Spinal cordinjuryisassociatedwithanincreasedriskof atrial
fbrillation: Apopulation-basedcohortstudy
Chun-Cheng
Wang,MD,
*†Chiz-Tzung Chang,MD,PhD,
‡§Cheng-Li Lin,MSc,
‡║Bor-Ren
Huang,MD,
*¶Chia-Hung Kao,MD
*#Introduction
Spinal cordinjury(SCI)isaseveremedicalconditionthat cancausevariousmotor,sensory,andautonomicdysfunc- tions.1 SCI
mayalsocauselong-lastingdysfunctionin several organsystems,leadingtohighmorbidityandan adverse prognosis.2 Cardiovascular diseaseistheleading cause ofdeathinpatientswithSCI.3–5
Vascularthrom- boembolism, acutemyocardialinfarction,posturalhypoten- sion, orthostatichypotension,andcardiacarrhythmiamay have beenreportedtooccurinpatientswithSCI.6–9 Cardiac arrhythmiaisfrequentlyobservedinpatientswith SCI andisgenerallyconsideredtobecausedbyautonomic dysfunction.Becausethespinalsympatheticpathwaysthat controltheheartandmaintainvasculartoneexitatthe frst thoracicvertebratofourththoracicvertebra(Th1-Th4)levels, anunopposedparasympathetictonemaybepresentatthe cervicalorhighthoracic(aboveTh5)levelinpatientswith SCI.Life-threateningbradyarrhythmiaorsinusarrestrequir- ingtheimplantationofapacemakerhasalsobeenreported.8 Atrial fbrillation(AF),themostcommonsustainedcardiac arrhythmiaencounteredinclinical
practicewithaprevalence of 1.0%–2.0%inthegeneralpopulation,canleadtosubstantial morbidity, reducedqualityoflife,andincreasedmortality.10,11 AF isacomplexdiseasecausedbymultipleunderlying
mechanisms,includingautonomicneuraldysregulation. Despite
theclinicalsignifcanceofAF,toourknowledge,no systemicanalysisofalargecohorthasbeenconductedto investigatetheassociationbetweenSCIandtheincidenceof AF.
Therefore,weconductedthisretrospectivecohortstudy, using
anationwidedatabase,toinvestigatethisissue.
Methods
Data source In thisretrospectivecohortstudy,weuseddatafromthe NationalHealthInsuranceResearchDatabase(NHIRD)of Taiwan, whichcontainscomprehensiveinformationon clinical visitsandadmissionsforeachinsurant,including demographic data,visitdates, International Classifcationof Disease, NinthRevision,ClinicalModifcation (ICD-9-CM) diagnostic codes,andprescriptionhistories.Thecompre- hensive NationalHealthInsurance(NHI)program12 was
established in1995andprovidescoverageto 499% ofthe population
of23.7millionpeople.TheTaiwanNational Health ResearchInstituteadministerstheNHIRD,and before theelectronic fles arereleased,personalidentifcation
informationisencryptedtoprotectpatientprivacy.Inthis study,
weusedclaimsdatafromtheLongitudinalHealth Insurance Database2000,asubsetoftheNHIRDthat contains allmedicalclaimsdataonarandomsampleof 1,000,000 benefciaries
