Author + information
- Received September 2, 2008
- Revision received January 20, 2009
- Accepted February 3, 2009
- Published online June 23, 2009.
- Alain Cohen-Solal, MD, PhD⁎,⁎ (, )
- Damien Logeart, MD⁎,
- Bidan Huang, PhD†,
- Danlin Cai, PhD†,
- Markku S. Nieminen, MD‡ and
- Alexandre Mebazaa, MD, PhD⁎
- ↵⁎Reprint requests and correspondence:
Prof. Alain Cohen-Solal, Département de Cardiologie, Hôpital Lariboisière, Universite Denis Diderot, INSERM U 942, 2 Rue A. Paré, Paris, 75475 Cedex 10, France
Objectives The purpose of this analysis was to examine whether decreases in B-type natriuretic peptide (BNP) levels during the first few days of hospitalization were associated with greater survival in patients with severe acutely decompensated heart failure (ADHF).
Background BNP level is a prognostic marker for all-cause mortality (ACM) in ADHF; whether early BNP changes can also help predict outcome in patients who need inotropes for treatment of severe ADHF is not known.
Methods We retrospectively assessed the association between changes in BNP levels and ACM in patients from the SURVIVE (Survival of Patients with Acute Heart Failure in Need of Intravenous Inotropic Support) trial—a ran-domized, controlled trial comparing levosimendan to dobutamine treatment in patients hospitalized with ADHF. BNP levels were measured at baseline and at days 1, 3, and 5. A patient was classified as a “responder” if the follow-up BNP level was ≥30% lower than baseline BNP. The relationship between early BNP response and subsequent ACM over short- (31-day) and long-term (180-day) intervals was evaluated.
Results Of 1,327 SURVIVE patients, this analysis included 1,038 who had BNP samples at both baseline and day 5. Responders at days 1, 3, and 5 had lower ACM than did nonresponders (p ≤ 0.001), with day-5 levels showing superior discriminating value. Short-term ACM (31-day) risk reduction was 67% in day-5 BNP responders compared with nonresponders, whereas long-term (180-day) ACM risk reduction was 47%.
Conclusions Patients with lowered BNP on treatment for ADHF had reduced mortality risks (31- and 180-day) compared to those with little or no BNP decrease. These results suggest that early lowering of BNP predicts both short- and long-term mortality risks. BNP reduction may therefore serve as a suitable prognostic marker of ACM. (Survival of Patients With Acute Heart Failure in Need of Intravenous Inotropic Support [SURVIVE]; NCT00348504)
Natriuretic peptides (B-type natriuretic peptide [BNP] or N-terminal pro–B-type natriuretic peptide [NT-proBNP]) have rapidly gained popularity as biomarkers of diagnostic and prognostic value in patients with acute and chronic heart failure (1–4). Previous studies found that discharge BNP values were strong predictors of subsequent outcomes in patients admitted for acute decompensation of heart failure (ADHF) (5). However, it is not known whether this association holds true for ADHF patients who have severe hemodynamic compromise and need inotrope therapy. Moreover, since such patients have high in-hospital mortality rates, particularly within the first few days of admission, BNP changes during the early treatment interval may help assess in-hospital responses to therapy and predict outcomes as well.
Our current report describes a retrospective analysis of results from patients in the SURVIVE (Survival of Patients with Acute Heart Failure in Need of Intravenous Inotropic Support) trial. SURVIVE is the largest randomized, double-blind study to date assessing the effect of treatment on short- (1-month) and long-term (6-month) mortality in patients who were hospitalized with severe ADHF (6). We sought to determine whether early changes in plasma BNP levels—within the first 5 days of hospitalization in the acute care ward—could predict short- and long-term mortality for severe ADHF patients.
SURVIVE was a randomized, multicenter, international, double-blind, double-dummy, parallel group, active-controlled trial of levosimendan versus dobutamine in patients with severe ADHF (6). The primary end point was all-cause mortality (ACM) at 180 days. Change in BNP concentration from initiation of study drug to 24 h was a secondary end point, with BNP levels at 72 and 120 h as additional efficacy parameters. Evaluation of mortality at days 1, 5, and 14, in addition to pre-specified analyses at days 31 and 180 were included as post hoc analyses. Patients enrolled in the SURVIVE trial had a left ventricular ejection fraction ≤30% within the prior 12 months and were in need of inotropic support, defined as an insufficient response to intravenous diuretics and/or vasodilators with at least 1 of the following: 1) oliguria; 2) dyspnea at rest or on mechanical ventilation; or 3) pulmonary-capillary wedge pressure ≥18 mm Hg or cardiac index ≤2.2 l/min. Exclusion criteria were a systolic blood pressure <85 mm Hg; heart rate ≥130 beats/min for ≥5 min; concurrent inotropic agents, except digitalis or dopamine (≤2 μg/kg/min); serum creatinine >5.0 mg/dl or on dialysis; significant hepatic impairment (investigator discretion); or severe obstruction of ventricular outflow tracts.
