The long-term use of β blockers has been shown to improve the outcomes of patients with heart failure (HF). However, it is still disputed whether this is a class effect, and, specifically, whether carvedilol or bisoprolol are superior to metoprolol. The present study was a comparative effectiveness study of β blockers for patients with HF in a population-based setting. We conducted an observational cohort study using the Quebec administrative databases to identify patients with HF who were prescribed a β blocker after the diagnosis of HF. We used descriptive statistics to characterize the patients by the type of β blocker prescribed at discharge. The unadjusted mortality for users of each β blocker was calculated using Kaplan-Meier curves and compared using the log-rank test. To account for differences in follow-up and to control for differences among patient characteristics, a multivariate Cox proportional hazards model was used to compare the mortality. Of the 26,787 patients with HF, with a median follow-up of 1.8 years per patient, the crude incidence of death was 47% with metoprolol, 40% with atenolol, 41% with carvedilol, 36% with bisoprolol, and 43% with acebutolol. After controlling for several different covariates, we found that carvedilol (hazard ratio [HR] 1.04, 95% confidence interval [CI] 0.97 to 1.12, p = 0.22) and bisoprolol (HR 0.96, 95% CI 0.91 to 1.01, p = 0.16) were not superior to metoprolol in improving survival. Atenolol (HR 0.82, 95% CI 0.77 to 0.87, p <0.0001) and acebutolol (HR 0.86, 95% CI 0.78 to 0.95, p = 0.004) were superior to metoprolol. In conclusion, we did not find evidence of a class effect for β blockers in patients with HF.
Randomized clinical trials and meta-analyses have clearly demonstrated that long-term use of β blockers improves the outcomes of patients with heart failure (HF). It has been well proved that metoprolol succinate, carvedilol, and bisoprolol are more effective than placebo in reducing mortality in the treatment of HF. Therefore, they have often been referred to as evidence-based β blockers (EBBBs). All the comprehensive guidelines for the treatment of HF published by the major medical societies have advocated the use of β blockers as a class 1 recommendation.
However, limited data have been available about the comparative effectiveness of β blockers for HF. With the exception of the Carvedilol Or Metoprolol European Trial (COMET) trial, all other randomized clinical trials have compared β blockers against a placebo; therefore, a possible class effect could not be discerned from these studies. In everyday practice, the variation in the type of β blockers prescribed to patients with HF has been wide. This variation might have resulted from the common belief that β blockers exert a class effect or because the published head-to-head comparisons of these drugs have not been convincing.
Our objective was to conduct a comparative effectiveness study of β blockers in the treatment of HF at the population level to determine whether they exert a class effect.
Methods
We conducted an observational cohort study using data from the Quebec government administrative database of hospital discharges (Maintenance et exploitation des données pour l’étude de la clientèle hospitalière [MED-ECHO]) and Quebec prescription claims database (from Régie de l’assurance maladie du Quebec [RAMQ]). We linked the MED-ECHO and RAMQ databases through the encrypted provincial health insurance numbers. By combining the information from both databases, we were able to obtain the complete survival data for the patients in Quebec (deaths that occurred in hospital would be recorded in the MED-ECHO database and those that occurred either in or out of the hospital would be recorded in the RAMQ database).
We gathered data on the treatment and clinical outcomes of all patients who had been admitted to the hospital because of HF ( International Classification of Diseases [ICD], revision 9, code 428 or ICD, revision 10, code I50) from January 1998 to March 2008. We retained only those patients (70,023) who had been admitted to the hospital with a primary diagnosis of HF. To increase the specificity of the HF diagnosis, we excluded those patients with an admission to a nonacute hospital setting and those <65 years old. We also excluded the patients discharged to a nursing home or a rehabilitation center (because information on outpatient medication use was not available for these patients) and admissions that represented a hospital transfer (to avoid counting patients twice).
