1 Introduction

Emotion dysregulation (ED) is a highly impairing emotional and behavioural manifestation in adolescents. Recently, clinical efforts have increased to screen for and identify ED independently of other comorbid psychiatric disorders with the aim to improve clinical outcomes for affected youth [1, 2]. The presence of ED is associated with clinical complexity, increased risk for self-harm, suicide attempts, aggression and poorer clinical outcomes than other symptom clusters [3].

It is defined as a pattern of emotional expression that interferes with effective goal-directed behaviour and, as such, is a common clinical presentation in children and adolescents requiring mental health assessment [4]. Symptoms of ED include intense and frequent experience anger in addition to other dysphoric symptoms like irritability, heightened sensitivity to cues of social threat or rejection, rapid shifts in emotions and difficulty returning emotions to baseline. Due to variability in its measurement, there is much variation in its prevalence [5]. While it is not uncommon in the general population, ED is common in clinical settings as it is a predominant reason for seeking help [6].

Intrinsic (i.e., temperament and biological factors) and extrinsic factors (i.e., adverse childhood experiences or suboptimal attachment styles) may interact to produce ED, and although ED is commonly associated with other psychiatric symptoms and disorders, it appears to be a distinct phenomenon [6]. ED presents similarly across comorbid disorders suggesting that it is transdiagnostic feature of psychopathology [7]. Whether its underlying etiologic processes are similar across individuals has not been demonstrated. This is not likely as emotion regulatory processes are numerous and differ depending on the emotion and the context in which the emotion is being regulated. ED can theoretically reflect one or more deficits in effectively recruiting or applying any of these processes during attempts at emotion regulation [8].

At time of writing, no articles have reviewed what is known about biologic or social processes that may underlie ED in adolescents. Part of the reason for this may be because the field of ED and its clinical operationalization are in early stages of development. Freitag et al. recently published a systematic review of questionnaire-based measures of ED. They identified 115 different measures of ED that have been cited in the literature [1]. It is important to note that this review included both measures of ED and emotion regulation in their study selection criteria. While there is no doubt that problems with emotion regulation are not distinct from ED, the current review considers ED and aberrant emotion regulation processes as distinct outcomes. We made this decision because we want to know what studies that have defined the clinical outcome as ED have identified about its underlying mechanisms. The practical application here is that clinical settings do not measure specific problems with emotion regulation directly, however they regularly report on whether ED is present or not.

The large number of available measures of ED is problematic for the field, as such large diversity in measurement combined with the lack of standardization of the definition of ED leaves room for misalignment between what we define as ED and what we may be measuring using these scales [1, 5]. Freitag et al. note that the majority of studies about childhood ED have included one of five measures of ED and each of these measures include between 10 and 36 items assessing both the “internal (cognitive or emotional) processes” and behavioural manifestations of these processes. These five most commonly used scales include the Children’s Emotion Management Scales (CEMS) [9,10,11], Cognitive Emotion Regulation Questionnaire (CERQ) [12, 13], Difficulties in Emotion Regulation Scale (DERS) [14,15,16], Emotion Regulation Checklist (ERC) [17], and the Emotion Regulation Questionnaire (ERQ). Internal processes refer to implicit and explicit regulation strategies like attention deployment and cognitive reappraisal. There is reason for guarded optimism in regards to the reliability of clinical measures of ED, however, as these five common measures of ED demonstrate moderate to high internal consistencies across different clinical samples of young people [1].

As such, this systematic review has two aims: First, to synthesize the available literature exploring neural correlates of ED. The second aim is to explore how the neural correlates of ED are similar or different across different clinical groups and healthy controls.

