1
Using The Triple Network Model to Find Neurobiological Differences
between Remitted Patients with Major Depressive Disorder and
Healthy Controls
An Internship Report
Date: August 2015 Author: Suzanne Martens Student number: 10194045 Supervisor: Caroline Figueroa Co-assessor: Steven Scholte
2 Using The Triple Network Model to Find Neurobiological Differences between Remitted Patients
with Major Depressive Disorder and Healthy Controls
Abstract Introduction
Depression is a prevalent mental disorder with a high probability of being recurrent. To reduce the impact of depression it is necessary to identify vulnerability factors in remitted MDD patients that increase risk of relapse. The Triple Network Model (TNM) states that dysfunctional cooperation of three core resting state networks (RSNs) in the brain causes depression. In this study, we examined whether TNM related aberrations found in MDD patients persist in remitted-MDD patients.
Methods
We examined TNM functioning in both neutral and sad mood, because cognitive vulnerability is thought to become more apparent in remission during sad mood or stress. By performing Independent Component Analysis on resting state data, intra- and internetwork functional connectivity of these three RSNs was investigated in 62 remitted MDD patients and 41 controls.
Results
No TNM related differences between remitted patients and controls were found. Instead, the hippocampus was found to be more connected with posterior areas of one of the RSNs (the Default Mode Network) during sad mood in controls, but not in patients.
Conclusions
The findings show that TNM related dysfunction is not present in remitted MDD patients. However, the altered functioning of the hippocampus might be related to the theory of Overgeneral
Autobiographical Memory (OAM). Further investigation of this alternative explanation is needed as well as investigation of the TNM using different methods to provide conclusive insights on the found effect.
§1. Introduction
Depression is a prevalent mental disorder associated with great personal suffering and adverse effects on society (Johnson, Weissman & Klerman, 1992). Between 23.5% and 25.6% of the population will suffer from at least one depressive episode (Kessler et. al, 2003; Richards, 2011). Because about 40% of these patients experience multiple episodes (Steinert, Hofmann, Kruse & Leichsenring, 2014; Kessler et. Al, 2003; Richards, 2011), about 10% of the whole population struggles with depression and related problems for a large part of their lives. To add to this, studies show that the chance of full recovery is negatively related to the amount of previous depressive episodes (Hardeveld, Spijker, De Graaf, Nolen & Beekman, 2010; Kessing, Andersen & Andersen, 2000; Kaymaz, Os, Loonen & Nolen, 2008). This realization invites a new way of thinking about reducing the impact of depression on society. Instead of a focus on prevention and treatment of depression in general, a focus on preventing relapses is necessary. Therefore, methods of identifying remitted patients at high risk of relapse need to be developed. This study therefore aspires to find differences between remitted patients and healthy controls that can be used for the development of these methods.
3 §1. 1 Psychological predictors of relapse
Previous research on predictors of relapse is mostly based on self-report questionnaires. Commonly found predictors in this line of research include an early onset of depression (Klein et. al., 1999; Giles, Jarrett, Biggs, Guzick & Rush, 1989), the amount of previous relapses (Hardeveld, Spijker, De Graaf, Nolen & Beekman, 2010; Kaymaz, Os, Loonen & Nolen, 2008; Kessing, Andersen &
Andersen, 2000), the presence of residual symptoms (Judd et. al., 2014; Hardeveld, Spijker, De Graaf, Nolen & Beekman, 2010; Steunenberg, Beekman, Deeg & Kerkhof, 2010; Dombrovski et. al., 2007) and comorbidity with (or earlier episodes of) other mental illnesses (Hart, Craighead & Craighead, 2008; Andreescu et. al., 2007; Giles, Jarrett, Biggs, Guzick & Rush, 1989). With the exception of the presence of residual symptoms, these predictors’ practical usefulness is to be debated because of their invariance to treatment (though their theoretical importance is clear).
Variables such as the presence of residual symptoms however do have the potential to be useful in preventing relapse because it’s possible to assess to what degree remitted patients exhibit them after treatment. For example, emotional dysregulation (Judd et. al., 2014; Hardeveld, Spijker, De Graaf, Nolen & Beekman, 2010; Steunenberg, Beekman, Deeg & Kerkhof, 2010; Dombrovski et. al., 2007) excessive rumination (Michalak, Hölz & Teismann, 2011) and negative cognitive bias (Bouhuys, Geerts & Gordijn ,1999) all predict relapse risk and can be measured at different time points. Assessing these kinds of variables might be useful for selecting individuals that need more treatment before discontinuation. Moreover, several studies show that receiving treatment while not currently depressed can prevent future relapse (Piet & Hougaard, 2011; Kuyken et. al., 2008; Bockting et. al., 2005). This suggests that the development of a follow up system where variables are assessed at different time points to monitor relapse risk over time could be helpful in preventing relapse.
However, using questionnaire data about the presence of residual symptoms for selecting vulnerable patients might not be the best choice because it’s predictive power is low. While a hazard ratio of 2.35 has been found for first time patients (meaning that people with high amounts of residual symptoms are 2.35 times as likely to have a relapse than people with low amounts of residual symptoms) (Judd et. al, 2014), a hazard ratio of only 1.15 has been found in a sample that contained both first time and recurrent patients (Dombrovski et. al., 2007). Furthermore, a problem with using self-report questionnaires is that patients with depressive disorders tend to underreport their symptoms (especially men: Sigmon et. al, 2005; Hunt, Auriemma & Cashaw, 2003). Gathering more versatile information on remission is therefore necessary, preferably in a way that bypasses the problem of underreporting symptoms.
§1. 2 Resting State Networks and Depression
One possible source of information to study is the brain activity of remitted patients. In recent years the field of functional Magnetic Resonance Imaging (fMRI) research has developed two major paradigms to investigate brain activity. Within the task-based paradigm researchers
investigate changes in brain activity in response to a certain stimulus as compared to some form of baseline activity. This approach has proven to be useful in identifying brain regions that are involved in certain functions. For example, the use of working memory tasks has consistently shown that the prefrontal cortex is crucial for retaining information for a short amount of time (Owen, McMillan, Laird & Bullmore, 2005). A disadvantage of this paradigm however is that it’s potential for diagnosing certain mental disorders is limited, because it’s unlikely that disorders associated with a large
4 spectrum of symptoms are related to just one or two dysfunctional brain regions. As for depression, it has indeed been observed that research trying to identify dysfunctional brain regions produces inconsistent results (Campbell, Marriott, Nahmias & MacQueen, 2014; Hamilton, Siemer & Gotlib, 2008; Videbech & Ravnkilde, 2004).
The resting state network paradigm might be of more use for investigating mental disorders. In this paradigm participants are scanned during rest instead of during task performance. Instead of activity of specific brain regions correlations between activity levels of different regions over time are investigated. This can be done by selecting regions of interest and correlating the activity of these regions, but it’s also possible to use all voxels in the brain (Damoiseaux et. al., 2006; Menon, 2011). This way brain regions that consistently change in activity levels at the same time can be discovered. These brain regions form networks called Resting State Networks (RSNs). Three RSNs have been identified that seem to be particularly useful when investigating mental disorders: the Default Mode Network (DMN), the Salience Network (SN) and the Central Executive Network (CEN) (Menon, 2011).
The default mode network.
