Supplementary Online Content€¦ · MRI-scanning contra-indications (a history of head trauma or...

©2015AmericanMedicalAssociation.Allrightsreserved.

Supplementary Online Content

Rive MM, Mocking RJT, Koeter MWJ, et al. State-dependent differences in emotion regulation between unmedicated bipolar disorder and major depressive disorder. JAMA Psychiatry. Published online May 6, 2015. doi:10.1001/jamapsychiatry.2015.0161.

eAppendix. Methods, Data Analysis, Results, and Discussion eTable 1. Mean Ratings of the Pilot Study Conducted to Obtain Valence, Arousal, Complexity, and Emotional Intensity Values for Matching Purposes eTable 2. Handling of Subjects Demonstrating an Increase in Emotion Intensity During Regulation eTable 3. Post-Scan Task: Ratings for Valence, Arousal, and Intensity of Corresponding Emotion eTable 4. Post-Scan Task: Ratings for Discrete Emotion Types eTable 5. Regulation Success Scores During Scanning (Scale: 0-100) eTable 6. Main Task Effects Across Subjects and Emotion Type eTable 7. Activity Differences for Regulate > Attend Between MDD and BD eTable 8. Activity Differences Between MDD and BD With Correction for the Number of Previous Episodes eFigure 1. Design of the Emotion Regulation Task eFigure 2. Main Task Effects Across Subjects and Emotion Type eFigure 3. Relationship Between Regulation Success for Happy Pictures and Sad Intensity Ratings for Happy Pictures in MDDd eFigure 4. Relationship Between rACC Contrast Estimates and Mean Happy Intensity Ratings Across Depressed Patients (MDDd + BDd) eReferences

This supplementary material has been provided by the authors to give readers additional information about their work.

Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 06/20/2020


eAppendix. Methods, Data Analysis, Results, and Discussion Methods Subjects: recruitment Included depressed subjects generally visited their physician (psychiatrist/general practitioner) because of depressive relapse after quitting medication. At that moment their physician either asked him/her to contact us, or asked him/her permission to refer to us. Subjects were assessed and scanned as soon as possible (if needed, within one week) to prevent treatment delay. Remitted subjects usually responded to advertisements or were referred to us by their treating physician. Subjects: in- and exclusion criteria

Inclusion criteria for patients were: MDD or BD-I/BD-II diagnosis according to the Structured Clinical Interview for DSM-IV (SCID-I)1; age 18-60 years; age at first mood episode ≤40 years; current Major Depressive Episode (MDE) or remission (according to SCID-I criteria); ≥2 MDEs; illness duration of ≥5 years since the first episode (reducing late MDD-to-BD conversion risk).

Exclusion criteria for patients were: current (hypo)mania; comorbid axis-I disorder (SCID-I), except for anxiety disorders; cluster-B personality disorder (if suspected assessed by SCID-II2); current use of antidepressants, anticonvulsants or mood-stabilizers (or stopped <1 month before scanning) or electroconvulsive therapy ≤2 months before scanning. We allowed incidental benzodiazepine use (stopped ≥1 day before scanning). MDD-subjects were excluded when they reported (hypo)manic derailment after antidepressant use or a first-degree family member with BD (excluding MDD subjects with BD-traits). BD-subjects were excluded when (hypo)manic (Young Mania Rating Scale (YMRS3) score >8) or when (hypo)manic episodes appeared only during antidepressant-use (ensuring inclusion of ‘true’ BD).

Exclusion criteria for HC (age 18-60 years) were: any lifetime psychiatric diagnosis; first-degree relative with psychiatric diagnosis, use of any psychopharmacological agent.

MRI-scanning contra-indications (a history of head trauma or neurological disease; severe general physical illness; claustrophobia or implanted metal objects) were a general exclusion criterion. Emotion regulation paradigm Paradigm The stimulus set consisted of pictures 4 stimulus types (sad, happy, fearful, neutral). Emotional pictures were viewed under two conditions (attend, regulate). The neutral trials provided a buffer to prevent habituation to the emotional pictures; they were only attended. This yielded 7 conditions (sad/happy/fear: regulate/attend, neutral: attend). Pictures were derived from the International Affective Picture System (IAPS4) and http://nl.dreamstime.com; each set was matched for valence, arousal and content (see below). The pictures were presented in a semi-blocked pseudo-randomized design. Each block started with the instruction presented in the middle of the screen (4s), followed by 3 successive pictures of the same emotional category (10s each). In total 72 pictures were presented. After each picture, subjects indicated the emotional intensity resulting from attending or regulating on a Visual Analogue Scale (VAS). After each block subjects also rated their performance (i.e. how well they



