Specificity in Etiology of Subtypes of Bipolar Disorder: Evidence From a Swedish Population-Based Family Study

Accepted Manuscript Specificity in etiology of subtypes of bipolar disorder: Evidence from a Swedish population-based family study Jie Song, Ph.D., Ralf Kuja-Halkola, Ph.D., Arvid Sjölander, Ph.D., Sarah E. Bergen, Ph.D., Henrik Larsson, Ph.D., Mikael Landén, M.D., Ph.D., Paul Lichtenstein, Ph.D. PII:

S0006-3223(17)32204-7

DOI:

10.1016/j.biopsych.2017.11.014

Reference:

BPS 13389

To appear in:

Biological Psychiatry

Received Date: 11 April 2017 Revised Date:

3 October 2017

Accepted Date: 10 November 2017

Please cite this article as: Song J., Kuja-Halkola R., Sjölander A., Bergen S.E., Larsson H., Landén M. & Lichtenstein P., Specificity in etiology of subtypes of bipolar disorder: Evidence from a Swedish population-based family study, Biological Psychiatry (2017), doi: 10.1016/j.biopsych.2017.11.014. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

ACCEPTED MANUSCRIPT Title: Specificity in etiology of subtypes of bipolar disorder: Evidence from a Swedish population-based family study Short title: Specificity in etiology of subtypes of bipolar disorder

Authors

RI PT

Jie Song1, Ph.D.; Ralf Kuja-Halkola1, Ph.D.; Arvid Sjölander1, Ph.D.; Sarah E. Bergen1,2, Ph.D.; Henrik Larsson1,3, Ph.D.; Mikael Landén1,4, M.D., Ph.D.; Paul Lichtenstein1, Ph.D. 1. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

SC

2. Stanley Center for Psychiatric Research, The Broad Institute of MIT and Harvard, Cambridge, MA, USA

M AN U

3. School of Medical Sciences, Örebro University, Örebro, Sweden

4. Institute of Neuroscience and Physiology, The Sahlgrenska Academy at Gothenburg University, Gothenburg, Sweden

Correspondence

TE D

Jie Song, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Box

EP

281, 17177 Stockholm, Sweden. Fax: + 46 8 314 975. Email: [email protected].

Key words: Bipolar I disorder; Bipolar II disorder; Family study; Etiology; Heritability;

AC C

Genetic correlation

Word count: 245 words in the abstract, 3981 words in the article body, 2 figures, 2 tables, 1 supplemental information

Jie Song

2

ACCEPTED MANUSCRIPT Abstract BACKGROUND: Uncertainty remains whether bipolar I disorder (BDI) and bipolar II disorder (BDII) differ etiologically. We used a population-based family sample to examine

RI PT

the etiological boundaries between BDI and BDII, by assessing their familial aggregation/coaggregation, and by assessing the co-aggregation between them and schizophrenia,

depression, attention-deficit/hyperactivity disorder, eating disorders, autism spectrum

SC

disorders, substance use disorders, anxiety disorders and personality disorders.

METHODS: By linking Swedish national registers, we established a population-based

M AN U

cohort (N=15,685,511) and identified relatives with different biological relationships. Odds ratios (ORs) were used to measure the relative risk of BDI and BDII in relatives of individuals diagnosed with BDI (N=4,309) and BDII (N=4,178). The heritability for BDI and BDII and the genetic correlation across psychiatric disorders were estimated by variance

TE D

decomposition analysis.

RESULTS: Compared with the general population, the OR of BDI was 17.0 (95% confidence interval (CI) 13.1-22.0) in first-degree relatives of BDI patients; higher than that

EP

of BDII patients (OR 9.8, 95% CI 7.7-12.5). The ORs of BDII were 13.6 (95% CI 10.2-18.2)

AC C

in first-degree relatives of BDII patients and 9.8 (95% CI 7.7-12.4) in relatives of BDI patients. The heritabilities for BDI and BDII were estimated at 57% (95% CI 32%-79%) and 46% (95% CI 21%-67%), respectively, with a genetic correlation estimated as 0.78 (95 %CI 0.36-1.00). The familial co-aggregation of other psychiatric disorders, in particular schizophrenia, showed different patterns for BDI and BDII. CONCLUSIONS: Our results suggest a distinction between BDI and BDII in etiology, partly due to genetic differences.

Jie Song

3

ACCEPTED MANUSCRIPT Introduction The discrete boundaries between current diagnostic classifications of psychiatric disorders have been increasingly challenged by the emerging evidence of common genetic

RI PT

determinants from recent family and molecular genetic studies (1, 2). This challenge is particularly relevant for subtypes of mood disorders. Bipolar I disorder (BDI) and bipolar II disorder (BDII), the two most common subtypes of bipolar disorder (BD), were formally differentiated decades ago and introduced in the fourth edition of the Diagnostic and

SC

Statistical Manual of Mental Disorders (DSM-IV) (3). Nevertheless, controversy exists about

M AN U

whether they only differ in severity or also in etiology. On the one hand, BDII can be seen as a milder form of BD that only requires hypomanic episodes as opposed to full-blown manic episodes required for a BDI diagnosis. On the other hand, thorough reviews have argued for a distinction between BDI and BDII (4, 5), and recent studies have reported more unfavorable

TE D

illness features in BDII (6) and genetic heterogeneity between them (7). Family studies help identify the etiologic boundaries between disorders. Despite solid evidence from previous family and genetic studies for familial aggregation of BD,

EP

investigations comparing BDI and BDII are scarce. Previous studies based on selected small samples in the 1980s and 1990s generally found familial aggregation among first-degree

AC C

relatives within each subtype but were inconclusive for the familial co-aggregation across subtypes (8-11). As the definitions of subtypes have changed over time, updated and more precise quantifications of the familial aggregation and co-aggregation are warranted. Recently, two studies using a community-based family sample have found specificity of familial transmission of BDI but not BDII (12, 13). Up to now, no assessments have been done thoroughly in a nationwide sample.

Jie Song

4

ACCEPTED MANUSCRIPT Examination of familial co-aggregation and genetic correlations between BD subtypes and

other psychiatric disorders also provide insights regarding their etiological entities. Despite a few studies examining the differences in psychiatric comorbid conditions between BDI and

risk factors contributed to these co-occurrence unknown.

