Developmental Effects in Physiological Stress in Early Adolescents with and without Autism Spectrum Disorder

Abstract

Humans place high value on how they are socially evaluated by others. The Trier Social Stress Test (TSST) is a well-established measure of social evaluative threat that promotes activation of the hypothalamic pituitary adrenal (HPA) axis and release of cortisol. Higher cortisol responses in typically developing (TD) adolescents are influenced by age and pubertal development especially in later stages. Children with ASD have been shown to exhibit blunted cortisol in response to the TSST although adults with ASD show a more prototypical response. The current study examined physiological stress in early adolescents with ASD and TD. It was hypothesized that TD youth would show elevated cortisol in response to the TSST influenced by age and pubertal stage. In contrast, youth with ASD would show a more diminished stress response yet still show effects for age and pubertal development.

Methods:

The sample included 241 youth, 138 with ASD (median age = 11.25) and 103 TD (median age=11.67). Standardized diagnostic and pubertal development (genital/breast (GB), and pubic hair (PH) stage) physical exams were performed. Salivary cortisol was collected before and after the TSST. Linear mixed effects models examined the effects of baseline cortisol, time, age, sex, pubertal stage, and diagnosis.

Results:

We did not find an effect of early pubertal development stage (GB or PH) on cortisol response. There was an interaction between age and TSST timepoint, showing stronger effects for older children across the timeline especially during the stressor. Finally, there was a significant diagnosis by TSST timepoint interaction characterized by a blunted cortisol stress response in youth with ASD compared to TD participants who showed higher cortisol.

Discussion:

We found evidence that age contributes to an increase in cortisol in response to social evaluative threat during early adolescence. TD youth exhibit an adaptive elevated stress response to psychosocial threat whereas youth with ASD do not. There may exist a developmental lag in the perception of and stress responsivity to social evaluation which may emerge in older adolescents with ASD.

1. Introduction

Humans are fundamentally social and place high value on how they are perceived by others. Circumstances in which people perceive they are being negatively evaluated can be threatening and result in increased physiological and psychological stress. This predisposition is heightened during adolescence when youth place even greater importance on peer and romantic relationships (Graber and Brooks-Gunn, 1996). Due to the dynamic psychological, social and physical changes that define it, adolescence has been described as a critical transition period (Dahl, 2004); yet, it can also be a time of remarkable opportunity (Masten et al., 2004; Romeo and McEwen, 2006).

The terms adolescence and puberty are often used interchangeably, in this context when distinctions are warranted, adolescence refers to the age-related developmental transition of juvenile social and cognitive maturation into adult forms. Adolescence involves ongoing brain-based structural and functional development especially with regards to emerging skills in executive functioning (e.g., self-regulation) and social cognition (e.g., theory of mind) (Blakemore and Choudhury, 2006) associated with maturation of the frontal lobes (Blakemore, 2008; Luna et al., 2010). Executive function or the ability to control and coordinate thoughts, emotions and behavior, continues to develop during the adolescent years (Luciana et al., 2005). Also, theory of mind, defined as the ability to understand mental states to predict one’s own behavior as well as the behavior of others (Premack and Woodruff, 1978), extends well into late adolescence (Dumontheil et al., 2010). Thus, these age-related maturational skills in sociocognitive functioning impact the way in which youth interact with and respond to the actions of others.

Puberty refers to biological maturation of sexual systems resulting in significant changes in morphology, cognition, emotion regulation and physiological stress responsivity (e.g., Chrousos et al., 1998; Spear, 2000). Such developmental changes are particularly evident in the regulation and responsivity of the hypothalamic–pituitary–adrenocortical (HPA) axis (Stroud et al., 2009). Cortisol, a glucocorticoid hormone, is the primary output of the HPA axis in humans and an established biomarker of physiological stress (Hellhammer et al., 2009). The biopsychosocial changes in puberty contribute to heightened activation of the HPA axis and increased cortisol in response to perceived stressors (Gunnar et al., 2009; Sisk and Foster, 2004). Activation during periods of affective change is normative and increased stress sensitivity arising during adolescence (e.g., Andersen and Teicher, 2008; Dahl, 2004; Sumter et al., 2010) is correlated with puberty (Romeo and McEwen, 2006). The puberty-HPA stress hypothesis proposes enhanced stress reactivity with the emergence of sexual maturation (Gunnar et al., 2009) such that a child’s responsivity to stressors begins to exhibit a more adult-like pattern (Spear, 2000), suggesting the HPA axis may be reset over the course of puberty. The increased stress response appears especially salient for anticipation of social evaluation, which has been associated with age and pubertal maturation (Sumter et al., 2010).

