Background

Depression and anxiety are associated with the greatest burden of disease of all neurological, psychiatric and substance use disorders [1]. The early onset of psychiatric illness plays a key role in such disability with approximately 75 % of these disorders occurring before the age of 25 years [2]. Despite this, our current capacity to provide tailored early interventions and prevent the progression of illness or slow the pathway to disability is lacking [3]. Whilst, some specific diagnoses have been successfully treated with certain interventions (e.g. CBT for social anxiety disorder; [4]), there are limitations to the optimal treatment for unipolar, bipolar and comorbid mood disorders. This is particularly evident for those with emerging illnesses who often experience mixed states and/or subthreshold symptoms [5]. Since these mood states typically arise during adolescence and young adulthood, a period of critical brain development and functional independence, the impact of illness can lead to greater disability and worse illness outcomes [6–8]. Thus, the search for objective markers of early risk states with predicative capacity in regards to disability and mortality requires rigorous investigation so that appropriate interventions can be trialled and delivered as early as possible to reduce the impact of disability [3, 9].

Traditionally, there has been a focus on the ‘clinical syndrome’ defined as identifying distinct clinical categories or disorders (based on ICD or DSM diagnostic criteria) with specific thresholds and the impact of these on functioning. However, even at a subthreshold symptom level significant contributors to disability and mortality include social and economic disability [10, 11], poor physical health (e.g. diabetes) [12], high suicide and self-harm behaviours [10, 13–15], and risky alcohol and substance use [16, 17]. Given the clinical impact of these factors for young people with emerging mood and anxiety disorders, we have identified them as key functional domains that, we argue, should be the focus of targeted personalised assessment and intervention. Although the term ‘functional domain’ has traditionally, often referred to outcomes relating to occupational (i.e. employment and education) status, here, we use the term to include other key factors that have significant (often concomitant) impacts on levels of functioning in young people. These domains largely align with the framework provided by the ‘International classification of functioning, disability and health’ [18] for conceptualising health and health related states. These include: (i) social and economic participation (i.e. engagement and stability in employment, education and social relationships); (ii) physical health; (iii) suicide and self-harm behaviours; (iv) alcohol and substance use; and (v) clinical syndrome (i.e. diagnostic category, stage of illness and severity of symptoms [3]). These domains are priority areas for service models in Australia (e.g. headspace [19, 20]), which recognise the need for early interventions that aim to target specific outcomes associated with illness persistence and greater disability [21]. Importantly, a focus on these five domains recognises the need to evaluate multiple (often interacting) aspects of an individual to better characterise their specific phenotype and, as a result, attempt to predict their potential illness trajectory.

To overcome some of the limitations associated with current diagnostic approaches that link poorly to neurobiological risk factors or patterns of treatment response it is important to characterise the neurobiology that may underlie or mediate observable functional impairment(s) [9, 22]. This emphasises the need to focus on the four remaining functional domains in addition to the traditional focus on the clinical syndrome to optimise personalised interventions. Models of psychopathology suggest that breakdowns in common brain circuits involved in cognition and behaviour are responsible for the development of psychopathology and general dysfunction [23]. In this view, quantifying the integrity of such brain systems (e.g. via neuroimaging, neurophysiology or circadian biology) along with their behavioural concomitants (e.g. neuropsychology, social cognition or sleep-wake patterns) may lead to the identification of objective markers of early risk states and also serve as treatment targets. For example, in a longitudinal study by our group, neuropsychological performance at baseline was the single best predictor of socio-occupational functioning at follow up, over and above diagnosis and symptom severity [24]. Such findings demonstrate the relevance of objective ‘brain’ markers (in this case, a cognitive phenotype) to provide important insights about a crucial functional domain, which cut across diagnostic categories to direct effective treatment strategies at the pathophysiological driver of poor patient outcomes.

Here, we present a systematic review of the neurobiological and neurocognitive correlates, of five functional domains in young people with mood and anxiety disorders. We focus specifically on major depression, bipolar disorder (I, II, not otherwise specified; NOS) and anxiety disorders (excluding post-traumatic stress disorder), since these most closely relate to the common developmental trajectories of emerging mood disorders in young people [25]. In this review we evaluate the relationship between the functional domains (described above), and evidence from neuropsychology, neuroimaging, sleep-wake and circadian biology, neurophysiology and metabolic studies. A wide age range was chosen (12–30 years) to focus on the adolescent and young adult population; referred to collectively as ‘young people’, to better understand the primary age group that are vulnerable and present to primary youth mental health services. The primary objective of this study is to establish the current status of the literature of young people with mood and anxiety disorder with respect to neurobiological investigations addressing any of the proposed five functional domains. Whilst, we expect that the large majority of identified studies would investigate clinical syndromes and a smaller number would investigate the remaining functional domains, it is expected that unique associations between neurobiological parameters and a functional domain, not accounted for by the clinical syndrome, will become clearer. The aim of our approach is to ultimately provide a framework for guiding the development of personalised assessment and interventions to prevent or delay significant disability in young mental health patients.

Methods

Methods of review regarding eligibility criteria, data collection and synthesis were specified in advance in the form of a review protocol. We followed the guidelines for conducting and reporting a systematic review set out by ‘the PRISMA statement’ [26], and the ‘Cochrane Handbook for Systematic Reviews’ [27].

Eligibility criteria

Report characteristics and information sources

We searched PubMed databases for unique records using the following criteria: (i) published in the last 20 years (i.e. between January 1994 and March, 2014, to coincide with the release of DSM-IV since this version introduced the use of clinical significance ratings related to the impact of illness on areas of functioning); (ii) the study was reported in English; and (iii) had keyword combinations (see Table 1 for full search terms). The reference lists of studies identified by our PubMed search were not utilised as an additional information source.

Table 1 Full list of search terms used according to each topic area
Full size table

Study characteristics and selection

Using a pro forma, the first author (FI) checked the abstract and/or full texts of each paper for the following inclusion criteria; i) a mean age between 12 and 30 years; ii) at least one group of subjects was reported as having a primary mood and anxiety (i.e. depression, bipolar, anxiety) disorder (according to DSM-IV or ICD-10 criteria) or syndrome (e.g. ‘at risk’, current depressive symptoms); iii) at least one of the following functional domains: (a) social and economic participation; (b) physical health; (c) suicide and self-harm behaviours; (d) alcohol and substance use; and/or (e) clinical syndrome, was measured/quantified; and iv) a statistical (i.e. correlational, regression, etc.) association between the functional domain and at least one neurobiological parameter (i.e. neuropsychology, brain imaging, sleep-wake and circadian biology, neurophysiology and metabolic) was reported. Review articles and case studies were excluded from the final synthesis. Studies were labelled ‘yes’ if they fulfilled all four criteria, ‘no’ if they failed to meet all four criteria or ‘possibly’ if it was unclear whether all criteria were fully met. Any disagreement with these rules was resolved by consensus with the senior author (DH).

