Neurocognitive deficits have been identified in essentially every psychiatric disorder and reflect one of the most common transdiagnostic features of child, adolescent, and adult psychopathology.Reference Millan, Agid and Brune1–Reference Snyder, Miyake and Hankin3 Neurocognition predicts a host of clinical outcomes in adult psychopathology, including long-term functional recovery,Reference Jaeger, Berns, Loftus, Gonzalez and Czobor4, Reference Gruber, Rosso and Yurgelun-Todd5 overall functioning,Reference Martino, Marengo and Igoa6 quality of life,Reference Cotrena, Branco, Shansis and Fonseca7 and social/occupational functioning.Reference Withall, Harris and Cumming8, Reference O’Donnell, Deldin and Grogan-Kaylor9 Furthermore, deficits in child/adolescent neurocognition have been associated with concurrent and future psychopathology, as well as academic, global, social, and functional outcomes.Reference Huang-Pollock, Shapiro, Galloway-Long and Weigard10–Reference Lee, Hermens and Redoblado-Hodge15 Despite the established importance of neurocognition, there is no nosology of neurocognition, and available interventions that specifically target neurocognition are limited (e.g., medication management).
Nearly, every research study to date has identified neurocognitive differences between a given psychiatric disorder and healthy controls, including the many meta-analyses documenting these differences. Despite this, there is limited evidence to suggest there are any neurocognitive differences between specific psychiatric disorders (as opposed to compared with healthy controls). Most recently, in one of the largest studies to date, Doyle et al.Reference Doyle, Vuijk and Doty16 examined intelligence, reaction time variability, and executive functioning (EF) in 486 youth referred for neuropsychiatric evaluation. All examined childhood psychiatric disorders (i.e., mood disorders, attention deficit hyperactivity disorder [ADHD], autism spectrum disorder [ASD], and psychosis) were associated with neurocognitive deficits, without significant differences between the diagnostic groupings. Furthermore, this pattern of deficits was not driven by comorbidity. The authors noted that commonly held notions of disorder-specific deficits (e.g., inhibition in ADHD) were not supported, and generally, no deficit was specific to any disorder. Within-disorder neurocognitive impairment also has significant variability: for example, bipolar disorder (adult; 27–93%),Reference Godard, Grondin, Baruch and Lafleur17–Reference Reichenberg, Harvey and Bowie19 depression (adult; 23–81%),Reference Godard, Grondin, Baruch and Lafleur17–Reference Lambek, Tannock, Dalsgaard, Trillingsgaard, Damm and Thomsen21 schizophrenia/schizoaffective (adult; 55–84%),Reference Reichenberg, Harvey and Bowie19 anxiety (adult; 18–50%),Reference Gualtieri and Morgan18 and ADHD (50–89%).Reference Lambek, Tannock, Dalsgaard, Trillingsgaard, Damm and Thomsen21, Reference Kofler, Irwin, Soto, Groves, Harmon and Sarver22
Related to above, Kofler et al.Reference Kofler, Irwin, Soto, Groves, Harmon and Sarver22 identified EF deficits in a majority of children with ADHD, yet deficits were divided across EF subdomains: 35% of the sample had only working memory deficits, 16% had only set shifting deficits, 13% had working memory and inhibitory control deficits, 11% had working memory and set shifting deficits, 11% had no deficits, 7% had inhibitory control and set shifting deficits, 4% had only inhibitory control deficits, and 4% had all three deficits. Additionally, a 2015 review paper found no clear neurocognitive impairments in child and adolescent depression,Reference Vilgis, Silk and Vance23 whereas an earlier meta-analysis (2014) did conclude that neurocognitive impairments were evident in depression.Reference Wagner, Müller, Helmreich, Huss and Tadić24 This highlights the significant variability that can occur in research methodology, measurement of cognition, and across individuals with the same psychiatric disorders.
The overall aim of this perspective is to discuss potential neurocognition-centric approaches to investigation and clinical care, specifically by: 1) summarizing findings from recent cluster analysis studies and 2) considering how these findings could guide improved neurocognition-focused diagnosis and treatment in psychopathology.
