You are on page 1of 9

Articles

Prediction of perinatal depression from adolescence and


before conception (VIHCS): 20-year prospective cohort study
George C Patton, Helena Romaniuk, Elizabeth Spry, Carolyn Coey, Craig Olsson, Lex W Doyle, Jeremy Oats, Stephen Hearps, John B Carlin,
Stephanie Brown

Summary
Background Perinatal depression is a neglected global health priority, aecting 1015% of women in high-income Lancet 2015; 386: 87583
countries and a greater proportion in low-income countries. Outcomes for children include cognitive, behavioural, Published Online
and emotional diculties and, in low-income settings, perinatal depression is associated with stunting and physical June 11, 2015
http://dx.doi.org/10.1016/
illness. In the Victorian Intergenerational Health Cohort Study (VIHCS), we aimed to assess the extent to which
S0140-6736(14)62248-0
women with perinatal depressive symptoms had a history of mental health problems before conception.
This online publication has been
corrected. The corrected version
Methods VIHCS is a follow-up study of participants in the Victorian Adolescent Health Cohort Study (VAHCS), which rst appeared at thelancet.com
was initiated in August, 1992, in the state of Victoria, Australia. In VAHCS, participants were assessed for health on August 28, 2015
outcomes at nine timepoints (waves) from age 14 years to age 29 years. Depressive symptoms were measured with the See Comment page 835
Revised Clinical Interview Schedule and the General Health Questionnaire. Enrolment to VIHCS began in Centre for Adolescent Health,
September, 2006, during the ninth wave of VAHCS; depressive symptoms at this timepoint were measured with the Murdoch Childrens Research
Institute, University of
Composite International Diagnostic Interview. We contacted women every 6 months (from age 29 years to age 35 years) Melbourne, Royal Childrens
to identify any pregnancies. We assessed perinatal depressive symptoms with the Edinburgh Postnatal Depression Hospital Melbourne, Parkville,
Scale (EPDS) by computer-assisted telephone interview at 32 weeks of gestation, 8 weeks after birth, and 12 months VIC, Australia
after birth. We dened perinatal depression as an EPDS score of 10 or more. (Prof G C Patton MD, E Spry BA,
H Romaniuk PhD, C Coey PhD,
S Hearps BPsych); Clinical
Findings From a stratied random sample of 1000 female participants in VAHCS, we enrolled 384 women with Epidemiology and
564 pregnancies. 253 (66%) of these women had a previous history of mental health problems at some point in adolescence Biostatistics Unit, Murdoch
or young adulthood. 117 women with a history of mental health problems in both adolescence and young adulthood had Childrens Research Institute,
and Department of
168 pregnancies, and perinatal depressive symptoms were reported for 57 (34%) of these pregnancies, compared with Paediatrics, University of
16 (8%) of 201 pregnancies in 131 women with no preconception history of mental health problems (adjusted odds ratio Melbourne, Royal Childrens
836, 95% CI 3342087). Perinatal depressive symptoms were reported at one or more assessment points in Hospital Melbourne, Parkville,
VIC, Australia
109 pregnancies; a preconception history of mental health problems was reported in 93 (85%) of these pregnancies.
(Prof J B Carlin PhD,
H Romaniuk); Psychological
Interpretation Perinatal depressive symptoms are mostly preceded by mental health problems that begin before Sciences and Paediatrics,
pregnancy, in adolescence or young adulthood. Women with a history of persisting common mental disorders before Murdoch Childrens Research
Institute, University of
pregnancy are an identiable high-risk group, deserving of clinical support throughout the childbearing years.
Melbourne, Parkville, VIC,
Furthermore, the window for considering preventive intervention for perinatal depression should extend to the time Australia (Prof C Olsson PhD);
before conception. Centre for Social and Early
Emotional Development,
School of Psychology, Deakin
Funding National Health and Medical Research Council (Australia), Victorian Health Promotion Foundation, Colonial
University, Geelong, VIC,
Foundation, Australian Rotary Health Research and Perpetual Trustees. Australia (Prof C Olsson); Royal
Womens Hospital and
Introduction symptom prole,9 a particular endocrine sensitivity,10 and Murdoch Childrens Research
Institute, University of
Despite being one of the most common complications of a better prognosis than aective disorders diagnosed
Melbourne, Parkville, VIC,
pregnancy, perinatal depression remains a neglected global outside of pregnancy.8 In the past two decades, views have Australia (Prof L W Doyle MD);
health priority.1 In high-income countries, this disorder shifted. Depressive symptoms are recognised as common Department of Obstetrics and
aects 1015% of women2,3 and can have physical, cognitive, during pregnancy and, in turn, predictive of postnatal Gynaecology, University of
Melbourne, Melbourne, VIC,
and emotional eects on their childrens development, depression.11,12 Perinatal depression is now commonly Australia (Prof L W Doyle);
continuing into later life.4 Both antenatal and postnatal used to encompass syndromes that emerge either during School of Population and
depressive symptoms have been associated with poor early pregnancy or after birth.13 Mental disorders before Global Health, University of
child health and development.5 In low-income and middle- pregnancy are also recognised as an important risk factor Melbourne, Melbourne, VIC,
Australia (Prof J Oats DM); and
income countries, estimates of prevalence vary from 15% for perinatal depression.8,1416 In a study of more than Healthy Mothers Healthy
to 50%.1 In these settings, associations extend to a failure to 1000 Italian women, in which perinatal depression was Families Research Group,
thrive in utero, childhood stunting, and childhood physical recorded from 3 months of gestation, 30% of women Murdoch Childrens Research
illness, with antenatal depressive symptoms also linked to with an incident perinatal episode reported a history of Institute, and General Practice
and Primary Health Care
preterm birth and low birthweight.3,6,7 depression.17 Yet, retrospective identication of episodes Academic Centre, University of
Postnatal depression was long held to be a discrete before pregnancy could lead to underestimates of earlier Melbourne, Parkville, VIC,
syndrome arising without previous history,8 with a unique mental disorder.18 Australia (S Brown PhD)

