J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
57
SEE PAGE 70
AB BR E V I A T I O N S
AND ACRONYM S
BNP = B-type natriuretic
peptide
CI = confidence interval
eGFR = estimated glomerular
filtration rate
HF = heart failure
HR = hazard ratio
NT-proBNP = N-terminal pro–
B-type natriuretic peptide
PENK = proenkephalin A
ROC = receiver-operating
characteristic
SBP = systolic blood pressure
WRF = worsening renal
ion
I n recent years, many advances have been madein the understanding of pathophysiology andthe management of chronic heart failure (HF).
However, the understanding and treatment of acuteHF has remained incomplete and broadly unchangedduring this period. Accordingly, prognosis remainspoor, with 1-year mortality rate exceeding 25% (1).Neurohormonal activation and worsening renal func-tion (WRF) play important roles in the pathogenesisof fluid redistribution, leading to acute decompensa-tion (2). Use of biomarkers might help characterizedifferent phenotypes in acute HF associated withdifferent outcomes that may prompt specific andexpedited therapies. Although activation of the natri-uretic peptide system is recognized, its value in pre-dicting death at first presentation with acute HF issuboptimal (3), and better tools are needed.
The endogenous opioids (enkephalins, endorphins,dynorphins), extensively studied in nociception andanesthesia, also have roles in cardiovascular regula-tion (4). Proenkephalin A (PENK) is widely expressed,and cardiac cells secrete enkephalins, which havelocal effects on opioid receptors. Cardiodepressivethrough a negative inotropic effect and lower bloodpressure and heart rate (5), opioid receptors, espe-cially the d receptor that binds enkephalins, are widelydistributed, with highest densities in the kidney (6).
The possible relationship between endogenousopioid systems and prognosis was suggested byprevious studies. Data from ADHERE (Acute Decom-pensated Heart Failure National Registry) demon-strated that opiate administration in acute HF hasbeen associated with poor outcomes (7). Fontana et al.(8) reported elevated met-enkephalin levels in severeacute HF compared with less severe acute HF.
In several acute disease conditions, elevatedplasma levels of a PENK fragment (amino acids 119through 159) have been associated with renaldysfunction and poor outcomes. For example, wepreviously demonstrated PENK to be an independentpredictor of major adverse cardiac events, includingdeath, reinfarction, and rehospitalization for HF inpatients presenting with acute myocardial infarction(9). This also has been shown more recently for stableambulatory patients with HF (10). PENK predicts
the European Union, the Cardiovascular Research Foundation Basel, the Univ
received speaker or consulting honoraria from Abbott, Alere, AstraZene
Ingelheim, BRAHMS, Cardiorentis, Eli Lilly, Novartis, Roche, Sanofi, Sieme
grants from Novartis and Servier; and has received speaker or consulting hon
that they have no relationships relevant to the contents of this paper to dis
Manuscript received September 7, 2016; revised manuscript received Septem
acute kidney injury after cardiac surgicalprocedures (11) and in patients with sepsis(12), and it has been linked to death and ma-jor adverse cerebrocardiovascular events inacute stroke (13).
In the present study, we investigated therelationship of the enkephalin system withWRF and worsening prognosis in acute HF.Renal impairment profoundly influencesprognosis in HF (14), and development ofacute kidney injury is common in acute HF,the so-called cardiorenal syndrome type 1(15). We therefore examined the utility ofPENK in assessing WRF in acute HF. Previousstudies were hindered by the instability of
met-enkephalin. we used a more recently developedassay (penKid assay, Sphingotec GmbH, Hennigsdorf,Germany) for PENK (16), with epitopes on the pro-enkephalin molecule that are stable in whole bloodfor at least 48 h, thus enabling a study of this systemin acute HF. The utility of PENK for prediction ofshort-term and long-term outcomes and inpatientmortality was examined in combination with variousclinical risk scores developed for inpatient mortality,namely ADHERE (17), GWTG-HF (Get With theGuidelines Heart Failure) (18), and OPTIMIZE-HF(Organized Program to Initiate Lifesaving Treatmentin Hospitalized Patients With Heart Failure) (19).
METHODS
Three cohorts of unselected patients with acuteHF who presented with acute dyspnea to the emer-gency department of the participating universityhospitals in 3 countries (United Kingdom, France, andSwitzerland) were recruited. Acute HF was defined,according to the guidelines of the European Society ofCardiology (20), as progressive worsening or new-onset of shortness of breath, along with clinicalsigns of pulmonary or peripheral edema and elevatedjugular venous pressure requiring intensification ofdiuretic and/or vasodilator therapy. Inclusion wasindependent of renal function, although patients with
Values are mean � SD, n (%), or median (IQR) unless otherwise indicated. *Quoted for the analysis of variance/Kruskal-Wallis or chi-square tests for continuous or categorical variables,respectively.
ACE ¼ angiotensin-converting enzyme; ARB ¼ angiotensin receptor blocker; BNP ¼ B-type natriuretic peptide; BP ¼ blood pressure; eGFR ¼ estimated glomerular filtration rate;IQR ¼ interquartile range; NS ¼ not significant; NT-proBNP [ N-terminal pro–B-type natriuretic peptide; PENK ¼ amino acids 119 to 159 of proenkephalin A.
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
58
terminal renal failure who were receiving establishedrenal replacement therapy were excluded. Thesestudies complied with the declaration of Helsinki,and ethics approval was granted from the respectiveresearch ethics committees. All patients providedwritten informed consent.
