Author + information
- Received February 21, 2014
- Revision received April 23, 2014
- Accepted April 25, 2014
- Published online July 22, 2014.
- Charlotte Andersson, MD, PhD∗,
- David Shilane, PhD†,
- Alan S. Go, MD‡,
- Tara I. Chang, MD, MS§,
- Dhruv Kazi, MD, MS‖,
- Matthew D. Solomon, MD, PhD‡,
- Derek B. Boothroyd, PhD† and
- Mark A. Hlatky, MD†,§∗ ()
- ∗Department of Cardiology, Gentofte University Hospital, Hellerup, Denmark
- †Department of Health Research and Policy, Stanford University School of Medicine, Stanford, California
- ‡Division of Research, Kaiser Permanente, Oakland, California
- §Department of Medicine, Stanford University School of Medicine, Stanford, California
- ‖Department of Medicine, University of California, San Francisco, California
- ↵∗Reprint requests and correspondence:
Dr. Mark A. Hlatky, Stanford University School of Medicine, HRP Redwood Building, Room T150, Stanford, California 94305-5405.
Background The effectiveness of beta-blockers for preventing cardiac events has been questioned for patients who have coronary heart disease (CHD) without a prior myocardial infarction (MI).
Objectives The purpose of this study was to assess the association of beta-blockers with outcomes among patients with new-onset CHD.
Methods We studied consecutive patients discharged after the first CHD event (acute coronary syndrome or coronary revascularization) between 2000 and 2008 in an integrated healthcare delivery system who did not use beta-blockers in the year before entry. We used time-varying Cox regression models to determine the hazard ratio (HR) associated with beta-blocker treatment and used treatment-by-covariate interaction tests (pint) to determine whether the association differed for patients with or without a recent MI.
Results A total of 26,793 patients were included, 19,843 of whom initiated beta-blocker treatment within 7 days of discharge from their initial CHD event. Over an average of 3.7 years of follow-up, 6,968 patients had an MI or died. Use of beta-blockers was associated with an adjusted HR for mortality of 0.90 (95% confidence limits [CL]: 0.84 to 0.96), and an adjusted HR for death or MI of 0.92 (CL: 0.87 to 0.97). The association between beta-blockers and outcomes differed significantly between patients with and without a recent MI (HR for death: 0.85 vs. 1.02, pint = 0.007; and HR for death or MI: 0.87 vs. 1.03, pint = 0.005).
Conclusions Use of beta-blockers among patients with new-onset CHD was associated with a lower risk of cardiac events only among patients with a recent MI.
- beta-adrenergic blocking agents
- comparative effectiveness research
- outcomes research
- treatment effectiveness
The clinical effectiveness of beta-blockers in reducing death and major adverse cardiac events among patients with coronary heart disease (CHD) has recently been questioned. Randomized clinical trials have shown that beta-blockers are effective in reducing cardiac events and mortality among patients who have had a recent myocardial infarction (MI) or who have heart failure with systolic dysfunction (1–5). In contrast, among patients with stable ischemic heart disease, only small, short-term randomized trials have tested whether beta-blockers reduce anginal symptoms compared with placebo, and no large trials have been performed to assess their effectiveness in reducing major cardiac events (6,7). Thus, the general belief that beta-blockers are cardioprotective for all patients with CHD is largely on the basis of extrapolating the results of clinical trials in patients with MI or heart failure to all patients with CHD.
A recent analysis of the observational data from the REACH (REduction of Atherothrombosis for Continued Health) registry found no significant association of beta-blocker therapy with major adverse cardiovascular events among patients with stable CHD (8). This finding initiated a vigorous debate about whether beta-blockers improve prognosis among patients with CHD. The effectiveness of beta-blockers as first-line therapy for hypertension in reducing cardiovascular events (9,10) and as cardioprotective agents for noncardiac surgery (11) also has been called into question. In light of these recent concerns, we sought to assess the association of beta-blocker therapy with major adverse cardiac outcomes among patients with incident CHD.
