Author + information
- Received February 19, 2009
- Revision received June 15, 2009
- Accepted June 29, 2009
- Published online November 10, 2009.
- John G.F. Cleland, MD⁎,⁎ (, )
- John J.V. McMurray, MD†,
- John Kjekshus, MD, PhD‡,
- Jan H. Cornel, MD§,
- Peter Dunselman, MD, PhD∥,
- Cândida Fonseca, MD¶,
- Åke Hjalmarson, MD, PhD#,
- Jerzy Korewicki, MD, PhD⁎⁎,
- Magnus Lindberg, MSc††,
- Naresh Ranjith, MD‡‡,
- Dirk J. van Veldhuisen, MD, PhD§§,
- Finn Waagstein, MD, PhD∥∥,
- Hans Wedel, PhD¶¶,
- John Wikstrand, MD, PhD##,
- CORONA Study Group
- ↵⁎Reprint requests and correspondence:
Prof. John G. F. Cleland, Department of Cardiology, Hull York Medical School at the University of Hull, Medical Research Building, Gate 2, Castle Hill Hospital, Kingston-upon-Hull, East Yorkshire HU16 5JQ, United Kingdom
Objectives We investigated whether plasma amino-terminal pro-brain natriuretic peptide (NT-proBNP), a marker of cardiac dysfunction and prognosis measured in CORONA (Controlled Rosuvastatin Multinational Trial in Heart Failure), could be used to identify the severity of heart failure at which statins become ineffective.
Background Statins reduce cardiovascular morbidity and mortality in many patients with ischemic heart disease but not, overall, those with heart failure. There must be a transition point at which treatment with a statin becomes futile.
Methods In CORONA, patients with heart failure, reduced left ventricular ejection fraction, and ischemic heart disease were randomly assigned to 10 mg/day rosuvastatin or placebo. The primary composite outcome was cardiovascular death, nonfatal myocardial infarction, or stroke.
Results Of 5,011 patients enrolled, NT-proBNP was measured in 3,664 (73%). The midtertile included values between 103 pmol/l (868 pg/ml) and 277 pmol/l (2,348 pg/ml). Log NT-proBNP was the strongest predictor (per log unit) of every outcome assessed but was strongest for death from worsening heart failure (hazard ratio [HR]: 1.99; 95% confidence interval [CI]: 1.71 to 2.30), was weaker for sudden death (HR: 1.69; 95% CI: 1.52 to 1.88), and was weakest for atherothrombotic events (HR: 1.24; 95% CI: 1.10 to 1.40). Patients in the lowest tertile of NT-proBNP had the best prognosis and, if assigned to rosuvastatin rather than placebo, had a greater reduction in the primary end point (HR: 0.65; 95% CI: 0.47 to 0.88) than patients in the other tertiles (heterogeneity test, p = 0.0192). This reflected fewer atherothrombotic events and sudden deaths with rosuvastatin.
Conclusions Patients with heart failure due to ischemic heart disease who have NT-proBNP values <103 pmol/l (868 pg/ml) may benefit from rosuvastatin.
- amino-terminal pro-brain natriuretic peptide
- heart failure
- randomized controlled trial
B-type natriuretic peptide and amino-terminal pro-brain natriuretic peptide (NT-proBNP) predict mortality in patients with heart failure, whether (1–5) or not (6) left ventricular ejection fraction (LVEF) is reduced. Indeed, B-type natriuretic peptide and NT-proBNP have usually proved superior to any other variable in predicting outcome (7,8), and their prognostic value may extend far beyond heart failure (8–10).
There is abundant evidence that statins reduce vascular events and improve prognosis in a broad range of patients with or at increased risk of atherosclerosis (10). However, the prognosis of some patient groups at very high risk of events may not be amenable to modification with statins (11–13). This may be because statins are ineffective in such patients or because disease progression is not driven predominantly by progression of atheroma. Most clinical trials of statins have excluded patients with overt heart failure, but several suggested that they could reduce the risk of the development of heart failure in a patient (14,15). The Heart Protection Study did not record which patients had heart failure at baseline but did measure NT-proBNP using a noncommercial assay (10,16). The risk of cardiovascular events increased as NT-proBNP increased, and, therefore, although the relative benefits of simvastatin declined with increasing plasma concentrations, the absolute benefits were similar regardless of the NT-proBNP level (10). However, unlike in CORONA, few patients in the Heart Protection Study had high plasma concentrations of NT-proBNP. In CORONA, rosuvastatin reduced the rate of atherothrombotic events and of cardiovascular hospitalizations, but no overall effect of rosuvastatin on mortality was demonstrated. We investigated the possibility of a transition point for NT-proBNP below which patients with heart failure had a more favorable outcome and greater benefit from rosuvastatin. We also investigated the ability of NT-proBNP to predict prognosis and mode of death in a large group of patients with chronic heart failure.