Following signed informed consent, venousblood was withdrawn from recumbent patients andcollected in pre-chilled tubes containing ethyl-enediaminetetraacetic acid as an anticoagulant. Theintervals for obtaining this admission sample wereup to 4 h (Paris), 1 h (Basel), and 12 h (Leicester).Plasma was stored at �80�C until analysis in asingle batch.
IMAGING AND ASSAYS. Transthoracic echocardiog-raphy was performed using standard techniques, andleft ventricular ejection fraction was calculated usingthe biplane method of discs formula.
The assay for stable PENK (molecular weight4,586 Da) was previously described (16), and it hassince been modified (9). In brief, 2 mouse monoclonalanti-PENK antibodies were developed by immuniza-tion with PENK peptide. Standards or samples (50 mlplasma) were immobilized by the capture antibody(2 mg coated on polystyrene tubes). The detectorantibody was labeled with methylacridinium ester,and bound chemiluminescence was measured. Thenormal range was mean � SEM of 46.6 � 14.1 pmol/l,with a median of 45 pmol/l (range 9 to 518 pmol/l)(12).
We used an immunoassay to measure troponin I,which has a coefficient of variation of 10% at 0.03 mg/lwith a 99th percentile of 0.04 mg/l. Plasma high-sensitivity troponin T was measured in patientsenrolled in Basel. The 99th percentile upper referencelimit was 0.014 mg/l. All samples were analyzed in acentral laboratory in a blinded manner.
TABLE 2 Patient Characteristics by PENK Quartiles on Admission
Values are mean � SD, n (%), or median (interquartile range) unless otherwise indicated. *p values are quoted for the analysis of variance/Kruskal Wallis or chi-square tests for continuous or categoricalvariables, respectively.
Abbreviations as in Table 1.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
59
In patients from Leicester, United Kingdom,plasma N-terminal pro–B-type natriuretic peptide(NT-proBNP) was quantified using a sandwichimmunoassay as described previously (21). In Paris,plasma brain natriuretic peptide (BNP) was measuredusing Abbott kits (Abbott Diagnostics, Rungis Cedex,France). The Elecsys NT-proBNP assay (Roche Di-agnostics GmbH, Mannheim, Germany) was used inBasel.OUTCOMES. The primary endpoint was 1-year all-cause mortality. Secondary endpoints includedwithin-hospital all-cause mortality in the entirecohort and all-cause mortality or HF rehospitalizationwithin 1 year, by using follow-up data for theLeicester and Basel cohorts only. HF rehospitalization
was defined as a hospital readmission for which HFwas the primary cause, requiring treatment with di-uretics, intravenous inotropes, or nitrates. Endpointswere obtained from hospital records and electronicdatabases. Patients surviving to discharge were fol-lowed up for at least 1 year after the initial hospital-ization. When patients had multiple events, the timeto first event was counted as the censored outcome.
Another endpoint was WRF, defined as a rise inplasma creatinine of $26.5 mmol/l or 50% higher thanthe admission value (whichever was smaller), within5 days of presentation (14,22). We did not use thedefinition with urine volumes because administrationof diuretics could cause large variations in thisparameter. Degree of WRF was analyzed as the
TABLE 3 Independent Predictors of PENK Levels
Variable F Statistic p Value*
eGFR 296.0 <0.0005
Plasma urea 166.0 <0.0005
Natriuretic peptide levels 74.0 <0.0005
Age 55.0 <0.0005
Female 33.0 <0.0005
Past history of renal failure 29.0 <0.0005
Systolic BP 8.2 0.0042
Heart rate 6.1 0.0133
Past history of diabetes 2.1 NS
Past history of heart failure 1.8 NS
Past history of hypertension 0.8 NS
Past history of IHD 0.1 NS
*p < 0.0005; adjusted R2 0.606.
IHD ¼ ischemic heart disease; other abbreviations as in Tables 1 and 2.
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
60
absolute increase in plasma creatinine from theadmission value and also as an ordinal scale definedas follows: 0 (rise in plasma creatinine <26.5 mmol/lor <1.5-fold); 1 (rise in plasma creatinine of$26.5 mmol/l or 1.5- to 2-fold); 2 (>2.0- to 3.0-fold);and 3 (>3.0-fold) (23). The modified diet in renaldisease formula was used to obtain the estimatedglomerular filtration rate (eGFR).
STATISTICAL ANALYSIS. Statistical analyses wereperformed using SPSS version 22 (IBM Corp., Armonk,New York) and Stata 13 (Statacorp LP, College Station,Texas) software. Assuming an event rate of 25% andthat the covariates predict up to 30% of thebiomarker variance, a sample size of 1,000 patientswould be powered (99% at p < 0.01) to detect ahazard ratio (HR) of the biomarker of 1.5, using thecommand stpower cox in Stata 13. All biomarkerlevels were log10 transformed and normalized to 1 SDincrement.
Gaussian data were analyzed using analysis ofvariance and general linear models, and nonpara-metric tests (Mann-Whitney U test, Kruskal-Wallistest, and Spearman [rs] correlations) were usedagainst non-Gaussian data. Chi-square tests wereused for categorical variables. Independent pre-dictors of PENK levels were assessed using generallinear models with coefficients and p values reportedfor 2,000 bootstrap samples.
Because all 3 cohorts used different natriureticpeptides assays, we combined the data in this com-posite study by normalizing BNP and NT-proBNP foreach center, by using log transformation and calcu-lating the z-transform (thus expressing natriureticpeptides normalized to 1 SD of the log-transformedbiomarker for each center) before pooling thez-transformed natriuretic peptides for all centers.
