maternal postnatal depression and the development of depression in offspring up to 16 years of age
TRANSCRIPT
NEW RESEARCH
Maternal Postnatal Depressionand the Development of Depressionin Offspring Up to 16 Years of Age
Lynne Murray, Ph.D., Adriane Arteche, Ph.D., Pasco Fearon, Ph.D., D.Clin.Psy.,Sarah Halligan, D.Phil., Ian Goodyer, M.D., Peter Cooper, D.Phil., D.Clin.Psych.
Objective: The aim of this study was to determine the developmental risk pathway todepression by 16 years in offspring of postnatally depressed mothers. Method: This was aprospective longitudinal study of offspring of postnatally depressed and nondepressedmothers; child and family assessments were made from infancy to 16 years. A total of 702mothers were screened, and probable cases interviewed. In all, 58 depressed mothers (95% ofidentified cases) and 42 nondepressed controls were recruited. A total of 93% were assessedthrough to 16-year follow-up. The main study outcome was offspring lifetime clinicaldepression (major depression episode and dysthymia) by 16 years, assessed via interview at 8,13, and 16 years. It was analysed in relation to postnatal depression, repeated measures ofchild vulnerability (insecure infant attachment and lower childhood resilience), and familyadversity. Results: Children of index mothers were more likely than controls to experiencedepression by 16 years (41.5% versus 12.5%; odds ratio � 4.99; 95% confidence interval �1.68–14.70). Lower childhood resilience predicted adolescent depression, and insecure infantattachment influenced adolescent depression via lower resilience (model R2 � 31%). Familyadversity added further to offspring risk (expanded model R2 � 43%). Conclusions: Offspringof postnatally depressed mothers are at increased risk for depression by 16 years of age. Thismay be partially explained by within child vulnerability established in infancy and the earlyyears, and by exposure to family adversity. Routine screening for postnatal depression, andparenting support for postnatally depressed mothers, might reduce offspring developmentalrisks for clinical depression in childhood and adolescence. J. Am. Acad. Child Adolesc.Psychiatry, 2011;50(5):460–470. Key Words: maternal depression, adolescent depression,attachment, resilience, adversity
iiktsefmt
A t least one-third of people experience amajor depressive episode during theirlifetime,1 and for many individuals the
experience is persistent.2 Understanding the de-velopment of depression is, therefore, an impor-tant public health issue. This is especially truewhen first onset occurs in the school-age years, assuch episodes are associated with particularlypoor outcome in terms of severity, chronicity,and recurrence.3-5 The association between life-time depression in mothers and offspring affec-tive disorders is strong.6,7 Studying the develop-
This article is discussed in an editorial by Dr. David Reiss on page
s431.
JOURN
460 www.jaacap.org
ment of children of depressed parents mighttherefore help to elucidate the mechanisms in-volved in transmission of disorder. Recent twinand adoption studies indicate that the associationbetween parent depression and depression injuvenile offspring is mediated predominantly byenvironmental mechanisms,8,9 and suggest anmportant role for family adversity (e.g., parent-ng difficulties, marital conflict).6 Nevertheless,ey questions remain. One question concernsiming of offspring exposure to maternal depres-ion and, specifically, whether risk might bespecially raised when exposure occurs in in-ancy, when dependency on the mother is maxi-
al. The second question is the need to identifyhe psychological pathways to disorder in off-
pring, especially those established early in de-AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
PND AND OFFSPRING DEPRESSION
velopment. Here, we examine these issues in alongitudinal study of the development of chil-dren of postnatally depressed mothers.
With regard to timing, three previous studieshave found exposure before 2 years of age to beassociated with increased offspring depressionby adolescence10-12; however, whereas onestudy11 showed this to be sufficient albeit notnecessary, the other two studies found overallduration of maternal depression to account forthe association between early maternal depres-sion and adolescent disorder.
With regard to psychological pathways to de-pression, although none of the three previousreports examined early child psychological func-tioning, wider research suggests that insecure at-tachment to the mother in infancy, well-establishedas associated with PND,13 might confer initialpsychological vulnerability. This is because at-tachment insecurity has been found to predict aconstellation of cognitive, affective and behav-ioral processes in school-aged children, termedlow “ego resiliency.”14,15 This constellation ischaracterised by inflexibility in the face of chang-ing and stressful circumstances, and difficulty inrecovering from challenge or failure.16 Notably, itshares core characteristics with a profile of cog-nitive vulnerability for depression (e.g., negativeaffect, feelings of unworthiness, low self-aspira-tions), as well as with temperamental dimensionsconsidered to raise risk for depressive disorder(e.g., negative emotionality, failure to persistwith endeavor, and low effortful control).17 Todate, however, although infant insecurity hasbeen found to predict clinical depression byage 17 years,18,19 this intervening developmen-tal pathway through which depression mightultimately emerge has not been examined.
Importantly, and consistent with evidence onfamily risk factors for offspring depression,6 lon-gitudinal research has shown that poor childoutcome following infant insecure attachment issubstantially more likely when the child is ex-posed to further family problems.19,20 BecausePND is commonly associated with family adver-sity, this suggests that any child psychologicalvulnerability for depression associated with PNDis similarly likely to be increased by ongoingenvironmental adversities. Studying longer-termadversity effects is not only important for im-proving the prediction of child outcome but alsofor clarifying the clinical significance of the post-
natal episode. As noted above,10-12 previous re-JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
search has suggested that the association be-tween PND and adolescent depression mayactually be mediated by the effects of later ma-ternal depression, a conclusion with importantimplications for intervention strategies.
