Amygdala and hippocampal substructure volumes and their association with improvement in mood symptoms in patients with mood disorders undergoing electroconvulsive therapy

Electroconvulsive therapy (ECT) demonstrates favorable outcomes in the management of severe depressive disorders. ECT has been consistently associated with volumetric increases in the amygdala and hippocampus. However, the underlying mechanisms of these structural changes and their association to clinical improvement remains unclear. In this cross-sectional structural MRI study, we assessed the difference in amygdala subnuclei and hippocampus subfields in n = 37 patients with either unipolar or bipolar disorder immediately after eighth ECT sessions compared to ( n = 40) demographically matched patients in partial remission who did not receive ECT (NoECT group). Relative to NoECT, the ECT group showed significantly larger bilateral amygdala volumes post-treatment, with the effect originating from the lateral, basal, and paralaminar nuclei and the left cortico-amydaloid transition area. No significant group differences were observed for the hippocampal or cortical volumes. ECT was associated with a significant decrease in depressive symptoms. However, there were no significant correlations between amygdala subnuclei volumes and symptom improvement. Our study corrobo-rates previous reports on increased amygdalae volumes following ECT and further identifies the subnuclei driving this effect. However, the therapeutic effect of ECT does not seem to be directly related to structural changes in the amygdala.


Introduction
Major depressive disorder (MDD) and bipolar disorder (BD) are a considerable source of disease burden throughout the world, with severe negative impacts on quality of life and cognitive function (Proudman et al., 2021).Electroconvulsive therapy (ECT) has been shown to be an efficacious treatment for treatment-resistant depression, with response rates of up to 80 % and results up to four times more successful than typical antidepressant pharmaceuticals (Gryglewski et al., 2021;Pagnin et al., 2004).However, ECT is largely underutilized due to stigma surrounding electrically induced convulsions, unclear mechanisms of action, and risk of cognitive side effects (Payne and Prudic, 2009).
ECT has been found to impact the structure and function of several of brain regions, with the most consistent findings observed in the hippocampus, and amygdala (Gbyl et al., 2021;Jorgensen et al., 2015).In particular, a stem cell producing subfield of the hippocampus, the dentate gyrus, has exhibited the most consistent and robust increase in volume following ECT (Gryglewski et al., 2021).Nevertheless, distributed prefrontal cortex (PFC) changes in the density or volume of grey matter as a result of ECT treatment have also been reported in regions such as the anterior cingulate cortex, dorsolateral PFC, frontopolar cortex, and insula (Gryglewski et al., 2019;Ousdal et al., 2020;Pirnia et al., 2016), as well as widespread changes in white matter integrity (Repple et al., 2020).The neurotrophic hypothesis of depression theorizes a reduction in brain-derived neurotrophic factor (BDNF) due to stress, with studies finding lower levels of BDNF in the periphery in MDD patients and an increase in both plasma and serum levels following ECT (van Buel et al., 2015).With MRI studies consistently reporting increased hippocampal volume after ECT-treatment, these findings support the hypothesis that ECT may reverse the reduction in hippocampal neurogenesis produced by depression (Ousdal et al., 2020).However, this neurogenic effect has only been observed as an increase in hippocampal volume in humans (van Buel et al., 2015).
A region associated with structural and functional dysregulation in depression, including reduced volume in unmedicated patients, the amygdala is another core region of interest in the investigation into the antidepressant mechanisms of ECT (Hamilton et al., 2008).Unlike in early depression, which has been shown to increase amygdala volume, decreased volumes have been found in recurrent MDD (Frodl et al., 2003).Electroconvulsive seizures (ECS) in rats have been repeatedly shown to enlarge amygdala volume, after as few as five sessions, by increasing the number of proliferating cells within the medial nucleus (Wennström et al., 2004).This is consistent with the theory that there is an antidepressant-mediated increase in growth factors levels such as BDNF which promotes neurogenesis and prevents glucocorticoid induced shrinkage in amygdala (Hamilton et al., 2008).Like the dentate gyrus, evidence of adult neurogenesis has been observed in the basolateral amygdala, making it another region of interest (Jhaveri et al., 2018).
Despite the well-established increases in the volumes of the amygdala and hippocampus following ECT, there is conflicting evidence on whether there is a straightforward relationship between these structural changes and therapeutic efficacy (Gryglewski et al., 2021).While some research has demonstrated a correlation between dentate gyrus volume and clinical improvement in depressed patients, the overall inconsistent evidence of many studies leaves the question of whether the structural changes associated with ECT have antidepressant effects or simply represents an epiphenomenon unanswered (Gbyl et al., 2021;Nordanskog et al., 2014).Therefore, how these structural changes may contribute to the antidepressant effects of ECT remains unclear.
Given consistent positive effects of ECT on total hippocampal and amygdala volumes demonstrated by previous ECT studies, the purpose of this cross-sectional structural MRI study was to compare MDD and BD patients after three-weeks of ECT (8 sessions) with demographically and diagnosis matched patients in remission who did not receive ECT (NoECT group) in hippocampal and amygdala subvolumes.In addition, we explored group differences in regional cortical thickness and volume and tested whether any of the identified structures showing group differences correlated with symptom improvement following ECT.We expected to observe larger hippocampi and amygdala in ECT-treatment patients with the highest effect observed for the dentate gyrus in the hippocampus and basolateral and medial amygdala subnuclei.

