Association of β-Blocker Use With Heart Failure Hospitalizations and Cardiovascular Disease Mortality Among Patients With Heart Failure With a Preserved Ejection Fraction

1Heart failure (HF) is a leading cause of hospitalizations and is associated with increased health care costs., More than half of the patients with HF have a preserved ejection fraction (HFpEF), defined as an ejection fraction (EF) of 50% or greater. Heart failure with a preserved EF continues to increase in prevalence and is associated with a high rate of hospitalization, yet, to our knowledge, evidenced-based therapies are lacking.,,,

2β-Adrenergic receptor blockers (β-blockers) provide an unequivocal benefit in the treatment of chronic HF with a reduced EF (HFrEF), with a strong foundation of evidence to support their use.,,,,,, Most patients with HFpEF enrolled in contemporary clinical trials or in published cohorts also receive β-blockers, despite an uncertain benefit.,,,,, A recent patient-based meta-analysis of 11 randomized β-blocker HF trials that enrolled patients with HFrEF also included a small number of patients with HFpEF. This analysis reinforced the benefit of β-blockers for patients with a reduced EF but did not demonstrate any benefit for patients with an EF greater than 50%.

3To extend our understanding of the role β-blockers play in HFpEF, we performed an analysis of data from participants randomized in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) clinical trial. The primary focus of this analysis was to investigate the association of β-blocker use with HF hospitalizations among patients at different EF thresholds.

4The TOPCAT trial and its design have been previously described in detail, as have its main results., The trial was an international, multicenter, double-blinded, randomized clinical trial of the aldosterone antagonist spironolactone for patients with HFpEF, defined as symptomatic HF at the time of screening and within the preceding 12 months in patients with a left ventricular EF of 45% or greater. Enrollment took place between August 10, 2006, and January 31, 2012, with mean follow-up of 3.3 years, and was based on either a hospitalization attributed to decompensated HF in the preceding year or elevated brain natriuretic peptide (BNP) or N-terminal pro BNP (NT-proBNP) levels (BNP level ≥100 pg/mL [to convert to nanograms per liter, multiply by 1.0] or NT-proBNP level ≥360 pg/mL) within 60 days of screening., The initial comprehensive baseline visit included a detailed medical history of comorbidities, social history, risk factor documentation, medication survey, quality of life questionnaire, and assessment of physical activity and medications. Not all follow-up visits included a medication inventory. The University of Vermont Institutional Review Board deemed this research to be exempt from review as it was a retrospective analysis performed on a deidentified data set. Enrolled patients provided written informed consent. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

5We used the deidentified TOPCAT database from the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repositories Information Coordinating Center. Because of reported trial quality concerns in some regions, we analyzed the data only from South America and North America.,, Patients without a baseline EF and those without a recorded baseline visit were also excluded. We used baseline and follow-up medication logs to chart β-blocker use.

6Statistical analysis was performed from January 31 to May 2, 2019. Baseline characteristics were tabulated by use of a β-blocker vs no use of a β-blocker. β-Blocker use itself was defined as the receipt of any β-blocker at the baseline visit. To assess the association between β-blockers and incident HF hospitalizations, hazard ratios were calculated comparing baseline β-blocker use vs no β-blocker use and stratified by an EF of 50% or greater or less than 50%. This threshold was chosen because it allows for a clinically meaningful separation between HF phenotypes as recommended by current guidelines., In a minimally adjusted covariate model, we corrected for age, sex, race/ethnicity, and treatment assignment (spironolactone or placebo). In the fully adjusted model, baseline myocardial infarction, atrial fibrillation, chronic obstructive pulmonary disease, asthma, and hypertension were added to the minimally adjusted model. This analysis was repeated for cardiovascular disease (CVD) mortality. Schoenfeld residuals were examined to confirm no violation of the assumptions of the Cox proportional hazards regression model.

7The association between level of baseline EF and relative hazard of HF hospitalization for those receiving β-blockers or not in the fully adjusted model was presented using restricted cubic spline models with 95% CIs relative to the median. Knots were not prespecified and were chosen using the Harrell method. The distribution of the baseline EF in both groups by HF hospitalizations was visualized using kernel density plots.

8We performed a sensitivity analysis among the participants whose β-blocker status was constant from baseline through each follow-up visit, censoring at the end of follow-up or at the first HF hospitalization. In this population with consistent documentation of β-blocker or no β-blocker use, we calculated relative hazards using the unadjusted, minimally adjusted, and fully adjusted models for HF hospitalizations overall and in the prespecified EF strata. In a second sensitivity analysis among the participants whose β-blocker status was constant, propensity scores were created using 30 baseline demographic, anthropometric, medication, and medical history patient characteristics. Up to 3 patients receiving β-blockers were matched to each patient not receiving a β-blocker. Heart failure hospitalization was analyzed for the propensity score–matched data using stratified Cox proportional hazards regression, with the matched sets as the stratifying variable.

