[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.147.196.37. Please contact the publisher to request reinstatement.
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Figure 1.
Brain Regions With Significant Differences in Activation Between Youth With Major Depressive Disorder (MDD) and Age-Matched Healthy Control Participants During Different Experimental Conditions
Brain Regions With Significant Differences in Activation Between Youth With Major Depressive Disorder (MDD) and Age-Matched Healthy Control Participants During Different Experimental Conditions

Brain regions reached significance at P = .05 (familywise error rate corrected) for the contrast of youth with MDD minus healthy control participants. Each brain map displays regions of peak difference shown in Talairach standard space according to neurological format. Positive activation values (warm colors) represent areas of hyperactivity (youth with MDD > healthy control participants) whereas negative activation values (cool colors) represent areas of hypoactivity (healthy control participants > youth with MDD).

Figure 2.
Neural Basis of Major Depressive Disorder (MDD) in Youth
Neural Basis of Major Depressive Disorder (MDD) in Youth

Brain regions with significant (P < .05, familywise error rate corrected) activation differences between youth with MDD and healthy control participants. Red circles indicate hyperactivity (youth with MDD > healthy control participants), and blue circles indicate hypoactivity (healthy control participants > youth with MDD).

Table 1.  
Included Whole-Brain fMRI Studies of Youth With MDD
Included Whole-Brain fMRI Studies of Youth With MDD
Table 2.  
Brain Regions With Significant Differences in Activation Between Youth With MDD and Age-Matched Healthy Control Participants During Different Experimental Conditions
Brain Regions With Significant Differences in Activation Between Youth With MDD and Age-Matched Healthy Control Participants During Different Experimental Conditions
1.
Birmaher  B, Brent  D, Bernet  W,  et al; AACAP Work Group on Quality Issues.  Practice parameter for the assessment and treatment of children and adolescents with depressive disorders. J Am Acad Child Adolesc Psychiatry. 2007;46(11):1503-1526.
PubMedArticle
2.
Lynch  FL, Clarke  GN.  Estimating the economic burden of depression in children and adolescents. Am J Prev Med. 2006;31(6)(suppl 1):S143-S151.
PubMedArticle
3.
Kerestes  R, Davey  CG, Stephanou  K, Whittle  S, Harrison  BJ.  Functional brain imaging studies of youth depression: a systematic review. Neuroimage Clin. 2014;4:209-231.
PubMedArticle
4.
Kessler  RC, Walters  EE.  Epidemiology of DSM-III-R major depression and minor depression among adolescents and young adults in the National Comorbidity Survey. Depress Anxiety. 1998;7(1):3-14.
PubMedArticle
5.
March  J, Silva  S, Petrycki  S,  et al; Treatment for Adolescents With Depression Study (TADS) Team.  Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: Treatment for Adolescents With Depression Study (TADS) randomized controlled trial. JAMA. 2004;292(7):807-820.
PubMedArticle
6.
Graham  J, Salimi-Khorshidi  G, Hagan  C,  et al.  Meta-analytic evidence for neuroimaging models of depression: state or trait? J Affect Disord. 2013;151(2):423-431.
PubMedArticle
7.
Chen  JJ, Liu  Z, Zhu  D,  et al.  Bilateral vs. unilateral repetitive transcranial magnetic stimulation in treating major depression: a meta-analysis of randomized controlled trials. Psychiatry Res. 2014;219(1):51-57.
PubMedArticle
8.
Mayberg  HS, Lozano  AM, Voon  V,  et al.  Deep brain stimulation for treatment-resistant depression. Neuron. 2005;45(5):651-660.
PubMedArticle
9.
Hulvershorn  LA, Cullen  K, Anand  A.  Toward dysfunctional connectivity: a review of neuroimaging findings in pediatric major depressive disorder. Brain Imaging Behav. 2011;5(4):307-328.
PubMedArticle
10.
Davey  CG, Allen  NB, Harrison  BJ, Yücel  M.  Increased amygdala response to positive social feedback in young people with major depressive disorder. Biol Psychiatry. 2011;69(8):734-741.
PubMedArticle
11.
Wager  TD, Lindquist  M, Kaplan  L.  Meta-analysis of functional neuroimaging data: current and future directions. Soc Cogn Affect Neurosci. 2007;2(2):150-158.
PubMedArticle
12.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 2000.
13.
Chantiluke  K, Halari  R, Simic  M,  et al.  Fronto-striato-cerebellar dysregulation in adolescents with depression during motivated attention. Biol Psychiatry. 2012;71(1):59-67.
PubMedArticle
14.
Colich  NL, Foland-Ross  LC, Eggleston  C, Singh  MK, Gotlib  IH.  Neural aspects of inhibition following emotional primes in depressed adolescents. J Clin Child Adolesc Psychol. 2015;0(0):1-10.
PubMedArticle
15.
Diler  RS, de Almeida  JRC, Ladouceur  C, Birmaher  B, Axelson  D, Phillips  M.  Neural activity to intense positive versus negative stimuli can help differentiate bipolar disorder from unipolar major depressive disorder in depressed adolescents: a pilot fMRI study. Psychiatry Res. 2013;214(3):277-284.
PubMedArticle
16.
Diler  RS, Pan  LA, Segreti  A,  et al.  Differential anterior cingulate activity during response inhibition in depressed adolescents with bipolar and unipolar major depressive disorder. J Can Acad Child Adolesc Psychiatry. 2014;23(1):10-19.
PubMed
17.
Gaffrey  MS, Barch  DM, Singer  J, Shenoy  R, Luby  JL.  Disrupted amygdala reactivity in depressed 4- to 6-year-old children. J Am Acad Child Adolesc Psychiatry. 2013;52(7):737-746.
PubMedArticle
18.
Halari  R, Simic  M, Pariante  CM,  et al.  Reduced activation in lateral prefrontal cortex and anterior cingulate during attention and cognitive control functions in medication-naïve adolescents with depression compared to controls. J Child Psychol Psychiatry. 2009;50(3):307-316.
