Fetal exposure to endogenous maternal glucocorticoids during gestation has broad and enduring effects on offspring development (Mina & Reynolds, Reference Mina and Reynolds2014). In animal models of prenatal stress, glucocorticoid exposure (both endogenous and exogenous) influences the development of the fetal hypothalamus–adrenal–pituitary (HPA) axis function, and this in turn leads to adaptation at every level of development from cells to behavior (Cottrell & Seckl, Reference Cottrell and Seckl2009; Harris & Seckl, Reference Harris and Seckl2011). In utero exposure to maternal glucocorticoids is believed to affect fetal development through structural and functional reorganization of physiological systems, such as the HPA axis, in a process that has been described as fetal or developmental programming (Barker, Reference Barker2004; Burton & Fowden, Reference Burton and Fowden2012; Sandovici, Hoelle, Angiolini, & Constância, Reference Sandovici, Hoelle, Angiolini and Constância2012). Such programming operates as a form of nongenomic microevolution that allows for rapid adaptation to changing environmental conditions.
According to the developmental origins hypothesis, glucocorticoids transmitted from mother to fetus during gestation convey useful information about the state of the external world, and this information is exploited by the fetus to optimize development in preparation for survival and success in the postnatal environment (Bateson et al., Reference Bateson, Barker, Clutton-Brock, Deb, D'Udine and Foley2004; Ellis & Del Giudice, Reference Ellis and Del Giudice2013; Gluckman et al., Reference Gluckman, Hanson, Bateson, Beedle, Law and Bhutta2009). Both animal and human studies suggest that male and female fetuses incorporate these signals in unique ways that result in sexually dimorphic outcomes (Sandman, Glynn, & Davis, Reference Sandman, Glynn and Davis2013). For example, maternal distress and elevated cortisol during pregnancy may result in accelerated development in males but more variable development (sometimes accelerated and sometimes reduced) in females (Doyle et al., Reference Doyle, Werner, Feng, Lee, Altemus and Isler2015). Furthermore, several recent studies have shown that maternal cortisol production during pregnancy is itself influenced by fetal sex, with higher levels associated with female sex, especially toward the end of pregnancy (DiPietro, Costigan, Kivlighan, Chen, & Laudenslager, Reference DiPietro, Costigan, Kivlighan, Chen and Laudenslager2011; Giesbrecht, Campbell, & Letourneau, Reference Giesbrecht, Campbell and Letourneau2015). Nevertheless, few human studies have evaluated the links between in utero cortisol exposure and infant HPA axis, and none have evaluated whether this association depends on levels of psychological distress or fetal sex. In the current study we address these gaps.
During pregnancy, cortisol passes through the placenta in sufficient quantity to alter fetal development, despite partial blockade by the enzyme 11 beta hydroxysteriod dehydrogenase type 2 (11βHSD2; Gitau, Cameron, Fisk, & Glover, Reference Gitau, Cameron, Fisk and Glover1998; Seckl, Reference Seckl2008). Psychological distress is a potent stimulant for the HPA axis in both nonpregnant and pregnant humans (de Weerth, Wied, Jansen, & Buitelaar, Reference de Weerth, Wied, Jansen and Buitelaar2007; Giesbrecht, Campbell, Letourneau, & Kaplan, Reference Giesbrecht, Campbell, Letourneau and Kaplan2013; Giesbrecht, Campbell, Letourneau, Kooistra, & Kaplan, Reference Giesbrecht, Campbell, Letourneau, Kooistra and Kaplan2012; Nierop et al., Reference Nierop, Bratsikas, Klinkenberg, Nater, Zimmermann and Ehlert2006), and cortisol is widely cited as a plausible biological mechanism whereby maternal experiences of psychological distress are transmitted to the fetus (Mina & Reynolds, Reference Mina and Reynolds2014). For these reasons, cortisol is proposed as a central mechanism for fetal programming, and animal models have convincingly shown that increases in maternal stress during pregnancy leads to behavioral and cognitive changes that influence the overall developmental potential of the offspring (Bale et al., Reference Bale, Baram, Brown, Goldstein, Insel and McCarthy2010; Mina & Reynolds, Reference Mina and Reynolds2014). It is important to note that the relationship between prenatal stress and offspring outcomes may depend on both the strength and the timing of the prenatal exposure. There is some evidence that mild elevations in stress and cortisol, at least within otherwise healthy populations, may be associated with advanced fetal and child maturation (DiPietro, Novak, Costigan, Atella, & Reusing, Reference DiPietro, Novak, Costigan, Atella and Reusing2006; Doyle et al., Reference Doyle, Werner, Feng, Lee, Altemus and Isler2015). Negative effects of prenatal stress and cortisol exposures are most consistently observed when exposures occur early in development and when the levels of exposure are high (Davis & Sandman, Reference Davis and Sandman2010; Laplante et al., Reference Laplante, Barr, Brunet, Galbaud du Fort, Meaney and Saucier2004), and potentially also at very low levels of exposure (DiPietro, Reference DiPietro2012). Many of the effects on offspring development are believed to result from or be exacerbated by hyperreactivity or hyporeactivity of the infant HPA axis following in utero exposure to maternal stress.
