Parsing ADHD with Temperament Traits

Abstract

It is widely agreed that the DSM-5, the handbook of psychiatric diagnosis, suffers from both high overlap among its putative disorders and high heterogeneity (variability) within each disorder. While these may appear to be opposite problems, in fact both may stem from failure to recognize transdiagnostic dimensions of emotion, cognition, and personality, among others, that inform psychopathology. These fundamental nosological challenges are exemplified in the case of attention-deficit/hyperactivity disorder (ADHD). In ADHD, broad clinical heterogeneity has defied easy clinical prediction of outcomes or clean statistical differentiation of meaningful, biologically informative sub-groups. Progress for ADHD heterogeneity looks promising, however, when we consider dimensions of trait affectivity such as surgency and negative affectivity, their constituent lower order traits such as irritability, and the integrative function of self-regulation. Focusing on developments in the study of temperament traits and ADHD as they relate to emotional dysregulation, several lines of investigation are proving useful. Utilization of selective computational models, biological validators, and longitudinal analyses points toward potential improvements in nosology and clinical assessment in the future by taking temperament traits into account.

ADHD is an under-rated problem

Mental disorders¹ collectively comprise the single greatest source of health burden in the world, accounting for nearly a quarter of total burden (Whiteford et al., 2013). Problems with inattention, disorganization, impulsivity, and hyperactivity are a surprisingly important player. When severe and impairing, these problems are usefully organized clinically into a syndrome called attention-deficit/hyperactivity disorder (ADHD). ADHD at first may seem a minor player in the saga of mental health related impairments alongside such crippling conditions as major depression, severe alcoholism, bipolar disorder, and schizophrenia. Indeed, some children under the ADHD umbrella do attain a relatively benign outcome with manageable remaining difficulties; and when stimulant medications work, as they do in a majority of cases, the short-term benefit is among the most dramatic in psychiatry.

However, a closer look reveals ADHD as in fact vastly under-appreciated in the matrix of mental-health related burden and cost. It is among the earliest emerging and most common of behavioral conditions in childhood. Despite the availability of ever more sophisticated treatments, long-term outcomes remain deeply concerning (Erskine et al., 2016; Hinshaw & Arnold, 2015; Swanson et al., 2017). These outcomes are striking. ADHD confers a 50% to 300% increased risk for serious secondary mental health problems including substance use disorders, depression, psychosis, and anxiety disorders, which appear to be often causally related to prior ADHD (Riglin et al., 2020; Treur et al., 2019). Especially when complicated by secondary depression or substance use, ADHD magnifies the chances of costly to disastrous life outcomes, too--school and occupational failure, poor health, homelessness, unemployment, injuries, and suicide (Forte et al., 2020). Whereas mental disorders are among the most substantial direct or indirect drivers of premature death worldwide (Walker, McGee, & Druss, 2015), ADHD is right in the mix. When it flowers into comorbid mood, rule breaking, and/or substance abuse (as is often the case), premature death due to accident, suicide, or early onset health problems is 20x more likely that for someone without this profile (Sun et al., 2019).

In short, ADHD is of more serious public health importance than often assumed. It represents a critical early risk phenotype for future severe outcome, albeit with the possibility of resilience or recovery. Understanding it can transform a developmental understanding of psychopathology more generally.

A central phenotype problem is the relation of emotion dysregulation to ADHD

One obstacle to progress on ADHD has been the remarkable heterogeneity of this population. Despite efforts to narrow the diagnostic criteria in the past 40 years (making it one of the most reliable clinical diagnoses in all of psychiatry), it continues to be assigned to children with widely varying profiles, histories, treatment responses, and outcomes. One clue may be variation in emotional dysregulation. Emotional dysregulation was actually part of the mid-20^th century clinical concept of “minimal brain dysfunction” that pre-dated ADHD. That concept became too vague and over-inclusive and so was abandoned for a more focused formulation of attention-deficit disorder (ADD) and then ADHD in last quarter of the 20^th century. But emotional dysregulation remains a major feature of ADHD for most children in this population.

