Attention and processing speed dysfunction

Attention and processing speed deficits on neuropsychological measures are some of the most consistently reported impairments in active state MDD. Individuals with MDD perform poorly relative to matched healthy controls, generally with small to medium effect sizes. Some studies suggest that MDD is associated with impairments of more complex, effortful attention (selective attention), but not sustained attention (e.g. Hammar, Lund, & Hugdahl, 2003). Other studies have reported that individuals with MDD demonstrate slower reaction times regardless of the complexity of the task (e.g. Pelosi, Slade, Blumhardt, & Sharma, 2000). These deficits in attention and processing speed are not restricted to the acute MDD state, as longitudinal studies report persistent cognitive difficulties over time, even in the relative absence of depressive symptoms (Weiland-Fiedler et al., 2004). One systematic review (Hasselbalch, Knorr, & Kessing, 2011) found that persistence of impairment in the remitted state may be particularly robust for selective attention. This persistence of impairment in these and other areas of cognitive functioning in remission has motivated debate as to whether neuropsychological deficits constitute a state effect (related to symptoms), scar effect (occurring after the “injury” of an MDE), or a trait that predates MDD onset. Low rates of MDD remission exist among middle-aged or older individuals with a history of repeated episodes (Hasselbalch et al., 2011). In contrast, younger remitted individuals in the early course of MDD are not generally impaired in attention or processing speed; the pattern of cognitive impairment in this population is specific to higher-order executive functions (Purcell, Maruff, Kyrios, & Pantelis, 1997). These findings raise the possibility that simple processing of over-learned tasks and sustained performance remain intact in early illness course but worsen with additional MDEs.

There is broad consensus that CCN dysregulation primarily underlines the processes of attention and processing speed. However, the heterogeneity of tasks and samples used in studies to date obfuscate whether the central dysfunction is one of prefrontal hyperreactivity or hyporeactivity, and whether directionality is specific to certain domains of cognitive tasks. In addition, whereas the CCN functions might be dissected into more basic (attention, processing speed) components as separate from more integrative components (e.g. working memory, executive functioning), there is less clear dissection of the regions that contribute to basic as opposed to more integrative functions in neuroimaging studies. Some define more posterior (e.g. inferior parietal) and motor regions in the network as more basic functions, yet a working memory task like the n-back demonstrates greater difficulty level modulation in the inferior parietal lobule than in the dorsolateral prefrontal cortex (DLPFC).

Neuroimaging studies of sustained and selective attention have noted hyperactivation of the left DLPFC and anterior cingulate (Wagner et al., 2006), although others have found decreased left mid-cingulate activation (George et al., 1997). Given that individuals with MDD do not consistently underperform on certain types of attention tasks relative to healthy controls, it has been proposed that when behavioral performance is preserved, depressed patients need greater activation within the same neural network to maintain this similar level of performance (Langenecker et al., 2007). With regard to processing speed, some studies have reported that attenuated activation of the left prefrontal cortex and anterior cingulate is associated with impaired verbal fluency (Okada, Okamoto, Morinobu, Yamawaki, & Yokota, 2003); however, largely, fMRI studies of simple processing speed are scarce. The use of a standardized fMRI battery to define which CCN areas contribute to both cognitive and emotion-related dysfunctions in MDD, and which reflect specific contributions in direction and location of effect may ameliorate some of these confounds.

Relationship of attention and processing speed to functional outcomes and treatment response