It’s thought that the DMN is involved in self-referential processing because of its inclusion of the medial temporal lobes (MTL), which are associated with episodic and autobiographical memory, and the anterior medial cortex (MC), which is associated with social cognition, emotion regulation and self-related thought (Menon, 2011). This idea fits well with studies showing that the DMN is active during rest but less active during tasks (Whitfield-Gabrieli & Ford, 2012; Sridharan, Levitin & Menon, 2008). Also included in the DMN are the posterior MC and the precuneus (PCN) (Anticevic, Cole, Murray, Corlett, Wang, & Krystal, 2012; Menon, 2011). Abnormal functioning of the DMN has been reported in a large number of mental illnesses, including depression. Specifically, depression is associated with hyperconnectivity of the DMN and persistent activity during task performance. These abnormalities are associated with at least one symptom of depression, namely excessive rumination (Marchetti, Koster, Sonuga-Barke, & De Raedt, 2012; Whitfield-Gabrieli & Ford, 2012; Greicius, 2008). Interestingly, Marchetti, Koster, Sonuga-Barke and De Raedt (2012) describe several studies that show that these abnormalities persist in remitted patients.
The central executive network.
The CEN is related to the DMN in that it can be thought of as the DMN’s opposite. The CEN is activated during task performance and less activated during rest (Sridharan, Levitin & Menon, 2008) and is involved in working memory and attentional control (Patel, Spreng, Shin & Girard, 2012; Menon, 2011). It’s therefore not surprising that the CEN consists of two areas that are related to attention, the dorsolateral prefrontal cortex (dlPFC) and the lateral posterior parietal cortex (PPC) (Chen et. al., 2013; Menon, 2011). Though the CEN has not been studied much from a network perspective in relation to depression, it is known that depressed patients often show reduced performance on difficult working memory tasks (Menon, 2011; Austin, Mitchell & Goodwin, 2001). Abnormal functioning of the CEN network might therefore also be of interest for research on depression.
5 The salience network.
The SN seems to function as a system for detecting salient stimuli and events of either emotional, cognitive or homeostatic value (Yuen et. al., 2014; Menon, 2011). It consists of the anterior cingulate cortex (ACC), substantia nigra (SNi), ventral tegmental area(VTA) and the anterior insula (AI) (Yuen et. al., 2014; Bonnelle et. al., 2012; Menon, 2011). These areas are associated with reward processing (SNi & VTA), error and conflict detection (ACC) and interoceptive and emotional processing (AI). Hypoconnectivity within the SN has been associated with apathy in depressed patients (Yeun et. al., 2014), showing that this network is also relevant for research on depression.
Combining all the data: The triple network model.
The three described networks and their relationships with depression could be treated as independent from each other. It might be more helpful however to combine all information on the networks into one framework. Menon (2011) proposed such a framework: the Triple Network Model (TNM). The TNM suggests that abnormal connectivity between or within the DMN, CEN and SN can explain many different types of symptoms of depression. At its core, the TNM models depression or other mental illnesses as a failure to appropriately engage and disengage the three networks in response to stimuli. The RSN of primary interest in this model isthe SN because of its inclusion of the (right) AI. The AI has been shown to initiate control signals in response to salient events that engage and disengage the DMN and CEN (Menon & Uddin, 2010). Hence, dysfunction of the SN should lead to (dis)engagement problems in the other networks. (Dis)engagement of networks can be disrupted in at least two ways. First, it’s possible that all networks function normally but that the SN fails to send appropriate control signals to the DMN and CEN because the connections between the networks are not optimal. Second, it’s possible that the connectivity between the three systems is normal but that abnormal functioning of one network limits the functioning of the others.
The TMN has already been investigated in both healthy and depressed samples. Two studies have found support for the existence of the TNM in healthy samples. Sridharan, Levitin and Menon (2008) used Granger Causality Analysis to confirm that the AI causally influences the DMN and CEN when salient stimuli are detected (note however that the validity of Granger Causality Analysis is sometimes debated: Poldrack, Mumford & Nichols, 2011). These findings were replicated by Goulden et. al. (2014). Support for the dysfunctions proposed by the TNM has also been found in depressed patients. Using Hurst components (measures of autocorrelation), Wei et. al. (2015) found that depressed patients showed more irregular DMN connectivity but more regular SN connectivity compared to healthy controls. Other studies have shown that depressed patients show decreased connectivity within the SN (Manoliu et. al., 2013), increased SN-DMN connectivity (Manoliu et. al., 2013), decreased DMN-CEN connectivity (Zheng et. al., 2015; Manoliu et. al., 2013) and an increased amount of overall connections with the insula (Zheng et. al., 2015) compared to healthy controls. However, few studies have investigated brain connectivity in remitted MDD. Furthermore, the TNM has not yet been examined in remitted patients.
When conducting research with remitted patients, cognitive reactivity needs to be taken into account. Cognitive reactivity, the activation of dysfunctional information processing by sad mood or stress, is related to remission in two ways. First, stronger cognitive reactivity has been identified as a risk-factor for relapse by itself (Segal et al. 2006). Second, it suggests that certain differences
between remitted patients and controls might stay hidden when studied solely in neutral mood. Possibly, all RSNs might function well under normal circumstances, but start functioning abnormally
6 in remitted patients but not in controls when in a sad mood or during stress. Several studies support this idea. First, research has shown that depressed people indeed react differently to negative emotions than controls. For example, depressed patients pay more attention to and react more strongly to negative information than healthy controls (Bouhuys, Geerts & Gordijn ,1999; Atchley, Stringer, Mathias, Ilardi & Minatrea, 2007). Furthermore, some fMRI and PET studies using mood induction techniques found that depressed people show different activation patterns than healthy controls when in a sad mood (Mayberg et. al., 2014; Keedwell, Andrew, Williams, Brammer & Phillips, 2005; Ramel, Goldin, Eyler, Brown, Gotlib & McQuaid, 2007). Second, research has shown that when patients are in remission, some vulnerabilities only show during sad mood or stressful situations (Ramel, Goldin, Eyler, Brown, Gotlib & McQuaid, 2007; Liotti, Mayberg, McGinnis, Brannan & Jerabek, 2014; Segal, Kennedy, Gemar, Hood, Pedersen & Buis, 2006; Farb, Anderson, Bloch & Segal, 2011). Finally, research with remitted-MDD patients using mood induction procedures has found differences between remitted-MDD patients and controls in sad mood that were not present in neutral mood. For example, Zamoscik, Huffziger, Ebner-Priemer, Kuehner and Kirsch (2014) investigated whether remitted patients and controls showed differences in connectivity between the para hippocampal gyri and the posterior cingulate cortex (both part of the DMN). They found no differences in neutral mood but observed more connectivity between these two regions in remitted patients compared to controls during sad mood. Furthermore, in the remitted patient group the connectivity between the two regions in sad mood was only positively correlated with rumination scores, adding a possible behavioral manifestation of their connectivity results. If mood was not included in this study, these differences would not have been found. This research therefore shows that it’s important to consider changes in connectivity according to mood as well as at baseline, especially when working with remitted patients.
