Developmental Changes in Baseline Cortisol Activity in Early Childhood: Relations with

advertisement
Sarah E. Watamura
Department of Human Development
Cornell University
Ithaca, NY 14853
E-mail: SKL24@cornell.edu
Bonny Donzella
Darlene A. Kertes
Megan R. Gunnar
Institute of Child Development
University of Minnesota
Minneapolis, MN 55455
E-mail: donze001@umn.edu
E-mail: kerte001@umn.edu
E-mail: gunnar@umn.edu
Developmental Changes in
Baseline Cortisol Activity in
Early Childhood: Relations with
Napping and Effortful Control
ABSTRACT: Development of the hypothalamic–pituitary–adrenocortical (HPA)
axis was examined using salivary cortisol levels assessed at wake-up, midmorning,
midafternoon, and bedtime in 77 children aged 12, 18, 24, 30, and 36 months, in a
cross-sectional design. Hierarchical linear modeling (HLM) analyses were used to
characterize cortisol production across the day and to examine age-related
differences. Using area(s) under the curve (AUC), cortisol levels were higher among
the 12-, 18-, and 24-month children than among the 30- and 36-month children. For
all five age groups, cortisol levels were highest at wake-up and lowest at bedtime.
Significant decreases were noted between wake-up and midmorning, and between
midafternoon and bedtime. Unlike adults, midafternoon cortisol levels were not
significantly lower than midmorning levels. Over this age period, children napped
less and scored increasingly higher on parent reports of effortful control. Among the
30- and 36-month children, shorter naps were associated with more adultlike
decreases in cortisol levels from midmorning to midafternoon. Considering all of
the age groups together, effortful control correlated negatively with cortisol levels
after controlling for age. These results suggest that circadian regulation of the HPA
axis continues to mature into the third year in humans, and that its maturation
corresponds to aspects of behavioral development. ! 2004 Wiley Periodicals, Inc.
Dev Psychobiol 45: 125–133, 2004.
Keywords: cortisol; infants; toddlers; children; development; temperament;
effortful control; napping
Cortisol is the primary hormonal product of the
hypothalamic–pituitary–adrenocortical (HPA) axis and
mediates a range of basal metabolic and stress-sensitive
processes in the body. Basal cortisol production follows a
circadian rhythm; however, the development, individual
variation, and context sensitivity of this rhythm are
areas of ongoing exploration. Given the interest in and
prevalence of studies examining salivary cortisol in young
Received 7 February 2004; Accepted 26 May 2004
Correspondence to: S. Watamura
Contract grant sponsor: University of Minnesota
Contract grant sponsor: NIMH
Contract grant number: MH 00946
Published online in Wiley InterScience
(www.interscience.wiley.com). DOI 10.1002/dev.20026
! 2004 Wiley Periodicals, Inc.
children in a range of populations and situations, an
understanding of the basal patterning of cortisol production among healthy, typically developing young children
is needed.
Twenty-four-hour cortisol rhythms are well established in adults, with the majority showing highest levels
approximately 30 min after wake-up, followed by a sharp
decrease over the next hour or two, and then a more
gradual decline over the remaining daytime and evening
hours (Kirschbaum & Hellhammer, 1989). This adultlike
pattern has been observed in children as young as 5 to
6 years of age (Davis, Bruce, & Gunnar, 2002). The
development of the diurnal pattern of cortisol production
has been examined in early infancy. Newborns exhibit
two peaks in daily cortisol production. These peaks are
approximately 12 hr apart and are not entrained to the
earth’s day–night cycle (Rivkees, 2003). By as early as
126
Watamura et al.
6 postnatal weeks, an early morning peak and an evening
nadir in cortisol production can be observed when
measures are averaged across infants (Larson, White,
Cochran, Donzella, & Gunnar, 1998). By 3 postnatal
months, this pattern is more obvious (Price, Close, &
Fielding, 1983), with individual differences becoming
increasingly more stable between the fourth and sixth
postnatal months (Gunnar, Brodersen, Krueger, & Rigatuso,
1996; Lewis & Ramsey, 1995). Despite the clear emergence of a diurnal rhythm over the first 6 postnatal months,
the infant’s daytime pattern of cortisol production is still
immature. Specifically, the average salivary cortisol level
midafternoon to late afternoon is as high as that obtained
midmorning (Larson et al., 1998). Infants retain this
pattern throughout most of the first year of life (Spangler,
1991). The current study assessed cortisol patterning
among older infants and toddlers to examine when the
transition from the infant pattern to a more adultlike
pattern might emerge.
