J A C C : C A R D I O V A S C U L A R I M A G I N G V O L . 5 , N O . 2 , 2 0 1 2

© 2 0 1 2 B Y T H E A M E R I C A N C O L L E G E O F C A R D I O L O G Y F O U N D A T I O N I S S N 1 9 3 6 - 8 7 8 X / $ 3 6 . 0 0

P U B L I S H E D B Y E L S E V I E R I N C . D O I : 1 0 . 1 0 1 6 / j . j c m g . 2 0 1 1 . 1 1 . 0 1 1

O R I G I N A L R E S E A R C H

Echocardiography, Natriuretic Peptides, andRisk for Incident Heart Failure in Older AdultsThe Cardiovascular Health Study

Andreas P. Kalogeropoulos, MD,* Vasiliki V. Georgiopoulou, MD,* Christopher R. deFilippi, MD,†John S. Gottdiener, MD,† Javed Butler, MD, MPH,* for the Cardiovascular Health Study

Atlanta, Georgia; and Baltimore, Maryland

O B J E C T I V E S This study sought to examine the potential utility of echocardiography and

N-terminal pro–B-type natriuretic peptide (NT-proBNP) for heart failure (HF) risk stratification in concert

with a validated clinical HF risk score in older adults.

B A C K G R O U N D Without clinical guidance, echocardiography and natriuretic peptides have sub-

optimal test characteristics for population-wide HF risk stratification. However, the value of these tests

has not been examined in concert with a clinical HF risk score.

M E T H O D S We evaluated the improvement in 5-year HF risk prediction offered by adding an

echocardiographic score and/or NT-proBNP levels to the clinical Health Aging and Body Composition (ABC)

HF risk score (base model) in 3,752 participants of the CHS (Cardiovascular Health Study) (age 72.6 � 5.4

years; 40.8% men; 86.5% white). The echocardiographic score was derived as the weighted sum of independent

echocardiographic predictors of HF. We assessed changes in Bayesian information criterion (BIC), C index,

integrated discrimination improvement (IDI), and net reclassification improvement (NRI). We examined also the

weighted NRI across baseline HF risk categories under multiple scenarios of event versus nonevent weighting.

R E S U L T S Reduced left ventricular ejection fraction, abnormal E/A ratio, enlarged left atrium, and increased

left ventricular mass were independent echocardiographic predictors of HF. Adding the echocardiographic score

and NT-proBNP levels to the clinical model improved BIC (echocardiography: �43, NT-proBNP: �64.1, combined:

�68.9; all p � 0.001) and C index (baseline: 0.746; echocardiography: �0.031, NT-proBNP: �0.027, combined:

�0.043; all p � 0.01), and yielded robust IDI (echocardiography: 43.3%, NT-proBNP: 42.2%, combined: 61.7%; all

p� 0.001), andNRI (based onHealth ABCHF risk groups; echocardiography: 11.3%; NT-proBNP: 10.6%, combined:

16.3%; all p � 0.01). Participants at intermediate risk by the clinical model (5% to 20% 5-yr HF risk; 35.7% of the

cohort) derived the most reclassification benefit. Echocardiography yielded modest reclassification when used

sequentially after NT-proBNP.

C O N C L U S I O N S In older adults, echocardiography and NT-proBNP offer significant HF risk

reclassification over a clinical prediction model, especially for intermediate-risk individuals. (J Am Coll

Cardiol Img 2012;5:131–40) © 2012 by the American College of Cardiology Foundation

From the *Emory Clinical Cardiovascular Research Institute, Emory University, Atlanta, Georgia; and the †Maryland HeartCenter, University of Maryland, Baltimore, Maryland. The research reported in this paper was supported by contract numbersN01-HC-85239, N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, N01-HC-45133, grant number HL080295 from the National Heart, Lung, and Blood Institute (NHLBI) and grant

number AG-023269 from the National Institute on Aging (NIA), with additional contribution from the National Institute ofNeurological Disorders and Stroke (NINDS). Additional support was provided through AG-15928, AG-20098, and AG-027058

S

p

sf5scpllsa12ad1oiaaa

sfcULnafi

A

A

pro–B-type natriuretic pepti

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

132

ubclinical changes in cardiac structure andfunction, including left ventricular (LV) hy-pertrophy, enlargement, wall motion abnor-malities, diastolic dysfunction, or reduced

ejection fraction (EF), and left atrial enlargement,often precede development of manifest (Stage C)heart failure (HF) (1). These changes, detectedmostly through echocardiography, have beenstrongly associated with HF risk in large cohortstudies (2–7). Similarly, elevated natriuretic peptidelevels are associated with structural and functionalcardiac abnormalities (8), and in turn, with HF risk(8–11). However, the value of echocardiography

See page 141

and natriuretic peptides for HF risk stratification has notbeen assessed in concert with a validated clinical risk

score. Moreover, due to suboptimal test char-acteristics from a screening perspective, andconcerns regarding costs and consequences ofunnecessary testing (4,12–16), these tests arenot recommended currently as standalonetools for identification of individuals at highrisk for HF. However, guidance by a clinicalrisk score, for example, the Health Aging andBody Composition (Health ABC) HF riskscore (17,18), may render these tests moreappealing as risk stratification tools for targetedpopulation groups.

We hypothesized that echocardiographyand N-terminal pro-B-type natriuretic peptide(NT-proBNP) levels, used individually or incombination (simultaneously or sequentially),will have incremental predictive value over the

clinical Health ABC HF risk score for 5-year HF riskassessment in older adults and yield clinically relevant riskreclassification. To evaluate this hypothesis, we analyzeddata from 3,752 participants of CHS (CardiovascularHealth Study) with available echocardiography and NT-proBNP data.

