Social Context in Stress and Autism: Comparing Physiological Profiles Across Two Social Paradigms in Youth with and without Autism Spectrum Disorder

Abstract

Background:

The social world is often stressful for individuals with autism spectrum disorder (ASD). Research shows youth with ASD demonstrate physiological hyperreactivity to some social stressors (e.g., interaction) but not others (e.g., evaluation); therefore, this study examined diagnosis (ASD or typical development (TD)), social context, perceived anxiety, and physiological responsivity across multiple stress systems; namely, the hypothalamic pituitary adrenal (HPA) axis and autonomic nervous system (ANS).

Method:

This study examined 244 ten-to-thirteen-year-olds with ASD (N = 140) or TD (N = 104). Physiological responses, measured by salivary cortisol, heart rate (HR), and respiratory sinus arrhythmia (RSA), were assessed before and after a social evaluative threat paradigm (Trier Social Stress Test; TSST) and social interaction (Trier Social Stress Test- Friendly; TSST-F). Mediation models examined the relationships between anxiety, diagnosis, and physiology.

Results:

Significant three-way interactions were observed for cortisol (p=0.007) and HR (p=0.002), suggesting diagnostic groups respond differently across context and time points. There was no significant interaction for RSA (p=0.149), although ASD youth had significantly lower RSA overall (p=0.038). State and trait anxiety did not mediate the relationship between diagnosis and physiology (all p>0.05).

Conclusions:

Findings emphasize the critical role of context and a multisystem approach in examination of physiological social stress in youth with ASD. Results provide a foundation to elucidate unique response patterns across physiological systems to more precisely identify those with heightened physiological arousal across social contexts. It is proposed that future identification of subtypes may ultimately inform approaches for enhancing social engagement.

Introduction

Stress is a subjective experience and can be difficult to define clearly and objectively (Levine & Ursin, 1991). One way in which psychological stress is conceptualized is as a physiological response to an unpleasant or threatening stimulus, often occurring when demands are perceived as too demanding for the individual (Cohen et al., 2007). Importantly, the physiological stress response can be adaptive, as it diverts attention to respond to or remove the source of the threat; however, prolonged, chronic, or severely acute physiological arousal can have significant consequences for physical (e.g., (Juster et al., 2010)) and psychological (e.g., (McEwen, 2003)) health. Individual differences also significantly influence one’s response to stress, including one’s appraisal of the situation or the severity of the stressor (Lacey & Lacey, 1958; Schneiderman et al., 2005). Given the potentially negative health consequences related to chronic or strong acute stress exposure, it is important to identify individual- and group-based factors influencing differential physiological response to everyday stressors in an effort to understand biological and psychological signs of risk and resilience.

The primary neuroendocrine stress system, the hypothalamic-pituitary-adrenal (HPA) axis, secretes cortisol – the main glucocorticoid in humans – from the adrenal glands in response to stress or to maintain homeostasis via a diurnal rhythm or negative feedback mechanism. Stress response involving the HPA axis is often conceptualized as either response to a physiological threat to homeostasis (‘systemic’ stressors) or as a response to perceived threat (‘processive’ stressors’) (Herman & Cullinan, 1997). Psychological or perceived stress response involves significant cognitive appraisal of context, emotion, and memory to assess a potentially threatening situation. This appraisal requires top-down modulation from several limbic regions, such as the prefrontal cortex, amygdala, and hippocampus. Indeed, the neural control of stress is a multifaceted process by which the HPA axis neuroendocrine cascade is modulated by several factors influencing the perception of threat, to include novelty/threat (amygdala), contextual cues (prefrontal cortex), and past experience (hippocampus) (Corbett et al., 2012; Herman et al., 2005; Van de Kar & Blair, 1999).

The Autonomic Nervous System (ANS) is responsible for regulating involuntary bodily processes (e.g., heart rate, blood pressure, respiration, digestion, etc.), and thus it’s function and reactivity plays an important role in regulating the body’s physiological response to stress. The ANS is separated into two branches with primarily opposing functions, the parasympathetic (PNS; ‘rest and relax’ branch) and sympathetic nervous systems (SNS; ‘fight or flight’ branch). Much like the HPA axis, functioning of the ANS is modulated by an extensive neural network, beginning with the prefrontal cortex at the top of the hierarchy and extending down into other forebrain regions such as the amygdala and hippocampus (Benarroch, 2012). As such, the ANS is often considered a ‘behavioral regulator’ in which balance and flexibility of the two branches significantly shapes one’s responses to environmental changes (Berntson et al., 2008), while an inflexible system may be associated with several psychopathologies (Beauchaine, 2015; Friedman, 2007). One oft used marker of ANS response to physical or perceived stress is cardiovascular functioning. Specifically, the sinoatrial node (SA), referred to as the pacemaker of the heart, is dually innervated by both branches of the ANS (Hamill et al., 2012), thus changes in heart rate (HR) or the beat-to-beat variability of heart rate (heart rate variability; HRV), may provide insight into the activity of the PNS and SNS systems.

The HPA axis and the ANS are two independent physiological stress systems that provide unique yet complementary mechanisms by which to examine and understand stress. These systems are also highly coordinated and interconnected with overlapping neural circuitry. As described above, both systems include top-down regulation from limbic structures, such as the prefrontal cortex and amygdala (e.g., (Ulrich-Lai & Herman, 2009)). As a result, any perceived threat is likely to trigger a coordinated response in the HPA axis and the ANS, where changes in the HPA axis and sympathetic nervous system largely promote excitatory, activational responses, whereas the parasympathetic system tends to promote a more calm, visceral state. Thus, consideration of both systems when examining physiological stress may provide greater insight into the overall impact of typical and atypical physiology in influencing psychosocioemotional behavior. Emphasizing the coordinated functioning of the HPA axis and ANS could have noteworthy psychological implications in cases of chronic stress or disruption within the shared neural circuitry. For example, an additive model proposes that reciprocal arousal between systems is optimal, while excessively low or high overall arousal will contribute to psychiatric risks (e.g., internalizing or externalizing symptoms). In contrast, the Interactive model (Bauer et al., 2002) emphasizes a balanced system in which response of one system mirrors the other, thereby asymmetrical activation (e.g., HPA hyperarousal, SNS hypoarousal and/or PNS hyperarousal) is associated with increased psychopathologies (e.g., internalizing disorders). Despite some investigations demonstrating that youth with hyperarousal of both the HPA axis and the ANS had the highest rates of perceived stress and internalizing or externalizing behaviors (El-Sheikh et al., 2011; El-Sheikh et al., 2008; Rotenberg & McGrath, 2016), there is relatively limited evidence to inform the impact of the combined physiological stress response on psychological states and behavior, particularly in those without typical development (TD).

