Heritability, phenotypic, and genetic correlations across dimensional and categorical models of bipolar disorder in a family sample

Abstract

Background:

Bipolar disorder (BD) presents with a wide range of symptoms that vary among relatives, casting doubt on categorical illness models. To address this uncertainty, we investigated the heritability and genetic relationships between categorical and dimensional models of BD in a family sample.

Methods:

This retrospective study included participants (n = 397 Females, n = 329 Males, mean age 47 yr) in the Amish-Mennonite Bipolar Genetics (AMBiGen) study from North and South America that were assigned categorical mood disorder diagnoses (“narrow” or “broad”) by structured psychiatric interview and completed the Mood Disorder Questionnaire (MDQ), which assesses lifetime history of manic symptoms and associated impairment. MDQ-dimensions were analyzed by Principal Component Analysis (PCA). Heritability and genetic overlaps between categorical diagnoses and MDQ-dimensions were estimated with SOLAR-ECLIPSE within 432 genotyped participants.

Results:

Individuals diagnosed with BD (n = 124) endorsed more MDQ items (61 %) than those with other mood disorders (26 %) or with no mood disorder (9 %), as expected. PCA suggested a three-component model for the MDQ, capturing 60 % of the variance. Heritability of the MDQ and its principal components was significant but modest (20–30 %, p < 0.001). Genetic correlations between MDQ measures and categorical diagnoses (ρG = 0.62–1.0; p < 0.001) were stronger than phenotypic correlations (ρP = 0.11–0.58; p < 0.001).

Limitations:

Recruitment through probands with BD resulted in increased prevalence of BD in this sample, limiting generalizability. Unavailable genetic data reduced sample size for some analyses.

Conclusion:

Findings support a genetic continuity between dimensional and categorical models of BD and suggest that the MDQ is a useful phenotype measure for genetic studies of BD.

1. Introduction

Bipolar disorder (BD) is a complex psychiatric illness characterized by significant fluctuations in mood and behavior that impair daily functioning and elevate the risk of suicide (Serra et al., 2022). Among adults in the United States of America (USA), 2.4–4.4 % are diagnosed with BD at some point in their lives (Merikangas et al., 2007; Nierenberg et al., 2023). BD often presents with a depressive episode that can be indistinguishable from unipolar depression, leading to misdiagnosis or delayed diagnosis for several years (Carta and Angst, 2016; Goes, 2023).

The current categorical classification of BD, while reliable, does not correspond well to the known neurobiology of mood, the growing body of established genetic risk factors, or the broad range of bipolar symptom manifestations (Gordovez and McMahon, 2020). For example, many individuals, including relatives of BD patients, suffer from subclinical symptoms that do not meet categorical diagnostic criteria, but still impact quality of life (Axelson et al., 2011; Woodward et al., 2023; Stanislaus et al., 2020).

Thus, there is a growing interest in dimensional measures of psychopathology that may better capture the full spectrum of BD and its neurobiological and genetic underpinnings (Bruce et al., 2019; Casey et al., 2013; Merikangas et al., 2011; Parker et al., 2016; Hosang et al., 2022; McCoy Jr and Perlis, 2024). The Mood Disorder Questionnaire (MDQ) is a commonly used dimensional measure to screen for BD and related mood disorders. As a self-report measure, it evaluates 13 manic symptoms, their co-occurrence during distinct episodes, and the extent to which these symptoms disrupt behavior (Hirschfeld et al., 2000). The MDQ has been shown to detect BD with sensitivity ranging from 0.60 to 0.71 and specificity ranging from 0.77 to 0.97 (Wang et al., 2015; Dumont et al., 2020; Sayyah et al., 2022). Several studies have shown that the MDQ is an effective screening tool for BD in both urban and rural clinical settings (Hardoy et al., 2005; Rouillon et al., 2011; Twiss et al., 2008; Dumont et al., 2020).

