The Evolving ADHD Phenotype in the Externalizing Context

Abstract

ADHD can be viewed as a precursor or liability for conduct problems and other externalizing conditions. Despite its familiarity, develoments in the past decade suggest that refining the phenotype will yield dividends for clinical prediction and treatment specification as well as mechanistic discovery and theory. ADHD is a useful category yet comprises multiple constitutive traits requiring further specification. The field is nowpoised to shift to a useful integration of novel dimensional traits into the ADHD phenotype. This shift will need to recognize the importance of both the established ADHD symptom dimensions and other dimensions, to create new presentations, types, or specifiers in the nosology. Clinical issues include the role of sluggish, disengaged cognitive style of inattention with hypo-actiivity. Dispositionasl traits related to emotional dysregulation will be key as well; these include negative affect and irritability (anger dysregulation). In In the cognitive sphere, alolng with ongoing work on executive functions and reward discounting, new insights related to cognitive information gain or neural gain (a mechanism by which relevant neural connections strengthen during rapid information uptake and decision making) may explain other cognitive findings and relate to dispositional traits and clinical features. Due to its early gateway role, clinical prediction algorithms to evaluate course of ADHD and advisability of intervention will be valuable. A neurodevelopmental understanding, with very early developmental origins and seeing ADHD as an early precursor of more serious complications later, opens opportunities for early preventive support and early detection drawing on non-clinical trait measures as risk indicators. An emerging focus on early origins from conception forward is informing understanding of etiology and will continue to do so in valuable ways. Clinical prediction algorithms should benefit from inclusion of low cost, readily deployable trait measures discussed here including selected temperament and cognition measures, which appear to enhance predictive power.

Severe problems with inattention/disorganization, impulsivity, and hyperactivity constitute the ADHD syndrome as currently understood. Yet this population is highly heterogeneious, and progress over the past decade commends further phenotype refinement to maximize clinical prediction, treatment matching, and mechanistic understanding. I illustrate this in work on dispositional trait ratings (i.e., temperament and personality), and attention/cognition in the laboratory. These domains provide unique opportunities for novel cross cutting phenotype dimensions using low cost, field-deployable measures.

The importance of progress on ADHD is difficult to overstate. Because when mild ADHD is difficult to distinguish from typical child behaviors and stimulant treatment often contains symptoms, its seriousness can be under-appreciated. Yet, when extensive enough to markedly interfere with developmental tasks (learning, peer relations) as required in DSM-5, ADHD is remarkably important due to early onset, role as a potential causal precursor to other mental health conditions (1–3), including elevated risk for future depression, anxiety disorders, psychosis (4–6) as well as cascading poor outcomes including elevated risk for homelessness, unemployment, and poor health (7), culminating in reduced average lifespan or premature death due to accident, suicide, or degraded physical health (8). In all, despite benign outcomes in a subset of cases, ADHD is a major public health concern (1, 6, 9). Poor outcomes typically reflect a developmental progression from ADHD to oppositional defiant disorder (ODD), rule breaking, aggression (conduct disorder), mood disorder, and addiction, with continued impulsivity and poor executive function (EF) (10). Thus, ADHD in childhood is a critical early risk phenotype.

Over the past two decades, ADHD is increasingly seen as a disorder of self-regulation rather than inattention per se, but that idea requires careful elucidation (11) as phenotype refinement is considered. In maintaining a developmental perspective, I focus on childhood with only passing reference to adulthood, where issues and considerations shift considerably. After summarizing ancillary issues and clinical phenotype concerns, I introducing a dynamic developmental lens, then I examine (a) key constitutive traits and (b) evolving ideas about neurocognition, before offering conclusions.

Issues in the field affected by phenotype selection and bypassed here.

Cross cutting trait dimensions are necessary to address heterogeneity and that in turn is necessary in nearly every area of ADHD research. Some of the most active areas include (a) The genetics (12, 13) and neuroscience (14–16) of ADHD, as well as the exploding field of early epigenetic changes and ADHD (17–20); (a) Concerns about stigma, the vocabulary of neurotypicality versus the vocabulary of disorder, and whether changes in nomenclature are in order; (c) Gender and sex differences, including hormone effects and differential sex and gender responses to environmental insults (21–24), including distinct risk profiles and clinical needs (25); (d) Race and ethnic differences in ADHD treatment availability, utilization, and biases in diagnosis and intervention even as the demographic picture evolves (26–29); (d) Interest in alternative treatments and new information on nutrition, lifestyle, and health linkages to psychological problems and ADHD (30, 31); (e) New work on ADHD course, including frequency of offset, onset, or recurrence as well as late onset (32, 33). (f) In the externalizing sphere, ongoing work on the dimension of callous-unemotional trait(s) and the degree of mechanistic relation to adult psychopathy (34, 35) also has relevance to ADHD profiles (36, 37) though space prevents elaboration herein. I have not even nodded to various developments in pharmacogenetics, psychosocial treatments, work on ADHD and autism spectrum disorder, and more. All of these testify to the dynamism in current ADHD research and all require clarity about the phenotype and an approach to heterogeneity in the population.