Deficits in attention and processing speed are important for understanding functioning and illness course. Cognitive deficits, broadly speaking, have been linked to considerable disability and limited functional recovery (Godard, Grondin, Baruch, & Lafleur, 2011). It was traditionally believed that functional disabilities in people with MDD normalize with remission of the depressive episode. However, in recent years it has become clearer that functional impairments persist longitudinally, even in the absence of depressive symptoms. Most studies linking cognitive deficits and functioning in major depression, however, are cross-sectional, and therefore minimally informative about temporal mechanisms that maintain impairment. Only select studies have demonstrated that cognitive deficits may predict later functioning. For example, one longitudinal study demonstrated that attention deficits (among additional cognitive impairments) were prospectively associated with global ratings of functional disability (Godard et al., 2011). Another found that attention/processing speed deficits specifically predicted impairment in social functioning over time (Sarapas, Shankman, Harrow, & Faull, 2013). A recent study of MDD and bipolar disorder (BD) reported that attention was predictive of maximum work status and relations with children (Godard et al., 2011). Thus, attention and processing speed may be viewed as part of a non-specific cognitively impaired profile at risk for global impairment, but a more specific vulnerability for depression-related occupational and social deficits. Although few studies have longitudinally evaluated the predictive value of attention and processing speed on depressive recurrences, one small study of patients with MDD or BD suggested that impaired divided attention at discharge may be at greater risk for future relapse (Majer et al., 2004). Impairment in functioning accumulates with repeated depressive episodes, thus additional research on the specificity of a cognitive profile predicting future relapse has the potential to narrow treatment targets and potential points of intervention.

In accord with the goal to better understand targets of treatment and in an effort to reduce the functional burden associated with cognitive impairment, recent studies have evaluated the role of neuropsychological functioning in treatment response. Although this remains a young area of research, a few studies have reported that baseline processing speed may interfere with antidepressant treatment (Gudayol-Ferré et al., 2013), which is thought to represent a dopaminergic deficit unresponsive to selective serotonin reuptake inhibitor (SSRI) treatment. However, another study did not replicate the specificity of this link (Gorlyn et al., 2008), indicating that responders outperformed non-responders across all cognitive domains. It has also been reported that simple cognitive tasks do not discriminate treatment responders from non-responders, but rather impaired baseline performance on complex cognitive tasks is associated with a greater likelihood of antidepressant response (Kampf-Sherf et al., 2004). Thus it remains unclear whether simple cognitive deficits such as processing speed represent a unique predictor of treatment responsiveness, or whether a more broadly impaired cognitive profile may constitute a treatment-resistant phenotype.

Dysfunction in working memory and executive functioning

Executive dysfunction, including working memory dysfunction, has been extensively reported in MDD. There are also underpowered studies that have reported negative effects. Unfortunately, in comparison to dementia and schizophrenia, the relatively smaller nature of difficulties with working memory and executive functioning in MDD has led to an underestimation of the degree of impact that these weaknesses might have.

Working memory difficulties in depression are less frequently reported than those observed in attention, processing speed, and executive functioning, yet as mentioned, evidence exists that WM is impaired in MDD and related mood disorders (McIntyre et al., 2013). A study by Pelosi et al. (2000) was illustrative of the working memory challenges that can be present in MDD. They used a Sternberg memory store task with memory probes from three to five digits. Untreated MDD patients demonstrated more errors as memory load increased, and this response pattern was associated with changes in ERPs at Fz, Cz, and Pz in early sensory and later memory storage time courses. Pelosi and colleagues also reported significantly poorer arithmetic subtest scores from the Wechsler Intelligence Scales in the patient group. Arithmetic has both working memory and executive functioning elements subsumed within it. The study highlighted how those with severe, untreated depression often do have working memory and executive functioning difficulties. Visual working memory difficulties are also present in those with recurrent MDD who are currently in the euthymic state (Weiland-Fiedler et al., 2004).

Several studies have reported executive functioning problems in MDD. In a review by Rogers and colleagues (2004), small sample sizes and heterogeneous measures prevented firm conclusions on whether executive functioning deficits were present, or if working memory deficits were also present. As effect sizes and a complete meta-analysis were not reported, it is difficult to estimate the extent of these cognitive difficulties and what impact they might have on functioning and outcomes. Research has found that verbal working memory was distinctly more impaired in a treatment-resistant depression sample, and that neuropsychological deficits were predictive of functional outcomes (Majer et al., 2004). Furthermore, a recent meta-analysis reported that working memory and executive functioning deficits in MDD were broad with moderate effect sizes, although significant heterogeneity was observed across studies (Snyder, 2013). An illustrative set of studies by Channon and Green (1999) demonstrated that problem-solving difficulties in MDD were present in more ambiguous contexts, and could be remedied by providing greater structure and contextual cues.

