Clinical and biological markers of electroconvulsive therapy effectiveness: a narrative review

Abstract

Electroconvulsive therapy (ECT) is the most effective treatment for several particularly severe or pharmacotherapy-resistant psychiatric disorders, and a growing number of studies have investigated factors influencing the effectiveness of ECT. The objective of this article is to review the current evidence on clinical and biological markers potentially related to response to ECT and to outline perspectives for future research. In depressive disorders, the presence of clinical characteristics such as higher age, psychotic and psychomotor symptoms, and the absence of comorbid personality disorders are associated with a particularly good response to ECT. However, these clinical factors alone explain only a part of the variance of treatment outcome. Biomarkers at the genetic/epigenetic, immune inflammatory, and brain imaging levels are now providing promising and, in part, converging findings on both the mechanism of action and predictors of response to ECT. Taken together, these findings suggest a close relationship between specific clinical symptoms, the immune inflammatory, and neuroplastic mechanisms of ECT. While genetic studies consistently indicate a higher genetic risk load in patients referred for ECT, findings regarding the predictive value of polygenic risk scores and also epigenetic markers for ECT response remain inconsistent. In the future, combining clinical features and biomarkers could help define subgroups of psychiatric disorders with distinct pathophysiology and reliably predict treatment outcomes at the individual patient level. For now, it may be assumed that different and possibly age-dependent prototypical forms of depressive disorders exist, which are characterized by specific clinical (psychotic and psychomotor symptoms) and inflammatory markers (higher levels of IL-6 and CRP) and show a differential response to treatment modalities.

Introduction

The search for reliable predictors of response to electroconvulsive therapy (ECT) is almost as old as the treatment itself. As ECT is still the most effective treatment for several severe clinical conditions [1] and yet largely underutilized [2], reliable prediction of treatment response could enable more timely access to ECT for those patients who will most likely benefit and avoid unnecessary treatments for those who will not. While research initially focused on clinical factors, a multitude of different methods has evolved during the last decades that allowed for a deeper investigation of the biological correlates of psychiatric disorders and the corresponding treatments. Biomarkers are commonly categorized as predictors, moderators, or mediators (Table 1) [3, 4]. Each category conveys distinct information: predictors are associated with overall outcomes, moderators indicate differential responses to treatments, and mediators reflect the underlying mechanisms of treatment effects. In this context, neuroimaging techniques and molecular biology approaches, including genetic and epigenetic methods, have substantially advanced our understanding of the pathophysiology of neuropsychiatric disorders, factors influencing treatment response, and longitudinal changes in biomarkers that provide insight into the mechanisms of action of ECT. Despite these considerable advances, a comprehensive integration of findings across methods, as well as their effective translation into clinical practice, remains lacking. In this review, we present the current state of research on factors influencing the effectiveness of ECT at various levels: clinical, (epi-)genetics, immune inflammatory pathway, and neuroimaging. We will focus primarily on findings related to depressive disorders, as this condition has been studied most extensively.

Table 1.

Types of biomarkers.

	Predictor	Moderator	Mediator
Time of Measurement	Before (or during) treatment	Before (or during) treatment	During or after treatment
Information Provided	Estimates overall treatment outcomes	Identifies differential response between treatments	Explains underlying mechanisms of treatment effects
Relevant Clinical Questions	“How likely is this patient to respond to ECT?” (prognosis)	“Which treatment is more appropriate for this patient–ECT or ketamine?” (treatment selection)	“How and why does ECT work?”(mechanistic insight)
Examples	Baseline CRP and IL-6 [81] Baseline levels of miR-223-3p [28]	CRP and differential response to antidepressant treatments [80, 84]	Volume and connectivity changes in the hippocampus and dentate gyrus [111, 113] DNA methylation and microRNA changes [59]

Clinical factors and effectiveness of ECT

Publications on clinical features and their influence on the effectiveness of ECT date back to the 1950s [5]. One early study showed that patients with endogenous depression (defined as the absence of relevant precipitating factors) responded better to ECT than patients with reactive depression. Depression severity did not differ between the two groups [6]. According to a retrospective study with computerized selection and weighting of multiple clinical features [7], endogenous features such as early morning wakening, psychomotor inhibition and delusions were positively associated with response, whereas neuroticism was negatively associated.