PubMedArticle
19.
Hall  LMJ, Klimes-Dougan  B, Hunt  RH,  et al.  An fMRI study of emotional face processing in adolescent major depression. J Affect Disord. 2014;168:44-50.
PubMedArticle
20.
Roberson-Nay  R, McClure  EB, Monk  CS,  et al.  Increased amygdala activity during successful memory encoding in adolescent major depressive disorder: an FMRI study. Biol Psychiatry. 2006;60(9):966-973.
PubMedArticle
21.
Sharp  C, Kim  S, Herman  L, Pane  H, Reuter  T, Strathearn  L.  Major depression in mothers predicts reduced ventral striatum activation in adolescent female offspring with and without depression. J Abnorm Psychol. 2014;123(2):298-309.
PubMedArticle
22.
Tao  R, Calley  CS, Hart  J,  et al.  Brain activity in adolescent major depressive disorder before and after fluoxetine treatment. Am J Psychiatry. 2012;169(4):381-388.
PubMedArticle
23.
Yang  TT, Simmons  AN, Matthews  SC,  et al.  Depressed adolescents demonstrate greater subgenual anterior cingulate activity. Neuroreport. 2009;20(4):440-444.
PubMedArticle
24.
Yang  TT, Simmons  AN, Matthews  SC,  et al.  Adolescents with major depression demonstrate increased amygdala activation. J Am Acad Child Adolesc Psychiatry. 2010;49(1):42-51.
PubMed
25.
Mingtian  Z, Shuqiao  Y, Xiongzhao  Z,  et al.  Elevated amygdala activity to negative faces in young adults with early onset major depressive disorder. Psychiatry Res. 2012;201(2):107-112.
PubMedArticle
26.
Yarkoni  T, Poldrack  RA, Nichols  TE, Van Essen  DC, Wager  TD.  Large-scale automated synthesis of human functional neuroimaging data. Nat Methods. 2011;8(8):665-670.
PubMedArticle
27.
Poldrack  RA.  Inferring mental states from neuroimaging data: from reverse inference to large-scale decoding. Neuron. 2011;72(5):692-697.
PubMedArticle
28.
Poldrack  RA.  Can cognitive processes be inferred from neuroimaging data? Trends Cogn Sci. 2006;10(2):59-63.
PubMedArticle
29.
Hamilton  JP, Etkin  A, Furman  DJ, Lemus  MG, Johnson  RF, Gotlib  IH.  Functional neuroimaging of major depressive disorder: a meta-analysis and new integration of base line activation and neural response data. Am J Psychiatry. 2012;169(7):693-703.
PubMedArticle
30.
Sundermann  B, Olde Lütke Beverborg  M, Pfleiderer  B.  Toward literature-based feature selection for diagnostic classification: a meta-analysis of resting-state fMRI in depression. Front Hum Neurosci. 2014;8:692.
PubMedArticle
31.
Price  JL, Drevets  WC.  Neural circuits underlying the pathophysiology of mood disorders. Trends Cogn Sci. 2012;16(1):61-71.
PubMedArticle
32.
Connolly  CG, Wu  J, Ho  TC,  et al.  Resting-state functional connectivity of subgenual anterior cingulate cortex in depressed adolescents. Biol Psychiatry. 2013;74(12):898-907.
PubMedArticle
33.
Drevets  WC.  Neuroimaging studies of mood disorders. Biol Psychiatry. 2000;48(8):813-829.
PubMedArticle
34.
Pizzagalli  DA.  Frontocingulate dysfunction in depression: toward biomarkers of treatment response. Neuropsychopharmacology. 2011;36(1):183-206.
PubMedArticle
35.
Cooney  RE, Joormann  J, Eugène  F, Dennis  EL, Gotlib  IH.  Neural correlates of rumination in depression. Cogn Affect Behav Neurosci. 2010;10(4):470-478.
PubMedArticle
36.
Utevsky  AV, Smith  DV, Huettel  SA.  Precuneus is a functional core of the default-mode network. J Neurosci. 2014;34(3):932-940.
PubMedArticle
37.
Menon  V, Uddin  LQ.  Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct. 2010;214(5-6):655-667.
PubMedArticle
38.
Leech  R, Sharp  DJ.  The role of the posterior cingulate cortex in cognition and disease. Brain. 2014;137(pt 1):12-32.
PubMedArticle
39.
Harvey  PO, Pruessner  J, Czechowska  Y, Lepage  M.  Individual differences in trait anhedonia: a structural and functional magnetic resonance imaging study in non-clinical subjects. Mol Psychiatry. 2007;12(8):703, 767-775.
PubMedArticle
40.
Vaidya  JG, Knutson  B, O’Leary  DS, Block  RI, Magnotta  V.  Neural sensitivity to absolute and relative anticipated reward in adolescents. PLoS One. 2013;8(3):e58708.
PubMedArticle
41.
Maddock  RJ, Garrett  AS, Buonocore  MH.  Remembering familiar people: the posterior cingulate cortex and autobiographical memory retrieval. Neuroscience. 2001;104(3):667-676.
PubMedArticle
42.
Foland-Ross  LC, Hamilton  P, Sacchet  MD, Furman  DJ, Sherdell  L, Gotlib  IH.  Activation of the medial prefrontal and posterior cingulate cortex during encoding of negative material predicts symptom worsening in major depression. Neuroreport. 2014;25(5):324-329.
PubMed
43.
Guyer  AE, Choate  VR, Detloff  A,  et al.  Striatal functional alteration during incentive anticipation in pediatric anxiety disorders. Am J Psychiatry. 2012;169(2):205-212.
PubMedArticle
44.
Craig  AD.  How do you feel--now? the anterior insula and human awareness. Nat Rev Neurosci. 2009;10(1):59-70.
PubMedArticle
45.
Chang  LJ, Yarkoni  T, Khaw  MW, Sanfey  AG.  Decoding the role of the insula in human cognition: functional parcellation and large-scale reverse inference. Cereb Cortex. 2013;23(3):739-749.
PubMedArticle
46.
Gotlib  IH, Joormann  J.  Cognition and depression: current status and future directions. Annu Rev Clin Psychol. 2010;6:285-312.
PubMedArticle
47.
Etkin  A, Wager  TD.  Functional neuroimaging of anxiety: a meta-analysis of emotional processing in PTSD, social anxiety disorder, and specific phobia. Am J Psychiatry. 2007;164(10):1476-1488.
PubMedArticle
48.
Insel  T, Cuthbert  B, Garvey  M,  et al.  Research domain criteria (RDoC): toward a new classification framework for research on mental disorders. Am J Psychiatry. 2010;167(7):748-751.
PubMedArticle
49.
Watson  D.  Rethinking the mood and anxiety disorders: a quantitative hierarchical model for DSM-V. J Abnorm Psychol. 2005;114(4):522-536.
PubMedArticle
Original Investigation
Meta-analysis
October 2015