Source of Data
We used electronic health records from Kaiser Permanente Northern California, a large, integrated healthcare delivery system that provides care to 3.2 million subjects. The membership is broadly representative of the local population, apart from slightly lower representation of the extremes of age and income. The electronic health records collect comprehensive data on hospitalizations, outpatient encounters, laboratory test results, and prescription medications. Diagnoses are coded according to the International Classification of Diseases (ICD)-9th version, and procedures are coded using ICD-9CM codes and Current Procedural Terminology codes, 4th edition.
We included all patients age 30 years and older who had an initial diagnosis of CHD between January 1, 2000, and December 31, 2008, which we defined as hospitalization for acute MI (non–ST-segment elevation myocardial infarction [NSTEMI] ICD-9CM codes 410.7 and 410.9 or ST-segment elevation myocardial infarction [STEMI] codes 410.0 to 410.6 and 410.8), unstable angina (ICD-9CM 411.1), or having undergone coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI). We excluded patients with any prior diagnoses of MI, unstable angina, or coronary revascularization. We also excluded patients with a history of chronic lung disease (ICD-9CM 491.x, 492.x, 493.x, 496, 518.1, 518.2), which is a relative contraindication to beta-blocker therapy. We adopted a new user design by excluding patients who received a prescription for a beta-blocker within the year before study entry. Patients were classified according to their initial presentation as having had a recent MI, defined as an index hospitalization for either STEMI or NSTEMI.
Definition of Comorbidity, Pharmacological Treatment, and Patient-Related Characteristics
We defined comorbid conditions on the basis of in-hospital and outpatient diagnosis codes in the year before the incident CHD event. Unless otherwise described, comorbidities were considered present if at least 1 prior diagnosis was available in the electronic health records. A diagnosis of heart failure required at least 2 outpatient visits with an ICD-9CM code for heart failure or an inpatient admission with a primary diagnosis of heart failure (12). Full codes for classification of the different diseases are available elsewhere (13).
We identified drug use (beta-blockers, statins, angiotensin-converting enzyme [ACE] inhibitors, angiotensin receptor blockers [ARB], and clopidogrel) on the basis of filled prescriptions from a Kaiser pharmacy in the year before the index hospitalization (for baseline exposure) and during subsequent follow-up. To account for partial adherence, dose reductions, and possible persistence of any drug effects after discontinuation, we considered patients to be on treatment with a given drug from the date they filled the prescription until 30 days after the date when the prescription would need to be refilled (i.e., allowing a 30-day “grace period”), on the basis of the dosage and the number of pills dispensed. The grace period is a conservative assumption in the analysis, as events that occur within the grace period are then counted as occurring on-treatment rather than off-treatment.
Study Start, End, and Definition of Outcomes
We followed patients from the index date (7 days after hospital discharge for their qualifying event) until December 31, 2008, or disenrollment from the health plan. We assessed 2 study endpoints: 1) all-cause mortality, identified from health plan administrative databases, Social Security Administration vital status files, and the California state death certificate registry; and 2) the composite of death from any cause or hospitalization for an acute MI (primary discharge diagnosis code of 410.x1).
This study was approved by the institutional review boards of the Kaiser Foundation Research Institute and Stanford University.
We used time-varying multivariable Cox proportional hazard regression models to assess the association of beta-blocker use with outcomes, which allows patients to have on-treatment and off-treatment periods during follow-up. We used a series of Cox regression analyses to adjust for successively more potential confounding factors. In the first model, we adjusted for baseline demographic factors, baseline clinical characteristics, and time-updated indicators of other cardiac drug treatments (statins, ACE/ARBs, and clopidogrel). In the second model, we added time-updated indicators of cardiac procedures (CABG, PCI) and adverse events (heart failure, unstable angina) in follow-up.
A priori, we specified the hypothesis that the presence of a recent MI would modify the association of beta-blocker use with outcomes. We tested this hypothesis using the probability value for the treatment-by-covariate interaction test (pint) in the models that controlled for time-varying beta-blocker exposure and recent MI, as well as for other baseline covariates and time-varying exposure to other cardiac drugs.
