Characterizing neurocognitive impairment in young people with major depression: state, trait, or scar?

Abstract Background Major depressive disorder (MDD) affects a quarter of adolescents and young adults and is associated with the greatest global burden of disease in this population. There is a growing literature, mostly in adults, showing that significant neurocognitive impairments are common in MDD. It remains unclear whether these impairments are pre‐existing trait markers of MDD, state‐related impairments that fluctuate with depressive symptoms, or ‘scar’ impairments that worsen with illness progression. The aim of this study is to provide a conceptual framework for understanding MDD and neurocognitive impairment in adolescence and young adulthood (ages 12–25 years). Method Examination of the evidence for neurocognitive deficits as trait, state, and scar features of MDD according to different study designs (family studies, premorbid studies, current depression, remitted depression, and longitudinal studies with repeated assessment) was conducted. Results The few premorbid and family studies conducted in youth provide equivocal evidence for neurocognitive impairments as trait markers of MDD. The presence of state‐based neurocognitive impairment remains unclear as evidence comes mostly from cross‐sectional studies. There are a limited, but growing number of longitudinal studies with repeated neurocognitive assessment in youth. Studies that examined neurocognition prior to the onset of MDD and with long‐term follow‐up provide tentative evidence for neurocognitive scarring. Conclusion Neurocognitive impairment is a feature of MDD in adolescents and young adults. To better understand the nature, timing, and pattern of impairment, longitudinal studies that examine neurocognition before and after the development of full‐threshold MDD, including following recurrence are needed. This knowledge will have important implications for mechanisms, prevention, and treatment of MDD in youth.


| INTRODUCTION
Major depressive disorder (MDD) is a common psychiatric disorder, which peaks in onset during adolescence and young adulthood (Kessler et al., 2005). By the age of 19 years, up to one in four young people in Western countries will have experienced a major depressive episode (Australian Institute of Health and Welfare, 2011;Fergusson, Boden, & Horwood, 2007;Lewinsohn, Rohde, & Seeley, 1998;Rohde et al., 2013). Adolescence and young adulthood represents a critical period due to the dynamic neurological and neurocognitive developmental processes occurring during this phase of life (Casey, Getz, & Galvan, 2008;Giedd, 2004;Paus, Keshavan, & Giedd, 2008) and because the formation and maintenance of social and intimate relationships and educational and vocational attainment are at the forefront. Major depression tends to be recurrent, particularly for those with a young age of onset (Birmaher et al., 2004;Fombonne et al., 2001). With each episode that is experienced there is an increased risk of recurrence (Lewinsohn et al., 2000;Solomon et al., 2000). Over the course of repeated episodes there is also a worsening pattern (Kessing, Hansen, & Andersen, 2004), with more frequent recurrence, greater severity, and more resistance to initially effective treatments (Kendler, Thornton, & Gardner, 2000). Furthermore, MDD confers significant risk for suicidal ideation, suicide attempts, and completed suicide (Cash & Bridge, 2009;Foley, Goldston, Costello, & Angold, 2006;Harrington et al., 1994;Rao, Weissman, Martin, & Hammond, 1993;Weissman et al., 1999). Thus, the onset of MDD during adolescence/ young adulthood is not only contemporaneously disruptive and potentially life threatening, if not adequately treated, it can be associated with lifelong impairment and disability (Fergusson & Woodward, 2002;Fergusson et al., 2007;Weissman et al., 1999). Indeed, because of the marked prevalence and impact of MDD, it accounts for the greatest global burden of disease in young adults (Gore et al., 2011) and is projected by the World Health Organisation (WHO) to be the leading cause of disability globally by 2030 (Mathers, Fat, & Boerma, 2008).
Thus, identifying markers of risk, illness progression and barriers to recovery, and tailoring evidence-based interventions accordingly to the clinical presentation are critical for effective early intervention in