2 Methods

2.1 Literature search and study selection

We operationalized ED as any pattern of emotional experience and/or expression that interferes with an individual’s ability to modulate behaviour in an appropriate and goal-directed manner [4]. The literature was assessed for studies that evaluated associations between ED and biological factors in adolescents aged 10–24, the recommended age for adolescence [18]. A search was conducted on ProQuest, Ovid, EBSCO, and PsycINFO for English-language articles published between database inception to March 11th, 2024 using the following medical search heading (MeSH) terms and search strings: emotion* dysregulation AND adolescen* AND (biological OR biomarker OR correlate OR predictor) OR (neuroimaging OR magnetic resonance imaging OR diffusion tensor imaging OR electroencephalo*). The complete search string from each database can be found in the supplementary materials. An additional search was performed in the reference list of identified articles. No additional articles were found.

2.2 Eligibility criteria

Our inclusion and exclusion criteria are outlined in Table 1 in the supplementary materials.

2.3 Data extraction

Four independent reviewers systematically screened all titles and abstracts for eligibility using Covidence, an online software tool for conducting systematic reviews. Reviewers were blinded to the choices of the other reviewers, and any conflicts between two reviewers were resolved by a third reviewer. Authors B.G. and N.S. reviewed the full-text of all articles that met inclusion criteria. Data was extracted using a standard data extraction form for: study identification (ID) or digital object identifier (DOI), title, author, year, specific aims, study design, population description (clinical/non-clinical, primary diagnosis/symptoms), sample size, gender distribution, age of participants, validated measure for ED, and biological correlate of ED.

2.4 Risk of bias assessment

The quality of included studies was assessed independently by author B.G. using the Cochrane risk of bias assessment [19]. Bias was assessed using the following six domains: bias from the randomization process, bias due to deviations from the intended interventions, bias resulting from missing outcome data, bias in the measurement of the outcome, bias in the selection of the reported results, and bias arising from conflicts of interest [19]. Complete results from the Cochrane risk-of-bias guidelines are presented in Table 2 in the supplementary materials.

3 Results

3.1 Search results and study characteristics

Our initial search revealed 1625 studies. After the removal of 132 duplicates, we reviewed the titles and abstracts of 1493 articles. The full texts of 57 articles were screened for eligibility, and 48 articles were excluded. The reasons for exclusion are as follows: 14 articles were excluded for not meeting the age requirement, 11 studies were excluded for not recording a validated measure of ED, 13 studies were excluded for not recording a biological correlate of ED, 10 studies were excluded for measuring a non-neural measure of ED. A complete summary of the search results is reported in Fig. 1. Findings from nine studies were reviewed. A summary of the study characteristics is displayed in Table 3 in the supplementary materials.

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) study selection flow diagram

Full size image

3.2 White and grey matter structural integrity

Two studies explored the association between ED and white matter (WM) microstructural integrity using diffusion imaging tractography, a non-invasive technique that measures water diffusivity in brain tissues. Outputs include fractional anisotropy (FA) and radial and axial diffusivity [20, 21]. Versace et al. assessed this relationship in youth with behavioural (attention deficit hyperactivity or disruptive disorder) and/or emotion dysregulation disorders (bipolar or unipolar mood disorder). Data from a sample of age- and sex-matched unaffected controls was also recruited for use in post-hoc analyses. Dimensional and categorical measures of ED were correlated to indicators of WM structural integrity, where lower FA and higher axial or radial diffusivity values implied better WM structural integrity. A categorical measure of ED was operationalized using diagnostic cutoff points for mania and/or depressive symptoms on the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children Mania Rating Scale (K-SADS-MRS), and Depression Rating Scale (K-SADS-DRS), and the dimensional measure was the Parent General Behaviour Inventory-10 Item Mania Scale (PGBI-10 M). WM structural integrity was measured in the cingulum, uncinate fasciculus, and forceps minor—regions consistently found to play a role in mood as opposed to behavioural disorders [20]. The axial diffusivity, radial diffusivity, and volume of each WM tract of interest was also extracted for each participant [20].