We expected that changes in cortisol production over
the daytime hours might be related to changes in sleep/
wake (i.e., napping) patterns with development. In humans
and other mammals, mutual influences between sleep and
the HPA axis have been observed (Daly & Evans, 1974;
Follenius, Brendenberger, Bandesapt, Libert, & Ehrhart,
1992). In infants, daytime patterns of cortisol production
fluctuate with daytime naps. Specifically, cortisol levels
decrease while the infant naps and then rebound after the
infant wakes up, returning to levels 45 min after the nap
that are equivalent to prenap levels (Larson, Gunnar, &
Hertsgaard, 1991). Longitudinal studies of sleep have
shown that the most prominent decline in napping during
childhood occurs between 18 months and 4 years
(Iglowstein, Jenni, Molinari, & Largo, 2003). If the lack
of an average decrease in cortisol levels between midmorning and midafternoon is related to napping during the
day, we might expect to observe the emergence of lower
midafternoon than midmorning cortisol concentrations as
some children begin to give up daytime naps.
Along with increasing periods of daytime alertness
and reduced periods of napping, the transition from
infancy to early childhood is marked by the development
of self-regulatory competencies that are hypothesized to
reflect the continuing maturation and myelinization of the
prefrontal cortex (Dawson, Panagiotides, Klinger, & Hill,
1992). These competencies include the ability to sustain
attention despite distractions and to inhibit prepotent
responses, responses that have been frequently executed
and are therefore easily recruited. These competencies
are supported by the capacity for effortful control, as
described by Posner and Rothbart (2000). Regions in the
prefrontal and cingulate cortex involved in the effortful
regulation of behavior also may contribute to regulation of
the HPA axis (Sullivan & Gratton, 2002). Thus, we might
expect that young children with a more developed
capacity for effortful control may exhibit more mature
patterns of cortisol production.
METHOD
Participants
Families were recruited by telephone from a list of potential
participants who indicated interest in research participation at
their child’s birth and who had children within 2 weeks of their
12-, 18-, 24-, 30-, or 36-month birthday. Of these, 77 families
returned some saliva samples, and full cortisol data were
available for 69 children: 18, 14, 14, 10, and 13 children per age
group, respectively. An additional 24 families returned questionnaire data. Saliva and/or questionnaire data were available
for 32 twelve-month-olds (16 female), 24 eighteen-month-olds
(18 female), 18 twenty-four-month-olds (8 female), 12 thirtymonth-olds (7 female), and 15 thirty-six-month-olds (8 female).
A total of 86 families were recruited by telephone but did not
return any data within the 6-week study time frame.
Measures
Parents were asked to collect saliva samples at wake-up,
approximately 10 a.m., approximately 4 p.m., and bedtime on
2 days within 2 weeks of their child’s birthday, without
disturbing their child’s normal schedule. Saliva sampling times
reported by parents were wake-up (mean sample time ¼ 7:43
a.m., range ¼ 5:01 a.m.–11:00 a.m.), 10 a.m. (mean sample
time ¼ 10:07, range ¼ 9:24 a.m.–12:27 p.m.), 4 p.m. (mean
sample time ¼ 4:11 p.m., range ¼ 3:18 p.m.–6:40 p.m.), and
bedtime (mean sample time ¼ 8:11 p.m., range ¼ 6:24 p.m.–
10:35 p.m.). We recognized that these guidelines would result in
a range of sample times, but felt it was important to capture
realistic ambulatory levels. All analyses excepting those using
hierarchical linear modeling (HLM) were performed twice, once
with the full range of samples and once restricted to samples
collected within 1 hr of the mean sampling time. Results for the
analyses used in this study did not differ between those conducted
on the full and time-restricted datasets, therefore only results
from the full dataset are presented. HLM analyses could not be
performed on the restricted dataset due to sample size.
Parents were asked to record sleeping and waking on the 2
saliva-sampling days in 15-min intervals using a 24-hr diary.
Parents also completed the 155-item Toddler Behavior Assessment Questionnaire (TBAQ; Goldsmith, 1996; with additional
scales by Rothbart, personal communication, June 8, 1997).
Measures of effortful control were obtained from this instrument using the inhibitory control, attention shifting, attention
focusing, and interest scales (four scales, a ¼ .70). Parents
reported on any health conditions, current medications, and
illness symptoms, and were asked to sample only on days when
their child did not have a fever. Data from 2 children using an
inhaler or nebulizer for asthma were not included. Finally, for
descriptive purposes, parents were asked to report the number of
hours of nonparental care per week, what type of child care was
used, and in cases where more than 10 hr per week were used,
when the child began the care arrangement.