Identification of studies

Figure 1 displays the series of steps undertaken as we identified studies for this systematic review. First, of the 3975 studies identified by the searches (see Table 1 for search terms), 565 titles and abstracts were examined for eligibility. At this stage, 188 studies were excluded on the basis of not meeting one or more of the eligibility criteria specified. The eligibility stage involved the assessment 377 full texts (i.e. the published manuscript) to evaluate whether these studies were suitable, which led to a further 243 studies being excluded. The remaining 134 studies were included in the final synthesis (see Table 2 for a summary of these studies).

Fig. 1
figure 1

This figure communicates the flow of studies through the systematic review process and identifies the number of studies excluded at each phase as well as the reason for exclusion

Full size image
Table 2 Overview of the included studies organised according to functional domain and neurobiological parameter investigated
Full size table

Synthesis of results

For each of the included studies, the reviewer (FI) collated data with respect to the study design (i.e. cross-sectional, longitudinal; see Table 2), sample characteristics (i.e. age, sample sizes), aims, key measures (e.g. neuropsychological, circadian, clinical) and key findings (presented in Tables 3, 4, 5, 6 and 7; one table per neurobiological parameter). To clarify, the key findings for each study were taken as any evidence of an association between a particular neurobiological measure and a functional domain. In order to achieve this, the various scales, tests, and assessments were collapsed into broader categories of key measures (e.g. specific neuropsychological subtests grouped into a cognitive domain; see Tables 3, 4, 5, 6 and 7). Given the variability in methodology and the large and varied outcomes of identified studies it was not appropriate to carry out a meta-analysis [28].

Table 3 Neuropsychological studies evaluating the five functional domains in young people (12–30 yrs) with a mood and/or anxiety disorder
Full size table
Table 4 Imaging studies evaluating the five functional domains young people (12-30 yrs) with a mood and/or anxiety disorder
Full size table
Table 5 Sleep-wake and circadian biology studies evaluating the five functional domains in young people (12-30 yrs) with a mood and/or anxiety disorder
Full size table
Table 6 Neurophysiological studies evaluating the five functional domains in young people (12-30 yrs) with a mood and/or anxiety disorder
Full size table
Table 7 Metabolic studies evaluating the five functional domains in young people (12-30 yrs) with a mood and/or anxiety disorder
Full size table

Results

A total of 134 studies were included in this systematic review (see Fig. 1); 10 of these studies were featured more than once in the data synthesis to make a total of 144 reported results. As summarised in Table 2, the included studies were categorized according to functional domain in the following proportions: 7.6 % (k = 11) investigated social and economic participation, 2.1 % (k = 3) physical health, 15.3 % (k = 22) suicide and self-harm behaviours, 6.9 % (k = 10) alcohol and substance use, and 68.1 % (k = 98) clinical syndrome. In regards to neurobiological parameters, 19.4 % (k = 28) focused on neuropsychology, 43.1 % (k = 62) on neuroimaging, 16 % (k = 23) on sleep-wake and circadian biology, 14.6 % (k = 21) on neurophysiology and 6.9 % (k = 10) on metabolic measures. The range of the mean ages for patient groups across all the included studies was 11.7 to 31.7 years, since those studies that had comparison groups both inside and outside the inclusion criteria of 12 to 30 were still included.

Neuropsychology

There were 28 studies (a total of 2877 participants; 58.5 % female) that utilised neuropsychology and across these studies 69 % (2037/2877) were patients and 29 % (804/2877) were healthy controls. Among the patient group 47 % (966/2037) had depression, 15 % (301/2037) had bipolar, 12 % (239/2037) had anxiety, and 26 % (531/2037) were classified as other.

Functional domains: social and economic participation, physical health, suicide and self-harm & alcohol and substance use

Our systematic search found an association between neuropsychology and three functional domains (i.e. social and economic participation, suicide and self-harm and alcohol and substance use). No studies that met our criteria investigated physical health. The relationship between global deficits in cognition and social and economic participation in mood disorders is unclear as there were only two studies; one reporting a positive relationship [24], and the other no relationship [29]. The former study utilised a mixed psychiatric sample that consisted of mood disorder and psychosis patients, which may have influenced the results given the well-supported relationship between social and economic participation and neuropsychology in patients with psychosis [30]. However, the latter study only investigated MDD (N = 16) and utilised the Global Assessment of Functioning scale (GAF) as a measure of participation, which could be problematic as the rating can be made on the basis of symptoms or functioning. Thus, it is clear that more studies are needed in this population to resolve or clarify such findings.

Studies exploring specific neuropsychological capabilities have provided greater insight into how these relate to functional domains that can be ambiguous when investigating global cognition. Of the neuropsychological studies reviewed, executive function appears to be particularly associated with social and economic participation as well as suicide and self-harm behaviours. More specifically, three of the five included studies have shown decision-making and conceptual flexibility impairments to be predictive of suicide and self-harm behaviours [31–33]. Taken together, these studies identify a shared pathophysiology among those who have previously attempted suicide, those who were current suicide attempters and those who were currently self-harming; that is, they all showed characteristic deficits in decision-making and cognitive inflexibility. There is also contrary evidence whereby suicide attempters performed better on the decision making task than depressed patients who hadn’t attempted suicide and healthy controls [34], and there were no neuropsychological differences between self-harmers and non-self-harmers [35]. However, these findings may be attributed to methodological differences and/or a modest sample size, especially since the latter study [35] did not distinguish between current self-harmers and previous self-harmers. The notion that impaired decision making and conceptual flexibility may predispose one to suicidal behaviours is supported by evidence showing neurobiological changes in the areas thought to subserve these functions. Namely, structural and functional dysfunction in the orbitofrontal prefrontal cortex have been identified which imply that suicidal behaviour may be associated with deficits in the attribution of importance to stimuli [36–38], although changes in other regions, such as the dorsolateral prefrontal cortex, have been implicated as well [38].