Meta-analyses and clinical studies

In recent years, meta-analyses have identified clinical factors that show an association with the effectiveness of ECT. A meta-analysis [8] investigated the impact of a previous non-response to pharmacotherapy and found a significantly lower remission rate (48 versus 64.9%) in patients with pharmacotherapy resistance (odds ratio (OR) 0.52). As this finding was largely based on observational studies, the authors pointed out that this effect could be influenced by confounding factors such as psychotic symptoms or episode duration. A subsequent meta-analysis [9] examined 15 clinical characteristics and found that ECT response was negatively associated with longer episode duration (standardized mean difference (SMD) -0.37), medication failure (OR 0.57). Weak associations with notable heterogeneity were found for older age and psychotic symptoms. Symptom severity and melancholic characteristics, on the other hand, showed no clear associations. Another meta-analysis [10] confirmed the positive influence of older age (SMD 0.35 for response, 0.26 for remission) and psychotic symptoms (OR 1.69 for response, 1.47 for remission). Additionally, severity of depression predicted response but not remission.

Recent clinical trials on the efficacy of intravenous ketamine versus ECT also investigated the influence of clinical characteristics on response. In a secondary analysis of the ELEKT-D study [11] patients with milder symptoms on a self-report scale (QIDS-SR) showed greater symptom reduction with ketamine vs. ECT, whereas patients with more severe symptoms had a faster response to ECT in the early course of treatment when compared to those receiving ketamine [12]. Patients with psychotic symptoms were excluded from the study, which may have biased the results [13, 14]. The KetECT study [15] included patients with psychotic symptoms and found a significant superiority of ECT. Interestingly, patients over the age of 50 showed a significantly higher remission rate and symptom reduction with ECT compared to ketamine, while there was no significant difference in patients under 50.

Other clinical characteristics have also been identified in individual studies and meta-analyses as potential factors influencing the response to ECT. For reasons of space, these are listed in Table 2.

Table 2.

Summary of clinical factors influencing ECT response in depression.

Positive (= better response to ECT)	Negative (= poorer response to ECT)
Higher age [9, 10, 17]	Longer episode duration [9, 16]
Psychotic symptoms [9, 10, 16–18, 20]	Pharmacotherapy resistance [8]
Psychomotor symptoms [16–18]	Comorbid personality disorder [158–160]
Depression severity [10]	Childhood trauma / maltreatment [161, 162]
Lack of capacity to consent [22, 23]	Suicidal ideation [21]
	Non-suicidal self-injury in young females [163]

Interdependence of clinical markers of treatment response

There is evidence that clinical predictors of ECT response are not independent of each other. Sensitivity analyses in a meta-analysis [10] suggest that the predictive effect of psychotic symptoms is stronger in older patients and those with lower pharmacotherapy resistance. The effect of older age was stronger with longer episode duration. In a subsequent study, the group investigated the interdependence and the direct versus indirect effects of clinical factors on the effectiveness of ECT [16]. The presence of psychomotor and psychotic symptoms correlated strongly with the treatment outcomes. A direct correlation between age and response was not found; rather, the association between age and efficacy was mediated by the aforementioned symptoms. Furthermore, there was a strong correlation between episode duration and pharmacological treatment failure. Mediation analyses from other studies also indicate that substantial parts of the age effect can be explained by other factors (psychotic symptoms, psychomotor symptoms, shorter episode duration) [17, 18]. Another study found that the effect of psychotic symptoms no longer remains significant when age and psychomotor symptoms are included [19]. Patients with late-life depression and psychotic symptoms showed significantly higher remission rates regardless of episode duration [20]. A Swedish registry study investigated the influence of suicidal ideation on the ECT response. Although suicidality can be considered an expression of high depression severity, a negative correlation was found. This could be explained by a higher rate of personality disorders and a lower age in the group with suicidal ideation [21]. A further example of a clinical factor that may serve more as a proxy than as an independent marker of ECT response is the capacity to provide informed consent. A large retrospective study [22] and a meta-analysis [23] showed superior clinical improvement in patients with depression and lack of capacity to consent. However, patients without capacity to consent were characterized by higher age, higher symptom severity at baseline, and more frequent use of bitemporal stimulation, all factors potentially associated with ECT response.

Clinical prediction models

Based on decades of clinical experience and the available evidence, several prediction models have been developed. One study applied Network Outcome Analysis including all 17 items of the Hamilton Rating Scale for Depression (HRSD) in 161 patients and found three baseline symptoms to be predictive of remission (positive: retardation, hypochondriasis; negative: suicidality) [24]. Another group trained a Bayesian network model with predictors derived from meta-analyses and a dataset of 248 treatment courses. Their model yielded an AUC of 0.686 for predicting remission and an AUC of 0.528 for non-response in the validation set [25]. A further study used multivariate linear regression analysis in 1892 patient datasets to develop a prediction model for depression outcome post-ECT [26]. The final model reached an adjusted R² of 19%. For the secondary outcome response, optimism-adjusted C-statistic (analogous to the AUC) was 0.682. Authors from Japan reported a machine learning approach in a retrospective chart review of 177 patients [27]. Their model predicted individual patient remission based on CGI-I with 71% accuracy, and shorter duration of the episodes was identified as the most predictive features in the model.