Detailed accounts of the design, enrollment, and patient characteristics of CORONA have been reported. In brief, patients 60 years of age or older with, in the investigators' opinion, stable, symptomatic heart failure, ischemic heart disease, and a reduced LVEF were eligible, provided they were not taking lipid-lowering drugs. Patients with a history of intolerance to statins; those experiencing a recent vascular event; those with severe uncorrected valve disease, a serum creatinine of >2.5 mg/dl (221 μmol/l), or any condition other than heart failure that would substantially reduce life expectancy or limit compliance with the protocol; and those taking <80% of dispensed placebo tablets during the run-in period were excluded. For most purposes, only patients who had a baseline measurement of NT-proBNP were included in this analysis.
The trial was approved by the ethics committee at each of the participating hospitals, and patients gave written informed consent. Eligible patients were treated with single-blind placebo to demonstrate compliance for as long as 4 weeks before randomization. Patients were randomly assigned, double-blind, to receive 10-mg rosuvastatin or matching placebo once daily.
Baseline assessments including symptoms, medical history, examination, and documentation of a low LVEF, usually by echocardiography, were recorded. Blood tests were done at the end of the run-in period before randomization. Patients were not required to rest or fast before testing. NT-proBNP was analyzed at a central laboratory using a commercially available assay (Roche Diagnostics, Basel, Switzerland).
The steering committee designed the trial and supervised its conduct in collaboration with representatives of the study's sponsor, AstraZeneca (Mölndal, Sweden). The sponsor collected the trial data.
Variables included in the analyses are shown by tertile of NT-proBNP in Table 1.Continuous variables that were not normally distributed, such as NT-proBNP, were expressed as median and interquartile range, and for modeling purposes, a natural log transformation was used.
The main objective was to investigate the effects of rosuvastatin according to the plasma concentration of NT-proBNP for various cardiovascular outcomes. The primary outcome was the composite of cardiovascular mortality, nonfatal myocardial infarction (MI), or nonfatal stroke, analyzed as the time to the first event. The secondary outcomes were (in listed order): all-cause mortality; any coronary event (defined as sudden death, fatal or nonfatal MI, coronary revascularization, ventricular defibrillation by an implantable defibrillator, resuscitation from cardiac arrest, or hospitalization for unstable angina); cardiovascular mortality (cause-specific cardiovascular death was also analyzed); and total number (episodes) of hospitalizations (for all causes, cardiovascular causes, unstable angina, and worsening heart failure). We also included 2 additional post hoc composite outcomes: death from any cause or hospitalization for worsening heart failure, which is commonly reported in heart failure trials, and an atherothrombotic end point (fatal or nonfatal MI or fatal or nonfatal nonhemorrhagic stroke) constituting events likely to be related to atherosclerothrombosis rather than myocardial disease. The definition and adjudication of all outcomes were previously described in detail (14).
Cox proportional hazards models were used to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs) comparing rosuvastatin and placebo according to tertiles of NT-proBNP (SAS version 8.2, Stata Corp., College Station, Texas), both unadjusted Cox and adjusted for the pre-specified risk factors that were included in the optimal allocation procedure at randomization (age; sex; ejection fraction; New York Heart Association functional class; history of MI, hypertension, and diabetes mellitus; total cholesterol; and beta-blocker use). For unadjusted Cox proportional hazards models, the p values are determined by the log-rank test. The total numbers of hospitalizations were analyzed with a permutation test.
The associations between [loge]NT-proBNP and the previously cited outcomes were investigated in multivariable models (17). Each analysis was conducted with and without the exclusion of NT-proBNP and high-sensitivity C-reactive protein to identify factors that they displaced or drew in when entered into the model. The relationship between [loge]NT-proBNP and the number of events per 100 patient-years of follow-up was also expressed using a cubic-spline function with HRs adjusted for age, history of diabetes, coronary bypass or claudication, New York Heart Association functional class, heart rate and rhythm, systolic blood pressure, and ejection fraction. The 95% CIs were estimated using the bootstrap method with 1,000 resubstitutions.
