Early Family and Child-Care Antecedents of Awakening Cortisol Levels in Adolescence

Abstract

Early caregiving experiences—including those with parents and in child-care—have been shown to play a role in the regulation of the hypothalamic-pituitary-adrenocortical (HPA) axis generally and one of its primary products, the stress-related hormone cortisol, in particular. No published study, however, has yet examined the predictive significance of early observed parenting or child-care experiences in relation to the functioning of the HPA axis over the longer term. Consistent with the attenuation hypothesis, this investigation—using data from the NICHD Study of Early Child Care and Youth Development—provides evidence that individuals who experienced: (a) higher levels of maternal insensitivity and (b) more time in child-care centers in the first three years of life had lower awakening cortisol levels at age 15. Longitudinal associations were small in magnitude. Nonetheless, results were (a) additive in the sense that both higher levels of maternal insensitivity and more experience with center-based care uniquely (but not interactively) predicted lower awakening cortisol levels, (b) not accounted for by later caregiving experiences measured concurrently with awakening cortisol at age 15 or by early demographic variables, and (c) not moderated by sex or by difficult temperament.

There is increasing consensus that the primary biological system subserving the mammalian stress response—the hypothalamic-pituitary-adrenocortical (HPA) axis—is “under strong social regulation in infancy and early childhood” (Tarullo & Gunnar, 2006, p. 632). For example, it has become clear that the absence of supportive experiences with primary caregivers can lead to an alteration of typical diurnal rhythms of cortisol. Specifically, among individuals who experience early social deprivation, the normative pattern of relatively high levels of cortisol just after awakening followed by a decrease in levels throughout the day appears to be altered. In at least two studies, children living in Russian and Romanian orphanages had relatively low early morning cortisol levels, with little or no systematic decreases in levels across the day (Carlson & Earls, 1997; Gunnar, 2000). Importantly, however, evidence for relatively low morning levels of cortisol has not been identified in all studies of institutionally reared children. For example, a study by Gunnar, Morison, Chisholm, and Schuder (2001) yielded evidence that post-institutionalized adoptees showed higher levels of cortisol across the day as compared to children reared in their families of origin or adopted early.

Nonetheless, emerging evidence suggests that lower basal cortisol levels also appear to characterize adults who have a history of maltreatment in childhood (Tarullo & Gunnar, 2006)—at least those without concomitant genetic loading for internalizing problems—leading a number of independent investigators to suggest what has become known as the attenuation hypothesis—that early interpersonal stressors ultimately result in the down-regulation of basal cortisol levels in later life (e.g., Susman, 2006). Key to this hypothesis, also discussed under the rubric of hypocortisolism (Gunnar & Vazquez, 2001), is that chronic activation of the HPA axis may ultimately lead to hyporesponsivity (see also Repetti, Taylor, & Seeman, 2002).

Such research has implications for our understanding of a potential biological mechanism by which relatively early life experiences may affect brain development and subsequent behavior, either conferring protection or risk to children (Talge et al., 2007; Weinstrock, 2005). However, to date, much research focused on the social regulation of the HPA axis generally—and cortisol levels in particular—has not provided prospective, longitudinal evidence as to whether such early effects are relatively enduring in humans. For example, although there is some evidence that early exposure to maternal stressors, such as depression in the early life course, predicts cortisol levels during early childhood (Essex, Klein, Cho, & Kalin, 2002) and adolescence (Halligan, Herbert, Goodyer, & Murray, 2004), no published study has yet examined the predictive significance of early observed parenting or child-care experiences in relation to the functioning of the HPA axis over the longer term. Indeed, in contrast with studies such as those reported by Essex et al (2002) and Halligan et al (2004) focused on maternal depression—a relatively common form of psychopathology often associated with an upregulation of the HPA axis—work on the basal attenuation hypothesis has more typically examined relatively non-normative and strikingly aversive experiences (e.g., maltreatment or the absence of primary caregivers).

A focus on the normative range of social experience in the early life course and its potentially enduring consequences for the basal functioning of the HPA axis is timely in light of evidence from non-human species (e.g., rats) of persistent effects of early rearing experiences on the HPA axis into adulthood that are in part accounted for by environment-driven changes in gene expression (Weaver et al., 2004) as well as evidence among humans that parenting in the normal range in part regulates the HPA axis specifically (Ahnert et al., 2004; Flinn & England, 1995; Gunnar et al., 1996b; Nachmias et al., 1996; Pendry & Adam, 2007; Spangler & Grossman, 1993; Spangler & Schieche, 1998) and neural systems implicated in stress reactivity more generally (Hane & Fox, 2006). For example, although children with insecure attachments and insensitive primary caregivers tend to show larger increases in cortisol when faced with social challenges (either reflecting reactivity or a return to baseline), there is increasing evidence that these larger increases in cortisol may be due in part to lower pretest levels (which therefore permit greater gains), consistent with the attenuation hypothesis (for a review see Gunnar & Vazquez, 2001).

