A lifespan model of interference resolution and inhibitory control: Risk for depression and changes with illness progression

Abstract

The cognitive processes involved in inhibitory control accuracy (IC) and interference resolution speed (IR) or broadly – inhibition – are discussed in this review, and both are described within the context of a lifespan model of mood disorders. Inhibitory control (IC) is a binary outcome (success or no for response selection and inhibition of unwanted responses) for any given event that is influenced to an extent by IR. IR refers to the process of inhibition, which can be manipulated by task design in earlier and later stages through use of distractors and timing, and manipulation of individual differences in response proclivity. We describe the development of these two processes across the lifespan, noting factors that influence this development (e.g., environment, adversity and stress) as well as inherent difficulties in assessing IC/IR prior to adulthood (e.g., cross-informant reports). We use mood disorders as an illustrative example of how this multidimensional construct can be informative to state, trait, vulnerability and neuroprogression of disease. We present aggregated data across numerous studies and methodologies to examine the lifelong development and degradation of this subconstruct of executive function, particularly in mood disorders. We highlight the challenges in identifying and measuring IC/IR in late life, including specificity to complex, comorbid disease processes. Finally, we discuss some potential avenues for treatment and accommodation of these difficulties across the lifespan, including newer treatments using cognitive remediation training and neuromodulation.

Cognitive functioning shows widespread variability across the lifespan. As we deepen our knowledge of cognitive functioning, we have come to appreciate that the “when” of measurement is nearly as important as the “what” of measurement. This review will cover several important considerations in the “what” and “when” of inhibition as a broad construct, including important measurement considerations across the lifespan and across disease states, using mood disorders as an example.

Within this review, we differentiate the process of inhibition from the outcome of successful inhibition. The process of inhibition involves increasing and decreasing levels of interference resolution (IR) via experimental manipulations of distractors, timing, instruction sets, and individual differences in inherent proclivity for responding. Successful inhibition, or inhibitory control (IC), is the ability to apply the correct response or inhibit unwanted responses. This important distinction is often glossed over in current literature. For example, an “impulsive” person may have poor IC because mechanisms regulating IR are inefficient or slow, or because they are unable to dynamically modulate IC goals with changing contexts. In contrast, a person with more “interference” may test quite well on tasks of IC due to over-regulation of behavioral responses from the increased IR context. Task descriptions, as well as interpretations of a given metric, have often lacked the clarity needed to make this distinction in the literature. Yet in many real-life and laboratory contexts, individuals are rewarded with both a (1) correct response, and (2) maximal efficiency in speed. As described below, this means a greater probability of selecting the correct response and inhibiting the incorrect response, with the least interference observed.

This review focuses on a relatively narrow set of skills (IC/IR) within the broader set of executive functions. Evidence from behavioral and neuroimaging models are covered, including models of key inhibitory nodes such as the inferior frontal gyrus (IFG), supplementary motor area (SMA) and more recent network-based models involving the cognitive control network (CCN). We provide an illustrative model of the relationship between IC and IR that clarifies how these factors can be distinct and yet still modestly correlated. There are notable distinctions in the development of IC and IR early in the lifespan, although both abilities are in place by early childhood. Yet, development of these skills continues into adulthood for implementation.

We apply this account of lifespan IC/IR to a specific model of mood disorders (Peters et al., 2017), with primary focus on major depressive disorder (MDD). Neuropsychological studies in mood disorders suggest that IC/IR measures are reflective of relatively stable risk features for depression (Snyder, 2013), only slightly perturbed by the active state or progression of illness (Langenecker, Caveney, et al., 2007; Ryan et al., 2017), although see others (e.g., Sheline, Wang, Gado, Csernansky, & Vannier, 1996). Complexities arise throughout the lifespan that contribute to difficulty measuring IC/IR in mood disorders. Co-occurring anxiety may change an individual’s valuation of errors, modifying response biases, and thereby masking IC/IR deficits (Crane et al., 2016). During the middle and later stages of life, multifactorial considerations are prevalent, such as diabetes, cardiovascular illness, and endocrine disturbances. Generalizability and clinical utility of study results is compromised by the frequent exclusion of these individuals with common medical comorbidities. Lastly, we briefly provide some ideas, strategies and previous studies aimed at minimizing the effect of IC/IR difficulties (and other neuropsychological skills) in mood disorders, such as interventions, work-place or school accommodations, and newer neuromodulation strategies.

Section I. Definition of Distinction between IC and IR

It has been suggested that the term inhibition may be too broadly used (Friedman & Miyake, 2004). Moreover, definitions of the subconstruct inhibitory control (IC) bear resemblance to descriptions of another subconstruct, interference resolution (IR; e.g., Hasher, Zacks, & May, 1999). Given the concern for various and overlapping definitions, there is need to further examine the relationship between IC and IR, overlapping and specific neural underpinnings, and relations to other cognitive functions (Friedman & Miyake, 2004; Nee, Wager, & Jonides, 2007).

Nigg (2000) proposed a taxonomy of inhibitory processes, which included four types of effortful inhibition of motor or cognitive response, including: (1) interference control, (2) cognitive inhibition, (3) behavioral inhibition, and (4) oculomotor inhibition. His model described interference control as the prevention of interference that occurs as a result of resource or stimulus competition. Cognitive, behavioral and oculomotor inhibition refer to the suppression of irrelevant dominant, automatic or prepotent responses, including cognitions, eye movements and behavior that may interrupt working memory or attention. This taxonomy distinguishes control and response during interference from the ability to inhibit more automatic responses. As outlined below, we agree with a distinction between the process and final outcome, regardless of modality.

IC and IR are processes that occur at different stages, with interference mechanisms often engaged relatively earlier in the processing stream and inhibitory mechanisms engaged later (e.g., to inhibit a motor response; Friedman & Miyake, 2004; Nee et al., 2007). This time course is an important factor in differentiating IC and IR. Metrics from a variety of cognitive tasks are useful for targeting these processes. However, later introduction of interfering stimulus sets is likely to dramatically decrease behavioral inhibition accuracy and confuses some of these temporal distinctions.

A. IC/IR Within Models of Executive Functioning

Executive functioning exhibits a latent structure based on cognitive performance tasks, involving three interrelated abilities. One ability is inhibition, deliberately stopping a prepotent behavioral response (Miyake et al., 2000). Inhibition is indicated primarily by antisaccade, Stop Signal (SST), and Stroop tasks. In the unity model, broad inhibition was compiled under a common executive function component (Friedman, Miyake, Robinson, & Hewitt, 2011; Miyake & Friedman, 2012). Examination of broad inhibition revealed two latent variables: (1) ability to deliberately suppress prepotent responses, using both reaction time and error percentages on several tasks, and (2) ability to resist or resolve interference from irrelevant external information, using differences in reaction times on several tasks (Friedman & Miyake, 2004). The ability to prevent intrusions from previously-relevant material was a separate factor, consisting of differences in performance accuracy metrics. These factors were partially replicated in a widely-used executive functioning inventory, the Delis-Kaplan Executive Function System (Latzman & Markon, 2010). Importantly, performance measures included in both analyses confound speed of task completion (IR) and accuracy of task completion (IC). Notably, many tasks fail to challenge individuals at more difficult levels to obtain sufficient variability in IC, yet even with lower variability in IC, this work suggests that IC and IR are separable factors.

The general term inhibition has multiple meanings and greater precision of terminology is needed to facilitate task advancement and precision clinical applications. We maintain a distinction between IC and IR because it clarifies important aspects of task design, interpretation, underlying neural systems supporting these processes, and understanding and treating clinical conditions.

