Psychoneuroendocrinology (2015) 56, 6278

Available online at www.sciencedirect.com

ScienceDirect

jou rn al h om epage: www.elsev ier .com/ locate /psyneuen

Prenatal maternal stress predicts stressreactivity at 2 years of age: The IowaFlood

Erin YoKatharSuzann

a Psychosocb DepartmeUniversity, c Departmed Departme

Received 15

KEYWOPrenatalCortisol Cortisol responseHPA axisChild de

Correspo6875 LaSalle

E-mail ad

http://dx.d0306-4530/ Study

ng Pinga, David P. Laplantea, Guillaume Elgbeili a,ina M. Hillererb, Alain Bruneta,c, Michael W. OHarad,e Kinga,c,

ial Research Division, Douglas Mental Health University Institute, Verdun, QC, Canadant of Obstetrics and Gynaecology, Salzburger Landeskrankenhaus (SALK), Paracelsus MedicalSalzburg, Austriant of Psychiatry, McGill University, Montreal, QC, Canadant of Psychology, University of Iowa, Iowa City, IA, United States

October 2014; received in revised form 19 February 2015; accepted 20 February 2015

RDS stress;reactivity;stress;;velopment

Summary Prenatal maternal stress (PNMS) predicts psychosocial development in offspring. Ithas been hypothesized that during PNMS, glucocorticoids pass the placenta, reaching the foetus,leading to a long-term reprogramming and dysregulation of the foetal hypothalamic-pituitary-adrenal (HPA) axis. However, results are inconsistent across PNMS studies. One problem may bethe confounding of objective degrees of hardship due to the stressor and subjective degrees ofdistress in the mother. The present study investigated the association between objective andsubjective PNMS due to a natural disaster, the June 2008 Iowa oods, and stress reactivity inthe offspring at 2 years of age. Women who were pregnant during the oods were recruited,on average, within three months of the oods and their stress levels assessed. Mothers and theirtoddlers (n = 94 dyads) participated in a brief mothertoddler separation to induce physiologicalstress responses in the offspring. Salivary cortisol samples were collected four times duringthe procedure. We computed absolute change in cortisol (baseline to 20-minute post-stressor;baseline to 45-minute post-stressor) and Area Under the Curve with respect to increase andground (AUCi; AUCg). Objective and subjective PNMS were positively correlated with AUCi,as was timing in gestation: the later in pregnancy the exposure occurred, the greater thecortisol increase. Controlling for objective hardship and other covariates, sex-by-subjectivePNMS interactions showed a signicant and positive association between subjective PNMS and

nding author at: McGill University & Principal Investigator, Douglas Mental Health University Institute, Perry Pavillion, E-3131, Blvd., Verdun, Quebec H4H 1R3, Canada. Tel.: +1 514 761 6131 x2353.dress: suzanne.king@mcgill.ca (S. King).

oi.org/10.1016/j.psyneuen.2015.02.015 2015 Elsevier Ltd. All rights reserved.

Prenatal maternal stress and toddler stress reactivity 63

Absolute Increase (45 min) and AUCi in females only, with little effect in males. These resultssuggest that PNMS leads to long-term alterations in the functioning of the HPA axis, evident asearly as 30-months of age. 2015 Elsevier Ltd. All rights reserved.

1. Introd

Prenatal m(Mulder etchosocial dchildren (OPNMS in Prof the abovof subjectcomes (Da(Laplante functioningsity (Dancasuch as nthese resuinfant deveadolescencdifferentia

The groative outcattention hypothalamcated in Pcommonly (Hunter etfoetus via et al., 199itively coris normallhydroxysteenzyme thaevidence omaternal s

Studies the develoimpact prerodents anwith highecontrols (C2003; Ematal stress ecortisol levstress, andand diurnaet al., 200cortisol lev

PNMS areactivity iand nonhumrelease moet al., 198Coe et al.,exposed tolower or b

li et al., 1994; Emack et al., 2008). In humans, dif- expeA axels ationn stStreEntri

201 levetisolted

2009 dam or noffs

sugoffs

nd, hi

ing to afis futer eted t off. Withaveratiot thaan ing nshing. Tay pr

pre et are a

preraturddleiatesubje

natudimed ofs to

that a pge, uction

aternal stress (PNMS) affects birth outcomes al., 2002), and subsequent cognitive and psy-evelopment in infants (Buitelaar et al., 2003) andConnor et al., 2003). Studying disaster-relatedoject Ice Storm, our group has replicated manye-mentioned ndings. For example, the severityive and/or objective PNMS predicts birth out-ncause et al., 2011), more immature play styleset al., 2007), decits in cognitive and language

(Laplante et al., 2008), higher rates of obe-use et al., 2012), and minor physical anomalies,gerprint asymmetry (King et al., 2009). Overall,lts suggest that PNMS has long-lasting effects onlopment that persist into late childhood and earlye, and that objective and subjective PNMS havel effects depending on the outcome of interest.wing body of literature linking PNMS to neg-omes has lead researchers to focus theiron the biological mechanisms of PNMS. Theicpituitary-adrenal (HPA) axis is often impli-NMS effects since altered HPA axis activity isassociated with many of the observed outcomes

al., 2011). Maternal stress is transmitted to thehigh levels of glucocorticoids (GCs) (Barbazanges6); maternal and foetal cortisol levels are pos-related (Gitau et al., 2001). While the foetusy protected from maternal cortisol via 11-roid dehydrogenase-2 (11-HSD2), a placentalt converts cortisol to its inactive form, cortisone,f 11-HSD2 downregulation in times of increasedtress has been reported (Mairesse et al., 2007).have investigated the impact of PNMS onpment of the foetal HPA axis, and how thissents itself throughout postnatal development. Ind nonhuman primates, PNMS has been associatedr basal levels of GCs in offspring, compared tolarke et al., 1994; Fameli et al., 1994; Coe et al.,ck et al., 2008). In humans, prenatal and postna-xposure, in the form of higher maternal morningels, greater worries during pregnancy, parenting

PTSD, have been associated with higher basall cortisol levels in infants and children (Gutteling4; Saridjan et al., 2010), as well as lower basalels (Yehuda et al., 2005).lso affects the HPA axis with respect to stressn animal offspring (Weinstock et al., 1992); ratsan primates exposed to maternal stress in utero

re GCs in response to stress than controls (Fride6; Takahashi and Kalin, 1991; Clarke et al., 1994;

(Fameferentthe HPsol levvaccinreunioSocial 2008; et al.,highernal corassociaet al.,2011),2013),els in resultsing of acrossologieseld.

