Self-esteem change and diurnal cortisol secretion in older adulthood

Self-esteem change and diurnal cortisol secretion in older adulthood

Sarah Y. Liu

, Carsten Wrosch

* , Gregory E. Miller

, Jens C. Pruessner

aConcordiaUniversity,MontrealH4B1R6,Canada

bNorthwesternUniversity,Evanston,IL60208-2710,USA

cMcGillUniversity,MontrealH4H1R3Canada

KEYWORDS Self-esteem;

Perceivedstress;

Depressivesymptoms;

Diurnalcortisol secretion;

Olderadulthood

Summary

Objective: Researchsuggeststhatself-esteemcandeclineinolderadulthood.Thisprocesscould removeabufferthatnormallyprotectsindividualsagainstdistress-relatedchangesincortisol secretion.Weexaminedthispossibilitybytestingwhetherchangeinself-esteemwouldpredict alterationsin cortisol secretion,particularly amongolder adultswho reportedhigh levelsof depressivesymptomsorperceivedstress.

Methods: 147olderadults(aged60+)completedthreedaysofdiurnalcortisolmeasurementsat threedifferenttimepoints,namelyeverytwoyearsoveratotalperiodoffouryears.Measuresof self-esteem,depressivesymptoms,andperceivedstresswereassessedatT1andT2.Potential demographicandhealth-relatedconfoundsweremeasuredatbaseline(partnershipstatus,SES, mortalityriskindex,andmedication).

Results: Linearregressionmodelsindicatedthatadeclineinself-esteemfromT1toT2predicted elevatedcortisoloutput(AUCG)fromT2toT3, F(1,137)=8.09,b=.25,R²=.05, p=.005.

Interactionanalysesrevealedthatthisassociationwasparticularlystrongamongparticipantswho experiencedhigherT1orT2levelsofdepressivesymptomsorperceivedstress,+1SD:bs=.34to .51,ps<.001,butnotsignificantamongtheircounterpartswhoreportedrelativelylowerlevels ofdepressivesymptomsorperceivedstress,1SD:bs=.03to11,ps>.43.

Conclusions: Declinesinself-esteemrepresentamechanismthatcontributestohigherlevelsof diurnal cortisol secretion if older adults experience psychological distress. Increases in self- esteem,bycontrast,canameliorateolderadults’cortisolregulationinstressfulcircumstances.

* Correspondingauthorat:DepartmentofPsychology,ConcordiaUniversity,7141SherbrookeStreetWest,Montreal,QCH4B1R6,Canada.

Tel.:+15148482424x2231;fax:+15148484523.

E-mailaddress:carsten.wrosch@concordia.ca(C.Wrosch).

Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-1fca4shp0dkj74

https://dx.doi.org/10.1016/j.psyneuen.2013.12.010

1. Introduction

Self-esteemisapsychologicalvariablereflectingaperson’s generalfeelings ofself-worthacross differentareasoflife (Rosenberg,1986).Researchhasidentifiedself-esteemasan adaptivepersonalitydimensionthatcanbeassociatedwith subjectivewell-being, effective biological regulation, and physicalhealth(Pruessneretal.,1999;Orthetal.,2012).

Sucheffectsofself-esteemmayoccurbecauseitameliorates the psychological consequences of difficult life circum- stances (Greenberg et al., 1992) and has adaptive down- stream implications for stress-related biological processes (Pruessner et al.,1999). Here we examine whether long- itudinalchangesinself-esteemcanalsobeassociatedwith stress-relatedalterations in theregulation ofcortisol pro- duced by the hypothalamic-pituitary-adrenocortical (HPA) axis.

1.1. Self-esteem andcortisolinthecontextof distress

Psychologicaltheoriessuggestthatsustainedperceptionsof stressandassociateddepressivesymptomscanelicitchanges inthefunctioningofindividuals’hormonalsystem(Folkman andLazarus,1986;Cohenetal.,2007).Alargenumber of studieshavesupportedthisassumptionbyshowingthatstress and negative emotions activate the HPA axis to release cortisolintothecirculation.Thisprocesshasbeendocumen- tedinlaboratorystudiesthatexaminethehormonaleffects ofinducedstress,aswellasinfieldstudiesthatobservethe biologicalconsequencesofnaturallyoccurringproblemsand emotions (Kirschbaum et al., 1993, 1995; Dickerson and Kemeny,2004;Milleretal.,2007).

The release of cortisol is thought to be an important process because it can mediate a variety of behavioral andphysiologicalresponsestostressfullifecircumstances.

On the one hand, cortisol is likely to facilitate effective behaviorsinthecontextofpressingdemands(Tayloretal., 2000).Ontheotherhand,cortisolcanhavedamagingeffects onhealth-relevantbiologicalprocesseswhenitisoverpro- ducedordysregulated.Particularly,prolongedactivationof the HPA axis may interfere with the regulation of other physiological systems, such as immune function, and has been associated with markers of systemic inflammation, physical health problems, and mortality (Sephton et al., 2000;Wroschetal.,2009;Rueggebergetal.,2012).How- ever,bluntedformsofcortisoldysregulation,mayalsopre- dicthealth-relatedproblems,perhapsasaconsequenceofa depletionof the system(Heim etal., 2000;Fries et al., 2005).

Despiteitsgeneraloccurrence,thereismuchvariabilityin the effect of stressful experiences on cortisol secretion (Kudielkaetal.,2009),whichmayinpartbeduetopsycho- logicalcharacteristicsthatareinvolvedintheadjustmentto stress (Wrosch et al., 2007; O’Donnell et al., 2008). One factorthatcouldfacilitateadjustmenttostressfulsituations relates to individual differences in self-esteem. Research suggests that self-esteem can promote effective coping (Baumeisteretal.,2003)andisassociatedwith lessthrea- teningappraisalsofproblematicsituations(Orthetal.,2009;

FordandCollins,2010).Suchbenefitsofself-esteem,inturn,

couldpreventstress-relateddisturbancesoftheHPAaxis.In supportofthisassumption,Pruessnerandcolleagues(1999) demonstrated that participants with low self-esteem secretedhigher levels of cortisolin response to astressor than their high self-esteem counterparts. Other research showed conceptually comparable findings by documenting that self-esteem modulates neuroendocrine responses to age-relatedchallenges, experiences ofshame,andthreats tothesocialself(Seemanetal.,1995;Gruenewaldetal., 2004;FordandCollins,2010).