Prenatal Maternal Distress and Infant HPA Axis Outcomes
Animal studies have shown that prenatal stress has both species-specific and sex-specific effects (Bale, Reference Bale2011). For example, among guinea pigs, chronic prenatal stress leads to blunted or hyporeactivity in the HPA axis response to stress exposure (Emack, Kostaki, Walker, & Matthews, Reference Emack, Kostaki, Walker and Matthews2008), whereas among rhesus monkeys (Coe et al., Reference Coe, Kramer, Czeh, Gould, Reeves and Kirschbaum2003) and rats, prenatal stress leads to HPA axis hyperreactivity, although in the case of rats these effects were found only for female offspring (Emack et al., Reference Emack, Kostaki, Walker and Matthews2008; Szuran, Pliska, Pokorny, & Welzl, Reference Szuran, Pliska, Pokorny and Welzl2000; Weinstock, Reference Weinstock2001; Welberg & Seckl, Reference Welberg and Seckl2001). A common theme across species and sex is persistent hyperactivity of basal HPA axis function, although the size of this effect is somewhat dependent on the timing and severity of the prenatal exposure (Welberg & Seckl, Reference Welberg and Seckl2001). Human studies have also shown elevations in basal cortisol along with flattening of the daytime cortisol trajectory following in utero exposure to maternal psychological distress (Brennan et al., Reference Brennan, Pargas, Walker, Green, Newport and Stowe2008; Laurent et al., Reference Laurent, Leve, Neiderhiser, Natsuaki, Shaw and Harold2013; O'Connor et al., Reference O'Connor, Ben-Shlomo, Heron, Golding, Adams and Glover2005; O'Donnell et al., Reference O'Donnell, Bugge Jensen, Freeman, Khalife, O'Connor and Glover2012; Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yehuda et al., Reference Yehuda, Engel, Brand, Seckl, Marcus and Berkowitz2005), although several studies have reported lower basal levels among infants whose mothers had a history of abuse or depression as a child (Azar, Paquette, Zoccolillo, Baltzer, & Tremblay, Reference Azar, Paquette, Zoccolillo, Baltzer and Tremblay2007; Brand et al., Reference Brand, Brennan, Newport, Smith, Weiss and Stowe2010). The later findings highlight the special significance of early life adversity not only for individuals but also for their offspring. Chronic elevation and flat daytime cortisol patterns are risk markers for psychopathology (Gunnar & Vazquez, Reference Gunnar and Vazquez2001; Tarullo & Gunnar, Reference Tarullo and Gunnar2006).
Studies assessing the associations between in utero exposure to maternal psychological distress and children's cortisol responses to stressors have not produced a consistent set of results. This is best exemplified in a study by Tollenaar et al. (Reference Tollenaar, Beijers, Jansen, Riksen-Walraven and de Weerth2011) who reported that prenatal maternal distress predicted infant cortisol reactivity, but the direction of the effect varied by child age. Pregnancy-specific anxiety, but not general measures of anxiety or stress, was associated with heightened cortisol response to a bathing routine at 5 weeks of age, but at 8 weeks and 12 months it was associated with reduced cortisol responses to a vaccination and maternal separation, respectively. Similarly, other studies have reported that prenatal depression and/or anxiety were associated (Azar et al., Reference Azar, Paquette, Zoccolillo, Baltzer and Tremblay2007; Brand et al., Reference Brand, Brennan, Newport, Smith, Weiss and Stowe2010; Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yong Ping et al., Reference Yong Ping, Laplante, Elgbeili, Hillerer, Brunet and O'Hara2015) or were not associated (Davis, Glynn, Waffarn, & Sandman, Reference Davis, Glynn, Waffarn and Sandman2011) with increased offspring cortisol reactivity. Of these studies, only two reported testing for Sex × Psychological Distress Exposure interactions, with both studies suggesting greater reactivity in females compared to males (Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yong Ping et al., Reference Yong Ping, Laplante, Elgbeili, Hillerer, Brunet and O'Hara2015). Nevertheless, previous studies have not examined sexually dimorphic effects of prenatal distress in the context of concurrent cortisol exposure, and this is a significant limitation because it is known that prenatal psychological distress and cortisol have additive effects on offspring outcomes (Davis & Sandman, Reference Davis and Sandman2010).
Prenatal Maternal HPA Axis and Infant HPA Axis Outcomes
Whereas many studies have examined the links between prenatal distress and children's cortisol, relatively few human studies have assessed the associations between prenatal cortisol exposure and children's HPA axis function. Of the six studies that we located, three reported positive associations between prenatal cortisol exposure and children's HPA axis function (Davis et al., Reference Davis, Glynn, Waffarn and Sandman2011; Gutteling, de Weerth, & Buitelaar, Reference Gutteling, de Weerth and Buitelaar2005; Gutteling, de Weerth, & Buitelaar, Reference Gutteling, de Weerth and Buitelaar2004), one reported a negative association (O'Connor, Bergman, Sarkar, & Glover, Reference O'Connor, Bergman, Sarkar and Glover2013), and two studies found no association (de Weerth, Buitelaar, & Beijers, Reference de Weerth, Buitelaar and Beijers2013; Tollenaar et al., Reference Tollenaar, Beijers, Jansen, Riksen-Walraven and de Weerth2011). One additional study found that maternal hair cortisol during pregnancy was positively associated with offspring hair cortisol at age 1 and 3, but not 5 and 8, although this finding is more directly relevant to basal HPA axis function than to HPA axis reactivity (Karlen, Frostell, Theodorsson, Faresjo, & Ludvigsson, Reference Karlen, Frostell, Theodorsson, Faresjo and Ludvigsson2013).
These studies were all of relatively high quality and included appropriate controls for potential confounding variables in statistical analyses. Nevertheless, differences with regard to timing of the maternal cortisol measure (both within the day and within pregnancy), differences in the type of prenatal cortisol exposure (e.g., area under the curve [AUC], cortisol awakening response [CAR], and amniotic fluid) and differences in the type of infant stressor may have contributed to the disparate findings. Furthermore, two key considerations have been neglected in these studies. First, child sex has not been included as a potential moderator. This is a significant limitation because sex is known to moderate the effects of prenatal distress on a variety of outcomes in children (Clifton, Reference Clifton2010; Davis & Pfaff, Reference Davis and Pfaff2014; Sandman et al., Reference Sandman, Glynn and Davis2013), including HPA axis function (Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yong Ping et al., Reference Yong Ping, Laplante, Elgbeili, Hillerer, Brunet and O'Hara2015), and because maternal cortisol secretion during pregnancy differs as a function of fetal sex (Giesbrecht et al., Reference Giesbrecht, Campbell and Letourneau2015). Second, all previous studies have considered prenatal distress and cortisol as independent predictors. Although this approach has yielded valuable insight regarding the unique effects of psychological and biological predictors on infant outcomes, it does not address the possibility that prenatal distress and cortisol have interactive (and possibly synergistic) effects on children's HPA axis. Because in utero cortisol exposure is a plausible biological mechanism by which maternal experiences of psychological distress become embedded in children's development and because HPA axis reactivity mediates the effects of the environment on children's development, studies assessing sex differences in the effects of prenatal cortisol and psychological distress exposures on infant cortisol reactivity have the potential to advance our understanding of the developmental origins of sex differences in developmental psychopathology.
