The analyses were based on data from the BleeMACS registry. BleeMACS is a retrospective, observational, multicenter cohort registry involving 15 401 consecutive patients. The inclusion and exclusion criteria, data collection, and variables have been described previously.17 Briefly, eligible patients were all consecutive adult patients (≥ 18 years) discharged with a definitive diagnosis of ACS, defined according to clinical guidelines1,2 as evidence of significant coronary artery disease on coronary angiography (stenosis ≥ 50% in left main coronary artery or ≥ 70% in the rest of the coronary arteries) and who underwent in-hospital PCI, with follow-up data for 1 year. Participants were recruited from 15 hospitals from North and South America (Canada and Brazil), Europe (Germany, Poland, the Netherlands, Spain, Italy, and Greece), and Asia (China and Japan). Enrollment was conducted from November 2003 through June 2014. More information about the BleeMACS design and study population is shown in the supplementary data (Information about BleeMACS registry). The study protocol conforms to the ethics guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval of the ethics committee of each center. The study was registered in ClincalTrial.gov (NCT02466854).

The primary outcome was all-cause mortality, with comparison of patients treated with ACEI/ARB at discharge vs those not treated. The prescription of ACEI/ARB was based on the clinical judgment of the treating physician. All patients were systematically followed up for 1 year to assess vital status, ascertained by trained research coordinators at each participating site. Data on vital status were obtained from hospital and/or administrative data records, and/or by contacting the patients or their relatives by telephone, when necessary.

The statistical analysis was performed with SPSS version 24.0, R version 3.2.2, and Stata MP64 version 14. Baseline characteristics according to treatment with ACEI/ARB were described by using number and percentage for categorical data and standardized mean difference for continuous data, respectively. Differences in characteristics were assessed by using chi-square tests and 2-sample Student t tests. The association between ACEI/ARB exposure and 1-year death was studied in robust Cox proportional hazards models with adjustment for potential confounders at baseline. The multivariable model risk adjustment was performed with all variables associated with postdischarge mortality based on clinical plausibility or P value <.05 in the univariate Cox analyses (univariate cox analysis and Table 1of the supplementary data). Because of important differences in key baseline characteristics depending on prescription of ACEI/ARB therapy (Table 1), we complemented this analysis using a propensity score (PS) analysis. PS were estimated using a nonparsimonious multivariable logistic regression model, with ACEI/ARB therapy as the dependent variables and those characteristics that differed between patients treated and not treated with ACEI/ARB (Table 1) as covariates. Subsequent PS matching was performed to assemble a cohort in which all the measured covariates would be well balanced across the comparator group (additional details are presented in propensity score analysis, Table 2 and Figures 2-4of the supplementary data). In the PS-matched population (constituted by 2 groups of 3 765 patients with similar characteristics according to prescription of ACEI or not), the reduction in death rate was compared using a robust stratified Cox regression model. Survival-time inverse probability weighting propensity score analysis (IPW) was also used to evaluate the association between ACEI/ARB use and mortality. The effect of ACEI/ARB was graphically represented in Kaplan-Meier curves adjusted by IPW, to balance the covariate distribution between the treatment and control observations, and by those covariates associated with postdischarge mortality in the univariate Cox analyses, to further mitigate residual confounding in the survival modeling. We complemented the analysis with an augmented IPW (AIPW) to estimate the average treatment effects (ATEs), in a doubly robust method that combined both the properties of the regression-based estimator and the IPW estimator. The ATE coefficient assesses the absolute risk difference in 1-year death rate between patients treated and not treated with ACEI/ARB. Given that propensity scoring adjusts for measured confounding only, an instrumental variable analysis with annual rates of each hospital for the prescription of guideline-indicated treatments (DAPT, beta-blockers, statins, and ACEI/ARB) as the instrument was used to further assess potential selection bias (more information in instrumental variable analysis and Table 3 of the supplementary data). The coefficient of instrumental variable analysis shows the relative risk reduction in death rate with ACEI/ARB. Analyses were undertaken for the overall ACS cohort and separately for cases with LVEF ≤ 40% and> 40% (the cutoff for LVEF of 40% was based on European guideline recommendations for acute coronary syndrome1,2). In the group of patients with LVEF> 40%, we separately analyzed patients with ST-segment elevation myocardial infarction (STEMI) and high-risk conditions (HF, renal failure, diabetes mellitus [DM], arterial hypertension) (Figure 1). These high-risk conditions were defined based on recommendations of European and American guidelines for treatment of ACS.1,2 An interaction test was carried out as part of the Cox regression models performed in the study to assess the treatment-by-subgroup interaction. A P-value <.05 was considered statistically significant.