While dimensional measures may provide an alternative approach to the assessment of BD, it is not clear how these measures relate to BD as a distinct disease entity. Few studies have examined the familial or genetic relationships between dimensional symptom measures and BD diagnoses (Bruce et al., 2019; Hosang et al., 2022; Mundy et al., 2023). The purpose of the present study is to characterize the heritability of MDQ measures and determine their phenotypic and genetic overlap with categorical BD diagnoses in families ascertained through a proband with BD. We hypothesized that categorical and dimensional models of BD have similar heritability and overlapping genetic risk.

2. Methods

2.1. Participants

The study sample was drawn from the Amish Mennonite Bipolar Genetics Study (AMBiGen), a long-term study of BD and related conditions among people in genetically isolated Amish and Mennonite families (Hou et al., 2013). The Amish and Mennonites offer unique advantages for genetic research due to their large family sizes, rural lifestyles, well-documented genealogy, and low prevalence of substance abuse (Gill et al., 2016). These advantages support the study of the genetic underpinnings of common disorders like BD in Anabaptist groups and may also have relevance to BD in the general population.

This retrospective study included participants that were recruited as part of the AMBiGen study from 2010 to 2022 (Hou et al., 2013; Gill et al., 2016). The study involved two primary populations: North American and South American participants. In North America, families from USA and Canada were recruited through probands with a confirmed diagnosis of bipolar type 1 (BD1), bipolar type 2 (BD2), or schizoaffective bipolar disorder (SABP). First- or second-degree relatives of these probands were also recruited. In South America, participants were recruited from three Mennonite settlements in Brazil. These settlements were selected due to their well-documented genealogical records and their descent from a limited number of founding couples (Lopes et al., 2016). All participants were at least 18 years of age. The study excluded individuals with major physical conditions (e.g., severe physical impairments or disabilities that could interfere with participation or affect the ability to complete mood and behavior assessments), neurological disorders (e.g., multiple sclerosis), or substance use disorders, as these factors could interfere with the accurate diagnosis of BD.

2.2. Data collection

All participants completed the MDQ either by mail or during an inperson assessment. The MDQ has previously been shown to have good sensitivity and specificity as a screen for BD in Anabaptist families ascertained through probands with BD (Gill et al., 2016; Dumont et al., 2020). Participants who had an MDQS greater than or equal to seven (Dumont et al., 2020), or who otherwise endorsed a history of mental health problems based on the Past History Schedule (McGuffin et al., 1986) were asked to undergo a direct assessment with the Diagnostic Interview for Genetic Studies (DIGS). The DIGS is a widely-used, semi-structured psychiatric examination that reliably elicits diagnostic criteria for major depression, mania, psychosis, alcohol and drug use, suicidal behavior, and anxiety disorders (Nurnberger Jr et al., 1994). The DIGS assesses lifetime symptoms as well as the most severe periods of major depression and mania. Following the DIGS, two clinicians independently assigned a best estimate final diagnosis (Leckman et al., 1982) based on the interview, available medical records, and reports from relatives.

Categorical diagnoses were grouped into “narrow” and “broad” diagnostic groups based on published family studies (Gershon et al., 1982). Participants diagnosed with BD1, BD2 with recurrent depression, and SABP were assigned to the “narrow” group. The “broad” group included the “narrow” group along with BD2 with a single episode of depression, schizoaffective depressive disorder (SAD), recurrent MDD, bipolar disorder not otherwise specified, and schizophrenia (SZ). SZ was included in the “broad” diagnostic group since SZ and BD co-occur in families and share over 80 % of common genetic risk factors (reviewed in Gordovez and McMahon, 2020). All other participants, including those with a single episode of major depression (n = 40), were assigned to the “unaffected” diagnostic group. We opted not to incorporate single episodes of major depression and anxiety disorders into the “broad” group since these disorders are very common and do not show strong familial co-aggregation with BD.