1.2. Self-esteemchangeinolderadulthood

Thedocumentedeffectsofself-esteemonlevelsofcortisol secretionmaybecomeparticularlyimportantinolderadult- hood, when many individuals encounter increasing age- related stressors (e.g.,onset of physical disease or social networkdeclines;Lang andCarstensen,1994; Heckhausen etal.,2010)andsecreteenhancedlevelsofcortisol(Sapolsky etal.,1986;Sapolsky,1992;Lupienetal.,2005).Inaddition, age-comparativeresearchsuggeststhatlevelsofself-esteem canchange in olderadulthood.For example,there is evi- denceforanincreasingvariabilityinself-esteematprogres- sively older ages (Trzesniewski et al., 2003). Moreover, research has documented age-related declines in older adults’self-esteem,whichcouldderivefromalossofsocial rolesoranincreaseinphysicalhealthproblems(Robinsetal., 2002;Shawetal.,2010;Orthetal.,2010).Otherresearch, however, indicates that levels of self-esteem can remain fairly stable in old age or gradually increase throughout adulthood (Gove et al., 1989; Collins and Smyer, 2005;

Wagneretal.,2013),whichsuggeststhateffectiveperson- alityfunctioningcouldalsobepreservedintooldage(Haase etal.,2013).

Although this mixed pattern of findings implies that there is inconsistencyregarding the direction of change in older adults’self-esteem, it makes it likelythat self- esteemcouldchangefordifferentolderadultsindifferent directions.Further,suchinter-individualdifferencesinthe direction of change in self-esteem could play a role in determining older adults’ diurnal cortisol secretion. In particular,ifolderadultsperceivehighlevelsofstressor depressivesymptoms,longitudinaldeclinesinself-esteem couldputthematanenhancedriskofexhibitinganincrease incortisolsecretion.Themaintenanceorincreaseofself- esteem, by contrast, could buffer cortisol increases in stressfulcircumstances.

1.3. Thepresentstudy

Weanalyzedassociationsbetweenself-esteem,psychologi- cal distress, and diurnal cortisol in three waves of data, collectedoverfouryears,fromaheterogeneoussampleof community-dwelling older adults. We expected that declines, as compared to increases, in self-esteem over thefirsttwoyearsof studywould predictconcurrentand subsequent increases in participants’ diurnal cortisol volume. In addition, we hypothesized that such effects wouldbecomeparamountifolderadultsperceivehigh,as compared to low, levels of stress or depressive sympto- matology.

2. Methods

Datawerecollectedaspartofalargerlongitudinalproject withcommunity-dwelling olderadultsknown asthe‘‘Mon- treal Aging and Health Study’’ (MAHS).¹ Participants were recruited through newspaper advertisements from the greaterMontrealarea.Thepopulationofinterestwasolder adults,thustheonlyeligibilitycriteriawasaminimumageof 60years.

ThebaselineassessmentoftheMAHSincluded215parti- cipants(T1) andsubsequentwavesofthe studywerecon- ductedeverytwoyears.Thisstudyreportsdatafromthefirst three waves of the MAHS, which included 181 and 164 participants in the two-year (T2) and four-year (T3) fol- low-ups, respectively. Study attrition from T1 to T3 was attributabletodeath(n=13),refusalinstudyparticipation (n=8),lostcontact(n=13),orwithdrawalduetopersonal reasons (n=17).Ofthe164 participants atT3,17partici- pants were further excluded because they either did not providedataoncortisol(n=13)orself-esteem(n=4).²Thus, thefinalanalyticsampleconsistedof147participants.Study attritionwasnotsignificantlyassociatedwithbaselinemea- sures of thestudy variables, except for participants’ age.

Older participants weremore likely than younger partici- pants to discontinue the study over the three waves (t[129.14]=2.49,p=.01)(fordistributionofstudyvariables, seeTable1).TheConcordiaUniversityResearchEthicsBoard approvedallprocedures.

2.2. Procedure

Participantswerescheduledforstudyvisitsduringeachwave ofassessment. Ifthey wereunableto visitthelaboratory, theywereassessedintheirhomes.Afterobtaininginformed consent, participants were asked to respond to a larger questionnaire that included all reported study measures.

Ateachvisit,theywerefurtherinstructedto collectsaliva samples over the course of three non-consecutive typical days.After completionofstudymeasuresat eachvisit,all materialswerecollectedandparticipantswerecompensated with$50.

2.3. Materials

Themainstudyvariablesincludedmeasuresofparticipants’

diurnalcortisolvolume,self-esteem,perceivedstress,and depressive symptoms. To minimize the possibility of con- founding associations with the main study constructs, the

analysisincludedsociodemographicandhealth-relatedcov- ariates(i.e.,partnershipstatus,socioeconomicstatus[SES], mortalityindex,andcortisol-relatedmedicationusage).

Diurnalcortisolvolume(AUCG)wasmeasuredatallthree waves. Participants were asked to collect saliva samples (usingcottonswabsinsterileplasticcontainerscalledsaliv- ettes,Sarstedt,QuebecCity,Canada)acrossthreenon-con- secutive typical days, at specific times of the day (awakening, 30-min, 2PM, 4PM, and bedtime). They col- lected the first sample when they woke up, and were Table 1 Means, standard deviations, and frequencies of mainstudyvariables(N=147).

