Utilizing NT-proBNP for Eligibility and Enrichment in Trials in HFpEF, HFmrEF, and HFrEF

JACC: HEART FAILURE

VOL.

-, NO. -, 2018

ª 2018 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION PUBLISHED BY ELSEVIER

Using NT-proBNP for Eligibility and Enrichment in Trials in HFpEF, HFmREF, and HFrEF Gianluigi Savarese, MD,a Nicola Orsini, PHD,b Camilla Hage, PHD,a Ola Vedin, MD, PHD,c Francesco Cosentino, MD, PHD,a,d Giuseppe M.C. Rosano, MD, PHD,e,f Ulf Dahlström, MD, PHD,g,h Lars H. Lund, MD, PHDa,d

ABSTRACT OBJECTIVES The purpose of this study was to assess the association between N-terminal pro–B-type natriuretic peptide (NT-proBNP) and cardiovascular (CV) versus non-CV events and between NT-proBNP and potential treatment effects in heart failure (HF) with preserved, mid-range, and reduced ejection fraction (HFpEF, HFmrEF, and HFrEF, respectively) and clinically relevant subgroups. BACKGROUND Optimizing patient eligibility criteria in HF trials requires biomarkers that enrich for CV but not for non-CV events and select patients most likely to respond to the tested intervention. METHODS In the Swedish HF registry population stratified by EF category, we used Kaplan-Meier curves to estimate unadjusted CV and non-CV risks (mortality or hospitalization); Poisson regressions to calculate crude event rates of CV and non-CV events according to NT-proBNP levels; and Cox regressions to calculate the adjusted hazard ratios for HF therapies according to NT-proBNP # or > median. RESULTS In a cohort of 15,849 patients (23% HFpEF, 21% HFmrEF, 56% HFrEF), median NT-proBNP was 2,037, 2,192, and 3,141 pg/ml, respectively. With increasing NT-proBNP, CV event rates increased more steeply than non-CV rates (range 20 to 160 and 30 to 100 per 100 patient-years in HFpEF; 20 to 130 and 20 to 100 in HFmrEF; and 20 to 110 and 20 to 50 in HFrEF, respectively). The CV-to-non-CV ratio increased with increasing NT-proBNP in HFpEF and HFrEF, but only in the lower range in HFmrEF. The association between treatments (e.g., angiotensin-converting enzyme-inhibitor, angiotensin II receptor blockers, and beta-blockers) and outcomes was consistent in NT-proBNP # and > median. CONCLUSIONS In HF trial design in different EF categories, NT-proBNP may be a useful tool for eligibility and enrichment for CV events, but its role in predicting a potential treatment response remains unclear. (J Am Coll Cardiol HF 2018;-:-–-) © 2018 by the American College of Cardiology Foundation.

From the aDivision of Cardiology, Department of Medicine, Karolinska Institutet, Stockholm, Sweden; bDepartment of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden; cDepartment of Medical Sciences, Uppsala University and Uppsala Clinical Research Center, Uppsala, Sweden; dHeart and Vascular Theme, Karolinska University Hospital, Stockholm, Sweden; e

Cardiovascular and Cell Sciences Research Institute, St. George’s University, London, United Kingdom; fIstituti di Ricovero e Cura

a Carattere Scientifico (IRCCS) San Raffaele Pisana, Rome, Italy; gDepartment of Cardiology, Linköping University, Linköping, Sweden; and the hDepartment of Medical and Health Sciences, Linköping University, Linköping, Sweden. This study was supported by grants 2013-23897-104604-23 and 523-2014-2336 from the Swedish Research Council; grants 20120321 and 20150557 from the Swedish Heart Lung Foundation; and a grant from Relypsa, Inc. No funding agency had any role in the design and conduct of the study, collection, management, analysis, or interpretation of the data, or in the preparation or approval of the manuscript. Dr. Savarese has received research grants from Merck Sharp & Dohme Italy and the Swedish Heart and Lung Foundation. Dr. Hage has received consulting fees from Novartis; and honoraria from Merck Sharp & Dohme. Dr. Vedin has received consultancy and lecture fees from AstraZeneca, Boehringer Ingelheim, Novartis, Fresenius Medicare, and Alnylam Pharmaceuticals. Dr. Cosentino has received research support from the Swedish Research Council, Heart and Lung Foundation, Karolinska Institute, European Foundation for the Study of Diabetes, and the Stiftelsen Frimurare Barnhuset Foundation; and has received honoraria from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Roche, Bristol-Myers Squibb, Merck Sharp & Dohme, Abbott, Bayer, and Novo Nordisk. Dr. Dahlström has received research grants from AstraZeneca; and honoraria from AstraZeneca and Novartis. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Manuscript received November 17, 2017; revised manuscript received December 28, 2017, accepted December 29, 2017.

ISSN 2213-1779/$36.00

https://doi.org/10.1016/j.jchf.2017.12.014

2

Savarese et al.

JACC: HEART FAILURE VOL.

T

-, NO. -, 2018 - 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

rials in heart failure (HF) have relied

first, the association between NT-proBNP and CV

on ejection fraction (EF) for inclu-

versus non-CV outcomes, and second, we examined

sion and demonstrated efficacy of

the association between HF treatments (used as sur-

multiple drug and device interventions in

rogates for treatments that may be tested in future

HF with reduced EF (HFrEF; EF<40%). In

trials) and CV and non-CV outcomes according to

blocker

contrast, trials in HF with preserved EF

NT-proBNP levels.

