Meta-analytic Evidence for Volume Increases in the Medial Temporal Lobe After Electroconvulsive Therapy

To the Editor:

Since its introduction to psychiatric practice more than 8 decades ago, electroconvulsive therapy (ECT) is widely recognized as a highly effective treatment for severe psychiatric disorders, but the exact mechanisms underlying treatment response have remained elusive. Neuroimaging research has put a spotlight on structural brain changes (1), initially inspired by the neurotrophic hypothesis of depression that postulates neuroneogenesis as essential for antidepressant treatment response (2,3). Intriguingly, some longitudinal magnetic resonance imaging studies indeed reported localized brain volume changes in patients after ECT. A straightforward interpretation of these findings, however, has been obscured, given the considerable heterogeneity of brain locations reported across studies. These inconsistent results might be attributable to several factors. Small sample sizes are a particular problem, as individuals undergoing ECT are among the most severely affected psychiatric patients (4) and, hence, are hard to recruit. Intra- and interstudy inhomogeneity of clinical samples—e.g., owing to diagnosis, age, and chronicity—and different statistical approaches for data analysis might further obscure the picture (5).

To investigate whether there is converging evidence for ECT-associated brain structural changes, we performed a quantitative coordinate-based meta-analysis using the activation likelihood estimation approach. This approach allows us to overcome the mentioned shortcomings, as it pools across large numbers of subjects from different cohorts and synthesizes results from different analysis approaches (6,7). Moreover, as activation likelihood estimation performs a voxelwise analysis for significant convergence across the entire gray matter, it does not require an a priori hypothesis (7). These two characteristics—pooling over different analysis approaches and the use of unrestricted inference spaces—yield the potential to provide an important additional perspective to recently published mega-analyses relying on region-of-interest analyses from large regions of interest, which would be insensitive to small, localized effects (8,9). Relevant studies were retrieved through PubMed and Google Scholar, review articles, and reference tracing. Inclusion criteria are reported in the legend to Table 1. In total, 12 studies published between 2014 and 2019 met inclusion criteria, with 2 of those reporting no significant changes (Table 1), enrolling a total of 308 patients (294 patients with major depression, 5 with bipolar disorder, 9 with schizophrenia). Our analysis pooled over all contrasts, regardless of their directionality (i.e., regional brain volume increases or decreases in the course of ECT). We then analyzed which studies contributed to the ensuing cluster to gain more insight into the effects that drove the observed convergence. Owing to the relatively small number of studies suitable for inclusion, we performed a jackknife (leave-one-out) analysis on the 10 studies reporting significant findings to assess the robustness of our results. All results were thresholded at a cluster-level familywise error cluster-corrected threshold of p(cluster-level familywise error) < .05 (cluster-forming threshold at p < .001) as recommended (7). Maps of the leave-one-out analyses were binarized and then averaged to yield the (per-voxel) probabilities of significant convergence when removing one of the studies from the analysis.

Table 1.

