ارتباط الکتروفیزیولوژیک پردازش خطا در اختلال شخصیت مرزی

Abstract The electrophysiological correlates of error processing were investigated in patients with borderline personality disorder (BPD) using event-related potentials (ERP). Twelve patients with BPD and 12 healthy controls were additionally rated with the Barratt impulsiveness scale (BIS-10). Participants performed a Go/Nogo task while a 64 channel EEG was recorded. Three ERP components were of special interest: error-related negativity (ERN)/error negativity (Ne), early error positivity (early Pe) reflecting automatic error processing, and the late Pe component which is thought to mirror the awareness of erroneous responses. We found smaller amplitudes of the ERN/Ne in patients with BPD compared to controls. Moreover, significant correlations with the BIS-10 non-planning sub-score could be demonstrated for both the entire group and the patient group. No between-group differences were observed for the early and late Pe components. ERP measures appear to be a suitable tool to study clinical time courses in BPD

1. Introduction Borderline personality disorder is characterized by a variety of symptoms like a pervasive pattern of instability of interpersonal relationships, self-image, and affects, recurrent suicidal behavior, affective instability and chronic feelings of emptiness (APA, 1994). Among these, increased impulsiveness has received special attention as one of the core symptoms of this disorder. There is still an ongoing debate on the neurobiological and neuropsychological mechanisms underlying BPD. Several positron emission tomography (PET) studies reported a hypometabolism in prefrontal cortical areas (De La Fuente et al., 1997 and Soloff et al., 2003) which reflects either a diminished serotonergic turnover (Hansenne et al., 2002) and/or dopamine dysfunction (Friedel, 2004), and/or an interaction between these resulting in emotional dysregulation and impulsive behavior. In a magnetic resonance imaging study, Tebartz van Elst et al. (2003) found reduced volumes of the hippocampus, amygdala, and orbitofrontal, dorsolateral prefrontal, and anterior cingulate cortex in BPD patients. Moreover, BPD patients often suffer from impairments in decision-making and planning again pointing to deficits in a circuitry encompassing the frontal lobes (Bazanis et al., 2002). According to Barratt (1985) impulsiveness is not a unidimensional trait but rather consists of three factors: (1) a motor impulsiveness sub-trait (Mot) involving acting without thinking (“I act on the spur of the moment”), (2) a cognitive impulsiveness sub-trait (Cog) that involves fast cognitive decisions (“I make up my mind quickly”), and (3) a non-planning impulsiveness sub-trait (NP) that involves lack of “futuring” (prospective reasoning) which heavily relies on social conventions and norms (“I am more interested in the present than the future”). The Barratt impulsiveness scale, Version 10 (BIS-10) has been administered to psychiatric inpatients and could clearly differentiate antisocial and borderline personality disorder from other diagnostic categories, like major depression and schizophrenia (Barratt, 1985). The BIS-10 has also consistently been rated as a reliable and valid instrument for measuring impulsiveness in patients and normal controls (Patton et al., 1995, Bayle et al., 2000, Someya et al., 2001, Preuss et al., 2003 and Spinella, 2004). As cognitive processing related to impulsiveness is fast by definition, the high time resolution of event-related potentials (ERP) in the range of milliseconds makes them an appropriate tool to investigate a time sensitive phenomenon like this. In this context, recent ERP-studies proved their suitability in that they showed that the error negativity (Ne; Falkenstein et al., 1990) or error-related negativity (ERN; Gehring et al., 1990) mirrors impulsive responding, e.g. in an Eriksen flanker task (Dikman and Allen, 2000, Luu et al., 2000 and Pailing et al., 2002). So in the present study we made use of ERP potentials to investigate error-processing in BPD patients, to test whether there are deviations from controls, and whether impulsivity may play a role mediating differences between patients and controls. The ERN/Ne is a negative going ERP component peaking between 100 and 150 ms after the onset of electromyographic (EMG) activity associated with an erroneous response (Scheffers et al., 1996) in forced choice reaction time paradigms like the Eriksen flanker task or Go/Nogo tasks (Falkenstein et al., 1999). As the present study uses button presses (instead of EMG activity) in order to define an error, the ERN/Ne is expected to be found within 100 ms after an erroneous button press. A source analysis of the ERN/Ne scalp potential with brain electric source analysis (BESA; Scherg and Berg, 1990) pointed to neural generators in medial prefrontal