Effect of Baseline or Changes in Adrenergic Activity on Clinical Outcomes in the β-Blocker Evaluation of Survival Trial
Background— Adrenergic activation is thought to be an important determinant of outcome in subjects with chronic heart failure (CHF), but baseline or serial changes in adrenergic activity have not been previously investigated in a large patient sample treated with a powerful antiadrenergic agent.
Methods and Results— Systemic venous norepinephrine was measured at baseline, 3 months, and 12 months in the β-Blocker Evaluation of Survival Trial (BEST), which compared placebo treatment with the β-blocker/sympatholytic agent bucindolol. Baseline norepinephrine level was associated with a progressive increase in rates of death or death plus CHF hospitalization that was independent of treatment group. On multivariate analysis, baseline norepinephrine was also a highly significant (P<0.001) independent predictor of death. In contrast, the relation of the change in norepinephrine at 3 months to subsequent clinical outcomes was complex and treatment group-dependent. In the placebo-treated group but not in the bucindolol-treated group, marked norepinephrine increase at 3 months was associated with increased subsequent risks of death or death plus CHF hospitalization. In the bucindolol-treated group but not in the placebo-treated group, the 1st quartile of marked norepinephrine reduction was associated with an increased mortality risk. A likelihood-based method indicated that 18% of the bucindolol group but only 1% of the placebo group were at an increased risk for death related to marked reduction in norepinephrine at 3 months.
Conclusions— In BEST, a subset of patients treated with bucindolol had an increased risk of death as the result of sympatholysis, which compromised the efficacy of this third-generation β-blocker.
Received April 23, 2003; de novo received December 30, 2003; revision received March 23, 2004; accepted April 29, 2004.
Most1–3 but not all4 studies indicate that adrenergic activity is a major determinant of outcome in chronic heart failure (CHF). In addition, cardiac adrenergic activity is the first neurohormonal marker that becomes elevated in subjects with left ventricular dysfunction.5 These observations form the cornerstone of the rationale for β-blocker therapy of heart failure.6
On the other hand, adrenergic support is an important compensatory mechanism in the failing heart, serving to maintain resting myocardial performance in a relatively normal range.7 When adrenergic drive is rapidly reduced in subjects with CHF, myocardial function may worsen,8 and treatments that substantially lower adrenergic drive may increase serious adverse events,9,10 including death.9 On the basis of these observations, it appears that “sympatholytic” pharmacological lowering of adrenergic activity may affect heart failure natural history differently from β-blockade.
Although baseline adrenergic activity has been examined in numerous CHF outcome studies1–4 as well as in clinical trials,11–14 until recently14 only relatively small numbers (typically hundreds) of subjects have been investigated in these studies. In addition, the relation of temporal behavior of norepinephrine as a potential determinant of natural history has been examined in only two other trials12,14 and never in a large CHF-cohort, placebo-controlled study that uses a powerful antiadrenergic agent. Thus, we investigated the effects of baseline levels and changes in adrenergic activity on clinical outcomes in the β-Blocker Evaluation of Survival Trial (BEST)15 and the interaction of bucindolol, a β-blocker with sympatholytic properties,15,16 on clinical outcomes.
The BEST protocol and the main outcomes have been previously described.15,17 Because of an initial delay in setting up the procedures, collection of blood samples for norepinephrine in all randomized patients began 6 months after trial initiation. As a result, 2126 of the 2708 randomly assigned subjects in BEST had at least a baseline norepinephrine sample collected and measured.
Norepinephrine Sample Collection and Measurements
Peripheral venous norepinephrine samples were drawn at baseline, 3 months, and 12 months by inserting a 21-gauge butterfly needle into an arm vein and placing the subject in a quiet room in a supine position for 30 minutes. The initial 3 mL of blood was discarded, and then 5 mL of blood was withdrawn and immediately transferred to prechilled 5-mL tubes containing EDTA. Within 30 minutes, plasma was separated and frozen at −70°C. Sites shipped samples on dry ice to a central laboratory (LabCorp) every 3 months, where the samples were stored at −85°C and assayed within 3 weeks. Norepinephrine was measured by HPLC-electrochemical detection, using the Bio-Rad HPLC method (Bio-Rad Laboratories). Quality control included remeasuring all samples with initial values of <200 pg/mL or >2000 pg/mL (from the second stored tube) and routinely (every 20 samples) measuring known amounts.
Means and standard deviations for continuous data and proportions or percentages for categoric data are presented. Wilcoxon rank sum tests or t tests were used for continuous data and χ2 or Fisher’s exact test for categoric data. An α level of 0.05 (2-tailed, unadjusted) was used to indicate statistical significance.