Similar deficits in executive function, particularly in conceptual flexibility, was associated with impaired social and economic participation [39, 40], whereas studies that investigated verbal learning and memory reported conflicting results regarding social and economic participation. Two studies [24, 41] identified a positive relationship with this functional domain, whilst another two studies did not find any association [29, 39]. Notably, logical memory retention (an index of structured learning and memory) was a common measure identified as significant in the positive studies, but it was not utilised in the other two studies; additional studies of structured learning and memory are needed however the evidence to date suggests that there may be an important role for this particular neuropsychological domain with regards to social and economic participation.

Functional domain: clinical syndrome

The association between cognitive function and the clinical syndromes of mood and anxiety disorders in adolescents and young adults has previously been reviewed extensively (see [42, 43]), and some of the findings of such reviews have been reiterated by the present systematic review (see below). Interestingly, it has been reported that impaired verbal memory is significantly associated with depression (not specified as MDD) [44] as well as the development of depressive symptoms, in a community sample [45]. These findings implicate a dysfunction in memory occurring earlier in the course of depressive syndrome development while, poor executive function may be associated with more persistent MDD [46]. In terms of delineating symptom severity, individuals diagnosed with unipolar depression with higher levels of depressive symptoms also show increased (i.e. delayed) choice reaction time [47], lower cognitive control of emotional processing [48], and processing speed deficits [49] suggesting that a broader range of cognitive (including social) measures are also associated with depression and may also be sensitive to the severity of illness.

Similar deficits in cognition have been observed among those with anxiety disorders. Greater Social Anxiety Disorder (SAD) symptom severity was associated with poor executive function, specifically cognitive inflexibility in SAD patients [39]. Most studies have identified cognitive deficits associated with OCD, such as, impaired verbal and visual memory [50], information processing [51, 52], and cognitive flexibility [51–53]. As observed in unipolar depression, increased symptom severity in those with OCD is associated with worse selective attention [52], however this has also not been uniformly reported [50]. Comorbidity has been identified as another factor related to the cognitive deficits in OCD, whereby comorbid depression was associated with executive function deficits in these patients [53]. It is clear that the literature for anxiety disorders is less consistent with regard to the pattern of cognitive deficits and their relationship to anxiety and the severity of illness. The findings regarding GAD and cognition are unsurprisingly similar to the cognitive deficits observed in depression, and provide support for a shared underlying neurobiology for these disorders [54].

Clinical trials utilizing neuropsychological function as an outcome measure have demonstrated that verbal memory improves following: (i) lamotrigine treatment in bipolar patients [55]; and (ii) Transcranial Magnetic Stimulation (TMS) treatment in depressed patients [56]. Furthermore, the improved verbal memory performance was significantly associated with improvements in clinical symptoms of mania and depression, in the former study, and with reductions in hallmark symptoms of depression, in the latter. This adds to the close and complex relationship observed between cognition and mood disorders in young people, and reiterates the need for future research to closely examine the direction of these relationships.

Neuroimaging

There were 62 studies (a total of 3069 participants; 55.5 % female) that utilised neuroimaging and across these studies 62 % (1894/3069) were patients and 38 % (1175/3069) were healthy controls. Among the patient group 28 % (534/1894) had depression, 27 % (520/1894) had bipolar, 15 % (288/1894) had anxiety, and 29 % (552/1894) were classified as other (i.e. mixed psychiatric samples, ADHD, alcohol use disorders, substance use disorder).

Functional domains: social and economic participation, physical health, suicide and self-harm & alcohol and substance use

Results indicate that neuroimaging is a particularly useful modality for investigating suicide and self-harm behaviours as well as alcohol and substance use. In contrast, the utility of neuroimaging for investigating social and economic participation [57] and physical health [58] outcomes is yet to be determined due to a lack of studies exploring these relationships. Moreover, consistent with the findings previously discussed (see 'Neuropsychology') linking poor cognitive flexibility and decision making to suicidal behaviours, reduced Anterior Cingulate Cortex (ACC) volume was associated with a higher number of suicide attempts in patients with comorbid MDD and borderline personality disorder [59]. Furthermore, a study investigating ACC function using fMRI demonstrated that individuals with a suicide attempt history had increased dorsal ACC activity when viewing angry faces, and reduced visual, sensory, prefrontal ACC activity to intense happy and neutral faces compared to both healthy and psychiatric controls [60]. It is suggested that the ACC is an important area involved in attentional control that regulates both cognitive and emotional processes [61]. Structural and functional abnormalities in this area may be indicative of attentional control deficits that affect normal cognitive and emotional processes that are associated with an increased risk for suicidal behaviours in this population. However further evidence for this theory is needed since some fMRI studies could not distinguish between suicide attempters and healthy controls using similar decision making and cognitive flexibility tasks. Psychiatric controls demonstrated increased activity in the ACC during the go-no go task [62] and hippocampus during the Iowa Gambling Task (IGT) [34], compared to suicide attempters who were comparable to healthy controls. It may be that the ACC is associated with attentional control related to the emotive processing that has been linked to suicide rather than the higher cognitive processes investigated in these latter studies.

Alcohol and substance use appears to affect multiple brain structures with most studies indicating that alcohol and substance use is associated with a pattern of reductions in brain volume and impairments in brain function. Cannabis use was investigated by two structural MRI (sMRI) studies where it was associated with reduced total white matter volumes [63], and reduced left fusiform gyrus grey mater volumes [64]. Whilst, the only fMRI study investigating cannabis use reported that lower amygdala reactivity was associated with higher rates of cannabis use in MDD patients [65]. One study [66] investigated alcohol use via sMRI and identified that lower prefrontal cortex white matter and overall grey matter volumes were associated with greater levels of alcohol consumption. Collectively, these findings are mostly consistent with the aforementioned neuropsychological studies (see 'Neuropsychology') and previous evidence regarding the neurobiological effects of alcohol and substance misuse on the structure and function of frontal and temporal brain regions [67].

Functional domain: clinical syndrome

From the fifty neuroimaging studies investigating clinical syndrome, multiple regions of interest have been studied using a variety of imaging methods.

