Global hippocampal atrophy in major depressive disorder : a meta-analysis of magnetic resonance imaging studies

INTRODUCTION
Major depressive disorder (MDD), an incapacitating mental disorder, is characterized by episodes of at least 2 weeks of apparent changes in mood, cognition, and neurovegetative functions. Many neuroimaging studies using magnetic resonance imaging (MRI) have examined morphometric changes in patients with MDD, but the results are not conclusive. This study aims to review the literature and perform a meta-analysis on hippocampal volume (HcV) in patients with MDD.


METHODS
Studies on HcV in patients with MDD diagnosis were identified from major databases (MEDLINE, EMBASE, The Cochrane Library, Scopus, PsycINFO, and SciELO) using the search terms depression, major depressive disorder, MDD, unipolar, magnetic resonance imaging, MRI, and hippocampus.


RESULTS
A meta-analysis of 29 studies fulfilling specific criteria was performed. The sample included 1327 patients and 1004 healthy participants. The studies were highly heterogeneous with respect to age, sex, age of onset, and average illness duration. However, the pooled effect size of depression was significant in both hippocampi. MDD was associated with right (-0.43; 95% confidence interval [95%CI] -0.66 to -0.21) and left (-0.40; 95%CI -0.66 to -0.15) hippocampal atrophy.


CONCLUSIONS
MDD seems to be associated with global HcV atrophy. Larger longitudinal follow-up studies designed to analyze the influence of sociodemographic variables on this relationship are required to yield better evidence about this topic.


Introduction
Major depressive disorder (MDD) is an incapacitating mental disorder characterized by episodes of at least 2 weeks of apparent changes in affection, cognition, and neurovegetative functions. 1Patients with MDD present a lower quality of life and higher prevalence of medical conditions.The global prevalence of MDD is approximately 5%.According to the World Health Organization, MDD is estimated to become the second most disabling condition by 2020. 2 Several neurobiological models have been proposed to explain the pathogenesis of MDD.Because MDD is primarily related to lower affection and humor, many authors have studied and emphasized the role of dysfunctional cortico-limbic networks in MDD. 3 Postmortem and animal model studies have reported lower hippocampal volume (HcV) in participants with depressive disorders.The hippocampus is involved in episodic and declarative memory, as well as in learning, areas that often present deficits in patients with depression. 4 The suggestion that HcV is lower because of depression has been influential in guiding neuroimaging studies on the analysis of the hippocampus. 5ny neuroimaging studies using magnetic resonance imaging (MRI) have examined morphometric changes in patients with MDD, but the results are not conclusive.Some studies have reported bilateral, 6 unilateral right, 7 or unilateral left hippocampal hypotrophy in these patients compared with healthy participants, 8,9 whereas others have reported no changes. 10,113][14] However, to date, no reviews have systematically examined the macro influence of depression on HcV.
This study aims to examine whether patients with MDD present hippocampal atrophy compared with nondepressive participants.We considered HcV reduction as the primary outcome.Two authors examined the abstracts of the articles retrieved against the defined inclusion criteria.All potentially relevant full-text articles were retrieved for quality and satisfaction assessment of inclusion criteria.

Methods
Figure 1 summarizes the study inclusion process.

For each data criterion, a data extraction table
was pilot-tested on five randomly-selected studies and adjusted accordingly.Two independent reviewers assessed the following data from the patient and control samples: number of participants, HcV means and standard deviations, and mean disease duration.
A third author double-checked the extracted data.
Disagreements were resolved by consensus between the authors; if no consensus could be reached, a third author made a decision.All volumes were converted to mm 3 before being entered into the meta-analysis.
To certify the quality and validity of the eligible studies, pairs of reviewers working independently assessed each study using the Strengthening The

Reporting of Observational Studies in Epidemiology
(STROBE) checklist. 16lculations were performed using the STATA software version 13.0 SE.The meta-analyses were performed using a random effects model weighting the studies by the inverse variance and calculating the Hedges' g effect size.The random effects model using the DerSimonian-Laird method was selected over the fixed effects approach because previous analyses have indicated considerable between-study heterogeneity.
To assess for between-study heterogeneity, the Cochran Q test was performed and I² statistic was recorded and further analyzed by meta-regression.
Begg's and Egger's tests were used to determine publication bias.