Focus herein.

I therefore focus here on the over-arching and cross-cutting issue of the integration of related dimensions, traits, and disorder categories, which informs the meta-issue of within-syndrome heterogeneity and across-syndrome relations. The issue of dimensions versus categories has been a fundamental concern of nosologists for a century including the RDoC and HITOP efforts more recently as well as historical and current empirically derived problem dimensions in child psychiatry (38–43). Categories are inescapable in a clinical decision context, yet ADHD symptoms are more reliable and valid when studied as dimensions than as a category (44). Subtyping by novel trait profiles is a promising hybrid approach for integrating category and dimension as I will describe.

The Clinical Level

ADHD and Externalizing.

ADHD is highly correlated with externalizing disorders (conduct (CD), oppositional defiant (ODD), and substance use but including key behavioral domains of aggression, rule breaking, and callous-unemotional traits). However, ADHD is best seen as developmental precursor rather than a parallel outcome or manifestation of externalizing (10, 45). Therefore, ADHD was moved in DSM-5 to a neurodevelopmental category due to early onset prior to externalizing disorders; neuroimaging findings showing alterations in brain development from early life forward, and overlap with neurodevelopmental disorders such as autism and learning disorders (46, 47); the past decade has seen massive literature exploring early neurodevelopmental roots and correlates consistent with that decision. Yet some externalizing features, like irritability and high aggression, are also apparent early in life, with neurobiological correlates (48, 49), and can be defended as neurodevelopmental (50), adding some complexity to this evolving picture.

Etiological distinctiveness.

Reflecting the developmental picture, ADHD has a partially distinct etiological structure from externalizing behaviors like conduct problems. For example, a substantial literature documents a social learning component for aggression and conduct problems, though not as much for callous-unemotional traits but see (35). Antisocial behavior tends to have only modest heritability of around .5 (51), with important influences of shared environment (although aggression and rule breaking appear to have distinct etiological influences (52)). ADHD in contrast has heritability in the .7–8 range and modest non-shared environment (as well as GxE interaction) as the suggested causal structure (12, 53, 54). Parenting behaviors are associated with ADHD even apart from co-occuring externalizing problems and participate in the dampening or amplification of ADHD symptoms (55, 56). This is partly due to parent ADHD as well as partly due to youth ADHD symptoms evoking parenting responses (genotype-environment correlations (57). However, the idea that parent behaviors are a primary cause ADHD has been refuted both by behavioral genetic studies showing high heritability and minimal shared environment effects (53), and clever experimental tests (58).

Yet, even with high heritability in view, early environmental effects are important in ADHD. They may mediate genetic susceptibility, interact with genotype (such interactions can be hidden in the heritability term), or account for non-shared environment variance. None are specific or unique to ADHD but among the candidates deserving further causal and mechanistic investigation are prenatal maternal stress and distress (59), pre-term birth and/or low birth weight (60–62), social disadvantage and adversity and its mediators (63), typical (relatively low) level lead exposure (64–66), and other early yet widespread environmental toxicants (67, 68). Although many such findings rely on observational designs vulnerable to unmeasured confounders including genetic transmission (genotype-environment correlation)(69), causal evidence can be sought in sibling comparisons, Mendelian randomization, and other designs. Thus, a causal role is called into question for maternal smoking (70) and pre-pregnancy maternal BMI (71), but supported for average level lead exposure (65), social disadvantage (72), and low birth weight (70). in ADHD. Studies also support an additive role of early insults and genetics (73) while confirming confounding by ancestry for many exposures (74).