Study design and procedures
For this study, we merged neurocognitive and structural MRI data derived from three study cohorts.Data were pooled from several clinical trials.The ECT group included all participants from Effects of Erythropoietin for Cognitive Side-effects of ECT (EPO-T) study (Schmidt et al., 2018).The NoECT group included age, sex, and diagnosis matched patients from the Bipolar Illness Onset (BIO) study (Kessing et al., 2017), and The Prefrontal Target Engagement as a biomarker model for Cognitive improvement studies (PRETEC-EPO and PRETEC-ABC) (Ott et al., 2021;Petersen et al., 2018).All studies were approved by the Danish Research Ethics Committee for the Capital Region of Denmark (EPO-T: H-16,038,506;BIO: H-7-2014-007;PRETEC-EPO: H-16,043, 370;PRETEC-ABC: H-16,043,480) and the Danish Data Protection Agency Capital Region of Denmark (EPO-T: RHP-2017-023; PRETEC: 2012-58-0004; BIO: RHP-2015-023).Written informed consent following oral and written study information was collected from all participants prior to inclusion.Studies were performed according to the Declaration of Helsinki.
ECT patient data were extracted from the EPO-T study (Schmidt et al., 2018) where all participants underwent ECT and either EPO or placebo treatment.The entire cohort was used as no significant group differences on any pre-defined outcomes were observed between EPOand placebo-treated patients (Miskowiak et al., 2023, in prep).Importantly, there was no significant effect of EPO vs placebo treatment on bilateral amygdala and hippocampus volumes (p > 0.4).Groups included (i) in-patients with MDD or BD who received ECT (ECT group), (ii) patients with MDD or BD in partial or full remission undergoing standard pharmacological treatment-as-usual (NoECT group).ECT patients were assessed twice, undergoing neuropsychological assessment prior to ECT (pre-ECT), repeated and appended with an MRI scan three days post-treatment (post-ECT).NoECT underwent an identical neuropsychological assessment and one identical MRI scan.All scans were conducted on the same MRI unit employing identical acquisition parameters.