M E T H O D S

Study population. The design of CHS has beenreviously published (19). Participants were nonin-

from the NIA, HL-075366 from NHLBI, and the University of PittsburgCenter P30-AG-024827. This project was also partially funded by anentitled “Novel Risk Markers and Prognosis Determination in Heart Failfrom the Clinical and Translational Science Award program, National InstituAll authors have reported that they have no relationships relevant to the cont

ion

ion

de

Manuscript received July 11, 2011; revised manuscript received October 20,

titutionalized persons 65 to 100 years old, recruitedrom Medicare eligibility lists. An original cohort of,201 persons was recruited in 1989 to 1990, and aecond cohort of 687 African Americans was re-ruited in 1992 to 1993. The present study includedarticipants with available baseline NT-proBNP

evels and echocardiographic data, obtained at base-ine for the original cohort and at year 2 for theecond cohort. From the 4,522 participants withvailable baseline NT-proBNP levels, we excluded:) participants with prevalent HF at baseline (n �12) and those from the second cohort with prev-lent HF at year 2 (n � 35); 2) those with missingata on Health ABC HF risk score variables (n �62); and 3) those with missing data on echocardi-graphy (n � 361). In all, 3,752 participants werencluded in the primary analysis. In a secondarynalysis, we examined 2,538 participants with avail-ble quantitative M-mode data on LV dimensionsnd mass.Definition of risk factors and Health ABC HF score. Wecalculated the Health ABC HF risk score andcategorized participants into �5%, 5% to 10%, 10%to 20%, and �20% 5-year HF risk, as previouslydescribed (17,18). To incorporate the second CHScohort, we calculated the score at baseline for theseparticipants; however, echocardiography for theseparticipants was available at year 2. For calculationof Health ABC HF risk score, classification ofprevalent coronary heart disease was based on self-report of coronary revascularization, myocardial in-farction, or angina (20). Smoking was classified ascurrent, past, or never. The Minnesota 3.1 code wasused to classify electrocardiographic LV hypertro-phy (21). The core laboratory at University ofVermont, Colchester, Vermont, analyzed fastingserum chemistry.Echocardiographic assessment in CHS. Two-dimen-ional echocardiography was performed at baselineor the original cohort and at year 2 for the secondohort. Studies were centrally interpreted at theniversity of California, Irvine (22). We classifiedV systolic function as normal, borderline, or ab-ormal, corresponding to EF �0.55, 0.45 to 0.54,nd �0.45, respectively (23). Based on previousndings from CHS showing values �0.7 and �1.5

laude D. Pepper Older Americans Independencery University Heart and Vascular Board grant

and Public Health Service Grant UL1 RR025008f Health, National Center for Research Resources.of this paper to disclose.

h CEmoure”tes oents

B B R E V I A T I O N S

N D A C R O N YM S

BIC � Bayesian information

criterion

EF � ejection fraction

HF � heart failure

IDI � integrated discriminat

improvement

LV � left ventricular

LVEF � left ventricular eject

fraction

NRI � net reclassification

improvement

NT-proBNP � N-terminal

2011, accepted November 17, 2011.

aEtLptpcdgSamL

cbctntpHnizafi

NA

pdtrpwS

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

133

to be associated with increased HF risk (4), themitral inflow E to A peak velocity ratio (E/A ratio)was classified into �0.7, 0.7 to 1.5, and �1.5. Lefttrial enlargement was based on American Society ofchocardiography recommended M-mode anteropos-

erior diameter of �23 mm/m2 (24). QuantitativeV M-mode measurements were available in 3,410articipants of the original cohort, with age beinghe strongest correlate of missing data (23), and 405articipants of the second cohort. LV mass wasalculated with the Devereux formula (25). Toefine LV hypertrophy and enlargement, we usedender-specific cutoffs proposed by the Americanociety of Echocardiography: 95 g/m2 for womennd 115 g/m2 for men for LV mass; and �32m/m2 for women and �31 mm/m2 for men forV diameter at diastole (24).

Natriuretic peptide assay. NT-proBNP was mea-sured with the Elecsys 2010 system (Roche Diag-nostics, Indianapolis, Indiana) in serum collected atbaseline. Coefficient of variation was 2% to 5%, andrange was 5 to 35,000 pg/ml. In a previous analysisfrom CHS (9), NT-proBNP �190 pg/ml optimallypredicted HF risk and cardiovascular mortality; wetherefore used this cutoff for additional analyses.Definition of incident HF. Methods used to assessardiovascular events in CHS, including HF, haveeen reported previously (26,27). An HF event wasonfirmed if, in addition to a physician diagnosis,here were: 1) documented symptoms (e.g., short-ess of breath, fatigue, orthopnea, paroxysmal noc-urnal dyspnea) and physical signs (e.g., edema,ulmonary rales, gallop rhythm) consistent withF; 2) supporting clinical findings such as pulmo-

ary edema on chest x-ray; or 3) therapy for HF,ncluding diuretics, digitalis, angiotensin-converting en-yme inhibitors, or beta-blockers. In the currentnalysis, incident HF was defined as adjudicatedrst hospitalization for HF.

Statistical analysis. NT-proBNP values were log-transformed before analyses because of lognormaldistribution. To identify echocardiographic vari-ables that predict HF independently of clinical riskfactors, we used the Health ABC HF risk score asa covariate in multivariable Cox proportional haz-ards models and followed an iterative, clinicallyguided approach to reach the final echocardiography-added model. The final set of echocardiographicpredictors was treated subsequently as a single score(weighted sum of covariates) to assess modelimprovement.