9To provide some pathomechanistical insights, we compared BNP and NT-proBNP levels at baseline by β-blocker status overall and stratified by an EF of 50%. We compared log-transformed baseline levels of BNP and NT-proBNP using unadjusted 2-tailed t tests. We visualized distributions of the quartiles of the BNPs and NT-proBNPs and the midrange of each quartile among those with an EF of 50% or greater using stacked bar graphs.

10The analyses were performed with Stata MP, version 15.1 (StataCorp) and the SAS, version 9.4 procedure PSMATCH (SAS Institute Inc) for the propensity score matching. We considered a 2-tailed P <.05 to be statistically significant.

11Among the 1767 TOPCAT participants from North America and South America, the mean (SD) age was 71.5 (9.6) years; 879 participants were women and 882 were men, and 1378 participants were white. Six participants were excluded because of missing EF data or missing baseline visit data (Figure 1). The final analytic population was 1761. Median follow-up was 2.4 years (interquartile range, 1.4-3.9 years).

12A total of 1394 participants (79.2%) were receiving β-blockers at baseline and 1567 (89.0%) had an EF of 50% or greater (Table 1). Irrespective of β-blocker use, most patients had preexisting hypertension. The prevalence of atrial fibrillation was 42.6% (594 of 1393) among those receiving a β-blocker and 40.7% (149 of 366) among those not receiving a β-blocker, and the prevalence of prior myocardial infarction was 22.1% (308 of 1395) among those receiving a β-blocker and 13.7% (46 of 336) among those not receiving a β-blocker. The proportion of patients with an advanced functional impairment (New York Heart Association class ≥3) was almost identical between groups.

13Overall, 399 patients (22.7%) underwent hospitalization for HF (β-blocker group, 344 of 1394 [cumulative incidence, 24.7%; unadjusted incidence rate, 96 per 1000 person-years; 95% CI, 86-106]; no β-blocker group, 55 of 367 [cumulative incidence, 15.0%; unadjusted incidence rate, 56 per 1000 person-years; 95% CI, 43-73]). In the fully adjusted model, there was a higher incidence of HF hospitalizations among patients with an EF of 50% or greater who were receiving β-blockers (Table 2) compared with those not receiving β-blockers (hazard ratio, 1.74 [95% CI, 1.28-2.37]; P <.001) (Figure 2). There was a significant interaction between an EF threshold, β-blocker use, and incident HF hospitalization (P =.03); higher EFs were associated with an increased risk for HF hospitalizations among patients receiving β-blockers (Figure 3). Patients with the highest EFs who did not receive β-blockers were least likely to be admitted for HF. There was a nonsignificant trend toward a lower incidence of HF hospitalizations among patients with an EF between 45% and 49% who were receiving β-blockers (hazard ratio, 0.68 [95% CI, 0.28-1.63]; P =.39). Cardiovascular disease mortality occurred among 229 participants (13.0%); β-blockers were not significantly associated with a change in CVD mortality (eTable 6 and eFigure 1 in the Supplement).

14The sensitivity analysis that considered HF hospitalizations among patients who either continued receiving a β-blocker or who never received a β-blocker through the end of follow-up or first HF hospitalization confirmed the principal findings (eTable 1 in the Supplement). In the fully adjusted model, patients with an EF of 50% or greater receiving β-blockers had a higher relative hazard of HF hospitalization compared with those not receiving β-blockers (hazard ratio, 1.76 [95% CI, 1.22-2.53]; P =.007). In a second sensitivity analysis of propensity score–matched cohorts, β-blocker use was also associated with more HF hospitalizations among patients with EFs of 50% or greater (eTable 2 and eTable 3 in the Supplement). This analysis also confirmed that patients receiving β-blockers who had higher EFs had an associated risk of being hospitalized for HF.

15β-Blocker use was associated with higher NT-proBNP and BNP levels in patients with an EF of 50% or greater (eTable 5 and eFigure 2 in the Supplement). This was not the case for patients with an EF between 45% and 49%.

16To date, the efficacy of β-blockers for patients with HFpEF is unknown. This post hoc analysis of the TOPCAT trial of spironolactone for patients with HFpEF suggests an association between β-blocker use and incident HF hospitalizations for patients with an EF of 50% or greater and an incremental positive association between β-blocker use and the risk for HF hospitalization at higher EF thresholds. There was no significant association between β-blocker use and CVD mortality.

17The TOPCAT trial corroborates the finding that most patients in contemporary HFpEF cohorts are treated with β-blockers. A total of 79.2% of the patients in the North American and South America cohort of this trial received β-blockers at baseline. This high prevalence of β-blocker use is similar to other HFpEF studies, as shown in eTable 4 in the Supplement.,,,, Although evidence for the benefits of β-blocker use for patients with HFpEF is lacking, this high rate of use is most likely explained by an assumption that β-blockers are efficacious for treating common comorbidities such as hypertension, coronary artery disease, and atrial fibrillation.