Autism spectrum disorder (ASD)¹ is a neurodevelopment disorder defined by challenges in social communication and interaction, as well as restricted, repetitive patterns of behavior and interests (APA, 2013). For many individuals with ASD, reciprocal social exchanges and interactions may be difficult, and these social experiences often induce significant psychological and physiological stress (Corbett et al., 2010; Corbett et al., 2014; Lopata et al., 2008; Schupp et al., 2013). In several studies of school-aged youth, children with ASD had higher cortisol reactivity relative to TD peers while interacting with others in a naturalistic playground paradigm (Corbett et al., 2010; Corbett et al., 2014; Schupp et al., 2013). HPA axis activation has also been observed in youth with ASD during engagement with unfamiliar children (Lopata et al., 2008) or school activities with same-aged peers (Richdale & Prior, 1992). In the ANS, hyperarousal in the form of reduced parasympathetic regulation has been observed in youth with ASD at rest (e.g., (Bal et al., 2010; Edmiston et al., 2016; Guy et al., 2014; Muscatello, Kim, et al., 2021; Muscatello et al., 2022; Neuhaus et al., 2016; Vaughan Van Hecke et al., 2009). Similarly, reduced parasympathetic responsivity has been observed in autistic youth following interaction with unfamiliar peers (e.g., (Muscatello, Vandekar, et al., 2021; Neuhaus et al., 2016; Vaughan Van Hecke et al., 2009). Additionally, other ANS differences have been reported, such as increased sympathetic reactivity to social (Edmiston, Muscatello, et al., 2017) and non-social (Lopata et al., 2008; Schaaf et al., 2015) stress, as well as elevated basal heart rate (Kushki et al., 2014) in ASD. However, it is also important to acknowledge mixed findings, particularly regarding autonomic responsivity, where some have found no differences between ASD and TD youth in cardiovascular reactivity to social stressors despite observed differences at rest (regulation) (Levine et al., 2012; Lory et al., 2020; Muscatello, Kim, et al., 2021).

Despite numerous studies reporting differences in the HPA axis (see (Taylor & Corbett, 2014) for review) or ANS (see (Benevides & Lane, 2015; Cheng et al., 2020) for review) in autism, very few have examined the unique interaction between these related physiological systems during resting-state or in response to stressors. Of those that have examined these interactions in autism (Corbett et al., 2019; Muscatello et al., 2020; Thoen et al., 2023), findings have thus far been mixed. In a preliminary examination of cortisol and HRV associations, youth with ASD showed a balanced, reciprocal system at rest by which increases in HPA axis regulation were associated with reduced parasympathetic activity (Corbett et al., 2019). Moreover, cardiovascular autonomic balance in which parasympathetic regulation is high while sympathetic activity is low has been reported to be particularly associated with fewer depressive symptoms in autistic youth (Muscatello et al., 2020). Finally, a recent study reported on autonomic and HPA axis regulation and reactivity at rest and following cognitive and social stress tasks in youth with and without ASD (Thoen et al., 2023). Researchers reported that while vagal (parasympathetic) regulation was reduced in ASD, there were no significant correlations between resting-state HPA and vagal modulation in youth with or without ASD (Thoen et al., 2023). Clearly, comprehensive examinations of multiple stress regulatory systems, and the ways in which they interact, remain an important area of investigation to more thoroughly elucidate the role of physiological regulation and reactivity in autism.

While social interactions or similar social experiences appear to be stress-inducing for many individuals with autism (e.g., (Corbett et al., 2014; Lopata et al., 2008; Schupp et al., 2013)), there are other social contexts in which these individuals demonstrate less reactivity or a more blunted response than would be expected. Social evaluation is considered one of the four primary inducers of HPA axis reactivity in humans (Dickerson & Kemeny, 2004), and evaluative threat protocols such as the Trier Social Stress Test (Kirschbaum et al., 1993) reliably induce cortisol and autonomic responses in adults and children (e.g., (Buske-Kirschbaum et al., 1997; Kirschbaum et al., 1995; Kirschbaum et al., 1993; Kudielka et al., 2004; Seddon et al., 2020)). However, the TSST does not provoke this expected stress response in many ASD individuals, whereas TD peers show adaptive physiological arousal (Corbett et al., 2019; Corbett et al., 2012; Edmiston, Blain, et al., 2017; Jansen et al., 2006; Lanni et al., 2012; Levine et al., 2012). Considering the role of context, two studies from our group have examined physiological responses when participants engaged in both a social interaction and social evaluation paradigm (Corbett et al., 2019; Corbett et al., 2012). In one study utilizing the TSST and the Peer Interaction Playground (PIP) paradigm (Corbett et al., 2010), cortisol response patterns followed a double-dissociation in which youth with ASD had increased HPA reactivity during social interaction and minimal reactivity to the TSST, whereas the opposite pattern held true for the TD group (Corbett et al., 2012). Another recent study in early adolescents showed a similarly blunted physiological response during the TSST; however, this study did not observe any differences during a more benign social interaction protocol (face-to-face conversation with same age peer) (Corbett et al., 2019). While both studies provided important preliminary information regarding the role of context in social stress for youth with ASD, they were limited by relatively small samples and primary focus on HPA axis reactivity with only preliminary exploration of autonomic response profiles (Corbett et al., 2019). Furthermore, the study team previously reported diagnostic differences to social evaluation in cortisol (Corbett et al., 2021) and RSA (Muscatello, Kim, et al., 2021). The current study therefore seeks to build on this work to examine the role of social context more comprehensively as it relates to physiological responsivity across multiple stress systems; specifically, the ANS and HPA axis in youth with ASD or TD. It is hypothesized that youth with ASD will show elevated cortisol, decreased heart rate variability, and increased heart rate during a social interaction task, while the opposite pattern will hold for a social evaluation paradigm. Furthermore, the study will investigate differences in psychological experiences (e.g., state anxiety) across contexts and diagnosis, as well as the relationship with physiology (ANS and HPA). We predict self-reported anxiety will mediate the relationship between task/context (interaction vs. evaluation) and physiology as well as between diagnosis (ASD vs. TD) and physiology.

Methods

Participants

The study included 244 youth with ASD (N = 140) or TD (N = 104) who were enrolled in a large, longitudinal study of pubertal development in autism (Corbett, 2017). The sample was based on power analyses performed for the source longitudinal study on pubertal development. Specifically, a priori power analyses determined that for longitudinal analyses including four groups (ASD vs. TD and Male vs. Female), a sample size of N = 60 per group (120 ASD males + females and 120 TD males + females) was needed to achieve a medium effect size (d = 0.51) based on Cohen’s d (Cohen, 1988) at 80% power. Data collected from Year 1 of the longitudinal study are included in the current analyses. Participants were enrolled between the ages of 10.0 to 13.9 years, and the sex distribution was 104 males/36 females with ASD and 59 males/46 females in the TD group. The ratio of males to females in the ASD group was approximately 3:1, which is equivalent to current sex ratio estimates in ASD (Loomes et al., 2017). Full demographic data, including age, race, ethnicity, and relevant diagnostics, is available in Table 1. Of note, in-depth analysis of cortisol (Corbett et al., 2021) and heart rate variability (Muscatello, Kim, et al., 2021) specifically during social evaluation were previously reported for this sample and are acknowledged accordingly in the discussion.