www.thelancet.com Vol 386 August 29, 2015 875


Articles

Correspondence to: We designed the Victorian Intergenerational Health Procedures


Prof George C Patton, Centre for Cohort Study (VIHCS) to ascertain the extent to which We began our study in September, 2006, during the
Adolescent Health, Murdoch
Childrens Research Institute,
parental health, lifestyle, and social adjustment before ninth wave of follow-up of VAHCS. We contacted all
University of Melbourne, Royal conception might aect maternal perinatal health and young women who were still active in VAHCS to obtain
Childrens Hospital, Parkville, neonatal development. Here, we aimed to assess the their verbal consent for participation in VIHCS. Every
VIC 3052, Australia extent to which women with perinatal depressive 6 months, we contacted participants by text message,
george.patton@rch.org.au
symptoms had a history of depression and anxiety email, or telephone to identify pregnancies; we asked
before conception. participants if they were currently pregnant, were
planning a pregnancy, or had a child younger than 1 year
Methods who was not enrolled in VIHCS. We invited women who
Participants armed they were pregnant or had an infant younger
We recruited female participants from the Victorian than 1 year to participate in our study.
Adolescent Health Cohort Study (VAHCS),19 a In VAHCS during waves 17, common mental
prospective cohort study of health in young people disorders were assessed with the Revised Clinical
living in the state of Victoria, Australia, which was Interview Schedule (CIS-R), a psychiatric interview
initiated in August, 1992. The VAHCS population designed to assess symptoms of depression and anxiety
consisted of a representative sample of male and female in non-clinical populations.20 Presence of a common
adolescents who were selected with two-stage cluster mental disorder was adjudicated as a score of 12 or more
sampling. In stage one, 45 schools were chosen at on the CIS-R, which is the level at which intervention by
random from a stratied frame of government, a family doctor would be appropriate. In our study,
Catholic, and independent schools with a probability we categorised the persistence of common mental
proportional to the number of year 9 students (typically disorders during adolescenceie, a score of 12 or more
aged 1415 years) in the schools in every stratum. In on the CIS-R in no waves, in one wave, or in two or
stage two, two distinct complete classes were selected at more waves.
random from every participating school. One class In VAHCS during waves 8 and 9, symptoms of
entered the study in the latter part of the ninth school depression and anxiety were assessed with the 12-item
year (wave 1; August, 1992) and the second class entered General Health Questionnaire (GHQ-12), with high
6 months later (wave 2; February, 1993). Participants symptoms of common mental disorder dened as a
were reviewed at four subsequent 6-month intervals score of 3 or more.21,22 Further, in wave 9, depression and
(waves 36; August, 1993, to February, 1995), with anxiety were assessed using the Composite International
three follow-up reviews in young adulthood: at age Diagnostic Interview (CIDI): major depressive disorder
2021 years (wave 7; 1998), 2425 years (wave 8; was measured with CIDI-auto23 and anxiety disorder with
200103), and 2829 years (wave 9; 200608). School CIDI-short form.24 We dened major depressive disorder
retention rates to year 9 in the year of sampling were and anxiety disorder according to the International
98%. One of the 45 schools did not continue beyond Classication of Diseases, 10th revision (ICD-10).
wave 1, with a loss of 13 participants. From a total of We classed participants as having anxiety disorder if they
2032 adolescents, 1943 (96%) participated at least once were diagnosed with generalised anxiety disorder, social
during the rst six (adolescent) waves, 1000 of whom phobia, agoraphobia, or panic disorder.
were young women. Figure 1 shows the ow of female We categorised the persistence of common mental
participants through VAHCS. disorders in young adulthood using all three measures
Written consent for participation in VAHCS was (CIS-R, GHQ-12, and CIDI) to identify symptoms of
provided initially by parents. At every survey point major depressive disorder and anxiety disorder in no
(wave), informed verbal consent was sought explicitly waves, one wave, or two or more waves. We dened
from every participant. The human research ethics continuity of mental health disorder from adolescence to
committee at the Royal Childrens Hospital Melbourne young adulthood as no disorder, adolescent disorder
approved data collection protocols for both VAHCS only, young adult disorder only, and both adolescent and
and VIHCS. young adult disorder.

Adolescent Young adult


Survey
timepoint Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8 Wave 9
Year 1992 1993 1993 1994 1994 1995 1998 200103 200608
Mean age 149 years 154 years 159 years 163 years 168 years 174 years 206 years 240 years 290 years
Sample n=466 n=907 n=888 n=875 n=854 n=848 n=866 n=824 n=806

2 entry points

Figure 1: Sampling and retention of female participants in VAHCS, 19922008


Participants were recruited during waves 1 and 2 and followed up at seven timepoints (waves 39). VAHCS=Victorian Adolescent Health Cohort Study.

