Introduction
Elucidating the meta-structure of mental disorders – how psychopathology can be more parsimoniously organized into higher-order dimensions – remains ongoing. Proponents argue that meta-structure can help to resolve comorbidity and identify better phenotypic targets for studies of etiology, pathophysiology, or treatment response (e.g. Andrews et al., Reference Andrews, Goldberg, Krueger, Carpenter, Hyman, Sachdev and Pine2009; Jonas et al., Reference Jonas, Cannon, Docherty, Dwyer, Gur, Gur and Nelson2024; Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes and Watson2020, Reference Kotov, Carpenter, Cicero, Correll, Martin, Young and Jonas2024; Krueger et al., Reference Krueger, Hobbs, Conway, Dick, Dretsch and Eaton2021; Watson et al., Reference Watson, Levin-Aspenson, Waszczuk, Conway, Dalgleish, Dretsch and Krueger2022) and may match subjective experience (Ringwald et al., Reference Ringwald, Norel, Bauer, Cecchi and Kotov2025), but others question evidence supporting this utility (e.g. First, Reference First2009; Tyrer, Reference Tyrer2018; Wittchen, Beesdo, & Gloster, Reference Wittchen, Beesdo and Gloster2009; Zimmerman, Reference Zimmerman2021) or propose other approaches that may provide more benefit (e.g. Cuthbert, Reference Cuthbert2022; Eaton et al., Reference Eaton, Bringmann, Elmer, Fried, Forbes, Greene, Krueger, Kotov, McGorry, Mei and Waszczuk2023; McGorry et al., Reference McGorry, Hickie, Kotov, Schmaal, Wood, Allan and Nelson2025; Robinaugh, Hoekstra, Toner, & Borsboom, Reference Robinaugh, Hoekstra, Toner and Borsboom2020). Despite the debate, contours of a higher-order meta-structure have been articulated and some of its major dimensions replicated (Kotov et al., Reference Kotov, Cicero, Conway, DeYoung, Dombrovski, Eaton and Wright2022). Even with active work in this area, however, the placement of mania – the defining feature of bipolar I disorder – in the meta-structure of psychopathology remains unresolved.
Bipolar I disorder has similarities with both depressive and psychotic disorders in regard to etiology, biomarkers, and treatment response (Goldberg, Andrews, & Hobbs, Reference Goldberg, Andrews and Hobbs2009; Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes and Watson2020; Watson et al., Reference Watson, Levin-Aspenson, Waszczuk, Conway, Dalgleish, Dretsch and Krueger2022) and could reasonably be placed in either chapter. In the last revision of the Diagnostic and Statistical Manual (DSM-5), bipolar disorders were given their own chapter, between that of depressive disorders and psychotic disorders, ‘in recognition of their place as a bridge’ between those two classes (Regier, Kuhl, & Kupfer, Reference Regier, Kuhl and Kupfer2013). Quantitative models of psychopathology have also struggled with mania’s placement. For example, the Hierarchical Taxonomy of Psychopathology (HiTOP; Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017) organizes symptoms hierarchically from narrow traits to at least six broader spectra (i.e. internalizing [emotional distress, fear], disinhibited externalizing [distractibility, risk taking], antagonistic externalizing [callousness, entitlement], thought disorder [unusual beliefs, perceptual disturbances], detachment [social withdrawal, inexpressivity], and somatoform [physical symptom preoccupation]). Mania’s placement in this structure, however, has been unresolved and so remains provisional on both internalizing and thought disorder. Table 1 summarizes structural analyses that included indicators of both internalizing and thought disorder dimensions, and shows that mania has been variously located on the internalizing dimension (Wolf et al., Reference Wolf, Schubert, Patterson, Grande, Brocco and Pendleton1988; Wright et al., Reference Wright, Krueger, Hobbs, Markon, Eaton and Slade2013), the thought disorder dimension (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Moffitt2014; Forbes et al., Reference Forbes, Sunderland, Rapee, Batterham, Calear, Carragher and Krueger2021; Kotov et al., Reference Kotov, Ruggero, Krueger, Watson, Yuan and Zimmerman2011; Levin-Aspenson, Khoo, & Kotelnikova, Reference Levin-Aspenson, Khoo and Kotelnikova2019), both simultaneously (Keyes et al., Reference Keyes, Eaton, Krueger, Skodol, Wall, Grant and Hasin2013), or on a separate factor entirely (Krabbendam et al., Reference Krabbendam, Myin-Germeys, Graaf, Vollebergh, Nolen, Iedema and Van Os2004; Stanton et al., Reference Stanton, Mostajabi, Myntti, Willis, Yun, Osborne and Brewer2023).
