Autonomic Disequilibrium at Rest in Autistic Children and Adults

Abstract

Autism Spectrum Disorder (ASD) symptoms have been proposed to be linked to Autonomic Nervous System (ANS) atypical functioning, in particular sympathetic hyper-arousal and parasympathetic under-activation. The objective of this study was to characterize autonomic functioning at rest in autistic and neurotypical children and adults. To characterize several aspects of autonomic functioning, we recorded simultaneously pupil diameter, heart rate and electrodermal activity during 5 min of rest in 44 children (6–12 years old, 22 autistic) and 42 adults (19–52 years old, 21 autistic). Several parameters allowed to characterize tonic and phasic indices of sympathetic and parasympathetic systems at rest. Autistic children exhibited the expected pattern of parasympathetic under-activation at rest compared to their typically developing (TD) peers, and with a tendency for a higher phasic sympathetic activity. Adults exhibited a reverse autonomic pattern, with autistic individuals showing higher sympathetic tonus and lower sympathetic phasic activity than their TD peers. In conclusion, we observed an autonomic disequilibrium at rest both in autistic children and adults, but with opposite patterns that could reflect adaptive compensation mechanisms during maturation. This disequilibrium in autistic children would switch from excessive phasic components to excessive tonic components in adults, possibly subtended by an atypical locus coeruleus functioning.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10484-025-09696-z.

Introduction

Autism spectrum disorder (ASD; DSM-5, 2013) is a complex neurodevelopmental disorder.1 With a prevalence estimated around 1% in the world (Maenner et al., 2021; Matson & Neal, 2009), it is defined by “persistent deficits in social communication and social interaction” and “restricted, repetitive patterns of behavior, interests, or activities” (DSM-5, 2013). Several clinical symptoms have been interpreted as possibly reflecting Autonomic Nervous System (ANS) atypical functioning in ASD, like social engagement differences (Porges, 2001), atypical sleep patterns (Liu et al., 2006), gastrointestinal disorders (Ferguson et al., 2017) or adverse health outcomes, such as cardiorespiratory dysfunction (Bricout et al., 2018; Heffernan et al., 2018; Kushki et al., 2013; Thapa et al., 2021).

The ANS maintains homeostasis (Cannon, 1929), mainly through a complex interaction between its two sympathetic and parasympathetic branches, involved respectively in high and low arousal states (McCorry, 2007), in response to a stimulus (autonomic reactivity) but also at rest (autonomic tone). Extending Cannon’s “Fight or Flight” theory (Cannon, 1929), Porges’ Polyvagal theory (1995, 2009) focused on adaptive mobilization of parasympathetic (PNS) and sympathetic (SNS) nervous systems in response to the context. In particular, the influence of the PNS has been evaluated by measuring cardiac vagal tone and vasomotor regulation, e.g. blood pressure variability (Kalfiřt et al., 2023; Ming et al., 2016; Tonhajzerova et al., 2021). These studies have revealed that, on top of its role in cardiac balance restoration, the PNS may play a crucial role in freezing (Friedman et al., 1993) but also in empathy and social engagement (Fabes et al., 1994; Hastings et al., 2008; Suess et al., 1994). Nonetheless, while increased PNS is associated with a variety of positive health outcomes, lower cardiorespiratory fitness levels seem to be connected to adverse health outcomes, such as chronic inflammation or cardiac dysfunction (Kushki et al., 2013).

Cardiac vagal tone has been described as downregulated in ASD (Porges et al., 2013), possibly related to social communication differences in autistic persons. Studies focusing on cardiac autonomic tone mainly suggest a hyper-arousal state in autistic children, with an elevated basal heart rate (Kushki et al., 2014) and a reduced cardiac parasympathetic activity (Kalfiřt et al., 2023; Ming et al., 2016; Thapa et al., 2019) compared to their neurotypical peers. In their meta-analysis, Arora et al. (2021) reported that 83% of studies using cardiac autonomic indices point towards a hyper-arousal in ASD. This hyper-arousal at rest could interfere with the generation of optimal ANS responses to external stimuli (Arora et al., 2021; Dalton et al., 2005; Dell’Osso et al., 2022; Porges, 1995; Schultz et al., 2006) while reduced cardiac parasympathetic activity in addition with high BMI (Bricout et al., 2018) seems to expose autistic children to a higher risk of arterial stiffening and cardio vascular diseases later in life (Heffernan et al., 2018). It has been proposed that repetitive behaviors and social avoidance may be adaptive strategies to regulate arousal in autistic persons (Hirstein et al., 2001; Hutt et al., 1964).

Most of the cardiac indices studied reflect mainly PNS regulation (e.g. Respiratory Sinus Arrhythmia—RSA, blood pressure variability or Low and High Frequencies in Heart Rate Variability—HRV analyses), while heartbeat is influenced by both SNS and PNS. Other autonomic measures could thus help to better understand the respective SNS and PNS modulations of autonomic functions in ASD. Indeed, in their review, Arora et al. (2021) showed that studies using electrodermal activity (EDA), under strict SNS control, mostly found hypo-arousal (in 56% of the studies) in ASD, while studies using pupil diameter (influenced by both SNS and PNS) are too few to properly conclude about arousal in ASD. It is important to note that SNS and PNS are not necessarily antagonistic, with possibilities of co-activation, co-inhibition, and uncoupled modes of functioning that could depend on the organ considered (Berntson et al., 1991, 2007).

The inconsistency in arousal differences between autistic and typically developing (TD) peers could thus arise from the specific autonomic indices and effectors chosen. This suggests that simultaneous measures of several ANS parameters could help resolve these inconsistencies. Moreover, ANS studies in ASD have targeted different age groups. While Arora et al. (2021) evidenced that differences (or absence of difference) between autistic and neurotypical populations could be found whatever the age, only one study directly tested the same paradigm in several age groups (Tessier et al., 2018) and showed hypo-PNS modulation in ASD, only in adults. Other factors could influence the overall pattern of results, like the experimental state (activation or rest), and Arora et al. (2021) pointed towards a larger difference between autistic and neurotypical populations during rest. At rest, most studies have focused on tonic values of ANS effectors, i.e. stable components characterizing the tonic ANS discharge at rest, while other phasic parameters can also be observed, like spontaneous EDA peaks. These phasic parameters are not usually studied but could reflect the lability of resting state.

The goal of this cross-sectional study was to characterize ANS functioning at rest in autistic populations. To this end, we tested both children and adults, comparing autistic and neurotypical age-matched participants, in a sitting position. As previously used to study ANS functioning at rest and its maturation process in TD populations (Bufo et al., 2022), we tested a 5 min rest paradigm (Bharath et al., 2019), recording simultaneously electrodermal activity, heart rate and its variability, and pupillary diameter. These three effectors being differently controlled by sympathetic and parasympathetic branches, this approach could help to better understand the hypo- and hyper-arousal suspected in ASD. Considering the literature, we hypothesized that, if ANS dysregulation was present, we should observe mainly hypo-PNS and hyper-SNS modulations of our parameters (Arora et al., 2021). As HRV–High frequency parameters are considered respiratory-linked heart rate fluctuations (Shaffer & Meehan, 2020; Shaffer et al., 2014), we expected to find lower values of HRV–High frequency in autistic compared to neurotypical populations. Knowing that recording position (e.g. supine, orthostatic, sitting) may affect HRV (Shaffer & Ginsberg, 2017; Tonhajzerova et al., 2021), and that resting HRV–Low frequency parameters specifically recorded in a sitting position seems not to reflect cardiac sympathetic fluctuations (Shaffer & Ginsberg, 2017), we expected lower HRV–Low frequency parameters in autistic compared to neurotypical populations. We also expected a sympathetic overarousal indicated by higher values in EDA parameters and higher median pupil diameter in autistic compared to neurotypical populations.

