Moeller SJ Supplementary 1
SUPPLEMENTARY MATERIALS AND METHODS
Subjects
Subjects from both samples were recruited using advertisements in local newspapers, by word-of mouth, and from local treatment facilities. All were fully informed of all study procedures and risks, and provided written consent in accordance with the local Institutional Review Board. Subjects were healthy and free of medications, as ascertained during a full physical and neurological examination by a neurologist and a diagnostic interview by a clinical psychologist. This interview included the Structured Clinical Interview for DSM-IV Axis I Disorders [research version1, 2], the Addiction Severity Index3, the Cocaine Selective Severity Assessment Scale4 and the Cocaine Craving Questionnaire5. Exclusion criteria were as follows: (A) history of head trauma or loss of consciousness (> 30 min) or other neurological disease of central origin (including seizures); (B) abnormal vital signs at time of screening; (C) history of major medical conditions, encompassing cardiovascular (including high blood pressure, cardiac arrhythmias apart from sinus bradycardia, or an abnormal electrocardiography at time of screening), endocrinological (including metabolic), oncological, or autoimmune diseases; (D) history of major psychiatric disorder (other than substance abuse or dependence for the cocaine subjects and/or nicotine dependence for both groups); (E) pregnancy as confirmed with a urine test in all female subjects; (F) contraindications to the MRI environment; and (G) except for cocaine in the cocaine subjects, positive urine screens for psychoactive drugs or their metabolites (amphetamine/methamphetamine, phencyclidine, benzodiazepines, cannabis, opiates, barbiturates and inhalants). For Sample 2, history of glaucoma was added as an additional exclusionary criterion because of methylphenidate administration.
Sample 1 comorbidities were as follows: one subject also met criteria for current marijuana dependence, and 23 cocaine subjects were current cigarette smokers. Past comorbidities were identified in 25 cocaine subjects and included fully sustained remission for alcohol (N=18), marijuana (N=11), opioid (N=1), hallucinogen (N=1), other stimulant (N=1) or polysubstance (N=2) use disorders. Sample 2 comorbidities were as follows: one cocaine subject also met criteria for current heroin dependence, and 10 cocaine subjects were current cigarette smokers. Comorbidities in remission among the cocaine subjects included alcohol (N=8) and marijuana (N=6) use disorders.
Task Training
Just before the fMRI task, subjects completed extensive training to decrease performance differences between the groups. Training consisted of at least two complete task runs with different color randomizations, once outside and once inside the scanner (one Sample 1 cocaine subject did not perform the training inside the scanner). After training, all subjects had achieved accuracies of ≥ 50%.
Calculation of Post-Conflict and Post-Error Slowing
Post-conflict slowing was calculated as ([iC+cI] – [cC+iI], where iC=congruent trial preceded by an incongruent trial, cI=incongruent trial preceded by a congruent trial, etc.; note the iI were null events, as in our task the incongruent events did not occur back-to-back) and post-error slowing (i.e., comparing RT for trials after errors versus RT for trials after correct responses)6. Following established procedures6, for the post-conflict slowing we excluded trials with an exact stimulus repetition (in color, word, or both) across consecutive trials, which could artifactually influence post-conflict slowing scores7. For post-error slowing, we calculated three scores: (A) slowing after an error when the current trial is a congruent event, (B) slowing after an error when the current trial is an incongruent event, and (C) their sum. Post-error slowing is an adaptive, corrective response that is thought to enable more controlled responding to prevent future errors8, 9.
Task-Related Ratings (Sample 1 Only)
To bolster the case for fatigue, we further examined task-related ratings that were obtained while these same Sample 1 subjects performed a drug Stroop task, which has been extensively described elsewhere10-13. All subjects performed the drug Stroop task before the color word Stroop task (drug Stroop data are not presented here). In particular, we inspected self-reported ratings of ‘sleepiness’ (a proxy for fatigue) (‘‘how sleepy are you right now?’’) (response scale: not at all to very much, 0 to 10) and parallel ratings for ‘interest’ (a proxy for vigilance and attention) (‘‘how interested are you in the task right now?’’) (same response scale). To obtain a measure of how much fatigue (or interest) increased during this drug Stroop task, we subtracted sleepiness ratings collected at the beginning of the drug Stroop task from the sleepiness ratings collected at the end of this task (the latter rating collected immediately before beginning the color word Stroop task). Both ratings were obtained using custom programs written in C++ and were presented through MRI compatible goggles.
