Originalarticle
The beneficial prognostic value of hemoconcentration is negatively affected by hyponatremia in acute decompensated heart failure:
Data from the Korean Heart Failure (KorHF) Registry
§JaewonOh(MD)a,Seok-MinKang(MD, PhD)a,*, In-Cheol Kim(MD, PhD)a,
SeongwooHan(MD, PhD)b,Byung-Su Yoo(MD, PhD)c,Dong-JuChoi (MD,PhD)d, Jae-Joong Kim(MD, PhD)e,Eun-Seok Jeon (MD,PhD)f, Myeong-ChanCho (MD, PhD)g, Byung-HeeOh(MD, PhD)h,ShungChullChae(MD,PhD)i,Myung-MookLee(MD, PhD)j, Kyu-HyungRyu(MD, PhD)b
aDivisionofCardiology,SeveranceCardiovascularHospitalandCardiovascularResearchInstitute,Seoul,RepublicofKorea
bDivisionofCardiology,HallymUniversityMedicalCenter,Hwaseong,RepublicofKorea
cDivisionofCardiology,YonseiUniversityWonjuSeveranceChristianHospital,Wonju,RepublicofKorea
dDivisionofCardiology,SeoulNationalUniversityBundangHospital,Seongnam,RepublicofKorea
eDivisionofCardiology,UlsanUniversityAsanMedicalCenter,Seoul,RepublicofKorea
fDivisionofCardiology,SungkyunkwanUniversitySamsungMedicalCenter,Seoul,RepublicofKorea
gDivisionofCardiology,ChungbukNationalUniversityHospital,Cheongju,RepublicofKorea
hDivisionofCardiology,SeoulNationalUniversityHospital,Seoul,RepublicofKorea
iDivisionofCardiology,KyungpookNationalUniversityHospital,Daegu,RepublicofKorea
jDivisionofCardiology,DonggukUniversityIlsanHospital,Goyang,RepublicofKorea
ARTICLE INFO
Articlehistory:
Received30May2016
Receivedinrevisedform1August2016 Accepted6August2016
Availableonline31August2016
Keywords:
Heartfailure Hemoglobin Sodium Prognosis
ABSTRACT
Background:Hemoconcentration(HC)isassociatedwithreducedmortality,whereashyponatremia(HN) hasbeenassociatedwithanincreasedriskofadverseoutcomesinpatientswithacutedecompensated heartfailure(ADHF).WesoughttodetermineifthepresenceofHNinfluencesthebeneficialprognostic valueofHCinADHFpatients.
Methods:Weanalyzed2046ADHFpatientsfromtheKoreanHeartFailureRegistry.WedefinedHCasan increasedhemoglobinlevelfromadmissiontodischarge,andHNassodium<135mmol/Latadmission.
Ourprimarycompositeendpointwasall-causemortalityand/orHFre-hospitalization.
Results:Overall,HCoccurredin889(43.5%)patientsandHNwasobservedin418patients(20.4%).HC offeredhigher2-yearevent-freesurvivalinpatientswithoutHN(73.2%vs.63.1%forno-HC,log-rank p<0.001), but not in patients with HN (54.2% vs. 58.7% for no-HC, log-rank p=0.879, p for interaction=0.003). In a multiple Cox proportional hazard analysis, HC without HN conferred a significant event-free survival benefit (hazard ratio: 0.703, 95% confidence interval 0.542–0.912, p=0.008)overno-HCwithHN.
Conclusions: OnlyHCoccurringinADHFwithoutHNwasassociatedwithimprovedclinicaloutcomes.
TheseresultsprovidefurthersupportfortheimportanceofHNasachallengingtherapeutictargetin ADHFpatients.
ß2016JapaneseCollegeofCardiology.PublishedbyElsevierLtd.Allrightsreserved.
§AnabstractbasedonthisstudywaspresentedatACC2014,the63rdAnnualScientificSessionoftheAmericanCollegeofCardiology,March29–31,2014,Washington,DC, USA.