RI PT

BDII, less is studied across family members, leaving the extent to which familial and genetic

In this study, we used genetically informative data from the Swedish national registers and the quality register of BD to investigate the potential etiological differences between BDI and

SC

BDII. Specifically, we assessed 1) the familial aggregation/co-aggregation of BDI and BDII,

M AN U

2) the heritability for BDI, BDII and their genetic correlation, and 3) the familial coaggregation and genetic correlations between BDI/BDII and other major psychiatric disorders.

Study population

TE D

Methods and Materials

Patients with BDI and BDII were identified in BipoläR, the Swedish quality register for BD.

EP

This register is described in details elsewhere (14-16). In brief, BipoläR was established in 2004. It records clinical data for patients diagnosed with BDI, BDII, schizoaffective disorder

AC C

of bipolar type and BD not otherwise specified. After a patient has been registered, followups are carried out annually. The BD diagnoses were made by the treating psychiatrists according to the DSM-IV-Text Revision (DSM-IV-TR). Since the psychiatrists have access to all clinical data including available longitudinal perspectives of patients’ course of illness, the validity of the diagnoses in BipoläR is likely to be high (16). The register includes patients with severe forms of BD that require hospitalization as well as patients treated in psychiatric outpatient healthcare units. Participation is voluntary for the clinicians and

Jie Song

5

ACCEPTED MANUSCRIPT patients. When we extracted data in December 2013, a total of 12,837 patients with BD were registered in BipoläR. The study population was established by linkage of Swedish national registers (1, 17). By

RI PT

identifying biological parents through the Multi-Generation Register, we constructed different cohorts of biological relatives: first-degree relatives (parent, offspring and full-

sibling) and second-degree relatives (grandparents, grandchildren, aunt/uncle, nephew/niece, maternal and paternal half-sibling). The lifetime diagnoses of other psychiatric disorders were

SC

extracted from the National Patient Register with records of psychiatric inpatient admissions

M AN U

(since 1973) and outpatient contacts (since 2001). Discharge diagnoses were coded according to the International Classification of Diseases (ICD) system (ICD-8: 1973–1986, ICD-9: 1987–1996, ICD-10: 1997–present). We obtained data on individuals’ sex and year of birth from the Total Population Register. Finally, we obtained information when they lived in Sweden from the Migration Register and the Cause of Death Register. All registers were

TE D

followed from their start until December 2013. The study was approved by the ethics committee at Karolinska Institutet.

EP

For analyses of heritability and genetic correlation, we established a sibling cohort born from 1955 to 1990 with follow-up time from 1973 to 2013. We excluded those who died,

AC C

immigrated or emigrated before age 25. We selected one sibling pair born within five years of each other (if multiple pairs were available, we selected the oldest pair) from each nuclear family. Our sibling samples consisted of 4,970 monozygotic twins, 7,488 dizygotic twins, 46,010 maternal half-sibling pairs, 30,985 paternal half-sibling pairs and 807,738 full sibling pairs.

Ascertainment of BDI, BDII and other psychiatric disorders Jie Song

6

ACCEPTED MANUSCRIPT In BipoläR, a total of 5,981 patients had a lifetime diagnosis with BDI and 5,868 had BDII during 2004-2013. In order to reduce misclassification, we excluded patients ever diagnosed with both BDI and BDII at different time of follow-up (N=1,195), which left 4,786 patients

with BDI and 4,673 patients with BDII. Furthermore, we excluded patients (N=972) who did

RI PT

not fulfill a validated algorithm for BD diagnosis (16) that we have applied in previous

studies (1), leaving a final sample of 4,309 BDI patients and 4,178 BDII patients. The first discharge date with an ICD code of a BD diagnosis from the National Patient Register was

SC

treated as the age at first diagnosis.

M AN U

The diagnostic codes used for other psychiatric disorders (schizophrenia, depression, attention-deficit hyperactivity disorder (ADHD), eating disorders, autism spectrum disorders (ASD), substance use disorders, anxiety disorders and personality disorders) are described in

Statistical analysis

TE D

Supplementary Table S1.

EP

The analyses were complicated by the fact that the National Patient Register, from which most of the BD diagnoses are retrieved, does not have information on BD subtype since it

AC C

uses the ICD diagnostic system. Consequently, we excluded all patients with unclear BD subtype status (from the National Patient Register) or other BD subtypes (from BipoläR) from the study population. As shown in the supplementary methods, this exclusion distorts the prevalence of BD in the analytic sample, but does not bias the estimated odds ratios (ORs). Familial aggregation and co-aggregation

Jie Song

7

ACCEPTED MANUSCRIPT To quantify the familial aggregation and co-aggregation for BDI and BDII, we conducted pairwise comparisons of related individuals. We constructed, for each type of biological relatedness, all possible pairs in the dataset. For each degree of relatedness we used logistic regression models to estimate the following four ORs: 1) the odds of BDI in a relative, given

RI PT

BDI in the proband, compared with the odds of BDI in a relative, given no BD in the proband; 2) the odds of BDI in a relative, given BDII in the proband, compared with the odds of BDI in a relative, given no BD in the proband; 3) the odds of BDII in a relative, given BDII in the

SC

proband, compared with the odds of BDII in a relative, given no BD in the proband; 4) the odds of BDII in a relative, given BDI in the proband, compared with the odds of BDII in a

M AN U

relative, given no BD in the proband. To increase power, we also estimated the ORs by combining the same degree of genetic relatedness (e.g., combining parent, offspring and fullsibling as the first-degree relative).

To assess the familial co-aggregation between BD subtypes and other psychiatric disorders

TE D

(schizophrenia, depression, ADHD, eating disorders, ASD, anxiety disorders and personality disorders), we estimated the ORs of a lifetime diagnosis of each disorder in full-siblings of

EP

patients with BDI and BDII compared with full-siblings of individuals unaffected with BD. Considering the significant differences in sex and calendar year of birth for patients with BDI

AC C

and BDII (see Table1), we adjusted for sex and year of birth (categorically, before 1955, 1955-1962, 1963-1969, 1970-1977, 1978-1988, after 1988) for the relative for whom we estimated the ORs. We calculated the 95% confidence intervals (CIs) with cluster robust standard errors to account for the non-independence between individuals due to familial clustering. For each psychiatric disorder, we tested the difference of ORs for the two exposures (i.e., BDI or BDII in the proband). Different ORs would suggest that common familial risks (shared genetic and environmental factors between relatives) contribute to the co-occurrences to a different degree. Jie Song