Timfound HPA axtrimesassocialescen(2008)spring to alteopmen(Sandmexaminrelatiooffspriings mseen inGuastiexposustudy.

Theof liteand toferentboth swith athese exposetunitieeventsship torace, a 2003). However, offspring whose mothers were stress prenatally have also been found to exhibitlunted GC levels in response to stressful stimuli

ently conta womansciated witrimental paradigms have been used to activateis, with the aim of eliciting a change in corti-in infants (noise burst, arm restraint, bathing,, still-face procedure, and maternal separation-ress), children (vaccination), and adults (Trierss Test) (Gutteling et al., 2004; Brennan et al.,nger et al., 2009; Grant et al., 2009; Tollenaar1; OConnor et al., 2013). From these studies,ls of prenatal anxiety, psychosocial stress, mater-

levels, and pregnancy-related anxiety have beenwith increases (Brennan et al., 2008; Entringer; Grant et al., 2009), decreases (Tollenaar et al.,pened or blunted changes (OConnor et al.,o change (Gutteling et al., 2004) in cortisol lev-pring in response to stress. Collectively, thesegest that PNMS leads to dysregulated function-prings HPA axis. Moreover, the inconsistenciesings, possibly attributable to differing method-ghlight the need for further research in this

of PNMS exposure and offspring sex have beenfect the association between PNMS and offspringnctioning. Huizink et al. (2008) reported secondxposure to the Chernobyl disaster of 1986 waswith higher basal levels of salivary cortisol in ado-spring, as was conrmed by Van den Bergh et al.h respect to sex differences, while female off-

commonly been reported to be more susceptiblens in their physiological and behavioural devel-n males as a result of prenatal stress exposure

et al., 2013), little literature exists specicallythe moderating effect of offspring sex on thep between PNMS and cortisol reactivity in humanhis is an important area of study as related nd-ovide a possible mechanism for the sex differencevalence rates of psychiatric illness (Fernandez-l., 2012). As such, the effects of timing of PNMSnd offspring sex were investigated in the present

sent study aimed to build on the existing bodye by examining the association between PNMSr cortisol reactivity. The Iowa Flood Study dif-

itself from other studies in that it examinesctive and objective levels of PNMS associatedral disaster, and subsequent in utero effects thatnsions of stress can have on the development offspring. Natural disasters provide unique oppor-

study PNMS because they are independent randomly distribute objective degrees of hard-opulation, irrespective of socioeconomic status,and/or maternal characteristics, thereby inher-

rolling for external factors that may inuence

predisposition to stress. Examining stress asso-h a natural disaster also allows for the study

64 E. Yong Ping et al.

of subjective distress while controlling for the objectivedegree of hardship to which the pregnant woman wasexposed.

The present study examined the role of objective andsubjective PNMS caused by the Iowa ooding of 2008 onstress reacIowa experIn total, 24their homein U.S. histand/or objof toddlersGiven thathyper-cortthat this stutive and sudirection ospring sex also tested

2. Metho

2.1. Part

This projeclongitudinaassociated opment. Mfrom an onmaternal pcomes. A d(Nylen et afrom threeand Blackhwere 18 yeEnglish-spe15, 2008).

In totalpleted theeligibility cTen dyads insufcientinvalid cor

Ninety-fnal sampthese moth(M = 29.3; married (Nried and limarried) (Nof Caucasiacan Indian Islander (Nor Hispanicwas found 3.2% indica16% betweto $50,000greater thauniversity or doctorat

technical, or associates degree, while 16% had completedsome college, and 3.2% a high school diploma.

Toddlers had been exposed to the ood at all trimesters:rst trimester (N = 24; 25.5%), second trimester (N = 37;39.4%), and third trimester (N = 33; 35.1%). At birth, the sam-

todd.2; S

in ase, 4

loo

ruitmetedmogre ahe monnaIowal, 66still pfter

3 m

Obtimases re t, losssion re tse onsenccouonnae tha

thatraliaeightscalprese. Bec

skewmaliz

Subrs sunt Sc

scnt tol, anms rded xtremscri

in ret stu

greatribul logtivity in a sample of 2 year old toddlers. In 2008,ienced its worst ooding in more than 50 years.

people were killed and 38,000 were driven froms; the ood was ranked among the top 10 disastersory. It was hypothesized that greater subjectiveective PNMS would predict greater dysregulation cortisol response to maternal separation stress.

past research has found evidence of hypo- andisol response patterns to stress provocation, anddy is the rst to test the relative effects of objec-bjective PNMS, we did not hypothesize a specicf cortisol change. The moderating effects of off-and the timing of in utero ood exposure were.

ds

icipants

t is part of the Iowa Flood Study: a prospectivel study that is investigating the effects of PNMSwith the Iowa ooding of 2008 on child devel-others were recruited into the Iowa Flood Studygoing study at the University of Iowa investigatingsychosocial characteristics and pregnancy out-escription of that study can be found elsewherel., 2013). Additional women were also recruited

severely ood-affected counties: Linn, Johnson,awk. Women were eligible to participate if theyars of age or over at the time of recruitment,aking, and pregnant at the time of the ood (June

, 104 mothertoddler dyads successfully com- protocol outlined in this paper and met allriteria for the current project as noted above.were removed from the sample due to providing

quantities of toddler saliva for assay (n = 9) ortisol assay results (n = 1).our women and their toddlers comprised thele. At recruitment into the Iowa Flood Study,ers ranged in age from 18 to 41 years of ageSD = 5.1) and described their marital status as

= 81; 86.2%), divorced (N = 2; 2.1%), not mar-ving with partner (N = 2; 2.1%), or single (never

= 9; 9.6%). Our sample consisted predominatelyn women (N = 87; 92.6%), in addition to Ameri-or Alaskan Native (N = 1; 1.1%), Asian or Pacic