They found a main effect of diagnostic group (i.e., youth with behavioural dysregulation disorders vs ED disorders vs both behavioural and ED disorders) on WM integrity in the forceps minor (P = 0.042) and the uncinate fasciculus (P = 0.009), and a main effect of manic symptoms on WM integrity in the cingulum (P = 0.043). Post-hoc analyses in the form of independent t-tests were conducted with the inclusion of a control group. Participants with ED disorders only demonstrated significantly lower FA in the forceps minor and uncinate fasciculus compared to typically developing controls (P = 0.006), and to those with both behavioural and ED disorders (P = 0.015), or with behavioural dysregulation disorders only (P = 0.025). Similar findings were seen in FA in the uncinate fasciculus, where participants with ED disorders demonstrated lower FA compared to both other diagnostic categories (both P = 0.004), and controls (P = 0.005). Researchers also found lower axial diffusivity in the forceps minor and uncinate fasciculus of youth with ED disorders compared to controls or youth from the other two diagnostic categories (all P < 0.004). Furthermore, a significant main effect of mania scores on FA (P = 0.043) and axial diffusivity (P = 0.05) in the cingulum was observed. There was a significant positive association between mania scores and FA in the cingulum across both control and clinical samples (P = 0.048) [20].

Tsai et al. assessed the association between ED and WM integrity (gauged using the generalized FA value [GFA]—a superior measure of FA) in 91 youth with a Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision (DSM-IV-TR) diagnosis of Attention Deficit Hyperactivity Disorder (ADHD), compared to 122 typically developing controls. ED severity was measured by scores on the Child Behavior Checklist—Dysregulation Profile (CBCL-DP). Major WM tracts across the brain were analyzed and mean GFA values from 76 tracts were derived from each participant. The GFA values of 19 of the 76 tracts were found to positively correlate with ED severity. The 19 tracts of interest have been found to play an integral role in connecting cortical regions associated with sensory and affective processing, emotion regulation, and cognitive control. In particular, WM integrity of the mid- and posterior corpus collosum was found to covary with ED, suggesting dysfunction in the interhemispheric integration of affective information. Interestingly, the direction of the relationship differed between youth with ADHD versus typically developing controls, wherein youth with ADHD demonstrated a negative correlation between ED severity and GFA values of the 19 tracts, and the control group exhibited a positive correlation. As such, in youth with ADHD, lower ED severity was associated with better WM integrity, while in typically developing controls, better WM integrity was associated with greater ED severity [21].

One study used voxel-based morphometry, a neuroimaging technique that segments the brain into cerebrospinal fluid (CSF), grey matter (GM), and WM, to examine the relationship between GM and ED [22]. Participants with ADHD were separated into two groups based on their ED scores [ADHD and high ED (n = 50), and ADHD and low ED (n = 68)], with the scores > 180 on the CBCL-DP selected as a cut-off for high ED. 104 healthy controls were also included. No significant differences in whole-brain total volume, or whole-brain GM/WM volumes were found across these groups. Youth with ADHD and low ED demonstrated significantly greater GM volume compared to youth with ADHD and high ED in the left anterior prefrontal cortex (PFC) (P = 0.033). In the left medial temporal pole (P = 0.023), GM volume decreased with ED severity in both groups. In the left medial prefrontal cluster (P = 0.019), GM volume increased with ED severity in ADHD and low ED groups, however, it decreased with ED severity in the ADHD and high ED group. The left cerebellum crus (P < 0.001) demonstrated the opposite relationship, where GM volume increased with ED severity in the ADHD and high ED group and decreased with ED severity in the ADHD and low ED group [22].

Healthy controls compared to youth with low ED did not demonstrate any significant differences in regional GM volumes. Both groups demonstrated a positive correlation between ED scores and GM volume in the left dorsolateral prefrontal cortex (dlPFC) (P = 0.022), and a negative correlation between ED scores and GM volume in the right medial temporal pole/inferior temporal gyrus (P = 0.004). In the left (P = 0.003) and right (P = 0.002) cerebellum posterior lobe, youth with ADHD and low ED demonstrated a negative relationship between ED severity and GM volume, while healthy controls demonstrated a positive relationship in the same regions [22].