To assess the prognostic value of biomarkers, a“base” model was generated using Cox survival anal-ysis, which included variables that were associatedwith study outcomes on univariable analysis atp < 0.10 or had been associated with poor outcomein other studies (age, sex, previous history of HF,ischemic heart disease, hypertension or diabetes,plasma urea, sodium, eGFR, hemoglobin, and bio-markers [log NT-proBNP or BNP]). Because differentnatriuretic peptide assays were used, these were lognormalized and then z-transformed (divided by 1 SDincrement) before analysis. PENK was added to thisbase model to evaluate its relative prognosticvalue with all variables entered simultaneously,with added value assessed using the increment in log-likelihood chi-square test for nested regressionmodels.
To predictWRF, we used logistic, linear, and ordinalregression models containing clinical variables,plasma biomarkers, and use of therapies associatedwith WRF such as diuretics and angiotensin-converting enzyme inhibitors or angiotensin receptorblockers.
A logistic regression model was used to assessthe relative prognostic power of these biomarkersand clinical risk scores (ADHERE, GWTG-HF, andOPTIMIZE-HF) to predict in-hospital all-cause mor-tality. The probability of an outcome calculated fromthe logistic base model and the added value ofnatriuretic peptides and/or PENK was evaluated byreclassification analysis with calculation of category-free net reclassification improvement as describedby Pencina et al. (24). A similar approach to assess netreclassification improvement was adopted for pre-diction of 1-year mortality or the composite of 1-yeardeath and/or HF.
We also constructed classification trees usingchi-square automatic interaction detection, whichchooses the biomarker at each step that has thestrongest interaction with the dependent variable.
RESULTS
PATIENT CHARACTERISTICS. According to patientcharacteristics seen in the cohorts (Leicester, Paris,and Basel) (Table 1), patients at the 3 sites weresimilar in sex, but varied in age, renal function, bodymass index, and comorbidities. The outcomes (deathand death or HF at 1 year) were similar among sites.The PENK levels were similar among sites, with nodifferences in the normalized natriuretic peptidedistributions.
The characteristics of the combined Leicester-Paris-Basel cohort are shown in Table 2, according to PENK
FIGURE 1 Predictors of WRF
0.25 0.5Odds Ratio
PENKNatriuretic Peptide
DiureticACE/ARB
Plasma Sodium
Plasma UreaHeart RateSystolic BP
PH DiabetesPH Renal Failure
PH HypertensionPH IHD
PH Heart FailureMale
Agep Value
NSNSNSNSNS
NS
NSNSNS
NSNSNS
.004
.009
<0.0005
.041Plasma Creatinine
1 2
Forest plots of a multivariable analysis shows odds ratio for clinical variables, natriuretic peptides, and amino acids 119 to 159 of proenkephalin
A for prediction of worsening renal function (WRF) during initial hospitalization. ACE ¼ angiotensin-converting enzyme; ARB ¼ angiotensin
receptor blocker; BP ¼ blood pressure; IHD ¼ ischemic heart disease; PENK ¼ proenkephalin A assay; PH ¼ past history; NS ¼ not significant.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
61
quartiles. Patients with higher PENK levels wereolder, had a lower body mass index, were more oftenfemale, and had comorbidities such as histories ofhypertension, ischemic heart disease, HF, and renalimpairment; their initial systolic blood pressures(SBP) and heart rates were also lower. With increasingPENK quartiles, renal function deteriorated, andnatriuretic peptide levels increased. Higher PENK also
TABLE 4 Significant Predictors
Predictor p Value
Male 0.026
Past history of renal failure <0.0005
Systolic BP 0.007
Plasma urea 0.014
Creatinine <0.0005
PENK 0.001
Ordinal regression for WRF stages
Systolic BP 0.016
Plasma urea 0.028
Creatinine 0.03
Sodium 0.037
Past history of renal failure 0.012
PENK <0.0005
WRF ¼ worsening renal function; other abbreviations as in Tables 1 and 2.
was associated with more frequent prescription ofloop diuretics and aldosterone antagonists.
CORRELATION ANALYSIS AND EFFECTS OF
CHANGES IN PROENKEPHALIN A. Spearman ana-lysis (rs, p value) showed that PENK was correlatedwith age (0.366; p < 0.0005), eGFR (�0.752;p < 0.0005), plasma creatinine (0.668; p < 0.0005),plasma urea (0.641; p < 0.0005), heart rate (�0.165;p < 0.0005), SBP (�0.100; p < 0.0005), troponin T(0.373; p < 0.0005), and z-score of log natriureticpeptide (0.419; p< 0.0005). There were nonsignificantcorrelations with plasma sodium. A univariate generallinear model indicated the following independentpredictors of PENK level, in descending order ac-cording to variance accounted for in the model(Table 3): eGFR, plasma urea, natriuretic peptidelevels, age, sex, past history of renal impairment, SBP,and heart rate. These variables accounted for 60.6%of the variance of PENK levels, and of these, 2 mea-sures of renal function (eGFR and plasma urea)accounted for 46.8% of the model.
Of the 1,714 patients with data on plasma creati-nine within 5 days of hospitalization, 264 had devel-oped a rise in plasma creatinine of $26.5 mmol/l or50% higher than the admission value. Using clinicalvariables, use of nephrotoxic drugs on admission
FIGURE 2 PENK and PENK-to-Creatinine Relationship With WRF
SampleAdmission Post-treatment
No WRF WRF
0
100
200
300
400A B
Plas
ma
PENK
(pm
ol/L
)
No WRFPE
NK/C
reat
inin
e Ra
tio (p
mol
/um
ol)
.0
.5
1.0
1.5
2.0
WRF
(A) Patients with worsening renal function (WRF) had higher proenkephalin A (PENK) levels on admission that increased post-treatment. (B) In patients
without WRF, PENK-to-creatinine ratios did not change between admission and post-treatment samples, whereas in patients with WRF, these ratios were
initially elevated and fell following treatment (p < 0.0005).
