Relationship of anxiety to core symptoms of autism

Several recent lines of work have explored the relationship between core symptoms of repetitive behavior in ASD and associated anxiety-related constructs. Rodgers and colleagues [6] found evidence for a relationship between parent reports of elevated anxiety in ASD children and increased repetitive behaviors, especially with ‘insistence on sameness’ in routines and interests. Rodgers and colleagueshypothesize that the association between anxiety and repetitive behavior in ASD may be mediated at least in part by an ‘intolerance of uncertainty’ that arises due to atypical information processing, including basic sensory processing, that is common in ASD. Intolerance of uncertainty (IU) underlies a perceptual bias associated with beliefs about worry that become more potent in ambiguous contexts [7]. A substantial literature implicates IU in the maintenance of generalized anxiety disorder (GAD). IU may arise from atypical information processing, including basic sensory processing, common to ASD. This atypical information processing in ASD impairs integration of environmental cues, leading to increased perception of ambiguity in ASD across multiple contexts. This in turn leads to increased worry and subsequent anxiety, including generalized anxiety.

Relatedly, two recent studies reported atypical responding in reversal learning paradigms. South et al. [8] paired one of two simple cues (colored squares) with a puff of air delivered to the neck in a sample of 30 older children and adolescents diagnosed with ASD and an age- and IQ-matched typically developing youth. Both groups demonstrated comparable acquisition of a classical fear conditioning measured through differential skin conductance responding to threat versus safety cues. However, the ASD group was significantly delayed in reacquiring fear to the reversed threat and safety cues. Moreover, significant correlations emerged between psychophyiological measures of reversal learning - when relatively simple rules for learning an experimental task are switched without warning - and questionnaire-based reports of repetitive behavior in ASD samples. These associations were unique to repetitive behaviors and were not found for other symptom domains. Similarly, D’Cruz et al. [9] reported behavioral results from a reward-based probabilistic learning task among 41 ASD participants including children and adults with matched controls. Participants chose a stimulus based on location with 80% of correct choices and 20% of incorrect choices randomly reinforced. After making the correct choice over multiple trials, the rewarded stimulus location changed without warning (reversal). D’Cruz et al. also observed intact contingency acquisition of the task in ASD but reported that, while the ASD participants tended to initially make correct post-switch choices, they ‘regressed’ to previous choices more often than controls.

Together these studies suggest that straightforward learning of initial rules or contexts is often intact, at least in high-functioning ASD samples, but that learning may break down under more complex situations, those with increased ambiguity and uncertainty. A recent review of the literature by Gaigg [10] argues that atypical emotion processing early in life in ASD disrupts a range of cognitive processes, including social cognition. As noted above, both the South et al. [8] and D’Cruz et al. [9] studies observed associations specifically with measures of everyday behavioral rigidity. Nonetheless it is possible these may also reflect indirect effects on social function, because social interactions are inherently dynamic. Attenuated social perceptual skills in ASD [11] may accentuate the unpredictability and aversiveness of social situations. For the present study, we utilized a paradigm that was created explicitly for the study of uncertainty in other clinical populations [12], in order to directly test the psychophysiological consequences of contextual uncertainty in ASD.

Associated variables

Participants completed several measures of anxiety and rigidity via a digital survey. These included the Spence Children’s Anxiety Scale-child report (SCAS-C; [24], a reliable and valid measure of child anxiety that reports both an overall total score and six subscale scores). We were most interested in the GAD subscale, which includes items such as ‘I worry about things’ and ‘When I have a problem, I feel shaky’. We used a shortened, 12-item version of the Intolerance of Uncertainty Scale (IUS) [S. Walker, ‘What do we know about the relationship between intolerance of uncertainty and worry in young children?’, unpublished PhD thesis] including questions such as ‘When things happen very suddenly, I get upset’ and ‘Feeling unsure stops me from doing most things’. One parent of each participant completed a parent-report version of the same measures (that is SCAS-P, IUS-P). We also examined a key measure of behavioral rigidity from the SRS, the Autistic Mannerisms subscale (for example, ‘when under stress, he or she shows rigid or inflexible patterns of behavior that seem odd’ and ‘has more difficulty than other children with changes in his or her routine’). In order to test specificity of these hypothesized associations, we also tested general associations with social disability (the SRS Total Score), IQ, and participant age.