To assess the incremental value of NT-proBNP

and echocardiographic score for 5-year HF risk

prediction, we used Health ABC HF risk score asthe base clinical model and calculated: 1) changes inthe Bayesian information criterion (BIC), alikelihood-based measure of model fit that penalizesfor complexity (28); 2) changes in the C index;3) the integrated discrimination improvement(IDI), which quantifies improvement in prognosticseparation (29,30), and relative IDI (rIDI), whichquantifies the relative contribution of an addedvariable to separation (30); and 4) the net reclassi-fication improvement (NRI), a clinically orientedmeasure of reclassification that considers any cor-rect upwards movement of events (in the new vs.the old model) and any downwards movement ofnon-events as model improvement and vice versa(29). To facilitate prospective interpretation, weused the censored data versions of IDI (31) andNRI (32). The Kaplan-Meier estimates for 5-yearHF incidence served as observed event rates. Wecalculated the continuous NRI, which counts allmovements of probabilities and does not rely oncategories (NRI0), and the NRI based on 5-yearrisk groups (i.e., �5%, 5% to 10%, 10% to 20%, and�20%) (17), which counts only movements ofpredicted probabilities that result in a change in riskcategory (NRIcat). Because event and non-eventNRI components are weighted by event incidence(32), and thus NRI is not comparable across riskcategories, we calculated the weighted NRI forevents and non-events (wNRI) (32) to comparereclassification across baseline risk categories. Cur-rently, no established thresholds for action regard-ing HF risk exist, and thus, the relative merit ofcorrectly classifying an event versus a non-event isunknown. We thus calculated wNRI by assigningweights from 1:1 (equal importance) to 16:1 (events16� more important). The NRI formulas (NRI0,

RIcat, and wNRI) are provided in the Onlineppendix and Online Table 1.To examine the incremental value of NT-

roBNP and echocardiographic score for risk pre-iction among participants with normal left ven-ricular ejection fraction (LVEF) at baseline, weeplicated the analyses in the subcohort of partici-ants with LVEF �0.55. Analyses were performedith STATA version 11 (StataCorp LP, Collegetation, Texas).

R E S U L T S

Clinical characteristics and HF incidence. Average agewas 72.6 � 5.4 years; 40.8% were men; 86.5% were

white. Table 1 summarizes the clinical characteris-

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

134

tics of participants by Health ABC HF risk group.There were 286 incident HF events during the first5 years of follow-up, corresponding to 8.0% 5-yearHF incidence.Echocardiographic findings and NT-proBNP levels.Figure 1 presents the prevalence of echocardio-

Table 1. Baseline Clinical Characteristics According to Health AB

<5%(n � 2,228)

Age, yrs 70.6 (3.9)

Men, % 34.0

White, % 88.5

Smoking, %

Current 6.3

Former 42.3

Coronary heart disease 3.4

Left ventricular hypertrophy, %* 1.2

Systolic blood pressure, mm Hg 129 (17)

Heart rate, beats/min 63 (10)

Creatinine, mg/dl 0.97 (0.23)

Fasting glucose, mg/dl 105 (22)

Albumin, g/dl 4.04 (0.28)

Total cholesterol, mg/dl 215 (38)

Values are n (%) or %. *By electrocardiography.ABC � Aging and Body Composition; HF � heart failure.

Figure 1. Abnormal Echocardiographic Findings According to P

Prevalence of abnormal echocardiographic findings according to pr(HF) at baseline. M-mode left ventricular diameter and mass were amm/m2 (24). †Indexed internal diameter at diastole �32 mm/m2 fo�95 g/m2 for women, �115 g/m2 for men (24). Parameters indexe

for all parameters.

graphic abnormalities according to clinically pro-jected HF risk by the Health ABC HF risk score.Left atrial diameter, LV diameters, and LV mass,all had strong increasing trends across Health ABCHF risk groups (Table 2). Abnormal EF (�0.45%)or restrictive-type diastolic (E/A �1.5) dysfunction

F Risk Group

Health ABC 5-yr HF Risk

5%–10%(n � 907)

10%–20%(n � 433)

>20%(n � 184)

74.4 (5.4) 76.5 (6.0) 79.3 (6.9)

48.2 56.1 51.6

83.7 83.8 82.6

15.3 16.4 22.8

42.1 41.8 33.2

25.3 50.8 66.9

3.9 11.5 24.5

143 (21) 149 (23) 158 (25)

66 (12) 68 (13) 74 (18)

1.09 (0.28) 1.21 (0.39) 1.38 (0.64)

117 (40) 125 (46) 139 (85)

3.97 (0.28) 3.96 (0.30) 3.86 (0.28)

211 (40) 206 (39) 206 (43)

cted HF Risk

ted Health ABC HF risk in 3,752 participants without heart failureble in 2,538 participants. *Indexed left atrial diameter �23men, �31 mm/m2 for men (24). ‡Indexed left ventricular massr body surface area. p � 0.001 for linear trend across categories

C H

roje

ojecvailar wod fo

tratasLpecrw(tetcf

aAiipad

scscmiioit

tpicNI

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

135

was present in 5.8%, 6.5%, 11.5%, and 16.8% ofparticipants in the escalating HF risk groups, re-spectively. Absolute NT-proBNP levels and preva-lence of increased (�190 pg/ml) NT-proBNP lev-els both had strong increasing trends across clinicalHF risk groups (Fig. 2).Echocardiographic parameters, NT-proBNP, and HF riskprediction. Among the echocardiographic parame-ers examined, EF, enlarged left atrium, and E/Aatio were independent predictors of HF risk whendded to Health ABC HF risk score (Table 3). Inhe subset with M-mode data on LV dimensionsnd mass available (n � 2,538), LV fractionalhortening was a stronger predictor than EF, andV mass index was an additional independentredictor of HF (Table 3). The weighted sum ofchocardiographic parameters constituted the echo-ardiographic score used subsequently to determineisk stratification improvement. For completeness,e have created separate scores for the main cohort

EF, enlarged left atrium, and E/A ratio) and forhe quantitative LV subset (fractional shortening,nlarged left atrium, and E/A ratio, LV mass) toest for incremental value, although only the mainohort parameters were converted to a point systemor practical use.