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
62
(angiotensin-converting enzyme inhibitors or angio-tensin receptor blockers, diuretics), natriuretic pep-tides, and PENK, the independent predictors of WRFwere past history of renal disease, plasma sodium,SBP, and PENK (Figure 1). The receiver-operatingcharacteristic (ROC) area for the full model was 0.69(95% confidence interval [CI]: 0.65 to 0.73) comparedwith 0.67 (95% CI: 0.63 to 0.71) for a model withoutPENK (p value for difference in ROC areas ¼ 0.054).
We also evaluated the relationship of PENK withthe severity of development of renal impairment byperforming linear regression of these variables andthe absolute change in plasma creatinine from theadmission level. The significant predictors are listedin Table 4.
Plasma samples taken before and after treatment ofthe acute episode of HF were available in 1,012 pa-tients of the Leicester and Basel cohorts, with post-treatment levels obtained approximately 5 daysfollowing admission. Overall, median levels (inter-quartile range) before and after treatment were 97.2pmol/l (66.9 to 146.8 pmol/l) and 98.7 pmol/l (66.6 to141.8 pmol/l), respectively, (p ¼ NS; Wilcoxon signedrank test). However, in comparing plasma PENK be-tween admission and after therapy for those patients
who did or did not develop WRF by using a repeatedmeasures design, there was a significant interactionbetween the pre-therapy and post-therapy PENKlevels and WRF development (p < 0.0005). Patientswith WRF showed a higher level of PENK on admis-sion, which increased further after therapy (Figure 2).
To explore whether PENK levels changed beforecreatinine levels, we calculated PENK-to-creatinineratios in patients according to WRF status. In pa-tients without WRF, PENK-to-creatinine ratios didnot change between admission and post-therapysamples, a finding suggesting that both analyteschanged in tandem, whereas in patients with WRF,ratios were initially elevated and fell significantlywith time (p < 0.0005). There was a significantinteraction between pre-therapy and post-therapyPENK-to-creatinine ratios and WRF development(p < 0.0005) (Figure 2).SURVIVAL ANALYSIS. During a minimum follow-upof 1 year, there were 518 deaths in the whole cohort(N ¼ 1,908) and 699 death or HF endpoints in theLeicester and Basel cohorts (N ¼ 1,694). Patients withelevated PENK levels had more deaths during follow-up (Table 2). Figure 3A illustrates a graded increase inthe cumulative incidence of all-cause mortality with
FIGURE 3 Outcomes According to PENK Levels
PENK QuartileTime (Days)
Even
t Rat
e %
A B
0
0
10
20
30
40
50
60
100 200 300 400
1 2 3 4
60
50
40
0 100 200Time (Days)
Even
t Rat
e %
300 400
30
20
10
0
Cumulative incidence of all-cause mortality (A) and the composite endpoint of death and/or heart failure hospitalization (B) rose with higher proenkephalin
A (PENK) quartiles.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
63
increasing PENK quartiles (p < 0.0005). Comparisonof PENK quartiles revealed significant differencesamong all of them (p < 0.001), except comparingquartile 1 versus quartile 2 (p ¼ 0.009). Figure 3Bshows a graded increase in event rates for deathand/or HF hospitalization with increasing PENKquartiles (p < 0.0005). Apart from quartile 1 versus 2(p ¼ 0.015), all other quartile comparisons were sta-tistically different (p < 0.0005).
Figure 4A illustrates the univariable hazard ratiosfor factors affecting the outcome of all-cause mor-tality at 1 year, by using Cox proportional hazardsurvival analysis. Model 1 (Figure 4B) included rele-vant clinical variables and z-transformed natriureticpeptide levels, with independent predictors beingage, past history of hypertension, SBP, plasma urea,sodium, eGFR, and natriuretic peptide levels. Addi-tion of PENK to this base model (Figure 4C) showedthat it had independent predictive value for death, itsadded value being statistically significant (p <
0.0005) using the increment in log likelihood ratiochi-square for nested regression models. For theendpoint of death at 3 and 6 months, the multivari-able adjusted HR for PENK remained significant forboth time points (3 months: HR: 1.49; 95% CI: 1.20 to1.85; p < 0.0005; 6 months: HR: 1.40; 95% CI: 1.17 to1.68; p < 0.0005).
The C statistic for 1-year mortality was 0.741 (in thebase model using the foregoing demographic andclinical chemistry variables), and it rose to 0.754(p ¼ 0.021) and 0.751 (p ¼ 0.051) with addition ofnatriuretic peptide and PENK, respectively, and to0.759 (p¼ 0.007) with the addition of both biomarkers.
Figure 5A reports the HRs for the outcome of deathor HF at 1 year in the Leicester and Basel cohorts.Model 1 (Figure 5B) is a multivariable model thatincluded the independent predictors: age, past his-tory of HF, hypertension, ischemic heart disease,diabetes, SBP, plasma urea, and natriuretic peptidelevels. Addition of PENK (model 2) showed that ithad independent predictive value for death or HF(p ¼ 0.003) (Figure 5C), and the increment in loglikelihood ratio chi-square was statistically significant(p ¼ 0.001). For the endpoint of death or HF at 3 and6 months, the multivariable adjusted HR for PENKremained significant for both time points (3 monthsHR: 1.27; 95% CI: 1.06 to 1.53; p ¼ 0.011; 6 months HR:1.32; 95% CI: 1.13 to 1.54; p < 0.0005).