Stimuli and apparatus

Stimulation was controlled by E-prime 2.0 Professional Software (Psychology Software and Tools, Inc., Sharpsburg, PA, USA). The acoustic startle stimulus was a 40-ms duration, 103 dB burst of white noise presented binaurally through headphones. Participants wore a vest that closed with Velcro straps and had a ½ inch (~2.54 centimeter) firm-yet-flexible hose threaded up through the vest and adjusted to point toward the junction of the neck and the chin, with a gap of about 8 centimeters between the end of the tube and the skin. The approximately 3-meter- long tube was connected to a tank of compressed ‘room’ air with a medical-grade valve and regulator set to deliver a puff of air to the neck at 75 psi (measured at the regulator valve). E-prime software controlled an electronic switch that delivered each puff for 100 ms. Prior to the start of the experiment, participants received two puffs of air to familiarize them with the stimulus and give them the chance to opt out if they wanted. Participants were also reminded they could stop the experiment at anytime if they chose to.

The startle reflex was measured using two 8 millimeter silver-silver chloride electrodes placed beneath the left eye, over the obicularis oculi muscle, one centered beneath the pupil and a second positioned immediately lateral to that toward the outer canthus. A third, grounding electrode was placed on the left forearm. Prior to placement of the electrodes, the skin was cleaned and then abraded using NuPrep skin gel (Weaver and Company, Aurora, CO, USA). Electrodes were attached to a BIOPAC MP150 EMG100C module (BIOPAC Systems, Inc., Goleta, CA, USA), and data was recorded on a Macintosh MacBook Pro laptop (Apple Inc., Cupertino, CA, USA) using AcqKnowledge 4.0 software (BIOPAC Systems, Inc.). The electromyogram was acquired with amplifier bandwidth set at 28 to 500 Hz with a sampling rate of 2,000 Hz.

Experimental design

The experimental design followed the protocol published by Schmitz and Grillon [12] in Nature Protocols. Testing consisted of a beginning habituation phase of nine startle probes (103 dB, 40 ms duration), followed by two recording blocks separated by a 5-minute break. Each block consisted of four initial habituating startle probes followed by three neutral (N), two predictable (P), and two unpredictable (U) conditions. These conditions were presented in one of the following orders: P N U N U N P or U N P N P N U. Block order was randomized across participants with 9 CON (of 21) and 8 ASD (of 18) participants receiving the block beginning with the P condition first. A diagram of the protocol is presented in Figure 1.

Open in a separate window

Figure 1

Schematic representation of sequences of stimulus presentation during each condition in one block of the NPU-threat test.The upper part of the figure represents a complete block including two P (predictable), two U (unpredictable) and three N (no shock) conditions. The lower part shows examples of each condition, including startle probes, cues (8 s duration) and shocks. Used by permission from [12].

Each condition (that is N, P, and U) lasted approximately 120 s during which time a cue (green circle for N condition; red square for P condition; blue triangle for U condition) was presented three times for 8 s each time. When the cue was absent, a white fixation cross was present on the screen the entire time to help maintain the participant’s attention. In addition to the cue or fixation cross, the top of the monitor displayed a reminder of the condition for the participant. These read ‘No puff’, ‘Puff only during red square’, and ‘Puff at any time’, for the N, P, and U conditions, respectively. Because this was not a conditioning experiment, the reminder was necessary to ensure that differences in response to the startle probe in each condition were not due to differences in learning across participants.

Startle probes were administered six times during each 120 s condition: during each of the three cues, and during each ITI. Startle probes were separated by at least 20 s. Air puffs were administered once or twice per P (0.5 s before the end of the cue) and U condition (pseudo-randomized during the absence of the cue) resulting in six air puffs per block and twelve air puffs throughout the course of the experiment. Schmitz and Grillon [12] have shown previously that absence of a shock during the cue in the U condition does not lead to safety learning.

Between blocks, participants had a 5-minute break to rest. During this time participants were asked to report their anxiety level retrospectively during each of the three conditions, with and without the cue present. Ratings were reported using Qualtrics online survey software (Qualtrics, Provo, UT, USA). Each question displayed the image associated with that condition (P, U, or N) either with or without the cue, and included a text-based numbered rating scale from 1 to 10, with 1 labeled as ‘not anxious/fearful’ and 10 labeled as ‘extremely anxious/fearful’. The same survey was repeated following the completion of the second block.