Constructs M(SD)or

percentage ^a

Range

Diurnalcortisolvolume(AUCG)(lognmol/lh)

T1 12.18(2.49) 6.16—18.70

T2 12.77(2.38) 5.72—24.25

T3 12.93(2.57) 5.97—19.96

Self-esteem(T1) 22.61(4.13) 12—30 Self-esteem(T2) 22.31(4.40) 9—30 Depressivesymptoms(T1) 5.74(4.32) 0—18 Depressivesymptoms(T2) 6.62(5.44) 0—23 Perceivedstress(T1) 2.44(0.65) 1—4.90 Perceivedstress(T2) 2.44(0.66) 1—4.30 Mortalityriskindex(T1) 5.66(2.35) 2—13

Age(y) 71.44(5.22) 64—90

Male(%) 49.70

Diabetes(%) 15.00

Cancer(%) 2.70

Lungorother

respiratorydisease(%)

11.60 Heartcondition(%) 18.40 BMI<25(%) 40.70 Currentsmoker(%) 11.00 Difficultybathing(%) 2.00 Difficultywalkingaround

thehome(%)

2.00 Difficultymanaging

finances(%)

2.70 Difficultydoingheavy

housework(%)

18.40 Married/livingwith

partner(T1)(%)

53.70

Socioeconomicstatus(T1) .01(.82) 1.83—2.12

Education^b 2.09(1.08) 0—4

Yearlyfamilyincome^c 1.54(1.28) 0—5 Perceivedsocialstatus 6.20(1.76) 0—10 Cortisol-relatedmedication

(T1)(%)

82.30

Notes: M,mean;SD, standarddeviation;AUC,area underthe curve.

aMandSDarepresentedforcontinuousvariables.

bEducationwasindexedas0=noeducation,1=highschool, 2=tradeorcollegiate,3=bachelors,and4=mastersordocto- rate.

cYearly familyincomewasindexedas0=lessthan$17,000, 1=upto$34,000,2=upto$51,000,3=upto$68,000,4=upto

$85,000,and5=morethan$85,000.

1NotethattheMAHSisanongoinglongitudinalstudyanddataon cortisol secretionhavebeenreportedinothermanuscripts (e.g., Wrosch et al., 2007; Jobin et al., 2013). However, none of the previouslypublishedstudies examined theeffectsof self-esteem onparticipants’cortisolsecretion.

2Missingdataforothervariableswerereplacedwiththerespec- tivesamplemeanpriortoconductingtheanalysesandwererelated toBMI(2missing),smoking(2missing),andT2depressivesymptoms (1missing).

providedwith atimerto collectthe 30-minmeasure. The researchassistantsubsequentlyremindedtheparticipantsby phonetocollecttheafternoonsamplesat 2and4PM.The participantscollectedthelastsampleofthedaythemselves, atbedtime. Time ofday was recorded forall samples. To preventcontaminationwithfoodorblood,participantswere askedtorefrainfromeatingorbrushingtheirteethbefore salivacollection.Theywereinstructedtoinsertasalivette intotheirmouthsforaperiodof30s,tocollectsaliva.The salivettes were stored in participants’ home refrigerators untiltheywerereturnedtothelaboratory.Uponcollectionof thesalivettes,sampleswerefrozenuntilcompletionofthe waveandanalyzedattheUniversityofTrier,Germany.The analysis involved the use of a time-resolved fluorescence immunoassaywith acortisol—biotin conjugate as atracer.

Cortisolanalysis fromthislaboratorytypically showsintra- assaycoefficientsofvariationthatarelessthan10%.

Allrawcortisolvalueswerelogtransformedtostabilize variance.Levelsofdailycortisolsecretionexhibitedatypical diurnalpattern.Cortisolvalueswerehighatawakening(Ms [SDs]=1.06—1.13[.19—.22]),peaked30minafterawaken- ing (Ms [SDs]=1.17—1.22 [.22—.24]), and continuously declined for the remainder of the day (2PM: Ms [SDs]=.77—.84 [.17—.18]; 4PM: Ms [SDs]=.72—.76 [.17—

.18]), with lowest cortisol output at bedtime (Ms [SDs]=.57—.60 [.17—.19]). Total diurnal cortisol volume wascalculatedbycomputingthearea-under-the-curvewith respectto ground(AUCG;Pruessner etal.,2003). Weana- lyzed AUCG because it represents a reliable measure of individuals’overallcortisoloutputacrossaday(forassocia- tionswithcortisolslopeandawakeningresponse,seeSection 4.1).AUCGwascalculatedseparatelyforeachofthethree assessmentdaysacrosswaves,basedonhoursafterawaken- ing.Becauseofpotentialcontaminationwithbloodorfood, cortisol values that were more than three SDs above the sample mean for a certain time of day were excluded.

Subsequently,we calculated AUCG onlyifparticipants had atleastfour offivepossiblecortisolvalues foragivenday (1300outof1323potentialdays;98.26%).Ondayswherea single cortisol value was missing, cortisol values were replacedbytherespectivesamplemeanbeforeAUCGcalcu- lation (for the 1300 days, 1.83% of cortisol values were replaced). The 30-min samples were excluded from the calculation of AUCG because the awakening response has beenshowntoberelativelyindependentfromotheraspects ofthediurnalcortisolrhythm(Pruessneretal.,2003;Chida andSteptoe,2009).ChangescoresforAUCGfromT1toT2 (andfrom T2to T3) wereobtainedin separateregression analysesbypredictingT2levelsofAUCGfromT1AUCG(and T3levelsofAUCGfromT2AUCG),andsavingthestandardized residualsforfurtheranalysis.

Self-esteemwasmeasuredatT1andT2byadministering theRosenbergself-esteemscale(Rosenberg,1986),whichis a10-itemself-reportquestionnaireusing4-pointLikert-type scales(strongly disagree=0 tostrongly agree=3).Sample itemsincludestatementssuchas‘‘IfeelthatIhaveanumber ofgoodqualities’’or‘‘Allinall,IaminclinedtofeelthatIam a failure.’’ Indicators of participants’ self-esteem were obtainedatT1andT2bycomputingasumscoreofthe10 items,afterreversecodingofnegativelyformulateditems (as=.79and.82).Individualdifferencesinchange ofself- esteemfromT1toT2wereobtainedinaregressionanalyses,

predictingT2self-esteemscoresfromT1self-esteemscores, andsavingthestandardizedresidualsforfurtheranalysis.

Depressive symptoms were measured at T1 and T2.

Participantsresponded to a 10-item versionofthe Center forEpidemiologicStudiesDepressionScale(CES-D;Andresen etal.,1994).Theywereaskedtoratehowfrequentlythey experienced10depressivesymptomsduringthepastweek, using 4-point Likert-type scales (rarely or none of the time=0to mostorallof thetime=3).Itemsincluded,‘‘I could not get going’’ and ‘‘I was bothered by thingsthat usuallydon’tbotherme.’’Scalescoresfordepressivesymp- tomswereobtainedatT1andT2bycomputingthesumofthe 10items(as=.72and.82).

