Professional Documents
Culture Documents
To cite this article: M.H. Bublitz, G. Bourjeily, C. Bilodeau & L.R. Stroud (2019): Maternal
circadian cortisol mediates the link between prenatal distress and breastfeeding, Stress, DOI:
10.1080/10253890.2018.1501023
ORIGINAL ARTICLE
Maternal circadian cortisol mediates the link between prenatal distress and
breastfeeding
M.H. Bublitza,b,c, G. Bourjeilya,b, C. Bilodeaua,b and L.R. Strouda,c
a
The Miriam Hospital, Providence, RI, USA; bDepartment of Medicine, Alpert Medical School of Brown University, Providence, RI, USA;
c
Department of Psychiatry and Human Behavior, Alpert Medical School of Brown University, Providence, RI, USA
LAY SUMMARY
Pregnant women who reported lower socioeconomic status in pregnancy were less likely to breast-
feed. This association was mediated by lower cortisol awakening responses, but not evening corti-
sol levels, over pregnancy.
CONTACT Margaret Bublitz margaret_bublitz@brown.edu Women’s Medicine Collaborative Suite 1F, 146 West River Street, Providence, RI, 02904, USA
ß 2019 Informa UK Limited, trading as Taylor & Francis Group
2 M.H. BUBLITZ ET AL.
Weerth & Buitelaar, 2005; Sandman, Wadhwa, Chicz-DeMet, (SD ¼ 1) weeks; Session 3: 36 (SD ¼ 1) weeks) and one time in
Dunkel-Schetter, & Porto, 1997), it is possible that psycho- the postpartum period (30 days after delivery). 100% of
social disturbances in pregnancy interrupt mammary secre- participants completed at least 1 study interview, 81% com-
tory differentiation and activation via changes in maternal pleted at least 2 study interviews, 72% completed all three
HPA activity, resulting in a decreased ability to breastfeed. pregnancy interviews, and approximately 85% completed the
Despite the importance of glucocorticoids secretion on postpartum interview. At the baseline interview, participants
mammary development in pregnancy, no prior studies have provided information on maternal characteristics (including
examined maternal HPA activity as a mechanism linking socioeconomic status) and pregnancy health information. At
psychosocial stress in pregnancy to breastfeeding. Therefore, each study session participants reported on symptoms of
in this study we aimed to examine if maternal circadian corti- stress, depression, and anxiety in pregnancy. For three days
sol mediated the association between psychosocial stress in following each interview, participants provided saliva samples
pregnancy and breastfeeding in the postpartum period. We (passive drool) upon awakening, 30 min after wake-up, and
hypothesized that the association between higher psychosocial bedtime. After participants completed the 3 days of saliva
stress and decreased likelihood to breastfeed would be medi- collection, study staff retrieved samples from participants’
ated by HPA dysregulation. An improved understanding of the
homes and provided payment.
pathways linking maternal distress in pregnancy to breastfeed-
ing is necessary for the effective development of interventions
to promote breastfeeding in the postpartum period. Measures
Breastfeeding status
Methods At approximately 1 month after delivery, participants were
asked to report their breastfeeding status. Women were then
Participants
categorized into three groups: those who reported exclu-
Participants were 197 pregnant women with singleton preg- sively breastfeeding for the first month after delivery, those
nancies whose breastfeeding status at 1 month after delivery who reported exclusively formula feeding for the first month
was known and who were part of a larger study of the after delivery, and those who did a mix of breast and formula
effects of maternal mood on fetal and infant development feeding. At each interview in pregnancy, participants were
(citation omitted for blind review). Women were excluded asked about their intentions to breastfeed using the follow-
from participating if they were less than 18 or greater than ing question: “How are you planning on feeding your baby?”
40 years old and if they were at risk for maternal or obstetric Response options were breastfeed only, breast and formula
complications. Maternal and infant characteristics, categorized feed, formula feed only, not sure, and refused to answer.
according to breastfeeding status, are presented in Table 1.
About 34% of women reported exclusively breastfeeding
in the first month after delivery, 18% reported exclusively Maternal cortisol
formula feeding, and 48% reported some combination of Up to 36 saliva samples were collected from each participant
both breast and formula feeding (“mixed feeding”). Women throughout the study and returned to the lab by study staff,
in the current study were on average 27 years old (SD ¼ 6), aliquoted, and stored at 80 C until analysis. Samples were
from racially and ethnically diverse backgrounds (45% non- then shipped to the laboratory of Clemens Kirschbaum, Ph.D.
