Does maternal care-giving behavior modulate the cortisol response to an acute stressor in 5-week-old human infants?

Abstract

In previous studies, a higher quality of care-giving behavior reduced the cortisol response to acute stressors in infants aged 3 months and older. Here, we investigated whether the quality of maternal care-giving behavior affected the cortisol response to being bathed in 5-week-old infants (N = 141). Mothers and infants were observed during a bathing routine. Infant saliva samples were collected before and after bathing to assess cortisol concentrations, and the quality of maternal care-giving behavior was scored from videotapes. Bathing elicited a significant increase in infant salivary cortisol level (reactivity), and cortisol concentrations returned to pre-stressor values 40 min after bathing (recovery). In contrast, with previous findings in older infants, the quality of maternal care giving was not associated with either cortisol reactivity or recovery. This finding suggests that the quality of maternal care-giving behavior is not effective in modulating 5-week-old infants' cortisol responses to a (mild) physical stressor. Although a satisfactory neurophysiological explanation for this inference is still lacking, current knowledge of the behavioral development of very young infants supports this suggestion.

Keywords::

• Infant
• cortisol
• stress
• maternal behavior
• reaction
• recovery

Introduction

Human infants generally respond to challenging physical stimuli (stressors) with an increase in secretion of the stress–responsive hormone cortisol (Gunnar et al. 2009; Jansen et al. 2010). Stressors investigated in the past range from painful procedures, such as vaccination (Davis and Granger 2009) to mild everyday stimuli, such as bathing routine (Albers et al. 2008). The cortisol response generally entails both the cortisol reaction (i.e. the increase in post-stressor cortisol concentrations), and cortisol recovery (i.e. the return to pre-stressor levels; Linden et al. 1997). The aim of the present study was to investigate whether higher quality maternal behavior could help infants regulate their cortisol response to a bathing routine.

Previous studies have shown that the amount of maternal soothing does not affect the infant cortisol response to stressors (Lewis and Ramsay 1999a,b; Braarud and Stormark 2006). However, as Lewis and Ramsay (1999b) suggest, this finding leaves open the possibility that qualitative aspects of maternal care-giving behavior may modulate infant distress. Two key measures of the quality of care-giving behavior are sensitivity and cooperation (Ainsworth et al. 1978). Sensitivity refers to the extent to which care givers timeously and appropriately respond to the infant's needs and signals. Cooperation refers to the extent to which care givers adjust their behavior to the infant's ongoing activity rather than interfering with the infant's actions. By being sensitive and cooperative with their infant, care givers can improve the controllability and predictability of a situation for an infant. There is a large body of both animal and human empirical evidence showing that predictability and controllability are key factors determining the stressfulness of a situation (Dickerson and Kemeny 2004). Thus, sensitive and cooperative care giving may provide the infant with an external source of stress regulation. This external regulation would be, especially, important when the infant's self-regulation capacities are insufficiently developed.

Indeed, several studies have shown that sensitive care giving reduces the cortisol reaction to a challenging stimulus in infants (Gunnar et al. 1992; Dettling et al. 2000). However, these studies were carried out on infants older than 6 months. In infants under 6 months of age, studies by Haley and Stansbury (2003) and by Albers et al. (2008) show that sensitivity does not affect the cortisol reaction to a stressor. However, these authors did find that recovery from stress is better when mothers are more sensitive. Although they did not report cortisol recovery scores, Haley and Stansbury (2003) provided evidence that in 5-month-old infants, both distress behaviors and heart rate decrease more after a stressor, when mothers are more sensitive. Albers et al. (2008) reported that in 3-month-old infants, cortisol concentrations return to pre-stressor levels faster when mothers are more sensitive.

Thus, the quality of maternal care giving appears to reduce cortisol reactivity after 6 months of age, and to promote cortisol recovery from 3 months onward, but it is yet unclear whether maternal care-giving quality is related to the cortisol response in even younger infants. Studies suggest that before 3 months of age, the infant brain may not be able to process complex social information, such as visual and auditory qualities of maternal care-giving behavior (Grossmann and Johnson 2007). Therefore, although clear forms of maternal soothing, such as holding the infant or use of a pacifier, may affect the infant stress reaction (Kawakami et al. 1996; Kleberg et al. 2008), it is unknown if more complex and subtle differences in maternal behavior are effective in modulating physiological stress in very young infants. Whether cortisol responses of young infants to stress and the quality of maternal care-giving behavior are associated is therefore unknown.

