|Home | About | Journals | Submit | Contact Us | Français|
Impulsivity is common in bipolar disorder, especially during mania. Understanding the functional neuroanatomy of response inhibition, one component of impulsivity, might clarify the neural substrate of bipolar disorder.
Sixteen DSM-IV first-episode, manic bipolar patients and 16 matched healthy subjects were examined during a first manic episode using fMRI while performing a response inhibition task. All subjects were studied using a 4 T Varian INOVA whole body MRI system. The response inhibition task was presented using nonferromagnetic goggles and task performance was recorded during scan acquisition. Imaging data were analyzed using AFNI. Group contrasts were made for the specific response inhibition measure.
The groups performed the task similarly, although both demonstrated relatively poor rates of target response, but high rates of successful ‘stops.’ Despite similar behavioral results, the groups showed significantly different patterns of fMRI brain activation. Specifically, during response inhibition the healthy subjects exhibited significantly greater activation in anterior and posterior cingulate, medial dorsal thalamus, middle temporal gyrus and precuneus. The bipolar patients exhibited prefrontal activation (BA 10) that was not observed in healthy subjects.
Bipolar and healthy subjects exhibit different patterns of brain activation to response inhibition; these differences may reflect different functional neuroanatomic approaches to response inhibition between the two groups.
Impulsive behaviors commonly occur in the course of bipolar disorder and in unaffected relatives of bipolar probands.1–3,24,25 Some measures of impulsivity may vary with affective state.3,26 However, impulsivity is not a singular behavior, but is a multi-faceted construct. One aspect of impulsivity is the inability to inhibit a prepotent behavioral response in order to make a more favorable or correct response. Recently, we observed that euthymic bipolar patients exhibited an impulsive response bias during a counting Stroop task consistent with these considerations.4 Specifically, bipolar subjects did not slow their responses during the interference (Stroop) condition, and consequently exhibited more errors; i.e., they did not inhibit the prepotent response (reading words) in favor of a correct response (counting words). Moreover, consistent with these behavioral data, brain activation differences between healthy and bipolar subjects while performing this task suggested a pattern of relatively decreased activation in the bipolar group in brain regions thought to be involved in error detection, e.g., temporal cortical areas, and ventrolateral prefrontal cortex (VLPFC). These findings suggest the hypothesis that dysfunction in brain networks that manage response inhibition may contribute to the expression of bipolar symptomatology.
The current study was performed to extend this work and test this hypothesis in two ways. First, we studied bipolar subjects during first-episode mania to clarify the functional neuroanatomy of response inhibition during that phase of illness, since impulsivity is a common feature of mania.1–3 Consequently, we predicted that the patients would demonstrate a more impulsive response bias on the task. Second, we used a task specifically designed to examine response inhibition. The basis of the task is the so called CPT-X, namely a simple continuous performance task which is purposefully undemanding (see reference 27 for a review of this approach). We elected this simple task as we did not want the attentional component to be primary, but it provided a substrate to compare to other CPT-based fMRI work.7 We then incorporated a ‘stop signal’ component; namely the target changed color, thereby requiring the subjects to inhibit the prepotent target response. The percent of ‘correct stops’ within the context of ‘correct target hits’ provides a measure of inhibitory control to extend the standard CPT-X design. We studied first-episode patients to minimize the effects of disease progression and prior medication exposure. Based on our overall hypothesis, we predicted that, compared with healthy subjects, bipolar patients would show decreased activation in brain regions associated with response inhibition, i.e. VLPFC and temporal cortical areas.4 Additionally, we hypothesized that this task would differentiate anterior cingulate activation between groups that was not observed with our Stroop task.4,15–19
Nineteen patients with DSM-IV type I bipolar disorder were recruited from consecutive admissions at the time of a first hospitalization for mania from the University Hospital and the Cincinnati Children’s Hospital Medical Center, Cincinnati, OH. To be included bipolar subjects: 1) met DSM-IV criteria for a current first manic or mixed episode that required hospitalization; 2) had no previous psychiatric hospitalizations; and 3) had less than one month of lifetime prior thymoleptic or antipsychotic treatment. Seventeen demographically similar healthy comparison subjects were recruited from the same community as the bipolar subjects by word-of-mouth and advertising in local media. Healthy subjects had no personal or first-degree family history of Axis I psychiatric disorders. Both bipolar and healthy subjects: 1) were 15 to 35 years old; 2) were medically and neurologically healthy; 3) did not meet DSM-IV criteria for a current substance dependence disorder within 3 months of the scan; and 4) had no contraindications to MRI (e.g., pregnancy, ferromagnetic implants). The patient and healthy subjects were closely matched on demographic variables (Table 1), although, by definition, the patients exhibited significantly more affective symptoms and co-occurring disorders. These subjects are a different sample than reported in previous fMRI studies from these investigators (e.g., 4–7). All subjects or legal guardians (if <18 years old) provided written informed consent (and assent if <18 years old) after procedures were explained in full. The study was approved by the institutional review boards of the participating institutions. One healthy and three bipolar subjects were not included due to technical problems with either scan or behavioral data acquisition or excessive motion (>5 mm); therefore, the final comparisons included 16 bipolar and 16 healthy subjects.