Hispanic White, 17% non-Hispanic Black, 26% Hispanic, 6% (Dresden University). Cortisol concentrations were analyzed
>1 race, 4% Asian, 1% “other,” 1% American Indian), 43% with an immunoassay with time-resolved fluorescence detec-
were married, 46% of pregnancies were planned, 37% had a tion. The intra- and inter-assay coefficients of variation were
high school education or less, and 17% of the sample less than 8%. A subset of the sample (17%) were given
reported an annual income of less than $10,000 USD. 34% of Medication Event Monitoring System (MEMS) caps (AARDEX,
the sample was categorized as “low” SES according to the Zurich, Switzerland) that time stamp every opening of bottles
Hollingshead Four Factor Scale. Infants were born on average containing tubes used to collect maternal saliva. For partici-
at 39 weeks gestation (SD ¼ 2), average weight was 3341 g pants with both self-reported and MEMS-recorded saliva
(SD ¼ 542), and APGAR (Appearance, Pulse, Grimace, Activity, sampling times, there was an average time discrepancy of
and Respiration) score at 5 min after delivery was 9 (SD ¼ 1) 8 min (SD ¼ 5) suggesting that, in a subset of participants,
demonstrating good health among infants born in this sam- women were adherent to the sampling protocol.
ple. The Institutional Review Boards of two hospitals
approved this study. All women provided written consent
prior to their participation. Research was conducted in an Maternal and infant conditions
ethical and responsible manner, and is in full compliance Maternal medical conditions were assessed via chart review
with all relevant codes of experimentation and legislation. at enrollment and following delivery. Medical conditions
included diagnosis of gestational diabetes mellitus (GDM),
preeclampsia, and hypertension. Infant outcomes, including
Procedure
gestational age at birth, birth weight, and APGAR score at
Pregnant women were interviewed one to three times during 5 min after delivery, were collected from medical chart review
pregnancy (Session 1: 24 (SD ¼ 3) weeks; Session 2: 30 following delivery.
STRESS 3
Table 2. The association between socioeconomic status (SES) and breastfeed- QIDS is a 16-item measure designed to assess depressive
ing is mediated by cortisol awakening response.
symptom severity in the past 7 days. The QIDS has demon-
Variable b SE p-value 95% confidence interval
strated acceptable psychometric properties in depressed
Model 1:
Formula vs. breastfeedinga
patients (Rush et al., 1986). Scores on the QIDS range from 0
SES 2.03 .67 .002 .04 to .49 to 27, with scores >15 indicating severe symptoms
Mixed vs. breastfeeding of depression.
SES 1.20 .57 .04 .10 to .92
Model 2:
b
Formula vs. breastfeeding
SES .93 .97 .34 .06 to 2.66 Maternal anxiety symptoms
CAR (24 weeks’ gestation) .19 .96 .84 .18 to 7.93 Maternal anxiety symptoms were measured by asking women
CAR (30 weeks’ gestation) .14 1.00 .89 .16 to 8.22
CAR (36 weeks’ gestation) .16 1.05 .88 .15 to 9.16 to rate the severity of symptoms across 14 items that meas-
Mixed vs. breastfeeding ure mental agitation, psychological distress, and somatic
SES 1.02 .76 .18 .08 to 1.60
CAR (24 weeks’ gestation) .72 .73 .32 .50 to 8.61 anxiety. Response options range from 0 ¼ not present to
CAR (30 weeks’ gestation) .65 .72 .36 .47 to 7.78 4 ¼ very severe. Scores >24 indicate severe symptoms
CAR (36 weeks’ gestation) .86 .77 .26 .52 to 10.64 (Hamilton, 1959).
Note. All analyses are adjusted for maternal age, pre-pregnancy BMI, and race.
Breastfeeding group has three categories: exclusive breastfeeding, exclusive
formula feeding, and mixed feeding. Breastfeeding is the reference category.
a
Log likelihood ¼ 260.56, df ¼ 8, p < .001. Maternal perceived stress
b
Log likelihood ¼ 153.08, df ¼ 14, p < .001. The perceived stress scale is a 10-item self-report measure
that asks participants to report the degree to which they
Socioeconomic status (SES) perceived situations as stressful over the past month on a
Participants were asked to self-report the current occupation scale of 0 ¼ never to 4 ¼ very often. Higher scores reflect
and highest level of education for themselves and another greater stress (Cohen, Kamarck, & Mermelstein, 1983).
contributing adult at their first study session. To measure SES
we utilized the Hollingshead Four Factor Scale (Hollingshead,
1975). This measure is an index of SES based on weighted Statistical approach
scores of education and occupation of contributing adults.