In order to investigate the effects of maternal care-giving quality on cortisol reactivity and recovery in younger infants, we observed 5-week-old infants and their mothers during a bathing routine and assessed whether the quality of maternal care giving was related to the infant salivary cortisol response. In line with the study by Albers et al. (2008), we hypothesized that cortisol reactivity in response to a bathing routine would not be associated with maternal care-giving behavior, but that cortisol recovery would be better with higher quality maternal care giving.

Material and methods

Participants

This study is part of an ongoing longitudinal project on the role of early care-giving factors in infant development. The Radboud University ethical committee for behavioral sciences gave ethical approval for the project. Prospective participants were recruited during pregnancy through flyers handed out to pregnant women by midwife clinics in and around the cities of Nijmegen and Arnhem. We accepted 219 women as participants. They all gave their written informed consent before the start of the project. All infants included in the project were healthy, born at full term and had a 5-min APGAR score ≥ 7.

The participants in this study were a subsample of this larger population and consisted of the 141 mothers and infants from whom sufficient data were collected (i.e. a pre-stressor and at least one post-stressor cortisol sample, in addition to video-material of the bathing routine). All participants were Dutch and Caucasian. Table I provides further demographic characteristics of our study population. Our subsample did not differ from the larger population in any of the presented characteristics (Mann–Whitney U-test; |Z| < 1.49, p>0.1).

Table I.  Demographics and characteristics of the mothers and infants for the study sample.*

	Study sample (n = 141)
Maternal characteristics
Mean age (in years)	32.4 (4.0)
Living with partner (%)	98.6
Employment outside the home (%)	83.7
Education level (%)
Low	3.6
Middle	19.8
High	75.9
Prenatal smoking (%)
Smoking	3.5
Non-smoking	96.5
Infant characteristics
Mean age at visit (in days)	34.5 (3.7)
Female (%)	46.8
First born (%)	39.7
Duration of gestation (in days)	267.8 (7.8)
Birth weight (in grams)	3629 (465)
APGAR score	9.7 (0.6)

* Note: Data are given as mean (SD) unless otherwise indicated.

Procedure

We contacted mothers by telephone within 3 weeks after delivery to schedule a home visit when the infants were approximately 5 weeks of age, and at the time of day when they usually bathed the infant. Visits were, therefore, carried out both in the morning and afternoon. During the home visit, we asked the mothers to bathe their infant as they would normally do and unobtrusively videotaped the routine (undressing, bathing, and dressing). Consequently, the procedure was not standardized, but differed in duration between participants. After the infant was dressed, we instructed the mothers to do what they would normally do after a bath (i.e. feed their infant, put the infant to sleep). Saliva samples were taken on arrival (T1) and 25 (T2) and 40 (T3) min after the infant was taken out of the bath. The first saliva sample (T1) was collected within 10 min after arrival of the experimenter. The bathing routine commenced around 10 min after T1 (median: 10 min, interquartile range, IQR: 5–21 min).

Measures

Quality of maternal care giving

Similar to Albers et al. (2008), the videotaped bathing routines were rated for maternal sensitivity and cooperation using two 9-point rating scales (Ainsworth et al. 1978); higher scores reflect more sensitivity/cooperation. Sensitivity refers to the extent to which care givers timeously and appropriately respond to the infant's needs and signals. Cooperation refers to the extent to which care givers adjust their behavior to the infant's ongoing activity rather than interfering with the infant's actions.

Trained observers, who did not know the mothers and infants they were observing and were blind with regard to the cortisol data, independently rated the interactions. Interobserver reliability was very good (Cohen's κ: 0.90 for both sensitivity and cooperation).