oftheNHIprogram.TheLongitu- dinal HealthInsuranceDatabase2000dataonsex,age,and health carecostshavebeenprovedtoberepresentativeofthe all insurants.Thisstudywasapprovedtofulfll thecondition for exemptionbytheInstitutionalReviewBoardofChina Medical University(CMUH-104-REC2-115).Theinstitu- tional reviewboardalsospecifcally waivedtheconsent requirement. Sampled participants Patients 20yearsandolderwhowerenewlyhospitalizedfor SCI (ICD-9-CM codes 806and952)between2000and2011 comprised theSCIcohort.TheSCIdiagnosisdatewas defned astheindexdate.Weclassifed patientswithSCI into 4subgroups:cervicalSCI(ICD-9-CM codes 806.0, 806.1, 952.0,and952.00–952.09); thoracicSCI(ICD-9-CM codes 806.2,806.3,952.1,and952.10– 952.19); lumbar, sacral, andcoccygealSCI(ICD-9-CM codes 806.4,806.5, 806.6,
806.7,806.8,806.9,952.2,952.3,952.4,952.8,and 952.9);
andmultilevelSCI.Weexcludedpatientsdiagnosed with AF(ICD-9-CM codes427.3and427.31);atrial futter (ICD-9-CM code 427.32);supraventriculartachycardia, ventricular
tachycardia,andheartblock(ICD-9-CM codes 426–427.5
and427.81);andpacemakerimplantationbefore the
indexdate.Forthenon-SCIcohort,werandomlyselected controls ataratioof4:1andfrequency-matchedthemto patients withSCIbyage,sex,andindexyear.Thesame exclusion criteriawereappliedtothenon-SCIcohortas applied totheSCIcohort. Outcomes andcomorbidities Each
studyparticipantwasfolloweduntiladiagnosisofAF, death,
losstofollow-up,withdrawalfromtheNHIprogram, or December31,2011,whicheveroccurred frst. We
usedinpatientdiagnosis fles todeterminebaseline comorbidities, includingdiabetes(ICD-9-CM code 250), hypertension (ICD-9-CM codes 401–405), hyperlipidemia (ICD-9-CM code
272),chronicobstructivepulmonarydis- ease (COPD)(ICD-9-CM codes 491,492,and496),heart failure (ICD-9-CM code 428),coronaryarterydisease (CAD) (ICD-9-CM codes 410–414), stroke(ICD-9-CM codes 430–438), cancer(ICD-9-CM codes 140–241), hyper- thyroidism (ICD-9-CM code
242),andchemotherapy. Statistical analysis We frst
comparedthedistributionsofsex,age,andbaseline comorbidities betweentheSCIandnon-SCIcohorts.Then, the χ2 test wasusedtoexaminecategoricalvariablesandthe Student t test
wasusedtoexaminecontinuousvariables.The cumulative incidenceofAFintheSCIandnon-SCIcohorts was assessedusingtheKaplan-Meiermethod,anddiffer- ences wereassessedusingalog-ranktest.Theincidence densities, stratifed accordingtosex,age,andcomorbidities, were
estimatedforbothcohorts.Therelativeriskof developing AFinpatientswithSCI,comparedwiththe non-SCI cohort,wasanalyzedusingunivariateandmulti- variate Coxproportionalhazardsregressionmodels.Hazard ratios (HRs)and95%confdence intervals(CIs)were estimated usingtheCoxmodels.ThemultivariateCox models wereadjustedforage,sex,andthecomorbidities of
heartfailure(CHF),CAD,stroke,andhyperthyroidism. Further
analysiswasperformedtoassesswhetherAFwas associated withthelevelofSCI.A2-tailed P value of o.05 was consideredtobestatisticallysignifcant. Allanalyses were
performedusingSASstatisticalsoftwareversion9.3 for Windows(SASInstitute,Inc.,Cary,NC).
Results
This studyincluded41,691patientswithSCIand166,764 patients withoutSCI(Table 1).Nostatisticaldifferencewas observed inthedistributionbetweentheSCIandnon-SCI cohorts.