areas, best to correspond to the anterior cingulate cortex (ACC; Dehaene et al., 1994 and Ruchsow et al., 2002). The involvement of ACC in error processing has been confirmed by several fMRI studies (e.g. Carter et al., 1998). The ERN/Ne was originally interpreted as an error detection signal resulting from a mismatch between the representation of the correct response and the representation of the actual (false) response (Falkenstein et al., 1990 and Gehring et al., 1993). Alternative accounts view the ERN/Ne as a brain potential reflecting the response evaluation process itself rather than the outcome of this process (Vidal et al., 2000). Rather contrary to these interpretations, Cohen and coworkers interpret the ERN/Ne to be associated with the detection of response conflict (Botvinick et al., 2001 and Carter et al., 1998). More recently, Holroyd and Coles (2002) proposed a theory combining elements of mismatch theory and reinforcement learning principles. According to their suggestion the ERN/Ne components are generated when a negative reinforcement learning signal is conveyed to the ACC via the mesencephalic dopamine system. This signal is used by the ACC to modify performance on the task at hand. During an Eriksen flanker task, Luu et al. (2000) found large ERN/Ne amplitudes at the beginning of the session in college students who were high on negative affect (NA) and negative emotionality (NEM). Moreover, a shift of response patterns was found during the time course of the experiment. By means of a post-task questionnaire part of the subjects were reported to have been bored and dissatisfied with their performance resulting in motivational problems and disengagement from the task. When EEG data were re-analyzed for members of the high-NA and high-NEM groups with motivational problems the amplitude of the ERN/Ne decreased with the time-course of the entire experiment. This pattern of results was strikingly different from results of participants who were low on NA and NEM and showed unaltered ERN/Ne amplitudes throughout the whole experiment. Similarly, Dikman and Allen (2000) demonstrated that individuals low on socialization exhibit smaller ERN/Ne amplitudes during tasks which penalize error responses. In the same vein, Pailing et al. (2002) found smaller ERN/Ne peak amplitudes and higher error rates in subjects with a tendency towards impulsive responding (i.e. with a less controlled response strategy). They defined impulsiveness based on a linear regression from correct individual reaction times (RTs) on reaction times from erroneous responses. Mean residual scores were defined as mean difference of observed RTs minus predicted RTs for error trials. Less negative mean residual RTs were regarded as indicating a more cautious (controlled) response strategy whereas more negative residuals were interpreted to indicate a less controlled (i.e. more impulsive) response style. Furthermore, ERN/Ne latencies were positively related with percentage of errors suggesting that individuals with faster ERN/Ne components should have more opportunity to catch an erroneous intention before fully committing the response than individuals with slower ERN/Ne components (Pailing et al., 2002). On the opposite side of a possible impulsivity–compulsivity dimension, Gehring et al. (2000) have reported enhanced ERN/Nes in patients with obsessive–compulsive disorder. In these patients ERN/Ne amplitude was positively related with symptom severity. In contrast to these documented relations between impulsivity/compulsivity and ERN/Ne amplitude, Luu et al. (2000) did not find correlations between ERN/Ne amplitudes and measures of impulsivity in a sample of healthy controls. Perhaps, these discrepant findings are due to different psychometric measures of impulsivity and subject populations in these studies. Another ERP component discussed in the context of error processing is the error positivity (Pe). The Pe is a slow positive wave with centro-parietal distribution which often follows the ERN/Ne in a time window between 300 and 500 ms after erroneous responses. The Pe can clearly be differentiated in time from the P300 (Falkenstein et al., 2000). A source localization analysis, using BESA revealed that the Pe consists of two components: an early Pe component with probable generators in the area of the caudal ACC and a late Pe component with probable generators in the area of the rostral ACC (Van Veen and Carter, 2002). The late Pe component was associated with awareness of erroneous responses, since it was more pronounced for perceived than for unperceived errors (Nieuwenhuis et al., 2001). In the present study we expected that error monitoring processes as expressed in the ERN/Ne and Pe components differ between BPD patients and controls. Additionally, we tested whether impulsiveness in BPD patients might have an impact on the ERN/Ne amplitudes.