Structural studies of mania
a Past-year symptoms were assessed at multiple points across early adulthood.
The mixed literature summarized above has left unclear the placement of mania in the meta-structure of psychopathology and speaks to the need for new methods to address the question (Jonas et al., Reference Jonas, Cannon, Docherty, Dwyer, Gur, Gur and Nelson2024; Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017, Reference Kotov, Cicero, Conway, DeYoung, Dombrovski, Eaton and Wright2022). Importantly, evidence, to date, has been based almost entirely on cross-sectional studies, with information about the course of mania or past episodes absent from these models. This represents a major limitation because mania is distinguished from both depression and psychosis by its tendency to follow an episodic course often with complete symptom remission between episodes (Gignac, McGirr, Lam, & Yatham, Reference Gignac, McGirr, Lam and Yatham2015;Judd et al., Reference Judd, Akiskal, Schettler, Endicott, Maser, Solomon and Keller2002; Solomon et al., Reference Solomon, Leon, Coryell, Endicott, Li, Fiedorowicz and Keller2010). Among those with bipolar I disorder, symptoms of mania are present for only 2%–11% of the time (Judd et al., Reference Judd, Akiskal, Schettler, Endicott, Maser, Solomon and Keller2002; Solomon et al., Reference Solomon, Leon, Coryell, Endicott, Li, Fiedorowicz and Keller2010). By comparison, depression and psychosis-related conditions tend to be more stable over time (e.g. Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes and Watson2020). For example, psychosis-related traits demonstrate remarkable 10-year stability (Hopwood et al., Reference Hopwood, Morey and Donnellan2013), and schizophrenia remission and recovery rates over time are lower than those observed following a manic episode (e.g. AlAqeel & Margolese, Reference AlAqeel and Margolese2012; Gignac, McGirr, Lam, & Yatham, Reference Gignac, McGirr, Lam and Yatham2015; Jaaskelainen et al., Reference Jaaskelainen, Juola and Hirvonen2013). Depression and other internalizing conditions showed moderate to strong continuity across years of assessment (Lahey et al., Reference Lahey, Zald and Hakes2014), and evidence from studies looking at the stability of underlying internalizing conditions related to depression and related disorders showed considerable stability in youth and middle age (Fergusson, Horwood, & Boden, Reference Fergusson, Horwood and Boden2006; Gustavson et al., Reference Gustavson, Franz and Panizzon2020). Yet, despite differences in stability across time, virtually all structural studies, to date, are based on symptoms present only at the time of assessment, so the resulting structures may reflect persistent internalizing and thought disorder symptoms, rather than symptoms of mania that have remitted. Assessing the longitudinal course of symptoms may be essential for disentangling mania from more stable forms of psychopathology.
Moreover, very few studies of mania’s placement in the meta-structure take the step of validating their final disposition. In other words, they consider only the co-occurrence of signs and symptoms using structural evidence, but not whether the structure aligns with other validators of pathophysiology, such as biomarkers or family history (cf. Forbes et al., Reference Forbes, Ringwald, Allen, Cicero, Clark, DeYoung and Wright2024). The one exception found was that thought disorder, which included mania, was associated with a greater family history of depression and neuropsychological impairment (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Moffitt2014). In considering potential validators of any final structural model, one can draw upon a robust literature showing that bipolar disorder is distinguished from other serious mental illness by elevation on bipolar polygenic risk score (PRS), good premorbid adjustment, treatment with mood stabilizers, and relatively intact cognitive functioning (Carlson et al., Reference Carlson, Kotov, Chang, Ruggero and Bromet2012; Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes and Watson2020; Watson et al., Reference Watson, Levin-Aspenson, Waszczuk, Conway, Dalgleish, Dretsch and Krueger2022). If mania is correctly situated within the meta-structure, the dimension on which it falls should be associated with these markers.