While all previous models have been developed retrospectively, efforts are currently underway to create prospective tools that may be feasible for clinical decision-making. For example, the Göttingen Response to ECT Assessment Tool (GREAT) is under development and has shown preliminary promise in a small sample (n = 45; manuscript accepted for publication, personal communication David Zilles-Wegner). Further validation in independent and larger samples is needed before clinical implementation can be considered.

Perspective

Large, well-characterized samples and elaborate statistical methods such as network analyses will hopefully help to clarify the reciprocal relationships between the various clinical predictors in the future. For current practice, it can be provisionally assumed that different and possibly age-dependent prototypical forms of depressive disorders exist. These subtypes may be characterized by partially differential symptoms (e.g. higher prevalence of psychotic and psychomotor symptoms in older patients), comorbidities (e.g. higher rate of personality disorders, trauma, and self-harm in younger patients) and, consequently, show a differential response to ECT. As even combinations of relevant clinical factors only explain a part of the variance in treatment outcome, future studies with large samples should include multimodal predictors in parallel. Preliminary examples of such combined approaches suggest associations between global or specific depressive symptom reductions post ECT and different biomarkers like baseline miRNA expression [28], amygdala volumes [29], and decreased IL-6 levels [30].

(Epi-)Genetics

Candidate gene studies

Genetic and epigenetic alterations have been widely studied in the context of psychiatric disorders [31–33]. However, evidence specific to ECT remains comparatively limited [34]. In the tradition of candidate gene studies with a priori selected targets based on their biological properties [35, 36], some studies have evaluated the role of specific genes in the context of ECT response. E.g., Brain-Derived Neurotrophic Factor (BDNF) has been investigated in various studies against the background of the neuroplasticity hypothesis. Though higher BDNF serum levels have been reported in carriers of the Val66Met (rs6265) polymorphism, no significant association was found with ECT response [37], corroborating earlier negative findings [38]. For FKBP51, which is an Hsp90-associated co-chaperone that modulates glucocorticoid receptor sensitivity [39], increased FKBP51 mRNA expression levels in peripheral blood mononuclear cells (PBMC) were found in TRD patients who responded to ECT, along with distinct DNA methylation changes [40]. A handful of studies have assessed other candidate genes in the context of ECT. For COMT Val158Met (rs4680) and DRD2 C957T (rs6277), a significant association with better ECT response was observed in mutant alleles compared to wildtype gene carriers [38, 41–43]. In contrast, studies on APOE ε4, SLC6A4 (5-HTTLPR), and P2X7R have found no significant association with ECT response [38, 44, 45].

Polygenic risk scores

Aside from candidate genes, polygenic risk scores (PRS) have garnered increasing interest in recent years [46, 47]. A polygenic risk score is a statistical estimate of an individual’s genetic risk to a specific medical condition by combining the effects of multiple genetic variants. In a recent and large genome-wide association study (GWAS), single nucleotide polymorphism (SNP) based heritability for the narrow case definition (patients receiving ECT in the context of MDD) was estimated at 29–34%, which is substantially higher than estimates from broader MDD cohorts. For instance, data from the Psychiatric Genomics Consortium (PGC) during the same period estimated SNP-based heritability for MDD at only 6.5–8% [48, 49]. In addition, the study found a genome-wide significant association of a single SNP (rs114583506) in an intron of HLA-B (a major histocompatibility locus on chromosome 6) with the broader case definition (major depressive episode (MDE) in patients with MDD and other diagnoses) compared to controls [48]. This aligns with an earlier yet smaller study which demonstrated a higher MDD-PRS in ECT-treated inpatients compared to controls [50]. However, no significant associations were found between MDD-PRS and ECT response in that study. A Swedish register study involving 2320 patients receiving ECT for MDE found that a higher MDD-PRS was associated with less clinical improvement (OR = 0.89, 95% CI 0.82–0.96), while a greater bipolar disorder-PRS was associated with a better improvement (OR = 1.14, 95% CI 1.05–1.23) in the Clinical Global Impression–Improvement scale (CGI-I) or self-reported MADRS scores [51]. In contrast, a European multi-center study involving 266 patients with depression reported that schizophrenia-PRS but not MDD-, cross-disorder-, or antidepressant response-PRS, was significantly associated with greater symptom improvement and higher remission rates following ECT [52].

In summary, even though these studies consistently indicate a higher genetic risk load and liability in patients referred for ECT, findings regarding the predictive value of PRS for ECT response remain inconsistent. Larger, rigorously designed prospective studies are needed. In this regard, the International Consortium on the Genetics of Electroconvulsive Therapy and Severe Depressive Disorders (Gen-ECT-ic) has pioneered a prospective GWAS study aiming for 30,000 ECT patients worldwide [53].