2.3. MDQ screening

Of the 1238 participants in the AMBiGen study as of September 2022, we included all 776 participants who had completed the MDQ. The MDQ rates 13 cardinal symptoms of mania, their temporal clustering, and associated impairment. Consistent with the literature, we assigned MDQ scores as a sum of the 13 cardinal symptoms. Thus, scores ranged from 0 to 13. The concurrent symptoms question (CQ) was scored separately: “Of the things we just talked about, have several of these ever happened during the same period of time?” and was converted from Yes or No to 1 or 0. The impairment question (PQ) was also scored separately: “How much of a problem did any of these cause you – like being unable to work; having family, money, or legal troubles; getting into arguments or fights?” and (for purposes of the PCA) was converted from an ordinal scale (“no problem,” _“minor problem,” “moderate problem,” and “serious problem”) to 0 for “no problem“ or “minor problem” and 1 for “moderate” or “serious” problem. We also explored the impact of adding the PQ value to the total MDQ score (MDQP), which expanded the maximum score to 14. As the MDQ is a self-report questionnaire, participants sometimes fail to answer all questions. The 50 participants whose MDQ responses contained >20 % unanswered questions were excluded from the study. For participants with <20 % unanswered questions, missing MDQ items among were treated as a zero (no).

2.4. Data analysis

A PCA was conducted using XLSTAT to explore the factor structure of the MDQ in this sample. The Kaiser-Meyer-Olkin test value was 0.93 indicating that this sample is adequate for PCA. Varimax rotations of two and three factors were run to determine which model fit best with the data. Varimax rotation simplifies item loadings by removing the middle ground and clarifying the factor on which data load is based (Dilbeck, 2017). Higher loading values show a factor’s importance in explaining the data, and these values represent the amount of variability in the data that each factor explains.

A genomic relationship matrix (GRM) was created using high-quality single nucleotide polymorphism (SNP) data. The GRM is a key tool in estimating heritability using genomic data, since it enables more accurate estimation of the genetic variance and covariance among individuals, as opposed to relying solely on pedigree information (Jiang et al., 2019).

$$h^2=\frac{{\sigma }_G^2}{{\sigma }_P^2}$$ (1)

Heritability analysis was performed with Solar-Eclipse (v9.0.0) (Kochunov et al., 2019; Seal et al., 2022; D’Amico et al., 2024). Heritability $\left({\text{h}}^2\right)$ represents the portion of the phenotypic variance $\left({\sigma }_{\text{P}}^2\right)$ accounted for by additive genetic variance $\left({\sigma }_{\text{G}}^2\right)$, as seen in Eq. (1). The sample size for the heritability analysis was reduced from 726 to 432 because GRM data were not available for 294 subjects. The traits chosen for the heritability analysis were MDQS, CQ, PQ, MDQP, and the rotated components (RC) from the three-factor PCA with varimax rotation (RC1, RC2, and RC3 referring exclusively to the rotated components of the three-factor varimax rotation). Before analysis, all traits underwent an inverse normal transformation to normalize their distributions and facilitate accurate statistical comparisons.

$$\rho G=\frac{cov(G1,G2)}{\sqrt{var(G1)\times var(G2)}}$$ (2)

$$\rho P=\frac{cov(P1,P2)}{\sqrt{var(P1)\times var(P2)}}$$ (3)

Genetic correlation analyses were performed with SOLAR-Eclipse (Almasy and Blangero, 1998; D’Amico et al., 2024) using the restricted maximum likelihood (REML) method, which estimates the variance-covariance matrix of the genetic effects on pairs of traits. The genetic correlation (ρG) was then calculated as the ratio of the estimated genetic covariance to the square root of the estimated genetic variances for the two traits, providing an estimate of the extent to which the genetic factors underlying the two traits are correlated. This is shown in Eq. (2). The phenotypic correlation (ρP) was estimated with a similar approach, shown in Eq. (3). A ρG value of zero means that the two traits do not share genetic factors, a ρG value of 1 suggests that the genetic factors are entirely shared, and a ρG value of − 1 would imply that all the genetic influences on one trait are opposite to those on another trait (Man et al., 2019). ρG values of exactly − 1 or 1 should not be overvalued since these values constitute a boundary constraint in the analysis. Results that report a value of − 1 or 1 should be considered as close to − 1 or 1, but not exactly − 1 or 1. Each pairwise correlation of MDQS, CQ, PQ, MDQP, and RC1, RC2, and RC3 with the “narrow” and “broad” groups were examined. To test the significance of the ρG values, we compared the ln likelihood of a restricted null model (with ρG fixed at zero) to that of an alternative model in which the ρG parameter was estimated (Glahn et al., 2012). P values < 0.05 were considered statistically significant.

