The large-scale placebo-controlled beta-blocker studies in systolic heart failure revisited: results from CIBIS-II, COPERNICUS and SENIORS-SHF compared with stratified subsets from MERIT-HF

Authors


Abstract

Aims

The four pivotal beta-blocker trials in heart failure (HF) had different inclusion criteria, making comparison difficult without patient stratifying. The aim of this study was to compare, in similar patients, the effects of bisoprolol, metoprolol controlled release/extended release (CR/XL), carvedilol and nebivolol on (i) total mortality, (ii) all-cause mortality or hospitalization due to cardiovascular causes (time to first event), (iii) all-cause mortality or hospitalization because of HF and (iv) tolerability, defined as discontinuation of randomized treatment.

Methods

We compared stratified (s) subsets in MERIT-HF with patients in CIBIS-II [New York Heart Association (NYHA) class III/IV and ejection fraction (EF) ≤35%] and COPERNICUS (NYHA III/IV and EF <25%) and in patients with systolic HF in SENIORS-SHF (age ≥70 years and EF ≤35%).

Results

The annual mortality rates in the placebo and beta-blocker arms were: (i) CIBIS-II (= 2647), 13.2% vs. 8.8% (relative risk reduction 34%, 95% CI: 19–46, < 0.0001) and MERIT-HFs (= 2002), 14.8% vs. 8.6% (relative risk reduction 42%, 95% CI: 24–56, < 0.0001); (ii) COPERNICUS (= 2289), 19.7% vs. 12.8% (relative risk reduction 35%, 95% CI: 19–48, = 0.0014) and MERIT-HFs (= 795), 19.1% vs. 11.7% (relative risk reduction 39%; 95% CI: 11–58, = 0.0086); (iii) SENIORS-SHF (= 1359), 11.3% vs. 9.7% (relative risk reduction 16%, NS) and MERIT-HFs (= 985), 14.8% vs. 10.1% (relative risk reduction 32%, 95% CI: 2–53, = 0.038). The effects on the other outcomes assessed were similar. Analyses indicated fewer discontinuations from randomized treatment on beta-blockers compared with placebo in COPERNICUS and the MERIT-HFs subsets.

Conclusion

The efficacy and tolerability of bisoprolol, carvedilol and metoprolol CR/XL are similar in patients with systolic HF, irrespective of NYHA class or ejection fraction. Nebivolol is less effective and not better tolerated.

Introduction

Beta-blockers have, without doubt, been one of the pivotal therapeutic advances in the management of patients with heart failure. In landmark clinical trials, the four agents bisoprolol, metoprolol succinate controlled release/extended release (CR/XL), carvedilol and nebivolol demonstrated improvement in outcome and were subsequently approved and marketed for the treatment of heart failure. The key trials were the second Cardiac Insufficiency Bisoprolol Study (CIBIS-II, = 2647) [1], the Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure (MERIT-HF, = 3991) [2, 3], the Carvedilol Prospective Randomized Cumulative Survival Trial (COPERNICUS, = 2289) [4, 5] and the Study of the Effects of nebivolol Intervention on Outcomes and Rehospitalization in Seniors with Heart Failure Trial (SENIORS, = 2128) [6, 7]. The characteristics of patients included, however, differed in each of these trials (i.e. due to differing inclusion and exclusion criteria), particularly with respect to age, symptom severity, as assessed by New York Heart Association (NYHA) class, and left ventricular ejection fraction. Notably, whereas three of the four trials enrolled only patients with systolic heart failure [1-5], SENIORS also included patients with preserved ejection fraction [6, 7]. SENIORS also had a different primary end-point [a composite of all-cause mortality or hospitalization due to cardiovascular causes (time to first event)] compared with the other trials (which had all-cause mortality as the primary outcome), and secondary outcomes also varied amongst these trials. Consequently, assessment of the relative efficacy and tolerability of these evidence-based beta-blockers is difficult as ‘like-with-like’ comparison is not possible from the published data. Therefore, we have used original data from the largest of the trials, MERIT-HF, to create stratified subsets of patients of sufficient size to enable such comparisons to be made; outcomes were compared between ‘CIBIS-II-like’ and ‘COPERNICUS-like’ patients in MERIT-HF and patients in the CIBIS-II and COPERNICUS trials, respectively. Similarly, original data from both MERIT-HF and SENIORS [7] were used to create a common subset of older patients with systolic heart failure for comparison. We focused on all-cause mortality (the primary end-point in CIBIS-II, COPERNICUS and MERIT-HF), the composite of all-cause mortality or hospitalization due to cardiovascular causes, analysed as time to first event (the primary end-point in SENIORS), all-cause death or hospitalization because of heart failure (a commonly used ‘disease-specific’ end-point in heart failure trials) and discontinuation of study drug (to assess tolerability), using original data from each trial to create this common set of outcomes for all comparisons.