MDD.
Neurocognitive impairments, such as poor concentration and memory, slowed speed of information processing, and difficulties organizing one's thinking (i.e., executive dysfunction), are a central feature of many psychiatric conditions, including MDD (Millan et al., 2012). Empirical studies on neurocognition in MDD have steadily increased over recent decades and have provided robust evidence for widespread neurocognitive impairments in MDD samples relative to healthy controls (Douglas & Porter, 2009;Rock, Roiser, Riedel, & Blackwell, 2014). Neurocognitive impairments in MDD have been associated both with poorer response to treatment (Bruder et al., 2014;Gallagher et al., 2007;Gordon et al., 2015;Potter et al., 2004) as well as poorer social and vocational functioning and greater disability in adults (Baune et al., 2010;Jaeger, Berns, Uzelac, & Davis-Conway, 2006;Woo et al., 2016) as well as young people  with MDD.
More recently, the field has moved toward investigating the precise nature and mechanisms of neurocognitive impairment in MDD, particularly in light of the possible progressive nature of the illness (Hasselbalch, Knorr, & Kessing, 2011). Specifically, an important question is whether neurocognitive impairments in MDD are pre-existing trait/vulnerability markers, state-related impairments that fluctuate with depressive symptoms, and/or 'scar' impairments that remain during periods of remission and worsen with illness progression. It is essential to understand the nature and source of neurocognitive impairments in MDD because this will advance current etiological theories and guide the specific prevention or intervention strategies that might be used (Allott et al., 2013).
The aim of this study is to provide a conceptual framework for understanding neurocognitive impairment in MDD and to examine the current evidence for neurocognitive deficits as trait, state, and/or scar features of MDD. Findings and discussion will be organized according to the various study designs that inform the trait, state, and/or scar hypotheses of neurocognitive impairment. There will be a specific focus on studies of adolescence and young adulthood (defined as ages 12-25 years) for two reasons. First, this a highly vulnerable developmental period of life and risk period for the first onset of MDD. Second, data gathered on youth are less confounded by factors associated with chronic MDD that may impact neurocognitive function, such as long-term use of medication and poly-pharmacy, substance use, or other psychiatric comorbidity, hospitalizations, electroconvulsive therapy, etc. Nevertheless, given the emerging nature of this field in youth cohorts, published reviews of studies in adults will also be included as a point of comparison and where findings in youth are lacking. This is not intended to be a systematic review of the literature on neurocognition in MDD, which is beyond the scope of this study. The overarching aim is to advance current theories of the origin and evolution of neurocognitive impairment in youth MDD. It is also hoped that this study will provide guidance as to areas for further research on neurocognition in youth MDD.

| CONCEPTUAL AND STUDY DESIGN FRAMEWORK FOR UNDERSTANDING NEUROCOGNITIVE IMPAIRMENT IN MDD
Neurocognitive impairment in MDD may be broadly interpreted as trait-, state-, or scar-based impairment. It is important to note that trait, state, and scar patterns of neurocognitive impairment may cooccur and different study designs may provide evidence for more than one of these mechanisms. Each of these patterns of impairment, including how different study designs can help to differentiate them, is described in more detail below.