Care provided by parents is not likely to be the exclusive source of potentially enduring early social influences that modulate HPA axis functioning throughout development. For example, the results of two recent meta-analyses (Geoffroy et al., 2006; Vermeer & van IJzendoorn, 2006) demonstrated that young children in primarily center-based child care (compared with those at home) experience increases in cortisol levels during the child-care day. Furthermore, it was demonstrated in one of these meta-analyses (Geoffroy et al., 2006) that the strongest associations were observed among temperamentally difficult children in relatively low quality child-care contexts. Again, however, Gunnar and Vazquez (2001) indicate that, at least in relation to Gunnar’s studies of children in relatively low quality child care, “the rising pattern is produced both by higher midafternoon and lower midmorning levels compared to home values at these times of the day on non-day-care days” (p. 528, italics in original).

Several caveats are nonetheless in order. First, although there is consistent evidence that young children in center-based care tend to show the atypical pattern of increasing cortisol levels across the child-care day—and that these associations while larger within relatively lower quality care contexts are nonetheless apparent even in extremely high quality child-care centers (Watamura, Kryzer, & Robertson, in press)—many studies do not find lower morning levels of cortisol among children in child care (i.e., across studies, it is primarily in the afternoon that children at home versus in child care can be discriminated; Vermeer & van IJzendoorn, 2006). Second, one study of children enrolled in half day preschool (an early peer context where atypical cortisol rhythms has thus far not been documented) actually found that it was well-liked, surgent (i.e., rather than difficult) children who showed relatively high levels of change in cortisol from morning to afternoon while at preschool, at least during the initial weeks of school (Gunnar, Tout, De Haan, Pierce, & Stansbury, 1997). Finally, studies of child care have typically contrasted midmorning with midafternoon cortisol levels (i.e., investigators in these studies tend not to acquire awakening cortisol data).

Drawing on data from the NICHD Study of Early Child Care and Youth Development (SECCYD), a longitudinal investigation of over 1,000 participants tracked prospectively from 1 month through the age of 15, the current report examines early family and child-care antecedents of awakening cortisol levels in human adolescents. Although individual differences in HPA functioning (e.g., as indexed by cortisol levels) have often been conceptualized as a product of genetically-based variations in temperament (Stansbury & Gunnar, 1996) and are indeed heritable (Wüst, Federenko, Hellhammer, & Kirschbaum, 2000), recent evidence that even normative stressors play a role in the development of the HPA axis led us to examine— consistent with the attenuation hypothesis—whether (a) quality of caregiving (insensitivity) and (b) child-care experiences (in particular, high quantity, low quality, and center-care exposure) over the first three years of life are associated with lower awakening levels of cortisol at age 15. The current investigation focused upon center care in particular given recent evidence that it is associated with lower morning cortisol levels among preschoolers (Dettling, Parker, Lane, Sebanc, & Gunnar, 2000).

The current work leveraged the considerable richness of the NICHD SECCYD dataset, which includes observational assessments of early and later caregiving experiences both inside and outside of the home, and supplemented these by acquiring awakening cortisol data from participants over a period of three days at age 15. Morning cortisol levels show intraindividual stability across time (Wüst et al., 2000). In addition, Gunnar and Vazquez (2001) claim that awakening levels are more indicative of long-term influences on the HPA axis whereas evening levels are more influenced by immediate environmental influences. That said, access to awakening cortisol data on the NICHD SECCYD cohort constrains our focus to a limited, though nonetheless important aspect of diurnal cortisol rhythms.

In sum, heeding Gunnar & Vasquez’s (2001) call for researchers to conduct “developmental studies … that help explicate the origins of low cortisol” (p. 515) we examined longer-term associations between normative caregiving experiences in the first three years of life and awakening cortisol levels in adolescence while controlling for (a) early demographic differences (i.e., child ethnicity, child sex, family income-to-needs ratio, and maternal education) that could function as potential selection factors or confounds as well as (b) later caregiving experiences measured concurrently with awakening cortisol at age 15. Finally, we examined whether observed associations differed by sex, were amplified by difficult temperament, or were moderated by either Parenting by Child Care or Child Care by Child Care (e.g., Quantity by Quality) interactions. Prior findings that child-care experiences may be especially stressful for temperamentally difficult children in lower quality care (Geoffroy et al, 2006) make such conditional statistical effects particularly important to examine, as does evidence suggesting that highly negatively emotional infants and toddlers are especially susceptible to rearing influences (Belsky, 2005; Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007).

METHOD

Participants

Participants were enrolled in the NICHD SECCYD, a prospective longitudinal study conducted at 10 research sites across the US. Of the 8,986 mothers who gave birth during selected 24-hour sampling periods, 5,416 (60%) agreed to be telephoned in two weeks and met the eligibility requirements (mother over 18, spoke English, mother healthy, baby not part of a multiple birth or to be released for adoption, family lives within an hour of research site, neighborhood not deemed too dangerous by police to visit). Of that group, a conditionally random sample of 3,015 was selected (56%) for a two-week phone call. The conditioning assured adequate representation (at least 10%) of single mothers, mothers without a high school degree, and ethnic minority mothers (not mutually exclusive). At the 2-week call, families were excluded if the baby had been hospitalized for more than 7 days (2%), they expected to move in the next three years (3%), they could not be reached in at least three attempts at telephone contact (17%), or they refused (21.3%; another 6.1% were lost for other reasons). A total of 1,525 families were selected for the call as eligible and agreed to an interview. Of these, 1,364 completed a home interview when the infant was 1 month old and became study participants. The recruited sample consisted of 52% boys, 24% participants of color, 11% mothers not completing high school, and 14% single-parent families. Additional details about recruitment and selection procedures are available in prior publications from the study (see NICHD ECCRN, 2005) and from the study web site (http://secc.rti.org). Nonetheless, it is important to note that while the NICHD SECCYD is a large, national study, it is not based on a nationally representative sample.