= 3; 3.2%), Black/African descent (N = 2; 2.1%), of any origin (N = 1; 1.1%). Total family incometo range from less than $10,000 to over $70,000:ted a net family income of less than $10,000,en $10,000 to $30,000, 17% between $30,001, 28.7% between $50,001 to $70,000, and 35.1%n $70,000. Over half of the sample (68%) was

educated, having completed a bachelor, master,e degree, 12.8% of the women had a vocational,

ple of (M = 39rangedof the

2.2. F

At reccompland deexposuing. Tquestiof the In totawhile naire awithin

2.2.1.We esresponexposuthreatdimenexposuresponand cointo aquestibelievdictiveof Auswas wof our in the pointsitivelyto nor

2.2.2.Motheof Eve22-itemrelevaarousasymptoresponto etion de7 dayspresenscoresthe disnaturadata.lers ranged in gestation age from 35 to 42 weeksD = 1.2). At the time of the assessment toddlersge from 29 to 34 months (M = 29.9; SD = 1.1), and8 were female (51.1%).

d-related (predictor) variables

ent into the Iowa Flood Study, participants a series of questionnaires about psychosocialraphic characteristics, and about their objectivend subjective distress levels related to the ood-ajority of women completed the ood-relatedires within three months of exposure to the peak

ooding (M = 9.7 weeks; range 4.724.9 weeks). women answered the recruitment questionnaireregnant, and 28 women answered the question-

delivery; moreover, 24 of the 28 women answeredonths of giving birth.

jective PNMS (IF100)ted objective maternal hardship using mothersto questions associated with four categories ofhat have been used in other disaster studies:, scope, and change (Laplante et al., 2007). Eachwas scored on a scale of 025, ranging from noo high exposure (see Table 1 ). Points for eachn each item were attributed following discussionsus among the IF100 committee members, takingnt each participants write-in comments on theires. Because there was no theoretical basis tot any one of the four dimensions was more pre-n another, and on the basis of McFarlanes studyn reghters (McFarlane, 1988), each dimensioned equally and summed to obtain the total scoree: The Iowa Flood 100 (IF100). Total IF100 scoresnt sample ranged from 0-50 out of a possible 100ause the distribution of objective PNMS was pos-ed, a natural log transformation was conductede the data.

jective PNMS (IES-R)bjective distress was assessed using the Impactale-Revised (IES-R; Weiss and Marmar, 1997). Theale describes symptoms from three categories

PTSD: intrusions (thoughts and images), hyper-d avoidance. Scale items were modied to reectelative to the Iowa ooding disaster. Participantson a ve-point Likert scale, from not at allely, the extent to which the item in ques-

bed symptoms experienced during the precedingsponse to the oods. The total score range in thedy was 060 out of a possible 88 points; IES-Rter or equal to 22 indicate possible PTSD. Becausetion of subjective PNMS was positively skewed, a

transformation was conducted to normalize the

Prenatal maternal stress and toddler stress reactivity 65

Table 1 Scoring of the Iowa Flood 100 (IF100).

1 VariableName

Max. points Label Scoring rules

THREAT 25 Sum(Hurt, Hurtso, Danger, Drown, waterlevel, Shocks,Bridge, Drink, Sewage, Food, Isolation, Otherdanger, time)

1.1 Hurt 1 Were youphysicallyhurt?

Yes = 1, no = 0

1.2 Hurtso 1 Someoneclose to yougotphysicallyhurt?

Yes = 1, no = 0

1.3 Danger 1 Were you inany kinds ofdangers?

Dangers = sum (drown, waterlevel, shocks, building, bridge,drink, sewage, food, isolation, otherdanger), recode1-highest to 1

1.4 Drown 2 Were you indanger ofdrowning?

Yes = 2, no = 0

1.5 Waterlevel 4 Were you indanger ofincreasingwater level?

Yes = 4, no = 0

1.6 Shocks 1 Were you indanger ofelectricalshocks?

Yes = 1, no = 0

1.7 Bridge 1 Were you indanger ofcollapse ofbridges?

Yes = 1, no = 0

1.8 Drink 3 Were you indanger oflack of safedrinkingwater?

Yes = 3, no = 0

1.9 Sewage 1 Were you indanger ofexposure toraw sewage?

Yes = 1, no = 0

1.10 Food 1 Were you indanger oflack of food?

Yes = 1, no = 0

1.11 Isolation 1 Were you indanger ofisolation?

Yes = 1, no = 0

1.12 Otherdanger 1 Were you indanger ofotherdangers?

Yes = 1, no = 0

1.13 Time 7 How muchtime wereyou given toleave home?

RECODE time (1 = 7) (2 = 4) (3 = 2) (4 = 1)

66 E. Yong Ping et al.

Table 1 (Continued )

2 LOSS 25 Sum(Damage, Heirloom, Property, Investment, Vehicle,Lossincome, Totalloss, Busiloss, damVSinsur)

2.1 Damage 4 Was your homedamaged?

RECODE damage (1 = 0) (2 = 1) (3 = 2) (4 = 3) (5 = 4)

2.2 Heirloom 2 Was there aloss of familyheirlooms?

Yes = 2, no = 0

2.3 Property 2 Was there aloss of personalproperty?

Yes = 2, no = 0

2.4 Investment 2 Was there aloss of personalinvestments?

Yes = 2, no = 0

2.5 Vehicle 1 Was yourvehicledamaged?

RECODE vehicle (1 thru 4 = 1)

2.6 Loss income 2 Did youexperience lossof personalincome?

Yes = 2, no = 0

2.7 Total loss 5 How much isyour total loss?

RECODE total loss (1 = 0) (2 = 1) (3 = 2) (4 = 3) (5 = 4) (6 = 5)

2.8 Busiloss 5 The total lossof yourpersonalbusiness?

RECODE busiloss (1 = 0) (2 = 1) (3 = 2) (4 = 3) (5 = 4) (6 = 5)

2.9 DamVSinsur 2 Damage of yourhome VS oodinsurance(home wasdamaged andno oodinsurance get 2points)

IF ((damage gt 1) and (ood insurance = 0))damVSinsur = 1

3 SCOPE 25 Sum(Daysaway, Neighbour, Electricity, phone)

3.1 Daysaway 9 Days peoplewereevacuatedfrom home

RECODE daysaway (1 thru 2 = 1) (4 thru 5 = 2) (6 thru7 = 3) (8 thru 9 = 4) (10 thru 11 = 5) (14 thru 20 = 6) (42thru 60 = 7) (61 thru 92 = 8) (208 thru 722 = 9)