Methods

Study patients and randomized treatment

Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure (n = 3991)

Important inclusion criteria

The important inclusion criteria in the MERIT-HF study were age 40–80 years, NYHA class II–IV for 3 months before randomization despite optimal standard therapy [defined as any combination of diuretics and an angiotensin-converting enzyme (ACE) inhibitor or an angiotensin receptor blocker (ARB)] and ejection fraction ≤0.40 [2, 3]. To achieve a large proportion of randomly assigned patients at high risk, the protocol stated that the investigators should aim to include less than 40% of patients in NYHA class II (i.e. more than 60% in NYHA III or IV). Patients with ejection fractions between 0.36 and 0.40 were included only if their maximum walking distance was ≤450 m (500 yards) during a 6-min walk test.

Important exclusion criteria

The exclusion criteria in this study were acute myocardial infarction or unstable angina within 28 days before randomization, indication or contraindication for treatment with beta1-blockade, heart failure secondary to systemic disease or alcohol abuse, unstable decompensated heart failure (pulmonary oedema or hypoperfusion), supine systolic blood pressure <100 mmHg, heart rate <68 beats per min (bpm) at enrolment and other serious disorders that might complicate management during follow-up. There were no exclusion criteria related to the level of serum creatinine at baseline [8].

Randomized treatment

After a single-blind placebo run-in phase of 2 weeks, patients were randomly assigned to metoprolol CR/XL [9] or placebo with starting doses of 12.5 or 25 mg once daily. The lower starting dose was recommended for patients in NYHA class III/IV. The study protocol stated that the dose should be doubled every second week to a target of 200 mg once daily, or the highest tolerated dose. The titration schedule could be modified according to written guidelines, and at the discretion of the investigator.

Cardiac Insufficiency Bisoprolol Study-II (n = 2647)

Important inclusion criteria

The important inclusion criteria in the CIBIS-II study were age 18–80 years, NYHA class III–IV for at least 3 months before enrolment, ejection fraction ≤0.35 and optimal standard therapy, defined as any combination of diuretics and an ACE inhibitor (or ARB) [1].

Important exclusion criteria

The exclusion criteria in this study were similar to those applied in MERIT-HF. In addition, according to the protocol, heart rate <60 bpm as well as renal failure (serum creatinine >300 μmol L−1) were exclusion criteria.

Randomized treatment

Patients were started on bisoprolol 1.25 mg or placebo once daily, and the dose of bisoprolol was increased successively to 2.5, 3.75, 5, 7.5 and 10 mg according to tolerance. Patients received the first three concentrations for 1 week each, and the higher concentrations for 4 weeks.

Carvedilol Prospective Randomized Cumulative Survival Trial (n = 2289)

Important inclusion criteria

The important inclusion criteria in the COPERNICUS study were symptoms of severe heart failure (equated in this analysis to NYHA class III or IV), ejection fraction <0.25 and optimal standard therapy, defined as any combination of diuretics and an ACE inhibitor (or ARB) [4, 5].

Important exclusion criteria

The exclusion criteria in this study were similar to those applied in MERIT-HF. However, contraindication to beta-blockade referred to nonselective beta-blockade and systolic blood pressure to <85 mmHg. In addition, severe pulmonary, renal (serum creatinine >247 μmol L−1) or hepatic disease were exclusion criteria.

