Neural Correlates of Error Monitoring in Adult Attention Deficit Hyperactivity Disorder After Failed Inhibition in an Emotional Go/No-Go Task

We enrolled 26 patients meeting DSM-IV criteria for adult ADHD (men, N=20; women, N=6; mean age=26.7 years [SD=5.7]; inattentive type: N=12; hyperactive/impulsive type: N=7; combined type: N=7) and 14 healthy controls (men, N=11; women, N=3; mean age=31.5 years [SD=11.4]), matched by age (SD=5 years), gender, and level of education. Participants provided written, informed consent in accordance with procedures approved by the Institutional Review Board of Semmelweis University, Budapest, Hungary. Patients were recruited from the adult ADHD outpatient clinic of the Department of Psychiatry and Psychotherapy of Semmelweis University. Controls were recruited from the office and medical staff at the University and their acquaintances. Participants in both the ADHD and the healthy control groups completed the 66-item version of the Conners’ Adult ADHD Rating Scales (CAARS).³¹ The 90-item Symptom Checklist³² (SCL-90R) was used to select controls with no current psychiatric comorbidity. Among the ADHD patients, two had depression in their medical history, while dysthymia occurred in one patient, somatization disorder occurred in one patient, and panic disorder also occurred in one patient. Patients taking stimulant treatment (N=10) were off medication at least 24 hours before testing. Lack of history of psychiatric disease was required for inclusion in the control group. The main exclusion criteria for participants in the control group were any present or past neurologic disorder and history of head injury with loss of consciousness.

Subjects performed the task in a dimly lit, sound-attenuated room. The computer screen for stimuli was placed at a viewing distance of approximately 100 cm. We applied an emotional go/no-go response inhibition task, which was presented by the Presentation 13.0 software (Neurobehavioral Systems, Inc., Albany, Calif.). We used pictures from the International Affective Picture System (IAPS [http://www4.ncsu.edu/~dgruehn/page7/page8/page8.html]) as stimuli; they comprised images with positive, negative, and neutral affective contents. The neutral, positive, and negative images were not different in physical characteristics (luminance, contrast and spatial frequency, confirmed by the Delplanque procedure³³), but they differed in terms of arousal (since negative pictures in the IAPS are associated with higher arousal than the neutral or positive ones).^34,35 Each participant was instructed to respond with a “go” button when a picture appeared on the screen and to withhold responding when the picture was repeated in the consecutive trial. Furthermore, participants were instructed to respond as quickly and accurately as possible. Each block consisted of 240 stimuli comprising 85% (160) of go stimuli and 15% (80) of no-go stimuli. All participants completed three experimental blocks. All stimuli were presented for 800 ms and were followed by an interstimulus interval of 600 ms.

The BioSemi recording system (sample rate=1024 Hz, band-pass filter=0.5−70 Hz) with average reference was used to acquire EEG. A standard BioSemi 128-electrode head cap system (https://www.biosemi.com/), with electrodes labeled in four blocks of 32 electrodes, was applied.

Data were analyzed off-line using Electro-magnetic Source Signal Imaging (EMSE Suite v.5.0, Source Signal Imaging, Inc., San Diego) and the Statistical Analysis System (SAS9.4) software. EEG data were filtered between 0.5 and 70 Hz using zero-phase shift-forward and reverse IR Butterworth-filter. Additionally, the 48–52 Hz Parks-McClellan stop-band notch filter was applied in order to remove any potential electrical interference from the 50-Hz line. Artifacts due to blinks and eye movements were removed manually and with the electrooculography artifact removal procedure. Epoch selection for the analyses was conducted manually, as well as applying automatic artifact rejection criteria. Response-locked data were segmented into epochs of 200 ms from before response to 400 ms after response. The time period immediately following the motor response was the main focus of our interest, and it was investigated in inferential statistical analyses. The difference between the two groups in terms of trial count (no-go trials with incorrect responses) was statistically significant (p<0.05). Subjects who committed at least six errors were included in the analysis.

The ERN was defined as the average amplitude in microvolts occurring in the window from 20 ms to 70 ms post-response. The Pe was defined as the average amplitude in microvolts peak within 100–300 ms of the response.⁴ The ERN was measured in FCz and Cz, and the Pe at Cz and Pz electrodes on the basis of previous research.² Performance was assessed with measures of the mean reaction time and commission error rate. Mean reaction time was calculated as the average of mean reaction times for correct go trials.