Neurocognitive Clusters
As noted above, the traditional approach to investigating disorder-specific neurocognitive profiles has significant limitations. An alternative is to take a neurocognition-centric approach by investigating neurocognitive profiles or phenotypes across psychiatric disorders. Such neurocognitive phenotypes may be more closely linked, for example, to underlying pathophysiological pathways and likely do not directly align with current psychiatric diagnostic criteria.Reference Robbins, Gillan, Smith, de Wit and Ersche25 The development of neurocognitive phenotype nosology would allow for: 1) improved information on the natural course and outcomes of neurocognitively homogeneous subgroups and 2) development and implementation of novel cognition-targeted treatments. This conceptualization fits well within the National Institute of Mental Health (NIMH) Research Domain Criteria (RDoC), which is a framework to investigate the cross-diagnostic mechanisms underlying psychiatric symptomatology. Cognitive systems (i.e., attention, perception, language, memory, and EF) is one of the five major RDoC domains, highlighting the importance of cognition in this approach.Reference Clark, Prior and Kinsella26 Research has already begun investigating neurocognitive phenotypes or homogenous clusters, primarily in adult psychiatric disorders (Table 1).
Table 1. Review of cluster analysis studies
ToM =; Theory of mind; EF =; Executive functioning; HC =; Healthy control; MDD =; Major depressive disorder; BP =; Bipolar disorder; Scz =; Schizophrenia; CANTAB =; Cambridge Neuropsychological Test Automated Battery; MATRICS =; Measurement and Treatment Research to Improve Cognition in Schizophrenia; VF: Verbal Fluency; NP: Neuropsychological; WM: Working Memory.
Adult
The cluster analysis approach has been successfully utilized in adults (13 studies) to detect neurocognitively homogenous subgroups in psychiatric disorders. Studies have examined neurocognitive clusters within specific disorders (i.e., schizophrenia, bipolar disorder, and gambling disorder)Reference Potter and Nestor27–Reference Mallorqui-Bague, Tolosa-Sola and Fernandez-Aranda31 or more commonly, across psychiatric disorders, such as examining across affective disorders,Reference Cotrena, Damiani Branco, Ponsoni, Milman Shansis and Paz Fonseca32 across psychotic disorders,Reference Lewandowski, Baker, McCarthy, Norris and Ongur33–Reference Lewandowski, Sperry, Cohen and Ongur35 and across the affect-psychosis spectrum.Reference Lee, Rizzo and Altshuler36–Reference Hermens, Redoblado Hodge, Naismith, Kaur, Scott and Hickie39 The majority of studies identified three to four neurocognitive clusters, including a neurocognitively intact cluster and a globally impaired cluster. In addition to intact/impaired clusters, most studies also identified one or two middle groups, described as either selective impairment, mixed profile, intermediate profile, or mild–moderately impaired groups, depending on the study. Some degree of variability in the affected domain was noted across studies due to the range of assessed neurocognitive domains, although visual memory deficits, psychomotor speed deficits, and executive deficits were specifically identified. Importantly, these distinct subgroups were associated with critical clinical/functional variables, in that generally the more impaired clusters were associated to worse outcomes/variables, such as symptomology,Reference Potter and Nestor27, Reference Mallorqui-Bague, Tolosa-Sola and Fernandez-Aranda31, Reference Cotrena, Damiani Branco, Ponsoni, Milman Shansis and Paz Fonseca32, Reference Lewandowski, Sperry, Cohen and Ongur35, Reference Reser, Allott, Killackey, Farhall and Cotton40–Reference Frias, Dickstein and Merranko43 psychiatric episodes,Reference Burdick, Russo and Frangou30, Reference Sauve, Malla, Joober, Brodeur and Lepage34 and response to clinical treatmentReference Gilbert, Merette and Jomphe28 as well as years of education,Reference Cotrena, Damiani Branco, Ponsoni, Milman Shansis and Paz Fonseca32, Reference Sauve, Malla, Joober, Brodeur and Lepage34 age,Reference Mallorqui-Bague, Tolosa-Sola and Fernandez-Aranda31, Reference Sauve, Malla, Joober, Brodeur and Lepage34 employment status,Reference Burdick, Russo and Frangou30, Reference Mallorqui-Bague, Tolosa-Sola and Fernandez-Aranda31 community functioning,Reference Lewandowski, Baker, McCarthy, Norris and Ongur33 and socioeconomic status.Reference Cotrena, Damiani Branco, Ponsoni, Milman Shansis and Paz Fonseca32
Youth–young adult
