Pubertal transition with current life stress and support alters longitudinal diurnal cortisol patterns in adolescents exposed to early life adversity

Abstract

Current and early life stress (ELS) are associated with diurnal cortisol patterns, which themselves are associated with mental and physical health. The pubertal recalibration hypothesis suggests that the social environment can impact dysregulated cortisol patterns for previously ELS-exposed youth as they transition through puberty. This study examined longitudinal change in cortisol awakening response (CAR) and diurnal slope (DS) across puberty as a function of ELS in infancy, current stress, and social support (N = 290, 7–17 years). The CAR and DS were examined thrice annually with an accelerated longitudinal design with nurse-assessed puberty to assess associations between diurnal cortisol and pubertal recalibration with ELS and the current social environment. Exposure to ELS was associated with less steep DS but not changes in CAR, and no evidence of pubertal calibration was found. The DS became less steep for youth in later pubertal stages and as youth progressed through puberty. The CAR was steeper for youth in later pubertal stages. Across the cohort, current life stress and support were associated with changes in the DS and the CAR through the pubertal transition. The pubertal stage and the peripubertal and pubertal social environment may have important implications for adrenocortical functioning with or without exposure to ELS.

1 |. INTRODUCTION

Exposure to psychosocial adversity early in life is associated with later negative health outcomes, including mental health issues such as externalizing behavior and depression, and physical health issues, such as cardiovascular disease and inflammation (Chrousos & Gold, 1992). The hypothalamic-pituitary-adrenal (HPA) axis, a key regulator of the body’s stress response and circadian rhythm (Knutsson et al., 1997; Selye, 1976), is thought to be a mediator of the relationship between early adversity and later health outcomes. The HPA axis is responsive to psychosocial stress (Adam, 2012; Dickerson & Kemeny, 2004; Gunnar, Frenn, et al., 2009; Gunnar, Talge, & Herrera, 2009), and its primary product, cortisol in humans, impacts the body and brain through arousal processes, metabolic processes, immune system functioning, mood, and sexual behavior (Lupien et al., 2009; Sapolsky et al., 2000). Early psychosocial experiences have a large influence on the development and function of the HPA axis in childhood and through the lifespan (Gunnar & Donzella, 2002; Lupien et al., 2009). Infancy is thought to be a sensitive period during which the HPA axis is programmed, and the alterations in cortisol regulation that arise from adversity in infancy have been proposed as a mechanism by which early stress exposure influence later outcomes (Davis et al., 2011; Heim et al., 2008).

The diurnal variation in cortisol may play an important role in the regulation of these systems (Adam & Kumari, 2009; Bass & Lazar, 2016; Man et al., 2016; Sephton & Spiegel, 2003) and a recent meta-analysis highlights the importance of diurnal cortisol variation on health outcomes such as fatigue, immune outcomes, obesity, and cancer (Adam et al., 2017). Cortisol levels exhibit a diurnal rhythm, with high levels in the morning upon waking, a surge approximately 30–45 min after waking, decreasing levels across the day, and a nadir around bedtime (Adam & Kumari, 2009; Pruessner et al., 1997). This pattern can be characterized by the cortisol awakening response (CAR), which assesses the slope of cortisol levels from wake to the peak of the morning surge, and the diurnal slope (DS), which measures the change in cortisol levels from wake to bed (Adam & Kumari, 2009). Research on the DS and the CAR find that they are both sensitive to psychosocial stress (Adam, 2006; Adam & Gunnar, 2001; Adam et al., 2006; Doane & Adam, 2010; Wilhelm et al., 2007). As cortisol exhibits an inverted “U” shape in relation to cognitive and behavioral functioning, both flatter and steeper slopes of the CAR and the DS have been associated with psychosocial adversity and poor health outcomes (Adam et al., 2010, 2014; Dietrich et al., 2013; Sephton et al., 2000; Wessa et al., 2006).

A number of studies have examined how stress, both early in life and current, relates to the CAR. Studies of early life stress (ELS) have examined two main contexts: youth who experienced neglect and deprivation early in life due to institutional (e.g., orphanage) care prior to adoption into supportive homes and youth who experienced neglect and abuse in their families of origin. Several of these studies have reported a blunted (smaller) CAR and a flatter DS associated with ELS (Flannery et al., 2017; Gunnar, 2001; King et al., 2017; Koss et al., 2014; Kumsta et al., 2017; Leneman et al., 2018; Quevedo et al., 2012; Shirtcliff et al., 2020; van der Vegt et al., 2010).

However, the impacts of ELS on diurnal cortisol rhythms may not be permanent. Puberty, due to neural and endocrine changes that may influence the HPA axis (Dorn & Biro, 2011; Marceau et al., 2015), is thought to be another sensitive period in HPA axis development. The pubertal recalibration hypothesis (Gunnar & Reid, 2019; Gunnar et al., 2019; Romeo, 2010) posits that, in addition to infancy, puberty is another sensitive period during which experiences have a larger impact on the regulation of the HPA axis than in childhood or later in development. This hypothesis was developed in relation to stress response (e.g., Romeo, 2010) but has also been applied to the CAR and DS diurnal cortisol rhythms. Thus, it is hypothesized that ELS-exposed individuals move from a dysregulated diurnal cortisol pattern to a normative diurnal cortisol pattern as they move through puberty if the current conditions are supportive and nonthreatening (Figure 1). Note, however, that there remains a risk that the individual will recalibrate and be at risk of moving to a dysregulated diurnal cortisol pattern if the conditions during the peripubertal period are harsh or threatening. Recent research replicating animal evidence (Romeo, 2010) has demonstrated that dysregulation of the stress response arising from ELS appears to be plastic to recalibrating as youth transition through puberty (Gunnar et al., 2019). A longitudinal study in children without an ELS history found evidence of attenuations in DS as youth transition through self-reported pubertal stages (Shirtcliff et al., 2012). A recent, cross-sectional study of a community sample of youth with varying levels of ELA exposure found no significant interaction between self-reported pubertal stage and ELA severity on cortisol DS but did find that higher ELS was associated with a blunted CAR for youth in early pubertal stages, and higher ELS was associated with a higher CAR for youth in later stages of puberty (King et al., 2017). Additionally, the social environment during puberty may influence diurnal cortisol rhythms: higher levels of social support predicted lower CAR in a cohort of young adolescents (Rickard et al., 2015), and a longitudinal study of adolescents found evidence of increasing CAR only for those with high depressive and anxiety symptoms (Nelemans et al., 2014). Cross-sectional analyses of youth exposed to ELS have suggested that the diurnal cortisol patterns might also recalibrate through puberty, if youth experience a supportive caregiving environment through the pubertal transition in some studies (Flannery et al., 2017; Quevedo et al., 2012) but not others (Leneman et al., 2018).