Salivary cortisol

Infant saliva samples were collected using sorbette eye sponges (Salimetrics, LLC, State College, PA, USA). Sorbettes are small absorbent sponges attached to a plastic shaft that can be used to collect saliva effectively. The sorbette can absorb small quantities of saliva, and cortisol recovery is good (De Weerth et al. 2007; Donzella et al. 2008). After the home visit, saliva samples were stored at − 25°C until further analysis. Salivary cortisol concentrations were assessed in duplicate at the Laboratory of Endocrinology of the University Medical Centre Utrecht in a competitive radioimmunoassay previously described in De Weerth et al. (2007). The lower limit of detection was 1.0 nmol/l, and inter-assay and intra-assay variation were below 10%.

Infant behavioral response (crying and fussing)

To substantiate the use of a bathing routine as a stressor, we also quantified the behavioral response to bathing. Crying and fussing were scored by two assistants as whether or not the infant had cried or fussed in 5-s intervals, using The Observer 5.0 (Noldus B.V., Wageningen, The Netherlands). We scored “crying” when the infant cried for more than 3 out of 5 s. If the infant cried less, or only whimpered, we scored “fussing”. For the statistical analyses, the total amount of infant crying and fussing was computed as a percentage per episode (undressing, bathing, and dressing) and a percentage of the total duration of the bathing routine. Interobserver reliability calculated on 15% of the videos was good (Cohen's κ: 0.85).

Statistical procedure

Analysis of the data was carried out in a two-step process

In the first step, presented as Preliminary results section, the infant response to bathing was examined, and cortisol measures and maternal sensitivity and cooperation were examined for intercorrelations. Furthermore, infant demographic variables and sampling factors (i.e. time of visit, duration of bathing routine) were tested against the infant cortisol variables to search for potential confounding effects on the infant cortisol response. All variables associated with cortisol reactivity or recovery scores at p < 0.1 were considered as potential confounders. Spearman tests for intercorrelations revealed no multicollinearity between demographic variables and/or sampling factors (r < 0.7; Tabachnick and Fidell 2007). Salivary cortisol concentration data were log-transformed to achieve normality of the data. Differences and correlations in cortisol concentrations between sampling times were tested with a one-way repeated measures ANOVA, with post hoc paired t-tests between sample differences, and Pearson correlations. Due to skewness of the infant crying and fussing scores, and the demographic and sampling variables, all analyses involving these variables were carried out using non-parametric tests (Spearman correlations, Kruskal–Wallis tests, and Mann–Whitney U-tests).

In the second step, presented as the main Results section, we examined the role of maternal care-giving quality in the infant cortisol response to bathing. From the original cortisol scores, a cortisol reactivity and recovery score were calculated from the raw cortisol data (see Preliminary results section). These scores were adequately distributed for use in parametric testing. Therefore, the role of maternal care-giving quality in the infant cortisol response was examined using a hierarchical multiple regression analysis, with variables identified as potential confounders entered in the first block, and maternal care-giving quality entered in the second block. For both regression models, missing data were replaced with the mean. This did not affect the overall outcome of the analyses. For both models, outliers were removed so that the residual error term was normally distributed, and Mahalanobis distances were within the norm (for the cortisol reactivity model, one participant was removed from the analysis; for the cortisol recovery model, three participants were removed).

Due to the skewness of the raw data, all data are presented as median scores and IQRs. The results of statistical tests were considered significant at p < 0.05.

Results

Preliminary results

Infant response to bathing

Table II shows the median percentages of infant crying and fussing during undressing, bathing, and dressing. The percentage of infant crying and fussing differed significantly between episodes [χ²(2,n = 138) = 79.96, p < 0.001]. In response to being taken out of the bath, only 24 infants (17.3% of participants, n = 139) did not cry or fuss at all.

Table II.  Infant crying and fussing and salivary cortisol concentrations in response to the bathing routine in 5-week-old human infants.

	Undressing	Bathing	Dressing
Crying and fussing (%)
Median	0^a	0^b	18.84^c
IQR	0–21.88	0–8.17	3.05–53.06
n	139	138	139
	Pre-stressor	Post-stressor (+25 min)	Post-stressor (+40 min)
Salivary cortisol (nmol/l)
Median	10.60^a	12.90^b	11.35^a
IQR	7.45–14.80	10.20–17.80	8.93–14.85
n	141	135	120

Notes: IQR, interquartile range. Post hoc analyses using Wilcoxon signed-rank tests for infant crying and fussing; paired t-tests for salivary cortisol (on log-transformed data; raw values reported here). Superscript letters (a, b, c) indicate between-group differences, with different letters denoting significant differences (all at p < 0.01).