Overview of the Included Studies

Reference	Patients, N	Diagnosis	Electrode Position	Age, Years, Mean ± SD	Number of ECT Treatments, Mean ± SD or Mean	Sex	Clinical Improvement and Depression Severity After ECT Index Series	Cognitive Side Effects	MRI Parameters	Preprocessing Pipeline
Biedermann et al., 2015 (19)	22	MDD (DSM-IV)	Not reported	49.1 ± 16.56	Not reported	5 M (23%), 17 F (77%)	HAMD-17 (pre- ECT: 28.0 ± 7.0, post-ECT: 9.2 ± 5.1) BDI (34.1 ± 13.6 vs. 16.2 ± 12.5) Mild depression	MMSE (pre- ECT: 26.7 ± 3.9, post- ECT: 28.1 ± 3.0)	Siemens Magnetom Vision Plus 1.5T TR = 1.35s, TE 1/2 = 40/32 ms, 1000 Hz bandwidth, spatial resolution 1.05 mm3, 154 slices	SPM8, VBM8 toolbox, smoothing with 8-mm3 FWHM Gaussian kernel
Boukaert etal., 2016 (20)	28	MDD (DSM-IV)	RUL, bitemporal	71.9 ± 7.8	11.2 ± 4	8 M (29%), 20 F (71%)	MADRS (pre- ECT: 35.04 ± 6.88, post- ECT: 9.61 ± 11.47) Mild depression	MMSE (pre-ECT: 23.64 ± 4.54, post-ECT: 26.32 ± 3.44)	3T Philips Intera MPRAGE, TR = 9.6 s, TE = 4.6 ms, flip angle = 8°, voxel size = 0.98 × 0.98 × 1.2 mm3, slice thickness = 1.2 mm, 182 slices	SPM8, VBM8 toolbox, smoothing with 8-mm3 FWHM Gaussian Kernel
Cano et al., 2019 (21)	24	MDD (DSM-IV)	12 RUL 12 bifrontotemporal (2 cohorts)	RUL: 42.25 ± 15.78 Bifrontotemporal: 59.17 ± 8.02	RUL: 10.33 ± 2.42 Bifrontotemporal: 11.08 ± 1.5	RUL cohort: 6 M (50%), 6 F (50%) Bifrontotemporal cohort: 6 M (50%), 6 F (50%)	RUL cohort: QIDS (pre- ECT: 17.42 ± 3.34, post- ECT: 11 ± 4.90) Mild depression Bifrontotemporal cohort: HAMD-21 (pre-ECT: 31.25 ± 9.21, post-ECT: 2.92 ± 2.54) Not depressed	Not reported	RUL cohort: Siemens Skyra 3.0T magnet scanner TR = 2530 ms; multiecho time = 1.69, 3.55, 5.41, and 7.27 ms; flip angle = 7°; field of view = 256 × 256 mm; matrix size = 256 × 256 pixels; in-plane resolution = 1 × 1 mm2; slice thickness = 1 mm; 156 slices Bifrontotemporal cohort: Philips Achieva 3.0T magnet scanner TR = 8.1 ms, TE = 3.7 ms, flip angle = 8°, field of view = 240 × 240 mm, matrix size = 256 × 256 pixels, in-plane resolution = 0.94 × 0.94 mm2, slice thickness = 1 mm, 160 slices	SPM12, smoothing with 6-mm FWHM isotropic Gaussian kernel
Dukart et al., 2014 (1)	10	5 unipolar MDD, 5 bipolar MDD—4 initially manic (DSM-IV)	RUL	53.9 ± 10.7	Not reported	4 M (40%), 6 F (60%)	HAMD-17 (pre- ECT: 21.8 ± 5.9, “significant improvement” after ECT) YMRS (pre-ECT: 25.0 only obtained for 1 patient) N/A Not reported	Not reported	1.5T Magnetom VISION scanner (Siemens) MPRAGE, TR = 11.4 ms, TE = 4.4 ms, flip angle = 30°, field of view = 269 mm, slice thickness = mm, pixel size = × 1.05 mm, slab thickness = 161 mm, matrix size = 256 × 256, 154 slices	SPM8, smoothing with 8-mm FWHM Gaussian kernel
Nickl-Jockschat et al., 2016 (22)	21	MDD (ICD-10: F32.1, F32.2, F32.3, F33.1, F33.2, F33.3)	RUL; in 12, switch to left anterior right temporal	55.5 ± 9.9	10.4	15 M (71%), 6 F (29%)	HAMD-17 (pre- ECT: 25.4 ± 5.1, post-ECT: 10.7 ± 6.5) Mild depression	Not reported	3T Tim Trio (Siemens) TR = 1900 ms, TE = 2.52 ms, flip angle = 9°, 1.00 mm isotropic resolution, image matrix = 256 × 256 × 176	SPM8, VBM8 toolbox, smoothing with 8-mm FWHM
Ota et al., 2015 (23)	15	MDD (DSM-IV)	Bilateral	52.17 ± 14.4	9 ± 2.3	9 M (60%), 6 F (40%)	HAMD-21 (pre- ECT: 19.7 ± 2.1, post-ECT: vs. 8.5 ± 6.5) Mild depression	Not reported	Magnetom Symphony 1.5T (Siemens) TR = 1580 ms, TE = 2.64 ms, flip angle = 15°, field of view = 256 × 315 mm2, slice thickness = 1.23 mm, slab thickness = 177 mm, matrix = 208 × 256, field of view = 256 × 315 mm2, 144 slices	SPM8, Easy Volume toolbox, smoothing with 8-mm FWHM Gaussian kernel