Current Study
The current prospective study, in a relatively high socioeconomic status (SES) and low-risk sample, set out to determine whether the associations between prenatal maternal cortisol and infant cortisol reactivity are moderated by infant sex and prenatal distress. Given evidence that the timing of exposure may affect infant outcomes (Tollenaar et al., Reference Tollenaar, Beijers, Jansen, Riksen-Walraven and de Weerth2011; Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012), we assessed pregnant women as early in pregnancy as possible, but in all cases prior to 22 weeks gestation and again near the end of pregnancy. Infant HPA axis function was assessed at 3 months postpartum using a blood draw as the stressor. Building upon the strength of the previous studies, we included appropriate covariates to rule out effects of the postnatal environment. Our aims were to replicate previous findings of sex differences in the association between psychological distress and infant cortisol reactivity, to determine whether the association between prenatal maternal cortisol and infant cortisol reactivity is moderated by infant sex, to determine whether the combined effects of prenatal cortisol and distress are moderated by infant sex, and to determine whether these associations are consistent across different points of gestation.
Methods
Sample
Participants were 294 women enrolled in an ongoing prospective cohort study, the Fetal Programming study, which is a community sample of volunteers recruited in 2011 and 2012. Women were included if they had a singleton pregnancy, were less than 22 weeks of gestation at the first study visit, and were 18 years of age or older. Women were excluded if they smoked or consumed alcohol during pregnancy, were being treated with a synthetic glucocorticoid, or had known fetal complications at time of study entry. Because gestational age at birth and birth weight are associated with infant HPA axis function (Buss et al., Reference Buss, Entringer, Davis, Hobel, Swanson and Wadhwa2012; Glover, Miles, Matta, Modi, & Stevenson, Reference Glover, Miles, Matta, Modi and Stevenson2005), we made an a priori decision to exclude from current analysis infants with low birth weight (<2500 g) or preterm birth (<37 weeks gestation). The sample was relatively well educated (85.7% had education beyond high school), financially stable (52.7% had household income ≥ CA $100,000/annum and 8.5% had household income < CA $39.999/annum), married or common law related (94.9 %), mature (mean age = 31.4 years, SD = 3.8 years), and primiparious (51.7%). The majority were non-Hispanic white (77.2%); the remainder were Asian (9.5%), Latin American (3.7%), Arab (2.0%), Japanese (1.4%), Filipino (1.0%), or other (5.2%). The sex ratio of infants favored boys (n = 118; 51.9%), consistent with the worldwide birth ratio of 51.7% males and 48.3% females (Hesketh & Xing, Reference Hesketh and Xing2006).
Procedures and measures
Saliva collection during pregnancy
Participants collected a series of diurnal saliva samples at two time points in pregnancy: Time 1 (T1) = 2 days in early pregnancy (mean = 14.9 weeks gestational age [GA], range = 5.9–21.9 weeks GA), and Time 2 (T2) = 2 days in late pregnancy (mean = 32.5 weeks GA, range = 27.1–36.7 weeks GA). At each time point, women self-collected saliva at home during weekdays on the following schedule: upon waking, 30 min after waking, at 11:30 a.m., and at 9:00 p.m. On each sampling day, participants collected the waking sample immediately upon waking and then initiated a 30-min timer on a personal digital assistant (PDA). This procedure allowed for precise timing of the 30-min postwaking sample while also allowing for individual waking times. With the exception of the waking sample, the PDA rang to indicate that a sample was to be collected. Each time the PDA rang, it provided a code corresponding to a prelabeled saliva tube and instructed the participant to place the saliva roll under her tongue. Participants then completed the PDA-administered psychological distress questionnaire during saliva collection. Time of each assessment was recorded by the PDA, permitting precise modeling of diurnal patterns. To facilitate adherence to the study protocol, the PDA was programmed to allow a 20-min response window following the signal, after which the questionnaire was no longer available.
Whole saliva was obtained from under the tongue using the SalivaBio Oral Swab (Carlsbad, CA). Mothers were asked to refrain from consuming food, caffeine, citric drinks, and dairy, and to avoid vigorous exercise (e.g., running or playing sports) or brushing their teeth within 30 min prior to saliva collection and to report adherence to these guidelines. Saliva samples were stored at –20 °C in participant home freezers until they were returned to the lab where they were they remained frozen at –80 °C until they were shipped frozen to Salimetrics (State College, PA).
Infant saliva collection
Infant cortisol was assessed at age 3 months before (baseline) and after (5 min, 20 min, and 40 min) a blood draw. Whole saliva was collected using sorbettes (BD Visispear™), which have been validated for cortisol collection (de Weerth, Jansen, Vos, Maitimu, & Lentjes, Reference de Weerth, Jansen, Vos, Maitimu and Lentjes2007).
Cortisol assay
All infant and maternal samples were assayed for salivary cortisol using the Salimetrics enzyme immunoassay. It has a lower limit of sensitivity of 0.007 μg/dl, standard curve range from 0.012 to 3.0 μg/dl, and average intra- and interassay coefficients of variation 3.5% and 5.1%, respectively. Method accuracies, determined by serial dilution, are 100.8% and 91.7%. A random 10% of samples were assayed in duplicate to confirm reliability; the intra-assay coefficient of variation and correlation coefficient between the duplicate tests for the maternal samples were .042 and .99, respectively, and for the infant samples, .035 and .99, respectively. Mean values from duplicate samples were used for analysis.