PerceivedstresswasmeasuredatT1andT2.Participants wereasked to respond to the 10-item version ofthe Per- ceived Stress Scale (Cohen et al., 1983). They rated how frequentlytheyexperienced10differentsituationsoverthe pastmonthbyusing5-pointLikert-typescales(never=1to veryoften=5). Itemsincluded,‘‘How often haveyou felt thatthingsweregoingyourway?’’and‘‘Howoftenhaveyou feltnervousandstressed?’’Positivelyformulateditemswere reversedcodedandindicatorsofperceivedstressatT1and T2wereobtainedbyaveraging theratingsof the10items (as=.87and.87).

Covariates weremeasured atbaseline (see Table1). A previously validatedrisk index forpredicting olderadults’

mortalitywascomputedbycountingparticipants’weighted riskfactors:age,beingmale,presenceofdiabetes,cancer, lungorotherrespiratorydisease,heartcondition,bodymass index<25,smokingandfunctionalaspectsofagingsuchas, bathing,walkingaroundthehome,managingfinances,and heavyhousework(forfurtherdetails,seeLeeetal.,2006).

SES was indexed by averaging the standardized scores of participants’ reportedannual familyincome, highestlevel of education, and perceived social status (rs=.40—.56, ps<.001).Self-reported partnership statuswas measured by categorizing participants into two groups: (1) single/

separated/widowed or(2) married/liveswith partner.The useof medicationthat could affectcortisol secretion was assessed by counting thenumber ofdifferent medications participants reported taking.Subsequently, avariable was computedindicatingwhetherornotparticipantstookmed- ications that could influence HPA axis activity(e.g., anti- depressants,beta-blockers,oranti-inflammatorydrugs).

2.4. Dataanalyses

Preliminaryanalyseswereconductedtodescribethesample (by calculating means), explore associations between the mainconstructs (bycalculatingcorrelations),andexamine meanleveldifferences overtime (bycalculatingANOVAs).

Thehypothesesweresubsequentlytestedusingstandardized predictorvariables inhierarchicallinear regressionmodels thatcontrolled for relevant sociodemographicandhealth- related covariates. The effects of self-esteem change, depressive symptoms, and perceived stress on changes in diurnalcortisolvolumeweretestedbyconductingtwosepa- rateregression analyses,using change scoresof AUCG: (1) fromT1toT2,and(2)fromT2toT3asdependentvariables.

In thefirst step ofthe analyses,the main effectsof self- esteemchange(T1toT2),levelsofdepressivesymptomsand perceivedstress(T1andT2),andthecovariates(partnership

status, SES, mortality index, and cortisol-related medica- tion) weretested for significance.The second step of the analysesexaminedseparatelywhethertheinteractionterms ofself-esteemchangewithlevelsof(1)T1depressivesymp- toms,(2)T2depressivesymptoms, (3)T1perceivedstress and(4)T2perceivedstresswouldpredictadditionalvariance inthedependentvariables.Becauseourinteractionanalyses involvedtestingmultipleeffectsoffourdifferentindicators ofpsychologicaldistress,weappliedaBonferronicorrection to the significance levels of the interactions (p<.0125).

Significantinteractioneffectswerefollowedupwithsimple slope analyses, examining the associations between self- esteem change and the outcome variables one standard deviationaboveandbelowthesamplemeanof depressive symptomsandperceivedstress.

3. Results

3.1. Preliminaryanalyses

The sample characteristics are presented in Table 1. At baseline,participantswereonaverage71yearsold,approxi- mately half of the sample was married or living with a partner,andhalf ofthesample werewomen.Thepartici- pants hadon average, a collegialor trade education, and

$34,000—$51,000yearlyincomes,indicatingthatthesample was of moderate SES. The majority of the sample used medicationthatcould influenceHPAaxisactivity.Between 3% and 18% of participants reported that they either had diabetes,cancer,respiratory disease,oraheartcondition.

Furthermore, approximately40%of theparticipants hada body mass index (BMI) of less than 25. A minority of the samplewassmokingandbetween2%and18%ofthesample had different functional limitations. Taken together, the mortality index rateforthe samplehadan averagerating of 6.10, which compared to Lee and colleagues’ (2006) validationsample,wouldcorrespondtoa9%mortalityrisk overfouryears.Thesociodemographicandhealthcharacter- istics of the sample were within the normative range of known distributions among older adults residing at home (NationalAdvisoryCouncilonAging[NACA],2006).

The zero-order correlations between the main study variablesarepresentedinTable2.Thesignificantassocia- tionsshowedpositivecorrelationsacrosswavesforcortisol volume,self-esteem,depressivesymptoms,andperceived stress, indicating some stability in these variables over time. Moreover, T1—T2 increases in cortisol were asso- ciated with higher T2 and T3 levels of cortisol volume, and T2—T3 increases in cortisol were associated with higher T3 levels of cortisol volume and lower T2 levels andT1—T2 reductionsofself-esteem. Inaddition,T1—T2 increasesinself-esteemwereassociatedwithlowerT1and T3 levels of cortisol as well as higher T2 levels of self- esteem and lower T2 levels of depressive symptoms and perceived stress. Finally, T1 and T2 levels of depressive symptomsandperceivedstresswerepositivelyassociated witheachother,aswellaswithlowerlevelsofself-esteem atT1andT2.

ANOVAs showed that cortisol volume significantly increased from T1 to T3, F(1, 146)=9.13, p=.003 (see Table1).Meanlevelsofdepressivesymptomsalsoincreased from T1toT2, F(1, 146)=5.20,p=.024,while levels of self-esteem and perceived stress did not significantly change in the entire sample from T1 to T2, Fs<1.04, ps>.31.

3.2. Mainanalyses

The results ofthe first regression analysis are reported in Table 3, predicting concurrent changes in diurnal cortisol volume(T1toT2).Inthefirststepoftheanalysis,themain effectsofself-esteemchange(T1toT2),levelsofdepressive symptoms andperceivedstress(T1and T2),or anyof the incorporated covariates were not significantly associated with changes in AUCG from T1to T2,Fs<1.03,ps>.31.

Inaddition,thesecondstepoftheanalysisshowedthatthe four interaction terms between self-esteem change with depressive symptoms andperceived stress (at T1and T2) didnotpredictsignificantchanges inAUCGfrom T1toT2, Fs<1.46,ps>.23.

