More than 70 years after its inception, no consensus has been reached on the mode of action of electroconvulsive treatment (ECT). We suggest that two features may guide the delineation of the involved mechanisms. First, the changes induced by ECT upon the brain systems that typically are altered in its primary recipient group, people with severe depression, and, second, the duration of clinical effects as measured by lowered scores on depression instruments.

Severe depression is characterized by pathophysiological heterogeneity; however, changes in four brain regions or systems are typical. First, the frontal lobes, with compromised activity in dorsal regions that include the anterior cingulate and lateral prefrontal cortical (PFC) areas, as well as altered processing of emotional stimuli in ventral and orbital frontal cortex (1, 2). Volumetric reductions have been described in dorsolateral, orbitofrontal, and ventral frontal cortex, including the subgenual anterior cingulate (3). Second, temporal lobe changes are seen in depression, including volume reductions, and functional alterations in the hippocampus and parahippocampal gyrus and hyperactivity of the amygdala (4, 5). Recent studies indicate that the structural changes in frontotemporal regions are associated with altered functional connectivity (FC), often increases, in ultraslow frequency bands (<0.025 Hz) (6). Third, altered structure and functioning is seen in the hypothalamic-pituitary-adrenal (HPA) stress axis, including hypersensitivity to stressors, chronically elevated levels of stress hormones and impaired feedback regulation by frontotemporal structures (7). Fourth, etiologically, the HPA axis alterations are consistent with stress-exposure as a significant factor in severe depression (7). Stress activates the mesocorticolimbic dopamine system as well as the HPA axis. Since dopamine is central to several functions that are impaired in depression, including pleasure, motivation, and concentration, dopamine function is also thought to be impaired in this disorder (8). The last decade, increased focus has been on the role of glutamate-GABA processes in the neurobiological changes seen in depression (9). We hypothesize that antidepressive effects of ECT reflect the impact of the intervention upon these systems that are altered in depression.

A temporal framework in the search for mechanisms of ECT is provided by the immediate remission rates and degree of enduring clinical effects and relapse rates in patients with severe depression. Remission rates for severely depressed patients with and without previous pharmacotherapy failure were reported at 48.0 and 64.9% respectively in a recent meta-analysis of seven studies (10). Another recent meta-analysis of six studies reported an overall remission rate of 50.9% in patients with severe depression (11) and a review of seven studies that compared ECT with transcranial magnetic stimulation in these patients reported a remission rate of 52% for ECT (12). Overall relapse rates in ECT responders tend to be high, reported as 15–20% 1 week after treatment and 50–80% after 6 months, commonly leading to continuation ECT and post-treatment pharmacotherapy (13, 14). Based on the remission and relapse rates, after 6 months, between 10 and 35% of patients can be expected to exhibit enduring beneficial effects from an ECT series. This estimate, however, is based on studies where ECT was not compared to simulated ECT (SECT). A consequence is the possible overestimation of effects that are actually due to the induced seizures rather than to non-specific or placebo-effects. Reviews of studies that compared ECT with SECT indicate only small immediate benefits for ECT and no difference from 1 month and onward post-treatment (15–17). In the SECT-controlled studies, the score reduction on rating scales in the SECT groups was typically between 25 and 50%, indicating the influence of placebo-effects, the procedures provided during an ECT course and spontaneous recovery from depression (16). Based on the remission and relapse rates and the placebo findings, effects induced by the seizures would be expected to typically endure for weeks to a few months, suggesting a similar temporal framework for the underlying mechanisms.

Our aim is to describe and integrate, in terms of a hypothesis, the available evidence on the effects of ECT upon frontal and temporal lobe activity, the HPA stress axis, the dopamine system, and glutamate/GABA functioning, and the association between these effects and changes in depression. Using keywords and combined-word strategies, multiple computer searchers were conducted in PubMed, Google Scholar, and Medline for relevant articles published in English from 1939 to June 2013.