FIGURE 1.

Hypothesized pubertal recalibration of diurnal cortisol patterns. Reproduced with permission from the authors from Gunnar and Reid (2019)

Thus, biological changes associated with puberty and social experiences during puberty may influence the development of the CAR during this developmental period. Unfortunately, there is little evidence nor consensus about normative changes in CAR or DS during the peripubertal period, hindering our understanding of how high-stress or low-support environments impact diurnal cortisol over time. Of note, many of the aforementioned studies relied on self-report of pubertal stage, are cross-sectional, or the measurement of ELS exposure co-occurs with the measurement of puberty and diurnal cortisol patterns. No study has yet examined the relationships between ELS, puberty, and diurnal rhythms in a longitudinal sample with nurse-assessed pubertal stage. It is therefore difficult to draw conclusions on either the programming effects of ELS on longitudinal diurnal cortisol patterns or a test of pubertal recalibration from the current literature.

The current longitudinal study sought to address these gaps in the literature. In addition to characterizing the normative changes in CAR and DS over a much wider age range than that of previous studies (from 7 to 17 years); this study examines youth who were and were not previously institutionalized (PI) as infants to examine longitudinal change in CAR and DS across puberty as a function of ELS, current life stress, and current social support. Importantly, this framework begins to tease apart the role of age- and puberty-related change, as within-individual increases in pubertal stage do not differ if, for example, an individual moves from pubertal stage 1 to 2 or stage 4 to 5. Furthermore, this framework directly replicates the pubertal-stage-focused analyses in a recent paper examining the pubertal recalibration hypothesis (Gunnar et al., 2019). Early institutional care provides a useful model for addressing questions related to the association between the quality of the early versus current environment and diurnal cortisol rhythms. PI children have a starkly different quality of environment in infancy, compared to adolescence. Thus, we can begin to tease apart the role of early adversity (represented by psychosocial neglect and unstable caregiving that abruptly ceases at adoption) and more recent experiences (e.g., social stressors and supports) on the development of the diurnal cortisol patterns during the peripubertal period.

We first examined whether exposure to institutional care in early life, compared to peers without deprivation exposure in infancy, would influence the magnitude or direction of change in the CAR and DS across the pubertal transition. We hypothesized that across all youth, CAR and the DS will become steeper as youth transition through puberty. We hypothesized that PI youth will exhibit a blunted CAR and less steep DS, relative to their non-adopted (NA) peers, especially in early pubertal stages. We also hypothesized that PI youth will exhibit a recalibration of diurnal rhythms such that the CAR and DS will become steeper and more similar to comparison youth as PI youth transition through puberty (Figure 1). We then examined whether self-reported current life stress and current social support (analyzed separately) would predict change in CAR and DS across pubertal stages. We predicted that all youth reporting higher levels of current stress and lower levels of current support would show larger increases in CAR (steeper slope) and a more blunted (flatter) DS and thus higher bedtime cortisol.

2 |. MATERIALS AND METHODS

2.1 |. Participants and study design

The current sample included 290 youth who were either internationally adopted from institutions (PI; n = 126, 85 females, 41 males) or born and raised in their families of origin in a Midwest region of the United States (NA; n = 164, 86 females, 78 males). These were the same participants whose reactivity to the Trier Social Stress Test was reported on previously (Gunnar et al., 2019). An accelerated longitudinal (cross-sequential) design was used, such that youth were recruited in 2-year age blocks starting at 7 to 15 years at the start of the study and participated in a total of three visits, each a year apart.

Children in the NA group were recruited from a registry of families potentially interested in participating in research, most of whom were recruited through a letter received soon after the child’s birth. PI children were recruited from a registry of families that responded to a survey sent out after adopting internationally through Minnesota agencies and ongoing recruitment via major local international adoption agencies. PI youth were selected to have been adopted prior to age 5 years (M = 19.43 month, SD = 12.64) and spent at least 50% of their pre-adoptive life in an institution (M = 95.19%, SD = 9.54). The adoptive families of PI youth and NA families were similar in socioeconomic status. Exclusion criteria for both PI and NA children were endocrine disorders and/or use of systemic glucocorticoid medication. Children intermittently using glucocorticoid-based nasal sprays were retained, but this was incorporated into a score of cortisol-impacting drugs and included as a covariate (Granger et al., 2009). PI children were further assessed and excluded for evidence of fetal alcohol syndrome. Exclusion criteria for NA children included autism spectrum disorder diagnosis. Only one child per household was included.

We assessed participants’ DS, CAR, and pubertal stage at three consecutive and annual data collection sessions (Figure 2). The present analysis uses data derived from saliva samples collected to assess the CAR and DS at all three sessions. At each session, the youth also completed a nurse exam to determine the pubertal stage and semi-structured interview assessing current life stressors and supports. Cross-sectional CAR data from session 1 was previously reported in Leneman et al. (2018). Participants who had data in any session were included in the analyses. Attrition did not differ by group (PI vs. NA), sex, race, family income, parent education, or youth-reported stress and support, though participants who did not complete session 3 were more likely to be older than those who did (t(272) = 2.03, p = .04). Consent was received from a caregiver for each participant and written and verbal assents were received from all participants. All study procedures were approved by the University of Minnesota Institutional Review Board.

FIGURE 2.

Salivary cortisol sampling collection procedures

2.1.1 |. Pubertal staging

Trained nurses conducted pubertal staging, scored according to Marshall and Tanner criteria (Marshall & Tanner, 1969, 1970). Breast and testicle Tanner scores, for girls and boys, respectively, were used as a measure of central puberty ranging from 1, indicating pubertal development not yet begun, to 5, indicating pubertal development is complete. Details on pubertal assessment procedures, quality control analyses, and data processing are available in Gunnar et al. (2019) and Reid et al. (2017).