were capable of attending or regulating). Blocks were separated by a fixation cross (4s) (eFigure 1). Instructions During the attend condition, subjects were instructed to experience any emotion elicited by the picture, without actively changing their feelings. During the regulate condition, subjects were instructed to distance themselves from the pictures (i.e. to become a detached observer) in order to achieve a neutral emotional state, by thoughts like: ‘This is only a picture’, ‘This will never happen to me’, ‘This situation has nothing to do with me’, ‘This picture is fake’, etcetera. To ensure that subjects adequately applied distancing, subjects were trained twice before scanning. Emotional stimuli IAPS pictures were preselected based on IAPS ratings (scale: 1-9) for valence (neutral: 4-6; positive, i.e. happy: >6; negative, i.e. fearful and sad: <4); arousal (neutral: <3; emotional, i.e. happy, fearful, or sad: >6) and furthermore, for ratings for emotion specificity as assessed by Mikels et al.5 (scale: 1-9: >7 for each specific emotion category; neutral: <3 for every emotion). In addition, we used stock photos from http://nl.dreamstime.com based on emotional content. In total, 110 pairs of pictures, matched for emotional content, were selected. To make matching between IAPS and Dreamstime pictures possible, an independent pilot study (N=41 healthy controls) was conducted. Subjects rated all pictures on valence and arousal (using the same Self Assessment Manikin (SAM)6 used for the IAPS database, ranging from unpleasant to pleasant for valence, and from calm to excited for arousal), emotion type (on a scale from 1 (emotion is not elicited at all) till 9 (emotion is elicited very strongly)) and complexity (on a scale from 1 (picture is very easy to interpret) till 9 (picture is very difficult to interpret)). Based on these ratings, eventually 36 sets of 2 pictures (9 sets for each emotional category) were selected. Within each pair, the pictures (one for the attend, one for the regulate condition) were matched for valence, arousal, complexity and emotional content. Mean ratings are presented in eTable 1. MRI scanning: acquisition parameters For the functional scan the following parameters were used: echo time 30 ms, repetition time 2300 ms, flip angle 80°, matrix 96x96, number of slices 40, slice gap 0 mm, slice thickness 3mm, ascending slice order, field of view 220x120x220 mm3, voxel size 2.29x2.29x3 mm3. For anatomical co-registration purposes, a 7-minute T1-weighted structural image was acquired (echo time 4.6 ms, repetition time 9.6 ms, flip angle 8°, matrix 256 x 256, number of slices 182, slice thickness 1.2 mm, field of view 218x256x256 mm3, voxel size 1x1x1.2 mm3). Data analysis Behavioral data ERT: Correction of behavioral data Within each emotion category, regulation success scores were calculated by expressing the decrease in in-scan VAS intensity scores (scale: 0=zero emotional intensity; 100=maximal emotional intensity) during regulation (i.e. attend-regulate scores) as a percentage of the VAS intensity scores during attending. Thus, regulation success scores represent the decrease in emotional intensity during regulation relative to baseline (attending) emotional intensity for a given emotion type (sad, happy, fear).



Because regulation is expected to decrease emotional intensity, the attend VAS score – regulate VAS score should be a positive value. However, some subjects (see eTable 2 below) demonstrated negative attend – regulate scores, indicating an increase in emotional intensity during regulation. Two explanations appear for this observation: 1. Subjects may have had trouble adequately using the VAS scale or did not correctly carry out the attend and regulation assignments (incidentally or structurally). If so, the scores of these subjects may be unreliable; 2.The scores accurately represent paradoxal increases in emotional intensity during regulation in these subjects. Then these scores cannot be used for the required arc sinus transformation, which is the method of choice to achieve a normal distribution of proportional data.

We therefore analysed the data in 2 ways: I. Assuming that the negative scores were due to incidental mistakes, we corrected

the negative scores (see below) and subsequently arcsine transformed the resulting values. The following correction criteria were used (see also eTable 2):

a. When these attend-regulate scores were in the range of -10 till 0, they were considered small and comparable to 0, indicating no difference between emotional intensity during regulating versus attending;

b. However, when they were <-10, the were considered flawed, because they indicated much greater emotional intensity while regulating than attending. Therefore, they were omitted.

The results of this analysis are described in the results section of the manuscript. II. Assuming that the large negative scores (<-10) were due to structural incorrect

performance, we ran the analysis without any of the subjects demonstrating large paradoxal emotional intensity increases (attend – regulate scores <-10).

Post-scan task To preclude that observed differences in emotion regulation success could be explained by differences in baseline emotional appraisal of the pictures, we investigated possible between-group differences. After scanning, each participant rated all pictures on on valence (ranging from unpleasant to pleasant), arousal (ranging from calm to exited) and intensity of five discrete emotions (sadness, happiness, fear, disgust, anger; ranging from 1 (emotion is not elicited at all) to 9 (emotion is elicited very strongly).

To assess whether groups differed in valence, arousal and intensity ratings regarding the emotion the picture was supposed to elicit (i.e. happiness intensity for happy pictures, sadness intensity for sad pictures, and fear intensity for fearful pictures), these scores were entered as dependent variables in separate linear mixed models with diagnosis (MDD, BD), state (depressed, remitted) and emotion (sad, fear, happy) as independent variables. Results were considered significant if surviving the Bonferroni correction for multiple (n=7) comparisons (i.e. p< 0.007).