B. Tasks to Measure IC and IR

A variety of tasks have been used to assess IC and IR, including Go/No-Go (GNG), SST, Flanker, Stimulus-Response Compatibility, Stroop, and Simon tasks (see Figure 1). Each varies in the specific parameters and cognitive demands placed on individuals to adequately respond. The nature and degree of IR and IC that is required for success also varies. Based on a quantitative meta-analysis of these various tasks, Nee et al. (2007) proposed that separable interference mechanisms are employed at three different stages of processing: (1) stimulus encoding, (2) response selection, and (3) response execution. Based on neural patterns during these tasks, they found: (1) the Stroop task involved resolution at the stage of stimulus conflict, consistent with IR; (2) GNG, Flanker, stimulus-response compatibility, and Stroop tasks involved IR during response selection; and (3) GNG and SST involved restraint of inappropriate responses during response execution, consistent with IC. Early processes employed to resolve conflict may be predominantly reflective of interference mechanisms and impact the speed with which individuals process information/stimuli and respond (e.g., response time metrics), thus a measurement of IR specifically. The final product of whether a correct response is made (i.e., accuracy metrics) may reflect the resulting inhibition and resolution of IR, leading to measurement of IC specifically.

Fig. 1.

Different components of task design and implementation that affect degree of interference resolution (IR), which in turn increases or decreases the likelihood of successful inhibitory control (IC). Panel A illustrates our model of IC/IR phases, demonstrating the hypothesized effect of IR upon IC at different stages of stimulus processing and response, with a more dramatic influence of IR on IC in late stages of processing and response. Panels B through I illustrate a few different types of IC/IR tasks, including (B) flanker, (C) multisource interference, (D) Stroop/Emotion Stroop, (E) Emotional Conflict, (F) contextual Go/No-Go, (G) Go/No-Go, (H) Stop Signal, and (I) Balloon Analogue Response Test. The color coding is used as a general marker of which features of task are modified to increase IR or decrease IC.

i. Prototypic measures of IC/IR

According to our view, Stop Signal tasks (SST) and Go/No-Go (GNG) tasks are quintessential measures of IC and IR, as both can separate IC (accuracy) from IR (speed) and allow for metrics of both (Congdon et al., 2012; Friedman & Miyake, 2004). Efficiency metrics combine IC and IR, with higher scores indicating increased accuracy for those who do not have to sacrifice speed.

In GNG tasks, individuals are instructed to respond to Go stimuli, while withholding responses to rule-based No-Go stimuli or Go stimuli interrupted with an auditory or visual stop signal. Common metrics include accuracy and reaction time to correct responses (e.g., hits on Go trials, correct inhibitions on No-Go and Stop trials) and errors (e.g., commissions to No-Go and Stop trials). While reaction times are easily obtained on hit and commission trials, successful inhibition has no overt reaction time – no response is a correct answer.

In contrast, SST enables assessment of reaction times through calculation of a metric known as the stop signal reaction time (Logan, 1994). The SST is based on the horse-race model that competing “go” and “stop” processes determine whether a response is executed or inhibited (Congdon et al., 2012; Logan, 1994; Logan & Cowan, 1984). A classic model of “late” interference, if the Stop signal comes too late in the response execution phase, then a cancellation of the “go” response is not obtained. If the “go” process finishes before the “stop” signal is perceived and the “stop” process is initiated, the behavioral response is executed. Stop signal reaction time is obtained by calculating the difference between the presentation time of the stop signal (stop signal delay) and completion time of the internal stopping process at 50% chance accuracy (Logan, 1994). The SST provides a model for potentially parsing IC and IR processes, particularly if the task were designed to have stepwise increments in accuracy rates for stop delays.

ii. Convergence of Behavioral and Self-Report Measures of IC/IR

Self-report measures are used to assess aspects of inhibition, but suffer from difficulty in parsing IC and IR. Moreover, such probes introduce an additional source of variance beyond neuropsychological testing. Degree of insight and mood disturbance might cloud perception in such self-reports. Indeed, several studies show weak convergence of self-report and neuropsychological measures in clinical and non-clinical samples (Duckworth & Kern, 2011; Gomide, Sergeant, Correa, Mattos, & Malloy-Diniz, 2014; Meyer et al., 2001). However, one study using the Barratt Impulsiveness Scale reported a significant relationship with neuropsychological measures in healthy adults (Enticott, Ogloff, & Bradshaw, 2006). We recently examined convergence between neuropsychological measures, including IC, and self-report measures of IC/IR among individuals with Bipolar Disorder (BD) and healthy controls (Crane et al., 2018), finding a modest convergence of neuropsychological tasks and self-reports. There was evidence of a general IC/IR construct across most self-reports and neuropsychological tests, which were measured equally well (invariance) across healthy controls and BD (Crane et al., 2018). Such hierarchical analyses may be one avenue for convergence of neuropsychological and self-reports to elucidate the dimensional continuum of healthy to unhealthy patterns of IC/IR functioning. Unfortunately, this work did not have sufficient measures to differentiate subconstructs of IC and IR.

C. Brain-based Models of IC/IR

We highlight just a subset of models and studies here to illustrate relevant nodes and networks for IC/IR. However, it is difficult for functional neuroimaging to separate brain functioning related to IR versus IC for numerous reasons. It is common practice to collapse interference components (IR) with final inhibition performance outcome (IC), which is inherently confounded by the slow pace of the hemodynamic response. Nonetheless, some studies have begun to distinguish circuit nodes involved in IC, IR, or both.

i. Specific nodes for IC/IR

Traditional models suggest the right IFG, and specifically Brodmann Area (BA) 44, as a specific inhibition module (Aron, Robbins, & Poldrack, 2014), and highlight the presupplemental motor area (pre-SMA) and dorsomedial prefrontal cortex as supportive for transduction of the Stop signal. This Stop signal is sent through fronto-basal-ganglia circuits, including a “hyperdirect pathway” between frontal cortex, basal ganglia and subthalamic nucleus, in order to stop a motoric response (Aron et al., 2007; Swann et al., 2012). Inferior parietal lobule, anterior insula and putamen have also been implicated in IC (Langenecker, Briceno, Hamid, & Nielson, 2007). Corroborating evidence comes from human frontal lesion studies (Aron, Fletcher, Bullmore, Sahakian, & Robbins, 2003), animal models (for review, see Schall & Godlove, 2012), and detailed functional neuroimaging studies employing a variety of response modality tasks (e.g., Chikazoe, Konishi, Asari, Jimura, & Miyashita, 2007; Rubia et al., 2001).

A separate neural model for IR, particularly during working memory tasks, points to the left IFG (BA 45; e.g., Berman et al., 2011; Langenecker, Nielson, & Rao, 2004; Nelson, Reuter-Lorenz, Persson, Sylvester, & Jonides, 2009; Wimber, Rutschmann, Greenlee, & Bauml, 2009). Tasks manipulate the degree of proactive or retroactive interference associated with accuracy in selectively updating memory, and typically use semantic or verbal stimuli, a potential reason for observed left lateralization. Greater activation in pre-motor and parietal regions are involved in specific interference, compared to fronto-striatal activation for action cancellation processes (Sebastian et al., 2013).

ii. Network approaches for IC/IR

Network-based approaches support evidence for a broader network of general control and regulation functions. IC correlates more strongly with connectivity of the entire network activated, not just cortico-basal ganglia circuits (for review, see Hampshire & Sharp, 2015). Frontal-parietal regions are reliably activated across tasks and studies, specifically bilateral dorsolateral prefrontal cortex (BA 9 & 46), IFG and insula, anterior cingulate, SMA, and inferior and superior parietal lobule, in addition to regions in other networks, such as the medial PFC in the default mode network (Criaud & Boulinguez, 2013; Li et al., 2017; Nee et al., 2007; Niendam et al., 2012; Zhang, Geng, & Lee, 2017). Several common regions are active during IC and IR across tasks, including right MFG, bilateral IFG, left inferior parietal lobule, left SMA, right insula, and right superior parietal gyrus (Li et al., 2017; Zhang et al., 2017). Current meta-analyses suggest ventral and dorsal attention networks in IR (Zhang et al., 2017), compared to fronto-parietal and cingulo-opercular networks in IC (Li et al., 2017). However, task selection has not yet aligned with our distinctions between IC and IR.