Table 2. Quantitative Echocardiographic Characteristics Accordi

N <5%

Left atrial diameter index, mm/m2 3,752 21.2 (3.5)

LV diastolic diameter index, mm/m2 2,538 27.3 (3.4)

LV systolic diameter index, mm/m2 2,538 15.6 (3.1)

LV mass index, g/m2 2,538 79 (20)

Values are n (%).LV � left ventricular; other abbreviations as in Table 1.

Figure 2. NT-proBNP Levels According to Projected HF Risk

Distribution of N-terminal pro–B-type natriuretic peptide (NT-proBNNT-proBNP levels (right) according to projected Health ABC HF risk

HF � heart failure.

NT-proBNP levels were independently associ-ted with 5-year HF risk when added to HealthBC HF risk score; hazard ratio per natural log-

ncrease of NT-proBNP was 1.61 (95% confidencenterval: 1.44 to 1.81; p � 0.001). Increased NT-roBNP as a binary variable (�190 pg/ml) wasssociated with 2.7-fold increased risk (95% confi-ence interval: 2.1 to 3.6; p � 0.001).Both NT-proBNP and echocardiographic score

ignificantly improved fit, discrimination, and risklassification when added to Health ABC HF riskcore (Table 4). Combined NT-proBNP and echo-ardiography had further incremental value over eachodality alone, achieving further improvements in

ntegrated discrimination (0.064, or 61.7% relativencrease) and net reclassification (35.9% for continu-us NRI and 16.3% for category-based NRI). Thempact of echocardiography was stronger when quan-itative LV parameters were considered (Table 4).

When the echocardiographic score was used sequen-ially after adding NT-proBNP to the base model, it stillrovided significant but less robust improvements inntegrated discrimination (0.019, or 12.8% relative in-rease) and net reclassification (20.0% for continuousRI and 7.9% for category-based NRI; both p � 0.01).

n the subset with quantitative LV data, echocardiogra-

o Health ABC HF Risk

Health ABC 5-Year HF Risk

5%–10% 10%–20% >20% p for Trend

21.5 (3.8) 22.2 (4.3) 23.0 (4.3) �0.001

27.4 (3.8) 27.8 (4.0) 28.2 (4.6) 0.024

16.2 (3.6) 16.6 (4.3) 17.2 (4.5) �0.001

85 (23) 93 (30) 94 (33) �0.001

vels (left) and prevalence of increased (�190 pg/ml) (9)0.001 for linear trend across categories in both analyses.

ng t

P) le. p �

n

ld1o�

ad

tions

relative integrated discrim

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

136

phy further improved classification compared to theNT-proBNP–added model (0.041, or 30.1% for IDI;31.0% for continuous NRI [p � 0.001]; and 13.2% forcategory-based NRI [p � 0.01]).Reclassification of HF risk according to baseline HF riskgroup. In participants with �5% projected risk,

either echocardiographic score nor NT-proBNP

Table 3. Echocardiographic Predictors of Incident HF

HR (95%

Main cohort (n � 3,752)

LV ejection fraction

Borderline (0.45 to 0.54) 1.20 (0.76–1

Abnormal (�0.45) 3.48 (2.35–5

E/A ratio

�0.7 1.61 (1.24–2

�1.5 3.23 (2.25–4

Enlarged LA (�23 mm/m2) 1.60 (1.27–2

Wall motion abnormalities (yes, no) 1.54 (1.13–2

Quantitative LV subset (n � 2,538)

LV ejection fraction

Borderline (0.45 to 0.54) 1.47 (0.84–2

Abnormal (�0.45) 3.78 (2.22–6

LV fractional shortening, per 5% decrease 1.23 (1.13–1

E/A ratio

�0.7 1.98 (1.42–2

�1.5 3.27 (2.08–5

Enlarged LA (�23 mm/m2) 1.55 (1.15–2

LV end-diastolic diameter, per cm/m2 1.69 (1.18–2

LV end-systolic diameter, per cm/m2 2.40 (1.71–3

LV mass index, per 10 g/m2 1.16 (1.10–1

*Adjusted for Health ABC HF Risk Score.CI � confidence interval; HR � hazard ratio; LA � left atrium; other abbrevia

of Echocardiographic Score and NT-proBNP for HF Risk Predicti

BICp Value(�BIC) C

p Value(�C)

2)

e 4,424.8 Reference 0.746 Reference

ded 4,381.8 �0.001 0.777 �0.001 0

4,360.7 �0.001 0.773 0.004 0

d 4,355.9 �0.001 0.789 �0.001 0

(n � 2,538)

2,607.5 Reference 0.744 Reference

ded 2,532.2 �0.001 0.798 �0.001 0

2,570.1 �0.001 0.774 0.014 0

d 2,520.1 �0.001 0.805 �0.001 0

improvement. NRI0 represents continuous NRI, which counts any movement of ply movements that result in change in risk category.on criterion; IDI � integrated discrimination improvement; NRI � net reclassifica

ination improvement; other abbreviations as in Tables 1 and 2.