Using the base model, the C statistic for 1-yeardeath or HF was 0.692, and it rose to 0.702(p ¼ 0.079) and 0.700 (p ¼ 0.09) with addition ofnatriuretic peptide and PENK, respectively, and to0.706 (p ¼ 0.039) with the addition of bothbiomarkers.
FIGURE 4 1-Year Mortality
0.25 0.5Hazard Ratio
AgeA
C
p Value<0.0005
<0.0005
<0.0005<0.0005<0.0005<0.0005<0.0005<0.0005
NS.001.057NSNS
NS
Male PH Heart Failure
PH IHDPH Hypertension
PH DiabetesSystolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic PeptidePENK
1 2
0.25 0.5 1 2Hazard Ratio
BAge
Male PH Heart Failure
PH IHDPH Hypertension
PH DiabetesSystolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic Peptide
p Value<0.0005
<0.0005
<0.0005<0.0005
<0.0005
NSNSNS
NS
0.25 0.5 1 2Hazard Ratio
NS
NS
.001
.015
AgeMale
PH Heart FailurePH IHD
PH HypertensionPH Diabetes
Systolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic PeptidePENK
p Value<0.0005
<0.0005
.011
<0.0005<0.0005
.054NSNS
.066
NSNS
.002
NS
<0.0005
Clinical variables, natriuretic peptides, and proenkephalin A (PENK) for prediction of 1-year all-cause mortality was studied in (A) univariable
analysis; (B) model 1, which included natriuretic peptides but excluded PENK; and (C) model 2, which included all variables and biomarkers.
eGFR ¼ estimated glomerular filtration rate; other abbreviations as in Figure 1.
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
64
FIGURE 5 1–Year Death and/or HF Hospitalization
Agep Value
<0.0005
<0.0005<0.0005
<0.0005
<0.0005
<0.0005<0.0005<0.0005<0.0005
NS
NSNS
NS
NS
p Value.001
.009
.005
.001
.000
NSNS
<0.0005
NS
.001.008
NS
NS
p Value.005
.010
.002
<0.0005
.001
NSNS
<0.0005.001
NS
.000
.004
NS
NS
Male PH Heart Failure
PH IHDPH Hypertension
PH DiabetesSystolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic PeptidePENK
AgeMale
PH Heart FailurePH IHD
PH HypertensionPH Diabetes
Systolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic Peptide
AgeMale
PH Heart FailurePH IHD
PH HypertensionPH Diabetes
Systolic BPHeart Rate
Plasma UreaPlasma Sodium
eGFRHemoglobin
Natriuretic PeptidePENK
0.25 0.5 1Hazard Ratio
A
B
C
2
0.25 0.5 1Hazard Ratio
2
0.25 0.5 1Hazard Ratio
2
Clinical variables, natriuretic peptides, and proenkephalin A (PENK) for prediction of 1-year all-cause mortality and/or heart failure (HF)
hospitalization was evaluated in (A) univariable analysis; (B) model 1, which excluded PENK; and (C) model 2, which included all variables.
Abbreviations as in Figures 1 and 4.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
*Model 1: Base model of clinical risk prediction score plus z-transformed log natriuretic peptide. †Model 2: Base model of clinical risk prediction score plus z-transformed log PENK. ‡Model 3: Base model ofclinical risk prediction score plus z-transformed log natriuretic peptide and PENK.
ADHERE ¼ Acute Decompensated Heart Failure National Registry; CI ¼ confidence interval; GWTG-HF ¼ Get With the Guidelines Heart Failure; OPTIMIZE-HF ¼ Organized Program to Initiate LifesavingTreatment in Hospitalized Patients With Heart Failure; OR ¼ odds ratio; ROC ¼ receiver-operating characteristic; other abbreviations as in Tables 1 and 2.
TABLE 6 In-Hospita
Endpoint NRI
Adding Natriure
No �12.4 (�Yes 39.5 (1
Total 27.0 (4
Adding P
No 25.9 (21
Yes 39.5 (17
Total 65.4 (4
Adding PENK to ADH
No 25.3 (20
Yes 26.3 (3.
Total 51.6 (28
NRI ¼ net reclassification i
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
66
PREDICTING INPATIENT MORTALITY. The use ofPENK and/or natriuretic peptide together with clin-ical risk scores for determining inpatient mortality(n ¼ 82) was examined using logistic regressionanalysis. Table 5 reports the odds ratios for z-trans-formed clinical scores individually and with theaddition of natriuretic peptide and/or PENK. Usingthe ADHERE score, PENK was an independent pre-dictor for inpatient mortality (p < 0.0005), with a Cstatistic of 0.709. PENK remained an independentpredictor when used together with either the GWTG-HF score (C statistic 0.718) or the OPTIMIZE score(C statistic 0.729). Natriuretic peptide levels were notindependent predictors in all 3 clinical scores, when
lity: NRI
(95% CI) p Value Endpoint NRI (95% CI)
tic Peptide to ADHERE Adding Natriuretic Peptide to OPTIM
17.4 to �7.5) <0.0005 No �31.9 (�36.8 to �26.9)
7.0–62.0) 0.001 Yes 51.4 (28.0–74.9)
.0–50.1) 0.021 Total 19.5 (�4.4 to 43.5)
ENK to ADHERE Adding PENK to OPTIMIZE
.0–30.9) <0.0005 No 8.8 (3.9–13.8)
.0–62.0) 0.001 Yes 20.0 (�3.4 to 43.4)
2.4–88.4) <0.0005 Total 28.8 (4.9–52.8)
ERE and Natriuretic Peptide Adding PENK to OPTIMIZE-HF and Nat
.4–30.3) <0.0005 No 15.2 (10.3–20.2)
8–48.8) 0.022 Yes 14.3 (�9.1 to 37.7)
.6–4.6) <0.0005 Total 29.5 (5.6–53.4)
mprovement; other abbreviations as in Tables 1, 2, and 5.
used with PENK. The increment in ROC areas whenbiomarkers were added to clinical scores did notachieve conventional levels of statistical significance.