A diagnosis of DSM-IV type I bipolar disorder currently manic or mixed in patients, or the absence of a psychiatric diagnosis in healthy subjects, was determined using the Structured Clinical Interview for DSM-IV, patient version (SCID-I/P).8 All subjects were administered the Young Mania Rating Scale (YMRS)10 and Hamilton Depression Rating Scale (HDRS)11 at the time of the MRI examination. We have established good reliability with these instruments.9 Substance use disorders were identified using the SCID-I/P. Additionally, urine toxicology screens were obtained to verify self-report information. Negative pregnancy tests were obtained on all female subjects prior to scanning. All subjects were right-handed as determined by the Crovitz Handedness Scale.12 At the time of the MRI scan, 8 (50%) of patients had received medications, although only for a few days or less (typically ≤ 4 days) since hospital admission, and all patients were symptomatic at the time of the scan. Medications received included: risperidone (n=4), aripiprazole (n=1), quetiapine (n=1), ziprasidone (n=1), divalproex (n=1), sertraline (n=1), venlafaxine (n=1), and mixed amphetamine salts (n=1); some patients were receiving more than one medication.
During the MRI scan, all subjects performed a response inhibition task, based on a standard ‘stop-signal’ design. In this task, the target (‘go’ cue) was a blue X presented in the center of the visual field. The ‘stop’ signal was a red X that replaced the blue X after variable time intervals (0, 50, 100 or 150 ms). The Xs were randomly interspersed with other blue letters that were not targets. All stimuli were presented for 450 ms, with a 50 ms interval between presentations. Subjects were instructed to press a button as quickly as possible every time a blue X occurred, but to resist responding if the color of the target changed to red. A total of 1919 trials in a 16-minute fMRI scanning run were presented, including 197 targets (blue X), and 64 ‘stop’ trials (red X). Responses were collected using a MRI-compatible button box.
Subjects were scanned at the University Of Cincinnati College Of Medicine’s Center for Imaging Research (CIR) using a 4.0 Tesla Varian Unity INOVA Whole Body MRI/MRS system (Varian Inc., Palo Alto, CA). The INOVA system is controlled by a SUN workstation running Varian’s VNMR-J™ and SPIN-CAD™ image processing and pulse-sequence development software under a Unix-based operating system. During the scan sessions, subjects reclined in a supine position on the scanner bed. Subjects held a button box in the right hand. Nonferromagnetic goggles (Resonance Technologies, Inc.) were positioned to provide clear visualization of the stimuli (i.e., the response inhibition task) with a visual resolution of 600 × 800 pixels and a 60 Hz refresh rate. The goggles were connected to a Macintosh computer running PsyScope13 to present the stimuli. Padding was inserted around the subject’s head to minimize movement. Headphones were provided to block background noise and so that investigators could communicate with the subjects during scan acquisition.
Following a three-plane gradient echo scan for alignment, a high-resolution, T1-weighted, 3-D brain scan was acquired for anatomic localization using a modified driven equilibrium Fourier transform (MDEFT) sequence (TMD=1.1 s, TR=13 ms, TE=6 ms, FOV=25.6 × 19.2 × 19.2 cm, matrix 256 × 192 × 96 pixels, flip angle=20 degrees). A midsagittal localizer scan was obtained to place 30 contiguous 5mm axial slices that extended from the inferior cerebellum to encompass the entire brain. While performing the response inhibition task, subjects completed an fMRI session in which scans were acquired using a T2*-weighted gradient-echo echoplanar imaging (EPI) pulse sequence (TR/TE=2000/30 ms, FOV=25.6 × 25.6 cm, matrix 64 × 64 pixels, slice-thickness=5 mm, flip angle=75 degrees).