Many different terms are used to describe the domain of emotional dysregulation. I use global term emotional dysregulation to mean problems with excessive and contextually inappropriate or maladaptive positive emotionality (e.g., extreme excitability or hope leading to impulsive action or inappropriate social behavior) and negative valence emotionality (anger, worry, fear, sadness). However, I distinguish irritability or anger dysregulation, from other negative affect (worry, sadness). Anger can be associated with approach or positive affect (goal frustration) as well as negative affect, and as noted later, behaves differently in relation to ADHD genetics than sadness or worry.

Of course, emotional dysregulation is not unique to ADHD—it occurs a wide range of conditions whether or not that is reflected in diagnostic criteria (oppositional defiant disorder, conduct disorder, anxiety disorders, mood disorders, bipolar disorder, borderline personality disorder). Likewise, irritability is an explicit part of the diagnostic criteria for oppositional defiant disorder (ODD) and disruptive mood dysregulation disorder (DMDD). Children can have ADHD and be quite irritable, however, without meeting criteria for ODD (e.g., can have extreme tantrums and be touchy without being generally defiant or argumentative).

To measure dysregulation, the field of emotion research boasts several sophisticated strategies. Yet clinical and developmental scientists have often anchored studies with simple rating scales derived from temperament trait models, due to their low cost and readiness for clinic deployability.

Temperament scales are in fact quite useful for exploring the ADHD phenotype. Our lab has utilized scales and theory from the work of Mary Rothbart and colleagues. In her work, temperament refers to patterns of emotional response and regulation measured as dimensional traits (Rothbart, 2011). It features three broad, well-validated super-ordinate domains: (1) negative affect, (2) surgency, and (3) effortful control. Negative affect is largely as just defined. Surgency reflects high levels of sociability, activity, positive affect, and high-intensity pleasure (enjoyment of highly stimulating activity). It thus maps largely onto the positive affectivity domain. Effortful control means the ability to use intentional or goal-oriented effort to overcome an immediate temptation. Both in item content and conceptually, it has considerable overlap with ADHD itself. For that reason, it is the least useful of the three super-ordinate domains in parsing heterogeneity in ADHD, and I do not much discuss it herein. Item overlap with ADHD is not a significant concern for the affective domains.

Consistent with strategies recommended by the NIMH Research Domain Criteria (RDoC) initiative (Kozak & Cuthbert, 2016), these broad affective reactive domains as well as the regulation (effortful control) domain have hypothesized neurobiological correlates in the interaction of amygdala-PFC circuitry, dopaminergic reward networks, and prefrontal-striatal and cortical-cortical control networks, respectively (Petersen & Posner, 2012; Posner & Rothbart, 2000; Whittle, Allen, Lubman, & Yücel, 2006). Those rich, increasingly well-supported theoretical formulations enable useful inferences linking conceptual neurobiology with ADHD variation as well, a topic that space requires we bypass herein.

Each major trait also includes lower-level component traits in a hierarchical trait model. Among the sub-domains, anger and soothability (i.e., inability to recover from emotional upset) are central to irritability as noted. We and others have used trait ratings to map the ADHD phenotype more deeply.²

Phenotypic heterogeneity and temperament

One line of work we have emphasized asks whether trait profiles have any value-added in understanding ADHD from a clinical or phenotypic perspective. Using a case control design and recruiting from the community to avoid clinic-referral selection bias, we identify cases of ADHD and identify typically developing children who might have had average rates of other psychopathology but clearly did not have ADHD.³ Parents rated the children’s typical behavior on Rothbart’s Temperament in Middle Childhood questionnaire (Simonds & Rothbart, 2004)).

In a first major study, a graph theory-based computation called a community detection analysis identified three separable ADHD profiles (Karalunas et al., 2014) (Figure 1A). These were labeled as Surgent (ADHD-S in the figure, connoting positive affectivity, or in everyday parlance, Risk-taking), Irritable (ADHD-N in the figure due to negative affectivity, or in everyday parlance Angry), and Mild (ADHD-M in the figure, normative temperament scores despite elevated ADHD symptoms). All three groups had elevated problems with effortful control and ADHD symptoms, but they differed markedly in their profile on the affective domains. They did not differ reliably in gender ratio, or in overall ADHD symptom level.