Category-free reclassification analyses, using thecontinuous net reclassification improvement index(>0) (Table 6), calculated for the biomarkers added todifferent clinical risk scores (ADHERE, GWTG-HF,and OPTIMIZE-HF) for predicting inpatient mortal-ity, showed that natriuretic peptides up-classifiedrisk in those patients endpoints, whereas PENKpredominantly down-classified risk in patientswithout endpoints, with a smaller effect on up-classifying risk in patients who died (for ADHEREand GWTG-HF risk scores). Adding PENK to models
p Value Endpoint NRI (95% CI) p Value
IZE-HF Adding Natriuretic Peptide to GWTG
<0.0005 No �26.4 (�31.7 to �21.0) <0.0005
<0.0005 Yes 38.2 (11.8–64.6) 0.005
NS Total 11.8 (�15.2 to 38.8) NS
-HF Adding PENK to GWTG
<0.0005 No 10.0 (4.6–15.3) <0.0005
NS Yes 30.9 (4.5–57.3) 0.022
0.018 Total 40.9 (13.9–67.9) 0.003
riuretic Peptide Adding PENK to GWTG and Natriuretic Peptide
<0.0005 No 17.3 (11.9–22.6) <0.0005
NS Yes 27.3 (0.8–53.7) 0.043
0.016 Total 44.6 (17.6–71.5) 0.001
FIGURE 6 Classification and Regression Tree: In-Hospital Death
PENK (pmol/L)
≤ 133.3 (133.3 - 211.3) >211.3
%92.4
7.6Total 19.9
%86.9
13.1Total 10.0
Key: Survivor In-Patient Death
%97.9
2.1Total 70.1
%95.7
4.3Total 100.0
As shown in this decision tree for in-hospital deaths, increasing levels of proenkephalin A
(PENK) were associated with higher in-hospital mortality rates. Abbreviation as in
Figure 1.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
67
consisting of risk scores and natriuretic peptideconfirmed that PENK predominantly down-classifiedrisk in survivors with less effect on up-classifyingrisk in patients with events (for ADHERE andGWTG-HF risk scores).
DECISION TREE ANALYSIS. To determine optimalcutpoints for PENK, we constructed decision treesusing PENK, natriuretic peptides, and clinical riskscores (ADHERE, GWTG-HF, or OPTIMIZE-HF) toclassify patients into survivors or those who died inthe hospital. Only PENK was selected from thesevariables using chi-square automatic interactiondetection, with the cutoff points and associated in-hospital death rates (Figure 6). Patients with PENKlevels <133.3 pmol/l had an in-hospital mortalityrate of 2.1%, whereas for patients with PENK>211.3 pmol/l, the mortality rate was 13.1%.
DISCUSSION
In this observational multicenter cohort study, wedescribed the use of a recent plasma PENK assay forassessing WRF and risk stratification following acuteHF, by measuring an analyte that is stable in plasmafor at least 48 h, unlike previous assays of labile en-kephalins. Plasma PENK was strongly correlated withrenal function (eGFR), and the majority of its variancewas accounted for by 2 measures of renal function,eGFR and plasma urea. Plasma PENK was an inde-pendent predictor of WRF, and levels increased overtime while renal function declined. The temporalpattern of change of PENK-to-creatinine ratiosdiffered in patients who had WRF compared withpatients who did not, a finding suggesting that thedynamics of PENK and creatinine differed.
During follow-up, PENK was a strong independentpredictor of death and the composite endpoint ofdeath or HF, for both short-term (3-month) andlonger-term (1-year) follow-up. We had previouslydemonstrated that PENK predicted outcomes such asdeath, myocardial infarction, and HF following acutemyocardial infarction and was strongly linked torenal function (9). This present work reinforced ourearlier findings and extended it to another acutecardiovascular presentation: acute HF. For predictionof in-hospital mortality, PENK remained significantwhen used with a variety of risk scores developed forthis purpose (ADHERE, OPTIMIZE-HF, and GWTG-HF).However, the increment in ROC areas was modest andnot significant. Previous analyses of data relying onincrement in ROC areas could wrongly conclude that abiomarker has no added value (because of its veryconservative power) even though logistic regression,
which relies on increment in log-likelihood ratio, couldindicate otherwise (25,26). We used reclassification asan additional method to demonstrate the utility ofPENK, especially in down-classifying risk in patientswithout endpoints.
Individual studies and meta-analyses of HF cohortshave indicated the importance of renal impairment indetermining prognosis (14), but there are few bio-markers that predict cardiorenal syndrome type 1 (15).Creatinine itself might weakly predict WRF in acuteHF, but our findings suggest that PENK levels mightcontribute to a history of renal impairment, SBP, andplasma sodium, to provide modest accuracy in WRFprediction. The temporal patterns of plasma PENK inpatients who developed WRF were different fromthose in patients who were spared this complication,with rising levels of PENK and falling PENK-to-creatinine ratios seen in patients with WRF.
