Dysfunctional processing of reward has been found to be prevalent in ADHD and in externalizing disorders. The “trait impulsivity” model, for example, postulates that early emerging impulsivity is a crucial vulnerability factor (externalizing liability) and indicative of a developmental pathway from ADHD to ODD and other externalizing disorders such as CD and substance use disorder (6). The impulsivity concept combines the two components of high subcortically mediated (bottom-up) approach motivation and low top-down inhibitory (cognitive) control (IC) mediated by the forebrain (7–9). In a further model, Blair et al. (2) proposed that deficient decision making, which comprises a high risk of impulsivity, represents the lowest common denominator for conduct problems. Similar to the trait-impulsivity model, the dysfunction is thought to involve subcortical bottom-up processes and prefrontal top-down control, and to be present in children showing multiple facets of externalizing problems, including children with ADHD, ODD/CD, and CU traits.

There is broad empirical evidence of low reward-related IC (RRIC) and other executive function deficits in children with ADHD as well as those with ODD/CD (10, 11). Low RRIC, dysfunctional reward-related decision making, and “delay aversion” have been assumed to characterize an ADHD subtype (12) with comorbid ODD/CD symptoms (13–15). CU traits, however, have sometimes been assumed to be associated with a rather good inhibitory control capacity and fewer cognitive deficits (4, 16, 17), making it possible that children with ADHD symptoms and comorbid CU traits show fewer RRIC deficits. Research on this issue is sparse. In particular, there are very few studies on the association between CU traits and IC in the context of ADHD development.

There is relatively broad evidence that children with ADHD show cortical arousal deficits. Cognitive deficits of the disorder have been assumed to be caused by difficulties in regulating arousal according to situational demands (1, 18, 19). Cortical and peripheral sympathetic arousal are linked via the locus coeruleus and the brain norepinephrine system. Measures of sympathetic electrodermal activity (EDA) have thus been taken as indicators of the arousal regulation dysfunction in ADHD (18). Bellato et al. (18) systematically reviewed the results of 55 studies on autonomic nervous system function in ADHD, and found that children and adolescents with ADHD showed hypoarousal (indicated, e.g., by low EDA). The authors concluded, however, that reactivity to rewarding, emotional stimuli, as well as the role of comorbidity, have not yet been sufficiently studied.

It has been shown that low resting-state sympathetic arousal in children with ADHD can be caused by comorbid ODD/CD problems, low anxiety, and psychopathic personality traits (20, 21). These characteristics overlap with CU traits (4, 22). Studies analyzing sympathetic arousal (using the cardiac pre-ejection period) during reward-related tasks found associations with comorbid ODD: Tenenbaum et al. (23) compared healthy children and children with ADHD, ADHD+ODD, and ADHD+CD regarding their sympathetic activity during a risky decision-making task and found the lowest activity in children with ADHD+ODD. In a sample of children with ADHD, Beauchaine et al. (24) found that sympathetic activity during a rewarded simple-matching task was associated with parent-reported conduct behavior problems. However, Conzelmann et al. (25) compared the electrodermal reactivity (EDR) to neutral, positive, and aversive stimuli between unmedicated boys with ADHD and healthy controls, and found lower EDR in the boys with ADHD in all three conditions, irrespective of comorbidity. As these studies did not assess CU traits, it is possible that comorbid CU traits and/or ODD symptoms explain low arousal in response to reward-related tasks.

Based on the research reviewed above, we examined the following hypotheses: (a) Low RRIC is associated with developing ADHD. This association overlaps with (i.e., can be explained by) comorbid ODD symptoms. We do not hypothesize an overlap with CU traits, as this top-down control component of impulsivity might not be impaired in children with CU traits. (b) High reward-related approach behavior is associated with developing ADHD. This association can be explained by ODD symptoms and CU traits, as this bottom-up component of impulsivity can be expected to be common to all three psychopathological domains. (c) Low autonomic reactivity to reward-related stimuli is linked to ADHD. This association overlaps with ODD symptoms and CU traits.

A sample of 198 preschool children (115 boys, 58%) was recruited from childcare facilities. The children were 4–5 years old (T1; m = 58, s = 6 months) at the first assessment wave and 8 years old (T2; m = 101.4, s = 3.65 months) at the second wave. Inclusion criteria were: IQ>80, lack of motor and sensory disabilities, lack of chronic physical and mental diseases, no indication of a trauma experienced by the child, and no continuous pharmacological treatment. To determine eligibility, a telephone interview and a screening questionnaire on the ADHD symptoms of the child [FBB-ADHS-V by (26), see below] were used. Children with high ADHD symptoms were oversampled. Of the 198 children, 179 participated in the 8-years assessment (retention rate of 89%). There were no differences between children who participated in the 8-years assessment and those who dropped out with respect to gender (Chi2 = 0.22; t = 0.00) and age of the child, ADHD symptoms, symptoms of anxiety/depression, and oppositional symptoms of the child (t-scores between −1.78 and 0.96).