Four studies have examined neurocognitive clusters in adolescents and young adults (12–30 years)Reference Reser, Allott, Killackey, Farhall and Cotton40–Reference Uren, Cotton, Killackey, Saling and Allott42, Reference Crouse, Moustafa, Bogaty, Hickie and Hermens44 with early or first-episode psychosis. In these young people, three to four clusters were identified, including intact neurocognition, globally impaired neurocognition, and one to two intermediate groups, such as moderately impaired,Reference Uren, Cotton, Killackey, Saling and Allott42 mixed,Reference Crouse, Moustafa, Bogaty, Hickie and Hermens44 or select visual memory and EF deficit groups.Reference Reser, Allott, Killackey, Farhall and Cotton40 In psychiatrically hospitalized adolescents/young adults with affective disorders, two clusters were identified, including one characterized by attention/memory deficits and another characterized by switching deficits.Reference Tickell, Scott and Davenport41 Across studies, these subgroups were associated with estimates of premorbid intelligence,Reference Reser, Allott, Killackey, Farhall and Cotton40, Reference Uren, Cotton, Killackey, Saling and Allott42, Reference Crouse, Moustafa, Bogaty, Hickie and Hermens44 symptomatology,Reference Reser, Allott, Killackey, Farhall and Cotton40–Reference Uren, Cotton, Killackey, Saling and Allott42 years of education,Reference Reser, Allott, Killackey, Farhall and Cotton40 and overall sociooccupational functioning.Reference Uren, Cotton, Killackey, Saling and Allott42, Reference Crouse, Moustafa, Bogaty, Hickie and Hermens44 Six additional studies have been conducted in children and adolescents with ADHD,Reference Fair, Bathula, Nikolas and Nigg45, Reference van Hulst, de Zeeuw and Durston46 learning disorders,Reference Poletti, Carretta, Bonvicini and Giorgi-Rossi47 bipolar disorder,Reference Frias, Dickstein and Merranko43 anorexia nervosa,Reference Rose, Stedal and Reville48 and affective/psychotic disorders.Reference Kavanaugh, Dupont-Frechette, Tellock, Maher, Haisley and Holler49 Three to four clusters were identified in each study, and when most identified the previously described intact and globally impaired subgroups,Reference Frias, Dickstein and Merranko43, Reference Rose, Stedal and Reville48, Reference Kavanaugh, Dupont-Frechette, Tellock, Maher, Haisley and Holler49 there was some variability in the intermediate subgroups. The intermediate group was characterized by a moderate level of impairment in bipolar disorder,Reference Frias, Dickstein and Merranko43 a verbal/visual discrepancy in anorexia nervosa,Reference Rose, Stedal and Reville48 and by both an organization deficit group and memory/inhibition deficit group in hospitalized children with affective or psychotic disorders.Reference Kavanaugh, Dupont-Frechette, Tellock, Maher, Haisley and Holler49
Studies on ADHD and learning disorders obtained more specific findings secondary to a narrowed neurocognitive focus. In learning disorders, identified subgroups were characterized by low verbal functions, low processing speed, low EF, and low reasoning/EF subgroups.Reference Poletti, Carretta, Bonvicini and Giorgi-Rossi47 In one ADHD study, subgroups were characterized by attentional variability, low EF (two sub-subgroups), low processing speed, and low arousal (two sub-subgroups),Reference Fair, Bathula, Nikolas and Nigg45 whereas in another, subgroups included intact neurocognition, low cognitive control, and variable response timing.Reference van Hulst, de Zeeuw and Durston46 Only in bipolar, ADHD and affective/psychotic disorders were these subgroups associated with clinical/demographic variables, such as symptomatology, overall functioning, and medication status.Reference Frias, Dickstein and Merranko43, Reference van Hulst, de Zeeuw and Durston46, Reference Kavanaugh, Dupont-Frechette, Tellock, Maher, Haisley and Holler49 When collapsed into two subgroups in one study (i.e., nonmild impairment vs moderate–high impairment), additional differences were detected in number of diagnoses/medications, age, and length of hospital stay.Reference Kavanaugh, Dupont-Frechette, Tellock, Maher, Haisley and Holler49 As exemplified here, the variability in utilized measures resulted in greater degree of inconsistency in child/adolescent studies compared with young adult/adult studies.
Diagnostic Consideration
In clinical practice, knowing the patient’s specific psychiatric diagnosis provides very limited information on his/her neurocognitive status, as the patient may theoretically possess one of the three to four possible neurocognitive phenotypes of psychopathology. Even when the psychiatrist, psychologist, or other mental health provider has results from the clinical neuropsychological evaluation, the lack of neurocognitive diagnostic nosology leaves clinicians dependent on qualitative interpretations of neurocognitive strengths and weaknesses. The field is moving toward improved quantitative interpretation of evaluation, but qualitative differences between neuropsychologists remain.Reference Schoenberg, Osborn, Mahone, Feigon, Roth and Pliskin50 A unified nosology for neurocognitive status in psychopathology is needed for improved assessment and care of neurocognitive deficits, although how this classification will manifest is still unclear.