| Trait neurocognitive impairment
For a characteristic to be considered a possible trait or risk marker, it must be clearly associated with the illness in question (i.e., predicts later onset of illness), but also independent of clinical state.
Specifically, it must be a stable persistent feature that is detectable | (3 of 12) e00527 prior to illness onset and also present during periods of symptomatic remission (Gottesman & Gould, 2003). Evidence for trait neurocognitive impairment comes from studies that investigate neurocognitive functioning of individuals (as early as childhood) before they develop full-threshold first-episode MDD. Evidence may also come from studies of remitted MDD samples, although cross-sectional studies of remitted MDD cannot differentiate trait from scar effects.
Only longitudinal studies assessing neurocognition before and after a major depressive episode can disentangle trait from scar impairments.
Trait impairments may occur through biological (heritable or nonheritable) or environmental mechanisms. For example, prenatal or early life stress might cause persistent neurocognitive deficits in childhood that are associated with later risk for MDD. For neurocognitive impairment to be considered a heritable trait putatively linked to a gene or genes (also referred to as an endophenotype), it should be more frequently present or more severe in relatives unaffected by MDD than in the general population, while still being more frequently present or severe in the affected versus unaffected family members (Gottesman & Gould, 2003). Thus, family studies showing similar, although attenuated impairments in unaffected relatives of individuals with MDD would provide evidence for neurocognitive deficits as heritable traits (endophenotypes) of MDD. If neurocognitive impairments are found to be traits of MDD, they could be used as early markers for identifying those at risk of developing MDD, including identifying specific thinking and coping styles related to depressive symptoms (e.g., rumination) (Han et al., 2016;Snyder, 2013). Trait neurocognitive impairments, therefore, may be an important focus of preventive interventions (Hetrick et al., 2008).

| Scar neurocognitive impairment
Progressive decline or attenuated development in neurocognitive functioning associated with the onset and progression of MDD would indicate scar-related neurocognitive impairment. As implied by the term 'scar', neurocognitive functioning would be expected to worsen with increased severity and chronicity of illness (e.g., number of depressive episodes, duration of illness, etc.). In adolescents and young adults who are still undergoing significant neurological (particularly fronto-temporal) and neurocognitive development, scar neurocognitive impairment may not necessarily manifest as a decline or progressive worsening course. In this instance, it may be that the adolescent still shows improved neurocognitive performance in line with ongoing development during this stage of life, but their development might be attenuated relative healthy peers (Anderson, Northam, Hendy, & Wrennall, 2001;Vijayakumar et al., 2016). The resulting neurocognitive pattern may then reflect a developmental 'lag' or even 'arrest', instead of decline per se. Furthermore, because of the dynamic phase of development during adolescence and young adulthood, impairments in neurocognition may not be proximally evident in relation to MDD and may only begin to emerge after passing through this neurodevelopmental period (Anderson et al., 2001).
Cross-sectional studies that compare individuals with less severe/ chronic MDD with more severe/chronic MDD, or that investigate the relationship between the degree of neurocognitive impairment and number of previous depressive episodes or length of illness in people who are in remission from MDD, may provide partial support for the scarring hypothesis. However, repeated neurocognitive testing of the same individuals via longitudinal designs provides the most rigorous evidence for scar-like effects. In adolescents and young adults, followup periods would need to span several years in order to capture the interaction between prolonged neurodevelopment, neurocognitive functioning, and MDD.
Possible causes of progressive neurocognitive scarring include pathways regulating inflammation, oxidative repair (Galecki et al., 2015), apoptosis and neurogenesis (Lee, Reif, & Schmitt, 2013), as well as dysregulation of the hypothalamic-pituitary-adrenal axis (Vreeburg et al., 2009). If neurocognitive scarring is a possible consequence of MDD, early identification through regular neurocognitive assessments in combination with routine blood biomarkers may be indicated.
Accordingly, psychological and biological interventions such as cognitive remediation therapy or specifically targeting potential pathomechanisms (e.g., inflammation) may be important components to the treatment armamentarium.