Of note in this review, mood disorders show dysfunction in these networks. Neural components, including those of the IR network, were activated during a parametric GNG and were associated with accuracy and degree of treatment response to antidepressants in actively depressed individuals (Crane et al., 2017).

D. A Modified Conceptualization of IC and IR

The multitude of competing models for IC/IR likely add confusion to the terminology of IC and IR processes. Within this review, we distinguish between IC and IR when possible, and combine as IC/IR when published study details do not yet allow clarity between these processes (see Table 1). Our contribution to models of IC, IR, and their interaction, is to highlight the many design features that influence the difficulty of these tasks. In most cases, increasing difficulty is due to specific task features that increase IR, and thereby reduce the likelihood of successful correct responses (IC), highlighted within Figure 1 along with common task designs.

Table 1.

Featured Tasks and Performance Metrics Corresponding with Inhibitory Control (IC) and Interference Resolution (IR).

	IC	IR	IC/IR
All Tasks	hits, commissions, correct inhibitions, % accuracy	rt	efficiency, accuracy across varying timeframes
Task Specific:
-Go/No-Go (GNG)	-	-	-
-Stop-Signal (SST)	accuracy @ specific stop delays	ssrt	-
-Flanker	incongruent, conditional accuracy functions	-	-
-Antisaccade	incongruent
-Stimulus-Response Compatibility (SRC)	incompatible	-	-
-Stroop	incongruent	completion within timeframe	-
-Simon/Spatial Conflict	incompatible, conditional accuracy functions	-	-
-Day-Night*	-	-	-
-Trails B**	errors	-	-

Notes. Tasks which report accuracy within a delineated timeframe are indicative of IR rather than IC, such as commonly reported Stroop accuracy. Abbreviations: IC = Inhibitory Control, IR = Interference Resolution, IC/IR = both inhibitory control and interference resolution, rt = reaction time, ssrt = stop signal reaction time.

^*Typically used with young children, see Section II.

^**Typically used with older adults, see Section IV.

Tasks can be broken down into three phases based upon the timing of the stimulus, rule, or context changes (Nee et al., 2007). We add a fourth phase for post-response adjustment (or error processing) to incorporate the role of post-error slowing in IR. Prior performance and associated affective reactions can functionally change the next trial, resulting in more interference and a need for slower and more careful responses.

We contend IR can be manipulated throughout most phases of a task. Panel A within Figure 1 is a conceptual model of how IC and IR are related, with different hypothetical relationships at different phases of the task. Many early manipulations (e.g., stimulus encoding, rule cues, and staging features) modify the degree of interference that must be resolved. Later manipulations worsen accuracy (IC) in a modest linear relationship. At the middle stage of response selection, the dynamic influence of IR on IC becomes more pronounced: minor modifications to rules, competing stimuli and other features increase IR dramatically, inducing a non-linear reduction in IC. Late manipulations, including later presentation of horse-race stimuli such as in the SST, can rapidly reduce IC accuracy to chance probability levels (goal of 50% accuracy for SST). If the Stop signal is introduced after response execution is well underway, the Stop signal arrives too late to compete and accuracy is at chance levels.

Smaller sources of interference in later stages of the stimulus-selection-execution pathway will have a greater effect on IC. Most IR manipulations are placed at the first or second stage. However, feedback about errors, even after accounting for awareness of errors without explicit feedback, can induce emotional interference. This is referred to as post-error slowing. Time between stimulus trials, and strategies to accentuate the cost of errors (noxious sounds and other feedback characteristics), can influence interference carryover from one trial to another. Indeed, prior response rules, stimulus types, and correct responses can facilitate or interfere with speed and accuracy in the next trial. If response rules or stimulus types match trial-to-trial with a correct response, there is also facilitation in speed (reduced IR) and sometimes an increase in IC accuracy.

E. Importance of IC/IR in Mood Disorders

For our purposes, distinguishing between IC and IR is very important in understanding mood-related disorders, and are related to course, severity, and treatment responsiveness. Of note, there may be relative shifts (by phase of illness) in the weight or value an individual places upon accuracy in responding. Increased weighting against errors in the actively depressed state can result in increased difficulty of IR, potentially-related to psychomotor slowing in MDD. Moreover, change in weighting preserves highly-accurate IC and can easily be mistaken as unchanged inhibitory functions if the effect on IR is not considered. General slowed mentation is also possible in MDD, and speed/accuracy tradeoffs function less effectively. Both speed and accuracy are important elements in understanding overall efficiency and broader inhibition functions, as ecologically-valid contexts favor rapid, error-free behaviors.

An additional feature to IR functioning in MDD is depression-related cognitions, such as repetitive negative automatic thoughts. These intrusive and distracting cognitions have the net effect of increasing IR (and RT variability) in individuals with MDD relative to healthy controls, and fit within our notion of IR. Negative automatic thoughts and ruminative style, while distinct from IC/IR, are indeed weakly negatively related to IC (Jacobs et al., 2016; Stange, Bessette, et al., 2017; Yang, Cao, Shields, Teng, & Liu, 2017). Finally, mood symptoms/states may have different effects upon IC and IR at different development stages across the lifespan (Section III).

Section II. Developmental Trajectory of IC/IR

Developmental models have typically intertwined the metrics used to examine inhibitory control (IC) and interference resolution (IR). We note here when distinctions based on published descriptions can be made. IC and IR show significant low to moderate correlations across developmental age spans (Khng & Lee, 2014). In addition to the clear interrelationship of IC and IR across development, both have a curvilinear fit with developmental age, showing rapid improvements in early childhood, slow improvement throughout adolescence, and a plateau at adulthood. Developmental models confirm IC/IR are separate from other executive functions, with a separate developmental trajectory of these functions beginning at age 6 and remaining distinct throughout the lifespan (McAuley & White, 2011). IC shows linear improvement in children after age 6 (Williams, Ponesse, Schachar, Logan, & Tannock, 1999). Adolescents appear to have the full capacities required for accurate IC, but have irregular and inefficient execution success of these abilities, showing increased susceptibility to interference demands (Luna, Garver, Urban, Lazar, & Sweeney, 2004; Luna, Padmanabhan, & O’Hearn, 2010; Ordaz, Foran, Velanova, & Luna, 2013). One stable interference is peer presence during the task.

A. Development of IC/IR

Task difficulty and type inherently affect the relationship between IC and IR throughout development. Adult IC/IR tasks require additional skills or abilities typically gained throughout early development in addition to IC/IR abilities, such as reading automaticity; thus, few studies have examined developmental changes across childhood, adolescence, and adulthood. Modified versions ensure other age-dependent abilities play at least a lesser role in measurement and modeling of inhibitory abilities. For instance, a nonverbal Day-Night task requires inhibition of a factual response and execution of an opposite response (e.g., respond “day” for moon or “night” for sun). The updated NEPSY-II utilizes black and white shapes and upward and downward arrows for congruent and incongruent trials (Korkman, Kirk, & Kemp, 2007).

Using such tasks, IC shows constant improvement in childhood (Macdonald, Beauchamp, Crigan, & Anderson, 2014; McAuley, Christ, & White, 2011), whereas tasks with greater interference show improvement up to young adolescence (Klenberg, Narhi, Korkman, & Hokkanen, 2015). Modest improvements continue from adolescence to adulthood, concentrated in tasks with greater interference and complexity (Huizinga, Dolan, & van der Molen, 2006; McAuley & White, 2011).

