Journal of Psychiatric Research 43 (2009) 1112–1117

Contents lists available at ScienceDirect

Journal of Psychiatric Research

journal homepage: www.elsevier .com/locate / jpsychires

Reduced amygdala volume in newly admitted psychiatricin-patients with unipolar major depression

Golo Kronenberg a, Ludger Tebartz van Elst b, Francesca Regen a, Michael Deuschle c,Isabella Heuser a, Michael Colla a,*

a Department of Psychiatry, Charité – University Medicine Berlin, Campus Benjamin Franklin, Eschenallee 3, 14050 Berlin, Germanyb Section for Experimental Neuropsychiatry, Freiburg Brain Imaging and South German Brain Imaging Center and Department of Psychiatry and Psychotherapy, Medical School,Albert-Ludwigs-University Freiburg, Hauptstrasse 5, 79104 Freiburg, Germanyc Central Institute of Mental Health, J5, 68159 Mannheim, Germany

a r t i c l e i n f o

Article history:Received 19 July 2008Received in revised form 16 March 2009Accepted 18 March 2009

Keywords:DepressionAmygdalaQuantitative MRIPlasticityCortisolAntidepressant

0022-3956/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.jpsychires.2009.03.007

* Corresponding author. Tel.: +49 30 8445 8704; faE-mail address: michael.colla@charite.de (M. Colla

a b s t r a c t

Structural neuroimaging studies investigating amygdala volumes in patients suffering from majordepression have yielded variable results. Discrepant findings across studies may be attributable in partto heterogeneity with respect to antidepressant medication and to lack of adequate control for the effectsof total brain volume and age. Here, 24 unipolar depressed in-patients newly admitted to a psychiatricunit and 14 healthy control participants matched for age, gender, and years of education underwentquantitative magnetic resonance imaging (MRI) toward the end of a one-week washout period. Salivacortisol was measured at 08.00 and at 16.00 h in patients during washout. Absolute amygdala volumeswere significantly reduced in the patient group (by 13% in left amygdala and 12% in right amygdala).The effect of reduced amygdala volumes in patients remained significant after correction for brain vol-ume (BV) and age. Furthermore, amygdala volume measurements in the patient sample showed a signif-icant inverse relationship to the number of preceding depressive episodes. In patients, severity of disease(baseline HAMD scores) and baseline cortisol levels were not related to amygdala volume. This study of asample of unmedicated depressed in-patients adds to the small, yet growing, body of evidence linkinguntreated major depression to reduced amygdala volume.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Located deep within the anterior inferior temporal lobes, theamygdalar nuclei constitute a unique part of the telencephalon.The amygdaloid complex receives projections from and returnsprojections to most cortical fields (e.g. Murray, 2007). Amygdalarprojections to the nucleus accumbens and ventral striatum as wellas lateral hypothalamus and brain stem structures like locus coeru-leus, among others, attest to the important role of the amygdala inthe elaboration of central aspects of emotion (e.g. Davis and Wha-len, 2001). Not surprisingly, this key limbic structure has beenimplicated in the etiopathogenesis of mood disorders. Studiesusing functional neuroimaging have reported increased amygdalarblood flow (Drevets et al., 1992) and glucose metabolism (Drevetset al., 2002) in unipolar major depression. Moreover, an associationbetween amygdala hyperactivity and negatively biased automaticemotion processing has been observed in major depression (Dann-lowski et al., 2007). Also, functional magnetic resonance imaginghas revealed greater functional connectivity of the right amygdala

ll rights reserved.

x: +49 30 8445 8365.).

with hippocampus and caudate-putamen in depressed individualsthan in control participants during encoding of subsequentlyremembered negative but not neutral or positive stimuli (Hamiltonand Gotlib, 2008).

Experimental evidence indicates that the amygdala is a highlyplastic brain structure in which new cells continue to be generatedin adulthood (e.g. Carrillo et al., 2007; Keilhoff et al., 2006). Whileprenatal stress is associated with a reduced density of proliferatingcells in the amygdala in the developing brain (Kawamura et al.,2006), electroconvulsive seizures have been shown to stimulateproliferation of glial progenitor cells in the adult rat amygdala(Wennström et al., 2004). In recombinant inbred strains of mice,natural variation in amygdala volume is associated with differ-ences in stress reactivity and fear learning (Yang et al., 2008).