| State neurocognitive impairment
State-based neurocognitive impairments are evident if they cooccur with depressive symptomatology and increase or decrease with the exacerbation or resolution of depressive symptomatology. They can also be expected to be more severe with greater symptom severity and may occur over and above existing trait or scar impairments. Cross-sectional studies of neurocognitive functioning of individuals with current MDD are the most common design for examining this, but are particularly nonspecific because they cannot distinguish state-like from trait-like neurocognitive deficits, or can they completely rule out scar-like effects. Better evidence for state-related impairment would come from longitudinal studies showing that neurocognition improves following depressive episode remission or negative findings from adequately powered cross-sectional studies of neurocognition in remitted MDD. Evidence could also include experimental findings showing that induced depressed mood produces poorer neurocognition, or studies using experience sampling methodology that can examine the temporal relationship between subjective cognitive and mood changes. 1 State-related neurocognitive impairment may be caused through alterations in neurotransmitter levels in the neurotransmitter systems that govern both mood and neurocognitive skills. By definition, many clinical symptoms of depression directly relate to neurocognitive function, including a diminished ability to think or concentrate, indecisiveness, sleep deprivation, and fatigue (American Psychiatric Association, 2013). Additionally, feelings of hopelessness and reduced motivation are likely to impair effort and performance on neurocognitively demanding tasks. If state-related neurocognitive impairments are a feature of MDD, then clinicians may need to adapt their therapy to ensure individuals derive benefit. For example, in the case of diminished ability to concentrate, clinicians may have shorter therapy sessions or focus more on behavioral activation rather than cognitive restructuring when delivering cognitive-behavior therapy. Neurocognitive functioning might also form part of the psychoeducation provided to clients, including the fact that neurocognitive impairments may contribute to poorer functioning while unwell. Additionally, neurocognitive impairment might also be used as a marker of treatment response (Gallagher et al., 2007). It is important to note that the model depicted in Figure 1 does not specifically include attenuated neurocognitive development as described earlier in relation to scarring in the context of neurodevelopmental processes. The aim of the figure is to simplistically illustrate the concepts of trait, state, and scar impairment. Table 1 outlines the various study designs that can provide evidence for trait, state, and scar patterns of neurocognitive impairment in MDD.
Evidence according to these different study designs is examined in the following sections.

| Family studies
Having a family member with MDD places an individual at increased risk of the disorder; neurocognitive impairment may be a heritable trait marker for MDD. However, there are relatively few studies that have examined the neurocognitive functioning of unaffected firstdegree relatives of people with MDD in general, let alone in young people. A study of unaffected children (mean age 9 years) (Micco et al., 2009)  Deficits co-occur with depressive symptomatology and increase or decrease with exacerbation or resolution of depressive symptomatology.
Deficits are present after remission from a depressive episode and are worse than before the episode. Deficits progress with increased severity and chronicity of MDD.
| (5 of 12) e00527 mean age 16 who had a parent with MDD (Belleau et al., 2013), and the other examined the relationship between mothers' self-reported depressive symptoms prospectively assessed over 4 years (when their children were aged 2-6 years) and the executive functioning of their children at age 6 (Hughes, Roman, Hart, & Ensor, 2013). Both studies showed a relationship between parental depressive pathology and neurocognitive impairments in their nondepressed children. However, it cannot be determined whether these impairments are related to biological or environmental factors or a combination of both.
Studies of twins raised in the same environment provide a more powerful avenue for examining this issue (Hsu et al., 2014). To date twin studies have only been conducted in adults. An early study compared the neurocognitive performance of unaffected twins from monozygotic and dizygotic twin pairs, where either their co-twin was affected by MDD (placing them at high risk for MDD) or their co-twin was also unaffected (placing them at low risk for MDD) (mean age ranged from 38 to 48 years across the twin combinations) (Christensen, Kyvik, & Kessing, 2006). After controlling for demographic and clinical variables, the healthy twins with a co-twin affected by MDD showed impairments in attention, working memory, executive function, language processing, and memory compared with those who did not have a twin affected by MDD, supporting the notion of genetic liability for neurocognitive impairment in MDD (Christensen et al., 2006). Similarly, a more recent study of adult (mean age 45 years) monozygotic twins discordant for MDD found that after controlling for age and gender there was evidence for impairments in attention and general knowledge in both the affected and unaffected twins, compared to twin pairs without a history of MDD (Hsu et al., 2014).
Furthermore, probands with early-onset depression (<18 years of age) performed more poorly on a visuoconstructional task than their unaffected co-twins. Current level of depressive symptoms was unrelated to neurocognitive performance (Hsu et al., 2014). Findings from these studies provide some evidence for neurocognitive impairment as a heritable trait in MDD, but the latter study also suggests that depression during adolescence may leave a neurocognitive scar independent of genetic vulnerability (Hsu et al., 2014).