Prenatal maternal psychological distress
Maternal distress during pregnancy was assessed simultaneous to saliva sample collection using nine indicators of psychological distress (e.g., I felt tense, I felt discouraged, and I had many worries) taken from the Profile of Mood States (McNair & Heuchert, Reference McNair and Heuchert2007) and adapted for administration on the PDA following procedures previously described (Cranford et al., Reference Cranford, Shrout, Iida, Rafaeli, Yip and Bolger2006; Giesbrecht et al., Reference Giesbrecht, Campbell, Letourneau, Kooistra and Kaplan2012). Participants rated each item on a 5-point Likert scale from not at all to extremely, based on their feelings during the previous 30 min. The mean of the nine items was used to calculate a psychological distress score for each person (possible range = 0–4), with higher scores indicting more distress. Reliability and validity of a short PDA-administered version of the Profile of Mood States is supported by previous research with participants experiencing a major life stressor (Cranford et al., Reference Cranford, Shrout, Iida, Rafaeli, Yip and Bolger2006) and with pregnant women (Giesbrecht et al., Reference Giesbrecht, Campbell, Letourneau, Kooistra and Kaplan2012). In the current study the nine-item distress measure had a Cronbach α of 0.87.
Infant stressor
The infant blood draw at 3 months was completed by a certified pediatric phlebotomist using either venipuncture or heel stick. Parents were allowed to choose the method to increase their willingness to complete this portion of the study. For venipuncture, a #2 butterfly (winged infusion 12 in.) needle or 25-gauge 0.75-in. standard needle was used in the forearm to collect 3–4 ml of blood. Heel stick was performed with a 1.0-mm (2.5-mm width × 1-mm diameter) lance, and enough blood was squeezed from the heel to blot a 1- × 1-cm strip of chromatography filter paper. Blood samples obtained using these procedures were not analyzed for the current report.
Data deduction
Maternal psychological distress
For both the early and the late pregnancy time points (T1 and T2, respectively), a mean psychological distress score was calculated by aggregating all distress items from all sampling moments. The Cronbach α for the distress measure was 0.88 at T1 and 0.85 at T2.
Maternal cortisol
Three summary measures of maternal pregnancy cortisol were derived from the diurnal saliva samples to capture different aspects of HPA axis function. Samples collected at waking and 30-min postwaking were used to calculate the CAR using the trapezoid method for AUC increase (Pruessner, Kirschbaum, Meinlschmid, & Hellhammer, Reference Pruessner, Kirschbaum, Meinlschmid and Hellhammer2003) as a measure of the morning increase in cortisol. Samples were excluded from this analysis if they were taken more than 15 min after waking (for the waking sample) or more than 50 min after waking for the second sample (Okun et al., Reference Okun, Krafty, Buysse, Monk, Reynolds III and Begley2010; Wust et al., Reference Wust, Wolf, Hellhammer, Federenko, Schommer and Kirschbaum2000). A measure of total cortisol secretion was obtained from samples collected at waking, 11:30 p.m., and 9:00 p.m. and calculated using the trapezoid method for the AUC from ground (AUCg; Pruessner et al., Reference Pruessner, Kirschbaum, Meinlschmid and Hellhammer2003). Finally, the daytime slope was calculated by fitting a regression slope through the cortisol data for each individual to describe the daytime cortisol decline (Fekedulegn et al., Reference Fekedulegn, Andrew, Burchfiel, Violanti, Hartley and Charles2007). The daytime slope is expected to be negative (as it was in all cases in our sample); thus, values closer to zero indicate a flatter slope, whereas more negative numbers indicate a stronger decline. The daytime slope included the 30-min postwaking, 11:30 a.m., and 9:00 p.m. samples.
Statistical procedures
Data screening
Of the 294 women enrolled during pregnancy, 255 returned for the 3-month infant assessment. Of these, 19 mothers or infants were excluded from analysis because of failure to follow the study protocol (n = 1), and either low birth weight (<2500 g) or preterm birth (<37 weeks gestation; n = 18). Thus, the final sample for analysis was 236. Women in the final sample did not differ from the original sample on any demographic or study variables, with the exception of daytime slope at T2; women not in the final sample had a flatter daytime slope compared to women in the final sample, t (292) = 2.36, p = .019. The data was screened for potential outliers, and values that exceeded a z score of 3.29 were adjusted according to recommendations (Tabachnick & Fidell, Reference Tabachnick and Fidell2012). No more than four values (1.5%) were adjusted for any variable.
Analytic models
Multilevel modeling was used to model trajectories of infant cortisol as a function of prenatal maternal cortisol, distress, and sex. Multilevel models have advantages over other repeated-measures approaches in their ability to accommodate individual timing of samples: they flexibly accommodate missing data, and they allow for control of initial values effects (Kristjansson, Kircher, & Webb, Reference Kristjansson, Kircher and Webb2007; Llabre, Spitzer, Siegel, Saab, & Schneiderman, Reference Llabre, Spitzer, Siegel, Saab and Schneiderman2004). A two-level model with random intercept and slopes was fit to the infant cortisol data using the Mixed procedure in SPSS 22.0. The outcome was the log of infant cortisol. At Level 1, both the linear and the quadratic effects of time were included to allow curvilinear trajectories in infant cortisol (cortisol increases are expected to slow or reverse between 20- and 40-min poststressor) and to allow for individual differences in peak cortisol (maximum levels are delayed in some infants). Time was centered at baseline; thus, the intercept in the mixed models refers to cortisol levels at baseline. The primary variables of interest were maternal prenatal cortisol, prenatal distress, and infant sex, which were included in the Level 2 submodel for the intercept and linear slope. Maternal prenatal cortisol and distress were centered at the sample means and sex was centered at males. Selection of covariates, random effects, and error structure was based upon model fit as assessed by –2 log likelihood. Separate analyses were conducted for the T1 and T2 exposures to determine if exposure timing influenced the results.