Baseline Cortisol Activity in Early Childhood
Procedures
Parents were contacted by telephone, at which time the purposes
and procedures were described. Those agreeing to participate
were mailed a packet with the parent-report questionnaires and
saliva-collection materials. Materials were sent to arrive 2 weeks
before the child’s birthday. Parents were contacted by telephone
weekly for a maximum of 6 weeks until saliva samples were
returned or participation was discontinued, and the first author
was available by telephone for questions during workday,
evening, and weekend hours. Of the 77 children who returned at
least partial cortisol data, 62 took saliva samples within 2 weeks
of their child’s birthday, 13 within 3 weeks, and 2 within 4 weeks.
Due to difficulties encountered with saliva collection in
this age range, two types of collection procedures were used.
At first, families were asked to sample saliva from their child
using the methods typically used in our laboratory with older
children. Thus, parents gave the child a very small amount of
premeasured, cherry-flavored Kool-Aid (.025 g), asked them to
swallow the crystals, and then mouth a dry cotton dental roll;
however, after attempting to sample 51 children using these
methods, we noted that nearly 60% of families were unable to
provide any of the requested samples. Of those children returning
some samples, an additional 30% were missing some of the
samples requested. Thus, it was apparent that these procedures
were not working well with this age group. A number of parents
reported that their children simply refused to mouth the dry
cotton roll, and eight samples (4.9%) were of insufficient volume
to assay or to assay in duplicate. At this point, we revised our
sampling procedure to one we hoped would increase our success
with toddlers. The remaining 136 families were asked to take
saliva samples using the procedures described next.
To provide color and flavor, we pretreated 6-in. cotton dental
rolls in a Kool-Aid and water solution in a commercial kitchen
following food-preparation guidelines, including wearing
gloves. Cotton rolls were bundled in groups of 50 and secured
with string 2 to 3 in. from the bottom. In a Pyrex 1-cup measure,
we dissolved 34 teaspoon of presweetened cherry flavored KoolAid in 6 oz. of water. No other flavor should be used as other
flavors have been found to compromise the assay. A bundle of 50
was then inserted into the cup and turned one full rotation.
This took less than 10 s and resulted in absorption of the KoolAid solution. Cotton rolls were then laid on disposable aluminum
oven liners so that they were not touching and placed in a 250" F
oven for approximately 1 hr or until dry. When cooled, the cotton
rolls were sealed in zip-lock bags and stored in the refrigerator
for up to 1 month. We also added a set of parent practicesampling materials so that they could experience and model the
sampling for their toddler. Finally, we prefitted the syringes with
½-in. cotton filters made from sterile dental rolls. This step was
added to allow additional filtering of the stimulant from the
saliva as a precaution. Using the new procedure, the overall
success rate was 43%, with 75% of families who returned some
data returning full saliva samples. Saliva sample volumes were
increased, with 13 (3.1%) insufficient to assay or insufficient to
assay in duplicate. This recipe does not compromise cortisol
determination by the CIBA magic assay using cherry-flavored
Kool-Aid and Richmond Dental (Charlotte, NC) cotton rolls
purchased through April 2004; however, as changes in the
127
formula used by manufacturers of sampling materials
and stimulants are possible, we recommend testing the sampling
procedure with the assay you plan to use prior to each study.
We conducted tests of the effects of pretreating the cotton
rolls with saliva from three adults using the CIBA magic assay,
and with pretreated cotton rolls which were sent to Dr. Douglas
A. Granger (College Park, PA) for testing with the electroimmunoassay (EIA) and the radioimmunassay (RIA) by his
Salimetrics laboratory. Mean cortisol levels in mg/dl using the
CIBA magic assay from the three adults using both a dry cotton
roll and a pretreated cotton roll expressed with a syringe fitted
with a filter were .22 (dry cotton roll) and .23 (pretreated and
filtered). There was no significant difference between the
methods, Wilcoxon signed rank test, z ¼ #27, n.s. Using the
EIA and RIA assays and cherry-flavored pretreated cotton rolls,
no significant differences were found; however, other flavors did
contaminate the samples (D. A. Granger, personal communication, November 3, 1998).
Regardless of collection method, parents were instructed to
trim the wetted part of the cotton roll, insert it into the syringe,
and express the saliva into the vial. The samples were refrigerated
until mailed to us in preaddressed, prestamped, bubble-padded
mailers, and samples were frozen until assayed. Samples were
assayed using a modified version of the Corning CIBA magic
cortisol RIA kit (Kirschbaum & Hellhammer, 1989). All samples
from an individual child were assayed in the same batch. Interand intraassay coefficients of variation (CV) using samples of
known cortisol determination were 12.33 and 5.53%, respectively. The mean intraassay CV for samples in this study was
9.8%, and the correlation between duplicate assays was
r(540) ¼ .99, p < .0001. To stabilize values, the samples were
averaged within time point across the 2 days of sampling. In 84%
of the cases, both samples were available and averaged. In the
remaining cases, the single value was used. When data were
analyzed using a Collection-Procedure Group (2) $ Cortisol
Sample Time (4) repeated measures analysis of variance, no
main effect of collection-procedure group was found, F(1,
69) ¼ .51, n.s. Thus, all further analyses were performed collapsed across collection-procedure type.