3.2 Neighbour 6 To whatextent wasyour neigh-bourhoodaffected?

RECODE neighbour (1 = 0) (2 = 1) (3 = 2) (5 = 6)

3.3 Electricity 5 Days peopleweredeprivedfromelectricity

RECODE electricity (0.04 thru 0.5 = 1) (1 thru 2 = 2) (3thru 7 = 3) (16 thru 30 = 4) (40 thru 140 = 5)

3.4 Phone 5 Days peopleweredeprivedfrom phone

RECODE phone (0.08 thru 0.5 = 1) (1 thru 3 = 2) (4 thru9 = 3) (15 thru 17 = 4) (42 thru 68 = 5)

Prenatal maternal stress and toddler stress reactivity 67

Table 1 (Continued )

4 CHANGE 25 Sum(Familyapart, Residence, Shelterdays, House,Changework, Changehour, Commute, Prenatalcare,Birthplan, Work, Spousework, Dietchange, skipmeal)

4.1 Familyapart 3 Did yourfamily staytogether?Days apart?

RECODE familyapart (1 thru 5 = 1) (7 thru 10 = 2) (18 thru45 = 3)

4.2 Residence 2 How manytimes wereyou requiredto changeresidencebecause ofood?

RECODE residence (0 = 0) (2 thru 4 = 2)

4.3 Shelterdays 1 Days stay inshelter

RECODE shelterdays (0 = 0) (1 thru 20 = 1)

4.4 House 4 Did you housepeople?(number ofpeople,number ofdays)

House = num of people housed * days house people,RECODE house (1 thru 6 = 1) (8 thru 14 = 2) (24 thru42 = 3) (90 thru 450 = 4)

4.5 Changework 2 Experiencedecrease/increasein physicalwork?

Yes = 2, no = 0

4.6 Changehour 1 Experiencedecrease/increasein work hours?

Yes = 1, no = 0

4.7 Commute 2 Experiencelongercommutingtime?

Yes = 2, no = 0

4.8 Prenatalcare 2 Experiencedifculty inaccessingprenatal care?

Yes = 2, no = 0

4.9 Birthplan 1 Experiencechange inbirth plan?

Yes = 1, no = 0

4.10 Work 1 Was yourplace of workdamaged?

Yes = 1, no = 0

4.11 Spousework 1 Was the placeof work ofyour spousedamaged?

Yes = 1, no = 0

4.12 Dietchange 2 Experiencechange indiet?

Yes = 2, no = 0

4.13 Skipmeal 3 Did you skipmeals andhow many?

RECODE skipmeal (1 thru 2 = 1) (3 = 2) (4 = 3)

Total IF100 100 Sum(Threat, Loss, Scope, Change)

68 E. Yong Ping et al.

2.3. Timing of in utero ood exposure

Timing of in utero ood exposure was dened as the num-ber of days between estimated date of conception and June15th, 2008, the date at which the ooding was recorded topeak; highlater in pretion, we sudate, whictional age

2.4. Matesaliva sam

At age 2 the Universsamples welevels throminutes aftaken appr15 minutessamples tatoddler waaration comthe two-wtoddlers a(M = 1.89 mwere given(Salimetrictongue andabsorptionical tube, samples weDouglas Insthey remai

2.5. Outc

2.5.1. TodSalivary cenzyme imrics. The in11.3 and 9standard dn = 3; 20-mstressor n =the effect Fidell (1989in cortisol stressor) wacute cortiafter admi(AUC) withground (AUcomputed 45-minute out the asssecretion, ues were s

by the amount of time elapsed between the rst and lastsample.

2.6. Control variables

Maruiting s, andnal aer Ho

assef the. Str

strend P

197sesse, 198ts eonnapto

vent et

Chir birstat

Time a

oddlt of

and.

roc

asesstitun infgreeto comentraphhert phars o

andonnaheighs) tor thl lev

ata

descere

les. er ood exposure days indicated ood exposuregnancy. To calculate estimated date of concep-btracted 280 days (40 weeks) from womens dueh was rst calculated using their babies gesta-and date of delivery.

rnal separation protocol and toddlerpling

years, toddlers were assessed in a laboratory atity of Iowa. To assess stress reactivity, four salivare collected from each toddler to assess cortisolughout the procedure: a buffer sample taken 10ter arriving in the laboratory; a baseline sampleoximately 45 minutes after the buffer sample and

before maternal separation (stressor); and twoken 20 and 45 minutes post-stressor (after thes reunited with his or her mother). Maternal sep-menced when a research assistant knocked on

ay mirror, signalling for mothers to leave theirlone in the monitored testing room for 2 minin, SD = 0.49). No instructions on how to leave. Collection of saliva involved placing a Sorbettes, product number 5029) under the toddlers

allowing it to absorb for 45 s to 1 min. After, the Sorbette was removed and placed in a con-and sealed with a cap. Directly after collection,re stored in a 80 C freezer until shipping to thetitute Research Centre, Quebec, Canada, wherened frozen until assayed.

ome variables

dler salivary cortisolortisol levels were assessed via competitivemunoassay (EIA) using kits provided by Salimet-ter- and intra-assay coefcients of variation were.6, respectively. Cortisol values two or moreeviations from the mean (buffer n = 1; baselineinute post-stressor n = 6; and 45-minute post-

3) were winsorized, a technique for reducingof extreme outliers outlined by Tabachnick and). Absolute Cortisol Change (the absolute changelevels from baseline to 20- or 45-minute post-as computed to thoroughly assess participantssol response, as well as their return to baseline,nistration of the stressor. Area Under the Curve

respect to increase (AUCi) and with respect toCg) were also analyzed. AUCi and AUCg wereusing the baseline, 20-minute post-stressor, andpost-stressor cortisol samples collected through-essment and reecting reactive and total cortisolrespectively (Pruessner et al., 2003). All AUC val-tandardized by dividing the total cortisol output