* Correspondingauthorat:DivisionofCardiology,DepartmentofInternalMedicine,YonseiUniversityCollegeofMedicine,134Shinchon-dongSeodaemun-gu, Seoul120-752,RepublicofKorea.Fax:+82222277732.
E-mailaddress:[email protected](S.-M.Kang).
ContentslistsavailableatScienceDirect
Journal of Cardiology
j our na l ho me pa g e : w ww . e l se v i e r . com / l oca t e / j j cc
http://dx.doi.org/10.1016/j.jjcc.2016.08.003
0914-5087/ß2016JapaneseCollegeofCardiology.PublishedbyElsevierLtd.Allrightsreserved.
Introduction
Heartfailure(HF)isassociatedwithhighmorbidity,mortality, andhealthcareexpendituresindevelopedcountries.IntheUSA, almostonemillionhospitalizationsforHFoccur annually.Fluid overloadisa mainreason forre-hospitalizations inHFpatients [1].Theeffectiveandsaferemovalofcongestionisanimportant therapeuticgoalinhospitalizedacutedecompensatedheartfailure (ADHF)patients. Evidence-based data regarding theextentand durationofdecongestioninADHFhavebeenlimited.Hemocon- centration(HC),therelativeincreaseinthecellularelementsin blood,isaclinicalparameterthatpredictseffectivediuresisandis relatedto aggressive fluidremoval. Severalrecent studieshave shown that HC was related to improved clinical outcomes in patientswithADHF[2–6].
Hyponatremia(HN)isthemostcommonelectrolyteabnormal- ityandisassociatedwithadverseclinicaloutcomesinhospitalized patientswithADHF[7–11].Therefore,ithasbeenacomponentof theriskfactorsusedinthepredictionofprognosisinHFpatients [12]. The pathophysiology of HN in ADHF is predominantly hypervolemic,accompaniedbyanexcessofbodywater,whichisa markerofcongestion.However,theexaggeratedsodiumlossbya highdoseofdiureticscouldleadtodepletionofsodium,resulting in depletional HN during decongestion therapy, as it were treatment-induced hyponatremia. Therefore, we set out to determine if the presence of HN influences the beneficial prognosticvalueofHCinADHFpatientsusingdatafromalarge nationwideregistryinKorea.Toourknowledge,therehavebeen nostudiesconductedtoevaluatetheassociationbetweenHCwith orwithoutHNandsubsequentclinicaloutcomesinADHF.
Methods
Studysampleanddesign
The primary results of the Korean Heart Failure (KorHF) Registryhavebeenpreviouslyreported[5,10,13,14].Briefly,KorHF Registrywasanationwide,prospective,observational,multicen- ter, online registry that investigated the etiology, clinical characteristics, treatment modalities, morbidity, mortality, and theprognosticmarkersofhospitalizedADHFpatients.Atotalof 3200patients withthe diagnosisof ADHF who wereadmitted within24haftersymptomonsetwereenrolledfrom24hospitals inKorea.TheADHFdiagnosiswasbasedonspecificsymptomsin thepatient’smedicalhistoriesandsignsonphysicalexamination, accordingtotheFraminghamcriteria.AconfirmeddiagnosisofHF wasrequiredalsoatdischarge[13].Fromtheinitialrecruitmentof 3200patients,357patientswithoutavailableechocardiographic data and 797 patients without other baseline and discharge laboratorydata[e.g.hemoglobin(Hb),serumsodium,etc.]were excluded. Thus, the final analysis included 2046 patients. The primary composite endpoint of all-cause mortality and/or re- hospitalizationduetoHFexacerbationwascollectedbyareviewof themedicalrecordsandfromtelephoneinterviewsconductedat theendofthestudy(1-yearfollow-uprate:78.4%,2-yearfollow- up rate: 65.5%). Research coordinators guided by documented definitionsusedstandardizedreportformstocollectthefollow-up events. Medical records were reviewed whenever patients requiredrepeathospitalization.Inadditiontopatienttelephone interviews, the referring physicians and institutions were con- tactedwhennecessaryforadditionalinformation.