8

ACCEPTED MANUSCRIPT Three additional analyses were performed to test the robustness of the results and to provide further insights. First, we varied the criteria defining BD subtype, including 1) a nonhierarchical diagnosis that allows one to be diagnosed with both BDI and BDII, 2) using the most recent diagnoses for each patient, and 3) assuming that patients from the National

RI PT

Patient Register without hospital admissions for mania may be ascribed to the BDII subgroup (for ICD codes, see supplements). Second, we did not remove the unspecified or other

subtypes from the study population and preformed analyses by directly estimating the ORs of

SC

BD subtypes (e.g., we estimated the odds of BDI in a relative, given BDI vs. non-BDI in the

transmission of the BD subtypes. Heritability and genetic correlation

M AN U

proband). Third, we performed stratified analyses to examine the sex-specific patterns of

We conducted variance decomposition analysis to investigate the relative importance of

TE D

genetic contributions to the liability of BDI and BDII. We fitted liability-threshold models presupposing that a continuous underlying liability for the subtype is normally distributed in the population and that individuals with a liability beyond a certain threshold develop the

EP

disorder. The variance in liability was decomposed into additive genetic, shared environment and non-shared environment components using structural equation modeling; model fitting

AC C

was done using maximum likelihood. The variance associated with the additive genetic components divided by the total variance was calculated as the heritability. It measures the proportion of the liability accounted for by genetic factors for each subtype. In addition, we estimated the genetic correlations between subtypes of BD as well as between BDI/BDII and the other psychiatric disorders. The correlation between the additive genetic components for two disorders is estimated to be the genetic correlation. Because no individual can have both diagnoses of BDI and BDII, we estimated their genetic correlation

Jie Song

9

ACCEPTED MANUSCRIPT by excluding contribution from the within-individual cross-disorder correlation (for details regarding model modifications and model assumptions, see the supplemental methods). We estimated the 95% likelihood-based CIs for heritability and genetic correlation.

RI PT

We used SAS 9.4 (18) for data management and Stata 13.0 (19) to perform logistic regression models. For analysis of genetic correlation we used OpenMx package (16) in R software

SC

version 3.5 (20).

M AN U

Results

The descriptive characteristics for BDI and BDII subgroups, the entire group from the quality register, and all patients with BD identified from the National Patient Register are shown in Table 1. Patients with BDI are significantly different from patients with BDII in several aspects, including a lower proportion of females, earlier calendar year of birth and first

TE D

diagnosis, and earlier age of diagnosis. The mean age at date of first discharge diagnosis is higher than the age at onset for BD reported in literature, which is probably due to the low

EP

coverage of Swedish National Patient Register in early calendar periods. When we only included patients who were born after 1973 when the psychiatric inpatient admissions were

AC C

recorded afterwards, the mean age at first diagnosis was 24.9 for BDI and 26.6 for BDII, which remained significantly different. The results of familial aggregation for BDI/BDII and co-aggregation between them are shown in Figure 1. We observed generally stronger familial transmission within each subtype than cross-subtype for each first-degree relative. Combined together, the overall OR of BDI was significantly higher in relatives of BDI patients than that of BDII patients (OR 17.0, 95% CI 13.1-22.0 vs. OR 9.8, 95% CI 7.7-12.5, P=0.002); the overall OR of BDII was higher,

Jie Song

10

ACCEPTED MANUSCRIPT

albeit not significantly different, in relatives of BDII patients compared with relatives of BDI patients (OR 13.6 (95% CI 10.2-18.2) vs. 9.8 (95% CI 7.7-12.4), P=0.09). Results using different definitions of BD subtypes and with inclusion of unspecified cases showed similar trends of a higher aggregation within each subtype than co-aggregation across them among

RI PT

first-degree relatives (see Supplementary Table S2). The familial risks for second-degree relatives varied across each relationship due to limited number of cases, and the combined ORs yielded no significant difference. Finally, analyses stratified on sex showed no specific

SC

patterns of transmission (Supplementary Table S3).

M AN U

The heritability was estimated at 57% (95% CI 32%-79%) for BDI and 46% (95% CI 21%67%) for BDII, with the remaining variance attributed to non-shared environmental factors. The genetic correlation between BDI and BDII was estimated at 0.78 (95% CI 0.36-1.00). To test if the estimate was largely affected by traditional hierarchical diagnosis, we applied a non-hierarchical classification for BDI and BDII (i.e., to allow for co-occurrence within the

TE D

same individual), and the estimate of genetic correlation remained similar 0.81 (95% CI 0.501.00).

EP

Figure 2 presents the ORs of other psychiatric disorders in full-siblings of BDI and BDII patients. The risk of schizophrenia was increased among full-siblings of BDI patients (OR 2.5,

AC C

95% CI 2.0-3.2) but not clear among full-siblings of BDII patients (OR 1.3, 95% CI 0.9-1.9), with a significant difference between them (P=0.003). Other psychiatric disorders (i.e., depression, ADHD, eating disorders, ASD, substance use disorders, anxiety disorders and personality disorders) had increased risks in siblings of patients with both BD subtypes compared with that of general population. Full-siblings of BDII patients had significantly higher ORs for substance use and anxiety disorders compared with full-siblings of BDI patients.

Jie Song

11

ACCEPTED MANUSCRIPT To evaluate whether the observed different patterns of familial co-aggregation were due to the core different features of BDI and BDII, i.e., mania and hypomania, we performed additional analyses by estimating the ORs of being diagnosed with schizophrenia in fullsiblings of patients who had ever experienced mania and hypomania. The results showed a

RI PT

positive association between schizophrenia and mania among siblings (OR 1.7, 95% CI 1.22.6), which is not the situation between schizophrenia and hypomania (OR 0.8, 95% CI 0.51.3).

SC

As shown in Table 2, the estimated genetic correlations between BDI/BDII and other major

M AN U

psychiatric disorders were generally considerable (0.23-0.49, though not always statistically significant), except for a non-obvious genetic correlation between BDII and schizophrenia.