Meta-analysis of Functional Neuroimaging of Major Depressive Disorder in Youth

Author Affiliations
  • 1Stanford Neurosciences Institute, Department of Psychology, Stanford University, Stanford, California
  • 2Laureate Institute for Brain Research, Tulsa, Oklahoma
JAMA Psychiatry. 2015;72(10):1045-1053. doi:10.1001/jamapsychiatry.2015.1376
Abstract

Importance  Despite its high prevalence and morbidity, the underlying neural basis of major depressive disorder (MDD) in youth is not well understood.

Objectives  To identify in youth diagnosed as having MDD the most reliable neural abnormalities reported in existing functional neuroimaging studies and characterize their relations with specific psychological dysfunctions.

Data Sources  Searches were conducted in PubMed and Web of Science to identify relevant studies published from November 2006 through February 2015. The current analysis took place from August 21, 2014, to March 28, 2015.

Study Selection  We retained articles that conducted a comparison of youth aged 4 to 24 years diagnosed as having MDD and age-matched healthy controls using task-based functional magnetic resonance imaging and a voxelwise whole-brain approach.

Data Extraction and Synthesis  We extracted coordinates of brain regions exhibiting differential activity in youth with MDD compared with healthy control participants. Multilevel kernel density analysis was used to examine voxelwise between-group differences throughout the whole brain. Correction for multiple comparisons was performed by computing null hypothesis distributions from 10 000 Monte Carlo simulations and calculating the cluster size necessary to obtain the familywise error rate control at P < .05.

Main Outcomes and Measures  Abnormal levels of activation in youth diagnosed as having MDD compared with control participants during a variety of affective processing and executive functioning tasks.

Results  Compared with age-matched healthy control participants (n = 274), youth with MDD (n = 246) showed reliable patterns of abnormal activation, including the following task-general and task-specific effects: hyperactivation in subgenual anterior cingulate cortex (P < .05) and ventrolateral prefrontal cortex (P < .05) and hypoactivation in caudate (P < .01) across aggregated tasks; hyperactivation in thalamus (P < .03) and parahippocampal gyrus (P < .003) during affective processing tasks; hypoactivation in cuneus (P < .001), dorsal cingulate cortex (P < .05), and dorsal anterior insula (P < .05) during executive functioning tasks; hypoactivity in posterior insula (P < .005) during positive valence tasks; and hyperactivity in dorsolateral prefrontal cortex (P < .001) and superior temporal cortex (P < .003) during negative valence tasks.

Conclusions and Relevance  Altered activations in several distributed brain networks may help explain the following seemingly disparate symptoms of MDD in youth: hypervigilance toward emotional stimuli from the overactivation of central hubs in the subgenual anterior cingulate cortex and thalamus that lead to a cascade of other symptoms; ineffective emotion regulation despite increased activation of the dorsolateral prefrontal cortex and ventrolateral prefrontal cortex during affective processing, which may reverse across development or the clinical course; maladaptive rumination and poor executive control from difficulties shifting from default mode network activity to task-positive network activity during cognitively demanding tasks; and anhedonia from hypoactivation of the cuneus and posterior insula during reward processing.

Introduction

Major depressive disorder (MDD) is recognized as a psychiatric illness with a course that frequently begins early in development and is characterized by enormous social, academic, and economic costs.1,2 Indeed, MDD is now the leading cause of morbidity and disability among adolescents,3 with a point prevalence rate in this age group of approximately 14%.4 Moreover, 30% to 65% of depressed adolescents fail to respond to treatment5; thus, it is critical we increase our understanding of the pathophysiology and etiology of MDD and translate these findings to more effective approaches to prevention and treatment.

Advances in neuroimaging techniques have led to a growing interest in investigating abnormalities in the brain structure and function in individuals diagnosed as having MDD.6 These advances provide promising opportunities to advance our understanding of the underlying pathophysiological effects of MDD and generate novel treatment targets for emerging technologies7,8 and more conventional pharmacotherapy and psychotherapy.

Although most studies using neuroimaging in MDD have focused on depressed adults, there is a growing body of research using functional magnetic resonance imaging (fMRI) to examine neural function in children and adolescents diagnosed as having MDD. Two qualitative reviews of this literature3,9 have noted findings largely consistent with prevailing neural models of adult depression that implicate frontal regions (eg, the prefrontal cortex and orbitofrontal cortex), striatal areas (eg, the caudate and putamen), and limbic structures (eg, the cingulate cortex, amygdala, and hippocampus) as well as other areas located outside this circuitry, such as the insula, cuneus, superior parietal cortex, and middle temporal cortex.