All analyses were performed with the R statistical package, version 2.12.2 (R Development Team, Vienna, Austria).
A total of 64,595 patients were hospitalized for an initial episode of CHD during the study period, and 28,785 were excluded for prior use of beta-blockers, 7,431 for a history of chronic lung disease, and 1,586 died or had a nonfatal MI before the index date (i.e., during the hospital admission or within 7 days after discharge). Of the 26,793 remaining patients, 19,843 (74%) initiated beta-blocker treatment before the index date, another 3,819 patients (14%) started beta-blockers during the subsequent year of follow-up, and 3,131 patients (12%) did not use beta-blockers during the average 3.7 years of follow-up. Patients who initiated beta-blockers within 7 days of discharge were younger, more likely to be male, and to have presented with a STEMI, but less likely to have diabetes, hypertension, heart failure, or atrial fibrillation than the remaining patient group (Table 1).
Most patients who initiated treatment with beta-blocker therapy subsequently had periods off-treatment, or discontinued beta-blockers entirely: 75% of patients who initiated beta-blockers within 7 days had a gap in treatment during follow-up, with a mean exposure of 2.5 person-years on-treatment and 1.1 person-years off-treatment. Of the patients who later initiated beta-blocker therapy, 77% subsequently had a gap in treatment, with 2.4 person-years on-treatment and 1.2 person-years off-treatment. Among all study patients, we observed 61,002 person-years of follow-up on beta-blocker treatment and 36,857 person-years off beta-blocker treatment.
During follow-up, 4,240 patients died and 6,968 had a cardiac event (death or MI). The crude, unadjusted mortality rate during periods when patients were on beta-blockers was lower than during periods when patients were off beta-blockers: 3.6 versus 5.6 per 100 person-years (Table 2). Similarly, the crude, unadjusted rate of death or MI when patients were taking beta-blockers was lower than the rate when patients were not taking beta-blockers: 6.4 versus 8.3 per 100 person-years (Table 2). After adjustment for baseline factors and time-varying exposure to other cardiac prescription drugs, beta-blocker treatment was associated with a hazard ratio (HR) for death of 0.90 (95% confidence limit [CL]: 0.84 to 0.96) in the overall population (Central Illustration). Additional adjustment for the occurrence of coronary revascularization, unstable angina, and heart failure during follow-up had relatively little effect on the association of beta-blockers with mortality, yielding an HR of 0.87 (CL: 0.82 to 0.93).
The use of beta-blockers also was associated with a significantly lower rate of death or MI (Central Illustration). The HR was 0.92 (CL: 0.87 to 0.97) after adjustment for baseline characteristics and time-varying exposure to other cardiac drugs (Central Illustration), and 0.90 (CL: 0.85 to 0.95) after additional adjustment for cardiac events and procedures in follow-up.
Treatment Effect Modification
We tested the hypothesis that the association of beta-blocker use with outcomes would differ among patients with and without a recent MI. The test for interaction between beta-blocker treatment and the presence of recent MI was significant both for all-cause mortality (HR, 0.85; CL: 0.79 to 0.92 vs. HR, 1.02; CL: 0.91 to 1.15; pint = 0.007), as well as for the composite endpoint of death or MI (HR, 0.87; CL: 0.82 to 0.93 vs. HR, 1.03; CL: 0.93 to 1.13; pint = 0.005).
We found that beta-blocker use was associated with lower overall rates of death or MI in a large, contemporary cohort of patients with newly diagnosed CHD. However, these associations appeared to be modified significantly by the presence of a recent MI: there was a strong association of beta-blocker use with lower cardiac event rates only among patients who presented with an MI (Central Illustration). Our findings are in broad agreement with a recent observational study from the REACH registry, which also found no significant association of beta-blockers with cardiac events among patients with stable CHD. The associations we found between beta-blockers and reduced cardiac events among patients with a recent MI also are concordant with the results of previous randomized trials (1,3,4), which support the validity of our observational comparisons. These results, as well as those of prior studies, suggest that beta-blockers may not reduce adverse cardiac events in patients who do not have a history of an MI.