The development of a formal nosology for psychopathology-related neurocognitive impairment should focus on characterizing the neurocognitive status of the patient, independent of the specific psychiatric disorder(s). As described by Schoenberg et al.,Reference Schoenberg, Osborn, Mahone, Feigon, Roth and Pliskin50 non-neuropsychologists prefer to see neuropsychological results as one of the three categories to most clearly communicate important findings: abnormal (and related to brain dysfunction), normal, or equivocal (may be mild impairment or normal variability). Too many (e.g., mildly-to-moderately impaired, moderately impaired, etc.) or too few (i.e., abnormal or normal) categories are reportedly undesirable to referring clinicians.Reference Schoenberg, Osborn, Mahone, Feigon, Roth and Pliskin50 To build upon clinical classifier and cluster analysis work, the field may benefit from clinical care and research studies universally classifying each patient/participant into one of the three to four categories: 1) intact neurocognition, 2) globally or definitively impaired, 3) mixed, intermediate or mild impairment, and 4) equivocal as to whether findings reflect normal variability or mild/early impairment.
Arguing for a universally administered battery of neuropsychological tests is beyond the scope of this article. However, universal reporting of domain composite scores and an overall composite (regardless of the preferred clinical tests) would provide a wealth of information in clinical/research work. The global deficit score (GDS) by Heaton and colleagues utilizes a well-defined, 5-point scale to categorize performance from intact (0) to severe impairment (5)Reference Carey, Woods and Gonzalez51 for each test performance. This is then averaged to create the composite GDS score. A GDS cutoff score of ≥0.5 has been long utilized in HIV research to accurately detect neurocognitive impairment.Reference Carey, Woods and Gonzalez51 More recently, the GDS has been paired with measurement of activities of daily living (ADL) to accurately classify HIV affected adults with and without HIV-associated neurocognitive disorder, including HIV-associated dementia (i.e., GDS ≥ 1.5 with severe ADL decline), asymptomatic neurocognitive impairment (i.e., GDS ≥ 0.5 without ADL decline), and minor neurocognitive disorder (i.e., GDS ≥ 0.5 with ADL decline).Reference Kamminga, Bloch, Vincent, Carberry, Brew and Cysique52 The approach Heaton and colleagues have taken in their work with HIV populations is the exemplar of classifying neurocognitive impairment with neuropsychological rigor in a digestible and generalizable manner. Calculating the GDS for a given battery of tests and utilizing 0–0.5 to indicate intact, 0.5–1.5 to indicate mild/intermediate impairment, and 1.5+ to indicate global or definitive impairment could be a promising approach in work with psychiatric populations.
Continued work is needed to investigate the prevalence of neurocognitively homogenous syndromes, and their associated clinical, functional, and neurobiological characteristics. Eventually, neurocognitive status could be listed as a specifier with the diagnosis, broken down into the previously described, empirically based three or four categories (with specific deficits then listed). This could be similar to currently available neurocognitive symptom codes within ICD-10, for example, EF deficit (R41.844), psychomotor deficit (R41.843), visuospatial deficit (R41.842). Alternatively, within the current DSM-5, mild and major neurocognitive disorders are currently utilized to capture the neurocognitive deficits that occur in a host of medical/neurological disorders. Currently, major neurocognitive disorder requires deficits to be two standard deviations below the mean, whereas minor neurocognitive disorder requires deficits to be one to two standard deviations below the mean. At present, neurocognitive deficits secondary to psychiatric disorders are a rule-out in these neurocognitive disorders, and thus, this diagnostic entity cannot currently be applied to psychopathology. As such, inclusion of psychopathology as an etiological entity for neurocognitive disorder (in the same manner as current medical/neurological etiologies) could provide a comprehensive nosology that would improve identification and subsequent treatment of psychopathology-related neurocognitive deficits.
Treatment
It is important to obtain data on neurocognitive deficits in specific psychiatric disorders as well as their association to symptom severity. However, the field must continue to develop an alternative, parallel pathway that takes a (transdiagnostic) neurocognition-centric approach. This approach is consistent with the NIMH’s new call for examination of constructs in heterogeneous populations. For example, instead of studying a specific diagnosis (e.g., ADHD), investigations could study a specific neurocognitive deficit (e.g., response inhibition deficit) in a range of participants, from healthy controls to those with a range of psychiatric disorders. Furthermore, intervention studies would only recruit those with documented neurocognitive deficits in a specific domain and randomize participants to active/sham intervention conditions (and thus, only treat those in need of intervention). Reduced reliance on diagnosis may create initial challenges in recruitment/enrollment, as our clinical referral sources have patients with diagnoses, but do not as often have patients with identified neurocognitive deficits.
