Polygenic scores for psychiatric disorders associate with year of first bipolar disorder diagnosis: A register-based study between 1972 and 2016

The diagnostic criteria of bipolar disorder (BD) have changed over time. To test if these changes are reflected in the polygenic profile in BD, we studied the association between first BD diagnosis year (during 1972 – 2016) and polygenic scores (PGS) for psychiatric disorders in BD patients ( N = 3,818). We found significant associations between diagnosis year and PGS for BD, depression, and attention deficit hyperactivity disorder (ADHD). The PGS remained largely stable over time in BD type 1, while changes were observed in BD type 2. These findings bear significance not only for genetic research but also for clinical practise, as shifts in patient characteristics can influence treatment response.


Introduction
Multiple studies have demonstrated the significant heritability of bipolar disorder (BD) (Song et al., 2015;Lichtenstein et al. 2009).Genome-wide association studies (GWAS) have in recent years unveiled its polygenic architecture, revealing a complex interplay of genetic factors contributing to BD susceptibility (Mullins et al. 2021).Polygenic scores (PGS) have emerged as a tool to summarize polygenic effects of categorical disorders such as BD (Martin et al. 2019).However, the results from GWAS are contingent upon the case definition.Changes in diagnostic criteria, or shifts in how these criteria are applied over time, not only influence the amount of variance explained by the PGS for a specific diagnosis, but also alter the phenotype most closely associated with a PGS.Conversely, examining changes in PGS over time for a diagnosis could offer insights into historical trends in diagnostic or clinical practice.
The modern conception of BD traces back to Emil Kraepelin's conceptualization of manic-depressive illness at the beginning of the twentieth century, but the definition has undergone changes since then.After first separating unipolar depression from bipolar disorders, the diverse clinical presentations of BD then prompted a splitting into two main BD subtypes in the DSM-IV (1994): BD type 1 (BD-1) presents with manic and often depression episodes, and BD type 2 (BD-2) that features less severe elated episodes (hypomania) alongside depressive episodes.In addition to these formal changes of diagnostic criteria, the interpretation and clinical thresholds of psychiatric diagnostic criteria are subject to variation over time (Angst and Marneros 2001).Together, this can lead to shifts not only in prevalence but also in characteristics of individuals diagnosed with BD.
The prevalence of individuals diagnosed with BD has shown an upward trend in Sweden since the 1990s (Carlborg et al. 2015).One possibility is that greater awareness has led to more accurate diagnosis and treatment.Alternatively, the increased prevalence may be due to a less stringent application of diagnostic criteria, encompassing cases with less severe illness, or divergent types of mental health problems, under the BD diagnosis (Akiskal et al. 2000).
The aim of this study was to investigate whether there have been changes over time in the polygenic traits of patients who have been diagnosed with BD.To this end, we investigated PGS for BD, schizophrenia (SCZ), major depressive disorder (MDD), and attention deficit hyperactivity disorder (ADHD) in individuals who received their first BD diagnosis between 1972 and 2016.Additionally, we examined whether similar trends were evident within subtypes of BD.

Population, phenotype measurements and polygenic scoring
We included people diagnosed with BD in the Swedish Bipolar Collection (SWEBIC) and the Predictors For ECT (PREFECT) study.The study participants were recruited between 2009 and 2013 in SWEBIC, and between 2013 and 2017 in PREFECT.Detailed descriptions of the SWEBIC (Song et al. 2024) and PREFECT (Clements et al. 2021) studies are found in the supplementary methods.To determine individuals' first inpatient diagnosis of BD, we used the year when patients received the first recorded inpatient BD diagnosis according to ICD-8/9/10 in the Swedish Hospital discharge register (HDR) up to December 31, 2016.The full list of ICD codes is found in table S1.A total of 3818 subjects had information regarding the first year of inpatient diagnosis in SWEBIC (N = 3277) and PREFECT (N = 541), including in total 1913 subjects with BD-1, 973 subjects with BD-2, and 800 subjects with BD-NOS (Table S2).

Statistical analyses
We conducted linear regression association analyses between the diagnosis year and PGS using R (v.4.2.3).Linear regression analyses were adjusted for age at diagnosis, sex, and the first six ancestry principal components.We meta-analysed the results from the genotyping batches using metafor package in R (rma.uni and method="FE").We used the Bonferroni method to correct for five PGSs (P < 0.010).For sensitivity analyses, we conducted linear regression meta-analyses 1) using year of diagnosis grouped into 5-year bins (Table S3) and 2) within each BD subtype including BD-1, BD-2, and not otherwise specified (BD-NOS).In the BD subtype meta-analyses, the Affymetrix 6.0 genotyping batch was not included in the BD-2 and BD-NOS meta-analyses due to limited number of BD-2 (N = 28) and BD-NOS (N = 4) in this batch.