RESULTS
Preliminary analyses assessed potential child care or sex
effects. Analyses were then conducted on the cortisol data
to examine age and time-of-day effects. Next, the relations
between napping and daily cortisol production were
analyzed, and finally, we explored effortful control–
cortisol relations. All available data were used for each
analysis, resulting in small fluctuations in sample size per
analysis.
Preliminary Analyses
Preliminary repeated measures analyses of the Cortisol
Sample (4) $ Child Care Group (exclusively by parent or
less than 10 hr/week nonparental care, n ¼ 41 vs. 10 hr/
week of child care or more, n ¼ 28) revealed no significant
128
Watamura et al.
effect of child care, F(1, 66) ¼ .64, n.s. Similarly, seven
t tests performed to examine potential sex differences in
cortisol values at any time point, cortisol AUC, effortful
control, or nap duration, revealed no significant differences, ts < 2, ps > .05. Thus, child’s sex or child care
attendance was not included as a factor in any subsequent
analyses.
Cortisol Patterning
The descriptive statistics for cortisol data are shown in
Table 1 and Figure 1.
Age Effects. Hierarchical linear modeling (HLM) was
used to characterize cortisol production over the day and to
examine differences in trajectories based on age. HLM
was considered appropriate because there was considerable variation in individuals’ cortisol production, and
HLM utilizes this individual variation in estimating
parameters. Moreover, this method is particularly useful
for these data because dependency of observations that
occurs when samples are collected only a few hours apart
is accommodated with HLM by specifying an autocorrelation covariance structure. Based on known changes in
the daytime pattern of cortisol production in infants and
adults, the nature of change in cortisol across the day was
expected to have linear and nonlinear components.
Table 1.
12 Months
Mean
Median
18 Months
Mean
Median
24 Months
Mean
Median
30 Months
Mean
Median
36 Months
Mean
Median
An unconditional model was tested including parameters
for intercept as well as linear, quadratic, and cubic change.
Missing data met criteria for being missing at random.
Parameters were estimated using maximum likelihood
estimation, which are valid under missing data. Because
the model was saturated, an autoregressive covariance
structure was utilized to minimize the number of parameters estimated.
The results indicated that the omnibus test for the
unconditional model was significant, F(3, 73) ¼ 144.45,
p < .0001. Significant effects for intercept and change
parameters are indicated in Table 2. The significant linear
change indicated that on average, cortisol declined over
the day. The significant cubic change parameter suggested
that while cortisol levels decreased during the morning
hours, they rose or leveled off midday and then continued
to decline towards the end of the day (see Figures 1 & 2).
A covariate model also was tested with age as a
predictor of intercept, linear, quadratic, and cubic change
parameters. The omnibus test indicated a significant
effect for age, F(4, 72) ¼ 7.10, p < .0001. Tests for
specific effects indicated an Age $ Intercept interaction,
suggesting that individual children’s average cortisol
levels were negatively related to age (see Table 2). An
Age $ Linear Change interaction also was observed,
suggesting that the younger the child, the steeper the drop
in cortisol from wake-up to bedtime.
Summary Statistics on Cortisol Values at Each Time Point and Nap Duration by Age Group
Wake-Up
Cortisol (mg/dl)
10 a.m.
Cortisol (mg/dl)
4 p.m.
Cortisol (mg/dl)
Bedtime
Cortisol (mg/dl)
AUC
Cortisol (mg/dl)
Average Nap
Duration (hr)
0.72 (0.30)
0.68
n ¼ 22
0.39 (0.19)
0.29
n ¼ 23
0.38 (0.29)
0.29
n ¼ 21
0.19 (0.21)
0.11
n ¼ 21
4.79 (2.48)
3.94
n ¼ 18
2.11 (0.65)
2.13
n ¼ 26
0.68 (0.28)
0.62
n ¼ 14
0.29 (0.13)
0.26
n ¼ 15
0.30 (0.22)
0.26
n ¼ 15
0.16 (0.19)
0.06
n ¼ 14
3.13 (0.77)
3.23
n ¼ 14
2.38 (0.53)
2.44
n ¼ 16
0.72 (0.37)
0.64
n ¼ 14
0.29 (0.16)
0.25
n ¼ 15
0.34 (0.33)
0.24
n ¼ 15
0.13 (0.19)
0.10
n ¼ 15
4.23 (3.99)
3.41
n ¼ 14
1.81 (0.72)
2.00
n ¼ 16
0.49 (0.10)
0.51
n ¼ 11
0.18 (0.08)
0.19
n ¼ 11
0.18 (0.08)
0.21
n ¼ 11
0.05 (0.03)
0.05
n ¼ 11
2.29 (0.52)
2.37
n ¼ 10
1.79 (0.93)
2.00
n ¼ 12
0.53 (0.16)
0.47
n ¼ 13
0.23 (0.08)
.22
n ¼ 13
0.20 (0.12)
0.16
n ¼ 13
0.11 (0.20)
0.05
n ¼ 13
2.92 (1.00)
2.54
n ¼ 13
0.86 (0.85)
1.00
n ¼ 13
AUC, area under the curve.