Discussion

TE D

Using a register-based sample of BD subtypes, which was larger than twice the combined size of previous studies (6, 8-13), and general population controls, we explored the difference

EP

between BDI and BDII in familial nature and etiology. Despite substantial genetic correlation, the higher familial aggregation within each BD subtype compared with across subtypes

AC C

suggests that BDI and BDII differ in etiology. Moreover, for the first time we extensively examined the familial co-aggregation and genetic correlations between BD subtypes and other major psychiatric disorders. The familial co-aggregation and genetic overlap between schizophrenia and BDI, which is not obvious between schizophrenia and BDII, further highlight the boundaries in the etiology between them. Notably, the main focus of this study is to compare the difference between BDI and BDII (by using a patient group that is more clinically relevant). Given the potential discrepancy of the sampling frame between the

Jie Song

12

ACCEPTED MANUSCRIPT

quality register and the patient population, the point estimate for each subtype may be biased but the differences between them are most likely to remain valid. We observed two different features between BDI and BDII subgroup. First, the proportion in

RI PT

females compared to males was higher in BDII, which is in line with several studies that have found a higher prevalence for females in BDII but no gender difference in BDI (21-23). Second, BDI vs. BDII patients had a younger mean age at first hospital visit. However,

whether this difference reflects an early age at onset of BDI than BDII is complicated by

SC

other factors, for example, more delayed accurate diagnosis for BDII and secular changes for

M AN U

ICD codes and DSM diagnoses. A recent review comparing the age at onset in patients with BDI and BDII also showed mixed findings (24).

Our results demonstrating specific familial aggregation among first-degree relatives for BDI and BDII confirm findings from early family studies (8-11). We further extended our

TE D

analyses to the second-degree relatives that were inconclusive in previous studies (11, 25). More importantly, the familial aggregation within each subtype was stronger than the coaggregation across them, suggesting that BDI and BDII to some extent are two entities with

EP

etiological heterogeneity. Our findings are in line with some family and genetic research on BD subtypes. Two recent family studies using a US nonclinical sample found familial

AC C

transmission only for BDI (12, 13), whereas our results support familial aggregation for both BDI and BDII. The recent genome-wide association study (GWAS) on BDII found several susceptibility loci not detected before using mixed or BDI subtypes, suggesting subtypespecific pathways (26). The estimates of heritability for BDI and BDII did not differ much in our study. However, a new GWAS using a large sample of BDI and BDII reported a significant difference in their “SNP-based” heritability (i.e., proportion of phenotypic variance explained by single-nucleotide polymorphisms) (7). Their estimate of genetic correlation also points towards a genetic heterogeneity between BDI and BDII. Jie Song

13

ACCEPTED MANUSCRIPT However, BDI and BDII are not completely etiologically different, which could be inferred from the considerable co-aggregation and strong genetic correlation between them. The

significant co-aggregation observed in our sample is consistent with family studies in 1980s but contrary to the two contemporary family studies that revealed no significant cross-

RI PT

aggregation between BDI and BDII (12, 13). The specific but overlapping etiology in

subtypes is somewhat expected in psychiatry and is congruent with recent molecular genetic evidence (7), suggesting that the genetic influences on the development of BDI and BDII are

SC

highly shared (25).

M AN U

The associations between BD subtypes and other psychiatric disorders further advance our understanding of their etiological boundaries and where they fall among psychiatric disorders. The familial co-aggregation could be due to shared genetic and environmental factors, but the latter was not supported in the heritability analyses. Taken together with the estimates of genetic correlations, we observed two interesting features. First, BDII compared with BDI

TE D

appears to share fewer etiological factors with schizophrenia. Combined with previous evidence that emphasizes the correlation between BD and schizophrenia (27), our findings

EP

suggest that the correlation is mainly driven by BDI. This was further supported by the familial clustering of schizophrenia and mania but not hypomania – the core difference

AC C

between BDI and BDII. Similarly, the recent GWAS also reported a greater load of polygenic risk alleles for schizophrenia in BDI patients compared with BDII patients (7). Second, familial co-aggregation suggested that BDII, in contrast to BDI, appears to have closer relationships with substance use and anxiety disorders. However, this is not obvious from the genetic correlations or from previous studies (6, 28). Of note, our results demonstrate both difference and overlap in the etiology between BDI and BDII, but we do not argue for or against a categorical or dimensional classification. The current diagnostic system adopts BDI and BDII as two separated entities mainly based on Jie Song

14

ACCEPTED MANUSCRIPT findings including diagnostic stability, family history and genetic loading (29, 30). This criterion has been challenged by some studies (31, 32), and an alternative continuum or spectrum has been suggested (4). Although family studies are useful in evaluating the

categorical or dimensional models, it could result in different interpretations (33). Whether to

etiology, but rather a clinical diagnostic convention (5).

RI PT

use a categorical or dimensional classification is not (only) based on shared or different

Our findings should be interpreted with caution due to the following reasons. First, our

SC

estimates are the familial aggregation/co-aggregation and genetic correlation of BDI and

M AN U

BDII identified from the quality register, which may differ from the real patient population if the quality register’s patient group is not representative. Such discrepancy is less likely to happen between the quality register and the National Patient Register, since both registers use clinical diagnoses and capture mainly treatment-seeking people. This, on the other hand, may occur in comparison with survey-based studies because registers could not optimally capture

TE D

less severe forms of BD and the sensitivity of diagnoses is relatively weaker. However, it is noteworthy that the quality register was established much later than the National Patient

diagnosis.

EP

Register, which indicates that we cover a younger population of patients with a more recent

AC C

Second, we used mixed diagnoses for ascertainment of disorders, with DSM-IV-TR for subtypes of BD and ICD codes for the other psychiatric disorders. As a consequence, the relative risks were affected by prevalence and cannot be compared across psychiatric disorders. The estimated ORs may be different from other studies using the same diagnostic system. Nevertheless, the comparisons of ORs for the same disorder between subtypes of BD as different exposures remained valid, which is the main focus of our study.

Jie Song

15

ACCEPTED MANUSCRIPT Third, estimates of genetic correlations are likely to be affected by different diagnostic definitions for BDI and BDII and different assumptions required for model fitting. Despite

this, the estimates of heritability and genetic correlation are less influenced by prevalence of disorders than relative risk measures (i.e., ORs) and are able to disentangle the contributions

RI PT

from genetic and environmental factors. Moreover, it yields a possibility for comparison

across disorders (e.g., genetic correlation for BDI/BDII vs. BDII/depression) that cannot be concluded from the familial co-aggregation. However, we failed to reach precise estimates of

SC

heritability and genetic correlations due to the relatively small samples of BD subtype.