The aim of the present study was to evaluate how data on the course of mania – not just current symptoms – would affect the placement of mania in the meta-structure of psychopathology and whether the resulting model aligned with classic validators of bipolar disorder. Current symptoms of internalizing, thought disorder, and mania were assessed by interviewer ratings as part of a 20-year longitudinal study. Additionally, the course of core symptoms related to these three constructs (i.e. history of depressive, manic, and psychotic episodes) was also ascertained during interviews conducted across the 20 years. Based on prior research, it was hypothesized that current and past markers of depression and psychosis would cohere and load on separate internalizing and thought disorder dimensions, respectively (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017). We then tested whether mania best fit with internalizing, thought disorder, or its own independent spectrum, and whether adding information from course (i.e. past episodes) altered this placement. Finally, we validated the best resulting structure against known correlates of bipolar disorder, including genetic risk, good premorbid adjustment, treatment with mood stabilizers, and smaller neuropsychological impairments relative to other psychotic disorders (Carlson et al., Reference Carlson, Kotov, Chang, Ruggero and Bromet2012; Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes and Watson2020; Watson et al., Reference Watson, Levin-Aspenson, Waszczuk, Conway, Dalgleish, Dretsch and Krueger2022).
Methods
Sample
Data were drawn from the Suffolk County Mental Health Project, a longitudinal study of first-admission psychosis. Participants were recruited from all 12 inpatient psychiatric hospitals in Suffolk County, New York (response rate 72%), between 1989 and 1995 (Bromet et al., Reference Bromet, Schwartz, Fennig, Geller, Jandorf, Kovasznay, Lavelle, Miller, Pato, Ram and Rich1992). Eligibility criteria were residence in Suffolk County, first admission within the past 6 months, no apparent organic etiology for psychosis, ability to speak English, IQ >70, and age between 15 and 60 years. Written consent was obtained from all study participants, or from parents in the case of minors. The Stony Brook University Committee on Research Involving Human Subjects and the review boards of participating hospitals approved the protocol annually.
During the baseline wave, 628 participants met the inclusion criteria. Participants were reassessed 6 months, 24 months, 48 months, 10 years, and 20 years after first admission. As of the 20-year follow-up, 81 participants had died. Of the 547 surviving participants, 70 were lost to follow-up, 33 could not be reached for varying reasons (moved to another country, institutionalized, or too ill to consent), and 71 declined to be interviewed. The remaining 373 participated in the 20-year follow-up in some form, but only 337 of these cases completed the full 20-year measures and are the focus of the present analyses. Research diagnoses were made by the consensus of study psychiatrists at year 20 using all available longitudinal information, including results of the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) (Spitzer, Williams, Gibbon, & First, Reference Spitzer, Williams, Gibbon and First1992) interviews with participants’ significant others, medical records, and observations and behavioral ratings by master’s-level interviewers. Diagnoses were made according to DSM-IV criteria, as previously described (Bromet et al., Reference Bromet, Kotov, Fochtmann, Carlson, Tanenberg-Karant, Ruggero and Chang2011). Table 2 reports the demographic characteristics of the sample, as well as diagnoses and descriptive statistics for symptoms. An analysis of attrition compared cases who participated in the 20-year assessment to surviving participants who did not. Participants were more likely to be Caucasian (81% vs. 69%) and have bipolar disorder (34% vs. 21%), but did not differ from non-participants in gender, age at psychosis onset, or age at the time of assessment (Supplementary Table 1).
Sample descriptive statistics
Note: 10Y, 10 years; 20Y, 20 years; 4Y, 4 years; 6mo, 6 months; BL, baseline; BP, bipolar disorder; Dep. Epi., Depressive Episode; HDRS, Hamilton Depression Rating Scale; Man. Epi., Mania Episode; MDD, major depression; Other Psy, other psychotic disorders; Psyc. Epi., Psychosis Episode; SA, schizoaffective disorder; SAPS, Scale for the Assessment of Positive Symptoms; SCID, Structured Clinical Interview for DSM-IV Axis I Disorders; SIP, substance-induced psychosis; SZ, schizophrenia; YMRS, Young Mania Rating Scale.