Randomized treatment

Patients were started on 3.125 mg carvedilol or matching placebo twice daily for 2 weeks, and the carvedilol dose was then increased at 2-week intervals (if tolerated), first to 6.25 mg, then to 12.5 mg and finally to a target dose of 25 mg twice daily.

SENIORS subgroup with impaired left ventricular function (n = 1359 with systolic heart failure, referred to as ‘SENIORS-SHF’)

Important inclusion criteria

The important inclusion criteria in the SENIORS-SHF study were age ≥70 years, ejection fraction ≤0.35 within the last 6 months and optimal standard therapy, defined as any combination of diuretics and an ACE inhibitor (or ARB) [6, 7].

Important exclusion criteria

The exclusion criteria in this study were similar to those applied in CIBIS-II, but systolic blood pressure <90 mmHg and renal failure denoted by serum creatinine ≥250 μmol L−1. In addition, significant hepatic dysfunction was an exclusion criterion in SENIORS-SHF.

Randomized treatment

Patients were started on 1.25 mg nebivolol or matching placebo once daily and, if tolerated, the nebivolol dose was increased to 2.5 and 5 mg, every 1–2 weeks, reaching a target of 10 mg once daily over a maximum of 16 weeks.

Statistical analyses

For CIBIS-II, COPERNICUS and SENIORS-SHF, P-values, point estimates and 95% confidence interval (CI) values for the prespecified end-points were taken from the published reports [1, 4, 5, 7]. For the stratified subsets in MERIT-HF, the log-rank test was used for the comparison of the two randomized groups, and a Cox proportional hazards model was used to calculate relative risk with 95% CI. The number of patients needed to treat for 1 year to avoid one event included in a prespecified end-point was estimated using the following formula: 1/(yearly event rate in placebo group – yearly event rate in the beta-blocker group). Yearly event rate (expressed in percentage) was defined as events multiplied by 100 divided by the total number of patient-years of follow-up for the event in question. Patient-years of follow-up were not available for combined end-points in SENIORS-SHF so these were estimated from reported data [6, 7]. Data indicated that the placebo mortality rate (risk) was unexpectedly low in SENIORS-SHF compared with the stratified MERIT-HF subset (Fig. 1). To determine whether this could be explained by a lower baseline risk as estimated from baseline risk factors, a prognostic index was compared between SENIORS-SHF and the stratified MERIT-HF subset. The index was based on data from the Controlled Rosuvastatin Multinational Trial in Heart Failure (CORONA) study, which used baseline risk factors to predict mortality [10]. All analyses were by intention to treat.

Figure 1.

Bar charts illustrating yearly risk and risk reduction for all-cause mortality in CIBIS-II (upper panel), COPERNICUS (middle panel) and SENIORS-SHF (lower panel) compared in each panel with the respective stratified subsets from MERIT-HF.

Results

Comparison of stratified MERIT-HF subgroup and CIBIS-II

Table 1 shows that 2002 of the 3991 patients randomly assigned to treatment in MERIT-HF fulfilled the CIBIS-II inclusion criteria (NYHA III or IV and ejection fraction ≤0.35). Mean age was slightly lower in CIBIS-II, compared with the MERIT-HF ‘CIBIS-II-like’ subset (61.0 and 64.2 years, respectively). Ejection fraction was 0.28 and 0.25, and 19% and 24% of patients were female, respectively. Mean beta-blocker dose at last visit was 6.2 mg bisoprolol once daily in CIBIS-II and 149 mg metoprolol CR/XL once daily in the MERIT-HF ‘CIBIS-II-like’ subset.