Epigenetics

Epigenetic processes, particularly DNA methylation patterns and both quantitative and qualitative changes in microRNA (miRNA) expression, have been widely implicated in psychiatric disorders, with numerous studies reporting alterations that correlate with symptom severity and treatment response [54–56]. These findings underscore their potential as novel diagnostic and predictive biomarkers and as targets for therapeutic intervention. For miRNA, both changes in brain tissue (e.g., miRNA-1202, miRNA-132) as well as in peripheral blood (e.g., miRNA-1202), have been demonstrated in patients with depression and schizophrenia [57, 58].

Despite this promising background, few studies have examined epigenetic changes in the context of ECT. A systematic review identified nine observational studies that investigated epigenetic mechanisms in peripheral blood samples of MDD patients receiving ECT. Though limited by small sample sizes and heterogeneity regarding the outcomes, the majority of studies detected ECT induced changes in DNA methylation, making them potential mediators for response to ECT [59]. Studies found BDNF promotor hypomethylation [60], S100A10 promotor methylation [61], and differential TNKS and FKBP51 methylation to be associated with ECT response [34]. In addition, a genome-wide study suggested differential methylation between responders and nonresponders in five protein-coding genes (RNF175, RNF213, TBC1D14, TMC5, WSCD1) and three non-coding RNAs (AC018685.2, AC098617.1, CLCN3P1) [62]. Another study reported a significant correlation between increased blood mRNA expression of three genes (BDNF, ERK1, Nr3C1) and DNA methylation of multiple CpG sites in the respective genes over the course of ECT [63]. A further study in TRD patients receiving ECT reported no baseline differences in methylation between remitters and non-remitters. However, methylation differences were found for tissue-type plasminogen activator (t-PA) in peripheral immune cells (B cells, NK cells, monocytes, and T cells, without any longitudinal changes throughout the treatment course [64]. An epigenome-wide association study involving 32 TRD patients undergoing ECT found longitudinal methylation changes at different CpG sites and gene regions (e.g., in FAM20C, IQCE, RUNX3 loci) that were associated with clinical improvement, though these results did not survive correction for multiple comparisons. A subgroup analysis revealed female-specific methylated regions enriched for transcriptional regulation, growth factor signalling, and immune pathways [65].

Studies investigating miRNA changes in the peripheral blood of patients treated with ECT are still limited but have yielded promising results. One study found that plasma levels of miRNA let-7b and let-7c were lower in TRD patients compared to healthy controls. Bioinformatic analyses revealed that these miRNAs are involved in the regulation of 27 genes within the PI3K-Akt-mTOR signalling pathway. However, neither predicted ECT response, nor did their expression change over the course of ECT [66]. Another study conducted a three-part investigation in MDD: a discovery-phase deep sequencing study (n = 16), a validation study comparing patients and controls (n = 37/34), and a larger cohort study examining VEGFA expression pre-/post-ECT (n = 97 patients, 53 controls). While no overall miRNA changes were found in the discovery phase, post hoc analysis indicated longitudinal changes in the psychotic depression subgroup. In the validation cohort, miR-126-3p and miR-106a-5p were elevated at baseline and normalized post-ECT in patients with psychotic depression (n = 7). Similarly, in the larger cohort, baseline VEGFA levels were elevated in both psychotic/nonpsychotic MDD patients and decreased significantly post-ECT only in the psychotic subgroup (n = 21) [67]. Further evidence for the involvement of miRNAs in the mechanism of ECT comes from a study involving 64 patients with TRD: At baseline, expression of miRNA-223-3p was significantly lower in responders compared to non-responders and demonstrated moderate predictive capacity for ECT response (AUC = 0.76, 95% CI = 0.6–0.91) [28]. Target mining revealed that miR-223-3p primarily regulates inflammation-related biological processes. In line with this finding, the expression of related inflammatory cytokines (IL-6, IL-1β, TNFα, and NLRP3 mRNA) were significantly higher in responders, making miR-223-3p-expression and related cytokine levels potential predictors of response to ECT.

According to a study using qPCR before the first and one month after the last ECT session in 27 TRD patients, eight miRNAs were downregulated after ECT. Significant correlations between reduced levels of miRNA-324-3p and miRNA-30c-5p and reduction in depressive symptoms were reported. Secondary bioinformatic analysis identified their gene targets as vascular endothelial growth factor (VEGF) and sirtuin-1 (SIRT1), which are important regulators of neurogenesis and synaptic plasticity [68].