Neuropathological studies of human postmortem brain speci-mens have yielded clear evidence for a decrease in glia in theamygdala of individuals suffering from major depression (Bowleyet al., 2002; Hamidi et al., 2004). However, so far, structural neuro-imaging studies of depressed patients have yielded conflicting re-sults. Possible confounding factors include methodical issues likeneuroanatomical definitions used and, most importantly, differ-ences in clinical variables such as age, severity of depression or

G. Kronenberg et al. / Journal of Psychiatric Research 43 (2009) 1112–1117 1113

use of antidepressant medication across different studies. A recentmeta-analysis of amygdala volume in major depressive disorderconcluded that although there is no aggregate-level difference be-tween depressed and nondepressed individuals in amygdala vol-ume, there is significant inter-study variability in the differencebetween depressed and nondepressed groups. Importantly, thegreater the proportion of medicated individuals in a depressedsample, the larger the difference between depressed and nonde-pressed participants in amygdala volume (Hamilton et al., 2008).Based on these findings, the authors speculated that abnormallylow amygdala volume may be a symptom or a consequence ofdepressive pathology (Hamilton et al., 2008).

In this investigation, we sought to examine right and left amyg-dala volumes in a sample of relatively severely depressed patientsin mid- and later life who had required hospital treatment for theircondition. Patients were unmedicated at the time of MRI. Further-more, we analyzed interrelationships between amygdala volumes,saliva cortisol concentrations and key clinical factors.

Fig. 1. (A) Example of single 3-D MRI slice showing delineation of in-slice volume ofamygdala and definition of boundaries according to Watson et al. (1992). Fordetailed description of amygdala volume measurements, see Section 2. (B)Comparison of absolute amygdala volume measurements between depressedpatients and matched controls.

2. Methods

2.1. Patients and controls

Data presented here were collected as part of a larger projectaimed at defining both cross-sectional and longitudinal brain vol-umetric changes associated with major depression (Colla et al.,2007). This is the second paper to be published from this work.In a previous manuscript, hippocampal volumes gained from thesample also used here were investigated with the same explorativetype of approach and using the same co-variables (Colla et al.,2007).

Inpatients suffering from unipolar depression who met DSM-IVdiagnostic criteria for Major Depressive Disorder with a minimumscore of 18 points on the Hamilton Rating Scale for Depression(HAMD, 21 items) were recruited for participation in this study(Day �8). Patients’ psychiatric family history, age at onset of illnessand length of current episode (index episode) were recorded. IQwas estimated with the Multiple Choice Word Fluency Test(MWT-B; Merz et al., 1975). Exclusion criteria included additionalaxis I comorbid psychiatric disorders, any current clinically rele-vant medical condition, history or evidence of stroke or transientischemic attack and alcohol/substance abuse within 6 months be-fore study entry. The control group consisted of healthy volunteersmatched for age, education and IQ estimates. A structured inter-view (SKID-I/II) (Wittchen et al., 1997) was administered to ruleout the presence of current or past psychiatric illness. Furtherexclusion criteria for healthy controls were first-degree relativeswith a psychiatric disorder, any relevant medical disorders as wellas history of alcohol/substance abuse. Neither patients nor controlswho had ever suffered a head injury were included.

Except for on-demand lorazepam and zolpidem patients werekept off psychotropic medication during the first week of the study(week �1). Baseline HAMD scores were assessed at the end of thisantidepressant-free week (Day �1). Patients with a score of lessthan 18 points at the end of this drug-free period were removedfrom the study as responders during washout (n = 3). Patients werethen randomized to treatment with either amitriptyline or paroxe-tine (Day +1). The protocol had been approved by the local ethicscommittee and all participants had given written informedconsent.

2.2. Image acquisition

Magnetic Resonance Imaging (MRI) scans were performed to-ward the end of week �1 when patients were still off psychotropic

medication. Data were acquired on a 1.5 T Magnetom VISIONTM

(Siemens, Erlangen, Germany) equipped with a standard circularlypolarized head coil. A vacuum-moulded head holder (Vac-PacTM,Olympic Medical, Seattle, WA) was employed to reduce motionof the subject’s head. Three-dimensional gradient echo imaging(magnetization prepared rapid gradient echo, MPRAGE) was per-formed in the sagittal plane, T1-weighted (TR = 11.4 ms, TE =4.4 ms, field of view (FOV) = 269 mm, flip angle = 30�, slice thick-ness = 1.05 mm, 154 contiguous slices, pixel: 1.05 � 1.05 mm, slab161 mm, matrix size = 256 � 256).

2.3. Image analysis

Measurements of amygdala volume were performed using theinteractive software program MRreg (Lemieux et al., 2000). Usingthis software the images were zoomed to a magnification of �4.ROIs were outlined manually using a mouse driven cursor. Amyg-dala volumes were measured by manually outlining the bound-aries of each structure separately following the establishedprotocol described by Watson et al. (1992) (Fig. 1), which we havealso employed in a number of previous studies (Tebartz van Elstet al., 1999, 2000, 2002, 2003). The in-slice volume was calculatedby multiplying the number of voxels contained within each traceby the voxel volume and dividing by the magnification factor.The volume of each amygdala was the sum of all in-slice volumes.‘Total’ amygdala volume represents the sum of left and right amyg-dala volumes (Wrase et al., 2008; Makris et al., 2004; Szeszko et al.,1999).