Table 1. Baseline characteristics and drug treatment in the three different study groups reviewed. Inclusion criteria are: CIBIS-II, NYHA class III or IV and ejection fraction ≤0.35 at randomization; COPERNICUS, NYHA class III or IV and ejection fraction <0.25; and SENIORS subgroup with impaired left ventricular function (SENIORS-SHF), age ≥70 years and ejection fraction ≤0.35. The MERIT-HF stratified subsets (MERIT-HFs) have been stratified for each of the other studies to accord with these different criteria, respectively
CharacteristicsCIBIS-II criteriaCOPERNICUS criteriaSENIORS-SHF criteria
CIBIS-IIMERIT-HFs COPERNMERIT-HFs SEN-SHFMERIT-HFs
= 2647= 2002= 2289= 795= 1359= 985
  1. COPERN, COPERNICUS; SEN, SENIORS; NYHA, New York Heart Association; SBP, systolic blood pressure; DBP, diastolic blood pressure; bpm, beats per min; ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker; ASA, acetyl salicylic acid (aspirin).

Mean age (SD) years61.0 (11)64.2 (9.7)63.3 (11)64.6 (9.7)76.0 (4.7)74.4 (2.8)
Female%192421233027
NYHA class%
I000030
II00005334
III8394NA914261
IV176NA925
Ejection fraction%28 (6)25 (6)20 (4)19 (4)29 (5)25 (6)
SBP mmHg130 (19)128 (17)123 (19)124 (17)136 (20)132 (19)
DBP mmHg80 (11)78 (9)76 (11)77 (9)79 (11)76 (9)
Heart rate bpm81 (15)84 (11)83 (12)85 (11)79 (14)81 (10)
Body mass index kg m−226.8 (4.1)27.2 (4.8)NA26.5 (4.9)26.7 (4.0)25.7 (3.9)
Serum creatinine μmol L−1104 (29)109 (34)134 (37)112 (35)106 (34)117 (38)
Previous myocardial infarction%5550NA484956
History of hypertension%4343NA415344
History diabetes mellitus%1228NA252724
Diuretics%998699969393
Digitalis%526966704362
ACE inhibitor%9688NA868186
ARB%NA8NA10108
ACE inhibitor or ARB%NA969795NA94
ASA%4145NA43NA49
Statin%NA22NA212620

In CIBIS-II, the placebo-corrected reduction in heart rate with bisoprolol at 2 months was 9.6 bpm; the corresponding reduction with metoprolol CR/XL in the stratified MERIT-HF subgroup at 2 months was 11.4 bpm.

In CIBIS-II, 228 patients died in the placebo group and 156 in the bisoprolol group corresponding to a yearly mortality rate of 13.2% and 8.8%, respectively (relative risk reduction 34%, 95% CI: 19–46, < 0.0001; Fig. 1). Corresponding data for the stratified MERIT-HF subgroup were 142 vs. 87 deaths and a yearly mortality rate of 14.8% vs. 8.6% (42% risk reduction, 95% CI: 24–56, < 0.0001). The number to treat for 1 year to save one life was 23 patients in CIBIS-II and 16 patients in the stratified MERIT-HF subgroup.

In CIBIS-II, 510 patients died or were hospitalized due to cardiovascular causes in the placebo group and 408 in the bisoprolol group, corresponding to a yearly event rate of 36.4% and 27.1%, respectively (risk reduction 25%, 95% CI: 14–34, < 0.0001; Fig. 2). Corresponding data for the stratified MERIT-HF subgroup were 365 and 285 events and a yearly event rate of 45.3% and 33.1%, respectively (27% risk reduction, 95% CI: 14–37, < 0.0001). The number to treat for 1 year to avoid one event included in this combined end-point was 11 patients in CIBIS-II and eight patients in the stratified MERIT-HF subgroup.

Figure 2.

Bar charts illustrating yearly risk and risk reduction for the combined end-point of all-cause mortality and hospitalization due to cardiovascular causes (time to first event) in CIBIS-II (upper panel), COPERNICUS (middle panel) and SENIORS-SHF (lower panel) compared in each panel with the respective stratified subsets from MERIT-HF.