Even though the body of evidence on the genetic and epigenetic mechanisms underlying the effects of ECT is growing and holds great promise, its translation into reliable predictors with clinical utility remains a long way off. To bridge this gap, large-scale and methodologically improved studies are needed to advance our understanding of ECT’s molecular mechanisms. Such efforts are essential to facilitate the integration of genetic and epigenetic marker as viable diagnostic or predictive biomarkers and incorporate them into multifactorial prediction models aimed at optimizing patient care.

ECT and the immune inflammatory pathway

Depression and neuroinflammation

Subgroups of MDD appear to be related to a dysregulated immune and inflammatory response [69]. A deeper understanding of these processes would improve our understanding of both the specific pathophysiology of MDD subgroups and the mechanisms of action of treatments that modulate inflammation. External stress can be a precipitating factor of MDD, which can increase levels of circulating cytokines, both peripherally and centrally [70, 71]. Increased levels of TNFα and IL-6 have frequently been reported while the findings regarding IL-1β and IL-8 are more contradictory [71]. In the central nervous system (CNS) these cytokines appear to result from activation of the microglia, secondary to external stress factors. The concept of microglial activation has been recognized since the discovery of microglia as the source of inflammatory mediators that are not normally expressed in the CNS [72–74]. This term is also employed in relation to reactive microgliosis at the site of a lesion [75]. Following a lesion in the CNS, the microglial cells rapidly activate an immune response to counter the infection or injury, and then rapidly restore CNS homeostasis. In the case of neuroinflammation, the release of inflammatory mediators and neurotoxins activates the microglia [76] and recruits them to the site of the lesion. In return, the activated microglia secrete cytokines and/or nitric oxide (NO) [77]. The microglia also take part in the resolution of inflammation by secreting anti-inflammatory factors. The different activation states of the microglia illustrate their morphological and functional plasticity [78].

During this immune cascade, cytokines exert an effect on the regulation of monoamine neurotransmitters. Foremost among these neurotransmitters is serotonin. In response to external stress, there is an inflammatory response that activates indole-amine 2,3-dioxygenase (IDO), which degrades tryptophan (TRP) to kynurenine (KYN). KYN is metabolized to kynurenic acid (KYNA) and 3-hydroxykynurenine (3-HK), which is itself metabolized to quinolinic acid (QUIN). Because TRP is a precursor of serotonin, the levels of serotonin are therefore reduced [71]. A similar effect has been described with regard to dopamine levels [69].

In addition, these cytokines appear to play an important role within the hypothalamic-pituitary-adrenal axis. Indeed, during MDD-related stress, there is an increase in the levels of nuclear factor-kappa-B (NF-kB), which controls the expression of pro-inflammatory genes in the CNS. There is an activation cascade at the level of the hypothalamus and then the pituitary, with increases in the concentrations of corticotropin-releasing hormone, and then the adrenocorticotropic hormone. In response, there is an increase in cortisol synthesis within the adrenal cortex. The increased plasma levels of cortisol, together with the parasympathetic nervous system (vagus nerve), serve to decrease the activation of NF-kB in the periphery and hence exert an anti-inflammatory action [71]. Several studies found associations between elevated peripheral inflammatory biomarkers (such as CRP, IL-6, and TNFα) and reduced effectiveness of conventional antidepressants [79]. Conversely, elevated inflammatory markers, especially IL-6, may predict better responses to treatments like ketamine and ECT [80].

Inflammatory markers as predictors of ECT response

A recent meta-analysis [81] reported a significant association with small effect sizes between higher baseline levels of CRP and IL-6 in serum/plasma samples and improvements in depressive symptoms following ECT (mean effect size of correlation 0.21 and 0.207, respectively). In addition, lower baseline levels of KYN, 3-HK, and TRP were associated with a better response to ECT. One study described a sex-related effect in the predictive value of baseline neuroinflammatory biomarkers. Female responders had lower baseline plasma levels of IL-8 as compared to female non-responders. Furthermore, levels of IL-8 changed differentially from baseline to post-treatment in relation to responder status and sex [82]. Another study that primarily investigated miRNA-expression (PAXgene Blood RNA Tubes) found higher corresponding baseline levels of NLRP3, IL-1b, IL-6, and TNF-α mRNA in ECT-responders [28]. In partial contradiction to this, a small study with 30 patients found a correlation between response and lower baseline plasma TNF-α levels [83]. To date, no study has identified any prognostic factor of ECT response related to plasma levels of IL-10 [84–87].