876 www.thelancet.com Vol 386 August 29, 2015


Articles

Adolescent risky behaviours were also measured in assessment measures nested within pregnancy and
VAHCS.19 Cigarette smoking was recorded with a pregnancies nested within women. For models assessing
self-report diary over the previous 7 days; daily smokers the persistence of perinatal mental disorders for every
were classied as those smoking on 6 or 7 days in the pregnancy, we used a two-level variance structure with
previous week. Risky alcohol use was assessed over a pregnancies nested within women. We used linear
period of 1 week in a beverage-specic and quantity- multilevel models with random intercepts for continuous
specic self-report diary; drinking ve or more standard measures of perinatal depression and logistic multilevel
drinks (one standard drink being equal to 10 g alcohol) models with random intercepts for binary measures.
on any day was dened as binge drinking, the most Initially, we adjusted models for perinatal assessment
common form of alcohol misuse in adolescents. point only (partial adjustment); in subsequent models
Participants who reported cannabis use at least once a we also adjusted for parental divorce or separation,
week were also identied by self-report. Antisocial pregnancy history, and adolescent risky behaviours.
behaviour during adolescence was measured with We estimated marginal means and probabilities from the
ten items on the Mott and Silva self-report early fully adjusted models at the mean values of the covariates.
delinquency scale25 relating to property damage, We used multiple imputation to handle incomplete data.
interpersonal conict, and theft in the previous 6 months. We obtained all proportions and model parameter
Antisocial behaviour was dened if one behaviour was estimates by averaging results across 20 imputed datasets
reported more than once or two dierent behaviours (appendix p 1),31 with inferences under multiple imputation See Online for appendix
were noted at least once. obtained using Rubins rules.32 We calculated frequency
We asked participants to complete computer-assisted estimates with imputed percentage estimates and
telephone interviews at three perinatal assessment total number of female participants or pregnancies.
points: at 32 weeks of gestation, 8 weeks after birth, and We assessed two-way interactions between perinatal
12 months after birth. We assessed perinatal depressive assessment point, persistence of disorder, and parity in
symptoms for every pregnancy with the Edinburgh fully adjusted models, but these were not retained in
Postnatal Depression Scale (EPDS).26 The EPDS is a nal models. We assessed all main eects and interactions
ten-item rating scale with high internal consistency, using p values from Wald tests. We did sensitivity analyses
designed for postnatal depression screening but validated with available case data from the current study for women
for antenatal use.26 Scores on the EPDS range from who had complete VAHCS data at waves 29. We analysed
0 to 30; we used a threshold of 10 or more to dene data with Stata version 13.
perinatal depression. This cuto generally indicates
minor depression in English-speaking women with a Role of the funding source
pencil and paper format,27,28 but on telephone interview The funding sources had no role in study design, data
it has been judged the optimum point at which to collection, data analysis, data interpretation, or writing of
identify depressive disorder assessed on a structured the report. The corresponding author had full access to
psychiatric interview.29 We also categorised the persis- all data in the study and had responsibility for the nal
tence of perinatal depressive symptoms for every decision to submit for publication.
pregnancy in the perinatal periodie, a score on the
EPDS of 10 or more at no or one perinatal assessment Results
point or at two or three assessment points. Between November, 2006, and July, 2013, women were
Every participants pregnancy history was recorded in screened for participation in VIHCS. Of 1000 young
VAHCS during young adulthood (waves 79), including women who had participated in VAHCS at least once
previous full-term pregnancies, miscarriages, termin- during adolescence, 872 (87%) were still active in
ations, and stillbirths. We also recorded pregnancy history VAHCS and were eligible for our study (gure 2).
at the rst perinatal assessment for every pregnancy. Of 128 women who were no longer active in VAHCS at
wave 9, two had died, 87 declined to participate further,
Statistical analysis and 39 were lost to follow-up. Baseline measurements
We estimated the frequency of preconception exposures gathered at the rst assessment did not dier between
for every woman, using data for their rst pregnancy (to women still active in VAHCS and those no longer
avoid double counting in case of multiple pregnancies). involved (appendix pp 23), with the exception of place of
For every pregnancy in our study, we estimated mean birth, with women not born in Australia less likely to
EPDS scores and the frequency of perinatal depressive remain active in VAHCS by wave 9 (odds ratio 35,
symptoms. We used multilevel models to investigate 95% CI 2158; p<00001).
the association between preconception disorders and Of the 872 women eligible for our study, 466 (53%)
perinatal mental health and to accommodate data reported 744 pregnancies during the study window
structured hierarchically.30 For models examining perinatal (median follow-up 63 years, IQR 5970). 12 infants
depressive symptoms at every perinatal assessment, we died in utero before 32 weeks of gestation, resulting in
used a three-level variance structure with repeated 732 livebirths to 465 women, similar to the expected

www.thelancet.com Vol 386 August 29, 2015 877


Articles

birth rate in Victoria by maternal age for the active points. The 67 women who did not participate in our
female sample (724 expected births).33 398 (86%) of study had a higher prevalence of depressive symptoms at
465 women who had 600 (82%) of 732 eligible pregnan- waves 7 and 9, and of anxiety at wave 9 (appendix pp 46).
cies participated at one or more perinatal assessment Furthermore, compared with those who did take part in
our study, these women were more likely to be born
outside Australia and to have used cannabis more often
1000 participants in VAHCS as a young adult.
128 no longer active in VAHCS 36 pregnancies were conceived around the time of the
87 declined
39 lost to follow-up
wave 9 VAHCS interview, and these were excluded from
2 died our analysis to clearly distinguish VAHCS wave 9 mental
872 screened for VIHCS entry health measures from perinatal mental health measures,
resulting in an analysis sample of 384 women and
406 ineligible (not pregnant)
564 pregnancies. Of women in the analysis sample,
Recruitment

466 woman had 744 pregnancies 256 (67%) had not had a previous child and 128 (33%)
had had a previous pregnancy resulting in a livebirth.
12 pregnancies were ineligible (death in utero)*

465 women were eligible for VIHCS, with


732 pregnancies
Participants Frequency
(n=384) (95% CI)
95 pregnancies were reported retrospectively*
37 pregnancies excluded (did not consent to Adolescence
participate)*
Persistence of mental health problems
398 women with 600 pregnancies gave consent No waves 187 49% (4354)
and participated in VIHCS One wave 68 18% (1422)
Two or more waves 129 34% (2938)
Parental divorce or separation 75 20% (1624)
36 pregnancies excluded
Risky behaviour
384 women with 564 pregnancies available
for analysis
Antisocial behaviour 112 29% (2534)
Analysis sample

289 women with 388 pregnancies at Alcohol misuse 135 35% (3040)
32 weeks of gestation
Daily cigarette smoking 84 22% (1826)
347 women with 499 pregnancies at
8 weeks after birth Regular cannabis use 28 7% (510)
368 women with 526 pregnancies at Young adulthood
12 months after birth
Persistence of mental health problems
No waves 211 55% (5060)
32 weeks of gestation

306 women with 414 pregnancies were assessed One wave 109 28% (2433)
128 pregnancies identified after eligibility
window Two or more waves 64 17% (1321)
24 pregnancies, mother was too busy Mental health problems, by wave
34 pregnancies, unable to contact mother
Wave 7* 71 18% (1422)
Wave 8 90 23% (1928)
1 pregnancy lost to follow-up Wave 9 44 12% (815)
8 weeks after birth

(withdrew consent) Wave 9, major depressive disorder 45 12% (815)