In CIBIS-II, 383 patients died or were hospitalized because of worsening heart failure in the placebo group and 264 in the bisoprolol group corresponding to a yearly event rate of 28.6% vs. 18.2% (risk reduction 36%, 95% CI: 25–45, < 0.0001; Fig. 3). Corresponding data for the stratified MERIT-HF subgroup were 285 and 202 events, with a yearly event rate of 33.2% and 21.8%, respectively (34% risk reduction, 95% CI: 21–45, < 0.0001). The number to treat for 1 year to avoid one event included in this combined end-point was 11 patients in CIBIS-II and nine patients in the MERIT-HF ‘CIBIS-II-like’ group.

Figure 3.

Bar charts illustrating yearly risk and risk reduction for the combined end-point of all-cause mortality and hospitalization because of worsening of congestive heart failure (CHF; time to first event) in CIBIS-II (upper panel), COPERNICUS (middle panel) and SENIORS-SHF (lower panel) compared in each panel with the respective stratified subsets from MERIT-HF.

In CIBIS-II, a similar number of patients discontinued randomized treatment in the bisoprolol group compared with the placebo group (Table 2). In the stratified MERIT-HF subgroup, 12% fewer patients discontinued metoprolol CR/XL compared with placebo (NS) (Table 2).

Table 2. Number of patients who discontinued study drug during follow-up in CIBIS-II (bisoprolol), COPERNICUS (carvedilol), SENIORS-SHF (nebivolol) and in the stratified (strat) subsets from MERIT-HF (metoprolol CR/XL)
End-point Placebo, nBeta-blocker, nRelative risk (95% CI) P-value
CIBIS-II1921941.00 (0.82–1.22)NS
MERIT-HF strat for CIBIS-II1741590.88 (0.71–1.09)NS
COPERNICUS2952310.77 (0.62–0.96)0.02
MERIT-HF strat for COPERNICUS86620.69 (0.50–0.96)0.027
SENIORS-SHF170170NANS
MERIT-HF strat for SENIORS-SHF91890.95 (0.71–1.27) NS

Comparison of stratified MERIT-HF subgroup and COPERNICUS

Table 1 shows that 795 of the 3991 patients randomly assigned to treatment in MERIT-HF fulfilled the COPERNICUS inclusion criteria (NYHA III or IV and ejection fraction <0.25). Mean age was similar in COPERNICUS and MERIT-HF: 63.3 years and 64.6 years, respectively; ejection fraction was 0.20 and 0.19, and 21% and 23% of patients were female, respectively (Table 1). Reported concomitant treatments at baseline were also similar. Mean beta-blocker dose at last visit was 18.75 mg carvedilol twice daily and 148 mg metoprolol CR/XL once daily, respectively.

Reduction in heart rate with carvedilol was not reported in COPERNICUS; the reduction with metoprolol CR/XL in the stratified MERIT-HF subgroup was 11.3 bpm at 2 months, 12.2 bpm at 3 months and 13.3 bpm at 6 months.

In COPERNICUS, 190 patients died in the placebo group and 130 in the carvedilol group, corresponding to a yearly mortality rate of 19.7% and 12.8%, respectively (risk reduction 35%, 95% CI: 19–48, = 0.0014; Fig. 1). Corresponding data for the stratified MERIT-HF ‘COPERNICUS-like’ patients were 72 and 45 deaths and a yearly mortality rate of 19.1% and 11.7% (39% risk reduction, 95% CI: 11–58, = 0.0086). The number needed to treat for 1 year to save one life was 14 patients in both studies.

In COPERNICUS, 395 patients died or were hospitalized due to cardiovascular causes in the placebo group and 314 in the carvedilol group corresponding to a yearly event rate of 41.6% and 30.2%, respectively (risk reduction 27%, 95% CI: 16–37, < 0.0001; Fig. 2). Corresponding data for the stratified MERIT-HF subset were 175 and 118 events, and a yearly event rate of 58.1% and 35.7%, respectively (39% risk reduction, 95% CI: 22–51, < 0.0001). The number to treat for 1 year to avoid one event included in this combined end-point was nine patients in COPERNICUS and five patients in the stratified MERIT-HF subgroup.