1114 G. Kronenberg et al. / Journal of Psychiatric Research 43 (2009) 1112–1117

Intrarater reliability figures were calculated from repeated mea-surements of a separate subset of 20 normal controls. Raters wereonly allowed to begin with the measurements after proving suffi-cient reliability in this subset of 20 images of healthy volunteersnot otherwise used in this study. The intraclass correlation coeffi-cient (ICC) was 0.91 for the right amygdala and 0.96 for the leftamygdala. Images of patients and controls were presented in a ran-dom sequence to insure that the rater was blinded to the identityand clinical status of the subject.

2.4. Brain volume

Brain volume was measured with an automatic segmentationroutine of the ANALYZE� 4.0. package (Mayo Clinic, Rochester,MN, USA), which uses a seed-growing threshold approach withconsecutive dilations and erosions. Total cerebral gray and whitematter (including brainstem, temporal lobes, the optic chiasm,the pituitary and cerebellum), CSF, dura mater and sinuses were in-cluded. The base of the cerebellum delimited the inferior border(Colla et al., 2007).

2.5. Hormonal measurements

Saliva cortisol measurements were performed as described pre-viously (Weber-Hamann et al., 2006). Briefly, once at baseline andevery other day of the treatment period, saliva for the estimation offree cortisol concentrations (SalivetteTM, Sarstedt, Germany) wascollected from all patients at 08.00 h before breakfast (morningsaliva) and the distribution of medication on the ward. Patientswere uniformly awakened at 07.45 h. Afternoon saliva sampleswere obtained at 16.00 h. A minimum of 15 saliva samples pertimepoint were required for inclusion in further statistical analysis.

Clear saliva was used for duplicate analyses of cortisol using atime-resolved immunoassay with fluorescence detection. The low-er limit of detection was 0.43 nmol/l with interassay- and intraas-say coefficients of variation of less than 10% across the expectedrange of cortisol levels (3–25 nmol/l). Saliva cortisol measurementshave been shown to reflect activity of the hypothalamus-pituitary-adrenal system in depressed patients (Weber-Hamann et al., 2005,2006).

2.6. Statistical analysis

Statistical analysis was performed with StatView for Macintosh,version 5.0.1 (SAS Institute Inc.). Neuroanatomical comparisonsbetween depressed patients and healthy control subjects were car-ried out by means of analysis of variance (ANOVA). The Pearsoncorrelation coefficient with level of significance set at 0.05 andtwo-tailed p values was used to determine the significance of cor-relations among and between MRI measures and clinical data. Nocorrection for multiple testing was performed. Effect size was mea-sured as the partial eta squared ðg2

pÞ. g2p describes the proportion of

the variability in the dependent measure that is attributable to afactor (Kirk, 1982; Tabachnick and Fidell, 1989). For univariate Ftests the formula for the partial eta squared is: g2

p = sums ofsquares for the effect/(sums of squares for the effect + sums ofsquares for the error term). Values are reported as mean ± STD.

In the literature, there are different ways in use to correct rawvolumetric data for differences in brain volume (e.g. van Petten,2004). As indicated in the text below, absolute amygdala volumesrepresent raw data with no correction. Measurements were alsonormalized through division by brain volume (‘‘BV-corrected vol-ume). Although ratios have been used as a normalizing procedurein the past (e.g. Eberling et al., 2004; Pinkhardt et al., 2006; Collaet al., 2007), this approach may cause spurious results. Therefore,volume measurements were also residualized against brain vol-

ume and age as sequential covariates (‘‘residualized volumes”; e.g.Maguire et al., 2003).

3. Results

3.1. Study population

The study population has been described in detail previously(Colla et al., 2007). Briefly, twenty-four patients who required hos-pitalization for an episode of unipolar major depression and four-teen healthy controls matched for age and education wereenrolled into the study. Patients and controls were similar in age(54.5 ± 11.9 vs. 53.8 ± 17.7 years), height (165.5 ± 8,2 vs. 170.9 ±7.1 cm), body weight (74.3 ± 18.4 vs. 81.1 ± 20.2 kg), gender(female/male: 15/9 vs. 8/6), years of education (12.3 ± 2.2 vs.12.6 ± 2.2) and IQ estimates (MWT-B: 106.1 ± 14.5 vs. 111.3 ±16.2). Only 4 of the 24 patients investigated here had received anti-depressant medication before being admitted to the hospital.

