β-Blocker Use and Clinical Outcomes in Stable Outpatients With and Without Coronary Artery Disease FREE

Treatment with β-blockers remains the standard of care for patients with coronary artery disease (CAD), especially when they have had a myocardial infarction (MI).^1- 2 The evidence is derived from relatively old post-MI studies, most of which antedate modern reperfusion or medical therapy, and from heart failure trials, but has been widely extrapolated to patients with CAD and even to patients at high risk for but without established CAD. It is not known if these extrapolations are justified. Moreover, the long-term efficacy of these agents in patients treated with contemporary medical therapies is not known, even in patients with prior MI.

β-Blockers are not without adverse effects and their tolerability is not ideal. Among 17 035 patients with MI, only 45% of patients were adherent to β-blocker use 1-year after an MI.³ The objective of the present longitudinal, observational study was to evaluate the differential association of β-blocker use on long-term cardiovascular outcomes in patients with known prior MI, in patients with known CAD without MI, and in patients with only risk factors for CAD.

The design, methods, and results of the Reduction of Atherothrombosis for Continued Health (REACH) registry, an international, prospective, observational registry, have been published elsewhere.^4- 7 Briefly, REACH enrolled consecutive eligible patients aged 45 years or older with established CAD, cerebrovascular disease, or peripheral arterial disease, or with at least 3 atherothrombotic risk factors during a 7-month recruitment period between December 2003 and June 2004; because of regulatory requirements in Japan, the enrollment in that country was delayed and occurred between August 2004 and December 2004. The last patient was enrolled in December 2004 and the date of final data collection was April 2009. Signed informed consent was obtained from all patients and the institutional review board in each country approved the protocol. The study participants were from 7 geographical regions.

Patients with data on β-blocker use were divided into 3 groups: known prior MI, CAD without known MI, and CAD risk factors only. Each of these groups was divided into 2 subgroups based on β-blocker use at the time of enrollment. The CAD cohort was identified by documented history of percutaneous coronary intervention, coronary artery bypass graft surgery, or ischemia but without a known history of MI. Data were collected centrally using standardized case report forms. Patients were followed-up prospectively for up to 4 years for the occurrence of cardiovascular outcomes, hospitalization, or vascular interventions. The primary outcome was a composite of cardiovascular death, nonfatal MI, or nonfatal stroke. The secondary outcome was the primary outcome plus hospitalization for atherothrombotic events or a revascularization procedure (coronary, cerebral, or peripheral). Tertiary outcomes were all-cause mortality, cardiovascular mortality, nonfatal MI, nonfatal stroke, and hospitalization, which were considered as separate outcomes.

All analyses were performed using SAS software version 9.2 (SAS Institute Inc). Power calculations were performed using accrual over 7 months and a median follow-up of 44 months. For the known prior MI cohort, which had 14 000 patients and 30% without β-blocker use, there was 80% power to detect a hazard ratio (HR) of 0.87. Similarly, for the known CAD without MI cohort, which had 12 000 patients and 40% without β-blocker use, there was 80% power to detect an HR of 0.87 to show benefit and 1.14 to show harm. For the cohort with only risk factors for CAD, which had 19 000 patients and 75% without β-blocker use, there was 80% power to detect an HR of 0.89 to show benefit and 1.13 to show harm.

The analysis was based on the intention-to-treat principle regardless of subsequent β-blocker use. Because of differences in key baseline characteristics (Table 1, Table 2, and Table 3), we used propensity score matching for the 3 cohorts to assemble a cohort for each comparison, in which all the measured covariates would be well balanced across comparator groups. The propensity score is the conditional probability of having a particular exposure (β-blocker use) given a set of measured baseline covariates.^8- 9 Propensity scores were estimated using a non–parsimonious multivariable logistic regression model,¹⁰ with the dependent variable of β-blocker use, and the 27 baseline characteristics entered as covariates. Matching was performed using a SAS macro with a greedy matching protocol (matching ratio of 1 to 1 without replacement) and a caliper width of 0.6 of the standard deviation. We estimated standardized differences for all covariates before and after matching to assess prematch imbalance and postmatch balance.¹¹ Standardized differences of less than 10% for a given covariate indicate a relatively small imbalance.¹¹

Paired comparisons were performed using conditional logistic regression analysis for categorical variables and paired t test for continuous variables. The risk of outcomes in the group with β-blocker use vs without β-blocker use was estimated using a Cox proportional hazard regression model stratified on the matched pairs. The analyses were exploratory in nature so no P value adjustment for multiple testing was applied.