Materials and Methods

Participants

Twenty-two autistic children (aged 6 to 12 years, mean age 8.1 years ± 1.7, 2 females) and 21 autistic adults (aged from 19 to 49, mean age 29.4 years ± 7.0, 11 females) were recruited through the University Hospital of Tours, EXcellence in Autism Center-Tours (see Table 1). Inclusion criteria were ASD diagnosis according to DSM-5 (DSM-5, 2013) or CIM-10 criteria, and age between 6 and 12 years old for children and above 18 years old for adults. The ASD diagnosis was made by a child or an adult psychiatrist by using ADI-R (Lord et al., 2016) and/or ADOS-2 (Lord et al., 1994) assessments, after a multidisciplinary evaluation personalized for each participant. Depending on the individual profile, several health professionals participated in the global evaluation, including detailed developmental history, psychiatric assessment and pediatric, psychological and neurological examinations. Non-inclusion criteria were the use of medical treatments affecting neural activity, in particular all medications that could affect sympathetic or parasympathetic measures. Age-matched neurotypical participants were recruited according to the following criteria: children aged between 6 and 12, no psychiatric disorders or neurologic diseases previously diagnosed, no learning disabilities or difficulties. The neurotypical groups were constituted of 22 children (aged 6 to 12 years, mean age 9.5 years ± 1.7, 16 females) and 21 adults (aged from 23 to 52, mean age 34.7 years ± 6.5, 17 females). All neurotypical recruitments were done locally in the University, University Hospital, and local schools. The neurotypical children were already part of the data described in a previous paper (Bufo et al., 2022).

Table 1.

Characteristics of neurotypical (TD) and autistic (ASD) children (C) and adult (A) participants

	Children		Children subgroups		Adults
	TDC	ASDC	TDC-PUP	ASDC-PUP	TDA	ASDA
N	22	22	8	8	21	21
Age (± SD)	9 ± 1	8 ± 1	8 ± 1	8 ± 1	34 ± 6	29 ± 7
Sex F:M	10:12	2:20	6:2	7:1	11:10	11:10
Verbal IQ [range]	124 [88–155]	68 [20–95]***	116 [88–135]	58 [20–86]***	111 [79–129]	117 [92–139]
Non verbal IQ [range]	117 [90–146]	82 [43–129]***	106 [90–114]	66 [43–90]***	106 [74–124]	105 [74–144]
CARS [range]		30 [24–40]		31 [26–40]
ADOS-2 [range]		Module [1–3] (n = 17)		Module [1–3] (n = 5)
Communication + Social Interaction		[4–20]		[13–20]		9 [1–17]
Stereotyped behaviors and Restricted interests		[1–6]		[2–5]		2 [0–5]
ADI-R [range]
Age of onset						2 [0–5]
Social						20 [11–30]
Communication						13 [4–25]
Repetitive restricted						6 [2–11]
AQ [range]					9 [1–17]	34 [22–47]***

For children, pupil data were obtained only in a subgroup of 8 autistic children, and we compared it with an age-matched subgroup of neurotypical children, noted C-PUP. Age is expressed in mean ± standard deviation (SD), all the other information is expressed in range from minimum to the maximum value. For autistic children, ADOS-2 scores ranges are reported in italics: they are given only for qualitative description of the population as different modules were evaluated

ADI-R Autism Diagnostic Interview Revised, ADOS-2 Autism Diagnostic Observation Schedule Second Edition, AQ Autism Quotient, CARS Childhood Autism Rating Scale, IQ Intellectual Quotient

***p < 0.001 for the comparison between TD and ASD groups

The children gave verbal consent. Their parents and the adult participants provided written informed consent according to institutional guidelines. The experiment conformed to the Code of Ethics of the World Medical Association (2013). The protocol was ethically approved by the Comité de Protection des Personnes (CPP; protocol PROSCEA 2017-A00756-47). All the participants were recorded in the same conditions, in the University Hospital of Tours, France.

For autistic persons, IQ and CARS scores were evaluated during their clinical evaluation outside of the research protocol. For autistic children, ADI-R scores were not communicated for the research and ADOS-2 scores were evaluated via different modules depending on the participant, with no calculation of a global score. For neurotypical participants, IQ scores were evaluated with 4 subtests of the WISC-IV in children (block design, matrix reasoning, vocabulary, and similarities) and 6 subtests of the WAIS-IV in adults (block design, matrix reasoning, visual puzzles, vocabulary, similarities, and information), before or after the experimental protocol for children and in a separate session for adults. For adults, AQ and questionnaires about weight, size, coffee drinking and smoking habits were filled before the protocol. Indeed, ANS basal state can also be sensitive to several parameters affecting directly the metabolism and that have been shown to differ between autistic and neurotypical populations (e.g. Body Mass Index BMI; Molfino et al., 2009). The impact of coffee drinking and smoking on ANS measures is presented in Supplementary Information.

Material

Cardiac and electrodermal tones were recorded using BIOPAC MP36® (BIOPAC Systems, Inc., Goleta, CA, USA), with a constant voltage of 0.5 V. The AcqKnowledge® 4.1 software was used to monitor these signals during the experiment. EDA was recorded using two 8 mm Ag/AgCl cup electrodes (EL258, BIOPAC Systems IN, Goleta CA, USA) and 0.5%-NaCl electrode paste (GEL101; BIOPAC Systems), positioned on the second phalange of the index and medium finger of the right hand for the children, and of the non-dominant hand for the adults (i.e. left hand in all adults except 3 ASD and 1 TD participants). The frequency of acquisition for EDA was 1 kHz (range band 0–5 Hz). The electrocardiogram (ECG) was recorded using two disposable vinyl electrodes (EL503, BIOPAC Systems, Inc., Goleta CA, USA) placed on the sternum and on the right shoulder. The frequency of acquisition for ECG was fixed at 1 kHz (range band 0–35 Hz). The pupil diameter was acquired with a frequency of 500 Hz by using the eye tracking system SMI RED500® synchronized with BIOPAC MP36®. Figure are created with Excel (v. 2024) and JAMOVI (v. 2.4.11), modified with GIMP (v. 2.10.14).