Study participants
Full details on inclusion and exclusion criteria across the studies from which data were collected are available on ClinicalTrials.gov;EPO-T: NCT03339596, BIO: NTC02888262 PRETEC-EPO: NCT03315897, PRETEC-ABC: NTC03295305.Every participant underwent assessment, including rating of mood symptoms, neuropsychological screening, and MRI-scanning, at CADIC, Psychiatric Centre Copenhagen and Copenhagen University Hospital, Rigshospitalet.
ECT group.Inclusion criteria included 18-70 years of age, diagnosis of MDD or BD verified with Mini International Neuropsychiatric Interview (M.I.N.I.) (Lecrubier et al., 1997) with present moderate to severe depressive symptoms, a Hamilton Depression Rating Scale 17-items version (HDRS-17) score of ≥17 (Hamilton, 1960) (Danish language fluency, and ability to provide informed content.Main exclusion criteria comprised being referred to compulsory ECT, having underwent ECT within three months prior to enrollment, current substance or alcohol use disorder, or acute significant suicide risk.See (Schmidt et al., 2018) for full exclusion criteria list.
NoECT group.Qualified patients were 18-65 years of age, diagnosed with an ICD-10 MDD or BD diagnosis, ascertained with the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) (Wing et al., 1990) and were in full or partial remission (defined as total score ≤14 on Hamilton Depression Rating Scale 17-items (HDRS-17) and ≤7 on Young Mania Rating Scale (YMRS) (Young et al., 1978).

ECT procedures
Participants underwent 8 ECT sessions (on consecutive Mondays, Wednesdays, Fridays) and followed the standard protocol of the Capital Region of Denmark (Videbech et al., 2020).Electrode configuration was bitemporal, pulse amplitude 0.9Ampere, width 0.5-1.5 ms, with anesthesia and muscle-relaxation achieved with thiopental and succinylcholine.Initial percentage charge was determined as half of the age of the patient, while charges in subsequent sessions were adjusted based on seizure quality.

Structural data processing
All participants were scanned with an identical structural MRI acquisition protocols presented in the Supplementary Materials.All scans were corrected for B0 field geometric distortions and visually inspected to rule out overt structural brain abnormalities or artifacts.Each T1-weighted image was processed at the subject-level using the FreeSurfer analysis suite v7.3.2 (http://surfer.nmr.mgh.harvard.edu/).The workflow included skull-stripping, intensity normalization, transformation to Talairach space, and automatic segmentation of cortical and subcortical gray-matter structures (Dale et al., 1999;Fischl et al., 2002).Post-processing, all outputs were individually assessed by two raters to ensure data quality.When warranted, smaller edits of the pial surfaces were completed and evaluated.After ensuring the quality of the main pipeline, an additional algorithm was employed to subdivide the amygdala and hippocampus in to 9 constituent subnuclei and 11 distinctive subfields, respectively (Iglesias et al., 2015;Saygin et al., 2017).Amygdalar subdivisions included the basal (Ba), accessory basal (AB), lateral (La), medial, central, cortical, and paralaminar nuclei, as well as the anterior amygdala and cortico-amygdaloid transition areas.Obtained hippocampal subfields comprised the hippocampal tail, subiculum, presubiculum, parasubiculum, cornu ammonis (CA1-4), heads and tails of the granule cell and molecular layers of the dentate gyrus (GCMLDG), fimbria, and hippocampal-amygdaloid transition area.To review the integrity of these added segmentations, a previously described quality control pipeline for hippocampal subfields was employed (Sämann et al., 2022).This specialized protocol was appended with a stem-and-leaf data exploration for amygdala subnuclei, after which all outliers were inspected, and inter-rater agreement on their utility was achieved.Outliers were defined according to an interquartile range of 1.5.

Statistical analyses 2.5.1. Demographic and clinical assessment
All analyses were conducted using Statistical Packages for the Social Sciences (SPSS) v28 (IBM Corporation, Armonk, NY), with an a-level of p < 0.05 (two-tailed).Data normality was ensured with the Shapiro-Wilk test (Shapiro and Wilk, 1965)and visual inspections of histograms and Q-Q-plots.Independent samples t-tests (normally distributed data), Mann-Whitney U tests (non-parametric data), and Pearson's Chi-square (χ 2 ) were conducted to investigate between-group differences in clinical and demographical characteristics.

