Author + information
- Received December 18, 2017
- Revision received June 18, 2018
- Accepted June 26, 2018
- Published online October 29, 2018.
- Cynthia A. Jackevicius, BScPhm, PharmD, MSca,b,c,d,e,∗ (, )@HeartRPh,
- Zunera Ghaznavi, MScf,
- Lingyun Lu, PharmD, MSca and
- Alberta L. Warner, MDf,g
- aDepartment of Pharmacy, Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, California
- bDepartment of Pharmacy Practice and Administration, College of Pharmacy, Western University of Health Sciences, Pomona, California
- cInstitute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
- dInstitute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
- eDepartment of Pharmacy, University Health Network, Toronto, Ontario, Canada
- fDivision of Cardiology, Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, California
- gDepartment of Medicine, David Geffen School of Medicine, University of California, Los Angeles, California
- ↵∗Address for correspondence:
Dr. Jackevicius, Western University of Health Sciences, College of Pharmacy, 309 East Second Street, Pomona, California 91766.
Objectives This study evaluated whether alpha-blocker (AB) use following an admission for heart failure (HF) was associated with an increased risk of HF readmission or death.
Background ABs, found to increase the risk of HF in the ALLHAT (Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial) trial, are commonly used for prostatic hypertrophy, including in those with or at risk for HF.
Methods This propensity score–matched cohort study included patients discharged from a Veterans Affairs hospital between January 2002 and September 2015 with a primary diagnosis of HF and ascertained AB use at discharge. Cox proportional hazards models were constructed to compare time to first HF readmission and death at 2 years between groups. Secondary analyses assessed effects by AB dose and type and by beta-blocker (BB) use.
Results Of 169,911 HF patients, 47,638 (28%) were prescribed an AB. Propensity score matching resulted in 35,713 matched pairs. In the propensity score–matched cohort, AB use was associated with fewer HF readmissions (39.8% vs. 41.7% at 2 years; hazard ratio: 0.95; 95% confidence interval [CI]: 0.92 to 0.97; p < 0.0001) and death (42.8% vs. 46.5%; hazard ratio: 0.93; 95% CI: 0.91 to 0.94; p < 0.0001). Nonselective ABs had fewer deaths and HF readmissions (p < 0.0001), while higher AB doses reduced mortality (p < 0.0001). AB treatment was associated with reduced deaths in both BB-treated and untreated patients, with no increase in HF.
Conclusions Treatment of HF patients with an AB was associated not with a higher but instead with a lower rate of HF readmission and death. Higher doses and nonselective ABs were also associated with lower mortality, regardless of BB use. ABs may be used safely in HF patients where clinically indicated. The finding of improved outcomes with ABs may warrant further study.
The use of alpha-blockers (ABs) to treat cardiovascular disease has been predominantly a story of failure. In the landmark ALLHAT (Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial) hypertension trial, incident heart failure (HF) was nearly doubled in subjects treated with doxazosin monotherapy compared with chlorthalidone, leading to premature discontinuation of the doxazosin arm (1), with similar increases also seen in smaller studies (2,3). The first V-HeFT I (Veterans Affairs Vasodilator Heart Failure Trial) trial found no benefit of prazosin monotherapy on mortality or ejection fraction (EF) compared with placebo in HF patients, albeit no harm was observed (4). As a result of these findings and the proliferation of more effective and better-tolerated antihypertensive agents, ABs have been relegated, from the perspective of the cardiovascular community, to obscurity as fourth-line antihypertensive agents (5).
However, ABs remain the first-line treatment for symptomatic prostatic hypertrophy (6). Patients with cardiovascular disease, including those with overt HF, often have concomitant prostatic hypertrophy and are frequently treated with ABs. It might be expected that the deleterious effect of ABs, as found in the ALLHAT trial with new-onset HF, might also precipitate HF in susceptible individuals, particularly in those with preexisting HF.
Only 1 small cohort study of 388 patients at a single Veterans Affairs (VA) medical center between 2002 and 2004 previously examined the safety of ABs in patients with HF (7). The researchers found no overall increase in HF readmissions or total mortality with ABs, although HF was more frequent in the subgroup not treated with beta-blockers (BBs).