Procedure

The recordings were all carried out in the same room, with constant conditions of luminosity (10 Lux for children and 19 Lux for adults), hygrometry (27% rh) and temperature (23 °C/76.4 °F). After EDA and ECG captors were installed, the participants were seated in an armchair opposite a computer screen (resolution 1920 × 1080 pixels, visual angle around 45 × 25°) on which the eye tracking illuminator (λ = 870 nm, norm compliance: CE, EMC, Eye Safety) and camera were fixed. Optimal distance and height were set up by using the IView-X software. Children’s position could be adjusted with a booster seat when necessary. The first experimental step was the eye tracking calibration, either in five points when the participants could concentrate on the white spot displayed in five sequential locations on the gray screen, or in a unique central point, in particular for autistic children. During the experiment, only a central black fixation cross on a gray background appeared on the screen. All participants were asked to stay still and to look at the screen during the experiment. The experimenter and participant were separated by a panel. After placement of the electrodes, the participants adjusted themselves to be comfortably seated in the armchair, and a period of 5 min was respected before starting the recording to stabilize the physiological constants. We then recorded simultaneously pupillary, cardiac and electrodermal tone during 5 min of rest in sitting position (Bharath et al., 2019). This time is considered to be long enough to allow for reliable analysis of HRV parameters while being short enough to ask children to sit still.

Signal Processing: Tonic and Phasic Parameters Extraction

Overall, the same methodological approach was used as in Bufo et al. (2022).

Pupil Tone

The pupil raw signal was preprocessed with MATLAB® (r2016a; MathWorks). The first step of preprocessing eliminated artifacts, such as blinking and brief signal losses, thanks to a velocity-based algorithm (Kret & Sjak-Shie, 2019; Nyström & Holmqvist, 2010). Afterwards, the resulting signal was smoothed using a median filter and a band pass filter (0.0004–0.0150 Hz). For each participant, we extracted three tonic parameters calculated on the 5 min of recording: median pupil diameter (in mm), hippus frequency (in Hz), and hippus amplitude (in mm), evaluated thanks to a Fourier transform method.

Cardiac Tone

The heart rate and heart rate variability were analyzed in Kubios (Tarvainen et al., 2014). On the 5 min of the recording, we first calculated the interbeat interval (RR interval) between each QRS complex expressed in milliseconds (ms). Artifacts were then detected and corrected based on the deviation of the RR interval value from a local average interval. The HRV analysis was performed on the whole 5 min window, with equidistant sampling interpolation at 4 Hz. The temporal domain was explored by calculating the RMSSD (Root Mean Square of Successive Differences between normal heartbeats), index of parasympathetic activity (Cheng et al., 2020; Shaffer & Ginsberg, 2017; Shaffer & Meehan, 2020; Shaffer et al., 2014). The frequency of the oscillations was evaluated by using a Fourier transform. We focused on high frequency (HF; 0.15–0.4 Hz), reflecting cardiac parasympathetic activity (Pomeranz et al., 1985), and low frequency (LF; 0.04–0.15 Hz), reflecting baroreflex activity at rest, index of vagal tone (Malliani et al., 1991; Pagani et al., 1986; Posada-Quintero et al., 2016; Shaffer & Ginsberg, 2017; Shaffer et al., 2014). For each participant, we thus obtained seven tonic parameters: RR interval (in ms), RMSSD (in ms), LF and HF peak frequency (in Hz), LF and HF absolute power (in ms²), and LF/HF ratio (absolute density LF/absolute density HF minus very low frequency; Berntson et al., 1997).

EDA Tone and Phasic Parameters

Preprocessing of the electrodermal activity was performed in Ledalab (Benedek & Kaernbach, 2010), an open-source software for MATLAB® (r2016a; MathWorks). The data were down-sampled to 10 Hz and bandpass-filtered with a first order Butterworth filter and cut-off frequencies of 5 Hz (Bach et al., 2009). Artifacts due to noise were corrected by using the spline interpolation. To ensure a conservative estimate of residual variance, we did not exclude non-responses (Staib et al., 2015). We used a Continuous Decomposition Analysis (CDA) to analyze four EDA parameters: tonic component of EDA (in µS), area under the curve (EDA AUC; in µS⋅s) for the whole 5 min, number of EDA peaks (the only phasic parameter), and amplitude of EDA peaks (in µS), based on Standard Deconvolution method (Benedek & Kaernbach, 2010; Boucsein, 2012).

Statistical Analyses

Considering the differences in acquisition (luminosity, location for EDA electrodes) and in clinical profiles between children and adults, we explored the effect of group (ASD vs. TD) on our measures, separately for children and adults.

All the statistical analyses were performed in JAMOVI® (version 2.2.1.0) and XLSTAT® (version 2020.1.2). The normality of the distribution of the data was verified by using the Shapiro–Wilk test and the homogeneity of the variance was verified by using the Levene test.

We evaluated the statistical difference between TD and ASD groups on pupil, cardiac and electrodermal parameters thanks to General Linear Models. As we had only a small sample size, especially for pupil parameters, we could not test all factors in the same model. We thus added age as a continuous factor in a separate model. We also tested sex as a categorical factor in a separate model. The influence of BMI, coffee drinking and smoking was also evaluated in a separate model in adults (results for the latter two are presented in Supplementary Information). Finally, we performed non-parametric Spearman correlations to verify the covariation of the different autonomic parameters with clinical data (autism severity: CARS in children and AQ in adults; verbal IQ) only in ASD children and adults.

Considering the number of participants, a sensitivity analysis was performed in G*Power ® 3.1 separately for children and adults, and for pupil, cardiac and electrodermal parameters. The statistical power considered was 80%. Overall, for each group (n = [39–44]) and each autonomic parameter, we could expect to detect large effects (f = [0.43–0.46], critical F = [4.07–4.10]; except for pupil parameters in children: n = 16, f = 0.75, critical F = 4.60).

As LH and HF absolute power, amplitude of EDA peaks and EDA AUC differed greatly among individuals, to meet the assumptions of inferential statistics they were logarithmically transformed (Benedek & Kaernbach, 2010; Sinnreich et al., 1998; Young & Leicht, 2011). All results were expressed in mean with standard error (S.E.); for significant results, effect size (expressed in partial eta squared η²p) and confidence interval (CI) of statistical estimates calculated at 95% are presented. p-values were corrected according to the False Discovery Rate (FDR) correction, following the Benjamini–Hochberg procedure, according to the number of tests performed on each parameter. Significant p values (p ≤ 0.05) and tendencies (0.05 < p ≤ 0.1) were showed along with their FDR threshold (p_FDR).

Results

Autonomic Parameters in Neurotypical and Autistic Children

Pupil Tone

Only eight autistic children (ASD_C-PUP) looked long enough at the screen to properly evaluate pupillary parameters. An age-matched neurotypical group (TD_C-PUP) was created to perform the statistical comparisons between the two groups. Age-matched TD participants were case by case selected.

As can be observed in Fig. 1A, none of the three pupillary parameters differed as a function of the group (median pupil diameter: ASD_C-PUP 5.26 ± 1.27 mm vs. TD_C-PUP 5.30 ± 0.34 mm, F₁₂ = 0.44, p = 0.51; hippus frequency: ASD_C-PUP 0.22 ± 0.08 Hz vs. TD_C-PUP 0.21 ± 0.04 Hz, F₁₂ = 1.11, p = 0.31; hippus amplitude: ASD_C-PUP 0.30 ± 0.08 mm vs. TD_C-PUP 0.30 ± 0.06 mm, F₁₂ = 0.88, p = 0.36).

Fig. 1.