359 women with 527 pregnancies were assessed Wave 9, anxiety disorder 45 12% (815)
50 pregnancies identified after eligibility Continuity of mental health problems
window
11 pregnancies, mother was too busy No disorder 131 34% (2939)
11 pregnancies, unable to contact mother Adolescent disorder only 80 21% (1625)
Young adult disorder only 56 15% (1118)
12 months after birth

5 pregnancies lost to follow-up (3 women Both adolescent and young adult 117 30% (2635)
withdrew consent) disorder

382 women with 560 pregnancies were assessed Estimates were obtained from imputed data for the rst child included in VIHCS.
12 pregnancies, mother was too busy Frequency estimates were calculated from imputed percentage estimates and
22 pregnancies, unable to contact mother total number of participants. CIS-R=Revised Clinical Interview Schedule.
GHQ=12-item General Health Questionnaire. CIDI=Composite International
Diagnostic Interview. *Mean age 207 years; mental health problems dened as
Figure 2: Sampling and ascertainment of pregnancies during the study period
CIS-R score 12. Mean age 241 years. Mental health problems dened as GHQ
Participants were recruited from the VAHCS population and screened for eligibility in VIHCS. VAHCS=Victorian
score 3. Mean age 291 years. Dened with CIDI.
Adolescent Health Cohort Study. VIHCS=Victorian Intergenerational Health Cohort Study. *128 pregnancies were
excluded, 12 because of death in utero, 79 because the pregnancy was reported retrospectively, and 37 because Table 1: Common mental disorders and health risks in adolescence and
consent was not given. Of these exclusions, 60 women were retained in VIHCS because of another pregnancy. young adulthood before conception in women with at least
36 pregnancies excluded from the analysis sample because the wave 9 assessment took place around the one pregnancy during the study period
estimated point of conception.

878 www.thelancet.com Vol 386 August 29, 2015


Articles

Of these 128 women, 82 had one previous livebirth, two-thirds of all participants reported a mental health
40 had two previous livebirths, four had three previous problem either during adolescence or young adulthood.19
livebirths, and two women had ve and six previous Table 2 shows data for depressive symptoms reported
livebirths each. On average, participants were 32 years during the three perinatal assessments. Of 256 primi-
old when their rst child included in VIHCS was born parous mothers, 56 (22%, 95% CI 1627) reported
(SD 2; range 2936). depressive symptoms at any perinatal assessment and
Table 1 shows data for preconception mental health 20 (8%, 411) had signs of depression or anxiety at two or
disorders and health risks in the 384 women included in three assessments. In the 265 multiparous women with
the analysis sample. 80 women (21%, 95% CI 1625) 308 second and subsequent pregnancies, depressive
reported an adolescent disorder only, 56 (15%, 1118) had symptoms were reported during 53 (17%, 1222)
a young adult disorder only, and 117 (30%, 2635) had pregnancies at any perinatal assessment and during
both an adolescent and young adult disorder. Overall, 14 (5%, 27) pregnancies at two or three assessments.

Primiparous (256 pregnancies Multiparous (308 pregnancies Total (564 pregnancies


in 256 women) in 265 women) in 384 women)
Perinatal depressive symptoms 56 (22% [1627]) 53 (17% [1222]) 109 (19% [1623])
Antenatal 34 (13% [917]) 32 (10% [714]) 65 (12% [915])
Postnatal (8 weeks) 26 (10% [615]) 17 (5% [38]) 43 (8% [510])
Postnatal (12 months) 22 (9% [512]) 22 (7% [410]) 44 (8% [610])
Persistence of depressive symptoms
None 200 (78% [7384]) 255 (83% [7888]) 455 (81% [7784])
One assessment 36 (14% [919]) 38 (12% [817]) 75 (13% [1017])
Two or three assessments 20 (8% [411]) 14 (5% [27]) 34 (6% [48])
Score on EPDS (range 030)
Perinatal 47 (39 [4451]) 43 (36 [4045]) 45 (37 [4347])
Antenatal 49 (39 [4455]) 47 (36 [4351]) 48 (37 [4452])
Postnatal (8 weeks) 51 (37 [4656]) 40 (35 [3644]) 45 (36 [4248])
Postnatal (12 months) 42 (40 [3748]) 41 (36 [3645]) 41 (38 [3845])

Data are either number of pregnancies (% [95% CI]) or mean (SD [95% CI]). Frequency estimates were calculated from imputed percentage estimate and total number of
pregnancies. Perinatal depressive symptoms dened as an EPDS score 10. Perinatal period was from 32 weeks of gestation to 8 weeks after birth. Antenatal assessment was
at 32 weeks of gestation. Postnatal assessments were at 8 weeks and 12 months after birth. EPDS=Edinburgh Postnatal Depression Scale.

Table 2: Perinatal depressive symptoms overall and by parity

Pregnancies (n=564) Antenatal (n=65) Postnatal Postnatal Perinatal (n=109)


8 weeks (n=43) 12 months (n=44)
Persistence of mental health problems
Adolescent phase (waves 26)
No waves 277 (49%) 21 (8% [411]) 13 (5% [28]) 11 (4% [26]) 34 (12% [816])
One wave 99 (18%) 11 (11% [418]) 8 (8% [215]) 7 (7% [213]) 19 (20% [1029])
Two or more waves 188 (33%) 33 (18% [1224]) 22 (11% [716]) 26 (14% [819]) 55 (29% [2236])
Young adult phase (waves 79)
No waves 320 (57%) 16 (5% [28]) 14 (4% [27]) 12 (4% [26]) 34 (11% [714])
One wave 155 (28%) 23 (15% [821]) 17 (11% [517]) 16 (10% [515]) 37 (24% [1632])
Two or more waves 88 (16%) 27 (30% [1942]) 12 (13% [621]) 16 (18% [1027]) 38 (43% [3055])
Continuity of mental health problems
No disorder 201 (36%) 8 (4% [17]) 7 (4% [17]) 4 (2% [04]) 16 (8% [412])
Adolescent disorder only 120 (21%) 7 (6% [111]) 7 (6% [111]) 8 (7% [211]) 18 (15% [822])
Young adult disorder only 76 (13%) 12 (16% [726]) 6 (8% [115]) 7 (9% [216]) 17 (23% [1333])
Adolescent and young adult disorder 168 (30%) 37 (22% [1530]) 23 (13% [819]) 25 (15% [921]) 57 (34% [2543])

Data are number of pregnancies (% [95% CI]). Frequency estimates were calculated with imputed percentage estimate and total number of pregnancies. Perinatal depressive
symptoms dened as an EPDS score 10. Perinatal period was from 32 weeks of gestation to 8 weeks after birth. Antenatal assessment was at 32 weeks of gestation.
Postnatal assessments were at 8 weeks and 12 months after birth. EPDS=Edinburgh Postnatal Depression Scale.