Given the lingering uncertainty about the safety of these agents, particularly in those with pre-existing HF, we undertook to study the association between AB treatment and recurrent HF and death. In addition, we sought to examine whether alpha-1a selective ABs, having minimal vascular effects, have a different profile of adverse events than do nonselective ABs, as well as the potential interaction between ABs and BBs in general, and specifically with carvedilol, a BB with weak alpha antagonist properties.
Study design and data sources
We conducted a population-based, propensity score–matched, retrospective cohort study. Clinical data originally recorded in the Computerized Patient Record System, the VA electronic health record system, were extracted from the VA Corporate Data Warehouse using the VINCI platform (8). Vital status was obtained from the VA vital status file, which combines death data from VA, Medicare, and Social Security Administration sources (9).
The study population consisted of all patients having an admission to a VA Medical Center with a primary discharge diagnosis of congestive HF (International Classification of Diseases-9th Revision [ICD-9] code: 428.xx) between January 1, 2002, and September 30, 2015. For patients with multiple HF admissions, the first admission was considered the index diagnosis.
AB treatment status was defined as any prescription within 90 days before admission, during admission, or within 7 days of discharge. Patients with AB dispensed before admission but not prescribed during the admission or dispensed any time after discharge were not included in the AB user group.
The primary outcomes measured were HF readmission, defined as a primary discharge diagnosis of HF, and all-cause mortality. Patients were followed for events and censored at the time of outcome occurrence, end of cohort follow-up (August 1, 2017), or last contact with the VA system. The last use of any VA services (outpatient, inpatient, pharmacy, or other services) was recorded and used for censoring.
Demographic variables collected included age, sex, and African American race at the time of discharge from the index HF admission. Comorbidities, as listed in Table 1, were identified using ICD-9 codes from the index admission as well as all outpatient encounters, prior hospitalizations, and problem lists before the index hospitalization. ICD-9 codes were used to define these variables (Online Table S1). We also calculated the Charlson-Deyo comorbidity score for each patient based on all comorbidity data (10). Concomitant medications, as listed in Table 1, were determined from outpatient and inpatient prescriptions from 90 days before, during, or within 7 days of discharge from the index HF admission. AB type and dose were recorded and classified as low daily dose (prazosin ≤6 mg, terazosin ≤10 mg, doxazosin <8 mg, tamsulosin ≤0.4 mg, or alfuzosin ≤10 mg) or high daily dose based on recommended starting dose and maximum doses for hypertension and obstructive uropathy. Tamsulosin and alfuzosin were classified as selective (nonvasoactive) ABs, while prazosin, doxazosin, and terazosin were classified as nonselective ABs. Left ventricular EF (LVEF) was extracted from clinical notes and procedure reports using natural language processing algorithms and classified as reduced (<40%) or not reduced EF (11). BBs were classified as to whether they were approved for use in HF (carvedilol, bisoprolol, metoprolol succinate) or not (other BB).
A propensity score–matched cohort was created for the primary analysis. The propensity score was the probability of receiving treatment with an AB accounting for relevant baseline clinical variables in a multivariate logistic regression model. Variables used in the propensity score model are shown in Table 1. Patients were matched on the logit of the propensity score using 1:1 greedy nearest-neighbor matching without replacement (12,13). A caliper width of 0.20 of the SD of the logit of the propensity score for the entire cohort was used (14). AB-treated patients without a suitable match were excluded from the cohort. Standardized differences were calculated for each of the baseline covariates to assess the degree of balance between groups, with a standardized difference of <0.10 indicating a good balance between groups for that covariate. Variables used for propensity score matching are shown in Table 1. Although LVEF and systolic blood pressure are also shown in Table 1 as baseline characteristics, these variables were not used for matching because of incomplete data.
Time to HF readmission and death was compared using the Kaplan-Meier method. A Cox proportional hazards regression model with robust SEs was used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for these outcomes in relationship to AB exposure (15,16).