Autonomic parameters in autistic (ASD) and neurotypical (TD) children. A Pupil parameters. Histograms represent the mean values (± standard error) of median pupil diameter (in mm), pupillary hippus frequency (in Hz), and hippus amplitude (in mm), extracted for both autistic children (dark grey/left columns) and neurotypical children (light grey/right columns) for the subgroups ASD_C-PUP and TD_C-PUP. B Cardiac parameters. Histograms represent the mean values (± standard error) of RR interval (in ms), RMSSD (in ms), LF peak frequency (in Hz), HF peak frequency (in Hz), LF absolute power (log scale, in ms²), and HF absolute power (log scale, in ms.²), extracted for both ASD_C (dark orange/left columns) and TD_C (light orange/right columns) groups. C EDA parameters. Histograms represent the mean values (± standard error) of the tonic component of EDA (in µS), EDA AUC (log scale, in µS s), number of EDA peaks, and amplitude of EDA peaks (log scale, in µS), extracted for both ASD_C (dark blue/left columns) and TD_C (light blue/right columns) groups. *p < 0.05 (Color figure online)

Age was introduced as a covariate of the group category, but no significant effect of age was found on any of the pupil parameters (median pupil diameter: F₁₃ = 0.49, p = 0.45; hippus frequency: F₁₃ = 0.0002, p = 0.99; hippus amplitude: F₁₃ = 0.003, p = 0.95). No significant interaction group × age was found on any of the pupil parameters (median pupil diameter: F₁₂ = 0.70, p = 0.41; hippus frequency: F₁₂ = 1.52, p = 0.24; hippus amplitude: F₁₂ = 0.41, p = 0.52).

No effect of sex was found on pupil parameters (median pupil diameter: F₁₂ = 0.58, p = 0.45; hippus frequency: F₁₂ = 0.71, p = 0.41; hippus amplitude: F₁₂ = 2.56, p = 0.13), nor any group × sex interaction (median pupil diameter: F₁₂ = 1.05, p = 0.32; hippus frequency: F₁₂ = 4.04, p = 0.57; hippus amplitude: F₁₂ = 0.04, p = 0.84).

Cardiac Tone

Cardiac parameters could be extracted for the whole group of 22 autistic children (ASD_C) and age-matched neurotypical children (TD_C).

RR interval was smaller (i.e., the heartbeat faster; Fig. 1B) for ASD_C (630 ± 71.2 ms) than for TD_C (706 ± 90.7 ms; F₄₀ = 4.63, p = 0.03, that stayed a tendency following FDR correction p_FDR = 0.07, η²p = 0.10, CI = [4.20 to 134]). For HRV parameters (Fig. 1B), a group effect was reported for LF peak frequency. LF peak frequency was lower for ASD_C (0.08 ± 0.006 Hz) than for TD_C (0.11 ± 0.004 Hz; F₄₀ = 4.68, p = 0.01, p_FDR = 0.05, η²p = 0.10, CI = [0.001 to 0.04]), but no group differences were found for other cardiac parameters as HF peak frequency (ASD_C 0.26 ± 0.06 Hz vs. TD_C 0.26 ± 0.07 Hz; F₄₀ = 0.00001, p = 0.99), HF absolute power [lower for ASD_C (in log scale), 2.73 ± 4.43 ms² than for TD_C 3.06 ± 4.407 ms²; F₄₀ = 098, p = 0.32], LF absolute power (in log scale, ASD_C 2.68 ± 0.10 ms² vs. TD_C 2.92 ± 0.07 ms²; F₄₀ = 0.08, p = 0.77) and RMSSD (ASD_C 39.1 ± 4.41 ms, TD_C 57.0 ± 6.32 ms; F₄₀ = 2.21, p = 0.14).

An effect of age was reported for RR interval (F₄₁ = 14.11, p < 0.001, p_FDR = 0.005, η²p = 0.25, CI = [0.82 to 2.75]) but not for the other cardiac parameters (LF peak frequency: F₄₁ = 0.005, p = 0.94, HF peak frequency: F₄₁ = 0.45, p = 0.50, HF absolute power: F₄₁ = 0.87, p = 0.35, LF absolute power: F₄₁ = 0.50, p = 0.48 and RMSSD: F₄₁ = 0.52, p = 0.43). No interaction group × age was reported for RR interval (F₄₀ = 1.10, p = 0.30), LF peak frequency (F₄₀ = 2.01, p = 0.16), HF peak frequency (F₄₀ = 0.60, p = 0.44, HF absolute power: F₄₀ = 0.30, p = 0.58, LF absolute power: F_40: = 0.24, p = 0.62 and RMSSD: F₄₀ = 0.62, p = 0.43).

No effect of sex was found for the cardiac parameters (RR interval: F₄₀ = 2.12, p = 0.15, LF peak frequency: F₄₀ = 0.00003, p = 0.99, HF peak frequency: F₄₀ = 0.03, p = 0.85, HF absolute power: F₄₀ = 0.57, p = 0.45, LF absolute power: F₄₀ = 0.38, p = 0.53 and RMSSD: F₄₀ = 0.47, p = 0.49). No group × sex interaction was observed either (RR interval: F₄₀ = 2.76, p = 0.10, LF peak frequency: F₄₀ = 0.61, p = 0.43, HF peak frequency: F₄₀ = 0.00006, p = 0.99, HF absolute power: F₄₀ = 1.26, p = 0.26, LF absolute power: F₄₀ = 3.26, p = 0.08 and RMSSD: F₄₀ = 0.77, p = 0.38).

EDA Tone and Phasic Parameters

EDA parameters could be extracted for the whole group of 22 neurotypical children (TD_C), but only in 19 autistic children (ASD_C) due to agitation for 3 children.

No significant difference between groups (Fig. 1C) was observed for three extracted EDA parameters (number of EDA peaks: ASD_C 75 ± 31 vs. TD_C 57 ± 35, F₃₇ = 2.07, p = 0.15; EDA AUC in log scale: ASD_C 1.58 ± 0.53 μS⋅s vs. TD_C 1.32 ± 0.75 μS⋅s, F₃₇ = 2.21, p = 0.14; amplitude of EDA peaks: ASD_C 0.84 ± 0.63 μS vs. TD_C 0.63 ± 0.74 μS, F₃₇ = 2.11, p = 0.15). There was a tendency for tonic component of EDA, with higher values in autistic children (ASD_C 4.12 ± 2.74 μS vs. TD_C 3.64 ± 2.68 μS, F₃₇ = 3.41, p = 0.08) that did not survive FDR correction (p_FDR = 0.28).

No effect of age was found on any of the EDA parameters (number of EDA peaks: F₃₈ = 0.002, p = 0.95; EDA AUC in log scale: F₃₈ = 0.05, p = 0.81; amplitude of EDA peaks: F₃₈ = 0.53, p = 0.47; tonic component of EDA: F₃₈ = 0.39, p = 0.53). No interaction group × age was found on any of the EDA parameters (number of EDA peaks: F₃₇ = 1.50, p = 0.22; EDA AUC in log scale: F₃₇ = 1.36, p = 0.24; amplitude of EDA peaks: F₃₇ = 2.22, p = 0.14; tonic component of EDA: F₃₇ = 0.52, p = 0.47).