The primary statistical analysis for group difference between ADHD and control subjects was based on the random regression hierarchical linear model. Amplitude (voltage) values within the time window of interest (20–70 ms, and 100–300 post-response for ERN and Pe, respectively) for error-related activity were used as dependent variables in the hierarchical linear model. We applied group, time (sampling point), and their interaction as independent variables; age, gender, and level of education served as covariates. The group main effect served as the principal interest in the analyses. A separate analysis was performed for each of the two time windows (ERN and Pe) in each brain region of interest (i.e., FCz, Cz, and Pz). For scalp areas that yielded a significant group difference in the primary analysis after correction for multiple testing, we conducted additional analyses to test whether psychopathological variables served as covariates in explaining the significant alterations in error-related activity. Covariates that we tested included the total score on the CAARS hyperactivity, impulsivity, inattention, and problems with self-concept domains. Group comparisons for behavioral data (error rates and reaction times), which had a skewed distribution, were investigated by generalized linear model analysis, since this method allows for the investigation of non-normally distributed variables. The Hochberg procedure was applied for correction for multiple testing. The statistical significance of the group difference was tested using Wald’s chi-square statistic. Continuous demographical variables (e.g., age) were tested with analysis of variance using F statistics; categorical variables (e.g., gender) were tested using chi-square analysis.

Basic demographic and clinical characteristics of the study population are presented in Table 1. As shown in the table, the study groups were similar on basic demographic variables, including age and gender. Approximately three-quarters of the sample consisted of males. As expected, the ADHD group had higher severity of general psychopathology as measured by the SCL-90R scale, and this group displayed higher severity on all specific symptom dimensions, including the CAARS factors of inattention, hyperactivity, impulsivity, and problems with self-concept.

The ADHD group made significantly more commission errors compared with the healthy control group for the neutral and negative stimuli, while the numeric difference of a similar magnitude did not obtain statistical significance for positive valence (Wald χ²=9.46, df=1, p=0.0040 for neutral stimuli; Wald χ²=5.00, df=1, p=0.0316 for negative stimuli; Wald χ²=1.52, df=1, p=0.2251 for positive stimuli). There were no group differences with regard to the reaction times.

To illustrate the group differences in error-related ERP waveforms, response-locked average ERPs to neutral and emotionally valenced stimuli are displayed in the frontal, central, and parietal areas (Figure 1). The shaded areas show the time windows in which we examined ERN and Pe. Error-related activity had similar waveforms in both groups. Significant group differences with negative stimuli were detectable for the ERN at the FCz, Cz, and Pz electrodes (for FCz: F=14.15, df=1, 37, p=0.0013; for Cz: F=288.74, df=1, 37, p<0.0001; for Pz: F=75.65, df=1, 37, p<0.0001). For neutral stimuli, we found a significant ERN amplitude difference only at the Cz electrode (F=12.17, df=1, 37, p=0.0028). We found no differences in ERN for the positive stimuli. The analysis of the Pe revealed a significant group difference for neutral stimuli at electrodes FCz, Cz, and Pz (for FCz: F=54.81, df=1, 37, p<0.0001; for Cz: F=109.29, df=1, 37, p<0.0001; for Pz: F=86.26, df=1, 37, p<0.0001). Group difference for Pe was not observable for the positive and negative stimuli. The interaction between group and time did not reach statistical significance in either of the above analyses. Numeric results for the error-related ERPs are summarized in Table 2.

For the ERP group differences that reached the level of statistical significance, we performed a follow-up analysis based on CAARS symptom dimensions, within the ADHD group. After adjustments for multiple comparisons, we found associations between ERN amplitude and impulsivity factor at the FCz electrode and between CAARS hyperactivity factor at the Cz electrode. Investigation of the direction of the relationship indicated larger ERN amplitude among those ADHD subjects who had higher severity on impulsivity compared with patients who had lower severity on it. Lower hyperactivity was associated with larger ERN amplitude compared with higher hyperactivity. We found no association with CAARS dimensions in terms of Pe amplitude. The relationships between error-related activity and CAARS symptom dimensions are presented in Table 3.