Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): III. Emotional dysfunction superspectrum

Internalizing spectrum

Internalizing is the largest and most complex of the HiTOP spectra. It consistently emerges as a distinct spectrum in structural analyses. However, the composition of this spectrum is critically dependent on the specific variables included in the analysis. Table 1 summarizes findings from the large number of studies that have modeled internalizing using diagnostic data^{8, 9, 45-87}. Internalizing clearly subsumes a very broad range of psychopathology, including content related to depressive disorders, anxiety disorders, obsessive-compulsive and related disorders, trauma- and stressor-related disorders, eating disorders, and personality disorders.

Several subfactors have been identified within internalizing. Table 1 presents findings related to the two broadest and best replicated subfactors². First, the distress subfactor includes disorders that involve pervasive negative emotionality⁶, such as MDD, dysthymic disorder, GAD and PTSD. Second, the fear subfactor is defined by disorders that involve more specific, context-delimited forms of distress and that frequently include behavioral avoidance, such as panic disorder, agoraphobia, social phobia, and specific phobia. These distress and fear subfactors are strongly correlated, and some studies have found them to be indistinguishable^{47, 52, 67}. Relatedly, some diagnoses – such as obsessive-compulsive disorder (OCD) – do not fall clearly into either subfactor.

Growing evidence indicates that eating pathology forms a third subfactor within internalizing^{2, 77, 78, 88}, although it is sometimes included in the distress subfactor (Table 1). At the syndrome level, this cluster is defined by disorders such as bulimia nervosa, anorexia nervosa, and binge eating disorder^{77, 78}. At the symptom level, structural/psychometric evidence has established the existence of eight specific dimensions: body dissatisfaction, binge eating, cognitive restraint, purging, excessive exercise, restricting, muscle building, and negative attitudes toward obesity. These eight dimensions have been replicated across a variety of populations^89-92.

Evidence has also emerged for a fourth subfactor of sexual problems^{2, 93-95}. This cluster is defined by multiple symptoms of sexual dysfunction, including low sexual desire, difficulties with arousal, low orgasmic function, and sex-related distress.

Finally, several studies have found that indicators of mania/bipolar disorder fall within the internalizing spectrum and often help to define its distress subfactor. However, other studies have linked mania to the thought disorder spectrum^{8, 47, 49}. Accordingly, mania is currently an interstitial construct in HiTOP, with important connections to both internalizing and thought disorder. Mania subsumes several distinct symptom dimensions, including emotional lability, euphoric activation, hyperactive cognition, reckless overconfidence, and irritability^96-100. These symptom dimensions have distinctive correlates, and more fine-grained analyses will likely reveal that they are located in different HiTOP spectra.

Somatoform spectrum

Somatoform is currently the most tentative of the HiTOP spectra². Early evidence suggested that somatoform psychopathology was subsumed within internalizing, based on data that somatization, hypochondriasis and neurasthenia loaded with depression and anxiety on a broader internalizing factor^{101, 102}. However, subsequent research has shown that, when a sufficient set of indicators is available, the somatoform spectrum is indeed separate from internalizing as well as the other HiTOP spectra^{46, 47, 102, 103, 105, 107-117}. These seemingly divergent sets of findings can easily be reconciled. Several studies^{46, 104, 106} have demonstrated convincingly that internalizing and somatoform do form a single spectrum at very broad levels of the hierarchy, but, as one moves further down in levels of abstraction, somatoform separates from internalizing.

Table 2 lists 16 studies^{46, 47, 102, 103, 106-117} conducted across a diverse range of countries – and using a wide range of populations and measurement modalities – that have yielded support for a higher-order somatoform factor. The indicators have mostly represented an array of bodily distress symptoms (e.g., pain, gastrointestinal, cardiopulmonary, chronic fatigue, functional neurological), akin to the bodily distress syndrome proposed by Fink and colleagues^{118, 119}. Although the broader categorical hypochondriasis diagnostic construct has loaded on the somatoform factor in the two studies in which it was included, this construct is multifactorial in nature¹²⁰; it therefore would be important to determine the degree to which the components of cognitive preoccupation, bodily perceptions, reassurance seeking, and hypochondriacal worry load on this somatoform factor. Indeed, absent from all these studies are specific indicators reflecting health anxiety, which clearly includes aspects of both internalizing (i.e., anxious apprehension and fearfulness) and somatoform (i.e., somatic preoccupation and disease conviction) pathology. Future studies need to elucidate the placement of health anxiety in the hierarchy.

Role of maladaptive traits

Negative affect/neuroticism (NA/N) is a fundamental trait domain in research on personality and personality pathology. It also is a key part of the DSM-5 alternative model of personality disorders, as well as a trait qualifier in the new ICD-11 personality disorder diagnosis¹²¹. NA/N cuts across and ties together propensities to experience diverse negative emotional experiences – because these experiences are highly correlated – and thereby represents the central feature of internalizing. Indeed, cross-sectional data show that individual differences in broadly conceptualized internalizing psychopathology and NA/N are very highly correlated and essentially fungible^121-123.