The analysis sample in the present report comprised 863 of the 1,364 participants, due to attrition and missing data (note that n = 863 is the subsample of participants who completed the cortisol component of the age 15 assessment and who also had early maternal sensitivity and/or early child-care data; ns vary as a function of missing data on this sub-sample from earlier assessments). Of the 1,364 participants originally enrolled in the study, 1,009 were available for recruitment at age 15 and were invited to participate. Of these, 873 participants provided cortisol samples but for 5 participants the samples were insufficient for analysis (5 additional participants were missing antecedent child care or maternal sensitivity data). Participants included in the analysis sample were primarily non-Hispanic White (77%); their mothers had a mean of 14.5 (SD = 2.5) years of education; their families had a mean income-to-needs ratio (from 6 to 36 months) of 3.7 (SD = 2.8). Participants who were in the analysis sample, compared with those who were not, were less likely be male (49% vs. 55%), χ² (1, N = 1364) = 4.59, p < .05; and had mothers with more years of education (M = 14.5, SD = 2.5 vs. M = 13.9, SD = 2.6), F (1, 1361) = 18.30, p < .001. The groups did not differ in their race/ethnicity (77% vs. 75% White/non-Hispanic), χ² (1, N = 1364) = 0.39, p = .53, or family income-to-needs (M = 3.69, SD = 2.78 vs. M = 3.47, SD = 3.03), F (1, 1300) = 1.70, p = .19. The participants in the analysis sample, compared with those who were not, also did not differ significantly on early child-care variables (see Measures section)—specifically, quantity (hours) of care (M = 24.11, SD = 16.27 vs. M = 22.39, SD = 16.51), F (1, 1196) = 2.82, p = .09, percentage of center care (M = 0.15, SD = 0.27 vs. M = 0.17, SD = 0.29), F (1, 1215) = 0.53, p = .47, or child-care quality (M = 2.88, SD = 0.44 vs. M = 2.86, SD = 0.46), F (1, 970) = 0.36, p = .55. However, the mothers of the participants in the analysis sample did have significantly higher scores on early maternal sensitivity (M = 3.19, SD = 0.41 vs. M = 3.09, SD = 0.45), F (1, 1304) = 13.51, p < .001.

Procedure

Participants were followed from birth through age 15. Assessments occurred when the participants were 1, 6, 15, 24, 36 and 54 months old; when they were in kindergarten and Grades 1, 2, 3, 4, 5, and 6; and at age 15. The following sections describe the specific measures used in the present analyses and the time points of administration. Additional details about all data collection procedures, psychometric properties of the instruments, and descriptions of how composites were derived and constructed can be found in the study’s Manuals of Operation and Instrument Documentation (http://secc.rti.org). Because the procedure for obtaining cortisol samples has not been described in our prior publications, specific details follow.

Cortisol

At an age 15 home visit, adolescents and parents were given detailed instructions and a demonstration on the saliva collection procedure. Adolescents collected saliva for cortisol assay upon morning awakening for three consecutive school days using a salivette (Sarstedt, Nümbrecht, Germany) provided by the research team, along with written collection instructions. They were instructed not to eat anything prior to saliva collection and to wash their mouth with water immediately on awakening using a cup of water and an empty cup placed by the bedside the previous evening. Adolescents were also told to keep the cotton roll from the salivette in their mouth for three minutes, then place the moist roll into the container, and place it in a freezer until taken to the data collection site.

Both the target adolescents and their parents were trained extensively in the use of the salivettes and the importance of carefully following the protocol provided by the researchers. Electronic monitoring of compliance (Kudielka, Broderick, & Kirschbaum, 2003) was not possible in the current study. Instead, after saliva collection, adolescents completed a “Daily Diary,” in which they recorded the date and time of collection, awakening time, medications taken, and quality of sleep the previous night. These data are shown in Table 1. Medications were not used as covariates in this analysis because we found very little evidence that any class of medication was associated with cortisol levels; of 26 classes of medication used in this cohort, only one statistically significant association between use of one class of medication (i.e., anti-infectives) and cortisol levels emerged in preliminary analyses, and the magnitude of this correlation was trivial. In addition, we compared adolescents who reported that they had smoked more than twice in the last year with those who reported that they had smoked once, twice, or never. The groups did not differ significantly in their cortisol values (M = .36, SD = .18 vs. M = .36, SD = .18), t (853) = 0.12, p = .91. Consequently, we did not include smoking as a covariate. Finally, two of the adolescents were pregnant at the time of assessment; we included their data because their cortisol values were not outliers in the distribution of values (one of these pregnant adolescents was in her third trimester, the other declined to provide information about her last menstruation).

Table 1.