M AN U

Our study has several other limitations. 1) The validity of subdiagnosis of BD in the quality register has not been evaluated. Although the diagnoses assigned by trained psychiatrists substantially minimized the proportion of false positives, accurate diagnoses of BD subtypes remained difficult in clinical practice, especially for BDII (33). The misdiagnosis between BDI and BDII might affect the magnitude of familial aggregation and co-aggregation;

TE D

nevertheless, the misclassification is more likely to be non-differential, which would produce bias towards the null and yield a conservative estimation of the difference. 2) Since the BD

EP

subtype information can only be obtained from the quality register, we removed the majority of the BD population from our analyses because they lacked subtype status. We are not sure

AC C

whether a bias would be introduced for the variance decomposition analysis due to this exclusion, and in what direction it would affect the genetic correlation. However, the estimation of ORs was unlikely to be strongly biased due to the relatively low prevalence of BD and the large sample size of the population. Moreover, the main aim of the study, i.e., to compare the familial aggregation and co-aggregation between BDI and BDII, would not be biased irrespective of the point estimates. 3) We were unable to further explore the source (i.e., illness features) of the etiological differences between BDI and BDII. For example, BDI and BDII differ not only in the occurrence of manic episodes; the diagnosis of BDII requires Jie Song

16

ACCEPTED MANUSCRIPT at least one major depressive episode, which is not needed (although typical) for the diagnosis of BDI. However, owing to the lack of reliable discrimination of the major components of mood disorders, we cannot further examine the specificity of major

depression in the etiology of BDI and BDII. Similarly, we were also unable to examine other

otherwise specified) due to the unreliable diagnosis.

RI PT

subtypes of BD (e.g., cyclothymia, schizoaffective disorder of bipolar type and BD not

In summary, this largest family study ever demonstrates different (albeit overlapping)

SC

etiological boundaries between the two most common subtypes of BD and further informs

M AN U

their genetic distinctions with schizophrenia. The emphasis of genetic distinction from our findings complements the growing evidences of shared biological pathways, indicating a complex etiology of BDI and BDII with both unique and common genetic determinants. From a clinical perspective, conventional division of BD into subtypes is necessary for employing different treatment strategies. From a research perspective, future genetic studies

TE D

focusing on a narrow bipolar phenotype may offer an opportunity to characterize novel

AC C

EP

biomarkers reinforcing the difference in the genetic architecture between subtypes.

Jie Song

17

ACCEPTED MANUSCRIPT Acknowledgments Supported in part by The Swedish Research Council, the Swedish Council for Health,

Working Life and Welfare, the Swedish foundation for Strategic Research (KF10-0039), and

RI PT

the China Scholarship Council. The project has also received funding from the European

AC C

EP

TE D

M AN U

SC

Union’s Horizon 2020 research and innovation program under grant agreement No. 667302.

Jie Song

18

ACCEPTED MANUSCRIPT Financial Disclosures The authors report no conflicts of interest including relevant financial interests, activities, relationships and affiliations.

RI PT

Henrik Larsson has served as a speaker for Eli-Lilly and Shire and has received research grants from Shire; Paul Lichtenstein has served as a speaker for Medice; Mikael Landén

declares that, over the past 36 months, he has received lecture honoraria from Lundbeck and

SC

AstraZeneca Sweden, and served as scientific consultant for EPID Research Oy; all outside the submitted work. All other authors report no biomedical financial interests or potential

AC C

EP

TE D

M AN U

conflicts of interest.

Jie Song

19

ACCEPTED MANUSCRIPT References

AC C

EP

TE D

M AN U

SC

RI PT

1. Song J, Bergen SE, Kuja-Halkola R, Larsson H, Landen M, Lichtenstein P (2015): Bipolar disorder and its relation to major psychiatric disorders: a family-based study in the Swedish population. Bipolar Disord. 17:184-193. 2. Cross-Disorder Group of the Psychiatric Genomics C, Genetic Risk Outcome of Psychosis C (2013): Identification of risk loci with shared effects on five major psychiatric disorders: a genomewide analysis. Lancet. 381:1371-1379. 3. Goodwin FK, Jamison KR (2007): Manic-depressive Illness, 2nd edn. New York: Oxford University Press. 4. Benazzi F (2007): Bipolar disorder--focus on bipolar II disorder and mixed depression. Lancet. 369:935-945. 5. Vieta E, Suppes T (2008): Bipolar II disorder: arguments for and against a distinct diagnostic entity. Bipolar Disord. 10:163-178. 6. Dell'Osso B, Holtzman JN, Goffin KC, Portillo N, Hooshmand F, Miller S, et al. (2015): American tertiary clinic-referred bipolar II disorder compared to bipolar I disorder: More severe in multiple ways, but less severe in a few other ways. J Affect Disord. 188:257-262. 7. Charney AW, Ruderfer DM, Stahl EA, Moran JL, Chambert K, Belliveau RA, et al. (2017): Evidence for genetic heterogeneity between clinical subtypes of bipolar disorder. Transl Psychiatry. 7:e993. 8. Coryell W, Endicott J, Reich T, Andreasen N, Keller M (1984): A family study of bipolar II disorder. Br J Psychiatry 145:49-54. 9. Heun R, Maier W (1993): The distinction of bipolar II disorder from bipolar I and recurrent unipolar depression: results of a controlled family study. Acta Psychiatr Scand. 87:279-284. 10. Endicott J, Nee J, Andreasen N, Clayton P, Keller M, Coryell W (1985): Bipolar II. Combine or keep separate? J Affect Disord. 8:17-28. 11. Gershon ES, Hamovit J, Guroff JJ, Dibble E, Leckman JF, Sceery W, et al. (1982): A family study of schizoaffective, bipolar I, bipolar II, unipolar, and normal control probands. Arch Gen Psychiatry. 39:1157-1167. 12. Vandeleur CL, Merikangas KR, Strippoli MP, Castelao E, Preisig M (2014): Specificity of psychosis, mania and major depression in a contemporary family study. Mol Psychiatry. 19:209-213. 13. Merikangas KR, Cui L, Heaton L, Nakamura E, Roca C, Ding J, et al. (2014): Independence of familial transmission of mania and depression: results of the NIMH family study of affective spectrum disorders. Mol Psychiatry. 19:214-219. 14. Tidemalm D, Haglund A, Karanti A, Landen M, Runeson B (2014): Attempted suicide in bipolar disorder: risk factors in a cohort of 6086 patients. PloS one. 9:e94097. 15. Karanti A, Kardell M, Lundberg U, Landen M (2016): Changes in mood stabilizer prescription patterns in bipolar disorder. J Affect Disord 195:50-56. 16. Sellgren C, Landen M, Lichtenstein P, Hultman CM, Langstrom N (2011): Validity of bipolar disorder hospital discharge diagnoses: file review and multiple register linkage in Sweden. Acta Psychiatr Scand. 124:447-453. 17. Lichtenstein P, Bjork C, Hultman CM, Scolnick E, Sklar P, Sullivan PF (2006): Recurrence risks for schizophrenia in a Swedish national cohort. Psychol Med. 36:1417-1425. 18. Inc. SI (2013): SAS Institute Inc. SAS® 9.4 Guide to Software Updates. Cary, NC: SAS Institute Inc. 19. StataCorp. (2013): Stata: Release 13. Statistical Software. College Station, TX: StataCorp LP. 20. Team RC (2015): R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria URL https://wwwR-projectorg/. 21. Benazzi F (2004): Factor structure of recalled DSM-IV hypomanic symptoms of bipolar II disorder. Compr Psychiatry. 45:441-446. Jie Song