Using the same oculomotor antisaccade task across three developmental stages (ages 8–30), Luna et al. (2004) found an inverse relationship between age and IC, with greatest gains in late childhood. Adolescents performed with adult-like accuracy. However, when saccades to inappropriate stimuli were stronger prepotent distractions (increasing IR), adolescents exhibited poorer IC than adults. In all, IC shows protracted development to efficient and consistent use, even if the skill set is available early in adolescence.

Similarly, specific IR has a comparable protracted course of development across childhood to adulthood. IR shows a rapid quadratic improvement, or faster successful reaction times, between early and late childhood and plateaus at adulthood (Luna et al., 2004; McAuley & White, 2011). There have been mixed findings regarding improvement during adolescence (Bedard et al., 2002; Klenberg et al., 2015; Luna et al., 2004).

There is clear evidence that IC/IR development in young children and adolescents translates to later psychosocial, academic, and health-related outcomes. Low childhood IC predicts poorer adolescent social and academic skills, self-competence, grades, and overweight status, representing a transdiagnostic risk marker (e.g., Anzman-Frasca, Francis, & Birch, 2015). Poorer IC is associated with increased exposure to childhood adversity (Marshall et al., 2016; Quinn et al., 2018). There is also strong evidence of poor IC in externalizing disorders (Oosterlaan, Logan, & Sergeant, 1998; Venables et al., 2018).

Adolescence is a time of increased risk-taking with poor outcomes associated with IC/IR decrements (Casey & Caudle, 2013). During adolescence, pressure is placed upon reliable use of IC/IR abilities due to increased expectations for independence, increased peer involvement, and reduced parental supervision. Considering that adolescents continue to develop consistent and efficient use of broad inhibition skills across contexts, especially emotionally-charged peer situations, IC/IR likely have a key role in expression and management of these contexts. Indeed, work has begun to show the interactive effects of rewarding peer situations, with peer presence as a source of increased IR difficulty (Breiner et al., 2018).

B. IC/IR in Emotional Contexts

Development of IC/IR is also influenced by the affective context(s) in which they develop, relevant to our discussion of mood disorders. Two theoretical positions dominate the literature, positing that: (1) neutral IC/IR functions develop prior to affectively-laden IC/IR (i.e., the dual-systems model), and (2) these functions exhibit linear improvement between childhood and adulthood. The relationship between “hot” (involving affective stimuli or manipulations) and “cold” IC/IR is non-linear over development, with better performance in both contexts in childhood compared to adolescence or adulthood (Aite et al., 2018; Carlson, 2005; Hongwanishkul, Happaney, Lee, & Zelazo, 2005; Prencipe et al., 2011). Effective probes to methodically measure these developmental differences in affective contexts are still under construction and evaluation (e.g., Tottenham, Hare, & Casey, 2011), yet most has not differentiated between IC and IR. When able to separate these constructs in affectively-laden contexts, adolescents show adult-like levels of IC accuracy, but slowed IR speed compared to children and adults (Aite et al., 2018; although see Schel & Crone, 2013; Somerville, Hare, & Casey, 2011).

The dual-systems model highlights a limbic affective system, associated with emotional processing, which develops prior to protracted maturation of cognitive control nodes, associated with control and regulation of the limbic system. This dichotomy and balance between “emotion” and “logic” has been a standard interpretive basis for most neurodevelopmental work (for reviews, see Casey, Jones, & Somerville, 2011; Pfeifer & Allen, 2012). Premotor regions and IFG show increasing activation with age in response to various IC tasks (Luna et al., 2010). A network of regions is activated during correct inhibitory responses (anterior cingulate cortex, and right dorsolateral prefrontal cortex or IFG) across development and appears fully mature at adolescence (e.g., Rubia, Smith, Taylor, & Brammer, 2007; Velanova, Wheeler, & Luna, 2008). Integrated function of these frontal regions through frontal-striatal-thalamic and frontal-parietal networks, correlates with better IC performance, suggesting these networks become more specialized and integrated throughout adolescence (Stevens, Kiehl, Pearlson, & Calhoun, 2007, 2009).

C. Challenges in Measurement Across Development (Multimethod Variance)

There are inherent methodological challenges to isolating the development of neural systems supporting cognitive control and emotion interference. In addition, the method by which IC/IR is measured must also be considered (as noted above with self-report accuracy, peers, emotion). Currently, many researchers and clinicians utilize standardized, performance-based tests along with executive functioning rating scales from observer- or self-report (Guy, Gioia, & Isquith, 2004). Standard dual administration of performance and report measures is intended to enhance ecological validity of measures, yet their concordance is strikingly low to modest (Boschloo, Krabbendam, Aben, de Groot, & Jolles, 2014; Silver, 2014). For example, performance-based IC predicted 4% of parent-reported IC/IR problems in children (Sorensen, Plessen, Adolfsdottir, & Lundervold, 2014). Performance-based tests may capture efficient cognitive function within pure laboratory environments that maximize performance and sacrifice ecological validity (Toplak, West, & Stanovich, 2013). Report-based questionnaires may be better at capturing ecological validity – the ability to apply these functions in everyday situations with complex, interactive and dynamic demands. Greater understanding of sources of variance and integration across modalities is needed to attain more representative models of IC/IR. For example, use of parent and teacher sources is critical to understanding IC/IR in multiple contexts, yet these individuals may not have enough training to understand what they are rating and how the scales work. Moreover, parents may have response biases in describing their children’s skills - demonstrated in other personality literatures around stigma (e.g., negative and positive impression biases) - and may be more accurate in reporting observed externalizing behaviors relative to internalizing behaviors. In addition, environmental context may influence children’s behavior such that behaviors observed in the home environment (and even by different parents) differs in important and meaningful ways from those observed in school or other settings (De Los Reyes, Thomas, Goodman, & Kundey, 2013). Finally, it is likely that older age and maturity may correspond with changing self-awareness of IC/IR difficulties.

D. Childhood Environmental Influences

While report and performance-based measures may both depend in unique ways on the observational context, a child’s environment may also have a direct influence on the development of these cognitive abilities. Indeed, IC, and potentially IR, appear to be impaired in maltreated and abused children as well as those affected by material deprivation (Finch & Obradović, 2017; Lawson, Hook, & Farah, 2017). Such impairment persists to adulthood (Irigaray et al., 2013), even when controlling for other significant psychiatric diagnoses (Malarbi, Abu-Rayya, Muscara, & Stargatt, 2017; Marshall et al., 2016; Quinn et al., 2018).

Stress is purported to be a primary pathway between adverse childhood experiences and lower IC/IR by “biologically embedding” early childhood experiences in altered neurodevelopment (Berens, Jensen, & Nelson, 2017). Early adversity and lack of enriched experiences are physical, social and emotional stressors that alter a multitude of biological mechanisms. Stress-mediators including cortisol and inflammatory cytokines, impede recovery of neurons via their oxidative-damaging properties (Danese & McEwen, 2012). Hypercortisolemic responses to stressors and disrupted circadian profiles are often present in those with early childhood adversity, exposing them to more intracerebral cortisol and neuronal damage over time (Blair & Raver, 2016; Ursache & Noble, 2016). In particular, fronto-limbic nodes are affected due to their dense concentration of glucocorticoid receptors (McEwen, Nasca, & Gray, 2016). The protracted development of fronto-parietal cortex throughout adolescence leaves these malleable areas more susceptible to negative exposures (Bick & Nelson, 2016). In addition, parents in such environments may be less capable of modeling effective IC/IR skills.