No effect of sex was reported for EDA parameters (number of EDA peaks: F₃₇ = 0.08, p = 0.77; EDA AUC in log scale: F₃₇ = 0.59, p = 0.44; amplitude of EDA peaks: F₃₇ = 1.29, p = 0.26), except a tendency on the tonic component of EDA (F₃₇ = 3.52, p = 0.06). Similarly, no group × sex interaction was reported for EDA parameters (number of EDA peaks: F₃₇ = 0.04, p = 0.84; EDA AUC in log scale: F₃₇ = 0.57, p = 0.45; amplitude of EDA peaks: F₃₇ = 0.69, p = 0.40), except a tendency on the tonic component of EDA (F₃₇ = 3.02, p = 0.09).

Autonomic Parameters in Neurotypical and Autistic Adults

Pupil Tone

Pupillary parameters could be evaluated in the 21 neurotypical adults (TD_A) and 19 out of the 21 autistic adults (ASD_A). Pupil acquisition failed in two participants due to technical problems.

As illustrated in Fig. 2A, we observed a significant difference for the median pupil diameter between the two groups (F₃₆ = 6.51, p = 0.01, η²p = 0.15, CI = [0.07 to 0.63], but only a tendency after FDR correction p_FDR = 0.07), with a larger pupil diameter in autistic (3.56 ± 0.11 mm) than in neurotypical participants (3.21 ± 0.07 mm). There was also a significant difference in hippus amplitude (ASD_A 0.21 ± 0.01 mm vs. TD_A 0.14 ± 0.0008 mm; F₃₆ = 11.06, p = 0.002, p_FDR = 0.01, η²p = 0.23, CI = [0.02 to 0.11]), but not in hippus frequency (ASD_A 0.22 ± 0.02 Hz vs. TD_A 0.22 ± 0.02 Hz; F₃₆ = 0.00004, p = 0.98).

Fig. 2.

Autonomic parameters in autistic (ASD) and neurotypical (TD) adults. A Pupil parameters. Histograms represent the mean values (± standard error) of median pupil diameter (in mm), pupillary hippus frequency (in Hz), and hippus amplitude (in mm), extracted for both ASD_A (dark grey/left columns) and TD_A (light grey/right columns) groups. B Cardiac parameters. Histograms represent the mean values (± standard error) of RR interval (in ms), RMSSD (in ms), LF peak frequency (in Hz), HF peak frequency (in Hz), LF absolute power (log scale, in ms²), and HF absolute power (log scale, in ms.²), extracted for both ASD_A (dark orange/left columns) and TD_A (light orange/right columns) groups. C EDA parameters. Histograms represent the mean values (± standard error) of the tonic component of EDA (in µS), EDA AUC (log scale, in µS s), number of EDA peaks, and amplitude of EDA peaks (log scale, in µS), extracted for both ASD_A (dark blue/left columns) and TD_A (light blue/right columns) groups. *p < 0.05 (Color figure online)

We did not find an effect of age as a covariate for median pupil diameter (F₃₇ = 1.71, p = 0.19), hippus amplitude (F₃₇ = 2.15, p = 0.15), or hippus frequency (F₃₇ = 0.29, p = 0.59), nor an interaction group × age for median pupil diameter (F₃₆ = 1.04, p = 0.31), hippus amplitude (F₃₆ = 0.59, p = 0.44), or hippus frequency (F₃₆ = 0.04, p = 0.83).

We observed no significant effect of sex on median pupil diameter (F₃₆ = 0.02, p = 0.86), hippus amplitude (F₃₆ = 2.36, p = 0.13), or hippus frequency (F₃₆ = 1.90, p = 0.17). No group × sex interaction was found either (median pupil diameter: F₃₆ = 0.06, p = 0.79; hippus amplitude: F₃₆ = 0.02, p = 0.86; hippus frequency: F₃₆ = 0.02, p = 0.87).

Cardiac Tone

Cardiac parameters could be evaluated in 19 out of the 21 neurotypical adults (TD_A) and the 21 autistic adults (ASD_A). Cardiac acquisition failed in two participants due to technical problems.

Tendencies for group differences were observed for two parameters (Fig. 2B): RMSSD and HF absolute power. RMSSD tended to be lower for autistic participants (29.2 ± 3.21 ms) than for neurotypical participants (40.1 ± 5.31; F₃₆ = 3.07, p = 0.08) but not after FDR correction (p_FDR = 0.18). Similarly, HF absolute power tended to be lower for autistic adults (2.45 ± 0.10 Hz) than for neurotypical adults (2.72 ± 0.10 Hz; F₃₆ = 3.03, p = 0.08, p_FDR = 0.20). RR interval was significantly lower (i.e., the heartbeat faster) for autistic adults (754 ± 23 ms) than for neurotypical adults (812 ± 22 ms; F₃₆ = 4.48, p = 0.04, η²p = 0.02, CI = [− 125 to − 2.70], but only a tendency after FDR correction p_FDR = 0.09).

LF peak frequency exhibited no difference between neurotypical (0.09 ± 0.005 Hz) and autistic (0.09 ± 0.005 Hz) participants (F₃₆ = 0.49, p = 0.48), nor did HF peak frequency (ASD_A 0.25 ± 0.01 Hz vs. TD_A 0.24 ± 0.01 Hz; F₃₆ = 0.27, p = 0.60), or LF absolute power (ASD_A 2.70 ± 0.07 ms² vs. TD_A 2.76 ± 0.09 ms²; F₃₆ = 0.18, p = 0.67).

A significant effect of age was found on RMSSD (F₃₇ = 7.15, p = 0.01, p_FDR = 0.04, η²p = 0.16, CI = [− 1.59 to − 0.21]) and HF absolute power (F₃₇ = 8.86, p < 0.01, p_FDR = 0.05, η²p = 0.19, CI = [− 0.03 to − 0.007]) showing decreased values with age for both groups, but not on RR interval (F₃₇ = 0.31, p = 0.57), LF peak frequency (F₃₇ = 0.13, p = 0.71), HF peak frequency (F₃₇ = 0.16, p = 0.68), or LF absolute power (F₃₇ = 0.75, p = 0.39). No significant interaction group × age was found on RMSSD (F₃₆ = 7.15, p = 0.40), on HF absolute power (F₃₆ = 0.30, p = 0.58), on RR interval (F₃₆ = 1.38, p = 0.24), LF peak frequency (F₃₆ = 0.05, p = 0.81), or LF absolute power (F₃₆ = 0.35, p = 0.55). A significant interaction group × age was found on HF peak frequency (F₃₆ = 5.85, p = 0.02, p_FDR = 0.04, η²p = 0.14, CI = [− 0.01 to − 0.009]), showing a statistical different evolution of HR peak frequency with age, when we control for the group. The interaction showed decreasing values with age for autistic adults and increasing values with age for neurotypical adults.