In COPERNICUS, 357 patients died or were hospitalized because of worsening heart failure in the placebo group and 271 in the carvedilol group corresponding to a yearly event rate of 37.9% and 25.5%, respectively (risk reduction 31%, 95% CI: 19–41, < 0.0001; Fig. 3). Corresponding data for the stratified MERIT-HF ‘COPERNICUS-like’ patients were 144 and 88 events, and a yearly event rate of 44.6% and 24.9%, respectively (44% risk reduction, 95% CI: 27–57%, < 0.0001). The number to treat for 1 year to avoid one event included in this combined end-point was eight patients in COPERNICUS and five patients in the MERIT-HF ‘COPERNICUS-like’ subgroup.

In COPERNICUS, 23% fewer patients discontinued randomized treatment in the carvedilol group compared with the placebo group (= 0.02, Table 2). Corresponding data for the matched MERIT-HF subgroup showed that 31% fewer patients discontinued metoprolol CR/XL compared with placebo (= 0.027, Table 2).

Comparison of stratified MERIT-HF subgroup and SENIORS-SHF

Table 1 shows that 985 of the 3991 patients randomly allocated to treatment in MERIT-HF fulfilled the SENIORS-SHF inclusion criteria (age ≥70 years and ejection fraction ≤0.35). Mean age in SENIORS-SHF and MERIT-HF was 76.0 and 74.4 years, respectively; ejection fraction was 0.29 and 0.25, and 30% and 27% of patients were female, respectively. Reported concomitant treatments at baseline were similar. Mean beta-blocker dose at last visit was 7.4 mg nebivolol and 141 mg metoprolol CR/XL once daily, respectively.

In SENIORS-SHF, the placebo-corrected reduction in heart rate with nebivolol at 4 months was 8.8 bpm; the corresponding reduction with metoprolol CR/XL in the stratified MERIT-HF subgroup at 3 months was 12.5 bpm and at 6 months was 12.8 bpm (heart rate was not recorded at 4 months in MERIT-HF).

In SENIORS-SHF, 135 patients died in the placebo group and 115 in the nebivolol group corresponding to a yearly mortality rate of 11.3% and 9.7%, respectively (risk reduction 16%, NS; Fig. 1). Corresponding data for the stratified MERIT-HF subgroup were 68 and 49 deaths, and a yearly mortality rate of 14.8% and 10.1%, respectively (32% risk reduction, 95% CI: 2–53, = 0.038). The number to treat for 1 year to save one life was 63 patients in SENIORS-SHF (NS) and 21 patients in the stratified MERIT-HF subgroup.

In SENIORS-SHF, 247 patients died or were hospitalized due to cardiovascular causes in the placebo group and 218 in the nebivolol group corresponding to a yearly event rate of 24.1% and 21.4%, respectively (risk reduction 14%, NS; Fig. 2). Corresponding data for the matched MERIT-HF subgroup were 173 and 143 events, and a yearly event rate of 44.0% and 34.1%, respectively (22% risk reduction, 95% CI: 3–38, = 0.026). The number to treat for 1 year to avoid one event included in this combined end-point was 37 patients in SENIORS-SHF (NS) and 10 patients in the stratified MERIT-HF subgroup.

In SENIORS-SHF, 181 patients died or were hospitalized because of worsening heart failure in the placebo group and 170 in the nebivolol group corresponding to a yearly event rate of 16.5% and 15.6%, respectively (risk reduction 6%, NS; Fig. 3). Corresponding data for the matched MERIT-HF subgroup were 132 and 101 events, and a yearly event rate of 31.4% and 22.4%, respectively (28% risk reduction, 95% CI: 7–45, = 0.012). The number to treat for 1 year to avoid one event included in this combined end-point was 111 patients in SENIORS-SHF (NS) and 11 patients in the stratified MERIT-HF subgroup.

In SENIORS-SHF, similar number of patients discontinued randomized treatment in the nebivolol group compared with the placebo group (170 patients in each group, Table 2). In the stratified MERIT-HF subset, 5% fewer patients discontinued metoprolol CR/XL compared with placebo (NS, Table 2).