Several lines of research bear out that adversity and deprived environments correspond with disrupted activation in fronto-limbic regions during IC (Bruce et al., 2013; Carrion, Garrett, Menon, Weems, & Reiss, 2008; Mueller et al., 2010), and diffuse widespread decrements in whole-brain structural integrity, as well as specific decrements in key nodes, such as prefrontal cortex, anterior cingulate, hippocampus, amygdala, cerebellum and corpus callosum (Blair & Raver, 2016; Merz, Harle, Noble, & McCall, 2016; Paquola, Bennett, & Lagopoulos, 2016; Quinn et al., 2018). Some white matter pathway reductions are correlated with abuse (Hart & Rubia, 2012), neglect (Hanson et al., 2013), or institutionalized care (Merz et al., 2016), suggesting a link with length of exposure to stress-mediators and white matter integrity, consistent with adolescent-onset mood disorders (Bessette, Nave, Caprihan, & Stevens, 2014). Both adversity and deprivation are thought to play a role in IC/IR impairments, possibly through the biological effects of (chronic) stress and neurodevelopmental compromise.

E. Developmental Plasticity

Cumulatively, there is rapid improvement in both IC and IR skills across childhood and adolescent development. Negative life experiences can impact neurodevelopmental changes and associated IC/IR abilities. While some effects appear diffuse across brain circuitry, key nodes with high concentrations of glucocorticoid receptors may be most greatly impacted. Negative exposures during development increase the risk for adverse cognitive sequelae, including IC and IR (Saleh et al., 2017). Fortunately, brain function and structure remain malleable throughout this developmental period, thereby making cognitive prevention and intervention efforts imperative for these youth. Cognitive training and neuromodulation targeting IC/IR skills may decrease risk for poorer outcomes (Section V) such as recurrent mood disorders.

Section III. IC/IR in Mood Disorders

In order to better understand trait vulnerability factors and intermediate phenotypes that may help identify individuals at-risk for mood disorders, tools to measure distinct inhibitory control (IC) and interference resolution (IR) processes across the lifespan and stages of illness are highly valuable. Cognitive functioning including IC and IR can be affected by state (e.g., significant symptoms, treatment), scar (i.e., persistent decrements in functioning after an episode), and burden (i.e., repetitiveor cumulative effects of the disease over time) effects of the disease. Each obscures our ability to understand how different aspects of cognitive functioning may be trait vulnerabilities for diseases like MDD (MacQueen & Memedovich, 2017), in turn confounding our ability to identify and potentially prevent disease.

However, relatively few studies have examined IC/IR during high-risk developmental periods in individuals in remission (rMDD) or prior to illness. Even in these few studies, protracted brain development and peak plasticity occurring during adolescence and young adulthood may complicate our understanding of the effects of illness through concealment, murky shifts, or even variable age maturation, such as those found in other developmental disorders (e.g., attention-deficit/hyperactivity disorder). Therefore, identifying stable risk factors is particularly challenging and yet highly important to the field (Bessette, Burkhouse, & Langenecker, 2018; Burkhouse et al., 2018).

A. Measurement of Trait IC/IR as Vulnerability for Mood Disorders

i. Remission

Accumulating evidence suggests poor IC is a persistent, trait vulnerability factor related to MDD, whereas IR weaknesses may be more strongly related to state effects of disease (Langenecker, Caveney, et al., 2007; Peters et al., 2017; Rao et al., 2015; Ryan et al., 2012). After first onset of MDD, weaknesses in IC persist in remission (Ahern & Semkovska, 2017). Remitted young and older adults demonstrate stable deficits in IC compared to healthy young adults with no history of MDD (Aker, Bo, Harmer, Stiles, & Landro, 2016; Peters et al., 2017). Impairments in IC may predispose individuals to engage in maladaptive emotion regulation strategies, such as less use of reappraisal, more ruminative repetitive tendencies, and higher expressive suppression, all related to worse symptoms of MDD (Jacobs et al., 2016; Joormann & Gotlib, 2010; Langenecker et al., 2005; Langenecker, Jacobs, & Passarotti, 2014; Stange, Alloy, & Fresco, 2017).

Aberrant neural processes have also been implicated during IC/IR among remitted adolescents and young adults. Young adults with rMDD demonstrate deficient engagement of the CCN and weaker integration of the CCN during both “hot” and “cold” IC/IR tasks compared to healthy young adults, stable over time (Bessette, Jenkins, et al., 2018; Peters et al., 2017; Stange et al., 2018). Disrupted CCN connectivity during rest in rMDD is associated with poorer IC, as well as greater ruminative tendencies, other negative cognitive styles, and the burden of multiple prior episodes (Bessette, Jenkins, et al., 2018; Bhaumik et al., 2017; Jacobs et al., 2014; Peters, Burkhouse, Feldhaus, Langenecker, & Jacobs, 2016; Stange, Bessette, et al., 2017). Of note, affective stimuli and acute stress may further disrupt fronto-cingulate functioning implicated in IR among rMDD (Vanderhasselt et al., 2012; Whitton et al., 2017). Taken together, these results suggest that active illness (i.e., threshold symptoms of depression or mania) is not the sole cause of poor IC (and potentially IR) in mood disorders. Notably, IC performance and CCN integrity are stable in rMDD, suggesting a robust signature, or candidate intermediate phenotype, of MDD (Bessette, Jenkins, et al., 2018).

ii. Vulnerability or Risk

To determine whether IC and IR pose a trait risk for mood disorders, it is necessary to test whether these abilities are affected prior to onset of illness. An obvious challenge to such high-risk studies is measuring neuropsychological functioning prior to the onset of any clinically-significant mood symptoms. Indeed, Micco et al. (2009) found worse IC performance on the Wisconsin Card Sort in subthreshold depressed high-risk adolescents. Such high-risk individuals may already have prodromal symptoms and complicating environmental impact, making baseline assessment difficult.

The most effective study designs to tease apart vulnerability from active state effects utilize high-risk children by virtue of a parent with the illness. Modified Stroop and Flanker tasks demonstrate worse IR in high-risk youth, particularly those with earlier onset of depression (Belleau, Phillips, Birmaher, Axelson, & Ladouceur, 2013; Vijayakumar et al., 2016). This finding is consistent with a meta-analysis of children and adolescents with a current mood disorder showing strongest deficits in Stroop IR performance (Wagner, Abramson, & Alloy, 2015).

IC may play a role in vulnerability, particularly in young children and in interaction with environmental components. Children’s IC abilities at age three longitudinally predicted later depressive symptoms (Bufferd et al., 2014), and IC impairments at age six were associated with increasing risk for MDD, by virtue of longer chronic exposure to a mother’s depressive symptoms (Hughes, Roman, Hart, & Ensor, 2013). Worse IC performance in affectively-laden tasks is associated with higher depressive symptoms in adolescents (Davidovich et al., 2016).

Taken together, IC and IR may represent vulnerability for mood disorders in young children and adolescents, modifiable or mediated by environmental factors. Those who later develop MDD show worse IC/IR in childhood. In Figure 2, we illustrate this vulnerability of IC/IR.

Fig. 2.

Development of IC/IR abilities rapidly increases over childhood and adolescence, plateauing at adulthood, before slowly declining with older age. Individuals with mood disorders show less rapid development, indicating risk (yellow dotted line) that does not reach the same levels at adulthood, prior to more rapid declination with age. In addition, mood episodes worsen IC/IR (red dotted line), potentially after onset of symptoms (blue line). With additional episodes comes a repetitive scar or burden (grey line) that furthers rapid declination in older adulthood.

iii. IC/IR effects in Active State

The field is replete with studies of MDD in the active state, summarized in a recent thorough meta-analysis of executive function in MDD, including combined IC/IR (d=.38-.60)(Snyder, 2013). IC may have a stronger impact than IR in the active state. As the literature still does not disentangle active-state IC and IR effects, we only make brief commentary on difficulties examining active state compared to other states.