Covariates
We included in our models covariates identified in previous studies (Buss et al., Reference Buss, Entringer, Davis, Hobel, Swanson and Wadhwa2012; Glover et al., Reference Glover, Miles, Matta, Modi and Stevenson2005; Granger, Hibel, Fortunato, & Kapelewski, Reference Granger, Hibel, Fortunato and Kapelewski2009) as potential confounders of the prenatal cortisol–infant cortisol association. In preliminary models, infant age at assessment, infant birth weight, blood draw method (venipuncture vs. heel stick), infant sleeping versus awake on the car ride to the appointment, maternal or infant medication use, and maternal prepregnancy body mass index were not associated with infant cortisol (all ps > .15) and were therefore excluded from further analysis. All models presented here included the following covariates: gestational week of pregnancy for maternal cortisol collection, arrival time at the laboratory, infant age at birth, infant negative affect, and maternal postnatal psychological distress (depression and anxiety), which was included to control for current level of distress. Infant negative affect in the period from arrival at the laboratory until the blood draw was rated by a trained research assistant using the negative affect scale of the Behavior Rating Scales of the Bayley Scales of Infant Development—Second Edition (Bayley, Reference Bayley1993), which assesses the amount and intensity of infant fussing/crying on a 5-point scale from no negative affect displayed to three or more intense, heightened, or prolonged displays of negative affect. Postnatal depression was assessed via the 10-item, self-report Edinburgh Postnatal Depression Scale (Cox, Holden, & Sagovsky, Reference Cox, Holden and Sagovsky1987). Scores range from 0 to 30, with scores above 12 strongly correlated with physician diagnosis of major depressive disorder (Jomeen & Martin, Reference Jomeen and Martin2007). In our study, the Edinburgh Postnatal Depression Scale Cronbach alpha was adequate (α = 0.78). Postnatal anxiety was assessed using the Symptom Checklist-90—Revised (Derogatis, Reference Derogatis1994), which is a multidimensional symptom inventory designed to reflect psychological symptom patterns. Only the 10-item anxiety scale was used in the current report; this anxiety scale has good convergent and divergent validity (Morgan et al., Reference Morgan, Wiederman and Magnus1998) and adequately discriminates between clinical and nonclinical samples (Bonicatto, Dew, Soria, & Seghezzo, Reference Bonicatto, Dew, Soria and Seghezzo1997; Holi et al., Reference Holi, Sammallahti and Aalberg1998). In our study, the Symptom Checklist-90—Revised anxiety scale Cronbach α was adequate (α = 0.79). The postnatal depression and anxiety measures were correlated (r = .50) and therefore were z transformed and combined to reduce the number of covariates included in the model. All covariates were centered at the sample mean, with the exception of infant negative affect that was centered at “no negative affect.”
Results
Descriptive results
Descriptive statistics displayed in Table 1 reveal several expected results. First, infant cortisol levels at baseline and 5-min post blood draw were not different,Footnote 1 t (70) = 1.19, p = .237, but there was a clear increase in cortisol 20 min poststressor, t (247) = 6.53, p < .001, and despite a small decline by 40-min poststressor infant cortisol remained significantly elevated over baseline, t (243) = 2.28, p = .024. The absolute cortisol increase from baseline to 20-min poststressor was 0.132 μg/dl. Second, maternal cortisol increased from T1 to T2, t (268) = 21.09, p < .001 for AUCg and t (268) = 5.20, p < .001 for CAR. There were no differences in distress or daytime slope between T1 and T2.
Descriptive statistics for study variables by time point
Note: BMI, Body mass index; GA, gestational age; T1, Time 1 at 14.9 weeks GA; T2, Time 2 at 32.5 weeks GA; CAR, cortisol awakening response; AUCg, area under the curve from ground.
Replication of Prenatal Maternal Distress × Sex Association
Maternal distress at T1 was associated with sex differences in baseline cortisol (β = 0.328, p = .033), but not with increase in cortisol in response to the stressor (β = –0.162, p = .508; see Model 1, Table 2). As shown in Figure 1, females exposed to low stress and all males had similar baseline and slopes whereas females exposed to high distress had elevated baseline cortisol resulting in overall increased levels of cortisol as compared to males F (1, 224) = 6.51, p = .011. At T2, these sex differences were reduced to a marginal effect on baseline cortisol (β = 0.116, p = .064) and a nonsignificant effect for infant cortisol increases (β = –0.134, p = .165; see Model 1, Appendix A, Table A.1).
Sex differences in infant cortisol reactivity as a function of prenatal distress exposure. Maternal psychological distress was modeled as a continuous variable, but for illustration purposes we present it at the mean of the upper and lower quartiles to represent high and low distress, respectively. Poststressor is 20 min after the blood draw. The high and low distress lines for males are overlapping.
Hierarchical linear modeling coefficients (standard errors) for prenatal cortisol T1, prenatal distress T1, and infant sex with infant cortisol reactivity as outcome
Note: T1, Time 1 at 14.9 weeks gestational age (GA); CAR, cortisol awakening response. Sex was centered at males. The results in bold are for the Sex × Prenatal Exposure interactions. All covariates were centered at the sample mean with the exception of infant negative affect, which was centered at “no negative affect.”
†p < .10. *p < .05. **p < .01. ***p < .001.
Effects of Prenatal Maternal Cortisol × Sex
CAR
The associations between the maternal CAR and infant cortisol reactivity at both T1 and T2 were moderated by infant sex (Model 2, Table 2). Note that because the results for T1 and T2 were similar, we have provided in Table 2 only the results for T1. Complete model results for T2 are available in Table A.1. At T1, the interaction between CAR and sex was significant both for baseline infant cortisol (β = –0.033, p = .002) and for the cortisol increase in response to the stressor (β = 0.056, p = .005). As shown in Figure 2, the effects of the maternal CAR were essentially opposite for males and females. For example, males exposed to a low CAR and females exposed to a high CAR both displayed large cortisol increases. Similar findings were observed for CAR T2, with significant CAR × Sex interactions both at baseline (β = –0.027, p = .007) and the increase over time (β = 0.062, p < .001; see Model 2, Table A.1).
Infant cortisol reactivity as a function of sex and maternal cortisol awakening response in early pregnancy (CAR T1). CAR T1 was modeled as a continuous variable, but for illustration purposes we present it at the means of the upper and lower quartiles to represent high and low CAR T1, respectively. Poststressor is 20 min after the blood draw.