Cortisol data reduction
Occupations were coded according to the Hollingshead Four
We calculated the cortisol awakening response (CAR) (Clow,
Factor occupational codes. Hollingshead scores range from
Thorn, Evans, & Hucklebridge, 2004) on each day of saliva
1 to 5, with scores of 4 and 5 indicating low SES. Scores
collection by taking the difference between cortisol values
were categorized into “low” (scores 4 and 5) and “high”
upon awakening and 30 min after wake-up and then averag-
(scores 1–3). The Hollingshead measure is one of the most
ing CAR across the three days of collection. Morning saliva
frequently used measures of socioeconomic position and has
samples that were less than 20 or greater than 40 min apart
demonstrated good inter rater reliability in past research
were omitted from analyses in order to accurately capture
(Cirino et al., 2002).
the cortisol awakening response. Valid maternal cortisol data
was available for 163 women at the first interview, 176
Maternal depressive symptoms women at the second interview, and 170 women at the third
At each study session, pregnant mothers reported their interview. CAR and evening cortisol are regulated by different
depressive symptoms using the Quick Inventory for biological processes (Wilhelm, Born, Kudielka, Schlotz, &
Depressive Symptomatology (QIDS) (Rush et al., 1986). The Wust, 2007) and thus were analyzed separately. CAR and
4 M.H. BUBLITZ ET AL.
25
20
Corsol (nmol/L)
15
10
evening cortisol values were log transformed due to skewed values >.10). Pre-pregnancy BMI, maternal age, and race/eth-
distributions. Raw values are presented in figures. nicity were included as covariates in subsequent analyses.
Hypothesis testing
To assess the influence of potential covariates, we performed
Breastfeeding and maternal distress in pregnancy
one-way analysis of variance (ANOVA) and chi-squared
analyses to examine differences in maternal and infant A multinomial logistic regression was performed to model
characteristics by the three breastfeeding groups (formula the relationship between maternal distress in pregnancy
feeding, breastfeeding, mixed feeding). Characteristics that (depression, anxiety, stress, SES) and membership in the three
differed by breastfeeding group were included as covariates breastfeeding groups (breastfeeding, formula feeding, or
in subsequent analyses. Repeated measures analysis of mixed). After adjusting for maternal age, race, and pre-preg-
covariance (ANCOVA) were performed to assess if maternal nancy BMI, results indicated that the addition of SES
cortisol values over pregnancy differed by breastfeeding (v2 ¼ 10.29, p ¼ .006, df ¼ 2) significantly improved the fit
groups. Multinomial logistic regression analyses were per- between model and data. The reference group was breast-
formed to examine if indices of maternal distress predicted feeding. Pregnant women who reported lower SES were
breastfeeding group, and if maternal cortisol mediated the more likely to formula feeding (b ¼ 2.03, SE ¼ .67, p ¼ .002)
association between maternal stress and breastfeeding. and mixed feeding (b ¼ 1.20, SE ¼ .57, p ¼ .04) than breast-
feed. Breastfeeding group membership did not differ by
Results perceived stress, depression, or anxiety symptoms in preg-
nancy (p values >.22).
Breastfeeding groups significantly differed on a number of
psychosocial characteristics including pre-pregnancy body
mass index (F(2,186) ¼ 5.79, p ¼ .004), maternal age
(F(2,194) ¼ 18.08, p < .001), marital status (v2 = 31.57, p < .001, SES and cortisol in pregnancy
df = 2), and race/ethnicity (v2 ¼ 9.11, p ¼ .01, df = 2). Women
who reported formula feeding in the first month after deliv- We next examined if maternal SES in pregnancy was associ-
ery were significantly younger and had higher body mass ated with maternal cortisol by performing multivariate
index prior to pregnancy. Women who breastfed were more analysis of variance. After adjusting for significant covariates,
likely to be Caucasian. Breastfeeding groups did not signifi- we found that SES was significantly associated with CAR at
cantly differ on maternal medical conditions (gestational dia- 30 weeks (b ¼ 3.96, p ¼ .05) and was marginally associated
betes mellitus, hypertension, or preeclampsia) (p values >.23), with evening cortisol at 24 (b ¼ 3.06, p = .08) and 36 weeks
parity (p ¼ .18), delivery mode (p ¼ .75), or postpartum (b ¼ 3.42, p = .07). Women who reported lower SES displayed
depression (p ¼ .42). Groups did not significantly differ on lower CAR and marginally higher evening cortisol, represent-
infant outcomes including gestational age at birth, small for ing a flattening of the diurnal rhythm across pregnancy
gestational age, and APGAR scores 5 min after delivery (p among lower SES pregnant women.