Methylphenidate Procedures (Sample 2 only)
During two fMRI scanning sessions on two separate study days, oral methylphenidate (20 mg) or placebo (lactose) was administered in a counterbalanced fashion across all subjects (note that there were no differences between the groups in number of days between the methylphenidate and placebo scans (cocaine: 8.8±2.9; control: 20.4±16.8; P>0.1). Oral methylphenidate or placebo was given 90 minutes prior to completion of the color word Stroop task, within the window of its peak effects (60-120 minutes)14. Measures of cardiovascular functioning (heart rate, blood pressure) and self-reports of methylphenidate effects were collected throughout the study. Heart rate was monitored at baseline (pre-medication), 45 min post-medication, 120 min post-medication, and post-fMRI as part of medical clearance. Blood pressure was taken at baseline and post-fMRI (in conjunction with the first and fourth heart rate measurements). Subjects also completed the Profile of Mood States, for which they provided self-report ratings (0-10, “How do you feel right now?”) for the dimensions of “high” and “methylphenidate desire.” These self-report measures were collected pre-medication, 45 min post-medication, and 120 min post-medication. Results of these measures are reported elsewhere15, showing that this oral dose of methylphenidate does not increase craving in cocaine subjects.
To comprehensively monitor the stimulant effects of methylphenidate (i.e., elevated cardiovascular reactivity, of special concern to a cocaine addicted population), study personnel were not blinded to the administered challenge during running of most subjects (N=25). Once it became clear that risks were minimal, we transitioned to double-blind MPH administration (N=4, two of whom were controls). However, even with single-blind administration, there were no differences between study days in post-fMRI guesses for the medication received (guess methylphenidate vs. placebo; χ2(1)=0.0, P>0.6), indicating that subjects were not fully aware of the exact type of medication received. Thus, we included these four subjects to maximize the number of subjects available for analysis. We also accounted for potential effects of single-blind versus double-blind medication administration as described below.
Covariate Analyses
SPSS analyses were used to control for variables that differed between the groups (Table 1, main text). For these covariate analyses, we inspected associations between the respective covariate and our dependent variables of interest [regions of interest (ROIs), behavioral measures]; if significantly correlated across all study subjects (P<0.05), these variables were entered as covariates in the relevant SPSS ANOVA or as control variables in partial correlations as appropriate16. All continuous and normally distributed variables were inspected with parametric tests (within groups: paired t-test; between groups: independent t-tests; correlations: Pearson r). Variables that were not normally distributed were inspected with the respective non-parametric tests (Wilcoxon, Mann-Whitney U, or Spearman r).
SUPPLEMENTAL RESULTS
Sample 1: Additional Behavior
Given our central interest in the neural response to error, in addition to the analyses described in the main text we performed a 2 (repetition) × 2 (congruency) × 2 (correctness) × 2 (group) mixed ANOVA for RT. This analysis revealed only a main effect of congruency as expected (P<0.001); no main effects or interactions were observed for correctness (i.e., RT for error trials versus RT for correct trials) (F<2.1, P>0.1). Thus, it is unlikely that the fMRI effects reported in the main text are attributable to differences in RT.
Although post-error slowing previously has been linked to adaptive, top-down control8, 9, we nonetheless endeavored to link it to task performance in the current study. Because the overall number of errors was not large (see Table 1, main text) – meaning it was not feasible to analyze back-to-back errors in this sample – we tested for correlations between post-error slowing with RT and task performance across the task. In the cocaine subjects only, there was a trend for those who showed the greatest increase in task errors (fourth repetition>first repetition) to also show the greatest decrease in respective post-error slowing (r=-0.48, P<0.05). Because this effect did not meet the nominal P<0.01 significance level established for correlations in this study, effects of post-error slowing in the current study should be interpreted with caution as also indicated in the main text.
Sample 1: Additional SPM
Across task repetitions and groups, and for both the ‘congruency’ and ‘correctness’ main effect contrasts, there were significant activations in multiple brain regions previously reported to be engaged by the color-word Stroop task6, 17 (Figure S1A-B). Such results contribute to a long-standing effort of using the color-word Stroop task to interrogate the prefrontal cortex in a range of subject populations that include healthy controls18, stimulant dependent populations19, schizophrenia patients20, and even relatives of affected probands21.