Neuroimaging studies of executive functioning and working memory indicate that there are nuanced differences in MDD subjects. As noted in our previous work (Langenecker et al., 2007), in the context of equal performance, there is often hyperactivation in the CCN regions for working memory and executive functioning tasks. When performance is decreased, there is often reduced activation in the MDD group. Task difficulty and complexity of control conditions can also drive these inconsistencies in between group differences. Furthermore, when emotion conditions are included, requiring either natural or explicit regulation, MDD individuals often exhibit hyperactivation in cognitive control regions (Hamilton et al., 2012).

Relationship of executive functioning and working memory to functional outcomes and treatment response

A sizeable number of studies have now suggested that executive functioning measures effectively predict treatment response (see Fu, Steiner, & Costafreda, 2013; Pizzagalli, 2011). In particular, neuroimaging studies with sustained attention and set-shifting parameters, both with and without the presence of emotional content, suggest that open-label treatment response can be effectively predicted (Pizzagalli, 2011). The rostral and dorsal anterior cingulate is a key nexus in these predictive studies. Intriguingly, following cingulotomy, patients with chronic, treatment-refractory depression were impaired on identification of dynamic emotional stimuli compared with a group of non-surgically treated patients with depression. These differences could be related to processing of emotions or sustained attention to salient cues (Ridout et al., 2007). On a behavioral level, slowed processing speed, poorer sustained attention, and greater difficulty with interference are strong predictive markers, but markedly less robust than the neural activation predictors. Processing speed with interference resolution, a factor integrating tests like Digit Symbol Substitution, Stroop, Trails, and Parametric Go/NoGo tests, is predictive of occupational status in those with BD and in healthy controls, including being unemployed or underemployed (Ryan et al., 2013). Another recent study, although with a small sample, demonstrated that executive functioning and verbal memory were predictive of occupational status, relations with children and global psychosocial functioning in MDD and BD (Godard et al., 2011).

Emotion perception dysfunction

Individuals with MDD demonstrate impairments in identifying emotional stimuli, particularly facial expressions of emotion, and show a negativity bias whereby they rate emotional stimuli less positively and/or more negatively than healthy controls (see Elliott, Zahn, Deakin, & Anderson, 2011), although not always (Ridout et al., 2007). A meta-analysis that included 20 studies of patients with MDD found a moderate effect size for emotion perception deficits (Kohler, Hoffman, Eastman, Healey, & Moberg, 2011). Observational studies have reported that poor facial affect discrimination in patients with MDD has been correlated with severity of negative affect (Elliott et al., 2011). Judgments of negative emotions in facial expressions have also been related to depression severity and persistence, although enhanced sensitivity for sad facial expressions and negatively biased automatic processing of neutral and happy faces has also been reported (Elliott et al., 2011).

Functional neuroimaging studies in individuals with active MDD have found that emotion processing is generally associated with enhanced activity in the EN and SN. For example, many studies have demonstrated in patients with MDD increased activation of the amygdala in response to emotional stimuli, particularly emotional facial expressions (Elliott et al., 2011), even those presented below awareness (Roiser et al., 2012), although increased amygdala activation in response to emotional stimuli is not always found (Elliott et al., 2011).

Differences between MDD and healthy controls have also been reported in other regions of the emotion and salience networks. Hyperactivation or lack of deactivation of the rostral anterior cingulate cortex (rACC) in patients with MDD has been reported in numerous affective tasks, including those involving self-referential processing of negative words, affective evaluation and cognitive reappraisal of negative pictures, and in response to sad words in emotional Stroop and affective Go/NoGo tasks (Pizzagalli, 2011). Subgenual ACC (SGAC) hyperactivity in response to emotional stimuli has been reported in MDD (Fitzgerald, Laird, Maller, & Daskalakis, 2008; Pizzagalli, 2011). In contrast, reduced rACC activation has also been found at rest and in response to both happy and sad stimuli in a meta-analysis of patients with MDD (Fitzgerald et al., 2008).