Structural magnetic resonance imaging (sMRI)

In depressive disorders the majority of studies examining brain structure have focused on frontal and limbic regions with mixed findings; although some promising patterns emerge when the severity and/or clinical course of specific disorders are considered. Reduced ACC volumes [68, 69], and increased ACC thickness [70], have been identified in MDD patients compared to healthy controls. Whilst, no significant association with clinical severity or symptoms was found in these studies, reduced ACC volume was associated with higher borderline personality disorder symptom severity but not depression, in patients diagnosed with comorbid MDD and borderline personality disorder [59]. Some further lines of inquiry provide greater detail about how the severity of the clinical syndrome may influence or be influenced by particular brain structures. Decreased grey matter volume in frontal brain regions were evident in patients with a discrete or persistent affective illness compared to those with attenuated syndromes and healthy controls [71], and MDD patients who experienced more than three untreated depressive episodes had reduced subcallosal gyrus volumes, an ACC sub-division [72]. Collectively, these studies reiterate the relationship between reductions in the ACC and depression, and suggest that greater reductions in the ACC may be associated with more severe illness.

Compared to healthy controls, MDD patients had lower amygdala volumes, and no association with clinical severity or illness duration [73], however MDD patients near the onset of their illness had increased amygdala-hippocampal volume ratios that were associated with higher severity of anxiety, but not depression severity [74]. This is consistent with evidence indicating that larger amygdala volumes in GAD patients are associated with greater symptom severity [75]. This suggests that common forms of depression and anxiety may share similar biological processes and genetic liability [76, 77], yet differ in their phenotypic expression. Although distinct pathophysiology may underlie the development of SAD, since lower amygdala grey matter density was associated with greater disease duration and earlier age of onset in this group [78].

Changes to areas of the brain following a course of treatment can provide valuable insight into the success of treatment and how this may have influenced the course of illness. Compared to healthy controls, treatment naïve patients with OCD had larger thalamic volumes, which normalised following paroxetine treatment. These reductions were also associated with a decrease in OCD symptoms. While, CBT treatment for OCD was not associated with change in thalamic volumes [79], greater symptom improvement following CBT was associated with a normalised metabolism in the ACC [80], and with increased prefrontal grey matter volumes [81]. These studies are consistent with the association between the lower symptom severity and greater prefrontal grey matter volumes [82]. Higher compulsive symptom severity was associated with reduced pituitary gland volume in OCD patients in males, compared to healthy controls [83], and larger corpus callosum area [84]. Collectively these OCD studies seem to indicate that pharmacological treatment for OCD may be particularly useful for targeting deficits in thalamus structure and function, whilst CBT may be better for targeting clinical features associated with prefrontal structures.

Functional magnetic resonance imaging (fMRI)

Studies investigating brain function using fMRI have also predominantly focused on frontal and limbic regions. Clinical improvement following lamotrigine treatment for bipolar disorder [85], and SSRI treatment for generalised social phobia [86], were associated with reductions in amygdala activation whilst viewing negative valanced emotional pictures. Similarly, higher activation of the amygdala to emotionally fearful faces compared to happy faces was associated with treatment response to CBT or medication in anxiety patients [87]. Bipolar disorder pharmacotherapy treatment responders compared to non-responders had greater amygdala functional connectivity within the frontolimbic network, and higher amygdala functional connectivity within this network after treatment was associated with greater improvements in mania symptoms [88]. These studies consistently demonstrate a relationship between higher amygdala activity and patterns of treatment response for both bipolar and anxiety disorders. This may be indicative of a shared neurobiological vulnerability for heightened amygdala reactivity associated with stress and the emergence of particular affective disorders [89].

Diffusion Tensor Imaging (DTI)

The evidence from these DTI studies collectively indicate that poorer white matter integrity is associated with affective disorders that may be an early marker of disorder. A diagnosis of a depressive disorder, compared to healthy controls was associated with lower fractional anisotropy, a measure indicating poorer white matter integrity, and higher mean and radial diffusivity in the corpus callosum, whilst higher fractional anisotropy and axial diffusivity and lower radial diffusivity in the uncinated fasciculus [90]. Lower fractional anisotropy in the ACC [91], and genu, body and splenium of the corpus callosum as well as the superior and anterior corona radiata [92] is evident in bipolar disorder compared to healthy controls. For those who experienced maltreatment during childhood, compared to healthy controls, MDD at follow up was associated with lower fractional anisotropy in the superior longitudinal fasciculi and the right cingulum-hippocampal projection, whilst substance use disorder at follow up was associated with lower fractional anisotropy in the right cingulum-hippocampal projection [93]. Greater obsession symptom severity in OCD patients was associated with higher fractional anisotropy in the splenium [94]. Lower white matter integrity in the genu of the corpus callosum, anterior thalamic radiation, anterior cingulum and sagittal stratum was associated with higher depression severity in MDD patients [95]. Having a discrete or persistent psychiatric illness was associated with lower fractional anisotropy in the left anterior corona radiata compared to healthy controls, whilst a similar pattern of lower fractional anisotropy within this region was associated with attenuated syndromes of psychiatric illness [96].

Magnetic Resonance Spectroscopy (MRS)

In terms of MRS studies, the major metabolites that were investigated include N-acetyl aspartate - a measure of neuronal integrity; choline - involved in cell membrane production, lactate - marks glycolysis has been initiated in an oxygen deficient environment; creatine - indicates metabolism of brain energy; glutamate/glutamine -an excitatory neurotransmitter involved in neural activation; and GABA - an inhibitory neurotransmitter involved in reducing neuronal excitability [97]. Lower total choline in the left striatum was associated with OCD and remained consistent over the course of illness [98]. Compared to healthy controls, bipolar disorder was associated with a higher lactate to N-acetyl aspartate and lactate to creatine ratios [99], and higher N-acetyl aspartate in the ACC and higher N-acetyl aspartate, choline and creatine in the ventral lateral prefrontal cortex [100]. Moreover, bipolar disorder responders had lower left ventral lateral prefrontal cortex glutamate/glutamine [101]. For both MDD patients and healthy controls, higher ACC GABA was associated with lower anhedonia scores, and lower ACC GABA was associated with MDD [68].

Sleep-wake and circadian biology

This section entails studies that have utilised either sleep physiology (e.g. sleep EEG) and/or sleep-wake monitoring (e.g. actigraphy) or indicators (e.g. cortisol secretion) to determine sleep-wake and circadian function. There were 23 studies (a total of 1609 participants; 59.3 % female) that utilised sleep-wake and circadian biology and across these studies 84 % (1352/1609) were patients and 16 % (257/1609) were healthy controls. Among the patient group 55 % (747/1352) had depression, 3 % (45/1352) had bipolar, 5 % (69/1352) had anxiety, and 36 % (491/1352) were classified as other.