To probe the Sex × CAR T1 effects, we calculated regions of significance using the web utility at www.quantpsy.org and described by Preacher, Curran, and Bauer (Reference Preacher, Curran and Bauer2006). Male and female reactivity differed (p < .05) when maternal CAR T1 was below the 26th percentile or above the 89th percentile.
AUCg
There were no effects of maternal AUCg T1 or T2 on infant baseline or change over time and no interactions between AUCg and infant sex.
Daytime slopes
The results for daytime slope were substantially the same as those for the CAR. Effects were observed for maternal daytime slope at both T1 (see Model 3, Table 2) and T2 (see Model 3, Table A.1). Females exposed to a flatter maternal daytime slope had elevated baseline (β = –0.052, p = .005) and smaller cortisol increases (β = 0.059, p = .046) compared to males. Similar findings were observed for daytime slope T2, although the effect at baseline was marginally significant (β = –0.033, p = .075), but the increase over time was significant (β = 0.063, p = .033).
Effects of Prenatal Maternal Cortisol × Distress × Sex
In the next set of models, we included both the maternal cortisol and distress variables and their interaction to determine if parental cortisol and distress interact in their effects on infant cortisol. Our interest was primarily in the interaction terms among cortisol, distress, and sex.
CAR
There was a three-way interaction among CAR T1, distress, and infant sex for infant cortisol reactivity (β = –0.377, p = .033), but not for baseline cortisol (β = 0.056, p = .608; Table 2, Model 4). For clarity of presentation, the findings for Model 4 are displayed in two figures: Figure 3 displays the interaction between sex and CAR T1 at low distress exposure and Figure 4 at high distress exposure. At low levels of prenatal distress (Figure 3), CAR T1 had opposite effects on male and female reactivity. That is, greater cortisol reactivity was associated with high CAR T1 exposure among females but with low CAR T1 exposure among males. Likewise, blunted cortisol reactivity was associated with low CAR T1 exposure in females but with high CAR T1 exposure among males. In contrast, at high levels of prenatal stress (Figure 4) CAR T1 exposure had little effect on reactivity among males, whereas females exposed to high CAR T1 had blunted reactivity and females exposed to lower CAR T1 had elevated reactivity. Exposure to high distress and low CAR T1 was associated with significantly higher cortisol levels in the poststressor period among females compared to males, F (1, 194) = 7.02, p = .009.
Sex differences in the effects of maternal cortisol awakening response in early pregnancy (CAR T1) on infant stress reactivity at low prenatal distress exposure (mean of the lower quartile). CAR T1 was modeled as a continuous variable, but for illustration purposes we present it at the mean of the upper and lower quartiles to represent high and low CAR T1, respectively. Poststressor is 20 min after the blood draw.
Sex differences in the effects of maternal cortisol awakening response in early pregnancy (CAR T1) on infant stress reactivity at high prenatal distress exposure (mean of the upper quartile). CAR T1 was modeled as a continuous variable, but for illustration purposes we present it at the mean of the upper and lower quartiles to represent high and low CAR T1, respectively. Poststressor is 20 min after the blood draw.
For T2, the three-way interactions among CAR, distress, and sex were not significant for either baseline or reactivity (Model 4, Table A.1). However, similar to the effects observed at T1, the two-way interactions between sex and cortisol were significant for both baseline (β = –0.046, p = .030) and reactivity (β = 0.090, p = .009), even after controlling for the effects of prenatal distress.
AUCg
There were no three-way interactions among AUCg, distress, and sex for either T1 or T2.
Daytime slopes
The three-way interactions among maternal daytime slope, prenatal stress, and infant sex were not significant at either T1 (Model 5, Table 2) or T2 (Model 5, Table A.1). Nevertheless, all significant two-way interactions observed in the individual models for psychological distress (Model 1) and daytime cortisol (Model 3) remained significant in the combined model. Similar effects were observed for the T2 model (Model 5, Table A.1); however, with the exception of the daytime Slope × Sex interaction, all other significant effects at T1 became trends at T2.
Discussion
In this prospective longitudinal study, we observed sexually dimorphic effects of prenatal cortisol and distress exposure on infant cortisol response to a pain stressor. The most striking feature of these findings is that prenatal cortisol and distress exposure had essentially opposite effects on male and female cortisol reactivity. That is, a smaller maternal CAR or flatter daytime slope was associated with blunted cortisol responses in females, whereas a larger CAR or steeper daytime slope was associated with blunted cortisol responses in males. When considering the combined effects of prenatal cortisol and distress exposure, males appeared to be less affected than females. Females exposed to a smaller CAR or flatter daytime slope combined with high distress exposure showed elevated cortisol responses, whereas females exposed to a large CAR or steeper daytime slope and high distress showed a blunted response. In contrast, males had similar cortisol increases regardless of the combination of cortisol and distress exposures. These findings suggest that prenatal cortisol and distress have interactive and sexually dimorphic effects on fetal HPA axis development in higher SES, lower risk populations.
Altered cortisol reactivity is associated with a variety of stress-related diseases (e.g., hypertension) and psychiatric disorders (e.g., depression; Bosch et al., Reference Bosch, Riese, Reijneveld, Bakker, Verhulst and Ormel2012). In utero exposure to maternal distress and cortisol are associated with lifelong alterations of neuroendocrine, behavioral and cognitive functions (Van den Bergh, Van Calster, Pinna Puissant, & Van Huffel, Reference Van den Bergh, Van Calster, Pinna Puissant and Van Huffel2008; Weaver, Reference Weaver2009), and elevated risk for developing psychiatric, cardiovascular, and metabolic disorders in later life, all of which are moderated by sex (Schlotz & Phillips, Reference Scholtz and Phillips2009). For example, a recent prospective investigation of prenatal cortisol exposure reported increased 10-year risk of coronary heart disease in adult offspring at age 42 (Stinson et al., Reference Stinson, Stroud, Buka, Eaton, Lu and Niaura2015). The risk, however, was only for women. In the context of other research suggesting that the HPA axis is a central mediator for many health and disease outcomes (Del Giudice, Ellis, & Shirtcliff, Reference Del Giudice, Ellis and Shirtcliff2011), our findings highlight the potential for lifelong effects related to prenatal cortisol exposure.