The current paper reports a prospective longi-tudinal study of a community sample in whichmothers either experienced PND, or else formeda control group of non–postnatally depressedmothers. Uniquely, the design incorporates di-rectly observed measurements of infant attach-ment and a childhood behavioral profile akin toprevious accounts of ego resilience, as well asassessment of family adversities (subsequent ma-ternal depression, poor maternal support for thechild, marital conflict), and repeated assessmentsof child psychiatric disorder (at 8, 13, and 16years).
The overarching aim was to identify a devel-opmental risk pathway from infancy to the emer-gence of depression by 16 years of age. Specifi-cally, we tested whether (i) relative to controls,offspring exposed to PND had an increased rateof depression (major depression or dysthymia);(ii) offspring depression was predicted by inse-cure infant attachment and expressions of lowerego resilience in childhood; and (iii) family ad-versity beyond the postpartum period increasedrisk for offspring depression. Because these dif-ferent processes likely act in concert and influ-ence each other over time, and could directlyaffect offspring outcome or operate by indirect,mediating, relationships, we took a developmen-tal approach, using structural equation modeling(SEM). SEM is ideal for measuring relationshipsbetween variables across time while controllingfor earlier influences, and for exploring multiplepathways simultaneously, including indirect ef-fects. Here, we aimed to test two indirect ef-fects: first, that insecure infant attachmentwould influence offspring depression via lowerchild resilience; and second, that longer-termfamily adversities would account for the asso-ciation between PND and offspring depression.
METHODParticipantsMothers (and children) were recruited at 2 monthspostpartum, and assessments conducted at 18 months,and 5, 8, 13 and 16 years. Initially, a communitysample of mothers on postnatal wards of the Cam-bridge (UK) maternity hospital (N � 702) was
screened with the Edinburgh Postnatal Depression461www.jaacap.org
sAsCEowdwdhtdwmc
iicnctutrn
wLn
MURRAY et al.
Scale21 at 6 weeks postpartum. The response rate was97%. Mothers were primiparous, aged 18 to 42 years,were cohabiting or married, and had healthy, full-terminfants. The Standardized Psychiatric Interview22
(SPI), modified according to Research Diagnostic Cri-teria,23 was administered to probable cases at 8 weeks.Of the women, 61 (9.4%) were identified as depressed,of whom 58 (95%) were recruited. A total of 42mothers, randomly selected from the remainingwomen, were similarly interviewed and recruited ascontrols if they had no history of depression.
We have previously reported that index groupoffspring showed more insecure attachment in in-fancy,24 signs of depressive cognitions at 5 years,25 andemergence of affective disorder, principally anxiety,by 13 years.10 Here, we provide novel data onadolescent psychiatric outcomes at age 16 years (53index and 40 control) and the trajectory to depres-sion from earlier psychological vulnerability andfamily adversity.
The study was approved by the Cambridge MedicalEthics Committee; participants gave written informedconsent.
Offspring MeasuresInfant Attachment. At 18 months, infant attachmentwas assessed using Ainsworth’s Strange Situation Pro-cedure,26 a standardized observational measure ofinfant responses to maternal separation and reunion inan unfamiliar environment. This is a reliable measurethat is used widely in research. Infants were classifiedas secure or insecure.24
Child Ego Resilience. At 5 and 8 years, the childrenwere videotaped, with a same-aged friend, during acompetitive card game, in which the aim is to accu-mulate “snap” deals (pairs of identical cards).25 Eachchild received the same number of losing and winningdeals, administered in a predefined order. Trainedresearchers who were blinded to other informationscored, on five-point scales, the study child’s behavioron losing deals. Ratings included emotional and be-havioral responses (i.e., expressions of sadness anddistress, failure to persist) and spoken cognitions ex-pressing negative self-attributions or pessimism (e.g.,“I always lose at card games”).25 The mean of negativeminus positive scores (range �5 to �5) was computedas an index of low ego resilience comparable to thatused in previous studies.Offspring Mental State. At 16 years, diagnostic inter-views were conducted by a clinical researcher blind tomaternal state using the Kiddie Schedule for AffectiveDisorders and Schizophrenia, Present and LifetimeVersion (KSADs).27 The same assessments had beenconducted at 8 and 13 years with mother and child.10
All interviews were discussed with a clinical team, anda consensus diagnosis made, with best estimate judge-
ments made based on all available clinical information.JOURN
462 www.jaacap.org
For the 8-year assessment, the current and previous 12months’ mental state was determined; and at 13 and 16years, both the current mental state and that since theprevious assessment were ascertained. For major de-pressive disorder, timing of onsets was recorded on amonth-by-month basis. Information from these threeinterviews was pooled to establish offspring lifetimeoccurrence of disorder.
Maternal and Family Adversity MeasuresMaternal Mental State. Maternal depression was as-essed using the SPI at recruitment, the Schedule forffective Disorder and Schizophrenia—Life-time ver-
ion28 at 18 months and 5 years, and the Structuredlinical Interview for DSM-IV29 at 8, 13, and 16 years.ach time, maternal mental state since the previousccasion was assessed; timing of onsets and offsetsas recorded by month and used to compute overalluration. Duration of child exposure exclusive of PNDas defined as the number of months the mother wasepressed beyond 4 months postpartum (when 55%ad remitted) up to the child’s onset of depression (or
o 16-year interview, if none). At 16 years, only 19.3%epressed mothers received treatment; of these, 64.3%ere prescribed antidepressants. Interviews were ad-inistered by trained researchers, and reviewed by a
linical team blinded to child diagnoses.Maternal Support. At 5 and 8 years, mother–childnteractions were video recorded in conditions requir-ng maternal support for the child. At 5 years, thisomprised a 10-minute snack, in which the childeeded help to manage the refreshments; at 8 years itomprised a 20-minute math problem task. Interac-ions were scored by trained raters, blinded to group,sing a scheme tailored to child age and task demands,
o measure maternal emotional support, and includedatings of warmth, acceptance, sensitive responsive-ess, and availability.30,31
Marital Conflict. At each assessment, marital conflictas assessed using a combination of interview (theife Events and Difficulties Schedule32) and question-aire (the Dyadic Adjustment Scale33). Each provided
a binary measure denoting significant conflict; thesewere summed to give a continuous measure of conflictup to 13 years. Conflict showed considerable continu-ity, the mean correlation between assessments being� � 0.46 (range, 0.32–0.63).