The present study therefore investigated whether abnormal functional connectivity related to the TNM is present in remitted-MDD patients, both in neutral and in sad mood. Two resting state scans were made, one after neutral and the other after sad mood induction by means of
autobiographical recall. The DMN, SN and CEN were then compared on intra- en internetwork connectivity. Based on the TNM and the findings reported earlier, the following hypotheses investigated:
1) Compared to healthy controls, remitted patients will show differences in intranetwork
connectivity. Specifically, they will show increased DMN and decreased SN and CEN connectivity. 2) Compared to healthy controls, remitted patients will show differences in internetwork
connectivity. Specifically, they will show increased SN-DMN and decreased SN-CEN and DMN-CEN connectivity.
3) Compared to healthy controls, negative mood will have a different impact on the connectivity of the DMN, CEN and SEN in remitted patients. Functional (intra- or internetwork) connectivity will change in a different way in patients and controls in response to sad mood.
7 §2. Methods
§2. 1 Participants.
62 remitted patients and 42 healthy controls were included in the study. The groups were matched on age, gender and education level. To be included patients had to be in a stable, medication free, remitted state at the moment of intake. This was characterized by a score below seven on the Hamilton Depression Rating Scale (HDRS-17, see Fleck, Poirier-Littre, Guelfi, Bourdel & Loo, 1995) for the last 8 weeks and a negative score on the Structural Clinical Interview for DSM Disorders (SCID, see First, 1995). Furthermore, patients had to have experienced at least two depressive episodes in their lifetime to ensure vulnerability to relapse (median = 3, range = 2:60) in this group. To be included controls needed to have no personal or first degree familial history of mental illness and no depressive symptoms at intake. For both groups several other exclusion criteria were also maintained: alcohol/drug dependency, psychotic or bipolar disorder, predominant anxiety disorder, severe personality disorder, standard MRI exclusion criteria, experienced electroconvulsive therapy within two months before scanning, history of severe head trauma, neurological disease, severe general physical illness and no Dutch/English proficiency.
§2.2 Materials. HDRS-17
The HDRS-17 questionnaire consist of 17 questions inquiring about depressive symptoms that were experienced in a certain week. For every question several answers are possible, all leading to follow-up question and/or resulting in a different score. For example, “Have you experienced tiredness this week?” can be answered with “no” (no points), “unclear” (one point) and “yes”. If the answer is “yes”, the interviewer tries to find out whether the patient experienced this tiredness less (one point) or more (two points) than half of the time. A score between zero and seven is considered normal, while a score of 20 or higher indicates depression.
SCID
The SCID is a structured interview that was used to determine whether participants were experiencing or had experienced a depressive episode in the past, how severe the episode was and whether circumstances before the start of the episode might disqualify it as a depressive episode. According to the SCID a depressive episode must last at least two weeks, in which the patient has felt depressed almost constantly or has not been able to experience joy. If a participant was physically ill, was using certain medication, drugs or alcohol or lost a loved one before the start of the episode, it is not counted as a depressive episode. Severity of the episode is determined by the amount of
symptoms and their severity.
LEIDS-R
The Leiden Index of Depression Sensitivity – Revised (LEIDS-R, Van der Does & Williams, 2003) questionnaire was used to assess cognitive reactivity. Participants are asked to imagine themselves on a day that they feel somewhat sad and to judge 34 statements about how they act in such a mood using a five point scale (0 = not at all, 4 = very strongly). For example, “When I feel down, I lose my temper more easily” is an item of the LEIDS-R. A higher score on the LEIDS-R indicates stronger cognitive reactivity, with a maximum score of 136.
8 Ruminative Response Scale-NL
The Ruminative Response Scale (RRS, Treynor, Gonzalez & Nolen-Hoeksema, 2003) questionnaire was used to assess rumination. Participants are asked to judge 22 statements that inquire about their general behavior using a four point scale (1 = almost never, 4 = almost always). For example, “think about how angry you are with yourself” is an item of the RRS. Higher scores on the RRS indicate that they ruminate more, with a maximum of 88.
Autobiographical Essays
For the mood induction procedure, participants were asked to write two small essays, one about a personal neutral experience (e.g. a daily routine) and one about a personal negative experience (e.g. the death of a loved one). Recordings of these essays were made. The recordings were used later on to induce a neutral or sad mood.
§2.3 Procedure.
Before getting in the scanner, an interview was conducted in which participants completed the HDRS and SCID questionnaires. After determining whether participants complied to the inclusion criteria, all other questionnaires were also completed prior to scanning. The procedure in the
scanner started with acquiring a T1 structural scan. Afterwards, participants listened to a recording of their neutral autobiographical essay to induce a neutral mood. Next, participants were told to relax and let their mind wander as they pleased without falling asleep for the duration of eight minutes. During this time, a functional resting state scan was made. Before playing the recording and after the resting state scan, participants’ mood was checked by asking them to rate their mood on a 1 to 10 scale (10 being best). This procedure was repeated (minus making the structural scan) using the sad autobiographical essay instead of the neutral one to induce a sad mood.
§2. 4 Image acquisition.
Structural magnetic resonance imaging was performed using a T1-weighted TFE sequence, obtaining data from 220 axial slices (TR =8.2s, TE = 3.6ms, flip angle = 8°, FOV = 240x188mm, matrix = 240x240mm, slice thickness = 1mm, interslice skip = 0mm, voxel size = 1x1x1, acquisition orientation = TRA, parallel imaging parameter method = SENSE in two directions (AP) 2 & 2.5, whole brain). Functional magnetic resonance imaging was performed on a Philips 3T Achieva system, using a T2*- weighted EPI sequence. For two sessions, functional images (210 volumes) were obtained from 37 axial slices (TR = 2s, TE = 27.63ms, flip angle = 76.1°, FOV = 240x240mm, matrix = 80x80mm, slice thickness = 6.5mm, interslice skip = 0.3mm, voxel size = 3x3x3, acquisition orientation = TRA, acquisition order = interleaved, parallel imaging parameter method = SENSE (AP) 2, single shot, whole brain).
§2. 5 Preprocessing
The data was preprocessed and analyzed using the Statistical Parameter Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm/) and related toolboxes. First, structural and functional images were manually reoriented parallel to the AC-PC plane. Next, functional images were realigned for each subject using rigid body transformations and the mean image as reference. Then, the realigned functional images were coregistered to the structural images for each subject using a normalized mutual information algorithm. Normalization of both the functional and structural scans was
9 performed using the customized group template and individual deformation flow fields to normalize the images to MNI space. Finally, the functional data was smoothed with a smoothing kernel of 10. Because the data was meant for resting state independent component analysis no slice time
correction, segmentation or filtering was performed. Also, because the regions affected by air-tissue interfaces were not of interest, no distortion correction was applied.
§2. 5 Independent Component Analysis
Using the GIFT toolbox (http://mialab.mrn.org/software/gift/index.html), an independent component analysis (ICA) was performed on the preprocessed functional data (maximum amount of components = 25, PCA type = standard, group PCA type = subject specific, ICA algorithm = infomax, back reconstruction type = spatial temporal regression, pre-processing: remove mean per timepoint, scaled to original data). The DMN, CEN and SN were found among the 25 resulting components by having two independent judges visually compare the components to results from Menon (2011).