Longitudinal changes associated with ECT

An intuitive hypothesis is that ECT may possess anti-inflammatory properties. There is a growing body of research investigating the longitudinal effects of ECT on immune-inflammatory pathways [88, 89]. In humans, ECT appears to trigger an acute immune activation, characterized by a transient surge in circulating pro-inflammatory cytokines [88, 90] following each ECT session. E.g., several studies have highlighted a transient (15–30 min) increase in the expression of TNF-α and IL-6 immediately following a single ECT session [90–93]. Furthermore, ECT is linked to the activation of innate immune cells in patients with major depressive disorder, including increased proliferation of granulocytes, natural killer (NK) cells, and monocytes, as well as enhanced NK cell activity in peripheral blood [94, 95]. Moreover, stimulus charge was correlated with the immediate IL-6 increase [92]. Targeted potentiation, rather than suppression, of inflammatory responses may be of therapeutic relevance to a subgroup of depressed patients [96]. This notion is in line with a recently suggested framework of the neurobiological effects of ECT, which comprises a sequence of short-term disruptive effects followed by a potentiation of neuroplasticity and, ultimately, rewiring of neural circuits [97].

A recent meta-analysis confirmed the above-mentioned immediate increase in IL-6 levels after one or two ECT sessions, followed by a decrease after the fourth session, with no difference from baseline levels at the end of the ECT series [89]. Furthermore, a trend toward decreased TNF-α levels was observed. A study investigating changes in cytokine levels during a course of ECT in elderly depressed patients found a non-significant decrease in CRP and IL-6 with small to medium effect size, although this reduction was not correlated with clinical response [98]. Another meta-analysis highlighted that increases in KYNA, AA (Anthranilic acid), and IL-8 levels during the ECT course were associated with an improvement in depressive symptoms [81].

Relationship between neuroinflammation and neuroplasticity

It is important to note that the immune system plays a central role in neuroplasticity and tissue remodeling [99]. In preclinical models, low-grade inflammation disrupts microglial function, which is also a key regulator of neuroplasticity [100, 101]. These factors impair neurogenesis and neuronal plasticity. ECT may counteract these effects by both rebalancing neuroprotective microglial activity and enhancing neurotrophic support through seizure induction and functional network stimulation [102, 103]. A recent study demonstrated that reductions of peripheral IL-6 and TNF-α correlated with increases of hippocampus volumes [85]. The authors suggested that the reduction in IL-6 may be an important factor mediating the central neuroplastic effect. The clinical relevance of hippocampal volume increases following ECT is discussed in the following section.

Neuroimaging

Neuroimaging predictors, moderators, and mediators

Neuroimaging techniques hold promise for identifying objective neurobiological markers at the whole-brain level that may be categorized into predictors, moderators, and mediators [3, 4] (see Table 1). A recent meta-analysis suggested a potential for pretreatment brain MRI features to predict MDD treatment outcomes [104]. Although we acknowledge the promise of predicting treatment response using neuroimaging methods, our review focuses on current insights into mediators, with an emphasis on findings from structural brain imaging. Evidence on mediators from structural MRI is more consistent than that from other neuroimaging modalities (e.g., functional MRI) or from studies on predictors. Furthermore, the complexity and cost of neuroimaging procedures, as well as the limited feasibility of performing MRI prior to ECT in patients who are often severely unwell, limit the widespread use of neuroimaging predictors in routine clinical practice. In contrast, identification of mediators that reflect changes occurring during treatment and may be associated with clinical outcomes can provide mechanistic insights into the pathophysiology of psychiatric disorders and inform new therapeutic strategies, such as the identification of novel brain targets for focal neuromodulation.

Neuroplastic effects of ECT on gray matter

A review article published in 1994 [105] concluded that there was “no evidence of ECT-induced (brain) structural changes”. This conclusion was framed in response to the question, “Does ECT cause brain damage?”. However, since a groundbreaking study reported ECT-induced hippocampal gray matter (GM) volume increases [106], the central research question has shifted to “Does ECT enhance brain’s neuroplasticity?” and, if so, “How do ECT-induced neuroplastic changes relate to its therapeutic mechanisms?”. The early finding of hippocampal volume increase was subsequently confirmed by meta-analyses [107, 108] and a mega-analysis using pooled data from international collaboration [109]. Further studies have identified the effects of ECT on dentate gyrus [110–114], a hippocampal subfield where neurogenesis occurs [115, 116]. Preclinical studies have consistently shown that electroconvulsive stimulation (ECS), an animal model of ECT, increases neurogenesis in the dentate gyrus in both rodents and nonhuman primates [117–120]. A post-mortem study also reported increased neuroplasticity markers in ECT-treated individuals [121]. These findings suggest that ECT-induced volume increases in the dentate gyrus may reflect enhanced neurogenesis. However, directly testing this hypothesis in humans remains challenging due to the lack of reliable in vivo measures of neurogenesis. A recent reverse translational MRI study revealed that ECS-induced MRI-detectable hippocampal volume increases was observed regardless of the presence of neurogenesis [120], indicating hippocampal volume increases may arise from neurogenesis-independent mechanisms. The study further demonstrated a significant relationship between volume increase measured by MRI and excitatory synaptic density measures by histological analysis, although other factors, such as increased dendritic arborization, microglia activity and myelination, may also contribute.

