Inhibition, flexibility, working memory and planning in autism spectrum disorders with and without comorbid ADHD-symptoms

General Procedure

The participants were recruited from our inpatient and outpatient department of child and adolescent psychiatry, while the healthy control group consisted of healthy siblings of the patients or were other children interested in taking part. All new referrals with suspected ADHD or ASD underwent an extensive child psychiatric examination, which was conducted by an experienced child and adolescent psychiatrist according to DSM-IV-TR criteria. Additionally standardized psychopathological measures were used (see diagnostic measures section). IQ was measured using the Culture Fair Intelligence Test, a non-verbal one-dimensional IQ-test [28]. The diagnosis of ADHD in the ASD+ group was given before recruitment. All children from the ASD+ group met full DSM-IV-TR criteria and were excluded if they had subthreshold ADHD characteristics. They furthermore fulfilled the age and the pervasiveness criterion as required in DSM-IV-TR.

Informed parental consent was obtained for all participants, and the study was approved by the Medical Ethical Committee of the University of Cologne. All children were tested individually in the Department of Child & Adolescent Psychiatry in a quiet room by one of the two researchers. The person testing was blind with regard to the ADHD diagnosis of the autistic participants. Testing was carried out within a larger study that comprised a two-hour session. EF tasks were presented in a fixed order (Inhibition, SWM, SOC and ID/ED) approximately after the implementation of the first half of the test. Due to the small simple size we decided not to counterbalance the order of the test. Participants were informed that they could discontinue testing at any time and were given positive comments throughout the testing. The parents or caregivers were sent detailed reports on their child's performance on the tests.

Diagnostic measures

The diagnosis of autistic disorder was made using the Autism-Diagnostic Interview-Revised (ADI-R; Cut-offs: Impairment of Social Interaction = 10; Impairment of Communication = 8; Stereotyped Behavior = 3) and the Autism Diagnostic Observation Scale (ADOS, Cut-offs: Communication and Social Interaction = 7 (Module 1), 12 (Module 2), 10 (Module 3+4)) [29–32]. Furthermore the Diagnostic Checklist for Pervasive Developmental disorders (DCL-TES) was applied, mainly to exclude ASD in the ADHD children and to differentiate between HFA and Asperger syndrome within the ASD group [33]. Additionally the Diagnostic Checklist for Oppositional Defiant or Conduct disorders (DCL-SSV) was used to have a dimensional description of ODD-symptoms in both the ASD and the ADHD group.

The diagnosis of attention deficit/hyperactivity disorder was made using the Diagnostic Checklist for Hyperkinetic Disorders/ADHD (DCL-HKS). Similar rating scales have been developed in the United States, based solely on DSM-IV criteria for ADHD [35]. The number of DSM-IV criteria fulfilled was provided, as was the severity score for each item ranging from 0 to 3 [34]. The checklists were applied as an interview with parents and teachers. All three checklists are made up of components of the Diagnostic System for Mental Disorders in Childhood and Adolescence (DISYPS-KJ) based on ICD-10 and DSM-IV and allow the assessment of a dimensional score and a categorical diagnosis [33]. The cut-offs of the checklists correspond with the criteria that have to be fulfilled according to ICD-10 and DSM-IV.

The groups ASD + and ADHD show an equal profile without statistically significant differences in the DCL-HKS scores with regard to the criteria of hyperactivity, inattention and impulsivity. Scores of the DCL-TES were as expected high for both ASD groups and low for the ADHD and the TD group. The scores are illustrated in Figure 1 and 2, separately for the four different groups.

Additional comorbid disorders (emotional disorders, OCD, enuresis and encopresis as well as ODD and OCD) were assessed using the Kiddie-SADS – Lifetime-Version (K-SADS-PL) [36]. No relevant comorbid disorders except of ODD, especially no learning disabilities, were found in any of the four groups. The K-SADS was also used to additionally confirm the ADHD diagnosis.

Experimental Procedure

Statistical Analysis

The statistical analysis was carried out using the SPSS for Windows Program Version 14.0.

In order to examine group differences between the groups a MANOVA with all EF parameters of the four paradigms as the dependent measures and the group as the between-subject variable and additional post hoc Scheffé tests were calculated. Due to the small sample sizes effect sizes, according to Cohen, were calculated in order to examine group differences between the four groups for the EF parameters of the four paradigms [42]. Since a large number of statistical tests were performed, significant results may have capitalised on chance and the overall probability of a type I error likely exceeded 5%. In the case of a priori predictions, Howell argues that correction for multiple comparisons is not warranted [43].

Next, Pearson correlations were carried out between the different executive variables and the values of the DISYPS ADHD scales (ADHD total score, inattention, hyperactivity and impulsivity) and the DISYPS ASD scales (ASD total score, mean impairment of social interaction, mean impairment of communication and mean stereotype behavior) for the four diagnostic groups.

Executive function tests

Planning Task (Stockings of Cambridge, SOC)

There was a medium effect size between the groups ASD- and ADHD (d = 0.6). All clinical groups needed more time between the subtasks. There was a high effect size between the groups ASD- and the control group (d = 0.6). Participants of the ASD+ group had a longer test duration than those of the TD group (d = 0.6).