Table 1.

Descriptive and Demographic Statistics Stratified by Diagnosis

	N	TD	ASD	Overall	Test Statistic
		(N=104)	(N=140)	(N=244)
Age	244	10.58 11.62 12.63	10.50 11.25 12.25	10.58 11.33 12.33	F1,242=2.713
Sex	244				χ2 (1)1=9.172**
Male		0.56 58/104	0.74 104/140	0.66 162/244
Female		0.44 46/104	0.26 36/140	0.34 82/244
Ethnicity	244				χ2(1)=0.912
Not Hispanic		0.95 99/104	0.92 129/140	0.93 228/244
Hispanic		0.05 5/104	0.08 11/140	0.07 16/244
Race	244				χ2(3)=12.062**
Caucasian		0.86 89/104	0.81 114/140	0.83 203/244
African American		0.02 2/104	0.12 17/140	0.08 19/244
American Indian		0.00 0/104	0.00 0/140	0.00 0/244
Asian/Pacific Islander		0.00 0/104	0.01 1/140	0.00 1/244
Biracial		0.12 13/104	0.06 8/140	0.09 21/244
BMI (Percentile)	239	30.00 53.00 87.67	39.00 69.00 96.00	34.00 63.00 93.00	F1,237=6.123*
ADOS Total	140	---	9.00 12.00 15.00	---
Full Scale IQ Score	243	107.00 117.00 127.83	86.42 103.00 117.58	96.00 110.00 122.00	F1,241=41.253***
SCQ Total Score	242	1.00 2.00 4.00	10.00 17.50 24.00	3.00 8.00 19.00	F1,240=485.703***
STAIC Total (State, Friendly)	214	26.00 29.00 31.00	26.00 30.00 33.00	26.00 29.50 32.00	F1,212=1.893
STAIC Total (State, Stressful)	213	30.00 33.00 38.00	28.00 32.00 37.00	29.00 32.00 37.33	F1,211=2.143
STAIC Total (Trait, Friendly)	214	26.00 31.00 35.00	27.92 34.50 40.00	26.00 33.00 38.00	F1,212=11.283**
STAIC Total (Trait, Stressful)	213	24.42 29.00 35.00	26.67 34.00 40.33	25.00 32.00 38.33	F1,211=9.243**

Note: N is the number of non-missing value.

¹Kruskal-Wallis.

²Pearson.

³Wilcoxon.

Q1 Median Q3

^*p<0.05;

^**p<0.01;

^***p<0.001

Participants were recruited from a broad community sample in the southern United States covering a 200-mile radius that targeted medical and health-related services, clinics, research registries, regional disability organizations, schools, and social media platforms. All included participants were required to have an intelligence quotient (IQ) score ≥ 70 due to task demands in the source longitudinal study. Participants in the ASD group had a confirmed diagnosis according to Diagnostic and Statistical Manual-5 criteria (APA, 2013), which were 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 (ADOS-2; (Lord et al., 2012)).

Considering nearly half of all youth with ASD are prescribed at least one psychotropic medication (Mire et al., 2014), the study did not require participants to be medication-naïve. However, because the current and source study examined hormonal responses, youth were excluded if taking medications that alter the Hypothalamic-Pituitary-Adrenal (HPA) axis (e.g., corticosteroids; see (Granger et al., 2009)) or HPG axis (e.g., growth hormone), or if they were diagnosed with a medical condition known to impact pubertal development (e.g., Cushing’s Disease). At enrollment, 49.6% of youth in the ASD group were taking at least one psychotropic medication compared to 9.8% in the TD group. Psychotropic medications included stimulants, selective-serotonin reuptake inhibitors, and central alpha-agonists. To account for diagnostic differences in psychotropic medication use, this was included as a covariate of interest in all models. As reported below, the main effect for medication was not significant in any of the resulting models (see Results).

Procedures

The research was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) and was approved by the Vanderbilt Institutional Review Board. Informed written consent and verbal assent was obtained from parents and youth, respectively, prior to inclusion and initiation of study protocols.

Diagnostic Procedures

The Autism Diagnostic Observation Schedule – Second Edition (ADOS-2; (Lord et al., 2012)) is a semi-structured play- and interview-based protocol to corroborate autism diagnosis. All participants with confirmed or suspected ASD were administered the Module 3 (fluent speech) and required a score of 7 or greater to be included in the study. The ADOS was administered by research reliable personnel. A licensed clinical psychologist (BAC) completed diagnostic measures for participants seeking a first-time diagnosis of autism.

The Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II; (Wechsler, 2011)) is a brief measure of cognitive ability designed to obtain a quick and reliable estimate of a child’s intellectual functioning. To be included in the study, participants could not have a co-occurring diagnosis of intellectual disability (IQ ≥ 70 required).

The Social Communication Questionnaire (SCQ; (Rutter et al., 2003)) is a parent-report questionnaire used to screen for symptoms of ASD. A score of 15 or more is suggestive of ASD, and all TD participants in the study had a total score of less than 10.

Social Stress Paradigms

Following diagnostic assessment, participants returned to the lab to complete two social paradigms – a friendly social interaction and a social evaluative threat task (described below). Of the 245 youth initially enrolled, 213 returned to the lab to complete the social stress paradigms. Twenty-nine participants did not return for the social stressors, while four participants were deemed ineligible following diagnostic procedures. There was no group difference in number of youths with ASD (n=17) or TD (n=12) who did not return for the social tasks (𝜒²(1)=0.02, p=0.87). Social task procedures were conducted in the afternoons to control for variations in diurnal cortisol rhythm. The order of the interaction and evaluation was consistent across participants, with all completing the friendly interaction first followed by the social evaluation. All participants completed both social paradigms in one visit to the lab.