The pathophysiological link between PENK andprognosis in acute HF may be related to the effectsof opioids on the cardiovascular system (CentralIllustration). Enkephalin excess may have a car-diodepressive effect (4), by lowering BP and reducingorgan perfusion (including kidney perfusion, whichmay explain the strong link between PENK andeGFR). Fontana et al. (8) demonstrated a rise in atrialnatriuretic peptide and norepinephrine levels,together with responses in heart rate and bloodpressure, following nonspecific opioid blockade usingnaloxone in severe cases of acute HF. However, inless severe cases of acute HF, naloxone showed an
CENTRAL ILLUSTRATION Proenkephalin in Acute Heart Failure
Kidneyfunction
Enkephalin secretion
Enkephalin secretion
Acute heart failure
Worseningrenal
function
IncreasedPENKlevels
CardiacoutputKidney
perfusion
DiuresisBlood
volume
Ng, L.L. et al. J Am Coll Cardiol. 2017;69(1):56–69.
Acute heart failure leads to reduced renal perfusion and worsening kidney function, which further exacerbates salt and fluid retention. Secretion of amino acids 119 to
159 of proenkephalin A (PENK) may be a counter-regulatory response to mitigate declining renal function, although extremely high levels may be cardiodepressive
and lead to further decline in renal perfusion.
Ng et al. J A C C V O L . 6 9 , N O . 1 , 2 0 1 7
Proenkephalin in Acute Heart Failure J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9
68
opposite depressive effect on atrial natriuretic pep-tide and no effect on heart rate and blood pressure.The data suggest that endogenous opioids could besuppressing atrial natriuretic peptide secretion insevere acute HF. Peacock et al. (7) also describedpoorer outcomes in patients with acute HF who wereadministered opiates within the ADHERE registry.Similarly, enkephalins may exert a direct effect onrenal function because Sezen et al. (27) reported thatd opioid agonists stimulate urine flow and sodiumexcretion, and, thus, in a situation where kidneyfunction declines, the increase in enkephalin releasemay be a counter-regulatory measure. Enkephalinsare widely expressed, and major sources of plasmaenkephalin and proenkephalin include the heart,adrenal glands, skeletal muscle, and kidney (6).
On the basis of our findings indicating links be-tween PENK and outcomes, there may be potential touse high PENK levels to select patients for intensifiedtherapy settings or low PENK levels to rule out suchcare. The links of elevated PENK levels with pooroutcome in acute HF agreed with findings in otheracute emergencies such as acute myocardial
infarction (9) and stroke (13). Better risk predictionmay allow clinicians to improve allocation of treat-ment and resources, including placement of patientswithin the hospital, frequency of monitoring of renalfunction, and initiation and speed of up-titration ofHF therapies associated with WRF (e.g., loop di-uretics, angiotensin-converting enzyme inhibitors,and aldosterone receptor blockers). Some of thesepossibilities may need to be addressed by well-designed clinical trials.STUDY LIMITATIONS. First, the basis of our findingswas a large number of patients prospectively enrolledand hospitalized in 3 European countries. Additionalstudies are warranted to validate these findings innon-European populations and in general practicesettings. Furthermore, different natriuretic peptideand cardiac troponin assays were used in the 3 sites.We attempted to mitigate this by z-transforming thelog-transformed peptide values. Additionally, wecannot comment on patients with terminal renalfailure who were undergoing long-term hemodialysisbecause they were excluded. Finally, troponin resultswere not available on all patients.
PERSPECTIVES
COMPETENCY IN MEDICAL KNOWLEDGE: PENK is
inversely related to prevalent renal function, predicts WRF in
acute HF, and provides an assessment of short-term and long-
term prognosis independent of renal function. It may be used in
conjunction with clinical scores for predicting in-hospital
mortality.
TRANSLATIONAL OUTLOOK: Further studies should assess
the utility of PENK to risk stratify and guide therapy for patients
with acute HF in comparison with other biomarkers.
J A C C V O L . 6 9 , N O . 1 , 2 0 1 7 Ng et al.J A N U A R Y 3 / 1 0 , 2 0 1 7 : 5 6 – 6 9 Proenkephalin in Acute Heart Failure
69
Prospective studies on the clinical effectiveness ofusing PENK for management strategies, whetherdirected at low-risk or high-risk groups, need to beperformed. Moreover, PENK as a predictor of WRFshould be compared with other potential biomarkers,such as neutrophil gelatinase–associated lipocalin orcystatin C, for which a multimarker panel could beinvestigated. Sequential sampling of PENK shouldalso be investigated to assess how quickly levels risebefore WRF development.
CONCLUSIONS
Following acute HF, circulating PENK levels reflectcardiorenal status and provide short-term and long-term prognostic information on both mortality andcardiovascular morbidity. PENK predicted WRF andcould be used in conjunction with different clinicalrisk scores for in-hospital mortality.
REPRINT REQUESTS AND CORRESPONDENCE: Dr.Leong L. Ng, Department of Cardiovascular Sciences,Clinical Sciences Wing, Glenfield Hospital, LeicesterLE3 9QP, United Kingdom. E-mail: [email protected].
RE F E RENCE S
1. Mebazaa A, Gayat E, Lassus J, et al., GREATNetwork. Association between elevated bloodglucose and outcome in acute heart failure: resultsfrom an international observational cohort. J AmColl Cardiol 2013;61:820–9.
