1159 and 1158 matched pairs, respectively.

Figures 1-3 of the supplementary data report standardized mean differences before and after propensity score matching. After propensity score matching, no differences among rates of in-hospital mortality were evident for groups with admission HR <100 bpm.

In the matched groups, 66 fatal events (5.7%) were observed among patients with HR <50 bpm, 65 (5.6%) among those with HR 50 to 75 bpm (OR, 0.98; 95%CI, 0.69-1.40; P=.936), 78 (6.7%) among those with HR 76-100 bpm (OR, 1.19; 95%CI, 0.85-1.68; P=.302), and 125 (10.7%) in group HR> 100 bpm (OR, 2.03; 95%CI, 1.49-2.78; P <.001).

Eighty-three (7.3%) MACCE were recorded among matched patients with HR <50 bpm, 78 (6.8%) in the group with HR 50 to 75 bpm (OR, 0.93; 95%CI, 0.67-1.28; P=.683), 96 (8.4%) in group with HR 76 to 100 bpm (OR, 1.17; 95%CI, 0.86-1.58; P=.312), and 147 (13.0%) in those with HR> 100 bpm (OR, 1.90; 95%CI, 1.43-2.52; P <.001). Mortality and MACCE rates along with associated ORs stratified by matched HR group are reported in figure 4. Additional details on outcomes observed in the matched groups are available in table 8 of the supplementary data.

Figure 4.

Propensity score matched analysis: in-hospital mortality rates and MACCE rates (bars) and associated odds ratios (whisker plots) among matched groups stratified by HR at admission. HR, heart rate; MACCE, major adverse cardiovascular and cerebrovascular events.

(0.18MB).

The availability of a large dataset of patients with MI and a high level of data granularity allowed us to investigate the prognostic role of bradycardia at admission, a parameter which was hypothesized to be associated per se with a deleterious prognostic impact in MI.

In line with reports from several previous registry-based analyses, we observed higher crude adverse event rates in patients with admission HR <50 bpm and> 76 bpm compared with patients with HR 50 to 75 bpm, replicating a bimodal distribution of events, also referred as the “J curve relationship”.7,9–12

However, the restricted cubic spline analysis highlighted a more complex relationship between admission HR and in-hospital mortality. According to our data, while this relationship cannot be oversimplified as a “J-shaped” curve, a certain degree of linearity was observed in patients with HR between 50 to 100 bpm, who account for the vast majority of real-life ACS patients. Interestingly, even when we analyzed HR as a continuous variable, a certain protective effect of HR roughly comprised from 50 to 75 bpm can be claimed, confirmed by negative log-odds ratio estimates for in-hospital mortality.

While the ominous impact of tachycardia (HR> 100 bpm) was strongly confirmed across multiple sensitivity analyses performed among our patients, the signal supporting an independent prognostic role for bradycardia became weaker when we adjusted for multiple clinically relevant covariates. Multiple logistic regression analyses performed after sequential inclusion/exclusion of patients on current medical therapy with drugs with a negative chronotropic effect, such as beta-blockers and calcium channel blockers, did not confirm the independent association between HR <50 bpm and in-hospital mortality, while remaining borderline significant after exclusion of patients with atrial fibrillation.

When we tested analogous groups matched for relevant clinical variables such as age, sex, the presence of renal impairment, ongoing medical therapy, and clinical presentations (OHCA, cardiogenic shock at admission and STEMI), rates of in-hospital events became equal among groups with HR <100 bpm. A progressive fading of any independent prognostic role attributable to HR <50 bpm, along with the “J curve relationship” between admission HR and both primary and secondary endpoints was thus evident after adequate correction of prognostically relevant baseline clinical differences by means of propensity score matching (figure 5).

Figure 5.

Central illustration. Overview of a multilevel method with 3 statistical approaches to assess the prognostic role of low HR in ACS including crude, restricted cubic splines and propensity-matched analyses. AMI, acute myocardial infarction; HR, heart rate; MACCE, major cardiovascular and cerebrovascular events.

(0.42MB).

The direct incremental relationship between HR and adverse events in patients with ACS adopted in the GRACE risk prediction model (which estimates a stepwise increase in the risk of events of 30% for every 30 bpm increase in HR)4–6 was originally challenged by Parodi et al.9 in 2010 while assessing the prognostic impact of HR at admission in 2477 STEMI patients treated with primary percutaneous coronary intervention between 1997 and 2007. After exclusion of patients presenting with AV block and atrial fibrillation, patients with HR <60 bpm (n=388) showed higher in-hospital and 6-month mortality rates compared with patients with HR 60 to 79 bpm. Nonetheless, the relationship between bradycardia and mortality could not be definitely confirmed, as bradycardia was not recognized as an independent predictor of mortality.