Table 3: Perinatal depressive symptoms according to preconception mental health problems

www.thelancet.com Vol 386 August 29, 2015 879


Articles

Of 44 pregnancies with postnatal depressive symptoms at


Partly adjusted model* Fully adjusted model*
12 months, 40 (91%, 81100) had a preconception history
Odds ratio (95% CI) p value Odds ratio (95% CI) p value of mental health problems. Of 109 pregnancies during
Continuity of mental health problems which perinatal depressive symptoms were reported at
No disorder 1 <00001 1 <00001 one or more assessments, 93 (85%, 7892) had a
Adolescent disorder only 232 (077700) 200 (069585) preconception history of mental health problems, with
Young adult disorder only 542 (1751683) 457 (1541353) 57 (52%, 4263) having mental health problems during
Adolescent and young adult disorder 1036 (4102617) 836 (3342087) both adolescence and young adulthood. Conversely, of
Perinatal assessment point 455 pregnancies in which no perinatal depressive
Antenatal 1 002 1 002 symptoms were reported, 271 (60%, 5564) had a pre-
Postnatal (8 weeks) 052 (029090) 052 (029090) conception history of mental health problems, with
Postnatal (12 months) 054 (031091) 054 (031091) 111 (24%, 2029) having mental health problems in both
Pregnancy history adolescence and young adulthood. Mean EPDS scores
Primiparous 156 (088275) 013 were higher at every perinatal assessment for women
Previous miscarriage or stillbirth 145 (072294) 030 with a preconception history of mental health disorders
Previous termination 330 (137798) 0008 than for those with no history, particularly when the
Parental divorce or separation 197 (088444) 010 depressive disorder continued from adolescence to young
Adolescent risky behaviours adulthood (appendix p 7).
Antisocial behaviour 063 (029140) 026
Table 4 presents data for multilevel models of the extent
Alcohol misuse 092 (044192) 082
to which preconception mental health problems predicted
Daily cigarette smoking 107 (043268) 089
perinatal depressive symptoms at 32 weeks of gestation,
8 weeks after birth, and 12 months after birth. Women
Regular (at least weekly) cannabis use 046 (011196) 029
with a history of mental health disorders in both
Variance
adolescence and young adulthood had the strongest risk
Woman 310 (SE 116) 228 (SE 116)
of perinatal depression, with an adjusted eight-fold
Pregnancy 073 (SE 086) 097 (SE 103)
increase in odds compared with women with no history
Odds ratios (95% CIs) are calculated from 564 pregnancies in 384 women. Perinatal depressive symptoms dened as of depression (odds ratio 836, 95% CI 3342087).
an EPDS score 10. Antenatal assessment was at 32 weeks of gestation. Postnatal assessments were at 8 weeks and Furthermore, participants who had a mental health
12 months after birth. EPDS=Edinburgh Postnatal Depression Scale. *Models adjusted for perinatal assessment point
(part) and divorce or separation, pregnancy history, and risky behaviours (full). Joint test of signicance.
disorder only in young adulthood had odds of perinatal
depression four times higher (457, 1541353) than
Table 4: Preconception predictors of perinatal depressive symptoms women with no history of mental health problems.
However, those with an adolescent mental health disorder
In all 384 women, 98 (25%, 2130) reported depressive only had no increased risk of perinatal depression (200,
symptoms at one or more perinatal assessment for one or 069585). Risks for depressive symptoms overall were
more pregnancies. Mean scores on the EPDS did not vary lower at both postnatal assessment points compared with
greatly between the three perinatal assessment points. the antenatal assessment (table 4). Women with a history
Levels of depressive symptoms were similar between of termination of pregnancy had odds of perinatal
primiparous and multiparous pregnancies at the depression more than three times higher than those
antenatal and 12-month postnatal assessments, but at without a previous termination (330, 137798),
8 weeks, women with their rst child reported somewhat independent of other covariates. Adjustment for potential
higher scores on the EPDS (51, 95% CI 4656 vs confounders and mediators had little eect on the overall
40, 3644; table 2). Miscarriages were reported before associations with preconception mental health disorders.
127 pregnancies (23%, 95% CI 1926) by 93 women, and Further analyses (appendix pp 89) showed similar
terminations were reported before 75 pregnancies (13%, patterns of ndings for the EPDS modelled as a
1016) by 57 women. continuous variable and for persisting perinatal
Table 3 shows the prevalence of depressive symptoms depressive symptoms.
across the perinatal period, according to preconception Figure 3 shows the continuity of mental health disorders
mental health problems. Of 65 pregnancies with antenatal from adolescence and young adulthood to the perinatal
depressive symptoms, 57 (87%, 95% CI 7897) had a period, by displaying the estimated mean EPDS scores
previous history of mental health problems during and probabilities of perinatal depression derived from the
adolescence or young adulthood, with 37 (57%, 4372) of multilevel models. It highlights the greater level of risk for
these having mental health problems during both perinatal depressive symptoms associated with a recent
adolescence and young adulthood. Of 43 pregnancies history of mental health disorders in young adulthood.
with postnatal depressive symptoms at 8 weeks, 36 (83%, We repeated our multiple imputation analyses with
7097) had a previous history of preconception mental complete case analyses (appendix pp 1013). Prevalence
health problems and 23 (52%, 3470) had mental health and mean estimates were similar to those under multiple
problems during both adolescence and young adulthood. imputation. The association between preconception