From the remaining 47,209 patients, we selected only those patients who had not been admitted within the 3 years before the study period. This 3-year exclusion period was imposed to ensure that almost all patients who had previously been admitted to the hospital because of HF had been excluded.
Next, we identified those patients with HF who had filled ≥1 prescription for a β blocker (n = 34,091). We recorded the type of β blocker, dose, and interval from discharge to the filling of the first prescription. The most commonly prescribed β blockers were metoprolol, bisoprolol, atenolol, carvedilol, and acebutolol. Most of the patients in the metoprolol group had received the short-acting, twice-daily tartrate form (>95%). Other β blockers were used in <5% of cases; thus, we restricted our sample only to those patients exposed to 1 of these 5 β blockers.
We also restricted the sample to only those patients (n = 26,787) who had had their β-blocker prescription dispensed within 90 days after discharge. This period was used to capture the patients who might not have filled their prescriptions immediately after discharge. We categorized the patients into groups according to the type of β blocker they had initially received. The data from the patients who subsequently switched β blockers were censored at the date of change.
Adherence to the therapy was measured with the aid of 4 additional variables. Persistence was determined as the percentage of days that a patient was covered by a β-blocker prescription during the year after discharge, or until death if the patient had died within that year. High persistence was defined as a persistence of ≥80%. Adherence after 275 days was computed as the percentage of patients who had filled a prescription for a β blocker within days 275 to 365 of follow-up. Adherence after 305 days was the percentage of patients who had filled a prescription for a β blocker within days 305 and 365 of follow-up.
Finally, we defined our primary outcome as mortality at any point during the follow-up period. The secondary outcome was defined as the combined end point of mortality or readmission to the hospital because of HF, whichever occurred first. For the secondary outcome of readmission to the hospital because of HF, we used the ICD-9 or ICD-10 code for HF, and, again, we recorded only those patients with a primary diagnosis of HF.
We also recorded the prescriptions filled for other medications. Data on acetylsalicylic acid prescriptions were ignored, because acetylsalicylic acid is available over-the-counter. Thus, we could not accurately assess its use by each patient. We also obtained co-morbidity information from the lists of secondary diagnoses in the hospital discharge records (using the specific ICD-9 or ICD-10 codes). The length of hospital stay, medical procedures received between admission and the first β-blocker prescription filled (e.g., catheterization, percutaneous coronary intervention, coronary artery bypass grafting, echocardiography), and the specialty of the admitting physician (i.e., general practitioner, internist, cardiologist, other) were also recorded.
We used descriptive statistics to summarize the demographic, clinical, and physician characteristics of the patients according to the type of β blocker prescribed at discharge. Unadjusted mortality for the users of each β blocker was calculated using the Kaplan-Meier method and compared using the log-rank test.
To account for the differences in follow-up duration and to control for differences among the patient characteristics, a multivariate Cox proportional hazards model was used. Among the β blockers, metoprolol had been available the longest and was therefore selected a priori (on the assumption that it would represent the largest group) as the first reference category. Adjusted hazard ratios (HRs) were estimated for the other β blockers relative to metoprolol. To allow for direct comparisons between bisoprolol and carvedilol, we conducted a second analysis with carvedilol as the reference category.
In all Cox models, the associations between the particular β blockers and mortality were adjusted for age, gender, initial dose, co-morbidities, in-hospital revascularization procedures, drugs prescribed at discharge, year of HF admission (to account for temporal trends), specialty of the admitting physician during hospitalization, interval to the filling of the first prescription, and length of stay. In addition, we created a covariate to characterize patients as being at less than, at, or greater than the target dose according to dose from the first prescription. We selected a target dose of 200 mg/day for metoprolol, 100 mg/day for atenolol, 50 mg/day for carvedilol, 10 mg/day for bisoprolol, and 400 mg/day for acebutolol, in accordance with medical data available when the study was initiated. We considered the time and the degree of β blockade for each patient using 2 binary time-dependent variables. One variable indicated the period of current use (according to the dates of the filled prescriptions and their duration), and the second specified whether the total daily dose reached the target dose.