§2. 6 Statistical Analysis
To test whether remitted patients and controls showed any differences in connectivity, a full factorial mixed model ANOVA with factors group (patients and controls) and mood (neutral and sad) was performed in SPM for all chosen components individually. This method was chosen because both intra and internetwork connectivity differences can be detected this way, because the connectivity of all voxels with the components is assessed, not just of the voxels within the component being investigated. Hence, only one test per component is needed to discover
differences in both intra and internetwork connectivity between patients and controls (main effect of group), between neutral mood and sad mood (main effect of mood) and between changes in
connectivity among groups as a result of mood induction (interaction effect). Because of the
exploratory nature of the research project, no correction for the amount of independent analyses (5) has been used. However, every analysis has used the relatively conservative Family Wise Error rate (calculated using random field theory) to correct for the amount of voxels that were tested.
§3. Results §3. 1 Participants
In total ten patients and three controls had to be excluded from analysis. Reasons for exclusion included abnormal brain anatomy (three patients, two controls), excessive motion during scanning (five patients), failed preprocessing (two patients) and belated discovery of noncompliance with our inclusion criteria soon after scanning (one control). All analyses were conducted with the data of the remaining 52 patients and 39 controls. For an overview of all drop-out, see Figure 1. The groups did not differ on demographical characteristics. For all relevant demographics, see Table 1. For medians and quartiles of non-normally distributed variables, see Appendix A.
§3. 2 Mood Ratings
Mood ratings taken during the scans were analyzed to test the effectiveness of the mood induction paradigm. All mood ratings were normally distributed. For the analyses of non-normally distributed variables Wilcoxon’s tests (for comparisons between time points) and Mann-Whitney tests (for comparisons between groups) have been used. All tests have been treated as two-sided. No multiple comparisons corrections were used.
10 Participants’ mood after the neutral mood induction scan was equal to their mood before the scan (Z = -1.602, p=.109). No differences between patients and controls were found on the mood ratings before the scan (U = 540.5, p=.073), after the scan (U = 440, p=.103) and on the difference scores between the scans (U = 466, p=.353). Participants’ mood ratings after the sad mood induction scan dropped quite strongly compared to the mood ratings before the sad induction scan (Z = -5.187, p <.001). Patients rated their moods lower than controls both before (U = 332, p<.05) and after sad mood induction (= 360.5, p<0.05). However, the difference scores of the patients did not differ from that of controls (U = 378.5, p=.334). For an overview of means and standard deviations, see Table 2. For an overview of medians and quartiles, see Appendix A.
§3. 3 Statistical Testing
Independent Component Analysis.
Five components were selected from the ICA for further analysis, see Figure 2. These components reflected the DMN (which was separated in a posterior and anterior component), CEN (which was separated in a left and right component) and SN.
Full factorial ANOVA.
Analyses comparing patients and controls in neutral mood revealed that no differences between patients and controls existed at baseline. Also, the mixed model analyses of the SN, anterior DMN and right CEN components showed no group or mood state differences in connectivity
anywhere in the brain as well as no interaction effects.
Analysis of the left CEN component however showed a difference in right PPC connectivity between the different mood states on the cluster level (x=18, y=-70, z=0, k=247, Z=4.32, D
FWE=0.039), see Figure 3. Follow up analyses (one sided t-tests) showed that connectivity of the network with the PPC was increased during sad mood as compared to neutral mood(x=18, y=-70, z=0, k=401, Z=4.47, D
FWE=0.011). No other effects were found for this network.
Furthermore, analysis of the posterior DMN showed an interaction effect on cluster level in the right hippocampus (x=38, y=-44, z=0, k=216, Z=4.22, D
FWE=0.038), see Figure 4. Follow up analyses (one sided t-tests) revealed that the effects of mood induction differed for patients and controls: controls showed an increase in connectivity of the network with the hippocampus as a result of sad mood induction that was absent in patients(x=38, y=-44, z=0, k=288, Z=4.38, D
FWE=0.023), see Figure 5. No other effects were found for this network.
Correlation analyses.
To investigate how these findings relate to behavior, cognitive reactivity and rumination were correlated with change in hippocampus connectivity. The change in hippocampus connectivity was negatively related to both cognitive reactivity (LEIDS-R scores) and rumination (RRS scores), r=-0.201 (F(1)=4.097, p<.05 ) and r=-0.262 (F(1)= 6,359, p<.05) respectively. Inspection of the
relationship showed that high scores on cognitive reactivity and rumination predicted low change in hippocampus connectivity in response to sad mood induction, see Figure 6.
11 Figure 1. Participant flowchart
Screening N> 500 Intake N= 65 controls Inclusion N= 46 controls NPR N= 44 controls fMRI N= 42 controls Intake N= 125 patients Inclusion N= 73 patients NPR N= 69 patients fMRI N= 62 patients
Not fulfilling inclusion criteria (e.g. age, current MDD)
Excluded:
- Refusal n= 8
- History of depression n= 5 - First degree familial
psychiatric disorder n= 5 - Unknown n= 1 Excluded: - Refusal n=10 - HDRS>7 n= 13 - <2 experienced MDD episodes n= 4 - No primary depression n= 9 - Bipolar disorder n= 5 - Current depression n= 5 - Current use AD n=2 - Severe head trauma n= 1 - Drug dependency n=1 - Unknown n= 2 Refusal Excluded: - Refusal n= 1 - Technical reason n= 1 Excluded: - Refusal n= 4
- MRI exclusion criteria n=2 - Current depression n=1 Analysis N= 39 controls Analysis N= 52 patients Excluded:
- Abnormal brain anatomy n=2 - Possibility psychiatric illness
n=1 Excluded:
- Failed preprocessing n=2 - Abnormal brain anatomy n=3 - Excessive motion n=5
12 Table 1. Sample Characteristics.
Between-Group Statistics Remitted Patients (n = 62) Healthy Controls (n = 42) χ2 T U p Female N (%) 43 (69) 28 (67) 0.084 0.77 Age Years; mean (SD) 53.7 (7.9) 51.9 (8.1) 1.13 0.26 Education Levels1 0/0/0/4/2 1/23/14 0/0/0/1/1 6/17/8 1.211 0.75 IQ Mean (SD) 109 (8.48) 106 (9.73) 897 0.125 Living situation Levels2 26/0/18/ 14/2/0/0 11/0/16/1 1/4/0/0 5.59 0.23 Employment status Levels3 24/23/15 /0 21/17/4/0 3.76 0.15 Handedness Levels4 4/50/4 4/34/2 0.422 0.81 Age of onset Years;
mean (SD) 27.18 (11.18)4 NA Episodes Mean (SD) 8.02 (11.7) 4 NA HDRS Mean (SD) Range 2.63 (2,21) 0 - 5 1.05 (1,41) 0 - 7 670 <0.001 LEIDS-R Mean (SD) Range 38.6 (14.1) 8 - 67 15.3 (15.5) 0 - 62 278.5 <0.001 RRS Mean (SD) Range 38.1 (12.4) 22 - 78 25.4 (7.3) 22 - 41 279.5 <0.001 1
Less than primary school/primary school/primary school and two years of further education/special higher education/low level higher education/higher education/high level higher education
2
Alone/with parents/with partner/with partner and children/alone with children/other/unknown 3
Unemployed/employed/retired/unknown 4
13 Table 2. Means, Standard Deviations and Test Statistics of Mood Ratings.
Patients (N = 52) Controls (N = 39) Test Statistics Mean SD Mean SD U Z p Neutral Mood Before scan After scan Difference scores 7.3 6.9 -0.5 0.873 1.214 0.940 7.6 7.3 -0.2 0.758 0.670 0.655 540.5 440 466 -1.601 0.109 0.073 0.103 0.353 Sad Mood Before scan After scan Difference scores 6.4 5 -1.2 1.240 1.733 1.356 7 6 -1 0.793 1.372 1.170 332 360.5 378.5 -5.192 <0.001 <0.05 <0.05 0.334 1
Comparison of mood before the neutral scan and after the neutral scan. 2
Comparison of mood before the sad scan and after the sad scan.