Amygdala and hippocampus volume analysis
Group differences in left and right amygdala and hippocampus volumes were tested separately using multivariate GLM models (MAN-COVA) including all segmented subvolumes within one of the structures as dependent variables, treatment group (ECT vs. NoECT) as fixed effect, and age, sex, and total intracranial volume (TIV) as covariates.This approach was implemented in order to consider the joint volumetric variation across substructures, capture potential differences in the patterns of regional volumetric changes, and minimize the number of comparisons.Upon significant effect of group, we assessed the post-hoc univariate tests of the individual subvolumes to clarify the regional contribution to the overall effect of group (Yildirim et al., 2023).In separate univariate GLM models, we further tested group differences in whole structure volumes using univariate tests controlling for age, sex, and TIV.Since the ECT group showed significantly higher depressive symptoms compared to NoECT, we rerun all statistical models further adjusting for depressive symptoms (HDRS-17 scores).

Exploratory cortical analysis
We performed a vertex-wise analysis across the entire cortex to assess group differences in regional cortical volume and thickness using a GLM model with group as fixed factor and age, sex and TIV as covariates.Cluster-wise significance was assessed at p < 0.05 following correction for vertex-wise multiple comparisons using Monte Carlo null-z simulations based on precomputed simulation data in FreeSurfer and a cluster forming threshold of p < 0.001.

Post-hoc analyses
To aid the interpretation of the group analysis adjusted by symptom severity, the volume data of the regions showing a significant effect of group (unadjusted for symptoms) was tested for correlations (Pearson's r or Spearman's Rho depending on normality) with post-ECT HDRS-17 scores within the ECT group and NoECT group separately.
Since the ECT group showed increased use of antipsychotics and medication in general we assessed the impact of these factors on bilateral amygdala and hippocampus subvolumes by setting up multivariate MANCOVA models analogue to the ones used for the main analysis above, but using use of either antipsychotics or any medication as a fixed factor.
Lastly, within the ECT group, we tested whether the EPO treatment had a significant effect on the subvolumes using analogue MANCOVA models with EPO treatment as fixed factor.All post-hoc analyses on volumetric data were adjusted for age, sex and TIV.

Group comparison on demographic and clinical characteristics
One participant from the ECT-treated group was excluded due to visible head motion artifacts, thus 37 ECT and 40 NoECT participants were included in all group comparisons (Table 1).The ECT patients underwent the MRI scan and post-ECT neuropsychological test on the same day, while NoECT patients completed these assessments 0-3 days apart.The ECT and NoECT groups were matched on age, sex,  1).Following the ECT treatment, patients showed a significant decrease in HDRS-17 scores of 36.9 % (p < 0.001).Nevertheless, post-ECT, at the time of the MRI investigation, the depressions symptoms in the ECTtreated group were still significantly higher compared to the NoECT group (Table 1, p < 0.001).There was a significant correlation between the pre-treatment HDRS-17 scores and both pre-to post treatment decrease in symptoms (r=− 0.398, p = 0.015), and the level of posttreatment symptoms (r = 0.826, p < 0.001) i.e., the more severe the pre-treatment symptoms the larger the symptom improvement and the lower the post-treatment symptoms.

Exploratory cortical analysis
The vertex-wise analyses across the entire cortex adjusted for age, sex, and TIV found no significant effect of treatment group on regional cortical volume or thickness.