HCwasdefinedas anincreasedHblevel fromadmission to discharge, i.e. Hb change (DHb)>0 [5]. We used the widely accepteddefinitionofHNasaserumsodiumlevel<135mmol/Lat admission and discharge [10]. We calculated the estimated glomerular filtrationrate(eGFR) usingthe Modificationof Diet
in Renal Disease (MDRD) equation: eGFR=175[standardized serum creatinine (mg/dL)] 1.154age 0.203(0.742 if female)(1.212ifblack)[14].
Statisticalanalysis
Continuousvariablesweredescribedusingmeansandstandard deviations(ormedianandinterquartilerangewhenitdistributes non-normally), and categorical variables were described using numbersorpercentages.Wecompareddifferencesamonggroups using Student’st-test,Chi-squaretest,and ANOVAifnecessary.
Kaplan–Meier(K–M)survivalanalysiswasusedtoestimateevent- free survival, and log-rank tests wereused tocompare clinical outcomesinpatientswithHCandno-HC.Independenteffectsof variables and p-value for interaction on clinical events were calculated usingCox multivariableproportional hazardsregres- sionanalysisandincorporatingcovariateswithp-valueslessthan 0.1 from unadjusted analyses. Hazard ratios (HRs) with 95%
confidenceintervals(CIs)demonstratedtheriskofclinicalevents.
Correlations between various laboratory value changes were examined byPearsoncorrelationanalysis.Valuesofpless than 0.05 were considered statistically significant and all reported probabilityvaluesweretwo-tailed.Allanalyseswereconducted using SPSS version 21.0 software (SPSS/IBM Corp., Chicago, IL, USA).
Results
Baselinecharacteristics
Thebaseline characteristicsand laboratoryfindings ofADHF patients (4 groups) are presented in Table 1 according to the presenceofHCandHN.Overall,inADHFpatients,themeanHb levelwassignificantlydecreased(12.42.3g/dLvs.12.12.1g/
dL, paired t-testp<0.001), but the serumsodium level wasnot significantly changed (138.15.2mmol/L vs. 138.24.6mmol/L, pairedt-testp=0.660)fromadmissiontodischarge.Nosignificant correlation was observed between Hb and serum sodium level changes between admission and discharge (r= 0.039, p=0.080, n=1991).BasedonourdefinitionofHCandHN,HCoccurredin889 (43.5%) patients and HN was observed in 418 (20.4%) patients.
PatientswhodevelopedHChadsignificantlylowerHb,glucose,total cholesterollevels,andhighereGFRatadmissioncomparedtono-HC patients. Patients with HN had significantly lower Hb, total cholesterol,andsodium,whiledemonstratinghigherglucose,blood urea nitrogen (BUN), and creatinine level compared to patients withoutHN.However,BUN/creatinineratio,leftventricularejection fraction,theprevalenceofhypertension,diabetesmellitus,HFwith reduced ejection fraction, and medication administration history werenotsignificantlydifferentamongthefourgroups.
Associationswithclinicaloutcomes
Theprimaryendpointofourstudywasthecompositeofall-cause mortality and/or re-hospitalization for HF aggravation. During medianfollow-upof371days(interquartilerange,85–872days),2- yearevents asa primary compositeendpointoccurredin 34.6%
(n=708)ofthepatients,includingin16.6%(n=339)ofdeaths.InK–
Msurvivalanalysis,theHCgrouphadasignificantlyhigher2-year event-free survival than the no-HC group (73.2% vs. 63.1%, respectively,log-rankp<0.001;Fig.1a).Basedontheunadjusted Cox proportional hazards analysis, bothHC (HR:0.759, 95% CI:
0.658–0.875,p<0.001)andHN(HR:1.471,95%CI:1.251–1.729, p<0.001) were prognostic predictors. However, based on the adjustedCoxregressionmodel,onlyHC(HR:0.74395%CI:0.632–
0.874, p<0.001),but not HN(HR: 1.088, 95% CI: 0.896–1.322,
p=0.393),was associatedwithsignificantlybetterevent-free survivalafter adjustmentfor heart failureadmission history, NewYorkHeartAssociationIII/IVstatus,diabetesmellitus,beta- blockers at discharge, age, body mass index, diastolic blood pressure,totalcholesterol, BUN,and creatinine(Table 2)and theseresultswereunchangedafterfurtheradjustingforknown riskfactorssuchassystolicbloodpressure,andleftventricular ejectionfraction.