Functional domains: social and economic participation, physical health, suicide and self-harm & alcohol and substance use

Sleep-wake and circadian biology appears to be useful for characterising two functional domains in young people, namely social and economic participation, and suicide and self-harm behaviours. Three sleep-wake and circadian studies investigated the relationship between salivary cortisol secretion and social and economic participation. Two of these studies [102, 103] were longitudinal investigations that identified that increased salivary cortisol at baseline [102], and before alprazolam treatment for patients with panic disorder [103] predicted poorer social and economic participation at follow up. Such results suggest that increased HPA activity, indexed by a greater salivary cortisol response, may be indicative of HPA axis deregulation with prognostic significance and not simply a cross sectional marker of stress and active illness. Similarly, the one cross-sectional study [104] found that increased cortisol secretion during a social interaction task was associated with poorer social functioning. Alone, this study would seem to demonstrate that increased cortisol secretion is a state marker of social stress, however in light of the previous longitudinal studies, it is possible that heightened HPA activity is indicative of a persistent dysregulated stress response to social specific cues. Notably, all three of these studies were published over 15 years ago, suggesting the need for new evidence to explore the relationships identified by the longitudinal studies.

Three sleep-wake and circadian studies were longitudinal and identified a relationship with suicide outcomes in MDD patients. At baseline, higher growth hormone secretion during the first 4 h of sleep [105], and higher cortisol secretion in the late hours of sleep [106] were both associated with the emergence of a suicide attempt at follow-up. The final study [107] identified that reduced delta sleep activity was associated with higher levels of suicidality as well as depression severity. Similarly, to the social and economic participation studies, these biological substrates also point to HPA dysregulation as a predictor of later suicide attempts in MDD patient groups.

Functional domain: clinical syndrome

Of the eighteen sleep-wake and circadian biology studies, nine studies investigated clinical syndrome utilising cortisol responses. Whilst, the timing of cortisol secretion varied between studies, the findings consistently indicate that increased cortisol response is associated with the development [108–110] and persistence [102, 111] of MDD, whilst remission [112] is associated with reductions in cortisol measures. The heterogeneity of depression becomes clearer with evidence of moderate-to-severe depression being associated with significant blunting of the cortisol response compared to those with mild depression, who had increased cortisol secretion during a stress task [113]. These results suggest that the neurobiological systems mediating the stress response are functioning very differently, and may indicate distinct pathophysiological drivers of depression for these two groups. Specifically, genetically mediated depression associated with increased severity and chronic stress leading to a desensitization of glucocorticoid receptors versus mild to moderate depression arising from predominately environment risk factors with typical HPA abnormalities [113]. This pattern of findings appears to differ for comorbid MDD and anxiety, which was associated with flatter diurnal cortisol slopes (daytime cortisol activity) [114], suggesting that the presence of anxiety influences cortisol function in a way that contrasts to depression alone.

Studies that investigated the relationship between the sleep-wake cycle and clinical syndrome reported quite consistent findings. When compared to unipolar depression, bipolar disorder is associated with delayed onset and lower levels of melatonin secretion [115], as well as increased rates of delayed sleep [116]. Similarly, poor sleep efficiency and lower sleep duration was reported in both unipolar depression [117] and hypomanic individuals [118] compared to healthy controls. Increased depression severity was associated with reduced delta sleep [107], while higher overnight dissipation of slow wave sleep predicted a reduction in depressive symptoms [119]. Similarly, greater high density REM and lower REM latency at baseline was associated with the development of depression at follow in healthy controls [111], whilst another found that greater high density REM was associated with fluoxetine treatment in patients with MDD [120]. Whilst these sleep-wake cycle deficits are evident in those with a full threshold disorders (both unipolar and bipolar disorder), increased rates of delayed sleep are also evident in individuals with either a discrete disorder or an attenuated syndrome compared to individuals with mild symptoms and healthy controls [121]. This indicates that a similar pattern of sleep-wake deficit are also evident in those with subthreshold disorders and highlight that such deficits can be identified earlier in the course of illness.

Neurophysiology

There were 21 studies (a total of 2034 participants; 69.6 % female) that utilised neurophysiology and across these studies 74 % (1510/2034) were patients and 26 % (524/2034) were healthy controls. Among the patient group 56 % (851/1510) had depression, 9 % (130/1510) had bipolar, 21 % (313/1510) had anxiety, and 14 % (216/2034) were classified as other.

Functional domains: social and economic participation, physical health, suicide and self-harm & alcohol and substance use

Evidence from the included studies indicated that neurophysiology may be particularly useful for characterising suicide and self-harm behaviours, and alcohol and substance use. Three separate neurophysiological studies found an association with specific types of suicide and self-harm behaviours. Specifically, lower contingent negative variation was associated with multiple episodes of deliberate self-harm [122], increased posterior EEG alpha asymmetry was associated with suicidal intent and the lethality of suicide attempt, but not depression severity, in suicide attempters [123], and slower reaction times during incentive based decision making tasks was associated with increased frequency for suicide and self-harm behaviours in remitted MDD patients [124]. Together these findings link abnormal brain functions implicated in decision making processes to different types of self-harm and suicidal behaviours [125, 126].

Functional domain: clinical syndrome

With regard to the fifteen neurophysiology studies, a number of these examined the relationship between treatment response and neurophysiological markers using a number of methods. Firstly, increased startle response was associated with multiple episodes of depression in depressed individuals [127]. Increased startle response was also associated with the presence of an anxiety disorder compared to healthy controls [128]. In this study, a reduction in acoustic startle response was associated with a reduction in anxiety symptoms following CBT, and a higher startle response at baseline predicted treatment response. Similarly, steep N1 of the Loudness Dependency of Auditory Evoked Potentials (LDAEP) at baseline and higher P2 LDAEP at week 1 predicted anti-depressant treatment response in patients with MDD [129]. These findings are consistent with the phenomenon described in this paper (see 'Neuroimaging') that link maladaptive processes associated with increased threat processing measured at a neural, psychological and behavioural level [130]. These findings implicate the amygdala as the predominate brain region involved in threat processing, however it functions as part of a wider brain circuit involving the dorsal medial prefrontal (anterior cingulate) cortex [131, 132].