SDs of the mean are given in parentheses.
Baseline Cortisol Activity in Early Childhood
129
With age, the variance of cortisol values also decreased (see Figure 2). Using Levene’s test for equality of
variance, this difference was significant at the wake-up,
F(4, 66) ¼ 2.62, p < .05, and midmorning sampling times,
F(4, 66) ¼ 3.25, p < .05. This may have been due, in part,
to the fact that the mean levels were decreasing with age;
however, coefficients of variation decreased as well,
particularly when comparing the three younger age groups
with the two older age groups. The mean CV for the three
younger age groups was 44.7 at wake-up and 49.6 at
midmorning while for the two older age groups the mean
CV was 25.3 at wake-up and 39.6 at midmorning.
FIGURE 1 Average cortisol patterning across the day for 12-,
18-, 24-, 30-, and 36-month-olds.
To further determine when between 12 and 36 months
significant decreases in cortisol production were observed,
we used the four cortisol samples obtained over the day to
compute area(s) under the cure (AUC) using the trapezoid
rule without baseline correction as described by Pruessner,
Kirschbaum, Meinlschmid, and Hellhammer (2003).
AUC were computed separately for each day and
averaged. A one-way ANOVA was then computed with
AUC as the dependent measure and age groups as the
independent measure. There was a significant main effect,
F(4, 62) ¼ 2.65, p < .05, and post hoc contrasts comparing
the three younger age groups to the two older age groups
revealed that AUC was significantly higher for the three
younger groups as compared with the two older groups,
t(62) ¼ 2.383, p < .05.
Table 2. Table of Fixed Effects for the Unconditional and
Age Covariate Models
Effect
Unconditional model
Intercept
Linear
Quadratic
Cubic
Age covariate model
Intercept
Age $ Intercept
Linear
Age $ Linear
Quadratic
Age $ Quadratic
Cubic
Age Cubic
%
p < .05.
p < .0001.
%%
Estimate
SE
t (df ¼ 75)
#0.6465
#0.1351
#0.0253
#0.0361
0.02
0.01
0.01
0.01
#26.80%%
#18.01%%
#2.01%
#7.75%%
#0.3951
#0.0113
#0.0959
#0.0017
#0.0148
#0.0004
#0.0264
#0.0004
0.06
0.00
0.02
0.00
0.03
0.00
0.01
0.00
#6.85%%
#4.68%%
#4.90%%
#2.13%
#0.43
#0.30
#2.02%
#0.80
Napping Effects. For each child, we calculated the
amount of sleep between 10 a.m. and 4 p.m., and averaged
this across 2 days (average nap duration, see Table 1).
While all of the 12- and 18-month-old children napped
both days, 4 of the 24-month-olds napped only 1 day
(25%). Among the 30- and 36-month children, 6 children
napped neither day (25%), 5 children napped 1 of 2 days
(21%), and 13 children napped both days (54%). In addition, average nap length and age in months were
negatively correlated among the 30- and 36-month-olds,
r(24) ¼ #.54, p < .01, but this was not true among the
younger children, r(37) ¼ .26, n.s. Using partial correlations to control for age, among the older children, nap
length was negatively correlated with the cortisol drop
from 10 a.m. to 4 p.m., pr(21) ¼ #.46, p < .05. That is,
older children who napped less exhibited more of a
decrease in cortisol between midmorning and midafternoon. In contrast, among the younger children, nap length
was not correlated with the drop in cortisol from 10 a.m. to
4 p.m., using either bivariate correlations, r(51) ¼ #.12,
n.s., or using partial correlations controlling age, pr(46) ¼
#.14, n.s. For the 30- and 36-month-olds, we calculated
the mean cortisol at 10 a.m. and 4 p.m. for the groups of
children who napped neither day, napped 1 day, and
napped both days. For the six 30- and 36-month-olds who
napped neither day, mean cortisol at 10 a.m. was .22(.03),
and at 4 p.m. was .15(.01). For the four 30- and 36-montholds who napped 1 day, mean cortisol at 10 a.m. was
.26(.07), and at 4 p.m. was .22(.09). For the remaining
thirteen 30- and 36-month-olds who napped both days,
mean cortisol at 10 a.m. was .19(.02), and at 4 p.m. was
.22(.03). Thus, those 30- and 36-month-olds who napped
both days showed increasing cortisol across the day while
those napping only 1 or neither day showed the more
mature decreasing cortisol pattern.