Typically, brief cross-sectional and pre/post treatment studies are employed to compare active versus remitted states. However, such studies pose multiple challenges to interpretation of cognitive dysfunction. Cross-sectional studies are very likely to be biased by the mood state of first measurement. Indeed, longitudinal depression symptoms in BD are heavily biased by the initial measure at intake (Cochran, McInnis, & Forger, 2016). Even for a disorder with significant fluctuations in depression and mania symptoms, future cognitive difficulties are strongly weighted based upon baseline symptoms (Ryan et al., 2017).

Additionally, the mechanisms by which individuals achieve remission (which are largely unknown) may or may not relate directly to the degree and type of cognitive dysfunction. Indeed, IC and IR skills are moderators of treatment response and disease course (Crane et al., 2017; Dawson et al., 2017; Langenecker et al., 2018; Peters et al., 2017). Second, the timeline for brain-based recovery of cognitive skills impacted in active mood disorders is currently unknown. Notably, current studies conflate overcompensation, recovery to healthy levels, treatment response, and the direct effect of treatment itself. Further complicating interpretation of such studies (Figure 2, panel inset), there is typically a longer delay to improvements in psychosocial, cognitive and occupational functioning relative to mood-based symptoms and hopefulness (Borson et al., 1992; Heald et al., 2004; Nebes et al., 2003; Peters et al., 2017; Teasdale, Lloyd, & Hutton, 1998; van der Voort et al., 2015). It would take even longer to confidently determine whether a particular individual’s cognitive function does not improve with treatment.

Carefully delineated studies are needed to dissect whether IC/IR are independent of state in MDD and the lingering state effects after transitioning from active to remitted MDD.

B. IC/IR and Scar Effects

Disrupted neuroplasticity and emergence of neuroprogression is implicated in chronic recurrent MDD, only briefly reviewed here (and more so in Section IV). Unfortunately, this literature rarely allows for distinctions of IC/IR from other executive functions. Broadly, the hypothalamic pituitary adrenal (HPA) axis, aided by and interacting with inflammation, may mediate neuroprogression. The accumulating cognitive morbidity associated with MDD has been referred to as scar burden (Weisenbach et al., 2014). We do not know whether neuroprogression in MDD might occur early in the illness; in a stop and start fashion like relapsing/remitting multiple sclerosis; with the emergence of comorbid conditions; or only after a particular dose, severity, or recurrence has transpired (Langenecker et al., 2014). Innovative longitudinal study models that separate IC/IR from other executive functions are needed to reveal the cross-pollination of scar burden with the emergence or worsening of cognitive morbidity.

C. Illustrative Example

Pooled data from 15 years of our studies suggest greater decremental effects of IR with greater age (linear and quadratic) and mood disorder diagnosis, explaining 20% of variance. IC decrements are seen in BD, and there are MDD-specific aging effects (See Table 2, Supplemental Materials for figures and more information).

Table 2.

Age-related changes in performance metrics from the Parametric Go/No-Go across healthy and mood disorder individuals.

	B	SE	t	95% CI	sr2
PCTT Overall Model	F (5, 1248) = 26.45, p < .001; Adj R2 = .09
Intercept	.92	.01	183.23*	.91, .93
MDD	−.03	.01	−4.01*	−.04, −.01	−.11
BD	−.03	.01	−4.34*	−.04, −.02	−.12
Age	−.02	.003	−5.85*	−.02, −.01	−.16
Age × MDD	.0001	.004	0.02	−.01, .01	.001
Age × BD	−.002	.01	−0.47	−.01, .01	−.01
PCIT Overall Model (IC)	F (5, 1247) = 5.14, p < .001; Adj R2 = .02
Intercept	.75	.01	81.29*	.73, .77
MDD	−.02	.01	−1.47	−.04, .01	−.04
BD	−.06	.01	−4.33*	−.08, −.03	−.12
Age	.004	.01	0.66	−.01, .01	.02
Age × MDD	−.01	.01	−1.73†	−.03, .002	−.05
Age × BD	.01	.01	1.04	−.01, .03	.03
RT Overall Model (IR)	F (8, 1244) = 40.26, p < .001; Adj R2 = .20
Intercept	460.84	4.27	107.87*	452.45, 469.22
MDD	24.58	5.04	4.88*	14.70, 34.46	.12
BD	17.40	5.70	3.05*	6.22, 28.58	.08
Age	11.62	2.00	5.81*	7.70, 15.55	.15
Age × MDD	0.94	2.69	0.35	−4.35, 6.22	.01
Age × BD	4.92	3.06	1.61	−1.09, 10.92	.04
Age2	2.14	0.98	2.18*	0.22, 4.07	.06
Age2 × MDD	−0.04	1.41	−0.04	−2.81, 2.73	−.001
Age2 × BD	1.94	2.01	0.97	−2.00, 5.88	.02

Notes. The intercept represents healthy individuals at the average age of the whole sample, M=36.90, SD=14.22. Abbreviations: BD = Bipolar Disorder, MDD = Major Depressive Disorder, PCTT = percent correct target trials, PCIT = percent correct inhibitory trials, RT = reaction time to Go trials.

^†p < .10, two-tailed.

^*p < .05, two-tailed.

Section IV. Late-Life Trajectory of IC/IR

The broad construct of inhibition declines with age (Ariza et al., 2015; Hedden & Gabrieli, 2004; Nielson, Langenecker, & Garavan, 2002; Yehuda, Halligan, Grossman, Golier, & Wong, 2002; Yehuda, Teicher, Trestman, Levengood, & Siever, 1996). Specific difficulty or inability to suppress irrelevant information (IR) may underlie the changes that tend to occur in normal aging (Hasher, Zacks, May, Gopher, & Koirat, 1999; Kane, Hasher, Stoltzfus, Zacks, & Connelly, 1994). There is ample evidence that these declines are mediated by prototypical decline in processing speed during aging (Craik, Jennings, Craik, & Salthouse, 1992; Salthouse, 1996; Salthouse & Babcock, 1991). White matter integrity, a strong correlate of processing speed (Gunning-Dixon & Raz, 2000), is often compromised with aging, particularly in the presence of chronic medical or psychiatric illness (Bender, Völkle, & Raz, 2016; Maillard, Carmichael, Reed, Mungas, & DeCarli, 2015; Mettenburg, Benzinger, Shimony, Snyder, & Sheline, 2012; Santiago & Potashkin, 2015; Zhang et al., 2018). Therefore, age-related alterations in brain structure that are often present in aging may lead to less efficient inhibitory control (IC) and IR processes. For example, functional neuroimaging work indicates that recruitment of brain networks during IC is more diffuse in older versus younger adults (Clapp, Rubens, Sabharwal, & Gazzaley, 2011; Langenecker & Nielson, 2003; Langenecker et al., 2004; Milham et al., 2002; Zhu, Zacks, & Slade, 2010). IC declines with age relative to IR – reorganization and dedifferentiation are key theories behind how declines in speed (IR) are greater, despite maintenance of some degree of accuracy (IC; Langenecker & Nielson, 2003; Park et al., 2012). Of note, relatively few aging studies have attempted to separate IC and IR processes. As such, this reviewed literature primarily addresses inhibition more generally.