Overall, the developmental picture remains that early emerging hyperactivity-impulsivity (and possibly inattention), perhaps in conjunction with irritability, form an early liability phenotype for the emergence of aggression/conduct problems in the presence of particular socialization context, and that ADHD is part of a causal chain for other later emerging psychological problems and conditions, despite bidirectional effects (75). This heightens ADHD’s importance for early primary and secondary prevention of mental health conditions.

ADHD presentations/subtypes.

Here, the key node for clinical phenotype refinement concerns the ADHD-inattentive presentation (ADHD-IN). The ADHD subtypes of DSM-IV were downgraded in DSM-5 to presentations (44) due to the developmental instability of the then-subtypes. Indeed, the inattentive and combained (ADHD-C) presentations of ADHD are, in many respects, essentially severity categories. (The hyperactive presentation, ADHD-H, is rare after preschool but also typically milder than ADHD-C on external correlates). Differential validity with relation to academic and externalizing problems holds better for the dimensions than the presentations, because presentations can change. However, ADHD-IN is critically heterogeneous.

In DSM-III, ADD-without hyperactivity (ADD-W/O) was described with low activity, sluggish tempo, daydreaming. The limited research on that group in the 1980’s suggested important differences from ADD+H (76). However, In DSM-IV, these symptoms were dropped (77); the new ADHD-IN type presumably encompassed the hypoactive group. Yet, ADHD-IN captures both children who are over-activite yet insufficient to meet ADHD-C, and those who are hypo-active. Although considered for DSM-5 under the rubtic of a so-called restrictive inattentive group, DSM-5 did not address this problem due to lack of data. In the past decade, a large body of work suggests the hyporactive, sluggish, daydreaming features indicate a different profile or group from the remainder of children with ADHD (78). This entire body outlines the potential validity of a dimension or a category of youth with sluggish cognitive tempo, now called cognitive disengagement syndrome (CDS), characterized by excessive daydreaming, mental confusion, and hypoactivity (Becker (79) provides a thorough review).

This group is distinct from ADHD-C and ADHD-IN with regard to associations with depression, anxiety, and particular kinds of sleep problems, and possibly distinct cognitive profiles (80). Yet comparative studies have generally not controlled well for co-occuring hyperactivity level or looked at a restrictive ADHD-IN group. When this has been done, differentiation from the restrictive ADHD-inattentive subgroup of ADHD-IN is unclear (81–83). Overall, this suggests the likelihood of a clinically important yet distinct profile hidden within the ADHD population as currently defined in DSM-5, which may be ultimately the same or related to the CDS syndrome, likely needs to either be recognize either as a separate disorder, a subtype, or a specifier. Clarifying the appropriate nosology of CDS and a potential ADHD restrictive inattentive profile, is needed to reduce clinical heterogeneity (79).

Development

ADHD as gateway phenotype.

As already alluded to, ADHD and externalizing problems are (like most psychological problems and psychopathology) not static entities but are developmentally dynamic, contextualized features of mental and behavioral life (for a theoretical elaboration see (42)). As noted, ADHD participates as an early feature in a developmental progression that sometimes continues into ODD, CD, and later depression, substance use, and other probolems. ADHD is a bellweather: a phenotypic precursor to externalizing –and often, internalizing--problems, depending on moderating developmental contexts.

Early Origins.

What of the origins of ADHD itself? The past decade has seen growing interest in the developmental origins of psychological problems including ADHD, suggesting the possibility of finding biomarkers or early risk factors and mechanisms during pregnancy. For example, a body of work relates maternal emotional distress during pregnancy to offspring psychological and behavioral problems including ADHD (84–86). Other work, including in our lab with Elinor Sullivan, suggests that maternal systemic inflammation is one mediator of this association (87–89) and that maternal exposures, such as poor diet or emotional duress or trauma appear to alter offspring temperament (90) (Non-human epigenetic studies also note paternal exposures can affect offspring later, but such research is nascent in humans). Infant negative affectivity, including irritability, is related to maternal systemic inflammation (88),and also in turn predicts elevated ADHD-related symptoms and problems prospectively by age 5–6 years in work from the same group (91). All are now being followed up in larger cohorts via the HBCD study.