Effortful Control. The 155-item TBAQ yields four
scales that were expected to reflect effortful control:
inhibitory control, attention shifting, attention focusing,
and interest. These scales were standardized and averaged to compute effortful control. As expected, reported
130
Watamura et al.
FIGURE 3 Partial correlation between Effortful Control and
Cortisol AUC, controlling age.
effortful control increased with age, F(4, 96) ¼ 14.28,
p < .001. Mean effortful control scores and SEs for all
12-, 18-, 24-, 30-, and 36-month-olds were #.56(.10),
#.02(.12), .18(.16), .44(.12), and .63(.15), respectively.
Next, we examined whether effortful control would be
related to cortisol production. Controlling statistically
for age, effortful control was negatively correlated with
cortisol AUC, pr(63) ¼ #.31, p < .05 (see Figure 3).
Examining the data within age group, effortful control and
cortisol AUC controlling age were significantly correlated
for 24-month-olds, pr(11) ¼ #.62, p < .05. At no other
age was the relation significant, although the sample sizes
are notably very small and partial correlations for the
12- and 18-month-olds were pr(15) ¼ #.27, n.s., and
pr(8) ¼ #.37, n.s., compared to those of the 30- and 36month-olds, pr(7) ¼ .02, n.s., and pr(10) ¼ #.10, n.s.,
respectively. Comparing the three younger age groups
with the two older age groups, as above for cortisol
patterning and napping, the relation between effortful
control and cortisol AUC controlling age was significant
for the three younger age groups, pr(40) ¼ #.39, p < .05,
but not for the two older age groups, pr(20) ¼ #.09, n.s.
DISCUSSION
The results of the present study suggest that the HPA
system is continuing to mature during the infant and
toddler periods in human children. Home baseline production of salivary cortisol decreased between 12 and
36 months, with children 30 to 36 months exhibiting
lower levels of cortisol, particularly near wake-up, than
children 12, 18, or 24 months of age. With age, we also
observed a decrease in the variability of cortisol values at
FIGURE 2 Cortisol patterning for individual children across
the day by age group.
Baseline Cortisol Activity in Early Childhood
wake-up and midmorning, even after correcting for the
age-related decreases in mean cortisol levels. Again,
among the 12-, 18-, and 24-month-olds, there was significantly greater variability in cortisol levels at wake-up
and midmorning than among the 30- and 36-month-olds.
This suggests that baseline cortisol production, at least
during the morning hours, may become more stable
during the toddler period.
Across all of these ages, however, there was a clear
daytime rhythm in cortisol production. On average, the
highest baseline values were obtained near wake-up and
the lowest values at bedtime. Across all age groups
studied, cortisol levels decreased significantly between
wake-up and midmorning, and again between midafternoon and bedtime. Nonetheless, even among the oldest
age group (36-month-olds), we did not obtain a significant
group difference in cortisol levels between midmorning
and midafternoon. On average, by the time children are
36 months of age, the daytime production of cortisol is
still not yet fully consistent with the mature pattern.
There was some suggestion that the mature pattern of
a decrease in cortisol levels between midmorning and
midafternoon might emerge as children give up their
daytime naps. All 12- and 18-month-old children in this
study were napping each day, and among the 24-montholds, a few napped only 1 of 2 days (25%); however, by
30 months, nearly half the children napped neither or only
1 day, and the duration of naps had begun to decrease.
Among the 30- and 36-month-olds, those who napped less
were more likely to exhibit a decrease in cortisol between
midmorning and midafternoon. We have previously
shown that among infants, napping is associated with a
significant decline in cortisol levels from pre- to immediately postnap, with a marked rebound in levels by 45 min
after the nap (Larson et al., 1991). Recently, we found that
at child care, cortisol levels for nearly all children were
lower immediately following their afternoon naps than
they were late in the morning and still later in the
afternoon (Watamura, Sebanc, & Gunnar, 2002).