Measures
Current symptoms
Master’s-level mental health professionals made ratings of symptoms based on their interviews with the participant, interviews with significant others, and medical records. Symptoms were rated using the SCID (First & Gibbon, Reference First, Gibbon, Hilsenroth and Segalaniel2004), the Scale for the Assessment of Positive Symptoms (SAPS; Andreasen, Reference Andreasen1984), the Hamilton Depression Rating Scale (Williams, Reference Williams1988), and the Young Mania Rating Scale (Young, Biggs, Ziegler, & Meyer, Reference Young, Biggs, Ziegler and Meyer1978).
Internalizing/depressive symptoms were captured from two composite scales. The first was derived from nine current symptoms of depression assessed by the SCID (scores range 9–27), administered without skip-outs, and the second was a composite of the 21 symptoms assessed by the Hamilton Depression Rating Scale (range 0–63) (Kotov, Guey, Bromet, & Schwartz, Reference Kotov, Guey, Bromet and Schwartz2010; Williams, Reference Williams1988). Thought disorder symptoms were measured with the reality distortion and disorganization scales of SAPS developed previously (Kotov et al., Reference Kotov, Foti, Li, Bromet, Hajcak and Ruggero2016). Mania symptoms were captured through three subscales – activation, cognition, and agitation – derived from an exploratory factor analysis of the YMRS. The factor loading matrix is reported in Supplementary Table 2. All scales for current symptoms had acceptable internal reliability (i.e. Cronbach’s alpha ≥ .67).
Course of symptoms
A timeline of manic, depressive, and psychotic episodes was generated at three time points across the 20 years (i.e. 4-year, 10-year, or 20-year assessment). Assessments were conducted 6 months, 24 months, 48 months, 10 years, and 20 years after the baseline. These timelines were created based on data from clinical interviews, medical records, and interviews with significant others, with coders coding whether episodes had occurred in the preceding interval. Manic and depressive episodes were defined according to DSM-IV criteria for each, respectively. Psychotic episodes were defined as the presence of any threshold psychotic symptom assessed by SCID Module B. Psychosis was counted from month 6 on, but not before 6 months, as nearly everyone was in a psychotic episode at baseline. The episode timelines were coded dichotomously to indicate whether a manic, depressive, or psychotic episode occurred during the three intervals: early (baseline/6-month to 4-year), middle (4-year to 10-year), and late (10-year to 20-year). Frequencies and descriptive statistics for all symptom marker variables are reported in Table 2.
Validators
Family history of bipolar disorder and family history of depression
Family history was assessed based on interviews with family members, the participant’s report, and medical records. Both were operationalized as dichotomous variables, rated 1 if any of the participant’s first-degree relatives had been diagnosed with bipolar disorder or depression, and 0 otherwise.
Polygenic risk scores
DNA was collected from 249 participants as part of the Genomic Psychiatry Cohort (Pato et al., Reference Pato, Sobell, Medeiros, Abbott, Skar, Buckley and Pato2013). DNA was extracted from peripheral lymphocytes and genotyped using the Illumina PsychArray-8 platform containing 571,054 markers. For details of quality control, see Jonas et al. (Reference Jonas, Lencz, Li, Malhotra, Perlman, Fochtmann and Kotov2019).