NA/N is a higher-order dimension that subsumes many more specific facets, which are also strongly related to various forms of internalizing. Specific facets of NA/N include anxiousness, depressivity, anger/irritability, separation insecurity, and emotional lability^{2, 124-126}, as well as social cognitive vulnerabilities such as anxiety sensitivity, self-criticism, rumination, hopelessness, and perfectionism. It is noteworthy that these social cognitive vulnerabilities show unique associations with internalizing syndromes^127-130. For example, anxiety sensitivity is associated with panic and other syndromes, net of the general NA/N association with internalizing¹²⁸. In addition, other major personality domains act synergistically with NA/N to affect the likelihood of experiencing specific forms of internalizing. For example, extraversion and conscientiousness mitigate the impact of NA/N on specific internalizing syndromes, such as depression^{131, 132}.

NA/N traits also are predictive of future episodes of internalizing disorders^133-135. Indeed, NA/N can be simultaneously conceptualized as a vulnerability for internalizing disorder, sharing causes with internalizing disorder, and lying within the same spectrum of human variation as internalizing disorder^{136, 137}. These connections may emerge from dynamic processes in which NA/N enhances stress, promoting internalizing symptomatology, and feeding back on general stress reactivity to further reinforce NA/N tendencies^{138, 139}.

The strong association between NA/N and internalizing has led to a focus on articulating shared mechanisms and specific points of continuity¹³⁷. Twin research shows that the close phenotypic overlap of NA/N and internalizing psychopathology is undergirded by shared genetic risk factors^{140, 141}. Distally, emerging molecular evidence also points to a genetic basis for NA/N-internalizing connections¹⁴². More proximally, shared neurocircuitry linking neuroticism to emotional dysregulation may constitute some of the manifest mechanisms underlying close NA/N-internalizing connections¹⁴³.

Finally, NA/N is broadly related to health complaints and somatic symptoms¹⁴⁴; in fact, some models include somatic complaints as a specific facet within this domain^{125, 145}. NA/N has also been shown to be substantially associated with overreporting of health complaints¹⁴⁴, medically unexplained symptoms^146-149, health anxiety and hypochondriasis^{120, 150-156}, and somatization/somatization disorder^157-160.

NA/N is broadly related to the symptoms, traits and disorders subsumed within the somatoform spectrum and, therefore, is partly responsible for its emergence in structural studies. Because NA/N is also broadly linked to the internalizing spectrum, it further helps to explain the existence of the emotional dysfunction superspectrum¹⁶¹, which reflects important commonalities between somatoform and internalizing psychopathology.

Overall model

Figure 1 summarizes the proposed model of the emotional dysfunction superspectrum and its constituent spectra. The sections for internalizing and somatoform build upon the current HiTOP model² in light of the literature reviewed in this paper – in particular, highlighting those areas whose placement within this superspectrum is ambiguous or tentative. The model also includes illustrative symptom and trait dimensions that populate the lower levels of the hierarchy; these are taken from Kotov et al² and subsequent studies.

Internalizing consistently emerges as a distinct dimension in structural models, but its boundaries are unclear. For example, internalizing is strongly characterized by personality pathology related to NA/N^121-123. However, personality disorders that load on internalizing (e.g., borderline and avoidant) often cross over into other spectra (externalizing and detachment, respectively^{46, 58}).

Table 1 demonstrates substantial support for subdividing internalizing into distress and fear subfactors, but evidence for the distress-fear distinction is not universal^{46, 52, 55, 56, 67}. Some studies have found evidence for additional subfactors of internalizing, including sexual problems^93-95 and eating pathology^{77, 78}, although eating pathology may form a separate structural dimension⁵⁵.

The somatoform spectrum is defined by a wide array of somatic complaints, as well as preoccupation with bodily symptoms. Somatoform problems covary substantially with internalizing psychopathology⁵² and, as with internalizing, somatoform psychopathology is strongly associated with individual differences in NA/N¹⁴⁴. Nevertheless, a somatoform spectrum can be distinguished from the internalizing one if a sufficient set of indicators is available^{46, 103, 105}.

Behavioral genetics

Twin studies suggest that the internalizing domain is moderately heritable and under shared genetic influences^{51, 140, 141, 162-167}. A substantial proportion of these genetic influences is also shared with externalizing, but the remaining vulnerability is specific to the internalizing spectrum. Importantly, these studies usually defined the internalizing spectrum as emotional problems, and the strongest genetic loadings were for MDD and GAD¹⁶³. Within this narrower conceptualization of internalizing, there is evidence for separate genetic influences on distress and fear^168-170.

No study has examined genetic and environmental influences on all of the symptoms and traits subsumed within internalizing. However, it is possible to piece together how different HiTOP internalizing syndromes are genetically related from the research that does exist across different combinations of disorders. Multiple forms of eating pathology have common genetic vulnerability^171-173. Moreover, twin studies indicate a shared genetic risk for eating pathology and emotional problems, including anxiety and depression symptoms^{51, 174-177}. There is also a substantial genetic correlation between anorexia nervosa and OCD¹⁷⁸. Finally, twin and family studies indicate a partial genetic overlap between mania and unipolar depression^179-181, although the genetic association between mania and schizophrenia is substantially stronger^182-185. Overall, there is prominent genetic overlap between different conditions within internalizing – except for mania – although there is no research on the genetic overlap with sexual problems.

In contrast, twin studies suggest that a significant proportion of genetic influences on somatoform spectrum symptoms are independent from internalizing problems^{186, 187}. For example, a common genetic factor contributes to four somatic symptoms: recurrent headache, irritable bowel syndrome, chronic impairing fatigue, and chronic widespread pain¹⁸⁸, independent of genetic influences shared with MDD and GAD. Nonetheless, the somatoform and internalizing spectra may share genetic underpinnings at a higher level of generality^{51, 186-191}.