We used motion correction parameters to calculate the total amount of motion in 6 directions of rotation and translation from the beginning to the end of each run. The maximum motion in any subject included in the analysis was <5 mm. There were no significant differences between groups in movement using this measure (p>.05). The average total displacement for all subjects was <2 mm. We also calculated the amount of displacement in all 6 directions between every TR, and the TR that was used as a reference point. The average displacement between any given TR and the reference was 0.45 mm.
The fMRI data were analyzed using AFNI (Analysis of Functional NeuroImages; http://afni.nimh.nih.gov/afni). The behavioral and demographic data were analyzed using SAS (Statistical Analysis System, SAS Institute, Cary, NC). Following acquisition, the MRI images were reconstructed using in-house software developed in IDL (Interactive Data Language), which converts raw FID files into AFNI format. In AFNI, MDEFT (structural) and EPI (functional) images were co-registered using scanner coordinates. Functional images were corrected for motion using a six-parameter rigid body transformation (14). Using tools in AFNI, anatomical and functional maps were transformed into sterotaxic Talairach space and spatially blurred to twice the voxel dimensions. Binary masking was applied to each image to remove pixels outside the brain. Individual activation maps were then created for each subject using a deconvolution algorithm that compares the actual hemodynamic response to a canonical hemodynamic response function (e.g., gamma function), creating voxelwise t-maps. Event-related hemodynamic response functions were calculated for correct ‘stops’ relative to non-targets as the direct measure of response inhibition. Motion correction parameters were included as regressors of no interest. Additionally, low frequency components of the signal, including linear, quadratic and cubic drift, were removed.
Analyses were performed voxelwise over the whole brain using AFNI. AFNI generates an estimate of the ‘fit coefficient’ (i.e., beta weight or scaling factor) describing the magnitude of the hemodynamic response relative to the average signal intensity for each event. Individual activation maps were compared across subjects and between groups. One-sample t-tests were performed for each subject based upon the null hypothesis of zero activation changes for correct ‘stops’ versus non-targets for manic versus healthy subjects. Group activation maps were thresholded at a corrected p<0.05, which was obtained with a voxel-level p<0.025 and a cluster of 44 contiguous voxels; these values were determined by Monte Carlo simulation tools in AFNI for both within and between group analyses. Regression analyses were performed to contrast healthy and manic subjects.
Behavioral response measures are shown in Table 1. No statistically significant differences were observed between groups on any response measure. Both groups showed relatively poor response rates to targets, with relatively high rates of correct ‘stops.’
Healthy subjects exhibited significantly increased regional activation (Table 2, Figure 1) during response inhibition in bilateral anterior cingulate (BA 24, 32), right inferior frontal gyrus (BA 9), bilateral cingulate (BA 23), bilateral medial dorsal thalamus, bilateral insula, bilateral middle temporal gyrus (BA 21), and bilateral precuneus/inferior parietal lobule (BA 7, 19, 39). Significantly decreased activation was observed in left middle temporal gyrus (BA 39) and posterior cingulate (BA 29), predominantly on the left and posterior and inferior to the region of increased activation noted previously.
In contrast, the bipolar subjects exhibited significantly increased activation (Table 2, Figure 2) only in left middle frontal gyrus (BA 10) and bilateral precuneus/inferior parietal lobule (BA7, 19, 39). Similar to healthy subjects, the patients demonstrated decreased posterior cingulate (BA 29) activation, predominantly on the left, although the activation extended somewhat superiorly to that observed in healthy subjects. They also similarly demonstrated decreased activation in left middle temporal gyrus (BA 39), although somewhat more extensive inferiorly than healthy subjects.
Healthy subjects demonstrated significantly greater activation than bipolar patients in left anterior cingulate (BA 32), medial dorsal thalamus predominantly on the left, precuneus (BA 7, 39) predominantly on the left, and left middle temporal gyrus (BA 21, 37). Healthy subjects also exhibited significantly less deactivation in bilateral posterior cingulate (BA 23, 29) than bipolar patients. None of the contrast effects result from sub-threshold group differences because all regions of significant contrast are sites of activation or deactivation in one or the other of the single group images. There were no areas of greater activation in the bipolar compared to healthy subjects.