Relatively little is known about the neurobiology of
daytime napping as it relates to regulation of the HPA
system; however, the Larson et al. (1991) and Watamura
et al. (2002) studies suggested that the process of waking
up from a nap stimulates the HPA system. This napawakening stimulation may obscure the diurnal decrease
in cortisol over the midportion of the day. It is thus possible
that as children sleep for shorter and shorter durations with
development, the underlying diurnal decline in activity of
the HPA system can be more readily detected. This view is
consistent with the argument that processes regulating the
HPA system in response to sleep and activity are somewhat
independent of the processes regulating the diurnal
rhythm (Czeisler & Klerman, 1999). Our evidence that
age was negatively correlated with napping time among
131
the older, but not younger, children was consistent with
previous data indicating that reduction in nap duration due
to giving up the afternoon nap occurs postinfancy, during
the later toddler and preschool years (Iglowstein et al.,
2003). This also may explain why we observed a negative
correlation between napping and cortisol levels among
only the older children, as they were in the age period
when children transition into the more adultlike day–
night sleep pattern.
The results of the present study also suggested an
inverse association between cortisol levels and children’s
self-regulatory abilities. Specifically, between 12 and
36 months, parents described children as exhibiting
greater effortful control competencies. That is, with age,
they were better able to sustain interest in tasks, regulate
their attention, and inhibit prepotent responses. These
abilities presumably involve the development of the anterior
cingulate and prefrontal cortices (Posner & Rothbart,
2000). Children who exhibited more mature effortful
control competencies also produced lower concentrations
of cortisol. This was the case even after controlling
statistically for age, suggesting that some of this effect
was due to individual variation within age groups. This
finding suggests several hypotheses, none of which can be
determined based solely on the present results. First,
increased effortful regulatory competence might reduce
the number of conflicts that children are involved in. To
the extent that such encounters activate the HPA axis, the
negative association between effortful control and cortisol
might reflect reduced stressful encounters for children
with greater effortful control abilities. Although the
sample size for the within-age analyses was small—and
thus the results tentative—the fact that the association
between cortisol AUC and effortful control was observed
among the younger, but not older, children cannot be
easily reconciled with the reduction of stressful interactions hypothesis. Effortful control competencies are
hypothesized to rely on the development of anterior
cingulate and prefrontal circuitry (Posner & Rothbart,
2000). Presumably, these circuits begin to exhibit marked
development in the second year of life (Dawson et al.,
1992). Thus, a second hypothesis might be that the
development of these circuits during the second year of
life is sufficient to begin to lower the set point of the HPA
axis while continued development beyond the second year
does not provide additional input to basal HPA functioning. This hypothesis, while admittedly highly speculative,
would place cortical maturation as the cause of altered
basal HPA axis regulation. A third, equally speculative
hypothesis is that decreasing levels of basal cortisol
production play a role, perhaps a permissive one, in the
initial development of the neural system underlying
effortful self-regulation. Of course, a fourth hypothesis
is that the changes in basal cortisol production observed
132
Watamura et al.
in this study and changes in effortful control reflect the
operation of variables that were not measured. Nonetheless, the literature on effortful control suggests that
more effective self-regulation is associated with the development of conscience (Kochanska & Knaack, 2003), and
with lower externalizing problems (Eisenberg et al., 2004;
Kochanska & Knaack, 2003), and the current results
suggest that lower basal cortisol concentrations may be
associated with this profile of adaptive self-regulation
during the toddler and preschool years.
The results of this study also shed light on methodological issues in the collection of salivary cortisol data on
children of this age. In previous work with infants and
with older children, the procedures used in the initial
phase of this study have been effective. That is, we used
dry cotton rolls and asked parents to swab their infant’s
mouth, and we asked children as young as 3 to first chew
and swallow a small amount of Kool-Aid and then chew
on a dry cotton roll. Both methods worked well; however,
we found that parents were having difficulty getting saliva
samples from these toddler-age children using dry cotton
rolls. We therefore pretreated the cotton rolls to add
color and flavor, and provided parent-practice samples.
The return rate for samples and the amount of saliva per
sample increased, and anecdotally, parents reported more
compliance and less distress. The pretreated cotton rolls
did not produce a difference in the results obtained;
therefore, this method might be useful in more easily
collecting saliva samples for cortisol determination with
children of this age.
Limitations
There are several limitations for the present study. First, a
cross-sectional rather than a longitudinal design was used.
It is possible that a different developmental picture would
emerge if children were followed across the toddler
period. Second, the sample was composed of middleclass, Caucasian children with parents willing and able to
complete the home-sampling and diary procedures. Thus,
caution is warranted in generalizing these findings. Third,
the correlational design does not allow a consideration of
the direction of effect for the napping and effortful control
findings. Finally, sleep and effortful control were assessed
using parent report. Using actigraphy for sleep measures
and laboratory assessments of effortful control would
allow more accurate assessments of these variables.