Norepinephrine levels at baseline or the change at 3 months was used to predict survival and the combined end point of death plus CHF hospitalization. Absolute and log-transformed data were initially analyzed. Because of skewness in norepinephrine levels, natural log (Ln)-transformed data were used in multivariate Cox proportional hazards regression models.
A maximum likelihood-based method18 was used to categorize changes in norepinephrine into 3 groups for prediction of death or death plus CHF hospitalization. This partitioning method finds the optimal split of norepinephrine values that maximize the likelihood of the resulting Cox proportional hazards model. In addition, a flexible cubic spline analysis19 was used to determine the shape and significance level of the relation of norepinephrine changes at 3 months to survival.
The baseline demographic and population descriptor data in subjects in whom at least a baseline norepinephrine was drawn were not different from the entire study population.15
Baseline norepinephrine mean values were 501±316 pg/mL in the placebo group (n=1061) and 529±370 pg/mL in the bucindolol group (n=1065, P=0.061 versus placebo). By paired t analysis at 3 months (P=0.0085) and 12 (P=0.0002) months, the placebo group exhibited a statistically significant increase in norepinephrine, whereas the bucindolol group exhibited significant decreases at 3 months (P=0.0001) and a trend (P=0.067) for a decrease at 12 months (Figure 1). Between-group changes in norepinephrine were statistically significant at 3 months (P<0.0001) and 12 (P<0.0001) months. Relative to changes in the placebo group, the decrease in norepinephrine in the bucindolol group was by 19% and 13% at 3 and 12 months, respectively.
Baseline Norepinephrine as a Predictor of Death or the Combined End Point of Death Plus CHF Hospitalization
In Figure 2A, hazard ratios (HR) are plotted for total mortality risk for baseline norepinephrine values, by quartiles relative to the 1st quartile assigned an HR of 1.0. For the entire cohort and for each treatment group, there is a progressive increase in mortality risk with increasing quartile. Similar results were obtained for the combined end point of death plus CHF hospitalization (Figure 2B).
Table 1 gives the univariate and multivariate analyses of baseline norepinephrine and other protocol-prespecified potential modifiers of mortality rates. Ln norepinephrine yielded a univariate HR (95% confidence limits) of 1.82 (1.58 to 2.09, P<0.001). On multivariate analysis, Ln norepinephrine was among the most powerful predictors of death.
Change in Norepinephrine as a Predictor of Death or the Combined End Point of Death Plus CHF Hospitalization
The relations of quartile changes in norepinephrine at 3 months to subsequent death or death plus CHF hospitalization are shown in Table 2, in which HRs are calculated relative to the 1st quartile of change. The quartile analysis was performed to keep the norepinephrine change/quartile the same in the placebo and bucindolol groups, with the cut-points derived from the entire cohort. This created 2 quartiles of norepinephrine reduction (1st and 2nd) and 2 of norepinephrine increase (3rd and 4th). Both absolute norepinephrine change in picograms per milliliter and percent change from baseline value are given in Table 2. Because of the sympatholytic effect of bucindolol, there were more bucindolol-treated patients in the 1st quartile and more placebo-treated patients in the 4th quartile.
As can be observed in Table 2, for absolute norepinephrine change versus death, the placebo group exhibited a trend for an increased risk in the 4th/1st quartile, with an HR of 1.38 (P=0.099) and no trends for differences in mortality rates in the 2nd or 3rd quartiles relative to the 1st. For death plus CHF hospitalization, in the placebo group, the 4th/1st quartile had a significant HR of 1.46 (P=0.011). In contrast, the bucindolol group exhibited no trends for an increased risk in the 4th/1st quartiles for either clinical outcome but a decreased risk for death in the 3rd quartile relative to the 1st (HR 0.66, P=0.046) and a trend (P=0.22) for a decreased risk in the 3rd/1st quartile for death plus CHF hospitalization.
For norepinephrine percent change, there were increases or trends for increases in risk in the placebo 3rd/1st and 4th/1st quartiles for both death and death plus CHF hospitalization. In contrast, in the bucindolol group, there were no such trends for an increased HR in the 3rd or 4th quartile relative to the 1st for either clinical end point, and, similar to the absolute norepinephrine change, there was a trend for a decreased HR (0.77) in the 3rd/1st quartile (P=0.21).
Table 2 also gives HRs by treatment group, expressed as bucindolol/placebo, for each norepinephrine quartile by absolute or percent change. For mortality rate, the bucindolol/placebo HR was significantly less than unity (reduction in mortality rate by bucindolol compared with placebo) in the 3rd quartiles for either absolute (HR=0.63) or percent (HR=0.56) norepinephrine change. For death plus CHF hospitalization, a similar pattern was observed, except that HRs in the 4th quartiles were also significantly reduced. In contrast to death, for death plus CHF hospitalization, the 2nd quartile yielded a nearly significant (P=0.067) increase in the bucindolol/placebo HR for absolute change and a significant (P=0.021) increase (HR=1.39) for percent change.