TheeffectofhyponatremiaonclinicaloutcomesinpatientswithHCor no-HC
TodetermineifthepresenceofHNinfluencesthebeneficial prognosticvalueofHC,weperformedsubgroupanalyses(Fig.2).
First, we conducted a subgroup analysis according to the presence or absence of HN at admission. HC was a good prognosticmarker inpatients without HN(2-yearevent-free survival,68.0%vs.59.1%forno-HC,log-rankp<0.001,n=1628, Fig.1b),butnotinpatientswithHN(2-yearevent-freesurvival, 54.2%vs. 58.7%fornon-HC, log-rankp=0.879, n=418, pfor interaction=0.003,Fig.1c).InK–M survivalanalysis,HNwas associatedwithasignificantlylower2-yearevent-freesurvival bothinpatientswithHC(54.2%vs.73.2%forwithoutHN,log- rankp<0.001,n=889)andno-HCpatients(58.7%vs.63.1%for withoutHN,log-rank p=0.037,n=1157).InCoxproportional hazardanalysis,HCwasanindependentpredictorforprimary compositeendpoint(HR0.723,95%CI0.601–0.871,p=0.001)in
patientswithoutHN,butnotinpatientswithHNatadmission (HR0.807,95%CI0.575–1.134,p=0.217),consistentlywiththe resultsofK–Manalyses.AndHCwithoutHNwasanindependent and additive marker of good prognosis for the primary compositeendpointinthesameadjustedCoxregressionmodel (HR0.703,95%CI0.542–0.912,p=0.008,Table3).Wealsohad similarresultsforeachcomponent(e.g.all-causemortalityorHF re-hospitalization) of primary composite endpoint (data not shown).
Second,whenwedividedsubgroupsofpatientsbasedonthe presenceofHNatdischarge,thebeneficialprognosticvalueof HCwaspreservedinpatientswithoutHNatdischarge(2-year event-free survival, 71.6% vs. 64.1% for no-HC, log-rank p=0.001, n=1639), but disappeared in patients with HN at discharge(2-yearevent-freesurvival,59.5%vs.53.8%forno-HC, log-rankp=0.103,pforinteraction=0.037,n=352),whichis similar to the results from the analysis of data obtained at admission.
Third,weanalyzedtherelationshipbetweenHCandsodium level changes, and we analyzed only ADHF patients with HN (Na<135mmol/L)atadmission.Wecategorizedpatientsintotwo subgroups:improvedHN(Na135mmol/Latdischarge,n=244) and persistent HN (Na<135mmol/L at discharge, n=168). In patientswithHC,nosignificantdifferenceswereobservedinterms of2-yearevent-freesurvivalbetween thetwogroups(54.3%vs.
54.9% for persistent HN, log-rank p=0.677, n=165, Fig. 3a).
However,forthosewithno-HC,patientswithimprovedHNhada Table1
Baselinecharacteristicsoffourgroupsaccordingtothepresenceofhemoconcentration(HC)orhyponatremia(HN).