CV = cardiovascular

(HFpEF, variably defined as EF of >40% to

ABBREVIATIONS AND ACRONYMS ACE = angiotensin-converting enzyme

ARB = angiotensin receptor

50%) have failed. The recent category HF

EF = ejection fraction

METHODS

with mid-range EF (HFmrEF; 40% to 49%)

HF = heart failure HFmrEF = heart failure with mid-range ejection fraction

HFpEF = heart failure with

has been characterized (1–3). Although EF is

STUDY PROTOCOL AND SETTING. The SwedeHF

not normal, there is no evidence-based ther-

(Swedish Heart Failure) registry has been previously

apy yet (4).

described (14). Briefly, it was created in 2000 and has

preserved ejection fraction

The failures of HFpEF phase II and III trials

been used widely in Sweden since 2003. The only

HFrEF = heart failure with

may be explained by wrong treatment strat-

inclusion criterion is HF as appraised by clinician

reduced ejection fraction

egy (e.g., neurohormonal antagonists may

judgement. Approximately 80 variables are entered

NT-proBNP = N-terminal

not work in HFpEF) or by wrong patient se-

at hospital discharge or after out-patient clinic visit to

pro–B-type natriuretic peptide

lection (may not have had confirmed HF or

complete a Web-based case report form. The Uppsala

may have had HFpEF but been poorly enriched, that

Clinical Research Center (Uppsala, Sweden) manages

is, low risk of cardiovascular [CV] events and/or high

the database. The coverage of prevalent HF in Swe-

risk of non-CV events, making HF-specific therapy

deHF was 54% (11), but the coverage of first onset

ineffective or requiring excessive sample sizes for

incident HF, which is more often encountered in

significant results) (2,5,6). HFmrEF resembles HFrEF

acute care or general medicine, is only 12% (15).

in many respects (1,3) but currently has no evidence-

The Swedish Board of Health and Welfare (Min-

based therapy and constitutes an important popula-

istry of Health and Social Affairs, Stockholm, Swe-

tion for future trials. The lack of treatments in

den)

HFmrEF and HFpEF could be amended through

provided the dates of death and the patient registry

several approaches. One strategy is novel therapy

that supplied baseline comorbidities beyond those

directed at novel targets (7,8), another method is to

available in SwedeHF, hospitalizations, and their

test existing generic HF drugs in HFpEF and HFmrEF,

causes, defined according to International Statistical

in trials where patients are selected based on optimal

Classification of Diseases, 10th edition, codes in the

potential for treatment effect (9). How should such

first position, and causes of death (where we used

selection be optimized?

underlying cause rather than immediate mode of

N-terminal pro–B-type natriuretic peptide (NT-

administers

the

population

registry

that

death).

proBNP) is diagnostic and prognostic in HFrEF (10–13)

Socioeconomic data were obtained by Statistics

and has been used in trials to ensure presence of HF

Sweden (Ministry of Finance, Stockholm, Sweden).

and to enrich for CV events. Cutoff values have varied

All permanent residents in Sweden have unique per-

depending on EF and atrial fibrillation but have been

sonal identification numbers that allow linking of

completely arbitrary. Simply requiring higher cutoff

disease-specific health registries and governmental

values would increase CV events (10) but reduce

health and statistical registries.

feasibility of including sufficient numbers of patients.

Establishment of the HF registry and this analysis

Furthermore, it is unknown how different NT-proBNP

with linking of the above registries were approved by

levels affect the CV-to-non-CV event ratios. Finally,

a multisite ethics committee. Individual patient con-

previous studies failed to demonstrate higher benefit

sent was not required, but patients were informed of

of treatments in patients with higher NT-proBNP

entry into national registries and allowed to opt out.

(10,13). Indeed, high levels of NT-proBNP may not

In the current study, both HF patients enrolled as

reflect the severity of HF but instead the presence of

outpatients and those enrolled at discharge from

several CV and non-CV comorbidities such as atrial

hospitals with no missing data for EF and NT-proBNP

fibrillation, chronic kidney disease, and cachexia,

were considered. That is, all patients were stable, but

which may not be treatable with HF-specific therapy.

those enrolled at discharge might have be considered

Trial design would be greatly aided by a better

enriched for increased risk, given acute HF hospital

understanding of the role of NT-proBNP in different

admission immediately before baseline, whereas

EF categories, comorbidity subgroups, and potential

those enrolled as outpatients might or might not

treatment responses. Thus, in HFpEF, HFmrEF, and

have had a recent HF hospitalization (unknown in

HFrEF and in relevant subgroups, we investigated,

our registry) and might have been less enriched.

JACC: HEART FAILURE VOL.

-, NO. -, 2018

3

Savarese et al.

- 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

When the same patient reported several registrations, the first registration, including NT-proBNP levels,

T A B L E 1 Baseline Characteristics

HFpEF HFmrEF HFrEF n ¼ 3,623 (23%) n ¼ 3,322 (21%) n ¼ 8,904 (56%)

was considered. Outcomes of the analysis were the composite of CV death and CV hospitalization (CV events) and the composite of non-CV death and non-

p Value

Investigated subgroups Location*

CV hospitalization (non-CV events). Patients were

Inpatient

1,960 (54)

1,383 (42)

3,645 (41)

classified as reporting a CV or non-CV event according

Outpatient

1,663 (46)

1,939 (58)

5,259 (59)

2,360 (65)

2,011 (60)

4,566 (51)

<0.001

creatinine clearance <60 ml/min

1,931 (58)

1,510 (50)

3,785 (45)

<0.001

creatinine clearance $60 ml/min

1,385 (42)

1,521 (50)

4,632 (55)

$75 yrs

2,890 (80)

2,310 (70)

5,163 (58)

<75 yrs

733 (20)

1,012 (30)

3,741 (42)

to the first event occurring after the index registra-

Atrial fibrillation*

tion. HFpEF was defined as having an EF of $50%,

Renal function*

HFmrEF as having an EF of 40% to 49%, and HFrEF as having an EF of <40% (4). STATISTICAL ANALYSIS. Baseline characteristics of

patients according to EF category were compared by using Student t-test or Wilcoxon-Mann-Whitney test for continuous variables and by chi square test for

<0.001

Age* <0.001

Sex*

categorical variables.

Male

1,690 (47)

2,041 (61)

6,476 (73)

OUTCOME ANALYSES. Unadjusted survivor func-

Female

1,933 (53)

1,281 (39)

2,428 (27)

1,564 (44)

1,725 (53)

4,627 (54)

<0.001

994 (27)

925 (28)

2,449 (27)

0.92 <0.001

tions by EF category were estimated by using the nonparametric Kaplan-Meier method, and unadjusted and adjusted hazard ratios (HR) with 95% confidence intervals (CI) were calculated using Cox proportional hazard models. Crude rates of CV and

Ischemic heart disease* Diabetes* Duration of HF, months* <6

1,640 (45.5)

1,521 (46.0)

4,366 (49.2)

$6

1,961 (54.5)

1,789 (54.0)

4,504 (50.8)

77  11

74  12

71  12

<0.001

Internal medicine or geriatrics

1,333 (44)

1,222 (41)

3,072 (36)

<0.001

Cardiology

1,730 (56)

1,761 (59)

5,474 (64)

1,700 (49)

2,015 (64)

6,638 (77)

