Study design
The present study is based on cross-sectional baseline data from the large-scale ongoing longitudinal clinical study, the Bipolar Illness Onset study (the BIO-study), initiated in 2015 [16].
Included participants
Patients newly diagnosed with bipolar disorder
We invited patients aged 18–70 at the Copenhagen Affective Disorder Clinic at Psychiatric Centre Copenhagen. The Centre covers all psychiatric centres in the Capital Region of Denmark and provides assessments of and treatment for patients newly diagnosed with BD in the catchment area. Patients are assessed and diagnosed by specialists in psychiatry according to the ICD-10 and DSM criteria upon referral to the Copenhagen Affective Disorder Clinic. Upon referral to the Copenhagen Affective Disorder Clinic, patients were routinely invited to participate in the BIO-study if a diagnosis of BD was set within the preceding two years, as this was chosen as the definition of being “newly diagnosed”. Patients aged 15–18 were recruited from the highly specialized Bipolar Team of the Child and Adolescent Mental Health Center, Capital Region of Denmark.
Unaffected first-degree relatives
If included patients consented, we invited their eligible siblings and children to participate in the study. The inclusion criteria were age 15–70 years and having a full sibling or a parent newly diagnosed with BD participating in the BIO-study. Exclusion criteria were being diagnosed with BD or schizophrenia.
Healthy control individuals
Healthy control individuals were included from the blood bank at Rigshospitalet in Copenhagen, Denmark. Random blood donors were approached on random days upon blood donation and invited to participate in the study. Inclusion criterion was age 15–70 years. Exclusion criteria were a current or previous medically treated psychiatric disorder in the subject or a first-degree relative.
Diagnostic and clinical assessments
Participants were included from June 2015 to January 2021. Following informed consent for participation, a PhD student (medical doctor or a psychologist) confirmed the BD diagnosis using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) at inclusion in the study [17]. Similarly, UR and HC underwent a SCAN interview. Clinical assessments for depressive and manic symptoms were performed using the Hamilton Depression Scale-17 (HDRS-17) and the Young Mania Rating Scale (YMRS) [18, 19].
Assessment of familial load of psychiatric disorders
Using a translated and modified Danish version of Family History Research Diagnostic Criteria (FH-RDC) [20], each participant was systematically interviewed about the presence of any current or previous psychiatric disorder in their first-degree relatives to the best of their knowledge. Information on parents, siblings and offspring were registered in six categories (0 = no psychiatric disorder, 1 = depression and/or suicide, 2 = BD, 3 = affective disorder not specified, 4 = schizophrenia, 5 = alcohol abuse, 6 = other psychiatric disorder (e.g., anxiety, autism or ADHD). Additionally, information on consummated suicides in first-degree relatives was registered. If a participant was uncertain about family members’ psychiatric symptoms or disorders, we conservatively noted the status of that family member as “0”, no psychiatric disorder. In cases where the patient was not genetically related to their first-degree relatives, e.g., due to adoption, the FH-RDC interview was not performed.
Estimating familial load (FL)
Different approaches have previously been used to analyze FL. Some studies have listed the number of family members with psychiatric disorders categorically, either as an ordinal categorical variable (number of diseases family members: 0, 1, 2, 3, 4+) [15] or a binary categorical variable (diseased family members: “yes”/“no”) [12, 14]. In contrast, other studies have argued that analyzing FL categorically might not capture the effect of FL on psychiatric disorders [21, 22]. It has therefore been suggested that a continuous “familial loading score” should be estimated and used for analyses of FL [23]. To ensure that we did not overlook a true association, we investigated FL in two ways: as a binary categorical and as a continuous variable.
Familial load as a categorical variable
Familial load was analyzed as a binary categorical variable. In patients, we investigated if having ≥ 1 first-degree relative with a psychiatric disorder was associated with impaired functioning compared with patients with no first-degree relative with a psychiatric disorder.
All included UR had at least one first-degree relative with BD, this being the proband patient. Therefore, in the UR group, we investigated if having ≥ 2 first-degree relatives with psychiatric disorders was associated with impaired functioning compared with UR, with only 1 first-degree relative with a psychiatric disorder.