. Results 3.1. Behavioral data Given the task, only false positive responses on Nogo trials were of interest (commission errors). Error rates were individually calculated as number of false positive reactions during Nogo trials. For healthy controls mean number of errors was 36.8 (S.D.: 38.9) corresponding to an error rate of 12.3%. BPD patients demonstrated a mean number of 39.7 errors (S.D.: 32.7) corresponding to an error rate of 13.2%. Error rates did not significantly differ between groups (t(22) = 0.42; p = 0.681). The mean reaction time (RT) was 237.7 ms (S.D.: 16.5) for correct Go trials and 222.2 ms (S.D.: 23.0) for incorrect Nogo trials in healthy subjects. In the patient group mean RT was 225.1 ms (S.D.: 32.7) for correct Go trials and 204.2 ms (S.D.: 44.8) for incorrect Nogo trials. We performed a 2(patients/controls) × 2(correct/incorrect) ANOVA for RTs which revealed that both groups were significantly faster on error trials than correct trials [F(1, 22) = 10.61, p < 0.001]. The group by condition interaction was not significant [F(1, 22) = 0.23, p = 0.634]. Mean BIS-10 scores were calculated for controls (NP: 23.1 (S.D.: 5.2); Mot: 21.5 (S.D.: 5.0); Cog: 22.6 (S.D.: 4.4)) and BPD patients (NP: 27.9 (S.D.: 5.9); Mot: 26.6 (S.D.: 5.7); Cog: 30.8 (S.D.: 4.8)). Significant group differences between patients and controls were observed for all three BIS-10 sub-traits (all p < 0.045). 3.2. Event-related potentials 3.2.1. Averaged waveforms for correct and incorrect trials: voltages With respect to the ERN/Ne component we found a significant condition effect for both central midline and central lateral electrodes (central midline: [F(1, 22) = 33.65, p < 0.001]; central lateral: [F(1, 22) = 36.63, p < 0.001]), and a significant interaction of group by condition (central midline: [F(1, 22) = 7.48, p < 0.012]; central lateral: [F(1, 22) = 8.32, p < 0.01], see Fig. 1, upper panel). Post hoc tests revealed that condition dependent differences were significant only for the control group (all p-values below p < 0.001) but not for the patient group (all p-values above p > 0.127). When considering between-group differences voltages in the error condition were significantly more negative in the control group compared to the patient group at all central electrodes (all p-values below p < 0.01). Voltages in the correct condition did not differ between groups (all p-values above p > 0.662). Grand averaged waveforms for correct and incorrect trials in BPD patients ... Fig. 1. Grand averaged waveforms for correct and incorrect trials in BPD patients (black) and healthy controls (grey). Potentials were collapsed across midline electrode positions (central: FCz, Cz; parietal: CPz, Pz). Figure options For parietal midline and parietal lateral electrodes we observed a main effect of condition (parietal midline [F(1, 22) = 71.71, p < 0.001]; parietal lateral [F(1, 22) = 58.48, p < 0.001]) as well as a condition by group interaction for parietal midline electrodes, only ([F(1, 22) = 5.88, p < 0.024]; see Fig. 1, lower panel). Post hoc tests revealed that condition dependent differences were significant within each group (all p-values below p < 0.006) whereas between-group differences did not reach significance neither for correct trials (all p-values above p > 0.486) nor for error trials (all p > 0.478). With regard to the early Pe component we found a significant condition effect for all electrodes (all F-values above F(1, 22) = 5.79; all p-values below p = 0.025). Voltages at central electrodes were more positive for error trials than for correct trials. At all parietal electrodes the reversed pattern was evident. Condition-by-group interactions for this component were not significant (all F-values below F(1, 22) = 0.665; all p-values above p = 0.423) regardless of electrode position (see Fig. 1). In the time window of the late Pe component we found a significant condition effect at both central and parietal electrodes (all F-values above F(1, 22) = 33.31; all p-values below p < 0.001) with voltages in the error condition more positive than voltages in the correct condition. Condition-by-group interactions were not statistically reliable (all F-values below F(1, 22) = 1.56; all p-values above p = 0.224; see Fig. 1). 