Post-hoc analyses
Adjusting the amygdala and hippocampus volume MANCOVA models for post ECT depressive symptom severity (HDRS-17 scores) revealed a reduced significance of the group effect on amygdala subnuclei volumes which only remained significant for the left amygdala (F (63,9) =2.03, p = 0.050), and not for the right (F (63,9) =1.39, p = 0.210) amygdala, or the right hippocampus (F (53,29) =1.18, p = 0.313).Post-hoc univariate tests revealed that the effect of treatment group prevailed for the paralaminar nucleus of the left amygdala (p = 0.015).
In the NoECT group, none of the amygdala subnuclei volumes that demonstrated a significant effect of treatment group showed significant correlation with HDRS-17 scores (p > 0.256, model adjusted by age, sex, and TIV).Similarly, in the ECT-treated group only the right basal nucleus volume showed a significant positive correlation with posttreatment HDRS-17 scores (r = 0.397, p = 0.020), and no subnuclei correlated with the pre-to post-treatment change with HDRS-17 scores (p > 0.081).
Medication use (medicated n = 60 vs not medicated n = 17) across patient groups was not associated with the amygdala or hippocampal volumes (p > 0.215) bilaterally.Similarly, use of antipsychotics (users n = 28, non-users n = 49) was not associated with any of the structures (p > 0.06).