Patients received either levosimendan or dobutamine as an intravenous infusion. Levosimendan was administered for 24 h, first as a loading dose of 12 μg/kg over 10 min, followed by a continuous infusion at 0.1 μg/kg/min for the next 50 min and 0.2 μg/kg/min for the remaining 23 h, as tolerated. Dobutamine was administered as long as clinically indicated (minimally 24 h) with a beginning infusion rate of 5 μg/kg/min and upward titration to a maximum of 40 μg/kg/min, as tolerated, to achieve clinical goals.
BNP methodology and analyses
Plasma samples for BNP assay were collected at baseline, 24 h, 3 days, and 5 days after the start of the study drug infusion. BNP values (pg/ml) were determined using the AxSYM BNP immunoassay (iBNP, Abbott Diagnostics, Abbott Park, Illinois). Within-run and total percent coefficient of variation for this test are between 6.0% and 8.5%. Change from baseline BNP concentration included only data from patients with BNP assay results available at both baseline and the specific sample time point evaluated.
To differentiate short-term variations in BNP levels from clinically meaningful changes (7–9), we examined the definition of responderas a patient with a decrease from baseline BNP value ≥30% at a given sample time point (day 1, 3, or 5) post hoc. This cutoff of ≥30% decrease in BNP to define and differentiate responders from nonresponders was tested by receiver-operator characteristic (ROC) analysis of data and indeed optimally separated patients with low and high mortality rates. The Cox proportional hazards model analysis was performed for time to ACM by 31 and 180 days with BNP status as the factor. BNP status by treatment interaction was examined in the model prior to fitting the above model and was found not statistically significant.
As previous studies have suggested that absolute BNP value at discharge also has high prognostic value (5,10), we repeated the analysis defining responders as those with BNP lowered below the cutoff level of 800 pg/ml at day 5; this cutoff was previously found to be associated at discharge with a 75% rate of death or rehospitalization (5). Similar unadjusted Cox proportional hazards model analysis was also performed for time to ACM by 31 and 180 days with the factor for BNP response status based on day 5 absolute BNP value.
We also assessed the relationship between BNP response status and ACM at days 31 and 180 while controlling for important baseline characteristics (systolic blood pressure, diastolic blood pressure, laboratory creatinine value, BNP, history of heart failure, angiotensin-converting enzyme inhibitor [ACEI] use, and beta-blocker use) and treatment allocation. To do so, we performed additional post hoc analyses using a Cox proportional hazards model for time to ACM and logistic regression model for mortality (yes/no) using BNP response status and these baseline covariates as the dependent variables in the model. The ROC curves from a logistic regression model were computed to assess the fit of the model.
Survival in responders and nonresponders was also analyzed using the Kaplan-Meier cumulative mortality method. Survival curves were compared using the log-rank test. The comparability of demographic characteristics between the day-5 BNP responders and nonresponders was assessed using a chi-square test for categorical variables and analysis of variance (ANOVA) with response status as a factor for continuous variables. The Kruskal-Wallis nonparametric test was used to evaluate differences between responders and nonresponders for baseline BNP to address the non-normality of the data. The Fisher exact test was used to compare responder and nonresponder groups for incidences of adverse events. SAS version 8.20 (SAS Institute Inc., Cary, North Carolina) was used. A value of p ≤ 0.05 was considered statistically significant.
The overall mortality rates in the SURVIVE population of 1,327 ADHF patients were 13% at 31 days and 27% at 6 months (6). Of 1,327 SURVIVE patients, our analyses included 1,038 patients who had BNP samples taken at both baseline and day 5. Of the 1,327 patients, 69 (5.2%) died prior to day 5; 217 did not have BNP samples at both baseline and day 5; and 3 had BNP sample reading errors of value zero. On day 1, 17 (1.3%) subjects died prior to BNP sampling, and 74 subjects did not have BNP samples at both baseline and day 1.
Assessment of BNP change versus mortality rates
The ROC analyses of cutoffs between −5% and −40% showed that a cutoff of ≥30% BNP decrease had the highest area under the curve value (0.786), and this cutoff value was thus selected to define and differentiate respondersfrom nonresponders.Mortality rates at 31 and 180 days according to various cutoffs of BNP decrease at day 5 are displayed in Figure 1.The threshold of 30% change in BNP optimally separated patients with low and high mortality rates. Conversely, patients having an increase in BNP between baseline and day 5 had a 31-day mortality rate 3 or more times higher than patients with a decrease in BNP of 30% or more.