These findings may also have relevance to our understanding of sex differences observed in the incidence and presentation of stress-related disorders. Davis and Pfaff (Reference Davis and Pfaff2014) and Sandman et al. (Reference Sandman, Glynn and Davis2013) have argued that sex-specific fetal growth strategies result in greater adaptive flexibility for females in the short term, especially under conditions of maternal distress, but increase the long-term risk for developmental psychopathology, with anxiety and depression as two examples of stress-related psychopathology for which there are clear sex differences in presentation and prevalence (Altemus, Sarvaiya, & Neill Epperson, Reference Altemus, Sarvaiya and Neill Epperson2014). That patterns of HPA axis function in offspring may link prenatal distress exposure to later psychopathology, at least for females, is supported by research showing that prenatal distress exposure is associated with a flattened diurnal cortisol profile in female but not male offspring during adolescence (Van den Bergh et al., Reference Van den Bergh, Van Calster, Pinna Puissant and Van Huffel2008).
The findings underscore the importance of including sex in studies assessing the effects of prenatal exposures on offspring HPA axis development. Although previous studies have not examined sex differences in offspring cortisol responses as a function of prenatal cortisol exposure, sex differences related to psychological distress exposure have been described; males tend to display blunted responses, whereas females display increased responses to stressors (Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yong Ping et al., Reference Yong Ping, Laplante, Elgbeili, Hillerer, Brunet and O'Hara2015). Our findings partially replicated these studies in that we observed elevated cortisol responses among females prenatally exposed to higher distress, but blunting of cortisol responses was observed only in reference to prenatal cortisol and was observed for both males and females, albeit under different levels of exposure. Our findings are also consistent with sexually dimorphic effects of prenatal exposure to exogenous glucocorticoids. Alexander et al. (Reference Alexander, Rosenlocher, Stalder, Linke, Distler and Morgner2012) found that exogenous glucocorticoids (administered prenatally because of threatened preterm birth) were associated with heightened cortisol responses observed among females.
Our findings offer insight into the inconsistent pattern of results that have been observed across previous studies assessing the association between in utero cortisol exposure and infant cortisol reactivity. Ignoring sex difference essentially takes the average of the effects for males and females, and because the effects we observed were, in some cases, opposite for males and females (see for example Figures 2 and 3), the results in any specific study will depend on the composition of the sample, including sex ratio. Of the previous studies reporting a positive association between prenatal cortisol and infant cortisol reactivity, the samples favored females (64%) although we note that the Davis et al. study (2011) had slightly more males (52.5%) even though they reported a positive association. Taken together, these findings suggest that sex differences in the effects of maternal cortisol and distress may essentially “wash out” the overall effects when sex effects are not explicitly modeled.
Our findings of sexually dimorphic effects of prenatal cortisol on infant cortisol reactivity extend previous findings by demonstrating both the independent effects of prenatal cortisol and distress exposures and their interactive effects on the developing HPA axis. Models 1–3 (Table 2) show the individual effects of distress and cortisol, whereas Models 4 and 5 show the unique and interactive effects. However, all of the individual effects of prenatal distress and cortisol remained when both exposures were included in the same model, although in the case of Model 4 (CAR) the unique effect of psychological distress was marginal. These findings confirm that distress and cortisol contribute in unique ways to development of the fetal HPA axis (Davis & Sandman, Reference Davis and Sandman2010). However, these findings also suggest that cortisol cannot fully account for the effects of maternal distress on infant cortisol reactivity. Although animal studies have shown that maternal cortisol during pregnancy mediates the association between maternal distress and infant outcomes, there is a substantial human literature that has not found support for the potential association between maternal cortisol and child outcomes (de Weerth & Buitelaar, Reference de Weerth and Buitelaar2005; Zijlmans, Riksen-Walraven, & de Weerth, Reference Zijlmans, Riksen-Walraven and de Weerth2015). Based upon the current findings, we argue that overlooking fetal sex as a potential moderator may help to explain the lack of empirical support for the effects of prenatal stress exposures on child outcomes. Future research will help to more fully explicate the effects of maternal distress on infant development by including sex within statistical models and by incorporating biomarkers for other stress response systems, such as the sympathetic and parasympathetic nervous systems.
The interactive effects of prenatal cortisol and distress exposures were complex. At low levels of prenatal distress, the effects of cortisol exposure were the same as those observed in the models where only cortisol was included (i.e., Models 1 and 2), perhaps reflecting that the sample as a whole had relatively low levels of psychological distress. In contrast, at high levels of prenatal distress exposure, variation in maternal cortisol had little effect on males, whereas females exposed to a small CAR and flatter daytime slopes had both elevated baseline and robust increases, whereas females exposed to a large CAR or steeper daytime slopes had essentially no response to the stressor (Figure 4).
Consistent with the regions of significance analysis reported by Young Ping et al. (2015), infant cortisol levels differed as a function of maternal cortisol exposure only at the low and high ends of the maternal cortisol distribution. These findings support the notion that both too little and too much cortisol during gestation have important implications for fetal development. It is important to note, however, that low and high cortisol exposures present differential risk for the males and females. As shown in Figures 1 and 3, and taking blunting of cortisol responses as an indicator of a potentially problematic adaptation, males appear to be vulnerable to high cortisol exposures, whereas females appear to be vulnerable to low exposures.