Finally, a post-hoc analysis of connectivity with the hippocampus was performed to
investigate whether the effect extends beyond the hippocampus. No interesting effects in light of the TNM have been found. However, the reader might be interested in changes in hippocampal
connectivity according to mood. For a discussing of this analysis, see Appendix B.
§4. Discussion
In this study, the Triple Network Model was evaluated in remitted MDD patients. Against our expectations, none of the hypothesized differences between remitted patients and controls were observed. Patients did not differ from controls in both intra- and internetwork connectivity in or between the DMN, CEN and SN in neutral or sad mood state. In sad mood state however, the
hippocampus was found to be more connected to the posterior DMN in controls, but not in patients. Furthermore, the change in connectivity was negatively related to trait and state rumination,
meaning that more connectivity predicted lower rumination scores.
§4. 1 The Triple Network Model
It was unexpected that no TNM related connectivity differences in neutral and sad mood-state were found. It’s possible that no TNM related abnormalities persist in a remitted mood-state. If this were true, it might concluded that while the TNM is useful on a theoretical level or for treatment of currently depressed patients, it has no value for developing ways to predict relapse. Possibly, dysregulation of the three networks only becomes apparent when someone is already depressed.
However, in this study only functional connectivity has been investigated. Possibly, aberrant functioning of the three networks is not a result of aberrant functional connectivity, but of aberrant activation of the networks. Perhaps cooperation between (or within) networks is not affected, meaning for example that the precuneus and medial frontal cortex still get activated at the same time, but are the structures activated more often in general. Though functional connectivity alone does not seem useful to predicting relapse, the TNM might be worth investigating further.
14
Figure 2. Components chosen from the ICA analysis. (a) Left CEN. This component represents the left CEN. It includes the left PPC and prefrontal cortex. (b) Right CEN. This component includes the right PPC and prefrontal cortex, forming the right CEN. (c) Anterior DMN. This component represents the anterior DMN. It includes the frontal medial cortex, precuneus and medial temporal lobes. (d) Posterior DMN. This component represents the posterior DMN. It includes the posterior medial cortex, precuneus and medial temporal lobes. In these results, the hippocampus is not part of the DMN (e) SN. This component represents the SN. The insula and ACC are clearly visible. Color values represent t-values.
15 Figure 3. Effect of mood on connectivity with the left
CEN. A cluster of 401 voxels covering the right posterior parietal cortex was found to be more connected with the CEN during sad mood as opposed to neutral mood. Color scales represent t-values (p<.001).
Figure 4. Interaction effect on connectivity with the posterior DMN. A cluster of 288 voxels covering the hippocampus and surrounding white matter was found to be more connected with the posterior DMN during sad mood in controls, but not in patients. Color scales represent t-values (p<.001).
Figure 5. Interaction effect on connectivity with the posterior DMN. Contrast estimates of the peak voxel of the cluster found during posterior DMN analysis are plotted for every group and mood state. Patients show no change in connectivity in response to mood alterations, while controls show an increase of connectivity of this voxel in response to sad mood.
16 Figure 7. Exploratory analysis of the connectivity with the hippocampus. (a) Left temporal pole and insular cortex. A cluster of 528 voxels covering the left temporal pole and insular cortex showed more connectivity with the hippocampus while in sad mood as opposed to neutral mood. (b) Right SMA. A cluster of 95 voxels covering the right SMA showed more connectivity with the hippocampus during sad mood. (c) Precuneus and left LOC. A cluster of 777 voxels and a cluster of 394 voxels covering the precuneus and left LOC respectively showed more connectivity with the hippocampus in neutral mood. (d) Left precentral gyrus A cluster of 134 voxels covering the left precentral gyrus showed more connectivity with the hippocampus in controls as opposed to patients. Color scales represent t-values (p<.001). Figure 6. The relationship between cognitive reactivity, rumination and change in hippocampus
connectivity. Both cognitive reactivity (LEIDS-R, left image) and rumination (RRS, right image) were found to be negatively related to the change in hippocampus connectivity.
17 §4. 2 Increased Connectivity Between the Hippocampus and the Posterior DMN
The increased connectivity between the posterior DMN areas and the hippocampus during sad mood in controls was also an unexpected finding. At first glance, this finding seems to be in direct contradiction with those of Zamoscik, Huffziger, Ebner-Priemer, Kuehner and Kirsch (2014)
mentioned in the introduction. However, they did not examine connectivity between the DMN and the rest of the brain, but between the parahippocampal gyrus and one ROI in the DMN (the posterior cingulate cortex). It is possible that the parahippocampal gyrus is more connected to this ROI in sad mood in patients while the hippocampus shows less connectivity with the DMN as a whole. The findings are therefore not necessarily contradictive.
Since the hippocampus is strongly related to autobiographical memory (Svoboda, McKinnon & Levine, 2006), our findings might be more counterintuitive for some than the findings of Zamoscik, Huffziger, Ebner-Priemer, Kuehner and Kirsch (2014). One might expect that (remitted) MDD patients would show stronger connectivity between the posterior DMN and hippocampus than controls because they tend to ruminate excessively (a behavior associated with DMN activity, see
introduction), which inherently requires autobiographical recall. However, the theory of Overgeneral Autobiographical Memory (OAM, Gibbs & Rude, 2004) would lead to the opposite hypothesis. OAM refers to the robust observation that depressed patients tend to recall categories of autobiographical memories instead of specific instances. For example, when asked to remember their last birthday, depressed patients would be more likely to not remember a specific party, but the concept of a party and their general experience. One explanation for this effect is based on research with mice showing that replay of a memory both during sleep and in an awake state will cause that memory to be consolidated in cortical areas beyond the hippocampus (Carr, Jadhav & Frank, 2011). The degree of consolidation of a memory is negatively related to both the specificity of the memory and the involvement of the hippocampus required during retrieval (Wiltgen et. al., 2010).
Rumination, which inherently requires retrieval of autobiographical memories, might therefore cause specific memories to be consolidated in cortical areas more strongly than they would have been if rumination was not present. This would both reduce specificity of the memory and involvement of the hippocampus during retrieval.
This theory is supported by several studies showing that rumination is positively related to behavioral measures of OAM in depressed patients (Raes et. al. 2005; Raes, Watkins, Williams & Hermans, 2008). For example, both Watkins, Teasdale and Williams (2000) and Park, Goodyer and Teasdale (2004) found that induced rumination compared to distraction produced less detailed descriptions of memories (more OAM) and a depressed mood in depressed patients. Furthermore, Young, Bellgowan, Bodurka and Drevets (2014) showed that OAM is also present in remitted
patients. The theory therefore seem to be able to account for both the differences in connectivity of the hippocampus with the DMN and the negative relationship of this difference with rumination found in this study. OAM might therefore be an interesting theory for further research with remitted MDD patients.