Discussion
We investigated if the polygenic profile of individuals diagnosed with BD has changed since the 1970′s.We found that the mean PGS for BD decreased over time whereas the mean PGS for MDD and ADHD increased.This means that the genetic predisposition towards bipolar disorder has shown a declining trend over the years, presumably reflecting shifts in diagnostic practices, while the genetic susceptibility to major depressive disorder and attention deficit hyperactivity disorder has been on the rise.When investigating polygenic changes stratified by BD subtypes, we found that the alterations in polygenic profiles varied by BD subtype: there were significant changes in the BD-2 group, but attenuated effect sizes in the BD-1 group.The relatively stable polygenic profile of BD-1 reflects the largely unchanged definition of this subtype (Larsson et al. 2021).
Within the BD-2 group, the MDD-PGS was higher at the end of the studied period.This finding is consistent with the notion put forward by prominent researchers that some patients diagnosed with depression should instead be considered as part of the bipolar spectrum (Angst et al. 2003;Ghaemi et al. 2002).This argument is based on the observation that some patients diagnosed with MDD also exhibit symptoms traditionally associated with BD, such as periods with elevated mood, irritability, and increased energy.Proponents of the bipolar spectrum view have therefore reasoned that the diagnostic criteria for BD are too narrow (Akiskal et al. 2000;Angst et al. 2003).Our findings suggest that clinicians have heeded these arguments by including individuals in the bipolar spectrum who would have previously received a diagnosis of MDD.
The PGS for ADHD also increased over the study period.One possible explanation is the comorbidity and overlap in symptoms between BD and ADHD (Schiweck et al. 2021), which may lead to mis-or over-diagnosis of BD in individuals with ADHD.For example, both ADHD and BD can present with impulsivity, distractibility, and mood instability.In addition, the prevalence of ADHD diagnosis has increased during the last decades (Atladottir et al. 2015), which could also have an impact on the comorbidity between the disorders.
Our findings have important clinical and scientifically implications.They first suggest that the increased prevalence of BD might not solely be explained by increased awareness of the disorder, but also by a broadened and changed profile of individuals who are diagnosed as suffering from BD with the non-prototypic cases aggregated in the BD-2 group.Second, shifts in patient characteristics may influence treatment response.For instance, if the polygenic profile in BD changes over time-as here reflected by for example increased MDD-PGS-this may affect the apparent effectiveness of lithium since MDD-PGS is also associated with worse lithium response (Amare et al. 2021;Schubert et al. 2021).Third, the results highlight that genetic studies will yield different results depending on when the diagnosis was made.This has important implications for the interpretation of genetic studies in the field of psychiatry.It suggests that genetic studies conducted at different time points may yield different results, emphasizing the importance of considering historical context in the interpretation of genetic data in psychiatric research.

Strengths and limitations
The strength of this study is a large cohort of BD with full coverage of psychiatric hospital discharge diagnoses since the 1970′s.An important limitation is that study participants had to be alive during the ascertainment period 2009-2013 and BD is associated with premature death (Pompili et al. 2013).If individuals who died before the study inclusion differ in their polygenic traits from those included, the results may not be fully representative.A second limitation is that patients diagnosed with BD for the first time during the earlier decades of the study period might have a more stable BD diagnosis by virtue of still having the diagnosis during the ascertainment period of 2009-2013 compared to those diagnosed later.Our analyses within the BD-1 subgroup, which only demonstrated a significant decline of PGS for BD-1 over time, could potentially be attributed to this.Finally, we only capture inpatient diagnoses, while many patients may have received an outpatient diagnosis earlier.
Taken together, we found that PGS for psychiatric disorders are stable over time within the BD-1 subgroup, but not within the BD-2 or BD-NOS subgroups.This likely reflects secular changes in the diagnostic criteria for BD-2 and BD-NOS, as opposed to BD-1.Furthermore, it likely mirrors shifts in clinical practice towards broader inclusion of patients who may have previously been diagnosed with MDD in the BD spectrum.These individuals are more likely to receive BD-2 or BD-NOS subdiagnoses rather than a BD-1 diagnosis.

Data availability
The data supporting findings in this study cannot be shared due to Swedish law that prohibits sharing of national register data to a third party.

Declaration of competing interest
ML declares that he has received lecture honoraria from Lundbeck pharmaceutical outside the scope of this work.EJ is currently employed at IQVIA, Mölndal, Sweden.