3.3 Functional connectivity and neural activity

Two studies used functional magnetic resonance imaging (fMRI) to examine changes in functional activity and connectivity in adolescents with ED [23, 24]. This type of imaging relies on blood oxygen level dependent changes in the brain, that index up- or down-regulation of metabolic activity in brain regions [23].

Bertocci et al. assessed task-based neural activity during the emotional-n-back (EFNBACK) task. They divided a subsample of youth from the LAMS study into two trajectories of ED, based on biannually collected scores across five years on the PGBI-10M: (1) youth with initially high ED that continued to decrease (HighD), and (2) youth with initially low ED that continued to decrease (LowD). Data collected from these two groups was compared to 24 healthy controls to determine functional abnormalities in neural circuitry associated with trajectories of ED by measuring recruitment of prefrontal cortical systems during the EFNBACK task, where emotionally salient stimuli are presented during a working memory task. Results indicated that, compared to healthy controls, LowD demonstrated greater activation in the bilateral dlPFC, but this group showed only greater activation in the left dlPFC when compared to HighD (both P = 0.001). Interestingly, participants from the HighD group also performed poorly on the EFNBACK task compared to both the LowD and control groups (P = 0.003). This suggests that greater activation in the dlPFC may be involved in compensating for mild decreases in task performance due to emotional and behavioural dysregulation. Analysis of functional connectivity (FC) between the amygdala and regions of the bilateral prefrontal-anterior cingulate found that, compared to LowD, HighD had reduced positive FC between bilateral amygdala and the left ventrolateral PFC (vlPFC) (P < 0.001). Similarly, compared to LowD, HighD also had reduced FC between bilateral amygdala and two clusters in the left dorsal anterior cingulate cortex (dACC) (P = 0.001). FC between the amygdala and dACC/vlPFC of healthy controls did not differ significantly from LowD and HighD. Healthy controls were recruited on the basis of having no psychiatric diagnoses themselves, along with no first-degree relatives with any mood disorders and psychosis, and any second-degree relatives with bipolar spectrum disorder and psychosis. It is unclear whether these parameters were determined on the basis of current or lifetime diagnoses, and no clinical measures were collected for healthy controls to substantiate their categorization in the control group. Including clinical measures for these individuals may assist in further explaining this finding [23].

Qin et al. (2021) studied the impact of a 12-week mindfulness-based cognitive therapy for children (MBCT-C) intervention on network-level neurofunction at rest, specifically network efficiency and characteristic path length, in ten adolescents with ED and a familial history of bipolar disorder. Dysregulation was determined by meeting any one of the following cut-offs: a score < 27 on the ERC, a score > 28 on the Children’s Depression Rating Scale- Revised (CDRS-R), or a score > 12 on the Young Mania Rating Scale (YMRS). No significant changes in clinical measures of depression (P = 0.474), global functioning (P = 0.059), mania (P = 0.553) or emotion regulation (P = 0.102) were seen following the MBCT-C protocol. Participants underwent resting state fMRI scanning before and after the intervention. The frontoparietal network (FPN), cingulo-opercular network (CON), and the default mode network (DMN) were the major networks of interest due to their previously demonstrated involvement in mindfulness-based interventions. Following MBCT-C, the FPN and the CON demonstrated lower characteristic path length (FPN P = 0.017, CON P = 0.023) and higher network efficiency (FPN P = 0.014, CON P = 0.020). No such differences were seen in the DMN. Notably, the reduction in characteristic path length of the CON was significantly associated with a change in scores on the ERC (P = 0.005). Moreover, increased functional connectivity was seen following MBCT-C in a region of the CON comprised of 14 nodes and 19 connections (involved mainly in the anterior cingulate cortex [ACC], dlPFC, basal ganglia, insula, and thalamus), as well as three nodes and two connections within the right fusiform and bilateral superior frontal gyrus of the DMN. Associations between neuronal function at the aforementioned regions of interest and scores on the ERC at baseline was not assessed [24].