We used forced-entry regression to include all these variables in all multivariate models to adjust the between-drug comparisons for potential confounders. All statistical analyses were performed using the SAS statistical software package, version 8 (SAS Institute, Cary, North Carolina).
Results
Our study population consisted of 26,787 patients, distributed as follows: 14,372 (54%) in the metoprolol group, 2,698 (10%) in the atenolol group, 2,140 (8%) in the carvedilol group, 6,524 (24%) in the bisoprolol group, and 1,053 (4%) in the acebutolol group.
Their median age was 77 years, and 49% of the patients were men. The most prevalent co-morbidities were hypertension (41%), cardiac dysrhythmias (40%, with atrial fibrillation accounting for 35%), diabetes mellitus (38%), dyslipidemia (29%), renal failure (acute or chronic, 39%), myocardial infarction (24%), and peripheral vascular disease (16%). Most of the patients had received diuretics (86%), mostly in the form of loop diuretics. Also, most of the patients (70%) had filled prescriptions for an angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker, or both. Other drugs were used in lesser percentages: nitrates (36%), statins (39%), warfarin (32%), calcium-channel blocker (26%), digoxin (25%), and spironolactone (16%). Patients from the carvedilol group had a greater rate of use of spironolactone (35%) and digoxin (42%). About 60% of the patients receiving acebutolol had been treated by a general practitioner. In contrast, a cardiologist had treated 50% of the patients in the carvedilol group ( Table 1 ).
| Variable | Prescribed β Blocker | All | ||||
|---|---|---|---|---|---|---|
| Metoprolol | Atenolol | Carvedilol | Bisoprolol | Acebutolol | ||
| All patients (n) | 14,372 | 2,698 | 2,140 | 6,524 | 1,053 | 2,6,787 |
| Median age (years) | 78 | 78 | 72 | 77 | 79 | 77 |
| Men (%) | 48% | 41% | 69% | 51% | 43% | 49% |
| Baseline co-morbidities (%) | ||||||
| Hypertension | 42% | 48% | 30% | 41% | 49% | 41% |
| Acute myocardial infarction | 26% | 20% | 28% | 23% | 24% | 24% |
| Cardiac dysrhythmias | 40% | 39% | 33% | 44% | 32% | 40% |
| Atrial fibrillation | 35% | 34% | 25% | 40% | 27% | 35% |
| Dyslipidemia | 28% | 27% | 32% | 32% | 29% | 29% |
| Diabetes | 38% | 35% | 39% | 39% | 39% | 38% |
| Chronic renal failure | 28% | 26% | 27% | 32% | 32% | 29% |
| Acute renal failure | 9% | 8% | 10% | 11% | 12% | 10% |
| Peripheral vascular disease | 16% | 15% | 16% | 15% | 21% | 16% |
| Cerebrovascular disease | 8% | 7% | 5% | 6% | 11% | 7% |
| Chronic obstructive pulmonary disease/asthma | 20% | 19% | 23% | 34% | 23% | 24% |
| Pulmonary edema | 1% | 1% | 1% | 1% | 1% | 1% |
| Shock | 0.2% | 0.1% | 0.4% | 0.2% | 0.0% | 0.2% |
| Malignancy | 4% | 5% | 4% | 5% | 4% | 4% |
| Dementia | 4% | 3% | 2% | 4% | 4% | 4% |
| Liver disease | 3% | 2% | 3% | 3% | 2% | 3% |
| Rheumatic disease | 2% | 3% | 2% | 2% | 3% | 2% |
| Peptic ulcer disease | 1% | 2% | 1% | 1% | 1% | 1% |
| Procedures between admission and first β-blocker prescription (%) | ||||||
| Cardiac catheterization | 15% | 10% | 21% | 15% | 11% | 15% |
| Percutaneous coronary intervention | 4% | 3% | 4% | 4% | 3% | 3% |
| Coronary artery bypass grafting | 3% | 1% | 1% | 1% | 1% | 2% |
| Other prescriptions between date of discharge and first β-blocker prescription (%) | ||||||
| Diuretics | 85% | 85% | 89% | 88% | 83% | 86% |