The impact of ECT upon frontotemporal activity has been investigated by both electroencephalography (EEG) and brain imaging techniques, in recent decades mainly by positron emission tomography (PET) and single photon emission computed tomography (SPECT), but also by functional magnetic resonance imaging (fMRI). We focus on studies that reported changes in brain activity level from before to after ECT as evidenced by increases (which we term activation) and decreases (deactivation) in the EEG power spectrum and cerebral blood flow. We also briefly summarize findings from recent fMRI studies that assessed functional network connectivity following ECT.

An ECT series strongly and repetitively activates the HPA stress axis. Measures of hormone activity in human recipients consistently show the significant increase after ECT in adrenocorticotrophin (ACTH), cortisol, and arginin vasopressin in the blood and saliva (58–61). The increases are abrupt and appear to normalize to baseline levels within 1 h post-treatment in the human studies (61, 62). More invasive rodent studies indicate a longer lasting increase in ACTH, with sustained elevated levels at least 24 h post-treatment (63). The increased ACTH and cortisol levels are more pronounced with high as compared to low intensity stimulation, particularly with the degree to which stimulus intensity exceeds the individual seizure threshold (61). In contrast, the level of hypothalamic corticotrophin releasing hormone (CRH) appears to be less affected by ECT, although rodent studies indicate an increase in CHR mRNA levels in the parvocellular area of the paraventricular nucleus at the heart of the HPA axis (60, 64, 65).

Takano et al. (34), using PET, observed a significant activation effect in the hypothalamus during ECT induced seizures as compared to baseline. Studies in rodents have demonstrated that the hypothalamic activation includes the paraventricular nucleus (66). The paraventricular nucleus activation is sufficiently strong to be accompanied by angiogenetic growth processes, including endothelial cell proliferation in the vasculature (66). The endothelian cell proliferation is particularly strong in a subset of hypothalamic nuclei that includes the parvocellular part of the paraventricular nucleus which produces CRH. The parvocellular angiogenesis correlates with the neural activation level caused by ECT and is a likely response to increased metabolic demand (66).

The dexamethasone challenge test has been used in an array of studies to investigate changes in HPA repression status after ECT (67, 68). The majority of studies have found a reduced cortisol response to dexamethasone in a proportion of patients after as compared to before treatment, by most authors interpreted as the regain of HPA suppression capabilities by the forebrain and normalization of functioning (67, 68).

Human studies indicate that ECT leads to dopamine system activation. Rudorfer et al. (69) reported an increased level of homovanillic acid (HVA), a measure of dopamine turn-over, in the cerebrospinal fluid (CRF) following ECT. Nikisch and Mathe (70) similarly reported a 60–70% increase of HVA in the CRF after a completed ECT series. That ECT triggers the dopamine system also is indicated by a human PET study (49). Although finding reduced activation in all assessed brain regions after as compared to before ECT, these researchers observed a relatively less decrease (that is, a relative increase) in regions with known dopaminergic innervation – the caudate nucleus and substantia nigra of the upper brain stem. This is consistent with a study of rhesus monkeys in which dopamine activity in the striatum was found to be increased by ECT and to last for 10 days following treatment end (71).

In rodents, repeated electroconvulsive stimulation (ECS) has consistently been found to enhance brain dopamine function (72). Stenfors et al. (73) found 30% increased dopamine concentrations in the frontal cortex and occipital cortex following six ECS treatments over 2 weeks. West and Weiss (74) reported that ECS in rats increased both spontaneous and bursting dopamine neuron activity in the ventral tegmentum. Moreover, a higher dose of ECS in rats (e.g., bilateral rather than unilateral) has been observed to lead to an increased dopamine release (75).

The impact upon the dopamine system is corroborated by findings of dopamine receptor changes. This includes upregulation of D1 and D3 receptors in the striatum as evidenced in preclinical studies (76, 77) and the possible decrease in D2 receptor binding in the rostral anterior cingulate in humans (78).