Following this initial evidence, Bangalore et al.12 published in 2010 an analysis on 135 164 NSTEMI patients enrolled in the early 2000s in the CRUSADE quality improvement initiative. The limited number of variables collected in CRUSADE did not allow the provision of relevant details regarding the presence of concomitant bradyarrhythmias, atrioventricular blocks, atrial or ventricular tachyarrhythmias or ongoing medical therapy. When stratified according to HR at presentation, significant differences emerged for all baseline characteristics. However, higher adjusted rates of the composite endpoint of all-cause mortality, nonfatal reinfarction and stroke were observed among admission HR <50 bpm compared with those with HR, 60 to 100 bpm (OR, 1.61; 95%CI, 1.23-2.10).

In 2016, a further subanalysis derived from the EURHOBOP project on 10 374 patients with ACS admitted from 2008 to 2010 across 58 European hospitals described a “J-shaped relationship” between admission HR and mortality with the lowest and highest HR conferring the greatest risk.11 This observation was supported by the significant associations observed between HR <40 bpm and crude mortality estimates either in STEMI and NSTEMI patients (STEMI crude OR, 4.3; 95%CI, 1.4-13.2; NSTEMI crude OR, 11.0; 95%CI, 2.3-51.5). Nevertheless, adjusted analyses strongly mitigated this association (STEMI OR, 1.7; 95%CI, 0.5-6.1; NSTEMI OR, 4.7; 95%CI, 1.0-23.6). Also in this case, no details were provided on the presence of associated AV conduction disturbances or other arrhythmias.11

More recently, Perne et al.7 reported the outcomes observed in 6168 patients prospectively enrolled between 2008 and 2014 in the German Chest Pain Unit Registry and stratified according to 4 different HR groups, confirming the “J curve relationship” between admission HR and crude estimates of mortality and MACCE. Again, lack of relevant details regarding concomitant medical therapy at admission did not allow this hypothesis to be confirmed, leading the authors to propose that lower HR might not be causative for the worse outcomes, but rather serves as a marker of increased risk.

All the above-mentioned data are difficult to interpret due to the lack of relevant information on the concomitant presence of atrioventricular conduction disturbances, a factor with a known impact on short-term outcomes.19 In addition, the external validity of the observations was limited by the lack of details regarding history or preadmission use of drugs with a negative chronotropic effect, the low number of patients included in different HR categories, the low event rate in the lowest HR groups, the use of noncontemporary treatment regimens, and the absence of adequate statistical adjustment to balance baseline clinical and ACS related differences among different HR groups.

Thus, our analysis, which reports on a large contemporary collective of patients with ACS analyzed after appropriate statistical correction for baseline differences among different HR groups, strengthened by the results of all sensitivity analyses also evaluating HR as a continuous variable, provides clinically relevant information to solve the “J curve” dilemma.

Patients with HR <50 bpm at admission show higher crude in-hospital adverse event rates compared with those with HR between 50 and 100 bpm. Nonetheless, the signal supporting an independent association between bradycardia at admission and short-term mortality is weak and was not confirmed after correction for relevant baseline differences by propensity score matching. The nonlinear, complex association between HR and inhospital mortality in the restricted cubic spline analysis challenges the predictive accuracy of risk prediction scores based on linear models, commonly used in ACS.

The AMIS Plus registry is funded by unrestricted grants from the Swiss Heart Foundation and from Abbot AG, Amgen AG, AstraZeneca AG, Bayer (Schweiz) AG, Biotronik AG, Boston Scientific AG, B. Braun Medical AG, Daiichi-Sankyo/Lilly AG, Cordis Cardinal Health GmbH, Medtronic AG, Novartis Pharma Schweiz AG, Sanofi-Aventis (Schweiz) AG, SIS Medical AG, Terumo AG, Vascular Medical GmbH, all in Switzerland, and the Swiss Working Group for Interventional Cardiology. The sponsors played no role in the design, data collection, analysis, or interpretation of the data.

L. Biasco, P. Dittli and G. Pedrazzini designed the study and drafted the manuscript. F. Foster-Witassek and D. Radovanovic acquired and analyzed the data, performed the statistical analysis, and critically reviewed the manuscript. G. Tersalvi, H. Rickli, M. Roffi, F. Eberli, R. Jeger and P. Erne critically reviewed the manuscript. All authors have read and approved the submission of the paper.

We hereby declare that all authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation. None of the authors have any conflicts of interest to disclose.