Overall, twin studies support shared genetic influences on the internalizing spectrum that are partially distinct from the genetic etiology of the somatoform spectrum. Future twin studies should assess a wider range of variables to test the genetic architecture comprehensively.

Molecular genetics

Genome-wide association studies (GWAS) detect genetic variants across the entire genome and allow one to compute molecular genetic correlations between traits¹⁹². Many genetic variants, each with a small effect size, have been found to contribute to the shared risk for internalizing. For example, depression shows high genetic correlations with generalized anxiety, NA/N, anhedonia, and PTSD (r_g>0.70)^193-196, as well as much smaller but significant genetic correlations with bipolar disorder, OCD, and anorexia nervosa (r_g=0.17-0.36)¹⁹⁷.

Genomic structural equation modeling (SEM) is another technique for investigating shared genetic influences across related conditions. It can extract common genetic dimensions from a set of molecular genetic correlations, and is thus useful for testing the genome-wide architecture of psychopathology. Using this approach, Waldman et al¹⁹⁸ identified a genetic internalizing factor, characterized by shared genetic influences on depression, anxiety and PTSD. However, bipolar disorder, OCD and anorexia nervosa were influenced by a genetic thought problems factor, rather than by internalizing. Lee et al¹⁹⁷ found that OCD and anorexia nervosa were influenced by a separate genetic factor from depression, whereas bipolar disorder had a uniquely strong association with schizophrenia (r_g=0.70). Finally, Levey et al¹⁹⁹ identified a genetic internalizing factor, which captured shared genetic influences on depression, NA/N, PTSD and anxiety.

Overall, genomic SEM supports a narrow internalizing factor that captures shared genetic influences on distress and fear disorders. Anorexia nervosa and OCD share a separate genetic factor in these studies, in line with the moderate genetic correlation between these conditions (r_g=0.45)²⁰⁰. Furthermore, the genetic vulnerability to bipolar disorder appears to align more closely with thought disorder than with internalizing. However, the high genetic overlap between schizophrenia and bipolar disorder is more specific to bipolar disorder I than bipolar disorder II (r_g=0.71 vs. 0.51), whereas depression is more closely correlated with bipolar disorder II than bipolar disorder I (r_g=0.69 vs. 0.30)²⁰¹. Similarly, bipolar disorder cases with psychosis have higher genetic risk for schizophrenia but lower risk for anhedonia, whereas bipolar cases with a suicide attempt have elevated genetic risk for depression and anhedonia²⁰².

Molecular genetic studies also provide evidence for a genetic distinction between distress and fear factors. Depression and generalized anxiety show a substantial genetic overlap (r_g=0.80), but are partly genetically distinct from fear disorders, such as specific phobia and panic (r_g=0.34 and 0.63, respectively)²⁰³. Moreover, depression and anxiety were influenced by two distinct but genetically correlated factors (r_g=0.80), while NA/N items were partitioned between them²⁰⁴. Likewise, the molecular genetic architecture of NA/N consists of two genetically correlated factors, corresponding to distress and fear^{142, 205, 206}.

As additional GWAS summary statistics become available, more fine-grained models of internalizing can be tested. Furthermore, although there is no GWAS of somatoform spectrum disorders, moderate genetic correlations between chronic pain and depression, anxiety and NA/N (r_g=0.40-0.59) suggest that there may be considerable genetic overlap between the internalizing and somatoform spectra, that is captured by the emotional dysfunction superspectrum^{207, 208}. Finally, genetic correlations can be affected by the heterogeneous psychiatric diagnoses used in GWAS. Homogeneous symptom dimensions can address this heterogeneity and enhance gene discovery^209-211.

Environmental risk factors

Environmental variation shapes the development of all forms of emotional disorder²¹². A vast literature attests to this fact, but studies focus primarily on a single diagnosis or a small cluster of disorders. Only recently has research begun to investigate environmental exposures in relation to quantitative dimensions that cut across traditional diagnostic boundaries.

Few risks are as potent as childhood maltreatment. Abuse and neglect confer long-lasting vulnerability to all types of emotional and somatic complaints. Keyes et al⁴⁹ created a model to explain this non-specificity in the US National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). They showed that maltreatment events predicted individual differences on an internalizing spectrum that represented the commonality among interview-based anxiety and depression diagnoses. Their model also allowed for the possibility of pathways from maltreatment to the unique part of each diagnosis that was independent of all other internalizing conditions. These diagnosis-specific effects were all comparatively weak, however, leading the authors to conclude that the relationship between maltreatment and emotional complaints could be represented solely by maltreatment's link with the internalizing spectrum. Several prospective studies have corroborated this finding^{8, 213-216}.

Adolescent stressors are often proximal triggers for first onsets of diagnosable emotional problems. Social disruption, such as peer victimization, is particularly salient during this period. Forbes et al²¹⁷ hypothesized that victimization's influence on the internalizing spectrum could explain its far-reaching effects. They found that victimization experiences, such as verbal abuse and relational aggression, were robustly linked to an array of self-rated emotional problems. They observed that these various effects were almost entirely mediated by an overarching internalizing factor. Other developmental research has documented the same pattern across a number of different challenges, including romantic problems, family discord, and financial difficulty²¹⁸. Moreover, it appears that differences on the internalizing spectrum predict the occurrence of future significant stressors, setting into motion a vicious cycle of stress exposure and worsening emotional problems^{1, 219}.