2.1.2 |. Salivary cortisol

At every annual session, participants were asked to provide saliva samples for 3 days. Participants were strongly encouraged to sample on weekdays, not necessarily consecutively, but on days with regular activity to increase the consistency of the wake-sleep cycle and sampling times. Samples were not to be taken on days in which the participant was sick with a fever. The instructions were to take samples immediately upon waking (wake), 30 min after waking (30 min), and at bedtime (bed) or 9:00 p.m., whichever came first. Directions included refraining from eating, drinking, using tobacco, or brushing teeth in between the two morning samples, especially relating to substances containing dairy or caffeine. Participants recorded wake time, sample time, sleep hours and quality, day of the week, month of the year, daily medications, sick or not, and daily events.

Participants collected saliva by mouthing an oral sponge (Sarstedt Salivettes for Cortisol Testing) until it was wet. This was then placed in a provided vial, which was refrigerated until returned by mail. Sponges were provided in a MEMS® 5 TrackCaps that recorded the time of opening to help check compliance. Samples returned to the lab were stored at −20°C until being shipped to Trier, Germany, for assay. Samples were assayed in duplicate using a time-resolved fluorescence immunoassay (DELFIA). The intra- and inter-assay coefficients of variation ranged from 4.0% to 7.9%. An index of medication usage that may affect adrenal cortisol production was created according to guidelines in Granger et al. (2009). Biologically implausible values were removed (> 2 μg/dl, n = 9 samples) as has been done in other studies (e.g., Bernard et al., 2015). Values were then log₁₀-transformed to resolve positive skew. All bedtime samples were used to examine the DS. To precisely examine the CAR, sample times were examined for compliance to the study protocol, and noncompliant samples were not included in the present analyses. Noncompliant samples (n = 380 samples, 6.8%) were defined as “wake” samples taken 10+ min after the reported wake up time; “30 min” samples taken less than 15 min after the reported wake time, less than 15 min after the “wake” sample, or 40+ min after reported wake time. For the CAR, the final number of samples was 1390 for session 1, 1180 for session 2, and 1202 for session 3. For DS, the final number of samples was 1478 for session 1, 1318 for session 2, and 1299 for session 3.

2.1.3 |. Youth-reported current stress and support

At the first session, we developed an expanded coding scheme with two scales, stress and support, to qualitatively re-code and a modified version of the Youth Life Stress Interview (Hammen & Rudolph, 1999), which is a 1-h semi-structured interview conducted by a trained experimenter inquiring about life events in six different relational domains over the past 12 months: same-sex peer, opposite-sex peer, romantic, sibling, parent-child, and parent-marital. While the interview was originally designed to score stressful life events, the expanded coding scheme takes into account youth’s self-reports of the amount of social support they received surrounding these stressful life events. Generally speaking, stress scores captured the presence or absence of negative social experiences, and support scores captured the presence or absence of positive social experiences. In order to tease apart these potentially unique dimensions of the current environment, we developed an expanded coding scheme with two scales, stress and support, which were scored by two trained raters for each of the six domains listed above (see Supplementary Material for new rating scales). Each scale ranges from 1 to 5, including half-points. For the stress scale, 1 = Superior/No Stress and 5 = Severe Stress, whereas for the support scale, 1 = Little to No Support and 5 = Superior/Above Average Support. Stress scores qualitatively represented frequency, duration, severity, content, and resolution of conflicts as well as experiences like teasing and bullying. Support scores encompassed (also qualitatively) the number and quality of friendships, level of trust in various relationships, and frequency of social and extracurricular activities outside of school. Generally speaking, stress scores captured the presence or absence of negative social experiences, and support scores captured the presence or absence of positive social experiences. See Supplementary Material for anchor descriptions of each point on the stress and support scales.

Raters were trained within one point of consensus codes from master coders on a training set of six interviews. Subsequently, all disagreements larger than one half-point for any domain were conferenced to consensus by the two raters (stress: intraclass correlation coefficient (ICC) (two-way mixed, average measures) = 0.95; support: ICC (two-way mixed, average measures) = 0.76). Youth-reported current stress and support were moderately inversely correlated (r = −.55), supporting the idea that they are measuring related but distinct constructs. See Supplementary Material for additional information and anchor descriptions of each point on the stress and support scales.

2.2 |. Data analysis

A hierarchical linear mixed-effects (HLM) model (Pinheiro & Bates, 2006) was used to examine change in the CAR and the DS over time using the “lmerTest” package in R (Kuznetsova et al., 2017; R Core Team, 2021). Time since wake (in minutes for CAR and hours for DS) and the linear effect of time were used as predictors to model the CAR and DS with random intercepts and random effects of time nested within-session and within-participant to account for the hierarchical nature of the data and allow for variability in cortisol across sessions and participants. Current stress and support were grand-mean-centered in all models. To account for the accelerated longitudinal design and disentangle between- and within-person effects of the pubertal stage as participants entered the study at a wide variety of ages, two puberty variables were calculated to capture both an individual’s change in puberty over time and capture individual’s mean pubertal score, compared with other individuals in the study. Within-individual change in puberty, or the longitudinal change in pubertal score for each individual, was modeled using pubertal stage centered around each participant’s mean across all sessions and named “pubertal change” for clarity. This is distinguished from between-individual change in puberty, which was modeled as each participant’s mean pubertal stage across all sessions and named “mean pubertal stage” for clarity.