The Trier Social Stress Test – Friendly (TSST-F; (Wiemers et al., 2013)) is an alternative form of the original Trier Social Stress Test (TSST; (Kirschbaum et al., 1993)) (described below). Unlike the original TSST, the TSST-F consists of a more ‘friendly’ paradigm in which participants are asked to describe their interests and hobbies with a novel, similar-age peer of the same sex who shows encouragement (smiles, nods, shows interest, maintains eye contact) and asks follow-up questions. This protocol does not result in measurable physiological response in typically developing individuals (Wiemers et al., 2013; Wiemers & Wolf, 2015). However, the TSST-F is conceptualized to parallel a play-based peer interaction paradigm (PIP; (Corbett et al., 2010), and therefore may be a more potent stressor for children with ASD. The 20-minute paradigm includes: 5-minute resting baseline, 5-minute preparation, 10-minute social interaction, and 5-minute debrief and recovery. To be consistent with the TSST (described below), we divided the social interaction into two, 5-minute segments in all statistical models. Therefore, the five time points of interest for the TSST-F were Baseline, Preparation, Interaction Part 1/Part 2 (abbreviated Interaction 1/2) and Recovery.

The Trier Social Stress Test (TSST; (Kirschbaum et al., 1993)) is a well-validated psychosocial stressor consistently shown to activate a physiological stress response in the adults and children (Buske-Kirschbaum et al., 1997; Kirschbaum et al., 1995; Kirschbaum et al., 1993; Kudielka et al., 2004; Seddon et al., 2020). The current study included the TSST-Child Version (TSST-C; (Buske-Kirschbaum et al., 1997)), which was developed for use in children and adolescents. In the TSST-C, 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 since this contributes to a stronger effect than using female judges only (Seddon et al., 2020). The 20-min task is divided into four subcomponents: 1) Preparation (5-minutes), 2) Speech (5-minutes), 3) Serial Subtraction (5-minutes), and 4) Debriefing/Recovery (5-minutes). In all models, the five time points of interest for the TSST were Baseline, Preparation, Speech, Math, and Recovery.

Peers and judges were thoroughly trained using a manualized protocol, review of videotaped TSST-F or TSST-C sessions, and practice with senior research personnel prior to working with a participant. Each administration was recorded for behavioral coding purposes, and videos were routinely checked to ensure 1) peers in the TSST-F maintained social interest without talking for >50% of the conversation, and 2) judges maintained neutral affect and did not deviate from scripted instructions. If deviations in the protocol were noted, booster training sessions were promptly provided. There was no overlap between peers/judges within a participant; therefore, a peer in the TSST-F could not also be a judge in the TSST.

Dependent Measures

Cortisol

Cortisol can be reliably and non-invasively measured through small amounts of saliva, which makes it ideal in studies of children and adolescents (Kirschbaum & Hellhammer, 1994). Samples were collected using established passive-drool procedures (Corbett et al., 2006; Corbett et al., 2008). Two children with ASD were unable to use the passive drool method due to oral motor difficulties and sensory aversions; therefore, an alternative method using dental cotton roll and syringes were used for these two participants (Tordjman et al., 2014). Previous studies reporting cortisol collected at-home in this sample showed results did not meaningfully change when excluding the two participants (Corbett et al., 2023), and therefore the data was included in the current study. Participants were instructed to avoid food and drink at-least one hour prior to the study procedures, and no food/drink other than water was offered throughout the duration of the visit. For all participants, protocols, including saliva sampling, began between the hours of 2pm – 3pm central time. Restricting protocols to mid-afternoon allowed for control of cortisol diurnal rhythm, consistent with expert recommendations (Clements, 2013; Kudielka et al., 2009), while the narrow window for start times assured that time-of-day of individual samples varied minimally between individuals. Furthermore, baseline cortisol prior to the stressor protocol was included as a covariate in all cortisol models. A total of ten samples were collected throughout the visit in 10- or 20-minute increments and timed to reflect cortisol at Baseline, Preparation, Interaction 1/Speech, Interaction 2/Math, and Recovery while accounting for the 20-minute lag in detection in saliva. A timeline of procedures is provided in Figure 1 (TSST-F) and Figure 2 (TSST).

Figure 1. Timeline of Cortisol Sampling and RSA Collection for TSST-F.

The dashed arrow represents the 20-minute lag in salivary cortisol detection.

Figure 2. Timeline of Cortisol Sampling and RSA Collection for TSST.

The dashed arrow represents the 20-minute lag in salivary cortisol detection.

Cortisol assays were performed using a Coat-A-Count^® radioimmunoassay kit (Siemens Medical Solutions Diagnostics, Los Angeles, CA) modified to accommodate lower levels of cortisol in human saliva. Samples stored at −80°C, were thawed and centrifuged at 3460 rpm for 15 minutes to separate the aqueous component from mucins and other suspended particles. The coated tube from the kit was substituted with a glass tube into which 100 ul of saliva, 100 ul of cortisol antibody (courtesy of Wendell Nicholson, Vanderbilt University, Nashville, TN), and 100 ul of ¹²⁵I-cortisol were mixed. After incubation at 4°C for 24 hours 100 ul 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 minutes, decanted, and counted (for details see (Corbett et al., 2014)). Serial dilution of samples indicated a linearity of 0.99. The intra-assay coefficient of variation was CV = 2.06%.

Heart Rate and Heart Rate Variability

Heart rate was collected using MindWare Mobile Impedance Cardiograph units (MindWare Technologies LTD, Gahanna, OH) for synchronized electrocardiography (ECG) and respiration data collection using a seven-electrode configuration. Prior to electrode placement, participants were told they would be wearing ‘stickers’ that measure their heart rate. They were offered the opportunity to place an electrode on their hand to acclimate to the sensory aspects of the stickers as well as having a 5-minute acclimation period following electrode placement. In total, four participants with ASD refused electrode placement because of significant sensory sensitivities.

ECG collection began with a five-minute baseline resting period. Participants were instructed to sit quietly without engaging in other tasks. Following the baseline period, ECG was continuously monitored throughout the TSST-F and TSST protocols (see Figures 1 and 2). ECG signal was sampled at 500 Hz and analyzed using the Heart Rate Variability Software Suite provided by MindWare Technologies (MindWare Technologies LTD, Gahanna, OH). Heart rate was calculated to provide an index of overall autonomic arousal. Additionally, a metric of the variation in beat-to-beat signal (heart rate variability; HRV), which can provide additional insight into autonomic influence on the heart, was calculated. Specifically, respiratory sinus arrhythmia (RSA) is an index of high-frequency heartbeat variability that fluctuates with respiration and is a commonly used metric for examining parasympathetic influence on the heart. RSA was derived in accordance with guidelines set forth by the Society for Psychophysiological Research committee on heart rate variability (Force, 1996) and quantified as the integral power within the respiratory frequency band (0.12 to 0.40 Hz). Based upon recommendations by Laborde and colleagues (2017), respiratory frequency was confirmed to lie within the high frequency/RSA band (0.12–0.40 Hz) for all participants and did not differ between diagnostic groups. Heart rate (HR) and RSA were calculated in one-minute epochs and averaged across 5-minute periods for baseline, preparation, social interaction part 1/speech, social interaction part 2/math, and recovery. Of the total available data, 4.2% were excluded due to excessive motion artifact, equipment error, or cardiac arrhythmias. HR was measured in beats per minute (bpm) and RSA in ms².