880 www.thelancet.com Vol 386 August 29, 2015


Articles

mental health disorders and perinatal depressive EPDS score


symptoms, although reduced, was consistent with the 70
estimates under multiple imputation. 65
60
Discussion 55
Perinatal depressive symptoms were reported in one in 50

Mean
13 pregnancies among women with no history of 45
depressive symptoms before conception. By comparison, 40
35
women with mental health problems persisting from
30
adolescence to young adulthood reported depressive
25
symptoms in more than one in three pregnancies.
In total, a history of mental health problems before 0
conception was noted in 85% of pregnancies in which
High perinatal depressive symptoms
perinatal depressive symptoms were evident. Most 020 No disorder Young adult disorder only
episodes were either a recurrence or continuation of Adolescent disorder only Adolescent and young
mental health problems that began before conception. adult disorder

This nding, in part, reects the prevalence of mental 015

health disorders in the whole population, in that


Probability

two-thirds of participants had at least one episode of 010


mental health problems in the 1420 years before the
index pregnancy. Nevertheless, these population
frequencies for common mental health disorders accord 005

with those reported for this age group elsewhere.18,34 Our


ndings further challenge the view that perinatal 0
depression is distinct from previous mental health 32 weeks gestation 8 weeks after birth 12 months after birth

problems. Figure 3: Continuity of mental health disorders from adolescence and young adulthood to the perinatal
Despite strengths in the prospective design of our period in 564 pregnancies of 384 participants
study and multiple points of assessment across (Upper) Estimated mean EPDS scores across adolescence and young adulthood. Error bars show 95% CI. (Lower)
Probability of depressive symptoms continuing from adolescence and young adulthood to the perinatal period.
adolescence and young adulthood in VAHCS, we might
Perinatal depressive symptoms dened as a score 10 on the EPDS. EPDS=Edinburgh Postnatal Depression Scale.
still have missed some earlier cases of depressive
disorder. Adolescent assessments focused on the Error in outcome measurement is a further potential
previous 7 days, potentially missing brief episodes that limitation. The EPDS is the most widely used
might have arisen between follow-up assessments. assessment of perinatal depressive symptoms but is
Moreover, assessments in young adulthood also focused likely to capture anxiety symptoms as well.28 Using a
on the period immediately before interview, with the threshold of 10 or more on the EPDS, we recorded
exception of the CIDI assessments at wave 9, which perinatal depression during at least one of the three
covered the previous 12 months. Thus, we might have perinatal assessments in around 20% of pregnancies.
underestimated the prevalence of perinatal depression Although this proportion is similar to those reported in
after earlier mental health problems before conception. meta-analyses of prevalence, the point estimates are
Although overall attrition was low in VIHCS, lower than in most studies using the EPDS in
non-response is a further limitation of our study. 87% community samples.29 Telephone administration of the
of the original sample from VAHCS were still active EPDS seems likely to be one explanation. A threshold of
participants at the start of our study, and they diered 10 or more generally corresponds to minor depression,
from non-responders mainly in the proportion who but by telephone administration, this score is the
were not born in Australia. However, 14% of women optimum cuto point for screening for depressive
who had pregnancies during the study window did not disorder.30 Diminished participation of women with
participate. This group diered from those who did mental health disorders before conception is another
take part not only in being more likely to be born possible contributor to lower prevalence estimates.
outside Australia but also in patterns of cannabis use Finally, women becoming pregnant at this time of life
before conception and, most importantly, previous (mean age 32 years) might have a better mental health
mental health problems. Use of multiple imputation prole than those who are pregnant at a younger or
should have reduced biases arising from missing older age.35 Although such factors might aect the
interviews in all phases of the study but it cannot strength of associations, the persistence of strong
address the dierential response in the perinatal phase associations when perinatal depression was dened
of the study. This drawback might have contributed to either as symptoms persisting across two assessments
some misspecication of associations with disorders or as a continuous variable suggests that our ndings
before pregnancy. are robust.