Figure 1:

Clinical validity and utility of a temperament-based taxonomy for ADHD. A: Community detection analysis strongly distinguishes three profiles schematically represented here despite similar levels of ADHD severity. (The size of the square is roughly equal to the standard error). N-negative valence/irritable. S=surgent/positive valence. M=mild. B: These profiles strongly differentiate in regard to clinical outcome and out-perform other baseline clinical features (Karalunas et al. 2014). For these figures, n=500 (ADHD=310, non-ADHD=190). 15 features were included in model, 3 are shown here for ease of presentation along with ADHD symptom totals. Data updated from Karalunas et al (2014).

These groupings had notable utility in clinical prediction and prognosis on 12–24 month follow up (Figure 1B). Figure 1B illustrates that ADHD-N (the irritable/angry group with high general negative affect) had a markedly higher rate of onset of new disorders at 1–2 years later. They gained new diagnoses of anxiety, mood, and oppositional defiant disorders. Several follow up studies in this cohort at other ages, and in another cohort, were also supportive. However, other replication attempts in the literature yielded results that tempered the differentiation of positive and negative valence dysregulation. It seems likely that there is a robust distinction within the ADHD population between those with fairly normative emotional regulation versus with marked emotional dysregulation. However, distinguishing between children with positive-versus negative-valence dysregulation is less robust.

Using polygenic scores to further clarify the role of emotional dysregulation in ADHD

A key question in the field is whether emotional dysregulation ought to be conceptualized as an intrinsic aspect of the ADHD syndrome, or when present really pertains to a nascent mood disorder. This question is reasonable because on the one hand, as noted earlier emotional dysregulation was commonly observed as part of the old minimal brain dysfunction syndrome. On the other hand, children with severe irritability, for example meeting criteria for disruptive mood dysregulation disorder, are also at risk of mood disorder (Stringaris, Zavos, Leibenluft, Maughan, & Eley, 2012). However, it may still be the case that when dysregulation or irritability occur in ADHD they are also part of the ADHD syndrome and part of the reason ADHD is a pathway to later depression. To clarify matters, we examined the structure of trait profiles in relation to polygenic risk.

The logic here was two-fold. First, if emotional dysregulation is really in some sense fundamental to the ADHD phenotype as a neurodevelopmental and heritable condition, then the ADHD polygenic risk should be related to emotional dysregulation to a similar degree as ADHD symptoms themselves. Second, if the if positive and negative valence domains are differentially important, then they might show differential association with ADHD genetic liability. As a competing test, we looked at the association of these valence domains with polygenic risk for depression.⁴

To maximize signal for the polygenic scores, we created robust latent variables using multiple (parent rated) measures of ADHD symptoms, surgency (positive valence), irritability, and negative valence domains other than irritable-angry (fear, sadness, worry). This was in 514 children of whom just over half had well-characterized ADHD and the others were “typically developing children” who could have other disorders but did not have ADHD, all recruited from community volunteers. We then fit these simultaneously in a structural equation model. Figure 2 is a schematic of these results (full data available in Nigg, Karalunas, et al. (2020)). When all relations are fit together in this way, each individual effect shows the association after controlling for all other associations in the model. Figure 2 shows that the genetic risk for ADHD is related equally to ADHD symptoms, to positive emotions/surgency, and to irritability or anger dysregulation. This suggests that from a genetic liability perspective, emotional dysregulation is part of an ADHD pathway and not explainable by depression risk.

Figure 2. Polygenic risk scores support association of emotion dysregulation with ADHD.

In this informal schematic, each circle is a latent variable with multiple indicators. For irritability, indicators are Oppositional Defiant Disorder irritable symptoms and Temperament in Middle Childhood Questionnaire (TMCQ) anger and soothability. For surgency they are relevant TMCQ scales. Negative affect indicators are TMCQ scales for sadness, worry, and fear. ADHD indicators are multiple parent ratings measures. Solid arrows represent statistically reliable non-zero associations. All relationships were tested; no line means no reliable association. N=514, covariates not shown. Based on data from Nigg, Karalunas et al., 2020.