3.2. Volumetric data

3.2.1. Reduced absolute amygdala volumes in depressed patientsWhile BV was similar in patients and controls, depressed pa-

tients displayed significantly reduced absolute total as well asabsolute right and absolute left amygdala volumes (Table 1;Fig. 1B). Two-factorial ANOVA (patient versus control status, side)revealed a significant reduction of absolute amygdala volume inpatients as an effect of patient status: F3,72 = 12.1, p < 0.001 (side:F3,72 = 0.06, p = 0.82; interaction: F3,72 = 0.04, p = 0.85).

3.3.1. Reduced amygdala volumes in depressed patients afternormalization for brain volume

Normalization through division by intracranial vault or brainvolume is often used for comparison of volume measurements(‘‘BV-corrected volumes”; e.g. Eberling et al., 2004; Colla et al.,2007). BV-corrected total amygdala volume was significantly re-duced in the patient group (Table 1). Two-factorial ANOVA (patientversus control status, side) also revealed a significant reduction ofabsolute amygdala volume/BV as an effect of patient status:F3,72 = 7.3, p < 0.009 (side: F3,72 = 0.03, p = 0.86; interaction:F3,72 = 0.1, p = 0.74).

3.3.2. Reduced amygdala volumes in depressed patients afterregressing on brain volume and age

As a further correction, we residualized absolute amygdala vol-umes against brain volume and age. Again, left and right amygdalavolumes were significantly reduced in depressed patients (Table1). Similarly, two-way ANOVA on residuals yielded a significant ef-fect of patient status: F3,72 = 8.7, p = 0.004 (side: F3,72 = 0.004,p = 0.95; interaction: F3,72 = 0.06, p = 0.81).

3.3.3. Correlations between amygdala volumes and clinical measuresPatients’ baseline HAMD scores (week �1) were 25.3 ± 5.9. We

did not observe a significant correlation between baseline HAMDscores and total amygdala volume (absolute total amygdala vol-ume: R = �0.2, p = 0.35, BV-corrected total amygdala volume:R = �0.14, p = 0.52; total amygdala volume residualized againstbrain volume and age: R = �0.16, p = 0.46). Similarly, amygdalavolume did not affect HAMD scores (week +4) after treatment(absolute total amygdala volume: R = 0.02, p = 0.93; BV-correctedtotal amygdala volume: R = �0.16, p = 0.46; total amygdala volumeresidualized against brain volume and age: R = �0.07, p = 0.76).Duration of the index episode [weeks] was not related to amygdalavolume (absolute total amygdala volume: R = 0.28, p = 0.18; BV-corrected total amygdala volume: R = 0.02, p = 0.94; total amygdala

Table 1Comparison of amygdala volumes between depressed patients and control subjects.

Volumes (ml) Depressed patients Control subjects F value df p value Effect size partial eta squared

Absolute left amygdala volume 1.71 ± 0.34 1.97 ± 0.26 5.9 36 0.02 0.14Absolute right amygdala volume 1.74 ± 0.27 1.97 ± 0.28 6.2 36 0.02 0.15Absolute total amygdala volume 3.46 ± 0.57 3.94 ± 0.43 7.6 36 <0.01 0.17Brain volume 1057.5 ± 96.1 1096.7 ± 113.5 1.3 36 0.26 0.03Left amygdala (BV-corr) � 1000 1.62 ± 0.3 1.81 ± 0.26 3.6 36 0.06 0.09Right amygdala (BV-corr) � 1000 1.65 ± 0.25 1.80 ± 0.16 3.8 36 0.06 0.10Total amygdala (BV-corr) � 1000 3.27 ± 0.51 3.60 ± 0.30 4.8 36 0.04 0.12Left amygdala (BV-age) �0.07 ± 0.32 0.13 ± 0.25 4.1 36 0.05 0.10Right amygdala (BV-age) �0.06 ± 0.25 0.11 ± 0.18 4.8 36 0.03 0.12Total amygdala (BV-age) �0.14 ± 0.53 0.23 ± 0.31 5.7 36 0.02 0.14

Tissue volumes are displayed as mean values ± STD in cubic centimeters (�ml). Absolute measurements are the uncorrected volumes. BV-corr = corrected for differences inbrain size (division by brain volume). BV-age = adjusted for differences in brain size and age (residualized after brain size and age). The effect size is measured as the partialeta squared ðg2

pÞ. g2p describes the proportion of the variability in the dependent measure that is attributable to a factor. For univariate F tests the formula for the partial eta

squared is: g2p = sums of squares for the effect/(sums of squares for the effect + sums of squares for the error term).