www.thelancet.com Vol 386 August 29, 2015 881


Articles

an episode of perinatal depression, suggesting that other


Panel: Research in context potentially modiable factorseg, preconception or
Systematic review perinatal stressorsare relevant.42,43 Groups of women
In April, 2015, we searched Medline, PubMed, and Google Scholar with combinations of could possibly be identied who are at augmented risk
terms that included maternal, perinatal, antenatal, postnatal, antepartum, for depressive symptoms continuing into pregnancy.
postpartum, adolescent, youth, young adult, mental disorder, depression, Factors such as experience of young adult social
emotional, depressive, natural history, and antecedent. We searched for published disadvantage, intimate partner violence, or previous loss
articles and did not restrict by language. Nearly all previous community-based studies of of pregnancy might aect this risk.4446
perinatal depressive symptoms began after recognition of a pregnancy. Thus, they relied The eects of psychosocial interventions for the
on recall to identify previous episodes, with the exception of one study39 that used clinical prevention of maternal perinatal depression have so far
case records in the period immediately before conception. been mixed. Most have focused on debrieng or coping
skills after childbirth, with proactive home visiting, peer
Interpretation support, and interpersonal psychotherapy seeming the
Most women with perinatal depressive symptoms have had previous episodes of most promising interventions.47,48 The ndings of our
depression and anxiety at levels that would have been of concern to their family doctor. study open two new avenues for prevention of perinatal
However, most women with a history of mental health problems in adolescence or young depression. First, in view of the high prevalence of
adulthood do not report perinatal depressive symptoms, suggesting that important perinatal depression in women with a previous history of
modiers of risk are acting either before pregnancy or during the perinatal period. Clinical mental health problems, greater attention to identifying
and preventive responses to perinatal depression should extend to the time before and targeting this group of womenparticularly those
pregnancy, particularly for women with a history of persisting or recurrent disorders. with a background of persistent or recurrent problems
across the perinatal period could have great merit. Second,
The idea that the postnatal period is a unique risk phase the extent of mental health problems in adolescence and
for mental disorders emerged in the mid-19th century young adulthood suggests that the window of intervention
and was a dominant view until the past couple of should extend to the years before pregnancy. Implemen-
decades.36,37 In recent studies, researchers have questioned tation of eective treatments for youth mental health
whether the perinatal period is truly a time of augmented disorders and addressing factors implicated in recurrence
risk for depression.38 Others have noted that mental before pregnancy seem promising strategies for the
health disorders before pregnancy are a major risk factor prevention of perinatal depression in the future.
for perinatal depression.14,15 In one report of treated Contributors
depression in members of a US health maintenance GCP and CO had the idea for the overall study, and GCP and CC had the
organisation, high rates of continuity of mental health idea for this report. GCP, HR, CC, ES, SH, and JBC undertook the data
analysis. GCP, CC, HR, SB, and ES prepared the rst draft of the report.
disorders were recorded from the period immediately All authors were involved in reviewing analyses and revising the
before pregnancy to antenatal and postnatal phases.39 manuscript.
In our study, with its much longer observation period Declaration of interests
before conception, even greater continuity was shown We declare no competing interests.
between preconception and perinatal mental health Acknowledgments
disorders. Thus, for most women, perinatal depressive Both the Victorian Adolescent Health Cohort Study (VAHCS) and the
symptoms are best regarded as a continuation or Victorian Intergenerational Health Cohort Study (VIHCS) have been
recurrence of problems beginning well before pregnancy. supported by a series of project grants from Australias National Health
and Medical Research Council (NHMRC), the Victorian Health
In view of this degree of continuity, it is perhaps Promotion Foundation, the Colonial Foundation, and Australian Rotary
surprising that this idea has not been reported previously. Health Research and Perpetual Trustees. The Murdoch Childrens
Most studies of perinatal depression have begun after Research Institute is supported by the Victorian Governments
recognition of pregnancy and, therefore, rely on retro- Operational Infrastructure Program. We thank the families who
participated in VAHCS and VIHCS; the study research team involved in
spective reports that underestimate previous mental data collection and management; and Prof Anthony Mann and
health (panel).18 A further explanation for increased Prof Sir Michael Rutter for advice on study design.
continuity of depression could be the growing delay References
in starting a family, particularly in high-income and 1 Wachs TD, Black MM, Engle PL. Maternal depression: a global
middle-income countries, extending the window in threat to childrens health, development, and behavior and to
human rights. Child Dev Perspect 2009; 3: 5159.
which a mental disorder might arise before pregnancy.40 2 OHara MW, Swain AM. Rates and risk of post-partum depression:
Together with an escalation in the prevalence of a meta-analysis. Int Rev Psychiatry 1996; 8: 3754.
adolescent and young adult mental disorders, risks for 3 Parsons CE, Young KS, Rochat TJ, Kringelbach ML, Stein A. Postnatal
depression and its eects on child development: a review of evidence
common mental disorders before conception might from low-and middle-income countries. Br Med Bull 2012; 101: 57.
possibly be higher than in previous generations.41 4 Kingston D, Tough S, Whiteld H. Prenatal and postpartum
Women with a history of mental health disorders maternal psychological distress and infant development:
a systematic review. Child Psychiatry Hum Dev 2012; 43: 683714.
persisting across adolescence and young adulthood had
5 Luoma I, Tamminen T, Kaukonen P, et al. Longitudinal study of
the highest frequencies of later perinatal disorders. Yet, maternal depressive symptoms and child well-being.
even in this group, only about a third of pregnancies had J Am Acad Child Adolesc Psychiatry 2001; 40: 136774.