The findings exemplify that the field of early origins of ADHD is rapidly expanding with the establishment of a multi-site, international consortium looking at biological, physiological, and genetic markers, neural development, as well as behavioral and emotional factors, during pregnancy and at delivery (92). Overall, we may speculate that ADHD’s earliest roots during pregnancy include a range of biological programming responses to emotional or physical stressors on the mother or fetus, on top of genotype. These in turn are mediated via placental gene expression and other mechanisms (87, 93), likely including epigenetic effects (18, 19, 94). This in turn sets the stage for early temperament precursors such as negative emotionality, irritability, and high activity. These in turn initiate a developmental cascade toward ADHD and secondary complications later. If true, such a cascade would open multiple opportunities for early supportive preventive intervention. We (45) provide an updated developmental account, which relies on and modestly elaborates work by others, particularly Rothbart e.g. (95).

Precursive Trait dimensions.

Here, just as ADHD is a precursor to externalizing (which in turn depends on a range of moderating environments), temperament is a precursor to the emergence of ADHD depending on moderating and mediating traits and socialization. (When temperament is defined in relation to behavioral traits or typically expressed emotions and incentive responses and their regulation, the fields of temperament and personality largely agree on the value of at least two major and partially orthogonal yet mutually inhibiting domains denoting positive and negative valence [sometimes generalized as extraversion and neuroticism, or as approach and avoidance incentive responding; see text], which then can be analyzed behaviorally into narrower component domains of behavioral and affective tendencies.) As to higher order traits, both negative and positive emotionally play a role empirically and conceptually. For example, anger reactivity in the infant and toddler periods predicts childhood ADHD symptoms longitudinally (96–99), possibly mediated by disrupted inhibitory control (96). (While anger reactivity is an approach emotion [frustrative approach] rather than avoidance, in factor analytic studies of temperament scales based on positive and negative emotions, it factors with negative emotions such as fear and sadness; thus I here link it to negative emotionality at the observational level.) Positive affect/high approach in infancy predicts child ADHD symptoms as well (98, 100–103). Also consistent with the model are findings that after toddlerhood, at 36 months of age, lower effortful control best predicts later ADHD (104, 105), though more studies are needed to confirm the mediational hypothesis. Figure 1 depicts one hypothesized developmental cascade.

Figure 1.

Hypothesized schematic of one developmental pathway to ADHD^a

^aPrenatal stressors on pregnant mother or preconception biology of mother or father can be emotional distress or physical insult/illness; biological pathways during gestation are more complex than schematized here; infant affect can be overlapping but could distinguish different pathways to temperament profiles later. Mediation by low effortful control needs more verification. Further discussion in Reference #(45).

Stages of development for ADHD are depicted in a series of blocks on the diagram. Arrows show developmental sequences.

Summary.

While moderation by caregiving is crucial to a final developmental model (reviewed in (45)), this picture fits conceptually with an idea of distinct developmental pathways leading to distinct developmental profiles within an ADHD population defined by variations in emotional regulation, with positive versus negative emotionality playing key roles in defining the profile. I consider these possible profiles within an ADHD-defined group next.

ADHD and dispositional traits: profiles and secondary outcomes

Consistent with the early developmental literature, over the past two decades an expanding literature has established that ADHD (like externalizing generally) is related to a mix of personality and temperament traits including low effortful control/conscientiousness or high impulsivity; high positive emotionality (or extraversion) in most studies, high negative emotionality (or neuroticism), and in some studies, low agreeableness (hostility); differential trait associations are observed for hyperactivity/impulsivity and inattention (106–113). To move forward with these findings and capitalize on them, careful consideration of heterogeneity of profiles (that is, trait combinations) within individuals meeting an ADHD criterion is needed.

Although ADHD is unsurprisingly and somewhat obviously related to traits like impulsivity, under-control, and low self-control, with which it shares overlapping conceptual and behavioral space, it is negative emotionality that most saliently moderates variation in ADHD outcome. The component trait of anger dysregulation or irritability is crucial here. Irritability often factors with negative emotions although neurobiologically it is conceived of as an approach not an avoidance response(114). Irritability cuts across multiple DSM-5 conditions including DMDD, CD, ODD and ADHD and others. It is a precursor in early life to multiple child behavioral problems future problems and psychopathology generally (49), including ADHD, depression, anxiety, and externalizing problems. Although DMDD is the prototypical disorder of irritability, many children with ADHD have marked problems with irritability yet fall short of a DMDD diagnosis.