We observed no significant effect of sex on RR interval (F₃₆ = 0.94, p = 0.33), RMSSD (F₃₆ = 2.60, p = 0.11), LF peak frequency (F₃₆ = 0.008, p = 0.92), or HF absolute power (F₃₆ = 1.82, p = 0.18). But we did find an effect of sex on HF peak frequency (F₃₆ = 6.01, p = 0.01, p_FDR = 0.04, η²p = 0.14, CI = [− 0.09 to − 0.008]) and LF absolute power (F₃₆ = 17.02, p < 0.001, p_FDR = 0.005, η²p = 0.33, CI = [0.22 to 0.63]). A group x sex interaction was found for RR interval (F₃₆ = 6.59, p = 0.01, only a tendency after FDR correction p_FDR = 0.06, η²p = 0.15, CI = [0.32 to 278], Fig. 3), but not on RMSSD (F₃₆ = 0.23, p = 0.62), LF peak frequency (F₃₆ = 0.17, p = 0.67), HF absolute power (F₃₆ = 0.32, p = 0.57), HF peak frequency (F₃₆ = 1.28, p = 0.26), or LF absolute power (F₃₆ = 0.32, p = 0.57). Post hoc tests revealed that RR interval was higher in neurotypical females compared to autistic females (p = 0.01), and that there was a tendency for higher RR interval in autistic males compared to autistic females (p = 0.08). No difference was observed between TD males and females (p = 0.99).

Fig. 3.

Interaction RR interval × Sex in adults. The graph represents the interaction between mean values of RR interval (in ms) and Sex, extracted for both ASD_A (in orange) and TD_A (in blue) groups. *p < 0.05 (Color figure online)

EDA Tone and Phasic Parameters

Skin conductance parameters could be evaluated in 19 out of the 21 neurotypical adults (TD_A) and the 20 out of the 21 autistic adults (ASD_A). Moreover, 1 neurotypical adult was a complete non-responder, so no number of EDA peaks and no amplitude of EDA peaks could be evaluated.

We found a group effect on tonic component of EDA (ASD_A 4.39 ± 0.64 μS vs. TD_A 1.91 ± 0.21 μS; F₃₅ = 12.32, p < 0.01, p_FDR < 0.05, η²p = 0.26, CI = [1.05 to 3.29]), with autistic adults exhibiting a higher tonus than neurotypical adults (Fig. 2C). Moreover, the number of EDA peaks was significantly higher for neurotypical adults (81 ± 12) than autistic adults (45 ± 7; F₃₅ = 4.81, p = 0.03, η²p = 0.12, CI = [− 62.5 to − 2.44], although marginally significant after FDR correction p_FDR = 0.07). There was no significant effect of the group for EDA AUC (log scale: ASD_A 1.49 ± 0.11 μS⋅s vs. TD_A 1.25 ± 0.14 μS⋅s; F₃₆ = 1.44, p = 0.23), nor for the amplitude of EDA peaks (log scale: ASD_A 0.76 ± 0.13 μS vs. TD_A 0.89 ± 0.10 μS; F₃₆ = 0.21, p = 0.64).

We found no effect of age on tonic component of EDA (F₃₇ = 0.21, p = 0.64), on the number of EDA peaks (F₃₆ = 0.07, p = 0.78), on EDA AUC (F₃₇ = 2.48, p = 0.12), nor for the amplitude of EDA peaks (F₃₆ = 0.81, p = 0.37). We found no interaction group × age on tonic component of EDA (F₃₆ = 0.65, p = 0.42), on the number of EDA peaks (F₃₅ = 0.002, p = 0.95), on EDA AUC (F₃₆ = 0.0006, p = 0.98), nor for the amplitude of EDA peaks (F₃₅ = 0.64, p = 0.42).

We found no effect of sex on tonic component of EDA (F₃₅ = 0.13, p = 0.71), on the number of EDA peaks (F₃₅ = 0.001, p = 0.96), on EDA AUC (F₃₆ = 0.11, p = 0.73), nor on the amplitude of EDA peaks (F₃₆ = 0.004, p = 0.94). We found no group × sex interaction on tonic component of EDA (F₃₅ = 0.34, p = 0.56), on the number of EDA peaks (F₃₅ = 0.14, p = 0.70), on EDA AUC (F₃₆ = 0.14, p = 0.70), nor on the amplitude of EDA peaks (F₃₆ = 0.005, p = 0.94).

Effect of BMI on Autonomic Parameters

There was no BMI difference between the autistic and neurotypical children (ASD_C: 21 [13–35], TD_C: 18 [12–26], p > 0.05; ASD_C-PUP: 18 [14–22], TD_C-PUP: 18 [13–22], p > 0.05; Table S1). As a consequence, we did not explore the influence of BMI on the autonomic parameters in these groups.

However, the autistic and neurotypical adults’ BMI differed (ASD_A: 24 [19–31], TD_A: 22 [17–29], p < 0.05; Table S1). We tested the effect of BMI as a continuous factor on the autonomic parameters within a General Linear Model (GLM). When there was an effect of the group on the autonomic parameters in the first ANOVA analysis, we explored the possible existence of a main effect of BMI or an interaction between BMI and group, completed, when necessary, by post hoc tests. When there was no effect of the group on the autonomic parameters in the first ANOVA analysis, we explored the possible existence of a main effect of BMI that could mask a group effect.

Pupil Tone

When controlling for BMI, we found a tendency of a main effect of BMI on hippus amplitude (F₃₅ = 3.89, p = 0.056, η²p = 0.10), which did not survive to the FDR correction (p_FDR = 0.19). No main effect of BMI or interaction with the group was observed for median pupil diameter or hippus frequency (p > 0.11).

Cardiac Tone

The RR interval was influenced by BMI (main effect: F₃₅ = 4.33, p = 0.044, η²p = 0.11, CI = [− 6.88 to − 0.21]), but without interaction with the group. This result became non-significant after FDR correction (p_FDR = 0.13).

Electrodermal Parameters

For all the EDA parameters, we observed no main effect of the BMI (p > 0.19), but a significant BMI × group interaction (tonic component: F₃₄ = 5.84, p = 0.02, η²p = 0.14, CI = [− 0.07 to − 0.91], marginally significant after correction p_FDR = 0.06; number of EDA peaks: F₃₄ = 6.51, p = 0.01, η²p = 0.16, CI = [− 2.26 to − 19.95], marginally significant after correction p_FDR = 0.06; EDA AUC: F₃₄ = 4.76, p = 0.03, η²p = 0.12, CI = [− 0.009 to − 0.24], but not after correction p_FDR = 0.18; amplitude of EDA peaks: F₃₄ = 9.27, p < 0.01, η²p = 0.21, p_FDR < 0.05, CI = [− 0.05 to − 0.25]), with BMI positively correlated with the EDA parameters in ASD_A group, and negatively in TD_A group.

Correlations of ANS Parameters with Clinical Data in Autistic Participants

In autistic children, no autonomic parameters correlated with CARS scores (p > 0.2). In autistic adults, there was a tendency for a positive correlation between AQ scores and two autonomic parameters: the median pupil diameter (r = 0.44, p = 0.062) and the LF peak frequency (r = 0.40, p = 0.074), but no correlation for the other parameters (p > 0.11).

In autistic children, no autonomic parameters correlated with verbal IQ scores (p > 0.2). In autistic adults, there was a positive correlation between pupil hippus amplitude and verbal IQ (r = 0.52, p = 0.03, CI = [0.06 to 0.79]) that did not survive after FDR correction (p_FDR = 0.14), and no correlation for the other parameters (p > 0.14).