20

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

22. Hantouche EG, Angst J, Akiskal HS (2003): Factor structure of hypomania: interrelationships with cyclothymia and the soft bipolar spectrum. J Affect Disord 73:39-47. 23. Angst J, Adolfsson R, Benazzi F, Gamma A, Hantouche E, Meyer TD, et al. (2005): The HCL-32: towards a self-assessment tool for hypomanic symptoms in outpatients. J Affect Disord. 88:217-233. 24. Dell'Osso B, Grancini B, Vismara M, De Cagna F, Maggi M, Molle M, et al. (2016): Age at onset in patients with bipolar I and II disorder: a comparison of large sample studies. J Affect Disord. 201:57-63. 25. Smoller JW, Finn CT (2003): Family, twin, and adoption studies of bipolar disorder. Am J Med Genet C Semin Med Genet. 123C:48-58. 26. Kao CF, Chen HW, Chen HC, Yang JH, Huang MC, Chiu YH, et al. (2016): Identification of Susceptible Loci and Enriched Pathways for Bipolar II Disorder Using Genome-Wide Association Studies. Int J Neuropsychopharmacol. 19(12): pyw064 27. Lichtenstein P, Yip BH, Bjork C, Pawitan Y, Cannon TD, Sullivan PF, et al. (2009): Common genetic determinants of schizophrenia and bipolar disorder in Swedish families: a population-based study. Lancet. 373:234-239. 28. Judd LL, Akiskal HS, Schettler PJ, Coryell W, Maser J, Rice JA, et al. (2003): The comparative clinical phenotype and long term longitudinal episode course of bipolar I and II: a clinical spectrum or distinct disorders? J Affect Disord. 73:19-32. 29. Peter HS, Brent MM, Phillip HW, Emily CB, Michele U (2005): Current Pathophysiological Findings in Bipolar Disorder and in its Subtypes. Curr Psychiatry Rev. 1:75-101. 30. Dunner DL (1993): A review of the diagnostic status of “Bipolar II” for the DSM-IV work group on mood disorders. Depression. 1:2-10. 31. Benazzi F (2007): Is there a continuity between bipolar and depressive disorders? Psychother Psychosom. 76:70-76. 32. Benazzi F (2006): Mood patterns and classification in bipolar disorder. Curr Opin Psychiatry. 19:1-8. 33. Phillips ML, Kupfer DJ (2013): Bipolar disorder diagnosis: challenges and future directions. Lancet. 381:1663-1671.

Jie Song

21

ACCEPTED MANUSCRIPT Table 1. Descriptive characteristics of patients with BDI and BDII

BDI

BDII

All BD patients in BipoläR

All BD patients in National Patient Register 65,644 40,577 (61.8)

AC C

EP

TE D

M AN U

SC

RI PT

N (%) 4,309 (33.6) 4,178 (32.6) 12,837 Female (%) 2,466 (46.9)* 2,797 (53.1) 7,997 (62.3) Calendar year of birth 1959 ± 16.3* 1965 ± 16.0 1962 ± 16.4 1953 ± 24.9 (mean ± SD) Calendar year of first 1999 ± 11* 2006 ± 7 2003 ± 10 1998 ± 13 diagnosis (mean) Total 39.2 ± 14.1* 40.5 ± 14.2 40.1 ± 14.2 45.0 ± 17.4 Age at first Born diagnosis after 24.9* 26.6 26.2 25.9 (mean ± SD) 1973 * P < .0001 in test of group difference between bipolar I and II disorder (t-test for continuous variable and χ2 test for binary varibale). Abbreviations: BDI, bipolar I disorder, BDII, bipolar II disorder; SD, standard deviation; BD, bipolar disorder

Jie Song

22

ACCEPTED MANUSCRIPT Figure 1. Odds ratios of BDI/BDII in relatives of patients with BDI/BDII

AC C

EP

TE D

M AN U

SC

RI PT

Abbreviations: BDI, bipolar I disorder; BDII, bipolar II disorder; OR, odds ratio; CI, confidence interval

Jie Song

23

ACCEPTED MANUSCRIPT Figure 2. Odds ratios of lifetime diagnosis of other psychiatric disorders in full-siblings of patients diagnosed with BDI/BDII

AC C

EP

TE D

M AN U

SC

RI PT

Abbreviations: BD, bipolar disorder; BDI, bipolar I disorder; BDII, bipolar II disorder; OR, odds ratio; CI, confidence interval

Jie Song

24

ACCEPTED MANUSCRIPT Table 2. Genetic correlations between other psychiatric disorders and BDI/BDII BDII Genetic correlation (95% CI)

Schizophrenia

0.34 (0.08-0.58)

-0.02 (-0.41-0.29)

Depression

0.25 (0.07-0.45)

0.49 (0.32-0.79)

Attention-deficit hyperactivity disorder

0.26 (0.07-0.48)

0.29 (0.13-0.46)

Eating disorders

0.24 (-0.03-0.51)

Autism spectrum disorders

0.24 (0.06-0.57)

Substance use disorders

0.23 (0.07-0.41)

Anxiety disorders

0.37 (0.22-0.61)

Personality disorders

0.45 (0.25-0.75)

RI PT

BDI Genetic correlation (95% CI)

0.28 (0.05-0.51)

0.25 (-0.10-0.45) 0.31 (0.18-0.45) 0.38 (0.25-0.55)

SC

Psychiatric disorder

0.28 (0.11-0.48)

AC C

EP

TE D

M AN U

Abbreviations: BDI, bipolar I disorder; BDII, bipolar II disorder; CI, confidence interval

Jie Song

ED

PT

M AN U

S

MANUSCRIPT OR of psychiatic disorders in ACCEPTED full siblings (95%CI)