A. Age Interactions with Mood Disorders in Late Life

Deficits in executive functioning may distinguish a subtype of late life depression (LLD; Alexopoulos, 2001), and IC and IR are particularly relevant. Worse IR or IC, respectively, are highly prognostic for poorer drug treatment response, greater improvement in cognitive remediation, and risk for dementia in older adults with depressive symptoms (Morimoto et al., 2011; Morimoto et al., 2016; Potter, Kittinger, Wagner, Steffens, & Krishnan, 2004; Sneed et al., 2010). Individuals with LLD exhibit aberrant patterns of activation during IC/IR tasks and rest (Aizenstein et al., 2006; Bobb et al., 2012; Dumas & Newhouse, 2015; Rao et al., 2015) in the dorsal anterior cingulate and dorsolateral prefrontal cortex (Aizenstein et al., 2009; Alexopoulos et al., 2012; Alexopoulos et al., 2013).

What remains unclear is whether poorer IC/IR in MDD continues into late life and becomes more severe with age, or whether later life cognitive dysfunction portends new onset disruption and risk for LLD (Weisenbach et al., 2014). A handful of studies examined interactions of age with depression status, finding that older adults with MDD are especially likely to experience IR decrements (Lockwood, Alexopoulos, & van Gorp, 2002) and aberrations to underlying neural circuitry (Rao et al., 2015), suggesting a double burden of cognitive aging and depression.

B. Potential Mechanisms of IC/IR Decline in Late Life

The basis of decreased IC/IR in LLD is likely multifactorial, and thus it is valuable to draw connections between neuropsychological presentations, changes in brain structure and functioning, other relevant biomarkers (i.e., markers of immune, endocrine, and vascular systems functioning) and comorbidities to suggest specific treatments and prevent cognitive decline and disability. We will briefly cover these cellular, hormonal, and neural processes that might underlie neuroprogression. Some of these processes are conceptualized as purely neurobiological processes underlying maintenance of neural networks, whereas others are more peripheral factors such as cardiovascular and pulmonary supply. Of note, there is strong evidence that the onset of MDD, either early or later in life, infers distinct etiology that, in turn, is relevant to prognosis and cognitive decline (Disabato et al., 2014).

Various hypotheses are proposed to explain the increased risk of cognitive decline and dementia in LLD and which are not mutually exclusive. As one example, Alexopoulos et al. (2005) proposed a model in which factors contributing to LLD could be separated into (1) mediating mechanisms (e.g., hypometabolism of dorsal cortical regions and hypermetabolism of ventral limbic regions), (2) predisposing brain abnormalities (e.g., abnormalities in frontostriatal and limbic circuitry, heredity and psychological vulnerability), and (3) etiological contributors (e.g., age-related brain changes, disease-related changes, and allostatic response to adversity). For example, vascular changes may be an etiological contributor, occurring as a result of disease-related changes and high-frequency comorbidities, such as hypertension or cardiovascular disease. These vascular changes may then predispose to disruption of metabolism in fronto-striatal regions (Aizenstein et al., 2005; Aizenstein et al., 2009). Under the umbrella of this model, one or more of these aforementioned factors creates a context in which the brain is more vulnerable to the impact of stressors.

i. HPA-Axis and Adversity

The HPA-axis plays a key role in the onset and maintenance of depression and may be a primary mediator of neuroprogression. Detailed models of neuroprogression related to HPA-axis dysregulation are based upon animal models, with a notable degree of concurrence in human experiments (Bruehl et al., 2009; Elgh et al., 2006; Greendale, Kritz-Silverstein, Seeman, & Barrett-Connor, 2000; Lupien et al., 1998; Lupien et al., 2005; McEwen, 2006; Starkman, Gebarski, Berent, & Schteingart, 1992). With age, the HPA-axis may become hyper- or hypoactive. Meta-analytic findings suggest that older adults experience a flattened diurnal slope (Lupien et al., 1996). The Glucocorticoid Cascade Hypothesis posits individuals with lifelong depression may experience increased risk for dementia due to frequent overexposure of the hippocampus to glucocorticoids, leading to the halting of dendritic neurogenesis (see Odaka, Adachi, & Numakawa, 2017 for review), potentially resulting in smaller hippocampi in older adults with early-onset recurrent depression (Sexton et al., 2012). The effects of HPA-axis disruptions are mediated by the emergence and experience of childhood adversity and associated with altered cognition and structural brain connectivity beyond the effects of depression, even in older adulthood (See Section II; Jaworska et al., 2014; Karstens, Ajilore, et al., 2017; Karstens, Rubin, et al., 2017; Petkus, Wetherell, Stein, Liu, & Barrett-Connor, 2012; Saleh et al., 2012; Stein, Kennedy, & Twamley, 2002). For some, alterations in the HPA-axis due to stress may lead to susceptibility for psychopathology, including changes in IC/IR, with persistent stress leading to further degradation in systems supporting memory and frontal control.

ii. Inflammation and Neuroprogression

The Inflammation Hypothesis of LLD (Alexopoulos & Morimoto, 2011) posits that an abnormal immune response in late life leads to chronic low-grade inflammation and subsequent neurotoxic and depressogenic effects. Pro-inflammatory states reduce the production of neurogenerative growth factors, stifle the clearance of neurotoxins, inhibit serotonin production and alter HPA-axis functioning (Black, 2002; Kohman & Rhodes, 2013; Myint, 2013). A meta-analysis examining longitudinal studies reported greater effect sizes for the relationship between C-reactive protein and future depressive episodes in individuals over the age of 50 (Valkanova, Ebmeier, & Allan, 2013), corroborated by longitudinal studies identifying associations between peripheral inflammatory markers and the onset and course of LLD (Baune et al., 2012; Gallagher, Kiss, Lanctot, & Herrmann, 2017). Yet, the directionality of the inflammation-LLD relationship may be complex (Kim et al., 2018). There is increasing cross-sectional evidence that inflammation is associated with poorer executive functioning composites including IC/IR functioning in older adults (Baune et al., 2008; Heringa et al., 2014), and that sex moderates the relationship between neutrophil gelatinase-associated lipocalin and functioning in IC/IR (Naude et al., 2014). Longer-term inflammation exposure via cancer trials are also associated with emergence of depression and cognitive related dysfunction (Felger et al., 2015; Goldsmith et al., 2016; Miller, Maletic, & Raison, 2009; Miller & Cole, 2012; Raison, Capuron, & Miller, 2006).

B. Mild Cognitive Impairment (MCI) and Depressive Symptoms

LLD is associated with the co-occurrence of mild cognitive impairment (MCI) or dementia syndromes (Steffens, 2016). MCI may subside following proper treatment, particularly with less severe impairment. However, treatment-resistant MDD is a high risk for progression into dementia (Leonard, 2017). Thus, it has been argued that geriatric MDD, particularly with an initial onset in late life, reflects a prodromal stage of dementia (Butters et al., 2008). According to the Vascular Depression Hypothesis, white matter lesions or reduced integrity (e.g., fractional anisotropy) in fronto-subcortical circuitry are associated with executive dysfunction in older adults including IC/IR (Dalby et al., 2012; Murphy et al., 2007) and depressive symptomatology (Sheline et al., 2008; Sheline, Price, Yan, & Mintun, 2010).

Various chronic illnesses that often present in mid-to-late life provide further evidence of the influence of vascular, metabolic, immune, and endocrine systems on IC/IR and increased risk and expression of depressive symptomatology (but are beyond detailed review here). Vascular risk factors and metrics of vascular burden including heart disease, hypertension, hypercholesterolemia, diabetes, obesity, atherosclerosis, and metabolic syndrome, increase risk for MDD (Mast et al., 2008).