| Premorbid studies
Premorbid studies include large birth or military conscription cohort studies that have examined the neurocognitive functioning of generally healthy individuals (children or adolescents) before they experience the first onset of psychiatric disorder, including MDD. Although these studies involve longitudinal designs, they are included in this section because neurocognitive functioning is only examined at one initial time point and change in neurocognitive functioning cannot be determined. Most of these studies have examined IQ rather than specific neurocognitive domains. Lower premorbid IQ has been associated with a mildly increased risk of a subsequent diagnosis of severe depression requiring hospital admission (OR = 1.22) (Zammit et al., 2004) and an ICD-8 diagnosis of 'depressive neurosis' (OR = 1.08) (David et al., 2008). Another study showed that for each standard deviation increase in childhood IQ there was a 23% reduction in the odds of having an adult diagnosis of major depressive episode, whereas lower IQ was associated with a greater risk of persistent depression between the ages of 18 and 32 (Koenen et al., 2009).
Finally, Glaser et al. (2011) investigated the association between IQ assessed at age 8 years with self-reported depressive symptoms at 11, 13, 14, and 17 years. Interestingly, they found that there was an association between childhood IQ and depressive symptoms in adolescence, but that the direction of the relationship varied according to age and pubertal stage (assessed using a scale of pubic hair and in females also breast development). Specifically, lower IQ was associated with higher depressive symptoms at age 11, but this association was reversed at ages 13 and 14, such that higher IQ was associated with higher depressive symptoms. This latter association persisted for 17-year-old males, but not females. These findings were broadly replicated by a similar recent longitudinal study (Weeks et al., 2014). The authors speculated that these complex findings may be associated with the onset and offset of pubertal changes, which are known to be associated with IQ (Ramsden et al., 2011) and to differ for females and males (Glaser et al., 2011;Weeks et al., 2014). Although MDD was not the outcome of either study, the findings highlight age, gender, and developmental factors (such as puberty) as important considerations when investigating neurocognitive functioning in the context of emerging and full-threshold MDD.
Whether specific neurocognitive deficits (as opposed to general IQ) are evident before the onset of a first episode of MDD, including during 'prodrome' periods, remains an open area of investigation, as defining the risk syndrome or prodromal phase of MDD has been a challenge for the field (Hetrick et al., 2008). A prospective birth cohort study that assessed individuals from the age of 3 found that there was no difference between healthy controls and individuals with a diagnosis of 'depression or anxiety disorder' at age 26, in their motor, expressive, and receptive language abilities assessed at ages 3, 5, 7, and 9 (Cannon et al., 2002). However, individuals who were diagnosed with 'depression or anxiety disorder' at age 26 displayed significant deficits in psychomotor speed and attention at age 13 (Cannon et al., 2006). A challenge in interpreting these findings is that it remains unclear when participants first started experiencing depressive (or anxiety) pathology and it is possible that subsyndromal depressive or anxiety symptoms were present during early adolescence. In a study of 192 young adolescent school students (mean age 12.4, range 9-15 years) Evans, Kouros, Samanez-Larkin, and Garber (2016)  prospective association between time 1 executive functioning and times 1 and 2 depressive symptoms (as well as other psychopathology) in a community sample of 220 adolescents (mean age 13.7; range 11-16 years). The sample was 'enriched' with respect to internalizing and externalizing problems. There was no association between time 1 executive functioning and concurrent self-reported depressive symptoms, but mother-reported symptoms of depression in their offspring was associated with executive functioning at time 1 (Han et al., 2016).
However, while controlling for IQ, gender, and age, executive functioning at time 1 was not associated with self-or mother-reported depressive pathology at 2-year follow-up (Han et al., 2016). An earlier study found no premorbid impairment in executive functioning, psychomotor speed, and attention in adolescents (mean age 14.8 years) who were later diagnosed with MDD in early adulthood (mean age of 21.7 years) (Meyer et al., 2004).
Whereas lower IQ appears to have an association with increased risk for MDD, the evidence regarding specific neurocognitive domains, particularly executive functioning impairments, as risk markers for MDD is limited and inconclusive.