No-HC HC
WithHN (n=252)
WithoutHN (n=905)
WithHN (n=166)
WithoutHN (n=723) Clinical
Malegender,n(%) 127(50.4) 475(52.5) 74(44.6) 345(47.7)
Age,years 7014a 6815 7114 6714
BMI,kg/m2 22.23.7a 23.74.2 22.53.4 23.34.1
NYHAIII/IV,n(%) 153(60.7) 528(58.3) 105(63.3) 460(63.6)
IschemicoriginofHF,n(%) 93(36.9)a 384(42.4) 66(39.8) 241(33.3)
HFadmhistory,n(%) 94(37.3)a 221(24.4) 58(34.9) 186(25.7)
Diabetes,n(%) 90(35.7)a 268(29.6) 62(37.3) 208(28.8)
Hypertension,n(%) 119(47.2) 430(47.5) 188(53.0) 314(43.4)
SBP,mmHg 12731a 13532 12828 13227
DBP,mmHg 7618a 8119 7516 7918
Heartrate/min 9526a 9326 9328 9026
Laboratory
Hbadm,g/dL 12.52.2a 13.22.0 10.92.1 11.82.3
Hbdc,g/dL 11.12.0a 11.92.0 12.12.0 12.92.2
Glucose,mg/dL 183.9103.8a 165.182.1 182.393.2 147.270.6
Cholesterol,mg/dL 161.445.7a 172.151.0 146.540.2 158.745.2
BUN,mg/dL 30.920.8a 23.113.7 30.318.2 23.114.1
Creatinine,mg/dL 1.721.31a 1.371.07 1.771.52 1.381.08
BUN/Cr 19.59.1 18.59.2 19.38.0 18.57.9
Sodiumadm,mmol/L 130.74.3a 139.93.3 130.44.7 140.23.2
Sodiumdc,mmol/L 135.45.4a 139.04.4 135.04.8 138.83.9
eGFRbyMDRD,L/min/1.73m2 50.630.0a 60.438.9 50.829.9 63.438.5
LogNT-proBNP(n=1409) 8.671.30a 8.111.53 8.981.23 8.331.38
Echocardiographic
LVEF,% 38.616.7 39.715.5 37.715.2 38.515.8
HFrEF,n(%) 171(67.9) 585(64.6) 116(69.9) 489(67.6)
Dischargemedications
ACEinhibitors/ARBs,n(%) 150(59.5) 597(66.0) 105(63.3) 459(63.5)
Beta-blockers,n(%) 85(33.7) 373(41.2) 55(33.1) 285(39.4)
Spironolactone,n(%) 63(25.0) 193(21.3) 37(22.3) 182(25.2)
Valuesarethemeanstandarddeviationorn(%).Hyponatremiaisdefinedasserumsodium<135mmol/L.BMI,bodymassindex;NYHA,NewYorkHeartAssociation functionalclass;HF,heartfailure;adm,admission;DBP,diastolicbloodpressure;SBP,systolicbloodpressure;Hb,hemoglobin;dc,discharge;BUN,bloodureanitrogen;eGFR, estimatedglomerularfiltrationratio;MDRD,ModificationofDietinRenalDisease;NT-proBNP,N-terminalproB-typenatriureticpeptide;LVEF,leftventricularejection fraction;HFrEF,heartfailurewithreducedLVEF(<45%);ACE,angiotensin-convertingenzyme;ARB,angiotensinIIreceptorblocker.
aIndicatesp-value<0.05byChi-squaretestorANOVA.
significantly higher 2-year event-free survival compared to patients withpersistentHN(63.3% vs.50.5% for persistentHN, log-rankp=0.002,n=247,pforinteraction=0.881,Fig.3b).Fig.4 showsforestplotsforadjustedCoxregressionanalysisforprimary compositeendpointaccordingtoeachsubgroup.
Discussion
Theprincipalfindingofthisstudyisthat(1)HNatadmissionor dischargeinfluencesnegativelytheassociatedclinicaloutcomesof thepatientswithHCandthat(2)HCwithoutHNatadmissionisan independentandadditivemarkerofgoodprognosisinADHF.