Predictive value of BNP changes over early intervals
At early time points (days 1, 3, and 5), better likelihood for survival was predicted by ≥30% lowering of plasma BNP from baseline level, that is, responderstatus; however, being a responderat day 5 provided numerically lower hazard ratios at both 31- and 180-day mortality than did being a responderat day 1 or 3 (Table 1).Based on these results, we used BNP decrease from baseline to day 5 as the preferred time interval for outcome prediction. With the Cox proportional hazards regression model, responders at day 5 had a 67% risk reduction in 31-day mortality compared with nonresponders (p < 0.001). For 180-day mortality, the risk reduction was 47% (p < 0.001). Significant differences were confirmed by Kaplan-Meier survival curves (log-rank test p < 0.0001) (Fig. 2).
BNP change versus absolute BNP value in prediction of outcome
In accord with the median BNP value at day 5 of 768 pg/ml and previous experience, an absolute BNP cutoff value 800 pg/ml (at day 5) was tested for prognostic significance (Table 2).BNP values below 800 pg/ml at day 5 had short- and long-term prognostic values similar to those observed with the ≥30% decrease from baseline BNP at day 5, based on comparison of ROC values from the logistic models or with −2 log likelihood values from the Cox proportional hazards model.
Baseline characteristics affecting BNP response
There was no difference in the baseline values in the groups of responders and nonresponders regarding age, sex, and rates of hypertension (Table 3).There were lower proportions of people with previous history of heart failure (p = 0.05) or myocardial infarction (p = 0.001) among nonresponders but a higher proportion of people with diabetes (p = 0.007). There was also no difference at baseline regarding the rate of prescription of ACEIs, beta-blockers, furosemide, spironolactone, or lipid-lowering agents. However, BNP responders more often received levosimendan than dobutamine treatment (63% levosimendan vs. 37% dobutamine; p < 0.001). Baseline plasma BNP was higher in responders than in nonresponders. Significant but modest differences were observed at baseline regarding blood pressure and sodium and potassium levels. There were no differences in heart rate, plasma creatinine, or transaminase levels.
To assess whether the BNP response status and 31-day mortality rate were affected by other factors, we assessed baseline parameters, treatment allocations, and BNP responses. A multivariate analysis examined whether the prognostic value of BNP response at day 5 was retained after stratifying by treatment allocation; baseline blood pressure; plasma creatinine; BNP value; or previous history of chronic heart failure and treatment with ACEI or beta-blockers. Only day-5 BNP response (chi-square: 17.6, p < 0.0001), baseline BNP (chi-square: 12.6, p = 0.0004), plasma creatinine (chi-square: 9.7, p = 0.0018), ACEI treatment (chi-square: 7.8, p = 0.005), and beta-blocker treatment (chi-square: 5.0, p = 0.03) were significantly associated with outcome at 31 days by Cox proportional hazards model analysis.
We also investigated whether BNP responses were associated with any greater rate of adverse events during the entire study period, related to important and/or rapid hemodynamic changes. There was no significant difference in the rates of acute coronary syndrome, hypotension, renal failure, or acute myocardial infarction between responders and nonresponders during the 180 days of the study (Table 4).Occurrence of cardiogenic shock, cardiac failure, and ventricular tachycardia was significantly less in the responder group (Table 4). There was a significant difference between responders and nonresponders regarding overall serious adverse events; 22% (119 of 549) of responders and 32% (158 of 489) of nonresponders had events (Table 4).
Our current analyses of the SURVIVE trial showed that a decrease in BNP of 30% or more during the first 5 days of in-hospital management of patients with ADHF was positively associated with improved survival at 31 and 180 days, regardless of the treatment group allocation. In SURVIVE patients, we observed that an early decrease in BNP level, whether assessed at day 1, 3, or 5, was associated with lower mortality over time. In particular, a BNP response (either defined by ≥30% lowering from baseline or by an absolute value <800 pg/ml) at day 5 appeared to best predict survival at days 31 or 180. Not surprisingly, the decrease in BNP at day 5 better predicted mortality at day 31 than at day 180; many aspects of care after the initial month of therapy can influence outcome beyond the first weeks, for example, initiation of beta-blocker therapy, ACEI dose increase, revascularization, and patient education. The BNP–survival relationship was independent of other factors known to affect prognosis, such as baseline plasma BNP or creatinine levels, blood pressure, or history of heart failure and certain heart failure treatment medications at entry. Most importantly, although more levosimendan patients than dobutamine patients belonged to the day-5 responders, the BNP–survival relationship was consistent regardless of whether patients were allocated for treatment with levosimendan or dobutamine. Notably, an increase in BNP during the first 5 days was associated with a much poorer outcome than a decrease in BNP with therapy.