Other aspects of the social milieu have demonstrated transdiagnostic effects on emotional complaints. For instance, racial discrimination is linked with a propensity to internalizing distress, but it is not specifically related to any particular type of emotional pathology²²⁰. Similarly, marital dissatisfaction is closely tied to a quantitative internalizing dimension rather than to individual forms of psychopathology⁸⁵. Other parts of the social environment also tend to have stronger effects on internalizing than on its constituent diagnostic categories¹.

It is not groundbreaking to find that environmental stressors are pathogenic. The key insight is that they seem to convey risk for such a broad range of emotional conditions because they operate primarily at the level of the higher-order internalizing spectrum, as opposed to specific manifestations thereof. This will not necessarily be the case across all environmental exposures, emotional phenotypes, or populations, but it is a robust trend thus far.

More research is needed to extend this paradigm to the full range of emotional dysfunction phenotypes. It is particularly important to investigate environmental variation relevant to the somatoform spectrum. Environmental events are implicated in the onset of somatoform disorders²²¹, but there is little research on this topic from a quantitative modeling perspective. Twin, adoption and quasi-experimental designs also are needed to explicate the causal nature of observed effects.

Cognitive and affective difficulties

The internalizing spectrum is associated with cognitive difficulties that can be broadly characterized as cognitive inflexibility and behavioral disinhibition. In addition, affective difficulties – such as hyposensitivity to reward and/or hypersensitivity to punishment – appear intertwined with impaired inhibition, attentional control and decision-making, and contribute to most internalizing disorders. In general, these cognitive-affective problems likely reflect a compromised ability to inhibit intrusive and perseverative thoughts and emotions governing responses such as reward seeking and/or aversion to punishment, thereby contributing to a pattern of aberrant emotional responses and maladaptive decision-making.

Cognitive and affective difficulties are common in disorders within the distress subfactor. MDD has been linked to cognitive difficulties encompassing aspects of psychomotor speed, attention, verbal fluency, visual learning and memory, and executive functioning^222-226. These problems become more severe as the disorder progresses. Similarly, PTSD is associated with temporal changes in severity of problems in attention, memory and executive functioning^{227, 228}. PTSD is also linked with attentional bias towards trauma-related stimuli²²⁹, general inhibitory control deficits²³⁰, and attenuated reward processing²³¹. These problems provide some evidence of reduced cognitive flexibility and behavioral disinhibition.

Cognitive and affective difficulties – which suggest cognitive inflexibility and behavioral disinhibition – are observed in all disorders within the fear subfactor, albeit to varying degrees of severity. There is evidence of mild executive functioning and memory problems in panic disorder, social phobia, specific phobias and GAD^232-236, whereas difficulties found in OCD tend to be more severe²³⁶. OCD is strongly associated with reduced cognitive flexibility, as well as difficulties in other cognitive domains^237-239. Unsurprisingly, anxiety-related disorders are linked to difficulties in social cognition^{239, 240}.

Disorders within the eating pathology subfactor are characterized by difficulties with attentional inhibition, biased attention to disorder-related stimuli, and attentional set-shifting; these are common indicators of reduced cognitive and behavioral flexibility^241-243 that likely underlie problems with emotional regulation and decision-making. There is additional evidence that individuals with eating disorders have compromised visuospatial ability, verbal functioning, learning and memory²⁴⁴. Other evidence suggests that eating disorders are associated with difficulties in integrative information processing, a cognitive perceptual-processing style termed weak central coherence²⁴⁵.

There are limited data related to objective measures of cognitive functioning in individuals with sexual disorders. However, there is evidence of perseverative cognitive schemas^{246, 247}, which are likely attributable to cognitive inflexibility and/or behavioral disinhibition.

Children, adolescents and college students with general internalizing symptoms show sluggish cognitive tempo^{248, 249}, which is linked with associated decrements in processing speed²⁴⁹. Internalizing is also associated with decreased cognitive flexibility in adolescents²⁵⁰, which is consistent with difficulties in executive functions across various internalizing subfactors.

Bipolar disorders I and II are associated with cognitive problems in attention, memory and executive functions^{224, 251-253}. Common with the other internalizing subfactors, there is evidence that bipolar disorder II is associated with reduced inhibitory control²⁵⁴. In contrast to most internalizing conditions, however, bipolar disorder is associated with hypersensitivity to rewards^{254, 255}.

Finally, few studies have explored cognitive difficulties in somatoform disorders. The available evidence suggests that the somatoform spectrum is associated with difficulties in attention and memory, and reduced attentional control in relation to threatening stimuli^{256, 257}. The limited available data suggest that this factor is linked with behavioral disinhibition, but more research is needed.

Neural substrates: neuroimaging

Across the internalizing spectrum, the neuroimaging literature varies by subfactor and modality to include magnetic resonance imaging (MRI) sequences of functionality (i.e., blood oxygen level-dependent activation, connectivity) and structure (i.e., volumetric, diffusion tensor imaging), as well as studies using nuclear imaging to reveal regional metabolic states – i.e., positron emission tomography (PET) and single photon emission computed tomography (SPECT).