3. Results

3.1. Sample demographics, psychiatric diagnoses, and dimensional ratings

A total of 776 AMBiGen study participants had completed an MDQ. Of these, 50 were excluded as they contained >20 % unanswered items. As a result, 726 participants were included in this study (484 from North America and 242 from Brazil, South America). Of these, 112 were assigned to the “narrow” diagnostic group (BD1 = 89, BD2 recurrent depression = 20, SABP = 3), and 212 were assigned to the “broad” diagnostic group (“narrow” plus BD2 with single episode of depression = 15, recurrent MDD = 43, SAD = 2, SZ = 6, other major mood disorders = 34). A total of 514 subjects did not meet the criteria for either the “narrow” or “broad” groups and were assigned to the “unaffected” diagnostic group for the purposes of this analysis. Sample characteristics are shown in Table 1.

Table 1.

Sample characteristics.

Variables	Total sample (n = 726)
Age in year, mean ± SD, (min, max)	47 ± 18, (18–99)
Female gender, n (%)	397 (54.7)
Geographic location, n (%)
North America	484 (66.7)
Brazil	242 (33.3)
Diagnosisa, n (%)
Bipolar 1	89 (12.3)
Bipolar 2	35 (4.8)
Major Depressive Disorder	83 (11.4)
Schizophrenia	6 (0.8)
Schizoaffective Disorder	5 (0.7)
Other mood disorders	73 (10.1)
No mood disorders	435 (59.9)

Note:: n = sample size, SD = Standard Deviation.

^aBased on DSM codes.

MDQS ranged from 0 to 13 (mean = 2.83, SD = 3.49). A total of 109 subjects (15.01 %) had an MDQS ≥ 7, the conventional cutoff for screening studies. MDQS scores differed between participants diagnosed with BD1 (mean 5.25, 95 % CI ± 0.925) or BD2 (mean 4.46, 95 % CI ± 1.24), and those with no mood disorders (mean 2.22, 95 % CI ± 0.284), as expected (Fig. 1). In the total sample, the endorsement rate of the 13 MDQ symptom items ranged from 7.71 % (item 13, “spending money got into trouble”) to 36.36 % (item 7, “easily distracted”). As expected, endorsement rates for subjects diagnosed with BD (n = 124) were higher than those without BD, ranging from 28.23 % (item 13) to 78.23 % (item 7) (Fig. 2). Subjects with a psychiatric diagnosis other than BD (n = 167) had an endorsement rate lower than those with a BD diagnosis, but higher than those with no mood disorders (n = 435; details in Table S1).

Fig. 1.

Mean MDQ scores for each categorical diagnosis. Error bars represent 95 % CI. Sample sizes can be found in Table 1. BD1 = Bipolar Disorder Type I, BD2 = Bipolar Disorder Type 2, MDD = Major Depressive Disorder, SZ = Schizophrenia, SAD = Schizoaffective Disorder, No dx = No diagnosis.

Fig. 2.