A series of papers using the Oregon-ADHD-1000 cohort of well-characterized community-recruited children with and without ADHD at intake (115) has clarified the importance of this domain for ADHD. Cross sectionally, irritability is strongly related to ADHD even in the absence of conduct disorder, even after adjusting for ODD symptoms, and when severity is insufficient to meet criteria for DMDD (116). A subgroup of children with ADHD with very high irritability+negative affect is reliably able to be identified through computational person-based strategies (117). Figure 2 shows a typical finding with explanatory notes. In Figure 2, the irritable group had worse clinical outcomes when followed up one to two years later than other children with equally severe ADHD. These include much higher rates of new onset of anxiety disorder, depression, and worsening ADHD and externalizing problems in the future—and this prediction was stronger than one relying only on the usual clinical indicators of ADHD severity or baseline comorbidity (117). Similar results were replicated in other groups—an emotionally dysregulated group is universally identified, with a high NA/irritable group most replicable and a surgent group nearly so (117–122).

Figure 2.

Three temperant profiles within ADHD^a

^aThese illustrative data illustrate the large differences between three temperament profiles of children all meeting criteria for ADHD, defined by mathematical grouping of similarity on a set of temperament scales from the TMCQ (Rothbart Temperament in Middle Childhood Questionnaire). A matched group of typically developing children are represented with a zero-score on the chart; all the ADHD group scores are thus standardized to that group. Note that both the Surgent and Irritable ADHD groups have elevated anger relative to typically developing children, but this is more extreme in the irritable group, and the irritable group has slow recovery (high unsoothability) along with more negative emotions generally. In contrast, the surgent group has low shyness or high sociability, as well as far higher scores on high intensity pleasure and assertiveness. These results were not fully explained by selection of ADHD severity (similar levels of ADHD symptoms), subtype, gender/sex, or comorbid disorders. The squares are approximately the size of the +/− 1 standard error range around the data point, thus group separation is large. Data from Reference #(117).

Three lines represent three groups of children with ADHD who differ in their emotional styles. The blue line is a group of typically developing children. The typical developing group is at zero and the squares represent the standard errors, showing that group differences are large on these trait measure.

Exemplifying the value of a multi-dimensional approach to clinical prediction and psychopathology (rather than static categories), longitudinal network modeling of a well characterized prospective sample enriched for ADHD (115) showed that dimensionally measured irritability, working memory weakness, and severity of ADHD symptoms emerged as independent and additive predictors of risk for future worse psychopathology, particularly for future depression and ODD (123). Further, we also demonstrated that the ADHD polygenic score was related to irritability and emotional dysregulation generally just as strongly as it was to ADHD symptoms; and a group with ADHD+ high irritability had a higher ADHD polygenic score than ADHD with typical emotional responding (124) or any other subgroup of youth with ADHD or ADHD sub-profile (including ADHD+ODD). In that sample selected to over-represent ADHD, the ADHD polygenic score but not the depression or bipolar polygenic score was related to the emotion regulation traits and ADHD (note that the polygenic score findings might not generalize to a sample not selected for ADHD).

As shown in Figure 2, this work also identifies a group of youth meeting criteria for ADHD who have only mildly elevated anger but extremes of positive approach emotions—activity level, sociability, excitability. This group was at only mildly elevated risk of worsened outcomes over and above other youth with or without ADHD. Various replication studies have suggested that an emotionally dysregulated subgroup with elevated negative emotions is highly replicable in ADHD samples, while the positive-emotion dysregulation group is only sometimes able to be observed in these cluster-analysis type of studies. Crucially, these groups are not necessarily showing more severe levels of ADHD symptoms—so these studies are not merely recovering mild versus severe cases.

All of this suggests that conceptualizing a refined ADHD phenotype to integrate emotional dysregulation, including negative emotionality and irritability, in some form, (as suggested long ago by Barkley, (125)), would dovetail with etiological structure as well as assist in clinical prediction. Such a refined analysis of the ADHD phenotype(s) may have etiological and clinical value as profiles, perhaps superior to existing DSM-presentations. It is helpful that the major traits have well-developed theory and literature related to neurobiology as well, helping advance integration.

Admittedly, introducing emotional dysregulation as a feature of ADHD in the nosology would not be without complications. For example, it would risk opening up ADHD again to be an over-inclusive category like the old category of minimal brain dysfunction or MBD. It could create confusion about when to add comorbid diagnoses for treatment. Yet a well-defined sub-profile approach could open new insights for research as well as clinical strategies and reduce over-assignment of multiple disorders when not necessary. It could reduce clinical confusion about diagnostic assignment in complex ADHD cases. To avoid the many complexities of overhauling the ADHD diagnostic criteria, an available interim step in a future DSM would be to include a specifier for ADHD with elevated levels of anger reactivity or extreme emotional dysregulation that is short of criteria for a DMDD diagnosis.