Familial load as a continuous variable
Additionally, FL was analyzed as a continuous variable, measuring the total load of psychiatric disorders among first-degree relatives. We used the Family Liability Index (FLI), which has also been used to analyze FL in high-risk individuals [24]:
$${\rm{Family\, Liability\, Index }}({\rm{FLI}}){\rm{ }} = \frac{{{\rm{BM}} + {\rm{BF}} + \sum {\rm{BS }} + {\rm{ }}(\sum {\rm{HS}} \cdot 0.5)}}{{2 + n\;{\rm{BS}} + (n\;{\rm{HS}} \cdot 0.5)}}{\rm{ }}.$$
The Family Liability Index (FLI) reflects the aggregation of psychiatric disorders among a participant’s first-degree relatives. The FLI is estimated based on the presence of psychiatric disorders in the biological mother (BM), the biological father (BF), biological siblings (BS) and biological half-siblings (HS). The index considers that half-siblings share only half the genetic information compared to parents and full siblings. The estimate indicates the load of diseased first-degree relatives in the numerator compared to the number of first-degree relatives (the family size) in the denominator. The FLI continuously assumes values from 0 to 1, with 0 reflecting “low familial load” and 1 reflecting “high familial load”.
Assessment of functioning
Functioning may be measured and estimated in various ways. In this study, we investigated functioning in three ways, as presented in the following.
The Functioning Assessment Short Test (FAST)
The FAST is a clinical observer-based interview to assess the participants’ overall functioning in the previous two weeks. It covers the subdomains of autonomy, occupation, cognition, financial issues, interpersonal relationships and leisure time. Total scores range from 0 to 72, the higher score, the more significant functional impairment [25]. FAST total scores between 0 and 11 indicate no impairment, 12–20 indicate mild impairment, 21–40 indicate moderate impairment and FAST total scores above 40 indicate severely impaired functioning [26].
The work and Social Adjustment Scale (WSAS)
The WSAS is a self-reported questionnaire in five subitems. The questionnaire covers work ability, practical housework, participating in social activities, having meaningful leisure time, and engaging in social relations. Participants rate their everyday life functioning from 0 to 8 on each item. Total scores range from 0 to 40, the higher score, the more significant impairment in self-reported functioning. The WSAS is sensitive in patients with BD and individuals without psychiatric disorders [27,28,29].
Socio-economic status (SES)
At baseline assessment, all participants reported on the following six domains of SES:
-
(1)
Educational achievement; measured continuously in total years of education.
-
(2)
Employment status; measured as “employed, student, pension or other” vs. “unemployed or disabled”.
-
(3)
Work ability; measured as “not on sick-leave” vs. “on sick-leave”.
-
(4)
Relationship status; measured as “being in a relationship” vs. “not in a relationship”.
-
(5)
Cohabitation status; measured as “living with someone in terms of shared address” vs. “living alone”.
-
(6)
Marital status; measured as “married, divorced, separated or widowed” vs. “never married”.
These six socio-economic domains were adopted from Statistics Denmark as proxies of overall SES [30] and have been used previously [4].
Statistical analyses
Descriptive data were analyzed to test assumptions of normal distribution. Continuous outcomes were analyzed in the student’s t-test, and categorical outcomes were analyzed using chi-square tests for pairwise comparisons of BD vs. HC and UR vs. HC, respectively. The correlations between FLI and FAST, WSAS and educational length were explored using 2-tailed Spearman’s correlation tests. The associations between FL and functioning were analyzed with multiple regression models with the continuous outcomes (FAST (total score and subdomains), WSAS and educational achievement and with binary logistic regression models with categorical outcomes (employment status, work ability, relationship, cohabitation and marital status). We performed analyses unadjusted (model 1) and adjusted for age, sex, HDRS-17 and YMRS (model 2). All analyses were performed separately on all outcomes on patients with BD and on UR. As multiple analyses were conducted, a Bonferroni correction for multiple testing was applied and the adjusted significance level was p < 0.00625. The Statistical Package for Social Sciences (SPSS Statistics 25) was used, and all model assumptions were met.