Of 5,011 patients enrolled in CORONA, a measurement of NT-proBNP was available for 3,664 (73%) and showed a highly skewed distribution (Fig. 1).The characteristics of patients with and without a measurement of NT-proBNP were similar (data not shown). Patients in the highest tertile of NT-proBNP (>277 pmol/l [2,348 pg/ml]) had, on average, many other markers of a worse prognosis (Table 1). Patients with plasma concentrations in the lowest tertile of NT-proBNP (<103 pmol/l [<868 pg/ml]) were younger; were more likely to have mild symptoms, a higher body mass index, systolic blood pressure, and cholesterol; were less likely to have rhythm disturbances, renal dysfunction, or increased high-sensitivity C-reactive protein; and were less likely to receive loop diuretics, aldosterone antagonists, or digoxin (Table 1). Sex, LVEF, and the proportions treated with angiotensin-converting enzyme inhibitors and beta-blockers were similar across tertiles. Within each tertile, the characteristics in the 2 randomization groups were very similar (data not shown).
NT-proBNP and rates of events
The risk of events increased exponentially when plotted against [loge]NT-proBNP as a continuous function (Fig. 2).The risk of death from heart failure was low in patients with NT-proBNP <50 pmol/l (423 pg/ml). The risk of sudden death began to increase at somewhat lower levels. There was a marked and progressive increase in the rate of each outcome of interest, apart from nonfatal vascular events, with each increasing tertile of NT-proBNP (numbers given in Fig. 3).
Multivariable analyses demonstrated that NT-proBNP was more strongly associated with each specified outcome than any other measured variable. The relationship between NT-proBNP was strongest for death or hospitalization for worsening heart failure, weaker for sudden death, and weakest for atherothrombotic events (Table 2).Few variables added substantially to the model fit for most outcomes. Eliminating NT-proBNP from the models generally increased the strength of association between outcome and LVEF, serum creatinine, body mass index, higher heart rate, and apolipoprotein A-I. Conversely, when NT-proBNP was included in the models, atrial fibrillation was more strongly associated with an adverse outcome and female sex with a more favorable outcome.
Change in lipids and lipoproteins during follow-up
For each tertile of NT-proBNP, rosuvastatin exerted a similar net percentage mean decrease versus placebo in low-density lipoprotein from baseline to the 3-month follow-up visit (47% [95% CI: 45% to 49%], 45% [95% CI: 43% to 48%], and 43% [95% CI: 41% to 46%]; p < 0.0001 for all) in low, middle, and upper tertiles, respectively; triglycerides (20% [95% CI: 15% to 24%], 22% [95% CI: 18% to 26%], and 20% [95% CI: 16% to 25%]; p < 0.0001 for all), and a similar percentage increase in high-density lipoprotein (7% [95% CI: 4% to 8%], 5% [95% CI: 3% to 7%], and 5% [95% CI: 3% to 7%]; p < 0.0001 for all). For each tertile of NT-proBNP, rosuvastatin exerted a similar net percentage median decrease in high-sensitivity C-reactive protein from baseline to the closing visit (37%, 37%, and 35%; p < 0.0001, p < 0.0001, and p = 0.0002, respectively).
Interaction between the effects of rosuvastatin and NT-proBNP
The Cox adjusted HR (0.65, 95% CI: 0.47 to 0.88; p = 0.005) for the primary end point among patients in the lowest tertile of NT-proBNP favored those assigned to rosuvastatin, and tests suggested heterogeneity for the effect of rosuvastatin (p = 0.0192) on the primary end point across NT-proBNP tertiles (Fig. 3). This effect was driven by a combination of fewer MIs, strokes, and cardiovascular deaths in the lowest tertile. Trends were similar for most other outcomes, except death from heart failure, for which there were very few events in the first tertile (Fig. 3). There was also a reduction in hospitalizations for cardiovascular reasons and for worsening heart failure in patients within the lowest tertile of NT-proBNP assigned to rosuvastatin (Table 3),with less certain evidence of benefit among patients in the upper 2 tertiles.
Discontinuation from randomized treatment increased with increasing NT-proBNP values in both the placebo and rosuvastatin groups, however, with fewer discontinuations of rosuvastatin compared with placebo in all 3 tertiles, significantly so in the lowest tertile (131 patients receiving placebo and 91 receiving rosuvastatin, 9.0 vs. 6.2 per 100 patient-years of follow-up, HR: 0.69, 95% CI: 0.53 to 0.91; p = 0.0068) compared with the middle tertile (122 patients vs. 113 patients, 8.9 vs. 8.2 per 100 patient-years of follow-up, HR: 0.92; p = NS) and upper tertile (157 patients vs. 153 patients, 14.5 vs. 12.7 per 100 patient-years of follow-up, HR: 0.88; p = NS).