To our knowledge, there is presently no quantitative integration of this work to address issues of reliability across studies and examine consistency in neural activations that have been reported in investigations of MDD in youth. Studies of neural function in youth with MDD have yielded contradictory findings and reported results that diverge from those documented in the adult depression literature. In addition, many studies of neural functioning in depressed youth have taken a region of interest approach, analyzing a specified group of brain regions (eg, the striatum) while ignoring other potentially important areas (eg, the cuneus); this strategy results in a targeted but biased search for neural abnormalities. Finally, although most studies in this area have included youth as old as 18 years, neuroscientists have argued that the developmental period used to study young participants should extend to individuals aged 24 years to capture important maturational changes in the prefrontal cortex.3,10

To address these issues, we conducted a meta-analysis of fMRI studies that used a voxelwise whole-brain approach (WBA) to identify brain regions characterized reliably by abnormal activation in depressed youth. This approach enabled us to combine results quantitatively from several studies to obtain greater statistical reliability and address conflicting findings in the literature. This approach also avoided a biased focus on brain regions that have received the greatest attention and helped to prioritize subsequent investigations of neural abnormalities in youth with MDD. It also allowed us both to identify abnormalities that were robust across a variety of experimental conditions and elucidate their relations with psychological dysfunction.

Methods
Overview

We applied the multilevel kernel density analysis11 to published fMRI studies that compared neural activation in groups of youth diagnosed as having MDD with age-matched healthy control participants.

Study Selection

We conducted an extensive literature search with the PubMed and Web of Science databases for fMRI studies of MDD in youth published from November 2006 through February 2015. All analyses in this study were completed from August 21, 2014, to March 28, 2015. We also examined the reference lists and study tables of relevant review articles to identify other primary studies to include that may have been missed in the original search. We then inspected each article generated by this search process and retained only articles that satisfied the following inclusion criteria: (1) used an fMRI voxelwise WBA of task-based activation data; (2) compared a group of participants aged 4 to 24 years (mean, 18.36 years) diagnosed as having MDD according to DSM criteria12 at the time of the scan with age-matched healthy control participants; and (3) reported coordinates of brain regions with abnormal activations in standard space, using the Talairach atlas or Montreal Neurological Institute template. Other details regarding this study selection process are provided in the eAppendix in the Supplement.

Data Analysis

For each included study, we first extracted published whole-brain activation coordinates presented in Talairach or Montreal Neurological Institute space of regions showing significant between-group differences. We then constructed indicator maps for each individual study in Talairach space at 1-mm isotropic voxel resolution using these reported coordinates. Next, we merged them to create a meta-analytic statistical map composed of global activation values computed as the weighted proportion of primary studies reporting statistically significant activation differences between MDD and control groups for each voxel throughout the whole brain.

To determine significance at a voxelwise level, we conducted 10 000 Monte Carlo simulations to generate null hypothesis distributions and then computed the cluster size necessary to obtain familywise error rate control at P < .05 for the comparison of multiple voxels across the whole brain. As an additional constraint, only clusters that were reported in at least 2 primary studies were retained to prevent a single study from generating a significant meta-analytic finding. Further details about this process are provided in the eAppendix in the Supplement.

Study Groupings

We applied the meta-analytic procedure described earlier to several groupings of reported contrasts from primary studies to elucidate the psychological functioning associated with observed MDD abnormalities in neural activation. At the broadest level, we combined reported findings from all included studies (aggregated, n = 14), excluding those with overlapping samples (n = 3) to maximize statistical power and identify abnormalities in neural activations that were robust across diverse experimental conditions. Next, we separately analyzed findings from studies that involved either emotional processing tasks (affective processing, n = 11) or executive functioning tasks (executive function, n = 5) to examine specific abnormalities in activation associated with these neurocognitive systems. We then further decomposed the emotional processing tasks and separately analyzed findings from positive vs neutral contrasts (positive valence, n = 5) and negative vs neutral contrasts (negative valence, n = 5). Finally, to examine whether observed differences between positive vs negative contrasts were statistically significant, we directly contrasted activation maps from these 2 conditions with one another (valence specific, n = 7). The primary studies that contributed to each of these 6 groupings along with a description of the experimental task and reported contrasts are presented in Table 1 and further details regarding these groupings are provided in the eAppendix in the Supplement.

Neurosynth Decoding

We used the Neurosynth (http://www.neurosynth.org) package (https://github.com/neurosynth; Python) to decode whole-brain maps from each of the 5 groupings of primary studies described earlier to provide a systematic quantitative basis for inferring psychological functioning from observed brain activity,26 which helped guide the interpretation of findings while avoiding many of the hazards of reverse inference.27 This decoding process compared whole-brain activation levels in submitted activation maps with those obtained in the Neurosynth database, which currently contains 9721 whole-brain human fMRI studies and automatically computes Pearson correlation coefficients between submitted whole-brain maps and 3160 key terms (features) used in titles and abstracts of published articles in the database. Importantly, the decoding process uses activation levels obtained throughout the whole brain rather than individual regions of interest to enable more comprehensive naturalistic inferences regarding underlying mental states.26,27

Results
Study Characteristics

Our literature search yielded 14 primary studies that satisfied the inclusion criteria described earlier and were used in 1 or more analyses. These studies collectively examined in-episode MDD (n = 246) and healthy control participants (n = 274) across a broad range of task conditions and ages (mean, 14.94 years) and also included a diverse set of other characteristics of MDD, such as youth with first-episode MDD and medication-naive youth. Table 1 and eTable 1 in the Supplement summarize the major characteristics of participants with MDD and neuroimaging tasks used in each primary study.

Neuroimaging Results

Results of our meta-analysis revealed that youth diagnosed as having MDD showed reliably different activation levels in several brain regions compared with age-matched healthy control participants, including both task-general (ie, aggregated) and task-specific (ie, affective processing, executive function, positive valence, and negative valence) effects. The main findings from these analyses organized by experimental task are described as follows and are also summarized in Table 2 and depicted graphically in Figure 1.