Furthermore, we assessed whether groups differed with regard to the type of emotion the pictures of each emotion category elicited. For example, one group could have experienced sad pictures as purely sad, whereas another group experienced sad pictures as both sad and fearful, which could have influenced emotion regulation success. Mean and 95% confidence intervals (95-CIs) of each emotional intensity score (i.e. sadness, happiness, fear, disgust and anger intensity) were calculated for the pictures of each emotion category (i.e. sadness, fear, happiness) for each group (HC, MDDr, BDr, MDDd, BDd). Subsequently, within each group, the type of emotion elicited by the sad, happy and fearful pictures was defined according to criteria formulated by Mikels et al.: 1. an elicited emotion was considered discrete if



the mean for that emotion was higher than the mean of all other emotions and there was no overlap with other confidence intervals; 2. an elicited emotion was considered blended if the means of two or three emotions were higher than the rest and CIs overlap each other; 3. an elicited emotion was considered undifferentiated if there was overlap of all CIs.

fMRI-data Preprocessing and first-level models

Statistical Parametric Mapping (SPM8, Wellcome Trust Center for Neuroimaging, London, UK; http://www.fil.ion.ucl.ac.uk/spm/software/SPM8) was used for pre-processing: slice timing; realignment and unwarping for motion-correction; co-registration of functional and structural data; segmentation; spatial normalization (into Montreal Neurological Institute (MNI) space) and resampling to 3 mm isotropic voxels; and smoothing (8mm full width at half maximum Gaussian kernel).

Individual data were analyzed using the general linear model, comprising regressors for each picture (attend/regulate sad/fear/happy; attend neutral) and regressors of no interest (rating scales, instruction periods and movement parameters). To provide a low level baseline, fixation cross periods were not modelled. To remove low-frequency noise, a high pass filter (cut-off: 128 s) was applied. Secondary analyses of group differences associated with behavioral (incorrect task performance) and illness (number of previous major depressive episodes) related differences

To make sure results were not influenced by possible incorrect performance of the task, as indicated above, we repeated the analyses without the subjects demonstrating any negative attend – regulate score <-10 (i.e. 3 MDDd, 1 BDr and 1 BDd subject; see eTable 2). Furthermore, we repeated the analysis correcting for the number of previous major depressive episodes, since there was a difference in this number between MDDd and BDd. Analyses using a 2x2 factorial design and post-hoc t-tests Results could also be analysed with a diagnosis x state full factorial design to investigate comparisons of diagnosis, state, and their interactions, using post-hoc tests to compare each cell with HC. To this end, we used two separate full factorial models, one for each contrast (regulate>attend for happy versus sad; and regulate>attend across all emotions). For comparisons with HC, we used two-sample t-tests. Correlation analyses To obtain more insight in our results, we performed several correlational analyses. For details, see supplemental results below. For normally distributed data we report Pearson’s r, otherwise Kendall’s is used. Results and Discussion Subjects: reasons for exclusion The 6 subjects (1 HC, 3 MDDr, 1 MDDd, 1 BDd) were excluded for the following reasons: one subject was so overwhelmed by the sad pictures she could not complete the task; to one subject pictures were presented upside-down; one subject told us during the debriefing afterwards that she regulated her emotions during all conditions,



including the attend conditions; during the scanning session of one subject the task and scanner did not start simultaneously; fMRI data of one subject were incomplete; data of one subject were lost because the data storage hard disk crashed. Behavioral data ERT: correction of behavioral data (eTable 2) There were 16 subjects with at least one negative attend-regulate score, of whom 5 subjects (3 MDDd, 1 BDr and 1 BDd) demonstrated an attend-regulate score of <-10 (see eTable 2). Scores between -10 and 10 were set to 0, scores below <-10 were considered missing values. Results of the behavioral analysis using these corrected data are described in the manuscript.

Repeating the model without any of the subjects showing any attend-negative score <-10 yielded a pattern of results similar to results of the model with all subjects included (results available on request). Post-scan task (eTables 3 and 4) Due to tiredness, 94% of included subjects (MDDr: N=21; MDDd: N=21; BDr: N=22; BDd: N=9) performed the post-scan task. Because there were no between-group differences in regulation of fear, we do not report on baseline appraisal results of fearful pictures here (results available on request).

Only in the depressed subgroup, MDD and BD subjects differed with regard to appraisal of the pictures. With regard to valence, there was a significant emotion (happy,sad)*diagnosis (MDDd,BDd) interaction (t=3.18, p=0.002), explained by a non-significant higher valence rating for happy pictures (t= 2.52, p= 0.012, d=1.1) and a non-significant lower valence rating for sad pictures (t= 0.71, p=0.087, d=0.6) in BDd compared to MDDd (eTable 3).

Regarding happy or sad intensity ratings, differences between MDDd and BDd did not reach significance, but effect sizes were large: BDd demonstrated higher happy and sad intensity ratings (happy: t=2.23, p=0.03; d=0.95; sad: t=1.97, p=0.05, d=0.73), indicating that BDd subjects appraised both happy and sad pictures as more intense than MDDd (eTable 3). Emotional intensity ratings furthermore revealed that whereas BDd subjects appraised happy pictures as discretely happy (mean 4.6, 95%CI 3.2-6.0), MDDd subjects perceived happy pictures as happy blended with sad (happy: mean 3.0, 95%CI 2.2-3.7; sad: mean 1.8, 95%CI 1.3-2.3) (eTable 4).