Discussion
The present findings show significantly larger bilateral amygdala volumes in mood disorder patients immediately after eight ECT sessions compared to demographically and diagnosis matched patients in remission in analyses adjusted for age, sex and total intracranial volume.The larger amygdala volumes were driven by specific subfields i.e., the lateral, basal, and paralaminar nuclei bilaterally, and the left corticoamygdaloid transition area (CTA).The total hippocampal volume and regional cortical volume and thickness were similar between the ECT and NoECT groups.However, there was a significant effect of the treatment group on the right hippocampal subfield volumes with both subthreshold larger and smaller subfields in the ECT vs NoECT group.As expected, we found significant decreases in depression symptoms (on average 10 ± 9 SD points on the HDRS) in the ECT group from before to after eight ECT sessions.However post-ECT, these scores were still significantly higher compared to the NoECT group (mean ± SD: ECT 17 ± 9, NoECT 8 ± 5) which prompted our post-hoc analysis adjusting for HDRS.We found no significant correlation between the amygdala subnuclei volumes showing an effect of group and degree of symptom improvement in the ECT group.
Our results are consistent with previous literature demonstrating increased volume of the entire amygdala and specific amygdala subnuclei volumes following ECT.Previous studies showed ECT and ECS (as used in preclinical models of ECT) increase volume in the right, left, and total amygdala in humans and rats, respectively (Gryglewski et al., 2021;Jorgensen et al., 2015;Joshi et al., 2016;Wennström et al., 2004).Specifically, the basal, lateral, and the paralaminar nuclei, and the CTA, have been found enlarged following ECT (Gryglewski et al., 2019;Xu et al., 2022), change that has generally been attributed to amygdala neurogenesis (Jhaveri et al., 2018).The basolateral complex is a key structure encoding emotional valence (Daviu et al., 2019) and together with the CTA may exhibit a negative association with the severity of depressive symptoms in MDD (Brown et al., 2019).Adjusting the statistical models testing for group differences in amygdala volumes for depressive symptom severity accounted to a significant degree for the observed group differences, which remained significant for the left amygdala.Following this adjustment for depression symptoms in the model including all subnuclei, the effect of the ECT group also remained significant only for the left paralaminar nucleus.This effect is not surprising since ECT may be responsible for both the symptom improvement and enlargement of the amygdala substructure i.e., these effects happen in parallel even though one may not necessarily have a causal effect on the other.
The absence of significant correlation between amygdala subnuclei volumes showing an effect of group and degree of symptom improvement in the ECT group aligns with a previous study (Daftary et al., 2019), suggesting that amygdala volume changes are not directly related to symptom improvement during the acute phase of ECT.Furthermore, corroborating this finding, one recent study found increases in amygdala volumes in MDD and BD patients undergoing ECT but no increase in patients that responded to treatment as usual, supporting the notion that amygdala plasticity is not a prerequisite for symptom improvement (Bracht et al., 2023).As a potential alternative mechanism, ECT may alter the functional connectivity between regions involved in affect without overt gray matter changes (Takamiya et al., 2021).Domke et al. found that increased connectivity between the left amygdala and the left DLPFC was associated with symptom improvement following ECT in a treatment-resistant depression population (Domke et al., 2023).Nevertheless, accumulating evidence suggests the role of amygdala in mood disorders, with smaller volumes observed in recurrent MDD (Frodl et al., 2003(Frodl et al., , 2002) ) and in BD patients with subsequent mood episodes compared with patients who remained mood stable (Macoveanu et al., 2023).These findings imply a potential role of the amygdala in a compensatory mechanism against mood episode onset.Thus, the larger amygdala volume in our ECT-treated patients with long-term illness and severe depression may still hold mechanistic importance, despite its lack of association with acute-phase depression changes.While further research may be needed to better understand the complex interplay between ECT, symptom improvement, and changes in amygdala volumes, the present findings suggests that the therapeutic effects of ECT are likely unrelated or indirectly related to the induced structural changes in the amygdala.
In contrast with our expectations, we found no significant differences in total hippocampal volumes in the ECT vs NoECT patients.While there was a significant effect of treatment group on the right hippocampal subfield volumes, none of the subfileds showed significant group differences.There is a preponderance of evidence of smaller hippocampal volumes in recurrent treatment-resistant severe depression (MacQueen et al., 2008(MacQueen et al., , 2003) ) which is consistent with the chronic stress-induced glucocorticoid toxicity affecting both hippocampus and amygdala (Hageman et al., 2008;Hamilton et al., 2008;Sapolsky, 2000).Given that longitudinal ECT studies have demonstrated consistent bilateral increases of the hippocampus (Argyelan et al., 2021;Gbyl et al., 2021;Gryglewski et al., 2021;Wilkinson et al., 2017), the most likely explanation is that the ECT treatment also increased hippocampal volumes in our cohort to a similar size to that of patients in remission.In line with our discussion of amygdala plasticity, it remains unclear whether and how ECT induced hippocampal plasticity mediates favorable antidepressant responses to ECT.Since increases in hippocampal gray matter volume have also been observed in ECT non-responders, it has been suggested that the location of ECT-related plasticity within the hippocampus may be critical to antidepressant outcome (Leaver et al., 2020).Consistent with this suggestion, several studies demonstrated ECT-induced volume increase in the dentate gyrus (Cao et al., 2018;Nuninga et al., 2020;Takamiya et al., 2019) which were associated with decreased depression scores (Nuninga et al., 2020).
Our exploratory analysis found no difference in cortical thickness or volume between the ECT and NoECT groups, which was expected based on recent findings demonstrating increases in cortical thickness following ECT (Gbyl et al., 2019;Sartorius et al., 2016;van Eijndhoven et al., 2016).Since significant frontal and temporal cortical thinning has been observed in treatment resistant depression compared to healthy controls (Chen et al., 2022), it is likely that the ECT treatment also increased cortical gray matter volumes in our cohort matching the level of our NoECT group.
Finally, it is relevant to consider possible neural mechanisms underlying ECT-induced plasticity.There is substantial evidence suggesting neurogenesis, gliogenesis, and angiogenesis following both acute and chronic ECS in rats, and the same mechanism has been proposed to underlie the observed structural changes in humans (for review see, Krishnan, 2016).This hypothesis is consistent with studies reporting increases in BDNF following ECT (Sorri et al., 2018).However, previous research has demonstrated structural changes can occur as soon as two hours following ECT.It appears implausible that neuroplasticity underlies such immediate volumetric effects (Brancati et al., 2021).Rather, such immediate changes may be the result of an inflammatory response.At the same time, studies have also demonstrated structural changes can persist for more than 6 months following ECT, which are more likely resulting from neuroplasticity than inflammation (Brancati et al., 2021).van Buel and colleagues propose that an inflammatory response induced immediately after single ECT sessions may mobilize neurotrophin expression and endogenous neuroprotection via microglial activation and microglia-derived BDNF (van Buel et al., 2015).In the present study the MRI scans were taken after three weeks (8 sessions) of ECT.Thus, it is difficult to conclude whether the asserted amygdala nuclei enlargement is a result of inflammation, neuroplasticity, or some other mechanism, and whether these changes are beneficial, especially since larger amygdala subnuclei volumes (except for the right basal nucleus) did not correlate with less severe symptoms.Future research should further investigate the complex interactions between ECT, depressive symptomatology, and amygdala subfield volumes and functional connectivity.
A significant limitation of our study was its cross-sectional design, which hindered our ability to infer a causal effect of ECT on amygdalae enlargement.Nevertheless, enlarged amygdalae is a consistent finding in previous longitudinal ECT trials, which we were able to replicate here when comparing to a demographically matched clinical control group, with the main aim being to expand previous findings by investigating the specific amygdala subnuclei affected.Secondly, the clinical control group in this investigation exhibited milder depressive symptoms and lower usage of antipsychotic and antidepressant medications, factors that might have impeded the detection of additional ECT-induced structural alterations, such as those within the hippocampus.
In conclusion, the current study brings indirect evidence supporting the prevailing notion that ECT increases amygdala volume, in particular in the lateral, basal, and paralaminar nuclei and in the left corticoamydaloid transition area.The observed effects in the amygdala were unrelated to ECT-related change in depressive symptoms, indicating that the efficacy of ECT is probably mediated by a different mechanism, e.g., increased cortico-limbic top-down regulation of mood.It remains to be determined whether changes in amygdala subnuclei volumes are clinically relevant for longer-term mood stability in line with a possible role of enlarged amygdala volume in compensatory mechanisms that may protect against new mood episodes.