Stress is a composite, multidimensional construct (Levine, 2005), defined here as the psychological and physiological reactivity to threatening events. When the threat is significant, the HPA axis neuroendocrine cascade is set into motion which releases corticotrophin-releasing hormone (CRH) and arginine vasopressin (AVP) from the paraventricular nucleus of the hypothalamus (Whitnall, 1993). Subsequently, CRH and AVP bind to receptors on the anterior pituitary stimulating the release of adrenocorticotropic hormone from the anterior pituitary gland, resulting in release of cortisol (Herman and Cullinan, 1997; Jacobson and Sapolsky, 1991).

The enhanced HPA responsivity during the adolescent period may help prepare the adolescent to adapt to increased cognitive and social demands and new challenges (Stroud et al., 2009). While such change during adolescence is normative it may increase vulnerability in populations prone to enhanced physiological arousal and poor adaption to change (Herbert, 1997; Stroud et al., 2009) such as individuals with autism.

Autism spectrum disorder (ASD) is characterized by core impairment in reciprocal social communication and a repertoire of repetitive, restricted interests and behaviors (APA, 2013). The Center for Disease Control estimates that 1 in 54 children 8-years of age are diagnosed with ASD in the United States with approximately 4.3 times as many boys as girls diagnosed (Maenner et al., 2020). As a result of core impairments in social cognition and behavior (APA, 2013), ASD has the potential to significantly impact how an individual perceives and physiologically responds to the environment. The adolescent period may be a particularly vulnerable time for youth with ASD due to an increase in social expectations, neural reorganization, and onset of puberty (Picci and Scherf, 2015).

The extant literature indicates that stress responsivity in individuals with ASD is influenced by a variety of factors to include the context. Higher cortisol levels are often found in children with ASD in response to non-social stimuli (e.g., Corbett et al., 2006; Corbett et al., 2008; Spratt et al., 2012) presumably as a result of novelty and uncertainty. Additionally, certain social scenarios result in enhanced activation of the HPA axis in children with ASD compared to typically developing (TD) children such as school integration (Richdale and Prior, 1992), social interaction with peers (Corbett et al., 2012; Corbett et al., 2010) and engagement with unfamiliar children (Lopata et al., 2008). Thus, it appears that social interaction with other children may be stressful for children with autism. However, other contexts, such as psychosocial stress in which the child is being socially evaluated by others have not consistently induced a stress response in children (Corbett et al., 2012; Jansen et al., 2000; Lanni et al., 2012; Levine et al., 2012) and adolescents (Edmiston et al., 2017; Hollocks et al., 2014) with ASD. While intriguing, these previous studies examining social evaluative threat were conducted on relatively small, predominantly male samples and findings were primarily based on within-group comparisons. Moreover, a nuanced consideration of what aspects of the stressor contributed to the differences (e.g., baseline, social stress or recovery) and the impact of developmental factors (e.g., age and puberty) have not been elucidated.

Interestingly, a direct comparison between adolescents (13-to-17 years) and adults (18-to-22 adults) with ASD, showed adults exhibit a more prototypical elevation in cortisol in anticipation of and in response to social evaluative threat (Taylor et al., 2018), suggesting developmental influences in stress responsivity to social evaluation that may relate to sociocognitive awareness. Indeed, age has been a critical moderating factor in the activation of the HPA axis such that older children with ASD between 8-to-12 years have shown higher cortisol during play interactions compared to same-age peers who did not show a cortisol response (Corbett et al., 2012; Schupp et al., 2013). The higher cortisol in the children with ASD during peer interactions was hypothesized to be the result of increased insight over time into their social difficulties or in reaction to a history of negative social encounters.

Although the research has been limited, pubertal development may also play an important role in HPA axis regulation and reactivity in ASD. For example, pubertal development and age were found to be significant predictors of higher evening cortisol in youth with ASD compared to TD youth (Muscatello and Corbett, 2018). Thus, considering the literature on the HPA axis in typical development (e.g., Gunnar et al., 2009; Shirtcliff et al., 2009; Stroud et al., 2009) and the emerging literature in ASD (e.g., Corbett et al., 2012; Corbett et al., 2010; Muscatello and Corbett, 2018), stress reactivity might be even more developmentally nuanced for adolescents on the autism spectrum.

The aim of the study was to examine hypothesized developmental effects (age and puberty) and physiological stress (cortisol anticipation and response) to social evaluative threat in early adolescents with ASD or TD. The current study builds on previous research through the inclusion of a large sample of youth, rigorous, between-subject statistical approach and consideration of developmental factors that may contribute to cortisol response. Based on the aforementioned research, the following hypotheses were made: 1) Higher pubertal status will predict higher stress cortisol response to the TSST in the TD and the ASD groups; 2) Older age will predict higher stress cortisol response to the TSST in the TD and ASD groups; and 3) TD adolescents will evidence elevated cortisol to social evaluation compared to adolescents with ASD who will show a blunted cortisol profile.