3.4 Electroencephalographic measures

Heffer and Willoughby assessed error-related negativity (ERN) in a subsample of 424 youth during the go/no-go task. ERN is a type of event-related potential (ERP) that occurs during salient events, for example, following an incorrect response during the go/no-go task. It is measured using electroencephalogram (EEG), a non-invasive measure of the brain’s electrical activity. Latent class analysis further subdivided the subsample into four dysregulation groups according to participant ratings on four items of the DERS: High Dysregulation/High Threat Sensitivity/Low-Moderate Impulsivity (group one), Moderate Dysregulation/Moderate Threat Sensitivity/High Impulsivity (group two), Low-Moderate Dysregulation/Moderate Threat Sensitivity/Low-Moderate Impulsivity (group three), and Low Dysregulation/Low Threat Sensitivity/Low Impulsivity (group four). With a 95% bootstrapped confidence interval, group one demonstrated the largest ERN during the task and group two demonstrated the lowest. Groups three and four did not demonstrate any significant differences [25].

Figuracion et al. also examined ERN, and error-related positivity (Pe) during the go/no-go task in youth with ADHD and ED (n = 116), compared to youth with ADHD and no ED (n = 117), and healthy controls (n = 121). A modified version of the EATQ-R was used to assess ED. Clinical and cognitive assessments were collected at baseline. Youth were followed up with annually thereafter, and EEG recordings were collected during years 5–8 of the study. Youth with ADHD and ED demonstrated blunted ERN in negative conditions of the go/no-go task compared to other two groups (P = 0.026), and youth with ADHD and no ED demonstrated blunted ERN in neutral conditions compared to the other groups (P = 0.043). Within-group comparisons revealed that youth with ADHD and ED produced blunted ERNs in negative and positive conditions compared to neutral conditions (P-0.019), while youth with ADHD and no ED (P = 0.313) and healthy controls had stable ERN across conditions (P = 0.981). With regards to Pe, there was no significant effect of the presence or absence of ED on Pe (P = 0.067). However, with the marginal p-value, authors noted that pairwise effects demonstrated that youth with ADHD and ED had a blunted Pe response across conditions, compared to healthy controls. Youth with ADHD without ED did not differ significantly from either of these two groups and demonstrated Pe that was in between that of the other two groups [26].

Zubovics et al. [27] assessed the role of neural responsivity to reward using ERPs in predicting ED. The Doors task and Monetary Incentive Delay (MID) tasks were used to measure sensitivity to reward. In the Doors task, participants were briefly presented with 120 sets of two doors and were asked to select either the left or right one. After each trial, they were presented with either an arrow facing up to indicate a gain of 100 Hungarian Forints (HUF), or an arrow facing down to indicate a loss of 50 HUF. For the MID, participants underwent 192 trials where they were presented with a series of cues in the form of full circles, full squares, and empty circles or squares to indicate the gain, loss, or neutral nature of each trial, respectively. Following a brief anticipatory period, participants were presented with a target stimulus and were required to press a button to gain or avoid losing money. They were informed of their success and failures, along with the total cumulative monetary reward they earned on the screen. Participants were informed that any virtual money earned from either task would be redeemable for snacks and fruits, where more desirable fruit and snack options were available at a higher cost than less desirable options.

Reward sensitivity consists of two components: initial responsiveness to the anticipation of the reward, and responsiveness to the reward after it is attained. As such, ERP components that preceded receipt of the reward were considered as measures of interest related to reward anticipation, and this included Cue P3, Target P3, and stimulus preceding negativity (SPN). Meanwhile ERP comonents that followed the receipt of the reward were classified as measures of interest related to responsivity to reward attainment, and this included reward positivity (RewP).