Redlich et al., 2016 (24)	23	MDD (DSM-IV)	Unilateral; in 3, switch to bilateral	45.7 ± 9.8	14 ± 3.8	9 M (39%), 14 F (61%)	HAMD-17 (pre- ECT: 26.0 ± 6.5, post-ECT: 13.1 ± 7.1) BDI (pre-ECT: 30.9 ± 10.2, post-ECT: 22.4 ± 10.4) Mild depression	Not reported	Gyroscan Intera 3T (Philips) 3D fast gradient-echo sequence (TFE), TR = 7.4 ms, TE = 3.4 ms, flip angle = 9°, 2 signal averages, inversion prepulse every 814.5 ms, acquired over a field of view of 256 × 204 × 160 mm, voxel size = 0.5 × 0.5 × 0.5 mm	SPM8, VBM8 toolbox, smoothing with Gaussian kernel of 8 mm FWHW
Sartorius et al., 2016 (25)	18	MDD (DSM-IV)	RUL	52 ± 14	11.3 ± 4.8	9 M (50%), 9 F (50%)	HAMD-21 (pre- ECT: 31.8 ± 8.2, post-ECT: 10.6 ± 7.3) Mild depression	MMSE (pre-ECT: ± 1.8, post-ECT: ± 4.75)	3T Tim Trio (Siemens) MPRAGE, TR = 1570 ms, TE = 2.75 ms, flip angle = 15°, voxel size =1 × 1 × 1 mm3, 256 mm field of view, 192 sagittal slices	SPM12, VBM8 toolbox, smoothing with 8-mm3 isotropic Gaussian kernel
Sartorius et al., 2019 (15)	92	MDD (DSM-IV)	69 RUL, 12 bilateral, 3 different electrode positions, 8 switch in electrode position	50 ± 12	12 ± 4.1	50 M (54%), 42 F (46%)	HAMD-21 (pre- ECT: 26.1 ± 7.0, post-ECT: 11.4 ± 7.4) Mild depression	MMSE (pre-ECT: ± 1.9, post-ECT: ± 3.5) Initial MMSE for 37, final MMSE for 35 patients reported	Cohorts from Mannheim, Heidelberg, Aachen: 3T Tim Trio (Siemens), MPRAGE, 192 sagittal slices, field of view = 256 mm, voxel size = 1 × 1 × 1 mm Münster: 3T (Intera), TFE, field of view = 256 × 204 × 160, voxel size = 0.5 × 0.5 × 0.5 mm Göttingen: MR: MPRAGE, field of view = 256 mm, voxel size = 1 × 1 × 1 mm	SPM12, VBM8 toolbox, smoothing with 8-mm kernel
Thomann et al., 2017 (26)	21	12 MDD, 9 SZ (DSM-IV)	RUL	SZ: 34.2 ± 12.3 MDD: 46.3 ± 11.3	11.3 ± 1.7	10 M (48%), 11 F (52%)	PANSS (sum) (pre-ECT min/ max: 60/119, post-ECT min/ max: 44/87) HAMD-17 (pre- ECT min/max: 7/38, post-ECT min/max 1/15) Not depressed to moderate	Not reported	3T TIM Trio (Siemens) MPRAGE, TR = 1.57 s, TE = 2.74 ms, flip angle = 15°, voxel size = 1 mm3 (image matrix = 256 × 256 × 192)	SPM12, VBM8 toolbox, smoothing with isotropic Gaussian kernel with FWHM of 8 mm
Wang et al., 2017 (27)	23	MDD (DSM-IV)	Bifrontal	38.74 ± 11.02	7.26 ± 3	11 M (48%), 12 F (52%)	HAMD-17 (pre- ECT: 22.22 ± 4.74, post- ECT: 3.83 ± 2.15) Not depressed	Not reported	3.0T whole-body MRI system (Signa HDxt, GE Healthcare) 3D inversion recovery prepared fast spoiled gradient recalled sequence, TR = 8.676 ms, TE = 3.184 ms, flip angle = 8°, voxel size = 1 × 1 × 1 mm3, field of view = 256 × 256 mm2, matrix size = 256 × 256, slice thickness = 1 mm, 188 sections	SPM8, VBM8 toolbox, smoothing with Gaussian kernels with different FWHM (FWHM 4, 6, 8, and 10 mm) to smooth the normalized gray matter images for the subsequent statistical analyses
Xu et al., 2019 (28)	11 (only 8 all MRIs)	MDD (DSM-IV)	Bilaterally on the forehead	39.27 ± 7.84	7.9 ± 1.2	5 M (45%), 6 F (55%)	HAMD-17 (pre- ECT: 23.27 ± 4.11, post- ECT: 4.13 ± 1.83) Not depressed	Not reported	MAGNETOM Skyra (Siemens) Rapid gradient-echo sequence, TR =1600 ms, TE = 2.6 ms, flip angle = 9°, field of view = 256 mm × 224 mm, 176 slices, matrix = 256 × 224, in-plane resolution = 1 × 1 mm2	SPM12, VBM toolbox, smoothing with Gaussian kernel with 8-mm FWHM