Together, this suggests that BDd subjects appraised happy pictures as more positive and sad pictures as more negative than MDDd, and furthermore, that happy pictures may elicit negative (i.e. sad) in addition to happy emotions in MDDd, but not in BDd. Regarding arousal, there were no significant between-group differences (eTable 3). fMRI data Secondary analyses of group differences associated with behavioral (incorrect task performance) and illness (number of previous major depressive episodes) related differences

Repeating the analyses without any of the subjects with a negative attend – regulate score, i.e. an increase in emotional intensity during regulating compared to attending, increased the significance of results (see the Supplemental Table).

Repeating the analyses correcting for the number of previous episodes yielded a comparable pattern of results, although the significance of the difference in DLPFC



activity between MDDr and BDr (p=0.013) and the difference in rACC activity between MDDd and BD slightly decreased (see eTable 8). Analyses using a 2x2 factorial design and post-hoc t-tests I. Regulation across emotions (eTable 7)

There was a diagnosis x state interaction in the DLPFC (uncorrected p<0.001); however, this effect did not survive after the multiple comparison correction. The interaction was driven by the MDDr versus BDr comparison: we found greater DLPFC activity in BDr (p=0.006). There were no main effects of diagnosis or state for the regulate>attend contrast collapsed over all emotions. Compared to HC, there were no significant differences. II. Regulation of sad versus happy emotions (eTable 7)

There a diagnosis x state x emotion (happy versus sad) interaction in the ACC (uncorrected p<0.001; rostral ACC); however, this effect did not survive after the multiple comparison correction. The interaction was driven by the MDDd versus BDd comparison (p=0.004). In BDd we found a decrease in rACC activity during the regulation of happy emotions and an increase during the regulation of sad emotions, whereas in MDDd there was no difference in rACC activity between happy or sad regulation. There were no main effects of diagnosis or state for for happy versus sad emotion regulation. Compared to HC, BDd subjects demonstrated significant different (lower) rACC activity for the (regulate>attend) (happy>sad) contrast (p=0.002). Correlation analyses I. Associations between regulation and brain activity First, we correlated regulation success scores and regional brain activity, i.e. overall regulation success scores with DLPFC activity in the remitted groups, and happy and sad regulation success scores with rACC activity in the depressed groups.

In remitted subjects, neither for MDDr nor BDr there was a significant correlation between the regulation success score (across emotions) and DLPFC activity (all p≥0.26). For HC, there was no correlation between DLPFC activity and regulation success either (p=0.90).

In the depressed subjects, neither in MDDd or BDd there was a significant correlation between happy versus sad regulation success and rACC activity (all p≥0.17). Separate correlations for happy and sad regulation success were also performed, but no significant correlations with rACC activity were found (all p≥0.12). In HC, no correlations were found either (all p≥0.53). Possible explanations for these unexpected null findings are: i) insufficient power; ii) the relationship between behavioral outcome and neural activity is more complex, e.g. non-linear; iii) the relationship between DLPFC activity and emotion regulation success differs between the subject groups (i.e., there is a group x regulation success score interaction for DLPFC/rACC activity. Testing this did not yield any significant result; however this could again be due to insufficient power); iv) our MDDr and BDr subjects represent two groups at opposite ends of the emotion regulation spectrum; consequently, due to insufficient variation of regulation success or DLPFC activity we could not detect any correlations. II. Relationship between happy regulation success and sad intensity scores for happy pictures



Results of the post scan task revealed that MDDd subjects appraised happy pictures as less intense happy than BDd subjects, and furthermore happy pictures elicited happiness blended with sadness in MDDd, but not BDd. We hypothesized that these differences in happy appraisal may be an explanation for the difficulties in happy emotion regulation found in MDDd, but not BDd: happy pictures elicited emotions more mood congruent in MDDd than in BDd, and consequently may have been more difficult to distance from in MDDd subjects. To test this hypothesis, we correlated mean sad intensity scores for happy pictures with mean happy regulation success scores.

In MDDd, subjective experience of sad intensity for happy pictures (as measured in the post-scan task) correlated negatively with regulation success, although this correlation was only trendwise significant (=-0.28; p= 0.07; eFigure 3). This result indicates that the stronger sad emotions elicited by happy pictures were, the more difficult regulation of emotions elicited by these happy pictures was. III. Associations between rACC activity and sad intensity scores for happy pictures Because the difference in happy emotion regulation between MDDd and BDd was furthermore associated with rACC activity differences, we also correlated mean sad intensity scores for happy pictures with rACC activity.