Data AnalysisWe first established whether PND was associated withoffspring depression, then used logistic regression andanalysis of variance to examine (i) main effects of PNDon insecure infant attachment and lower ego resilienceat 5 and 8 years, and (ii) the association between PNDand family adversity (duration of maternal depression,
poor maternal support, marital conflict). In turn, weAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
GPPrl(5
a(W1vCoeOwsN1
ostnata
PND AND OFFSPRING DEPRESSION
investigated main effects of these variables on off-spring depression, using logistic regression. We thenconducted path analyses, using SEM, in two stages.First we examined the effects of PND on offspringdepression, taking into account child vulnerabilityfactors; second, we included family adversity fac-tors. SEM analyses were performed using the max-imum likelihood estimation with the Mplus 4.2software.34
RESULTSMean offspring age was 16.06 years (SD � 0.18years; range, 15.75–17.00 years). There were min-imal differences between index and controlgroups regarding socioeconomic status (I, II, andIII nonmanual35: 61.5% for control versus 67.9%for PND), child gender (male 50% for control vs47.2% for PND), and current family status (childresident with both biological parents, 87.5% for
TABLE 1 Prevalence of Lifetime Psychiatric Diagnosis by
Total Sample
Total(N � 93)
Female(n � 48)
Male(n � 45)
Axis 1 diagnosisAny 41.9 47.9 35.6
Depressive disorderAny 29.0 35.4 22.2MDE 26.9 33.3 20.0Dysthymia 4.3 6.3 2.2
Anxiety disorderAny 24.7 33.3 15.6Specific phobia 12.9 14.6 11.1GAD 5.4 6.3 4.4OCD 5.4 8.3 2.2Social phobia 3.2 4.2 2.2Separation anxiety 2.2 2.1 2.2Agoraphobia 2.2 2.1 2.2Panic 1.1 2.1 0.0PTSD 1.1 2.1 0.0
Behavioral disorderAny 8.6 4.2 13.3ADHD 5.4 2.1 8.9ODD 3.2 2.1 4.4Conduct Disorder 2.2 0.0 4.4
Note: ADHD � attention-deficit/hyperactivity disorder; GAD � generalcompulsive disorder; ODD � oppositional defiant disorder; PND � p
control versus 73.6% for PND). t
JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
roup Differences in Adolescentsychiatric Disordersychiatric diagnoses occurring by 16 years areeported in Table 1. Index children were moreikely than controls to experience Axis 1 disorder52.8% versus 27.5%, respectively; Wald statistic �.83, odds ratio [OR] � 2.95, 95% confidence
interval [CI] � 1.23–7.11, p � .05), and this wasccounted for by elevated rates of depressionMDE and dysthymia) (41.5% versus 12.5%;
ald statistic � 8.39, OR � 4.99, 95% CI �.68–14.70, p � .01), and anxiety disorders (32.1%ersus 15%; Wald statistic � 3.43, OR � 2.68, 95%I � 0.94–7.59, p � .10). Although depressionccurred in more girls than in boys, the differ-nce was not significant (Wald statistic � 1.93,R � 1.92, 95% CI � 0.77–4.81, NS), and neitheras the interaction PND by child gender (Wald
tatistic � 0.75, OR � 0.32, 95% CI � 0.03–4.12,S). Average age of first onset of depression was
3.81 years (SD � 1.40), the earliest being at 11.92;
ernal Group and by Child Gender
%
Control PND
l40)
Female(n � 20)
Male(n � 20)
Total(n � 53)
Female(n � 28)
Male(n � 25)
5 35.0 20.0 52.8 57.1 48.0
5 20.0 5.0 41.5 46.4 36.05 20.0 5.0 37.7 42.9 32.05 5.0 0.0 5.7 7.1 4.0
0 20.0 10.0 32.1 42.9 20.05 10.0 5.0 17.0 17.9 16.00 0.0 0.0 9.4 10.7 8.05 5.0 0.0 7.5 10.7 4.05 0.0 5.0 3.8 7.1 0.00 0.0 0.0 3.8 3.6 4.00 0.0 0.0 3.8 3.6 4.00 0.0 0.0 1.9 3.6 0.05 5.0 0.0 0.0 0.0 0.0
5 5.0 10.0 9.4 3.6 16.00 0.0 10.0 5.7 3.6 8.05 5.0 0.0 3.8 0.0 8.00 0.0 0.0 3.8 0.0 8.0
nxiety disorder; MDE � major depressive episode; OCD � obsessivel depression; PTSD � post-traumatic stress disorder.
Mat
Tota(n �
27.
12.12.
2.
15.7.0.2.2.0.0.0.2.