| Loop diuretics | 82% | 80% | 85% | 85% | 80% | 82% |
| Metolazone | 1% | 1% | 2% | 1% | 0.5% | 1% |
| Other diuretics | 4% | 7% | 2% | 4% | 7% | 4% |
| Angiotensin-converting enzyme inhibitor | 57% | 52% | 70% | 53% | 47% | 56% |
| Angiotensin II receptor blocker | 12% | 16% | 12% | 17% | 13% | 14% |
| Nitrate | 38% | 36% | 31% | 31% | 45% | 36% |
| Hydralazine | 3% | 2% | 4% | 3% | 3% | 3% |
| Spironolactone | 13% | 9% | 35% | 19% | 8% | 16% |
| Digoxin | 25% | 20% | 42% | 22% | 14% | 25% |
| Warfarin | 31% | 28% | 35% | 34% | 21% | 32% |
| Statin | 37% | 34% | 43% | 43% | 34% | 39% |
| Calcium channel blocker | 26% | 32% | 11% | 25% | 44% | 26% |
| Clopidogrel or ticlopidine | 11% | 9% | 11% | 13% | 10% | 11% |
| Amiodarone | 9% | 5% | 13% | 9% | 4% | 8% |
| Sotalol | 0.2% | 0.2% | 0.1% | 0.1% | 0.2% | 0.2% |
| Specialty of treating physician (%) | ||||||
| Family/general practice | 51% | 54% | 39% | 49% | 59% | 50% |
| Cardiology | 34% | 32% | 50% | 41% | 28% | 37% |
| Internal medicine | 11% | 10% | 8% | 7% | 10% | 10% |
| Other | 3% | 4% | 2% | 3% | 3% | 3% |
| Median hospital length of stay (days) | 7 | 7 | 7 | 7 | 7 | 7 |
The number of β-blocker prescriptions filled within the first year after discharge was similar across all groups and consistent with a pattern of 1 prescription filled per month ( Table 2 ). Most of the patients were discharged from the hospital with a β-blocker prescription (median interval from date of discharge to date of the first β-blocker prescription was 0 to 1 day for all groups). The initial dose was less than the recommended target dose (¼ to ½ of the target dose in all groups), although, overall, 14% of patients reached the target dose from the first prescription in all groups. This was an expected finding for this study population. With a length of stay in the hospital of only 7 days and most patients filling their prescription on the day of discharge, it would be expected that most patients would have a low starting dose that would then be titrated to the target dose within the next few months. The acebutolol and atenolol groups had a greater percentage of patients reaching the target dose at the first prescription (27% and 19%, respectively) compared to the metoprolol and carvedilol groups (12% and 10%, respectively).
| Variable | Prescribed β Blocker | All | ||||
|---|---|---|---|---|---|---|
| Metoprolol | Atenolol | Carvedilol | Bisoprolol | Acebutolol | ||
| Patients (n) | 14,372 | 2,698 | 2140 | 6,524 | 1,053 | 26,787 |
| Median prescriptions filled within 1 year of discharge (per patient) (n) | 12 | 12 | 13 | 12 | 12 | 12 |
| Median interval from date of discharge to date of filling prescription (days) | 0 | 1 | 1 | 0 | 1 | 0 |
| Patients who reached target dose at first prescription (%) | 12 | 19 | 10 | 16 | 27 | 14 |
| Median initial dose (mg) | 50 | 50 | 13 | 5 | 200 | 50 |
| Mean persistence (% of time) ⁎ | 81% | 77% | 86% | 83% | 78% | 81% |
| Mean high persistence † | 71% | 67% | 79% | 75% | 67% | 72% |
| Adherence after 275 days (%) ‡ | 80% | 73% | 85% | 82% | 74% | 80% |
| Adherence after 305 days (%) § | 77% | 70% | 83% | 81% | 72% | 78% |
| Patients switched to different drug at any time during follow-up | 16% | 32% | 15% | 8% | 33% | 17% |
⁎ Persistence defined as percentage of time for which patients were covered by a β-blocker prescription during year after discharge or until death if patient died within that year.
