Given the concern that beta-blocker use may be associated with an increased risk for heart failure (HF) in populations with normal left ventricular systolic function, we evaluated the association between beta-blocker use and incident HF events, as well as loop diuretic initiation in the Systolic Blood Pressure Intervention Trial (SPRINT). SPRINT demonstrated that a blood pressure target of <120 mm Hg reduced cardiovascular outcomes compared with <140 mm Hg in adults with at least one cardiovascular risk factor and without HF. The lower rate of the composite primary outcome in the 120 mm Hg group was primarily driven by a reduction in HF events. Subjects on a beta blocker for the entire trial duration were compared with subjects who never received a beta blocker after 1:1 propensity score matching. A competing risk survival analysis by beta-blocker status was performed to estimate the effect of the drug on incident HF and was then repeated for a secondary end point of cardiovascular disease death. Among the 3,284 propensity score–matched subjects, beta-blocker exposure was associated with an increased HF risk (hazard ratio 5.86; 95% confidence interval 2.73 to 13.04; p < 0.001). A sensitivity analysis of propensity score–matched cohorts with a history of coronary artery disease or atrial fibrillation revealed the same association (hazard ratio 3.49; 95% confidence interval 1.15 to 10.06; p = 0.028). In conclusion, beta-blocker exposure in this secondary analysis was associated with increased incident HF in subjects with hypertension without HF at baseline.
Hypertension is prevalent and a major driver of cardiovascular disease (CVD) morbidity and mortality, including heart failure (HF), stroke, and coronary artery disease (CAD). , The Systolic Blood Pressure Intervention Trial (SPRINT) was designed to evaluate whether a systolic blood pressure (BP) target of <120 mm Hg in patients without HF and ≥1 CVD risk factor would reduce cardiovascular events. The trial was stopped early after interim analyses indicated that the primary composite outcome was reduced in the lower BP arm. Although patients with symptomatic HF were generally excluded from enrollment, the main benefit of a lower BP was a marked reduction in incident HF ( Figure 1 ). The SPRINT protocol encouraged participating investigators to use antihypertensive medications supported by evidence from large randomized trials as first-line agents. The use of beta blockers was reserved for the presence of a secondary indication. Nonetheless, at the conclusion of SPRINT, more than a third of the participants were receiving beta blockers. Data from randomized trials and secondary analyses suggest an association between beta-blocker use and decompensated HF in patient populations with predominantly normal ejection fractions (EFs). The objective of our study was to investigate whether beta-blocker exposure was a risk factor for incident HF in a trial cohort with predominantly normal EFs and cardiac risk factors.
Methods
The SPRINT design and results have been described in detail. In brief, the trial was a multicenter, open-label, randomized trial enrolling 9,361 subjects in the United States and Puerto Rico with a systolic BP of ≥130 mm Hg who had no history of diabetes or stroke but who had ≥1 CVD risk factor. Subjects were enrolled to either (1) intensive BP control with a target systolic BP of ≤120 mm Hg (lower BP arm), or (2) standard BP control with target systolic BP of ≤140 mm Hg (higher BP arm). Included subjects were ≥50 years of age and met prespecified BP criteria with accompanying requirements for number of antihypertensive medications.
CVD risk factors could include 1 or more of the following: (1) known presence of clinical or subclinical CVD (other than stroke), (2) stable chronic kidney disease with an estimated glomerular filtration rate 20 to 59 ml/min/1.73 m 2 , (3) Framingham Risk Score for 10-year CVD risk ≥15%, and/or (4) age ≥75 years. Clinical CVD included previous myocardial infarction (MI) or acute coronary syndrome (ACS), previous vascular intervention or surgery for obstructive arteriosclerosis, an established ≥50% arterial stenosis, or an abdominal aortic aneurysm ≥5 cm with or without repair. Subclinical CVD included patients with a coronary artery calcium score ≥400 Agatston units within the previous 2 years, ankle brachial index <0.90 within the previous 2 years, or left ventricular hypertrophy by electrocardiography (ECG), echocardiogram report, or other cardiac imaging modality within the previous 2 years. Participants with symptomatic HF in the 6 months preceding enrollment or who were known to have EF <35% were excluded.