evel provided clinically relevant reclassification un-er a wide range of scenarios (Fig. 3). In the 5% to0% category, clinically relevant reclassification wasbserved both with echocardiography (wNRI15%) and NT-proBNP (wNRI �10%) across

ssigned weights. The 10% to 20% HF risk categoryerived the most reclassification benefit both from

variate* Multivariate*

p Value HR (95% CI) p Value

0.43 1.17 (0.74–1.84) 0.50

�0.001 2.91 (1.96–4.33) �0.001

�0.001 1.56 (1.19–2.03) 0.001

�0.001 2.70 (1.87–3.09) �0.001

�0.001 1.50 (1.18–1.91) 0.001

0.007 — —

0.17 — —

�0.001 — —

�0.001 1.16 (1.07–1.27) 0.001

�0.001 1.72 (1.22–2.42) 0.002

�0.001 2.92 (1.84–4.63) �0.001

0.004 1.42 (1.04–1.93) 0.028

0.005 — —

�0.001 — —

�0.001 1.11 (1.06–1.17) �0.001

as in Tables 1 and 2.

I (rIDI) NRI0p Value(NRI0) NRIcat

p Value(NRIcat)

ference Reference Reference Reference Reference

4 (42.2%) 31.8% �0.001 10.6% 0.007

5 (43.3%) 31.7% �0.001 11.3% 0.007

4 (61.7%) 35.9% �0.001 16.3% �0.001

ference Reference Reference Reference Reference

6 (53.7%) 39.7% �0.001 18.8% 0.001

3 (32.5%) 26.6% 0.001 12.1% 0.024

5 (72.8%) 50.2% �0.001 24.5% �0.001

ted probabilities; NRIcat is based on �5%, 5% to 10%, 10% to 20%, and �20%

improvement; NT-proBNP � N-terminal pro–B-type natriuretic peptide; rIDI �

Uni

CI)

.89)

.15)

.10)

.63)

.03)

.11)

.57)

.46)

.34)

.77)

.14)

.08)

.43)

.36)

.21)

Table 4. Added Value on

Statistic ID

Main cohort (n � 3,75

Health ABC HF Scor Re

Echocardiography ad .04

NT-proBNP added .04

Echocardiography anNT-proBNP added

.06

Quantitative LV subset

Health ABC HF score Re

Echocardiography ad .05

NT-proBNP added .03

Echocardiography anNT-proBNP added

.07

rIDI expresses the relative rediccategories, and counts onBIC � Bayesian informati tion

LdtaoWsp

tvhifvq(d

ighte

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

137

echocardiography and NT-proBNP; wNRI was�20% for both modalities regardless of weightsassigned to events versus nonevents. In the �20%risk category, reclassification gain was evident onlywhen 1:1 weights were assigned. Modest synergisticgains were observed when echocardiography andNT-proBNP were used simultaneously. When riskwas first refined by NT-proBNP, echocardiographyprovided gains over the NT-proBNP–added model(�10% wNRI) in the 5% to 10% and 10% to 20%categories.Echocardiography and NT-proBNP in participants withnormal LVEF. Among the 3,480 participants with

VEF �0.55, there were 236 incident HF eventsuring the first 5 years of follow-up, correspondingo 7.1% 5-year HF incidence. Both echocardiographynd NT-proBNP conferred significant reclassificationf HF risk in these participants (Online Table 2).

hen measured by the IDI, the magnitude of reclas-ification benefit by echocardiography or NT-

Figure 3. NRI by Echocardiography and NT-proBNP According t

Net reclassification improvement (NRI) across the Health ABC heartscore (A), N-terminal pro–B-type natriuretic peptide (NT-proBNP) (Bscore is added after refinement of risk with NT-proBNP. Five differecorrect reclassification of events versus non-events, representing inis equally important for events and non-events, whereas 16:1 denothan correct reclassification of a non-event. Clinically relevant correwith echocardiography (A) and NT-proBNP (B) across assigned weig20% (yellow bars) heart failure risk as clinically determined by the HeaproBNP and echocardiography were used simultaneously (C). When risincremental benefit. wNRI indicates the average weighted NRI, the we

roBNP in this subcohort was diminished relative to

hat observed in the entire study cohort, because riskariance was lower in this relatively healthier subco-ort. However, both the continuous and the categor-

cal NRI for echocardiography were practically unaf-ected. NT-proBNP exhibited slightly smaller NRIalues relative to the entire cohort. In the subset withuantitative LV measurements and LVEF �0.55n � 2,384), the reclassification properties of echocar-iography were enhanced.

D I S C U S S I O N

In this study, we demonstrated that echocardiogra-phy and NT-proBNP could have incremental valueover a validated clinical model for HF risk predic-tion in older adults. Improvements in HF riskprediction were not limited to formal improvementsin model fit and discrimination, but also resulted inclinically relevant risk reclassification among the35.7% of participants deemed to be at intermedi-

seline HF Risk

re (HF) risk categories with addition of the echocardiographicboth (C) to the Health ABC HF model; (D) echocardiographicenarios are presented with varying weights (from 1:1 to 16:1) forsing importance of events: 1:1 denotes that correct reclassificationhat correct reclassification of an event is 16 times more importantclassification (indicated by �10% positive bars) was observedfor participants with baseline 5% to 10% (green bars) or 10% toABC HF risk score. Modest synergistic gains were observed when NT-as first refined by NT-proBNP (D), echocardiography provided modestd percentage of participants correctly reclassified.

o Ba

failu), ornt sccreates tct rehtslthk w

ated risk (5% to 10% and 10% to 20% predicted

t8wdLet8gfvg

coLAfetsfc

tuipte

cehniboieOsfsudnarHa

d(lsp

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

138

5-year HF risk) by the clinical score. Of note, LVdiastolic function proxy parameters (E/A ratioand left atrial size) were independent determi-nants of HF risk in this cohort. The risk reclas-sification potential of echocardiographic param-eters and NT-proBNP was also evident amongparticipants with normal (�0.55) LVEF at base-line. Quantitative assessment of LV mass furtherstrengthened the value of echocardiography for riskreclassification; this was more pronounced in par-ticipants with normal LVEF, potentially becauseLV mass plays a central role in HF risk determina-tion in these individuals.