In the propensity score–matched analysis, many patients remained unmatched and were thus excluded from the analysis (which may lead to slightly reduced efficiency). Therefore, a regression adjustment with the propensity score (as a continuous variable) was also performed,¹² in which all the patients in the cohort were analyzed.

For both the propensity score–matched and propensity score–adjusted analyses, the proportional hazards assumptions were violated for hospitalization-related outcomes because the exact dates of hospitalization were not available. For these outcomes, odds ratios (ORs) are reported rather than HRs. The proportional hazards assumption was tested using Schoenfeld residuals, which help determine nonproportionality over time either graphically or by testing for a nonzero slope in a regression model of residuals on time.

The differential association of β-blocker use across the 3 cohorts was tested using a test for interaction with a P value of less than .10 considered statistically significant. All tests were 2-tailed and (aside from the test for interaction) a P value of less than .05 was considered statistically significant.

Given the well-known, beneficial association of β-blocker use with reduction in morbidity and mortality among patients with heart failure, a sensitivity analysis was conducted after excluding patients with known heart failure. In addition, analyses were performed using β-blocker use as a time-dependent covariate that incorporates changes in β-blocker use over time. In addition, analyses based on inverse probability weighting were performed.¹³ For this purpose, a propensity score weight, also referred to as the inverse probability of treatment weight, was calculated as the inverse of the propensity score. Analyses were conducted to test for internal validity of the data set using one cohort with recent MI (≤1 year) and another with known heart failure, both of which have shown utility of β-blocker use in prior studies.

From the REACH registry, 44 708 patients satisfied the inclusion criteria of whom 14 043 patients (31%) had prior MI, 12 012 patients (27%) had documented CAD but without MI, and 18 653 patients (42%) had CAD risk factors only. In the included cohort, 96% of patients had 2-year follow-up and 74% of the cohort had 4-year follow-up. The overall median follow-up was 44 months (interquartile range [IQR], 35-45 months). Among the 44 708 patients, 21 860 were included in the propensity score–matched analysis.

In the prior MI cohort, there were 9451 patients (67%) with β-blocker use. There were significant differences between patients with β-blocker use and those without β-blocker use in the baseline characteristics prior to matching (Table 1). Propensity score matching matched 3379 patients with β-blocker use (36% of the cohort) with 3379 patients without β-blocker use (74% of the cohort) with similar propensity scores. After propensity score matching, there were no significant differences in baseline variables (Table 1), with absolute standardized differences of less than 10% for all variables, indicating an adequate match (eFigure 1). The median follow-up was 43 months (IQR, 30-45 months) in the group with β-blocker use vs 43 months (IQR, 29-45 months) in the group without β-blocker use.

The event rates were not significantly different in those with β-blocker use (489 [16.93%]) vs those without β-blocker use (532 [18.60%]) for the primary outcome (HR, 0.90 [95% CI, 0.79-1.03]; P = .14; Figure 1), the secondary outcome (30.96% vs 33.12%, respectively; OR, 0.91 [95% CI, 0.82-1.00]; Figure 2), or any of the tertiary outcomes (Figure 2; eFigures 2-4) of cardiovascular death (9.68% vs 10.27%; P = .31), MI (5.50% vs 5.51%; P = .42), and stroke (4.44% vs 5.21%; P = .28) in the matched cohort. The results were similar in the propensity score–adjusted model (eFigure 5).

In the CAD without MI cohort, there were 6864 patients (57%) with β-blocker use. There were significant differences between patients with β-blocker use vs those without β-blocker use in the baseline characteristics prior to matching (Table 2). Propensity score matching matched 3599 patients with β-blocker use (52% of the cohort) with 3599 patients without β-blocker use (70% of the cohort) with similar propensity scores. After propensity score matching (Table 2), the absolute standardized differences were less than 10% for all matched variables, indicating an adequate match (eFigure 6). The median follow-up was 43 months (IQR, 31-45 months) in the group with β-blocker use vs 43 months (IQR, 31-45 months) in the group without β-blocker use.

The event rates were not different in those with β-blocker use (391 [12.94%]) vs those without β-blocker use (405 [13.55%]) for the primary outcome (HR, 0.92 [95% CI, 0.79-1.08]; P = .31; Figure 1), for cardiovascular death (5.90% vs 6.97%, respectively; P = .32; eFigure 7), for stroke (4.84% vs 4.79%; P = .39; eFigure 8), and for MI (3.79% vs 2.98%; OR, 1.24 [95% CI, 0.91-1.69]; P = .16; eFigure 9). The event rates were higher in those with β-blocker use (1101 [30.59%]) vs those without β-blocker use (1002 [27.84%]) for the secondary outcome (OR, 1.14 [95% CI, 1.03-1.27]; P = .01) and for hospitalization (870 [24.17%] vs 773 [21.48%], respectively; OR, 1.17 [95% CI, 1.04-1.30]; P = .01) in the propensity score–matched model (Figure 2). The results were similar in the propensity score–adjusted model (eFigure ).