TheSCIcohortwassignifcantly olderthanthenon- SCI cohort(meanage52.0 _ 18.1 yearsvs51.5 _ 18.2 years; P o .001). Overall,patientswerepredominantlymen (63.4%)
andolderthan50years(52.5%).Comparedwiththe non-SCI cohort,theSCIcohorthadsignifcantly higher proportions ofdiabetes,hypertension,hyperlipidemia, COPD,
heartfailure,CAD,stroke,andhyperthyroidism (P o .001 forall). The
meanfollow-upperiodswere5.69and6.17yearsfor the SCIandnon-SCIcohorts,respectively(P o .001). The median durationfromthestudytothe frst occurrenceofAF in
theSCIcohortwas2.66yearsascomparedwith3.51years in thenon-SCIcohort.Themedianend-pointtimeratiowas 0.76. Asshownin Figure 1, theprobabilityofAFwas signifcantly
higherintheSCIcohortthaninthenon-SCI cohort (log-ranktest, P o .001). The
overallincidenceofAFwas1.36-foldhigherinthe SCI cohortthaninthenon-SCIcohort(2.70casesper1000 person-years vs1.99casesper1000person-years),repre- senting anadjustedHRof1.28(95%CI1.17–1.40)
(Table 2). Comparedwiththenon-SCIcohort,theSCI cohort exhibitedsignifcantly
higherrisksofdevelopingAF in bothmaleandfemalesubgroups(adjustedHR1.40;95% CI 1.23–1.58 andadjustedHR1.16;95%CI1.01–1.32). When stratifed byage,theSCIgrouphadsignifcantly higher risksofdevelopingAFinthe50-to65-yearage subgroup (adjustedHR1.41;95%CI1.11–1.78) andinthe 465-year agesubgroup(adjustedHR1.21;95%CI1.09– 1.34) ascomparedwiththenon-SCIgroup.Therewas no signifcant differenceintheriskofAFbetweenSCIand non-SCI groupsinthe r49-year
agesubgroup.Stratifed according tocomorbidities,therelativeriskofAFwas signifcantly
higherintheSCIcohortthaninthenon-SCI cohort forbothpatientswith(crudeHR1.03;95%CI0.92– 1.15 andadjustedHR1.25;95%CI1.11–1.40) andwithout (crude HR0.88;95%CI0.76–1.02 andadjustedHR1.33; 95% CI1.14–1.54) comorbidity. In themultivariateanalysis,theadjustedHRforAFwas 1.12-fold
higherforwomenthanformen(95%CI1.04– 1.21), whichincreasedwithage(every1year)(adjustedHR 1.10; 95%CI1.09–1.10). Inaddition,thecomorbiditiesof diabetes,
hypertension,COPD,heartfailure,andCADwere signifcantly associatedwithincreasedrisksofAF(Table 3). Compared withpatientswithoutSCI,patientswithmulti- level
SCIwere1.83-foldmorelikelytodevelopAF(95%CI 1.45–2.31), followedbypatientswithlumber,sacral,coccy- geal spinalcordinjury(L-S-Co-spinalinjury)(adjustedHR 1.31; 95%CI1.14–1.50) andthoracicspinalinjury(T-spinal injury) (adjustedHR1.31;95%CI1.09–1.57) (Table 4).
Tofurtherinvestigatetheinteractionbetweencancerand
SCIforthedevelopmentofAF,wedividedtheSCIgroupinto
associatedSCIsubgroupexhibitedasignifcantlyhigherrisk of developingAF(adjustedHR1.29;95%CI1.17– 1.41). Comparisons ofdifferencesinbaselinecharacteristics between
cancer-andnon–cancer-associatedSCIsubgroups are shownin Table 6. Comparedwiththenon–cancer- associated
subgroup,thecancer-associatedSCIsubgroup was signifcantly older,waspredominantlywomen,andhad signifcantly higherproportionsofdiabetes,hypertension, hyperlipidemia,
hyperthyroidism,stroke,CAD,COPD,and CHF. Accordingtothelevelofinjury,thecancer-associated SCI subgrouphadsignifcantly higherproportionsof T-spinal andL-S-Co-spinalinjuriesthandidthenon–cancer-associated SCIsubgroup. Differences inbaselinecharacteristicsbetweendif- ferent
levelsofSCIwerefurthercompared(Table 7). Compared withpatientswithcervicalspinalinjury (C-spinal injury),patientswithT-spinalinjury,L-S-Co- spinal injury,andmultilevelSCIwerefemalepredominant, were older,andhadhigherproportionsofdiabetes,hyper- tension, hyperlipidemia,COPD,heartfailure,CAD,and stroke.
