This study was based on data from KAMIR, which is a nationwide multicenter prospective registry of patients with AMI in Korea [14]. The data were collected from each participating hospital and entered into the KAMIR database via a web-based electronic system. From November 2011 to November 2015, 13,104 consecutive patients with AMI were enrolled in KAMIR. Among the patients, those who died in the hospital (n = 503); who were not prescribed any β-blocker at discharge (n = 2,058); who were prescribed β-blockers other than carvedilol and nebivolol (vasodilating) as well as bisoprolol and metoprolol (conventional) at discharge (n = 85); who did not undergo PCI but had undergone coronary artery bypass graft, thrombolysis, or medical therapy alone (n = 1,376); who did not revisit the hospital after discharge (n = 95); and whose prescribed β-blocker was either changed to a different class of β-blockers (from vasodilating to conventional, or vice versa) or was discontinued during the 3-year follow-up (n = 1,718) were excluded. As a result, 5,835 patients were excluded and 7,269 patients were analyzed in this study. The participants were divided into two groups according to the type of prescribed β-blocker: vasodilating β-blocker group (carvedilol, n = 3,251; nebivolol, n = 357) and conventional β-blocker group (bisoprolol, n = 3,533; metoprolol, n = 128). As this study was performed on the basis of registry data, propensity score matching was conducted to adjust for differences in baseline characteristics; finally, 3,079 pairs of propensity score-matched patients were created (Fig. 1). Our study was conducted according to the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of all centers (Seoul National University Hospital; SNUH IRB No. 1603-041-747), and written informed consent was obtained from all patients.

Previous history of cardiovascular risk factors, medication at discharge were obtained. Diabetes mellitus was defined as previously diagnosed untreated diabetes mellitus or current treatment with oral hypoglycemic agents or insulin. Hypertension was defined as previously diagnosed untreated hypertension or current treatment with antihypertensive medications. Dyslipidemia was defined as previously diagnosed untreated dyslipidemia or current treatment with lipid-lowering agents. Chronic kidney disease was defined as estimated glomerular filtration rate less than 60 mL/min/1.73 m² at the time of admission.

PCI was performed according to the current procedural guideline. All patients received 300 mg loading dose of aspirin and 600 mg loading dose of clopidogrel, 60 mg loading dose of prasugrel or 180 mg loading dose of ticagrelor before or during PCI, unless they had previously received these antiplatelet drugs. During PCI, weight-adjusted unfractionated heparin was administered to patients to keep the activated clotting time between 250 and 350 seconds. There was no limitation on the selection of stent manufacturers, number of stents, use of glycoprotein IIb/IIIa receptor inhibitors, use of intravascular imaging device or post-dilatation after stent implantation. Pre- and post-PCI coronary flows were graded based on the thrombolysis in myocardial infarction (TIMI) grading system. Coronary lesions were classified according to the American College of Cardiology/American Heart Association coronary lesion classification system.

The primary outcome was a composite of cardiac death, recurrent non-fatal myocardial infarction (MI), or hospitalization for heart failure (HF) at 3 years. Secondary outcomes were each component of the primary outcome. All deaths were considered to be associated with a cardiac cause, unless a definite non-cardiac cause could be established. Recurrent MI was defined as recurrent symptoms of myocardial ischemia with new ischemic electrocardiographic changes compatible with MI or elevation of cardiac troponin values with at least one value above the 99th percentile upper reference limit. Hospitalization for HF was defined as re-hospitalization as a result of worsening HF requiring more intensive care than continuation of usual treatment in the outpatient department. Clinical follow-up was routinely performed 6, 12, 24, and 36 months from the index hospitalization by visiting hospitals and whenever any clinical event took place. The patient’s physician identified all clinical events and the principal investigator of each hospital confirmed them.

The results were expressed as mean ± standard deviation (SD) for continuous variables and as percentage for categorical variables. For continuous variables, Student’s t test was used to test the differences between the two treatment groups; for categorical variables, Pearson’s chi-square test was used to compare the distribution between the two treatment groups. Because the β-blocker strategy was not randomized, a propensity score was used to adjust for potential selection or predisposition bias. The propensity score was estimated using multiple logistic regression analysis, with all variables shown in Supplementary Table 1. On the basis of the propensity score, the patients were selected by 1:1 matching without replacement using a greedy algorithm and the nearest available pair matching methods. A caliper width of 0.1 SD of the logit of the estimated propensity score was used for matching. The covariate balance achieved by propensity score-matching was assessed by calculating the absolute standardized differences in covariates between the two groups. After propensity score matching, the study end-points were demonstrated using the Kaplan-Meier survival curves and compared using the log-rank test. The Cox proportional hazards model was used to estimate the hazard ratio (HR) and 95% confidence intervals (CI) for between-group comparison. To determine independent associations between the type of β-blocker and clinical outcomes, a multivariable Cox regression analysis was performed. In the multivariable Cox regression analysis, age, sex, hypertension, diabetes mellitus, dyslipidemia, smoking status (current or ex-smoker), chronic kidney disease, left main or left anterior descending artery as infarct-related artery, multivessel disease, left ventricular ejection fraction (LVEF), ST segment elevation MI (STEMI) or non-ST segment elevation MI (NSTEMI), and use of renin-angiotensin system blocker and statin were included as covariates, which were significant on univariable analysis and were generally considered clinically relevant. All analyses were two-tailed, and clinical significance was defined as p < 0.05. All statistical analyses were performed using the statistical package SPSS version 22.0 (IBM Co., Armonk, NY, USA), and R programming language version 3.13 (R Foundation for Statistical Computing, Vienna, Austria).

We calculated the HR for primary outcome in various subgroups in both propensity score-matched and whole population. No significant interactions between the use of vasodilating β-blockers and primary outcome were observed in most subgroups except the mode of presentation (Fig. 3 and Supplementary Fig. 2). Interestingly, the efficacy of vasodilating β-blockers was significantly prominent in STEMI patients than in NSTEMI patients. Of note, the benefit of vasodilating β-blockers was not affected by the presence of left ventricular dysfunction.