Bipolar and schizophrenia PRSs were calculated for each participant based on the summary statistics from the most recent Psychiatric Genomics Consortium (PGC) genome-wide association study (GWAS) results (available at http://ldsc.broadinstitute.org/ldhub/; Mullins et al., Reference Mullins, Forstner, O’Connell, Coombes, Coleman, Qiao and Andreassen2021; Trubetskoy et al., Reference Trubetskoy, Pardiñas, Qi, Panagiotaropoulou, Awasthi, Bigdeli and O’Donovan2022). single nucleotide polymorphisms (SNPs) were clumped to a more significant SNP if they were in linkage disequilibrium (LD) (r 2 > 0.10) within a 500 kb window. PRS calculation was carried out in PRSice (Eusden, Lewis, & O’Reilly, Reference Eusden, Lewis and O’Reilly2015). PRS used all available SNPs and weights a priori to minimize multiple testing (i.e. a p-value threshold of 1). Results at other thresholds were similar. Analyses of PRS were completed in the subset of European ancestry participants (N = 201). European ancestry was defined as those participants determined to be ≥80% European ancestry according to ADMIXTURE analyses (Alexander, Novembre, & Lange, Reference Alexander, Novembre and Lange2009), and within three standard deviations of the mean for the first three principal components of population stratification (Privé et al., Reference Privé, Luu, Blum, McGrath and Vilhjálmsson2020). All analyses were covaried on the first 10 principal components of population stratification in order to control for confounding due to ancestry (Price et al., Reference Price, Patterson, Plenge, Weinblatt, Shadick and Reich2006).
Premorbid adjustment
Premorbid psychosocial functioning was assessed using an interview based on the Premorbid Adjustment Scale (PAS; Cannon-Spoor, Potkin, & Wyatt, Reference Cannon-Spoor, Potkin and Wyatt1982; Rabinowitz, Levine, Brill, & Bromet, Reference Rabinowitz, Levine, Brill and Bromet2007), which was administered to participants and significant others at the 6-month follow-up. PAS ratings were based on these interviews and school records. Rating periods covered childhood (up to age 11), early adolescence (ages 12–15), and late adolescence (ages 16–18). The last available PAS composite prior to psychosis onset was used. The PAS consists of five domains (sociability and withdrawal, peer relationships, scholastic performance, adaptation to school, and social–sexual relationships) rated on a 7-point scale (scores range from 0 to 6, with 0 representing good functioning). Composite scores were calculated as the mean of the five PAS items, with scores reversed so that higher numbers indicate better functioning.
Medication history
Medication data were documented based on self-report, pill bottles and medication lists brought to follow-up appointments, and medical records. Lifetime history of treatment with mood stabilizers was operationalized as the sum of intervals during which the participant took this class of medication, weighted by the length of the interval.
Neuropsychological performance
Cognitive tests performed at the 20-year follow-up included performance on the word reading subtest of the WRAT-3 (a proxy for premorbid academic achievement; Wilkinson, Reference Wilkinson1993), Trails B (Reitan, Reference Reitan1955), and the Controlled Oral Word Association Test (Bechtoldt, Benton, & Fogel, Reference Bechtoldt, Benton and Fogel1962). These tests were selected from a larger neuropsychological battery because they distinguish bipolar disorder from first-episode schizophrenia (Bora & Pantelis, Reference Bora and Pantelis2015).
Supplementary Table 3 reports the descriptive statistics for all validators.
Analyses
Code for all analyses is available from the corresponding author on request. Parallel analysis was performed in R using the ‘paran’ package (Dinno, Reference Dinno2018; R Core Team, 2021). All latent variable models were estimated in Mplus, version 7 (Muthen & Muthen, Reference Muthén and Muthén2007), using maximum likelihood estimation with robust standard errors. This estimator uses all available data, including partial cases, and is robust to non-normality of the indicators. Model comparisons were evaluated based on the Bayesian Information Criterion, as well as the interpretability of factors (Clark & Watson, Reference Clark and Watson1995). Latent variable models of symptoms were estimated using confirmatory factor analysis. Validators were regressed on the latent factors in structural equation models.
Results
Mania in the meta-structure
The first model (Model A or the internalizing model) specified that all mania markers (i.e. current symptoms and past episodes) load with the internalizing dimension. The second model (Model B or the thought disorder model) specified that all mania markers load with the thought disorder dimension. The third model (Model C or independent mania model) specified a third independent dimension separate from internalizing and thought disorder factors and composed of both current and past episode mania markers.
We next tested two split models based on the hypothesis that current and past markers of mania would capture different information. Therefore, a separate mania dimension composed of only past markers would form a factor split off from current symptoms. One model had current symptoms load onto internalizing (Model D1), and the other had current symptoms load onto thought disorder (Model D2).