Median cortisol concentrations in response to the bathing routine are presented in Table II. The bathing routine provoked a significant change in salivary cortisol concentrations (ANOVA: F_2,112 = 23.98, p < 0.001). In response to the bathing routine, 60.7% of the infants showed a cortisol reaction to bathing (T2 − T1>0), while 39.3% of the infants did not (T2 − T1 ≤ 0).

We found marginally significant correlations for infant crying and fussing between all three episodes [undressing × bathing: r(139) = 0.11, p = 0.09; bathing × dressing: r(138) = 0.35, p < 0.001; undressing × dressing: r(139) = 0.28, p < 0.001], and, therefore, calculated a total percentage of infant crying and fussing.

We found no correlations between T1 and T2 salivary cortisol concentrations [r(135) = 0.02, p = 0.99], indicating that the law of initial value (Wilder 1957; Jin 1992) did not apply to our data. Therefore, the cortisol reactivity score could be calculated as a delta score of T2 − T1 (Jessop and Turner 2008). T2 and T3 cortisol concentrations were significantly correlated [r(114) = 0.75, p < 0.001]. We calculated a cortisol recovery score by computing T3 − T1 (Albers et al. 2008). This recovery score was correlated with T2 [r(114) = 0.40, p < 0.001], indicating that the recovery score depended on the cortisol reactivity in response to the bathing routine. Distributions of both the cortisol reactivity and recovery score were sufficiently normal to include them as the dependent variables in the hierarchical regression analyses.

The total percentage of infant crying and fussing was positively correlated to infant cortisol reactivity [r(132) = 0.22, p < 0.01], and the correlation between the total percentage of infant crying, and fussing and cortisol recovery tended toward significance [r(117) = 0.13, p = 0.08].

Maternal care-giving behavior

Maternal sensitivity and maternal cooperation were highly correlated [r(141) = 0.81, p < 0.001]. Because this correlation was so high, an overall maternal care-giving quality score was computed by averaging the scores on both scales.

Demographic variables

Associations between infant demographic variables, infant cortisol reactivity, and infant cortisol recovery were examined. Parity was not associated with the infant cortisol response (reactivity Z = − 0.05, p = 0.96 and recovery Z = − 0.37, p = 0.71). Infant sex was not associated with the infant cortisol response (reactivity Z = − 0.49, p = 0.62 and recovery Z = − 0.09, p = 0.93). The duration of gestation was correlated with the cortisol response to bathing [reactivity r(97) = − 0.22, p < 0.05 and recovery: r(86) = − 0.20, p < 0.05]. Infant birth weight was correlated with the infant cortisol response to bathing [reactivity r(128) = − 0.19, p < 0.05 and recovery r(114) = − 0.17, p < 0.05]. The correlations between infant age at the home visit and the infant cortisol response were not significant [reactivity r(135) = 0.01, p = 0.48 and recovery r(120) = 0.02, p = 0.43]. Finally, infant APGAR scores were not significantly correlated to the cortisol response to bathing [reactivity r(121) = 0.09, p = 0.17 and recovery r(108) = 0.05, p = 0.32].

Sampling factors

The duration of the entire bathing routine was not correlated with the infant cortisol response [reactivity r(133) = − 0.06, p = 0.26 and recovery r(108) = − 0.12, p = 0.11]. The correlation between the duration of the dressing episode and infant cortisol reactivity was not significant [r(133) = − 0.07, p = 0.22]. The correlation between the duration of the dressing episode and recovery tended towards significance [r(118) = − 0.13, p = 0.08].

Time of day of the visit was not associated with the cortisol response (reactivity Z = − 0.95, p = 0.34 and recovery Z = − 0.41, p = 0.68).

The correlation between the duration of the latency from T1 to bathing (see Procedures section) and the cortisol response to bathing was not significant [reactivity r(132) = − 0.10, p = 0.13 and recovery r(118) = − 0.08, p = 0.18].

Main results

Maternal behavior and infant salivary cortisol concentration

For the regression model assessing the association between maternal behavioral quality and infant cortisol reactivity, duration of gestation, infant birth weight, and the amount of infant crying and fussing were included as potential confounders. For the regression model assessing the association between maternal behavioral quality and infant cortisol regulation, these potential confounders were also included, in addition to the duration of the dressing episode and T2 cortisol concentrations.