Perceived Anxiety

The State-Trait Anxiety Inventory for Children (STAIC; (Spielberger, 1973)) is a self-report measure of anxiety, completed by participants, in which an individual describes how he/she is currently feeling (state) and how he/she usually feels (trait). The STAIC was completed twice, once immediately following the TSST-F and again after the TSST-S to assess perceived anxiety after the two stress paradigms. To keep protocols consistent across both stressors, and to account for previous research showing that autistic youth consistently report elevated trait, but not state, anxiety following tasks that invoked a stress response (Mertens et al., 2017; Simon & Corbett, 2013), participants completed both the state and trait measures following each stressor. Previous studies have consistently shown that youth with ASD and without intellectual disability, are able to reliably self-report anxiety (Corbett et al., 2017; Lanni et al., 2012; Simon & Corbett, 2013), and previously-reported internal consistency has been good (α = 0.90 – 0.91) (Corbett et al., 2019).

Statistical Analysis

Characteristics of participants were stratified by diagnosis and summarized in Table 1, using proportions & frequencies for categorical measures, and the 1st, 2nd (median), and 3rd quartiles for numerical measures. Additionally for categorical variables, Pearson’s χ² tests of homogeneity were conducted to identify whether the frequency distributions differed between the two groups (TD vs. ASD). Wilcoxon rank-sum tests were conducted for continuous variables to test whether the probability distributions differed by diagnosis. The null hypothesis in each test purported that there was no difference in distributions between the two groups.

We also investigated the relationship between the physiological response measures by calculating within subject correlations between each pair of the responses across stressor type and time points. We then tested whether these correlations differed by diagnostic groups using Wilcoxon rank-sum tests.

To investigate the dynamics of physiological response between diagnostic groups and stress paradigms we fit linear mixed effects models separately for each of the physiological response measures: cortisol, heart rate, and RSA. Cortisol values were positively skewed and were log10 transformed prior to statistical analyses. We modelled the physiological measures including a task subcomponent indicator (time) by diagnosis by stress paradigm interaction, controlling for baseline physiological response, sex, whether the subject was on psychotropic medication, and age. Age was allowed to have a nonlinear effect via natural cubic splines with 2 degrees of freedom. We also considered a diagnosis by sex interaction, but it was not shown to be significant in any of the models and did not change conclusions. To account for within subject correlation, we included a random intercept by subject. We implemented heteroskedasticity consistent standard errors and tested using type 2 sum of squares analysis of variance. After evaluation of the significance of the three-way interaction, we conducted pairwise contrasts between diagnosis and stress paradigm separately at each task subcomponent to investigate how the diagnostic groups differed in terms of their physiological response during stressors at each task subcomponent time. In testing the pairwise contrasts, we used Holm’s p-value adjustment for the eight multiple comparisons.

To evaluate the relationship between anxiety and the physiological responses, we used similar models used for the diagnostic groups differences but added an additional term adjusting for anxiety (state or trait). Finally, we tested the differences in psychological responses with another mixed effects model on the STAIC anxiety scores (state and trait separately). Similarly, to the physiological response models, we included a diagnosis by stress paradigm interaction, while controlling for sex, whether the subject was on psychotropic medication, and age, again allowed to have a nonlinear effect using natural cubic splines. We then compared the diagnostic groups for each stressor with pairwise contrasts.

In task subcomponents where there were physiological response differences between diagnostic groups, we investigated whether state or trait anxiety mediated the diagnosis by context/task effect, separately. Each mediation analyses separately tested whether (state or trait) anxiety mediated the effect of diagnosis separately for the TSST and TSST-F tasks.

All mixed effects models were fit using the lme4 package, and mediation analyses were done using the mediate package in R version 4.2.1.

Results

Aim 1. Diagnosis, Social Context, and Physiological Response

Cortisol

There was a significant three-way interaction between diagnostic group, stressor type, and task time points on cortisol response during the social stressors (χ²=12.07, df=3, p=0.0072; Table S1), To investigate what task timepoints were driving the effect we tested the diagnosis difference at each time point of each task and found evidence that those with ASD show blunted cortisol response during social evaluative tasks (TSST) during the Speech and Math with uncorrected p-values of 0.0031 and 0.0138, respectively (Table 2). TD youth are predicted to have 41.74% and 33.51% higher cortisol levels at Speech and Math respectively, during the social evaluative task (TSST) than an otherwise identical youth with ASD (Figure 3B). No significant differences between the diagnostic groups were observed during the social interaction tasks (TSST-F) at any time point (Table 2; Figure 3A). We further examined the effect of context (task type) at each time point within diagnostic groups, and pairwise comparisons show ASD youth did not have significantly different cortisol in the TSST versus TSST-F at any timepoint (Table S2). In terms of social evaluative threat, TD youth have higher cortisol levels during the Speech and Math portions of the task, while ASD youth demonstrated a blunted cortisol stress response to social evaluation. In contrast, there were no diagnostic differences during the social interaction (TSST-F), suggesting that ASD youth did not have a significantly higher physiological stress response to social interaction with a novel peer.

Table 2.

Cortisol Pairwise Contrasts Between Diagnosis for Different Stressors at all Time Points.

Social Task	Task Time	Estimate	SE	df	t Ratio	p	Holm p
Friendly/TSST-F (Social Interaction)	Preparation	0.0046	0.0326	420	0.1422	0.8870	1.0000
	Interaction 1	0.0118	0.0353	420	0.3326	0.7396	1.0000
	Interaction 2	−0.0089	0.0322	422	−0.2767	0.7821	1.0000
	Recovery	−0.0455	0.0313	426	−1.4545	0.1466	0.7328
Stressful/TSST (Social Evaluative)1	Preparation	0.0484	0.0389	423	1.2441	0.2142	0.8566
	Speech	0.1515	0.0508	423	2.9795	0.0031	0.0244
	Math	0.1255	0.0508	425	2.4716	0.0138	0.0969
	Recovery	0.088	0.0454	425	1.938	0.0533	0.3197

Note: Contrasts (TD – ASD)

¹Cortisol data from the included sample are also reported in (Corbett et al., 2021).

Figure 3. Average log10 Cortisol by Diagnosis and Timepoint for TSST-F (3A) and TSST (3B).

Note: Data presented in 3B overlaps in part with data previously published as part of a larger study on developmental effects of cortisol response to the TSST (Corbett et al., 2021).