2. Methods

2.1. Participants

Data were collected as part of the SENSE longitudinal study on pubertal development and stress (Corbett, 2017). The current study includes data from Year-1 enrollment when the children were between 10-years-0-months to 13-years-11-months of age. The total sample that completed the social stress protocol included 241 youth, 138 with ASD (median age = 11.25, IQR= 10.5, 12.25) and 103 TD (median age = 11.67, IQR= 10.58, 12.65). The ASD group included 36 (26.1%) females and the TD group included 46 (44.7%) females.

Inclusion in the study required an intelligence quotient (IQ) score ≥ 70 and either a confirmed diagnosis of ASD (see below) or determined to be TD with no known developmental delay or neurodevelopmental diagnosis. Exclusion criteria included current use of medications known to alter the Hypothalamic-Pituitary-Adrenal (HPA) axis (e.g., corticosteroids; see (Granger et al., 2009)) or HPG axis (e.g., growth hormone), or medical condition known to impact pubertal development (e.g., Cushing’s Disease). Also, participants taking oral contraceptives, growth hormones, or nicotine all known to influence the HPA axis, were excluded (Foley and Kirschbaum, 2010; Kirschbaum et al., 1995). Regarding medication, 65.2% of youth in the ASD group were taking at least one medication, while 17.5% of TD participants reported taking a daily medication. Medication use included stimulants, melatonin, selective-serotonin reuptake inhibitors, antihistamines, and alpha-agonists.

The research was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). The Vanderbilt Institutional Review Board (IRB) approved the study. Informed written consent/assent was obtained from all care providers and study participants, respectively, prior to inclusion in the study. Participation required two research visits to the University. On visit 1, the diagnostic and cognitive measures were administered. The participants were trained on the salivary collection methods for home sampling. On visit 2, participants were exposed to the TSST (Buske-Kirschbaum et al., 2003; Kirschbaum et al., 1993)

2.2. Diagnostic Procedures

The diagnosis of ASD was based on the Diagnostic and Statistical Manual-5 (APA, 2013) and confirmed by: (1) a previous diagnosis by a psychologist, psychiatrist, or behavioral pediatrician with autism expertise; (2) current clinical judgment, and (3) corroborated by the Autism Diagnostic Observation Schedule Module 3 (ADOS-2; (Lord et al., 2012), a semi-structured interview-based instrument administered by research-reliable personnel. The Social Communication Questionnaire (SCQ; (Rutter et al., 2003), was used to screen ASD. A score of 15 is suggestive of ASD; thus, TD youth with a score ≥ 10 were excluded. The Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II, (Wechsler, 2011)) was used to obtain an estimate of the child’s intellectual functioning (IQ ≥ 70 required).

2.3. Psychosocial/Developmental Measures

While different assessment approaches are available to assess pubertal timing (Mendle et al., 2019), recent research with the sample comparing physical exam to parent- and self-report demonstrated that physical exam was the optimal gold standard approach for more objective, concurrent pubertal measurement (Corbett et al., 2019b) especially for youth with ASD. Therefore, physical exam scores were used in the current study.

Physical Examination (PE).

The PE was completed to reliably identify pubertal development and assign Tanner stage (Marshal and Tanner, 1970; Marshall and Tanner, 1969). The exam ascertained two measures with 5 stages for Genitals (G1–G5 for boys) and Breasts (B1-B5 for girls; GB stage) and Pubic hair (P1-P5 for both genders; PH stage). The exam consisted of visual inspection and categorization of pubertal and genital maturation. To be consistent with the original Tanner staging and to maximize participation, palpation of breasts or measurement of testes was not conducted.

Pubertal assessment consisted of a brief, standardized physical exam conducted by trained, licensed study physicians. A male physician conducted most of the exams, but a female physician provided same-gender exams as requested. Study physicians established rapport, explained the rationale for the exam and address any questions or concerns, which normalized the experience. During the exam, the adolescent was requested to loosen clothing to fully expose breast and lower genital region, rather than disrobing, which aided in the level of comfort for the participants. A companion (e.g., parent or same-gender research member) was offered to accompany the participant during the exam, which was conducted in a clinic exam room. Physicians were blinded to parental- and self-reports. As noted, previous research has shown significant differences between gold standard physical exam and parental- and self-reports especially for children and adolescents with ASD (Corbett et al., 2019b).