| Current depression
Many cross-sectional studies have examined neurocognitive functioning in currently depressed individuals, with numerous reviews having been conducted, showing evidence for small-to-moderate impairments in several neurocognitive domains (Austin, Mitchell, & Goodwin, 2001;McDermott & Ebmeier, 2009;Rock et al., 2014;Snyder, 2013). Significant executive functioning impairments were evident in the currently depressed sample, but not in those in remission (Maalouf et al., 2011). Although these findings suggest that impairments are most likely state related, the sample size was small and cross-sectional comparison of independent groups does not rule out the possibility of pre-existing trait-related impairments in the currently depressed group, or of the later emergence of 'scarring' in the remitted adolescent group. As discussed previously, in addition to longitudinal studies, large studies of remitted MDD that demonstrate intact neurocognitive functioning would provide better support for the state-related model.

| Remitted depression
Several cross-sectional studies have assessed neurocognitive functioning in adults in remission from MDD (usually defined by a cut-off score on clinician-rated scales maintained over a specified duration). A systematic review and meta-analysis of 27 studies by Bora, Harrison, Yucel, and Pantelis (2013)) found a broad range of neurocognitive impairments of small-to-medium effect size (d = 0.39-0.59) in adults in remission from MDD relative to healthy controls (Bora et al., 2013).
This review included a subanalysis of early-versus late-adult onset MDD samples, and found that late-onset MDD (onset after age 60) was associated with significantly larger neurocognitive impairments compared with early-onset MDD, suggesting different etiological mechanisms possibly involving vascular and neurodegenerative factors in late-onset MDD (Bora et al., 2013). As the 'early-onset' | (7 of 12) e00527 samples included people up to the age of 59 years, specific interpretations regarding the impact of MDD in adolescence or early adulthood cannot be made. Nevertheless, their meta-regression analysis examining the impact of early-and late-onset showed that the number of previous episodes and duration of illness had no significant influence on the degree of neurocognitive deficits observed in euthymic MDD patients (Bora et al., 2013). This finding does not support a progressive 'scarring' pattern of neurocognitive impairment associated with illness chronicity.
A second systematic review and meta-analysis by Rock et al. (2014) focused only on studies that used the Cambridge Neuropsychological Test Automated Battery (CANTAB) to assess neurocognition; six studies of people in remission from MDD were included. In this meta-analysis, remitted MDD was associated with significant moderate deficits compared to healthy controls in executive functioning (d = 0.53-0.61) and attention (d = 0.52). The authors did not examine whether there was a relationship between neurocognitive impairments and illness characteristics such as number of episodes, age of onset, and depression severity.

Research involving youth in remission from MDD remains limited.
Of the included studies in the Bora et al. (2013) meta-analysis, the lowest mean age of participants was 34 years, with no studies focusing on youth specifically, and in the Rock et al. (2014) meta-analysis only one study involved young people (Maalouf et al., 2011). This was the study by Maalouf et al. (2011) described earlier with their finding of no evidence for impairments during remission in 20 people aged 15 years on average. In contrast, Kyte et al. (2005) found that adolescents (mean age 15 years; n = 30) who had a diagnosis of first-episode MDD in the previous year and whose current mean mood ratings were not significantly different from n = 49 healthy controls (suggesting many were likely in remission), displayed impulsivity on a decision-making task.
Medication use and previous depression severity did not influence the findings (Kyte et al., 2005). Another study of 42 young adults (mean age 21.3 years) who had been in remission from recurrent MDD for at least 1 month found significant moderate-to-large impairments in immediate verbal memory, processing speed, and executive functioning relative to healthy controls (n = 33) (Smith, Muir, & Blackwood, 2006). Without data on neurocognitive functioning prior to MDD, it remains unclear whether the findings of the latter two studies reflect trait-and/or scar-related impairments. Furthermore, residual (still resolving) state-based neurocognitive effects may have been present in these samples.