A specific antihypertensive medication regimen was not prescribed. Instead, a treatment algorithm emphasizing initiation of a 2- or 3-drug regimen preferentially using a combination of a thiazide-type diuretic and/or an angiotensin-converting enzyme inhibitor (ACE-I) or angiotensin receptor blocker (ARB) and/or calcium channel blocker (CCB) was recommended based on clinical trial data. The use of beta blockers as part of the hypertension treatment regimen was only advised in specific clinical scenarios including the setting of underlying CAD, as a heart rate–controlling agent with atrial fibrillation (AF), impaired renal function, or electrolyte abnormalities that would preclude ACE-I/ARB/thiazide diuretics. Hypertension remained the primary indication for the use of all medication classes, with further selection based on concomitant co-morbidities. For study follow-up visits, titration of agents already in use or addition of preferred agents was recommended.
The baseline visit included laboratory testing, vital signs, anthropomorphic data, ECG, medical history, demographic characteristics, medication inventory, and quality of life surveys. For the next 3 months, participants returned for monthly follow-up visits, which included a comprehensive medical history, assessments of vital signs, and medication inventories for the first 3 months. Thereafter, interval study visits occurred every 3 months for the remainder of the first year, and then annually. A medication inventory was included at each visit.
The deidentified SPRINT database was obtained from the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repositories Information Coordinating Center. The database included 9,361 randomized subjects. Of those, 17 subjects were excluded based on incomplete medication inventories. Because it was our goal to examine incident HF, we also excluded the 332 subjects with a history of HF (enrolled in the trial based on a history of HF without symptoms in the preceding 6 months and with EF documented above 35%). The remaining analyzed population included 9,012 subjects as shown in Figure 2 . Baseline and follow-up medication inventories were used to determine the use of beta blockers and other major antihypertensive classes such as ACE-I/ARB, thiazides, and CCBs.
The primary end point of this secondary analysis was incident HF defined as hospitalization or emergency department visit for HF requiring intravenous loop diuretics. The secondary end point was death from CVD with causes including MI, stroke, or postmortem findings of an acute CVD event.
For the primary analysis, we compared subjects on a beta blocker for the entire trial with subjects who never received a beta blocker to assess the risk of incident HF and beta-blocker status. To correct for a maximal number of confounding variables, we created propensity score–matched (PSM) cohorts. After estimation of the propensity score, subjects within the subgroups were matched in a 1:1 ratio to the controls accounting for 27 confounding variables: age, gender, race, previous MI, previous ACS, carotid artery disease, peripheral artery disease (PAD), aortic stenosis, abdominal aortic aneurysm ≥5 cm, calcium score >400, low ankle brachial index, left ventricular hypertrophy on ECG, study arm, Framingham risk, systolic BP, diastolic BP, number of antihypertensive agents at enrollment, smoking status, aspirin use, estimated glomerular filtration rate, high-density lipoprotein, total cholesterol, body mass index, statin use, and antihypertensive class use. We used the “nearest neighbor” matching algorithm with a caliper size of 1% of the SD of the estimated propensity score to construct a matched-paired sample. This caliper size is more stringent than typically recommended for observational studies and allows for optimal matching of co-morbidities, that is, MI and antihypertensive medication use. After the PSM, we conducted a competing risk survival analysis by beta-blocker status to estimate the effect of the drug on incident HF accounting for acute MI and all-cause death as competing risk. We repeated the analysis for the secondary end point of CVD death.
The PSM analysis was repeated for the subjects with a potential beta-blocker indication such as history of MI, ACS, coronary revascularization, and history of AF or AF on the baseline ECG. To compare the impact of the other major antihypertensive classes on HF, we repeated the PSM analyses for patients who did not change ACE-I/ARB, thiazide, or CCB status during the trial. The impact of beta blockers on objective variables, such as heart rate and systolic BP, was analyzed using linear mixed-effect models controlling for multiple co-morbidities (fixed effects). As multiple data points were available for each subject, a patient identifier was used as random effect.