Demographics

non-CV events according to the continuous levels of

Age, yrs

NT-proBNP were estimated by univariate Poisson

Specialty*

regression models. The Poisson models were not adjusted because the primary aim was to estimate event rates by EF and NT-proBNP “as is” when selected for trials. That is, in trial selection, there is no “adjustment” when including patients. In these analyses, we flexibly modeled NT-proBNP by using restricted cubic splines with 4 knots at fixed percentiles of the distribution. Cox regression models were performed to calculate the HR for HF therapies (e.g., angiotensin-converting

enzyme

[ACE]

inhibitors

and/or angiotensin receptor blockers [ARBs] and of beta-blockers), according to NT-proBNP levels greater or lesser than or equal to the median value. Because

<0.001

Follow-up referral specialty* Cardiology or Internal medicine Primary care or other care Follow-up referral to outpatient HF nurse clinic*

<0.001

1,756 (51)

1,152 (36)

1,941 (23)

1,187 (34)

1,493 (47)

4,931 (57)

<0.001

0.55

Year of registration* 2001–2009

1,669 (46)

1,557 (47)

4,197 (47)

2010–2012

1,954 (54)

1,765 (53)

4,707 (53)

Clinical NYHA functional class* <0.001

treatment decisions are subject to selection bias and

I

375 (16)

341 (13)

600 (8)

the aim here was to assess a potential independent

II

1,119 (46)

1,372 (53)

3,330 (45)

III

869 (36)

827 (32)

3,154 (43)

IV

60 (2)

59 (2)

280 (4)

27.9 (6.2)

27.6 (5.8)

26.7 (5.3)

<0.001 <0.001

association between NT-proBNP and treatment effect, these analyses were adjusted. Variables included in all multivariate models are the 40 variables labeled

BMI, kg/m2* Blood pressure, mm Hg

with an asterisk in Table 1. In all multivariate models,

Systolic

132  21

130  21

123  20

missing data were managed by multiple imputation

Diastolic

73  12

74  12

73  12

0.008

Arterial blood pressure, mm Hg*

93  13

92  13

90  13

<0.001

Heart rate, beats/min*

73  15

73  15

74  15

<0.001

2,037 (912–4,420)

2,192 (930–4,899)

3,141 (1,370–7,080)

<0.001

(n ¼ 10). Statistical analyses were performed by Stata version 14.2 (StataCorp, LLC, College Station, Texas). A p value of <0.05 was considered statistically significant.

RESULTS PATIENTS. Between May 11, 2000, and December 31,

2012, 80,772 registrations were recorded from 51,060 unique patients. Of these, 15,849 patients had no

Laboratory values NT-proBNP, pg/ml* Creatinine clearance, ml/min

61  30

67  33

70  34

<0.001

Hb, g/l

128  17

132  17

135  17

<0.001

Continued on the next page

Savarese et al.

4

JACC: HEART FAILURE VOL.

-, NO. -, 2018 - 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

and HFrEF for age, sex, and several other character-

T A B L E 1 Continued

istics but was notably more similar to HFrEF for HFpEF HFmrEF HFrEF n ¼ 3,623 (23%) n ¼ 3,322 (21%) n ¼ 8,904 (56%)

p Value

ischemic heart disease. There were no differences for diabetes.

Treatments ACE inhibitor or ARB*

<0.001

Median

NT-proBNP

values

in

HFpEF

(2,037 pg/ml; IQR: 912 to 4,420 pg/ml) and HFmrEF

2,766 (76)

2,868 (86)

8,262 (93)

656 (18)

545 (16)

1,480 (17)

0.088

(2,192 pg/ml; IQR: 930 to 4,899 pg/ml) were similar but

3,113 (86)

2,553 (77)

7,219 (81)

<0.001

much lower than in HFrEF (3,141 mg/ml; IQR: 1,370 to

596 (16)

510 (15)

1,207 (14)

<0.001

7,080 mg/ml) (Online Figure 2).

Platelet inhibitor*

1,550 (43)

1,558 (47)

4,278 (48)

<0.001

Oral anticoagulant agent*

1,500 (42)

1,396 (42)

3,786 (43)

0.53

ASSOCIATIONS BETWEEN EF CATEGORY AND CV

Digoxin* Diuretic agent* Nitrate*

Statin*

1,443 (40)

1,567 (47)

4,313 (49)

<0.001

VERSUS NON-CV EVENTS. Both the unadjusted and

Beta-blocker*

2,879 (80)

2,850 (86)

8,118 (91)

<0.001

the adjusted risks of CV events were higher in HFrEF

MRA*

1,078 (30)

932 (28)

3,408 (38)

<0.001

Device therapy*

than in HFpEF and HFmrEF (Figure 1A). The crude risk of non-CV events differed considerably and

None

3,562 (99)

3,208 (97)

8,040 (91)

CRT-P

13 (0.4)

33 (1.0)

250 (2.8)

CRT-D

8 (0.2)

21 (0.6)

283 (3.2)

ICD

26 (0.7)

45 (1.4)

283 (3.2)

<0.001

remained significant after adjustment by EF category, with HFmrEF intermediate between HFpEF and HFrEF (Figure 1B).

Comorbidities

ASSOCIATIONS

Smoking* <0.001

BETWEEN

NT-proBNP

AND

CV

VERSUS NON-CV OUTCOMES ACCORDING TO EF

Never

1,461 (51)

1,232 (45)

3,027 (40)

Previous

1,132 (40)

1,177 (43)

3,396 (45)

Current

243 (9)

300 (11)

1,145 (15)

2,646 (73)

2,176 (65)

5,053 (57)

<0.001

NT-proBNP levels regardless of EF, but there were

Coronary revascularization*

787 (22)

1,073 (32)

2,908 (33)

<0.001

numerous differences among EF categories (Figure 2).