Several studies, primarily those using magnetic resonance spectroscopy, have indicated reduced levels of glutamate and glutamine (the glial cell version of glutamate) in the anterior cingulate, dorsolateral PFC, and dorsal and ventral medial PFC in depression (79–84). Changes in glutamate/glutamine appear to vary with severity and course of illness progression, with a negative correlation indicated between frontal cortical glutamate levels and depression severity level (80) and with decreased levels in later phases as compared to early phases in regions such as the ventromedial PFC (including the ACC) and the hippocampus (85, 86). Moreover, GABA levels may be reduced in frontal regions such as the dorsomedial/dorsal anterolateral PFC in depression, although the findings are not consistent (82, 87, 88). Investigations of plasma and CRF levels have found reduced GABA levels in acutely depressed patients (89).

Two human studies reported increased glutamate levels in dorsolateral PFC and anterior cingulate shortly after an ECT series (80, 90), while a third found no changes in the frontal cortex and anterior cingulate following a mean of 20 ECTs (81). Conflicting results also arise from preclinical studies, with increased and decreased glutamate levels in the hippocampus shortly after ECS (91, 92). Regarding GABA, Esel et al. (93) reported increased serum levels in a human study acutely following an ECT series. In rodents, ECS has been reported to increase GABA levels in a range of brain regions, with the nucleus accumbens as one possible exception (94). Increased GABA following ECT would be in line with increased glutamate levels since glutamate is the precursor of the formation of GABA (92). Since glutamate and GABA exerts their functional effects in conjunction, a more informative picture may result from studies that have assessed effects of ECS upon the glutamate to GABA ratio. Two studies using rat models of depression found that ECS down-regulated the glutamate to GABA ratio in the hippocampus (95, 96). A decreased glutamate to GABA ratio by ECT would be in line with the anticonvulsant theory of Sackeim et al. (97) where a relatively increased GABAergic tone is postulated.

The literature we have reviewed indicates, first, that ECT reduces activity in the frontal cortex and temporal lobes for weeks to months, indicated by EEG-deceleration with increased theta and delta wave activity and findings from brain imaging studies of functional deactivation. Second, ECT acutely activates the HPA stress axis, an activation that is sufficiently strong to be accompanied by structural changes in the command center of the paraventricular hypothalamic nucleus. Third, the mesocorticolimbic dopamine system is activated, which may last at least for several days. Fourth, ECT impacts upon glutamate-GABA processes in frontotemporal networks, including acute activations and a reduced glutamate to GABA ratio. That ECT has these effects is in line with previously suggested models focusing on frontotemporal deficits (26, 45), anticonvulsive effects (98), and diencephalon processes (99, 100).

Based on its pattern of brain effects, we suggest that ECT can be conceptualized and understood as severe stress or brain trauma. The most direct support for this conceptualization is the strong activation of the HPA stress axis. Also consistent with this notion is the activation of the dopamine system by ECT since dopamine system engagement is a typical correlate to HPA activation during stress (101). Moreover, the impaired frontotemporal functioning after ECT can be understood as a stress effect. ECT induced frontotemporal impairment may result both from direct cortical effects of the current and from activation of the HPA axis. As for the latter, highly elevated, enduring, or repetitive HPA axis activation and stress hormone secretion generally are associated with the dismantling of frontotemporal activity, including the deactivation of frontal cortical and hippocampal regions and disruption of memory and its underlying component processes such as long term potentiation (LTP) (7, 102–104). The association between HPA axis activation, cognitive deficits, and frontotemporal changes is seen also within groups of patients with depression (7). This is consistent with reports of a correlation between learning impairment and maximum post-ECT cortisol and ACTH levels, and with findings that ECT disrupts hippocampal LTP (58, 105–108).