C. Aging Conclusions

In conclusion, the neuroprogression of IC/IR difficulties and depression in late life, particularly with later onset, may result from a confluence of factors that influence fronto-striatal and fronto-temporal neurocircuitry in older adults. Continuing to identify relevant pathological features, cognitive and biological markers may help guide treatment planning in LLD. With a greater understanding of mechanisms involved in neuroprogression of IC/IR difficulties, preventative treatments could be developed. Given the etiological heterogeneity of this disorder, nontraditional treatments or augmentation that promotes neuroplasticity may be beneficial for those resistant to standard psychotropic medications.

Section V. Addressing and Accommodating IC/IR to Avoid or Reduce Disability

Our ethical obligation to individuals who experience mood disorders is to identify whether these weaknesses may be amenable to intervention and accommodation such that quality of life might improve for those with IC/IR difficulties and mood disorders. Little is currently known regarding neural or biological bases for increased disability burden due to IC/IR decrements, although the disorders are marked with high comorbidity of diabetes, obesity, and cardiovascular disease (Arterburn et al., 2012). Additive stress on neural systems from chronic MDD and related inflammatory and neuroendocrine disruption may lead to instantiation of permanent and compounded memory, affective dysfunction, and IC/IR decrements.

Notably, studies of late life and midlife do little to inform treatments that could stop or reverse cognitive disability processes (Alexopoulos, 2002; Alexopoulos et al., 2005). To our view, scar burden from mood disorders either results in neurological age acceleration or earlier instantiation of typical aging processes. The byproduct of such processes is functional disability, underemployment, and related morbidity issues. In our model, Figure 2 exemplifies what can be summarized as precocious aging in mood disorders. Alternatively, scar burden may be independent to typical aging processes (Weisenbach et al., 2014; Wright & Langenecker, 2008; Wright et al., 2009).

A. Accommodations in Educational and Vocational Contexts

Accommodation and interventions in mood disorders, especially for cognitive deficits, have traditionally been short-term, and long-term benefits have not been carefully tested to date. What is particularly worrisome for providers is that IC/IR skills may need to be constantly used and challenged to maintain flexibility and efficiency. Long periods of time where these systems are weakened or underutilized could potentially result in more rapid age progression.

Adolescents and young adults are typically prescribed psychotropic medication or talk therapy services for mood symptoms, but cognitive dysfunction is not directly targeted. Schools and workplaces could incorporate early prevention and identification programs, as well as special education services or accommodations under a Section 504 plan of the Rehabilitation Act. Modifications, as recommended by neuropsychologists, could include schedule changes, extra test-taking time, and substitutions or alternative requirements for assignments or deadlines, and can be stratified based on school, college, or workplace setting. For example, access to financial aid at ¾ course load in college could be helpful in providing those with weaker IC/IR skills a greater chance for higher levels of achievement. Moreover, a workplace’s openness around rules regarding limits to hours worked (e.g., allowing them to work extra hours, mid-day breaks) could allow an employee the time needed to accomplish their workload (and facilitating stronger self-efficacy). Also, those with poor IC/IR and mood disorders may be less proficient at managing hostile work environments. Our experience is that a supportive management style and work environment can go a long way toward providing accommodations that enable individuals with IC/IR difficulties (and potentially mood disorders) to contribute at a meaningful level. Avoiding functional disability can have positive benefits on self-esteem and long-term functioning for individuals with IC/IR difficulties or mood disorders.

B. Remediation of Dysfunction in IC/IR

Importantly, it is possible that IC/IR is amenable to interventions. Because this literature has become extensive and is updated continually, we would like to briefly comment on potential avenues for IC/IR plasticity.

i. Remediation through Cognitive Training

The field has become more open to cognitive remediation techniques to target various deficiencies in mood disorders, with the idea that improvements in more global cognitive functions including IC/IR, via neuroplasticity, will also improve disorder-relevant symptoms. Training on tasks similar to the SST in healthy controls has shown within-domain improvement in conjunction with greater activation in right IFG and shifting of top-down PFC activation during cues rather than during implementation of stopping behavior (Berkman, Kahn, & Merchant, 2014). Cognitive training in mood disorders has shown moderate to large effects on attention, working memory and global functioning, but only small to moderate effects for symptom severity and daily functioning (Motter et al., 2016; Vanderhasselt et al., 2015). Moreover, cognitive training appears to reduce vulnerabilities associated with MDD relapse, such as stress reactivity and brooding in high trait ruminators (Koster, Hoorelbeke, Onraedt, Owens, & Derakshan, 2017) or treatment-resistant MDD (Morimoto, Wexler, & Alexopoulos, 2012).

ii. Remediation through Neuromodulation

Some studies have used repetitive transcranial magnetic stimulation (rTMS) or transcranial direct current stimulation (tDCS) alone or during cognitive training to amplify effects. Excitatory rTMS to prefrontal regions, typically left dorsolateral prefrontal cortex, is thought to induce increased cognitive control activity, not just in the direct area of stimulation but across wide-spread anatomically connected regions. This neuromodulation technique appears to conjointly improve depressive symptoms with cognitive function, including trend-level associations with speeded IR performance such as in the Stroop (Ilieva et al., 2018; Watanabe et al., 2015). In the latter sample, this positive effect was predicted by rTMS-induced modulation within right pre-SMA, striatum and globus pallidus interna. The effects of neuromodulation on cognitive remediation appear promising, and an important area for continued investigation and intervention.

Related, tDCS is a technique that applies a weak electrical current directly on the scalp and theoretically through brain tissue by creation of a circuit with a battery and two electrodes, thus hyperpolarizing neurons in a particular direction. Most tDCS studies have placed these electrodes in various regions of prefrontal cortex, particularly to induce neuronal depolarization leading to increased activity from the left dorsolateral prefrontal cortex. A steady treatment of active tDCS appears to have a moderate effect on current depressive symptoms (Shiozawa et al., 2014), and single-session of active tDCS can improve affectively-laden IC/IR in actively depressed individuals (Wolkenstein & Plewnia, 2013). It remains to be seen whether tDCS can enhance cognitive remediation for IC/IR specifically and in turn reduce mood symptoms in clinical or high-risk samples.

Section VI. Overall Conclusions

Speed (IR) and accuracy (IC) in behavior selection/execution remain important cognitive skills with implications for educational, vocational, and social/interpersonal contexts. Those who can best modulate IR to enhance IC can fine-tune their behavior to current contextual circumstances and are less reliant upon fixed behavioral repertoires. There are important implications for IC/IR in the context of mood disorders since the neural circuitry implicated in emotion regulation spatially overlaps with IC, and the region targeted for rTMS treatment of depression is an important node for IR. Measurement of the dynamic interplay between IC and IR, particularly across the lifespan of mood disorders may help the field to disentangle processes of vulnerability, disease, and scar burden, with potential implications outside the field of mood disorders. Dimensional transdiagnostic modeling of these cognitive processes may also help to tease apart issues apparent in other fields of neuropsychiatry. Therefore, we call for more longitudinal studies of developmental and lifespan approaches to the study of both IC and IR. Within these and related studies, we advocate for more attention to measurement and reporting of task and metric variations, as well as greater attention to within-person variability in IC. Modulating IR in more contexts can give a greater appreciation for individual strengths and weaknesses, with implications for interventions and accommodations. Finally, our conceptualization, and that of many within the field, is that the neural systems supporting IC and IR are sensitive to acute and chronic perturbations – early intervention and preventative treatments, including those in mood disorders, could reduce morbidity and mortality.

Abbreviations

BA

Brodmann’s Area

BD

Bipolar Disorder

CCN

cognitive control network

GNG

Go/No-Go task

HPA axis

hypothalamic pituitary adrenal axis

IC

inhibitory control

IFG

inferior frontal gyrus

IR

interference resolution

MDD

Major Depressive Disorder

PFC

prefrontal cortex

SMA

supplemental motor area

SST

Stop-Signal Task