Activity in other regions of the EN and SN, including the ventral striatum, insula, caudate, and orbitofrontal cortex (OFC), has also differentiated MDD and healthy control groups (Elliott et al., 2011). For example, reduced activation has been reported in the insula in MDD patients at rest and during negative stimulus presentation (Elliott et al., 2011; Fitzgerald et al., 2008). Researchers have also reported increased activity at rest in the caudate and in response to sad stimuli in the putamen (Fitzgerald et al., 2008), enhanced OFC responses to negative pictures (Elliott et al., 2011), and reduced OFC responses to happy stimuli in patients with MDD (Fitzgerald et al., 2008). Harmer and colleagues (Harmer, Cowen, & Goodwin, 2011) argue that anhedonia, which is a key symptom of MDD, could not only be related to deficits in emotion perception, but that it could also involve deficits in the reward processing regions of the EN. Indeed, unmedicated patients with remitted MDD have also demonstrated reduced ventral striatal response to rewarding stimuli and enhanced responses in the caudate to aversive stimuli (Harmer et al., 2011).

Intriguingly, the CCN is often under- or over-activated in affective paradigms that may or may not have an explicit interference or cognitive control condition. Hypoactivation in patients with MDD has been reported in the CCN, particularly the DLPFC and dorsal cingulate, including during affective tasks, whereas other studies have reported hyperactivation in these same regions for emotion processing paradigms (reviewed in Bricen∼o et al., 2013). Recently, Spielberg and colleagues (2014) offered some insight into why this may occur, positing that different subtypes of MDD, based upon anxiety symptoms and personality traits, may demonstrate hyper or hypoactivation of DLPFC, OFC, and amygdala, including differences in coupling between these regions based upon anxiety and depression symptoms.

Relationships of emotion perception to functional outcomes and treatment response

Emotion perception has not been frequently evaluated when considering functional outcomes and treatment response, although emotion perception task performance has been significantly correlated with relapse (Bouhuys, Geerts, & Gordijn, 1999). While some studies have reported improvement in emotion perception following stabilization of depression symptoms, others have reported persistence of emotion recognition deficits following remission of MDD (Joormann & Gotlib, 2007; Kohler et al., 2011; Roiser et al., 2012). Persistence of ventromedial prefrontal cortex (vmPFC) hyperactivation to sad stimuli is observed in adults with remitted MDD and is predictive of relapse (Farb, Anderson, Bloch, & Segal, 2011).

Amygdala response to emotional stimuli has been found to predict long-term MDD prognosis (Fu et al., 2013), and this amygdala hyperactivation has been reported to normalize after antidepressant treatment (Elliott et al., 2011). However a meta-analysis of medication-free, active MDD patients reported conflicting results concerning the role of the amygdala in predicting treatment response (Fu et al., 2013). Increased reactivity in the amygdala and caudate in response to happy facial expressions has also been reported following antidepressant treatment, in the absence of behavioral change (Harmer et al., 2011). Pizzagalli (2011) conducted a meta-analysis of 23 studies (n = 426 MDD patients) of resting rACC activity and treatment response. The study included multiple imaging and treatment modalities, and found a robust relationship whereby increased rACC activation was linked to better treatment response across cognitive control, affective perception, affective bias, and cognition/emotion interactive paradigms. A more recent meta-analysis of medication-free patients with active MDD also supported the well-replicated finding that increased baseline rACC activity as well as increased activation in the medial PFC predicted higher likelihood of clinical improvement (Fu et al., 2013). In contrast, another meta-analysis found that reduced activation in insula activity post-treatment decreased (Fitzgerald et al., 2008). This same meta-analysis found both increased and decreased activity in the putamen following treatment. Additional support for the importance of the striatum in MDD comes from research reporting that electrical stimulation of the striatum has been effective for treatment-resistant depression (Fu et al., 2013).

Associations between poor emotion perceptual abilities and worse psychosocial outcomes such as social functioning/competence, relationship well-being, and occupational functioning have been reported in healthy individuals (Kohler et al., 2011; Ryan et al., 2013), patients with BD (Ryan et al., 2013), and schizophrenics (Ridout et al., 2007). Thus it is likely that a similar relationship occurs in MDD; however, there is a lack of investigation into the effects that emotion perceptual abilities have on such functional outcomes in MDD.