The following five subscales of the DERS were used to assess ED: nonacceptance of emotional response, difficulty engaging in goal directed behaviour, impulse control difficulites, lack of emotional awareness, and limited access to emotion regulation strategies. Regression analyses revealed that during gain trials of the Doors task, RewP was negatively associated with the DERS awareness subscale (P = 0.046), and during loss trials, RewP was negatively associated with scores on the DERS strategies subscale (P = 0.046). During both the gain and loss trials, there was a negative correlation between Cue P3 and scores on the DERS strategies subscale (both P = 0.046). Meanwhile, SPN during loss anticipation on the MID task was negatively associated with scores on the DERS impulse subscale (P = 0.046). Results from multiple regression analyses revealed that 10.9% of the variance in DERS strategies scores is related to the Cue P3 gain trials of the Doors task, while self-reported scores on the reinforcement sensitivity theory—personality questionnaire behavioural inhibition scale (RST-PQ BIS) described 29.8% of the variance in scores on the DERS strategies subscale (P < 0.001). Furthermore, 9.8% of the variance in DERS impulse scores was related to SPN during loss trials of the MID, and an additional 18.2% of the variance was related to RST-PQ-BIS scores (P = 0.008) [27].

3.5 Multimodal neuroimaging studies

One study used multimodal neuroimaging techniques to build on findings from Versace et al. (2015), and assess the predictability of neuroimaging measures combined with clinical measures to predict ED severity in a transdiagnostic sample of LAMS youth. The PGBI-10M was used to measure ED 14.2 months apart at time 1 and time 2. Neuroimaging data was also collected at time 1. This includes functional neuroimaging data collected during a validated block-design card guessing reward task, where participants were asked to use a button to guess whether the value of the cards presented will be more or less than 5 (values ranged from 1 to 9). After this, the card was presented along with an upward facing green arrow to indicate a correct guess by the participant, or a downward facing red arrow to indicate an incorrect guess. There was a focus on the ventral striatum and its functional connections to whole-brain reward circuitry, due to its key role in reward processing. Structural data was also collected during the fMRI using diffusion imaging to measure radial and axial diffusivity, as well as volume and length of WM tracts.

Results demonstrated that the combination of the following seven predictors optimized the fit of the model: time 1 PGBI-10M scores, scores on the mania rating scale (KMRS) and depression rating scale (KDRS), sex, right and left cingulum legnth, and ventral striatum-right parietal connectivity. In particular, greater right and left cingulum length was predictive of better emotion regulation, indicated by lower time 2 PGBI-10M (compared to time 1 PGBI-10M scores). In contrast, greater functional connectivity between the right and left ventral striatum was predictive of higher time 2 PGBI-10M scores, compared to time 1. Female sex and higher scores on the KMRS or KDRS were also predictive of higher time 2 PGBI-10M scores [28].

4 Discussion

The aim of this systematic review was to review the current literature on the neural correlates of ED in the adolescent population. Nine studies were identified that assessed associations between ED and structural neuroimaging measures (n = 3), functional neuroimaging/FC (n = 2), neural electrical activity (n = 3), and combined structural and functional neuroimaging data (n = 1). Studies on WM and GM microstructural integrity demonstrate ED-related differences in areas related to affective and cognitive processing in clinical samples of youth [20,21,22]. Studies measuring ERPs demonstrate task-based differences in typically developing youth [25, 27], and findings from ERP studies in youth with ADHD demonstrate ED-dependent variation in the presentation and neural signature of ADHD [20,21,22]. Functional neuroimaging studies highlight the role of the FPN and CON in the regulation of emotion [23, 24]. Findings for these neural correlates are discussed in more detail.

4.1 White and grey matter microstructural integrity

Neuroimaging studies measuring WM and GM microstructural integrity demonstrated reduced connectivity between regions involved in the interhemispheric integration of affective information, the regulation and processing of affect, and cognitive control in youth with ED and behavioural or emotional psychopathology compared to controls [20, 23]. In youth with ADHD, higher ED severity was associated with higher WM integrity in regions related to cognitive control and affective processing and regulation, whereas in typically developing controls, higher severity of ED was associated with reduced WM integrity in the same regions [21]. Moreover, youth with ADHD and low ED demonstrated greater GM volume in regions related to affective processing and cognitive control, while those with high ED demonstrated a negative correlation between ED scores and GM volume [22].