7.5.2.0.
ized a
here was no gender difference in age of first
463www.jaacap.org
tgssTsaltscaawtp8rvet3spt
ap[R
MURRAY et al.
episode (boys’ mean � 14.20, SD � 1.53; girls’mean � 13.58, SD � 1.32; t(25) � 1.12, NS).Among those who were depressed, 59.3% (n �16) had also had an anxiety diagnosis at somepoint. Girls more often experienced anxiety thanboys (Wald statistic � 3.79, OR � 2.71, 95% CI �0.99–7.41, p � .05), but there was no interactionPND by gender on anxiety (Wald statistic � 0.06,OR � 1.33, 95% CI � 0.15–12.07, NS). Behaviordisorders were uncommon and were comparablein the two groups (Wald statistic � 0.11, OR �1.28, � 95% CI � 0.29–5.72, NS).
Group Differences in Child Vulnerability andFamily Adversity, and Their Prediction toAdolescent DisorderTable 2 shows the effects of PND, child gender,and their interaction on child vulnerability andfamily adversity measures.Infant Attachment (Secure Versus Insecure). As re-ported previously,24 index group infants weremore likely to be insecurely attached than werecontrols, and boys were more likely to be inse-cure than were girls. There was a trend for a PNDby gender interaction, with boys of PND mothersbeing more frequently insecure.Child Ego Resilience (Lower Versus Higher). PNDwas associated with expressions of lower egoresilience at 5 years,25 and this tended to be thecase at 8 years, with index group boys againbeing particularly affected.Prediction of Depression by 16 Years. As shown inTable 2, both infant insecure attachment andlower ego resilience at 8 years predicted off-spring depression; lower ego resilience at 5 yearstended to do the same.Family Adversity. As shown in Table 2, PNDtended to be associated with poorer maternalsupport for the child at 5 years and was alsoassociated with the duration of maternal depres-sion and marital conflict. Poorer maternal sup-port at 8 years was not predicted by PND.Prediction of Depression by 16 Years. As shown inTable 2, poor maternal support at 5 years pre-dicted offspring disorder, but 8-year support didnot, and so was not considered further. Maritalconflict and duration of maternal depressionwere significantly associated with offspring life-time depression.
Path AnalysesThe first model focused on the child’s psycholog-
ical trajectory. Figure 1a shows the hypothesizedJOURN
464 www.jaacap.org
relationships between model variables: insecureinfant attachment and lower ego resilience at 5and 8 years were each expected to increase risk ofoffspring depression. PND was hypothesized tohave an adverse effect on infant attachment, andon 5- and 8-year ego resilience. Finally, insecureattachment was expected to have an adverseeffect on ego resilience. The correlations be-tween these variables, as well as those concern-ing family adversity, supported these predic-tions (Table 3).
Model acceptability was estimated from the �2
statistic, the comparative fit index (CFI), theTucker-Lewis coefficient (TLI), and the rootmean-square error of approximation (RMSEA).36
CFI and TLI values closer to 1.0, and RMSEAvalues of 0.08 or less, indicate acceptable modelfit.37
The initial model including child vulnerabilityfactors showed good fit [�2(1) � 0.02, p � .89;CFI � 1.0; TLI � 1.27; RMSEA � 0.00]. Never-heless, three paths were weak (r � 0.10), sug-esting their low contribution; therefore we ran aubsequent model without them. This showedimilarly good fit [�2(3) � 0.43, p � .93; CFI � 1.0;LI � 1.24; RMSEA � 0.00]. Figure 1b shows theecond model’s standardized regression weights,s well as the three excluded paths (dashedines). Notably, we had hypothesised that one ofhese, from insecure infant attachment to off-pring depression, would be mediated by lowerhild ego resilience. Consistent with this, thessociation between insecure infant attachmentnd offspring depression dropped substantiallyhen lower ego resilience was included (Sobel’s
est Z � 1.72, p � .08). In turn, and in line withredictions, the relationship between lower-year ego resilience and offspring depressionemained substantial (0.32). In combination, childulnerability factors accounted for 19% of theffect of PND on offspring depression, increasinghe explained variance from R2 � 18% to R2 �1%; and the effect of PND on offspring depres-ion dropped from 0.42 (direct association, nootential mediators) to 0.34 when insecure at-
achment and lower resilience were included.We next extended the model to include family
dversity. This showed good fit, and the ex-lained variance in depression increased by 12%
�2(9) � 12.66, p � .18; CFI � 0.92; TLI � 0.84;MSEA � 0.060; offspring depression R2 � 43%].