2.6.1.At recincludhabitsmaterit as p

Wetime oof agetion of(LES) aet al.,was asAbidinticipanQuestinal symthe InWatson

2.6.2.Toddleand ge

2.6.3.The timfrom tpercen1200 h1922 h

2.7. P

All phthe InWrittewho aasked assessdemog

Motpresen2 yemotor,questi(e.g., sampleever, fcortiso

2.8. D

First, tion) wvariabternal characteristicsment, we obtained demographic information,ocioeconomic status (SES), smoking and alcohol

relationship status. We computed SES based onnd paternal education and occupation, and codedllingshead criteria (Hollingshead, 1973).ssed the remaining maternal covariates at the

child assessment, when toddlers were 2 yearsessful life events and the participants percep-ss were assessed using the Life Experience Surveyerceived Stress Scale (PSS), respectively (Sarason8; Cohen et al., 1983). Parenting-related stressd using the Parenting Stress Index (PSI; Loyd and5). The presence of help and/or support in a par-nvironment was assessed with the Social Supportire (SSQ; Sarason et al., 1983). Finally, mater-ms of depression and anxiety were measured byory of Depression and Anxiety Symptoms (IDAS;al., 2007).

ld characteristicsth outcomes, such as birth weight, birth length,ional age, were obtained from hospital records.

e of day of cortisol samplingt which baseline cortisol samples were collecteders was recorded for each participant. Sixty-sixsamples were taken before 1200 h, 23% between

1659 h, and 11% of samples between 1700 h and

edure

of the Iowa Flood Study were approved bytional Review Board at the University of Iowa.ormed consent was collected from all mothersd to participate. At recruitment, mothers weremplete a questionnaire packet that included the

of objective and subjective PNMS, as well asic information.toddler dyads were invited to participate in thease of the study when their children reachedf age. The study protocol consisted of cognitive,

stress assessments for the toddlers, self-reportires for the mothers, and physical measurementst and weight) and biological samples (e.g., saliva

aken from both the toddlers and mothers. How-e purpose of this study, only results relating toels and the stressor paradigm are presented.

analysis

riptive analyses (mean, range, standard devia- conducted on outcome, predictor, and controlTo assess acute changes in cortisol levels in

Prenatal maternal stress and toddler stress reactivity 69

response to maternal separation, paired samples t-testswere run between baseline and 20-minute post-stressor and45-minute post-stressor cortisol levels; these were rerun formales and females separately. Independent samples t-testswere also run comparing male and female toddler corti-sol levels aPearson covariables. were condbaseline toline to 45-AUCg. To ahardship (Lentered intively. Theof objectivsubjective and sex wrespectiveoutcome, aates changstress effebaseline coing step 51962). Nexthe outcomwise, into t.05 and extion of posample sizeof timing obetween Paction termExposure, vidually inanalyses wtion of all pThe same regression predictor vwere cond

3. Resul

3.1. Desc

Descriptivevariables asol levels wpost-stressp < .05), astressor sap < .01. Whfemales fr(t(40) = 2ple (t(42) =shown). Inlevels at bpost-stressfemales.

A correlation matrix between predictor and outcomevariables is presented in Table 3. Lower toddler birthweight (r = 0.300, p < .01) and baseline cortisol levels(r = 0.599, p < .01) were related to greater increases incortisol from baseline to 20-minute post-stressor. Higher

ive PNMS scores (r = 0.258, p < .05) and subjectivecores (r = 0.246, p < .05), and later timing of in uterore (rion af dayuch se in

asee

of in Tuteroo theing lain

20-mweretion icate wplainteracode20-m

asee

of tle 4bute expl

timiot srtisof thitioMS ute nal scoreonse

werebing

abss weog-t

(noodel45-mcross the three sampling times. Next, bivariaterrelation coefcients were computed between allHierarchical multiple linear regression analysesucted for each of the four outcome variables:

20-minute post-stressor absolute change, base-minute post-stressor absolute change, AUCi, andssess the effects of PNMS, the level of objectiveogIF100) and subjective distress (LogIES-R) wereto the model during step 1 and step 2, respec-

ordering of these steps allowed for the controle hardship in the assessment of the effect ofdistress. Timing of the ood exposure in uteroere entered into the model in steps 3 and 4,ly, to examine their respective main effects on thend to determine if controlling for these covari-ed the magnitude and signicance of the prenatalcts. For both absolute change outcome variables,rtisol values were entered into the model dur-

to adjust for the law of initial values (Wilder,t, covariates that signicantly correlated withe variable at p < .05 were allowed to enter, Step-he following step of the model (p-value for entryit .10). This method allowed for the conserva-wer considering the limitations imposed by our. Subsequently, to assess the moderating effectsf exposure and toddler sex on the relationshipNMS and cortisol reactivity, we entered an inter-

(IF100 Timing of Exposure, IES-R Timing ofIF100 Sex, or IES-R Sex) into the model indi-

the nal step. Repeated measures mixed modelere also conducted that allowed for the reten-articipants, regardless of missing cortisol values.predictor variables as those entered in the AUCmodels were used for these analyses. All of theariables were tested as xed effects. All analysesucted using SPSS 20.0.

ts

riptives

statistics for outcome, predictor, and controlre presented in Table 2. On average, toddler corti-ere found to increase from baseline to 20-minuteor sampling (by 40% on average; t(78) = 2.093,s well as from baseline to 45-minute post-mpling (by 63% on average), t(81) = 2.796,en split by sex, results remained signicant forom baseline to 20-minute post-stressor sampling.019, p < .05) and to 45-minute post-stressor sam-

2.386, p < .05), but not for males (data notdependent samples t-tests revealed that cortisolaseline, 20-minute post-stressor, and 45-minuteor did not differ signicantly between males and

objectPNMS sexposusecretTime oAUCi, sincrea

3.2. Bchang

Resultssentedof in lated texplainels expof thelevels separato signcovariaand exthe innal mdlers levels.