Average awakening cortisol by medication

	Used Medication			Did Not Use Med
Medication Categories	n	M	SD	n	M	SD	Difference
Inhaler/ Nasal Spray with Steroids	16	0.33	0.17	817	0.36	0.18	0.0308
Inhaler/ Nasal Spray without Steroids	12	0.35	0.17	821	0.36	0.18	0.0151
Anti-Infectives	30	0.29	0.14	803	0.36	0.18	0.0713*
Asthma Nonsteroidals	9	0.33	0.31	824	0.36	0.18	0.0291
Gastrointestinals	10	0.32	0.20	823	0.36	0.18	0.0385
Genitourinals	2	0.26	0.01	831	0.36	0.18	0.1001
Cardiac	5	0.28	0.10	828	0.36	0.18	0.0814
Neuro/Seizure, ADHD Nonstimulants	9	0.33	0.17	824	0.36	0.18	0.0317
Hormones/Thyroid	5	0.39	0.32	828	0.36	0.18	−0.0277
Antipsychotics	9	0.36	0.17	824	0.36	0.18	0.0005
Antidepressants	7	0.30	0.18	826	0.36	0.18	0.0568
SSRIs	16	0.38	0.17	817	0.36	0.18	−0.0232
CNS Stimulant Meds	44	0.34	0.17	789	0.36	0.18	0.0234
Other Psychotropics	1	0.26	--	832	0.36	0.18	0.0960
OTC Antiacids	3	0.30	0.09	830	0.36	0.18	0.0568
OTC Antihistamines	24	0.37	0.22	809	0.36	0.18	−0.0062
Cold/Cough	10	0.34	0.17	823	0.36	0.18	0.0238
Acne	6	0.50	0.35	827	0.36	0.18	−0.1364
Analgesics (Pain)	64	0.35	0.19	769	0.36	0.18	0.0104
Nutritional Supplements/Vitamins	15	0.35	0.12	818	0.36	0.18	0.0127
Hypnotics	2	0.50	0.11	831	0.36	0.18	−0.1404
Diabetes	3	0.36	0.20	830	0.36	0.18	0.0002
Allergy Steroidals	12	0.32	0.16	821	0.36	0.18	0.0359
Allergy Nonsteroidals	24	0.34	0.17	809	0.36	0.18	0.0211
Contraceptives	17	0.37	0.23	816	0.36	0.18	−0.0118
Other	14	0.34	0.16	819	0.36	0.18	0.0188

Note.

^*p < 0.05; -- = not applicable; OTC = Over the Counter

In preliminary analyses, we also examined whether average awakening cortisol levels were correlated with a set of five potential covariates that have previously been demonstrated to be correlated with awakening cortisol levels: (a) general sleep problems, (b) average time of awakening, (c) average minutes elapsed from awakening to cortisol acquisition, (d) morning/evening preference, and (for female participants) (e) days since first day of last menstruation (note that a small number of participants recorded in their daily diaries that they did not acquire cortisol immediately upon awakening or that they woke up relatively late on the collection days—although these cases were retained in the current analysis, in supplementary analyses we confirmed that the regression results reported in this manuscript were unchanged after dropping these participants). Only general sleep problems and average time of awakening were significantly associated with awakening cortisol. As such, we used these two variables as covariates in order to examine whether these two variables accounted for any association between either early sensitivity or child care and awakening cortisol levels (see Results). Gender and race/ethnicity (White, non-White) differences in cortisol levels also were evaluated via t-tests. Males had higher cortisol values than females (M = .38, SD = .19 vs. M = .33, SD = .16), t (844) = 4.28, p < .001, but White and non-White participants did not differ significantly (M = .34, SD = .20 vs. M = .36, SD = .18), t (302) = 1.36, p = .18. (Note that dfs for this latter analysis derive from a Satterthwaite t-test because of significant differences in the variances of the white versus non-white groups).

In general, teens returned their cortisol samples when they came to campus to complete a subsequent age 15 laboratory visit; in a small number of cases, research assistants returned to the teen’s home to pick up the samples. Upon arrival from each of the sites, the samples were stored in ultra low freezers (−80 C) and then shipped to Salimetrics (State College, PA) on dry ice for assay. All samples were assayed using a highly sensitive enzyme immunoassay specifically designed for use with saliva (Cat. No.1–0102/1–0112 Salimetrics, PA; www.salimetrics.com). The test has a calibrator range of 0.012 to 3.000 μg/dl and a sensitivity of <0.003 μg/dl. Average intra- and interassay coefficients of variation (CV) were 5.34% and 9.86%, respectively. The standard curve was highly reproducible (mean R²= .998). Samples were assayed in duplicate (.8%, 1.2%, and 1.3% of participants lacked duplicates on days 1, 2, and 3, respectively) and the average was used in analyses.

Measures

Demographic Characteristics

Child gender and maternal education level were obtained by maternal report at 1 month. Family income was reported by mothers at each major data collection point and converted to an income-to-needs ratio by dividing total family income by the US Census-based poverty-level income for that family size. Ratios were averaged from data obtained at 6 through 36 months. Child race/ethnicity was coded as 1 = White/non-Hispanic, 0 = all others.