Theresultsofthesecondanalysis,predictingsubsequent changes indiurnalcortisol volume fromT2to T3,arealso

Table2 Zero-ordercorrelationsbetweenmainstudyvariables(N=147).

1 2 3 4 5 6 7 8 9 10 11

1.DiurnalcortisolvolumeAUCG(T1)

2.DiurnalcortisolvolumeAUC_G(T2) .37^**

3.DiurnalcortisolvolumeAUCG(T3) .30^** .43^**

4.DDiurnalcortisolvolumeAUCG(T1toT2) .00 .93^** .34^**

5.DDiurnalcortisolvolumeAUCG(T2toT3) .15 .00 .90^** .06

6.Self-esteem(T1) .01 .08 .02 .09 .01

7.Self-esteem(T2) .13 .00 .15 .06 .17^* .66^**

8.DSelf-esteem(T1toT2) .16^* .06 .22^** .00 .21^** .00 .75^**

9.Depressivesymptoms(T1) .07 .09 .02 .07 .02 .57^** .46^** .11 10.Depressivesymptoms(T2) .01 .00 .02 .00 .03 .35^** .46^** .31^** .56^**

11.Perceivedstress(T1) .03 .03 .07 .02 .06 .49^** .39^** .08 .71^** .54^**

12.Perceivedstress(T2) .06 .07 .02 .05 .05 .38^** .45^** .26^** .51^** .69^** .63^**

*p<.05.

**p<.01.

reportedinTable3.Thefirststepoftheanalysisshowedthat thecovariatesandthemaineffectsofdepressivesymptoms andperceivedstresswerenotsignificantly associatedwith changeinAUCGfromT2toT3,Fs<3.75,ps>.05.However, changeinself-esteemfromT1toT2significantlypredicted changeinAUCGfromT2toT3,F=8.09,p=.005.Thenega- tivesignoftheregression coefficientdemonstratesthatto the extent participants experienced a steeper decline in their self-esteem over the first two years of study, they exhibited larger increases in diurnal cortisol volume over thesubsequenttwoyears(seeTable3).

Thesecondstepoftheanalysisshowedsignificantinter- actioneffectsinpredictingchange inAUCGfromT2 toT3 betweenself-esteemchange(T1toT2)with(1)T1levelsof depressive symptoms, F=6.68, p=.011, (2) T2 levels of depressivesymptoms, F=10.01, p=.002, (3) T1 levels of perceived stress,F=11.32, p=.001,and (4) T2 levels of perceivedstress,F=10.83,p=.001.³

Toinvestigatethesignificantinteractioneffects,wecal- culated the simple slopes for the associations between declines in self-esteem from T1 to T2 and subsequent

increases in AUCG (T2 to T3), separately for participants who scored one standard deviation above and below the sample meansof depressive symptomatologyor perceived stressatT1orT2.Theobtainedresultsare documentedin Table4andshowedthatself-esteemdeclinesoverthefirst twoyearsofstudyweresignificantlyassociatedwithsubse- quentincreases inAUCGamong participants who reported highT1orT2levelsofdepressivesymptoms,orhighT1orT2 levels of perceived stress. By contrast, declines in self- esteemwerestatisticallyunrelatedtosubsequentincreases inAUCGamongparticipantswhoreportedlowT1orT2levels ofdepressivesymptoms,orlowT1orT2levelsofperceived stress.

Because of the similarity of the observed interaction effects, werepeated thesecond regressionanalysis, using a psychological distress composite of averaged depressive symptomsandperceivedstressacrossthefirsttwowavesasa predictorvariable(insteadofthefourseparatemeasuresof distress).Weconductedthissupplemental analysisto esti- mate the most reliable association between self-esteem changeandsubsequentcortisolincreaseamongparticipants whoreported high,average,orlowlevels ofpsychological distress.Inthefirststepofthisanalysis,onlyhighlevelsof the mortality index, F(1, 140)=4.25, R²=.03, b=.18, p=.041, and declines in self-esteem from T1 to T2,F(1, 140)=7.92, R²=.05,b=.24, p=.006, weresignificantly associated with subsequent increases in AUCG. Moreover, similar to the previous analysis, the second step of the analysisshowedthatself-esteemchangesignificantlyinter- actedwith thepsychological distresscomposite topredict changeinAUCGfromT2to T3,F(1,139)=15.37,R²=.09, b=.34,p<.001. Fig. 1illustrates theobtainedassocia- tionsbetweenself-esteemchange(T1toT2)andsubsequent increaseinAUCGforparticipantswithlow(1SD),average Table3 HierarchicalregressionanalysespredictingchangesindiurnalcortisolfromT1toT2,andfromT2toT3,bychangesin self-esteemfromT1toT2andT1andT2levelsofperceivedstressanddepressivesymptoms(N=147).

Predictors DDiurnalcortisolvolume

AUCG(T1toT2)

DDiurnalcortisolvolume AUCG(T2toT3)

R² b R² b

Maineffects

Married/livingwithpartner(T1) .00 .06 .00 .06

Socioeconomicstatus(T1) .01 .09 .00 .06

Cortisolrelatedmedication(T1) .00 .03 .01 .11

Mortalityindex(T1) .00 .01 .02 .17

Depressivesymptoms(CES-D)(T1) .00 .07 .01 .13

Depressivesymptoms(CES-D)(T2) .00 .00 .00 .03

Perceivedstress(PS)(T1) .00 .04 .00 .07

Perceivedstress(PS)(T2) .01 .13 .01 .14

DSelf-esteem(T1toT2) .00 .00 .05^** .25^**

Interactions

DSelf-esteemCES-D(T1) .00 .02 .04^* .22^*

DSelf-esteemCES-D(T2) .01 .09 .06^** .28^**

DSelf-esteemPS(T1) .01 .11 .07^** .28^**

DSelf-esteemPS(T2) .01 .09 .07^** .29^**

Notes:R²valuesrepresenttheuniqueproportionofvarianceexplainedineachstepoftheanalyses.brepresentsstandardizedregression coefficientsineachstepofanalyses.Dfsformaineffects=1,137;dfsforinteractions=1,136.

* p<.05.