Four studies utilised Event Related Negativity (ERN) to characterise the clinical syndrome of OCD patient groups. All four studies found that increased ERN was associated with OCD compared to healthy controls [133–136], however none identified a relationship with symptoms severity, treatment status or medication use. Conversely the N100 ERP had an inverse relationship with symptom severity, whilst both the N100 and P200 had an inverse relationship with illness chronicity [52]. These studies indicate that ERN may be a trait measure associated with OCD independent of symptoms severity, diagnosis and/or treatment effects, while the ERP N100 and P200 may be measures sensitive to illness specific factors (e.g. chronicity).

Metabolic

There were 10 studies (a total of 1,385 participants; 43 % female) that utilised metabolic measures and across these studies 80 % (1133/1385) were patients and 18 % (252/1385) were healthy controls. Among the patient group 22 % (254/1133) had depression, 35 % (400/1133) had bipolar, 6 % (72/1133) had anxiety, and 36 % (407/1133) were classified as other.

Functional domains: social and economic participation, physical health, suicide and self-harm & alcohol and substance use

Three metabolic studies investigated the relationship between suicide and self-harm behaviours and cholesterol. Lower total cholesterol levels were associated with a suicide attempt history [137], and increased severity of suicidal behaviour (e.g. ideations, gestures) among those who were currently suicidal [138], while lower high density lipid cholesterol was associated with higher suicide ideation [139]. Together these findings seem to indicate that lower cholesterol is associated with the spectrum of suicidal behaviours, particularly in males. However, this is an area of contention since higher total cholesterol was associated with current suicidal behaviour versus those not currently suicidal. These findings highlight the complex association between suicidal behaviours and cholesterol, and indicate that its usefulness a biological marker needs further clarification.

Two studies examined the correlates of the risk of obesity, indexed by BMI, among young people with mood disorders. Substance use was associated with an 2.8 fold increased risk of overweight/obesity among bipolar patients and increased BMI was associated with better social and economic participation over a 4-year longitudinal study in mood disorder patients (unipolar and bipolar) [140]. The latter study also reported an increase in the prevalence rates for overweight/obesity, consistent with the finding of the former study. There are a number of illness related factors that could explain this relationship which include a return of normal appetite, medication use, self-modulation of mood by overeating [141] as well as biological factors implicated in mood, and metabolic function and weight maintenance, such as leptin [142] and neurotransmitter abnormalities [143].

Functional domain: clinical syndrome

See Table 7 for individual results.

Discussion

As expected, there is a predominate focus in the literature on clinical syndrome in young patients compared to the other four functional domains (i.e. social and economic participation, physical health, suicide and self-harm behaviours and alcohol and substance use). Whilst the neurobiology of these clinical syndromes have been extensively reviewed previously [131, 144], we provide an overview of these findings. Typically biomarkers of the clinical syndrome alone do not readily provide a complete understanding of disability and the risk factors that put young people at greater risk for a worse illness trajectory [3]. This review demonstrates the use of these biomarkers to investigate multiple functional domains in addition to the clinical syndrome. It is clear that the nature of the relationship between the underlying neurobiology and functional domains is an issue that needs to be resolved in this area. However, overall this review exhibits the usefulness of neurobiological parameters to assess these additional functional domains and identify treatment targets, in addition to the traditional focus on clinical syndrome, to optimise interventions and improve illness trajectories.

Implications for personalised psychiatry

The search for gold standard screening or diagnostic tests has ultimately been unsuccessful, however the increasing emphasis on personalised (or ‘stratified’) psychiatry has the potential to make significant advances in terms of clinical validity and applicability [145]. Whilst, conventional diagnostic methods remain entirely relevant, the addition of the neurobiological markers that indicate prognosis or potential treatment targets are essential for advancing personalised psychiatry. Utilising individual characteristics to guide treatment decisions is key to personalised medicine and providing person-centred care. This review exhibits the utility of the RDoC approach to investigate individual characteristics that extend to functional domains that contribute to ongoing disability and poorer outcomes. By collating the available evidence, this systematic review provides a basis for future investigations to further evaluate the clinical utility of specific neurobiological markers and their relationship with these functional domains (Fig. 2).

Fig. 2
figure 2

The potential neurobiological targets associated with mental illness trajectory organised by functional domain. This is a model illustrating the potential neurobiological targets for each functional domain identified by this systematic review. At the centre is ‘mental illness trajectory’ that is a term employed here to encompass all the characteristic features of illness, and its associated short term and long term outcomes. Surrounding this are the five functional domains investigated by this review to illustrate how a young person’s mental illness trajectory is made up of and influenced by each of these functional domains. Within the respective sections of the figure, the potential neurobiological targets for each functional domain identified by this systematic review are highlighted in the boxes. The absence of any neurobiological targets in the physical health section of the figure reflects the lack of investigation into this particular functional domain emphasising the need for more research into this area. The clinical syndrome section is the most populated which reflects the predominant focus in the literature on this particular domain. Despite less focus on the remaining three functional domains, the evidence for potential neurobiological targets for these areas is promising for clinical utility and future research

Full size image

Social and economic participation

Our findings suggest that those with more severe impairments in memory and executive functioning are at the greatest risk for diminished participation in education, employment and social settings. The use of cognitive tests that characterise memory and executive functions of young people with emerging illnesses may be particularly useful for identifying individuals who are at risk of poor social and economic participation outcomes. Unlike diagnosis, neuropsychological function independently predicts social and economic participation outcomes in young people with emerging mood disorders [146]. The evidence suggests that neuropsychology, specifically memory and executive function, may mediate these outcomes in young people. Therefore, those with identified weaknesses in these areas are likely to require more intensive intervention targeted at these deficits to improve this functional domain. Such treatment practise has demonstrated clinical utility for improving memory performance in adult groups with affective disorders [147]. Moreover, cognitive training was comparable to pharmacological treatment for improving the depression state, whilst also associated with additional benefits such as, no adverse side effects, improved levels of anxiety, and better academic performance [148]. Cognitive training may slow or prevent the impact of cognitive impairment on the social and economic participation domain for young people with an emerging affective disorder.