To further explore the treatment-associated differential mortality risk associated with norepinephrine change, a likelihood-based method18 was used. As shown in Figure 3, separate likelihood analysis within each treatment group identified 11 subjects in the placebo group and 153 subjects in the bucindolol group who were at respective higher risks (HR=3.31, P=0.004; HR=1.69, P=0.002) of subsequent death with norepinephrine reduction at 3 months. The reductions in norepinephrine in these risk groups were by ≥783 pg/mL in the placebo group and ≥244.5 pg /mL in the bucindolol group. Figure 3 also illustrates that subgroups with an increase in norepinephrine at 3 months were identified to be at higher mortality risk in both treatment groups.
Because the likelihood-based method provides maximal optimization of norepinephrine change cut-points predictive of increased mortality rates, we used less discriminatory fitting using flexible cubic spline fitting.19 The best fit by this method was a U-shaped, nonlinear curve with 5 knots and 3 degrees of freedom, with respective χ2 values for the bucindolol-treated group, placebo-treated group, and entire cohort of 13.2 (P=0.0042), 11.1 (P=0.011), and 32.5 (P<0.0001).
Characteristics of Subjects With an Increase or Decrease in Norepinephrine Associated With Increased Mortality Risk
Characteristics of the high-risk-mortality subgroups identified at both ends of the norepinephrine change spectrum by likelihood-based analysis, compared with the respective intermediate change groups serving as controls, are shown in Table 3. The 153 subjects in the bucindolol subgroup identified at higher mortality risk with norepinephrine reduction had high baseline norepinephrine levels and an average decrease in norepinephrine at 3 months of 529 pg/mL. These subjects also had lower left ventricular and right ventricular ejection fractions (LVEF and RVEF; data not shown) and higher heart rates compared with the intermediate-change control group, which had little or no norepinephrine change (−44 pg/mL). The 153 bucindolol-treated subjects with marked norepinephrine reduction also had a higher percentage of class IV subjects and a trend (P=0.088) toward more black versus nonblack subjects as compared with the intermediate-change group. Of the 52 deaths that occurred in these 153 subjects, 79% were classified as cardiac and 63%, 27%, and 2% were attributed to sudden cardiac death, pump failure, and myocardial infarction, respectively. In contrast, the subgroup treated with bucindolol that had a higher mortality risk associated with an increase in norepinephrine (n=137) had lower baseline RVEFs and similar baseline LVEFs but a significantly lower LVEF increase at 3 months compared with the intermediate-change group. In this subgroup, the percentage of class IV and nonblack/black distribution did not differ from the intermediate group. In this subgroup, 35 of the 43 deaths were cardiovascular, but the minority were sudden (34% versus 51% pump failure and 6% caused by myocardial infarction).
Baseline norepinephrine data from the BEST Trial confirm and extend previous reports1–3,11–14 of a positive relation between level of adrenergic activation and adverse clinical outcomes. The data on baseline norepinephrine indicate that this parameter is as strong a predictor of clinical outcomes as has been identified in a CHF population. Surprisingly, in BEST, the increased risk conferred by a higher baseline norepinephrine level was not substantially lowered by antiadrenergic therapy, as death or death plus CHF hospitalization hazard ratios progressively increased with increasing norepinephrine quartile in both the bucindolol and placebo treatment groups. One possibility for this lack of protective effect by bucindolol in the higher baseline norepinephrine quartiles was a sympatholytic effect occurring in subjects with the most advanced CHF and the greatest degree of myocardial dysfunction.
On the other hand, bucindolol conferred a clinically protective effect in quartiles of patients exhibiting an increase in adrenergic activity at 3 months. No such reduction in clinical end points was observed in quartiles of norepinephrine reduction. In fact, for death plus CHF hospitalization, the 2nd quartile of norepinephrine reduc-tion exhibited evidence of increased risk in bucindolol-treated patients. Moreover, when quartiles of norepinephrine 3-month change were referenced to the 1st quartile (which had the greatest degree of reduction), the 3rd/1st quartile relation exhibited evidence of an increase in mortality rate in the 1st quartile for the bucindolol group but not for the placebo group. These suggestions of an adverse effect of bucindolol in patients exhibiting a reduction in norepinephrine at 3 months prompted additional analyses of the sympatholytic effects of this unique β-blocking agent.