Only within the total depressed patient group (MDDd and BDd together), there was a significant negative correlation between post-scan happy intensity and rACC activity (=-0.28, p=0.02, one-tailed) (eFigure 4A) as well as between in-scan happy intensity ratings and rACC activity (=0.35, p=0.003, one-tailed) (eFigure 4B), indicating that the happier the emotion that was elicited, the lower rACC activity while regulating this emotion.



eTable 1. Mean Ratings of the Pilot Study Conducted to Obtain Valence, Arousal, Complexity, and Emotional Intensity Values for Matching Purposes

Emotion type (picture)

Valence mean (SD)

Arousal mean (SD)

Complexity mean (SD)

Intensity of corresponding emotion

mean (SD) Sad 2.4 (0.4) 5.4 (0.6) 2.6 (0.4) 5.8 (0.7) Fear 3.1 (1.1) 5.4 (0.5) 2.0 (0.8) 5.0 (0.7) Happy 7.0 (0.6) 4.8 (0.3) 1.7 (0.2) 6.0 (0.2) Neutral 5.0 (0.2) 2.2 (0.3) 2.1 (0.7) Not applicable; ratings

of each discrete emotion were < 2.1

Abbreviations: SD: standard deviation



eTable 2. Handling of Subjects Demonstrating an Increase in Emotion Intensity During Regulation subject group Emotion type Attend – regulate

VAS score Corrected score

1003 HC Fear -5.8 0 2004 MDDd Sad -11.33 Missing 2009 MDDd Happy -2.67 0 2012 MDDd Sad/happy/fear -2.22/-5.44/-3.56 0/0/0 2025 MDDr Sad -12 Missing 2027 MDDd Fear -21.56 Missing 2030 MDDd Happy -11.78 Missing 2038 MDDd Happy -0.89 0 2039 MDDr Sad -0.22 0 3016 BDr Happy/ fear -0.22/-0.67 0/0 3020 BDr Happy -36.67 Missing 3025 BDr Fear -0.44 0 3031 BDd Sad -1.33 0 3034 BDr Sad/happy/fear -4/-2.67/-7.78 0/0/0 3037 BDr Sad/fear -7.33/-5.11 0/0 3041 BDr Sad/happy -3.11/-1.56 0/0 Because regulation is expected to decrease emotional intensity, the difference between attend intensity and regulate intensity score (calculated as the attend VAS score (scale: 0=zero emotional intensity; 100=maximal emotional intensity) – regulate VAS score (scale: 0=zero emotional intensity; 100=maximal emotional intensity)) should be a positive value.Negative scores were corrected to enable arcsine transformation needed to achieve a normal distribution:

‐ Attend – regulate VAS scores < -10 were considered flawed, indicating much greater emotional intensity while regulating than attending; therefore, they were omitted

‐ Attend – regulate VAS scores > - 10, but < 0, were considered comparable to 0, indicating no difference between emotional intensity during regulating versus attending

Abbreviations: BDd: bipolar disorder, depressed; BDr: BDremitted; HC: healthy control; MDDd: major depressive disorder, depressed; MDDr: MDD remitted.



eTable 3. Post-Scan Task: Ratings for Valence, Arousal, and Intensity of Corresponding Emotion Emotion type (picture)

HC (N=35) M

(SD)

MDDr (N=21) M

(SD)

BDr (N=26) M

(SD)

MDDd (N=21) M

(SD)

BDd (N=9)

M (SD)

MDDr vs BDr*

MDDd vs BDd* t p d t p d

A. Valence (scale 1-9: 1 = very negative, 5= neutral; 9=very positive) Sad 2.9 (0.9) 2.7 (0.7) 2.8 (0.2) 3.3 (1.3) 2.6

(0.5) 0.12 0.90 0.04 1.71 0.09 0.58

Happy 6.7 (0.9) 6.5 (0.9) 6.2 (1.0) 5.3 (0.8) 6.3 (0.9)

1.03 0.31 0.34 2.52 0.01 1.11

Emotion (happy, sad)*diagnosis

3.18 0.002

B. Arousal (scale 1-9: 1=feeling very calm; 9=feeling very aroused) Sad 4.4 (1.9) 4.4 (1.7) 4.5 (1.7) 3.8 (1.9) 4.7

(2.0) 0.30 0.77 0.06 1.18 0.24 0.46

Happy 4.2 (1.8) 4.5 (1.7) 4.3 (1.8) 3.3 (1.9) 3.5 (1.2)

0.52 0.61 0.12 0.53 0.60 0.13

Across emotions (including fear)

0.20 0.84 0.05

Emotion (happy, sad)*diagnosis

1.19 0.24

C. Intensity of corresponding emotion (scale 1-9: 1= emotion is not elicited by the picture; 9=emotion is strongly elicited by the picture) Sad 4.9 (2.0) 5.0 (1.6) 4.7 (1.9) 3.8 (2.0) 5.3

(1.9) 0.49 0.63 0.16 1.97 0.05 0.73

Happy 5.1 (2.0) 5.3 (1.6) 5.1 (1.7) 3.0 (1.7) 4.6 (1.9)

0.23 0.82 0.08 2.23 0.03 0.95

Across emotions 0.02 0.98 0.05



(including fear) Emotion (happy, sad)*diagnosis

0.28 0.80

Only comparisons relevant to explain happy versus sad emotion regulation differences between MDDd and BDd, and overall emotion regulation differences between MDDr and BDr are reported. There were no significant emotion x diagnosis x state interactions for valence, arousal or intensity of corresponding emotion. Only for valence, there was a significant (happy versus sad) x diagnosis interaction within the depressed subgroup, explained by more extreme ratings in BDd than in MDDd subjects. Results were significant at p<0.007 (Bonferroni-corrected for 7 comparisons), indicated in bold. Abbreviations: BDd: bipolar disorder, depressed state; BDr: BD, remitted state; HC: healthy control; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state; d=effect size expressed as Cohen’s d. * results of the post hoc tests



eTable 4. Post-Scan Task: Ratings for Discrete Emotion Types HC (N=35) MDDr (N=21) BDr (N=22) MDDd (N=21) BDd (N=9) Sad pict