Discussion

Reduced Parasympathetic Tonus in Autistic Children

In children, the only robust difference between autistic and neurotypical individuals was observed on a cardiac parameter, LF peak frequency. LF band parameters interpretation is not consensual as to whether they reflect sympathetic influence, parasympathetic influence or both (Electrophysiology Task Force, 1996; Reyes del Paso et al., 2013; Shaffer & Ginsberg, 2017). Our results will not help shed light on this debate. Several recent reviews have pointed towards mostly a parasympathetic interpretation, in particular when measured at rest, during a short period and in a sitting position (Billman, 2013; Reyes del Paso et al., 2013; Shaffer & Ginsberg, 2017). A lower LF peak frequency in autistic children in our study might thus reflect a reduced PNS regulation on cardiac dynamics. Autistic children would present a reduced parasympathetic regulation compared to neurotypical children, in line with several studies (Cheng et al., 2020; McCormick et al., 2018; Thapa et al., 2019) and summed up in a recent review (Arora et al., 2021) when based on cardiac indices. However, as other PNS indices do not significantly decrease in our results (e.g. HF peak frequency), this finding should be interpreted with caution and confirmed in further studies. Moreover, a more balanced SNS/PNS interpretation of LF peak frequency changes would also be reflected in purely SNS indices (e.g. EDA), which is not the case either.

Authors usually conclude that a PNS under-activation would reflect a SNS hyper-arousal, with the simplistic hypothesis of a SNS/PNS antagonistic balance. Results from other autonomic effectors allow us to refute this hypothesis in our children population. Indeed, we observed no significant difference in SNS tonus (contrary to Bujnakova et al., 2016), indexed by EDA tonic component and median pupil diameter. Pupil diameter at rest in autistic children is inconsistently reported as higher or unchanged compared to their TD peers (e.g. Anderson et al., 2013; Arora et al., 2021; Zhao et al., 2022). The autonomic pattern we describe would thus suggest rather a SNS/PNS co-inhibition functioning (Berntson & Cacioppo, 2007; Berntson et al., 1991) in our autistic children group. It could also point towards a disharmony in the autonomic mobilization of different effectors in autism.

High Sympathetic Tonus in Autistic Adults

In adults, a different ANS pattern was observed. First, cardiac parameters were not significantly different between adults with and without autism, suggesting a rather similar PNS regulation in these populations. However, we observed significant differences for SNS parameters extracted from EDA and pupil recordings. For EDA, the tonus was higher in autistic adults than in their neurotypical peers, while the spontaneous events (phasic) tended to be less numerous. The difference in EDA tonus was confirmed by a significant difference in pupil tonus between these two groups, with globally a higher SNS tonus in autistic adults. As this elevated SNS tonus was accompanied by a tendency of an increased PNS regulation (indexed by RMSSD), our results suggest a SNS/PNS co-activation in autistic adults (Berntson & Cacioppo, 2007; Berntson et al., 1991, 1997). Our results in autistic adults thus do not correspond to the majority of results described in the literature (Arora et al., 2021), even if ANS dysregulation in autistic adults is less consensual (Ben Shalom et al., 2006; Joseph et al., 2008). These results have to be mitigated by the influence of some other factors, not frequently reported. Indeed, we observed in adults an effect of BMI on electrodermal parameters, that might explain our adult groups’ differences.

Atypical Autonomic Maturation in Autism

While in the literature (see Arora et al., 2021 for review) there is no substantial influence of age (children studies vs. adolescents/adults studies) on the overall autonomic pattern in autism, our study is one of the few that tested both age groups in rather similar conditions. While not a longitudinal study, our observations in children and adults could point towards hypothetical maturational trajectories. In a previous paper (Bufo et al., 2022), we showed a higher PNS tone in neurotypical children than in neurotypical adults. This maturation of PNS tonus is less clear in the autistic population, as autistic children already exhibited a lower PNS tone than their peers. On the contrary, we observed a clear maturation effect for SNS parameters. The comparison between adults and children (Figs. 1 and 2) suggests that EDA tonic component would decrease in the neurotypical population while it would stay constant in the autistic population through maturation. While we cannot deny the difference between the two adult groups, we have to be careful about the interpretation of the maturation of EDA values. Indeed, for neurotypical adults, we observed a lower tonic component of EDA (and a higher number of EDA peaks) than in our previous study (Bufo et al., 2022). The slight experimental differences between the two studies were the level of luminance (not expected to influence EDA, but having an impact on median pupil diameter), and the site of EDA recordings (non-dominant hand in the present study but only for adults, right hand—so mainly dominant hand—in Bufo et al., 2022 and in the present children’s data). Picard et al. (2016) have shown that EDA signal is higher when recorded on the right hand, which is coherent with the difference we observe between the present study and Bufo et al. (2022). In neurotypical populations, the tonic component of EDA was comparable in children and adults (Bufo et al., 2022). This would suggest that the difference between autistic and neurotypical adults result from an increased EDA tonus in autistic adults. We would expect to observe an overall higher EDA tonus in adults with EDA recordings on the right hand. For pupil parameters, the only maturation effect we can describe considering the experimental differences between children and adults, is the divergence from neurotypical population with age, going towards a higher SNS tonus in autistic adults. In autism, there would thus be an evolution from a low parasympathetic tonus in children to a high sympathetic tonus in adults.

Compensatory Autonomic Plasticity in Autism: The Tonic and Phasic (Dis)equilibrium Hypothesis

Autonomic functioning at rest is mainly characterized by its tonus. However, spontaneous phasic responses can also occur, indicating an autonomic lability of the individual. This lability has already been described in children compared to adults in neurotypical populations (Bufo et al., 2022). A higher number of spontaneous EDA peaks in neurotypical individuals has been correlated with better allocation of attention towards environmental events in adults (Schell et al., 1988) and better social skills in children (Neuhaus et al., 2015).

The present study is one of the first to assess phasic autonomic responses during rest in autism (Neuhaus et al., 2015). Interestingly, there seems to have a balance between EDA tonus and EDA spontaneous phasic activity at the group level. This is particularly visible in adults, with a higher tonus associated with a lower number of EDA peaks in autistic adults, and the reverse pattern in neurotypical adults. This observation suggests an exacerbated autonomic maturation in autism, as neurotypical adults already present a higher EDA tonus and smaller number of EDA peaks than neurotypical children (Bufo et al., 2022). To sum up with what has been observed on the tonic parameters, there would thus be a maturation in autism from an exaggeratedly low autonomic tonus/high phasic reactivity in children, to an exaggeratedly high autonomic tonus/low phasic reactivity in adults.