Heart Rate

The same mixed model framework was used to investigate differences in heart rate response during each task. There was a significant three-way interaction between diagnostic group, stressor type, and task time point indicating that the groups respond differently across the context and time points (χ²=15.01, df=3, p=0.0018; Table S3). Differences between diagnostic groups were investigated at each timepoint of each task to identify where the heart rate response differed. Individuals with ASD show elevated heart rate during social interaction tasks (TSST-F) during the Recovery timepoint (p=0.0068; Table 3). On average, youth with ASD are predicted to have a 3 bpm higher heart rate than otherwise identical TD individuals at recovery following social interaction tasks (TSST-F) (Figure 4A). There is no significant difference between the diagnostic groups in heart rate during social evaluative tasks (TSST) at any timepoint (Table 3; Figure 4B). Within-group contexts show significantly elevated heart rate in the TSST compared to the TSST-F at interaction time points for both TD and ASD youth (Table S4). In sum, there was a statistically significant elevated heart rate in ASD subjects compared to TD subjects at recovery during social interactive (TSST-F) tasks, but this difference was not clinically meaningful (3 bpm). There was no statistically significant difference between groups during social evaluative tasks. Thus, for heart rate responses, youth with ASD and TD had relatively similar physiological responses to both social evaluation and interaction.

Table 3.

Heart Rate Pairwise Contrasts between Diagnoses for Different Stressors at all Time Points.

Social Task	Task Time	Estimate	SE	df	t Ratio	p	Holm p
Friendly/TSST-F (Social Interaction)	Preparation	−0.4821	0.9297	220	−0.5185	0.6046	1.0000
	Interaction 1	−1.7807	0.9936	221	−1.7921	0.0745	0.5215
	Interaction 2	−1.3638	0.8851	221	−1.5408	0.1248	0.5332
	Recovery	−2.5991	0.9512	220	−2.7324	0.0068	0.0544
Stressful/TSST (Social Evaluative)	Preparation	−0.4152	1.1357	221	−0.3656	0.7150	1.0000
	Speech	2.4786	1.4503	221	1.7090	0.0889	0.5332
	Math	1.1592	1.3267	224	0.8737	0.3832	1.0000
	Recovery	−1.7416	1.0208	222	−1.7062	0.0894	0.5332

Note: Contrasts (TD – ASD)

Figure 4.

Average Heart Rate by Diagnosis and Timepoint for TSST-F (4A) and TSST (4B).

RSA

The three-way interaction between diagnostic group, stressor type, and task time point was tested to investigate whether there were group differences in the RSA response to the tasks. There was not sufficient evidence for the three-way interaction (χ²=5.32, df=3, p=0.1499; Table S5); therefore, pairwise comparisons between or within diagnosis based on stressor or timepoint are not reported in the main text nor discussed. However, data has been made available in Supplemental Tables S6 and S7 for completeness. Tests of the main effect of diagnosis, however, were significant (χ²=4.32, df=1, p=0.0378), with evidence for significantly lower RSA overall in the ASD group (Figure 5). While the diagnosis main effect indicates lower RSA regulation in general in ASD, the lack of significant three-way interaction suggests that youth with ASD and TD did not differ in RSA reactivity during either social evaluation or interaction.

Figure 5. Average RSA by Diagnosis and Timepoint for TSST-F (5A) and TSST (5B).

Note: Data presented in 5B was previously published as part of a larger study on developmental and physical effects of RSA response to the TSST (Muscatello, Kim, et al., 2021).

Physiological Response Relationships

Within-subject cortisol and heart rate were significantly more positively correlated (p=0.0038) in the TD group (median=0.357, IQR=[0.012,0.619]) compared to the ASD group (median=0.119, IQR=[−0.143,0.405]). Correlations between cortisol and RSA showed significant differences between diagnostic groups (p=0.0468), with the TD group (median=−0.178, IQR=[−0.440,0.095]) showing a more negative relationship between the two outcomes compared to the ASD group (median=0.000, IQR=[−0.357,0.390]). There was no significant difference in distribution of heart rate/RSA correlations between the diagnostic groups (p=0.2133).

Aim 2. Mediating Effects of Anxiety

State anxiety levels were observed to be significantly elevated in the ASD group during social interaction tasks (TSST-F) (p=0.0178; Table 4) but were not significantly different between the groups during the social evaluative task (TSST) (p=0.5631; Table 4). Trait anxiety levels were not observed to differ depending between groups depending on social context, and individuals with autism were expected to have elevated trait anxiety (Table 4). Additionally, we observed that state (χ²=7.94, df=1, p=0.0048) anxiety was positively related with cortisol levels such that for every one unit increase in anxiety, cortisol levels are expected to increase by 1.62%. We performed a mediation analysis to see if state or trait anxiety mediated the diagnosis differences in any of the time points (interaction 1 & 2/Speech & Math) where we had found significant diagnosis differences on cortisol in the social interaction (TSST-F) and social evaluative (TSST) tasks. We did not find anxiety to significantly mediate the impact of diagnosis cortisol responses during the social interaction or social evaluation tasks (Table 5). In addition, trait anxiety was not associated with higher cortisol levels (χ²=3.59, df=1, p=0.058) and there was no association between anxiety and heart rate or RSA (Table 6).

Table 4.

Diagnostic Group Differences in State Anxiety by Social Task

Outcome	Stressor	Estimate	SE	df	t Ratio	p
State Anxiety	Friendly/TSST-F	−2.2016	0.9245	330	−2.3815	0.0178
	Stressful/TSST	0.6663	1.1511	331	0.5788	0.5631
Trait Anxiety	Friendly/TSST-F	−3.4973	1.1898	226	−2.9394	0.0036
	Stressful/TSST	−3.8319	1.2769	226	−3.0010	0.0030

Note: Contrasts (TD – ASD)

Table 5.

Average Causal Mediation Effects for Anxiety State Models.

Physiological Response	Task Time	Social Task	Estimate	95% CI	p
Cortisol	Interaction 1/Speech	Friendly/TSST-F (Social Interaction)	0.0064	(−0.0026, 0.0200)	0.190
		Stressful/TSST (Social Evaluative)	−0.0018	(−0.0108, 0.0000)	0.610
	Interaction 2/Math	Friendly/TSST-F (Social Interaction)	0.0077	(−0.0014, 0.0200)	0.120
		Stressful/TSST (Social Evaluative)	−0.0020	(−0.0112, 0.0100)	0.600
Heart Rate	Recovery	Friendly/TSST-F (Social Interaction)	0.0012	(−0.2202, 0.2400)	0.970
		Stressful/TSST (Social Evaluative)	−0.0005	(−0.1137, 0.1100)	0.980

Table 6.

Average Causal Mediation Effects for Anxiety Trait Models.