Aggregated

Across all experimental conditions combined, youth with MDD exhibited hyperactivity compared with healthy control participants in neural clusters centered at the left dorsolateral prefrontal cortex (dlPFC; P < .03), left subgenual anterior cingulate cortex (sgACC; P < .05), right anterior insula (P < .005), bilateral thalamus (P < .01), left parahippocampal gyrus (P < .003), and left superior temporal cortex (STC; P < .03) as well as hypoactivity centered at the right caudate (P < .03).

Affective Processing and Executive Function

When the analyses were limited to affective processing tasks, youth with MDD continued to exhibit hyperactivity in clusters centered at the left dlPFC (P < .03), bilateral thalamus (P < .03), and left parahippocampal gyrus (P < .003). In contrast, during executive function tasks, youth with MDD showed hypoactivity centered at the right dorsal cingulate cortex (P < .05), right dorsal anterior insula (P < .05), and left cuneus (P < .003).

Positive and Negative Valence

During positive vs neutral conditions, youth with MDD exhibited hypoactivity centered at the right posterior insula (P < .005). In contrast, during negative vs neutral conditions, depressed youth showed hyperactivity in the left dorsolateral prefrontal cortex (P < .001) and left STC (P < .003). Moreover, when positive and negative valence activation maps were contrasted directly with one another, each region reached significance (P = .03 to P = .001), strongly suggesting these effects are specific to youth with MDD in particular valence conditions.

Neurosynth Analysis

The Neurosynth features related to psychological functioning that most strongly correlated with each of the 5 whole-brain maps (youth with MDD minus healthy control participants) are presented in the eFigure and eTable 2 in the Supplement. For the aggregated and affective processing maps, the most highly correlated terms were predominantly related to general affective processing (eg, emotion and emotional), negative valence emotions (eg, unpleasant and fear), or high arousal (eg, arousal and threat). For the positive valence map, the same terms described earlier were also highly correlated as well as others more directly related to MDD (eg, depressed and sad) and a novel term (conflicting) whereas for the negative valence map, similar terms were highly correlated in addition to a novel term (avoidance). For the executive function map, general terms related to cognitive processing (eg, cognition, cognitive control, and memory) as well as terms more directly related to attentional shifting (eg, shifting and competition), error monitoring (eg, monitoring), and response inhibition (eg, inhibition and competition) were strongly and inversely correlated.

Discussion

In this meta-analysis, we identified several brain regions in which youth diagnosed as having MDD reliably exhibited abnormal levels of activation compared with age-matched healthy control participants during affective processing and executive functioning tasks. In formulating a theoretically informative and clinically useful model of MDD, it is important to consider both the independent functioning of each brain region and how regions may interact with each other as key components of 1 or more distributed networks while acknowledging that this process of reverse inference is necessarily speculative.28

Analysis of Individual Brain Regions

We found that several brain regions were reliably under- or overreactive to specific stimuli in youth diagnosed as having MDD. Importantly, each region identified in this meta-analysis received substantial attention in neuroimaging studies of both clinical and nonclinical samples using a variety of experimental tasks designed to probe different cognitive functions. Thus, there is a relatively rich literature concerning the possible dysfunctions that might be associated with each of these regions.

We observed reliable overactivation centered in the sgACC across experimental tasks and the thalamus during affective processing tasks in youth with MDD. These neural regions have been implicated in emotional attention and salience attribution and both have shown reliable overactivation at baseline or wakeful rest in meta-analyses of adult MDD.29,30 In particular, the sgACC is a surgical target for deep-brain stimulation in treatment-refractory MDD8 because of its extensive functional connectivity with other limbic regions that together form a medial prefrontal network.31 Elevated activity in the sgACC and reduced functional connectivity between the sgACC and the cuneus and insula as well as elevated functional connectivity with the dlPFC have been associated with adolescent MDD32; importantly, these patterns correspond closely to our findings of hyperactivity in the sgACC and dlPFC and hypoactivity in the cuneus and insula. Similarly, the thalamus has been postulated to potentiate activity in other distributed regions that together form a salience network for emotional processing.29

We found that the dlPFC was selectively overresponsive to negative stimuli and the ventrolateral prefrontal cortex was hyperactive across tasks in youth with MDD. These brain regions are among the most prominent components of frontolimbic models of MDD and have been implicated consistently in neuroimaging studies of depressed adults as centers of cognitive control and emotional regulation.33,34 However, studies of MDD in adults have typically found abnormally low levels of activation in these regions, a finding that was reversed in the present meta-analysis of youth with MDD. This age difference in activation suggests a possible biomarker or compensatory mechanism that changes during development and/or across the course of the disorder itself.

We also found reliable overactivity centered at the parahippocampal gyrus during affective processing tasks and in the STC during negative valence tasks in youth with MDD. These regions along with the sgACC and ventrolateral prefrontal cortex have been identified as components of the default mode network, a macroscale network that appears to subserve self-referential processing, particularly during wakeful rest. In the context of MDD, overactivation of these regions has been associated with maladaptive rumination and the inability to disengage in self-reflective thinking, particularly during affective processing and negatively valence conditions,29,34,35 which is consistent with observed hyperactivation in these same conditions in the present meta-analysis.