As seen in Figure 2, the direct effect of PND on
offspring depression dropped from 0.34 whenAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
TABLE 2 Postnatal Depression, Gender, and Postnatal Depression by Gender Effects on Child Vulnerability and Family Adversity, and Effects of Each of These onChild Depression Outcome
Control (n � 40) PND (n � 53)PND, Gender, and PND � Gender Effects on Child Vulnerability
and Family Adversity
Effects of ChildVulnerability and
Family Adversity onChild DepressionOutcome (KSADs)Male Female Male Female PND Effects Gender Effects
PND � GenderEffects
a. Child vulnerabilityInfant insecure attachment
(% of insecure)25.0% 25.0% 84.0% 46.4% Wald � 13.11***
OR � 5.37 CIb �
2.16-13.33
Wald � 3.78* OR� 0.44 CI �
0.19–1.01
Wald � 3.33†
OR � 0.16 CI� 0.02–1.14
Wald � 3.64* OR �
2.45, CI �
0.98–6.18Ego resilience, 5 y, mean
(SD)1.17 (0.66) 0.67 (0.89) 0.74 (0.83) 0.43 (0.69) F(1, 88) � 4.22* F(1, 88) � 6.30** F(1, 88) � 0.35 Wald � 3.36† OR �
1.07, CI �
0.99–1.16Ego resilience 8 y, mean
(SD)1.40 (0.65) 1.35 (0.55) 0.83 (0.79) 1.34 (0.85) F(1, 80) � 3.17† F(1, 80) � 2.08 F(1, 80) � 2.94† Wald � 7.67** OR �
1.07, CI �
1.02–1.13b. Family adversityPoor maternal support, 5
y, mean (SD)a2.54 (1.05) 2.88 (0.89) 3.06 (0.93) 3.09 (1.02) F(1, 84) � 2.91† F(1, 84) � 0.81 F(1, 84) � 0.55 Wald � 3.68* OR �
1.62, CI �
0.99–2.65Poor maternal support, 8
y, mean (SD)1.36 (0.11) 1.40 (0.14) 1.36 (0.10) 1.41 (0.12) F(1, 81) � 0.15 F(1, 81) � 2.48 F(1, 81) � 0.004 Wald � 2.00 OR �
0.08, CI �
.002-4.17Duration of MD (mo),
mean (SD)3.00 (4.49) 2.65 (4.54) 13.84 (12.50) 20.61 (15.57) F(1, 89) � 37.90*** F(1, 89) � 1.88 F(1, 89) � 2.31 Wald � 7.54** OR �
1.05, CI �
1.01–1.08Marital conflict, mean (SD) 1.10 (1.25) 1.10 (1.02) 2.08 (1.32) 2.46 (1.34) F(1, 89) � 19.82*** F(1, 89) � 0.53 F(1, 89) � 0.53 Wald � 6.78** OR �
1.58, CI �
1.12–2.24
Note: CI � confidence interval; KSADs � Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version; MD � maternal depression; PND � postnatal depression; Wald � Wald statistic.aHigher scores indicate poorer maternal support.bAll confidence intervals at 95%.*p � .05; **p � .01; ***p � .001; †p � .10.
PND
AN
DO
FFSPRING
DEPRESSIO
N
JOU
RNA
LO
FTH
EA
MERIC
AN
AC
AD
EMY
OF
CH
ILD&
AD
OLESC
ENT
PSYC
HIA
TRY
VOLU
ME
50N
UM
BER5
MA
Y2011
465
ww
w.jaacap.org
MURRAY et al.
only the child’s psychological trajectory was con-sidered to 0.08 when additional family adversitywas included, suggesting that child vulnerabilityand family adversity accounted for 81% of theeffect of PND on offspring depression. Because
FIGURE 1 (a.) The hypothesized relationship betweenincluding infant attachment and child resilience. (b.) The fiaccount infant attachment and child resilience. Note: Dasmodel. Coefficients represent standardized regression we
TABLE 3 Relationship Among All Variables
2 3
1 PND 0.39*** 0.22*2 Insecure attachment 0.083 Lower ego resilience, 5 y4 Lower ego resilience, 8 y5 Poor maternal support, 5 y6 Marital conflict7 Duration of MD8 Offspring lifetime depression
Note: Coefficients represent Pearson or � correlations, as appropriate. Poas 0 � secure/1 � insecure. Offspring lifetime depression scored as
*p � .05; **p � .01; ***p � .001; †p � .10.
JOURN
466 www.jaacap.org
this model took the intercorrelations among thethree adversity factors into account, we exam-ined their individual paths to identify the contri-bution of each. This showed that duration ofmaternal depression (Z � 2.61, p � .008) and
atal depression (PND) and offspring depression,model linking PND to offspring depression, taking intolines represent weak paths excluded from the final. Offspring depression n � 93.
4 5 6 7 8
.17 0.18† 0.28** 0.64** 0.32**
.25* 0.15† 0.31** 0.21* 0.20*
.14 0.14 0.12 0.28** 0.19†
0.14 0.14 0.10 0.31**0.23* 0.29** 0.20*
0.33** 0.28**0.31*
l depression (PND) scored as 0 � control/1 � index. Attachment scoredo depression/1 � depression.
postnnal
hedights
000
stnata0 � n
AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
fnw
PND AND OFFSPRING DEPRESSION
marital conflict (Z � 2.23, p � .03) were partialmediators of PND on offspring depression, withPND still being marginally significant when theformer was entered in the model (p � .074), andsignificant (p � .03) when marital conflict wasincluded; no such effect obtained for poor mater-nal support at 5 years (Z � 1.28, NS).
Duration of Exposure to Maternal DepressionGiven the effect on offspring depression of theoverall duration of maternal depression, wewere interested in identifying the threshold (ifany) whereby such exposure starts to increaserisk, over and above the PND effect itself. Thuswe disaggregated the total duration of depres-sion beyond 4 months postpartum into quar-tiles: never subsequently depressed, and de-pressed between 1 and 7, 8 and 16, and 17 ormore months. Only four PND mothers had nosubsequent depression (first quartile), andtherefore only the remaining quartiles wereexamined. Effects on offspring depression ofexposure versus nonexposure to each of thesedepression durations was examined in terms of
FIGURE 2 The relationship between postnatal depressiovulnerability and family adversity. Note: Coefficients reprn � 93.
direct and indirect (controlling for PND) ef-
JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
ects. The second and third quartiles showedo effects; however, a main effect of the fourthas observed (Wald statistic � 10.21, OR �
5.20 (95% CI � 1.88 –14.29), p � .01). Further-more, when controlling for PND, the fourthquartile was still significantly associated withoffspring depression (Wald � 3.67, OR � 3.03,95% CI � 0.97–9.44), p � .05; PND Wald �3.12, OR � 3.00 (95% CI � 0.89 –10.16), p �.078).