Autism spectrum disorders

Substance use disorders

Anxiety disorders

Personality disorders

BDI: 1.7 (1.3 − 2.2)

SC

P = 0.07

BDII: 2.3 (1.9 − 2.9) BDI: 2.0 (1.4 − 2.7)

P = 0.14

BDII: 1.4 (1.0 − 2.0) BDI: 1.7 (1.1 − 2.6) BDII: 2.4 (1.8 − 3.1) BDI: 1.2 (1.1 − 1.4)

BDII: 1.5 (1.4 − 1.8) BDI: 1.6 (1.4 − 1.8)

BDII: 1.9 (1.7 − 2.1) BDI: 1.8 (1.5 − 2.2) BDII: 1.9 (1.6 − 2.3)

RI PT

P = 0.11

BDII: 2.1 (1.9 − 2.4)

M AN U

Eating disorders

BDI: 1.9 (1.7 − 2.1)

TE D

Attention−deficit/hyperactivity disorder

P = 0.003

BDII: 1.3 (0.9 − 1.9)

EP

Depression

BDI: 2.5 (2.0 − 3.2)

AC C

Schizophrenia

BD subtype in probands BDI

P = 0.20

BDII

P = 0.03

P = 0.03

P = 0.88 1.0

1.5

OR

2.0

3.0

Song et al.

Supplement

ACCEPTED MANUSCRIPT

Specificity in Etiology of Subtypes of Bipolar Disorder: Evidence From a Swedish Population-based Family Study Supplementary Information

Schizophreniaa Schizoaffective disordera Bipolar disorder

a

a

Depression Attention-deficit/hyperactivity disorder

ICD 8 (1969-1986) 295.0-295.4, 295.6, 295.8, 295.9 295.7

ICD 9 (1987-1996) 295A-295E, 295G, 295W, 295X 295H

ICD 10 (1997-)

296.0-296.3, 296.8, 296.9

296A-296E, 296W, 296X

F30, F31

296.2, 300.4

296B, 300E, 311

F32-F39

314

F90

-

SC

Psychiatric disorder

RI PT

Supplementary Table S1. Diagnosis of psychiatric disorders (ICD code in Swedish version)

F20

F25

F50.0-F50.3, F50.9 Autism spectrum disorders 299 F84 F10-F19 except Substance abuse disorders 303, 304 303, 304, 305A, 305X x.5 F40-F42, F44-F45, Anxiety disorders 300 except 300.4 300 except 300E F48 Personality disorders 301 301 F60-F62, F69 a The definitions of schizophrenia, schizoaffective disorder, bipolar disorder and depression require at least two inpatient or outpatient admissions. A hierarchical structure of diagnosis was used (Individuals with diagnosis of bipolar disorder but not schizophrenia were regarded as having bipolar disorder. Individuals with depression but neither schizophrenia nor bipolar disorder were coded as having depression). For these four disorders, individuals with admissions with ICD codes 295.5, 295F (latent schizophrenia) were excluded from those diagnosed with schizophrenia. 306.5, 784.0

307B, 307F

TE D

M AN U

Eating disorders

AC C

EP

Abbreviations: ICD, International Classification of Disease

1

Song et al.

Supplement

ACCEPTED MANUSCRIPT Supplemental Methods Familial aggregation and co-aggregation As an individual cannot be diagnosed with both BDI and BDII, the true BD diagnosis status of an individual is either (BDI=1, BDII=0), (BDI=0, BDII=1) or (BDI=0, BDII=0), where “0” stands for “no diagnosis” and “1” stands for “diagnosis”. However, as the National Patient Register lacks

RI PT

information on BD subtype, the measured BD diagnosis status is either (BDI=1, BDII=0), (BDI=0, BDII=1), (BDI=0, BDII=0) or (BDI=NA, BDII=NA), where NA stands for “missing”. The true diagnosis of subjects with (BDI=NA, BDII=NA) is either (BDI=1, BDII=0), (BDI=0, BDII=1) or (BDI=0, BDII=0). By restricting the analysis to those with available information on BD subtype, we alter the distribution of BDI and BDII in the sample. However, estimated odds ratios will still be

SC

unbiased; demonstrated here. Let 𝑌𝑌 be the indicator of a particular BD status (BDI or BDII) in the

person under consideration. Let 𝑋𝑋 be the indicator of a particular BD status (BDI or BDII) in the 𝑂𝑂𝑂𝑂 =

M AN U

person’s relative. The target parameter is the odds ratio

𝑝𝑝(𝑌𝑌 = 1, 𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 = 0, 𝑋𝑋 = 0) 𝑝𝑝(𝑌𝑌 = 0, 𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 = 1, 𝑋𝑋 = 0)

Let 𝑌𝑌 ∗ and 𝑋𝑋 ∗ be the indicators of whether 𝑌𝑌 and 𝑋𝑋 are observed, respectively. The estimated parameter is the observed odds ratio

=

𝑝𝑝(𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1|𝑌𝑌 = 1, 𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1|𝑌𝑌 = 0, 𝑋𝑋 = 0) × 𝑂𝑂𝑂𝑂 𝑝𝑝(𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1|𝑌𝑌 = 0, 𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1|𝑌𝑌 = 1, 𝑋𝑋 = 0)

𝑝𝑝(𝑌𝑌 ∗ = 1|𝑌𝑌 = 1)𝑝𝑝(𝑋𝑋 ∗ = 1|𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 ∗ = 1|𝑌𝑌 = 0)𝑝𝑝(𝑋𝑋 ∗ = 1|𝑋𝑋 = 0) × 𝑂𝑂𝑂𝑂 𝑝𝑝(𝑌𝑌 ∗ = 1|𝑌𝑌 = 0)𝑝𝑝(𝑋𝑋 ∗ = 1|𝑋𝑋 = 1)𝑝𝑝(𝑌𝑌 ∗ = 1|𝑌𝑌 = 1)𝑝𝑝(𝑋𝑋 ∗ = 1|𝑋𝑋 = 0)

EP

=

TE D

𝑝𝑝(𝑌𝑌 = 1, 𝑋𝑋 = 1|𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1)𝑝𝑝(𝑌𝑌 = 0, 𝑋𝑋 = 0|𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1) 𝑝𝑝(𝑌𝑌 = 0, 𝑋𝑋 = 1|𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1)𝑝𝑝(𝑌𝑌 = 1, 𝑋𝑋 = 0|𝑌𝑌 ∗ = 1, 𝑋𝑋 ∗ = 1)