Early Maternal Sensitivity

Our measure of maternal sensitivity was calculated from ratings of mothers’ behavior during a videotaped interaction between mother and child under semi-structured, free-play conditions at 6, 15, 24, and 36 months (NICHD ECCRN 1998; 2002). A summed composite score of maternal sensitivity was created at each age from coding of the videotapes for sensitivity to non-distress, positive regard, and intrusiveness (reversed) at 6, 15, and 24 months, each on a four-point scale with a possible range of 3–12 for the composite; and supportive presence, respect for autonomy, and hostility (reversed) at 36 months, each on a seven-point scale with a possible range of 3–21 for the composite. These composites were transformed to the same scale and averaged from 6–36 months. The early maternal sensitivity score thus represented child-centered interactions that reflected emotional support for the child that was responsive to the child’s needs and warmly encouraging of the child’s positive engagement with the toys and tasks in the interaction procedures (sensitivity to non-distress, supportive presence), respectful of the child’s interests and efforts (lack of intrusiveness, respect for autonomy), and affectively positive (positive regard, lack of hostility). Internal consistency of the sensitivity composite at each age was moderate, ranging from α = .70 to .78, and interrater reliabilities of the composites, calculated with Winer (1971) intraclass correlations across raters, ranged from .83 to .87. The standardized alpha of the final sensitivity composite = .73.

Later Maternal Sensitivity

At age 15, we video-recorded an 8-minute home discussion of one or two areas of disagreement between the adolescent and mother (e.g., chores, homework, money), selected by the adolescent. Seven-point rating scales of the interaction were used (Owen, Klausli, Aultman, Brown, Little, & Milling, 2006), based on adaptations of the more micro-analytic coding systems of Allen and his colleagues (Allen, Hauser, Bell, McElhaney, Tate, 1996; Allen, Hall, Insabella, Land, Marsh, & Porter, 2003) and coding systems used at earlier ages in the NICHD SECCYD (e.g., Owen, Klausli, & Murrey, 2000). The age-15 maternal sensitivity composite comprised the sum of the mother’s ratings for validation (enthusiastic, positive reactions to and agreement with the teen’s expressed points of view), engagement (expressed interest in the listening to the teen’s thoughts and feelings), inhibiting relatedness (reversed; cutting off and devaluing the teen’s point of view), hostility/devaluing (reversed; expressions of anger, discounting or rejection of the teen or the teen’s ideas), respect for autonomy (encouragement of and respect for the teen’s own ideas and points of view), and valuing/warmth (expressions of positive regard, warmth, and affection). Possible scores could range from 6 to 42. Internal consistency was moderately high, with Cronbach’s alpha of .81. Interrater reliability (intraclass correlation) was .86.

Child Care

Child-care information was obtained through phone calls with the mother and observations of the child’s primary child-care setting. Mothers were called every 3 months between the time the baby was 1 month and 36 months and asked to list the various places the child received care and the hours per week that the child spent in each arrangement. The quantity of care is the average hours per week from 3–36 months that the child spent in regular nonmaternal care; the amount of center care is expressed as the proportion of phone calls from 3–36 months in which the child was reported to be attending a child-care center.

Quality of care was assessed via live observation of primary nonmaternal arrangements that were used for 10 or more hours per week at 6, 15, 24, and 36 months, focusing on caregiver behaviors towards the study child. Observations were conducted during two half-day visits scheduled within a 2-week interval. At each half-day visit, observers completed two 44-minute cycles of the Observational Record of the Caregiving Environment (ORCE) (NICHD ECCRN 1996; 2002). The main measure of child-care quality, the positive caregiving composite, was calculated by averaging aggregated positive caregiving ratings across the four ages. At 6, 15, and 24 months, the aggregated score at each age comprised the average of five four-point ratings: Sensitivity to child’s nondistress signals (the extent to which caregiver-child interaction is characterized by prompt and appropriate responses to the child’s social gestures, expressions and signals, and is generally child-centered), stimulation of cognitive development (the quality and frequency of caregiver effort to facilitate the child’s cognitive development), positive regard for the child (the quality and quantity of expressions to the child that connote the caregiver’s positive feelings toward the child), detachment (reversed; the degree to which the caregiver is emotionally and physically uninvolved with the child and unaware of the child’s needs for appropriate interaction to facilitate involvement with objects or people), and flatness of affect (reversed; the frequency with which the caregiver lacks animation in facial and vocal expression and tone). At 36 months, the aggregated score comprised the average of seven four-point ratings—the same one used at the earlier ages, plus intrusiveness (reversed; the degree to which the caregiver imposes her agenda on the child as opposed to interacting in a way that provides a sense of control to the child), and fostering exploration (the quality of opportunities for, and encouragement of the child’s exploration of objects and the environment). Internal consistency of the aggregated score at each age ranged from α = .82 to .89. Interobserver reliability for the aggregated scores using “gold standard” videotapes ranged from .80 to .94, and with live observations, from .89 to .90. The aggregated scores from 6, 15, 24, and 36 months were averaged to form the positive caregiving composite.

Temperament

When the study infants were 1 month old, mothers completed a 39-item version of the Early Infancy Temperament Questionnaire (Medoff-Cooper, Carey, & McDevitt, 1993); at 6 months they completed the 56-item Revised Infant Temperament Questionnaire (Carey & McDevitt, 1978). At both ages, the dimensions of activity, approach, adaptability, mood, and intensity were assessed and a mean score (with appropriate item reflection) was calculated to denote the level of difficult infant temperament. The 1- and 6-month mean scores were significantly correlated in the full sample, r (1279) = .32, p < .001, and therefore were averaged.