There were no significant group differences for any of the tasks.

The Z score plots with medium and high effect sizes between the four groups are shown in Fig. 3.

Relationship between EF and ADHD/ASD symptoms

In addition, Pearson product-moment correlations separated for the two groups affected by ADHD-symptoms were used to examine the relationship between all dependent measures of the neuropsychological paradigms and the clinically observed ADHD symptoms (inattention, hyperactivity, impulsivity and ADHD total score) measured with the DCL-HKS.

The ADHD group showed only small, but significant correlations for the variable "flexibility errors" with inattention (r = -0.5, p < .03) and for the variable "flexibility test duration" with inattention (r = -0.5, p < .02). In the ASD+ group, the variable "inhibition median" correlated significantly with the total score ADHD (r = -0.6, p < .01) and impulsivity (r = -0.6, p < .03), the variable "inhibition hits" with inattention (r = -0.6, p < .01), the variable "inhibition false alarms" with hyperactivity (r = -0.5, p < .02) and the total score ADHD (r = -0.5, p < .02) as well as the variable "inhibition omissions" with inattention (r = -0.5, p < .04).

Furthermore the ASD+ group showed small, but significant correlations for the variable "flexibility errors" with inattention (r = -0.6, p < .03) and the variable "working memory test duration" and the total score ADHD (r = 0.5, p < .02).

There were no significant correlations between different executive variables and the values of the ADHD scales in the ASD- and in the TD group.

In contrast to the ADHD symptomatology, we tested the relationship between EF and autistic symptoms (mean impairment of social interaction, mean impairment of communication, mean stereotype behaviour and total score ASD) with the help of Pearson product-moment correlations for the four groups. We found significant correlations in the ASD – group for the variable "inhibition hits" with impairment of social interaction (r = -0.7, p < .000) and ASD total score (r = -0.5, p = .02), for the variable "inhibition omissions" and all ASD subscale scores (impairment of social interaction: r = 0.6, p = .004; impairment of communication: r = 0.4, p = .03; stereotype behaviour: r = 0.5, p = .02; ASD total score: r = 0.5, p = .006) as well as for the variable "flexibility test duration" and "flexibility stages" with stereotype behaviour (r = 0.5, p = .03; r = .51, p = .02).

The ASD+ group showed significant correlations for the variable "flexibility test duration" and "flexibility stages" with stereotype behaviour (r = -0.5, p = .03; r = -0.4, p = .04).

In the ADHD group we found significant correlations for the inhibition paradigm (hits/stereotype behaviour: r = 0.4, p = 0.4, errors/impairment of social interaction: r = -0.4, p = .03; errors/ASD total score: r = -0.5, p = .02; omissions/stereotype behaviour: r = 0.6, p = .007).

There were no significant correlations in the TD group.

Inhibition task (Go/NoGo-Task)

The expectation that the ADHD children would be more impaired in inhibitory control, especially with regard to errors of omission and commission, was confirmed for the Go/NoGo-task for all variables. Our initial prediction that ASD+ children would show inhibition deficits according to an additivity hypothesis was partly confirmed, as these children showed worse performance making more errors of omission and less hits compared to the healthy control children. Our results are partly in line with findings of previous studies. Ozonoff and Happé found more deficits in response inhibition for ADHD children than for autistic children comparing ADHD and ASD groups, whereas the Nyden, Johnson and partly the Geurts study revealed also deficits for ASD children [12–17]. However less severe inhibition deficits in children with ASD were consistently found in all the above mentioned studies, except of the one by Ozonoff et al. and Goldberg et al. who applied a stroop task [12, 15]. Our results confirm a suggestion of Goldberg et al. to better use non-verbal measures (e.g. a Go-NoGo task) to differentiate between ADHD and ASD. A study by Christ et al. assessing children with ASD with different inhibitory tasks revealed that the stroop task didn't lead to inhibition deficits compared to a flanker and a GoNo/Go task [47]. The authors argue that referring to a model by Casey et al., it is possible that the integrity of some but not all neural circuits subserving inhibitory control are compromised in children with ASD [48, 49].

An interesting finding of our study is that comorbid ADHD symptoms seem to worsen inhibition performance in ASD children with comorbid ADHD, as pure ASD children performed rather well in comparison to the pure ADHD group in our study. Previous studies including comorbid ADHD symptoms couldn't find differences between ASD and ADHD children and vice versa. This underlines the importance of taking into account severe inattention and hyperactivity problems warranting an ADHD diagnosis in ASD children when interpreting inhibition data. If comorbid ADHD symptoms are not statistically referred to individual inhibition problems of children with ASD might not be detected due to a too large heterogeneity of the samples.