This autonomic functional pattern, even outside of the context of autism, raises the question of balance between tonic and phasic SNS mobilization. This pattern could be linked to what has been described of the functioning of the locus coeruleus (LC), the main source of norepinephrine in the brain (Usher et al., 1999), whose neuronal discharge has been correlated to SNS regulation (Mathew, 1995). Indeed, Aston-Jones and Cohen (2005) suggest that optimal behaviour in response to external stimuli would be correlated to phasic LC neuronal discharge, possible only when the LC tonus is intermediate. If the LC tonus is too low or too high, the phasic discharge would disappear (Aston-Jones et al., 1999). Autistic adults would thus exhibit an inappropriately high SNS tonus, that may be linked to non-optimal behaviours, for example hypo- or hyper-responsiveness to external sensory stimulations (Janitzky, 2020). It also questions the possible SNS maturation from autistic children to adults: if autistic children exhibit low SNS tonus but high spontaneous SNS phasic activity, and autistic adults high SNS tonus, could we consider this maturation as a maladaptive neuronal plasticity or an exacerbated compensatory trajectory? Autonomic homeostasis is supposed to be adjusted to our environment and its requirements. Hyperarousal during childhood is costly in energy but fosters developmental acquisitions by seizing environmental opportunities (Mayes et al., 2000), in order to build an adjusted physiological, cognitive, emotional and social behavior. With maturation, autonomic system converges towards a low-cost functioning (lower ANS tonus) associated with more integrated responses to the environment (higher ANS phasic discharges). The pattern we describe both in autistic children and adults suggest an unbalance from the equilibrium observed in the neurotypical population. In autistic children, the tendency for higher spontaneous phasic discharge could reflect atypical reactions to the environment. Maturation processes are the result of a complex interaction between genetics and environmental factors including life experience (Bonnet-Brilhault et al., 2018), associated with more frequent stressful events during autistic developmental trajectory. It has been shown that contextual factors like stress could influence the switch between a phasic and a tonic mode of functioning of the LC (Valentino & Van Bockstaele, 2008), influencing neuroinflammation and potentially neurodegenerative disorders (Janitzky, 2020). This maturational process could thus result in a compensatory atypical autonomic equilibrium in autistic adults. The question of atypical tonic LC activity in autism should thus be studied (Zhao et al., 2022), and its potential link to sensory responsiveness (Janitzky, 2020).

Potential Links with Clinical Symptoms

Fenning et al. (2019) highlight that interactions between SNS and PNS reactivity play a significant role in the development of externalizing problems in children with ASD. Specifically, children are more likely to show externalizing problems when they exhibit low reactivity in both systems or high reactivity in both systems, suggesting that a lack of balance in autonomic regulation leads to greater behavioral challenges. We observed very few links between our autonomic and clinical parameters, possibly due to the heterogeneity of the autism spectrum. One aspect that should be further investigated is the presence of comorbidities that could be linked to autonomic functioning. In particular, autonomic dysregulation in autism has been previously studied in the context of comorbidity with anxiety (e.g. Chiu et al., 2016; Panju et al., 2015; Parma et al., 2021), with autonomic SNS reactivity lower in anxious ASD children. Some authors even proposed that autonomic dysregulation in autism would only be the result of comorbid anxiety (Barbier et al., 2022). To our knowledge, only one study focused on autonomic tonus at rest and anxiety in ASD. By comparing EDA and cardiac measures in groups of children with ASD and anxiety, ASD without anxiety, anxiety without ASD, and TD children, Parma et al. (2021) suggested that a reduced PNS regulation at rest would characterize autism, while SNS regulation differences would characterize anxiety. Our absence of difference in EDA tonus between autistic and neurotypical children would thus suggest that our results are not explained by a difference in anxiety physiological level. To conclude, more and more studies focus on autonomic dysregulation and its possible links with mental disorders, such as major depressive disorder, generalized anxiety disorder, substance use, schizophrenia, etc. (Alvares et al., 2020; Cheng et al., 2020; Huang et al., 2020; Koch et al., 2022; Thapa et al., 2021). All these studies point towards the necessity of systematic analysis of comorbidities in future studies.

Limitations

The ambition of this study was to obtain an overview of autonomic maturation at rest both in neurotypical and autistic populations thanks to simultaneous multiple physiological measures. However, as pointed throughout the manuscript, several limitations must be kept in mind. The first one is that it is a cross-sectional study, limiting the interpretation in terms of maturation. Only a longitudinal study would allow a real description of individual autonomic trajectories, difficult to implement in research to cover such a large age range. Including an adolescent population would help to better characterize the developmental trajectories (while also adding supplementary parameters to consider and thus requiring a very large sample). Our sample size is also limited, reflecting the challenge of clinical research, in particular in vulnerable children populations. Our sample size for each group was based on our final autistic children group. While we describe 22 autistic children, 15 more were included but could not be properly recorded even after familiarization sessions. Difficulties in cooperation depend on age, individual ASD characteristics and comorbidities, and our clinical sample reflects this heterogeneity, in particular when comparing our children and adult groups. Future studies should try to include participants covering the whole autism spectrum across age to eliminate possible confounds, while also controlling for individual parameters such as BMI or smoking status that could impact autonomic measures.

Conclusion

Our study revealed autonomic dysregulation in autism observable at rest. The distinctive profiles in autistic children and adults, both for their SNS and PNS tonic components but also their sympathetic phasic component, could illustrate the two endpoints of an atypical maturation trajectory. During maturation in autistic individuals, compensatory mechanisms would induce autonomic disequilibrium to switch from excessive phasic components to excessive tonic components, possibly subtended by an atypical locus coeruleus functioning. This would require to be confirmed via a longitudinal protocol allowing to directly assess the maturation trajectory. Studying autonomic regulation in neurodevelopmental disorders like autism is thus crucial to understand these disorders, but should always consider the developmental trajectory and possible compensatory mechanisms.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADI-R

Autism diagnostic interview-revised

ADOS

Autism diagnostic observation schedule

ANS

Autonomic nervous system

AQ

Autism quotient

ASD

Autism spectrum disorder

AUC

Area under the curve

BMI

Body mass index

CDA

Continuous decomposition analysis

CIM-10

International Classification of Diseases and Related Health Problems-10th Revision

DSM-5

Diagnostic and Statistical Manual of Mental Illnesses-5th Revision

ECG

Electrocardiogram

EDA

Electrodermal activity

GLM

General linear model

HF

High frequency

HRV

Heart rate variability

IQ

Intellectual quotient

LC

Locus coeruleus

LF

Low frequency

PNS

Parasympathetic nervous system

RMSSD

Root mean square of successive differences between normal heartbeats

RSA

Respiratory sinus arrhythmia

SNS

Sympathetic nervous system

TD

Typically developing

Author Contributions

Maria Rosa Bufo: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing—original draft and review and editing, Visualization; Marco Guidotti: Resources, Writing—review and editing; Mathieu Lemaire: Conceptualization, Methodology, Investigation, Resources, Project administration, Funding acquisition; Joëlle Malvy: Resources; Emmanuelle Houy-Durand: Resources, Writing—review and editing; Frédérique Bonnet-Brilhault: Conceptualization, Resources, Writing—review and editing; Frédéric Briend: Investigation, Data curation, Writing—review and editing, Project administration; Nadia Aguillon-Hernandez: Conceptualization, Validation, Writing—original draft and review and editing, Supervision, Project administration, Funding acquisition; Claire Wardak: Conceptualization, Validation, Writing—original draft and review and editing, Supervision, Project administration, Funding acquisition.

Data Availability

Data available upon request to the corresponding author.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical Approval

The protocol was ethically approved by the Comité de Protection des Personnes (CPP; Protocol PROSCEA 2017-A00756-47). All the participants were recorded in the same conditions, in the University Hospital of Tours, France.