3.3. Bchang

Resultsin Tab45-minicant, PNMS,were nlute co4.3% oan addtive PN45-minadditioPNMS in respscoresBy procance,femalePNMS l1.4194the mdlers levels. = 0.330, p < .01) were related to greater cortisolcross the assessment relative to baseline (AUCi).

of baseline cortisol sampling was also related tothat earlier testing was associated with a smaller

cortisol.

line to 20-minute post-stressor absolute

the hierarchical multiple regression are pre-able 4a. Objective and subjective PNMS, timing

ood exposure, and toddler sex were unre- 20-minute absolute cortisol change, collectivelyonly 4.3% of the variance. Baseline cortisol lev-ed an additional 36.6% of the variance (p < .01)inute absolute cortisol change: higher baseline

related to smaller increases in 20-minute post-cortisol levels. Given that birth weight was foundntly correlate with the outcome variable, theas allowed to enter Stepwise in the next step,ed an additional 3.7% of the variance. None oftion terms explained additional variance. Thel explained 45.2% of the variance of the tod-inute post-separation absolute cortisol change

line to 45-minute post-stressor absolute

he hierarchical multiple regression are presented. The association between objective PNMS andabsolute cortisol change was marginally signif-aining 4.5% of the variance (p < .1). Subjectiveng of in utero ood exposure, and toddler sexignicantly associated with the 45-minute abso-l change, collectively explaining an additionale variance. Baseline cortisol levels explainednal 4.0% of the variance (p < .1). The subjec-

sex interaction was signicantly related to theabsolute cortisol change and accounted for an6.9% of the variance (p < .05): higher subjectives predicted a greater change in cortisol levels

to stress in females, whereas subjective PNMS not related to cortisol change in males (Fig. 1a).

the interaction to nd the regions of signi-olute cortisol change values between males andre found to signicantly differ when subjectiveransformed levels were equal to or greater thann-transformed IES-R level of 3.1346). Overall,

explained 19.7% of the variance of the tod-inute post-separation absolute cortisol change

70 E. Yong Ping et al.

Table 2 Descriptive analysis of variables.

Variables N Mean Range SD

Buffer cortisol (g/dL) 84 0.105 0.010.48 0.084Baseline c20-minute45-minutePercent chPercent chAbsolute cAbsolute cAUC grounAUC increObjectiveObjectiveSubjectiveSubjectiveTiming of SES Number oAlcohol unPerceivedLife ExperParenting Social supDepressioBirth weig 3Birth lengGestationTime of ba

Note: AUCIDAS = Inve

3.4. Area

Results of sented in greater AU(p < .05): hwith a greThe additexplained exposure evariance (pciated witdid not exthe baselinwise, but PNMS sexand accouas subjecttisol secrelevels were(Fig. 1b). Bnicance, found to dlevels wertransformeOverall, thAUCi.ortisol (g/dL) 85 post-stressor (g/dL) 87 post-stressor (g/dL) 91 ange cortisol (baseline to 20 min post-stressor) 78 ange cortisol (baseline to 45 min post-stressor) 81 hange cortisol (baseline to 20 min post-stressor) 79 hange cortisol (baseline to 45 min post-stressor) 82 d (g/dL) 71 ase (g/dL) 71

PNMS (LogIF100) 94 PNMS (IF100) 94

PNMS (LogIES-R) 94 PNMS (IES-R) 94 exposure in utero (days) 94

92 f cigarettes/day 92 its/week 92

Stress Scale 90 ience Survey 90 Stress Index 90 port 87 n and anxiety (IDAS) 89 ht (g) 92 th (cm) 83 al age at birth (weeks) 92 seline cortisol sampling 94 = Area Under the Curve. PNMS = Prenatal Maternal Stress. SES = Socioentory of Depression and Anxiety Symptoms.

under the curve increase (AUCi)

the hierarchical multiple regression are pre-Table 5a. Greater objective PNMS predictedCi in step 1, explaining 6.7% of the varianceigher objective PNMS scores were associatedater increase in cortisol following the stressor.ion of subjective PNMS in step 2 increasedvariance to 8.8%. When timing of in utero oodntered in step 3 it explained 7.3% of unique

< .05), with exposure later in pregnancy asso-h a greater increase in cortisol. Toddler sexplain additional variance. Next, time of day ofe cortisol sampling was allowed to enter Step-was not retained in the model. The subjective

interaction was signicant in the nal stepnted for 7.0% of additional variance (p < .05):ive PNMS scores increased, so did total cor-tion in females; conversely, subjective PNMS

not related to total cortisol secretion in malesy probing the interaction for the regions of sig-AUCi values between males and females wereiffer signicantly when IES-R log-transformed

e less than 0.393 or greater than 3.463 (non-d IES-R levels of 0.482 and 30.919, respectively).e nal model explained 23.4% of the variance of

3.5. Area(AUCg)

Table 5b pregression dictor nor ttotal cortisthe varianc

3.6. Addi

A repeatedfurther exptive PNMS ainteractionsince startsex was fop = .08; seecortisol levnal subjecslope as a f

4. Discu

The goal oation betw0.097 0.010.38 0.0700.108 0.010.25 0.0620.130 0.010.54 0.124

39.844 58.06477.23 107.08663.048 95.371039.89 165.7340.017 0.170.24 0.0730.038 0.250.46 0.1230.176 0.060.60 0.090

0.002 0.500.23 0.1131.850 03.93 0.8898.57 050 10.6651.268 04.11 1.0665.453 060 8.799

148.266 0280.98 81.99953.77 2166 9.8520.27 014 1.5290.07 02 0.252

13.50 129 5.0754.43 014 2.888

66.44 4097 13.7584.39 19 1.805

32.73 2270 8.005546.12 22204740 444.29251.70 4558 2.51739.24 3542 1.24912:12 8:5519:22 2:58conomic Status (higher scores indicate higher SES).

under the curve with respect to ground

resents the results of the hierarchical multipleanalyses conducted for AUCg. Neither the pre-he control variables were signicantly related tool secretion during the assessment. Only 5.5% ofe in AUCg was explained by the nal model.

tional analyses

measures mixed model analysis was conducted tolore the signicant interaction between subjec-nd toddler sex on cortisol reactivity. A three-way

between cortisol sampling time (i.e., minutes of procedure), subjective PNMS, and toddlerund to be marginally signicant (t(193) = 1.72,

Fig. 2): in females, the magnitude of the slope ofels across time increased as a function of mater-tive distress exposure, while no differences inunction of subjective PNMS was found for males.

ssion

f the present study was to determine the associ-een disaster-related PNMS and toddler cortisol

Prenatal maternal stress and toddler stress reactivity 71

Table 3 Pearson correlation coefcients for all variables (n).