Thus, as we look ahead, elaborating on the relative importance of emotional dysregulation in ADHD even in the absence of a frank comorbid anxiety or mood disorder, and of irritability (anger dysregulation) as a component of that general emotional dysregulation, will be helpful in formulating the optimal clinical as well as descriptive strategy, though it requires further definitional work. As the field works to integrate trait dimensions into the existing nosology (with its categories intended to facilitate clinical decision making), traits of temperament related to emotional dysregulation and regulation appear to be particularly helpful in the case of ADHD, and warrant close consideration.

Cognition and ADHD: Self-Regulation, Alertness/Arousal, and Executive Function

A second fruitful arena for low cost and mechanistically valuable phenotype refinement lies with lab assays of cognition and attention. I consider lab assays because they balance the generality and reliability of trait ratings with controlled measurement not subject to rater bias. Executive functioning, reward discounting, and arousal are central and rapidly evolving construcs here. If ADHD is a disorder of self-regulation, then dysregulation of impulse, emotion, and attention makes sense--the neural circuits that support regulation of action and of emotion are closely related (126, 127) and in turn rely on the regulatory role of attention (128, 129).

Executive functions (EF) include response suppression (response inhibition), working memory, planning, task switching, and others effortful or deliberate cognitive functions. Cognitive control--ignoring distraction and adapt attention to task demands; sometimes referred to as self-control or effortful control to include suppressing an impulse to sustain goal behavior. This control function emerges toddlerhood in precursive form and is a foundation for more complex executive functions later in childhood (130). Self-regulation is broader than EF because it encompasses both deliberate or effortful strategies to regulate behavior and emotion like suppressing an impulse or planning and (b) relatively automatic regulatory processes. It also involves (c) social supports (indeed, in early life regulation is heavily dependent on extrinsic supports such as caregiver behaviors). For conceptual elaboration see (11).

Laboratory-measured EFs like working memory and response inhibition/suppression have modest yet reliable associations with ADHD in meta-analyses. Impulsivity is multicomponential, but one aspect measured in the lab is temporal discounting of future reward. Meta-analysis suggests a modest but reliable ADHD-effect of steepened reward discounting gradient (131) and problems with neural reward systems (typically conceived in relation to meso-limbic dopamine circuits) are among the most common proposals for a mechanistic breakdown in ADHD (10, 132).

Most striking in the past decade is a resurgence of computational studies related to arousal. The linkage to personality traits is of interest before proceeding. Arousal was of interest in cognition and central to theories of personality traits in the 20^th century (135, 136). For example, Zuckerman’s (137) sensation seeking, like Eysenck’s extraversion, was related to low cortical arousal leading to stimulation-seeking behavior to optimize arousal state. Gray (138, 139) linked arousal directly to incentive-specific arousal systems--the behavioral inhibition (BIS) and behavioral approach (or activation) system (BAS). Most of these arousal-based trait theories were related to psychopathy or antisocial behavior or conversely to anxiety, with little mention of ADHD (or its predecessors)(140). However, Quay (141, 142) proposed that ADHD was related to underactivity of Gray’s BIS; various reward response theories are relatead to Gray’s BAS though do not always make the connection. In the 1970’s, Satterfield, Zentall, and others (143–145) proposed that ADHD was characterized by sub-optimal central arousal leading to both hyperactivity and inattention, but without reference to personality theory. This century, Soderlund and colleagues (146–148) proposed an updated optimal stimulation/moderate brain arousal theory of ADHD that has been generative.

However, arousal, which encompasses autonomic systems (sympathetic and parasympathetic) as well as multiple neural circuits, has historically been critiqued due to under-specification and inconsistent physiological measurement results (140). The concept is clarified for ADHD when defined as alertness, as suggested by Petersen and Posner (151). Alertness here is modulated by ascending NE neurons from the locus coeruleus to the cortex (the LC/NE circuit). (Noted that alertness is also modulated by dopamine, phasic cholinergic responding and glutamatergic states (152, 153) and other neural circuits (154–156)). The ascending LC/NE system plays a critical role through a process of neural gain, reflected neurally in selective firing and suppression of neural circuits, and reflected cognitively in improved signal: noise ratio during fast decision tasks (153, 157, 158). Neural gain reflects the strength of neuronal response to signal and is correlated with informational gain during attention and decision tasks. It can be proposed as an integrative step for ADHD cognition and attention (159).