G. Kronenberg et al. / Journal of Psychiatric Research 43 (2009) 1112–1117 1115

volume residualized against brain volume and age: R = 0.06,p = 0.80). In first episode patients, duration of illness equals theduration of the index episode whereas in all others, this variablerepresents the time [weeks] since the first occurrence of an episodeof major depression irrespective of disease-free intervals in the in-terim period. We did not detect a significant association betweenduration of illness and amygdala volume measurements (absolutetotal amygdala volume: R = �0.33, p = 0.11; BV-corrected totalamygdala volume: R = �0.15, p = 0.50; total amygdala volumeresidualized against brain volume and age: R = �0.15, p = 0.47).However, the number of preceding depressive episodes showed asignificant inverse relationship with amygdala volume measure-ments (absolute total amygdala volume: R = �0.53, p = 0.02; BV-corrected total amygdala volume: R = �0.45, p = 0.03; total amyg-dala volume residualized against BV and age: R = �0.47, p = 0.02;(Fig. 2)).

3.3.4. Correlations between amygdala volumes and cortisol measuresA complete set of cortisol measurements was available for 23

patients. We did not detect a correlation between baseline morn-ing cortisol concentrations and amygdala volume measurements

Fig. 2. Correlation between absolute amygdala volume meas

(absolute total amygdala volume: R = �0.04, p = 0.84; BV-correctedtotal amygdala volume: R = 0.2, p = 0.30; total amygdala volumeresidualized against BV and age: R = 0.2, p = 0.38). Similarly, base-line afternoon saliva cortisol levels were not related to amygdalavolume (absolute total amygdala volume: R = 0.04, p = 0.86; BV-corrected total amygdala volume: R = 0.3, p = 0.16; total amygdalavolume residualized against brain volume and age: R = 0.26,p = 0.24).

4. Discussion

The amygdala is a key brain structure integrating emotional sal-ience with experience and perception. Using quantitative struc-tural MRI, we here investigated amygdala volumes in newlyhospitalized patients with major depression. Our study yieldedthe following major findings: (1) As compared to matched controlparticipants, patients displayed smaller absolute left and absoluteright amygdala volumes. Importantly, patients investigated herewere off psychotropic medication at the time of MRI and did notsuffer from relevant comorbid medical or neurological disorders.(2) Amygdala volumes in depression remained robustly reduced

urements and number of preceding depressive episodes.

1116 G. Kronenberg et al. / Journal of Psychiatric Research 43 (2009) 1112–1117

after correction for brain volume and age effects (BV-corrected vol-umes, residualized volumes). (3) We did not detect a relationshipbetween severity of depression (HAMD scores) at baseline andamygdala volume measurements. Similarly, baseline cortisol con-centrations were not related to amygdala volume. However, thenumber of preceding depressive episodes was significantly inver-sely related to amygdala volume.

So far, structural neuroimaging studies in depression haveyielded conflicting results. Notably, both unchanged, increasedand decreased amygdala volumes have been reported.

Bremner et al. (2000) report increased right amygdala volumesin patients with major depression in remission. However, this ef-fect did not hold when differences in brain volume were controlledfor. Another recent study found increased amygdala volumes onlyin first episode depression whereas amygdala sizes did not differbetween patients with recurrent major depression and healthycontrols (Frodl et al., 2003). However, in that study, patients hadalready received antidepressant pharmacotherapy during hospital-ization before MRI scans were performed. Similarly, two other re-cent studies by one group found enlarged amygdala sizes in youngwomen with major depression on antidepressant medication(Lange and Irle, 2004; Weniger et al., 2006). We have previously re-ported increased amygdala volumes in dysthymic patients suffer-ing from temporal lobe epilepsy (Tebartz van Elst et al., 1999).Furthermore, an investigation of fronto-limbic brain structures in7 suicidal and 10 non-suicidal medication-free unipolar female pa-tients recently reported larger amygdala volumes in the suicidalsubgroup (Monkul et al., 2007). In that study, amygdala volumeswere not correlated with the number of previous episodes, lengthof illness or age at onset. However, at the time of study participa-tion, this study sample was relatively inhomogeneous with regardto presence and severity of depression (Monkul et al., 2007).

Additionally, two early studies of the amygdala–hippocampalcomplex in mood disorders did not find volumetric changes com-pared to healthy volunteers (Axelson et al., 1993; Coffey et al.,1993).

By contrast, a number of other studies report a reduction inamygdala volume in major depression consistent with the resultsreported here. Caetano et al. present unadjusted mean amygdalavolumes and report a non-significant trend toward decreased leftamygdala size in depression (reduction by �10% as compared tohealthy control subjects; Caetano et al., 2004). Importantly, onlya subset of patients investigated in that study was currently de-pressed at the time of MRI examination. Also, information on med-ication status and hospitalization is not provided.