As shown in Table III, the overall regression model for cortisol reactivity significantly predicted 9% of all variance. However, of the potential confounders, only the percentage of infant crying and fussing was associated with cortisol reactivity, and maternal behavioral quality did not significantly add to the percentage of explained variance.

Table III.  Parameters for two hierarchical regression models for cortisol reactivity and cortisol recovery.

	β	t	ΔR^2
Model 1: cortisol reactivity (n = 140)
R^2 = 0.09, F = 3.49^**
Duration of gestation	− 0.07	− 0.76
Infant birth weight	− 0.13	− 1.39
Infant crying and fussing	0.23	2.59^*	0.09
Maternal behavioral quality	− 0.06	− 0.72	0.00
Model 2: cortisol recovery (n = 138)
R^2 = 0.24, F = 6.90^***
Duration of gestation	− 0.06	− 0.66
Infant birth weight	0.04	0.51
Duration of dressing episode	− 0.20	− 2.64**
Infant crying and fussing	− 0.01	− 0.07
T2 cortisol concentration	0.46	5.57^***	0.23
Maternal behavioral quality	0.10	1.20	0.08

Notes: ***p < 0.001, **p < 0.01, *p < 0.05. Indentations indicate successive regression steps.

The regression model for cortisol recovery significantly explained 24% of the variance (Table III). However, maternal behavioral quality did not significantly add to the percentage of explained variance of cortisol recovery. In our model, cortisol recovery is significantly associated with cortisol concentrations 25 min after being taken out of the bath (T2), and the duration of the dressing episode.

Discussion

In this study, we sought to investigate whether the quality of maternal care giving modulates infant cortisol responses to a mild physical stressor at 5 weeks of age. Specifically, we expected a higher quality of maternal care giving to be associated with a quicker infant cortisol recovery after a bathing routine. Our data showed that infants responded to the bathing routine with a significant increase in salivary cortisol concentration, followed by a significant decrease. However, the quality of maternal care giving was not associated with either cortisol reactivity or cortisol recovery.

In a previous study, a bathing routine resulted in an increase in cortisol level (Albers et al. 2008) in at term, healthy infants. Our study provides further evidence that a normal everyday situation that is characterized by considerable amounts of handling and a metabolic (i.e. cold) challenge can be stressful for very young infants, and may therefore be a “participant-friendly” paradigm to further investigate the stress response in very young infants.

More relevant for the aim of this paper, however, is our finding that cortisol reactivity and recovery in 5-week-old infants are not associated with maternal care-giving quality. These results contradict our hypothesis and differ from those of Albers et al. (2008), who found that 3-month-old infants showed a quicker cortisol recovery with higher quality maternal care giving.

As our study was partially a replication of the study by Albers et al. (2008), we utilized the same methodology; the same bathing situation, the same procedure to assess cortisol secretion, and the same scales to rate maternal care-giving behavior. The magnitude of the cortisol response in both studies was similar, with effect sizes of 0.47 vs. 0.45 for reactivity, and 0.80 vs. 0.72 for recovery for our study and the study by Albers et al. (2008), respectively. Furthermore, the median scores on both maternal sensitivity and maternal cooperation were comparable, as was their correlation [0.81 vs. 0.82 for our study and Albers et al. (2008), respectively]. Taken together, these similarities suggest that, with the age of assessment being the only substantial difference between the two studies, the younger age of our participants is the most probable explanation behind our non-significant findings.

However, our study was naturalistic in nature, and therefore a degree of variability existed in the manner in which each infant was bathed. Unfortunately, we were unable to assess all of these variables. Potential effects of external factors, such as room and water temperature, and infant hunger and sleep, that may have affected the infant cortisol response were not taken into account. Nonetheless, this was also the case in the study by Albers et al. (2008), and they did report an effect of maternal behavioral quality on cortisol recovery. Therefore, in spite of the naturalistic nature of our study, and the potential pitfalls associated with it, our findings likely indicate that the infant cortisol response and maternal care-giving quality are dissociated at this age.