We contrasted fMRI activation in the 8 bipolar patients who were medication free at the time of the scan to the 8 patients who were not, using the same analytic methods as previously described, except with a more liberal uncorrected p<.05 to define areas of significant activation differences. Due to the small numbers of subjects, this secondary analysis should be viewed as exploratory only in order to interpret the primary findings (Figures 4). Patients with no medication exposure exhibited greater activation in the right amygdala (32 -6 -15), right insula (40 -1 9), and left inferior parietal lobule (BA 40; -57 -31 33). Patients receiving medications (though, as noted, for only a few days) demonstrated increased activation in posterior left middle temporal gyrus (BA 39; -34 -70 24) and bilateral precuneus (BA 19; -22 -78 27, 25 -82 19). Activation coefficients from AFNI are plotted in Figure 5 among subgroups to illustrate these differences between groups. The unmedicated and medicated patients demonstrated similar reaction times [573 (26) vs. 577 (37) ms respectively; t=0.24, p=.81] and discriminability [0.89 (0.08) vs. 0.86 (0.04); t=1.1, p=.31] on the task; however, the unmedicated patients exhibited nonsignificantly lower bias [0.08 (0.05) vs. 0.13 (0.05); t=2.1, p=.06].
In this study, we examined fMRI regional brain activation during a response inhibition task in first-episode manic bipolar patients and demographically similar healthy subjects. Significant differences between groups were observed in regional brain activation, despite no differences in task performance. The lack of differences in task performance failed to support our initial prediction of greater impulsive responding in the manic patients. Specifically, both groups exhibited relatively poor rates of response to targets (about 50% correct hits) with correspondingly high rates of successful ‘stops’ (i.e., low rates of false hits). This pattern of responses suggests two likely behavioral strategies. The first is that subjects excessively delayed responses to targets in order to avoid failing ‘stops,’ resulting in high rates of missed targets. Supporting this interpretation is the observation that the response times were slightly longer than the presentation interval, suggesting that subjects in both groups were delaying the response until the possibility of the target changing had passed. Alternatively, subjects may have inconsistently attended to the task, thereby increasing rates of non-response to targets, which would also decrease the rates of false hits during ‘stops.’ Our task design does not allow us to determine whether one of these behavioral strategies predominated in either group. Moreover, subjects were not questioned regarding the cognitive strategy they used when completing this task. These possibilities become relevant when interpreting fMRI data.
The pattern of regional activation in healthy subjects suggested active response inhibition, consistent with the first of the paradigms discussed in the previous paragraph. Namely, activation in regions of the cingulate (BA 24, 32, 23), insula, thalamus, inferior frontal gyrus and precuneus is consistent with previous studies of similar response inhibition tasks15–19 We also observed relative deactivation within the posterior cingulate; this may reflect that the posterior cingulate is more activated during search for targets (i.e., when viewing the series of non-targets), than during response inhibition, as has been suggested in other studies.18 The activation pattern observed in healthy subjects is consistent with an activation of response inhibition networks and an active response inhibition behavioral strategy during this task.