Another study which normalized to total cerebral volume foundsignificantly reduced right amygdala volumes in female depressedpatients (compared with healthy female volunteers, female de-pressed subjects had a 23% smaller right amygdala and showed atrend toward 15% smaller left amygdala volume; Hastings et al.,2004). In that study, no significant correlations were observed be-tween normalized amygdala volume and clinical variables includ-ing age, age at first episode of major depression, and number ofmajor depressive episodes. Importantly, patients did not receiveantidepressant medication at the time of the MRI scan. However,the major thrust of that investigation was to compare effects ofgender on brain volumetric measurements, the size of the samplewas relatively small (10 female and 8 male patients), and patientsdid not require hospitalization. Another recent study which onlyprovides uncorrected volumes also found significantly decreasedtotal amygdala size in depression (Hickie et al., 2007).

A meta-analysis of magnetic resonance imaging studies ofamygdala volume in major depressive disorder recently concludedthat depression, as an unmedicated disorder, is associated with de-creased amygdala size (Hamilton et al., 2008). The present study ofa sample of unmedicated depressed in-patients investigated after

admission to a psychiatric unit toward the end of a one-weekwashout period bears out this proposition. Only four of the 24 pa-tients investigated here had received antidepressant medicationbefore being admitted to the hospital. Importantly, we have takena well-established approach both in measuring size of amygdala(e.g. Tebartz van Elst et al., 1999, 2000) and in correcting for pos-sible confounders such as age and BV (e.g. Colla et al., 2007). Also,the data from responders during wash-out were dropped from thisstudy which may have increased the size of the observed effects.However, one important limitation of our study is the still rela-tively small sample size which precludes systematic differentiationof effects of previous medication, affective status, hospitalizationand disease history.

Interestingly, we did not detect a relationship between eitherbaseline depressive symptomatology (HAMD scores) or baselinesaliva cortisol concentrations and amygdala volume measure-ments. Obviously, baseline psychopathology and baseline salivacortisol reported here only present a narrow snapshot of precedingdisease severity. Also, saliva cortisol levels measured at the time ofMRI examination need not necessarily reflect past cortisol levelsthat may have exerted untoward effects on the brain.

It is noteworthy that in our sample, previous depressive epi-sodes were negatively related to amygdala size. This may be inter-preted as evidence that depression decreases amygdala size.However, at least as relates to cortisol, experimental and clinicalevidence linking neuroplastic volume changes of the amygdala tocirculating glucocorticoids and stress is largely lacking.

Another interpretation of the negative correlation between pre-ceding depressive episodes and amygdala volume is that reducedamygdala volume constitutes a risk factor for depression. In linewith this interpretation, a recent prospective longitudinal 3-yearstudy of patients with major depression failed to detect an effectof the clinical course (i.e. remission versus non-remission) onamygdala volume in the patient sample (Frodl et al., 2008). How-ever, compared to controls, patients showed significantly more de-cline in gray matter density of the left amygdala among other brainregions (Frodl et al., 2008).

In conclusion, our established methodical approach with dis-tinct volume corrections adds to the small, yet growing, body ofevidence linking untreated major depression to reduced amygdalavolume.

Contributors

All authors contributed to this study and approved the finalmanuscript.

Role of the funding source

This work was funded by DFG De 660/1-1 (to M.D. and I.H.) andby a grant from the Ministry of Science, Research and the Arts ofBaden-Württemberg AZ: 23-7532.22-11/1 (to L.TvE). The fundingagencies had no further role in study design; in the collection, anal-ysis and interpretation of data; in the writing of the report; and inthe decision to submit the paper for publication.

Conflict of interest statement

None declared.

Acknowledgements

This work was funded by DFG De 660/1-1 (to M.D. and I.H.) andby a grant from the Ministry of Science, Research and the Arts ofBaden-Württemberg AZ: 23-7532.22-11/1 (to L.TvE).

G. Kronenberg et al. / Journal of Psychiatric Research 43 (2009) 1112–1117 1117

References

Axelson DA, Doraiswamy PM, McDonald WM, Boyko OB, Tupler LA, Patterson LJ,et al. Hypercortisolemia and hippocampal changes in depression. PsychiatryResearch 1993;47(2):163–73.

Bowley MP, Drevets WC, Ongür D, Price JL. Low glial numbers in the amygdala inmajor depressive disorder. Biological Psychiatry 2002;52(5):404–12.

Bremner JD, Narayan M, Anderson ER, Staib LH, Miller HL, Charney DS. Hippocampalvolume reduction in major depression. American Journal of Psychiatry2000;157(1):115–8.