Beta-blockers reduce heart rate and have a negative inotropic effect, thereby reducing myocardial oxygen demand and ameliorating myocardial ischemia. The efficacy of beta-blockers in reducing angina and objective evidence of myocardial ischemia has been demonstrated in randomized trials that required relatively small sample sizes and relatively short follow-up of a few weeks (14,15). Because of their salutary effect on symptoms, beta-blockers became a cornerstone of medical management of angina. Subsequently, clinical trials with larger sample sizes showed that beta-blockers reduced hard cardiac events among patients with a recent MI (16), and additional trials demonstrated the effectiveness of beta-blocker treatment for patients with heart failure and reduced left ventricular function (17). Clinical guidelines have therefore given Class I indications to beta-blocker use for patients after an MI and for patients with heart failure with reduced systolic function (18).
Although beta-blockers are effective in reducing anginal symptoms among patients with CHD, there is scant evidence that they reduce cardiac events among patients who have neither a recent MI nor heart failure with systolic dysfunction. Despite a general belief that beta-blockers may reduce cardiac events in these lower-risk patients, this indication for use of beta-blockers only has a Class IIB recommendation, on the basis of “Level C” evidence—expert opinion only (18). In fact, several lines of evidence suggest there may be no effect of beta-blockers on hard outcomes for all patients with CHD. Our study, and that of the REACH investigators (8), showed the association of beta-blocker treatment with outcomes to be significantly stronger in the presence of a recent MI. Similarly, a recent Japanese observational study of patients with STEMI also found that the long-term beneficial effect of beta-blockers was limited to patients who were at a high baseline risk of adverse outcomes after discharge (19). Additional evidence comes from recent studies of noncardiac surgery, which have consistently shown that the protective effect of beta-blocker therapy is limited to patients at high risk of cardiovascular adverse events (20,21). Finally, several hypertension trials have failed to show any greater effectiveness of beta-blockers on cardiac events than other classes of antihypertensive agents (9,10).
There are several possible mechanisms for a difference in the effectiveness of beta-blockers on the basis of the presence of a prior MI. Patients with myocardial damage from an MI have activation of the sympathetic nervous system, and beta-blockers may be more effective in such patients because of their higher levels of sympathetic tone and circulating catecholamines. Use of coronary revascularization in patients with CHD would be expected to reduce episodes of myocardial ischemia, which would thereby reduce any potential benefits of beta-blockers on the basis of their anti-ischemic properties. This mechanism may explain why metoprolol had no greater effect than placebo in reducing mortality during the acute phase of MI in the COMMIT (Clopidogrel and Metoprolol in MI Trial) (22). Finally, patients with CHD who have never had an MI are at relatively lower risk, and there may be little room for further risk reduction in the presence of concomitant guideline-based therapy with statins, antiplatelet agents, and ACE/ARBs. In this study, we found a substantial attenuation of the association of beta-blockers with outcomes after statistical adjustment for the time-varying use of these other cardiac medications (Central Illustration).
The present study differs from prior studies in several ways. We adopted a “new user design” to avoid biases that arise from studying patients on long-term drug therapy. We also analyzed the data on the basis of time-varying exposure to beta-blockers, so the same patient could contribute to the analyses as being “on beta-blockers” for some intervals and “off beta-blockers” for other periods. This approach differs from the intention-to-treat framework used in randomized trials, in which patients originally assigned to a therapy are assumed to be on that therapy throughout follow-up, regardless of whether they are adherent or cross over to alternative treatments. The time-varying analysis method should yield greater risk reductions for treatment with an effective drug, as the intention-to-treat estimate is diluted by periods of nonadherence, whereas the time-varying estimate of effect is not. However, patients who are adherent with drug treatment tend to have more favorable clinical characteristics, which may bias estimates of treatment effectiveness. In this study, we found that adjustment for time-varying use of statins, ACE/ARBs, and clopidogrel attenuated the association of (time-varying) exposure to beta-blockers with mortality. We believe that controlling for time-varying exposures to all cardiac medications is therefore the best approach, as it also controls for patient factors related to overall adherence with prescribed therapies.