Changes in glutamate and GABA levels following ECT may be in line with those seen after stress and trauma. Severe psychosocial stress and glucocorticoid administration acutely increase glutamate and GABA levels in frontotemporal regions, which is consistent with similar increases reported from several but not all ECT studies (109, 110). Chronic stress, in contrast, is indicated to reduce glutamate and GABA levels over time, in line with the reductions seen in depression (86). Informed by the stress literature, a differentiation between acute and enduring effects may help explain the conflicting reports of ECT effects upon these amino acids. At the same time, a central component of the changes that ECT induce may be down-regulating of the glutamate to GABA ratio, leading to increased inhibitory tone, cortical silencing, and reduced cortical excitability (92, 111). This type of change may give rise to the increased threshold for convulsions that Sackeim (98) has hypothesized to result from ECT. Notably, rodent studies indicate that also repeated stress increases the convulsion threshold (112). Finally, glutamate – GABA functioning is central to FC, and changes in FC by ECT may be consistent with a stress or trauma impact. For example, mild brain injury is typically associated with reduced FC (113, 114).

The notion that effects of ECT can be equaled to severe stress or trauma may seem to contrast with the oft suggested view that the intervention reestablishes both normal HPA axis suppression and hippocampal structure and functionality, the latter through dendate gyrus neurogenesis and cell proliferation (115, 116). However, also these types of changes may be consistent with the impact of trauma. First, a reduced cortisol response to dexamethasone following ECT may not reflect the regain of suppressor status but instead a progression of HPA axis dysfunction. Interpretation of results from the dexamethasone challenge test is problematic since the test here is given within the context of a treatment that significantly activates the HPA axis. A part of this is the profound changes induced by ECT in the paraventricular hypothalamus. Both this and the general administration of dexamethasone in the HPA activating context of ECT may contribute to lowered post dexamethasone cortisol levels, providing alternative explanations than restored suppressor status (117). It is of particular concern that chronic stressors may move the HPA axis from an over-responsive system to one that becomes under-responsive or non-responsive, in line with exhaustion (118, 119). Second, psychosocial stress is generally found to inhibit hippocampal neurogenesis, and enriched environments increase it. Neurogenesis, however, also is a typical consequence of brain trauma such as ischemia, hemorrhage, and, of particular relevance, status epilepticus (120–122). Indeed, the cascade of vascular growth, neurogenesis, and cell proliferation that is seen in the hippocampus after ECT also are commonly observed after brain insults. These may be restorative processes that protect against neuronal loss following trauma (120, 122).

To sum up, ECT seems to share several features with those of severe psychosocial stress, including acute effects such as HPA axis and dopamine system activation and subsequent effects that include reduced frontotemporal activation, blocked LTP, and impaired memory, in addition to possibly exhausted HPA axis regulation. Moreover, ECT may share several effects with brain insult as well, and we have noted vascular growth, neurogenesis, and cell proliferation. Hence, we suggest that ECT represents a particular, perhaps unique type of severe stressor or trauma (Table 1).

We suggest that the temporarily improved scores on depression instruments following ECT reflect the combination of frontal and temporal lobe functional impairments and activation of the HPA axis and the mesocorticolimbic dopamine system. These effects as well as other detailed changes observed in structures such as the hippocampus appear consistent with those typically seen after severe stress-exposure and/or brain trauma. Hence, we conjecture that central to the effect mechanisms of ECT is the impact upon the brain in a manner that is consistent with a unique type of severe stress-exposure or trauma.

To test our hypothesis of mechanisms of action, future animal and human studies may focus on the association between improved scores on depression instruments and changes in the component processes that we have delineated. Since each of the component processes may show considerable individual variability, and since the effect of changes in one process may depend on changes in the other processes, measuring changes in several components within the same study may be particularly fruitful. The notion of trauma and stress oriented effects by ECT also may be tested by investigating whether other types of acute and enduring changes than those we have addressed here and that typically are associated with stress and trauma are seen following ECT. A reverse strategy naturally is to attempt to block the various component processes that we have outlined and to investigate the consequence for therapeutic response, a strategy that already is implied in the use of alternative physical treatments such as transcranial magnetic stimulation. In addition, since studies that have compared ECT to SECT indicate the involvement of substantial placebo-effects, future studies should attempt to control for and partition out placebo-effects in order to reduce noise in the data. Psychological factors such as patient beliefs that ECT will help may interact with neurobiological changes instigated by the treatment to determine mood effects.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.