Across the three measure of cortisol during pregnancy, those that measured diurnal changes (i.e., the CAR and diurnal slope) were associated with infant HPA axis function, whereas AUCg, a measure of total cortisol, was unrelated to infant cortisol. In the context of the developmental origins hypothesis, we speculate that the total amount of prenatal cortisol exposure is a less useful signal to the fetus compared to the pattern of changes occurring within the maternal HPA axis. Absolute levels of cortisol may provide less information about the postnatal environment than do the fluctuations in cortisol. Some support for this notion can be derived from the frequently reported observation that cross-sectional correlations between psychological distress and cortisol are often null or very low, whereas robust cortisol increases are observed following exposure to a stressor. This difference seems to indicate that overall measures of cortisol carry little information about an individual's psychological experience, but dynamic changes in cortisol strongly reflect experience. During pregnancy, cortisol is dynamically related to interactions with the environment such that maternal experiences of psychological distress are related to both momentary increases in cortisol (Giesbrecht et al., Reference Giesbrecht, Campbell, Letourneau, Kooistra and Kaplan2012) and overall changes in diurnal patterns (Obel et al., Reference Obel, Hedegaard, Henriksen, Secher, Olsen and Levine2005). In this way, dynamic measures of cortisol represent signals to the developing fetus about maternal experience within her environment and the environment into which the fetus will be born. These findings suggest that researchers interested in the effects of prenatal stress exposures on offspring HPA axis function should plan to collect dynamic measures of maternal HPA axis function, as opposed to single samples of blood, saliva, or hair that can only provide overall estimates of secretory output.
In general, the timing of exposure during gestation had little effect on the outcomes. Nevertheless, the findings for earlier pregnancy were consistently stronger compared to those observed for later pregnancy, and the interaction among cortisol, distress, and sex was significant only for earlier pregnancy. This is consistent with a common theme among developmental studies that biological systems are most sensitive to external influences during periods of rapid development (Bornstein, Reference Bornstein1989). From this perspective, infant HPA axis function should be most strongly associated with exposures during the period around midgestation when the fetal HPA axis becomes functional (Gitau, Fisk, Teixeira, Cameron, & Glover, Reference Gitau, Fisk, Teixeira, Cameron and Glover2001). Such effects were observed in a large population-based cohort of 1.1 million births in which the risk for stress-related disorders among offspring, especially males, was highest when exposure to severe life events occurred in the second trimester (Khashan et al., Reference Khashan, McNamee, Henriksen, Pedersen, Kenny and Abel2011). Nevertheless, several studies have found that stress and cortisol exposures in later gestation have more pronounced effects on the developing HPA axis compared to earlier exposures (Davis et al., Reference Davis, Glynn, Waffarn and Sandman2011; Vedhara et al., Reference Vedhara, Metcalfe, Brant, Crown, Northstone and Dawe2012; Yehuda et al., Reference Yehuda, Engel, Brand, Seckl, Marcus and Berkowitz2005). The reasons for these discrepancies are not clear; however, we note that the cortisol and distress exposures in our study are continuous over the period of gestation and not discrete as might be observed for a significant life event. For example, the experience of a natural disaster and associated stress would be confined to the third trimester if it initially occurred in the third trimester. Thus, in the context of our research, we do not expect strong timing effects.
Because the postnatal environment also appears to shape infant behavioral and physiological responses to stress (Bergman, Sarkar, Glover, & O'Connor, Reference Bergman, Sarkar, Glover and O'Connor2010), it is important to note that the effects we observed were independent of postnatal exposure to maternal anxiety and depression. Furthermore, the effects we observed were not explained by sociodemographic factors such as household income, maternal education, age, or prepregnancy body mass index, or by characteristics of the infant such as negative affect immediately prior to the stressor or gestational age at birth. These findings strengthen the hypothesis that prenatal exposures were independently associated with the sex differences we observed.
Strengths and limitations
The strengths of this study include its prospective design, multiple prenatal assessments, the inclusion of both psychological measures of distress and cortisol, the inclusion of potential confounders as covariates in our models, the consideration of previously overlooked sex differences, and adequate sample size to test for sex-specific effects. Nevertheless, the interpretation and implications of these findings are subject to several limitations. First, the participants comprised the community sample with relatively high family income and education. It should be noted that although the median household income in the sample was high, it is equivalent to the median household income of the greater metropolitan region in which the sample was recruited. Second, although a range of distress scores was observed in this sample, the sample as a whole reported overall levels of distress that were low. We note that the majority of previous studies assessing the links between maternal prenatal stress and infant outcomes have done so with community samples comparable in terms of SES and distress to the current sample (see, e.g., DiPietro, Kivlighan, Costigan, & Laudenslager, Reference DiPietro, Kivlighan, Costigan and Laudenslager2009; Jacobs et al., Reference Jacobs, Myin-Germeys, Derom, Delespaul, van Os and Nicolson2007). Third, it was not possible to schedule the blood draw at the same time of day across our sample or to use the same blood draw method for all infants. Nevertheless, neither time of day nor blood draw method was associated with infant cortisol in any of our models, and this decreases the possibility that it may have contributed to the results. Fourth, a longer period of acclimation to the lab and multiple baseline samples would have increased confidence that the baseline measures were an accurate reflection of true basal levels. Fifth and finally, it will be important in future studies to account for the effects of sex-specific placental responses to maternal distress. Placental sex differences in the expression of 11ßHSD2 (lower when the fetus is female; Mericq et al., Reference Mericq, Medina, Kakarieka, Marquez, Johnson and Iniguez2009) and glucocorticoid receptors (higher in males; Saif et al., Reference Saif, Hodyl, Stark, Fuller, Cole and Lu2015) suggest that the placenta may regulate cortisol metabolism in sex-specific ways, and thereby contribute to sex differences in development of the fetal HPA axis.
Conclusion
The results of the present study showed that the associations between maternal cortisol during pregnancy and infant cortisol reactivity are moderated by sex and by maternal psychological distress. These findings suggest that the effects of in utero cortisol exposure on HPA axis function in children is different for boys and girls and that these effects depend on simultaneous in utero exposure to maternal distress, at least for girls. Further investigation is needed to determine whether sex-specific alteration of infant HPA axis function consequent to in utero cortisol and distress exposure links these prenatal exposures to sex differences in the development of psychopathology.
Appendix A
Hierarchical linear modeling coefficients (standard errors) for prenatal cortisol T2, prenatal stress T2, and infant sex with infant cortisol reactivity as outcome