Peripheral artery disease*

366 (10)

368 (11)

843 (9)

0.028

Non-CV events were higher in HFpEF and HFmrEF

Stroke/TIA*

708 (19)

552 (17)

1,368 (15)

<0.001

throughout the NT-proBNP range; CV event rates

Anemia*

1,412 (39)

1,157 (35)

2,635 (30)

<0.001

Valvular disease*

1,166 (33)

804 (25)

1,851 (21)

<0.001

Lung disease*

1,108 (31)

940 (28)

2,229 (25)

<0.001

551 (15)

482 (15)

1,051 (12)

<0.001

Hypertension*

Cancer within 3 yrs*

CATEGORIES. Event rates per 100 patient-years for

both CV and non-CV events increased with increasing

increased with increasing NT-proBNP more steeply than non-CV event rates in all EF groups, but the increase occurred at lower NT-proBNP levels with higher EF, such that it was steeper at lower NT-

Socioeconomics

proBNP and flatter at higher NT-proBNP concentra-

Family type* Living alone

2,022 (56)

1,671 (50)

4,272 (48)

Married/cohabitating

1,596 (44)

1,649 (50)

4,601 (52)

Compulsory school

1,831 (51)

1,552 (47)

3,890 (44)

Secondary school

1,247 (35)

1,227 (37)

3,503 (40)

503 (14)

520 (16)

1,418 (16)

<0.001

from 20 to 160 and 30 to 100 per 100 patient-years in

Education*

University Income below median*

2,120 (59)

1,697 (51)

4,057 (46)

Number of children*

2.0  1.4

2.0  1.4

2.0  1.4

tions. Crude rates for CV and non-CV events ranged

<0.001

HFpEF; 20 to 130 and 20 to 100 in HFmrEF; and 20 to 110 and 20 to 50 in HFrEF, respectively. The CV-tonon-CV

<0.001 0.001

event

ratio

increased

together

with

increasing NT-proBNP levels in HFpEF and HFrEF but only in the lower range of NT-proBNP in HFmrEF. ASSOCIATIONS BETWEEN HF THERAPY AND CV

Values are n (%), mean  SD, or median (interquartile range). *Variables included in multivariable models. ACE ¼ angiotensin-converting enzyme; ARB ¼ angiotensin receptor blocker; BMI ¼ body mass index; CRT-D ¼ cardiac resynchronization therapy defibrillator; CRT-P ¼ cardiac resynchronization therapy pacemaker; Hb ¼ hemoglobin; HF ¼ heart failure; HFmrEF ¼ heart failure with mid-range ejection fraction; HFpEF ¼ heart failure with preserved ejection fraction; HFrEF ¼ heart failure with reduced ejection fraction; ICD ¼ implantable cardiac defibrillator; MRA ¼ mineralocorticoid receptor antagonist; NT-proBNP ¼ N-terminal pro–B-type natriuretic peptide; NYHA ¼ New York Heart Association; TIA ¼ transient ischemic attack.

VERSUS NON-CV OUTCOMES ACCORDING TO EF CATEGORY. Treatment with ACE inhibitor and/or

ARB was associated with a significant reduction in risk of CV and also non-CV events in HFmrEF and HFrEF, whereas in HFpEF, the reduction in risk observed was statically significant for CV but not for non-CV events

missing values for EF, NT-proBNP concentration, and

(Figure 3). Beta-blockers were significantly associated

follow-up $1 day (Online Figure 1).

with improved CV and also non-CV outcome in

BASELINE CHARACTERISTICS. In the overall popu-

lation, mean age was 73  12 years; 36% were women; 23% had HFpEF, 21% had HFmrEF, 56% had HFrEF;

HFmrEF, with a significantly reduced risk of CV but not of non-CV events in HFrEF and no change in risk of either CV or non-CV outcome in HFpEF.

and median NT-proBNP was 2,640 pg/ml (inter-

ASSOCIATIONS BETWEEN HF THERAPY AND CV

quartile range [IQR]:1,140 to 5,914 pg/ml). Table 1 re-

VERSUS NON-CV OUTCOMES ACCORDING TO EF

ports

EF

CATEGORY AND NT-proBNP. The associations be-

category. HFmrEF was intermediate between HFpEF

tween ACE inhibitor or ARB and outcomes (both CV

patients’

characteristics

according

to

JACC: HEART FAILURE VOL.

-, NO. -, 2018

Savarese et al.

- 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

F I G U R E 1 Kaplan-Meier Curves Fitted for First CV and Non-CV Event According to the Ejection Fraction Category

(A) CV events (B) non-CV events. CI ¼ confidence interval; CV ¼ cardiovascular; HFmrEF ¼ heart failure with mid-range ejection fraction; HFpEF ¼ heart failure with preserved ejection fraction; HFrEF ¼ heart failure with reduced ejection fraction; HR ¼ hazard ratio.

and non-CV) observed in the overall population

outcome than outpatients regardless of EF, but the

were consistent in patients with NT-proBNP # and

risk of CV events was more increased in HFpEF versus

> median throughout the EF spectrum (p interaction

HFmrEF versus HFrEF, whereas HFpEF and HFmrEF

>0.05) (Figures 4 and 5). Similarly, no significant

inpatients reported worse non-CV outcome versus

interaction was observed between treatment with

those with HFrEF. The risk of non-CV events was

beta-blockers and NT-proBNP (# and > median) for

increased in HFrEF patients with versus without

CV and non-CV events in any EF category (Figure 5).

atrial fibrillation and $75 years of age versus

SUBGROUPS. The risk of CV events was increased in

<75 years of age, in those with versus without renal

all EF categories in patients with versus those

disease regardless of EF (even though the association

without atrial fibrillation, ischemic heart disease,

was not statistically significant after adjustments in

renal disease, and diabetes. Age $75 years versus <75

HFpEF and HFmrEF), in HFmrEF and HFrEF women

years and women versus men were associated with

versus men, and diabetic patients versus non-diabetic

improved CV outcome in HFrEF but not in HFpEF and

patients; however, no significant interaction between

HFmrEF, even though no significant interaction be-

these

tween these subgroups and EF was reported. In-

duration $6 months versus <6 months was associated

patients reported higher risk of CV but also of non-CV

with increased risk of CV events in HFmrEF and

subgroups

and

EF

was

observed.

HF

F I G U R E 2 Association Between Continuous NT-proBNP Levels and Risk of Outcomes in the Different Ejection Fraction Categories

The histogram reports the distribution of patients according to the NT-proBNP cutoff levels. NT-proBNP ¼ N-terminal pro–B-type natriuretic peptide; other abbreviations as in Figure 1.

5

6

Savarese et al.

JACC: HEART FAILURE VOL.

-, NO. -, 2018 - 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

F I G U R E 3 Associations Between HF Therapies and Outcomes

by Ejection Fraction

disease, in inpatients versus outpatients, in men versus women, in age <75 years versus $75 years, in creatinine clearance $60 ml/min versus <60 ml/min, and in HF duration $6 months versus <6 months. Therefore, in patients at higher risk, the increase in CV risk was steeper than in non-CV risk (Online Figures 4 to 11).