How can the contemporary idea of neural gain and alertness be assayed with low cost, deployable measures? Late 20^th century work on ADHD response readiness featured clever studies about the modulation of performance by event rate (160, 161). Subsequently, interest shifted to reaction time variability and then to ex-Gaussian decomposition of reaction times on fast computer tasks, and the discovery of excess very slow responses associated with ADHD (162). Informative as well are well-replicated contemporaneous findings of reduced signal detection efficiency in ADHD, measured using the computational parameter d-prime on continuous performance tests (163).

Arguably the best measure in the past decade is a well established computational decomposition of fast reaction time data called the diffusion model (164). It separates the components of decision threshold, non-decision (e.g., sensory-motor) response time, and rate of information gain (called drift rate), and thus also handles speed-accuracy trade-off elegantly. The drift rate parameter reflects the rate or efficiency of information gain (related to neural gain) via a presumed process of sequential sampling of the environment by the nervous system as seen in the recurrent electrical oscillations in the brain at rest and in activity.

A series of papers in the past decade by Huang-Pollock and colleagues highlighted the utility of this approach. Karalunas and Huang-Pollock (165) showed that the drift rate parameter (slow information gain) statistically accounted for both ADHD slow reaction time and excess very slow responses, and reaction time variability, subsequently confirmed in a meta-analysis (162). Its explanatory power was somewhat better than the earlier method of looking at ex-Gaussian parameters of the RT distribution. The diffusion model also explained poor performance on executive function tasks and accounted for patterns of response on event rate tasks (166). As with other measures of cognitive function, effects are not specific to ADHD. Alertness/gain efficiency appears to be a general marker of psychopathology (167) but as a cross cutting trait may have particular utility in ADHD.

This computational approach links directly to the neural gain models in terms of microsecond information gain efficiency. Information gain is critical to moment-to-moment task as well as social adaptation. Low alertness can manifest as slower information gain during moment-to-moment adaptation to context behaviorally and cognitively, leading to mis-aligned responses in both learning and social contexts, to poor attention, and to impulsivity in conjunction with other systems.

Despite difficulties in psychophysiological measures of arousal-related measures, this perspective is broadly consistent with key findings. For example, ADHD is associated with elevated electroencephalogram (EEG) slow wave (theta band) activity suggesting lower cortical arousal and related to slow and variable responding (though in a subset and with effect size too small for diagnostics (168–170)). A more apt EEG metric for measuring neural gain may be the aperiodic exponent of the (EEG) power spectrum--it is thorught to reflect the ratio of excitation and inhibition in cortical circuits (171) or simplistically, neural noise during task. It is correlated with drift rate in the diffusion model as well as d-prime in the signal detection model (two top cognitive markers of gain), associated with ADHD, becomes less “noisy” with methylphenidate, and marks family history of ADHD when recorded in infants (172–175). In peripheral measures, pupillary dilation and pupillary oscillation studies are also consistent with alterations and LC/NE neural gain alteration in ADHD (176–178), again implying “noiser” information uptake.

Cognitive mechanisms gain support as endophenotypes using both EF and alertness. For example, we (180) asked whether EF or alertness /information gain mediate genetic effects (measured by the ADHD polygenic score) on ADHD itself. We maximized measurement reliability by creating statistical latent variables for (a) alertness (that is, the weighted combination of d-prime, drift rate, and response variability across tasks) and (b) working memory (the weighted combination of several common task) as well as response inhibition. Alertness and working memory (but not response inhibition) mediated the association of the ADHD polygenic score with ADHD symptoms.

Overall, the emergence of strong findings for alertness and neural gain opens a new avenue for ADHD cognition. In conjunction with work on EF and delay reward discounting, it may enable integration and a path for linkage to traits as well.

Conclusion and Recommendations

ADHD is critical as a potential gateway into other more serious behavioral and emotional complications. Identifying mechanisms and predicting its course remains critical. Heterogeneity within the ADHD population affects all areas of research as well as treatment decision making. For clinical decision making and clinical prediction, enriching and refining the ADHD phenotype with novel phenotypes measures appears both necessary and tractable.