§4. 3 Limitations
Finally, some limitations of the study need to be addressed. First, no multiple comparisons correction has been used for independent analyses. This decision has been made due to the explorative nature of the research project. In order to be sure of the findings’ robustness, research investigating more specific hypotheses about the found effect needs to be performed.
18 Second, this paper has been referring to the hippocampus as being separate form the DMN. This is because in this specific sample, the ICA has not found the hippocampus to be included in both the posterior and anterior DMN. Therefore the found effect has also not been treated as a within network connectivity difference. However, as mentioned in the introduction, most studies find a DMN that does include the hippocampus. Some might therefore argue that the found effect reflects a within-network difference between remitted patients and controls. While this might be true, it does not change the conclusions drawn about the TNM: the hypothesis was that, if anything, patients would show more DMN connectivity than controls, not the other way around. Nonetheless, this distinction might be relevant for future research.
Third, as already briefly mentioned, the TNM has only been investigated using functional connectivity. However, an approach that combines RSNs and activation based paradigms might reveal differences in TNM functioning between remitted patients and controls that were missed in this study. For example, a new study could use resting state scans for finding the three networks and then use mood induction and a task to investigate whether remitted patients activate their networks in different ways than controls do. Another possibility for further investigation of the TNM in
remitted patients might be examining activation of the anterior insula under certain conditions. Finally, it is not certain that this sample experienced OAM because no behavioral measures of it were taken. Conclusions based on this theory are therefore somewhat suggestive. Also, it is not known whether OAM causes depression or depression causes OAM. Depression might be caused by OAM when categorical memories have negative connotations. For instance, systematically recalling the categorical memory “everyone betrays my trust” instead of the specific memory “this specific person has done something to harm me” might negatively affect mood. It’s also possible however that OAM is caused by depression and is a leftover symptom in remitted patients, perhaps without any predictive relationship with relapse. Future research should include behavioral measures of OAM, ideally in a follow up system so it can be investigated whether increases in OAM over time increase the likelihood of relapse.
§4. 4 Conclusions
To conclude, this study suggests that the TNM might be more relevant in currently depressed patients than in remitted patients. Perhaps TNM aberrations present a state factor instead of a trait factor. However, the preliminary results from this study also indicate that the hippocampus might play an important role in processing of sad mood in remitted depression. Possibly, this is related to OAM. Future research on this theory might deliver results of more practical use for the prediction, and possibly prevention, of depressive relapse.
References
Andreescu, C., Lenze, E. J., Dew, M. A., Begley, A. E., Mulsant, B. H., Dombrovski, A. Y., … & Reynolds, C. F. (2007). Effect of comorbid anxiety on treatment response and relapse risk in late-life
depression: controlled study. The British Journal of Psychiatry, 190(4), 344-349.
Anticevic, A., Cole, M. W., Murray, J. D., Corlett, P. R., Wang, X. J., & Krystal, J. H. (2012). The role of default network deactivation in cognition and disease. Trends in cognitive sciences, 16(12), 584-592.
19 Atchley, R. A., Stringer, R., Mathias, E., Ilardi, S. S., & Minatrea, A. D. (2007). The right hemisphere’s
contribution to emotional word processing in currently depressed, remitted depressed, and never-depressed individuals. Journal of Neurolinguistics, 20(2), 145-160.
Austin, M. P., Mitchell, P., & Goodwin, G. M. (2001). Cognitive deficits in depression possible
implications for functional neuropathology. The British Journal of Psychiatry, 178(3), 200-206. Bhagwagar, Z. U. B. I. N., & Cowen, P. J. (2008). ‘It’s not over when it’s over’: persistent
neurobiological abnormalities in recovered depressed patients.Psychological medicine, 38(03), 307-313.
Bockting, C. L., Schene, A. H., Spinhoven, P., Koeter, M. W., Wouters, L. F., Huyser, J., & Kamphuis, J. H. (2005). Preventing relapse/recurrence in recurrent depression with cognitive therapy: a randomized controlled trial. Journal of consulting and clinical psychology, 73(4), 647-657. Bonnelle, V., Ham, T. E., Leech, R., Kinnunen, K. M., Mehta, M. A., Greenwood, R. J., & Sharp, D. J.
(2012). Salience network integrity predicts default mode network function after traumatic brain injury. Proceedings of the National Academy of Sciences, 109(12), 4690-4695.
Borsboom, D., Cramer, A. O., Schmittmann, V. D., Epskamp, S., & Waldorp, L. J. (2011). The small world of psychopathology. PloS one, 6(11), e27407.
Bouhuys, A. L., Geerts, E., & Gordijn, M. C. (1999). Depressed patients’ perceptions of facial emotions in depressed and remitted states are associated with relapse: a longitudinal study. The Journal of nervous and mental disease,187(10), 595-602.
Buysse, D. J., Frank, E., Lowe, K. K., Cherry, C. R., & Kupfer, D. J. (1997). Electroencephalographic sleep correlates of episode and vulnerability to recurrence in depression. Biological psychiatry, 41(4), 406-418.
Campbell, S., Marriott, M., Nahmias, C., & MacQueen, G. M. (2014). Lower hippocampal volume in patients suffering from depression: a meta-analysis. American Journal of Psychiatry, 161(4), 598-607.
Carr, M. F., Jadhav, S. P., & Frank, L. M. (2011). Hippocampal replay in the awake state: a potential substrate for memory consolidation and retrieval. Nature neuroscience, 14(2), 147-153. Chen, A. C., Oathes, D. J., Chang, C., Bradley, T., Zhou, Z. W., Williams, L. M., … & Etkin, A. (2013).
Causal interactions between fronto-parietal central executive and default-mode networks in humans. Proceedings of the National Academy of Sciences, 110(49), 19944-19949.
Damoiseaux, J. S., Rombouts, S. A. R. B., Barkhof, F., Scheltens, P., Stam, C. J., Smith, S. M., & Beckmann, C. F. (2006). Consistent resting-state networks across healthy
subjects. Proceedings of the national academy of sciences, 103(37), 13848-13853. Dombrovski, A. Y., Mulsant, B. H., Houck, P. R., Mazumdar, S., Lenze, E. J., Andreescu, C., … &
Reynolds, C. F. (2007). Residual symptoms and recurrence during maintenance treatment of late-life depression. Journal of affective disorders, 103(1), 77-82.
Farb, N. A., Anderson, A. K., Bloch, R. T., & Segal, Z. V. (2011). Mood-linked responses in medial prefrontal cortex predict relapse in patients with recurrent unipolar depression. Biological psychiatry, 70(4), 366-372.
First, M. B. (1995). Structured Clinical Interview for the DSM (SCID). The Encyclopedia of Clinical Psychology, 1-6.
Fleck, M. P. D. A., Poirier-Littre, M. F., Guelfi, J. D., Bourdel, M. C., & Loo, H. (1995). Factorial structure of the 17-item Hamilton Depression Rating Scale. Acta Psychiatrica
20 Gibbs, B. R., & Rude, S. S. (2004). Overgeneral autobiographical memory as depression vulnerability. Cognitive Therapy and Research, 28(4), 511-526.
Giles, D. E., Jarrett, R. B., Biggs, M. M., Guzick, D. S., & Rush, A. J. (1989). Clinical predictors of recurrence in depression. American Journal of Psychiatry, 146(6), 764-767.