The primary analysis was the occurrence of HF readmission or all-cause death in the propensity score–matched cohort. Secondary analyses were performed in the entire cohort after adjusting for baseline covariates using a multivariate Cox proportional hazards model. Secondary analyses included examining the relationship between AB exposure and outcomes in several subgroups: higher-dose versus lower-dose AB treatment, selective versus nonselective AB treatment, AB agent (terazosin, doxazosin, prazosin), BB versus no BB treatment, carvedilol versus other BB, and HF-approved BB versus other BB. In these analyses, each baseline variable was compared between groups using chi-square or Student's t tests, and those significantly different at the p < 0.10 level and those deemed clinically important were used in a multivariate Cox proportional hazards model. To further validate our multivariate methods, we conducted an analysis comparing selective versus nonselective AB in a propensity score–matched cohort created specific to this analysis and using the aforementioned propensity score methods approach. To evaluate the possibility of residual confounding, we conducted a falsification analysis by examining the relationship between AB treatment and outcomes presumably unrelated to treatment, with hospitalizations with a discharge diagnosis of non–skin cancer as the outcome in the propensity-matched cohort using a Cox proportional hazards model. To analyze the potential differences between patients with a prior HF diagnosis in their response to ABs, we constructed a separate propensity score–matched cohort of only those with a prior diagnosis of HF and evaluated the effect of ABs on outcomes in this cohort using a Cox proportional hazards model. Continuous variables are expressed as mean ± SD, unless indicated otherwise. All statistical analysis was done using SAS Enterprise Guide version 7.1 (SAS Institute, Cary, North Carolina). The study was reviewed and approved by the Greater Los Angeles Research and Development Committee and the Human Subjects Subcommittee.
Of the 169,911 subjects having a qualifying hospital admission for HF, 47,638 (28%) received an AB. ABs used were terazosin (n = 24,764, mean dose 7.3 ± 5.8 mg/day), tamsulosin (n = 17,182, mean dose 0.46 ± 0.40 mg/day), doxazosin (n = 3,468, mean dose 4.1 ± 2.7 mg/day), prazosin (n = 2,058, mean dose 4.6 ± 3.5 mg/day), and alfuzosin (n = 166, mean dose 9.0 ± 3.2 mg/day).
Propensity score matching on AB use resulted in 35,713 matched pairs. The baseline characteristics and standardized differences in the matched cohort are reported in Table 1 and show that all standardized difference for variables used in the model were <0.10. The baseline clinical characteristics of the entire unmatched cohort are presented in Online Table S2. Mean LVEF was 42.0 ± 15.5%, with 41.5% of those with an EF value available having an LVEF <40%, and 62% with an EF <50%. This results in approximately 38% to 58% of the cohort representing HF with preserved EF patients, depending on the EF threshold for HF with preserved EF.
Median duration of follow-up in the propensity score–matched cohort was 834 days in AB-untreated patients and 915 days in AB-treated patients. At 2 years, there was a total of 12,428 HF readmissions in AB-untreated patients and 12,047 in AB-treated patients, and 16,289 deaths in AB-untreated patients and 14,930 deaths in AB-treated patients. The readmission rate for HF was lower in AB-treated patients, 39.8% versus 41.7% at 2 years (HR: 0.95; 95% CI: 0.92 to 0.97; p < 0.0001). The rate of death was also lower in AB-treated patients, 42.8% versus 46.5% (HR: 0.93; 95% CI: 0.91 to 0.94; p < 0.0001) (Figure 1). A separate propensity score–matched cohort of only patients with a prior diagnosis of HF yielded consistent results (data not shown).