2.4. Social Stress Paradigm

During the second visit, participants completed the social evaluation paradigm (described below). Due to cortisol diurnal rhythm, the stress paradigm was conducted in the afternoon (e.g., 2:30–5:00 pm), when there are stronger, more reliable effects with less inter-individual baseline variability (Goodman et al., 2017; Seddon et al., 2020).

Trier Social Stress Test (TSST)-Child Version; (Buske-Kirschbaum et al., 2003; Kirschbaum et al., 1993) is a well-validated, experimentally induced psychosocial stressor known to reliably activate the HPA axis in TD populations (Kudielka et al., 2004a) including children and adolescents (Seddon et al., 2020). The TSST combines several elements shown to activate the HPA axis including social evaluative threat, unpredictability, and uncontrollability to produce moderate stress in most children, adolescents and adults (Dickerson and Kemeny, 2004; Kirschbaum et al., 1993). The TSST is a 20-min task divided into four subcomponents: 1) Intro/Preparation, 2) Present Speech, 3) Serial Subtraction, and 4) Debriefing. The protocol involves a scenario in which the participant must deliver the ending to a short story in front of a panel of judges (unresponsive judges showing neutral facial expressions) who will, purportedly, be judging the child’s performance against that of other children. For the protocol, mixed-age (adult and peer), as well as mixed-sex judges were used which contributes to a stronger effect than females only (Seddon et al., 2020). The 5-minute speech task is followed by a 5-minute serial subtraction task. The TSST results in a profound increase of salivary cortisol in 70–80% of participants (Kirschbaum et al., 1993; Kudielka et al., 2004b).

2.5. Dependent Measures

2.5.1. Salivary cortisol.

Salivary cortisol can be measured reliably and non-invasively utilizing small amounts of saliva, making it an ideal measure in studies of children and youth (Kirschbaum and Hellhammer, 2000). Basal salivary cortisol was collected 4 times per day (Edwards et al., 2001; Wilhelm et al., 2007), from the home over 3 consecutive week-days using established protocols (Corbett et al., 2006; Corbett et al., 2008); however, these data are not part of the current study. Participants were provided with direct instruction, a mini-manual and a DVD of step-by-step passive drool procedures. They were instructed to avoid food and drink consumption for one hour prior to sample collection. If the participant became ill, the visit was rescheduled after the participant was healthy. For females, the menstrual cycle was documented, and basal and stress cortisol was collected during the Luteal phase to reduce variability across the menstrual cycle (Kirschbaum et al., 1999).

Cortisol has a 20-min lag of detection in saliva; thus, samples were collected in intervals to capture discrete aspects of the stressor paradigm. Specifically, six samples were collected (see Figure 1): S1 Arrival (following a 20-min acclimation period); S2 Baseline; S3 Preparation period; S4 “Speech” Stressor 1; S5 Stressor 2, “Math”; and S6 Recovery.

Figure 1.

Timeline for salivary collection for the Trier Social Stress Test. Six samples were collected: S1 Arrival; S2 Baseline; S3 Preparation period; S4 “Speech” Stressor 1; S5 Stressor 2, “Math”; and S6 Recovery.

2.5.2. Cortisol assay.

Prior to assay, samples were stored at −20°C. Salivary cortisol assay was performed using a Coat-A-Count® radioimmunoassay (RIA) kit (Siemens Medical Solutions Diagnostics, Los Angeles, CA) modified to accommodate lower levels of cortisol in human saliva relative to plasma. Saliva samples were thawed and centrifuged at 3460 rpm for 15 min to separate the aqueous component from mucins and other suspended particles. All samples were duplicated. The coated tube from the kit was substituted with a glass tube into which 100ml of saliva, 100ml of cortisol antibody (courtesy of Wendell Nicholson, Vanderbilt University, Nashville, TN), and 100ml of ¹²⁵I-cortisol were mixed. After incubation at 4°C for 24h, 100ml of normal rat serum in 0.1% PO4/EDTA buffer (1:50) and precipitating reagent (PR81) were added. The mixture was centrifuged at 3460 rpm for 30 min, decanted, and counted. Serial dilution of samples indicated a linearity of 0.99. Interassay coefficient of variation was 2.06%.

2.6. Statistical Analyses:

Cortisol values were skewed; therefore, we applied a log10 transformation of the cortisol measurements prior to statistical analyses. We used linear mixed effects models to fit log10 cortisol measurements as a function of age, sex, diagnosis, pubic hair (PH) Tanner stage, arrival log10 cortisol (S1 Arrival) and a covariate indicating timepoint within the TSST protocol (5 samples). We included a random intercept for subject to account for correlation across the cortisol measurements. We used classical/parametric standard error estimates for all hypothesis tests. For null results we reported a robust effect size index (S) that can be computed directly from the chi-squared statistics, and numerator and denominator degrees of freedom (Vandekar et al., 2020). For tests of mean differences with equal proportions, S is equal to ½ Cohen’s d.