STRESS 5
Breastfeeding and cortisol in pregnancy among women who reported low SES may have been
insufficient for mammary epithelial cells to differentiate into
Results from repeated measures analysis of covariance
lactocytes, potentially resulting in breastfeeding difficulty
revealed a main effect of breastfeeding group on maternal
and greater rates of formula feeding in the first
CAR across pregnancy (F(2) ¼ 4.29, p ¼ .02). Tukey’s post hoc
month postpartum.
comparison revealed that women who formula fed had
Conversely, CARs were highest over pregnancy among
significantly lower CAR over pregnancy compared those who
women who mixed fed. Elevated CARs may have provided
mixed fed (p ¼ .03). We did not find evidence of a main effect
levels of glucocorticoids necessary for mammary differenti-
of breastfeeding group differences in evening cortisol in
ation and activation. Alternatively, elevated CAR may repre-
pregnancy (F(2) ¼ 1.13, p ¼ .33). Univariate analysis of covari-
sent maternal HPA dysregulation, leading to difficulties with
ance revealed that differences were significant at 24 weeks’
breastfeeding (particularly among women who did some
gestation (F ¼ 6.45, p ¼ .002). Post hoc paired comparisons
degree of formula feeding). Past research suggests that CARs
indicated that cortisol awakening responses were significantly
that fail to decline over pregnancy are associated with
lower among women who formula fed compared with those
who mixed fed (p ¼ .03). We did not observe significant dif- adverse pregnancy outcomes including shortened gestational
ferences in cortisol awakening responses by breastfeeding length (Buss et al., 2009). Therefore, results from this study
group at 30 weeks’ gestation (F ¼ 2.62, p ¼ .08) or at 36 suggest that both low and high awakening responses may
weeks’ gestation (F ¼ 5.78, p ¼ .35) (Figure 1). indicate HPA dysregulation and negatively affect breastfeed-
ing success in the postpartum period. However, the mecha-
nisms through which elevated maternal cortisol would
Cortisol as a mediator between maternal distress and impact mammary secretory differentiation and activation are
breastfeeding unclear, and we do not yet understand if there is an “ideal”
cortisol level to promote breastfeeding success.
Given the significant associations between SES and CAR, and
In contrast to previous studies, we did not observe associ-
breastfeeding group and CAR, we examined maternal CAR in
ations between maternal symptoms of depression, anxiety, or
pregnancy as a mediator in the association between maternal
SES and breastfeeding group membership by performing perceived stress and breastfeeding. The lack of an association
multinomial logistic regression. Breastfeeding was again the may have been due to the acuity of symptoms, as women
reference group. We found that, after adjusting for CAR at were asked to report on mental health symptoms within the
24, 30, and 36 weeks’ gestation, SES no longer predicted past week or month. The observed association between low
breastfeeding group membership. Specifically, after adjusting SES and breastfeeding could have been due to the chronicity
for CAR there were no differences in formula feeding of low SES, which likely pre-dated pregnancy and continued
(b ¼ .93, SE ¼ .97, p ¼ .40) or mixed feeding (b ¼ 1.02, into the postpartum period, which may have influenced
SE ¼ .76, p ¼ .36) by maternal SES (Table 2). maternal HPA activity. Consistent with our findings, past
research indicates that chronic stressful life events in early
pregnancy are associated with lower CAR (Obel et al., 2005),
Discussion indicating that low SES in pregnancy may have resulted in
Results from this study indicated that women who report low blunted CAR and in turn disrupted mammary development.