This evidence indicates a range of functional disruptions (i.e., diminished or accentuated activity and connectivity) or aberrations (i.e., decreased white matter integrity and reduced volume) in neuroanatomical regions and pathways. The severity of these disruptions and aberrations is influenced by issues involving methodology, disorder comorbidity, illness phase/severity, genetics, pharmacology, and pathophysiology. Nevertheless, most studies show mild-to-moderate differences in comparison to controls or other clinical groups. Overall, the findings highlight the shared underlying neurobiology of the internalizing spectrum, which commonly includes fronto-striatal and fronto-limbic circuitry implicated in compromised self-regulation of behavior and processing of emotions in response to salient reward or punishment.

The literature on the distress subfactor is well established. Borderline personality disorder and PTSD share common neuropathological pathways, namely those included in cognitive-limbic circuitry²⁵⁸. MDD is associated with reduced volume of both cortical and limbic regions²⁵⁹. PTSD and MDD show altered activation in regions associated with cognition and emotion^{260, 261}. PTSD is associated with alterations in white matter tracts involved in executive functions, context learning and memory, salience processing, and emotional control²⁶². MDD and PTSD both show reduced brain volume of specific regions, with PTSD showing greater reductions overall²⁶³. In MDD, there are also significant reductions in white matter tracts involved in cognition, memory and emotion²⁶⁴. For GAD, there is functional and structural evidence of alterations in frontal-limbic neurocircuitry²⁶⁵. Overall, the findings suggest compromised fronto-limbic-striatal circuitry in this subfactor.

There is substantial evidence of compromised functioning and structural differences within the fear subfactor. Most data come from studies of OCD and social anxiety, followed by phobias, with less evidence for other fear disorders. Overall, there appears to be consistent hyperactivation of regions implicated in cognitive-emotional responses to threat^266-272. Alterations in connectivity are shared between fear disorders (e.g., panic disorder and social phobia); although these might include disruptions (e.g., hypoconnectivity) within various interdependent neural networks, most often there are alterations in fronto-striatal connectivity^{273, 274}. Alterations within the sensorimotor network are observed primarily in panic disorder. The limited structural evidence shows compromised white matter integrity, and differences in cortical and subcortical volume^{269, 275}.

The eating pathology subfactor is characterized by compromised self-regulation and aberrant reward processing^276-279. Studies show compromised connectivity and abnormal regional activation in response to reward²⁷⁸. There is also evidence of underlying neuroendocrine dysfunction²⁸⁰. In terms of structural evidence, there are inconsistencies in findings from volumetric studies and a small but growing literature indicating compromised white matter tracts^281-283. Overall, findings provide evidence to implicate disrupted functioning of fronto-striatal circuits involved in cognitive-emotional control.

There is little neuroimaging research related to sexual problems. However, the handful of papers are consistent in showing altered neural activity, namely hypoactivation of areas associated with cognition, motivation and autonomic arousal, and increased activation of the self-referential network^{284, 285}. Few studies have investigated structural differences or white matter integrity in this subfactor.

The mania subfactor is interstitial between internalizing and thought disorder, sharing a number of neural abnormalities with psychotic disorders¹¹. However, in line with the theme observed in internalizing, bipolar disorder is associated with disrupted fronto-limbic circuitry as evidenced by altered white matter tracts and abnormal regional activation^286-289.

There is evidence of structural and functional aberrations in the somatoform spectrum. Due to methodological confounds, the literature is not as strong as in areas such as distress and fear. Nevertheless, the findings suggest disruptions or alterations in the fronto-striatal-limbic network²⁹⁰.

Neural substrates: neurophysiology

Neurophysiological measures provide more direct indicators of neural activity that have greater temporal sensitivity. Internalizing conditions most frequently have been examined using electroencephalography (EEG), including both spectral power and event-related potentials (ERPs), which index a number of different cognitive, emotional and motivational processes.

Frontal EEG asymmetry is a relative difference in alpha power between the right and left frontal regions^{291, 292}. Alpha activity has been shown to index inhibition of cortical activity, and lower frontal EEG asymmetry scores (right alpha minus left alpha) are posited to reflect relatively less left than right cortical activity. Frontal EEG asymmetry has primarily been interpreted via an approach-withdrawal model²⁹³, such that less relative left cortical activity is thought to reflect reduced approach motivation and increased withdrawal motivation.

The distress subfactor has demonstrated the most substantial association with frontal EEG asymmetry²⁹⁴, although the evidence is inconsistent²⁹⁵. MDD and depression symptoms have been associated with a lower relative left frontal EEG asymmetry, both at rest and during emotional and motivational tasks^296-302. Panic disorder³⁰³ and OCD³⁰⁴ have also been associated with a lower relative left frontal EEG asymmetry. In contrast, onset of bipolar disorder is predicted by greater relative left frontal EEG asymmetry³⁰⁵.

The reward positivity (RewP), also known as the feedback negativity, is an ERP component reflecting reinforcement learning and reward system activation³⁰⁶. The RewP has demonstrated the most consistent association with the distress subfactor^{307, 308}. MDD and depression symptoms have been associated with a more blunted RewP in both adolescents and adults^309-316. GAD symptoms have also been associated with a more blunted RewP³¹⁷. The RewP has been associated with risk for, and family history of, MDD^{318, 319}, and has been shown to predict major depressive episodes, first-onset depressive disorder, and greater depression symptoms prospectively^{320, 321}.