With advances in neuroimaging, in vivo assessments of brain microstructure are now feasible in human ECT research. A recent clinical study reported increased synaptic density measured by positron emission tomography imaging targeting synaptic vesicle protein 2 A in elderly depressed individuals who achieved remission following ECT [122]. Using diffusion MRI, another study found increased neurite density and dispersion in the dentate gyrus following ECT [123]. Importantly, edema is unlikely to account for the observed volume increases [124, 125], and there is no evidence of excessive cell death in either animal models [126] or human post-mortem studies [121, 127, 128]. Taken together, the current evidence suggests that ECT induces neuroplastic changes, including increased gray matter volume, most likely reflecting increased dendritic arborization and synaptic density, rather than neurogenesis or edema.

Clinical relevance of neuroimaging findings regarding ECT-induced gray matter changes

While the longitudinal effect of ECT on GM has been well characterized by multiple lines of research, the clinical relevance of ECT-induced GM changes remains inconsistent and complex. Studies focusing on the dentate gyrus have reported volume increases associated with clinical improvement [110, 111, 113, 129] or cognitive outcomes [114], but not all studies support associations with clinical effectiveness [112, 130]. These inconsistencies may be due to differences in statistical approaches—namely, non-significant results in delta-delta correlations vs. significant within-subject correlations [131]. Briefly, this discrepancy suggests that larger neuroplastic changes are not necessarily therapeutic, but recovery from depression may require an optimal degree of neuroplastic change for each individual. A preclinical study revealed that the antidepressant-like effects of ECS were abolished in mice lacking neurogenesis [119]. Therefore, neuroplastic changes in the dentate gyrus may be necessary but not sufficient for clinical improvement. Rather, subsequent downstream changes may be critical for clinical improvement. Supporting this notion, clinical outcomes were associated with both structural [132] and functional [130, 133–135] connectivity between the hippocampus and other brain regions. Additionally, changes in cognitive function following ECT have been correlated with volumetric changes in the hippocampus [136] and dentate gyrus [114]. This aligns with the concept of “memory clearance,” which suggests that modifying the strength of existing synaptic connections may impact existing memories [137]. Thus, ECT-induced neuroplastic changes may underlie both its antidepressant effects and its transient cognitive side effects.

Another important consideration is that ECT-induced GM changes are not specific to the hippocampus but occur across widespread brain regions [138], with the largest GM changes in the right medial temporal regions (Fig. 1). Critically, not all of the longitudinal changes observed were directly related to clinical effectiveness [139]. Indeed, principal component analysis has revealed that the second component was associated with clinical improvement, while the first component explaining the largest variability of longitudinal changes was associated with the number of ECT sessions [140]. These findings suggest that the most prominent longitudinal change following ECT may be driven by direct effects of electrical stimulation [141, 142] and/or induced seizures [131], and that these effects may be too large or nonspecific to identify true clinical mediators through simple linear correlation analysis. However, clinical evidence has shown that both clinical and cognitive outcomes are associated with ECT-related components (i.e., electrical charge and the quality of induced seizures) [143], suggesting that these factors should not be overlooked in analyses of mediators of ECT effectiveness.

Fig. 1. ECT-induced GM volume change.

ECT induces statistically significant GM volumes in widely distributed brain regions. However, the extent of volume increases was regionally variable, with the largest change observed in the right medial temporal regions, including the amygdala and hippocampus. The figure was reproduced from a previous publication [131]. All participants in this study received right unilateral ECT.

Neuroplastic effects of ECT on white matter

Compared to conventional structural MRI (e.g., T1-weighted imaging), diffusion MRI uses a specific MRI sequence that is more sensitive to the diffusion of water molecules and allows inferences on the organization of WM fibers. Diffusion tensor imaging (DTI) and the frequently used parameter fractional anisotropy (FA) serve as indicators of WM integrity. Whole-brain analyses of diffusion MRI have yielded mixed findings. One study reported increased FA of the anterior cingulum, forceps minor and left superior longitudinal fasciculus after ECT [144], while another found increased FA of the right splenium and left cortico-spinal tract [145]. In contrast, other studies using the same analysis framework (i.e., tract-based spatial statistics) found no significant longitudinal FA changes [146–148]. Although these studies also assessed other DTI-metrics, including mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), consistent patterns have not emerged. Another line of research targeting specific tracts have reported reduced FA and AD in the left parahippocampal cingulum [149], reduced MD, AD, and RD in hippocampus-related tracts [132], and reduced structural connectivity between the subcallosal cingulate and ventral striatum, assessed using probabilistic tractography [150]. Taken together, findings on the effects of ECT on WM remain inconsistent. These inconsistencies may be partly attributed to the limitations of DTI, particularly its sensitivity to intravoxel WM fiber complexity [151, 152]. Advanced diffusion MRI models such as diffusion kurtosis imaging [153] and neurite orientation dispersion and density imaging [123] may provide more reliable insights into the clinical relevance of ECT-related WM changes.