DISCUSSION NT-proBNP is useful to ensure the diagnosis of HF and to predict prognosis in HF, and it is thus used as an eligibility criterion in HF trials. Here we confirm that in a real-world registry setting, NT-proBNP is considerably elevated and prognostic in HF, regardless of EF. NT-proBNP is also used to enrich for CV events, but little is known about its influence on CV versus non-CV event rates in different EF categories. Here we observed complex relationships among EF, NT-proBNP, and CV and non-CV events. Namely, we observed greater non-CV risk with higher EF; and increasing CV-to-non-CV event ratio with increasing NT-proBNP concentrations in all EF categories but a steeper increase with higher EF. Finally, how NTACE-I ¼ angiotensin-converting enzyme inhibitor;

proBNP affects treatment response is unknown. We

ARB ¼ angiotensin receptor blocker; other abbreviations

did not observe any differences in associations be-

as in Figure 1.

tween treatment and outcomes according to NTproBNP concentration.

HFrEF and of non-CV events only in HFrEF (Online

ASSOCIATIONS BETWEEN EF CATEGORY AND CV VERSUS NON-CV EVENTS. Previous studies have re-

Figure 3). In all EF categories in all the subgroups explored,

ported conflicting data regarding prognosis in HF

crude CV event rates were higher than non-CV event

across the EF spectrum. In OPTIMIZE-HF (Organized

rates regardless of NT-proBNP levels. Consistent with

Program to Initiate Lifesaving Treatment in Hospital-

the main analysis, most of the increase in CV and non-

ized Patients with Heart Failure) study no differences

CV event rates occurred at lower NT-proBNP levels

in mortality and hospitalization risk were observed

with higher EF, but the splines for CV and non-CV

across EF categories (16). On the other hand, in the

event rates diverged at lower NT-proBNP in HFrEF

GWTG-HF (Get With The Guidelines - Heart Failure)

versus HFpEF and HFmrEF concentrations. In partic-

program, similar unadjusted and adjusted risk of

ular, the increase in both CV and non-CV event rates

mortality was reported in HFrEF versus HFmrEF

associated with the increase in NT-proBNP levels was

versus HFpEF, whereas crude and adjusted rates for

steeper in all subgroups with otherwise higher risk

CV and HF hospitalization were higher in HFrEF

profiles, that is, patients with versus without atrial

versus HFmrEF versus HFpEF, but those for overall

fibrillation, ischemic heart disease, and diabetes, in

readmissions were higher in HFmrEF versus HFpEF

age $75 years versus <75 years, in men versus women,

versus HFrEF (17). A previous analysis from SwedeHF

in inpatients versus outpatients, in creatinine clear-

showed higher crude mortality in HFpEF versus

ance <60 ml/min versus $60 ml/min, and in HF

HFmrEF and HFrEF but higher adjusted mortality

duration $6 months versus <6 months. Therefore, in

rates in HFrEF versus HFpEF and HFmrEF (1). Data for

patients

NT-proBNP

causes of death are limited, with candesartan in

conferred a more rapid increase in both CV and non-

CHARM (Candesartan in Heart Failure-Assessment of

CV risk patients. Similarly, splines diverged at lower

Reduction in Mortality and Morbidity) program

NT-proBNP levels in patients with higher risk (except

reporting highest CV mortality rates in EF <50% and

for younger patients and those with creatinine

highest non-CV death risk in EF $50% (18). We

clearance <60 ml/min), i.e., in those with versus

observed higher crude and adjusted CV event rates

without atrial fibrillation, diabetes, and ischemic heart

(CV mortality or first CV hospitalization) in HFrEF

at

higher

risk,

increasing

JACC: HEART FAILURE VOL.

-, NO. -, 2018

Savarese et al.

- 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

F I G U R E 4 Kaplan-Meier Curves Fitted for First CV and Non-CV event According to the Ejection Fraction Category, the Use of ACE Inhibitors or ARB and

NT-proBNP # and > Median Value

(Top panel) CV events; (bottom panel) non-CV events. Abbreviations as in Figures 1 to 3.

versus HFpEF and HFmrEF. This is consistent with

concentration is lower number of eligible patients

HFrEF having a greater CV risk profile (higher

and feasibility of adequate enrollment. However, this

NT-proBNP levels and New York Heart Association

differs by EF category. For example, an NT-proBNP

functional class and lower systolic blood pressure). In

concentration of 1,000 pg/ml would be able to pre-

contrast to previous studies, we have also reported

dict a CV event rate of 30 per 100 patient-years

non-CV event rates, which increased with higher EF in

regardless of EF. However, the CV-to-non-CV events

unadjusted and adjusted analyses. This is consistent

ratio would approximate 1.5 in HFrEF but only 1 in

with higher age, more comorbidities, and for example,

HFmrEF and 2/3 in HFpEF. Thus, 1,000 pg/ml could

lower hemoglobin and creatinine clearance with

be considered appropriate in HFrEF, but a higher

higher EF. However, even after we adjusted for these

cutoff would yield relatively more CV events in

factors, patients with higher EF had higher non-CV

HFmrEF and HFpEF. Of course, regardless of EF,

risk, suggesting there are other and unmeasured risk

given the shape of the splines, a higher NT-proBNP

factors for non-CV events in HFpEF in particular.

cutoff value would ensure very high event rates and

ASSOCIATIONS

BETWEEN

NT-proBNP

AND

CV

optimized CV-to-non-CV event ratio, but it would be

VERSUS NON-CV OUTCOMES IN HFpEF, HFmrEF,

more difficult to identify patients with such high NT-

AND HFrEF. NT-proBNP has been shown to be diag-

proBNP levels, at least in the outpatient setting. This

nostic and prognostic regardless of EF (4,11,12), but

was also shown when cutoff values commonly used in

less is known regarding its prediction of CV versus

clinical trials were used to calculate the correspond-

non-CV events. The splines for CV and non-CV event

ing CV and non-CV event rates in SwedeHF (Table 2).

rates according to NT-proBNP levels in the different

Regardless of trial and EF, a high proportion of Swe-

EF categories and relevant subgroups are complex to

deHF patients were eligible with trials’ NT-proBNP

interpret but provide comprehensive and detailed

cutoff values. In contrast, when many patients are

data that sponsors and trialists may wish to consider

screened for trials, many are not eligible. This may be

in trial design. Higher NT-proBNP levels will yield

explained by the tendency to screen younger and

more CV events and more non-CV events but, most

healthier patients with less comorbidity, who have

importantly, a higher ratio of CV to non-CV events.

lower NT-proBNP levels, and thus, are less likely to be

The

eligible.

trade-off

with

increasing

NT-proBNP

7

8

Savarese et al.