† High persistence defined as persistence ≥80%.
‡ Percentage of patients who filled a prescription for β blocker within days 275 to 365 of follow-up.
§ Percentage of patients who filled a prescription for β blocker within days 275 to 365 of follow-up.
The overall adherence rates were high. Most patients (72%) had β-blocker coverage for ≥80% of the year. The patients in the carvedilol group had the greatest adherence to the therapy (83% adherence at 305 days), and those in atenolol group had the lowest adherence rates (70% adherence at 305 days). During the follow-up period, the patients prescribed carvedilol or bisoprolol were least likely to be switched to a different β blocker (15% and 8%, respectively, vs 33% in the acebutolol group).
The median follow-up was 1.8 years per patient. During the follow-up period, 43% of the patients died. The crude (unadjusted) annual incidence of death ranged from 15% in the atenolol group to 18.4% in the metoprolol group, with an overall unadjusted yearly rate of death of 17.8 deaths per 100 person-years.
The unadjusted mortality and combined mortality-readmission rates were slightly lower for the patients prescribed carvedilol than for those prescribed metoprolol. The difference in survival was not significant when bisoprolol or acebutolol was compared to metoprolol.
After adjustment ( Table 3 ), the fixed-exposure multivariate Cox proportional hazard model indicated that carvedilol and bisoprolol were associated with a similar mortality rate (HR 1.04, 95% confidence interval [CI] 0.97 to 1.12, p = 0.22, and HR 0.96, 95% CI 0.91 to 1.01, p = 0.16, respectively) relative to metoprolol. Carvedilol and bisoprolol were also associated with a similar combined mortality and readmission rate (HR 1.08, 95% CI 1.01 to 1.16, p = 0.01, and HR 0.97, 95% CI 0.93 to 1.01, p = 0.26, respectively) relative to metoprolol. Acebutolol appeared associated with lower mortality (HR 0.86, 95% CI 0.78 to 0.95, p = 0.004) and lower mortality-readmission (HR 0.87, 95% CI 0.80 to 0.95, p = 0.002) rates relative to metoprolol.
| Variable | HR | 95% CI | p Value |
|---|---|---|---|
| Prescription filled | |||
| Metoprolol (referent) | — | — | 1 |
| Atenolol | 0.84 | 0.79–0.89 | <0.0001 |
| Carvedilol | 1.08 | 1.01–1.16 | 0.0122 |
| Bisoprolol | 0.97 | 0.93–1.01 | 0.2606 |
| Acebutolol | 0.87 | 0.80–0.95 | 0.002 |
| Co-morbidity factors | |||
| Male gender | 1.06 | 1.03–1.10 | 0.0002 |
| Age | 1.02 | 1.02–1.02 | <0.0001 |
| Hypertension | 0.92 | 0.89–0.96 | <0.0001 |
| Dyslipidemia | 0.90 | 0.87–0.94 | <0.0001 |
| Chronic obstructive pulmonary disease/asthma | 1.09 | 1.05–1.13 | <0.0001 |
| Atrial fibrillation | 1.06 | 0.98–1.15 | 0.1106 |
| Cardiac dysrhythmias | 0.99 | 0.92–1.06 | 0.8553 |