Notably, the common indicators of social and economic functioning in the reviewed studies varied greatly with some measures (e.g. SOFAS, academic grades) being more useful than others (e.g. GAF) that conflate symptoms and functioning. Of the reviewed studies, there was a lack of focus on social functioning and the relationship with certain neurobiological parameters. The only study to specifically examine this relationship identified that HPA dysregulation, indexed by an enhanced cortisol response during a social task was predictive of social functioning [104]. Although the aforementioned study was cross sectional, longitudinal evidence supporting the predictive validity of HPA dysregulation and greater disability (including social disability), provides further support for the role of HPA dysregulation in social and economic participation [102, 103]. However, it is clear that the role of HPA functioning is complicated given its implications in multiple functional domains including suicide and self-harm behaviours and clinical syndrome.

Physical health

Physical health as a functional domain for young people with mood disorders has been notably understudied (only three studies met the inclusion criteria) compared to the other four functional domains. This is not completely unexpected considering that the majority of neurobiological modalities (neuropsychology, neuroimaging, sleep-wake and circadian biology and neurophysiology) are not recognised as being the traditional method for investigating this particular domain [149]. Of note, given that metabolic measures are used to classify physical health outcomes (i.e. BMI is both a physical health outcome and a metabolic measure) it was not deemed appropriate to carry out searches for metabolic measures and physical health outcomes. Future studies may look to improve the methods and/or key terms used to investigate the best available measures to assess and track physical health outcomes in this population.

Suicide and self-harm behaviours

Studies investigating suicide and self-harm behaviours spanned across all five neurobiological parameters, and yielded consistent findings across these parameters. Across studies, those with specific executive function impairments in the domains of decision-making and conceptual flexibility appear to be more likely to engage in suicidal thinking or behaviours. Therefore, identifying these deficits may be particularly important for recognising at risk patients and providing effective interventions. Furthermore, our findings provide converging evidence from different measures of brain structure and function that suicide and self-harm behaviours are associated with significant disruptions in decision-making ability. The reduced ACC volumes and activity in the identified neuroimaging studies as well as neurophysiological evidence of lower accuracy at differentiating correct and incorrect responses on a decision-making task corroborate the neuropsychological findings regarding the relationship between suicide and self-harm behaviours and impaired decision making and conceptual flexibility.

Importantly, these findings have major implications for the assessment and intervention of suicide and self-harm behaviours. Firstly, decision making ability should be incorporated into the assessment of young people with mood and anxiety disorders to stratify young people on the basis of risk for suicidal and self-harms behaviours. Whilst, there are certainly other risk factors involved in suicide and self-harm behaviours [150, 151], the findings from this review indicate that decision-making may be a mediator of suicide and self-harm outcomes. Secondly, cognitive remediation interventions aimed at improving decision-making ability in those young people that are identified as having significant impairments may prove to be a successful early intervention to reduce or prevent elevated risk for suicide and self-harm behaviours [152].

The relationship between metabolic studies investigating blood cholesterol and suicide and self-harm behaviours was another major area identified by this review. Three studies investigating cholesterol consistently identified an association between suicide and self-harm behaviours and lower cholesterol levels. Such findings are consistent with evidence that decreased cholesterol levels in the brain may be associated with reduced synaptic plasticity and impaired neurobiological functioning [153]. Furthermore, lower serotoninergic activity has also been associated with reduced cholesterol and implicated in the affective disorders [154], which was evident in another metabolic study identified by this review [155]. It has been suggested that these changes to cholesterol levels may be the result of HPA axis dysfunction, however further studies are needed to explore these associations and investigate the treatment implications.

Alcohol and substance use

Converging evidence from neuropsychology, neuroimaging and neurophysiology indicate that alcohol use is associated with global impairments. Unsurprisingly, all the neuroimaging studies suggest that widespread impairment across frontal and temporal areas of the brain are associated with alcohol and substance use. Specifically, reduced brain volume and function are particularly prominent in these areas and these results reflect the findings of neuropsychological and neurophysiology studies that suggest alcohol and substance use to be associated with cognitive impairments and poor attention. Independently each individual study’s limitation of small sample sizes and lack of replication limit their clinical applicability, however together the evidence points to similar phenomena.

An important issue is to determine the how these findings can be used to understand the risk factors associated with alcohol and substance use. Many of the impairments may be the result of alcohol and substance use rather than a mediating factor involved in the risk of engaging in these behaviours. These problems have major implications in terms of assessment and intervention. Whilst, the aforementioned neuroimaging, neuropsychology and neurophysiology studies may be useful for tracking changes in the effects of these behaviours overtime, it is still unclear what assessment measures may be particularly useful in the early identification of individuals at risk for engaging in these harmful behaviours. More specific longitudinal studies before young people engage in alcohol and substance use are needed to differentiate between pre-existing and subsequent effects of alcohol and substance use. These longitudinal investigations will help model the role that risk factors such as risk taking, impulsivity, social occupational factors and decision-making play in the development of poorer alcohol and substance use outcomes.

Clinical syndrome

The final functional domain addressed by this review has been extensively reviewed elsewhere [131, 144], and therefore we provide an overview of these findings specific to young people and in the context of the other four functional domains. The primary focus of this particular domain is to identify features of the clinical syndrome that may help characterise a young person’s illness phenotype and/or stage of illness [25, 156]. The majority of studies identified by this review utilised case-control methods to investigate the neurobiological characteristics that separate discrete diagnostic cases from healthy controls, however these distinctions often provide limited clinically useful information. Those studies that sought to describe how specific neurobiological characteristics are related to illness severity (or perhaps more importantly, employed a case-case control method to delineate between cases) were particularly useful for identifying neurobiological or neurocognitive risk factors of poorer illness trajectories. Using a multi parameter approach, this review has been able to collate evidence that covers a number of ‘columns’ in the RDoC matrix (i.e. circuits, physiology, behaviours) to study the pathophysiological drivers of each functional domain, which is crucial for identifying treatment targets [157].