NOTES
The authors thank the parents and children for their participation and feedback, Darcy List for 30-month data collection,
Erika Riggs for assistance with data collection, and Dr. Douglas
Granger for assistance in establishing the reliability of the
pretreated cotton rolls. This research was supported by a
University of Minnesota Undergraduate Research Opportunities
Program award to Sarah E. Watamura, and a National Institute of
Mental Health Research Scientist Award (MH 00946) to Megan
R. Gunnar.
REFERENCES
Czeisler, C. A., & Klerman, E. B. (1999). Circadian and sleepdependent regulation of hormone release in humans. Recent
Progress in Hormone Research, 54, 97–130.
Daly, J. R., & Evans, J. I. (1974). Daily rhythms of steroid and
associated pituitary hormones in man and their relationship
to sleep. In M. H. Briggs & G. A. Christie (Eds.), Advances
in steroid biochemistry and pharmacology (pp. 61–109).
New York: Academic Press.
Davis, E. P., Bruce, J., & Gunnar, M. R. (2002). The anterior
attention network: Associations with temperament and neuroendocrine activity in 6-year-old children. Developmental
Psychobiology, 40, 43–56.
Dawson, G., Panagiotides, H., Klinger, L. G., & Hill, D. (1992).
The role of frontal lobe functioning in the development of
infant self-regulatory behavior. Brain and Cognition, 20,
152–175.
Eisenberg, N., Spinrad, T. L., Fabes, R. A., Reiser, M.,
Cumberland, A., Shepard, S. A., Valiente, C., Losoya, S. H.,
Guthrie, I. K., & Thompson, M. (2004). The relations of
effortful control and impulsivity to children’s resiliency and
adjustment. Child Development, 75, 25–46.
Follenius, M., Brandenberger, G., Bandesapt, J. J., Libert, J. P.,
& Ehrhart, J. (1992). Nocturnal cortisol release in relation to
sleep structure. Sleep, 15, 21–27.
Goldsmith, H. H. (1996). Studying temperament via construction of the Toddler Behavior Assessment Questionnaire.
Child Development, 67, 218–235.
Gunnar, M. R., Brodersen, L., Krueger, K., & Rigatuso, J.
(1996). Dampening of adrenocortical responses during
infancy: Normative changes and individual differences.
Child Development, 67, 877–889.
Iglowstein, I., Jenni, O. G., Molinari, L., & Largo, R. H. (2003).
Sleep duration from infancy to adolescence: Reference
values and generational trends. Pediatrics, 111, 302–307.
Kirschbaum, C., & Hellhammer, D. H. (1989). Salivary cortisol
in psychobiological research: An overview. Neuropsychobiology, 22, 150–169.
Kochanska, G., & Knaack, A. (2003). Effortful control as a
personality characteristic of young children: Antecedents,
correlates, and consequences. Journal of Personality, 71,
1087–1112.
Larson, M., Gunnar, M., & Hertsgaard, L. (1991). The effects
of morning naps, car trips, and maternal separation on
adrenocortical activity in human infants. Child Development,
62, 362–372.
Larson, M. C., White, B. P., Cochran, A., Donzella, B., &
Gunnar, M. (1998). Dampening of the cortisol response to
handling at 3 months in human infants and its relation to
sleep, circadian cortisol activity, and behavioral distress.
Developmental Psychobiology, 33, 327–337.
Baseline Cortisol Activity in Early Childhood
Lewis, M., & Ramsey, D. (1995). Developmental change in
infants’ responses to stress. Child Development, 66, 657–670.
Posner, M., & Rothbart, M. (2000). Developing mechanisms
of self-regulation. Development and Psychopathology, 12,
427–442.
Price, D. A., Close, G. C., & Fielding, B. A. (1983). Age of
appearance of circadian rhythm in salivary cortisol values in
infancy. Archives of Disease in Childhood, 58, 454–456.
Pruessner, J. C., Kirschbaum, C., Meinlschmid, G., & Hellhammer,
D. H. (2003). Two formulas for computation of the area
under the curve represent measures of total hormone
concentration versus time-dependent change. Psychoneuroendocrinology, 28, 916–931.
133
Rivkees, S. A. (2003). Developing circadian rhythmicity in
infants. Pediatrics, 112, 373–381.
Spangler, G. (1991). The emergence of adrenocortical circadian
function in newborns and infants and its relationship to sleep,
feeding, and maternal adrenocortical activity. Early Human
Development, 25, 197–208.
Sullivan, R., & Gratton, A. (2002). Prefrontal cortical
regulation of hypothalamic–pituitary–adrenal function in
the rat and implications for psychopathology: Side matters.
Psychoneuroendocrinology, 27, 99–114.
Watamura, S. E., Sebanc, A. M., & Gunnar, M. R. (2002).
Rising cortisol at child care: Relations with nap, rest, &
temperament. Developmental Psychobiology, 40, 33–42.
Download