At T1, all children were medication naïve. At 8 years (T2), three children were medicated with methylphenidate, and were therefore excluded from the analyses of the 8-years data. Table 1 contains descriptive data of the sample. Parents gave their written informed consent to participate in the study, and received an expense allowance of 50 Euros at T1 and 70 Euros at the T2 assessment. The study was approved by the Ethics Committee of the Medical Faculty, University of Marburg.

Correlation coefficients among the study variables were calculated for descriptive purposes. In those cases where gender of child was significantly associated with a neuropsychological/physiological predictor variable, we adjusted for gender in all respective analyses.

To test the hypotheses on the associations of (a) T1 and T2 RRIC, (b) T1 and T2 approach tendency and (c) T2 electrodermal reactivity with T2 ADHD diagnosis, we conducted logistic regression analyses. In analyses (a) and (b), we adjusted for T1 ADHD symptoms (model 2) to assess whether the predictor variables predict T2 ADHD over and above T1 ADHD symptoms. In all analyses (a, b, c), ODD and CU symptoms were covaried in model 3. For control purposes we additionally adjusted for anxiety/depressive symptoms and the approximated verbal IQ of the child in model 4.

In a next step, for the significant predictor-ADHD links, we tested whether ODD and/or CU symptoms significantly explain this link (i.e., the common variance between the predictor variable with the T2 ADHD diagnosis). For this purpose, we partitioned the total predictor-ADHD link (common variance) into the direct link (common variance between predictor and ADHD not explainable by the comorbid dimension) and the indirect link (i.e., common variance explainable by the comorbid dimension) using ordinary least squares (OLS) regression, and tested these links using the bootstrapping method recommended by Preacher and Hayes (44). The path-analytic procedure is suitable for analyzing the role of third variables (e.g., mediators, confounders, suppressor variables) in relationships between two variables (45). Calculations were conducted using the SPSS macro “Indirect” (44) and IBM SPSS Statistics software (IBM Corp.).

Consistent with our hypothesis, the T1 Snack-Delay task and the T2 Gift-Bag task were significantly associated with the T2 ADHD diagnosis (Table 3A, model 1). Associations remained statistically significant after adjusting for the T1 ADHD symptoms score (Table 3A, model 2). We further hypothesized that the link between RRIC and ADHD is shared with ODD symptoms. However, the T1 Snack-Delay task was only marginally significantly associated with the T2 ODD score, and the T2 Gift-Bag task was not correlated with the T2 ODD score (Table 2). Moreover, covariation of ODD and CU scores (Table 3A, model 3) did not change the significant associations between the RRIC tasks and T2 ADHD diagnosis. Hence, the results indicate a unique association between low RRIC and ADHD.

The T1 SWT task and the T2 Int-AT task were significantly associated with the T2 ADHD diagnosis (Table 3B, model 1) indicating high approach motivation in children with ADHD. After adjusting for T1 ADHD symptoms, the prediction by the T1 SWT task was no longer significant (Table 3B, model 2). This finding indicates significant common variance between ADHD symptoms and high approach motivation already at T1. We further hypothesized that the link between approach motivation and ADHD can be explained by comorbid ODD and CU symptoms. Adjustment for the ODD and CU scores led to a reduction in the associations of T1 and T2 approach motivation task with T2 ADHD (Table 3B, model 3). Next, we analyzed whether T2 ODD and CU scores explain the link between T1 SWT task and T2 ADHD (Table 4). The indirect links via the ODD and via the CU score were significant. After accounting for the indirect link via the comorbidity scores, the SWT task-ADHD association was no longer significant. Hence, T1 approach motivation predicted ADHD with comorbid ODD and CU problems. Regarding the link of the T2 Int-AT task with T2 ADHD, indirect effects by CU and ODD scores were not statistically significant (Table 4).

As expected, children with ADHD showed lower mean SCRs to the questions on the attractiveness of the toys than did the other children (Table 2; Table 3C, model 1). Mean SCRs were not significantly correlated with ODD symptoms and CU traits (Table 2). Adjusting for ODD and CU scores did not change the significant association with the ADHD diagnosis (Table 3C, model 3). Hence, low sympathetic reactivity to the stimuli was uniquely associated with ADHD.

For the purpose of comparison, we additionally assessed baseline SCL. As shown in Table 2, the score was significantly associated with preschool ADHD symptoms. Associations with preschool and school-age ODD symptom scores just failed to reach statistical significance (p's < 0.10). Children with high symptoms showed a lower baseline/resting arousal level.