Endorsement rate of the Mood Disorder Questionnaire items. Whole sample (n = 726) is represented by the blue bars, Bipolar Disorder sample (n = 124) is represented by the green bars, Other Mood Disorders (n = 167) are represented by the red bars, No Mood Disorders (n = 435) are represented by the gray bars. MDQ items: 1 So hyper I got into trouble, 2 Irritable, 3 More self-confident, 4 Needed less sleep, 5 More talkative, 6 Thoughts raced, 7 Easily distracted, 8 Much more energy, 9 Much more active, 10 Much more social, 11 Much more interested in sex, 12 Got involved in excessive, foolish, or risky things, 13 Spending money got me into trouble. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.2. Principal component analysis of MDQ items supports a 3-factor model

PCA identified two principal components (PC) that accounted for 53.05 % of the variance (Fig. 3). PC1 explained 45.05 % of the variance with an eigenvalue of 5.85, while PC2 explained 8.00 % with an eigenvalue of 1.04. Although PC3 had an eigenvalue of 0.93, below the usual cutoff of 1, it was included in the analysis since it explained nearly as much variance (7.12 %) as PC2. Factor loadings for PC1 were significant for all MDQ items. PC2 had notable loadings for items related to irritability (item 2) and distractibility (item 7), while PC3 showed a significant loading for overspending (item 13). Details can be found in Table S3.

Fig. 3.

Scree Plot comparing the first 13 principal components (PC) of the Mood Disorder Questionnaire (MDQ). Eigenvalues are shown by gray bars. The dotted line represents eigenvalue = 1.0. The green line represents the variability (%) of each PC before varimax rotation. The variability (%) of each rotated component after a two-factor or three-factor varimax rotation is indicated by the blue and red lines, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To further clarify the relationships between these PCs, we applied varimax rotation. The two-factor varimax rotation retained 53.05 % of the variance but distributed it more evenly across the first two components (34.01 % and 19.04 %, respectively), with more distinct factor loadings (details in Table S4). The three-factor varimax rotation captured 60.18 % of the variance (Fig. 3), with a balanced distribution of variance across all three components; the loadings broadly represented themes of energy and self-attitude (RC1), irritability and cognition (RC2), and behavioral disinhibition (Fig. 4). See Table S5 for additional details of loadings for the three-factor varimax rotation.

Fig. 4.

Three-factor varimax rotation Principal Component Analysis model of Mood Disorder Questionnaire (MDQ) items 1 to 13, with factor loading scores for each rotated component. Factor loadings > 0.5 are shown. RC = rotated component.

3.3. MDQ factors are modestly heritable

Results of the heritability analyses are shown in Table 2. The MDQP and the MDQS had the highest h², 36 % and 30 % respectively, followed by the PQ with 26 %; all were significantly different from zero (p ≤ 0.001). The CQ did not show any significant heritability. Heritability results for the three-factor varimax rotation ranged from 0.21 (RC1, p < 0.05) to 0.13 (RC3, p = 0.056). Results from the two-factor varimax rotation were similar (Table S6).

Table 2.

Heritability estimates and correlations among dimensional and categorical measures.

Traits	Heritability				Correlations
					Narrow		Broad
	n	h2	Std. E	p-value	ρG	ρP	ρG	ρP
MDQS	432	0.30	0.10	<0.001	1.00a	0.16**	0.90***	0.21***
CQ	432	0.09	0.08	0.12	1.00a	0.47***	0.84	0.58***
PQ	432	0.27	0.10	<0.001	1.00**a	0.37***	0.64***	0.40***
MDQP	432	0.36	0.11	<0.001	1.00a	0.23***	0.95***	0.28***
RC1	432	0.21	0.10	<0.05	1.00a	0.32***	0.62	0.37***
RC2	432	0.20	0.09	<0.05	1.00a	0.30***	1.00a**	0.42***
RC3	432	0.13	0.09	0.056	1.00a	0.35***	1.00**a	0.11***

Note: n = sample size, h² = heritability, Std E. = standard error, ρG = genotypic correlation, ρP = phenotypic correlation, MDQS = MDQ score, CQ = concurrent symptoms question (“…have several of these ever happened during the same period of time?”) PQ = impairment question (“How much of a problem did any of these cause you…?”), MDQP = MDQS + PQ, RC1 = first rotated component, RC2 = second rotated component, RC3 = third rotated component.