Regarding outliers, there were two participants in the ASD group that seemed to be outliers, one demonstrated a strong downward trend across the TSST while the other demonstrated a strong upward trend; however, no cortisol values were >3 SD from the mean. Further, for both participants, the cortisol values were consistent with their observed behavioral response to the TSST (e.g., did not show any distress or change in behavior before or during the paradigm). As a sensitivity analysis, we removed both subjects, but there was no qualitative difference in the result and our conclusions remained the same.

All hypotheses were tested within this model: To investigate whether pubertal stage was associated with a higher cortisol response, we tested the main effect of PH stage and subsequently tested the PH stage by TSST timepoint interaction to see whether Tanner stage differentially affects timepoints of the TSST. To investigate whether older age predicted higher cortisol response, we tested the linear main effect of age and subsequently tested the age by TSST timepoint interaction in order to examine whether age differentially affected each timepoint of the TSST task. As with the Tanner stage variable, the first test investigates a mean effect of age in the response through the duration of the task and the interaction tests for a possible difference in the trajectory of the task. Finally, to study whether cortisol is related to diagnosis, we tested the main effect of diagnosis as well as the TSST timepoint by diagnosis interaction to study differential effects of diagnosis during each timepoint of the task. For each model we assessed the intraclass correlation (ICC). The ICC is the between subject variance over sum of the within and between subject variance.

We performed Wald tests using Type 2 sum of squares, so that main effects terms control for all other main effects, but not interactions, and interaction terms are tested controlling for all other terms in the models. All analyses were repeated using the genital/breast stage (GB) Tanner measure instead of the PH stage (see Supplementary Material). Analyses were performed and reported using R 3.5.3.

3. Results

The total sample included 241 youth. However, 203 youth were included in the analysis controlling for GB stage, while 202 youth were included in the analysis controlling for PH stage due to missing data. There were expected differences between IQ, with the TD group having a median IQ of 117 (IQR= 107, 128) and ASD group having a median IQ of 103 (IQR= 89.62, 118). The ASD group had a median Total ADOS of 12 (IQR= 9, 15). Demographic data are presented in Table 1 for age, sex, IQ, ADOS, medication use, PH stage, GB stage, and log10 cortisol across the TSST.

Table 1.

Sample descriptive statistics. Cortisol values are log10 transformed.

		TD		ASD
		(n = 103)		(n = 138)
	N	Md	IQR	Md	IQR
Age	241	11.67	10.58, 12.65	11.25	10.50, 12.25
Full Scale IQ	240	117.00	107.00, 128.00	103.00	86.92, 118.00
ADOS Total Score	138	--	--	12.00	9.00, 15.0
Cortisol: Arrival	207	0.37	0.19, 0.59	0.38	0.21, 0.55
Cortisol: TSST Baseline	204	0.19	0.06, 0.38	0.23	0.04, 0.37
Cortisol: Prep	207	0.36	0.15, 0.56	0.28	0.07, 0.48
Cortisol: Stressor Speech	205	0.46	0.24, 0.83	0.35	0.09, 0.56
Cortisol: Stressor Math	206	0.38	0.17, 0.79	0.30	0.06, 0.54
Cortisol: Recovery	206	0.29	0.11, 0.60	0.20	0.03, 0.36
	N	Proportion		Proportion
Sex: Female	241	0.447 46/103		0.261 36/138
Taking Medication: Yes	241	0.175 18/103		0.652 90/138
GB development stage	235
Stage 1		0.380 38/100		0.407 55/135
Stage 2		0.350 35/100		0.311 42/135
Stage 3		0.180 18/100		0.133 18/135
Stage 4		0.090 9/100		0.126 17/135
Stage 5		0.000 0/100		0.022 3/135
PH development stage	235
Stage 1		0.485 48/99		0.537 73/136
Stage 2		0.202 20/99		0.147 20/136
Stage 3		0.182 18/99		0.169 23/136
Stage 4		0.111 11/99		0.125 17/136
Stage 5		0.020 2/99		0.022 3/136

Note. ADOS = Autism Diagnostic Observation Scale Total Score; TD = typically developing; ASD = autism spectrum disorder, N = number; Md = median; IQR = interquartile range; IQ = intelligence quotient, TSST = Trier Social Stress Test; Prep = preparation; GB = genital/breast Tanner stage; PH = pubic hair Tanner stage.