The error-related negativity (ERN) is an ERP component that occurs in response to an error of commission and is posited to reflect the increased need for cognitive control and threat sensitivity³²². An enhanced ERN has been associated with both fear and distress subfactors³²³. OCD, GAD and social anxiety all have been characterized by an enhanced ERN^324-330. The ERN has been associated with risk for, and family history of, OCD^{325, 331, 332}, and has been shown to predict the development of first-onset anxiety disorders and GAD prospectively^{333, 334}. Within the somatoform spectrum, initial evidence suggests that health anxiety is associated with an enhanced ERN³³⁵.

The P3 is a widely studied ERP component that is posited to index attentional allocation. Distress, eating and somatoform disorders all have been associated with a reduced P3^336-341. These findings suggest that P3 alterations may be shared across the internalizing and somatoform spectra. Because P3 reductions have also been widely reported in psychosis and externalizing psychopathology^{11, 12}, they may simply represent a marker of general psychopathology³⁴². Enhanced P3, however, has also been associated with the internalizing spectrum, especially with its fear and eating pathology subfactors^343-346.

The late positive potential (LPP) is a later ERP component reflecting elaborative and sustained attention toward motivationally salient stimuli. The distress subfactor has been associated with a reduced LPP to emotional stimuli^347-351, whereas the fear subfactor has been associated with an enhanced LPP to aversive and unpleasant stimuli^{349, 352-355}.

Other biomarkers

Disorders within the internalizing and somatoform spectra share several peripheral biomarkers related to stress reactivity. First, brain-derived neurotrophic factor (BDNF) assessed in blood serum and plasma indexes neuronal survival, synaptic signaling, and synaptic consolidation. Meta-analyses support reduced expression of BDNF in depression, bipolar disorder, suicide behavior, and eating pathology^356-361.

Second, cortisol productivity is a biomarker of hypothalamic-pituitary-adrenal axis function. Increased cortisol levels have been associated with distress^362-365, fear^{233, 366}, and somatoform³⁶⁷ conditions. Blunted cortisol, however, has also been reported^{368, 369}, especially in PTSD³⁷⁰. Mixed findings exist for eating pathology^{371, 372} and may be explained by the heterogeneity in sample composition and symptom severity.

Third, elevated levels of pro-inflammatory markers in peripheral tissues are evident in emotional dysfunction disorders. Meta-analyses found elevated levels of C-reactive protein, interleukin (IL)-6, and tumor necrosis factor (TNF)-α in depression^373-376; IL-6, IL-1β, TNF-α and interferon (IFN)-γ in PTSD^{377, 378}; IL-6 and TNF-α in bipolar disorder³⁷³; and IL-6 and TNF-α in anorexia nervosa³⁷⁹. However, there were no significant associations with bulimia nervosa³⁷⁹. Although it transcends diagnostic boundaries, inflammation might nonetheless be attributable to specific symptoms such as sleep problems, appetite changes, and fatigue^{380, 381}.

Finally, the gut-brain-microbiota axis is closely linked to the stress response, and a differential abundance of gut bacterial groups has been identified in depressive, anxiety, PTSD, bipolar, eating and pain-related psychopathology^{382, 383}. Some bacteria have been implicated across multiple conditions. For example, there is a reduction in the abundance of Faecalibacterium in patients with MDD³⁸⁴, bipolar disorder³⁸⁵, GAD³⁸⁶, and irritable bowel syndrome³⁸⁷.

Overall, peripheral biomarker studies indicate common biological signatures for disorders within the emotional dysfunction superspectrum. However, existing research is constrained by methodological limitations, including small sample sizes and a focus on a limited number of disorders. Moreover, the implicated biomarkers are also associated with other forms of psychopathology, such as schizophrenia³⁸⁸. Studies assessing multiple forms of psychopathology are needed to clarify the specificity versus non-specificity of these biological correlates.

Childhood temperament antecedents

Models of childhood temperament consistently highlight three dimensions that capture tendencies towards negative emotionality, approach-sociability (or surgency), and effortful control (or low impulsivity and disinhibition). These dimensions have close ties with basic traits of normative personality and maladaptive personality pathology^389-392.

Given that NA/N is the core of internalizing psychopathology, it is unsurprising that negative emotionality in childhood predicts subsequent internalizing^{389, 393}. This prospective association has been found not only for core internalizing dimensions, such as depression and anxiety symptoms, but also for eating pathology^394-396 and somatic symptoms³⁹⁷. However, other evidence suggests that youth negative emotionality is a non-specific risk for subsequent psychopathology broadly⁸, particularly externalizing psychopathology^{398, 399}.

Individual differences and behavior genetics research both suggest that low levels of approach-sociability (fearfulness, social withdrawal, behavioral avoidance) together with high levels of negative emotionality may be a combination of traits that differentiates internalizing from externalizing psychopathology^{397, 400, 401}. Interestingly, this combination of high negative emotionality and low approach-sociability may predict anxiety, but not depression⁴⁰². For example, a nationally representative cohort study of 4,983 Australian children followed from age 5 to 13 found that high negative emotionality in early childhood represented a broad risk for subsequent psychopathology, but low approach-sociability only uniquely predicted higher levels of anxiety⁴⁰³. This is consistent with the research finding that behavioral inhibition – a combination of negative emotionality and low approach – is a robust predictor of anxiety^{404, 405}. By contrast, high negative emotionality and high approach-sociability (and extraversion) were found to predict subsequent purging behaviors in adolescence³⁹⁴, which is more consistent with patterns seen with externalizing disorders^{403, 406}.