We also examined whether the effect of longduration of maternal depression might actuallybe explained by recent exposure (defined asepisodes in the year before offspring onset).Although there was a modest association be-tween the two measures (r � .23, p � .028), recentmaternal episodes were unrelated to offspringdepression (t � �1.20, NS); and when bothvariables were considered jointly, only the fourthquartile duration was significant, indicating thatits effects were not explained by recent exposure(recency: Wald statistic � 0.43, OR � 1.07, 95%CI � 0.87–1.31, NS; fourth quartile duration:Wald statistic � 9.57, OR � 5.27, 95% CI �
ND) and offspring depression, accounting for childstandardized regression weights. Offspring depression
n (Pesent
1.83–15.12, p � .002).
467www.jaacap.org
ue
mseietdfteiirccM
sapabeht
MURRAY et al.
Anxiety � Depression Lifetime DiagnosisGiven the association between PND and off-spring anxiety, we explored whether our modelalso applied to anxiety, testing each target pre-dictor in relation to this outcome. Only maritalconflict was significant (Wald statistic � 8.34,OR � 1.73, 95% CI � 1.19 –2.52, p � .01),suggesting that this form of adversity mightcontribute to the development of both disor-ders, whereas the child developmental trajec-tory elucidated in the present study may berelatively specific to depression.
DISCUSSIONWe report the occurrence of clinical depression ina community sample of adolescents within aprospective longitudinal study that uniquely in-cluded direct investigation of interpersonal psy-chology (mother–infant attachment) and indicesof low resilience under social stress. Further-more, we examined the role of ongoing familyadversity to establish the extent to which itadded to child risk, and accounted for anyassociation between maternal PND and childdisorder.
In line with other studies,11,12 we found off-spring of postnatally depressed mothers to beat substantially increased risk for depressionthemselves; furthermore, their rate of life-timedepression by age 16 was slightly more than 40%,consistent with evidence on adolescent offspringof depressed parents in general,6,7 as was theirage at first onset (typically after age 12), andfrequent anxiety disorder.3,4
As predicted, we found insecure attachment tothe mother in infancy to predict depression inadolescence; furthermore, this association aroselargely via lower child ego resilience, as ex-pressed in a constellation of cognitive–affective–behavioral features, elicited under conditions ofsocial challenge, some years before first onset ofdepression. To our knowledge, this representsthe first demonstration of this trajectory frominfancy through to adolescence. Notably, how-ever, continuing family adversity (poor maternalsupport, marital conflict and prolonged maternaldepression) added substantially to offspring risk.In our sample, boys of depressed mothers wererather more likely than girls to be insecure andless resilient, possibly reflecting a more generalvulnerability to PND.38 This might account forthe somewhat smaller than usual gender differ-
ence in the rate of depression in our sample.39JOURN
468 www.jaacap.org
We found marital conflict and further mater-nal depression to be partial mediators of theeffects of PND on adolescent depression. Withregard to the latter, however, only maternaldepression accumulating for more than 17months beyond the postnatal period added sig-nificantly to offspring risk, and even such pro-longed subsequent maternal depression did notentirely eliminate the association between thepostnatal episode and offspring depression. Inour study, therefore, the presence of PND doesseem to have been of particular importance foroffspring outcome. It should be noted, though,that we recruited a relatively low-risk sample;given greater socio-economic adversity,12 such
nique effects of the postnatal episode might beclipsed by subsequent family difficulties.
Although our study focussed on the develop-ent of depression, we investigated whether the
ame trajectory held for anxiety. Offspring anxi-ty was also predicted by PND, and it occurredn more than half those offspring who experi-nced depression. Nevertheless, apart from PND,he only predictor associated with both offspringisorders was marital conflict, an established risk
or child anxiety, as well as depression.40 The facthat neither insecure attachment nor lower childgo resilience predicted anxiety was not surpris-ng: With regard to attachment, insecure infantsn our sample showed either an avoidant or aesistant pattern (with avoidant being far moreommon (86% of insecure infants),24 the sameombination that predicted depression in the
innesota study.19 Anxiety, by contrast, hasbeen linked more commonly to resistant insecureattachment.41 Moreover, our measure of low egoresilience, as in the wider literature,15,16 sub-umed cognitions and temperamental featuresssociated with vulnerability to depression (e.g.,essimism, low self-worth, lack of persistence,nd flexibility), as well as sad affect, rather thanehaviors and cognitions more specific to anxi-ty. Thus, although anxiety and depression areighly comorbid, our findings, in common with
hose of others,42 do suggest some etiologicalspecificity.
Our study had a number of strengths, includ-ing direct assessments through childhood, re-peated diagnostic interviews, and high sampleretention. Nevertheless, some limitations ap-plied. First, our relatively small sample size re-stricted the analyses that could be conducted
(e.g., insecure attachment subtype effects). Fur-AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
sowamamm
PND AND OFFSPRING DEPRESSION
thermore, although recent research has shownthat the association between parent and juveniledepression is mediated predominantly by theenvironment,8,9 it was a study limitation that wecould not address the issue of possible geneticcontributions to offspring disorder; and informa-tion was not collected on maternal antenatal risk(e.g., alcohol/drug use) or mental state, factorsincreasingly recognized as important in under-standing effects of postnatal disorder.43
The substantially raised risk for depressionamong offspring of postnatally depressed moth-ers underlines the importance of screening forPND and of delivering early interventions. More-over, as this risk could, in part, be traced back toinfant insecure attachment to the mother, and astreatments directed only at alleviating maternaldepression do not appear to benefit the mother–child relationship,44 interventions focussing onpromoting good parental care are desirable. Ev-idence for the benefits of these is accumulat-ing.45-47 Our findings regarding subsequent ma-ternal depression also suggest, however, thatwhere mothers experience postnatal depression,long-term monitoring and support might also be
important. Furthermore, as marital conflict con-sion and psychiatric outcomes in adolescent offspring: a 13-yearlongitudinal study. J Affect Disord. 2007;97:145-154.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011
tituted an important component of risk for bothffspring depression and anxiety and, in lineith wider research,32 was significantly associ-
ted with maternal depression, interventionsight need to be tailored accordingly, possibly
lso encompassing the paternal depression com-only associated with maternal disorder andarital conflict.48 &
Accepted February 3, 2011.