M (95% CI)

Happy pict M

(95% CI)

Sad pict M (95%

CI)

Happy pict M

(95% CI)

Sad pict M (95%

CI)

Happy pict M

(95% CI)

Sad pict M (95%

CI)

Happy pict M

(95% CI)

Sad pict M (95%

CI)

Happy pict M (95%

CI) Emotion (rating)

sadness 4.9 (4.2-5.6)

1.1 (1.0-1.3)

5.0 (4.2-5.7)

1.3 (1.0-1.7)

4.7 (3.9-5.5)

1.5 (1.2-1.8)

3.8 (2.9-4.7)

1.8 (1.3-2.3)

5.3 (3.8-6.7)

1.5 (1.0-2.1)

happiness 1.0 (1.0-1.1)

5.1 (4.4-5.8)

1.0 (1.0-1.5)

5.3 (4.5-6.0)

1.2 (1.0-1.4)

5.1 (4.4-5.9)

1.1 (1.0-1.1)

3.0 (2.2-3.7)

1.1 (1.0-1.2)

4.6 (3.2-6.0)

Fear 2.8 (2.2-3.3)

1.3 (1.1-1.4)

2.5 (1.8-3.2)

1.4 (1.0-1.8)

3.0 (2.-3.8)

1.4 (1.2-1.7)

2.5 (1.7-3.2)

1.4 (1.1-1.8)

2.9 (1.4-4.4)

1.2 (1.0-1.5)

digust 3.0 (2.4-3.6)

1.1 (1.0-1.2)

3.1 (2.4-3.8)

1.2 (1.0-1.5)

2.8 (2.1-3.5)

1.1 (1.0-1.3)

2.5 (1.6-3.4)

1.3 (1.0-1.5)

3.2 (1.6-4.9)

1.1 (1.0-1.1)

anger 3.2 (2.5-3.8)

1.0 (1.0-1.1)

3.0 (2.3-3.7)

1.2 (1.0-1.4)

2.6 (1.9-3.3)

1.1 (1.0-1.3)

2.5 (1.8-3.2)

1.4 (1.1-1.7)

3.3 (1.8-4.8)

1.0 (1.0-1.1)

conclusion Discretely sad

Discretely happy

Discretely sad

Discretely happy

Discretely sad

Discretely happy

Sad, blended with fear, disgust

and anger

Blended happy

and sad

Sad, blended with fear, disgust

and anger

Discretely happy

For each picture, the intensity of each discrete emotion was rated on a scale from 1 - 9: 1= emotion is not elicited by the picture; 9=emotion is strongly elicited by the picture. The type of elicited emotion was categorized according to the criteria by Mikels et al.5: 1. an elicited emotion was considered discrete if the mean for that emotion was higher than the mean of all other emotions and there was no overlap with other confidence intervals (CI); 2. an elicited emotion was considered blended if the means of two or



three emotions were higher than the rest and CIs overlap each other (indicated in bold); 3. an elicited emotion was considered undifferentiated if there was overlap of all CIs. For both happy and sad pictures, there were no differences between MDDr and BDr, nor for sad pictures between MDDd and BDd regarding the type of emotion elicited. However, happy pictures elicited happy emotions that were blended with sadness in MDDd subjects, in contrast to in BDd subjects, who rated happy pictures as discretely happy. Abbreviations: BDd: bipolar disorder, depressed state; BDr: BD, remitted state; HC: healthy control; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state; pict: picture.



eTable 5. Regulation Success Scores During Scanning (Scale: 0-100) Emotion type (picture)

HC (N=36) mean (SD)

MDDr (N=21) mean (SD)

BDr (N=26) mean (SD)

MDDd (N=21) mean (SD)

BDd (N=9) mean (SD)

MDDr vs BDr*

MDDd vs BDd*

t p d t p d

Sad 71 (20) 61 (31) 38 (31) 53 (42) 38 (35) 2.39 0.02 0.73 1.62 0.11 0.40 Fear 75 (23) 70 (23) 44 (43) 50 (54) 58 (31) 2.35 0.02 0.73 0.02 0.99 0.15 Happy 78 (26) 77 (24) 50 (51) 44 (57) 85 (19) 2.61 0.01 0.66 2.52 0.01 0.82 Across emotions 75 (23) 69 (27) 44 (42) 49 (51) 60 (34) 3.39 0.001 0.70 Emotion (happy,sad)*diagnosis