Funding sources
The study was partially funded by the Lundbeck Foundation grant number: R215-20154121, Augustinus Fonden, and Ivan Nielsens Fonden for Personer med Specielle Sindslidelser.The foundations have not been involved in writing the present manuscript or in the design of the study, nor in the data collection, analysis, or interpretation of data.

Declaration of competing interest
KWM has received consultancy fees from Lundbeck, Janssen, Angelini Pharma and Richter Gedeon in the past three years.JZP has within the last three years received honoraria from Lundbeck Pharma.LVK has within recent three years been a consultant for Lundbeck and Teva A/S.JM, SC, EBKS, ATYN, MBJ report no conflicts of interest.

Fig. 1 .
Fig. 1.Differences in mean amygdala subnuclei volumes in patients undergoing ECT relative to a matched patient control group (NoECT).Error bars represent 95 % confidence intervals.

Fig. 2 .
Fig. 2. Differences in mean total left and right amygdala and hippocampus volumes in patients undergoing ECT relative to a matched patient control group (NoECT).Error bars represent 95 % confidence intervals.

Table 1
Group comparison of ECT patients and standard pharmacological-treated (NoECT) patients in demographic and clinical variables.years, diagnosis, age of illness onset, and illness duration.Fewer patients in the ECT group were unmedicated, and treatment with antipsychotic medication was more frequent in this group compared to the NoECT group (65% vs. 10 %; Table Abbreviations: EPO=erythropoietin; M=mean; SD=standard deviation; Mdn=Median; IQR=Interquartile range; MDD=major depressive disorder; BD=bipolar disorder; HDRS-17=Hamilton Depression Rating Scale 17-items.Independent samples t-tests for normally distributed data (mean (SD)), Mann-Whitney U test for non-parametric data (median (IQR)), chi-square for categorical variables.*Paired t-test comparing pre-vs.post ECT HDRS-17 scores showed a 37 % improvement after treatment (p < 0.001). 1 value missing for one participant 2 values missing for two participants 3 values missing for nine participants.educational