Sleep-Related Covariates

We assessed general sleep problems at age 15 with a 7-item measure developed by Owens, Spirito, and McGuinn (2000). (Higher scores indicated more difficulties getting to and staying asleep). Time of awakening was assessed as minutes after midnight, averaged across the three days of cortisol acquisition.

Cortisol Levels

Cortisol values (in μg/dL) were averaged over the three days of data collection. (As expected, the correlations among the three samples were large by Cohen’s criteria—r’s ranged from .38 to .52 across the three days of sampling; all ps < .001). Distributions of the cortisol values were examined to ascertain the cut-off at 3 SDs above the mean. In order to adjust for outliers, values above this cut-off were assigned the next highest value that was less than or equal to the mean plus 3 SDs. Because the average cortisol data were only moderately skewed (skew = 1.08), we elected to use raw, untransformed data in analyses.

RESULTS

Analytic Plan

Descriptive statistics for all measures used in analyses presented below are shown in Table 2. To examine whether early maternal sensitivity and child-care experiences (i.e., quality, quantity, and center-care exposure) are associated with awakening cortisol levels in adolescence, we conducted a series of regressions analyses. First, we examined the prediction of awakening cortisol values from early sensitivity and early child-care experience considered separately. In each analysis we controlled for sleep-related covariates (sleep-related problems and average time of awakening) and a set of demographic covariates (child ethnicity, child sex, family income-to-needs ratio, and maternal education; see Table 3, Models 1–2). Although not reported in Table 3, significant associations identified in Models 1 and 2 were not moderated by sex. Second, we added concurrent maternal sensitivity to Models 1 and 2 to determine whether the statistical effects observed in either of the first two models could be accounted for by maternal sensitivity observed concurrently with the assessment of awakening cortisol at age 15 (Models 3–4). Third, we added early temperament to Models 1 and 2 to see whether early temperament moderated the predictive significance of early maternal sensitivity and child-care experiences (Models 5–6). Finally, we included early maternal sensitivity and early child-care experiences in the same model to predict awakening cortisol at age 15 (Model 7). Model 7 also included interactions between early sensitivity and early child-care experiences. All associations reported are standardized betas, as well as the overall R² of each model.

Table 2.

Descriptives for predictors and outcome

Variable	n	M	SD	Min	Max
Average Cortisol (μg/dL)	863	0.36	0.18	0.015	1.12
Gender (percent male)	863	49%	--	--	--
Ethnicity (percent white)	863	77%	--	--	--
Income-to-Needs (6 – 36 months)	862	3.69	2.78	0.19	20.31
Mother’s Education (in years)	863	14.45	2.45	7.00	21.00
General Sleep Problems Score	856	23.98	5.56	9.00	45.00
Average Time of Awakening in Minutes	829	414.38	73.78	255.00	756.67
Early Maternal Sensitivity (6 – 36 months)	863	3.19	0.41	1.54	4.00
% in Center Care (3 – 36 months)	843	0.15	0.27	0	1.00
Hours in Non-Maternal Care (3 – 36 months)	838	24.11	16.27	0	58.88
Quality of Child Care (6 – 36 months)	675	2.88	0.44	1.62	3.88
Difficult Temperament (1– 6 months)	863	3.25	0.43	1.86	4.93
Maternal Sensitivity (Age 15)	810	31.19	5.00	9.00	42.00

Note. Average Time of Awakening in Minutes refers to how many minutes the participant reported waking up after midnight, averaged over the three days of cortisol data collection.

Table 3.

Regression models predicting average awakening cortisol levels at age 15

	Model 1: Early Sensitivity		Model 2: Child Care		Later Sensitivity Controlled				Temperament Interactions				Model 7: Across - Domain
Predictors					Model 3: Early Sensitivity		Model 4: Child Care		Model 5: Early Sensitivity		Model 6: Child Care
	(n = 821)		(n = 621)		(n = 771)		(n =588)		(n = 821)		(n = 621)		(n = 621)
Intercept	0.00	****	0.00	****	0.00	****	0.00	****	0.00	****	0.00	****	0.00	****
Gender (1 = male, 0 = female)	−0.11	**	−0.13	**	−0.12	***	−0.13	**	−0.11	**	−0.13	**	−0.12	**
Ethnicity (1 = white, 0 = non-white)	0.03		0.06		0.02		0.05		0.02		0.05		0.02
Income-to-Needs (6–36 mo.)	0.03		0.04		0.04		0.05		0.03		0.04		0.03
Mother’s Education	−0.04		−0.02		−0.04		−0.03		−0.04		−0.02		−0.05
General Sleep Problems Score	0.08	*	0.09	*	0.08	*	0.10	*	0.08	*	0.08	*	0.09	*
Average Time of Awakening in Minutes	−0.09	**	−0.10	*	−0.10	**	−0.10	*	−0.09	**	−0.10	*	−0.10	*
Early Maternal Sensitivity (6–36 mo.)	0.12	**			0.12	**			0.11	*			0.12	*
% in Center Care (3–36 mo.)			−0.09	*			−0.09	*			−0.09	*	−0.11	*
Hours in Non-Maternal Care (3–36 mo.)			0.03				0.04				0.03		0.03
Quality of Child Care (6–36 mo.)			0.03				0.03				0.03		0.01
Later Maternal Sensitivity (Age 15)					−0.04		−0.01
Difficult Temperament (1–6 mo.)									0.00		−0.02
Difficult Temp * Early Sensitivity									0.04
Difficult Temp * % in Center Care											0.01
Difficult Temp * Hours in Child Care											0.02
Difficult Temp * Quality of Child Care											0.05
Early Sensitivity * % in Center Care													0.01
Early Sensitivity * Hours in Child Care													0.00
Early Sensitivity * Quality of Child Care													−0.05
Model R2	0.05	****	0.05	***	0.05	****	0.05	***	0.05	****	0.05	**	0.07	****