In contrast, we found reliable underactivity centered on the dorsal cingulate cortex, dorsal anterior insula, and cuneus during executive functioning tasks, the posterior insula during positive valence tasks, and the caudate across all tasks combined in youth with MDD. These regions have been identified as major nodes of the task-positive network, another macroscale network typically anticorrelated with the default mode network. Activation of these regions appears to facilitate executive functioning and attentional shifting; in the context of MDD, underactivation in these regions is associated with difficulty transitioning from maladaptive rumination to adaptive processing.34,3638

However, each task-positive network region has also been implicated in other types of cognitive processing related to deficits in MDD. In particular, the cuneus/precuneus has been linked to reward processing and underactivation in this region has been associated with anhedonia,39,40 one of the cardinal symptoms of MDD. In addition, dorsal cingulate cortex activation has been linked to the retrieval of autobiographical memories in healthy individuals41 and negative memory bias in MDD42; activation of the caudate has been related to deficits in anticipated incentives and behavioral inhibition in adolescents with anxiety disorders.43 Finally, activation of the insula has been associated with emotional salience and arousal as well as interoceptive awareness.29,44 Examination of its functional subdivisions indicates the importance of the dorsal anterior region in executive control and the posterior region in the processing of painful or unpleasant stimuli,45 findings that are congruent with the low arousal deficiencies in cognitive control and the persistence of negative affect in patients with MDD.

Analysis of Whole-Brain Activation

Using the Neurosynth decoding framework, we found that youth with MDD compared with healthy control participants showed abnormal whole-brain activation patterns associated with altered cognitive processing. In particular, depressed youth showed brain activation patterns linked to increased engagement with negative emotions that persisted across conditions and during the presentation of positive stimuli. They also showed brain activation patterns associated with corresponding deficits in executive functioning, particularly in cognitive control, attentional shifting, and error monitoring. Importantly, these results are consistent with cognitive biases documented in depression46 but were independently derived from quantitative analysis of whole-brain activation patterns.

Neural Basis of MDD Symptoms in Youth

These findings show consistent anomalies in several distributed brain networks in youth diagnosed as having MDD (Figure 2) that might help explain the expression of specific symptoms of depression in this population. First, the hypervigilance to emotional stimuli exhibited by youth with MDD may be owing to overactivity in the sgACC and thalamus, regions that serve as central hubs of emotional salience. Furthermore, the altered functional connectivity of the sgACC with other brain regions in depressed youth may result in the cascade of depressive symptoms described later. Second, the ineffective attempts at emotional regulation that have been documented in depressed youth may result from their elevated recruitment of the dlPFC and ventrolateral prefrontal cortex, which, nevertheless, ultimately fail to modulate responses in emotional circuitry. This pattern of activation may reverse with increasing age and/or the course of depression as functional connectivity with these regions is down-regulated, a possibility that should be examined in future longitudinal studies. Third, the rumination consistently associated with depression may be linked to maladaptive regulation of default-mode and task-positive network activity. In particular, excessive self-referential processing supported by the activation of default mode network structures, such as the parahippocampal gyrus and STC, appears to persist in depressed youth despite competing attentional demands that would normally disengage these regions and recruit task-positive network structures, such as the cuneus and caudate. Finally, the combination of diminished cuneus activation and posterior insula response to positive stimuli and the heightened activation in the dlPFC and STC during the processing of negative stimuli likely contributes to anhedonic symptoms and the maintenance of negative affect.

Limitations

To our knowledge, this meta-analysis is the first to provide a quantitative synthesis of neuroimaging studies of MDD in youth and yielded several statistically robust and theoretically meaningful neural abnormalities. Nevertheless, we should note 3 limitations of this study. First, the number of primary studies (n = 14) included in this meta-analysis meant that we had relatively limited statistical power for detecting brain regions with abnormal activations. However, the total number of participants with MDD (n = 257) and healthy control participants (n = 274) was large and yielded several findings that satisfied highly stringent requirements for statistical significance (P < .001, familywise error rate corrected), demonstrating robust differences even with this relatively small number of primary studies. Second, the primary studies included in this meta-analysis did not use a standardized set of task conditions nor did they specifically examine the effects of participant characteristics, such as sex, pubertal status, or medication; they also included a relatively wide range of participant ages. While this increased the generalizability and clinical relevance of our results, it is possible that differences in these variables were related systematically to the observed effects in neural activity and we encourage investigators to examine these potentially important variables in future studies. Third, because we used the standard approach of other coordinate-based meta-analyses in which many of the primary studies did not report inferential statistics or cluster sizes for their obtained clusters, we could not compute effect sizes. We assigned a radius of 10 mm for each reported cluster in the initial indicator maps. This value has been used in other coordinate-based meta-analyses and is considered a suitable standard value for the desired sensitivity and spatial resolution of fMRI.29,47

Conclusions

Although MDD is defined diagnostically as a clinical syndrome of co-occurring symptoms, clinical researchers are moving toward a biologically driven dimensional system based on neural models of psychological dysfunction48,49 capable of linking together seemingly disparate symptoms to a coherent neural explanation. Accordingly, in this meta-analysis, we identified several brain regions in which youth with MDD reliably exhibited abnormal activations compared with age-matched healthy control participants during affective processing and executive functioning tasks. Based on these findings and the existing neuroscience literature, we propose that altered activation of several distributed brain networks described previously may help explain seemingly disparate symptoms in this population as well as develop targeted interventions for prevention and treatment.

Back to top
Article Information

Corresponding Author: Chris H. Miller, BS, Stanford Neurosciences Institute, Department of Psychology, Stanford University, Jordan Hall, Bldg 420, Room 163, Stanford, CA 94305-2130 (chmiller@stanford.edu).

Submitted for Publication: April 2, 2015; final revision received June 10, 2015; accepted June 14, 2015.

Published Online: September 2, 2015. doi:10.1001/jamapsychiatry.2015.1376.

Author Contributions: Mr Miller had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Miller, Sacchet, Gotlib.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: All authors.

Obtained funding: Miller, Gotlib.

Administrative, technical, or material support: Miller, Gotlib.