Comparison of baseline risk in SENIORS-SHF and the stratified MERIT-HF subgroup

The observed yearly total mortality rate in the placebo arm was 11.3% in SENIORS-SHF and 14.8% in the stratified MERIT-HF subset; that is, a 24% lower rate in SENIORS-SHF compared with the MERIT-HF subset. A number of analyses were performed to determine whether differences in baseline risk factors (Table 1) could explain this difference in yearly mortality rate observed during follow-up (Table 3). The results showed that the analysed baseline risk factors indicated an expected 28% lower mortality rate in the placebo group during follow-up in SENIORS-SHF compared with the stratified MERIT-HF placebo subset.

Table 3. Estimation of relative follow-up risk of total mortality in placebo group comparing SENIORS-SHF with MERIT-HF subgroup stratified for SENIORS-SHF inclusion criteria (age ≥70 years and ejection fraction ≤0.35). Follow-up risk has been defined from several predefined baseline risk factors (see Methods)
VariablesSEN-SHFMERIT-HFDelta-valueβ-coeffaProductb
= 1 359= 985
  1. SEN, SENIORS; coeff, coefficient; NYHA, New York Heart Association; bpm, beats per min; BMI, body mass index.

  2. a

    From the Controlled Rosuvastatin Multinational Trial in Heart Failure (CORONA) study, which used baseline risk factors to predict mortality (10).

  3. b

    Delta-value multiplied by β-coefficient.

Age/10 years7.67.440.160.2580.0413
Female sex0.300.270.03−0.166−0.00497
NYHA class III/IV0.440.66−0.220.286−0.0629
Ejection fraction29254.00−0.0321−0.128
Systolic blood pressure/10 mmHg13.613.20.400.004550.0182
Heart rate/10 bpm7.98.1−0.200.117−0.0235
BMI kg m−226.725.71.00−0.0554−0.0554
Serum creatinine/10 μmol L−110.611.7−1.100.0782−0.0860
Diabetes mellitus0.270.240.0300.3090.00927
Global sum    −0.329
Hazard ratio    0.72

Discussion

Our attempt to stratify patients from MERIT-HF with those in the other pivotal beta-blocker trials resulted in subsets of patients with good correspondence with respect to baseline characteristics and background treatment in CIBIS-II and COPERNICUS. However, the baseline characteristics of patients in SENIORS-SHF indicated that they were at lower risk of death compared with the stratified MERIT-HF subgroup (see below). The MERIT-HF subsets were sufficiently large to provide robust estimates of event rates and treatment effects. We found important similarities and differences in both the clinical outcomes examined and in the effects of the different beta-blockers tested.

First with regard to clinical outcomes, the most objective end-point (and the one least influenced by local practice), all-cause mortality, normalized for duration of follow-up, was remarkably similar in the subsets of stratified patients in MERIT-HF compared with CIBIS-II and COPERNICUS. However, mortality rate in the placebo group was lower in SENIORS-SHF (11.3%) than in the stratified MERIT-HF subset (14.8%). One explanation for this discrepancy is very likely to be differences in baseline risk factors (Tables 1 and 3). Our analyses using predefined important baseline risk factors for mortality indicated a 28% lower baseline risk in SENIORS-SHF patients (see Table 3), which is consistent with the 24% lower mortality rate in the placebo arm compared with the stratified MERIT-HF subset observed during follow-up (Fig. 1, lower panel).

There was less agreement between event rates in the matched ‘CIBIS-II-like’ and ‘COPERNICUS-like’ cohorts within MERIT-HF and the patients in CIBIS-II and COPERNICUS, respectively, for the composite of all-cause death or hospitalization due to cardiovascular causes, possibly because of varying definitions of what was meant by ‘cardiovascular’ in each trial and different practices in the participating countries. There was much more consistency for the composite of all-cause mortality or hospitalization because of heart failure amongst these three trials, reflecting the more homogenous definition of hospitalization for heart failure. However, once again SENIORS-SHF had much lower event rates in the placebo arm than the stratified MERIT-HF placebo subset (e.g. the rate in the placebo arm of the composite of death or hospitalization because of heart failure, time to first event, was 47% lower than in the stratified MERIT-HF subset, 16.5% vs. 31.4%, Fig. 3, lower panel). The close correlation between the event rates in matched patients in MERIT-HF, CIBIS-2 and COPERNICUS suggests that SENIORS-SHF is the outlier, for reasons that are commented upon above (lower baseline risk).