Another way to see this is to stratify the children with ADHD in different ways and then compare their polygenic risk. In this analysis, it is crucial to first adjust the scores for the ADHD symptom severity—otherwise, we may unwittingly merely compare children with more versus fewer ADHD symptoms, not a very useful comparison for ADHD polygenic risk in this case. With that statistical correction in place, most ways of partitioning the ADHD group do not differentiate them on polygenic risk. ADHD-combined versus ADHD-primarily inattentive profiles barely differed (d<.08). The ADHD irritable group did not differ from the rest of the ADHD children (d<.06). However, lumping ADHD-S and ADHD-N together so that we can compare all dysregulated to all other ADHD, the effect size was d=.224, p<.05. While that effect size is still modest, it is reliable and tells us that this last comparison is etiologically meaningful. This result was consistent with and extended an earlier report (Riglin et al. (2017).

These results together help resolve two longstanding scholarly and clinical debates about ADHD. The first debate has been whether emotional dysregulation in ADHD should be viewed as a sign of an incipient comorbid mood disorder, or should be considered part of the ADHD syndrome as much as any other symptom domain. Note that a core feature need not characterize all children with a condition. For example, in the ADHD population not all are considered hyperactive. Likewise, not all are expected to have marked emotionally dysregulation. These data support the idea that it in cases of ADHD marked emotional dysregulation may often be part of the condition, and not an indicator of comorbidity.

The second debate is whether the kind of emotional dysregulation seen in ADHD primarily pertains to irritability/anger dysregulation, or to both positive and negative emotional dysregulation. The latter position was proposed many years ago by Barkley (1997). These results support that position with the caveat that surgency and irritability were related to ADHD genetic risk, but sadness/anxiety/worry were not. A further caution is that although we distinguished positive (ADHD-S) and negative valence (ADHD-N) ADHD groups, the surgent or positive valence group still had somewhat elevated anger scores (but with normative soothability, or recovery) as seen in Figure 1A. Overall, this perspective allows for integration with an understanding of ADHD as a disorder of self-regulation, encompassing dysregulation of attention, activity level, impulsivity, and emotion, rather than a disorder of inattention or hyperactivity per se.

Using prospective trait studies to examine etiology

But how does this fit in with development? A major question is the developmental process by which negative emotionality or disrupted emotional regulation fits into the ADHD risk process. A clue rests in theories put forward by several leaders in the developmental and psychopathology fields regarding the interplay of emotional and cognitive regulation in early life. Perhaps the earliest and clearest articulation has come from work of Mary Rothbart (see in particular, her summary in Rothbart (2011)). The idea here, which several authors in the field have pursued, is that excess early negative affect disrupts top down regulatory processes (e.g., precursors to executive function, such as effortful control). In turn, this sets the state for a profile that begins to take shape as ADHD by late preschool or early elementary school.

Nigg, Goldsmith, and Sachek (2004) were among the first to sketch this temperament-based multiple-pathway perspective for ADHD. They suggested that ADHD might emerge for some children primarily via excess early negative emotionality, leading to a phenotype of ADHD with high impulsivity, anger reactivity, and perhaps reactive aggression. For other children, ADHD would emerge via excess approach (“exuberance” or positive affectivity), leading to excessive risk-taking and impulsivity. Both of these would also lead developmentally to disrupted development of effortful control or cognitive control. Finally, cognitive control could be weakened later in preschool development for other reasons, with a phenotypic expression of inattention and executive dysfunction but not salient dysregulated emotion. Their model was further elaborated by Nigg, Sibley, Thapar, and Karalunas (2020).

Developmental studies increasingly provide some support for that perspective. With regard to negativity, infant (Miller, Hane, Degnan, Fox, & Chronis-Tuscano, 2019) and toddler (Goldsmith, Lemery, & Essex, 2004; Rabinovitz, O’Neill, Rajendran, & Halperin, 2016) anger reactivity predicted ADHD symptoms in middle childhood, mediated in one study by disrupted inhibitory control (related to effortful control, cognitive control, and impulsivity) at age 5 years (Miller, Hane, et al., 2019). Yet on the positive reactivity side, infant positive reactivity or high approach (along with activity level) also correlate with subsequent child ADHD symptoms (Goldsmith et al., 2004; Jonas & Kochanska, 2018; Miller, Degnan, Hane, Fox, & Chronis-Tuscano, 2019). As a caution, however, sex differences in some of these effects were also seen. Given the higher incidence of ADHD, ODD, and severe tantrum problems in boys than girls, sex differences in developmental pathways and mediators clearly need further examination.