Study supervision: All authors.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work is supported by a Stanford University Graduate Fellowship and National Science Foundation Graduate Research Fellowship DGE-1147470 (Mr Miller) and grants MH74849 and MH101545 from the National Institute of Mental Health (Dr Gotlib).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Additional Contributions: We thank Samuel M. McClure, PhD, Arizona State University; Luke J. Chang, PhD, University of Colorado; Steven Kasparek, BA, Middlebury College; Chandana Karipineni, MS, Foothill College; Natalie L. Colich, BA, Stanford University; and Meghan Goyer, BA, Stanford University for their contributions to this project. They did not receive financial compensation for their contributions.

References
1.
Birmaher  B, Brent  D, Bernet  W,  et al; AACAP Work Group on Quality Issues.  Practice parameter for the assessment and treatment of children and adolescents with depressive disorders. J Am Acad Child Adolesc Psychiatry. 2007;46(11):1503-1526.
PubMedArticle
2.
Lynch  FL, Clarke  GN.  Estimating the economic burden of depression in children and adolescents. Am J Prev Med. 2006;31(6)(suppl 1):S143-S151.
PubMedArticle
3.
Kerestes  R, Davey  CG, Stephanou  K, Whittle  S, Harrison  BJ.  Functional brain imaging studies of youth depression: a systematic review. Neuroimage Clin. 2014;4:209-231.
PubMedArticle
4.
Kessler  RC, Walters  EE.  Epidemiology of DSM-III-R major depression and minor depression among adolescents and young adults in the National Comorbidity Survey. Depress Anxiety. 1998;7(1):3-14.
PubMedArticle
5.
March  J, Silva  S, Petrycki  S,  et al; Treatment for Adolescents With Depression Study (TADS) Team.  Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: Treatment for Adolescents With Depression Study (TADS) randomized controlled trial. JAMA. 2004;292(7):807-820.
PubMedArticle
6.
Graham  J, Salimi-Khorshidi  G, Hagan  C,  et al.  Meta-analytic evidence for neuroimaging models of depression: state or trait? J Affect Disord. 2013;151(2):423-431.
PubMedArticle
7.
Chen  JJ, Liu  Z, Zhu  D,  et al.  Bilateral vs. unilateral repetitive transcranial magnetic stimulation in treating major depression: a meta-analysis of randomized controlled trials. Psychiatry Res. 2014;219(1):51-57.
PubMedArticle
8.
Mayberg  HS, Lozano  AM, Voon  V,  et al.  Deep brain stimulation for treatment-resistant depression. Neuron. 2005;45(5):651-660.
PubMedArticle
9.
Hulvershorn  LA, Cullen  K, Anand  A.  Toward dysfunctional connectivity: a review of neuroimaging findings in pediatric major depressive disorder. Brain Imaging Behav. 2011;5(4):307-328.
PubMedArticle
10.
Davey  CG, Allen  NB, Harrison  BJ, Yücel  M.  Increased amygdala response to positive social feedback in young people with major depressive disorder. Biol Psychiatry. 2011;69(8):734-741.
PubMedArticle
11.
Wager  TD, Lindquist  M, Kaplan  L.  Meta-analysis of functional neuroimaging data: current and future directions. Soc Cogn Affect Neurosci. 2007;2(2):150-158.
PubMedArticle
12.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 2000.
13.
Chantiluke  K, Halari  R, Simic  M,  et al.  Fronto-striato-cerebellar dysregulation in adolescents with depression during motivated attention. Biol Psychiatry. 2012;71(1):59-67.
PubMedArticle
14.
Colich  NL, Foland-Ross  LC, Eggleston  C, Singh  MK, Gotlib  IH.  Neural aspects of inhibition following emotional primes in depressed adolescents. J Clin Child Adolesc Psychol. 2015;0(0):1-10.
PubMedArticle
15.
Diler  RS, de Almeida  JRC, Ladouceur  C, Birmaher  B, Axelson  D, Phillips  M.  Neural activity to intense positive versus negative stimuli can help differentiate bipolar disorder from unipolar major depressive disorder in depressed adolescents: a pilot fMRI study. Psychiatry Res. 2013;214(3):277-284.
PubMedArticle
16.
Diler  RS, Pan  LA, Segreti  A,  et al.  Differential anterior cingulate activity during response inhibition in depressed adolescents with bipolar and unipolar major depressive disorder. J Can Acad Child Adolesc Psychiatry. 2014;23(1):10-19.
PubMed
17.
Gaffrey  MS, Barch  DM, Singer  J, Shenoy  R, Luby  JL.  Disrupted amygdala reactivity in depressed 4- to 6-year-old children. J Am Acad Child Adolesc Psychiatry. 2013;52(7):737-746.
PubMedArticle
18.
Halari  R, Simic  M, Pariante  CM,  et al.  Reduced activation in lateral prefrontal cortex and anterior cingulate during attention and cognitive control functions in medication-naïve adolescents with depression compared to controls. J Child Psychol Psychiatry. 2009;50(3):307-316.
PubMedArticle
19.
Hall  LMJ, Klimes-Dougan  B, Hunt  RH,  et al.  An fMRI study of emotional face processing in adolescent major depression. J Affect Disord. 2014;168:44-50.
PubMedArticle
20.
Roberson-Nay  R, McClure  EB, Monk  CS,  et al.  Increased amygdala activity during successful memory encoding in adolescent major depressive disorder: an FMRI study. Biol Psychiatry. 2006;60(9):966-973.
PubMedArticle
21.
Sharp  C, Kim  S, Herman  L, Pane  H, Reuter  T, Strathearn  L.  Major depression in mothers predicts reduced ventral striatum activation in adolescent female offspring with and without depression. J Abnorm Psychol. 2014;123(2):298-309.