| Longitudinal studies with repeated neurocognitive assessment
A systematic review of 30 studies by Douglas and Porter (2009) that examined the longitudinal course of neurocognitive functioning and its relationship to symptomatology in adults with MDD revealed that improvements in verbal memory, verbal fluency, and psychomotor speed were closely related to improvements in mood, whereas attention and executive functions remained impaired despite clinical status-suggestive of trait-and also possibly scar-based impairment (Douglas & Porter, 2009). However, a drawback of many of the studies included in the Douglas and Porter (2009)  They examined whether any impairments observed during remission resolved, remained stable, or got worse at follow-up assessment 3-15 weeks later (Peters et al., 2015). Only a single small-tomoderate deficit (d = 0.38) in cognitive control (the ability to regulate and inhibit responding based on previously presented information) was observed in remitted youth at time 1; this impairment remained stable at time 2. The impairment in cognitive control was unrelated to residual depressive or anxiety symptoms, number of prior depressive episodes, age at onset, number of hospitalizations, longest episode duration, years since last episode, or ever being prescribed medication (Peters et al., 2015).
An earlier prospective study by Schmid and Hammar (2013) assessed the neurocognitive functioning of slightly older young adults (mean age 26 years; n = 28) during their first episode of MDD and again at 12 months follow-up, compared to a healthy control group (Schmid & Hammar, 2013). Similar to Petersen et al., there was again evidence for persistent executive functioning impairments (specifically in inhibition, switching, and semantic fluency), despite a reduction in depressive symptoms over the 12-month period.
Furthermore, Schmid and Hammar (2013) found that participants with poorer performance in inhibition/switching in the acute phase of illness (time 1) had a greater tendency to experience a relapse within the first year (Schmid & Hammar, 2013). Another recent study administered a range of subtests from the CANTAB to adolescents aged 13-18 years (n = 13) who had recently recovered from a major depressive episode and again at 2-month follow-up, compared to healthy controls (Bloch et al., 2015). They found that while attention improved over the 2-month period, suggestive of a state-related impairment, stable visual working memory impairments were evident over the follow-up period (Bloch et al., 2015).   (Belleau et al., 2013;Hughes et al., 2013) and two did not (Klimes-Dougan et al., 2006;Micco et al., 2009). Unaffected twins of adult twin pairs discordant for MDD were found to be impaired in various neurocognitive domains (Christensen et al., 2006;Hsu et al., 2014), but it was not clear when the affected adult twins first experienced MDD, thus developmental factors could not be considered. To our knowledge there are no twin studies conducted in youth discordant for MDD, or are there studies that examine neurocognitive functioning in unaffected first-degree relatives longitudinally. Premorbid studies are similarly equivocal, with some evidence that lower premorbid IQ or executive functioning is associated with a mildly increased risk for MDD (David et al., 2008;Evans et al., 2016;Zammit et al., 2004), whereas other studies found no such link (Han et al., 2016;Meyer et al., 2004 and there was no assessment prior to remission (Kyte et al., 2005;Smith et al., 2006).
Prospective longitudinal studies that assess neurocognition prior to, during, and after initial and subsequent episodes of major depression are the most definitive for informing the type of neurocognitive patterns observed in MDD. This is particularly the case for trait-versus scar-related impairments. A growing number of longitudinal studies with repeated neurocognitive assessment have been conducted in young people (Bloch et al., 2015;Peters et al., 2015;Schmid & Hammar, 2013), but only two have assessed neurocognition prior to the onset of MDD (Beaujean et al., 2013;Vijayakumar et al., 2016).