Note.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001;

^****p < 0.0001

Regression Models

As can be seen in Table 3, the results of the regression analyses described above are straightforward. Models 1 and 2 reveal that, controlling for a set of early demographic and sleep-related variables, early maternal sensitivity (Model 1) and center-care exposure (Model 2) predicted awakening cortisol levels at age 15 in ways consistent with the attenuation hypothesis—both insensitivity and more center-care experience were significantly associated with lower levels of awakening cortisol. In unreported follow-up regressions of Model 2 we examined whether any significant interactions among child-care variables were evident; none were apparent. Note also that the set of control variables accounted for 4% of the variance in awakening cortisol. Thus, early sensitivity and the child care block (in Models 1 and 2, respectively) each accounted for approximately 1% additional variance in awakening cortisol.

Neither the effect of early sensitivity nor of center-care exposure was accounted for by later observed sensitivity at age 15 (measured concurrently with awakening cortisol; see Models 3 and 4). Finally, Models 5 and 6 reveal no significant moderating role for early difficult temperament, and Model 7 indicates that the statistical effects of child-care and early sensitivity were unique, but not interactive. Simplified regression analyses were conducted (though are not presented) to examine whether interactive statistical effects were evident when examining each of the child-care variables with early maternal sensitivity separately (e.g., Child-Care Quality X Early Sensitivity; Child-Care Quantity X Early Sensitivity; Percent Center Care X Early Sensitivity). No significant interactions emerged in these follow-up analyses. Finally, note that the overall R²’s of Models 1–7 reveal that, in aggregate, variables explained a modest amount of variance in awakening cortisol levels (regressions accounted for between 5% and 7% of the variance).

DISCUSSION

Although there is increasing evidence that the HPA axis is regulated in part by interpersonal experiences in infancy and early childhood (Tarullo & Gunnar, 2006), this study is to our knowledge the first reported in the literature to prospectively examine whether the observed quality of parent-child relationships and child care in the first three years of life in the normative range predict and thus potentially influence human HPA-axis functioning in relatively enduring ways. Specifically, consistent with the attenuation hypothesis, this investigation provides evidence that individuals who experienced higher levels of maternal insensitivity and spent more time in child-care centers in the first three years of life had lower awakening cortisol levels at age 15.

The results provide additional evidence that early experiences with primary caregivers and in child care might have enduring developmental sequelae (Belsky, Vandell, Burchinal, Clarke-Stewart, McCartney, Owen, & the NICHD Early Child Care Research Network, 2007; Roisman, Collins, Sroufe, & Egeland, 2005; Sroufe, Egeland, Carlson, & Collins, 2005). Moreover, we found no evidence that the predictive significance of early caregiving experiences was attributable to continuity in interpersonal experiences with mother from the first three years of life to age 15. That is, even after controlling for the observed quality of maternal sensitivity when participants were 15 years of age (i.e., concurrent with the assessment of awakening cortisol levels), the association between awakening cortisol in adolescence and both early maternal insensitivity and experience with child-care centers in the first three years of life persisted. Similarly, significant associations were additive in the sense that both higher levels of maternal insensitivity and more center-based care uniquely (but not interactively) predicted awakening cortisol levels—and remained significant after controlling for a host of early demographic variables, including broad markers of early risk (child ethnicity, child sex, family income-to-needs ratio, and maternal education).

Although associations remained after adjusting for early demographic variation, later maternal sensitivity, and sleep-related variables associated with awakening cortisol levels, it could well be that—while beyond the scope of this investigation—the predictive significance of early experience with caregivers and center-based care is in fact more proximally mediated through the kinds of environmental stressors (e.g., poorer peer experiences) engendered by suboptimal early experiences. Furthermore, even if the predictive relations identified in this report are in fact completely neurobiologically mediated effects of early experience per se, they need not be viewed as ineluctable in light of increasing evidence for the efficacy of interventions in ameliorating the effects of early experience on HPA axis functioning (i.e., cortisol reactivity and diurnal cortisol cycles; Brotman et al., 2007; Fisher, Stoolmiller, Gunnar, & Burraston, 2007).

Having established that the predictive significance of maternal insensitivity and center-based care was robust, even if small in magnitude in this large though non-representative sample, we also evaluated theoretically expected moderators of these associations. We found that neither sex nor difficult temperament conditioned our findings, with caveat that the NICHD ECCRN did not assess difficulty between 6 and 54 months, a period when temperament may become somewhat more stable. Difficult temperament was examined in particular given prior meta-analytic evidence that child-care experience may be especially stressful for such already “biologically compromised” children (Geoffroy et al., 2006), as well as evidence that difficult temperament functions in the NICHD SECCYD dataset as a reliable index of differential susceptibility to rearing influences (Belsky, 1997), at least in some domains (e.g., Belsky, 2005; Belsky et al., 2007; Bradley & Corwyn, 2007). Importantly, we also found no evidence for Maternal Sensitivity by Child Care or Child Care by Child Care interactions, suggesting, for example, that center-care experience was not associated with lower awakening cortisol levels only for those participants with insensitive mothers or only for participants in low quality child care.