Figure 2.

Unadjusted impact of ACEI or ARB at discharge oin mortality reduction according to LVEF. 95%CI, 95% confidence interval; ACEI, angiotensin- converting enzyme inhibitors; ARB, angiotensin receptor blockers; LVEF, left ventricular ejection fraction.

(0.17MB).

Figure 3.

Adjusted survival Kaplan-Meier curves for the prescription of ACEI or ARB at discharge according to LVEF (left ventricular ejection fraction > 40% or ≤ 40%). 95%CI, 95% confidence interval; ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers; LVEF, left ventricular ejection fraction. Numbers at risk. In blue patients with left ventricular ejection fraction > 40%, and in red patients with LVEF ≤ 40%. In the box with green dashed line, patients treated with ACEI/ARB.

(0.3MB).

Figure 4.

Adjusted survival Kaplan-Meier curves for the prescription of ACEI or ARB at discharge in patients with LVEF > 40%, according to: A) presence of STEMI or NSTEACS; B) high-risk conditions (heart failure, renal failure, diabetes mellitus, hypertension). 95%CI, 95% confidence interval; ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers; HR, hazard ratio; LVEF, left ventricular ejection fraction; NSTEACS, non–ST-segment elevation acute coronary syndrome; STEMI, ST-segment elevation myocardial infarction.

(0.64MB).

Figure 1.

Study population according to LVEF (> 40% or ≤ 40%), type of ACS (STEMI or NSTEACS), and high-risk conditions (heart failure, renal failure, diabetes mellitus, hypertension). ACS, acute coronary syndrome; LVEF, left ventricular ejection fraction; NSTEACS, non–ST-segment elevation ACS; STEMI, ST-segment elevation myocardial infarction.

(0.23MB).

Missing data were handled by multiple imputations using the fully conditional specification method (an iterative Markov Chain Monte Carlo algorithm), generating 10 imputed datasets using all applicable adjustment variables and all outcome variables as predictors (supplementary data, Multiple Imputation). The effect of ACEI/ARB therapy on the mortality rate was computed separately in the 10 imputed data sets and then averaged over data sets using Rubin's combination rules (Imputed Datasets Analysis and Tables 4-13 of the supplementary data). Several sensitivity analyses were performed (additional details in Complete Case Analysis, Tables 14-16 and Figures 5-10 of the supplementary data).

ACEI/ARB were prescribed in 75.2% (n=11 581) of the 15 401 patients. There were significant differences in baseline characteristics between patients with and without ACEI/ARB (Table 1). In particular, patients who received ACEI/ARB tended to be at higher ischemic risk (including diabetes, hypertension, and dislipemia) compared with those not receiving ACEI/ARB. Patients treated with ACEI/ARB had worse renal function, but higher LVEF values, than patients not treated with ACEI/ARB, and the concomitant use of DAPT, beta-blockers, and statins was more frequent with ACEI/ARB prescription.

Of the entire cohort (N=15 401), there were 569 deaths (3.7%) during the first year after hospital discharge. Unadjusted 1-year mortality was significantly lower in patients who received ACEI/ARB compared with those who did not (3.2% vs 5.1%, P <.001). After weighting and adjustment, using multivariable Cox analysis adjustment for those variables associated with mortality in the univariate analysis (Table 1 of the supplementary data), and for PS techniques (PS matching, IPW, and AIPW), treatment with ACEI/ARB continued to be associated with lower 1-year mortality (Table 2). Specifically, there was a significant absolute risk difference of 0.8% in 1-year postdischarge mortality between patients treated and not treated with ACEI/ARB, and a significant relative risk reduction of 1-year mortality of 23.4%, according to the AIPW regression adjustment and instrumental variable analysis, respectively (Table 2).