^*p ≤ 0.05

^**p ≤ 0.01

^***p ≤ 0.001. P-values indicate correlations that are significantly different from zero.

^aρG values of 1.00 represent a boundary constraint in the analysis method and should not be construed as complete genetic overlap between traits.

3.4. Strong genetic and weaker phenotypic correlations between MDQ factors and BD

Genetic overlaps between MDQ traits and categorical diagnoses were estimated as genetic correlations (Table 2). All genetic correlations were above 0.6 and many reached the upper boundary of the estimator, indicating substantial genetic overlaps. Although the ρG values were all close to 1.0, genetic correlations within the “narrow” diagnosis group were significant only for impairment (PQ) (ρG = 1.00, p < 0.01). The larger “broad” diagnostic group displayed strong and significant genetic correlations with most MDQ-derived traits, including MDQS (p < 0.01), PQ (p < 0.05), MDQP (p < 0.001), and RC3 (p < 0.001). Both the “narrow” and “broad” diagnostic groups also showed significant phenotypic correlations with MDQ traits, although these were smaller in magnitude than the genetic correlations. The maximal phenotypic correlation was observed between the concurrent symptoms (CQ) item of the MDQ and the “broad” diagnostic group (ρP = 0.58, p < 0.001).

4. Discussion

This study is the first to explore the heritability of the MDQ and genetic overlap between MDQ measures and categorical diagnoses in a family sample. By examining the heritability of the MDQ, we aimed to enhance understanding of the genetic influences on BD and its constituent signs and symptoms. As expected, individual endorsed more MDQ items than those with other or no mood disorders. PCA suggested that a three-component model of the MDQ captured most of the variance, with loadings representing distinct symptom themes. The MDQ score and its principal components were mostly heritable, but the component representing behavioral inhibition was not. Genetic overlaps between MDQ measures and categorical diagnoses were strong; stronger even than the phenotypic correlations. These results indicate substantial genetic overlaps between dimensional and categorical models of BD and support a genetic continuity across a broad spectrum of BD symptoms.

The strong genetic correlations between MDQ and categorical diagnoses in this sample suggests substantial overlap of genetic risk factors, while the weaker phenotypic correlations suggest that the shared genetic risk is expressed differently in different individuals. The genetic correlations we observed are generally higher than those previously reported between SZ and BD (ρG = 0.68) and between BD and MDD (ρG = 0.47) (Fabbri, 2021). In another study of the AMBiGen sample, D’Amico et al. (2024) found no significant genetic correlations between BD diagnostic groups and cognitive performance; our study emphasizes the relevance of focusing on mood-related traits, which might be more genetically aligned with BD risk.

Previous MDQ studies have shown that either a two factor (Carta et al., 2014; Chung et al., 2008; Ouali et al., 2020; Sanchez-Moreno et al., 2008) or three factor (Chung et al., 2009; Jon et al., 2009; Massidda et al., 2016; Mundy et al., 2023; Yang et al., 2011) model best fits the data structure of the instrument. We put this to the test by performing a PCA with varimax rotation on both models, comparing the results to a PCA without varimax rotation. The three-factor model performed slightly better in this sample. These findings align with similar three-factor models observed in Korean (Jon et al., 2009), Chinese (Yang et al., 2011), Hong Kong (Chung et al., 2009), and UK (Mundy et al., 2023) populations, suggesting a cross-cultural consistency in the underlying dimensions of the MDQ. While specific dimension labels may differ slightly, the overall three-factor model seems robust, consistent with core dimensions that capture typical constructs of mania.