Table 3 reports the results of these five models, and Supplementary Table 4 reports their fit indices. The internalizing model (Model A) had the worst fit, with current mania markers loading only weakly and past mania loading not at all or negatively onto the internalizing dimension. The thought disorder model (Model B) had improved fit over Model A. Current mania markers had loadings on this dimension (i.e. all loadings >.35), but past mania episodes failed to load onto this factor (i.e. all loadings <.03).
Model factor loadings and correlations
Note: Scales assessing mania are bolded. 10Y, 10 years; 20Y, 20 years; 4Y, 4 years; 6mo, 6 months; BL, baseline; Dep. Epi., Depressive Episode; HDRS, Hamilton Depression Rating Scale; INT, internalizing; MAN HX, history of mania; MAN, mania; Man. Epi., Mania Episode; Psyc. Epi., Psychosis Episode; SAPS, Scale for the Assessment of Positive Symptoms; SCID, Structured Clinical Interview for DSM-IV Axis I Disorders; TD, thought disorder; YMRS, Young Mania Rating Scale.
The independent mania model (Model C) showed the best relative fit of all models. However, this model had two major limitations that led to its rejection. First, there was an excessively high overlap between the thought disorder dimension and the independent mania dimension (i.e. r = .89). This extremely high correlation indicated that these two factors are not distinct. Second, past mania markers had negative loadings onto this dimension, suggesting that the course of mania and current mania do not fit on the same factor.
For both reasons, Models D1 and D2 (the split model) were run. Here, current mania symptom markers were allowed to load with either the internalizing (D1) or the thought disorder (D2) dimension, but past mania markers formed an independent mania history factor.
Model D2 (current mania loading onto Thought Disorder) showed a better fit than D1 (mania on Internalizing) and was selected as the final model. It also had, by far, the lowest chi-square value of all the models analyzed (Supplementary Table 4). Most importantly, this solution resolved the limitations of Model C: factor dimensions were all largely independent in Model D2, with good discrimination from one another (i.e. r’s ≤ .20), and all mania markers loaded onto their respective factors (i.e. all loadings >.35). Moreover, Model D2 showed superior fit to Models A and B. Figure 1 displays Model D2.
Validation
Validators were regressed on the three latent factors simultaneously of the final model selected (i.e. Model D2 or the split model). The results of these analyses are reported in Table 4 and supported the selection of Model D2 as the final model. Course markers reflecting a past history of mania factor showed the most robust association with mania-related validators. This factor was associated with family history of bipolar disorder, family history of depression, and higher bipolar PRSs. It was also associated with better premorbid adjustment and better cognitive function. This factor was also associated with mood stabilizer treatment more strongly than the other factors.
Regression of bipolar disorder validators on the final model’s three latent factors
Note: Statistically significant associations are bolded. BP PRS, bipolar polygenic risk score; COWAT, Controlled Oral Word Association Test.
a Premorbid adjustment and Trails B scores were reversed such that higher scores indicate better functioning.
Internalizing was associated with a greater family history of bipolar disorder and depression, but not other mania-related validators. Thought disorder was linked to greater genetic liability for schizophrenia, worse premorbid function, worse cognitive performance (i.e. higher scores on Trials B), and less treatment with mood stabilizers. This dimension included current mania symptoms, but its external correlates were largely opposite of what’s expected for mania. Overall, validation largely supported the final model selection that had past mania markers independent from thought disorder and internalizing factors.
Discussion
In a study of 337 individuals with psychotic disorders, current and past symptoms of mania diverged in their placement in the meta-structure of psychopathology. Current symptoms aligned with the thought spectrum, but markers of past mania formed their own, independent spectrum. Most critically, this dimension reflecting course (i.e. past episodes of mania) was the only one closely related to known correlates of bipolar I disorder – such as genetic risk, good premorbid functioning, and treatment with mood stabilizers. Three overarching conclusions may be drawn from this study. First, cross-sectional assessment of symptoms represents a major limitation of structural research on mania’s placement in quantitative nosology. This finding is not surprising, given that cross-sectional assessment methods are more likely to miss clinical phenomena like mania that are better identified through longitudinal assessment. Yet it is nevertheless important for models like HiTOP to keep in mind, given that they may misrepresent the meta-structure whenever they rely solely on cross-sectional evidence. More generally, episodicity may represent an independent feature for some forms of psychopathology, and so may need to be explicitly integrated into structural models. Second, evidence from this study suggests that mania’s placement in the meta-structure of psychopathology is in its own spectrum, independent from either the internalizing or thought disorder dimensions. Third, current symptoms of mania per se were not validated in the present study as related to key markers of bipolar disorder and instead captured the thought disorder spectrum.