** p<.01.

3Weobtainedthesamepatternofsignificantfindingsifwedidnot use change scores in our analyses, but instead operationalized change by using levels of predictor and outcome variables and controllingtheanalysesforpreviouslevelsoftheseconstructs.This patternalsoremainedstable ifwe includedT3measures ofself- esteem,depressivesymptom,andperceivedstress(andT1levelsof cortisolforpredictingcortisolchangefromT2toT3)asadditional covariatesintotheanalyses.Finally,follow-upanalysesshowedthat themaineffectsofbaselineself-esteem,andinteractionsincluding baselineself-esteemwithdepressivesymptomsorperceivedstress, didnotpredictchangesincortisolvolumefromT1toT2orT2toT3.

(M), and high (+1SD) levels of the psychological distress composite. Simple slope analyses demonstrated that the associationbetweenself-esteemdeclinesandsubsequently enhanced cortisol volume significantly increased to the extent that participants experienced higher levels of the psychological distress composite (+1SD: b=.41, p<.001;M:b=.10,p=.25;1SD:b=.21,p=.14).

4. Discussion

The results from this study suggest that changes in self- esteem are associated with older adults’ diurnal cortisol secretion. Although our study showed a net stability of self-esteem levels overthe first two years of study, there was considerablevariability in self-esteem over time and individualdifferencesinself-esteemchangewereasignifi- cantpredictorofalterationsinparticipants’cortisoloutput.

Specifically,olderadultswhoexperiencedadeclineinself- esteem overthefirst twoyears ofstudyexhibitedsteeper increasesindiurnalcortisolvolumeoverthesubsequenttwo years,ascomparedtoparticipantswhoreportedincreasesin self-esteem.Moreover,thisassociationwasenhancedamong olderadultswhoperceivedhighlevels ofdepressivesymp- tomsorperceivedstressatbaselineortwo-yearfollow-up, butabsentamongtheircounterpartswithcomparablylower levels of perceived stress or depressive symptoms. This

patternofresultswassignificantaftercontrollingforpoten- tialconfounds,suchasSES,partnershipstatus,mortalityrisk factors,orcortisol-relatedmedication.

Thesefindingssuggestthatincreasesinself-esteemcan protect older adults from exhibiting distress-related increases in diurnal cortisol secretion. By contrast, older adults who experience adecline in theirself-esteem may be morelikely to exhibitelevated cortisol outputin such circumstances. We think that such a process may occur because self-esteem can facilitate adaptive coping with stress(Baumeisteretal.,2003).Inthecontextofage-related stressors, an increasein self-esteem could likely result in more positive appraisals of challenging life circumstances and through this process buffer stressful experiences and subsequentincreasesincortisolsecretion.Participantswho encounteradeclineintheirself-esteem,however,maybe more likelyto appraise challengesas threats (Orthetal., 2009)andthus,exhibitincreasesindiurnalcortisolsecretion.

Notethatourresultsshowedthatchangesinself-esteem wereassociatedonlywithsubsequent,butnotconcurrent, changes in diurnal cortisol secretion. Although we were surprised by the absenceof a concurrent association, one potential explanation for this finding may relate to the substantial time gaps between study assessments (i.e., two years). Giventhat cortisol changemay haveoccurred atanytimeduringthetwo-yearinterval,itispossiblethat Table4 Associationsbetweenself-esteemchange(T1toT2)andsubsequentchangesindiurnalcortisolvolumeAUCG(T2toT3) forparticipantswithhigh(+1SD)andlow(1SD)levelsofdepressivesymptomsandperceivedstressatT1andT2(N=147).

DDiurnalcortisolvolumeAUC_G(T2toT3)

Depressivesymptoms Perceivedstress

T1 T2 T1 T2

High Low High Low High Low High Low

DSelf-esteem(T1toT2) .49^** .03 .34^** .10 .51^** .11 .35^** .11

**p<.01.

Figure1 Associationsbetweenself-esteemchanges(T1toT2)andsubsequentchangesindiurnalcortisolvolumeAUCG(T2toT3), separatelyforparticipantswhoexperiencedlow(1SD),average(M),andhigh(+1SD)levelsofpsychologicaldistress(averagedT1 andT2scoresofdepressivesymptomsandperceivedstress).Standardizedcoefficients(b)areindicatedforeachgroup.**p<.001.

increasesincortisolhaveprecededchangesinself-esteem, weredrivenbyfactorsotherthanself-esteem,andtherefore notconcurrentlyassociatedwithself-esteemchanges.How- ever, theprediction of subsequent cortisol changes inour second analysis ensured that declines in self-esteem occurredbeforetheobservedincreasesincortisol,demon- strating a directional association between self-esteem changeandcortisolvolume.

Oursupplementalanalysesfurthershowedthatbaseline levels of self-esteem did not significantly contribute to increasesincortisolsecretion(seeFootnote3).Consistent withotherwork,thisresultmayimplythatdeviationsfrom individuals’ levels of self-esteem are more impactful in predicting stress-related biological consequences than levels of self-esteem alone (Ross et al., 2013). Although more research is needed to substantiate this conclusion, suchpatternsmayoccurifcortisolsecretionhabituatesover timetoindividuals’typicalperceptionsaboutthemselvesor theirlives(Jobinetal.,2013;Wu¨stetal.,2005;Milleretal., 2007).Insuchcases,deviationsfromaccustomedlevelsof self-esteem could exert a more reliableeffect on indivi- duals’biologicalstress responses.Further,such aprocess maybe observable particularlyin older adulthood, when variabilityin self-esteem tends to increase (Trzesniewski etal.,2003).

Theresultsfromthepresentstudyhaveimportantimpli- cations for theory and research on stress-related distur- bances of cortisol secretion. First, they extend previous research examining the protective functions of levels of self-esteem in the stress-related cortisol link (Pruessner etal.,1999).Tothisend,ourresultssuggestthatanincrease ofself-esteemovertimecanalsobeanadaptiveandinde- pendentcontributor to olderadults’ HPAaxis functioning.