We first fit a model with all the main effects, and the age by TSST timepoint and PH stage by TSST timepoint interactions to investigate whether age or pubertal status are associated with differences in cortisol at any stage during the task (see Supplementary Table S1 for parameter estimates). Wald tests using Type 2 sum of squares showed no evidence for a main effect of PH stage or the interaction (p=0.99), suggesting that pubertal development stage does not have an effect on cortisol response controlling for sex, diagnosis, baseline cortisol and age (Table 2).

Table 2.

Type 2 sum of squares ANOVA table to test main effects and interactions for age and PH stage. Results indicated linear main effects of age and an age by TSST timepoint interaction.

Factor	X2	Df	p	Effect Size (S)
Sex	2.7	1	0.098	0.10
Diagnosis	3.6	1	0.059	0.12
PH stage	0.0	1	0.98	0.00
Arrival cortisol	29.5	1	<0.0001	0.39
TSST	166.3	4	<0.0001	0.94
Age	3.1	1	0.078	0.11
PH stage*TSST	0.1	4	0.998	0.00
Age*TSST	10.1	4	0.04	0.18

Note. Cortisol was measured across the five timepoints of the TSST (Baseline, Prep, Stressor 1, Stressor 2, Recovery). Significant p-values in bold. Df = Degrees of freedom.

There was a non-significant main effect of age (p=0.078), but a significant TSST timepoint by age interaction (p=0.0396) when controlling for all main effects and the interaction between puberty stage by timepoint. These results indicated that the age effect was stronger during the stressors of the task relative to the TSST baseline timepoint (Figure 2). Because PH stage and age are expected to be correlated, we tested the main effect and interactions of PH again without including age in the model. This did not substantially change the conclusions of the results for the main effect (χ²=1.6, df=1, p=0.20, S=0.06) or the interaction (χ²=6.03, df=4, p=0.197, S=0.1) (Supplementary Table S2).

Figure 2.

Estimated mean trajectories in log10 cortisol for ages 10–13 controlling for covariates (sex, PH stage, arrival cortisol (S1 Arrival), and diagnosis). Values are estimates in ASD females with the median baseline cortisol averaging across the 5 PH stages.

Because there was little evidence for a PH stage interaction, we did not include it in the model for subsequent analyses. We then investigated whether diagnosis was related to stress response, by separately testing the main effect of ASD diagnosis and the interaction between ASD diagnosis and TSST timepoint (Table 3; see Supplementary Table S3 for parameter estimates). There was no main effect of diagnosis (p=0.059), but a significant diagnosis by TSST timepoint interaction characterized by a blunted cortisol response in ASD to the social evaluation portion of the stressor and a quicker return to baseline cortisol levels (Figure 3). Across the models we compared, the intraclass correlation for the random effect term was approximately 0.72 indicating that the between subject variance accounted for approximately 72% of the variance. In order to test whether these results were related to IQ, we included IQ as a covariate in a supplementary analysis (Table S4). There was no qualitative change in the findings and IQ was not significantly associated with cortisol response. We also considered the possibility that the effect of age on TSST cortisol response differed by diagnosis by testing a 3-way interaction (Table S5). We did not find evidence for a differing effect of age between the two diagnostic groups.

Table 3.

Type 2 sum of squares ANOVA table to test main effects and interactions diagnosis with TSST timepoint. Results indicate a significant diagnosis by TSST interaction indicating that ASD children have similar TSST baseline cortisol to TD children, but a blunted response during the stressors and an earlier return to baseline levels (see Figure 2).

Factor	X2	Df	p	Effect Size (S)
Sex	2.7	1	0.098	0.10
PH stage	0.0	1	0.98	0.00
Arrival Cortisol	29.5	1	<0.0001	0.39
Age	3.1	1	0.079	0.11
TSST	173.0	4	<0.0001	0.96
Diagnosis	3.6	1	0.059	0.12
Age*TSST	10.6	4	0.031	0.19
Diagnosis*TSST	32.2	4	<0.0001	0.39

Note. Significant p-values in bold. Df = Degrees of freedom.

Figure 3A.

ASD children have a blunted cortisol response to the stressors during the TSST and a quicker return to TSST baseline values. Figure 3B. The same result with individual trajectories showing substantial between subject variability in mean cortisol. Triangles denote the estimated mean for each timepoint in TD and ASD after controlling for covariates. Gray lines are the raw data after removing each subjects’ offset for the random effect.

To investigate whether there is a difference based on the type of Tanner stage, we repeated all aforementioned analyses presented using the genital/breast-based (GB) Tanner stages (see Supplementary material and table S6). The analyses using GB Tanner stage produced the same conclusions as with the PH Tanner stage analyses.