Goulden, N., Khusnulina, A., Davis, N. J., Bracewell, R. M., Bokde, A. L., McNulty, J. P., & Mullins, P. G. (2014). The salience network is responsible for switching between the default mode network and the central executive network: replication from DCM. Neuroimage, 99, 180-190.
Greicius, M. (2008). Resting-state functional connectivity in neuropsychiatric disorders. Current opinion in neurology, 21(4), 424-430.
Hamilton, J. P., Siemer, M., & Gotlib, I. H. (2008). Amygdala volume in major depressive disorder: a meta-analysis of magnetic resonance imaging studies.Molecular psychiatry, 13(11), 993- 1000.
Hardeveld, F., Spijker, J., De Graaf, R., Nolen, W. A., & Beekman, A. T. F. (2010). Prevalence and predictors of recurrence of major depressive disorder in the adult population. Acta Psychiatrica Scandinavica, 122(3), 184-191.
Hart, A. B., Craighead, W. E., & Craighead, L. W. (2001). Predicting recurrence of major depressive disorder in young adults: A prospective study. Journal of Abnormal Psychology, 110(4), 633 – 643.
Hunt, M., Auriemma, J., & Cashaw, A. C. (2003). Self-report bias and underreporting of depression on the BDI-II. Journal of Personality Assessment,80(1), 26-30.
Johnson, J., Weissman, M. M., & Klerman, G. L. (1992). Service utilization and social morbidity associated with depressive symptoms in the community. Jama, 267(11), 1478-1483. Judd, L. L., Paulus, M. J., Schettler, P. J., Akiskal, H. S., Endicott, J., Leon, A. C., … & Keller, M. B.
(2014). Does incomplete recovery from first lifetime major depressive episode herald a chronic course of illness?.
Kaymaz, N., Os, J. V., Loonen, A. J., & Nolen, W. A. (2008). Evidence that patients with single versus recurrent depressive episodes are differentially sensitive to treatment discontinuation: A meta-analysis of placebo-controlled randomized trials. Journal of Clinical Psychiatry, 69(9), 1423-1436.
Keedwell, P. A., Andrew, C., Williams, S. C., Brammer, M. J., & Phillips, M. L. (2005). A double dissociation of ventromedial prefrontal cortical responses to sad and happy stimuli in depressed and healthy individuals. Biological psychiatry, 58(6), 495-503.
Kessing, L. V., Andersen, E. W., & Andersen, P. K. (2000). Predictors of recurrence in affective disorder—analyses accounting for individual heterogeneity. Journal of affective disorders, 57(1), 139-145.
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D., Merikangas, K. R., … & Wang, P. S. (2003). The epidemiology of major depressive disorder: Results from the National Comorbidity Survey Replication (NCS-R). Jama, 289(23), 3095-3105.
Klein, D. N., Schatzberg, A. F., McCullough, J. P., Dowling, F., Goodman, D., Howland, R. H., … & Keller, M. B. (1999). Age of onset in chronic major depression: relation to demographic and clinical variables, family history, and treatment response. Journal of Affective
21 Kuyken, W., Byford, S., Taylor, R. S., Watkins, E., Holden, E., White, K., … & Teasdale, J. D. (2008).
Mindfulness-based cognitive therapy to prevent relapse in recurrent depression. Journal of consulting and clinical psychology, 76(6), 966.
Liotti, M., Mayberg, H. S., McGinnis, S., Brannan, S. L., & Jerabek, P. (2014). Unmasking disease- specific cerebral blood flow abnormalities: mood challenge in patients with remitted unipolar depression. American Journal of Psychiatry.
Maldjian, J .A., Laurienti, P. J., & Burdette, J. H. (2003). Precentral gyrus discrepancy in electronic versions of the Talairach atlas. NeuroImage, 21(1), 450-455.
Maldjian, J. A., Laurienti, P. J. , Burdette, .J. B. , & Kraft, R. A. (2003). An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. NeuroImage, 19, 1233-1239.
Manoliu, A., Meng, C., Brandl, F., Doll, A., Tahmasian, M., Scherr, M., … & Sorg, C. (2013). Insular dysfunction within the salience network is associated with severity of symptoms and aberrant inter-network connectivity in major depressive disorder. Frontiers in human neuroscience, 7.
Marchetti, I., Koster, E. H., Sonuga-Barke, E. J., & De Raedt, R. (2012). The default mode network and recurrent depression: a neurobiological model of cognitive risk factors. Neuropsychology review, 22(3), 229-251.
Mayberg, H. S., Liotti, M., Brannan, S. K., McGinnis, S., Mahurin, R. K., Jerabek, P. A., … & Fox, P. T. (2014). Reciprocal limbic-cortical function and negative mood: converging PET findings in depression and normal sadness.
Menon, V. (2011). Large-scale brain networks and psychopathology: A unifying triple network model. Trends in cognitive sciences, 15(10), 483-506.
Menon, V., & Uddin, L. Q. (2010). Saliency, switching, attention and control: a network model of insula function. Brain Structure and Function, 214(5-6), 655-667.
Michalak, J., Hölz, A., & Teismann, T. (2011). Rumination as a predictor of relapse in
mindfulness-based cognitive therapy for depression. Psychology and Psychotherapy: Theory, Research and Practice, 84(2), 230-236.
Owen, A. M., McMillan, K. M., Laird, A. R., & Bullmore, E. (2005). N-back working memory paradigm: A meta-analysis of normative functional neuroimaging studies. Human brain mapping, 25(1), 46-59.
Park, R. J., Goodyer, I. M., & Teasdale, J. D. (2004). Effects of induced rumination and distraction on mood and overgeneral autobiographical memory in adolescent major depressive disorder and controls. Journal of Child Psychology and Psychiatry, 45(5), 996-1006.
Patel, R., Spreng, R. N., Shin, L. M., & Girard, T. A. (2012). Neurocircuitry models of posttraumatic stress disorder and beyond: a meta-analysis of functional neuroimaging
studies. Neuroscience & Biobehavioral Reviews, 36(9), 2130-2142.
Piet, J., & Hougaard, E. (2011). The effect of mindfulness-based cognitive therapy for prevention of relapse in recurrent major depressive disorder: A systematic review and meta-
analysis. Clinical Psychology Review, 31(6), 1032-1040.
Poldrack, R. A., Mumford, J. A., Nichols, T. E. (2011). Handbook of functional MRI data analysis. New York: Cambridge University Press.
Raes, F., Watkins, E. R., Williams, J. M. G., & Hermans, D. (2008). Non-ruminative processing reduces overgeneral autobiographical memory retrieval in students. Behaviour Research and Therapy, 46(6), 748-756.
22
Raes, F., Hermans, D., Williams, J. M. G., Demyttenaere, K., Sabbe, B., Pieters, G., & Eelen, P. (2005). Reduced specificity of autobiographical memory: A mediator between rumination and ineffective social problem-solving in major depression?. Journal of affective disorders, 87(2), 331-335.
Ramel, W., Goldin, P. R., Eyler, L. T., Brown, G. G., Gotlib, I. H., & McQuaid, J. R. (2007). Amygdala reactivity and mood-congruent memory in individuals at risk for depressive
relapse. Biological psychiatry, 61(2), 231-239.
Richards, D. (2011). Prevalence and clinical course of depression: A review, Clinical Psychology Review, 31(7), 1117-1125.