Caetano SC, Hatch JP, Brambilla P, Sassi RB, Nicoletti M, Mallinger AG, et al.Anatomical MRI study of hippocampus and amygdala in patients with currentand remitted major depression. Psychiatry Research 2004;132(2):141–7.

Carrillo B, Pinos H, Guillamón A, Panzica G, Collado P. Morphometrical andneurochemical changes in the anteroventral subdivision of the rat medialamygdala during estrous cycle. Brain Research 2007;1150:83–93.

Coffey CE, Wilkinson WE, Weiner RD, Parashos IA, Djang WT, Webb MC, et al.Quantitative cerebral anatomy in depression. A controlled magnetic resonanceimaging study. Arch Gen Psychiatry 1993;50(1):7–16.

Colla M, Kronenberg G, Deuschle M, Meichel K, Hagen T, Bohrer M, et al.Hippocampal volume reduction and HPA-system activity in major depression.Journal of Psychiatric Research 2007;41(7):553–60.

Dannlowski U, Ohrmann P, Bauer J, Kugel H, Arolt V, Heindel W, et al. Amygdalareactivity to masked negative faces is associated with automatic judgmentalbias in major depression: a 3 T fMRI study. Journal of Psychiatry andNeuroscience 2007;32(6):423–9.

Davis M, Whalen PJ. The amygdala: vigilance and emotion. Molecular Psychiatry2001;6(1):13–34.

Drevets WC, Price JL, Bardgett ME, Reich T, Todd RD, Raichle ME. Glucosemetabolism in the amygdala in depression: relationship to diagnostic subtypeand plasma cortisol levels. Pharmacology Biochemistry and Behaviour2002;71(3):431–47.

Drevets WC, Videen TO, Price JL, Preskorn SH, Carmichael ST, Raichle ME. Afunctional anatomical study of unipolar depression. Journal of Neuroscience1992;12(9):3628–41.

Eberling JL, Wu C, Tong-Turnbeaugh R, Jagust W. Estrogen- and tamoxifen-associated effects on brain structure and function. NeuroImage2004;21:364–71.

Frodl TS, Koutsouleris N, Bottlender R, Born C, Jäger M, Scupin I, et al. Depression-related variation in brain morphology over 3 years: effects of stress? Arch GenPsychiatry 2008;65(10):1156–65.

Frodl T, Meisenzahl EM, Zetzsche T, Born C, Jäger M, Groll C, et al. Larger amygdalavolumes in first depressive episode as compared to recurrent major depressionand healthy control subjects. Biological Psychiatry 2003;53(4):338–44.

Hamidi M, Drevets WC, Price JL. Glial reduction in amygdala in major depressivedisorder is due to oligodendrocytes. Biological Psychiatry 2004;55(6):563–9.

Hamilton JP, Gotlib IH. Neural substrates of increased memory sensitivity fornegative stimuli in major depression. Biological Psychiatry2008;63(12):1155–62.

Hamilton JP, Siemer M, Gotlib IH. Amygdala volume in major depressive disorder: ameta-analysis of magnetic resonance imaging studies. Molecular Psychiatry2008;13(11):993–1000.

Hastings RS, Parsey RV, Oquendo MA, Arango V, Mann JJ. Volumetric analysis of theprefrontal cortex, amygdala, and hippocampus in major depression.Neuropsychopharmacology 2004;29(5):952–9.

Hickie IB, Naismith SL, Ward PB, Scott EM, Mitchell PB, Schofield PR, et al. Serotonintransporter gene status predicts caudate nucleus but not amygdala orhippocampal volumes in older persons with major depression. Journal ofAffective Disorders 2007;98(1–2):137–42.

Kawamura T, Chen J, Takahashi T, Ichitani Y, Nakahara D. Prenatal stress suppressescell proliferation in the early developing brain. Neuroreport2006;17(14):1515–8.

Keilhoff G, Becker A, Grecksch G, Bernstein HG, Wolf G. Cell proliferation isinfluenced by bulbectomy and normalized by imipramine treatment in aregion-specific manner. Neuropsychopharmacology 2006;31(6):1165–76.

Kirk RE. Experimental design: procedures for the behavioral sciences. 2nded. Belmont, CA: Brooks/Cole; 1982.

Lange C, Irle E. Enlarged amygdala volume and reduced hippocampal volume inyoung women with major depression. Psychological Medicine 2004;34(6):1059–1064.

Lemieux L, Liu RS, Duncan JS. Hippocampal and cerebellar volumetry in seriallyacquired MRI volume scans. Magnetic Resonance Imaging 2000;18(8):1027–33.

Maguire EA, Spiers HJ, Good CD, Hartley T, Frackowiak RSJ, Burgess N. Navigationexpertise and the human hippocampus: a structural brain imaging study.Hippocampus 2003;13:250–9.