In the risk factors alone cohort, there were 4854 patients (26%) with β-blocker use. Prior to propensity score matching, there were significant differences between patients with β-blocker use vs those without β-blocker use in the baseline characteristics (Table 3). Propensity score matching matched 3952 patients with β-blocker use (81% of the cohort) with 3952 patients without β-blocker use (29% of the cohort) with similar propensity scores. After propensity score matching (Table 3), the absolute standardized differences were less than 10% for all matched variables, indicating an adequate match (eFigure 10). The median follow-up was 43 months (IQR, 32-45 months) in the group with β-blocker use vs 43 months (IQR, 31-45 months) in the group without β-blocker use.

The event rates were higher in those with β-blocker use (467 [14.22%]) vs those without β-blocker use (403 [12.11%]) for the primary outcome (HR, 1.18 [95% CI, 1.02-1.36]; P = .02; Figure 1), for the secondary outcome (870 [22.01%] vs 797 [20.17%], respectively; OR, 1.12 [95% CI, 1.00-1.24]; P = .04) but not for MI (89 [2.82%] vs 68 [2.00%]; HR, 1.36 [95% CI, 0.97-1.90]; P = .08; eFigure 11), and for stroke (210 [6.55%] vs 168 [5.12%]; HR, 1.22 [95% CI, 0.99-1.52]; P = .06; eFigure 12). In the propensity score–matched model, there were similar event rates for cardiovascular death (6.41% in patients with β-blocker use vs 6.40% in those without β-blocker use; P = .66; eFigure 13) and for hospitalization (14.62% vs 13.74%, respectively; P = .26; Figure 2). The results were similar in a propensity score–adjusted model (eFigure ).

A significant interaction was found for secondary outcome risk both in the propensity score–matched (P = .003 for interaction; Figure 2) and the propensity score–adjusted models (P = .006 for interaction; eFigure ), such that the effect of β-blocker use was not uniform across the cohorts. The secondary outcome risk was not different in the prior MI cohort, whereas in the CAD cohort and the risk factors alone cohort, patients with β-blocker use had a 12% to 14% higher secondary outcome risk compared with the group without β-blocker use. A similar interaction was noted for the outcomes of MI (P = .08 for interaction) and hospitalization (P = .03 for interaction; Figure 2), such that the effect of β-blocker use was not uniform across the cohorts.

A sensitivity analysis that was performed in the cohort of patients without known heart failure yielded largely similar results (eFigure 14-15). Similar results were seen for the primary outcome across various statistical models including models using β-blocker as a time-dependent covariate for both propensity-adjusted and inverse probability weighting (eTable). The analysis performed in the cohort with recent MI (≤1 year) showed an association of β-blocker use with a lower incidence of the secondary outcome (OR, 0.77 [95% CI, 0.64-0.92]) and hospitalization (OR, 0.77 [95% CI, 0.62-0.95]), but not with the primary outcome (HR, 0.79 [95% CI, 0.60-1.04]). Propensity score–adjusted analysis restricted to the cohort with heart failure (n = 6056) showed no benefit for the primary outcome (HR, 0.89 [95% CI, 0.79-1.01]; P = .08).

In this analysis of 44 708 patients from the REACH registry, β-blocker use was not associated with a lower incidence of cardiovascular events among individuals with a prior history of MI, among individuals with CAD but no MI history, or among individuals with risk factors only for atherosclerotic disease.

In patients after MI, there is evidence that β-blocker use reduces mortality.^14- 16 In a meta-analysis published in 1999,¹⁷ a 23% reduction in death in trials of β-blocker use was observed. However, in this meta-analysis,¹⁷ the mean follow-up was only 1.4 years, the median publication date of the 82 randomized trials included was 1982, and most of the trials were performed in the era before modern reperfusion or medical therapy was routine for MI. A reperfused, viable myocardium has little in common with a necrotic or scarred myocardium, which generates arrhythmias because of reentry mechanisms, enhancing the capability of β-blockers to prevent sudden death. In the present study, more than 80% of patients with prior MI and with β-blocker use also used aspirin or lipid-lowering agents and more than 50% used an angiotensin-converting enzyme inhibitor. In a subgroup analysis from a cohort with prior MI from the International Verapamil-Trandolapril Study (INVEST), a non–β-blocker−based strategy (verapamil-trandolapril) did not significantly differ from a β-blocker–based strategy for the prevention of cardiovascular events, and was also associated with a greater subjective feeling of well-being and a trend toward lower incidence of angina pectoris and stroke.¹⁸