PubMedArticle
22.
Tao  R, Calley  CS, Hart  J,  et al.  Brain activity in adolescent major depressive disorder before and after fluoxetine treatment. Am J Psychiatry. 2012;169(4):381-388.
PubMedArticle
23.
Yang  TT, Simmons  AN, Matthews  SC,  et al.  Depressed adolescents demonstrate greater subgenual anterior cingulate activity. Neuroreport. 2009;20(4):440-444.
PubMedArticle
24.
Yang  TT, Simmons  AN, Matthews  SC,  et al.  Adolescents with major depression demonstrate increased amygdala activation. J Am Acad Child Adolesc Psychiatry. 2010;49(1):42-51.
PubMed
25.
Mingtian  Z, Shuqiao  Y, Xiongzhao  Z,  et al.  Elevated amygdala activity to negative faces in young adults with early onset major depressive disorder. Psychiatry Res. 2012;201(2):107-112.
PubMedArticle
26.
Yarkoni  T, Poldrack  RA, Nichols  TE, Van Essen  DC, Wager  TD.  Large-scale automated synthesis of human functional neuroimaging data. Nat Methods. 2011;8(8):665-670.
PubMedArticle
27.
Poldrack  RA.  Inferring mental states from neuroimaging data: from reverse inference to large-scale decoding. Neuron. 2011;72(5):692-697.
PubMedArticle
28.
Poldrack  RA.  Can cognitive processes be inferred from neuroimaging data? Trends Cogn Sci. 2006;10(2):59-63.
PubMedArticle
29.
Hamilton  JP, Etkin  A, Furman  DJ, Lemus  MG, Johnson  RF, Gotlib  IH.  Functional neuroimaging of major depressive disorder: a meta-analysis and new integration of base line activation and neural response data. Am J Psychiatry. 2012;169(7):693-703.
PubMedArticle
30.
Sundermann  B, Olde Lütke Beverborg  M, Pfleiderer  B.  Toward literature-based feature selection for diagnostic classification: a meta-analysis of resting-state fMRI in depression. Front Hum Neurosci. 2014;8:692.
PubMedArticle
31.
Price  JL, Drevets  WC.  Neural circuits underlying the pathophysiology of mood disorders. Trends Cogn Sci. 2012;16(1):61-71.
PubMedArticle
32.
Connolly  CG, Wu  J, Ho  TC,  et al.  Resting-state functional connectivity of subgenual anterior cingulate cortex in depressed adolescents. Biol Psychiatry. 2013;74(12):898-907.
PubMedArticle
33.
Drevets  WC.  Neuroimaging studies of mood disorders. Biol Psychiatry. 2000;48(8):813-829.
PubMedArticle
34.
Pizzagalli  DA.  Frontocingulate dysfunction in depression: toward biomarkers of treatment response. Neuropsychopharmacology. 2011;36(1):183-206.
PubMedArticle
35.
Cooney  RE, Joormann  J, Eugène  F, Dennis  EL, Gotlib  IH.  Neural correlates of rumination in depression. Cogn Affect Behav Neurosci. 2010;10(4):470-478.
PubMedArticle
36.
Utevsky  AV, Smith  DV, Huettel  SA.  Precuneus is a functional core of the default-mode network. J Neurosci. 2014;34(3):932-940.
PubMedArticle
37.
Menon  V, Uddin  LQ.  Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct. 2010;214(5-6):655-667.
PubMedArticle
38.
Leech  R, Sharp  DJ.  The role of the posterior cingulate cortex in cognition and disease. Brain. 2014;137(pt 1):12-32.
PubMedArticle
39.
Harvey  PO, Pruessner  J, Czechowska  Y, Lepage  M.  Individual differences in trait anhedonia: a structural and functional magnetic resonance imaging study in non-clinical subjects. Mol Psychiatry. 2007;12(8):703, 767-775.
PubMedArticle
40.
Vaidya  JG, Knutson  B, O’Leary  DS, Block  RI, Magnotta  V.  Neural sensitivity to absolute and relative anticipated reward in adolescents. PLoS One. 2013;8(3):e58708.
PubMedArticle
41.
Maddock  RJ, Garrett  AS, Buonocore  MH.  Remembering familiar people: the posterior cingulate cortex and autobiographical memory retrieval. Neuroscience. 2001;104(3):667-676.
PubMedArticle
42.
Foland-Ross  LC, Hamilton  P, Sacchet  MD, Furman  DJ, Sherdell  L, Gotlib  IH.  Activation of the medial prefrontal and posterior cingulate cortex during encoding of negative material predicts symptom worsening in major depression. Neuroreport. 2014;25(5):324-329.
PubMed
43.
Guyer  AE, Choate  VR, Detloff  A,  et al.  Striatal functional alteration during incentive anticipation in pediatric anxiety disorders. Am J Psychiatry. 2012;169(2):205-212.
PubMedArticle
44.
Craig  AD.  How do you feel--now? the anterior insula and human awareness. Nat Rev Neurosci. 2009;10(1):59-70.
PubMedArticle
45.
Chang  LJ, Yarkoni  T, Khaw  MW, Sanfey  AG.  Decoding the role of the insula in human cognition: functional parcellation and large-scale reverse inference. Cereb Cortex. 2013;23(3):739-749.
PubMedArticle
46.
Gotlib  IH, Joormann  J.  Cognition and depression: current status and future directions. Annu Rev Clin Psychol. 2010;6:285-312.
PubMedArticle
47.
Etkin  A, Wager  TD.  Functional neuroimaging of anxiety: a meta-analysis of emotional processing in PTSD, social anxiety disorder, and specific phobia. Am J Psychiatry. 2007;164(10):1476-1488.
PubMedArticle
48.
Insel  T, Cuthbert  B, Garvey  M,  et al.  Research domain criteria (RDoC): toward a new classification framework for research on mental disorders. Am J Psychiatry. 2010;167(7):748-751.
PubMedArticle
49.
Watson  D.  Rethinking the mood and anxiety disorders: a quantitative hierarchical model for DSM-V. J Abnorm Psychol. 2005;114(4):522-536.
PubMedArticle
×