HCisaclinicalparameterthatcanpredicteffectivediuresisand agoodprognosisinpatientswithADHF.HChasbeendefinedasa measurement of non-plasma components of the intravascular spacesuchasHb,hematocrit,protein,andalbumin[2–5,15,16].Of these,themostwidelyuseddefinitionofHCisthechangeinHb [3,5,15,16].However,whenwedefineHCasthechangeinHb,we havetoassumethatthechangeinHbispredominantlydrivenby changes in volume status rather than bleeding, bone marrow dysfunction, and subsequent transfusion[16].Until now, there havebeenfewreportsdescribingthesubgroupofpatientsinwhich HCisnotrelatedtogoodprognosis.Inourstudy,HCwasagood prognosticpredictor,irrespectiveofanemiaandrenaldysfunction, whichcouldbeawidelyusedsubgroupinanADHFstudy(datanot shown). However, there have been few reports describing the subgroupofpatientsinwhichHCisnotrelatedtogoodprognosis.
ArecentlypublishedstudyreportedthatonlylateHC(HCinthe latterhalfofthehospitalization)wasassociatedwithimproved survivalcomparedtoearlyHC[16].Inthepresentstudy,wefound thatHCwasnotagoodpredictorinADHFpatientswithHN(at admissionanddischarge).
The pathophysiology of HN in ADHF is complex, generally considered as hypervolemic (dilutional) HN, and a problem of impairedwaterexcretionratherthanimpairedsodiumexcretion.
However,thecombinationofsodium-restricteddietandexagger- atedsodiumlossesbyahighdoseofdiureticagents,especiallyin patientswithseverehyperglycemia,mightleadtoseveredepletion of sodium, resulting in depletional HN. So over-decongested patientsbyhighdoseofdiureticscouldbehyponatremiarather thandecongestion-inducednormonatremiaorhypernatremia.But it is hard to distinguishwhether HN comes fromdilutional or depletional HN even though N-terminal pro B-type natriuretic peptidelevelwashigherinpatientswithHNthanthosewithout HN,suggestingdilutionalHNplaysaroleinHNdevelopmentofour study.In addition,ourresultsmayimply thatboth HCandHN influencetheprognosisofADHFindependentlyandthedifferent pathophysiologicalmechanismmayexistbetweenHCandHNin ADHFand HNmayresultfromothermechanismsbeyond fluid overload. Therefore,we can speculate thebeneficial prognostic value of HC might be influenced according to the situation of volumeorsodiumbalance.Assuch,theclinicalinterpretationsof HC in HN settings might be cautious, in particular with consideration to the balance between sodium and free water.
Maybe,adifferentdecongestiontherapymaybeneededbeyond HC-guidedtherapy(e.g.ultrafiltrationoravasopressinantagonist) for patients with HN [17,18]. Further study is warranted to consideranddistinguishtheHNorigininADHF.
IntheclinicalsettingofADHF,therehavebeensomeconflicting reportsregardingwhetherimprovingHNduringhospitalizationis associatedwitha goodoradetrimentalprognosis.Inmostpost hoc, retrospectively designed studies, improved HN and the correctionofHNatdischargehavenotbeenfoundtoberelated to improved clinical outcomes in ADHF, even though HN at admissionisrelatedtoincreasedshort-termmortality[7,8,10].Ina randomizedclinicaltrial,Efficacyof VasopressinAntagonismin HeartFailureOutcomeStudyWithTolvaptan(EVEREST),tolvaptan improvedonlyHFsignsandsymptoms,buthadnoeffectonlong- termmortalityinpatientshospitalizedwithHF[19–21].However, intherecentposthocanalysisofEVERESTstudy,tolvaptanwas found to be associated with improved long-term outcomes in patientswithsevereHN(Na<130mmol/L)[22,23].
Fig.1.Theprimarycompositeendpointforhemoconcentrationinallpatients(Panel A),patientswithouthyponatremia(PanelB),andwithhyponatremiaatadmission (PanelC).HC,hemoconcentration;Adm,admission;ADHF,acutedecompensated heartfailure.