The third temperamental domain, (low) effortful control, appears to have an inconsistent association that is not specific to internalizing after controlling for concurrent levels of externalizing psychopathology⁴⁰⁴. Similarly, both high and low effortful control (persistence) in early childhood have been found to predict eating pathology in adolescence^{407, 408}. This domain seems to be a more specific and robust predictor of subsequent externalizing¹².

Illness course

Data from the US National Comorbidity Study Replication suggest that anxiety disorders generally have an earlier age of onset (50% by age 11) than depressive disorders (50% by age 32). However, this distinction is largely driven by disorders within the fear subfactor^409-411. Age of onset for somatoform disorders appears to fall in between (50% by age 19⁴¹²). Rates for both anxiety and depressive diagnoses decline in midlife (e.g., >55 years⁴¹³).

Although traditionally discouraged as a diagnosis before adulthood, borderline personality disorder frequently emerges in late childhood or early adolescence⁴¹⁴. Within eating disorders, anorexia nervosa appears to have a mean age of onset between 16 and 19 years, with bulimia nervosa slightly later between 17 and 25 years⁴¹⁵.

Internalizing and somatoform diagnoses follow an episodic, oftentimes chronic, course. Within a hierarchical framework, there are three primary ways of conceptualizing course: homotypic (i.e., course within a single condition), heterotypic (i.e., relations between different conditions over time), and latent liability (i.e., the course exhibited by a shared underlying factor). Psychiatric research traditionally has emphasized homotypic course. For example, using the NESARC dataset, which has two waves separated by approximately three years, Lahey et al⁴¹⁶ found moderate to strong homotypic continuity of six internalizing diagnoses (tetrachoric r = .41-.56). Bruce et al⁴¹⁰ showed that the probability of recovery was only moderate for GAD, social phobia, and panic disorder with agoraphobia, but high for MDD and panic disorder without agoraphobia; however, risk for recurrence was high for all disorders over a 12-year span. Shea and Yen⁴¹⁷ found that MDD showed high rates of both remission and recurrence over a two-year follow-up; in contrast, anxiety disorders had very low recovery rates, even after five years. Similar findings emerge in epidemiological samples, although more individuals appear to recover without recurrence⁴¹⁸.

Two studies of large clinical samples found high rates of remission (85-99%) for borderline personality disorder over the course of 10-16 years, with moderate rates of relapse (10-36%)^{419, 420}. A review suggested that anorexia and bulimia nervosa both show high remission (70-84%) over 10-16 years, with those who have not remitted often transitioning to an eating disorder not otherwise specified⁴²¹.

High rates of comorbidity raise questions of how this covariation manifests across time. Heterotypic continuity frames the question of course in terms of whether a given form of psychopathology (e.g., MDD) at one point in time conduces to another (e.g., GAD) at a later point⁴²². Lahey et al⁴¹⁶ found that heterotypic continuity was widespread within and across internalizing and externalizing diagnoses, although somewhat stronger within spectra. In fact, heterotypic continuity was comparable in magnitude to homotypic continuity, with significant heterotypic effects persisting after adjusting for all other diagnoses. Likewise, heterotypic developmental trajectories are the rule rather than the exception across childhood and adolescence, with childhood symptoms such as emotion dysregulation and irritability considered markers of a broad vulnerability for subsequent mental illness^{423, 424}. Relatedly, Moffitt et al⁴²⁵ found that neither GAD nor MDD preferentially preceded the other, and ordering effects were symmetrical. Few studies have examined the stability of somatoform disorders, but four-year stability in early adulthood was high when considering heterotypic continuity⁴²⁶.

Given this widespread heterotypic continuity, it becomes important to chart the course of the shared liability attributable to the higher-order spectra. In early adulthood (ages 18-25), longitudinal continuity among diagnoses was best accounted for by the stability of a general internalizing factor⁴²⁷. The same appears true in later adulthood, as latent internalizing factors were significantly correlated between age 41 and ages 56 (r=.51) and 61 (r=.43); these associations could largely be explained by genetic factors⁴²⁸. Relatedly, the substantial heterotypic continuity of depression and anxiety symptoms, and of different eating pathology symptoms, was largely attributable to stable, common genetic influences^{173, 429, 430}. Finally, Wright et al⁴³¹ found that an interview-assessed, disorder-based internalizing factor strongly predicted a symptom-based internalizing factor (beta=.60) assessed via daily diary 1.4 years later. Overall, the evidence suggests that spectra represent the primary pathways of illness course, and constitute liabilities for the development of multiple conditions across the lifespan.

Treatment response

Given the high rates of comorbidity and the ubiquitously positive treatment response to cognitive behavior therapy (CBT) across various internalizing disorders^432-434, there has been a focus on testing treatments that were designed to be transdiagnostic (i.e., target multiple disorders). Meta-analyses of transdiagnostic theory-based CBT protocols for internalizing have demonstrated medium to large effect sizes for anxiety and depression, that were maintained at post-treatment follow-up^432-435. There are particularly large effects for CBT in youth when parents are more involved in treatment⁴³⁶.