The lack of unified neurocognitive nosology to be used in conjunction with psychiatric disorder classification has limited the development and implementation of novel neurocognition-targeted interventions. As has been described above, not everyone with a given psychiatric disorder will have neurocognitive deficits, and therefore, not everyone with a psychiatric disorder will need or respond to a specific neurocognitive intervention. For example, repetitive transcranial magnetic stimulation (rTMS), which is an evidence-based, clinically available treatment for treatment-resistant depression, may have an effect on neurocognition. Its effects have been examined as a secondary variable, not as the target of an intervention, in prior research and has been limited and/or inconsistent in neuropsychiatric conditions.Reference Martin, McClintock, Forster and Loo53 However, when studies have more specifically targeted underlying brain activity (e.g., oscillations) and a clearly defined neurocognitive function (e.g., working memory), preliminary results have been promising. Studies have found that rTMS (compared to sham) at the dorsolateral prefrontal cortex has led to enhanced frontoparietal gamma and theta oscillations during working memory demands and subsequently improved working memory performance.Reference Hoy, Bailey and Michael54, Reference Barr, Farzan, Rusjan, Chen, Fitzgerald and Daskalakis55
For cognitive training (CT), clinical results in ADHD have found a select effect on neurocognition but not a corresponding improvement in clinical symptomatology.Reference Cortese, Ferrin and Brandeis56 However, in other work, thetaReference Anguera, Boccanfuso and Rintoul57 and gamma powerReference Dale, Brown and Fisher58 oscillations increased after CT within frontoparietal regions during cognitive demands.Reference Anguera, Boccanfuso and Rintoul57, Reference Dale, Brown and Fisher58 Such oscillatory changes were associated with enhanced cognitive control performance.Reference Dale, Brown and Fisher58 Consistent with the foundational principle that those with deficits will be the ones to respond to cognitive interventions,Reference Diamond and Ling59 one reason for limited outcomes of neurocognitive interventions (e.g., TMS/CT) could be due to the fact that many of the patients had an intact neurocognitive profile at baseline, with limited room to improve, that is, there was a ceiling effect.
Conclusions
Neurocognitive deficits are one of the most significant predictors of outcomes in child/adolescent and adult psychopathology. There is no consistent evidence of significant across disorder neurocognitive deficits, yet there is a significant degree of within-disorder variability in the rate of neurocognitive impairment. A novel approach to investigating psychopathology-related neurocognition is to identify the neurocognitively homogenous clusters or phenotypes. A wealth of adult research has consistently identified intact neurocognition, globally impaired neurocognition, and mixed/intermediate impairment phenotypes, with these phenotypes associated with distinct clinical/functional variables. Recent clinical work has similarly recommended classification into intact, impaired/abnormal, or equivocal labels. Importantly, no studies have identified a direct association between a specific diagnosis and a specific neurocognitive subgroup; neurocognitive subgroups often consist of individuals with a range of different diagnoses as well as healthy control participants. Thus, these subgroups are not merely a reflection of the already established psychiatric diagnostic classification, but alternatively, reflect their own neurocognitively homogenous subgroups of patients. Preliminary work in child/adolescent psychiatry has been consistent with adult findings, but further work is needed.
These findings have critical implications for neurocognitive assessment, research, and treatment. The field would benefit from more formalized terminology or classification of these within-disorder neurocognitive phenotypes. Universal utilization of indicators such as intact, impaired, mixed, or equivocal for clinical evaluations could move the field forward, whereas research could subsequently examine clinical and/or neurobiological characteristics of these subgroups. Future work might include providing a specifier in the psychiatric diagnosis. Finally, we must begin to develop and test possible neurocognition-targeted treatments in neurocognitively impaired phenotypes (such as the utilization of rTMS for EF) in order to improve long-term clinical and functional outcomes of these patients.
Funding
BCK’s effort on this manuscript was supported by the APA Division 40 (Early Career Award), Thrasher Research Fund (Early Career Award), and Rhode Island Foundation (Medical Research Grants).
Disclosures
The authors declare that they have nothing to disclose.