882 www.thelancet.com Vol 386 August 29, 2015


Articles

6 Surkan PJ, Kennedy CE, Hurley KM, Black MM. Maternal 27 Rowe HJ, Fisher JR, Loh WM. The Edinburgh postnatal depression
depression and early childhood growth in developing countries: scale detects but does not distinguish anxiety disorders from
systematic review and meta-analysis. Bull World Health Organ 2011; depression in mothers of infants. Arch Womens Ment Health 2008;
89: 60715. 11: 10308.
7 Grote NK, Bridge JA, Gavin AR, Melville JL, Iyengar S, Katon WJ. 28 Johnstone SJ, Boyce PM, Hickey AR, Morris-Yates AD, Harris MG.
A meta-analysis of depression during pregnancy and the risk of Obstetric risk factors for postnatal depression in urban and rural
preterm birth, low birth weight, and intrauterine growth restriction. community samples. Aust N Z J Psychiatry 2001; 35: 6974.
Arch Gen Psychiatry 2010; 67: 1012. 29 Figueiredo FP, Balarini BF, Silva APC, et al. P-1465: EPDS by
8 Cooper PJ, Murray L. Course and recurrence of postnatal telephonewhat cuto point use. Eur Psychiatry 2012; 27: 1.
depression: evidence for the specicity of the diagnostic concept. 30 Goldstein H. Multilevel statistical models, 4th edn. Chichester:
Br J Psychiatry 1995; 166: 19195. John Wiley & Sons, 2010.
9 Pitt B. Atypical depression following childbirth. Br J Psychiatry 31 Romaniuk H, Patton GC, Carlin JB. Multiple imputation in a
1968; 114: 132535. longitudinal cohort study: a case study of sensitivity to imputation
10 Bloch M, Daly RC, Rubinow DR. Endocrine factors in the etiology methods. Am J Epidemiol 2014; 180: 92032.
of postpartum depression. Compr Psychiatry 2003; 44: 23446. 32 Rubin DB. Multiple imputation for nonresponse in surveys.
11 Heron J, OConnor TG, Evans J, Golding J, Glover V. The course of New York: Wiley, 1987.
anxiety and depression through pregnancy and the postpartum in a 33 Australian Bureau of Statistics. 3301.0Births, Australia, 2013.
community sample. J Aect Disord 2004; 80: 6573. Oct 23, 2014. http://www.abs.gov.au/ausstats/abs@.nsf/mf/3301.0
12 Lancaster CA, Gold KJ, Flynn HA, Yoo H, Marcus SM, Davis MM. (accessed July 10, 2014).
Risk factors for depressive symptoms during pregnancy: 34 Copeland W, Shanahan L, Costello EJ, Angold A. Cumulative
a systematic review. Am J Obstet Gynecol 2010; 202: 514. prevalence of psychiatric disorders by young adulthood: a prospective
13 Gavin NI, Gaynes BN, Lohr KN, Meltzer-Brody S, Gartlehner G, cohort analysis from the Great Smoky Mountains Study.
Swinson T. Perinatal depression: a systematic review of prevalence J Am Acad Child Adolesc Psychiatry 2011; 50: 25261.
and incidence. Obstet Gynecol 2005; 106: 107183. 35 Woolhouse H, Gartland D, Hegarty K, Donath S, Brown S.
14 OHara MW, Wisner KL. Perinatal mental illness: denition, Depressive symptoms and intimate partner violence in the
description and aetiology. Best Pract Res Clin Obstet Gynaecol 2014; 12 months after childbirth: a prospective pregnancy cohort study.
28: 312. BJOG 2012; 119: 31523.
15 Meltzer-Brody S, Boschloo L, Jones I, Sullivan PF, Penninx BW. The 36 Marc L-V. Trait de la folie des femmes enceintes, des nouvelles
EPDS-Lifetime: assessment of lifetime prevalence and risk factors accouches et des nourrices, et considrations mdico-lgales qui se
for perinatal depression in a large cohort of depressed women. rattachent ce sujet. Paris: JB Balliere et Fils, 1858.
Arch Womens Ment Health 2013; 16: 46573. 37 OHara MW, McCabe JE. Postpartum depression: current status
16 Rich-Edwards JW, Kleinman K, Abrams A, et al. Sociodemographic and future directions. Annu Rev Clin Psychol 2013; 9: 379407.
predictors of antenatal and postpartum depressive symptoms 38 Vesga-Lopez O, Blanco C, Keyes K, Olfson M, Grant BF, Hasin DS.
among women in a medical group practice. Psychiatric disorders in pregnant and postpartum women in the
J Epidemiol Community Health 2006; 60: 22127. United States. Arch Gen Psychiatry 2008; 65: 805.
17 Banti S, Mauri M, Oppo A, et al. From the third month of 39 Dietz P, Williams S, Callaghan W, Bachman D, Whitlock E,
pregnancy to 1 year postpartum: prevalence, incidence, recurrence, Hornbrook M. Clinically identied maternal depression before,
and new onset of depressionresults from the Perinatal during, and after pregnancies ending in live births. Am J Psychiatry
DepressionResearch & Screening Unit study. Compr Psychiatry 2007; 164: 151520.
2011; 52: 34351. 40 Gluckman PD, Hanson MA. Evolution, development and timing of
18 Mott TE, Caspi A, Taylor A, et al. How common are common puberty. Trends Endocrinol Metab 2006; 17: 712.
mental disorders? Evidence that lifetime prevalence rates are 41 Collishaw S, Maughan B, Goodman R, Pickles A. Time trends in
doubled by prospective versus retrospective ascertainment. adolescent mental health. J Child Psychol Psychiatry 2004;
Psychol Med 2010; 40: 899909. 45: 135062.
19 Patton GC, Coey C, Romaniuk H, et al. The prognosis of common 42 LaCoursiere DY, Hirst KP, Barrett-Connor E. Depression and
mental disorders in adolescents: a 14-year prospective cohort study. pregnancy stressors aect the association between abuse and
Lancet 2014; 383: 140411. postpartum depression. Matern Child Health J 2012; 16: 92935.
20 Lewis G, Pelosi AJ, Araya R, Dunn G. Measuring psychiatric 43 Rubertsson C, Wickberg B, Gustavsson P, Radestad I. Depressive
disorder in the community: a standardized assessment for use by symptoms in early pregnancy, two months and one year
lay interviewers. Psychol Med 1992; 22: 46586. postpartum-prevalence and psychosocial risk factors in a national
21 Goldberg DP, Gater R, Sartorius N, et al. The validity of two versions Swedish sample. Arch Womens Ment Health 2005; 8: 97104.
of the GHQ in the WHO study of mental illness in general health 44 Giannandrea SA, Cerulli C, Anson E, Chaudron LH. Increased risk
care. Psychol Med 1997; 27: 19197. for postpartum psychiatric disorders among women with past
22 Donath S. The validity of the 12-item general health questionnaire pregnancy loss. J Womens Health 2013; 22: 76068.
in Australia: a comparison between three scoring methods. 45 Olsson CA, Horwill E, Moore E, et al. Social and emotional
Australas Psychiatry 2001; 35: 23135. adjustment following early pregnancy in young Australian women:
23 WHO. Composite international diagnostic interview, CIDI-auto 2.1: a comparison of those who terminate, miscarry, or complete
administrators guide and reference. Geneva: World Health pregnancy. J Adolesc Health 2014; 54: 698703.
Organization, 1997. 46 Hall M, Chappell LC, Parnell BL, Seed PT, Bewley S. Associations
24 Kessler RC, Andrews G, Mroczek D, Ustun B, Wittchen HU. The between intimate partner violence and termination of pregnancy:
World Health Organization composite international diagnostic a systematic review and meta-analysis. PLoS Med 2014; 11: e1001581.
interview short form (CIDI SF). Int J Methods Psychiatr Res 1998; 47 Austin MP, Frilingos M, Lumley J, et al. Brief antenatal cognitive
7: 17185. behaviour therapy group intervention for the prevention of
25 Mott TE, Silva PA. Self-reported delinquency: results from an postnatal depression and anxiety: a randomised controlled trial.
instrument for New Zealand. Aust N Z J Criminol 1988; 21: 22740. J Aect Disord 2008; 105: 3544.
26 Cox JL, Holden JM, Sagovsky R. Detection of postnatal depression: 48 Dennis C, Dowswell T. Psychosocial and psychological
development of the 10-item Edinburgh postnatal depression scale. interventions for preventing postpartum depression.
Br J Psychiatry 1987; 150: 78286. Cochrane Database Syst Rev 2013; 2: CD001134.

www.thelancet.com Vol 386 August 29, 2015 883

You might also like