Intra-Dimensional/Extra-Dimensional Shift Task (ID/ED)

The results for the flexibility task show differences on the basis of effect sizes. The ASD- group made less errors than the control group. As also discussed by Happé one reason for the absence of differences between the ASD and the TD group might be the high proportion of participants with Asperger syndrome [16]. With regard to the variable stages the ASD+ group even performed better than the TD group. To the authors' knowledge there are to date no studies showing that children with Asperger syndrome are better in cognitive flexibility than typically developing children. The ASD+ group, as was the case in the planning task, showed difficulty with test duration compared to the control and the ASD- group. As also the ADHD group had a longer test duration it seems that time is a key problem for those children affected by ADHD-symptoms. Studies using the same task of the CANTAB also failed to find significant differences in post-hoc tests [15, 16]. Interestingly, Ozonoff et al. pointed out that not all types of attention-shifting are impaired in ASD, only those that require prefrontal cortical function [12]. An analysis of performance at different cognitive levels of the same flexibility task of the CANTAB revealed no impairment on higher levels when shifting between categories or when rules were required. This kind of analysis was not applied neither in our study nor in the Goldberg or Happé study. Geurts et al. describe for the HFA group slower mean reaction times on a different flexibility measure (change task) and a higher percentage of perseverative responses in the WCST (Wisconsin Card Sorting Test) [14]. Perseveration itself is not measured with the ID/ED Task of the CANTAB.

Spatial Working Memory Task (SWM)

Even though both groups affected by ADHD needed more time to perform the working memory task, especially in the ADHD-group medium to high effect sizes are apparent with poorer performances in comparison to the control group, whereas pure autistic children also made more errors than the healthy children. Happé et al. also describe deficits on the same working memory task for the ADHD group but not for the ASD group, whereas Goldberg et al. found deficits for both groups with poorer performance of the autistic children [15, 16]. One reason why our results are different from the Goldberg study might be that in their study, the ASD children had significantly lower IQs than the ADHD children, whereas we used IQ-corrected z-scores. The result by Geurts et al. who found no differences between an ADHD, ASD and a control group for a different working memory task (self-ordered pointing task) could not be replicated [14].

As we found only small problems of working memory in both ASD groups, comorbid ADHD symptoms don't seem to play the key role in working memory performance. One could argue that there were no effects in the ASD groups due to a wide age range in our study (6–18 years) as age-related improvement for the working memory task was described in particular for ASD children in an analysis of developmental age [16]. This difficulty was eliminated by using age-corrected z-scores in our study.

Planning Task (Stockings of Cambridge, SOC)

With regard to the performance in the planning task, our results are difficult to interpret. Even effect sizes revealed only medium effects for the ASD+ group concerning duration of the task and medium effects for the mean initial thinking time and the mean subsequent thinking time especially for the ASD- group. ASD+ children also showed deficits on these variables. These results are partly in line with a study by Ozonoff et al., who applied the CANTAB planning task in a large sample of 79 autistic individuals, finding no differences for the mean initial thinking time, but for the mean subsequent thinking time [12]. The number of solved problems was not affected by this and replicates the results of a study by Goldberg et al., who also failed to find group differences in the number of problems solved using the same paradigm [15]. Although the Geurts group used the Tower of London as a different measure of planning abilities they also found significant differences for execution time, with worse performance in a pure autistic group [14]. Thus, planning difficulties for ASD individuals might be less a problem of comprehension than of speed.

Relationship between EF and ADHD/ASD symptoms

Our second aim was to examine relationships between EF and clinical symptoms of ADHD and ASD. In our study clinically observed ADHD symptoms don't correlate with EF deficits in ADHD children, whereas inhibition performance shows an interaction of comorbid ADHD symptoms in ASD children on all measures, especially with the symptom of inattention. Also test duration seems to be influenced by ADHD symptoms in this group.

Interestingly even though there were low ASD scores in the ADHD sample, inhibitory parameters are associated with ASD symptoms in the ADHD group. Correlations of ASD symptoms and EF don't seem to follow a fixed pattern in the ASD groups.

These results underline the difficulty of bringing together clinically observed behaviour and neuropsychologically measured EF functions. This indicates a need for caution when attempting to transfer laboratory outcomes to daily life.

Due to the small sample size, this investigation can only be seen as a descriptive attempt to approach the problem of influence of attention disorders and increased impulsivity and hyperactivity as postulated for example by Geurts et al. [14]. However the fact that though some aspects of group differences could not be shown by the analysis of variance, it is obvious that there is a high amount of medium and high effect sizes describing to a certain degree differences between the four groups. Therefore it can be hypothesized that the statistical power in our study is too small and that within a larger sample existing differences might be proved as being statistically significant.

Finally the characteristics of our study sample (age- and IQ-correction, inclusion of ADHD in the ASD group) limit the comparability of these findings with respect to other research reports with differently characterized samples.

One reason for our decision to control for IQ, well knowing that there is a current controversy about this topic, was the fact that we partly wanted to avoid issues concerning late maturation of the frontal lobes and the close overlap between the constructs of EF and fluid intelligence [50, 51]. Thus, as also argued by Happé et al., findings from studies not controlling for IQ are difficult to interpret [16]. Furthermore especially in ASD samples a number of high EF tasks have shown deficits in low-but not in high-functioning groups with ASD. Finally Hill & Bird point out that the approach of comparing data between single groups is problematic since individuals differences are large and requires that all individuals are homogeneous [3]. Controlling for IQ thus reduces the heterogeneity of the participants.