Physiological Response	Task Time	Social Task	Estimate	95% CI	p
Cortisol	Interaction 1/Speech	Friendly/TSST-F (Social Interaction)	−0.0126	(−0.0303, 0.0000)	0.074
		Stressful/TSST (Social Evaluative)	−0.01173	(−0.0320, 0.0000)	0.072
	Interaction 2/Math	Friendly/TSST-F (Social Interaction)	−0.0102	(−0.0272, 0.0000)	0.142
		Stressful/TSST (Social Evaluative)	−0.00914	(−0.0253, 0.0000)	0.130
Heart Rate	Recovery	Friendly/TSST-F (Social Interaction)	−0.0919	(−0.4339, 0.2200)	0.540
		Stressful/TSST (Social Evaluative)	−0.0709	(−0.3810, 0.2000)	0.548

Discussion

Previous research in youth with ASD has shown that seemingly benign peer interactions trigger a heightened physiological response (e.g., (Corbett et al., 2012; Corbett et al., 2010; Lopata et al., 2008; Richdale & Prior, 1992; Schupp et al., 2013)), yet more prototypical stressors such as social evaluation may not trigger the expected and adaptive reactivity (Corbett et al., 2012; Edmiston, Blain, et al., 2017; Edmiston et al., 2016; Jansen et al., 2000; Lanni et al., 2012; Levine et al., 2012). The current study therefore sought to further explore the importance of context on multiple key physiological stress markers for youth with and without autism spectrum disorders. Furthermore, the extent to which perception of threat (anxiety) mediated physiology for youth with ASD was examined. It was hypothesized that diagnostic differences in cortisol would show a double dissociation, with TD youth demonstrating cortisol reactivity during the TSST but not the TSST-F, while ASD youth would have increased reactivity to the TSST-F but no response to the TSST. Hypotheses were partially supported, though a full double dissociation was not observed. As expected from reported findings with the sample (Corbett et al., 2021), TD youth had significantly more cortisol reactivity in response to the TSST, while ASD youth lacked a HPA response to the social evaluation. While one may question whether this blunted stress response is actually adaptive, previous research directly examining emotion recognition, and in particular recognition of neutral faces, suggests that for some autistic youth, a diminished response to social evaluation may be due, in part, to an inability to perceive the task as stressful or to recognize the neutral facial expressions of the raters (Corbett et al., 2019). While the current study did not consider ability to recognize neutral faces as a mediating factor, it could be hypothesized that differences in affect recognition may have also at least partially contributed to the similar lack of cortisol reactivity to the TSST in the current study, which serves as an important and promising line of future research.

When further considering findings from the TSST-F and comparing responses across the two conditions, TD youth had a differential stress response, with no reactivity to the more friendly social interaction. Inconsistent with hypotheses, however, youth with ASD showed both a lack of stress response to the TSST as well as the TSST-F interaction, and cortisol values did not significantly differ across conditions. Previous research has demonstrated that some youth with ASD experience enhanced stress responses to benign social encounters (Corbett et al., 2012; Corbett et al., 2010; Lopata et al., 2008; Richdale & Prior, 1992; Schupp et al., 2013). For example, studies have reported elevated cortisol in youth with ASD during the PIP involving periods of solo and cooperative play with a trained, same-age research helper and another TD child (Corbett et al., 2012; Corbett et al., 2010; Corbett et al., 2014; Schupp et al., 2013). Interestingly, another study by Corbett and colleagues (2018) did not observe any group differences between ASD and TD adolescents in a pilot study of a videogame-based interaction paradigm. The authors hypothesized that the nonsignificant findings may have been a result of several key differences between the play-based and videogame paradigm, such as increased structure in the videogame interaction and preference for videogames in many youth with ASD ((Corbett et al., 2018; Mazurek & Engelhardt, 2013a, 2013b)). Similar factors may also explain the lack of cortisol reactivity differences in the TSST-F in the current study. Specifically, unlike both the PIP (Corbett et al., 2010) and videogame-based (Corbett et al., 2018) interactions, the TSST-F involves only one novel peer rather than two. Additionally, the TSST-F provides structure, with participants instructed to spend the time sharing their interests, hobbies, and favorite things with the same-aged child. Thus, for youth with ASD, the explicit instructions and decreased social demands of the TSST-F compared to the PIP may explain the lack of significant cortisol reactivity. Future research aiming to better understand stress experiences of autistic youth may examine more nuanced differences in social context, such as the number of peers involved or whether the task involves a specific activity versus a largely unstructured interaction.

Autonomic responses showed a different pattern of findings from those observed in the HPA axis. As predicted, TD youth showed elevated heart rate and decreased RSA during the TSST, consistent with vagal withdrawal/decreased PNS inhibition and possibly, sympathetic activation. Surprisingly, youth with ASD showed a similar response profile, with both elevated HR and lower RSA despite no significant HPA axis response. Furthermore, when examining within-group correlations, the ASD and TD groups did not differ in the strength of the correlation between HR and RSA. This is consistent with a previous study by Levine and colleagues (2012), which reported significantly higher cortisol in a non-autistic comparison group but no differences between children with and without autism on measures of sympathetic and parasympathetic function. However, others have reported suppressed HR in ASD during social evaluative threat (Hollocks et al., 2014; Jansen et al., 2000). Hollocks and colleagues (2014) noted that despite the differences in heart rate in their sample, no significant group differences in those with or without ASD were noted for HF-HRV or the LF/HF ratio, which may serve as an indirect index of sympathetic activation.

The HR reactivity observed in both groups could be attributed to either vagal withdrawal (removal of PNS inhibitory control), sympathetic reactivity, or a combination of the two. Indeed, one barrier to using the HR metric in studies of the ANS is its limited utility as an index of autonomic dysfunction (Schmitz et al., 2011). In particular, conclusions regarding the balance of the SNS and PNS cannot be determined. The SA node of the heart is dually innervated by both the sympathetic and parasympathetic systems; therefore, changes in heart rate could be due to changes in activation of either branch. Furthermore, heart rate can be influenced by factors not attributed to direct autonomic innervation of the SA node, such as indirect sympathetic modulation via catecholamine transmission (Berntson et al., 1997). Some studies have tried to circumvent this issue of dual innervation by calculating the ratio of low-frequency HRV relative to high-frequency HRV, with the assumption that the ratio can provide a measure of the influence of the sympathetic system on low-frequency HRV while accounting for parasympathetic activity. Yet the utility of the LF/HF ratio as a true index of sympathetic activation remains controversial (Force, 1996). In the context of the current study and interpreting group differences, or lack thereof, it will be important for future studies to directly examine a more accurate index of sympathetic activity, such as pre-ejection period (Brenner & Beauchaine, 2011) to identify whether different social stressors induce SNS reactivity similarly in ASD and TD. While the RSA findings point to no differences between the groups in parasympathetic withdrawal, it remains unclear whether any changes in sympathetic reactivity would have been observed in response to the TSST in ASD youth.