This review has provided evidence to indicate that verbal memory problems may be an early indicator associated with the emergence of depression since these deficits were evident in young people with depression, not identified as MDD, and the emergence of depressive symptoms in a community sample) [44, 45]. However, for those young people with more persistent MDD and more severe depression or anxiety symptoms executive function deficits seem to be more prominent. Similarly, another line of evidence emerging from this review indicates that greater illness severity is associated with greater reductions in the ACC, a prefrontal brain region associated with executive function. Considering the previously discussed results regarding the relationship between executive function, and another two functional domains (i.e. social and economic participation and suicide and self-harm behaviours), this reiterates the role of executive function as a mediating factor that is associated with poorer illness outcomes across multiple functional domains that should be a clear treatment target at both a brain circuit and behaviour level. The benefits of targeting neuropsychological function are made clear by treatment studies that have shown that improved cognitive function following a form of treatment, namely, TMS or pharmacological, has improve hallmark symptoms in both bipolar and depression. Again, this is reiterated by evidence from neuroimaging that treatments increasing ACC function are associated with clinical improvements in depression and bipolar.

The role of comorbid anxiety in depression and bipolar disorder is notable and associated with a substantial increase in morbidity and mortality [158–160]. From a circuitry point of view, the amygdala is one of the primary brain regions in a broader network that is involved in depression and anxiety [131, 132]. Specifically, evidence from this review indicates that increased amygdala volumes are associated with increased anxiety symptom severity in MDD as well as a diagnosis of GAD, reiterating theories that these conditions share similar pathophysiology that cuts across diagnostic boundaries [161]. Importantly, other neuroimaging work utilising fMRI has demonstrated that heightened amygdala activity is associated with both depression and anxiety [132]. This is evidenced by the neurophysiology study identified by this review whereby successful CBT treatment for anxiety disorder was associated with a reduction in anxiety symptoms and the startle response, another index of amygdala reactivity [128]. Together these findings implicate the amygdala as a treatment target that may reduce neurobiological substrates of anxiety, commonly implicated in the emergence of depression and a problematic feature in bipolar disorder. To reiterate the value of amygdala reactivity as a treatment target for anxiety in affective disorders, studies in the present review demonstrated that reduced amygdala activation to negative emotional stimuli, either using pharmacological treatments or CBT, were associated with clinical improvement in both bipolar and anxiety disorders.

From the sleep-wake and circadian biology studies, sleep dysfunction and cortisol secretion have consistently demonstrated a relationship with the clinical syndrome features that allude to differential illness trajectories. For example, sleep dysfunction, characterised by a number of different sleep parameters, was not only associated with the presence of a discrete depressive or bipolar illness, but similar dysfunctions were also evident in those with hypomanic or attenuated syndromes. This is a critical finding since it presents sleep dysfunction as a primary treatment target to prevent illness progression in these affective illnesses. Similarly, increased cortisol secretion was consistently implicated in the emergence of depression [102, 108–110, 112], whilst two metabolic studies suggest that increased BMI is associated with the chronic course of mood disorders [140, 162]. Neurobiologically these findings implicate HPA dysfunction as being a core feature involved in mood disorders, and so interventions aimed at improving the deficits in these brain circuits may be useful to address these clinical outcomes [163].

Moving towards greater clinical translation

One of the clear problems identified by this systematic review is the lack of consistently used patient groups and assessment measures. For any given functional domain, multiple self-report or clinician rated scales, neuroimaging techniques, and cognitive tests were implemented that often assess the same or similar outcomes, whilst the selection of patient groups varies dramatically for each study. This fundamentally limits the capacity for strong comparisons to be made between studies, or arrive at meaningful and clinically relevant conclusions. For the field of psychiatry to make new ground regarding the underlying neurobiology of psychopathology and its associated outcomes major consolidation of the common standardised measures for these key functional domains and neurobiological parameters should be implemented. Admittedly differences in scientific or clinical motivations will affect the widespread adoption of common measures, however much like the RDoC initiatives focus on key neurobiological domains of interest similar efforts should be made to maximise the standardised measurement of the key functional domains of interest across multiple diagnostic groups in clinical and research settings. In doing this review, we have provided an overview of the essential functional domains and current neurobiological evidence associated with these domains, the next step will be to outline the standard measures that should be drawn upon to promote better consolidation of findings in the psychiatric and neurobiological study of mood and anxiety disorders.

The lack of emphasis on all of these key functional domains and their role in disorder onset, persistence and impact from a publication, reporting and/or research priority point of view is problematic for the clinical translation of psychiatric research. Our study highlights this pertinent issue so that future research may better account for these factors and their relationship to key neurobiological parameters and disorders to improve our understanding of how these functional domains interact or relate to mental illness trajectories.

Limitations and future directions

Some limitations of this review should be considered. First, few studies investigated a particular functional domain (i.e. physical health) or utilised particular neurobiological parameters, namely sleep-wake and circadian biology or metabolic, which limits the synthesis of these findings and caution is advised when interpreting these results. Secondly, the restricted use of search terms for the functional domains or neurobiological parameters may have limited the identification of key studies, particularly favouring studies reporting current primary disorders rather than lifetime diagnoses. Best efforts were made to be inclusive of as many studies as possible to carry out a complete overview of the literature, however future studies should look to expand on this work by adding key search terms that may have been missed to further the advancement of this growing literature focusing on the functional domains of mood and anxiety disorders in young people. While we focused on RDoC levels of analysis that correspond to ‘circuits’, ‘physiology’, ‘behaviour’, and ‘self-report’, future studies may want to include levels of analysis that include genetics, molecules and cells for a deeper understanding of these neurobiological factors.

Furthermore, future researchers may want to consider including borderline personality disorder into these investigations, given its significant affective component and relationship with these primary mood and anxiety disorders. Limitations associated with the systematic review process should be considered when interpreting the present findings, namely the use of one independent researcher for study selection and lack of a systematic risk of assessment bias since these may have impacted on the reliability and validity of data synthesis. Finally, the wide age range selected for this study is an important caveat that should be addressed in future studies since there are significant developmental/neurobiological changes during this dynamic period of brain development with respect to grey (e.g. pruning) and white (myelination; connectivity) matter processes.

Conclusions

Mood and anxiety disorders are especially difficult to characterise and treat in young people (age 12 – 30 years) when confounds of normal development and changing environmental influences are prominent. This review identified a predominant focus in the literature on the clinical syndrome, which in our view does not adequately address key individual characteristics, such as suicide and self-harm behaviours or alcohol and substance use, that are involved in disability and persistent illness. Based on the synthesis of results from multiple neurobiological modalities, we provide a detailed summary of how the clinical utility of neurobiological measures may be improved by focussing on personalised assessment of these additional functional outcomes. We suggest that a shift in focus towards characterising the mechanisms that underlie and/or mediate multiple functional domains will optimise personalised interventions and improve illness trajectories.