Results
[12][13][14] One study that found lower left HcV was excluded because volume measurements were unavailable. 38A report by Andreescu et al. 39 was also excluded because only global HcV measurements were available.Pantel et al. 40 reported a study in German that was excluded from the analysis because of language limitations.
All studies selected for this review were crosssectional studies published in English, except one cohort study. 30The main inclusion criteria entailed adults (≥18 years).The Hamilton Depression Scale was the main two of these measurements markedly deviated from the mean. 9,14Sample characteristics of the included studies are presented in Table 1.Some of the studies included patients with recurrent depression, 6 first-episode depression, 12,17,26 and late-onset depression. 7,13,21Eight studies evaluated illness duration, which varied from 7 to 57 months. 6,12,18,25,27,28,33,35Table 2 shows a summary of the included studies.
Because the study performed by MacQueen et al. 26 analyzed two samples (first-episode versus multiple episodes of depression) that did not overlap, it was considered as having two distinct samples for statistical calculation.Therefore, the degrees of freedom (df), which is typically the number of studies minus 1, was 29 instead of 28.
The meta-analysis was repeated, omitting one study at a time, to ensure that the result was not skewed by a single study.This analysis did not change the random-effects estimate, and the results continued to be statistically significant.
Because the studies had significant heterogeneity, the data were analyzed using meta-regression.We assumed that differences in age, sex, duration of illness, and year of publication could explain some of the variation.These variables were analyzed separately and together, but were not significantly correlated with the random-effects estimate in either the right or left hippocampus.

Discussion
The studies included in the analysis yielded highly heterogeneous results.Notably, this heterogeneity was expected because there were marked differences among the patient groups with respect to age and sex distribution, type of MDD (e.g., first episode, dysthymia, refractory, late onset), illness duration, age at the first episode, and factors related to treatment.Moreover, an increase in HcV variation was expected considering various protocols for the scanning and delineation of hippocampal structures.
However, a meta-analysis plays an important role in the analysis of scientific evidence because these possible confounding factors are diluted or neutralize  each other based on the large number of participants analyzed.Meta-analyses are conducted to determine whether an effect is present, and summarize the data to determine if this effect is positive or negative.This process increases the external validity of the studies as well as the extendibility of results to the general population of patients with MDD.
It is established that HcV losses are expected as a natural process of aging 41 ; however, Sheline et al. 42 demonstrated that illness duration, not age, predicted hippocampal loss in women with recurrent MDD.
Only eight studies included in our analysis presented information about depression duration.However, in the meta-regression, neither illness duration nor age were significantly correlated with the random-effects estimate.
Seven studies failed to find significant HcV differences in patients with depression compared with healthy participants. 11,18,19,28,29,32,33Six studies demonstrated significant reduction only in the right hippocampus. 7,20,21,30,31,35Six other studies showed similar findings in the lower left hippocampus. 8,9,12,13,36,37][25][26][27]34 MacQueen et al. 26   According to the results of Vakili et al., 32 a smaller volume of the right hippocampus was associated with poor responses to antidepressants.This result is incipient, as investigating this association was not the primary objective of the study.However, if confirmed, that finding would be clinically interesting as a potential predictor of treatment response.

The pathophysiological pathways that explain
HcV reduction in MDD remain unclear.Some authors theorize that HcV reduction is associated with disturbed hypothalamic pituitary adrenal axis function and adrenal hypersecretion of glucocorticoids, particularly cortisol.According to this hypothesis, cortisol leads to neural atrophy and the inhibition of neurogenesis in the hippocampus. 43Another important supporting mechanism is the glutamate N-methyl-D-aspartate (NMDA) channel present in inhibitory neurons that comprise this pathway.Subiculum neurons make a synapse with a hypothalamic neuron, inhibiting it through NMDA receptors.The hypothalamic neuron modulates the corticotropic cell, stimulating it through gamma-aminobutyric acid (GAMA) liberation, which culminates in the production and release of the adrenocorticotropic hormone (ACTH).[46] It is important to address the potential protective properties of selective serotonin reuptake inhibitors (SSRIs) in patients with MDD.As demonstrated by Frodl et al., 19 after a follow-up of 3 years, patients with MDD who used SSRI antidepressants showed protective effects against hippocampal atrophy as well as an increase in the left HcV.Although this is the only study to demonstrate this association, many studies have indicated that SSRIs reduce functional deficits in inflammatory and ischemic events.

Study limitations
In principle, cross-sectional studies such as those included in the present analysis do not allow conclusions about causality to be drawn.
Further, socioeconomic and lifestyle characteristics from specific population groups can act as confounders.
For example, the habit of regular exercise training is a protective factor against hippocampal atrophy and may be responsible for hippocampus hypertrophy in healthy participants. 48her factors can also act as confounders, such as the comorbidity of MDD and anxiety disorders, 49 which was not assessed in any of the studies.
Longitudinal follow-up studies with large samples are the best design to allow the drawing of conclusions about the association between hippocampal atrophy and MDD.