During a friendly social interaction, heart rate responses largely mirrored the cortisol findings, with minimal group differences between ASD and TD youth. Although only significant during the recovery period, ASD youth tended to have higher HR overall during the TSST-F (see Figure 4B), potentially indicative of some increased ANS arousal to social interaction. Furthermore, both groups demonstrated an increase in RSA to the social interaction. Notably, however, while the ASD group had an increase in RSA to the social interaction, the magnitude was less than the TD group, particularly during the first half of the social interaction. As demonstrated in Figure 5, mean average RSA was often lower in the ASD group compared to the TD group in both the TSST and TSST-F. Nevertheless, this reduction in RSA, particularly during social interaction, is consistent with a growing literature to support parasympathetic differences in autistic individuals (see (Patriquin et al., 2019) for review). This continues to raise speculation as to whether reductions in parasympathetic regulation are related to social engagement challenges in ASD (e.g., (Porges, 2005)). Further examination into the relationship between autistic characteristics, social behavior, and HRV is necessary before more conclusions can be drawn regarding the role of the ANS in autism. As the current study reports diagnostic differences in magnitude (mean RSA) but not slope (change in RSA to stress over time), it will be important for next steps to consider whether ASD is characterized by an intrinsic hypoarousal of the PNS and what the clinical implications may be for a regulatory system that is overall less active (magnitude) but equally responsive (slope) in this population.

In addition to examining group differences in physiological response patterns for each of the three outcomes (cortisol, HR, and RSA), the study also included a preliminary examination of within-diagnosis relationships between the physiological metrics. For both groups, cortisol and HR were positively related, where higher cortisol was associated with higher heart rate. In contrast, a negative correlation was observed with cortisol and RSA responsivity. Notably, these correlations were significantly stronger in the TD group. These findings may be partially explained by the blunted cortisol response to social evaluation (TSST) in the ASD group. In response to the TSST, autistic youth demonstrated HR and RSA responses similar to their TD peers, while cortisol reactivity was significantly lower. Thus, weaker correlations between cortisol and HR or cortisol and RSA in the ASD group may be driven largely by the unusual stress response in one system (HPA axis/cortisol) but not the other (ANS/HR and RSA). However, these preliminary analyses focused on physiological response overall, rather than differentiating by task type and/or time point. Therefore, future research will benefit from close examination of relationships between physiological systems to provide a more nuanced understanding of stress system interactions and profiles and the role they play in psychosocial functioning.

Finally, the study aimed to explore the relationship between physiological response and self-reported anxiety. Following the TSST, youth with ASD compared to their TD peers did not report significantly different state anxiety despite a significantly blunted cortisol response. Interestingly, they did report elevated state anxiety following the TSST-F, suggesting that the youth with ASD may have perceived the friendly encounter as more anxiety-provoking, even if the physiological response, particularly within the HPA axis, did not match their psychological perception. Despite that interesting distinction, there were no mediating effects of state or trait anxiety on the relationship between diagnosis and physiological response. Previous literature linking anxiety and physiology (cortisol, HR, RSA) has been mixed, particularly within ASD (see (McVey, 2019) for review). For example, some have shown that although youth with ASD have significantly less cortisol response compared to their non-autistic peers during the TSST, their reports of state anxiety immediately following the stressor did not differ (e.g., (Corbett et al., 2019; Lanni et al., 2012; Simon & Corbett, 2013)). Indeed, other studies examining the role of anxiety in autism and physiological stress reveal a complex profile in which those with autism plus clinically significant anxiety had significantly blunted physiological arousal compared to those without anxiety (both ASD and TD) (Hollocks et al., 2014). Certainly, the relationship between physical response and psychological perception of threat and stress is a complicated one in which future research of large samples of those with and without ASD and with and without clinical anxiety will be necessary to elucidate the nature of these complex relationships.

Limitations and Future Directions

The current study had many strengths, including a large, well-characterized sample and multimethod examination of physiology across systems (HPA, ANS). Furthermore, the inclusion of two paradigms – one examining social evaluation and the other social interaction – allowed for unique examination of stress and arousal in varying social contexts. Nevertheless, the study was not without its limitations. The study sample was primarily White (83%). Due to the cognitive and language demands of the larger study, all participants had to have an IQ above the threshold for intellectual disability (≥70). An estimated 37.9% of youth with ASD has a co-occurring intellectual disability (Maenner et al., 2023); therefore, our results may not be generalizable to the broader autism spectrum. Future research in more diverse samples that is also inclusive across the broader range of the intellectual spectrum are necessary. Additionally, as mentioned, more precise measures of ANS function, particularly within the sympathetic branch, are critical to examine the unique effects of autonomic balance more comprehensively in social functioning and ASD. The current study also relied on RSA as the primary measure of parasympathetic function; however, the utility of this metric as a valid measure of parasympathetic/vagal control is debated (e.g.,(Grossman & Taylor, 2007; Thomas et al., 2019)). While other investigators strongly support using RSA as an outcome measure for vagal tone (Porges, 2023), it is critical that future studies be pursued in which multiple autonomic measures, including both linear and nonlinear metrics, are examined to provide a more comprehensive understanding of autonomic function or dysfunction in ASD. Regarding the TSST social evaluation paradigm, numerous social factors may influence one’s perception of the task as stressful (see (Allen et al., 2017) for review). Although the study utilized a mixed-age/mixed-sex approach (one adult rater and one same-aged peer), comparisons of the extent to which age- or sex- matching of the judges with the participant influenced stress response was beyond the scope of the study. Therefore, future research including an in-depth exploration on the role of age- and sex-matching of judges in the TSST would be informative. Finally, while the current study primarily focused on group- or diagnostic-level differences, future studies would benefit from in-depth examination of individual-level or within-group responses to identify subtypes based upon responder profiles (e.g., activational versus reciprocal response) and individual characteristics, such as sex, autistic characteristics/traits, social functioning, medication use, and psychiatric co-occurring conditions.

Conclusions and Implications

To conclude, the study findings show that context matters when seeking to understand physiological reactivity to social stressors in youth with ASD. Moreover, the current report uniquely demonstrates that utilization of a multimethod design to simultaneously examine several stress systems, including the HPA axis and the ANS, will provide an enhanced, nuanced view into physiological reactivity within ASD. These findings provide a guide in which future research may elucidate unique response patterns across physiological systems to more precisely identify those with heightened physiological arousal when interacting in different social contexts. In turn, these response subtypes may be examined with social behavior and psychological functioning (e.g., anxiety) to identify directionality of these relationships (e.g., physiology influences behavior, or vice versa) and thereby better inform approaches for enhancing social engagement.

Highlights.

• Children with ASD experience stress differently across social contexts.

• Stress and arousal within context differ between physiological systems.

• A multisystem approach in examination of physiological reactivity is important.

• Perceived anxiety does not mediate the relationship between diagnosis and stress.

• Autism may often be characterized by parasympathetic hypoarousal.