Introduction
Schizophrenia is among the most severe and debilitating psychiatric disorders, significantly impacting the quality of life and general functioning (Dong et al., Reference Dong, Lu, Zhang, Zhang, Ng, Ungvari, Li, Meng, Wang and Xiang2019; Green, Kern, & Heaton, Reference Green, Kern and Heaton2004; Holm, Taipale, Tanskanen, Tiihonen, & Mitterdorfer-Rutz, Reference Holm, Taipale, Tanskanen, Tiihonen and Mitterdorfer-Rutz2021; Saarni et al., Reference Saarni, Viertiö, Perälä, Koskinen, Lönnqvist and Suvisaari2010). It is characterized by a combination of neuropsychological mechanisms that result in disturbed mental processes (Palmer, Dawes, & Heaton, Reference Palmer, Dawes and Heaton2009). Cognitive impairment is one of the most robust endophenotypes in schizophrenia and is also common in patients with other psychotic disorders (including schizoaffective disorder, bipolar disorder, and psychotic depression) (Barch & Ceaser, Reference Barch and Ceaser2012; Green & Harvey, Reference Green and Harvey2014; Lepage, Bodnar, & Bowie, Reference Lepage, Bodnar and Bowie2014; Li et al., Reference Li, Zhou, Zhang, Ng, Ungvari, Li and Xiang2020).
Patients with psychotic disorders often experience sleep problems, including insomnia and hypersomnia symptoms (Freeman, Sheaves, Waite, Harvey, & Harrison, Reference Freeman, Sheaves, Waite, Harvey and Harrison2020; Krystal, Reference Krystal2012; Reeve, Sheaves, & Freeman, Reference Reeve, Sheaves and Freeman2019, Reference Reeve, Sheaves and Freeman2021). In our previous study, patients with affective psychotic disorders exhibited more insomnia symptoms than patients with schizophrenia, who more often experienced excessively long sleep (37% of patients with schizophrenia, compared with 24% of patients with psychotic depression, as derived from previously published data). These sleep problems are associated strongly with worse subjective health (Cederlöf et al., Reference Cederlöf, Holm, Lähteenvuo, Haaki, Hietala, Häkkinen, Isometsä, Jukuri, Kajanne, Kampman, Kieseppä, Lahdensuo, Lönnqvist, Männynsalo, Niemi-Pynttäri, Suokas, Suvisaari, Tiihonen, Turunen and Paunio2022). Both sleep problems and cognitive impairments have been linked to negative outcomes, including a worse quality of life (DeRosse, Nitzburg, Blair, & Malhotra, Reference DeRosse, Nitzburg, Blair and Malhotra2018; Ritsner, Kurs, Ponizovsky, & Hadjez, Reference Ritsner, Kurs, Ponizovsky and Hadjez2004).
Genome-wide association studies (GWAS) have identified multiple risk variants for schizophrenia (Trubetskoy et al., Reference Trubetskoy, Pardiñas, Qi, Panagiotaropoulou, Awasthi, Bigdeli, Bryois and O’Donovan2022) and other psychiatric disorders and traits, leading to the development of polygenic scores (PGSs). PGS for schizophrenia (PGSSZ) has a considerable explanatory effect for schizophrenia and weaker but significant associations also with schizoaffective disorder and bipolar disorder type 1 (Allardyce et al., Reference Allardyce, Leonenko, Hamshere, Pardiñas, Forty, Knott, Gordon-Smith, Porteous, Haywood, Di Florio, Jones, McIntosh, Owen, Holmans, Walters, Craddock, Jones, O’Donovan and Escott-Price2018; Sullivan et al., Reference Sullivan, Agrawal, Bulik, Andreassen, Børglum, Breen, Cichon, Edenberg, Faraone, Gelernter, Mathews, Nievergelt, Smoller, O’Donovan, Daly, Gill, Kelsoe, Koenen, Levinson and Sklar2018). A high PGSSZ has in patients with schizophrenia been previously associated with poorer quality of life (Pazoki et al., 