4. Discussion

Adolescence and the onset of puberty is characterized by significant changes in biopsychosocial factors to include response to perceived social threat (e.g., Gunnar et al., 2009; Stroud et al., 2009; Sumter et al., 2010). The current study examined the impact of pubertal status, age and diagnosis in a group of early adolescents with ASD and TD. Based on prior research, the following hypotheses pertaining to HPA axis activation to social evaluative threat (TSST) were proposed: 1) Advanced pubertal status will predict higher stress cortisol response in the TD and the ASD groups; 2) Older age will predict higher stress cortisol response in the TD and ASD groups; and 3) TD adolescents will evidence elevated cortisol compared to adolescents with ASD who will show a blunted cortisol profile.

The initial hypothesis related to pubertal development was not confirmed as there was no main effect or interaction for GB or PH stage in response to the social evaluative stressor. Effect sizes were small suggesting that pubertal stage did not have an effect on cortisol response for this early adolescent sample. It is essential to note that the majority of the adolescents in both groups fell within the initial Tanner 1 and 2 stages (73% TD and 72% ASD). Most of the studies showing an impact of puberty on increased HPA axis reactivity examined a much broader age and developmental range (e.g., 9–17 years, (Sumter et al., 2010)). It is likely that as the sample matures over the course of the four-year longitudinal study, pubertal development will have more impact on stress responsivity allowing more definitive conclusions (Gunnar et al., 2009). Previous research supports this notion as higher cortisol responses in TD adolescents have been influenced by pubertal development especially during later stages (Gunnar et al., 2009). While previous research has shown increased activity of the HPA axis coinciding with sexual maturation (Stroud et al., 2009); earlier findings reveals non-linear trajectories and coincide with pubertal scores in the mid-to-upper range contributing to higher stress cortisol (Gunnar et al., 2009). However, a large recent meta-analysis of mostly TD youth reported pubertal status was not a significant moderator of stress response to the TSST although they reported limited power to detect an effect (Seddon et al., 2020).

The scale used to measure puberty may be relevant. The majority of the studies examining pubertal development and stress responsivity used the Pubertal Development Scale (Petersen et al., 1988), a self- or parent-report questionnaire for pubertal stage (e.g., Gunnar et al., 2009; van den Bos and Westenberg, 2015; although see Stroud et al., 2009). In contrast, the current study used Tanner staging based on standardized physical exam which has been shown to be a more gold standard measure of puberty than parent or self-report in youth with and without ASD (Corbett et al., 2019b). Measures of pubertal timing can vary based on the a) degree of objectivity (objective, subjective, or perceived), b) temporal association with development (concurrent or retrospective) and study design (cross-sectional or longitudinal) (Mendle et al., 2019). It is unclear the extent to which the use of different pubertal scales may have contributed to the discrepant findings across studies.

In regards to age, there was a significant main effect providing evidence of an increase in cortisol response across the whole task in older youth confirming the second hypothesis. There was also a significant age by TSST timepoint interaction revealing the age effect was strongest during the social stressor than during anticipation or recovery. The findings are broadly consistent with many studies in TD youth (e.g., Gunnar et al., 2009; Stroud et al., 2009; Sumter et al., 2010) including a recent meta-analysis indicating that age is a strong moderator of cortisol response to social-evaluative threat especially during early adolescence (Seddon et al., 2020). Stroud (2009) also demonstrated that younger TD children (<13 years) show lower cortisol levels than children 13 years and higher, who exhibit a more pronounced HPA response. The age-related changes coincide with the developmental transition of adolescence (e.g., Chrousos et al., 1998; Spear, 2000) and corroborate the proposed age-related stress sensitivity during adolescence (Dahl, 2004).

There are a variety of factors that may contribute to the age-related cortisol response to psychosocial stress, which include the role of peers, increased awareness of social appraisal, and social cognitive development. During adolescence, youth place greater value on peers including increased sensitivity to peer evaluation (Gunnar et al., 2009; Stroud et al., 2009). This likely played a role in the current study, as peers served as raters during the TSST and therefore may have contributed to the higher cortisol levels in the older youth. Developmental changes in social cognition may also have had an impact. During late childhood and adolescence, youth transition from non-recursive thinking to recursive thinking (ability to think about other people’s thoughts and awareness that people think about and evaluate them) also known as theory of mind. It has been hypothesized that the transition of recursive thinking may coincide with enhanced fear of social appraisal by others (Bokhorst et al., 2008). Indeed, it has been shown that the transition to recursive thinking is associated with the timing of cortisol response to social evaluation (Van den Bos et al., 2016). Therefore, the heightened importance of peers, greater awareness of social appraisal and sociocognitive maturation during the adolescent transition are plausible contributors to the age-related increase in cortisol responsivity.