Our study results are consistent with findings of another study group. Kazanegra et al. (11) were among the first to demonstrate a close relationship between a decrease in pulmonary wedge pressure and decreased plasma BNP during in-hospital treatment of patients with ADHF; this observation was thought to reflect the important treatment goal of achieving a favorable cardiac pre-load in this population, usually by decreasing left ventricular filling pressures by use of diuretics, nitrates, phosphodiesterase inhibitors, or nesiritide. Subsequent studies, however, were not consistent for this correlation between BNP and pulmonary wedge pressure (12–15). Our findings are also consistent with prior results in ambulatory patients with congestive heart failure, which showed that a strategy aimed at decreasing BNP levels was more effective for predicting prognosis than was a traditional strategy based on clinical judgment (16,17).
Previous studies also reported that natriuretic peptide (BNP or NT-proBNP) decreases (from baseline to discharge) or low absolute BNP values at discharge (between 7 and 15 days) were associated with better outcomes in single-center study populations of heart failure patients who had moderately compromised hemodynamic status (5,10,18,19). Our findings in SURVIVE patients extended such observations to a larger multicenter, multinational population of patients with heart failure so severe that the patients needed intravenous inotrope therapy.
Prompt treatment is vital to rapid correction of compromised hemodynamic status in patients hospitalized with ADHF. A major concern, however, is that overly aggressive strategies might result in adverse events such as treatment-related hypotension, renal failure, or arrhythmias. Notably, SURVIVE study treatments marked by BNP decreases ≥30% from baseline to day 5 were not associated with increased frequencies of these side effects.
Our results show that the BNP decrease in the first days of treatment has a similar prognostic value to a lower absolute value of BNP at day 5. However, the high prognostic value of day-5 BNP changes (or lower absolute levels of BNP) applies only to patients still alive at this time point. In the SURVIVE trial, 69 patients (5.2%, representing 40% of 31-day mortality) died before day 5. These early deaths could not have been predicted by the day-5 BNP response. In this regard, the day-1 BNP response (hazard ratio: 0.57, p < 0.0001), although less predictive than the day-5 response, may be of value for the clinicians in charge of patients in acute-care settings.
Our findings thus suggest that BNP response (lower levels) with treatment can predict ACM prognosis; however, these findings and the prognostic value of day-1 or day-5 BNP response with regard to treatments should be tested in prospective studies of hospitalized patients receiving treatment for ADHF.
Taken together, our findings and those of others support the concept that BNP changes during the first 5 days of hospitalization have a very high prognostic value in patients with heart failure. In our study of SURVIVE data, BNP lowering by more than 30% from baseline during the first 5 days predicted both 31- and 180-day mortality; BNP change by day 5 was a better predictor than was change on the first day. Day-1 BNP change, however, also has high value by providing early prognostic information. Such early and reliable prediction information offers the advantage of achieving early optimization of treatment. Consistent with our current findings in patients hospitalized with ADHF, we hypothesize that survival can be improved when the in-hospital treatment algorithm is aimed at decreasing BNP more than 30% or lowering BNP to <800 pg/ml within 5 days of admission; however, this strategy requires further study to determine whether it remains valid when specific treatments are used.
Abbott Laboratories and Orion Pharma provided funding for this study. Drs. Huang and Cai, project statisticians, are employees of Abbott Laboratories. Drs. Cohen-Solal, Nieminen, and Mebazaa have received honoraria from Abbott. Dr. Nieminen is a consultant for OrionPharma. Dr. Logeart has received fees from Biosite and Roche Diagnostics.
- Abbreviations and Acronyms
- angiotensin-converting enzyme inhibitor
- all-cause mortality
- acute decompensation of heart failure
- analysis of variance
- B-type natriuretic peptide
- receiver-operator characteristic
- Received September 2, 2008.
- Revision received January 20, 2009.
- Accepted February 3, 2009.
- American College of Cardiology Foundation
- Moe G.W.,
- Howlett J.,
- Januzzi J.L.,
- Zowall H.
- Logeart D.,
- Saudubray C.,
- Beyne P.,
- et al.
- Logeart D.,
- Thabut G.,
- Jourdain P.,
- et al.
- Daniels L.B.,
- Maisel A.S.
- Parsonage W.A.,
- Galbraith A.J.,
- Koerbin G.L.,
- Potter J.M.
- Jourdain P.,
- Jondeau G.,
- Funck F.,
- et al.
- Gackowski A.,
- Isnard R.,
- Golmard J.L.,
- et al.
- Bettencourt P.,
- Azevedo A.,
- Pimenta J.,
- Frioes F.,
- Ferreira S.,
- Ferreira A.