| Acute myocardial infarction | 1.04 | 1.00–1.08 | 0.0202 |
| Diabetes | 1.25 | 1.21–1.30 | <0.0001 |
| Diabetes with complications ⁎ | 1.14 | 1.07–1.21 | <0.0001 |
| Chronic renal failure | 1.3 | 1.24–1.35 | <0.0001 |
| Acute renal failure | 1.05 | 0.99–1.11 | 0.0569 |
| Cerebrovascular disease | 1.08 | 1.02–1.15 | 0.0046 |
| Peripheral vascular disease | 1.19 | 1.14–1.24 | <0.0001 |
| Malignancy | 1.47 | 1.36–1.58 | <0.0001 |
| Dementia | 1.21 | 1.11–1.31 | <0.0001 |
| Rheumatic disease | 1.07 | 0.96–1.19 | 0.221 |
| Liver disease | 1.12 | 1.02–1.24 | 0.0169 |
| Peptic ulcer disease | 0.99 | 0.85–1.16 | 0.9742 |
| Pulmonary edema | 0.96 | 0.82–1.11 | 0.5946 |
| Shock | 1.23 | 0.85–1.78 | 0.2554 |
| Specialty of treating physician | |||
| Cardiologist | 0.92 | 0.89–0.96 | <0.0001 |
| Internist or other | 1.03 | 0.99–1.09 | 0.1216 |
| Hospitalization characteristics | |||
| Total hospital length of stay, including transfers | 1.003 | 1.002–1.005 | <0.0001 |
| Admission in 1998 | 1.02 | 0.94–1.10 | 0.6056 |
| Admission in 1999 (referent) | — | — | 1 |
| Admission in 2000 | 0.95 | 0.88–1.02 | 0.155 |
| Admission in 2001 | 0.92 | 0.86–0.99 | 0.036 |
| Admission in 2002 | 0.94 | 0.88–1.01 | 0.1391 |
| Admission in 2003 | 0.93 | 0.87–1.00 | 0.0757 |
| Admission in 2004 | 0.90 | 0.84–0.98 | 0.0135 |
| Admission in 2005 | 0.87 | 0.81–0.95 | 0.0012 |
| Admission in 2006 | 0.95 | 0.87–1.03 | 0.2344 |
| Admission in 2007 | 0.97 | 0.88–1.07 | 0.6193 |
| Prescription characteristics † | |||
| Angiotensin-converting enzyme inhibitor use | 0.86 | 0.83–0.90 | <0.0001 |
| Calcium channel blocker use | 1 | 0.96–1.04 | 0.7002 |
| Digoxin use | 1.08 | 1.03–1.12 | 0.0001 |
| Amiodarone use | 1.23 | 1.16–1.30 | <0.0001 |
| Hydralazine use | 1.23 | 1.13–1.35 | <0.0001 |
| Nitrate use | 1.17 | 1.13–1.212 | <0.0001 |
| Angiotensin receptor blocker use | 0.90 | 0.85–0.95 | 0.0002 |
| Spironolactone use | 0.96 | 0.91–1.01 | 0.1329 |
| Loop diuretic use | 0.97 | 0.92–1.01 | 0.1803 |
| Metolazone use | 1.43 | 1.22–1.67 | <0.0001 |
| Other diuretic use | 0.96 | 0.88–1.04 | 0.3695 |
| Statin use | 0.94 | 0.90–0.97 | 0.0019 |
| Warfarin use | 0.96 | 0.92–1.00 | 0.1262 |
| Clopidogrel or ticlopidine use | 1.13 | 1.06–1.20 | <0.0001 |
| Sotalol use | 1.42 | 0.97–2.09 | 0.0655 |
| Interval from date of discharge to date of prescription | 0.99 | 0.99–1 | 0.084 |
| Percentage of patients reaching target dose at first prescription | 0.94 | 0.90–0.99 | 0.0366 |
| Procedures during index admission | |||
| Coronary artery bypass grafting | 0.62 | 0.53–0.74 | <0.0001 |
| Percutaneous coronary intervention | 0.94 | 0.83–1.06 | 0.3284 |
| Cardiac catheterization | 0.73 | 0.69–0.78 | <0.0001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