The models were built sequentially. We examined whether the longitudinal change in CAR and DS differed as a function of (1) early institutionalization and mean pubertal stage and pubertal change (independently), (2) pubertal recalibration (the interaction of group, and mean pubertal stage, and pubertal change), and (3) and the interaction of mean pubertal stage and pubertal change with current stress and support via separate cross-level interaction models (represented by Equation 1). All significant interactions were examined graphically. Wake time (minutes since midnight for CAR or hours since midnight for DS) was included as a time-varying covariate in all analyses. All available sampling days for each participant were included without averaging across days to take advantage of all variability in repeated measures (Mikolajczak et al., 2010). Sampling day (1–3), weekend versus weekday sampling day, sex, and the daily medication index (Granger et al., 2009) were examined as potential covariates.

$$\begin{array}{l}\text{Level }1(\text{sample}):{\text{log}}_{10}{\text{cortisol}}_{ijk}={\pi }_{0jk}+{\pi }_{1jk}{\text{sample time}}_{ijk}+{\pi }_{2jk}\text{wake\hspace{0.17em}}{\text{time}}_{ijk}+{\pi }_{3jk}{\text{day}}_{ijk}+{\text{e}}_{ijk}\\ \text{Level }2(\text{session}):{\pi }_{0jk}={\beta }_{00k}+{\beta }_{01k}\text{pubertal\hspace{0.17em}}{\text{change}}_{jk}+r_{0jk}\\ {\pi }_{1jk}={\beta }_{10k}+{\beta }_{11k}\text{pubertal\hspace{0.17em}}{\text{change}}_{jk}+r_{1jk}\\ {\pi }_{2jk}={\beta }_{20k}+r_{2jk}\\ {\pi }_{3jk}={\beta }_{30k}+r_{3jk}\\ \text{Level\hspace{0.17em}}3(\text{individual}):{\beta }_{00k}={\gamma }_{000}+{\gamma }_{001}\text{mean\hspace{0.17em}pubertal\hspace{0.17em}}{\text{stage}}_k+{\gamma }_{002}{\text{group}}_k+{\gamma }_{003}\text{group}*\text{mean\hspace{0.17em}pubertal\hspace{0.17em}}{\text{stage}}_k+u_{00k}\\ {\beta }_{01k}={\gamma }_{010}+{\gamma }_{011}{\text{group}}_k+u_{01k}\\ {\beta }_{10k}={\gamma }_{100}+{\gamma }_{101}{\text{group}}_k+u_{10k}\\ {\beta }_{11k}={\gamma }_{110}+{\gamma }_{111}{\text{group}}_k+u_{11k}\\ {\beta }_{20k}={\gamma }_{200}+u_{20k}\\ {\beta }_{30k}={\gamma }_{300}+u_{30k}.\end{array}$$ (1)

Note. This is an example equation using group and pubertal stage as a level 3 (individual-level) predictor. Two additional models were run replacing the group with session 1 stress and session 1 support, respectively. Variables denoted “e,” “r,” or “u” are error terms.

3 |. RESULTS

Demographics of the study are displayed in Table 1 and correlations between key variables are shown in Table 2.

TABLE 1.

Demographics

	Previously institutionalized (N = 126)	Non-adopted (N = 164)	Overall (N = 290)
Sex, n (%)
Female	85 (67.5%)	86 (52.4%)	171 (59.0%)
Male	41 (32.5%)	78 (47.6%)	119 (41.0%)
Age at session 1 (years), Mean (SD) [min, max]	11.3 (2.42) [7.08, 15.1]	11.1 (2.27) [7.27, 15.0]	11.2 (2.33) [7.08, 15.1]
Annual family income, n (%)
< $25,000	0 (0%)	1 (0.725%)	1 (0.407%)
$25,001–40,000	0 (0%)	5 (3.62%)	5 (2.03%)
$40,001–55,000	9 (8.33%)	6 (4.35%)	15 (6.10%)
$55,001–70,000	12 (11.1%)	12 (8.70%)	24 (9.76%)
$70,001–85,000	13 (12.0%)	12 (8.70%)	25 (10.2%)
$85,001–100,000	14 (13.0%)	30 (21.7%)	44 (17.9%)
$100,001–150,000	40 (37.0%)	41 (29.7%)	81 (32.9%)
$150,001–200,000	20 (18.5%)	31 (22.5%)	51 (20.7%)
Primary caregiver education, n (%)
Some high school	0 (0%)	1 (0.645%)	1 (0.362%)
High school or equivalent	1 (0.826%)	1 (0.645%)	2 (0.725%)
Some college	10 (8.26%)	17 (11.0%)	27 (9.78%)
Community college/2-year degree	7 (5.79%)	8 (5.16%)	15 (5.43%)
Bachelor’s degree	33 (27.3%)	58 (37.4%)	91 (33.0%)
Some graduate school	9 (7.44%)	12 (7.74%)	21 (7.61%)
Master’s degree	52 (43.0%)	49 (31.6%)	101 (36.6%)
Doctoral degree	8 (6.61%)	9 (5.81%)	17 (6.16%)
Other	1 (0.826%)	0 (0%)	1 (0.362%)
Race, n (%)
American Indian/Alaskan Native	13 (10.6%)	0 (0%)	13 (4.55%)
Asian	51 (41.5%)	1 (0.613%)	52 (18.2%)
Black	6 (4.88%)	2 (1.23%)	8 (2.80%)
Native Hawaiian/Pacific Islander	0 (0%)	0 (0%)	0 (0%)
White	49 (39.8%)	147 (90.2%)	196 (68.5%)
Multiracial	4 (3.25%)	13 (7.98%)	17 (5.94%)
Current Support, mean (SD)*	−0.045 (0.429)	0.055 (0.428)	0.011 (0.430)
Current Stress, mean (SD)*	0.091 (0.580)	−0.07 (0.521)	0.001 (0.553)

^*Current stress and current support are standardized values.

TABLE 2.