2. Mentz RJ, O’Connor CM. Pathophysiology andclinical evaluation of acute heart failure. Nat RevCardiol 2016;13:28–35.
3. Cohen-SolalA,Laribi S, IshiharaS,et al. Prognosticmarkers of acute decompensated heart failure: theemerging roles of cardiac biomarkers and prognosticscores. Arch Cardiovasc Dis 2015;108:64–74.
4. Holaday JW. Cardiovascular effects of endoge-nous opiate systems. Annu Rev Pharmacol Toxicol1983;23:541–94.
5. van den Brink OW, Delbridge LM, Rosenfeldt FL,et al. Endogenous cardiac opioids: enkephalins inadaptation and protection of the heart. Heart LungCirc 2003;12:178–87.
6. Denning GM, Ackermann LW, Barna TJ, et al.Proenkephalin expression and enkephalin releaseare widely observed in non-neuronal tissues.Peptides 2008;29:83–92.
7. Peacock WF, Hollander JE, Diercks DB, et al.Morphine and outcomes in acute decompensatedheart failure: an ADHERE analysis. Emerg Med J2008;5:205–9.
8. Fontana F, Bernardi P, Pich EM, et al. Rela-tionship between plasma atrial natriuretic factorand opioid peptide levels in healthy subjects andin patients with acute congestive heart failure. EurHeart J 1993;14:219–25.
9. Ng LL, Sandhu JK, Narayan H, et al. Pro-enkephalin and prognosis after acute myocardialinfarction. J Am Coll Cardiol 2014;63:280–9.
10. Arbit B, Marston N, Shah K, et al. Prognostic use-fulness of proenkephalin in stable ambulatory patientswith heart failure. Am J Cardiol 2016;117:1310–4.
11. Shah KS, Taub P, Patel M, et al. Proenkephalinpredicts acute kidney injury in cardiac surgerypatients. Clin Nephrol 2015;83:29–35.
12. Marino R, Struck J, Hartmann O, et al. Diag-nostic and short-term prognostic utility of plasmapro-enkephalin (pro-ENK) for acute kidney injuryin patients admitted with sepsis in the emergencydepartment. J Nephrol 2015;28:717–24.
13. Doehner W, von Haehling S, Suhr J, et al.Elevated plasma levels of neuropeptide pro-enkephalin a predict mortality and functionaloutcome in ischemic stroke. J Am Coll Cardiol2012;60:346–54.
14. Damman K, Valente MA, Voors AA, et al. Renalimpairment, worsening renal function, andoutcome in patients with heart failure: an updatedmeta-analysis. Eur Heart J 2014;35:455–69.
15. Ronco C, Haapio M, House AA, et al. Car-diorenal syndrome. J Am Coll Cardiol 2008;52:1527–39.
16. Ernst A, Köhrle J, Bergmann A,Proenkephalin A. 119-159, a stable proenkephalinA precursor fragment identified in human circula-tion. Peptides 2006;27:1835–40.
17. Fonarow GC, Adams KF Jr., Abraham WT, et al.,ADHERE Scientific Advisory Committee, StudyGroup, and Investigators. Risk stratification for in-hospital mortality in acutely decompensated heartfailure: classification and regression tree analysis.JAMA 2005;293:572–80.
18. Peterson PN, Rumsfeld JS, Liang L, et al.American Heart Association Get With theGuidelines-Heart Failure Program. Circ CardiovascQual Outcomes 2010;3:25–32.
19. Abraham WT, Fonarow GC, Albert NM, et al.,OPTIMIZE-HF Investigators and Coordinators. Pre-dictors of in-hospital mortality in patients hospi-talized for heart failure: insights from the OrganizedProgram to Initiate Lifesaving Treatment in
Hospitalized Patients with Heart Failure (OPTIMIZE-HF). J Am Coll Cardiol 2008;52:347–56.
20. Ponikowski P, Voors AA, Anker SD, et al.2016 ESC guidelines for the diagnosis andtreatment of acute and chronic heart failure: theTask Force for the Diagnosis And Treatment ofAcute and Chronic Heart Failure of the EuropeanSociety of Cardiology (ESC). Eur Heart J 2016;37:2129–200.
21. Omland T, Persson A, Ng L, et al. N-terminalpro-B-type natriuretic peptide and long-termmortality in acute coronary syndromes. Circula-tion 2002;106:2913–8.
22. Damman K, Tang WH, Testani JM,McMurray JJ. Terminology and definition ofchanges renal function in heart failure. Eur Heart J2014;35:3413–6.
23. Kellum JA, Lameire N, KDIGO AKI GuidelineWork Group. Diagnosis, evaluation, and manage-ment of acute kidney injury: a KDIGO summary(Part 1). Crit Care 2013;17:204.
24. Pencina MJ, D’Agostino RB Sr., Steyerberg EW.Extensions of net reclassification improvementcalculations to measure usefulness of new bio-markers. Stat Med 2011;30:11–21.
25. Seshan VE, Gönen M, Begg CB. ComparingROC curves derived from regression models. StatMed 2013;32:1483–93.
26. Vickers AJ, Cronin AM, Begg CB. One statisti-cal test is sufficient for assessing new predictivemarkers. BMC Med Res Methodol 2011;11:13.
27. Sezen SF, Kenigs VA, Kapusta DR. Renalexcretory responses produced by the delta opioidagonist, BW373U86, in conscious rats.J Pharmacol Exp Ther 1998;287:238–45.
![B] SECONDARY SECURITIES MARKET Price …territories have been few and far between. This portends well for the Indian markets as it possibly indicates more of This portends well for](https://static.documents.pub/doc/80x56/5e675cf66856793f6975df10/b-secondary-securities-market-price-territories-have-been-few-and-far-between.jpg)