The third hypothesis predicting differences between the groups based on stress response was also confirmed. Specifically, the TD adolescents showed the expected rise in cortisol in response to the TSST, which is found in most youth studies (Seddon et al., 2020). In contrast, the HPA response in youth with ASD was characterized by a blunted cortisol response to the speech and math task and faster return to basal cortisol levels. Previous studies have reported diminished cortisol response to the TSST in children (Corbett et al., 2012; Lanni et al., 2012; Levine et al., 2012) and adolescents (Edmiston et al., 2016). The current study demonstrates definitive and significant between-group differences for the cortisol stress response, especially during the middle timepoints of the task which reflect the social stress portion of the protocol.

In order to activate the HPA axis, the stressor must be perceived to be threatening (Evans et al., 2013). Current and previous studies suggest that children with ASD do not perceive social evaluative threat during the TSST to be stressful (Corbett et al., 2012; Lanni et al., 2012; Levine et al., 2012). The TSST requires the raters (judges) to show neutral, unsupportive affect during the public speech which most people perceive as threatening. To better understand why many children and adolescents with ASD evidence an atypical stress response to the TSST, the influence of social cognition was investigated because many individuals with ASD demonstrate impairment in facial affect recognition, including to neutral faces (Berggren et al., 2016; Jarvinen et al., 2015; Tottenham et al., 2014). Corbett and colleagues examined the extent to which the perception of neutral faces mediates the stress response to the TSST. Findings revealed that the perception of threat in the form of neutral faces mediates the stress response to the TSST in TD adolescents and youth with ASD (Corbett et al., 2019a). In other words, individuals better able to perceive emotions of others exhibit a stronger cortisol response to social evaluative threat (Bechtoldt and Schneider, 2016). Similarly, recursive thinking has been associated with cortisol response to social evaluation (Van den Bos et al., 2016). Importantly, individuals with ASD often demonstrate diminished sociocognitive skills, including theory of mind (Baron-Cohen et al., 1997; Kleinman et al., 2001). Therefore, the lack of awareness of the appraisal by others may impact the blunted cortisol response observed in the ASD children and adolescents. It may also be the case that comorbid conditions, such as anxiety, contribute to differences in physiological response to psychosocial stress in ASD (Hollocks et al., 2014).

It is important to acknowledge that notable variability in the stress response was observed especially in the ASD group, which is a frequent finding in ASD samples (Corbett et al., 2010; Schupp et al., 2013). Previous research has highlighted age as an important factor in determining the strength of the stress response to social stressors such that older children with ASD express elevated cortisol compared to younger children with ASD and same age TD children (Corbett et al., 2010). The few studies in adolescents have also suggested a diminished cortisol response to the TSST (Edmiston et al., 2017; Hollocks et al., 2014). However, adults with ASD appear to demonstrate a more normative increase in cortisol in response to social evaluation when compared to adolescents with ASD (Taylor et al., 2018) suggesting a developmental lag in social awareness, physiological response or both. In other words, it is plausible that as youth with ASD mature and progress through adolescence into adulthood they may acquire more normative sociocognitive skills, such as executive functioning and theory of mind, shown to impact cortisol response to social evaluative threat (Van den Bos et al., 2016).

5. Limitations, Future Directions and Conclusions

The study includes a rigorously characterized sample of youth with ASD and TD adolescents, inclusion of standardized physical exam for more gold standard assessment of pubertal development, use of established social stress protocol and collection methods, and a robust statistical approach. However, limitations involved the lack of enrollment of youth with ASD with accompanying intellectual impairment and a predominantly Caucasian sample, which is not representative of the broader autism or diverse racial and ethnic community. Finally, the current study included a narrow range of early adolescents limiting the interpretation of the impact of pubertal development on HPA axis activation. The parent longitudinal study (Corbett, 2017), which follows the sample over four years will be positioned to more fully examine the puberty-HPA stress hypothesis (Gunnar et al., 2009). Additionally, other factors, such as the presence of anxiety (Gunnar et al., 2009) or differences based on sex (Kudielka and Kirschbaum, 2005) that can play a role in stress responsivity during pubertal development will to be explored in next steps.

In conclusion, the current study replicated and extended research showing increased stress sensitivity arising during adolescence for TD and ASD youth which is driven by age rather than puberty at least in early adolescents. The impact of pubertal effects on physiological stress as measured by cortisol may increase as the sample advances through stages of sexual maturation. Due to the atypical trajectories, variability and magnitude of cortisol in youth compared to adults with ASD, there may be a developmental lag in the perception of or response to social stress in autism that warrants careful investigation using a longitudinal design.

Highlights.

Cortisol is elevated during social evaluation which increases during adolescence

Age not pubertal stage contributed to higher cortisol during early adolescence

Early adolescents with autism show blunted cortisol response compared to typical youth

There may be a developmental lag in stress response to social evaluation in autism