3.2.2. Difference waves: voltages Similar to the analyses of the raw ERPs to correct and incorrect trials, significant group differences for the ERN/Ne component as measured by difference waves were locally constrained at central electrode positions (central midline: [F(1, 22) = 6.93, p < 0.015]; central lateral: [F(1, 22) = 5.92, p < 0.024]; see Fig. 2, upper panel). At these positions, patients demonstrated significantly smaller amplitudes than controls. At parietal electrode positions, there were no significant group differences (parietal midline: [F(1, 22) = 2.16, p = 0.156]; parietal lateral: [F(1, 22) = 1.13, p = 0.300]; see Fig. 2, lower panel). Grand averaged difference waves (incorrect minus correct) in BPD patients ... Fig. 2. Grand averaged difference waves (incorrect minus correct) in BPD patients (black) and healthy controls (grey). Potentials were collapsed across electrode positions (midline, see legend in Fig. 1; lateral electrode positions: central: C1/C2; parietal: P1/P2). Figure options With regard to early and late Pe amplitudes group differences were not significant, neither at central nor at parietal electrode positions (all p-values above p > 0.137; see Fig. 2). 3.2.3. Difference waves: latencies With regard to the ERN/Ne, the early and the late Pe component there were no group differences for latencies (all p-values above p > 0.315). 3.3. Component dissociation In order to substantiate the differential effects of group on ERN/Ne, early and late Pe components statistically we performed a 2(group) × 3(component: ERN/early Pe/late Pe) ANOVA. As dependent variables we calculated mean amplitudes across relevant electrodes (e.g. the mean of the peak amplitudes from four central electrodes for the ERN/Ne effect; Fig. 3). We found a significant main effect of component [F(2, 44) = 87.97, p < 0.001] and a significant component by group interaction [F(2, 44) = 3.99, p < 0.026]. Post hoc tests revealed that ERN/Ne amplitudes were significantly more negative in healthy controls compared to BPD patients (p = 0.013), while early Pe (p = 0.801) and late Pe (p = 0.348) components did not differ between groups. To control, a second MANOVA on averaged peak amplitudes from parietal electrodes for the ERN/Ne and early Pe components, and central electrodes for the late Pe component did not yield a significant component-by-group interaction (F(2, 44) = 0.73; p = 0.486; data not shown). Component-by-group interactions. In case of the ERN/Ne and early Pe component ... Fig. 3. Component-by-group interactions. In case of the ERN/Ne and early Pe component data points reflect mean peak amplitudes (errors bars: standard error of the mean) on incorrect trials averaged across the four central electrodes. For late Pe mean voltages were averaged across the four parietal electrodes. Asterisk (*) denotes a significant group difference with p = 0.019 (post hoc test). Figure options 3.4. Correlational analyses To control for the direction of correlation coefficients peak amplitudes for the ERN/Ne component during erroneous trials (raw waves) were multiplied with minus 1, so that a positive correlation would indicate an increase in ERN/Ne amplitude (i.e. a more negative potential) with increasing BIS-10 score. 3.4.1. Baratt imulsiveness scale and ERN/Ne amplitudes To test whether ERP data were correlated with clinical symptoms we calculated correlational analyses for the BIS-10 sub-traits with ERN/Ne amplitudes. Because we did not find significant group differences on early and late Pe amplitudes, correlations of these parameters with the BIS-10 were not analyzed. Similarly, as there were no significant group differences on error rates correlations of these measures with ERN/Ne and Pe latencies or BIS-10 sub-traits were not calculated. 3.4.1.1. Entire group For the entire group (patients and controls: n = 24) we found a significant negative relation between the ERN/Ne amplitude and the NP sub-trait of the BIS-10 at all central electrodes (FCz: r = −0.42; p = 0.041; Cz: r = −0.46; p = 0.025; C1: r = −0.52; p = 0.009; C2: r = −0.45; p = 0.029). For scores of the Mot and Cog sub-traits from BIS-10 no significant correlations were observed (all p-values above p > 0.110). 3.4.1.2. Within-group correlations For patients with BPD (n = 12) we found a significant negative relation between ERN/Ne amplitudes and the NP sub-trait of the BIS-10 at central electrodes Cz (r = −0.66; p = 0.020), C1 (r = −0.69; p = 0.013), and C2 (r = −0.68; p = 0.014). The correlation coefficient was short of significance at electrode FCz (r = −0.57; p = 0.051). For the Mot and Cog sub-traits from BIS-10 no significant correlations were observed (all p-values above p > 0.469). For healthy controls (n = 12) no significant correlations between the ERN/Ne amplitude and BIS-10 sub-traits were found (all p-values >0.384). 3.4.2. Medication and ERN/Ne amplitudes in patients with BPD To control for whether group differences in ERP data were compromised by the fact that only patients received medication we correlated dosages of tricyclic antidepressants, chlorpromazine, and benzodiazepines (see Table 1) with ERN/Ne amplitudes in the BPD patient group. These analyses did not reveal significant correlations (all r-values with r < 0.20; all p-values with p > 0.526). Thus, it is very unlikely that our observed group differences were due to medication effects in the BPD patients group.