Although it is never advisable to be completely confident in interpreting null interactions substantively, the NICHD SECCYD dataset is large, thereby maximizing statistical power, and, where possible, we examined multiple operationalizations of key moderators. In unreported analyses, for example, we found no evidence either that the more reliable mother-reported difficult temperament assessment at 6 months on its own, or an index of observed susceptibility to distress (assessed in the context of a maternal separation and reunion procedure at 15 months; Ainsworth, Blehar, Waters, & Wall, 1978), conditioned any of the links between morning cortisol and either early child-care or maternal sensitivity reported here. It seems noteworthy, too, that the additive predictive significance of maternal insensitivity and center-care experience mirror just what was found when the NICHD SECCYD predicted externalizing problems through 6^th grade from parenting and early child-care experience (Belsky et al., 2007). Furthermore, at least one prior study has already implicated center-based care experiences in particular in lower morning cortisol levels among preschoolers (Dettling, Parker, Lane, Sebanc, & Gunnar, 2000). Nonetheless, additional research is necessary to explain the precise mechanisms through which center-based care per se may be associated with awakening cortisol levels—and why, for example, lower quality care in the early life course was not similarly associated with later awakening cortisol levels (keeping in mind, of course, that other aspects of child care quality not examined here [e.g., peer experiences, rules, routines, chaos] could well be associated with the functioning of the HPA axis).

Although the results reported here have important theoretical implications vis-à-vis the relatively enduring significance of early interpersonal experiences, it is not clear whether there are policy-related implications of these findings. On the one hand, it is important to stress the small magnitude of the associations identified in this report. These findings suggest that, although the predictive significance of early insensitivity and center-care experience may be apparent through mid adolescence, normative experiences may not strongly influence awakening cortisol levels—or that intervening experiences between infancy and adolescence not examined here moderate the longer-term impact of early experience. Another possible explanation of the modest associations is that our measures of child-care experience and early sensitivity, though state-of-the-art, are imperfect. To our knowledge, the NICHD SECCYD is the first investigation to aggregate four separate observations of maternal sensitivity in the first three years of life in order to examine the predictive significance of the quality of early parenting. However, the average of four ratings of maternal sensitivity based on observations of approximately 15 minutes each is, of course, unlikely to perfectly reflect the cumulative nature of a child’s early experiences with his or her primary caregiver. As such, the associations identified in this paper could well underestimate the importance of early experiences in the functioning of the HPA axis.

Second, as described earlier, awakening cortisol was acquired without the use of a device that allows for the electronic monitoring of compliance. Thus, if participants who experienced more early insensitivity and more center-based care experiences are also more likely to lead the kinds of chaotic lives that made it difficult for them to have accurately followed cortisol data collection procedures at age 15—and to misrepresent the actual time that cortisol was collected—it could be that their relatively lower levels merely reflected their saliva being acquired later in the morning (i.e., when cortisol levels are normatively lower compared to awakening). Although possible, we view this as unlikely in that all statistical effects of early experiences held controlling for a host of variables presumably correlated with family chaos, including concurrent assessments of maternal sensitivity and early demographic measures (e.g., family income-to-needs).

Finally, the NICHD SECCYD did not acquire measures of awakening cortisol in the first three years of life. Thus, we could not control in this analysis for any initial covariation between center-based care or sensitivity and awakening cortisol levels in the first three years of life prior to examining the predictive significance of early experiential inputs. As such, we can not be certain the predictive associations documented here necessarily reflect downward deflections in awakening cortisol levels. In addition, although this study relied on awakening cortisol levels gathered on three consecutive days in a large cohort, for practical reasons we were unable to measure cortisol levels multiple times per day across multiple days, a best practice that not only maximizes the reliability of the assessment of cortisol (Hellhammer, Fries, Schweisthal, Schlotz, Stone, & Hagemann, 2007) but also, of course, provides the only means of examining individual trajectories associated with diurnal rhythms of cortisol. This is a non-trivial limitation of our study in that we were unable to determine whether individuals who experienced relatively high levels of maternal insensitivity or center-based child-care experience showed flattened diurnal cycles, as has been demonstrated in some studies among children who experience more profound caregiving stressors in the early life course. Because we can only conclude that individuals who experience more insensitivity and center-based care had awakening cortisol levels that were relatively lower than their counterparts in adolescence, it would be speculative at best to suggest that these data indicate more general basal down-regulation of the HPA axis.

That said, there is reliable evidence that individuals who showed persistently elevated antisocial have lower cortisol levels during laboratory visits than do individuals who never show clinically significant antisocial behavior (see, e.g., McBurnett et al., 1991; McBurnett et al., 2000). Ongoing analyses of the current cohort are examining whether lower levels of awakening cortisol similarly functions as a hormonal marker of risk in the NICHD SECCYD dataset.