This post hoc analysis of CORONA confirms, in patients with heart failure and ischemic heart disease, that elevated plasma concentrations of NT-proBNP are associated with many other variables that predict an adverse prognosis; that NT-proBNP is a powerful predictor of many cardiovascular outcomes including atherothrombotic vascular events, sudden death, and especially heart failure; and that few other clinical or laboratory variables add substantially to its ability to predict outcome. Patients with lower plasma concentrations of NT-proBNP had a lower event rate but seemed to obtain greater benefit from rosuvastatin. This occurred despite similar effects of rosuvastatin on lipid profile and high-sensitivity C-reactive protein in each tertile of NT-proBNP. Fewer patients withdrew from rosuvastatin than placebo in the lowest tertile. The rates of discontinuation of rosuvastatin and placebo in patients in the upper 2 tertiles of NT-proBNP were similar and seemed unlikely to account for the lack of benefit.
There is extensive literature on the potential diagnostic and prognostic utility of natriuretic peptides, although few studies are as large as CORONA (8). Plasma concentrations of NT-proBNP <50 pmol/l (approximately 423 pg/ml), suggesting mild and/or well-controlled heart failure (18), were recorded in <25% of patients. Most patients in CORONA had grossly elevated levels of NT-proBNP, similar to those observed in studies of moderate or severe heart failure such as the CARE-HF (Cardiac Resynchronization-Heart Failure) trial (19,20), although the relationship to New York Heart Association functional class was not strong, perhaps reflecting the variability of symptom classification provided by clinicians and protocol-driven patient selection (21). Previous reports suggest that plasma concentrations of natriuretic peptides in patients with known cardiovascular disease are determined primarily by cardiac and renal function with superadded effects due to body mass index, fluid balance, age, sex, and heart rhythm, many of which are thought to influence prognosis (8). It is perhaps because NT-proBNP reflects so many prognostic variables that it is itself such a powerful predictor of prognosis. Conversely, the multivariable models suggest that the association between NT-proBNP and outcome is only partially explained by the previously cited factors because the model has less predictive power when NT-proBNP is dropped. Of particular interest was the strengthening of the relationship between a favorable outcome and female sex when NT-proBNP was included in the model. Healthy women are known to have higher plasma concentrations of NT-proBNP (22), and women may have a better prognosis than men for a given value of NT-proBNP. Patients with atrial fibrillation also have higher plasma concentrations of NT-proBNP, whether or not they have clinical heart failure, and they too seem to have a better prognosis than suggested by their plasma NT-proBNP level (23). LVEF differed little across tertiles of NT-proBNP. This could reflect the inaccuracy of echocardiographic measurements, but the numbers in each group are large and the lack of accuracy should increase variability but not affect mean values for LVEF. Plasma concentrations of NT-proBNP may also reflect diastolic left ventricular function (6), atrial fibrillation, valve dysfunction, and, most importantly in the context of CORONA, mitral regurgitation. Accordingly, a strong relationship between LVEF and NT-proBNP should not be expected. This has important implications for the diagnostic use of natriuretic peptides, which may not perform well when used solely to predict the presence of left ventricular systolic dysfunction rather than the more general problem of cardiac dysfunction.
Because natriuretic peptides reflect many aspects of cardiac function, it is not surprising that the strongest relationship is with hospitalization or death due to worsening heart failure. Because such events account for a high proportion of cardiovascular events, it is also not surprising that natriuretic peptides predict cardiovascular hospitalization and death. However, NT-proBNP was also a powerful predictor of sudden death. Assuming that such events predominantly reflect ventricular arrhythmias and knowing that the severity of left ventricular dysfunction is a powerful predictor of arrhythmias, this is also not surprising. Other predisposing factors such as myocardial electrical re-entry pathways that are not entirely dependent on the severity of ventricular damage, may weaken its relationship to NT-proBNP. However, some sudden deaths may be vascular in origin. NT-proBNP was a relatively poor predictor of atherothrombotic events compared with other measured outcomes. Few patients with advanced heart failure are likely to survive a further substantial vascular event, which may instead be recorded as sudden death, either because it occurred rapidly or because patients were found dead, having been unable to call for help after the onset of symptoms.