2020), more involuntary hospitalizations (Meier et al., Reference Meier, Agerbo, Maier, Pedersen, Lang, Grove, Hollegaard, Demontis, Trabjerg, HjorthØj, Ripke, Degenhardt, Nöthen, Rujescu, Maier, Werge, Mors, Hougaard, BØrglum and Mattheisen2016), and clozapine use (Lin et al., Reference Lin, Pinzón-Espinosa, Blouzard, Van Der Horst, Okhuijsen-Pfeifer, Van Eijk, Guloksuz, Peyrot and Luykx2023). Poor cognitive functioning has been associated with PGSSZ in the general population and first-episode psychosis cohorts, but not in schizophrenia(Jonas et al., Reference Jonas, Lencz, Li, Malhotra, Perlman, Fochtmann, Bromet and Kotov2019; Mallet, Strat, Dubertret, & Gorwood, Reference Mallet, Strat, Dubertret and Gorwood2020; Richards et al., Reference Richards, Pardiñas, Frizzati, Tansey, Lynham, Holmans, Legge, Savage, Agartz, Andreassen, Blokland, Corvin, Cosgrove, Degenhardt, Djurovic, Espeseth, Ferraro, Gayer-Anderson, Giegling and Walters2020). Studies on associations between PGSSZ and sleep in adulthood are scarce. One study with a general population sample, including a subsample of patients with schizophrenia, found associations between PGSSZ and early morning awakenings, reduced sleep efficiency, and increased napping (Wainberg et al., Reference Wainberg, Jones, Beaupre, Hill, Felsky, Rivas, Lim, Ollila, Hanna and Tripathy2021).
Well-established PGSs for sleep traits include those for insomnia (PGSINS) (Lane et al., Reference Lane, Liang, Vlasac, Anderson, Bechtold, Bowden, Emsley, Gill, Little, Luik, Loudon, Scheer, Purcell, Shaun, Kyle, Lawlor, Zhu, Redline, Ray, Rutter and Saxena2017) and sleep duration (PGSSD) (Dashti, Redline, & Saxena, Reference Dashti, Redline and Saxena2019), and for diurnal preference (PGSME) (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019; Wray et al., Reference Wray, Lee, Mehta, Vinkhuyzen, Dudbridge and Middeldorp2014; Zhang, Privé, Vilhjálmsson, & Speed, Reference Zhang, Privé, Vilhjálmsson and Speed2021). PGSs for short (PGSSS) and long sleep duration (PGSLS) have also recently been created (Austin-Zimmerman et al., Reference Austin-Zimmerman, Levey, Giannakopoulou, Deak, Galimberti, Adhikari, Zhou, Denaxas, Irizar, Kuchenbaecker, McQuillin, Concato, Buysse, Gaziano, Gottlieb, Polimanti, Stein, Bramon and Gelernter2023). These PGSs have been derived from self-report data in the general population, in whom they have been shown to have a relatively modest impact on sleep-related measures. Sleep PGSs have in genetic correlation studies been found to correlate with various psychiatric disorders and traits, including both affective disorders and PGSSZ. (; Dashti, Redline, & Saxena, Reference Dashti, Redline and Saxena2019; Jansen et al., Reference Jansen, Watanabe, Stringer, Skene, Bryois, Hammerschlag, de Leeuw, Benjamins, Muñoz-Manchado, Nagel, Savage, Tiemeier, White, Agee, Alipanahi, Auton, Bell, Bryc, Elson and Posthuma2019; Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019; O’Connell et al., Reference O’Connell, Frei, Bahrami, Smeland, Bettella, Cheng, Chu, Hindley, Lin, Shadrin, Barrett, Lagerberg, Steen, Dale, Djurovic and Andreassen2021). Few studies on sleep or diurnal preference PGSs in psychiatric disorders have been done. In one study, high PGSSD was not associated with schizophrenia or bipolar disorder (Dashti et al., Reference Dashti, Jones, Wood, Lane, van Hees, Wang, Rhodes, Song, Patel, Anderson, Beaumont, Bechtold, Bowden, Cade, Garaulet, Kyle, Little, Loudon, Luik and Saxena2019).