Conclusions
Although the studies available in the literature are quite heterogeneous, MDD seems to be associated This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement.Details of the study protocol were registered on the International Prospective Register of Systematic Reviews (PROSPERO) and can be accessed at www.crd.york.ac.uk/PROSPERO/ display_record.asp?ID=CRD42018086196.The electronic databases MEDLINE, EMBASE, The Cochrane Library, Scopus, PsycINFO, and SciELO were searched for papers published between January 1960 and October 2017.The search terminology included the terms depression, major depressive disorder, MDD, unipolar, magnetic resonance imaging, MRI, and hippocampus.Only studies written in English, Portuguese, or Spanish were reviewed.At least two of the authors performed each search.Furthermore, all reference lists of the obtained papers were checked for studies not indexed in electronic databases.The complete search strategy was as follows: Depression; Major depressive disorder; MAJOR DEPRESSIVE DISORDER (MeSH); MDD; (unipolar) AND (depression); Magnetic resonance imaging; MAGNETIC RESONANCE IMAGING (MeSH); MRI; Hippocampus; HIPPOCAMPUS (MeSH); (hippocampal) AND (volume); (1 OR 2 OR 3 OR 4 OR 5) AND (6 OR 7 OR 8) AND (9 OR 10 OR 11).All observational studies and clinical trials that evaluated the relationship between MDD and HcV measured through MRI were included in this review.Studies with sample groups in which patients presented any other associated neuropsychiatric or metabolic condition were excluded.Inclusion criteria were as follows: patients with a primary diagnosis of MDD assessed using international diagnostic criteria (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [DSM-IV] or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10]); a healthy comparison group; participants screened for neurological, psychiatric, and other medical disorders that could affect brain structure, including alcohol and substance abuse; MRI as the primary measurement tool; and a continuous measure of HcV as the dependent variable.

Figure 1 -
Figure1-PRISMA flow diagram for the meta-analysis of hippocampal volume in major depressive disorder.Adapted from Moher et al.15

[
95%CI] 77.8 to 88.1, p < 0.00001) for the right hippocampus and 88% (95%CI 83.4 to 90.7, p < 0.00001) for the left hippocampus, which provides high evidence of between-study heterogeneity.Therefore, the effect size was calculated under the assumption of a random effects model.The DerSimonian-Laird pooled effect size revealed bilateral statistical significance: = Beck Depression Inventory II; CDRS = Childhood Depression Rating Scale; CESD = Center for Epidemiologic Studies Depression scale; DDES = Duke Depression Evaluation Schedule; DF = drug-free; DR = drug-resistant; EOS = early-onset depression; FE = first episode; GCI = Global Clinical Impression Scale; GD = geriatric depression; GDS = Geriatric Depression Scale; HADS = Hospital Anxiety and Depression Scale; HAM-D = Hamilton Depression Scale; LOS = late-onset depression; MADRS = Montgomery-Åsberg Depression Rating Scale; MDD = major depressive disorder; MRI = magnetic resonance imaging; RMD = recurrent major depression; ZDS = Zung Depression Scale.* Left hippocampal volume increased after a 3-year follow-up in use of antidepressants.

Figure 2 -
Figure 2 -Standardized mean difference of right hippocampal volume in patients with depression relative to comparison subjects from a meta-analysis of 29 magnetic resonance imaging studies.95%CI = 95% confidence interval; Std = standard.

Figure 3 -
Figure 3 -Standardized mean difference of left hippocampal volume in patients with depression relative to comparison subjects from a meta-analysis of 29 magnetic resonance imaging studies.95%CI = 95% confidence interval; Std = standard.
with global HcV atrophy.Many confounding factors may have influenced the divergence between studies, in particular sociocultural variables, which are responsible for the construction of social identity and affect the way in which stressful and depressive situations may contribute to changes in hippocampal neuronal circuits.Nevertheless, idiosyncratic biological factors may influence neuronal circuitry development and plasticity and play a role in the fine adjustment of hippocampal dynamics.Larger longitudinal follow-up studies that consider these aspects are needed to yield better evidence about this topic.

Figure 4 -
Figure 4 -A, B) Begg's funnel plots, and C, D) Egger's funnel plots showing no obvious signs of publication bias for the meta-analysis results.

Table 1 -
Studies of hippocampal volume in patients with major depressive disorder

Table 2 -
Summary of included studies evaluating hippocampal atrophy in MDD