The same general pattern was apparent with respect to the beneficial effects of treatment. The large reductions in all-cause mortality with metoprolol CR/XL in ‘CIBIS-II-like’ and ‘COPERNICUS-like’ patients in MERIT-HF were similar to those of bisoprolol and carvedilol in CIBIS-II and COPERNICUS, respectively. This finding was consistent for the other efficacy outcomes examined, with particularly large relative and absolute risk reductions in ‘COPERNICUS-like’ patients.

However, these findings contrasted strikingly with those for the comparison between ‘SENIORS-SHF-like’ patients in MERIT-HF and patients in SENIORS-SHF. Whereas treatment with metoprolol CR/XL reduced each of the efficacy outcomes significantly, no outcome was significantly improved with nebivolol. This difference could not be explained on the basis of inadequate statistical power as the number of patients experiencing each outcome was considerably larger in the subset from the SENIORS-SHF; there were 250 deaths amongst patients in SENIORS-SHF compared with 117 in the stratified subgroup of patients from MERIT-HF. One potentially relevant difference, however, was in the reduction in heart rate obtained with nebivolol (placebo-corrected decrease of 8.8 bpm at 4 months in SENIORS) compared with metoprolol CR/XL (decrease in 12.5 bpm at 3 months and 12.8 bpm at 6 months in ‘SENIORS-like’ patients in MERIT-HF) as the degree of reduction in heart rate seems to be associated with the magnitude of clinical benefit of beta-blockers [11].

Study-drug tolerability differed between trials but less so amongst the patient cohorts; for example, discontinuations from randomized treatment in the placebo group in the ‘COPERNICUS-like’, ‘CIBIS-II-like’ and ‘SENIORS-SHF-like’ subsets from MERIT-HF were very similar. Remarkably, the discontinuation rates for bisoprolol in CIBIS-II, carvedilol in COPERNICUS and metoprolol CR/XL in the corresponding stratified patient cohorts were lower than in the corresponding placebo groups. Of note, discontinuation of active therapy was significantly less than that of placebo in those with the most advanced heart failure (i.e. ‘COPERNICUS-like’ patients). However, again, nebivolol proved to be the exception, with a similar discontinuation rate in the active therapy and placebo groups in SENIORS-SHF.

Our study has a number of limitations. First, direct comparisons are preferable to indirect comparisons; realistically, however, a trial of comparison to evidence-based beta-blockers is unlikely to be conducted. Secondly, stratification of patients in MERIT-HF to those in the other trials was based only on two variables (NYHA class and ejection fraction in CIBIS-II and COPERNICUS; age and ejection fraction in SENIORS-SHF), although the patients in the resultant MERIT-HF cohorts were in fact quite similar to those in CIBIS-II and COPERNICUS and, for three of these trials, the mortality rates were also similar. Finally, we were only able to compare patients in MERIT-HF with the subgroup of SENIORS patients with systolic heart failure. Arguably, this could have resulted in reduced power to observe an effect of nebivolol but the number of events in SENIORS-SHF was still larger than in the corresponding subset of patients from MERIT-HF in which a clear benefit of metoprolol CR/XL was demonstrated.

In conclusion, the efficacy and tolerability of bisoprolol, carvedilol and metoprolol CR/XL are similar in patients with systolic heart failure, irrespective of NYHA class or ejection fraction. Nebivolol is less effective and is not better tolerated.

Conflict of interest statement

JW was formerly a scientific advisor on cardiovascular research to AstraZeneca, Mölndal, Sweden. JJVMcM was involved in other trials sponsored by AstraZeneca (CHARM and CORONA) for which his employer, Glasgow University, was paid for his time spent as an Executive Committee member. HW was a member of the Executive Committees of MERIT-HF and CORONA for which he received consulting fees. DC has no conflicts of interest to declare.

Ancillary