Drs. Murray, Arteche, Fearon, Halligan, and Cooper are with theUniversity of Reading, School of Psychology and Clinical LanguageSciences, Reading, UK. Dr. Goodyer is with University of Cambridge,Cambridge, UK.
The study was supported by grants from the Medical Research Council(G9324094) and the Tedworth Charitable Trust (TED76).
The authors thank Sheelah Seeley, Janet Edwards, and JosephMurray of the University of Cambridge for assistance in manuscriptpreparation.
Disclosure: Drs. Murray, Arteche, Fearon, Halligan, Goodyer, andCooper report no biomedical financial interests or potential conflicts ofinterest.
Correspondence to Dr. Lynne Murray, School of Psychology,University of Reading, Reading, RG6 6AL, UK; e-mail: [email protected]
0890-8567/$36.00/©2011 American Academy of Child andAdolescent Psychiatry
DOI: 10.1016/j.jaac.2011.02.001
REFERENCES1. Andrews G, Poulton R, Skoog I. Lifetime risk of depression:
restricted to a minority or waiting for most. Br J Psychiatry. 2005;187:495-496.
2. Judd LL, Akiskal HS, Maser JD, et al. A prospective 12-year studyof subsyndromal and syndromal depressive symptoms in unipo-lar major depressive disorders. Arch Gen Psychiatry. 1998;55:694-700.
3. Kovacs M, Feinberg TL, Crouse-Novak M, Paulauskas SL, PollockM, Finkelstein R. Depressive disorders in childhood II. A longi-tudinal study of the risk for subsequent major depression. ArchGen Psychiatry. 1984;41:643-649.
4. Harrington R, Fudge H, Rutter M, Pickles A, Hill J. Adultoutcomes of childhood and adolescent depression: I. Psychiatricstatus. Arch Gen Psychiatry. 1990;47:465-473.
5. Jaffee SR, Moffitt TE, Caspi A, Fombonne E, Poulton R, MartinJ. Differences in early childhood risk factors for juvenile-onsetand adult-onset depression. Arch Gen Psychiatry. 2002;59:223-224.
6. Beardslee WR, Versage EM, Gladstone TRG. Children of affec-tively ill parents: a review of the past 10 years. J Am Acad ChildAdolesc Psychiatry. 1998;37:1134-1141.
7. Weissman MM, Wickramaratne P, Nomura Y, et al. Families athigh and low risk for depression: a 3-generation study. Arch GenPsychiatry. 2005;62:29-36.
8. Silberg JL, Maes H, Eaves LJ. Genetic and environmental influ-ences on the transmission of parental depression to children’sdepression and conduct disturbance: an extended Children ofTwins study. J Child Psychol Psychiatry. 2010;51:734-744.
9. Tully EC, Iacono WG, McGue M. An adoption study of parentaldepression as an environmental liability for adolescent depres-sion and childhood disruptive disorders. Am J Psychiatry. 2008;165:1148-1154.
10. Halligan SL, Murray L, Martins C, Cooper PJ. Maternal depres-
11. Hammen C, Brennan P. Severity, chronicity, and timing ofmaternal depression and risk for adolescent offspring diagno-ses in a community sample. Arch Gen Psychiatry. 2003;60:253-260.
12. Hay DF, Pawlby S, Waters CS, Sharp D. Antepartum and postpartum depression: different effects on different adolescentoutcomes. J Child Psychol Psychiatry. 2008;49:1079-1088.
13. Martins C, Gaffan E. Effects of early maternal depression onpatterns of infant-mother attachment: a meta-analytic investiga-tion. J Child Psychol Psychiatry. 2000;41:737-746.
14. Arend R, Gove FL, Sroufe AL. Continuity of individual adapta-tion from infancy to kindergarten: a predictive study of ego-resilience and curiosity in preschoolers. Child Dev. 1979;50:950-959.
15. Block JH, Block J. The role of ego-control and ego-resiliency in theorganization of behavior. In Collins WA, ed. Development ofcognition, affect, and social relations: the Minnesota Symposia onchild psychology. Hillsdale, NJ: Erlbaum; 1980.
16. Eisenberg N, Spinrad T, Fabes R, Reiser M, Cumberland A,Shepard S. The relations of effortful control and impulsivity tochildren’s resiliency and adjustment. Child Dev. 2004;75:25-46.
17. Goodyer IM. Emanuel Miller Lecture: early onset depressions-meanings, mechanisms and processes. J Child Psychol Psychiatry.2008;49:1239-1256.
18. Duggal S, Carlson EA, Sroufe LA, Egeland B. Depressive symp-tomatology in childhood and adolescence. Dev Psychopathol.2001;13:143-164.
19. Sroufe LA, Carlson EA, Egeland B, Collins WA. The developmentof the person: the Minnesota Study of risk and adaptation frombirth to adulthood. New York: Guilford; 2005.