Longitudinal studies have primarily focused on executive functioning due to the developmental relevance of this domain in adolescence and young adulthood (Anderson et al., 2001), as well as the established neurobiological and psychological links between executive functions and depression (Snyder, 2013). The longitudinal studies without a premorbid neurocognitive assessment have suggested mild trait-related neurocognitive impairments in youth with MDD because the impairments remained relatively stable and unrelated to illness severity (Bloch et al., 2015;Peters et al., 2015;Schmid & Hammar, 2013). In contrast, the two studies with premorbid assessment did not support the trait model (Beaujean et al., 2013;Vijayakumar et al., 2016). These two latter studies by Beaujean et al. (2013) and Vijayakumar et al. (2016) suggested that MDD experienced during adolescence may be associated with neurocognitive scarring (specifically attenuated neurocognitive development). Importantly, these studies had sufficiently long follow-up periods (ranging from 4 to 8 years) to enable the detection of possible scarring effects of MDD, whereas the other longitudinal studies only had follow-ups of only weeks to months (Bloch et al., 2015;Peters et al., 2015;Schmid & Hammar, 2013), which is likely an insufficient time period for a scarring effect to be observed, particularly in adolescents who are still developing. As discussed earlier it may be that scar-related neurocognitive impairments in youth may not become clearly evident until the young person has passed through the critical neurodevelopment that occurs during adolescence. Whereas the examination of neurocognitive functioning prior to the first onset of depressive symptoms and full-threshold MDD remains a significant research challenge, this must be a priority for definitive findings regarding progressive neurocognitive scarring as a consequence of the illness. The studies by Beaujean et al. (2013) and Vijayakumar et al. (2016) provide a positive first step in achieving this.
What emerged from examination of this literature was the importance of taking into consideration factors that may confound the neurocognitive findings. Particularly, age of onset and pubertal stage seem to be important considerations in the severity of neurocognitive impairments early in the course of illness. Two reviews showed that older age of onset is associated with poorer neurocognition (Bora et al., 2013;Lee et al., 2012) and indicated that neurocognitive impairment in older people with MDD may have a different underlying mechanism. These findings highlight the importance of investigating youth populations separately and to take into account developmental factors, such as pubertal stage. The important moderating role of factors such as age of onset, puberty, IQ, family history, gender, medication status, duration of illness, and number of episodes need to be factored into the designs of future studies. A limitation of many extant studies in youth MDD is that neurocognitive assessment was limited to a small number of domains. Gleaning from the literature, it may be that state-, trait-, and scar-based impairments occur differentially in different neurocognitive domains. For example, attention (i.e., concentration, working memory) and processing speed domains may be more effected by clinical state (in line with the clinical symptoms of depression) (Bloch et al., 2015;Lee et al., 2012), whereas higher-order executive functions may be more permanently impacted (Belleau et al., 2013;Evans et al., 2016;Vijayakumar et al., 2016). It is recommended that future studies examine a broad range of neurocognitive domains to test this possibility.
A clinical staging heuristic has been proposed as a potentially useful framework for characterizing individuals in the early stages of MDD and providing the opportunity for early intervention (Hetrick et al., 2008;McGorry et al., 2007). Neurocognition has been suggested as a possible important biomarker in the staging model broadly, although research on the staging of neurocognition in MDD has received limited investigation (Lin, Reniers, & Wood, 2013;McGorry et al., 2014). We hope this overview is a catalyst for neurocognition to be an import- We would like to thank graphic designer, Bernie Cram, for designing

FUNDING INFORMATION
The University of Melbourne (Grant/Award Number: 'Ronald Philip Griffiths Fellowship').

CONFLICT OF INTERESTS
None declared.

ENDNOTE 1
Evidence for these latter two study designs are beyond the scope of this article as they do not focus specifically on MDD.