Summary on neuroimaging markers

In summary, neuroimaging research over the past decade has greatly advanced our understanding of the longitudinal effects of ECT on the human brain, particularly GM. However, not all ECT-induced changes are clinically relevant. Distinguishing between changes that mediate clinical improvement and those that do not is crucial. Moreover, identifying neurobiological mediators that are differentially associated with improvements in depressive symptoms vs. cognitive outcomes is essential. Such insights will pave the way for refining ECT techniques (e.g., novel electrode placements with optimized stimulation parameters) and developing novel neuromodulation approaches that maximize therapeutic efficacy while minimizing cognitive effects, ultimately leading to more personalized and optimized care for individuals with severe depression. It should be noted, however, that the complexity and cost of neuroimaging methods may limit their widespread use in routine clinical practice. Additionally, the possibility of acquiring several MRI sequences prior to ECT is often limited given the symptom severity of individuals referred for ECT.

Conclusion

Various clinical and biological markers may influence the effectiveness of ECT in depression. At the clinical level, specific symptom patterns have a significant impact on treatment outcomes. Older patients with psychotic and/or psychomotor symptoms mostly show a favorable response, while younger patients with comorbidities and longer episode duration tend to benefit less from ECT. The increasingly recognized inflammatory subtype of depressive disorders [80, 154] appears to be associated with a particularly good response to ECT. Neuroimaging studies show (transient) structural changes in various brain regions that are only in part related to the effectiveness of ECT. Nevertheless, the neuroplastic processes expressed therein may be a necessary condition for the therapeutic effect [155].

While most previous studies focused on only one domain, more recent studies and reviews have begun to elucidated potential links between the levels of symptoms, (epi-)genetics, the immune inflammatory pathway, and neuroimaging. Higher baseline levels of inflammatory factors are associated with greater depressive symptom reduction post ECT [81], psychomotor retardation (a clinical predictor of response to ECT) has been linked to peripheral inflammation [154, 156], and one recent neuroimaging study suggests that peripheral immune changes may be associated with increases in hippocampal volume following ECT [85]. Taken together, these findings suggest a close relationship between specific clinical symptoms and the immune inflammatory and neuroplastic mechanisms of ECT [157]. In line with this, recent studies provided preliminary evidence for associations between global or specific depressive symptom reductions post ECT and baseline miRNA expression [28], amygdala volumes [29], and decreased IL-6 levels [30]. However, a uniform model that can resolve the inconsistencies in the results is still lacking, most likely due to the heterogeneous samples and methods used. Moreover, it is necessary to determine which of the multiple changes observed in the context of ECT represent independently occurring processes, rather than merely different levels of observation of a common underlying mechanism. From a clinical perspective, there are currently no biomarkers available that can be used, either individually or in combination, to guide clinical decision-making, and patients should not be excluded from clinically indicated ECT on the basis of the absence of the markers discussed here.

In the future, adequately powered studies with in-depth characterization of the clinical phenotype in combination with multimodal molecular and neuroimaging markers will hopefully facilitate the identification of (neurobiologically) distinct subgroups of patients with MDD and ultimately lead to a reliable response prediction for the individual patient (Fig. 2). Stratification according to clinical characteristics could reduce the heterogeneity of patient samples and facilitate the identification of biological predictors. While our review focused on depression, markers for the effectiveness of ECT are equally important for other ECT-responsive diagnoses like schizophrenia.

Fig. 2. Multimodal prediction of treatment response to ECT for depression.

Graphical depiction of the multimodal investigation of clinical and biological markers to improve the prediction of therapeutic response to ECT.

Author contributions

DZW developed the concept for the article, coordinated the project, reviewed the literature, and wrote the section on clinical markers. IvMH reviewed the literature and wrote the section on (epi)genetics. AY reviewed the literature and wrote the section on the immunological inflammatory pathway. AT reviewed the literature and wrote the section on neuroimaging. All authors reviewed, edited, and approved the final version of the entire manuscript.

Competing interests

The authors declare that there are no competing financial interests in relation to this work.