The Heart Protection Study suggested that the relative, but not the absolute, benefits of simvastatin declined as NT-proBNP increased (10). The CORONA data are consistent with this observation and extend the observation to patients with much higher plasma concentrations of NT-proBNP (16). Low plasma concentrations should trigger a review of the clinical evidence for the diagnosis of heart failure and its associated treatment. Perhaps patients who have heart failure, left ventricular systolic dysfunction, and ischemic heart disease with NT-proBNP of a magnitude similar to that found in the first tertile of NT-proBNP in CORONA should receive a statin, but treatment may be futile in patients with heart failure and markedly elevated plasma concentrations (24). However, the analysis also suggests that statins are not harmful in patients with advanced heart failure and might be used in such patients when the patient and physician consider it appropriate.
This is a retrospective analysis, and, ideally, the results should be confirmed in a prospective study. It is unlikely that patients or physicians would be willing to participate in further placebo-controlled trials of statins in patients with ischemic heart disease and NT-proBNP <103 pmol/l, but trials of statins or other agents that alter the lipid profile in patients with higher levels might be considered. The biological variability in plasma concentrations of NT-proBNP is large, and the relationship between NT-proBNP and therapeutic benefit from rosuvastatin is probably modified by other patient characteristics. Accordingly, plasma values of NT-proBNP should be used for general guidance rather than for precise rule setting. Plasma concentrations of NT-proBNP cannot be equated precisely with those of brain natriuretic peptide, for which a variety of different assays exists. However, an NT-proBNP plasma concentration of 100 pmol/l (846 pg/ml) equates to a brain natriuretic peptide value of approximately 200 pg/ml using the Biosite assay (Biosite, San Diego, California) or ADVIA Centaur assay (Bayer Diagnostics, Tarrytown, New York) (25).
The CORONA and the Heart Protection Study findings are consistent and suggest that the relative benefits of statins decline as the plasma concentration of NT-proBNP increases so that an effect is no longer apparent above values of approximately 103 pmol/l (868 pg/ml). These data support the use of statin therapy in patients with coronary disease and less advanced heart failure.
This study was supported by AstraZeneca. Drs. Cleland, Hjalmarson, Kjekshus, Korewicki, McMurray, Ranjith, van Veldhuisen, Waagstein, and Fonseca have been paid lecture fees by AstraZeneca. Drs. Cornel, Dunselman, Hjalmarson, Kjekshus, McMurray, van Veldhuisen, Waagstein, Korewicki, and Wedel have received consulting or advisory board fees from AstraZeneca. Dr. Fonseca has received consulting or advisory board fees from AstraZeneca and Roche. Drs. Hjalmarson, McMurray, and Wikstrand have received research grants from AstraZeneca. Mr. Lindberg is an employee of AstraZeneca. Dr. van Veldhuisen has received research grants from AstraZeneca and Roche. Dr. Waagstein reports owning shares in AstraZeneca. Dr. Wikstrand is a former Senior Medical Advisor at AstraZeneca, now part time consultant. For disclosures for the other members of the CORONA study group see Kjekshus et al. (11).
- Abbreviations and Acronyms
- hazard ratio
- left ventricular ejection fraction
- myocardial infarction
- amino-terminal pro-brain natriuretic peptide
- Received February 19, 2009.
- Revision received June 15, 2009.
- Accepted June 29, 2009.
- American College of Cardiology Foundation
- Anand I.S.,
- Fisher L.D.,
- Chiang Y.-T.,
- et al.
- Masson S.,
- Latini R.,
- Anand I.S.,
- et al.,
- Val-HeFT Investigators
- Hartmann F.,
- Packer M.,
- Coats A.J.S.,
- et al.
- Cleland J.G.,
- Freemantle N.,
- Ghio S.,
- et al.
- Doust J.A.,
- Pietrzak E.,
- Dobson A.,
- Glasziou P.
- McDonagh T.A.,
- Cunningham A.D.,
- Morrison C.E.,
- et al.
- Heart Protection Study Collaborative Group
- Khush K.K.,
- Waters D.D.,
- Bittner V.,
- et al.
- Cleland J.G.,
- Squire I.,
- Ng L.
- Wedel H.,
- McMurray J.J.,
- Lindberg M.,
- et al.
- Dickstein K.,
- Cohen-Solal A.,
- Filippatos G.,
- et al.
- Cleland J.G.F.,
- Daubert J.-C.,
- Erdmann E.,
- et al.
- Morello A.,
- Lloyd-Jones D.M.,
- Chae C.U.,
- et al.
- Cleland J.G.,
- Tavazzi L.,
- Daubert J.C.,
- Tageldien A.,
- Freemantle N.
- Austin W.J.,
- Bhalla V.,
- Hernandez-Arce I.,
- et al.