Second,theyshedlightonsomeoftheinconsistenciesinthe extant literature on the associations between distressing experiencesandcortisol secretion.Similar to ourfindings, maineffectsofpsychologicaldistressoncortisoldisturbances havenotbeenreportedconsistentlyacrossstudies(Kudielka etal.,2009). However,certain characteristics thatenable individualstocopewithdistressingexperiencesmayobstruct such an association. Thus, a link between psychological distress and cortisol output may be observed particularly amongindividuals who have difficulty adjusting to critical life circumstances (Wrosch etal., 2007). Our findings are consistent with this conclusion by suggesting that to the extent participants experienced a greater decline in self- esteem, psychological distress became increasingly asso- ciatedwithsubsequentlyenhancedlevelsofcortisolsecre- tion(seeFig.1).

However, wenotethattheflip sideof thelattereffect suggeststhat distressbecame increasinglyassociated with fewer increases in cortisol among participants who had experienced an increase in self-esteem (see Fig. 1). One potential explanation of this effect is that self-esteem increases could, under some circumstances, also enhance cortisoloutput.Suchanassociationmayoccurifself-esteem activates maladaptive behaviors tendencies, a possibility thathasbeen indicatedinprevious research(e.g.,Narcis- sism, Neff, 2011). Alternatively, effects of distress on declinesincortisoloutputamongparticipantswhoincreased inself-esteem could bepartially related to thepossibility that these participants wereexhausted because they had

experiencedaperiodofhighdistressandlowself-esteemat baseline(cf.Topsetal.,2008).

Finally,the study’s findings contributeto the emerging literatureonself-esteemchangeinolderadulthood(Robins etal.,2002;CollinsandSmyer,2005;Shawetal.,2010;Orth etal.,2010).Whileourstudycannotprovideafirmanswerto thequestionofwhetherornotself-esteemdeclinesinold age,it points to the conclusion thatthere is considerable variabilityinolderadults’self-esteemovertime.Moreover,it demonstratesthatsuchvariabilityinpersonalityfunctioning representsmeaningfulpsychologicalchangesthatrelateto trajectoriesofahormonethathaswide-rangingregulatory influencesinthebody(Weiner,1992;Lupienetal.,2009).

4.1. Limitationsandfutureresearch

Therearelimitationstothepresentstudy.First,ouranalyses were focused on predicting AUCG of cortisol because it represents a reliable indicator of overall cortisol volume acrosstheday.However,otherresearchhasstudiedtheslope ofcortisolfromawakeningtobedtime(Sephtonetal.,2000) orthecortisolawakeningresponse(CAR;Vrshek-Schallhorn etal.,2013).Supplementalanalysesofourdatashowedthat increasedself-esteemwasalsoassociatedwithsubsequently (but not concurrently) more normative (i.e., declining) changes in cortisol slope, F(1, 137)=6.45, b=.23, p=.010.However,therewerenomaineffectsofself-esteem changeonCAR,anddepressivesymptomsorperceivedstress did not moderate the associations between self-esteem changeandcortisolslopeorCAR,Fs<2.23,ps>.14.While thispatternlendssomefurthersupporttoourconclusionthat changeinself-esteemisanimportantpersonalityprocessin oldage,italsosuggeststhatcortisolslopeandCARmaybe lesssensitivetodifferencesinself-esteemanddistressthan AUCGofcortisol.⁴

Second,although weused a mortality indexas a parsi- moniouscovariate,thismeasuredidnotaddresstheassocia- tionswith the single variables ofthe index. Supplemental correlation analyses,linking theseparate variables ofthe mortalityindexwithT1—T2andT2—T3changesincortisol, showedthatnoneofthesinglevariablesweresignificantly associated with cortisol change, all jrjs<.17, all ps>05, exceptforsex. Inparticular,menexhibitedlarger cortisol increasesthanwomenfromT2toT3,r=.17,p=.038.This resultisconsistentwithsomepreviousstudies(Kirschbaum etal.,1992),andfutureresearchmayidentifythevariables thatcouldunderliesex-specifictrajectoriesofcortisolsecre- tionamongoldermenandwomen.

Third,whilethereportedresultssuggestthatchangesin self-esteemprecededchangesincortisoloutput,ourstudyis

4NotapplyingaBonferronicorrection,thesupplementalanalyses wouldhaveshownasignificantinteractioneffect(i.e.,T1—T2self- esteemchangeT2perceivedstress)inpredictingchangeincorti- solslopefromT1toT2,F(1,136)=4.84,R²=.03,p=.029.Although this interaction wasnot found for thethree other indicators of distress,andthusmaybeattributabletochance,wenotethatits pattern indicated a concurrent association between self-esteem increaseandincreasinglyflattercortisolslopesamongparticipants withhigh,b=.19,p=.048,butnotlow,b=.13,p=.37,T2levels ofperceivedstress.

based on a naturalistic design andtherefore cannot draw causal inferences regarding the observed associations. In addition,ourdatastemfromarelativelysmalllongitudinal project,whichlimitsthegeneralizabilityofthestudy’scon- clusions.Thus,futureresearchshouldreplicatethereported findings in larger andrepresentative studies.Such studies shouldalsoexaminechangesinotherpersonalityconstructs (e.g.,coping tendencies,optimism,or broadertraits) and biological processes (e.g., inflammatory cytokines) that couldinfluenceanumberofage-relateddiseases.Giventhat cortisolsecretioncouldinfluenceimmunefunctionandphy- sicalhealth(Bjo¨rntorpandRosmond,1999;Sapolskyetal., 2000;Lupienetal.,2009;Rueggebergetal.,2012),research alongthese linesmayrevealhowadaptivechanges inper- sonality functioning can protect quality of life in older adulthood.

5. Conclusion

Theresultsfromthisstudyidentifydeclinesinself-esteem asamechanismthatmaycontributetoelevatedcortisol volumeamongolderadultswhoexperiencepsychological distress.Increasesinself-esteem,bycontrast,arelikelyto ameliorate older adults’ cortisol regulation in stressful circumstances.Thesefindingsmaybeusedininterventions that target self-esteem to improve older adults’ quality oflife.

Conflict of interest

Noneoftheauthorshaveaconflictofinteresttodeclare.

Role of the funding source

ThisstudywasfundedbygrantsfromCanadianInstitutesof HealthResearchawardedtoCarstenWrosch.Itwasfurther presentedatthe71stAnnualScientificMeetingoftheAmer- icanPsychosomaticSocietyinthecontextofSarahY.Liu’s receptionoftheYoungScholarAward.

