Sample

All participants were recruited in the context of a therapy study on ADHD at the department of clinical psychology and psychotherapy, Philipps University, Marburg, Germany for details on this study, see [27]. Families with a child suspected of having ADHD or parents of children who were already pre-diagnosed by a local pediatrician registered their child at the outpatient clinic of the department. Healthy control children were recruited from a local elementary school. Control children were required to have no formal or suspected ADHD diagnosis and no family history of ADHD. All children had to be 7–16 years old at the beginning of the study. Exclusion criteria included autism, IQ below 80, brain disorders, and any genetic disorder that mimics ADHD. Ethical approval for this study was obtained from the Department of Psychology, Philipps University Marburg, Germany. All parents and children gave written and verbal informed consent.

In the ADHD group, 91% of the children had a diagnosis of the combined subtype according to the Diagnostic and Statistical Manual of Mental Disorders DSM-IV, American Psychiatric Association; [28], 9% fulfilled the criteria for the predominantly inattentive subtype, and none of the children had the predominantly hyperactive/impulsive subtype. Comorbidity rates in the ADHD group were high, with 46% of children having at least one comorbid disorder. Twenty-nine percent were diagnosed with Oppositional Defiant Disorder (ODD), 9% with Enuresis Nocturna, 7% with anxiety disorders, and 3% with tic disorders.

The ADHD group significantly differed from the non-affected sibling and healthy control groups with respect to gender (χ ² ₍₂₎  = 10.5, p = .005). In the ADHD group, 78% of the children were male, while only 59% and 44% were male in the non-affected sibling and control groups, respectively. In addition, the non-affected sibling group differed from the other two groups in age (χ ² ₍₂₎  = 50.3, p < .01), as siblings tended to be older than the ADHD and control children. We therefore used age- and gender-adjusted QbTest values (Q-values) to calculate the QbTest factor scores, and also included both variables as covariates.

Procedure and materials

ADHD children and their sibling were assessed by experienced and well-trained staff at the department of psychology. Children were motivated with small breaks between tasks and a game at the end of the testing session. Children who were diagnosed with ADHD were offered to participate in a current therapy study [27]. Assessment of control children was performed in a similar way. All control children received a cinema gift coupon (10 € value) for participation.

Parents and teachers of all children were asked to fill out the German version of the Conners’ 3rd rating scales [29]. Parent and teacher ratings (T-scores) greater than 63 on the total symptom subscale are regarded as clinically relevant. A biographical parent questionnaire covering demographic characteristics, family history of psychological or psychiatric disorders, and information on child development was completed for all children. Furthermore, a semi-structured, standardized clinical interview Kiddie-SADS; German version; [30] was conducted with all ADHD children. The Kiddie-SADS covers all psychological/psychiatric disorders according to the DSM-IV, and if symptom areas were screened as positive, the long version of the interview was performed. The Kiddie-SADS was not administered to non-affected siblings or controls, who were regarded as non-affected if they scored lower than 63 on either the Conners’ parent or teacher rating scale and if no history of behavior problems was reported in the biographical parent questionnaire.

The neuropsychological QbTest was administered to all participating children. Children in the ADHD group that were on stimulant medication were required to be off medication at least 48 hours prior to testing. The QbTest is a combined continuous performance (CPT) and activity test [22] that assesses the three core symptoms of ADHD. In general, the three QbTest factors—Hyperactivity, Inattention, and Impulsivity—show adequate construct and discriminant validity [23]. Qbtech© provides two different test versions of the test, one for children aged 6–12 years, and an adolescent/adult version 12-60 years. For both test versions, separate norms for male and female participants, and age-groups (per year in the children’s version, per decade in the adult version) are available. According to the test manual [22], the children’s version is based on normative data from 576 children (262 male, 314 female). For the adult version, normative data is based on data from 731 adults (360 male, 371 female).

Statistical analysis

All analyses were conducted using SPSS 19.0. All variables were examined for accuracy of data entry and missing values prior to the analyses. No missing values were found. The Kolmogorov-Smirnov test did not reach significance for any of the three factors used as dependent variables in the multivariate analyses, thus normal distribution of the data can be assumed.

To control classification and separability of the three groups, we conducted t-tests for each group to determine whether group means were significantly different from our cut-off score in the Conners’ Parent questionnaire (T = 63; percentile = 90.3).

Even though effects of age and gender should have minimal influences because age- and gender-normed Q-values were used to calculate factor scores, we still performed a MANCOVA with the QbTest factor scores as dependent variables, group (ADHD, non-affected sibling, healthy control) as a between subject factor, and age and gender as covariates.

Finally, to further explore effects of familiality, we conducted trend analyses across the three groups to test whether non-affected siblings show intermediate impairments between ADHD children and healthy controls. A linear trend in the absence of a residual quadratic trend would indicate familiality of the given construct. A residual quadratic trend would indicate that the non-affected siblings were more similar to either the ADHD or control group [15].

Group differences in Hyperactivity, Impulsivity, and Inattention QbTest factor scores

The ADHD group scored significantly higher than 63 on the Conners’ Parent ADHD index (Mean = 70.0; SD = 7.3; p < .001). The sibling group (Mean = 56.2; SD = 11.3; p = .014), and control group (Mean = 51.3; SD = 7.0; p < .001) both scored significantly lower than 63.