18Extensive, high-quality evidence supports the use of β-blockers for patients with HFrEF.,,,,,, In addition, several recent analyses have investigated whether guideline-directed HF therapies have a utility for patients with EFs between 40% and 49%, an entity termed HF with midrange EF.,, A recent individual patient-level meta-analysis of 11 major HFrEF trials investigating the effects of β-blockers at different ranges of EF identified a reduction in CVD mortality among patients with an EF between 40% and 49%. Analogous to our findings, the same meta-analysis suggested that patients with an EF of 50% or greater did not see any benefits from being randomized to receive β-blockers. However, only 244 patients fell into this category.

19Our observations contrast with prior randomized β-blocker trials, which did not report increased HF hospitalizations among patients with HFpEF. However, there are several points to consider. First, to our knowledge, only 2 randomized clinical outcome trials have been performed that studied β-blockers in patients with HFpEF. The larger SENIORS trial (Randomized Trial to Determine the Effect of Nebivolol on Mortality and Cardiovascular Hospital Admission in Elderly Patients With Heart Failure) considered an EF greater than 35% to define HFpEF and analyzed 752 participants with an EF in this range, among whom approximately half had an EF between 35% and 50%. In addition, HF hospitalizations were not specifically recorded. The open-label Japanese Diastolic Heart Failure (J-DHF) trial of carvedilol (mean dose, 8.5 mg/d) randomized 245 patients with an EF greater than 40%. It appears possible that assessing HF hospitalizations in mixed populations with a reduced and normal EF may result in an overall neutral or even beneficial effect associated with β-blockers., Furthermore, patients with a history of HF and a recovered EF have recently been shown to gain sustained benefits from guideline-directed HFrEF therapies, which include β-blockers. Their inclusion in HFpEF cohorts—although likely small in numbers—could also be associated with beneficial HF outcomes.

20Evidence that may help explain these findings comes from related patient populations with preserved EFs. Specifically, the LIFE (Losartan Intervention For Endpoint Reduction in Hypertension) trial and contemporary randomized myocardial infarction trials,, have raised concerns that β-blocker use is associated with adverse cardiovascular outcomes, including an increased risk of developing HF. Mechanistically, this risk was explained by an increase in central blood pressure by reflected pressure waves. In addition, prolonged diastolic filling increases ventricular volumes and pressures, increasing the ventricular load., These mechanisms combine to increase myocardial wall stress, which may explain why, in historical hypertension trials, BNP and NT-proBNP levels were found to be elevated in patients receiving β-blockers.

21Similarly, in the TOPCAT trial, β-blocker use was associated with higher levels of circulating BNPs and NT-proBNPs in patients with a normal EF. The same was seen in the ELANDD (Effects of the Long-term Administration of Nebivolol on the Clinical Symptoms, Exercise Capacity, and Left Ventricular Function of Patients With Diastolic Dysfunction) study, which assessed the effect of nebivolol on clinical symptoms and exercise capacity in patients with diastolic dysfunction, and in the CIBIS-ELD (Titration to Target Dose of Bisoprolol vs Carvedilol in Elderly Patients With Heart Failure) study, which evaluated the tolerability of β-blocker up-titration in patients with HFpEF., In the SWEDIC (Swedish Doppler-Echocardiographic) study of patients with HFpEF randomized to receive carvedilol or placebo, BNP levels also increased, and the authors noted an unexpected worsening in HF in the patients treated with carvedilol. As has been recently demonstrated, β-blocker cessation for patients with stable HFpEF is safe and leads to marked reductions in NTpro-BNP levels.

22Although incremental increases in BNP and NT-proBNP levels have been associated with worse outcomes for patients with HFpEF—including in another secondary analysis of the TOPCAT trial—it is not clear whether this predictive capacity is preserved if indeed modified by β-blockers. However, in contrast with the observations of patients with HFpEF, BNP and NT-proBNP levels are markedly lowered by sustained β-blocker use among patients with HFrEF.,

23Our examination has several limitations. Participants in the TOPCAT trial were randomized to receive spironolactone and not β-blockers. Our adjustments may not sufficiently correct for all confounding variables, and some confounders may be unidentified. We also cannot account for both duration and intensity of β-blocker exposure. This secondary analysis can be viewed only as explorative and hypothesis generating because it does not establish cause and effect.

24These results demonstrate that β-blocker use in the TOPCAT trial cohort was associated with a higher risk for incident HF hospitalization among patients with an EF of 50% or greater, without an associated change in CVD mortality. Future studies are needed to prospectively assess the effects of β-blockers in HF populations with a normal EF.