There is a fundamental difference betweenscreening for subclinical disease and risk stratifica-tion for future disease. The detection of the formeris by definition the “end product,” for example,breast cancer, which necessitates treatment. Thelatter needs further evidence for effective preventionstrategies and action thresholds to be clinicallyrelevant. Currently, besides treatment of asymp-tomatic LV systolic dysfunction, there are no spe-cific HF prevention strategies beyond usual riskfactor management, for example, hypertension.However, most participants who developed HF inCHS did not have reduced EF at baseline (4), andthe prevalence of asymptomatic LV systolic dys-function in the community is too low to justifypopulation-wide screening (16). In fact, even com-bining abnormal systolic (EF �0.45) or restrictive-ype diastolic (E/A �1.5) dysfunction would miss0% of CHS participants who developed HFithin 5 years. On the other hand, if all echocar-iographic predictors in our study (any degree ofV systolic or diastolic dysfunction or left atrialnlargement) were considered as subclinical disease,he yield would be too high (43%, 53%, 65%, and0% for the ordered Health ABC HF risk scoreroups) to be clinically meaningful. Thus, screeningor asymptomatic abnormalities is unlikely to beery effective by itself for HF prevention in theeneral population of older adults.

In line with previous studies (4,6,33), the echo-ardiographic markers associated with HF risk inur study were related to both systolic and diastolicV function. Despite adjustment for the HealthBC HF risk score, which incorporates clinical risk

actors associated with diastolic dysfunction (17),chocardiographic surrogates of diastolic dysfunc-ion were strongly associated with HF risk in ourtudy, highlighting the importance of diastolic dys-unction when considering incorporation of echo-

ardiography in HF prevention efforts (34). Al-

hough we cannot suggest that echocardiography besed for HF risk stratification at this point, consider-ng that a number of patients undergo echocardiogra-hy for other indications, we provide a modificationable for the Health ABC HF risk score based onchocardiographic findings (Online Table 3).

Similar to echocardiography, most studies haveoncluded that natriuretic peptides are not costffective for detection of asymptomatic structuraleart disease (35,36). This is rooted in the fact thatatriuretic peptides are conceptually used to select

ndividuals for echocardiographic screening (16),earing all the limitations noted above. Based onur results, the effect of echocardiographic screen-ng on HF prevention would still be modest at best,ven after NT-proBNP–based selection (16,37).n the other hand, NT-proBNP refined risk as-

essment in our study, in line with observationsrom younger cohorts (38,39). Currently, no dataupport use of NT-proBNP as a HF risk tool innselected older adults. However, our findings un-erscore the potential of a strategy incorporatingatriuretic peptides for intermediate-risk individu-ls when considering HF prevention efforts. In thisespect, we provide a practical approach to refine

F risk based on NT-proBNP levels if these arevailable (Online Table 4).

Abnormal systolic or restrictive-type diastolic LVysfunction often has causes amenable to treatmentischemia, cardiomyopathy, LV hypertrophy, valvu-ar disease, and so on). Therefore, it might be worthcreening with echocardiography older adults withrojected 10% to 20% or �20% 5-year HF risk. In

these groups, which combined constituted 16.5% ofour cohort, abnormal systolic or restrictive-typediastolic dysfunction was present in 11.5% and16.8%, respectively. Based on analyses showing thatprevalence rates of �5% in the target population arelikely to result in cost-effective echocardiographicscreening (37), the Health ABC HF risk scoremight serve as a useful selection tool. Similarly,incorporation of NT-proBNP in the Health ABCHF risk score for screening selection appears to bean attractive target for future investigations.

What are the practical implications of our find-ings? First, there are currently no prevention strat-egies for individuals at increased HF risk, barringtreatment of asymptomatic LV dysfunction or in-dividual risk factors. Thus, it is difficult to translateimprovement in HF risk classification into action-able items or cost-effectiveness gains at this point.This gap highlights the need for prospective data on

HF prevention strategies. Second, we demonstrated

bmaucoatwbrlcag

cinocwapb

EC

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

139

that echocardiographic screening with a “subclinicaldisease” approach alone would fail to detect mostpatients at risk for HF, even for a 5-year horizon. Astrategy providing HF risk assessment and riskfactor modification based on projected HF risk, andpotentially novel preventive interventions, wouldlikely be more effective from a population perspec-tive. Echocardiography and natriuretic peptide test-ing might be reserved for selected individuals. Sucha strategy, however, would need prospective valida-tion. Third, extrapolating from the coronary heartdisease prevention paradigm, it may be reasonableat this point to follow-up more closely individuals at�10% 5-year HF risk, with the aim to intensify riskfactor control, monitor for symptoms, and potentiallyscreen with echocardiography for LV dysfunction basedon the rationale described in the previous text.Study limitations. First, definition of incident HF wasased on admissions and this may have underesti-ated HF incidence. However, because of central

djudication, underestimation is most likely distrib-ted evenly across risk groups and thus our reclassifi-ation estimates should be relatively unbiased. On thether hand, we cannot exclude the possibility ofscertainment bias in determining the relative impor-ance of the echocardiographic measures. Participantsith possible clinical HF were perhaps more likely toe admitted if a clinically performed echocardiogramevealed LV systolic dysfunction (as opposed to iso-ated left atrial dilatation or increased LV mass). Thisould partly explain why LVEF was strongly associ-ted with HF risk compared with other echocardio-

stroke, congestive heart failure, and Heart J 2008;29:74

ardiography was assessed by the E/A ratio and,ndirectly, by the left atrial volume. These measures doot reflect contemporary assessment or classificationf diastolic dysfunction. Third, our study populationomprised of white and black participants, and blacksere underrepresented. Thus our findings may not be

pplicable to other races. Finally, we studied partici-ants age 65 years or older; thus our findings may note applicable to younger populations.