Makris N, Gasic GP, Seidman LJ, Goldstein JM, Gastfriend DR, Elman I, et al.Decreased absolute amygdala volume in cocaine addicts. Neuron2004;44(4):729–40.

Merz J, Lehrl S, Galster V, Erzigkeit H. The multiple selection vocabulary test (MSVT-B) – an accelerated intelligence test. Psychiatry Neurolology MedicalPsychology (Leipz) 1975;27(7):423–8.

Monkul ES, Hatch JP, Nicoletti MA, Spence S, Brambilla P, Lacerda AL, et al. Fronto-limbic brain structures in suicidal and non-suicidal female patients with majordepressive disorder. Mol Psychiatry 2007;12(4):360–6.

Murray EA. The amygdala, reward and emotion. Trends in Cognitive Science2007;11(11):489–97.

Pinkhardt EH, Tebartz van Elst L, Ludolph AC, Kassubek J. Amygdala size inamyotrophic lateral sclerosis without dementia: an in vivo study using MRIvolumetry. BMC Neurology 2006;6:48.

Szeszko PR, Robinson D, Alvir JM, Bilder RM, Lencz T, Ashtari M, et al. Orbital frontaland amygdala volume reductions in obsessive-compulsive disorder. Arch GenPsychiatry 1999;56(10):913–9.

Tabachnick BG, Fidell LS. Using multivariate statistics. 2nd ed. New York: Harper &Row; 1989.

Tebartz van Elst L, Woermann F, Lemieux L, Trimble MR. Amygdala enlargement indysthymia – a volumetric study of patients with temporal lobe epilepsy.Biological Psychiatry 1999;46:1614–23.

Tebartz van Elst L, Woermann FG, Lemieux L, Thompson PJ, Trimble MR. Affectiveaggression in patients with temporal lobe epilepsy: a quantitative MRI study ofthe amygdala. Brain 2000;123(Pt 2):234–43.

Tebartz van Elst L, Baeumer D, Lemieux L, Woermann FG, Koepp M, KrishnamoorthyS, et al. Amygdala pathology in psychosis of epilepsy: a magnetic resonanceimaging study in patients with temporal lobe epilepsy. Brain 2002;125(Pt1):140–9.

Tebartz van Elst L, Hesslinger B, Thiel T, Geiger E, Haegele K, Lemieux L, et al.Frontolimbic brain abnormalities in patients with borderline personalitydisorder: a volumetric magnetic resonance imaging study. BiologicalPsychiatry 2003;54(2):163–71.

van Petten C. Relationship between hippocampal volume and memory ability inhealthy individuals across the lifespan: review and meta-analysis.Neuropsychologia 2004;42(10):1394–413.

Watson C, Andermann F, Gloor P, Jones-Gotman M, Peters T, Evans A, et al.Anatomic basis of amygdaloid and hippocampal volume measurements bymagnetic resonance imaging. Neurology 1992;42:1743–50.

Weber-Hamann B, Kopf D, Lederbogen F, Gilles M, Heuser I, Colla M, et al. Activity ofthe hypothalamus-pituitary-adrenal system and oral glucose tolerance indepressed patients. Neuroendocrinology 2005;81(3):200–4.

Weber-Hamann B, Werner M, Hentschel F, Bindeballe N, Lederbogen F, Deuschle M,et al. Metabolic changes in elderly patients with major depression: evidence forincreased accumulation of visceral fat at follow-up. Psychoneuroendocrinology2006;31(3):347–54.

Weniger G, Lange C, Irle E. Abnormal size of the amygdala predicts impairedemotional memory in major depressive disorder. Journal of Affective Disorders2006;94(1–3):219–29.

Wennström M, Hellsten J, Tingström A. Electroconvulsive seizures induceproliferation of NG2-expressing glial cells in adult rat amygdala. BiologicalPsychiatry 2004;55(5):464–71.

Wittchen H, Zaudig M, Frydrich T. Strukturiertes Klinisches Interview für DSM-IV(SKID-I und SKID-II) Achse I: Psychische Störungen/Achse II: Persönlichkeits-störungen. Hogrefe Verlag, 1997.

Wrase J, Makris N, Braus DF, Mann K, Smolka MN, Kennedy DN, et al. Amygdalavolume associated with alcohol abuse relapse and craving. American Journal ofPsychiatry 2008;165(9):1179–84.

Yang RJ, Mozhui K, Karlsson RM, Cameron HA, Williams RW, Holmes A. Variation inmouse basolateral amygdala volume is associated with differences in stressreactivity and fear learning. Neuropsychopharmacology, 2008 [Epub ahead ofprint].