Of the 2316 patients (15.0%) with LVEF ≤ 40%, 1861 (80.4%) were treated with ACEI/ARB. Of the 13 085 patients (85.0%) with LVEF> 40%, 9720 (74.3%) were treated with ACEI/ARB. One-year mortality was 6.9% (n=161) and 3.1% (n=409) in patients with and without LVEF ≤ 40%, respectively. Therapy with ACEI/ARB was associated with a higher mortality reduction with lower LVEF (Figure 2). ACEI/ARB was strongly associated with lower mortality in patients with LVEF ≤ 40% after weighting and adjustment for the different methods (Table 2, Figure 3), with an absolute and relative mortality reduction of 2.6% (P value for ATE according to AIPW regression adjustment=.076) and 46.1% (P value for instrumental variable analysis <.001), respectively. For patients with LVEF> 40%, ACEI/ARB therapy was not associated with lower mortality in the various Cox analyses (Table 2, Figure 3), or with the AIPW regression adjustment (relative risk reduction of 0.5%, 95%CI,−1.2% to 0.1%). However, the coefficient of instrumental variable analysis showed a significant relative risk reduction of 15.7% (95%CI,−26.4% to−4.9%). The interaction P value for treatment-by-LVEF was significant (.008), indicating that the clinical benefit of ACEI/ARB was concentrated in patients with LVEF ≤ 40%. These results (from the imputed data cohort) were consistent with those observed in the case-complete cohort (Tables 14-16 and Figure 5-10 of the supplementary data).

Benefit of ACEI/ARB in STEMI and high-risk patients with LVEF> 40%

In the group of patients with LVEF> 40% (n=13,085), 7477 (57.1%) had STEMI (Figure 3). In these patients with STEMI, a specific robust Cox analysis, adjusted by IPW and by those variables associated with mortality in the univariate analysis, showed a lower 1-year mortality in patients treated with ACEI/ARB vs those not treated with ACEI/ARB (HR 0.44; 95%CI, 0.21-0.93; P=.031), which was not observed in patients with non–ST-segment elevation ACS (NSTEACS) (Figure 4A). The interaction P value for treatment-by-ACS type was significant (.001), indicating that the clinical benefit of ACEI/ARB in patients with LVEF> 40% was observed only in patients with STEMI (not in those with NSTEACS). Of the total number of patients with LVEF> 40%, 1678 (12.8%) had a history of HFdefined as prior admission for HF, and/or Killip class> I at ACS admission, and/or de novo HF during ACS hospitalization, 1991 (15.2%) had renal failure, defined as estimated glomerular filtrate rate by isotope dilution mass spectrometry (IDMS)—traceable Modification of the Diet in Renal Disease (MDRD) formula <60 mL/min/1.73 m2 at admission—, 3011 (20.0%) had DM, and 7627 (50.6%) had arterial hypertension. Of the subgroup of patients with LVEF> 40%, 9215 patients (61.1%) had at least 1 of these high-risk conditions. In this high-risk group, treatment with ACEI/ARB showed a trend to lower 1-year mortality (HR, 0.76; 95%CI, 0.55-1.05; P=.092) in comparison with the nonhigh-risk group (Figure 4B). However, the interaction P value for treatment-by-risk conditions was not significant (0.238), indicating that there are no differences in the clinical effect of ACEI/ARB in patients with LVEF> 40% by the presence or absence of high-risk conditions. The adjusted Kaplan-Meier curves for each high-risk condition are shown in Figures 11-14 of the supplementary data, together with the case-complete analysis (Tables 14-16 and Figures 5-10 of the supplementary data).

This study has some limitations. First, only patients who survived hospital stay were studied and, consequently, we did not investigate the role of in-hospital ACEI/ARB. Several investigators in the thrombolysis era have shown that the benefit of ACEI/ARB occurs during the first few days after ACS, suggesting that mechanisms other than benefits on the remodeling process may play a role.3,5,7 Second, the survival benefits of ACEI/ARB were compared based on medications at discharge. In addition, the prescription dose, long-term adherence, discontinuation, incidence of adverse events, and drug information after discharge were not available in the present study. Moreover, we do not have data about prior use of ACEI/ARB before hospital admission. Third, we did not differentiate between ACEI and ARB. Although several authors have shown that both ACEI and ARB show an equal benefit, other authors have reported marked differences between the 2 drugs,26,27 showing different long-term tolerance and adherence.28 Fourth, our study included a selected and nonrandomized sample; in addition, although propensity scoring and instrumental variable analysis were adjusted for confounding by indication, and further adjustments were made for many additional confounders in the survival models, residual confounding is probable.