Segal, Z. V., Kennedy, S., Gemar, M., Hood, K., Pedersen, R., & Buis, T. (2006). Cognitive reactivity to sad mood provocation and the prediction of depressive relapse. Archives of General
Psychiatry, 63(7), 749-755.
Sigmon, S. T., Pells, J. J., Boulard, N. E., Whitcomb-Smith, S., Edenfield, T. M., Hermann, B. A., … & Kubik, E. (2005). Gender differences in self-reports of depression: The response bias hypothesis revisited. Sex Roles, 53(5-6), 401-411.
Sridharan, D., Levitin, D. J., & Menon, V. (2008). A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks. Proceedings of the National Academy of Sciences, 105(34), 12569-12574.
Steinert, C., Hofmann, M., Kruse, J., & Leichsenring, F. (2014). Relapse rates after psychotherapy for depression–stable long-term effects? A meta-analysis.Journal of affective disorders, 168, 107- 118.
Steunenberg, B., Beekman, A. T., Deeg, D. J., & Kerkhof, A. J. (2010). Personality predicts recurrence of late-life depression. Journal of Affective Disorders, 123(1), 164-172.
Svoboda, E., McKinnon, M. C., & Levine, B. (2006). The functional neuroanatomy of autobiographical memory: a meta-analysis. Neuropsychologia,44(12), 2189-2208.
Treynor, W., Gonzalez, R., Nolen-Hoeksema, S. (2003). Rumination Reconsidered: A Psychometric Analysis. Cognitive Therapy and Research, 27(3), 247-259
Van der Does, AJW & Williams, JMG (2003). Leiden Index of Depression Sensitivity – Revised (LEIDS-R). Leiden University.
Videbech, P., & Ravnkilde, B. (2004). Hippocampal volume and depression: a meta-analysis of MRI studies. American Journal of Psychiatry, 161(11), 1957-1966.
Watkins, E., Teasdale, J. D., & Williams, R. M. (2000). Decentring and distraction reduce overgeneral autobiographical memory in depression. Psychological Medicine, 30(04), 911-920.
Wei, M., Qin, J., Yan, R., Bi, K., Liu, C., Yao, Z., & Lu, Q. (2015). Association of resting-state network dysfunction with their dynamics of inter-network interactions in depression. Journal of affective disorders, 174, 527-534.
Whitfield-Gabrieli, S., & Ford, J. M. (2012). Default mode network activity and connectivity in psychopathology. Annual review of clinical psychology, 8, 49-76.
Whitfield-Gabrieli, S., and Nieto-Castanon, A. (2012). Conn: A functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connectivity.
doi:10.1089/brain.2012.0073
Wiltgen, B. J., Zhou, M., Cai, Y., Balaji, J., Karlsson, M. G., Parivash, S. N., ... & Silva, A. J. (2010). The hippocampus plays a selective role in the retrieval of detailed contextual memories. Current Biology, 20(15), 1336-1344.
23 Young, K. D., Bellgowan, P. S. F., Bodurka, J., & Drevets, W. C. (2014). Neurophysiological correlates of autobiographical memory deficits in currently and formerly depressed subjects.
Psychological medicine, 44(14), 2951-2963.
Yuen, G. S., Gunning-Dixon, F. M., Hoptman, M. J., AbdelMalak, B., McGovern, A. R., Seirup, J. K., & Alexopoulos, G. S. (2014). The salience network in the apathy of late-life
depression. International journal of geriatric psychiatry, 29(11), 1116-1124.
Zheng, H., Xu, L., Xie, F., Guo, X., Zhang, J., Yao, L., & Wu, X. (2015). The Altered Triple Networks Interaction in Depression under Resting State Based on Graph Theory. BioMed Research International.
Zamoscik, V., Huffziger, S., Ebner-Priemer, U., Kuehner, C., & Kirsch, P. (2014). Increased involvement of the parahippocampal gyri in a sad mood predicts future depressive symptoms. Social
cognitive and affective neuroscience, nsu006.
Zobel, A. W., Nickel, T., Sonntag, A., Uhr, M., Holsboer, F., & Ising, M. (2001). Cortisol response in the combined dexamethasone/CRH test as predictor of relapse in patients with remitted
24 APPENDIX A
Table 3. Medians and First and Third Quartiles of Sample Demographics per Group
Patients (N = 62) Controls (N = 42)
Median Quartiles Median Quartiles
IQ 109 104, 113 105 101.25, 113
Amount of previous episodes 4 2, 7 NA NA
HDRS at intake* 2 1, 4.75 1 0, 1.25
RRS score* 36 30, 47 25 22.5, 26.5
LEIDSR score* 42 35, 38 9 6, 25.5
* significant difference between patients and controls (p<.001).
Table 4. Medians and First and Third Quartiles of Mood Ratings. Patients (N = 52) Controls (N = 39) Whole Sample (N = 91) Median Quartiles Median Quartiles Median Quartiles Neutral Mood Before scan After scan Difference scores 7 7 0 7, 8 6, 8 0, 1 7 7 0 7, 8 7, 8 0, 0.125 7 7 NA 7, 8 7, 8 NA Sad Mood* Before scan** After scan** Difference scores 7 5 1 6, 7 4, 6.5 0, 2.125 7 7 1 6, 7 5, 7 0, 2 7 6 NA 6, 7 4, 7 NA * significant difference between time points (p<.001).
25 APPENDIX B
To understand the role of the hippocampus in this context better, a seed-to-voxel mixed model analysis using the hippocampus as seed region has been performed using the CONN toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012). Seed-to-voxel analysis requires different preprocessing steps than ICA analysis. The data have been treated the same up until the smoothing step. First, using SPM the data was smoothed with a kernel of 6 instead of 10 to avoid loss of resolution. Using the CONN toolbox, the data was then subjected to segmentation, detrending (linear), despiking and filtering (band pass filter, 0.01-0.1 Hz) before first level analyses were conducted. The analysis used bivariate correlation as the measure of connectivity and used the hemodynamic response function for within condition weighting. The right hippocampus was selected as seed using the anatomic automatic labeling atlas of the WFU Pickatlas (Maldjian, Laurienti & Burdette, 2003; Maldjian, Laurienti, Burdette & Kraft, 2003)
The analysis showed that four regions differed in connectivity with the hippocampus according to mood: the precuneus (cluster size = 534 voxels, p<.001), the right Lateral Occipital Cortex (LOC, luster size = 260 voxels, p<.001), the left temporal pole (cluster size = 353 voxels, p<.001) and the right Supplementary Motor Area (SMA, cluster size = 95 voxels, p<.05). Follow up analyses (one sided t-tests) showed that connections between the hippocampus and the precuneus (cluster size = 777 voxels, p<.001) and left LOC (cluster size = 394 voxels, p<.001) were stronger during neutral than sad mood, while the connections between the hippocampus and the left
temporal pole (including the left insula) (cluster size = 528 voxels, p<.001) and right SMA (cluster size = 136 voxels, p<.05) were stronger during sad mood, see Figure 5. One effect of group was also found (cluster size = 134 voxels, p<.005): controls showed more connectivity between the hippocampus and the left precentral gyrus than patients (cluster size = 200 voxels, p<.01), see Figure 5. No interaction effects were found.[SM1]