Variable Absolutechange(baseline to

Absolutechange(baseline to

in-stre

Area under thecurve(increase)

Area under thecurve (ground)

Objective 13 (8Subjective 93 (8Timing of 57 (8Toddler se 23 (8Time of d 79 (8SES 88 (8Number o 58 (8Alcohol un 22 (8Perceived 39 (7Life Exper 14 (7Parenting 46 (7Social sup 86 (7Depressio 63 (7Birth weig 22 (8Birth leng 02 (7Gestation 31 (8Baseline c 96 (8

Note: PNM es inAnxiety Sy p < .1.* p < .05.** p < .01.

reactivity method indegree of ofrom the to test thePNMS expecompared iety, we wof timing omoderating

On averstudy was fminutes fowith previoinfant andexperimenical examiet al., 200average, athe brief mdifferenceresponse tthe female

We testdler stress to 20 or 45tion relativ(AUCi). Though inveschange, re

ion wch oer o20 minpost-stressor)

45 mpost

PNMS (LogIF100) 0.190 (79) 0.2 PNMS (LogIES-R) 0.150 (79) 0.1exposure in utero (days) 0.080 (79) 0.1x (0 = boy; 1 = girl) 0.127 (79) 0.1ay of baseline cortisol sampling 0.043 (79) 0.1

0.221 (77) 0.1f cigarettes/day 0.141 (77) 0.0its/week 0.103 (77) 0.1

Stress Scale 0.102 (76) 0.1ience Survey 0.130 (76) 0.1

Stress Index 0.031 (76) 0.0port 0.005 (75) 0.0n and anxiety (IDAS) 0.094 (76) 0.0ht (g) 0.300 (77)** 0.0th (cm) 0.198 (70) 0.1al age at birth (weeks) 0.070 (77) 0.0ortisol (g/dL) 0.599 (79)** 0.1S = Prenatal Maternal Stress. SES = Socioeconomic Status (higher scormptoms.

in response to brief maternal separation. Ourcluded the ability to distinguish between thebjective exposure of the mother to the hardship

secretto whiof powooding and her degree of subjective distress, and relative contributions of these two aspects of therience. Given the sudden onset of the ooding,to other forms of prenatal stress such as anx-ere also able to test, and control for, the effectsf the stressor in utero. Finally, we also tested the

effects of toddler sex.age, the stressor paradigm used in the presentound to elevate toddler cortisol levels for up to 45llowing the stressor. Our nding is in accordanceus studies documenting a signicant increase in

toddler cortisol levels following exposure to antal stressor, including maternal separation, phys-nation, and inoculations (for review see Gunnar9). Females in the present sample exhibited, on

signicant increase in cortisol levels followingaternal separation, whereas males did not. This

suggests that the overall elevation in cortisol ino maternal separation was largely inuenced bys in our sample.ed the effects of PNMS on four aspects of tod-reactivity: absolute cortisol change from baseline

minutes post-stressor, and total cortisol secre-e to ground (AUCg) or to increase from baselinee broad scope of our analyses allowed for a thor-tigation of cortisol reactivity, including acuteturn to baseline, total secretion, and reactive

of PNMS woMore severgreater AUof change jective PNthe trend utes. Althogroup as a and 45-minand b, thelute cortisonot in malemeasures msignicantlPNMS levelgreater subsecretion inentire asselevels did nchange in rin basal codecade (fogirls commtions than 2004; OCosex differemental stressor)

2) 0.258 (71)* 0.053 (71)2) 0.246 (71)* 0.030 (71)2) 0.330 (71)** 0.100 (71)2) 0.076 (71) 0.124 (71)2) 0.240 (71)* 0.005 (71)0) 0.072 (70) 0.097 (70)0) 0.027 (69) 0.075 (69)0) 0.024 (69) 0.030 (69)9) 0.024 (68) 0.014 (68)9) 0.160 (68) 0.005 (68)9) 0.023 (68) 0.107 (68)6) 0.009 (67) 0.155 (67)8) 0.099 (68) 0.127 (68)0) 0.137 (69) 0.055 (69)3) 0.094 (64) 0.044 (64)0) 0.047 (69) 0.047 (69)2) 0.028 (71) 0.790 (71)**dicate higher SES). IDAS = Inventory of Depression and

ith respect to baseline. We analyzed the extentbjective (e.g., amount of nancial loss, lengthutage) and subjective (PTSD symptoms) aspects

uld explain variance in these cortisol measures.e objective PNMS from the oods correlated withCi and, at the trend level, with absolute levels20 and 45 minutes post-stressor. Similarly, sub-MS was positively correlated with AUCi and, atlevel, with greater absolute change at 45 min-ugh the effects of objective PNMS applied to thewhole, the effect of subjective distress on AUCiute change was limited by sex: as seen in Fig. 1a

effect of subjective PNMS on 45-minute abso-l change and AUCi was signicant in females buts, a nding that was supported by the repeatedixed model analysis (Fig. 2). Thus, toddler sex

y moderated the relationship between subjectives and these two aspects of cortisol reactivity:jective PNMS levels predicted greater cortisol

response to maternal separation and across thessment in females; conversely, subjective PNMSot predict cortisol levels or the degree of cortisolesponse to stress in males. While sex differencesrtisol levels have been documented over the lastr review see Jessop and Turner-Cobb, 2008), withonly having higher baseline cortisol concentra-boys (Watamura et al., 2003; Gutteling et al.,nnor et al., 2013), few studies have documentednces in cortisol reactivity in response to experi-ssors in children (Jessop and Turner-Cobb, 2008).

72 E. Yong Ping et al.

Table 4 Results of hierarchical multiple regression for (a) absolute change baseline to 20-minute post-stressor (n = 77) and (b)absolute change baseline to 45-minute post-stressor (n = 82).