We performed PubMed and Google Scholar searches for structural MRI studies involving ECT. Additional studies were determined by reference tracking and from reviews. We applied the following inclusion criteria: 1) human studies; 2) whole-brain comparison of gray matter volumes pre- and post-ECT; 3) published until August 5, 2020, in original peer-reviewed journals; and 4) peak coordinates of the activation foci available for all reported clusters in Montreal Neurological Institute or Talairach stereotactic space. If two studies used nearly identical patient populations, we excluded one. Based on the PubMed and Google Scholar searches as well as the reference tracking, we screened a total of 11,916 articles. We had to exclude 11,904 articles. A total of 12 studies met the inclusion criteria. In two of those studies, whole-brain analysis yielded no significant result (17,20). The vast majority of studies included patients with MDD according to DSM-IV or ICD-10 criteria. However, there were two studies that included patients who had schizophrenia (24) or were initially manic (1).

3D, 3-dimensional; BDI, Beck Depression Inventory; ECT, electroconvulsive therapy; F, female; FWHM, full width at half maximum; HAMD, Hamilton Depression Rating Scale; M, male; MADRS, Montgomery–Åsberg Depression Rating Scale; MDD, major depressive disorder; MMSE, Mini-Mental State Exam; MPRAGE, magnetization prepared rapid acquisition gradient-echo; MRI, magnetic resonance imaging; PANSS, Positive and Negative Syndrome Scale; QIDS, Quick Inventory of Depressive Symptomatology; RUL, right unilateral; SZ, schizophrenia; TE, echo time; TFE, turbo field echo; TR, repetition time; YMRS, Young Mania Rating Scale.

Our main analysis yielded two clusters. Only contrasts indicating regional brain volume increases contributed to these clusters. The larger cluster was located in the right medial temporal lobe (18, −6, −18) comprising mainly the amygdala, with a smaller portion extending into the right hippocampus and basal forebrain. A smaller left hemispheric cluster (−14, −4, −20) was also located in the amygdala, while smaller parts extended into the parahippocampal gyrus, hippocampus, and entorhinal cortex (Figure 1A). The right cluster was mainly driven by studies employing exclusively right unilateral stimulations (these studies contributed 49% to this cluster), with smaller contributions from experiments with only bilateral (20%) and mixed (right unilateral and bilateral) stimulation (31%). Conversely, the majority of contributions to the left hemispheric cluster came from experiments with bilateral (40%) and mixed stimulations (38%), while experiments with right unilateral stimulations (22%) played a minor role. The leave-one-out analysis corroborated the bilateral spatial convergence in the amygdala and hippocampus. Results were more robust for the right cluster (100% convergence across the leave-one-out iterations for large parts of the cluster) than for the left one (80%–90% convergence) (Figure 1B). Exploratory analyses did not suggest an influence of affective response, age, gender, or number of treatments on brain structural changes, while too low power did not allow a similar analysis for cognitive impairment.

Figure 1.

Brain volume increases in the course of electroconvulsive therapy. (A) Our meta-analysis found convergent evidence for brain volume increases in the medial temporal lobes of both hemispheres. The larger cluster was located in the right medial temporal lobe (18, −6, −18; z score = 6.99) comprising mainly the amygdala and—to a lesser degree—the hippocampus. A smaller left hemispheric cluster (−14, −4, −20; z value = 4.84) was also located in these two medial temporal lobe structures. Results are displayed at a cluster-level familywise error cluster-corrected threshold of p(familywise error) < .05 (cluster-forming-threshold at a p < .001). (B) Our leave-one-out analysis showed bilateral spatial convergence in the amygdala and hippocampus. Results were more robust for the right cluster (100% peak convergence across the leave-one-out iterations) than for the left (80%–90% peak convergence). Results are displayed at a cluster-level familywise error–corrected threshold of p < .05.

To our knowledge, this is the first observer-independent coordinate-based meta-analysis of magnetic resonance imaging studies on brain structural changes during ECT. Our results provide robust evidence for a volume increase in the medial temporal lobes during ECT across different samples, sites, and analysis pipelines. This is in line with the results of previous original studies and mega-analyses in the field (1,8–10) and further highlights the role of the amygdala that, common to the contrary perception (8), was found to be more affected in our study than the hippocampus. The amygdala is a central hub within an emotion-processing network, while the hippocampus plays a key role in memory formation (10). Consequently, the observed brain volume increases are located in brain regions that are plausibly related to the therapeutic effect of ECT (11) and the main side effects. The amygdala closely relates to depressive (12) and psychotic (13) symptoms, while the hippocampus appears as linked to mnestic impairments (14). A straightforward interpretation of these findings as macroscopic correlates of neuroplastic effects, however, is challenged by ambiguous results regarding an association of regional brain volume changes and clinical response (9,15,16). Dysregulated ion intake during seizures with subsequent increase of neuronal volumes could also be a contributing factor (17). The strong contribution of electrode position might indicate a significant role also of the electrical field (18).

In sum, our findings corroborate brain volume increases in the course of ECT, in particular in two brain regions closely related to the therapeutic and the main side effects of ECT (amygdala and hippocampus). Electrode position seems to significantly influence the laterality of these findings. Once more studies are available for inclusion, future meta-analysis will have to focus on potential brain structural correlates of the main and side effects, as well as on the influence of clinical and treatment parameters.