In a search for early mechanistic factors, we have begun to extend this work to the prenatal period, initially in a prospective study in a small cohort of 62 children from women followed since the second trimester of pregnancy. Figure 3 illustrates those results. Infant negative reactivity observed in the lab did indeed relate to ADHD symptoms rated by mothers (and teachers) at ages 4–6 years of age and mediated maternal distress to later offspring ADHD, consistent with the proposed developmental cascade. This effect held even after controlling for concurrent offspring negative affect, maternal ADHD symptoms, or family history of ADHD in other first or second degree relatives.

Figure 3.

Maternal prenatal distress is related to offspring ADHD symptoms in this figure, but the effect is mediated both by maternal inflammation measured in blood in 3^rd trimester (a cytokine composite) and by 6 month negative affect observed in a laboratory temperament task. The serial mediation path (from distres→inflammation→6m negative affect→4–6 year old ADHD was reliable but barely (B=.18, p=.05). Overall model fit was excellent. Circles represent latent variables with multiple indicators (not shown). Dashed line indicates a path that was estimated but was not statistically significant. ADHD Symptoms R² = .70. N=62. Note: *p<0.05, **p<0.01.

Previous work in animals has shown that maternal prenatal cytokines influence brain development and behavior in offspring (Schaafsma et al., 2017), and prior human studies have shown associations to offspring behavior in toddlerhood (Graham et al., 2018). However, the relation to offspring ADHD symptoms at an age when they could be reliably assessed in early childhood had not previously been examined. We found that maternal systemic inflammation (indexed by a set of related cytokines assayed from her blood in the 2^nd trimester), also mediated the association of maternal distress with offspring negative affect and ADHD. Figure 3 shows the serial mediation path. That result was bolstered by other findings (not shown) that sufficiently high intake of omega-3 long chain fatty acids protected against the inflammation effect on infant negative affect (Gustafsson et al., 2019). This could point the way to new kinds of early identification and intervention.

These findings, while striking, require caution—the human studies are all small and preliminary and clearly need more exploration before we can be sure of them. That effort is currently underway in a much larger prospective sample. That work will pursue the hypothesis that maternal risk factors operate through elevated systemic inflammation that in turn disrupts neural development in the offspring and contributes to early negative affect dysregulation and then cascades to subsequent poor effortful control and ADHD symptoms.

However, even if these particular results later require revision, these data illustrate the kind of logic that we think will be helpful in coming to understand how emotion regulation participates developmentally in the more general dysregulation of cognition, behavior, and affect that is involved in ADHD.

Conclusion

Overall, the line of work sketched here supports the value of integrating trait measures of temperament or emotional reactivity into our understanding of ADHD. In particular, high level traits of surgency (positive valence) and negative affect valence and their constituent lower order traits, particularly anger regulation (irritability) are useful in organizing ADHD clinical and biological heterogeneity. At least in our hands, they outperform existing ADHD subtypes or profiles that are represented in DSM-5 on both genetic signal differentiation and future clinical prognosis prediction. Effortful control is less useful because alterations in effortful control are so widespread in the ADHD population. Notably, temperament traits may also be helpful in identifying mechanisms of comorbidity—for example, high irritability may contribute to overlap of ADHD with ODD, even though not all irritable ADHD youth will have ODD.

As I noted at the outset, this logic of using neurobiologically supported domains to address heterogeneity and comorbidity will not be unique to ADHD. Rather, it is an approach that is being widely considered across many areas of developmental psychopathology (Beauchaine & Hinshaw, 2020; Smith & Pollak, 2021).

Simple measures of emotion regulation, such as parent ratings of temperament traits could prove to be valuable adjuncts to clinical assessment. More work would be needed to establish the appropriate normative data, but the conceptual evidence has become quite notable. The prospect that an early developmental cascade involving different kinds of temperament trajectories leads to ADHD is tantalizing as well. It could open up new possibilities for early intervention, for example caregiving interventions in infancy, that would support the development of at-risk children.