References

AgingNACA,2006.SeniorsinCanada:2006ReportCard.National AdvisoryCouncilonAging.

Andresen,E.M.,Malmgren,J.A.,Carter,W.B.,Patrick,D.L.,1994.

Screeningfordepressioninwellolderadults:evaluationofashort formoftheCES-D.Am.J.Prev.Med.10,77—84.

Baumeister,R.F., Campbell,J.D.,Krueger,J.I., Vohs,K.D.,2003.

Doeshighself-esteemcausebetterperformance,interpersonal success,happiness,orhealthierlifestyles? Psychol.Sci.Publ.

Interest4,1—44.

Bjo¨rntorp, P.,Rosmond,R.,1999.Hypothalamicoriginofthemeta- bolicsyndromeX.Ann.N.Y.Acad.Sci.892,297—307.

Chida,Y.,Steptoe,A.,2009.Cortisolawakeningresponseandpsy- chosocialfactors:asystematicreviewandmeta-analysis.Biol.

Psychol.80,265—278.

Cohen,S.,Kamarck,T.,Mermelstein,R.,1983.Aglobalmeasureof perceivedstress.J.HealthSoc.Behav.24,385—396.

Cohen, S., Janicki-Deverts, D., Miller, G.E., 2007. Psychological stressanddisease.J.Am.Med.Assoc.298,1685—1687.

Collins,A.L.,Smyer,M.A.,2005.Theresilienceofself-esteeminlate adulthood.J.AgingHealth17,471—489.

Dickerson, S.S., Kemeny, M.E.,2004.Acute stressorsandcortisol responses:atheoreticalintegrationandsynthesisoflaboratory research.Psychol.Bull.130,355—391.

Folkman,S., Lazarus,R.S.,1986.Stressprocesses anddepressive symptomatology.J.Abnorm.Psychol.95,107.

Ford,M.B.,Collins,N.L.,2010.Self-esteemmoderatesneuroendo- crineandpsychologicalresponsestointerpersonalrejection.J.

Pers.Soc.Psychol.98,405—419.

Fries,E.,Hesse,J.,Hellhammer,J.,Hellhammer,D.H.,2005.Anew view onhypocortisolism.Psychoneuroendocrinology30,1010—

1016.

Gove,W.R.,Ortega,S.T.,Style,C.B.,1989.Thematurationaland roleperspectivesonagingandselfthroughtheadultyears:an empiricalevaluation.Am.J.Sociol.94,1117—1145.

Greenberg,J.,Solomon,S.,Pyszczynski,T.,Rosenblatt,A.,Burling, J.,Lyon,D.,Simon,L.,Pinel,E.,1992.Whydopeopleneedself- esteem?Convergingevidencethatself-esteemservesananxiety- bufferingfunction. J.Pers.Soc.Psychol.63,913—922.

Gruenewald,T.L.,Kemeny,M.E.,Aziz,N.,Fahey,J.L.,2004.Acute threattothesocialself:shame,socialself-esteem,andcortisol activity.Psychosom.Med.66,915—924.

Haase, C.M., Heckhausen, J., Wrosch, C., 2013. Developmental regulation across thelife span: toward a newsynthesis.Dev.

Psychol.49,964—972.

Heim,C.,Ehlert,U.,Hellhammer,D.H.,2000.Thepotentialroleof hypocortisolism inthepathophysiologyofstress-relatedbodily disorders.Psychoneuroendocrinology25,1—35.

Heckhausen,J.,Wrosch,C.,Schulz,R.,2010.Amotivationaltheory oflife-spandevelopment.Psychol.Rev.117,32—60.

Jobin, J., Wrosch, C., Scheier,M.F., 2013. Associations between dispositionaloptimismanddiurnalcortisolinacommunitysam- ple: when stress is perceived as higher than normal. Health Psychol.,http://dx.doi.org/10.1037/a0032736.

Kirschbaum, C., Pirke, K., Hellhammer, D.H.,1993. The ‘‘trier social stress test’’: a tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology 28,76—81.

Kirschbaum,C.,Pru¨ssner,J.C.,Stone,A.A.,Federenko,I.,Gaab,J., Lintz,D.,Schommer,N.,Hellhammer,D.H.,1995.Persistenthigh cortisolresponsestorepeatedpsychologicalstressinasubpopu- lationofhealthymen.Psychosom.Med.57,468—474.

Kirschbaum, C., Wu¨st, S., Hellhammer, D., 1992. Consistent sex differencesincortisolresponsestopsychologicalstress.Psycho- som.Med.54,648—657.

Kudielka, B.M., Hellhammer, D.H., Wu¨st, S., 2009. Why do we respondsodifferently?Reviewingdeterminantsofhumansalivary cortisolresponsestochallenge. Psychoneuroendocrinology34, 2—18.

Lang,F.R.,Carstensen,L.L.,1994.Closeemotionalrelationshipsin latelife:furthersupportforproactiveaginginthesocialdomain.

Psychol.Aging9,315—324.

Lee,S.J.,Lindquist,K.,Segal,M.R.,Covinsky,K.E.,2006.Develop- mentandvalidationofaprognosticindexfor4-yearmortalityin olderadults.J.Am.Med.Assoc.295,801—808.

Lupien,S.J.,Fiocco,A.,Wan,N.,Maheu,F.,Lord,C.,Schramek,T., Tu, M.T., 2005. Stress hormonesand human memory function acrossthelifespan.Psychoneuroendocrinology30,225—242.

Lupien,S.J.,McEwen,B.S.,Gunnar,M.R.,Heim,C.,2009.Effectsof stressthroughoutthelifespanonthebrainbehaviourandcogni- tion.Nat.Rev.Neurosci.10,434—445.

Miller,G.E.,Chen,E.,Zhou,E.S.,2007.Ifitgoesup,mustitcome down?Chronicstressandthehypothalamic-pituitary-adrenocor- ticalaxisinhumans. Psychol.Bull.133,25—45.

Neff,K.D.,2011.Self-compassion,self-esteem,andwell-being.Soc.

Personal.Psychol.Compass5,1—12.

O’Donnell,K.,Badrick,E.,Kumari,M.,Steptoe,A.,2008.Psycho- logicalcopingstylesandcortisoloverthedayinhealthyolder adults.Psychoneuroendocrinology33,601—611.