Foremost, results from this study underscore how illness course is an essential feature of mania that may be independent from current symptoms. Our findings are consistent with the long-standing view that assessing illness course is critical to the assessment of bipolar disorder (Frank, Nimgaonkar, Phillips, & Kupfer, Reference Frank, Nimgaonkar, Phillips and Kupfer2015) and align with a long-standing validator for psychiatric disorders within traditional nosology (Kraepelin, Reference Kraepelin and Barclay1899; Robins & Guzé, Reference Robins and Guzé1970). Illness course has been missing from most past structural work attempting to adjudicate mania’s placement in the meta-structure of psychopathology, and may be one reason for conflicting findings on whether mania is better conceptualized as part of internalizing, thought disorder, or its own dimension (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017, Reference Kotov, Cicero, Conway, DeYoung, Dombrovski, Eaton and Wright2022). Our results underscore that better incorporation of course features may be necessary for quantitative models of nosology, such as HiTOP. The extent of this need beyond mania remains an open question since course is not likely to be universally critical for placing psychopathology within a meta-structure. For example, internalizing, thought disorder, and externalizing dimensions exhibit high rank-order stability over many years (Eaton, Krueger, & Oltmanns, Reference Eaton, Krueger and Oltmanns2011; Jonas et al., Reference Jonas, Cannon, Docherty, Dwyer, Gur, Gur and Nelson2024; Krueger, Caspi, Moffitt, & Silva, Reference Krueger, Caspi, Moffitt and Silva1998) and so are less likely to show divergence between current and past symptoms. Indeed, in the present study, current and past symptoms were placed on the same spectrum for internalizing and thought disorder. Mania likely showed divergence because it is considerably less stable over time.
Beyond implications with respect to course, the present study found that current symptoms of mania not only failed to load with past episodes, but also loaded onto a factor that was not associated with traditional validators of bipolar disorder. For example, the factor they loaded onto was much more strongly related to genetic risk for schizophrenia than for bipolar disorder. Part of this finding may be related to measurement. Our results are consistent with previous research findings that interviewer-rated mania is often linked to thought disorder (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Moffitt2014; Kotov et al., Reference Kotov, Ruggero, Krueger, Watson, Yuan and Zimmerman2011). However, an important factor here is that symptoms of mania overlap with symptoms of schizophrenia, so that differential diagnosis is often based on course and outcome rather than presenting symptoms. Notably, genetic risk for bipolar disorder has been associated with the number of hospitalizations over the course of illness (Kalman et al., Reference Kalman, Papiol, Grigoroiu-Serbanescu, Adorjan, Anderson-Schmidt, Brosch and Andlauer2022), but not cross-sectional phenotypes reflecting clinical profiles (Dwyer et al., Reference Dwyer, Kalman, Budde, Kambeitz, Ruef, Antonucci and Koutsouleris2020) nor the distinction between affective and non-affective psychosis (Rodriguez et al., Reference Rodriguez, Alameda, Quattrone, Tripoli, Gayer-Anderson, Spinazzola and Vassos2022). In short, when it comes to mania, the content of symptoms may be less important than their course over time.