The MANOVA performed on the QbTest data showed a large and significant main effect of group (F_(2,109) = 7.75, p < .01, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .177), indicating that groups differed in mean hyperactivity, inattention, and behavioral impulsivity scores. There were significant differences between groups for all three factors, and the effect sizes were large (Inattention: F_(2,109) = 15.47, p < .01, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .221; Impulsivity: F_(2,109) = 14.57, p < .01, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .211; Hyperactivity: F_(2,109) = 9.94, p < .01, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .154; see Table 2). As predicted, ADHD children had the highest scores (greatest impairment) for all three symptom dimensions, followed by their non-affected siblings, and control children had the lowest scores. These results were confirmed by non-parametric analyses (Inattention: χ²₍₂₎ = 24.4, p < .01; Impulsivity: χ²₍₂₎ = 22.7, p < .01; Hyperactivity: χ²₍₂₎ = 16.5, p < .01).

Pair-wise comparisons (post-hoc Scheffé tests) confirmed significant group differences between the ADHD and control group for all three factors (p s < .01). This was supported by results from the non-parametric Mann-Whitney-U test and effect sizes ranged from medium to large (Hyperactivity: U = 521.00, p < .01, r = −0.42; Inattention: U = 433.00, p < .01, r = −0.49; Impulsivity: U = 426.00, p < .01, r = −0.5). As expected, ADHD children were significantly more hyperactive than control children while performing the QbTest, and also showed more impulsive and inattentive behaviors (see Figure 1).

Hyperactivity factor scores were significantly different between non-affected siblings and controls (p = .039), such that non-affected siblings showed an intermediate level of motor activity between their ADHD siblings and healthy controls. This effect was confirmed by non-parametric results that had a medium effect size (U = 313.5, p = .015, r = −0.3), and can thus be interpreted as preliminary evidence for the utility of the Hyperactivity factor as a candidate intermediate phenotype for ADHD. The difference between the non-affected sibling and healthy control groups did not reach significance for the Impulsivity factor (p = .128). The non-parametric Mann-Whitney-U test revealed a trend towards a significant difference in behavioral impulsivity between non-affected siblings and healthy controls (U = 360.0, p = .071, r = −0.22). Furthermore, a post-hoc Scheffé test revealed no significant difference between the non-affected sibling and control groups on Inattention factor scores (p = .962), and group means for those two groups were very close, as shown in Figure 1. There were significant group differences between ADHD children and their non-affected siblings (p < .01), indicating that non-affected siblings were more similar to the healthy controls than to their biological ADHD siblings in terms of inattention.

Effects of age and gender

The MANCOVA revealed the following results for age and gender. Hyperactivity factor scores were not significantly influenced by either age (F_(1,107) = 0.61, p = .435, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .006) or gender (F_(1,107) = 1.66, p = .201, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .015), nor were Inattention (age: F_(1,107) = 1.79, p = .184, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .016; gender: F_(1,107) = 1.46, p = .230, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .013) or Impulsivity factor scores (age: F_(1,107) = 0.48, p = .226, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .014; gender: F_(1,107) = 0.43, p = .514, ${\mathrm{\eta }}_{\mathsf{p}}^2$ = .004).

Trend analyses

Trend analyses for Hyperactivity factor scores revealed a strong linear trend (F_(2,109) = 19.09, p < .01) without a quadratic trend (F_(2,109) = 0.79, p = .38) indicating that, as predicted, the non-affected sibling group showed an intermediate level of motor activity compared to the more hyperactive ADHD group and the less hyperactive healthy control group. Hence, motor activity showed evidence of familiality. For Inattention factor scores, there was a linear trend (F_(2,109) = 26.95, p < .01), and the quadratic trend also reached significance (F_(2,109) = 3.98, p = .05). Inattention levels were very similar between the non-affected sibling and control groups (see Table 2), and thus there was no evidence of familiality for inattention. Finally, there was a linear trend for Impulsivity factor scores (F_(2,109) = 29.11, p < . 01) in the absence of a quadratic trend (F_(2,109) = 0.02, p < . 88), indicating that the three groups showed the typical intermediate phenotype pattern across behavioral impulsivity levels.

Limitations

The following limitations of this study should be noted. First, as discussed above, age and gender distributions differed between the groups in our study. Boys were overrepresented in the ADHD group, and mean age was higher in the non-affected sibling than ADHD and control groups. However, analyses controlling for age and gender did not reveal significant influences of age and gender, and the higher number of boys in the ADHD group reflects the male to female ratio in ADHD [2–4].

Second, we did not administer the clinical interview to non-affected siblings and controls. Additionally, Conners’ Teacher Questionnaires were missing for some children in the non-affected sibling and control groups. However, known formal diagnosis of ADHD and other childhood disorders were assessed in the biographical parent questionnaire and participation in the control group was explicitly advertised as seeking children without any ADHD-related behavior problems. Nevertheless, five children in the non-affected sibling group and two children in the control group showed high Conners’ Parent ratings (T > 63). As parent ratings of ADHD behavior and QbTest factor scores were significantly correlated (Pearson’s Correlation: .24 < r > .44), it is likely that non-affected siblings and controls who show elevated QbTest scores also have higher ADHD behavior ratings than children with lower QbTest scores. As described above, we controlled classification and separability of the three groups by testing group means against our Conners’ Parent questionnaire cut-off score (T = 63). The ADHD group scored significantly higher, while the other two groups scored significantly lower, than the cut-off score. Thus, overall, the groups adequately differed with respect to parent ratings of ADHD-related behavior.

Finally, due to relatively small groups, the reported results are preliminary and need to be confirmed in larger samples. Future studies should further explore (1) the utility of technically assessed motor activity as an intermediate phenotype in ADHD, and (2) the advantage of neuropsychological factor scores over single variable scores.