Our observation that β-blocker use was not associated with a lower rate of cardiovascular events among patients with MI and among those with CAD but no history of MI are consistent with the recently updated American Heart Association secondary prevention guidelines in which β-blocker therapy is a class I recommendation for heart failure, MI, or acute coronary syndrome for up to 3 years after MI, but it is a class IIa recommendation for longer-term therapy.¹⁹ Moreover, β-blocker therapy for all other patients with coronary or other vascular disease was downgraded to a class IIb recommendation.¹⁹ Similarly, the updated 2011 European Society of Cardiology secondary prevention guidelines recommend long-term β-blocker therapy only in patients with reduced left ventricular systolic function (class I).²⁰

In our analyses, we did not have data on the type of β-blocker used. However, our results are important for several reasons. First, although guidelines recommend β-blocker therapy, there are no clear recommendations as to the β-blocker type (except for heart failure). Second, in the only randomized trial (Carvedilol Acute Myocardial Infarction Study; CAMIS)²¹ comparing carvedilol with atenolol in patients with acute coronary syndrome (left ventricular ejection fraction of 53.9%), there was no superiority of the newer β-blocker compared with the older one for composite cardiovascular events (P = .99). Third, atenolol is still the second most commonly prescribed β-blocker in the United States with more than 33.8 million prescriptions,²² and a total retail cost of more than $260 million in 2010²³; even if atenolol was the most commonly prescribed β-blocker in the study, the results would be widely applicable.

Ever since the landmark studies of β-blocker use in patients after MI or heart failure, β-blockers have been considered to be a cardioprotective therapy. Their benefits have been extrapolated to patients with CAD without MI or heart failure and also to patients at high risk for CAD.

In the CAD and risk factors only cohorts in the present analysis, we observed higher event rates for certain outcomes with β-blocker therapy. The paucity of evidence for β-blocker therapy is described in the European Society of Cardiology guidelines on stable angina. As noted in these guidelines,²⁴ a cardioprotective benefit of β-blockers has not been demonstrated in randomized controlled trials of patients with stable coronary disease.

Prior studies in patients with hypertension^25- 34 have shown that when compared with placebo, β-blocker use is not associated with reduction in all-cause mortality or MI.^35- 36 β-Blocker use is associated with a 19% to 20% reduction in stroke,^35- 36 but this reduction is less than that observed with other antihypertensive agents.³⁷ Based on this evidence, β-blocker therapy has been downgraded by the European Society on Hypertension, the American Heart Association, and others^38- 41 to a fourth-line agent for the treatment of hypertension.

Some of the purported mechanism for lack of efficacy of β-blockers include reduced efficacy at reducing central aortic pressure^42- 43 and metabolic adverse effects, including the risk of new diabetes and dyslipidemia.^30,44 In addition, β-blockers are often not well tolerated and adherence may be suboptimal.^45- 47

Available evidence suggests that β-blocker use is associated with benefit in patients with acute MI (without impending shock or heart block),⁴⁸ and may be efficacious in the short-to-intermediate term duration for patients after MI and in those with chronic heart failure.

We did not have data on the type of β-blocker, the medication dosage, or the reason patients were without β-blocker use. In addition, we did not have data on type of MI or prior β-blocker use. Anginal status and history of heart failure were controlled for, although ejection fraction was not recorded. However, the results were largely similar in a sensitivity analysis excluding patients with known heart failure. Although propensity score matching adjusts for known confounders, other unmeasured confounders cannot be accounted for and the possibility of residual confounding by indication cannot be completely ruled out. Although our data are from a registry and are limited by the lack of nonrandomized design, contemporary data from a well-designed large registry such as the CRUSADE (Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the ACC/AHA guidelines) registry,⁴⁹ have been used to modify guideline recommendations (morphine use in non–ST-segment elevation acute coronary syndrome downgraded from a class I to class IIb). In addition, in an analysis from the REACH registry,⁷ statin use was associated with a significant reduction in the risk of events (HR, 0.73 [95% CI, 0.69-0.77]; P < .001), providing another internal validation of the data set.

Among patients enrolled in the international REACH registry, β-blocker use was not associated with a lower event rate of cardiovascular events at 44-month follow-up, even among patients with prior history of MI. Further research is warranted to identify subgroups that benefit from β-blocker therapy and the optimal duration of β-blocker therapy.