SES in pregnancy are less likely to breastfeed in the postpar- While we tested for a number of demographic characteristics
tum period. The association between SES and breastfeeding that may have accounted for differences in breastfeeding,
was mediated by maternal CAR over pregnancy. These there are likely to be other un-measured sources of psycho-
findings are consistent with prior research indicating an social stress or un-measured societal constraints (such as
association between low SES in pregnancy and reductions in early return to work, paid maternity leave, accommodations
breastfeeding. Findings are also in line with past evidence for pumping breastmilk at work) that may have also contrib-
that glucocorticoids in mid-gestation are necessary for mam- uted to both attenuated CAR and the likelihood of formula
mary differentiation, activation, and milk production. Results feeding. Future research is needed to address this limitation.
from this study extend our understanding of maternal HPA Cortisol awakening responses significantly differed by
activity as a mechanism in the association between maternal breastfeeding group, but evening cortisol levels did not sig-
distress in pregnancy and breastfeeding behavior in the post- nificantly differ by group. Cortisol awakening response and
partum period. evening cortisol are under different neuroregulatory control
We can articulate several possible explanations for study (Wilhelm et al., 2007). The findings from this study suggest
findings. We speculate that lower CAR observed in women that neural regions that regulate the awakening response
with low SES may have interfered with mammary secretory including the hippocampus and hypothalamic suprachias-
differentiation that precedes secretory activation and milk matic nucleus (Clow, Hucklebridge, & Thorn, 2010) may also
production. Others have also proposed that there may be be important in mammary secretory differentiation and acti-
common neuroendocrine pathways between perinatal stress vation. This has not yet been studied. However, evidence has
and breastfeeding (Stuebe et al., 2012). Past research has demonstrated differences in patterns of maternal neural acti-
found that hormonal shifts necessary for secretory differenti- vation to infant cues among women who are breastfeeding
ation, including maternal glucocorticoid production, begin as vs. those who are not (Kim et al., 2011). More research is
early as 20 weeks’ gestation. Thus, lower CARs observed needed to understand the links between patterns of HPA
6 M.H. BUBLITZ ET AL.
Obel, C., Hedegaard, M., Henriksen, T.B., Secher, N.J., Olsen, J., & Levine, Thulier, D., & Mercer, J. (2009). Variables associated with breastfeeding
S. (2005). Stress and salivary cortisol during pregnancy. duration. The Journal of Obstetric, Gynecologic, and Neonatal Nursing,
Psychoneuroendocrinology, 30, 647–656. doi:10.1016/j.psyneuen.2004.11.006 38, 259–268. doi:10.1111/j.1552-6909.2009.01021.x
Pang, W.W., & Hartmann, P.E. (2007). Initiation of human lactation: Secretory Whalen, B., & Cramton, R. (2010). Overcoming barriers to breastfeeding
differentiation and secretory activation. Journal of Mammary Gland continuation and exclusivity. Current Opinion in Pediatrics, 22, 655–663.
Biology and Neoplasia, 12, 211–221. doi:10.1007/s10911-007-9054-4 doi:10.1097/MOP.0b013e32833c8996
Rush, A.J., Giles, D.E., Schlesser, M.A., Fulton, C.L., Weissenburger, J., & Wijndaele, K., Lakshman, R., Landsbaugh, J.R., Ong, K.K., & Ogilvie, D.
Burns, C. (1986). The Inventory for depressive symptomatology (IDS): (2009). Determinants of early weaning and use of unmodified cow’s
Preliminary findings. Psychiatry Research, 18, 65–87. doi:10.1016/0165- milk in infants: A systematic review. Journal of the American Dietetic
1781(86)90060-0 Association, 109, 2017–2028. doi:10.1016/j.jada.2009.09.003
Sandman, C.A., Wadhwa, P.D., Chicz-DeMet, A., Dunkel-Schetter, C., & Wilhelm, I., Born, J., Kudielka, B.M., Schlotz, W., & Wust, S. (2007). Is the corti-
Porto, M. (1997). Maternal stress, HPA activity, and fetal/infant out- sol awakening rise a response to awakening? Psychoneuroendocrinology,
come. Annals of the New York Academy of Sciences, 814, 266–275. 32, 358–366. doi:10.1016/j.psyneuen.2007.01.008
doi:10.1111/j.1749-6632.1997.tb46162.x Zhu, P., Hao, J., Jiang, X., Huang, K., & Tao, F. (2013). New insight into
Stuebe, A.M., Grewen, K., Pedersen, C.A., Propper, C., & Meltzer-Brody, S. onset of lactation: Mediating the negative effect of multiple perinatal
(2012). Failed lactation and perinatal depression: Common problems biopsychosocial stress on breastfeeding duration. Breastfeeding
with shared neuroendocrine mechanisms? Journal of Womens Health Medicine, 8, 151–158. doi:10.1089/bfm.2012.0010
(Larchmt), 21, 264–272. doi:10.1089/jwh.2011.3083