C O N C L U S I O N S

Echocardiography and NT-proBNP significantlyreclassify 5-year HF risk in older adults when addedto a clinical model, mainly among intermediate-riskindividuals. A strategy incorporating echocardiog-raphy or NT-proBNP for these individuals may beworth investigating prospectively. Echocardiogra-phy at this point might be reserved for those at�10% projected 5-year HF risk because of theprevalence of potentially treatable underlying ab-normalities in this group. However, a strategy basedsolely on screening for asymptomatic abnormalitiesis unlikely to have dramatic impact on HF preven-tion in older adults. An alternative strategy, basedon risk stratification and tiered preventive interven-tions, might be more effective and warrants furtherinvestigation.

Reprint requests and correspondence: Dr. Javed Butler,mory Clinical Cardiovascular Research Institute, 1462lifton Road NE, Suite 504, Atlanta, Georgia 30322.

raphic measures. Second, diastolic function by echo- E-mail: javed.butler@emory.edu.

R E F E R E N C E S

1. Hunt S. ACC/AHA 2005 guidelineupdate for the diagnosis and manage-ment of chronic heart failure in theadult: a report of the American Collegeof Cardiology/American Heart Associ-ation Task Force on Practice Guidelines(Writing Committee to Update the2001 Guidelines for the Evaluation andManagement of Heart Failure). J AmColl Cardiol 2005;46:e1–82.

2. Vasan RS, Larson MG, Benjamin EJ,Evans JC, Levy D. Left ventriculardilatation and the risk of congestiveheart failure in people without myo-cardial infarction. N Engl J Med1997;336:1350–5.

3. Gardin JM, McClelland R, KitzmanD, et al. M-mode echocardiographicpredictors of six- to seven-year inci-dence of coronary heart disease,

mortality in an elderly cohort (theCardiovascular Health Study). Am JCardiol 2001;87:1051–7.

4. Aurigemma G, Gottdiener J, Shem-anski L, Gardin J, Kitzman D. Pre-dictive value of systolic and diastolicfunction for incident congestive heartfailure in the elderly: the Cardiovascu-lar Health Study. J Am Coll Cardiol2001;37:1042–8.

5. Gottdiener JS, Kitzman DW, Au-rigemma GP, Arnold AM, ManolioTA. Left atrial volume, geometry, andfunction in systolic and diastolic heartfailure of persons � or �65 years of age(the Cardiovascular Health Study).Am J Cardiol 2006;97:83–9.

6. de Simone G, Gottdiener J, ChinaliM, Maurer M. Left ventricular masspredicts heart failure not related toprevious myocardial infarction: theCardiovascular Health Study. Eur

1–7.

7. Bluemke DA, Kronmal RA, Lima JA, etal. The relationship of left ventricular massand geometry to incident cardiovascularevents: the MESA (Multi-Ethnic Studyof Atherosclerosis) study. J Am Coll Car-diol 2008;52:2148–55.

8. Wang TJ, Larson MG, Levy D, et al.Plasma natriuretic peptide levels and therisk of cardiovascular events and death.N Engl J Med 2004;350:655–63.

9. deFilippi C, Christenson R, GottdienerJ, Kop W, Seliger S. Dynamic cardio-vascular risk assessment in elderly peo-ple. The role of repeated N-terminalpro-B-type natriuretic peptide testing.J Am Coll Cardiol 2010;55:441–50.

10. McKie P, Cataliotti A, Lahr B, et al. Theprognostic value of N-terminal pro-B-type natriuretic peptide for death and car-diovascular events in healthy normal andstage A/B heart failure subjects. J Am Coll

Cardiol 2010;55:2140–7.

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

hr

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 2 , 2 0 1 2

F E B R U A R Y 2 0 1 2 : 1 3 1 – 4 0

Kalogeropoulos et al.

Echocardiography and NT-proBNP for HF Risk

140

11. Emberson J, Ng L, Armitage J,Bowman L, Parish S, Collins R.N-terminal Pro-B-type natriureticpeptide, vascular disease risk, andcholesterol reduction among 20,536patients in the MRC/BHF HeartProtection Study. J Am Coll Cardiol2007;49:311–9.

2. Wang T, Levy D, Benjamin E, Va-san R. The epidemiology of “asymp-tomatic” left ventricular systolic dys-function: implications for screening.Ann Intern Med 2003;138:907–16.

3. Redfield M, Rodeheffer R, Jacobsen S,Mahoney D, Bailey K, Burnett JJ.Plasma brain natriuretic peptide to de-tect preclinical ventricular systolic ordiastolic dysfunction: a community-based study. Circulation 2004;109:3176–81.

4. Bibbins-Domingo K, Ansari M,Schiller NB, Massie B, Whooley MA.Is B-type natriuretic peptide a usefulscreening test for systolic or diastolicdysfunction in patients with coronarydisease? Data from the Heart and SoulStudy. Am J Med 2004;116:509–16.