Correlation matrix

			1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32
		Tanner stage S1	-
	2	Tanner stage S2	0.895***	-
	3	Tanner stage S3	0.808***	0.902***	-
	4	Current stress	−0.003	−0.052	−0.038	-
	5	Current support	0.042	0.099	0.084	−0.509***	-
Cortisol: wake	6	S1 day 1	−0.109	−0.150*	−0.088	0.004	−0.048	-
	7	S1 day 2	−0.121	−0.120	−0.081	0.094	−0.067	0.501***	-
	8	S1 day 3	−0.125*	−0.173**	−0.126	0.126	−0.109	0.298***	0.404***	-
	9	S2 day 1	−0.050	−0.120	−0.047	−0.081	−0.007	0.193**	0.244***	0.193**	-
	10	S2 day 2	0.052	−0.008	0.026	−0.007	−0.026	0.233***	0.246***	0.169*	0.380***	-
	11	S2 day 3	0.070	0.043	0.047	−0.080	−0.023	0.219**	0.157*	0.254***	0.327***	0.365***	-
	12	S3 day 1	−0.068	−0.049	−0.034	−0.037	0.017	0.249***	0.307***	0.163*	0.253***	0.165*	0.180*	-
	13	S3 day 2	0.110	0.079	0.108	−0.016	0.024	0.301***	0.197**	0.175*	0.280***	0.170*	0.274***	0.304***	-
	14	S3 day 3	0.015	−0.014	−0.005	0.024	−0.013	0.139	0.186**	0.176*	0.313***	0.253***	0.358***	0.299***	0.411***	-
Cortisol: 30 min	15	S1 day 1	0.035	0.001	0.102	−0.073	−0.069	0.294***	0.299***	0.264***	0.185**	0.094	0.139	0.113	0.123	0.058	-
	16	S1 day 2	0.123	0.094	0.165*	−0.031	0.028	0.157*	0.282***	0.228***	0.097	−0.022	0.006	0.118	0.138	0.010	0.507***	-
	17	S1 day 3	0.062	0.008	0.065	−0.104	0.048	0.210**	0.252***	0.207**	0.183*	0.099	0.184*	0.220**	0.205**	0.208**	0.394***	0.484***	-
	18	S2 day 1	0.014	−0.036	0.026	−0.032	0.000	0.068	0.087	0.016	0.379***	0.217**	0.255***	0.046	0.182*	0.109	0.293***	0.219**	0.249***	-
	19	S2 day 2	0.113	0.078	0.059	−0.125	−0.011	0.075	0.132	0.070	0.259***	0.316***	0.269***	0.151*	0.074	0.148*	0.310**	0.368***	0.298**	0.397***	-
	20	S2 day 3	0.087	0.061	0.072	−0.034	0.013	0.171	0.081	0.126	0.110	0.098	0.374***	0.099	0.089	0.220	0.313***	0.252***	0.250***	0.362***	0.404***	-
	21	S3 day 1	−0.021	0.020	0.015	−0.081	0.037	0.108	0.112	0.090	0.204**	0.133	0.123	0.313***	0.315***	0.316***	0.158*	0.151*	0.285***	0.231**	0.155*	0.180*	-
	22	S3 day 2	−0.036	−0.053	0.008	−0.058	−0.058	0.166*	0.213**	0.131	0.314***	0.125	0.180*	0.204**	0.283***	0.190**	0.390***	0.316***	0.283***	0.391***	0.192*	0.130	0.504***	-
	23	S3 day 3	0.057	0.046	0.083	0.004	0.016	0.145*	0.047	0.164*	0.221**	0.181*	0.137	0.166*	0.184**	0.303***	0.353***	0.270***	0.208**	0.297***	0.255***	0.205**	0.402***	0.505***	-
Cortisol: bed	24	S1 day 1	−0.043	−0.028	−0.034	0.039	−0.071	0.215***	0.163*	0.146*	0.089	0.228	−0.009	0.075	0.061	0.053	0.217***	0.150	0.132*	0.067	0.221**	0.094	0.042	0.139	0.146*	-
	25	S1 day 2	0.084	0.055	0.069	0.028	−0.068	0.162*	0.213***	0.149*	0.103	0.155*	0.004	0.071	0.051	0.119	0.154*	0.196**	0.156*	0.086	0.075	0.052	0.045	0.037	0.063	0.491***	-
	26	S1 day 3	0.019	0.020	−0.050	0.023	−0.006	0.095	0.230***	0.129*	−0.034	0.039	0.058	0.025	0.012	0.032	0.168*	0.162*	0.149*	−0.008	0.036	0.033	0.045	0.182*	0.098	0.398***	0.468***	-
	27	S2 day 1	0.195**	0.126	0.051	0.014	0.026	0.004	0.047	0.009	0.123	0.168*	0.171*	0.059	0.034	0.043	0.051	0.093	0.117	0.084	0.282***	0.120	0.150*	0.001	0.063	0.131	0.099	0.211**	-
	28	S2 day 2	0.246**	0.226***	0.163*	−0.065	0.024	−0.041	−0.008	0.016	0.115	0.169*	0.169*	0.074	0.051	0.090	0.030	0.085	0.142	0.119	0.304***	0.166*	0.163*	0.045	0.147*	0.146*	0.214**	0.314***	0.525***	-
	29	S2 day 3	0.168**	0.088	0.048	−0.029	−0.055	0.058	0.044	0.127	−0.022	0.135*	0.159*	0.007	−0.005	0.038	0.066	0.037	0.024	0.037	0.141*	0.136*	0.060	0.095	0.040	0.274***	0.205**	0.297***	0.457***	0.472***	-
	30	S3 day 1	0.258***	0.230***	0.195**	−0.072	0.008	0.036	−0.077	0.099	0.141*	0.129	0.215**	0.160*	0.175**	0.133*	0.048	0.005	0.222**	0.057	0.135	0.121	0.172*	0.135	0.121	0.009	0.063	0.133	0.306***	0.391***	0.315***	-
	31	S3 day 2	0.291***	0.282***	0.268***	0.053	−0.010	0.020	0.039	0.092	0.213**	0.110	0.159*	0.156*	0.199**	0.190**	0.197**	0.182*	0.194**	0.111	0.169*	0.114	0.217**	0.202**	0.256***	0.137*	0.076	0.137	0.245***	0.260***	0.242***	0.415***	-
	32	S3 day 3	0.217***	0.199**	0.167*	0.058	−0.028	0.070	0.061	0.006	0.028	0.084	0.034	0.150*	0.074	0.129	0.022	−0.016	0.069	−0.060	0.196**	0.040	0.144*	0.031	0.143*	0.173*	0.152*	0.152*	0.259***	0.304***	0.329***	0.386***	0.462***	-

Note: Correlations are Pearson correlation coefficients.

^*p < .05

^**p < .01

^***p < .001.