Findings indicate no significant differences between transdiagnostic CBT and disorder-specific CBT protocols, which supports the efficacy of transdiagnostic CBT for internalizing^{434, 435}. Moreover, although there has been concern about including certain diagnoses (e.g., OCD and PTSD) in transdiagnostic CBT treatments, Norton et al⁴³⁷ showed that transdiagnostic treatments for DSM-IV anxiety disorders were not associated with differential outcome by diagnosis.

Similar to transdiagnostic CBT, the unified protocol (UP) for the transdiagnostic treatment of emotional disorders was specifically designed to target co-occurring internalizing disorders^{438, 439}. Studies show that the UP is equivalent in effectiveness to gold-standard treatments designed to target single disorders^{438, 440}. The UP is much more efficient than single-disorder treatments, because clinicians only need to learn one protocol to treat internalizing disorders. Preliminary efficacy data show that, across diagnostic categories, the UP results in significant improvements in daily functioning, mood, depression, anxiety, and sexual functioning^441-444. Treatment benefits from the UP were maintained at 6- to 12-month follow-up^443-445. Transdiagnostic interventions are now being extended to flexible modular protocols in adults⁴⁴⁶, mirroring efficacious modular transdiagnostic treatments across the internalizing spectrum in youth⁴⁴⁷.

Interpersonal psychotherapy (IPT) is efficacious for treating certain internalizing disorders, such as depression and bulimia nervosa^{448, 449}, although results were less pronounced and slower to emerge for the latter condition⁴⁴⁹. One review indicated that IPT was superior to CBT in treating depression⁴⁴⁸. Variants of IPT, including interpersonal social rhythm therapies (IPSRT), are beneficial as acute and maintenance treatments for both unipolar and bipolar depression^450-452, but have not been studied extensively in other forms of internalizing. Thus, there is support of IPT as a treatment for some, but not all, forms of internalizing, with the majority of research showing that it may be a useful treatment for distress and eating disorders, with limited efficacy for fear-based disorders, such as social phobia⁴⁵³.

The limited available evidence indicates that treatments used for internalizing disorders (i.e., CBT and antidepressants) also are efficacious for somatic symptom disorders^{221, 454}. Although findings are mixed, CBT has been found to have lasting benefits for up to 12 months post-treatment^455-458.

Turning to pharmacological treatments, selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs) are efficacious for the treatment of several internalizing disorders compared to placebo^{459, 460}; however, SSRIs are associated with an increased risk for sexual dysfunction⁹³. Meta-analyses showed that atypical antipsychotics were significantly more efficacious for treating unipolar and bipolar depression and PTSD compared to placebo^461-464. Another meta-analysis of off-label uses of antipsychotics found that quetiapine resulted in significant improvements in GAD symptoms, whereas risperidone significantly reduced OCD symptoms⁴⁶⁵. A large clinical trial found that olanzapine significantly increased weight gain in the treatment of anorexia nervosa compared to placebo⁴⁶⁶. However, atypical antipsychotics had limited benefits for improving quality of life in people with depression⁴⁶⁷ and did not impact psychological symptoms in individuals with anorexia nervosa⁴⁶⁶. Overall, substantial data indicate that SSRIs and SNRIs are beneficial for treating most internalizing conditions, with accumulating evidence that atypical antipsychotics may be useful adjunctive medications. The available evidence for the efficacy of pharmacological treatments for somatoform disorders appears mixed and of low quality⁴⁶⁸.

Summary of validity evidence

Table 3 summarizes the validity evidence reviewed in previous sections. It is noteworthy that virtually all associations are transdiagnostic in nature. That is, the studied variables are not simply related to a single form of psychopathology, but rather are associated with multiple conditions within the emotional dysfunction superspectrum (and, in many cases, to other forms of psychopathology as well). Studies have shown that multiple dimensions falling within the superspectrum share genetic diatheses, environmental risk factors (e.g., childhood maltreatment, financial difficulty, racial discrimination), cognitive and affective deficits (e.g., cognitive inflexibility, behavioral disinhibition), neural substrates (e.g., impaired fronto-striatal and fronto-limbic circuitry, blunted RewP, enhanced ERN) and other biomarkers (e.g., pro-inflammatory markers), as well as childhood temperamental antecedents (e.g., high negative emotionality, low surgency). Not surprisingly, therefore, dimensions within this spectrum respond to the same transdiagnostic treatments (including CBT and SSRIs) and are substantially related to one another both concurrently and prospectively.

These validity data are quite congruent with the structural evidence reviewed earlier. That is, many variables are related to both internalizing and somatoform conditions, and these shared factors can be captured by the emotional dysfunction superspectrum; other variables are more clearly linked to one spectrum than the other, thereby accounting for their emergence as distinct spectra at a lower level of the hierarchy. Similarly, some variables show relatively non-specific associations with all major forms of internalizing, which helps to account for its coherence as a structural dimension; in contrast, other variables show stronger links to some types of internalizing than to others, consistent with the emergence of distinct subfactors within internalizing.

Two caveats are important to mention. First, several validators were also linked to other spectra (e.g., the psychosis superspectrum also responds to antipsychotics, the externalizing superspectrum also shows high childhood maltreatment, and all three superspectra are positively associated with pro-inflammatory markers)^{11, 12}, such that the specificity of these associations is uncertain. Second, some internalizing conditions show a distinct profile on certain validators, which underscores the value of the lower levels of the HiTOP hierarchy. Mania, in particular, is distinct with regard to genetic liability, affective deficits, and episodic course.