The MDQP showed the highest heritability among the dimensional traits we tested (36 %). This is lower than most heritability estimates of categorical BD diagnoses (40–85 %) (Bruce et al., 2019; Fabbri, 2021; Lee et al., 2013; McGuffin et al., 2003), which suggests a larger role for non-genetic factors in dimensional measures, including environmental influences and measurement error. The PQ itself was also significantly heritable (26 %). The heritability of the MDQP may reflect the increased specificity and reduced noise associated with problematic behaviors, as well as the extra information conveyed by impairment. While interesting, the MDQP should be regarded as an exploratory trait. More studies are needed to validate its usefulness as a dimensional measure of bipolarity.

In the three-factor model, RC1 and RC2 were significantly heritable while RC3 was not. This finding suggests that the largely non-behavioral symptoms that load on RC1 and RC2 are more heritable than the behavioral symptoms that load on RC3. Put into a clinical context, there may be fewer paths to persistent increased energy, self-confidence, sleeplessness and elation than there are to erratic behavior, hence one might expect to see less genetic overlap between behavior and BD diagnosis. However, behavioral symptoms of mania were endorsed by relatively few participants (which probably reflects cultural and environmental factors characteristic of the Anabaptist communities we studied), reducing power to detect heritability for these traits in this sample.

To our knowledge four previous studies have examined the relationship between dimensional and categorical models of BD. Fears et al. (2014), examined a large range of traits measured in extended pedigrees ascertained through probands with BD. Among their five dimensional measures of “affective temperament,” one (TEMPS Hyperthymia) showed significant genetic correlation with BD. Bruce et al. (2019) found that a quantitative measure of bipolar symptoms was significantly heritable and concluded that “bipolarity trait assessment may be used to supplement the diagnosis of BD in future genetic studies and could be especially useful for capturing subclinical genetic contributions to a BD phenotype.” Those results largely agree with the present study, even though we employed a different measure of “bipolarity.” Hosang et al. (2022) explored the relationship of subsyndromal hypomania and BD, SZ, and MDD in twins using the parent-rated MDQ. They found moderate genetic and nonshared environmental correlations between hypomania and BD; hypomania was significantly associated with the polygenic risk scores for SZ and MDD but not for BD. Their findings are consistent with our study, which also found a significant genetic correlation between the MDQS and the “broad” diagnostic group that includes MDD and SZ. Mundy et al. (2023) conducted a genome-wide association study of MDQ scores on a sample of the UK population in order to determine the validity of the MDQ as a screening tool for BD. The authors found that the MDQS was not significantly heritable and was most strongly associated with symptoms of general distress or psychopathology, not just manic symptoms. These results contrast with the significant heritability of the MDQ and genetic overlap with BD in the present study. This contrast likely reflects differences in study populations. Our study was based on individuals ascertained through family members with BD. Manic symptoms reported by our participants are thus more likely to share common genetic determinants with BD than similar symptoms reported by the general population.

5. Limitations

This study has several limitations. First, while the total sample was well-powered, owing to unavailable genetic data the heritability and genetic correlation analyses had a reduced sample size, reducing power to detect weak effects and precision of the apparently stronger effects. Second, the recruitment of participants through probands with BD, while typical of family studies, resulted in an increased prevalence of BD in this sample, reducing the generalizability of the genetic associations we report to populations with lower baseline rates of BD. Third, while this study sheds light on the genetic and phenotypic relationships between dimensional and categorical measures of BD, it does not point to specific genes.

6. Conclusions & future research directions

Mental health among relatives of people with BD varies broadly, ranging from completely well to severely affected. Categorical diagnoses of BD fail to capture this complexity. Our findings suggest that a brief screening tool like the MDQ can effectively identify heritable traits that have a strong genetic overlap with BD itself. This means that the MDQ is a measure of quantitative phenotypes that may be well suited for the discovery and characterization genes that confer risk for BD. These results also support the value of the MDQ as a screening tool for BD, especially among individuals at high a priori risk. Further studies are needed to determine if these results generalize to larger, more diverse populations. More work is also needed to fully elucidate the genetic relationships between dimensional and categorical models of bipolar disorders and related conditions.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jad.2024.12.030.