3.1 |. CAR

Table 3 displays the results of the hierarchical linear mixed-effects models fit to model the CAR across the whole sample. There was a significant increase in cortisol from the wake to 30 min sample on average for all participants, indicating the presence of the CAR. Sex, day of the sample, and the medication Granger score were not significant predictors and thus were excluded from the model. There was no effect of ELS on the CAR (model 1 in Table 3, Figure 3(A)). Mean pubertal stage predicted the CAR, such that participants later in puberty overall displayed steeper CAR (model 1 in Table 3, Figure 3(B)). There was no evidence of pubertal recalibration, Table 1 (model 2), such that the interaction terms, puberty and adversity exposure, were not statistically significant in predicting CAR. Model 3, examining current social stress, found a significant interaction between current social stress, pubertal change, and the CAR (Table 3). Higher levels of social stress were associated with steeper CAR cortisol slopes as all youth progressed through puberty (Figure 5(A)). Model 4, examined current social support and showed a significant interaction between current social support, pubertal change, and the CAR (Table 3). Higher levels of social support were associated with less steep CAR cortisol slopes as all participants progressed through puberty (Figure 5(B)).

TABLE 3.

Predicting cortisol awakening response (CAR)

	CAR model 1Early adversity and puberty			CAR model 2: Pubertal recalibration			CAR model 3:Puberty and current stress			CAR model 4:Puberty and current support
Predictors	Estimates	CI	p	Estimates	CI	p	Estimates	CI	p	Estimates	CI	p
(Intercept)	−0.557	−0.631–−0.484	<.001	−0.570	−0.651–−0.489	<.001	−0.559	−0.633–−0.484	<.001	−0.557	−0.631–−0.482	<.001
Waketime	−0.007	−0.016–−0.002	.122	−0.007	−0.017–0.002	.114	−0.008	−0.017–0.002	.101	−0.008	−0.017–0.001	.093
Time since wake	0.295	0.231–0.359	<.001	0.283	0.192–0.373	<.001	0.312	0.247–0.378	<.001	0.313	0.247–0.379	<.001
Group [NA]	0.007	−0.020–0.033	.619	0.029	−0.032–0.089	.349	0.012	−0.015–0.039	.398	0.011	0.016–0.038	.439
Tanner stage: longitudinal Δ (within-individual)	−0.018	−0.037–0.001	.066	−0.008	−0.022–0.006	.289	−0.020	−0.039–−0.000	.047	−0.018	−0.038–0.001	.065
Tanner stage: individual mean (between-individual)	−0.012	−0.021–−0.003	.008	−0.024	−0.052–0.004	.093	−0.011	−0.020–−0.002	.018	−0.011	−0.020–−0.002	.014
Time*group [NA]	0.022	−0.028–0.072	.384	0.042	−0.072–0.156	.473	0.002	−0.049–0.053	.934	0.005	−0.046–0.056	.843
Time*Tanner stage: longitudinal Δ	−0.004	−0.039–0.031	.834	−0.006	−0.058–0.046	.826	0.005	−0.031–0.040	.798	0.005	−0.030–0.040	.787
Time*Tanner stage: individual mean	0.026	0.009–0.043	.002	0.030	0.003–0.057	.027	0.023	0.005–0.040	.010	0.022	0.005–0.040	.011
Group [NA]*Tanner stage: longitudinal Δ				0.011	−0.027–0.049	.564
Group [NA]* Tanner stage: individual mean				−0.007	−0.025–0.011	.423
Time*group [NA]* Tanner stage: individual mean				−0.007	−0.041–0.028	.709
Time*group[NA]* Tanner stage: longitudinal Δ				0.004	−0.067–0.074	.920
Stress (model 3) or support (model 4)							0.028	−0.039–0.096	.407	−0.013	−0.097–0.072	.765
Time*stress (model 3) or support (model 4)							−0.088	−0.218–0.042	.185	0.111	−0.054–0.276	.188
Group [NA]*stress (model 3) or support (model 4)							0.041	−0.010–0.091	.113	0.026	−0.039–0.091	.436
Stress (model 3) or support (model 4)* Tanner stage: longitudinal Δ							−0.040	−0.076–−0.005	.026	0.048	0.003–0.094	.037
Stress (model 3) or support (model 4)* Tanner stage: individual mean							−0.012	−0.030–0.005	.157	−0.003	−0.024–0.018	.801
Time* group [NA]*stress (model 3) or support (model 4)							−0.028	−0.126–0.069	.567	−0.101	−0.226–0.024	.112
Time*stress (model 3) or support (model 4)* Tanner stage: longitudinal Δ							0.095	0.028–0.162	.005	−0.119	−0.204–−0.033	.007
Time* stress (model 3) or support (model 4) * Tanner stage: individual mean							0.017	−0.017–0.050	.322	0.000	−0.040–0.041	.984
Random effects
σ 2	0.04			0.04			0.04			0.04
τ 00 SUBID:session	0.02			0.02			0.02			0.02
τ 00 session	0.00			0.00			0.00			0.00
ICC	0.30			0.30			0.29			0.29
N SUBID	289			289			281			281
N session	3			3			3			3
Observations	3772			3772			3670			3670
Marginal R2/conditional R2	0.156/0.406			0.157/0.406			0.159/0.405			0.157/0.405

Note: Estimates provided are standardized coefficients. “Time since wake” CAR slope as the variable used to model the change in cortisol production from the wake sample to the 30 min sample. “Day” is a categorical variable, dummy coded with day 1 as the reference point. Pubertal (Tanner) stage, stress, and support were grand-mean-centered. CI = confidence interval. NA = Non-adopted youth.

FIGURE 3.

There was a significant increase in cortisol from the wake to 30 min sample, indicating the presence of the cortisol awakening response (CAR). (A) There was no effect of early life adversity on the CAR. (B) Between-individual change in puberty (mean pubertal stage) predicted the CAR, such that participants later in puberty overall displayed steeper CAR (p = .002), but there was no evidence of pubertal recalibration of the CAR

FIGURE 5.