Sample 1: Effects of Covariates
Covarying out smoking history, which differed between the study groups (Table 1, main text), did not attenuate the post-error slowing repetition main effect or repetition × group interaction (P<0.05), the dACC repetition main effect (P<0.01), or the midbrain repetition × group interaction (P<0.01). The latter suggests that our results cannot be attributed to the desensitizing effects of cigarette smoking on midbrain dopamine neurons22. Covarying out age, which also differed between the groups (Table 1), did not attenuate the post-error slowing main effect of repetition (P<0.05) or the midbrain repetition × group interaction (P<0.001); however, covarying out age did attenuate the post-error slowing repetition × group interaction (P>0.07). Other effects did not require covariate analyses, as they were not associated with cigarette smoking or age16. In addition, note that the midbrain repetition × group interaction remained significant when dividing the cocaine subjects into those testing positive or negative for cocaine in urine23, 24, or when dividing the cocaine subjects into those with or without current cocaine dependence (with the latter encompassing cocaine abuse, remission, and polysubstance abuse; see above).
Sample 1: Effects of Second and Third Repetitions
Here we reanalyzed our main results from the main text (increased errors, decreased post-error slowing, decreased dACC activity to error, repetition × group interaction in the midbrain) while also including the second and third task repetitions, therefore testing for graded effects as a function of time-on-task.
Errors
For task errors, we expected significant repetition-related increases (Repetition 1 < Repetition 2 < Repetition 3 < Repetition 4) across all subjects. Consistent with hypotheses, the linear contrast was significant across the congruent and incongruent trials, and across all subjects, indicating that subjects committed progressively more errors throughout the task [F(1,51)=20.0, P<0.001]. There was also a congruency × repetition linear contrast interaction [F(1,51)=22.0, P<0.001], such that this linear increase in errors was significant only during the congruent trials [F(1,51)=21.8, P<0.001] (Figure S2A). In addition to supporting the effects above, this finding again speaks against alternative explanations of our results (e.g., practice effects).
Post-Error Slowing
For post-error slowing, we expected a repetition × group interaction, such that repetition-related decreases (Repetition 1 > Repetition 2 > Repetition 3 > Repetition 4) would be more pronounced in the healthy control subjects. Consistent with hypotheses, the linear contrast was significant for the repetition × group interaction [F(1,41)=7.9, P<0.01]. In particular, the descending linear contrast was significant only in controls [F(1,17)=16.9, P<0.01] (Figure S2B), indicating that only this group progressively decreased their post-error slowing throughout the task (although inspection of the means indicates a possible floor effect in the cocaine subjects).
SPM
To examine all four repetitions in the dACC/supplementary motor area and midbrain, we estimated a 4 (repetition: first, second, third, fourth) × 2 (group: control, cocaine) mixed ANOVA in SPM. Analyzing all repetitions resulted in 27/33 cocaine subjects and 14/20 controls included in this model. We then extracted the error-induced BOLD signal in the same peak coordinates in the dACC/supplementary motor area and midbrain that were found for the main analyses (i.e., those that emerged when analyzing the first and last task repetitions; see Table 2, main text, for peak coordinates) for subsequent analysis in SPSS.
dACC. For the dACC, we expected significant repetition-related activation decreases (Repetition 1 > Repetition 2 > Repetition 3 > Repetition 4) across all subjects. Consistent with hypotheses, the descending linear contrast was significant across all subjects for all three dACC/supplementary motor area peak coordinates [Fs(1,39)>8.4, P<0.01] (Figure S2C). These results indicate that activity in the dACC (and supplementary motor area) progressively decreased with repetition throughout the task in all subjects, further buttressing the idea of mental fatigue (which occurred in a graded fashion as a function of time-on-task).
Midbrain. For the midbrain, we expected a repetition × group interaction, such that repetition-related activation decreases would emerge in the cocaine subjects, while repetition-related activation increases (Repetition 1 < Repetition 2 < Repetition 3 < Repetition 4) would emerge in controls. Indeed, the repetition × group linear contrast interaction in the midbrain was significant [F(1,39)=9.4, P<0.01], such that the cocaine subjects progressively decreased response in this region with repetition as hypothesized [F(1,26)=16.3, P<0.001] (Figure S2D). Although the healthy controls did not progressively increase response in this region with repetition [F(1,13)=0.8, P>0.3), the significant omnibus interaction directly supports the findings for the first and last task repetition.
Sample 2: Behavior
Because methylphenidate previously has been shown to facilitate task performance in cocaine addicted individuals15, 25 and even in healthy controls15, here we tested its impact on performance of the color word Stroop task. We were specifically interested in testing for methylphenidate modulation of task repetition effects [note that main effects of methylphenidate (i.e., collapsed across all task conditions and repetitions) will be reported separately].