The altered directionality of the relationship between WM and ED in youth with ADHD may reflect the role of distinct neural correlates underlying behaviours associated with ED in ADHD, and these differential neural underpinnings may be associated with unique emotion regulation strategies [21]. This is further underscored by findings from Tsai et al. where youth with ADHD and low versus high ED demonstrated differential neural patterns associated with ED, depending on the severity of ED [22]. Further research is needed to examine potential differences in the processing of emotional information in relation to attentional control in a developing sample.

4.2 Functional neuroimaging

Functional neuroimaging studies demonstrated changes in regions of the FPN and CON in participants with ED, with the most robust changes being observed in the dlPFC [23, 24]. During the EFNBACK task, the lateralization of these changes varied according to the level of ED severity demonstrated by participants; participants with low ED demonstrated a bilateral increase in dlPFC activity, and those with higher ED demonstrated a left-sided increase only. Coupled with the general finding of lower task performance on the EFNBACK by individuals with high ED, this suggests that individuals with low ED can effectively upregulate dlPFC activity to aid in task performance whereas those with high ED are unable to sufficiently compensate for poorer task performance. Changes in functional connectivity across regions regulating affect may also underly ED, as response to psychotherapy was associated with improved functional connectivity in the CON (which includes the dACC, insula, and thalamus) in youth at risk for bipolar disorder [24], and reduced FC between the bilateral amygdala and the left vlPFC, and the bilateral amygdala and the left dACC was also observed in highly dysregulated adolescents compared to controls [23].

4.3 Neural electrical activity

In healthy controls, task-based EEG studies demonstrated ED-related differences in ERP (a measure of neural activation related to cognitive/motor/sensory processing), and ERN (a type of ERP, follows incorrect responses in classification tasks) [25, 27, 29, 30]. Higher levels of ED in typically developing youth were associated with higher ERN during the go/no-go task measuring response inhibition [25]. This suggests that ERN potency—an indicator of self-monitoring—is not only logically associated with threat sensitivity as shown by others [31], but also ED. Whether this process occurs for all youth with ED remains to be studied as the youth with ED and highest levels of impulsivity demonstrated the lowest ERN values. Interestingly, youth with ADHD demonstrated ED-based variance in self-monitoring, where youth with ADHD and low ED had differential neural responses to errors made during the go/no-go task compared to youth with ADHD and high ED [21, 22].

During the Doors task measuring reward processing, significant differences were found in RewP (measure of reward processing), SPN (measure of reward anticipation and cognition), and CueP3 (measure of attentional orienting) [27] in youth with higher ED scores compared to typically developing controls. ED was measured with the DERS subscales. Subscale score of deficits in remaining aware of one’s own affective state were correlated with (RewP) in gain trials, difficulties in improving mood were correlated with RewP and CueP3 during loss trials [27]. These results suggest that difficulties in reward processing and attention allocation may underlie ED.

5 Limitations and future directions

This systematic review identified several key themes related to the neural correlates of ED that may benefit from further exploration. Literature on youth with ADHD and ED reflects the role of distinct neural correlates underlying behaviours associated with ED [21]. The findings of this review suggest that difficulties in reward processing and attention allocation may underlie ED [27]. In task-based studies, participants with high ED are unable to sufficiently compensate for poorer task performance [23]. This may be related to changes in functional connectivity across regions that regulate affect and have also been shown to correlate with ED. Prior literature has also highlighted the role of ERN potency—an indicator of self-monitoring—in ED [25, 31], making this a promising avenue for future research.

Two methodological limitations are worthy of mention to improve future studies. We did not adhere to the complete PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Including additional MeSH terms such as MRI or EEG specifically may have helped improve our search results.