Dysfunction via emotion distortion and biases

Individuals with depression demonstrate biased memory processes, including preferential recall of negative information (Mathews & MacLeod, 2005) and a tendency toward over-general memory rather than recalling specific events (Williams et al., 2007). Individuals with MDD may have the ability to perform in environmentally structured situations, but may experience more difficulty in unconstrained situations (Gotlib & Joormann, 2010).

As an example, rumination is a perseverative negative thought pattern that involves passively dwelling on negative feelings (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008) and may also incorporate biases and distortions. Rumination impairs functional behavior by leading an individual to become stuck in thought and blocking approach actions (Watkins & Moulds, 2005), leading to the proposal that rumination represents a trans-diagnostic mechanism in the development of psychopathology (Nolen-Hoeksema & Watkins, 2011). Rumination hinders the reciprocal and dynamic interchange between cognitive and emotional states. For example, rumination interferes with effective problem-solving (Watkins & Moulds, 2005) and instrumental behavior (Lyubomirsky & Nolen-Hoeksema, 1993). Prospective research conducted with 200 adolescents enrolled at age 12 or 13 found that higher levels of baseline rumination were associated with decreases in selective attention and attentional switching at follow-up (Connolly et al., 2014). As such, rumination processes can detract from, interfere with, or undermine adaptive outcomes requiring cognitive control.

Relationships of emotion biases to functional outcomes and treatment response

Limited work has directly explored the relation between emotion distortions, biases, rumination, and functional outcomes, although one longitudinal study has found that part of the observed relationship between rumination and interpersonal stress is accounted for by discontent with social support (Flynn, Kecmanovic, & Alloy, 2010). This is in line with integrated cognitive-interpersonal vulnerability to depression models, which suggest that individuals who are prone to depression generate more stress, are more reactive to rejection, and drive their social support away through negative habits such as rumination and excessive reassurance seeking.

Rumination also prospectively predicts suicidal ideation (Miranda & Nolen-Hoeksema, 2007), predicts both chronic and acute work-related fatigue, and appears to function through reduced sleep quality (Querstret & Cropley, 2012). In contrast, frequent use of cognitive strategies such as cognitive reappraisal is associated with the experience and expression of greater positive and less negative affect, increased well-being, and better interpersonal functioning (Gross & John, 2003). Similarly, in a series of studies, Bouhuys and colleagues (1999) found that a tendency to label ambiguous schematic faces as more negative was related to depression persistence six weeks following hospital admission. Furthermore, facial emotion recognition biases are not restricted to those with an MDD diagnosis, as they have also been reported in individuals who have never experienced a depressive episode but have an increased familial risk of depression, suggesting that this may be a trait risk factor for MDD (Harmer et al., 2011; Roiser et al., 2012).

Memory biases are also predictive of the course of MDD, particularly for autobiographical and emotionally salient events or information. For example, intrusive memories of life events have been found to predict recovery, and severe life events prior to episode onset are associated with greater changes in cognitive biases over time (Gotlib & Joormann, 2010). Moreover, the extent to which individuals retrieve over-general memories predicts delayed recovery (Gotlib & Joormann, 2010). Biases and deficits in cognition such as impaired emotion regulation (Ehring, Fischer, Schnulle, Bosterling, & Tuschen-Caffier, 2008) and rumination (Nolen-Hoeksema et al., 2008) have been observed following MDE and may make individuals vulnerable to relapse. Rumination remains elevated following remission from depression, is associated with lower levels of treatment response, prospectively predicts severity and duration of depressive episodes (see Nolen-Hoeksema et al., 2008 for a review), and mediates the effect of negative life events on subsequent affect (Moberly & Watkins, 2008).

There are tremendous opportunities for understanding the relationships between emotion perception, bias, and distortion with real-world functional outcomes. As the focus has been heavily toward mechanistic understandings of disease, these opportunities have not yet been approached or realized. In many ways, though, a focus on functional outcomes may result in increased opportunities for industry partnerships, including alternative interventions and primary and secondary prevention options that would be integrated into educational and vocational settings. There are also opportunities to use network-driven models for better understanding of points for intervention for both behaviors and functional outcomes. We discuss these networks and ways they have been targeted to predict treatment response or measure functional outcomes, given the current National Institute of Mental Health (NIMH) focus on defining and demonstrating change in disease targets.