JACC: HEART FAILURE VOL.

-, NO. -, 2018 - 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

F I G U R E 5 Kaplan-Meier Curves Fitted for First Cardiovascular and Non-CV Event According to the Ejection Fraction Category, the Use of Beta-Blockers and

NT-proBNP < and > Median Value

(Top panel) CV events; (bottom panel) non-CV events. BB ¼ beta-blocker; other abbreviations as in Figures 1 and 2.

ASSOCIATIONS BETWEEN HF THERAPY AND CV

In I-PRESERVE (Irbesartan in Heart Failure with

VERSUS NON-CV OUTCOMES ACCORDING TO EF

Preserved Ejection Fraction) and TOPCAT (Treatment

CATEGORY AND NT-proBNP. Benefits of ACE inhibi-

of Preserved Cardiac Function Heart Failure with an

tor/ARB and beta-blocker therapy in HFrEF are well

Aldosterone

established (19–21), and 2 previous analyses in

ronolactone, respectively, were beneficial in patients

SwedeHF reported reduced mortality associated with

with low but not high natriuretic peptide levels

ACE

but

(10,24). This has been hypothesized to be associated

unchanged all-cause death and HF hospitalization

with high NT-proBNP levels linked to more severe

risk associated with beta-blockers in HFpEF and

and less reversible fibrosis and structural disease as

HFmrEF (22,23). The current analysis showed ACE

well as to several other comorbidities (more atrial

inhibitors were associated with reduced adjusted risk

fibrillation, worse renal function) that may be less

of CV events regardless of EF and beta-blocker ther-

responsive to HF treatments. On the other hand, in

apy associated with improved CV outcome in HFmrEF

HFrEF, valsartan in Val-HeFT (Valsartan Heart Failure

and HFrEF but not in HFpEF. Notably, similar results

Trial) and carvedilol in COPERNICUS (Carvedilol

were also reported for non-CV outcomes in HFmrEF

Prospective Randomized Cumulative Survival) were

and HFrEF for ACE inhibitor/ARB and in HFmrEF for

associated with similar prognosis in patients with

beta-blockers. The reasons are not readily apparent,

high versus low NT-proBNP levels, whereas in the

and observational data cannot demonstrate any

Australia-New Zealand HF study, carvedilol improved

potential treatment effect. Indeed, although we

outcome more in those with higher BNP and lower

adjusted extensively for comorbidities and risk

norepinephrine levels (25–27). In PARADIGM-HF

markers, there may have been remaining unmeasured

(Prospective Comparison of ARNI with ACEI to

selection bias and, thus, confounding explaining

Determine Impact on Global Mortality and Morbidity

better non-CV outcomes in treated patients. It is also

in Heart Failure), sacubitril/valsartan was effective

conceivable that HF therapy improves non-CV out-

regardless of NT-proBNP levels (13). Finally, in our

comes, for example, through overall better well-being

comprehensive analysis of HF patients, the associa-

and reduced frailty and complications from CV

tion between ACE inhibitor/ARB or beta-blockers and

morbidity and hospitalizations.

prognosis was not influenced by NT-proBNP levels

inhibitor/ARB

and

reduced

mortality

Antagonist),

irbesartan

and

spi-

JACC: HEART FAILURE VOL.

-, NO. -, 2018

Savarese et al.

- 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

T A B L E 2 CV and Non-CV Event Rates, Their Ratio, and Number of Eligible Patients in Swedish Heart Failure Registry According to

NT-proBNP Cutoffs Commonly Used in Trials

Trial (Ref. #)

EF Category

CV/non-CV ratio Eligible patients CV ER in Non-CV ER in in SwedeHF Calculated in SwedeHF NT-proBNP SwedeHF Calculated SwedeHF Calculated Considering the Calculated Considering AF Cutoff Considering the Considering the NT-proBNP the NT-proBNP Status Used, pg/ml NT-proBNP Cutoff* NT-proBNP Cutoff* Cutoff Used Cutoff Used,† n (%)

Dapa-HF (Online Ref. 1)

HFrEF

–

$600

47.2

28.8

1.6

Emperor-Reducedk(Online Ref. 2)

HFrEF

–

$600

47.2

28.8

1.6

7,977 (90)

Emperor-Preserved (Online Ref. 3)

HFpEF

SR

>300

35.6

43.9

0.8

1,032 (82)

AF

>900

53.6

48.2

1.1

2,012 (85)

SR

>300

34.0

35.2

1.0

1,091 (83)

AF

>900

49.6

39.1

1.3

1,720 (86)

SR

$400

39.6

25.5

1.6

3,900 (90)

AF

$1,200

57.7

33.9

1.7

3,848 (84)

HFmrEF GALACTIC-HF (Online Ref. 4)

HFrEF

7,977 (90)

PARADIGM-HF (Online Ref. 5)

HFrEF

–

$600

47.2

28.8

1.6

7,977 (90)

PARAGON-HF (Online Ref. 6)

HFpEF

SR

>300

35.6

43.9

0.8

1,032 (82)

AF

>900

53.6

48.2

1.1

2,012 (85)

SR

>300

34.0

35.2

1.0

1,091 (83)

AF

>900

49.6

39.1

1.3

1,720 (86)

HFmrEF‡ SPIRRIT-HFpEF (Online Refs. 6 and 7)

HFpEF HFmrEF

TOPCAT (Online Ref. 9) VICTORIA (Online Ref. 10)

SR

>300

35.6

43.9

0.8

1,032 (82)

AF

>750

52.0

47.7

1.1

2,091 (89)

SR

>300

34.0

35.2

1.0

1,091 (83)

AF

>750

48.5

38.4

1.3

1,786 (89)

HFpEF

–

$360

45.5

46.8

1.0

3,265 (90)

HFmrEF‡

–

$360

42.2

36.9

1.1

2,984 (90)

HFrEF HFmrEF§

SR

$1,000

43.2

26.2

1.6

3,285 (76)

AF

$1,600

60.6

34.5

1.8

3,532 (77)