(A) Model 3, examining current social stress, found a significant interaction between current social stress, the within-subject pubertal stage, and the CAR (Table 3). Higher levels of social stress were associated with steeper CAR cortisol slopes as they progressed through puberty. (B) Model 4 examined current social support and showed a significant interaction between current social support, the within-subject pubertal stage, and the CAR: Higher levels of social support were associated with less steep CAR cortisol slopes as participants progressed through puberty

3.2 |. Diurnal (wake to bed) cortisol slope

Table 4 displays HLM results for models predicting DS. ELS predicted diurnal cortisol slope (DS), such that PI youth had a flatter (less steep) DS, compared to comparison youth (Figure 4(A)) Puberty was associated with DS both in terms of the mean pubertal stage and pubertal change. Higher Tanner stages overall were associated with higher bedtime cortisol levels, and bedtime cortisol increased as participants progressed through pubertal stages (model 1 in Table 4 and Figure 4(B)). There was no evidence of pubertal recalibration in the DS (model 2 in Table 4). In examining current social stress, we found a significant interaction between current social stress, mean pubertal stage, and the DS (model 3, Table 4). Higher levels of social stress were associated with higher bedtime cortisol values for individuals in later stages of puberty (Figure 6(A)). In examining current social support we found a significant interaction between current social support, pubertal change, and the DS (model 4, Table 4). The highest levels of social support were associated with lower bedtime cortisol levels as participants progressed through stages of puberty, compared to the lowest levels of social support (Figure 6(B)).

TABLE 4.

Predicting wake to bed cortisol

	Diurnal model 1: Early adversity and puberty			Diurnal model 2: Pubertal recalibration			Diurnal model 3: Puberty and current stress			Diurnal model 4: Puberty and current support
Predictors	Estimates	CI	p	Estimates	CI	p	Estimates	CI	p	Estimates	CI	p
(Intercept)	−0.500	−0.605–−0.395	<.001	−0.514	−0.629–−0.399	<.001	−0.498	−0.603–−0.392	<.001	−0.497	−0.602–−0.392	<.001
Day 2	−0.036	−0.059–−0.013	.002	−0.036	−0.059–−0.013	.002	−0.037	−0.060–−0.014	.002	−0.037	−0.060–−0.014	.002
Day 3	−0.027	−0.050–−0.004	.024	−0.027	−0.050–−0.004	.024	−0.029	−0.052–−0.005	.016	−0.029	−0.053–−0.006	.015
Waketime	−0.011	−0.024–0.002	.092	−0.011	−0.024–0.002	.093	−0.012	−0.025–0.001	.063	−0.012	−0.025–0.001	.063
Time since wake	−0.080	−0.083–−0.076	<.001	−0.078	−0.083–−0.073	<.001	−0.080	−0.083–−0.076	<.001	−0.080	−0.083–−0.076	<.001
Group [NA]	0.009	−0.029–0.046	.649	0.031	−0.054–0.116	.476	0.014	−0.024–0.052	.474	0.014	−0.025 – 0.052	.485
Tanner stage: longitudinal Δ (within-individual)	−0.022	−0.051–0.008	.148	−0.027	−0.068–0.015	.214	−0.021	−0.050–0.008	.154	−0.021	−0.049–0.008	.158
Tanner stage: individual mean (between-individual)	−0.013	−0.025–−0.000	.047	−0.008	−0.028–0.012	.415	−0.011	−0.024–0.002	.086	−0.012	−0.025–0.001	.073
Time*group [NA]	−0.004	−0.007–−0.002	.002	−0.007	−0.013–−0.000	.041	−0.004	−0.007–−0.002	.002	−0.005	−0.007–−0.002	.001
Time*Tanner stage: longitudinal Δ	0.003	0.001–0.005	.002	0.005	0.002–0.007	.002	0.003	0.001–0.005	.001	0.003	0.001–0.005	.001
Time*Tanner stage: individual mean	0.004	0.003–0.005	<.001	0.004	0.002–0.005	<.001	0.004	0.003–0.005	<.001	0.004	0.003 – 0.005	<.001
Group [NA]*Tanner stage: longitudinal Δ				0.009	−0.044–0.063	.735
Group [NA]* Tanner stage: individual mean				−0.007	−0.033–0.018	.571
Time*group [NA]* Tanner stage: individual mean				0.001	−0.001–0.003	.432
Time*group [NA]* Tanner stage: longitudinal Δ				−0.003	−0.007–0.001	.146
Stress (model 3) or support (model 4)							0.044	−0.049–0.137	.355	−0.023	−0.140–0.094	.695
Time*stress (model 3) or Support (model 4)							−0.012	−0.019–−0.005	<.001	0.007	−0.001–0.016	.097
Group [NA]*stress (model 3) or support (model 4)							0.033	−0.038–0.103	.364	0.021	−0.069–0.112	.642
Stress (model 3) or support (model 4)*Tanner stage: longitudinal Δ							−0.035	−0.086–0.015	.167	0.044	−0.020–0.108	.175
Stress (model 3) or support (model 4) * Tanner stage: individual mean							−0.016	−0.040–0.008	.185	0.001	−0.028–0.030	.944
Time*group [NA]*stress (model 3) or support (model 4)							−0.002	−0.007–0.003	.363	0.006	−0.000–0.013	.066
Time*stress (model 3) or support (model 4)*Tanner stage: longitudinal Δ							0.002	−0.002–0.006	.273	−0.001	−0.006–0.003	.565
Time*stress (model 3) or support (model 4) ‘Tanner stage: individual mean							0.004	0.002–0.005	<.001	−0.003	−0.005–−0.001	.014
Random effects
σ 2	0.09			0.09			0.09			0.09
τ 00 SUBID:session	0.03			0.03			0.03			0.03
τ 00 session	0.00			0.00			0.00			0.00
ICC	0.25			0.25			0.25			0.25
N SUBID	290			290			282			282
N session	3			3			3			3
Observations	3976			3976			3868			3868
Marginal R2/conditional R2	0.666/0.751			0.666/0.751			0.669/0.752			0.670/0.751

Note: Estimates provided are standardized coefficients. “Time since wake” CAR slope as the variable used to model the change in cortisol production from the wake sample to the 30 min sample. “Day” is a categorical variable, dummy coded with day 1 as the reference point. Pubertal (Tanner) stage, stress, and support were grand-mean-centered.

FIGURE 4.

(A) Wake to bed cortisol slope and early life adversity. (B) Wake to bed cortisol slope and pubertal stage

FIGURE 6.