0.19 0.85 3.86 0.000

Emotion (sad, fear)*diagnosis

0.06 0.95 1.53 0.13

Emotion (happy, fear)*diagnosis

0.26 0.80 2.33 0.02

Happy versus sad t=0.70; p=0.49

t=4.19; p=0.000

Within the remitted group, BDr subject demonstrated significant less successful emotion regulation across emotions. Within the depressed group, there was a significant emotion (happy versus sad) x diagnosis interaction: MDDd subjects showed no differences between the regulation of sad and happy emotions, whereas BDd subjects showed a significant stronger regulation of happy compared to sad emotions. Results were significant at p<0.003 (Bonferroni-corrected for 16 comparisons), indicated in bold. Abbreviations: HC: healthy controls; BDd: bipolar disorder, depressed state; BDr: BD, remitted state; HC: healthy control; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state; SD: standard deviation; d=effect size expressed as Cohen’s d. * results of the post hoc tests



eTable 6. Main Task Effects Across Subjects and Emotion Type Area Side Cluster

size Peak voxel (MNI)

x (mm)

y (mm)

z (mm)

t

A. Attend>neutral Amygdala R 2 26 0 -20 5.08 Hippocampus, extending to L 117 -20 -10 -14 7.14 amygdala Inferior frontal gyrus (BA45/47)

L 346 -50 30 6 6.80

Inferior frontal gyrus (BA45/47), extending to

R 881 54 26 2 7.03

insula Middle frontal gyrus, extending to

L 110 -40 12 44 5.67

superior frontal gyrus and precentral gyrus (BA8)

Middle frontal gyrus, extending to precentral gyrus (BA8)

R 363 38 4 42 6.27

Superior frontal gyrus (BA10)

L 22 -28 54 0 5.94

Posterior cingulate cortex (BA31/23), extending to

L/R 1268 2 -48 26 8.32

precuneus (BA7/31/32) Medial superior frontal gyrus (BA9/10)

L/R 725 4 54 38 7.46

Supplemental motor area extending to

L/R 167 8 26 58 6.42

superior frontal gyrus Inferior occipital gyrus (BA 18) extending to

L 2371 -48 -74 -4 14.25

middle occipital gyrus, inferior and middle temporal gyrus (BA 37/39) and fusiform gyrus

R 2507 46 -72 -10 14.16 Superior temporal gyrus R 5 28 12 -24 5.11 Supramarginal gyrus (BA40)

L 183 -58 -40 34 6.41

Inferior parietal lobule L 62 -30 -50 46 5.85 Angular gyrus (BA40) extending to

R 153 56 -52 30 5.47

middle temporal gyrus (BA13) and supramarginal gyrus (BA40)



eTable 6. Main Task Effects Across Subjects and Emotion Type (Continued) B. Regulate > attend Supplemental motor area (BA6) L/R 376 6 0 64 8.77 Mid-cingulate cortex (BA32) R 2 6 8 42 5.18 Supramarginal gyrus (BA40) R 8 58 -44 36 5.22 Postcentral gyrus L 23 -46 -10 50 5.93 Precentral gyrus (BA9) R 71 50 -4 50 6.14 Lingual gyrus R 6 14 -86 -6 5.26 Medial occipital gyrus L 48 -16 -94 10 6.97 Regional brain activity associated with emotional reactivity (A) and emotion regulation (B) across subjects and emotion type (p<0.05, whole brain FWE corrected). Abbreviations: BA: Brodmann area; L: left; MNI: Montreal Neurological Institute; R: right. *small volume correction using a bilateral anatomical mask



eTable 7. Activity Differences for Regulate > Attend Between MDD and BD Comparison Area Side Cluster size Peak voxel (MNI) x (mm) y (mm) z

(mm) F/t

A. Across emotions

Full factorial (diagnosis (MDD, BD) x state (remitted, depressed)

Main effects of diagnosis none

Main effects of state none

Diagnosis x state interaction

Inferior frontal gyrus, pars triangularis/opercularis (BA46)

R 28 44 22 26 15.7*

BDr>MDDr Inferior frontal gyrus, pars triangularis/opercularis*** (BA46)

R 31 40 16 34 4.8**

Comparison with HC

-

B. Happy>sad, MDDd versus BDd Full factorial (diagnosis (MDD, BD) x state (remitted, depressed)

Main effects of diagnosis none

Main effects of state none

Diagnosis x state interaction

Rostral anterior cingulate cortex (BA32)

R 10 8 36 8 14.9 *

MDDd>BDd Rostral anterior cingulate cortex (BA32)

R 48 10 36 6 4.8**

Comparison with HC



HC>BDd Rostral anterior cingulate cortex (BA32)

R 60 8 40 6 5.25**

*Presented are the uncorrected results (diagnosis x state across emotions: uncorrected p<0.001; diagnosis x state x emotion (happy versus sad): rACC: uncorrected p<0.001): interaction effects were not significant after multiple comparison correction. **Results are significant at p<0.009, FWE corrected for the bilateral anatomical ROI for post-hoc comparisons,. ***also significant at p<0.05 FWE corrected across the whole brain (k=2; t=5.41) Abbreviations: BA: Brodmann area; BDd: bipolar disorder, depressed state; BDr: BD, remitted state; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state.



eTable 8. Activity Differences Between MDD and BD With Correction for the Number of Previous Episodes Comparison Area Side Cluster size Peak voxel (MNI) x (mm) y (mm) z (mm) t A. Across emotions, MDDr versus BDr