Compared with placebo, bucindolol reduced norepinephrine by 19% at 3 months. This compares to a 24% relative reduction in norepinephrine at 3 months by the central sympatholytic agent moxonidine in the Moxonidine Congestive Heart Failure (MOXCON) Trial.9 As in MOXCON,9 the sympatholytic effects of bucindolol appeared to be associated with an increased risk for adverse clinical outcomes, particularly for sudden death. In addition to the evidence within quartiles of norepinephrine reduction discussed above, likelihood-based analysis identified 18% of the bucindolol group with a marked norepinephrine reduction (by >224 pg/mL), who had a 1.69-fold increased risk for death, whereas only 1% of the placebo-treated patients were identified as being at increased risk for death with marked norepinephrine reduction. This analysis also revealed an increased risk for death in patients with an increase in norepinephrine but in similar numbers of bucindolol-treated and placebo-treated patients. The increased risk of death at both ends of the spectrum of 3- month norepinephrine change was confirmed by flexible cubic spine fitting, which yielded a statistically significant U-shaped curve for both the bucindolol-treated and placebo-treated groups.
The subgroup of bucindolol-treated subjects with a reduction in norepinephrine identified by likelihood analysis to be at increased risk of death comprised patients with more advanced (class IV versus III) heart failure, higher baseline norepinephrine levels, more depressed LV and RV function, and a trend for a greater proportion of blacks versus nonblacks. Thus, the sympatholytic effects of bucindolol probably led to adverse outcomes in a subset of subjects with severe myocardial dysfunction who were dependent on adrenergic activity for cardiac functional support; however, such a mechanism has not been proved by our data, and other explanations are possible.
The only previously published clinical trial data on the relation of changes in systemic adrenergic activity to outcomes are from Cooperative New Scandinavian Enalapril Survival Study (CONSENSUS),12 in which neurohormonal changes at 6 weeks were unrelated to outcome in 239 subjects, and the Valsartan Heart Failure Trial (Val-HeFT),14 in which absolute changes in norepinephrine at 4 months in 4301 patients did not predict differences but percent changes did predict differences in subsequent deaths in both the placebo-treated and valsartan-treated groups. However, unlike in Val-HeFT,14 we found a positive relation between increasing absolute levels of norepinephrine and increasing rates of death or death plus CHF hospitalization risk. The major new finding of the current study is that both decreases and increases in adrenergic activity can be associated with adverse clinical outcomes in a CHF population. The adverse effects of increases in norepinephrine can be abrogated by bucindolol, as opposed to the risks conferred by baseline norepinephrine.
In summary, a comprehensive investigation of systemic adrenergic activity as estimated from peripheral venous norepinephrine levels measured in the BEST Trial indicates that in advanced CHF, (1) baseline norepinephrine is a predictor of adverse clinical outcomes but not therapeutic response, (2) both increases and decreases in norepinephrine at 3 months predict adverse outcomes, and (3) bucindolol mitigates the risk of increases in norepinephrine, but through its sympatholytic properties places certain types of patients at clinical risk from reductions in norepinephrine.
This trial was sponsored by the Department of Veterans Affairs Cooperative Studies Program (DVACSP) and the National Heart, Lung, and Blood Institute through an interagency agreement. Additional support was provided by Incara Pharmaceutical Company. Lori Planting of the DVACSP provided extensive administrative assistance in the preparation of the manuscript.
While this manuscript was in review, bucindolol was licensed from Incara and Indevus Pharmaceutical Company to Arca Discovery, Inc, in which Dr Bristow and the University of Colorado have equity interests.
Rundqvist B, Elam M, Bergmann-Sverrisdottir Y, et al. Increased cardiac adrenergic drive precedes generalized sympathetic activation in human heart failure. Circulation. 1997; 95: 169–175.
Bristow MR. β-Adrenergic receptor blockade in chronic heart failure. Circulation. 2000; 101: 558–569.
Swedberg K, Bristow MR, Cohn JN, et al. The effects of moxonidine SR, an imidazoline agonist, on plasma norepinephrine in patients with congestive heart failure. Circulation. 2002; 105: 1797–1803.
Benedict CR, Shelton B, Johnstone DE, et al. Prognostic significance of plasma norepinephrine in patients with asymptomatic left ventricular dysfunction. Circulation. 1996; 94: 690–697.
Swedberg K, Eneroth P, Kjekshus J, et al. Hormones regulating cardiovascular function in patients with severe congestive heart failure and their relation to mortality. Circulation. 1990; 82: 1730–1736.
Anand IS, Fisher LD, Chiang YT, et al. Changes in brain natriuretic peptide and norepinephrine over time and mortality and morbidity in the Valsartan Heart Failure Trial (Val-HeFT). Circulation. 2003; 107: 1278–1283.
Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. New York, NY: John Wiley and Sons; 1980.
Green PJ, Silverman BW. Nonparametric Regression and Generalized Linear Models: A Roughness Penalty Approach. London: Chapman and Hall; 1994.