Discussion
Interpretation ofthestudyresult OurresultsshowedthattheriskofAFafterlong-term follow-up
washigherintheSCIgroupthaninthenon-SCI group, particularlyinpatientswiththeT-spinalinjury,L-S- Co-spinal injury,andmultilevelspinalinjury. Association
betweentheriskofAFandSCI Cardiovascular dysfunctionshavebeenproposedtobethe most commoncausesofdeathinpatientswithchronicSCI. Among
thesedysfunctions,cardiacfailure,ventricular
tachycardia,AF,CAD,andcerebrovasculardiseasearethe most commoncausesofdeath.13,14 However, despitethe clinical
importanceofAFinpatientswithchronicSCI,
epidemiologicalevidencelinkingtheincidenceofAFtoSCI is lacking.Twocasereports15,16 have demonstratedthatAF occurs
duringepisodesofautonomicdysrefexia. Prakash et al17
comparedelectrocardiographic fndings frompatients with chronicSCIwiththosefromable-bodiedcontrolsand found thattheprevalenceofAFwassimilarbetweenthe2 groups. However,inourstudy,theriskofAFwashigherin the
chronicSCIgroup.Thedifferencesbetweenthestudy results maybeduetothedifferentbaselinedemographic characteristics ofthecontrolgroups.Inourstudy,theSCI group hadsignifcantly
higherproportionsofhypertension, CHF, CAD,andCOPDthandidthenon-SCIgroup. However, Prakashetalreportedsignifcantly higherpropor- tions
ofhypertension,CHF,CAD,andCOPDintheirable- bodied controlgroupthanintheSCIgroup.Chronichyper- tension,
CHF,CAD,andCOPDhavebeensuggestedtobe risk factorsforAF.11,18–22 To
clarifywhetherSCIisan independent riskfactorforpredictingtheriskofAFafter controllingforthesepotentialconfoundingfactors,we performed
additionalsubgroupanalyses.Theresults revealed
thatinthesubgroupwithoutanycomorbidities, patients withSCIstillhadasignifcantly higherriskofAFin the long-termfollow-upperiod. Electrocardiographic
changesinpatientswithSCI Previous studies23–25 have investigatedtheelectrocardio-
graphic changesinpatientswithchronicSCI.Ravensbergen et al23
investigatedtheelectrocardiographicpredictorsfor arrhythmia
inpatientswithSCIandfoundthatP-wave dispersion (PWD)wassignifcantly higherinpatientswith SCI thanintheable-bodiedgroup.Akbaletal24 reported that
PWD wassignifcantly higherinpatientswithSCIfor Z1 year thaninthosewithSCIfor o1 year.Yasaretal25 compared PWDinpatientswithSCIduringthepostinjury period (SCIfor
46 months)andable-bodiedhealthy controls. TheauthorsconcludedthatPWDinpatientswith an injurylevelofTh6orhigherwassignifcantly higherthan in patientswithaninjurylevelofTh7orlower.Inaddition, PWD inbothgroupswashigherthaninthehealthycontrol group, implyingthatautonomicdysfunction(usuallyseenin an injurylevelofTh6orabove)mayinfuence PWD.Higher PWD
hasalsobeenproposedtobeassociatedwithan increased riskofAF.26 Proposed
mechanismsofcardiacarrhythmia in patientswithSCI The
autonomicnervoussystemplaysanimportantrolein cardiovascular control.AfterSCI,disruptionofthecommu- nication
betweenthebrainstemandtheautonomicnervous system canoccur.8,13
Sympatheticpreganglionic fbers exit at
theTh1-Th4leveltoformsynapsesinthethoracicganglia with postganglionic fbers,
whichmayinnervatetheheart and beresponsibleforincreasingheartrateandcardiac output. Parasympathetic fbers exitthecentralnervous system
atthebrainstemlevelthroughthevagalnerveto innervate theheartandantagonizetheeffectofthesym- pathetic nervoussystem.Therefore,patientswithSCIatthe cervical
orhighthoraciclevel(aboveTh6)couldhave impaired sympatheticimpulsesandunopposedparasympa- thetic activity,therebyleadingtoautonomicdysfunction. Previous studies9,27 reported
thatmostcardiacarrhythmias (eg, sinusbradycardiaandsinusarrest)thatoccurinpatients with SCImainlydosobecauseofautonomicdysfunction and
thattheriskisusuallyattenuatedaftertheacutestage.In our study,thecumulativeincidenceofAFseemedto increase steadilywithtime,implyingthatthepathogenesis of
AFinpatientswithSCIcannotsimplybeexplainedby autonomic dysfunction.Inthechronicstage,cardiacdys-rhythmia suchasAFmayoccursecondarytoautonomic dysrefexia, asyndromeofrefex
sympatheticdischargefrom the preganglionicneuronsinthethoracolumbarspinalcord in responsetoviscerosensitivestimulioriginatingbelowthe level
ofinjuryinpatientswithSCIaboveT6.However,we observed anincreasedriskofAFinpatientswithT-spinal injury, L-S-Co-spinalinjury,andmultilevelSCI,while patients
withC-spinalinjuryshowedonlyanonsignifcant trend towardanincreasedriskofAF.Thereareseveral possible explanationsforthis.First,somepatientswithC- spinal injurymayhavemultilevelinjuryandweretherefore assigned tothemultilevelSCIsubgroup;hence,theassoci- ation
betweenC-spinalinjuryandAFmaybeunderesti- mated inourcohort.Second,comparedwithpatientswithC- spinal injury,patientswithT-spinalinjury,L-S-Cospinal injury, andmultilevelSCIweresignifcantly olderandhad higher proportionsofcancer,hypertension,diabetes,chronic kidney
disease,COPD,hyperthyroidism,heartfailure,CAD, and stroke,allofwhichmaycontributetoanincreasedrisk of AF.WeproposedthatthepathogenesisofAFinpatients with
chronicSCIinvolvesbothautonomicdysfunctionand other coexistingcardiovascularriskfactors.