As a theorized transdiagnostic construct, ED sets a tall order for scientists who wish to study it. Three natural limitations to doing this work are choice of behavioural measurement approach, choice of sampling approach and selection of biological measures indexing (hypothesized mechanisms underlying) ED. Perhaps the largest limitation of the field currently is that ED is measured using a plethora of measures, which may index different aspects of ED. To this point, reports of the prevalence of ED in clinical samples of youth is quite variable (7–45%), depending on the measure or sample used [5, 32]. A recent review of the literature recommended using the combination of a broadband measure of ED (ex: the CBCL-DP) along with a measure of chronic issues with mood regulation (ex: the affective reactivity index), plus a measure that captures the longitudinal course of emotional outbursts to index ED most accurately in children and adolescents [5]. Moreover, ED is an index of difficulties in regulation and functional impairment, thus making a measure of social dysfunction a critical component for the measurement of ED [33]. These components of measurement are important in ensuring that the construct measured is indeed ED, and not severe psychopathology.

Only two of the six studies in this review used more than one measure of ED [20, 24]. One study (Versace et al.) used the presence of ED disorders and mood symptoms to infer ED [20]. As the concept of ED disorders has not yet been established, nor has the distinctiveness of ED from behavioural symptoms consistent with depression in youth, this approach introduces confusion as to whether ED is the outcome or a covariate in the research study. Future studies should aim to measure the clinical phenotype of ED clinically using multiple measures covering intensity, course, and functional impairment.

The sampling approach for a study about ED may be as critical as selecting the measure of ED. ED, like its biological correlates, is proposed to have transdiagnostic properties, yet ED may manifest differently in participants depending on their clinical phenotype. It remains an important question whether all aspects of ED are similar across diagnoses, so to test this, researchers reviewing ED in the context of systematic review and meta-analysis would ideally compare ED findings within and between different clinical groups as well as between clinical groups and youth identified to have very little propensity for ED. Researchers designing empirical studies need to balance the accessibility of clinical samples which may not be as phenotypically diverse as anticipated with the imperative to consider that ED and biological processes are likely transdiagnostic and not disorder-specific and therefore should not be treated as such [7, 34]. The key question remains: in what aspects is the physiology of ED similar across different clinical groups AND is in what ways is it different? Future studies should aim to recruit transdiagnostic samples to better highlight correlates specific to ED. It is also important to examine state-based measures (i.e., examining neural measures during heightened emotional vs neutral state). There is also a need for studies examining neural markers of ED in youth with ADHD, as differing levels of ED in ADHD can have clinical implications in the presentation and treatment of ADHD, as well as its neural underpinnings [21, 22]. Furthermore, considering the equifinality of ED, whole-brain analyses are important in gaining an accurate representation of the neural correlates underlying ED. Given the differences in ED-related neural activity in youth with disorders spanning the internalizing/externalizing framework, like ADHD, there is also a need for multimodal studies that integrate multiple different neural, clinical, and demographic measures to assess ED [21, 22]. There is also a need for studies examining sex and gender differences in ED, especially during developmental ages due to the differing impacts of gonadal hormones on brain development in children and adolescents [35, 36].

We focus on neural correlates of ED in this review, however there are other important physiological correlates that deserve consideration for future studies in adolescents. Currently, this research includes adult only samples predominantly including participants with PTSD or BPD, or are in the preclinical phase. Examples of such correlates include peripheral markers of inflammation [37, 38], measures of allostatic load (cardiovascular markers, lipids, hypothalamic-pituitary adrenal hormones) [39], oxytocin, vasopressin and endogenous opioids [40].

6 Conclusion

In conclusion, ED is a construct associated with psychiatric disorders in youth and is strongly predictive of negative clinical outcomes and suicidality. Our ability to identify ED and provide ED targeted interventions will be an important improvement for clinical psychiatric care. This systematic review identified studies describing neural correlates that may be related to ED including weaker and less integrated WM connections in emotion decision making regions of the brain as well as weaknesses in ERP and differences in FC indexing such processes. Future studies should aim to use more than one measure of ED that includes a longitudinal index of ED severity, transdiagnostic samples to assess the accuracy and generalizability of these results to diverse patient populations, and one or more biological correlates.