There was no significant difference with higher AB doses (13%, n = 5,932) compared with lower AB doses (87%, n = 41,706) for HF readmission (HR: 0.98; 95% CI: 0.94 to 1.02; p = 0.35); however, there was a lower incidence of death (HR: 0.94; 95% CI: 0.91 to 0.97; p < 0.0001) (Table 2). Those patients treated with nonselective (64%, n = 30,290) versus selective (nonvasoactive) (36%, n = 17,348) ABs had a lower rate of HF readmission (HR: 0.90; 95% CI: 0.88 to 0.93; p < 0.0001) and a lower incidence of death (HR: 0.80; 95% CI: 0.78 to 0.82; p < 0.0001). Compared with untreated patients, those treated with nonselective ABs experienced a lower rate of death and HF readmission, while those treated with selective agents experienced a slightly higher rate of death, and no significant difference in HF readmission (Table 2). In a separate analysis, propensity score matching for receipt of selective versus nonselective ABs, we found results consistent with our main analysis for both endpoints (data not shown). In a secondary analysis according to individual AB agent, we found that each agent was associated with reduced mortality—terazosin (HR: 0.97; 95% CI: 0.95 to 0.98; p < 0.0001), doxazosin (HR: 0.95; 95% CI: 0.92 to 0.99; p = 0.01), prazosin (HR: 0.95; 95% CI: 0.90 to 0.99; p = 0.02)—while only terazosin was associated with a decrease in HF readmissions—terazosin (HR: 0.93; 95% CI: 0.91 to 0.95; p < 0.0001), doxazosin (HR: 0.99; 95% CI: 0.94 to 1.01; p = 0.80), prazosin (HR: 0.96; 95% CI: 0.90 to 1.03; p = 0.23).
BB treatment was received by 84.3% of patients. AB treatment was associated with a lower incidence of readmission for HF in patients receiving a BB (HR: 0.95; 95% CI: 0.93 to 0.97; p < 0.0001) and a similar but nonsignificant trend toward lower HF readmission rate in patients not receiving a BB (HR: 0.95; 95% CI: 0.90 to 1.00; p = 0.06; without an AB-BB interaction; p = 0.13). Death was lower with AB treatment in both BB-untreated (HR: 0.93; 95% CI: 0.90 to 0.96; p < 0.0001) and BB-treated (HR: 0.91; 95% CI: 0.89 to 0.92; p < 0.0001; for AB-BB interaction; p = 0.60) patients (Figure 2).
In patients receiving a BB other than carvedilol, usually metoprolol, AB treatment was associated with a lower incidence of readmission for HF (HR: 0.94; 95% CI: 0.92 to 0.96; p < 0.0001; p = 0.10 for AB-BB interaction vs. no-BB group) and death (HR: 0.90; 95% CI: 0.88 to 0.91; p < 0.0001; for AB-BB interaction; p = 0.96) with no interaction compared with BB-untreated patients for either outcome. In patients treated with carvedilol compared with BB-untreated patients, although there was a significant AB-carvedilol interaction (p = 0.02), there was no difference in HF readmissions with AB therapy (HR: 0.99; 95% CI: 0.95 to 1.02; p = 0.47. However, AB reduced the risk of death in the carvedilol-treated patients (HR: 0.95; 95% CI: 0.92 to 0.97; p < 0.0001) and BB-untreated patients (HR: 0.93; 95% CI: 0.90 to 0.96; p < 0.0001); with the effect slightly more pronounced in the BB untreated group (p = 0.01 for AB/carvedilol interaction). In those treated with HF-approved agents (carvedilol, bisoprolol, metoprolol succinate), there was no difference in HF readmissions with AB use (HR: 0.99; 95% CI: 0.96 to 1.01; p = 0.37). However, HF readmissions were lower in those treated with other BB agents (HR: 0.96; 95% CI: 0.93 to 0.98; p = 0.003; p = 0.06 for AB-BB–type interaction). AB use was associated with a lower incidence of death both with a HF-approved BB (HR: 0.94; 95% CI: 0.92 to 0.97; p < 0.0001) and other BB (HR: 0.91; 95% CI: 0.89 to 0.93; p < 0.0001; p = 0.002 for AB-BB interaction). In the falsification analysis, there was no difference in cancer admissions between AB-treated and untreated patients (HR: 0.98; 95% 0.92 to 1.06; p = 0.64).