Psychophysiological response

T scores of blink magnitude during the cue phases showed the expected main effect of condition, (F(2,70) = 62.03, P <0.001, η² = 0.64). The a priori prediction (see [13]) of P >U >N was confirmed and the linear contrast was significant (F(1,37) = 120.983, P <0.001, η² =0.77). As can be seen in Figure 2, responses were identical for both groups following the 2 × 3 ANOVA. There was no main effect for group (F(1,35) = 0.00, P = 0.96, η² = 0.00) or group x condition interaction (F(2,70) = 0.18, P = 0.83, η² = 0.01). A similar 2 × 3 ANVOA for the ITI (uncued) phases likewise demonstrated a significant main effect for condition (F(2,70) = 15.45, P <0.001, η² = 0.31). As noted above, for the ITI phase it is expected that the contextual anxiety will be highest in the between-cue times, when the timing of the air puff is uncertain, U >P >N. This expected pattern was confirmed (Figure 3) and the linear contrast was significant (F(1,37) = 32.61, P <0.001, η² = 0.47). There was again no main effect for group (F(1,35) = 0.41, P = 0.53, η² = 0.01) or group x condition interaction (F(2,70) = 0.84, P = 0.44, η² = 0.02). Because of the imbalance in the ratio of males to females across groups we reran these analyses for main effects and interactions with only males in each group; results were identical in showing significant main condition effects but no group x phase interactions for both cued and ITI conditions.

Open in a separate window

Figure 2

EMG magnitude to startle probes during cued phases shows expected main effect for condition type (P >U >N) with a similar between-group response.

Open in a separate window

Figure 3

EMG magnitude to startle probes during the uncued intertrial intervals shows the expected main effect for condition type (U >P >N) with a similar between-group response.

Next, we performed the same analyses using raw blink magnitude instead of t scores. The relative pattern of responses for EMG magnitude across the N, P, and U phases was identical for the t score analyses of both cued and ITI conditions. However, there was an additional, significant main effect for diagnostic group in both cued and ITI phases. Interestingly, the ASD group showed overall increased magnitude relative to controls: cued condition F(1,35) = 5.50, P <0.05, η² = 0.14; ITI F(1,35) = 4.59, P <0.05, η² = 0.12. In order to ascertain whether this group difference was due to differing task-related response or was present more generally, we analyzed both baseline and post-break time habituation phases: as shown in Figure 4, the ASD group showed substantially increased responses during both habituation phases (baseline t(35) = 2.32, P <0.05; post-break t(35) = 2.92, P <0.01).

Open in a separate window

Figure 4

Absolute startle magnitude during habituation phases shows the ASD group is significantly more reactive to the white noise probes outside of the task threat context.

Given these differences in baseline, we redid the analyses of raw blink magnitude using response to the first habituation phase as a covariate (suggested by [13] as an alternate to using standardized t scores). No results changed in this analysis. For the cued phase, the main effect held (F(2,70) = 3.29, P <0.05, η² = 0.09) with no main effect for group (F(1,35) = 0.64, P = 0.43, η² = 0.02) or group x condition interaction (F(2,70) = 0.20, P = 0.82, η² = 0.01). This was likewise true for the ITI phase: main effect (F(2,70) = 5.09, P <0.01, η² = 0.13) with no main effect for group (F(1,35) = 0.25, P = 0.62, η² = 0.01) or group x condition interaction (F(2,70) = 0.12, P = 0.89, η² = 0.00).

Latency

Results for the analyses of latency-to-peak showed the identical pattern as findings from the cued magnitude scores: condition (F(2,70) = 6.87, P <0.01, η² = 0.16), P >U >N linear contrast (F(1,37) = 16.17, P <0.001, η² = 0.30); group effect (F(1,35) = 0.03, P = 0.86, η² = 0.00); group x condition interaction (F(2,70) = 1.28, P = 0.28, η² = 0.04). In the ITI condition, latency-to-peak analyses showed no significant effects: condition (F(2,70) = 1.41, P = 0.25, η² = 0.04, group (F(1,35) = 1.10, P = 0.30, η² = 0.03); group x condition interaction (F(2,70) = 0.37, P = 0.68, η² = 0.01).