MDDr>BDr none

BDr>MDDr Inferior frontal gyrus, pars triangularis/opercularis (BA46)

R 30 40 16 34 4.86

B. Happy>sad, MDDd versus BDd

MDDd>BDd Rostral anterior cingulate cortex (BA32) R 51 10 36 6 4.82

BDd>MDDd none Results are significant at p<p0.009, FWE corrected for the bilateral anatomical ROI. Correction for previous major depressive episodes slightly decreased the significance of the DLPFC activity difference between MDDr and BDr (compare to table 2). Abbreviations: BA: Brodmann area; BDd: bipolar disorder, depressed state; BDr: BD, remitted state; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state.



Supplemental Table. Activity Differences Between MDD and BD in a Sample Without Subjects Showing a Paradoxical Increase in Emotional Intensity During Regulation Comparison Area Side Cluster

size Peak voxel (MNI)

x (mm)

y (mm) z (mm)

t

A. Across emotions, MDDr versus BDr

MDDr>BDr none

BDr>MDDr Inferior frontal gyrus, pars triangularis/opercularis* (BA46)

R 31 40 16 34 4.65

B. Happy>sad, MDDd versus BDd

MDDd>BDd Rostral anterior cingulate cortex (BA32)* R 71 10 36 6 5.67

BDd>MDDd none Results are significant at p<p0.009, FWE corrected for the bilateral anatomical ROI. Excluding subjects with a large paradoxal increase in emotional intensity during regulation (i.e. 3 MDDd, 1 BDr and 1 BDd subject) increased the significance of results (compare to table 2). Abbreviations: BA: Brodmann area; BDd: bipolar disorder, depressed state; BDr: BD, remitted state; MDDd: major depressive disorder, depressed state; MDDr: MDD, remitted state. * also significant at p<0.05 FWE corrected across the whole brain (DLPFC: k=1; t= 5.00; ACC: k=6; t=5.67; )



eFigure 1. Design of the Emotion Regulation Task

Each block started with the instruction presented in the middle of the screen (4s), followed by 3 successive pictures of the same emotional category (sad/happy/fear/neutral) (10s each). After each picture, subjects indicated the emotional intensity resulting from attending or regulating on a Visual Analogue Scale (VAS). After each block subjects also rated their performance. Blocks were separated by a fixation cross (4s).



eFigure 2. Main Task Effects Across Subjects and Emotion Type

Regional brain activity associated with (A) emotional reactivity (attend) and(B) emotion regulation across subjects and emotion type (p<0.05, whole brain FWE corrected).



eFigure 3. Relationship Between Regulation Success for Happy Pictures and Sad Intensity Ratings for Happy Pictures in MDDd

There was a trendwise negative correlation (=-0.28, p=0.07, one-tailed) between behavioral in-scan regulation success scores and post-scan intensity ratings for sad emotions elicited by happy pictures, indicating that the stronger sad emotions were elicited by happy pictures, the more difficult emotion regulation for MDDd subjects was.



eFigure 4. Relationship Between rACC Contrast Estimates and Mean Happy Intensity Ratings Across Depressed Patients (MDDd + BDd)

(A) There was a significant negative correlation between post-scan happy intensity and the contrast estimates for the regulate happy>baseline contrast at 10/36/6 (‘rACC activity’) (=-0.28, p=0.02, one-tailed) as well as (B) between in-scan happy intensity ratings and rACC activity (=-0.35, p=0.003, one-tailed), indicating that the happier the emotion elicited, the lower rACC activity



while regulating this emotion. An explanation, although speculative, may be that regulation of happy emotions resolves an emotional conflict (happiness is mood-incongruent for depressed subjects), thus leading to deactivation of the rACC.



eReferences

1. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured Clinical Interview for DSM-IV Axis I Disorders - Patient Edition (SCID I/P), Version 2.0. New York: Biometrics Research Department, New York State Psychiatric Institute; 1996.

2. First MB, Gibbon M, Spitzer RL, Williams JBW, Benjamin LS. Structured Clinical Interview for DSM IV Axis II Personality Disorders (SCID-II). New York: Biometrics Research Department, New York State Psychiatric Institute; 1997.

3. Young RC, Biggs JT, Ziegler VE, Meyer DA. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978;133:429-435.

4. Lang PJ, Bradley MM, Cuthbert BN. International affective manual. Technical Report A-8.; 2008:University of Florida, Gainesville, FL.

5. Mikels JA, Fredrickson BL, Larkin GR, Lindberg CM, Maglio SJ, Reuter-Lorenz PA. Emotional category data on images from the International Affective Picture System. Behav Res Methods. 2005;37(4):626-630.

6. Bradley MM, Lang PJ. Measuring emotion: the self-assessment manikin and the semantic differential. J Behav Ther Exp Psychiatry. 1994;25:49-59.


Supplementary Online Content€¦ · MRI-scanning contra-indications (a history of head trauma or...

Documents

Transcript of Supplementary Online Content€¦ · MRI-scanning contra-indications (a history of head trauma or...