20. Belsky J, Fearon RMP. Early attachment security, subsequentmaternal sensitivity, and later child development: does continuityin development depend upon continuity of caregiving? Attach
Hum Dev. 2002;4: 361-387.469www.jaacap.org
MURRAY et al.
21. Cox J, Holden J, Sagovsky R. Detection of postnatal depression:development of the 10-item Edinburgh Postnatal DepressionScale. Br J Psychiatry. 1987;150:782-786.
22. Goldberg DP, Cooper B, Eastwood MR, Kedward HB, ShepherdMA. A standardised psychiatric interview for use in communitysurveys. Br J Prev Soc Med. 1970;24:18-23.
23. Spitzer RL, Endicott J, Robbins E. Research diagnostic criteria:rationale and reliability. Arch Gen Psychiatry. 1978;35:773-782.
24. Murray L. The impact of postnatal depression on infant develop-ment. J Child Psychol Psychiatry. 1992;33:543-561.
25. Murray L, Woolgar M, Cooper P, Hipwell A. Cognitive vulnera-bility to depression in 5-year-old children of depressed mothers.J Child Psychology Psychiatry. 2001;42:891-899.
26. Ainsworth M, Blehar M, Waters E, Wall S. Patterns of attachment.Hillsdale, NJ: Erlbaum; 1978.
27. Kaufman J, Birmaher B, Brent D, et al. Schedule for AffectiveDisorders and Schizophrenia for school-age children—presentand lifetime version (K-SADS-PL): initial reliability and validitydata. J Am Acad Child Adolesc Psychiatry. 1997;36:980-988.
28. Endicott J, Spitzer R. The Schedule for Affective Disorders andSchizophrenia. Arch Gen Psychiatry 1978;35:837-844.
29. Spitzer R, Williams J, Gibbon M. Structured Clinical Interview forDSM-IV (SCID). New York: New York State Psychiatric Institute,Biometrics Research; 1995.
30. Murray L, Sinclair D, Cooper P, Ducournau P, Turner P, Stein A.The socio-emotional development of five year old children ofpostnatally depressed mothers. J Child Psychol Psychiatry. 1999;40:1259-1272.
31. Murray L, Woolgar M, Martins C, Christaki A, Hipwell A, CooperP. Conversations around homework: links to parental mentalhealth, family characteristics and child psychological functioning.Br J Dev Psychol. 2006;24:125-149.
32. Brown GW, Harris TO. Social origins of depression. London:Tavistock; 1978.
33. Spanier GB. A new measure for assessing the quality of marriagesand similar dyads. J Marriage Fam. 1976;38:15-28.
34. Muthén LK, Muthén BO. Mplus: user’s guide (5th ed.). LosAngeles, CA: Muthén & Muthén; 2007.
35. Registrar General. Classification of occupation. London: HMSO;1980.
36. Markland D. The golden rule is that there are no golden rules: acommentary on Paul Barrett’s recommendations for reporting
JOURN
470 www.jaacap.org
model fit in structural equation modelling. Personal Indiv Diff.2007;42:851-858.
37. Bentler PM. Comparative fit indexes in structural models. PsycholBull. 1990;107:238-246.
38. Murray L, Halligan S L, Cooper PJ. Effects of postnatal depressionon mother-infant interactions, and child development. In WachsT, Bremner G, eds. Handbook of infant development. Oxford, UK:Wiley-Blackwell; 2010.
39. Cyranowski J, Frank E, Young E, Shear K. Adolescent onset of thegender difference in lifetime rates of major depression. Arch GenPsychiatry. 2000;57:21-27.
40. Murray L, Creswell C, Cooper PJ. The development of anxietydisorders in childhood: an integrative review. Psychol Med.2009;39:1413-1423.
41. Warren S, Huston L, Egeland B, Sroufe LA. Child and adolescentanxiety disorders and early attachment. J Am Acad Child AdolescPsychiatry. 1997;36:637-644.
42. Kessler RC, Gruber M, Hettema JM, Sampson N, Yonkers KA.Co-morbid major depression and generalized anxiety disorders inthe National Comorbidity Survey follow-up. Psychol Med. 2008;38:365-374.
43. O’Connor TG, Heron J, Golding J, Beveridge M, Glover V.Maternal antenatal anxiety and children’s behavioral/emotionalproblems at 4 years. Br J Psychiatry. 2002;180:502-508.
44. Nylen K, Moran T, Franklin C, O’Hara M. Maternal depression: areview of relevant treatment approaches for mothers and infants.Infant Ment Health J. 2006;27:327-343.
45. Bakermans-Kranenburg MJ, van Ijzendoorn M, Juffer F. Less ismore: meta-analyses of sensitivity and attachment interventionsin early childhood. Psychol Bull. 2003;129:195-215.
46. Cooper PJ, Tomlinson M, Swartz L, et al. Improving quality ofmother-infant relationship and infant attachment in socioeco-nomically deprived community in South Africa: randomisedcontrolled trial. Br Med J. 2009;338:974.
47. Kitzman H, Olds D, Cole R, et al. Enduring effects of prenatal andinfancy home visiting by nurses on children: follow-up of arandomized trial among children at age 12 years. Arch PediatrAdolesc Med. 2010;164:412-418.
48. Cummings EM, Keller PS, Davies PT. Towards a family processmodel of maternal and paternal depressive symptoms: exploring
multiple relations with child and family functioning. J ChildPsychol Psychiatry. 2005;46:479-489.AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY
VOLUME 50 NUMBER 5 MAY 2011