The possible association between AB and HF is an important clinical question that has remained unresolved for nearly 2 decades, despite the frequent use of these agents in patients with cardiovascular disease and concomitant prostatic hypertrophy. The present study provides strong and consistent evidence of a lack of harmful effect of AB on recurrent HF admissions and death at the doses typically used in contemporary practice. AB treatment was in fact associated with a decrease in recurrent HF and all-cause death in our propensity-matched cohort with well-balanced baseline characteristics. The lower rate of death was confined to those treated with agents having vasoactive properties (prazosin, doxazosin, and terazosin) and was not seen with the alpha-1a receptor–specific, selective AB tamsulosin and alfuzosin. Likewise, patients treated with higher doses of AB also experienced a lower incidence of death than those who were treated with lower doses, with no increase in HF readmissions.
These findings differ from our original hypothesis and from those of the ALLHAT trial and smaller earlier trials (1–3). The ALLHAT trial showed nearly double the rate of new-onset HF in high-risk hypertensive patients treated with doxazosin compared with chlorthalidone, which resulted in the early termination of the doxazosin arm of the trial (17,18). This has been widely considered to represent a harmful effect of AB therapy, likely mediated by neurohormonal activation resulting from unopposed alpha receptor antagonism (19–21). However, without a placebo group, the findings could equally have resulted from a relative lack of the now well-established cardioprotective effect of thiazide diuretics (22) and the other active antihypertensive treatments rather than to a harm by the AB. There was no increase in incidence of HF with AB in other hypertension studies, including a large VA Cooperative hypertension study (23). In an HF population similar to that of our study, the V-HeFT I trial found that prazosin monotherapy for HF resulted in no difference in all-cause hospitalization, mortality, or exercise capacity, although HF admissions were not a study endpoint and the sample size would not have been sufficient to exclude a small increase in event rates (24,25). The neutral effect on HF in our study would also tend to support the idea that the ALLHAT trial findings were the result of a relative lack of cardioprotective effect of the AB doxazosin in relation to the other trial therapies, rather than a harm.
AB treatment was associated with a lower incidence of recurrent HF admissions and death both in BB-treated and BB-untreated patients in the entire HF cohort, in addition to only those patients with a prior HF diagnosis. Although AB therapy was associated with a lower rate of both HF readmissions and mortality with noncarvedilol BB therapy, most commonly metoprolol, only mortality and not HF readmissions was reduced with AB therapy in carvedilol-treated patients. While we did not investigate the mechanism related to these findings, it is possible that the additive alpha-blocking effect of carvedilol may not provide additional benefit to those patients already taking an AB, but also no harm.
A small prior study by Dhaliwal et al. (7) of 388 patients at a single VA center found no overall increase in HF admissions or death in patients treated with an AB, although there was a slightly higher incidence of HF readmission in the small subset not receiving a concomitant BB. However, this study was underpowered and estimates were imprecise, as shown by wide CIs around estimates. A preliminary study from our institution in a small cohort of 503 HF patients also found a lower rate of death in AB-treated patients, although a slightly higher rate of HF readmission (26). Given the marked baseline differences between the unmatched groups (see the Online Table S2), it is likely that baseline differences between groups not accounted for in the prior preliminary analysis likely resulted in residual confounding and explain the difference in results for HF readmission. The current study findings using a larger, propensity score–matched cohort better adjust for a greater number of clinical covariates with balanced baseline characteristics after matching and provide a more precise and stable estimate of treatment effect given the large number of outcome events.
A noteworthy finding in the current study is a lower rate of HF readmission and death in AB-treated individuals. Furthermore, the relationship between AB use and mortality was more pronounced with higher doses. This effect on mortality also only occurred in those treated with a vasoactive AB. Together, these observations lend strength to a hypothesis that ABs may in fact have additional clinical benefits in HF, potentially mediated via their vasodilating properties. The hypothesis was first proposed in response to the positive results of the COMET (Carvedilol Or Metoprolol European Trial) trial, which found a significant reduction in total and cardiovascular mortality with carvedilol, which has weak alpha-adrenergic blocking activity, compared with metoprolol tartrate (27). We also found that HF patients receiving a non–HF-approved BB had the same or greater association between AB and improved mortality outcomes. The finding of improved outcomes in AB-treated patients may warrant further study.