It is possible that the experience of bathing was not a sufficient acute stressor to warrant the infant's need for seeking external regulation. However, the effect size of the cortisol reaction suggests that the infants were stressed by the bathing routine. This was also evident from the significant increase in infant crying after being taken out of the bath. Infant crying is generally seen as an attachment behavior aimed at eliciting regulation of homeostasis (Zeifman 2001; Coan 2008).

Another, possibly more plausible, explanation for the lack of association between infant cortisol level and maternal care giving is that the observed maternal behavior (i.e. sensitivity and cooperation) is not effective in reducing physiological stress in infants at this early age. This may be because at 5 weeks post partum the infant brain is not yet at a developmental stage where it can process complex social information in the form of refined caregiver verbal and non-verbal cues, especially, when under stress. Neurobiological research suggests that a milestone for normal development of the infant brain occurs in the third month after birth, with a rapid change in maturation of what is called the “social brain” (Grossmann and Johnson 2007). This developmental change could explain the emergence of caregiver–child face-to-face interactions that are characterized by mutual attunement of behavior and emotions (Feldman et al. 1999; Schore 2001). Due to methodological limitations, it is difficult to assess how these brain areas function before 3 months of age, but behavioral development suggests immaturity of the social brain prior to 3 months. Before 2 months of age, infants only display endogenous smiles, mainly during rapid eye movement (REM) sleep (Wolff 1987), while after 2 months of age, they develop social smiles (Emde et al. 1976). This coincides with a number of other changes (face perception from outline to internal features: Wolff 1987; Nelson 2001; emotion expression: Lavelli and Fogel 2005) that taken together suggest an increase in susceptibility to refined social cues. A final strand of evidence for a rapid change of the brain in the third month of life that might underlie the emergence of social regulation of infant arousal comes from research on infant crying. During the first months post partum, infant daily crying increases until approximately 6 weeks post partum, and then declines to relatively stable levels at 3 months post partum (Barr 1990). In an extensive review on infant crying, Zeifman (2001) associates this development with changes in emotional lability and environmental control of both positive and negative emotions.

Taken together, these considerations suggest that by itself, the quality of maternal care giving may not modulate the infant cortisol response to a mild stressor before 3 months of age, possibly because the infant may not be able to process complex interactional information. However, this does not imply that the young infant's cortisol response cannot be modulated at all by care-giving factors. Findings by Kawakami and colleagues (1996, 1997) suggest that tactile and olfactory stimuli associated with maternal care can reduce cortisol reactivity in response to a heelstick procedure in newborn infants. Kleberg et al. (2008) showed that even in pre-term infants (26 weeks gestational age), higher quality care giving improves cortisol recovery after an acute painful clinical stressor. However, in the Kawakami studies, the stimuli offered are of a basic nature, and recognizable even to newborns (Macfarlane 1975; Fifer and Moon 1995; Mizuno et al. 2004). In the study, Kleberg et al. (2008), the better care provided during the painful clinical procedure included a more comprehensive set of positive stimuli and interventions, including pacifier use and holding of the infant, as well as stopping the procedure (stressor) whenever the infant became too upset. Therefore, although it is clear that this whole cluster of care-giving interventions affected cortisol recovery, it is unclear whether the quality of the interaction between caregiver and infant by itself was significantly associated with it. Our findings merely suggest that, by itself, the quality of maternal care giving may not be associated with the infant cortisol response to an acute stressor, without implying that care-giving factors in general are not associated.

In conclusion, in 5-week-old infants a bathing routine resulted in a significant increase in salivary cortisol, followed by a significant decrease in cortisol concentrations. With the significant increase in infant crying and fussing after being taken out of the bath, this indicates that the bathing routine constitutes a mild physical stressor at this age. At 5 weeks of age, maternal care-giving quality was not associated with the extent of the cortisol increase or decrease. Although replication of our findings is warranted, it appears that the role of maternal care-giving quality in buffering the infant cortisol response may be age-dependent, as is also suggested by infant (social) development. We, therefore, stress the relevance of future research exploring the (development of) social buffering mechanisms of the infant stress response.

Acknowledgements

This study was supported by a grant to CDW from The Netherlands Organization for Scientific Research (NWO, grant 452-04-320).

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.