Association betweenfemalesexandAF Previous
studieshavesuggestedthattheriskofAFishigher in menthaninwomen.17β-Estradiol mayattenuatethe shortening ofatrialrefractorinessthatoccursafterrapidatrial pacing,
therebydecreasingthesusceptibilitytoAF.28 How-ever, inourstudy,womenhadahigherriskofAFthandid
men. Thepossibleexplanationisthatasignifcant numberof patients
inourcohortwereinpostmenopausalperiod.A previousstudy29 suggestedthatthoughAFislessprevalentin
women, theoverallabsolutenumberofAFishigherinwomen. This maybeexplainedbythefactthatpostmenopausalwomen
havehigherrisksofheartfailureandischemicheartdisease, which maycontributetotheincreasedriskofAF.
Association betweencancerandtheriskofAF It
hasbeenreportedthatpatientswithcancerhadan increased riskofAF.Theexplanationsincludedirectcancer invasion, paraneoplasticmanifestations,physicaloremo- tional stress,increasedinfammation,cancer-relatedmeta- bolic orelectrolytedisorders,andcomplicationsinherentto
chemotherapyorsurgicaltreatment.30 In ourstudy,patients with
SCIwithcancerdidnothaveanincreasedriskofAF compared withthenon-SCIcohort.However,ourresultsare limited byarelativelysmallsamplesizeandinadequate statistical
powerforthepatientswithSCIwithcancer.We cannot concludewhetherpatientswithSCIwithcancerhave a higherriskofAFonthebasisofourresults. Study limitations There
areseverallimitationsofthisstudy.First,allthe informationwasobtainedfromamedicalclaimsdatabase and allpatientswereanonymous.Therefore,wecouldnot review medicalchartsorimagingstudiestoverifythe diagnoses.Theaccuracyof ICD-9-CM codes isamajor concern.
However,thediagnosesofmostcovariatesinour study havebeenvalidatedbefore.31 Second, becausesome
AF casesmaybeasymptomatic,thetrueincidenceofAF may
beunderestimated.SomeasymptomaticAFcasesmay not havebeenidentifed inthenon-SCIgroupbecauseofa lack ofphysicianvisits.However,theestimatedincidence rate
(0.8–28.3 casesper1000person-years).11 Our dataare
comparablewiththosefrompreviousstudies.Third,the NHIRD
doesnotcontaindetailedinformationonother potential confounderssuchassmokinghabits,alcohol consumption,bodymassindex,typesofheartdisease,heart failure
functionalclass,leftventricularejectionfraction, and metabolicorelectrolyteabnormalities,whichmayalso be involvedintheoccurrenceofAF.Fourth,thisisan
observationalstudy;therefore,wecannotmakedirectcon- clusions
aboutcausalrelationships.Finally,weexcluded patients youngerthan20years;therefore,ourconclusion cannot beextrapolatedtotheyoungerpopulation.
Conclusion
Thelong-termriskofAFwashigherintheSCIgroupthanin the non-SCIgroup,particularlyinpatientswithT-spinal injury, L-S-Cospinalinjury,andmultilevelspinalinjury.