In contrast, the bipolar subjects exhibited limited activation in cingulate, inferior frontal gyrus, and thalamus. Like the healthy subjects, the patients exhibited similar patterns of decreased activation in regions of the posterior cingulate (BA 29) and left middle temporal gyrus (BA 39). Unlike the healthy subjects, the bipolar patients also exhibited activation in middle frontal gyrus (BA 10), although the difference between groups in this region did not meet a priori statistical significance. Since task performance was similar to healthy subjects, several possibilities are suggested by the differences in activation patterns. First, as noted previously, the differences in activation may reflect a different underlying reason for the relatively low rate of target hits; namely, rather than choosing a conservative strategy as suggested by the healthy subjects (i.e., delayed response to prevent false hits during ‘stops’), the similar task performance in the patients may have resulted from poorer attention to the task. The lack of cingulate activation supports this notion. On the other hand, the similar reaction times and measures of bias (B″) suggest that the two groups were approaching the task similarly vis-à-vis their behavioral response strategy. Bias represents factors other than the ability to discriminate targets from non-targets during the CPT that affect performance. These factors include motivation, fatigue or tendencies toward either more liberal or conservative task performance. The recruitment of BA 10 by the patients suggests instead that patients may have used a compensatory neural strategy to manage the demands of this task, even while using a similar behavioral strategy (i.e., delaying target response in order to maximize ‘stops’). This brain region has been associated with decision making, affective modulation, and conflict resolution, and may be activated as tasks become more difficult.22, 23 Alternatively, medication effects may account for some of the differences between bipolar and healthy subjects. The patients who had not received medication demonstrated greater activation in brain areas that have been associated with affect modulation, namely amygdala and insula,6 than patients who had received medications, even though they had only received treatment for a few days. The patients who received medications showed increased activation in posterior attentional areas. In previous work in unmedicated euthymic patients, we suggested that activation of posterior attentional areas compensates for interference of cognitive networks by over-activated mood networks.6 Therefore, the increased activation noted here, even after only a few days of medication, may be an early marker toward clinical improvement. Notably, in this secondary analysis between medicated and unmedicated patients there were no differences in activation in brain regions that differed between healthy and bipolar subjects in the primary analysis. Nonetheless, the possibility that medication-related effects might have indirectly contributed to differences between bipolar and healthy subjects cannot be excluded.
Several limitations to this study should be considered when interpreting results. Additional cognitive tasks (e.g., an attentional task such as a continuous performance taske.g.7) may have helped to better differentiate the underlying reasons for similar task performance with different brain activation patterns. Related to this limitation, the task selected did not differentiate between healthy and manic subjects. An alternative design, or additional or more difficult tasks, might have better identified specific behavioral deficits related to response inhibition in bipolar disorder, which would then be reflected in brain activation patterns. Specifically, this task appeared to be less cognitively demanding than those used in other studiese.g.3. On the other hand, the lack of task performance differences controls for this potential confound when interpreting differences in fMRI activation patterns. An alternative approach would be to use a Logan stop-signal task,28 which dynamically adjusts the stop-signal based on individual subject response times. Doing so would control for inhibition success across groups; however, in this study, no differences in successful stop rates were observed. Another limitation was that the number of subjects was relatively small, particularly for sub-analyses (i.e., patients on and off of medication). Consequently, these secondary analyses in particular should be interpreted cautiously and additional differences between groups may have been observed with larger numbers of subjects. In the absence of larger numbers of subjects, specific medication effects could not be delineated. Nonetheless, the preliminary contrasts suggest that medication exposure does not explain differences observed between the bipolar and healthy subjects. The bipolar subjects were more likely to have histories of co-occurring psychiatric disorders. However, the numbers of subjects with each precluded meaningful analyses. Similarly, many of the patients exhibited psychotic symptoms as is common during mania. Whether some differences could be ascribed to specific symptoms or co-occurring syndromes cannot be discounted; larger samples might be able to address these possibilities. Balancing these limitations is the unique, early course patient sample scanned while manic and while using a well-defined response inhibition task. Future studies incorporating additional cognitive tasks with this patient population may extend this work to clarify the functional neuroanatomy of response inhibition in bipolar disorder, in general, and mania, specifically.
This study was funded by support from the Stanley Medical Research Institute, the NIMH (MH071931, MH066626), and the Craig and Francis Lindner Center of Hope Foundation.
Disclosures: In addition to the funding that supported this work directly, one or more of the investigators received grant support to the University of Cincinnati Academic Health Center in the past year for other projects from Abbott Laboratories, AstraZeneca, Eli Lilly, Forest, GlaxoSmithKline, Orexigen, Ortho-McNeil, Johnson and Johnson, Shire, Janssen, Pfizer, Bristol Myers Squibb, Repligen, Martek, NIDA, NIAAA, NARSAD, Thrasher Foundation. Additionally, although not directly related to any of the results, in the past year the following investigators received honoraria for speaking or consulting from: Dr. Strakowski – Pfizer (Kendle), Eli Lilly, Tikvah, France Foundation (CME company), DiMedix (CME company); Dr. DelBello - Bristol-Myers Squibb, AstraZeneca, France Foundation, GlaxoSmithKline, Eli Lilly, Kappa Clinical, NIDA, and Pfizer; Dr. Adler – AstraZeneca.