For example, emotional profiles moderate cognitive and physiological effects (181). No bio-measure will be isomorphic with the DSM-concept of ADHD unless sub-profiles are developed using enriched trait sets as tools. But the salience of the trait profiles and in cognition of the low alertness/adaptive gain model points toward the potential for low cost, deployable and novel phenotypes to enhance understanding of ADHD. Indeed, statistical studies suggest meaningful subgroups of children within an ADHD population based on trait profiles. These in turn appear to modulate a range of cognitive and physiological findings, in some instances clarifying long-standing controversies (182). In the cognitive domain, recent advances in attentional neuroscience related to computational models of neural gain provide an integrative opportunity to unify various cognitive findings under an umbrella better anchored than was the case with older but flawed ideas of arousal.

Treatment and medication effects.

With regard to trait irritability, methylphenidate appears be helpful (183) although other work suggests it suppresses symptoms of inattention and hyperactivity more than irritability and a small trial showed that there was better improvement with combined stimulant and citalopram for severe irritability (184). Further, parent management and interventions for self-regulatoin remain helpful in this domain (114). Thus, treatment for ADHD with an eye on emotional dysregulation likely will require modifications—particularly when first line treatment disappoints--exemplifying the importance of this type of specifier or sub-presentation in a future nosology. With regard to the neurocognitive parameters, treatment studies support mechanistic utility. Methylphenidate to enhances neural gain (measured by decision performance under different conditions) (185), improves the EEG aperiodic exponent (186), and the ERP P-300 amplitude (187). It also improves classic parameters like d-prime and reaction-time variability, somewhat specifically (188).

Clinical Prediction.

Overall, a key advantage of both trait ratings and computationally decomposed reaction time measures is their low cost and hence field deployability for prediction analytics and clinical treatment matching. In addition to potential specifiers for ADHD using clinical features such as cognitive disengagement syndrome or callous-unemotional traits, the potential for novel phenotypes to enhance ADHD conceptualization and clinical care is highlighted in two domains.

First, temperament measures of emotional regulation, including both negative emotions, positive emotions, and irritability (conceived either as a negative valence emotion or as frustrative approach) can characterize both early precursors as well as important course modifiers (relevant clinical subtypes or profiles) and predictors for ADHD. This has implications for the entire developmental course of externalizing and mood disorders in this population. Further, work on alternative ratings measures of impulsivity, ego-under-control, ego resilience, and related traits are also likely to be useful here. Inclusion of temperament traits or profiles enhances prediction of clinical outcome (117–119, 189). These efforts can be aided by recent measure refinement made to be available reliable, valid, and brief scales with established predictive power. These include recent work updating the Rothbart temperament scales by Karalunas and colleagues (190, 191), as well as established work such as Eisenberg’s simplification of the Block and Block undercontrol and resilience scales (192).

Second, cognitive measures could use a fresh look for clinical prediction. Although the historical and inevitable lower correlation between laboratory measures and trait or behavior ratings (relative to ratings-to-ratings correlations) will cause these measures to have lower correlations with clinical outcomes than trait ratings do, they may be hypothesized to add important value particularly in edge cases where decision making is difficult (193). The available low-cost measures here include those for selected executive functions such as working memory and response inhibition, as well as venerable computational measures of reward temporal discounting. More emphasis should be given, however, to renewed interest in measures of information gain or (presumed) neural gain parameters such as d’ and drift rate discussed earlier. These can be recovered from reaction time sequences in brief tasks and open integrative opportunities for process understanding.

Conclusion.

All of the cross cutting clinical, trait, and cognitive features are non-specific to ADHD or any other disorder. Rather, they are useful for refining phenotype description into unique profiles and specifiers that enhance clinical prediction and understanding. For this same reason, they are less useful in isolation and will be most useful when combined with with ADHD ADHD symptom profiles to enhance forward prediction of clinical course and treatment response decisions for clinical-level problems. In practical application in a future DSM, it is not necessarily clear that new clinical diagnostic criteria are needed although in a longer time frame that may prove appropriate. At this juncture, refinemnets to specifiers and presentations may be a prudent next step yet dramatically enhance understanding and care delivery.

In conclusion, ADHD phenotypes characterization and refinement is needed to further energize and clarify the burgeoning clinical and mechanistic work on ADHD and to inform the coming wave of clinical prediction models using advanced analytics. Phenotype refinement using simple dispositional trait measures should help clarify mechanistic and physiological findings, assist in early neurodevelopmental studies of developmental pathways and early outcome targets for etiological studies, account for heterogeneity of findings in mechanistic and treatment studies, and assist in the next generation of A/I powered clinical prediction and machine learning models.