SR

$1,000

44.2

42.2

1.0

759 (58)

AF

$1,600

55.7

41.1

1.4

1,392 (69)

*Per 100 patient-years. †Percent of the overall HFpEF or HFmrEF or HFrEF SwedeHF population with/without AF. ‡PARAGON and TOPCAT enrolled patients with EF of $45%. However, in our calculations, patients with EF of 40% to 45% were included. §VICTORIA enrolled patients with EF of <45. However, in our calculations, patients with EF 45% to 50% were included. kEmperor-Reduced had a different NT-proBNP cutoff for each EF strata (36% to 40%, 31% to 35%, #30%). However, we considered the same cutoff (600 pg/ml) for all strata. AF ¼ atrial fibrillation; CV ¼ cardiovascular; Dapa-HF ¼ Study to Evaluate the Effect of Dapagliflozin on the Incidence of Worsening Heart Failure or Cardiovascular Death in Patients With Chronic Heart Failure; Emperor-Preserved ¼ Empagliflozin outcome trial in patients with chronic heart failure with preserved ejection fraction; Emperor-Reduced ¼ Empagliflozin outcome trial in patients with chronic heart failure with reduced ejection fraction; EF ¼ ejection fraction; ER ¼ event rate; GALACTIC-HF ¼ Global Approach to Lowering Adverse Cardiac Outcomes Through Improving Contractility in Heart Failure; PARADIGM-HF ¼ Prospective Comparison of ARNI with ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure; PARAGON-HF ¼ Prospective Comparison of ARNI with ARB Global Outcomes in HF With Preserved Ejection Fraction; SPIRRIT-HFpEF ¼ Spironolactone Initiation Registry Randomized Interventional Trial in Heart Failure with Preserved Ejection Fraction; SR ¼ sinus rhythm; SwedeHF ¼ Swedish Heart Failure; TOPCAT ¼ Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist; VICTORIA ¼ Vericiguat Global Study in Subjects With Heart Failure With Reduced Ejection Fraction; other abbreviations as in Table 1.

across the EF spectrum. Differences in sample size,

STUDY LIMITATIONS. Observational data are subject

in median levels of NT-proBNP, in definitions of

to selection bias and confounding. Our analyses

HFpEF and HFrEF, and in selection of patients

investigating the associations between treatments

may contribute to explain the inconsistence of the

and prognosis were included primarily as an analogy

results among these studies. Thus, currently, whether

to the question of expected treatment effect that is

HF treatments may be more or less or similarly

posed by trialists. Although these analyses were

effective in patients with higher versus lower

extensively adjusted, we cannot rule out the influence

NT-proBNP

unresolved.

of unmeasured confounders and in no way imply that

Unrelated to the aim of the current analysis, we

our data represent any potential treatment effect. In

observed a low use of insertable cardioverter-

addition to EF category, numerous factors affect

defibrillator and cardiac resynchronization therapy

levels of NT-proBNP and its interaction with cova-

in our population, which is expected in HFpEF and

riates such as renal function, body mass index, and

HFmrEF but not in HFrEF patients. This low use in

cause-specific outcomes; thus, we cannot exclude the

Sweden has previously been shown to be a result of

fact that complex interactions among these may affect

primarily demographics and organizational factors

outcomes and/or potential treatment response. Pa-

and a lag in guideline implementation, such as poor

tients were included in the SwedeHF study based on

access to HF specialists (28–31).

clinician-judged HF; thus, we cannot exclude the fact

concentration

remains

9

10

Savarese et al.

JACC: HEART FAILURE VOL.

-, NO. -, 2018 - 2018:-–-

NT-proBNP for Eligibility and Enrichment in HF Trials

that some patients, particularly those with HFpEF, may not have HF. However, our analysis also has several strengths. We analyzed a large and unselected cohort of HF patients who reported NT-proBNP assessment. This allowed us to investigate the role of NT-proBNP in prognosis throughout the EF spectrum. Notably, in our analysis, we were able to consider cause-specific death and hospitalization.

Patient selection in HF trials requires optimization, particularly in HFpEF and HFmrEF. NT-proBNP concentration may have a key role for eligibility and enrichment in trials, but it is unknown which event rates, a high CV-to-non-CV event ratio, and not to dramatically reduce sample size. We observed

In HF trial design in different EF categories, NTproBNP may be a useful tool for eligibility and enrichment for CV events, but its role in treatment response remains unclear. When designing HF trials, it is important to recognize the facts that non-CV events will be higher with higher EF, that potential CV treatment effect will be lower with higher EF, even if the pathophysiological treatment target is appropriate, and that higher NT-proBNP criteria will enrich for CV events but also CV-to-non-CV event ratios. However, it remains unknown which NTproBNP concentration range may have the greatest HF treatment response.

complex relationships between NT-proBNP and CV and non-CV events, namely: greater non-CV risk with higher EF; increasing CV-to-non-CV event ratio with increasing NT-proBNP in all EF categories; but a steeper increase with higher EF. Finally, we did not observe any differences in associations between treatment and outcomes according to NT-proBNP concentration. TRANSLATIONAL OUTLOOK: Our analysis provides comprehensive and detailed data that sponsors and trialists may wish to consider in trial design. Higher NT-proBNP concentration will yield more CV events and more non-CV events, but most importantly, a higher CV-to-non-CV event ratio. The trade-

ADDRESS FOR CORRESPONDENCE: Dr. Gianluigi

Savarese, Department of Medicine, Cardiology Unit, Institutet, S1:02,

COMPETENCY IN MEDICAL KNOWLEDGE:

NT-proBNP value should be used in order to get high

CONCLUSIONS

Karolinska

PERSPECTIVES

171

76,

Stockholm,

off with increasing NT-proBNP concentration is lower number of eligible patients and feasibility of adequate enrolment, and this differs by EF category.

Sweden. E-mail: [email protected].

REFERENCES 1. Koh AS, Tay WT, Teng THK, et al. A comprehensive population-based characterization of heart failure with mid-range ejection fraction. Eur J Heart Fail 2017;19:1624–34. 2. Lund LH. Heart failure with “mid-range” ejection fraction-new opportunities. J Card Fail 2016; 22:769–71. 3. Vedin O, Lam CSP, Koh AS, et al. Significance of ischemic heart disease in patients with heart failure and preserved, midrange, and reduced ejection fraction: a nationwide cohort study. Circ Heart Fail 2017;10. pii:e003875. 4. Ponikowski P, Voors AA, Anker SD, et al. 2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur J Heart Fail 2016;18:891–975. 5. Butler J, Fonarow GC, Zile MR, et al. Developing therapies for heart failure with preserved ejection fraction: current state and future directions. J Am Col Cardiol HF 2014;2:97–112. 6. Sharma K, Kass DA. Heart failure with preserved ejection fraction: mechanisms, clinical features, and therapies. Circ Res 2014;115:79–96.