Predictor variables B Std. error R R2 R2 F F

(a) Absolute change baseline to 20-minute post-stressor (n = 77)Step 1

Objective PNMS .141 .011 .009.141 .020 .020 1.531 1.531

Step 2Objective PNMSSubjective PNMS

.118

.058.009.004

.010

.009

.151 .023 .003 .864 .213

Step 3Objective PNMSSubjective PNMSTiming in utero

.105

.052

.065

.008

.0045.797E05

.010

.009

.000

.164 .027 .004 .670 .298

Step 4Objective PNMSSubjective PNMSTiming in uteroSex (0 = boy; 1 = girl)

.123

.035

.081

.131

.010

.0027.231E05.019

.010

.009

.000

.017

.208 .043 .017 .815 1.244

Step 5Objective PNMSSubjective PNMSTiming in uteroSex (0 = boy; 1 = girl)Baseline cortisol

.116

.004

.000

.156

.613**

.009

.0002.553E07.022

.677**

.008

.007

.000

.013

.102

.640 .409 .366 9.826** 43.926**

Step 6Objective PNMSSubjective PNMSTiming in uteroSex (0 = boy; 1 = girl)Baseline cortisolBirth weight

.073

.028

.026

.096.584**.208*

.006

.0022.356E05.014.645**3.171E05*

.008

.007

.000

.013

.101

.000

.668 .446 .037 9.410** 4.740*

Step 7Objective PNMSSubjective PNMSTiming in uteroSex (0 = boy; 1 = girl)Baseline cortisolBirth weightSubjective PNMS Sex

.006.0021.093E05.001.650**3.082E05*.011

.008

.009

.000

.021

.101

.000

.013

.672 .452 .005 8.124** .672

(b) Absolute change baseline to 45-minute post-stressor (n = 82)Step 1

Objective PNMS .213 .029 .015.213 .045 .045 3.810 3.810

Step 2Objective PNMSSubjective PNMS

.157

.119.021.014

.017

.015

.238 .057 .011 2.367 .929

Step 3Objective PNMSSubjective PNMSTiming in utero

.135

.115

.112

.018

.014

.0001

.017

.015

.0001

.262 .069 .012 1.912 1.001

Step 4Objective PNMSSubjective PNMSTiming in uteroSex (0 = boy; 1 = girl)

.153

.100

.120

.142

.021

.012

.0001

.035

.017

.015

.0001

.027

.297 .088 .020 1.865 1.675

Prenatal maternal stress and toddler stress reactivity 73

Table 4 (Continued)

Predictor variables B Std. error R R2 R2 F F

Step 5ObjectiSubjectTiming Sex (0 =Baseline

.358 .128 .040 2.231 3.455

Step 6ObjectiSubjectTiming Sex (0 =BaselineSubject

Note: PNM p < .1.* p < .05.** p < .01.

Nevertheletivity in thof increaseexposed toet al., 201with increapal neurog(Fameli etthese resuity to prendevelopmepocampus differencefactors, suof 11-HSD2010). Expwith prenaas testostecritical to the activitbetween sels than mand femalunknown, notion thatual dimorpwhich placdisorders (tion is warreported iespecially stress have(Schopper tibility masuch as typthe presenipants willto determover, to ex

lateres.

resmminmenriodsvelssed cortilevataccoal HPod.S ereat5a, t

5.4ectiv

predeve

effeve PNMSive PNMSin utero

boy; 1 = girl) cortisol

.155

.089

.116

.165.201

.021

.011

.0001

.040.349

.017

.014

.0001

.027

.188

ve PNMSive PNMSin utero

boy; 1 = girl) cortisolive PNMS Sex

.024.018.0001.039.392*.064*

.016

.018

.0001

.040

.182

.025

S = Prenatal Maternal Stress.

ss, the nding of greater female cortisol reac-e present project is in accordance with ndingsd diversity of physiological outcomes in females

early life adversity compared to males (Sandman3). In animal studies, PNMS has been associatedsed abnormal brain morphology and hippocam-enesis in female offspring compared to males

al., 1994; Hillerer et al., 2013). Collectively,lts suggest that the sexes differ in susceptibil-atal insults, which have the potential to modifynt of integrated brain systems, including the hip-and HPA axis. The reason for these observed sexs is currently unknown; however, biological foetalch as androgen levels and activity and expression2, may be implicated (for review see Charil et al.,osure to elevated maternal GC levels associatedtal stress may modify foetal androgen levels, suchrone, for which levels and timing of release arebrain development. With respect to 11-HSD2,y level of this enzyme has been found to differexes, with females commonly having higher lev-

with outcom

Theprograenvirocal pehigh leincreavated that eways, of foetchildho

PNMwith gTable explainby objhalf ofbrain timingales. While the exact mechanism by which maleses differ in their response to PNMS exposure isthe results from the present study support the

dysregulated HPA activity may underlie the sex-hism seen in psychiatric illness prevalence rates,es females at a greater risk for developing moodFernandez-Guasti et al., 2012). Further investiga-ranted to determine whether the sex differencen this paper exists beyond toddlerhood. This isimportant given that animal models of prenatal

reported greater susceptibility in male offspringet al., 2012), suggesting that sex-specic suscep-y change depending on experimental variables,e of stressor or timing of exposure. Given thatt project is part of a longitudinal study, partic-

continue to be followed throughout childhoodine whether the ndings persist and, more-amine whether cortisol reactivity is associated

to prenatawith greatmothers whigher basaadolescent(OConnor et al., 200possess evto offsprinskills, and reproduce 2010).

There arst is therienced thefrom very was compleThirdly, the.444 .197 .069 3.074* 6.487*

cognitive and behavioural developmental

ults presented here lend support for the foetalng hypothesis, the process by which the externalt reprograms foetal development during criti-

(Glover et al., 2010). Given that exposure to of disaster-related PNMS was associated withactivity of the HPA axis, evident through ele-sol secretion in response to stress, this suggestsed maternal GCs, or other physiological path-mpanying prenatal stress led to reprogrammingA axis, affecting the axis functioning throughout

xposure later in pregnancy was also associateder cortisol secretion in toddlers. As seen iniming of foetal exposure to PNMS was found to% of the variance in AUCi values unexplainede and subjective PNMS. Therefore, the secondgnancy may represent a critical period for foetallopment, including the HPA axis. Our reportedct is supported by past studies where exposure

l stress mid- to late-pregnancy was associateder changes in cortisol levels in infants born toith PTSD (Yehuda et al., 2005), and predictedl levels of salivary cortisol in pre-adolescent and

offspring exposed to prenatal maternal anxietyet al., 2005; Huizink et al., 2008; Van den Bergh8). Reprogramming of foetal development mayolutionary value if resulting modications leadg with increased resiliency and efcient copingoffspring that are more likely to survive andin their external environment (Glover et al.,

re several strengths of the present study. The fact that all of the women in the study expe-

same stressor, a ood, to degrees that variedlittle to extreme. Secondly, the stressor itselftely independent of the womens own inuence.se facts allowed us to distinguish between the

74 E. Yong Ping et al.

Figure 1 showing regof signican

mothers otheir leveltive PNMS mIce Storm the Iowa Frisk factoruncorrelateas SES (r