This was a study of prevalent, not incident, AB treatment. Whether there is an early hazard associated with initiation of AB treatment, particularly in the absence of concomitant neurohormonal blocking agents, cannot be ruled out. The index event for patients entering our cohort was the first VA hospital admission for HF. Although most patients had a prior HF diagnosis, either from symptoms and events not requiring hospital admission or from an admission at an outside facility, and recorded in VA clinical notes, it was the first VA HF admission. Therefore, it is difficult to compare the present study with a prospective study of patients without prior HF, such as the ALLHAT trial. Data on LV function, which may be related to prognosis, was incomplete and not used for the propensity score–adjusted analysis. Whether ABs have differential effects in patients with HF with preserved compared with those with reduced LV systolic function could not be determined accurately. Cause of death for patients dying outside of a VA facility was not available. How many of these deaths would have been counted as a HF endpoint, and the effect on the results, is not known.
The finding of improved outcomes in HF patients treated with an AB is novel. The fact that the association with lower mortality is more marked with higher doses of AB and is limited to vasoactive agents further supports this finding. Nevertheless, this subgroup finding should be viewed with caution because clinicians may have preferentially treated higher-risk patients with selective (nonvasoactive) AB out of concern for hypotension or worsening HF. Although our overall cohort was well-matched using propensity score methods on a large number of clinical and treatment characteristics, as with any observational study, there is the possibility of residual confounding caused by unobserved differences, for example, in the quality of care.
In this large, propensity score–matched cohort, we found that concomitant treatment with an AB in patients following hospitalization for HF was not associated with an increased incidence of HF readmission or death. In contrast, rates of HF readmission and death were lower. AB treatment was associated with a reduction in total mortality, with a greater reduction in those receiving higher doses, and was confined to those receiving vasoactive, nonselective AB agents, together suggesting that there may be a modest clinical benefit from AB use. Patients with HF may be safely treated with these agents where otherwise clinically indicated. The finding of improved outcomes in AB-treated patients may warrant further study.
COMPETENCY IN MEDICAL KNOWLEDGE: Patients with HF treated with an AB had lower rates of readmission for HF and all-cause death. Mortality was lower in those receiving a vasoactive, nonselective AB and higher doses of AB, and was seen regardless of concomitant BB use. ABs may be safely given to patients with HF, including those not on concomitant BB therapy.
TRANSLATIONAL OUTLOOK: Addition of an AB to conventional HF therapy is a potential therapeutic target that may warrant further investigation.
The authors thank Drs. Jeesun Cho and Tina Lee for their assistance with literature review.
Dr. Jackevicius is an associate editor for Circulation: Cardiovascular Quality and Outcomes. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- ejection fraction
- heart failure
- hazard ratio
- International Classification of Diseases-9th Revision
- left ventricular ejection fraction
- Veterans Affairs
- Received December 18, 2017.
- Revision received June 18, 2018.
- Accepted June 26, 2018.
- 2018 American College of Cardiology Foundation
- ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group
- ↵VA Informatics and Computing Infrastructure (VINCI), VA HSR HIR 08-204, U.S. Department of Veterans Affairs (2008). Available at: https://www.hsrd.research.va.gov/for_researchers/vinci/. Accessed October 12, 2016.
- Sohn M.W.,
- Arnold N.,
- Maynard C Hynes D.M.
- Kim Y.,
- Garvin J.H.,
- Goldstein M.K.,
- et al.
- Austin P.C.,
- Chiu M.,
- Ko D.T.,
- Goeree R.,
- Tu J.V.
- ↵Bergstralh, EJ Kosanke JL. GMATCH Macro. Mayo Foundation. Section of Biostatistics, 2004. Available at: http://www.mayo.edu/research/documents/gmatchsas/doc-10027248. Accessed October 12, 2016.
- Davis B.R.,
- Cutler J.A.,
- Furberg C.D.,
- et al.
- Jackevicius C.A.,
- Lu L.,
- Lee T.,
- Cho J.,
- Ghaznavi Z.,
- Warner A.