7. Paulus WJ, Tschope C. A novel paradigm for heart failure with preserved ejection fraction: comorbidities drive myocardial dysfunction and remodeling through coronary microvascular endothelial inflammation. J Am Coll Cardiol 2013; 62:263–71. 8. Lam CS, Lund LH. Microvascular endothelial dysfunction in heart failure with preserved ejection fraction. Heart 2016;102:257–9. 9. Lund LH, Oldgren J, James S. Registry-based pragmatic trials in heart failure: current experience and future directions. Curr Heart Fail Rep 2017;14: 59–70.

12. Savarese G, Musella F, D{’}Amore C, et al. Changes of natriuretic peptides predict hospital admissions in patients with chronic heart failure. A meta-analysis. J Am Col Cardiol HF 2014;2: 148–58. 13. Zile MR, Claggett BL, Prescott MF, et al. Prognostic implications of changes in N-terminal pro–B-type natriuretic peptide in patients with heart failure. J Am Coll Cardiol 2016;68:2425–36. 14. Jonsson A, Edner M, Alehagen U, Dahlstrom U. Heart failure registry: a valuable tool for improving the management of patients with heart failure. Eur J Heart Fail 2010;12:25–31.

10. Anand IS, Rector TS, Cleland JG, et al. Prognostic value of baseline plasma amino-terminal pro–brain natriuretic peptide and its interactions with irbesartan treatment effects in patients with heart failure and preserved ejection fraction:

15. Lund LH, Carrero JJ, Farahmand B, et al. Association between enrolment in a heart failure quality registry and subsequent mortality: a nationwide cohort study. Eur J Heart Fail 2017;19:

findings from the I-PRESERVE trial. Circ Heart Fail 2011;4:569–77.

16. Fonarow GC, Stough WG, Abraham WT, et al.

11. Savarese G, Hage C, Orsini N, et al. Reductions in N-terminal pro–brain natriuretic peptide levels are associated with lower mortality and heart failure hospitalization rates in patients with heart failure with mid-range and preserved ejection fraction. Circ Heart Fail 2016; 9. pii:e003105.

1107–16.

Characteristics, treatments, and outcomes of patients with preserved systolic function hospitalized for heart failure: a report from the OPTIMIZE-HF Registry. J Am Coll Cardiol 2007;50:768–77. 17. Shah KS, Xu H, Matsouaka RA, et al. Heart failure with preserved, borderline, and reduced ejection fraction: 5-year outcomes. J Am Coll Cardiol 2017;70:2476–86.

JACC: HEART FAILURE VOL.

-, NO. -, 2018

Savarese et al.

- 2018:-–-

18. Henkel DM, Redfield MM, Weston SA, Gerber Y, Roger VL. Death in heart failure: a community perspective. Circ Heart Fail 2008;1: 91–7. 19. Young JB, Dunlap ME, Pfeffer MA, et al. Mortality and morbidity reduction with candesartan in patients with chronic heart failure and left ventricular systolic dysfunction: results of the CHARM low-left ventricular ejection fraction trials. Circulation 2004;110:2618–26. 20. Group CTS. Effects of enalapril on mortality in severe congestive heart failure. Results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med 1987;316:1429–35. 21. Chatterjee S, Biondi-Zoccai G, Abbate A, et al. Benefits of beta blockers in patients with heart failure and reduced ejection fraction: network meta-analysis. BMJ 2013;346:f55. 22. Lund LH, Benson L, Dahlstrom U, Edner M. Association between use of renin-angiotensin system antagonists and mortality in patients with heart failure and preserved ejection fraction. JAMA 2012;308:2108–17.

NT-proBNP for Eligibility and Enrichment in HF Trials

24. Anand IS, Claggett B, Liu J, et al. Interaction between spironolactone and natriuretic peptides in patients with heart failure and preserved ejection fraction: from the TOPCAT trial. J Am Col

29. Lund LH, Braunschweig F, Benson L, Stahlberg M, Dahlstrom U, Linde C. Association between demographic, organizational, clinical, and socioeconomic characteristics and

Cardiol HF 2017;5:241–52.

underutilization of cardiac resynchronization therapy: results from the Swedish Heart Failure Registry. Eur J Heart Fail 2017;19: 1270–9.

25. Hartmann F, Packer M, Coats AJ, et al. Prognostic impact of plasma N-terminal pro– brain natriuretic peptide in severe chronic congestive heart failure: a substudy of the Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS) trial. Circulation 2004;110: 1780–6. 26. Richards AM, Doughty R, Nicholls MG, et al. Neurohumoral prediction of benefit from carvedilol in ischemic left ventricular dysfunction. Australia-New Zealand Heart Failure Group. Circulation 1999;99:786–92. 27. 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;107:1278–83.

23. Lund LH, Benson L, Dahlstrom U, Edner M, Friberg L. Association between use of beta-

28. Lund LH, Svennblad B, Dahlstrom U, Stahlberg M. Effect of expanding evidence and evolving clinical guidelines on the prevalence of

blockers and outcomes in patients with heart failure and preserved ejection fraction. JAMA 2014;312:2008–18.

indication for cardiac resynchronization therapy in patients with heart failure. Eur J Heart Fail 2017 Sep 26 [E-pub ahead of print].

30. Thorvaldsen T, Benson L, Dahlstrom U, Edner M, Lund LH. Use of evidence-based therapy and survival in heart failure in Sweden 2003-2012. Eur J Heart Fail 2016;18:503–11. 31. Lund LH, Benson L, Stahlberg M, et al. Age, prognostic impact of QRS prolongation and left bundle branch block, and utilization of cardiac resynchronization therapy: findings from 14,713 patients in the Swedish Heart Failure Registry. Eur J Heart Fail 2014;16: 1073–81.

KEY WORDS eligibility, heart failure, N-terminal pro–B-type natriuretic peptide, registry, trials

A PP END IX For supplemental references and figures, please see the online version of this paper.

11