Research and clinical relevance of present findings, and meta-structure broadly, requires further investigation (e.g. Tyrer, Reference Tyrer2018; Zimmerman, Reference Zimmerman2021). However, if replicated, findings carry potential implications. With regard to research, dimensional studies of psychotic disorders have focused on symptom severity and may overlook heterogeneity related to characteristics of illness course. Indeed, patients with the same psychotic symptom profiles may either remit quickly or suffer from a protracted illness that does not fully resolve. Adding the mania history factor to the study design can reduce this heterogeneity, allowing studies to disentangle the effects of chronicity and symptom severity. This is relevant for research ranging from etiology to treatment response. For example, we found that mania history and mania severity (thought disorder) dimensions have very different – and sometimes opposite – associations with genetic vulnerabilities, cognitive functioning, and utilization of mood stabilizers. With regard to clinical relevance, longitudinal course has long been recognized as essential for the identification of bipolar disorder in traditional nosologies, and the present study extends this to suggest that quantitative classifications likewise need to capture mania history when conducting assessments. With regard to prognosis, mania history factor may be informative, as given the same severity of thought disorder symptoms, patients elevated on this history factor may be more likely to follow an episodic course, whereas patients without this elevation may be more chronic. We do not have data to test this possibility directly, and further research is needed to evaluate it.
Despite the importance of our findings for structural models of psychopathology and its possible research and clinical relevance, important limitations need to be acknowledged. People with bipolar I disorder who were never hospitalized were not included in this study, nor were those who did not experience at least some psychotic symptoms. An estimated 63% of manic episodes require hospitalization (de Zelicourt et al., Reference de Zelicourt, Dardennes, Verdoux, Gandhi, Khoshnood, Chomette and Fagnani2003), and psychosis is observed in 58% (Goodwin & Jamison, Reference Goodwin and Jamison2007), suggesting that our sample is reasonably representative of bipolar I disorder. Nevertheless, the sample reflects more severe cases, so it will be important to replicate these findings among individuals who have not been hospitalized or experienced psychosis. Moreover, the present study was confined only to understanding mania’s placement relative to internalizing and thought disorder spectra. This focus was motivated by the existing literature, as previous structural studies placed mania only in these two spectra. It will nevertheless be necessary and useful to examine mania in the context of all six HiTOP spectra, for example. The study also still mostly relied on cross-sectional assessments (albeit done repeatedly over time) and did not have a robust set of validators, such as a functional recovery or quality of life. As such, results should be considered preliminary until a prospective design with more robust validators can be used to validate this structure. In addition, assessment relied on a specific clinical interview, and although this measure is an established standard in bipolar disorder research (Young, Biggs, Ziegler, & Meyer, Reference Young, Biggs, Ziegler and Meyer1978), it will be important to confirm results with broader instruments. Additionally, this study was limited by attrition over follow-ups, and, like other studies of this kind, the point prevalence of mania was low. Both weaknesses were ameliorated by the lengthy follow-up and thorough phenotyping, which allowed for a rigorous assessment of premorbid characteristics, course, and treatment history. Finally, the overarching aim of the study was to understand mania’s placement in the meta-structure of psychopathology; however, validating this overarching structure and its clinical utility will require considerably more work, with no definitive evidence yet that this will lead to better care for individual patients (cf. First, Reference First2009; Tyrer, Reference Tyrer2018; Wittchen, Beesdo, & Gloster, Reference Wittchen, Beesdo and Gloster2009; Zimmerman, Reference Zimmerman2021).
Conclusions
While current symptoms of mania overlapped with thought disorder psychopathology in the present study, the course of mania formed a distinct spectrum, which was not the case for past episodes related to thought disorder or internalizing. The dimension based on course was related to known correlates of bipolar disorder, including family history, genetic risk, and relatively preserved cognitive function, supporting the importance of distinguishing the course of mania from thought disorder and internalizing. If replicated, these results would support the addition of an independent mania spectrum to the HiTOP model. Incorporating course features is a necessary step for refining models of psychopathology.
Supplementary material
The supplementary material for this article can be http://doi.org/10.1017/S0033291725101712.
Data availability statement
Genotype data are available through dbGap (Accession no. phs001020.v2.p1). The phenotypic data that support the findings of this study are available from the corresponding author upon request.
Funding statement
This research was supported by NIMH grant R01 MH044801 awarded to E.B. and grant R01 MH110434 awarded to R.K.
Competing interests
All authors, except E.B. and R.K., declare that they have no competing interests.
Open access