5. de Lemos J, McGuire D, Khera A, etal. Screening the population for leftventricular hypertrophy and left ven-tricular systolic dysfunction using na-triuretic peptides: results from theDallas Heart Study. Am Heart J 2009;157:74653.e2.

6. Atherton J. Screening for left ventric-ular systolic dysfunction: is imaging asolution? J Am Coll Cardiol Img2010;3:421–8.

7. Butler J, Kalogeropoulos A, Georgio-poulou V, et al. Incident heart failureprediction in the elderly: the HealthABC Heart Failure Score. Circ HeartFail 2008;1:125–33.

8. Kalogeropoulos A, Psaty BM, VasanRS, et al., Cardiovascular HealthStudy. Validation of the Health ABCheart failure model for incident heartfailure risk prediction: the Cardiovas-cular Health Study. Circ Heart Fail2010;3:495–502.

9. Fried LP, Borhani NO, Enright P, etal. The Cardiovascular Health Study:design and rationale. Ann Epidemiol1991;1:263–76.

0. Psaty BM, Kuller LH, Bild D, et al.Methods of assessing prevalent car-diovascular disease in the Cardiovas-cular Health Study. Ann Epidemiol1995;5:270–7.

1. Blackburn H. Classification of theelectrocardiogram for population

studies: Minnesota Code. J Electro-cardiol 1969;2:305–10.

2. Gardin JM, Wong ND, Bommer W,et al. Echocardiographic design of amulticenter investigation of free-livingelderly subjects: the CardiovascularHealth Study. J Am Soc Echocardiogr1992;5:63–72.

3. Gardin JM, Siscovick D, Anton-Culver H, et al. Sex, age, and diseaseaffect echocardiographic left ventricu-lar mass and systolic function in thefree-living elderly. The CardiovascularHealth Study. Circulation 1995;91:1739–48.

4. Lang RM, Bierig M, Devereux RB, etal. Recommendations for chamberquantification: a report from theAmerican Society of Echocardiogra-phy’s Guidelines and Standards Com-mittee and the Chamber Quantifica-tion Writing Group. J Am SocEchocardiogr 2005;18:1440–63.

5. Devereux RB, Alonso DR, Lutas EM,et al. Echocardiographic assessment ofleft ventricular hypertrophy: compari-son to necropsy findings. Am J Car-diol 1986;57:450–8.

6. Ives DG, Fitzpatrick AL, Bild DE, etal. Surveillance and ascertainment ofcardiovascular events. The Cardiovas-cular Health Study. Ann Epidemiol1995;5:278–85.

7. Schellenbaum GD, Heckbert SR,Smith NL, et al. Congestive heartfailure incidence and prognosis: caseidentification using central adjudica-tion versus hospital discharge diagno-ses. Ann Epidemiol 2006;16:115–22.

8. Volinsky CT, Raftery AE. Bayesianinformation criterion for censored sur-vival models. Biometrics 2000;56:256–62.

9. Pencina MJ, D’Agostino RB Sr.,D’Agostino RB Jr., Vasan RS. Evalu-ating the added predictive ability of anew marker: from area under theROC curve to reclassification and be-yond. Stat Med 2008;27:157–72, dis-cussion 207–12.

0. Pencina MJ, D’Agostino RB, VasanRS. Statistical methods for assessmentof added usefulness of new biomark-ers. Clin Chem Lab Med 2010;48:1703–11.

1. Chambless L, Cummiskey C, Cui G.Several methods to assess improve-ment in risk prediction models: exten-

sion to survival analysis. Stat Med2011;30:22–38.

32. Pencina MJ, D’Agostino RB Sr.,Steyerberg EW. Extensions of netreclassification improvement calcula-tions to measure usefulness of newbiomarkers. Stat Med 2011;30:11–21.

33. Takemoto Y, Barnes M, Seward J, etal. Usefulness of left atrial volume inpredicting first congestive heart failurein patients � or � 65 years of agewith well-preserved left ventricularsystolic function. Am J Cardiol 2005;96:832–6.

34. From AM, Scott CG, Chen HH. Thedevelopment of heart failure in pa-tients with diabetes mellitus and pre-clinical diastolic dysfunction apopulation-based study. J Am CollCardiol 2010;55:300–5.

35. Nielsen O, McDonagh T, Robb S,Dargie H. Retrospective analysis ofthe cost-effectiveness of using plasmabrain natriuretic peptide in screeningfor left ventricular systolic dysfunctionin the general population. J Am CollCardiol 2003;41:113–20.

36. Redfield M, Jacobsen S, Burnett JJ,Mahoney D, Bailey K, Rodeheffer R.Burden of systolic and diastolic ven-tricular dysfunction in the community:appreciating the scope of the heartfailure epidemic. JAMA 2003;289:194–202.

37. Heidenreich P, Gubens M, FonarowG, Konstam M, Stevenson L, Shek-elle P. Cost-effectiveness of screeningwith B-type natriuretic peptide toidentify patients with reduced leftventricular ejection fraction. J AmColl Cardiol 2004;43:1019–26.

38. Smith J, Newton-Cheh C, AlmgrenP, et al. Assessment of conventionalcardiovascular risk factors and multi-ple biomarkers for the prediction ofincident heart failure and atrial fibril-lation. J Am Coll Cardiol 2010;56:1712–9.

39. Velagaleti R, Gona P, Larson M, et al.Multimarker approach for the predic-tion of heart failure incidence in thecommunity. Circulation 2010;122:1700–6.

Key Words: epidemiology yeart failure y risk prediction yisk score y risk stratification.

‹ A P P E N D I X

For a supplemental data section and tables,please see the online version of this article.