(A) Wake to bed cortisol slope: the interaction between puberty and current stress. (B) Wake to bed cortisol slope: the interaction between puberty and current support

4 |. DISCUSSION

This study was designed to investigate the longitudinal impact of ELS, the pubertal transition, and current life stress and support on morning awakening and diurnal cortisol slopes. These results suggest that early adversity, current experiences of stress and social support, and the pubertal transition can independently and jointly influence diurnal cortisol slope, notably associated with higher levels at bedtime and increases in the CAR. Though pubertal recalibration was found in this sample of adolescents in response to an acute stressor in the lab (Gunnar, 2019), the present study found no evidence to support pubertal recalibration in diurnal cortisol patterns of the CAR or DS. Flannery et al. (2017) found that older PI youth exhibited DS similar to that of comparison youth, and Quevedo et al. (2012) showed more similar CAR between PI and comparison youth with increasing age. However, both of these studies were cross-sectional and thus were unable to test the necessary longitudinal piece of the pubertal recalibration hypothesis.

In our study, ELS predicted flatter DSs, compared to NA peers, across all pubertal stages. However, ELS was not associated with changes in the CAR. Our finding in regard to DSs is similar to a study of adult adoptees, where those who had reported experiencing severe neglect early in life also had a flatter DS (van der Vegt et al., 2010). Accordingly, flatter DS is also associated with a number of mental and physical health outcomes (Adam et al., 2017), which is consistent with a pattern of mental and physical health outcomes seen in international adoptees who have experienced neglect in the form of institutional care (Gunnar & Reid, 2019). For DS, it is possible we did not get an accurate measure of bedtime cortisol because we instructed them to take the sample at bedtime or at 9:00 p.m., whichever came first. However, only 10.1% of all participants had a bedtime past 9:00 p.m., and no group differences were found, suggesting that many of the cortisol values for bedtime reflect the before sleep cortisol value. This study on the CAR extends cross-sectional work in this sample (Leneman et al., 2018), which found that ELS was associated with a flatter CAR only for late-adopted children but was unable to examine longitudinal effects. While significant effects of ELS on the CAR have been observed in other studies (King et al., 2017; Mangold et al., 2010; Quevedo et al., 2012), those findings have been mixed. For instance, Quevedo et al. (2012) showed more similar CAR between youth with and without ELS, while King et al. (2017) found that that youth who experienced ELS (across childhood) show a larger CAR later in puberty than those without ELS. Continued exposure to stress across childhood may be necessary to see the effects of ELS on the CAR. The present study is notable in that it examines time-limited ELS in a large, longitudinal sample with the gold-standard measurement of puberty. It is possible that Quevedo et al.’s findings were confounded by their cross-sectional nature, whereas these findings are able to tease apart within-individual and between-individual changes. Additionally, the King et al.’s study used the self-reported pubertal stage, which may influence a youth’s experience of stress as they compare their pubertal development to their peers.

Similar to previous work, we found that youth exhibited smaller diurnal decreases as they developed physically (Shirtcliff et al., 2012). However, unlike Shirtcliff et al., we found no evidence of sex differences (2012). Unlike Adam et al. (2006), we found that higher stages of pubertal development were associated with higher bedtime cortisol levels and less steep decrease across the day. Older adolescents may be engaged in more vigorous physical sport in the evening period, which can keep cortisol higher later into the evening (Kertes & Gunnar, 2004). We also found no evidence of within-individual change of CAR across puberty. This may be due to a combination of reasons. First, the CAR is hypothesized to reflect the anticipatory challenge of the day ahead (Clow et al., 2010; Fries et al., 2009). Second, an increasing CAR with higher pubertal stages may reflect the maturation of cortico-limbic pathways and a tighter integration of these pathways with HPA regulatory mechanisms in the hypothalamus (Gogtay et al., 2004; Romeo, 2010). Taken together, we might see an effect of the mean pubertal stage but not pubertal change because participants who had higher mean pubertal stages (and are thus often older at the start of the study) may experience higher anticipatory stress in response to salivary saliva collection and thus higher CAR. In contrast, participants who overall had a lower mean pubertal score (and were younger at the start of the study) may have acclimated to the anticipation of collecting saliva samples. In contrast, the DS, which is not thought to reflect anticipatory stress but has shown signs of change with maturation, exhibited changes in both mean pubertal stage and pubertal change.

Similar to studies of adults, we found that current life stress and social support influenced the CAR and the DS. Current life stressors, the availability of social resources, and anticipatory stress have previously been associated with the CAR (Fries et al., 2009; Gartland et al., 2014; Lovell et al., 2012; Proulx et al., 2017; Stalder et al., 2016; Williams et al., 2005). Similarly, DS is sensitive to emotional and psychosocial stress, with adverse psychosocial experience associated with a flatter DS (Adam & Gunnar, 2001; Adam et al., 2007; Doane et al., 2013). These studies, along with the current study, suggest that current environmental circumstances can influence the CAR and DS, pointing to potential social drivers of change in diurnal cortisol patterns over time.

This study is strengthened by its longitudinal approach and the examination of the role of both ELS and current stress/support; however, there are a number of limitations that should be considered. First, current stress and support were measured via youth-report, the quality of which can vary by age even in a structured interview format. Also, there is evidence that, later in pubertal development, a sex difference in the CAR emerges such that, for females, cortisol peaks at 45 min rather than 30 (Oskis et al., 2009; Platje et al., 2013). Presumably, though, this would affect the absolute level of cortisol at peak (with females reaching a higher peak) but not substantially impact the slope of cortisol production across the first 30 min post-awakening, which was measured here. Results for the DS also suggest that future work should account for systematic differences across sampling days. Last, this study only examines the longitudinal change in CAR and DS over 2 years, though the ages of participants spanned a wide range of late childhood and early adolescence. Future studies should examine diurnal cortisol patterns over a longer time span to better understand the developmental trajectory across the peripubertal period. Nonetheless, this study is strengthened by the ability to disentangle within-individual and between-individual changes in puberty’s influence on diurnal cortisol patterns. In the future, it will also be important to contextualize these developmental trajectories by investigating their associations with measures of physical, mental, or emotional well-being.

In conclusion, this study was the first to our knowledge to examine longitudinal changes in the CAR and DS during the peripubertal period and its associations with both early adversity and current stressors/supports. Results suggested that aspects of the quality of the current social environment, but not necessarily experiences of early institutional care, predicted changes both in measures of diurnal cortisol patterns across 2 years of late childhood/early adolescence, a period of enhanced biological plasticity and social/relational changes.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.