Results
This secondary analysis of SPRINT included 9,012 participants with a median age of 67 years (interquartile range 61 to 75). Median follow-up was 3.3 years, with 8,736 subjects (96.93%) followed after 1 year, 8,736 subjects followed after 2 years, and 6,076 subjects followed after 3 years. Of these, 4,501 (50%) had been randomized to the lower BP arm, 3,218 (36%) were women, and 3,803 (42%) were non-White ( Table 1 ). A history of acute MI was present in 576 subjects (6%), and 397 (4%) had a history of ACS. Most of the trial participants were using between 1 and 3 antihypertensive medications at baseline. Overall, 3,248 subjects (36%) were on a beta blocker at the beginning of the trial, and 2,813 (31%) received beta blockers for the entire trial. In addition, 3,284 subjects were matched using a propensity score.
| Variable | Overall cohort n (%) |
|---|---|
| (n = 9,012) | |
| Age, median [IQR] | 67.0 [61.0 to 75.0] |
| Women | 3,218 (36%) |
| Black | 2,669 (30%) |
| Hispanic | 965 (11%) |
| Other | 169 (2%) |
| White | 5,209 (58%) |
| Lower BP target arm | 4,501 (50%) |
| 10 y Framingham risk, median [IQR] | 17.7 [12.0, 25.6] |
| SBP, median [IQR], mm Hg | 138.0 [130.0, 149.0] |
| DBP, median [IQR], mm Hg | 78.0 [70.0, 86.0] |
| Number of antihypertensive agents | |
| 0 | 868 (10%) |
| 1 | 2,700 (30%) |
| 2 | 3,178 (35%) |
| 3 | 1,795 (20%) |
| 4 | 455 (5%) |
| 5 | 15 (0%) |
| 6 | 1 (0%) |
| Current smoker | 1,192 (13) |
| Aspirin | 4,530 (50) |
| Statin | 3,850 (43) |
| Cholesterol, median [IQR] | 187.0 [162.0 to 215.0] |
| BMI, median [IQR], kg/m 2 | 29.0 [25.8 to 32.9] |
| History of acute myocardial infarction | 576 (6%) |
| History of acute coronary syndrome | 397 (4%) |
| Coronary revascularization | 790 (9%) |
| LVH on ECG | 396 (4%) |
| Beta blocker | |
| At baseline | 3,248 (36%) |
| Sometimes | 4,626 (51%) |
| Never | 4,386 (49%) |
| Entire trial duration | 2,813 (31%) |
| ACE-I/ARB | |
| At baseline | 5,853 (65%) |
| Sometimes | 7,705 (85%) |
| Never | 1,307 (15%) |
| Always | 5,128 (57%) |
| CCB | |
| At baseline | 3,612 (40%) |
| Sometimes | 6,184 (69%) |
| Never | 2,828 (31%) |
| Always | 3,152 (35%) |
| Thiazide | |
| At baseline | 4,516 (50%) |
| Sometimes | 6,754 (75%) |
| Never | 2,258 (25%) |
| Always | 3,260 (36%) |
An analysis of subjects who received beta blockers for the entire trial compared with subjects who never received a beta blocker was well balanced in terms of concomitant BP medications ( Table 2 ). The types of beta blockers used and dosing was similar to the overall cohort ( Supplementary Tables 1 and 2 ). In the matched group that never received beta blockers, there were 6 incident HF events, and in the group that received beta blockers for the entire trial, there were 41 HF events. The cumulative incidence of HF and competing events are shown in Figure 3 . This analysis confirmed a positive association between beta blocker use and incident HF (hazard ratio [HR] 5.86, confidence interval [CI] 2.63 to 13.04; p <0.001) ( Table 3 ).