Although we had information on a wide variety of clinical variables, the datasets did not include information on several factors, such as the extent of coronary artery disease, left ventricular ejection fraction, and patient’s functional capacity. Furthermore, we do not have any information on why a patient was on or off beta-blockers at any particular time. Patients may have discontinued their use of beta-blocker treatment because of side effects, such as fatigue, bradycardia, hypotension, depression, or sexual dysfunction, which may identify higher-risk patients. Furthermore, because this study was observational, we cannot exclude residual confounding that might have affected our estimates of treatment effectiveness. Use of time-dependent Cox models are a form of “on-treatment” analysis in which medications can appear to be more effective because a claimed prescription is associated with compliance and perhaps a “healthy user effect.” Despite these potential biases, we found no significant association of beta-blockers with outcomes among patients without a recent MI.
Among patients with newly diagnosed CHD, beta-blockers were associated with a lowered rate of cardiac events only among patients at higher risk because of a recent MI. A randomized clinical trial is warranted to test whether beta-blockers reduce the risk of hard endpoints among lower-risk patients who have CHD, but have not had an MI.
COMPETENCY IN MEDICAL KNOWLEDGE: Both randomized trials and observational studies have shown that beta-blockers reduce mortality and reinfarction in patients with recent MI, but other studies suggest that beta-blockers may not prevent cardiac events in patients with CHD who have not sustained recent MI.
COMPETENCY IN SYSTEMS-BASED PRACTICE: Clinical practice guidelines recommend beta-blockers for survivors of recent MI (Class I), and that this form of therapy be considered for patients with coronary disease who have not sustained recent MI (Class IIB).
TRANSLATIONAL OUTLOOK: Randomized trials are needed to assess the safety, efficacy, and net clinical benefit of beta-blockers among patients with coronary disease who have not had recent MI.
This research was funded in part by a grant (0875162) from the American Heart Associationhttp://dx.doi.org/10.13039/100000968, Dallas, Texas, with additional support by a grant (FSS - 11-120873) from the Danish Agency for Science, Technology and Innovationhttp://dx.doi.org/10.13039/501100001825, Copenhagen, Denmark. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- angiotensin receptor blocker
- coronary artery bypass graft surgery
- coronary heart disease
- 95% confidence limits
- hazard ratio
- myocardial infarction
- non–ST-segment elevation myocardial infarction
- percutaneous coronary intervention
- ST-segment elevation myocardial infarction
- Received February 21, 2014.
- Revision received April 23, 2014.
- Accepted April 25, 2014.
- American College of Cardiology Foundation
- Packer M.,
- Fowler M.B.,
- Roecher E.B.,
- et al.,
- Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS) Study Group
- Huang H.L.,
- Fox K.A.A.
- Pepine C.J.,
- Handberg E.M.,
- Cooper-DeHoff R.M.,
- et al.,
- INVEST Investigators
- Hlatky M.A.,
- Solomon M.D.,
- Shilane D.,
- Leong T.K.,
- Brindis R.,
- Go A.S.
- Fox K.M.,
- Mulcahy D.,
- Findlay I.,
- Ford I.,
- Dargie H.J.
- von Arnim T.
- Freemantle N.,
- Cleland J.,
- Young P.,
- Mason J.,
- Harrison J.
- Chatterjee S.,
- Biondi-Zoccai G.,
- Abbate A.,
- et al.
- Fihn S.D.,
- Gardin J.M.,
- Abrams J.,
- et al.,
- American College of Cardiology Foundation/American Heart Association Task Force
- Nakatani D.,
- Sakata Y.,
- Suna S.,
- et al.,
- Osaka Acute Coronary Insufficiency Study (OACIS) Investigators
- Chen Z.M.,
- Pan H.C.,
- Chen Y.P.,
- et al.,
- COMMIT (ClOpidogrel and Metoprolol in Myocardial Infarction Trial) collaborative group