= 𝑂𝑂𝑅𝑅

AC C

where the first equality follows from Bayes rule and the second equality follows from the facts that the missingness of BD status in two relatives are unrelated to each other (so that 𝑝𝑝(𝑌𝑌 ∗ , 𝑋𝑋 ∗ |𝑌𝑌, 𝑋𝑋) =

𝑝𝑝(𝑌𝑌 ∗ |𝑌𝑌, 𝑋𝑋)𝑝𝑝(𝑋𝑋 ∗ |𝑌𝑌, 𝑋𝑋)), and that missingness of BD status in one person is unrelated to the true BD

status of that person’s relative (so that 𝑝𝑝(𝑌𝑌 ∗ |𝑌𝑌, 𝑋𝑋)𝑝𝑝(𝑋𝑋 ∗ |𝑌𝑌, 𝑋𝑋) = 𝑝𝑝(𝑌𝑌 ∗ |𝑌𝑌)𝑝𝑝(𝑋𝑋 ∗ |𝑋𝑋)).

2

Song et al.

Supplement

ACCEPTED MANUSCRIPT Estimate of heritability and genetic correlation Univariate model to estimate heritability For these analyses, different prevalence of BDI and BDII were allowed in the five different types of siblings. The prevalence of sex effects were assumed to be identical between all sibling types and were adjusted for. The genetic relatedness was fixed to 1 for MZ twin pairs (they are genetically

RI PT

identical), to 0.5 for DZ twins and full siblings (they share on average 50% of their segregating genes), and to 0.25 for half-siblings. We assumed that the family environment is shared between MZ twins, DZ twins, full siblings and maternal half-siblings (the family environmental correlation was fixed to 1), but unshared between paternal half-siblings.

SC

We assumed no dominant genetic components contribute to the liability and that there are no epistasis or dominance deviations between genes and no interactions or correlations between genetic and environmental components. We decompose the variance into three components: A, additive genetic

M AN U

factors which represents linear effects of all alleles; C, shared environmental factors (between family members); and E, non-shared environmental factors.

Bivariate model to estimate genetic correlation between disorders

Unfortunately, only the full siblings had concordantly affected cross-sibling cross-subtype pairs. Therefore, we could only estimate two parameters (AE) rather than three parameters (ACE). However,

TE D

analyses using the univariate model showed that shared environment components (C) contributed very little to the phenotypic variance. Thus, only AE models were considered in subsequent bivariate models to estimate the genetic correlations. Because no individual can have both diagnoses, we performed the analyses by excluding contribution to the likelihood from the within-individual cross-

EP

disorder correlation. Since estimation of unshared environmental contributions to the overlap between the disorders relies on the within-individual cross-disorder correlation, we assumed no contribution of

AC C

unshared environment between disorders. When we applied a non-hierarchical structure of classification for diagnosis, we let the correlations between disorders to be estimated without additional assumptions of no contribution from unshared environmental factors. Three different optimizers, “NPSOL”, “SLSQP” and “CSOLNP”, were compared in the OpenMx package. We used “MxTryhard” and “MxTryhardOrdinal” syntax to check that the result remains stable.

3

Song et al.

Supplement

ACCEPTED MANUSCRIPT Supplementary Table S2. Odds ratios of BDI/BDII in first-degree relatives of patients with BDI/BDII (using different diagnosis criteria and study population)

Non-hierarchical diagnosisa

BDI BDII BDI BDII BDI BDII BDI BDII

Most recent diagnosisb Diagnosis using ICD codesc Cohort with inclusion of unspecified casesd

BDI in first-degree relative Test of OR (95% CI) difference 13.9 (12.0-16.2) P=0.005 9.3 (7.7-11.2) 16.0 (13.0-19.9) P=0.001 10.1 (8.3-12.3) 7.9 (7.2-8.8) P<0.001 4.3 (3.9-4.6) 15.9 (12.2-20.5) P=0.003 9.3 (7.3-11.8)

BDII in first-degree relative Test of OR (95% CI) difference 9.2 (7.6-11.1) P=0.05 12.3 (10.4-14.5) 10.0 (8.3-12.2) P=0.44 11.4 (8.8-14.7) 6.1 (5.7-6.5) P=0.002 7.1 (6.8-7.5) 9.2 (7.3-11.8) P=0.08 13.0 (9.7-17.3)

RI PT

BD subtype in probands

SC

a. Non-hierarchical diagnosis allows one individual to be diagnosed with both BDI and BDII, including 5981 BDI and 5868 BDII, among which 1195 were with mixed diagnosis from the quality register.

M AN U

b. Most recent diagnosis defined BDI and BDII with each person’s most recent diagnosis during the follow-up, including 5313 BDI and 5175 BDII patients. c. The presence and absence of manic episodes to identify BDI and BDII patients from the National Patient Register (BDI: ICD-9: 296A, 296C; ICD-10: F30.1, F30.2, F31.1, F31.2; BDII: ICD-10: F30.0, F31.0, F31.3, F31.4, F31.8 and not fulfill the diagnosis of BDI). By adding the patients of BDI and BDII from the quality register, we finally identified 16, 413 patients with BDI and 25,635 patients of BDII.

AC C

EP

TE D

d. The analyses were performed including patients with unclear BD subtype status from the National Patients Register. For example, for familial aggregation of BDI, the analysis was directly estimated the odds of BDI in a relative, given BDI vs. no BDI in the proband.

4

Song et al.

Supplement

ACCEPTED MANUSCRIPT Supplementary Table S3. Odds ratios of BDI/BDII in first-degree relatives of patients with BDI/BDII (stratified by sex)

Male

AC C

EP

TE D

M AN U

SC

Female

BDI BDII Test of difference BDI BDII Test of difference

RI PT

BD subtype and sex in probands

Odds ratio (95% CI) of BD subtype in first-degree relatives Male Female BDI BDII BDI BDII 17.8 (10.1-31.5) 11.7 (6.7-20.7) 16.6 (11.5-24.1) 8.6 (5.4-13.8) 11.9 (6.8-21.0) 20.8 (10.3-41.8) 10.2 (6.0-17.2) 12.3 (7.7-19.9) 0.33 0.21 0.13 0.29 16.4 (11.3-23.8) 10.1 (6.0-17.1) 17.2 (11.2-26.5) 9.8 (6.6-14.5) 8.6 (5.3-13.8) 12.5 (7.8-20.1) 9.8 (6.7-14.6) 13.2 (8.6-20.2) 0.03 0.55 0.05 0.32

5