In this study, we aimed to compare genetic influences on objective measures of disease severity and subjective measures related to life quality and sleep in patients with psychotic disorders in a systematically collected nationwide sample. Specifically, we aimed to clarify the relationship between PGSs of sleep, diurnal traits, and schizophrenia, with (1) self-reported measures on sleep, subjective health, and cognitive functioning and (2) objective measures including involuntary hospitalizations and cognitive performance) in a large sample of patients with psychotic disorders. We hypothesized that genetic predisposition for sleep and diurnal traits are primarily related to subjective measures, while that for schizophrenia is mainly related to objective measures. Finally, to potentially shed light on the mechanisms of sleep problems in psychotic disorders, we (3) explored differences in these PGSs in non-affective and affective psychotic disorders, as compared with individuals with no psychiatric disorders from the general population.
Materials and methods
Study sample
This study is part of the SUPER research project, which examines psychotic disorders. The SUPER project is part of the international Stanley Global Neuropsychiatric Genomics Initiative. In Finland, the Institute for Molecular Medicine Finland (FIMM), the Finnish Institute of Health and Welfare (THL), and the University of Helsinki oversaw the research project. The project was carried out in cooperation with all hospital districts in Finland.
SUPER cohort
Adult Finnish patients with schizophrenia spectrum disorders (ICD-10 codes: F20–F29), bipolar disorder, or psychotic depression (ICD-10 codes: F30.2, F31, F32.3, and F33.3) with a history of at least one psychotic episode were eligible to participate. The diagnoses were retrieved from the Finnish Care Register for Health Care (HILMO). Finnish registry data have been shown to be of high-quality, including for mental health diagnoses (Sund, Reference Sund2012). In the present study, patients with (1) schizophrenia, (2) schizoaffective disorder, (3) bipolar disorder, or (4) psychotic depression were included. These diagnoses were categorized into diagnostic groups of (1) schizophrenia and (2) affective psychotic disorders, including schizoaffective disorder, bipolar disorder, and psychotic depression. The questionnaire, interview, cognitive tests, and genetic sampling were all conducted during the same session with a study nurse. The complete study protocol and further information about the study population were published previously (Ahola-Olli et al., Reference Lähteenvuo, Ahola-Olli, Suokas, Holm, Misiewicz, Jukuri, Männynsalo, Wegelius, Haaki, Kajanne, Kyttälä, Tuulio-Henriksson, Lahdensuo, Häkkinen, Hietala, Paunio, Niemi-Pynttäri, Kieseppä, Veijola, Lönnqvist and Palotie2023).
The total sample size was 10,411 patients. Patients with an unknown or other diagnosis than the four diagnostic groups, those who did not complete the questionnaire, were older than 80 years (or no age registered), or had no or subpar genotypic information, were excluded. After exclusion 8,232 patients remained. For the study sample flow chart, see Supplementary Figure 1.
PGS calculation
All PGSs were based on PRS-CS (Ge, Chen, Ni, Feng, & Smoller, Reference Ge, Chen, Ni, Feng and Smoller2019), using the largest publicly available summary statistics for each trait. These included the PGS for insomnia (Lane et al., Reference Lane, Liang, Vlasac, Anderson, Bechtold, Bowden, Emsley, Gill, Little, Luik, Loudon, Scheer, Purcell, Shaun, Kyle, Lawlor, Zhu, Redline, Ray, Rutter and Saxena2017), sleep duration (Dashti, Redline, & Saxena, Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019), schizophrenia (Trubetskoy et al., Reference Trubetskoy, Pardiñas, Qi, Panagiotaropoulou, Awasthi, Bigdeli, Bryois and O’Donovan2022), as well as the PGSs for long and short sleep duration (Austin-Zimmerman et al., Reference Austin-Zimmerman, Levey, Giannakopoulou, Deak, Galimberti, Adhikari, Zhou, Denaxas, Irizar, Kuchenbaecker, McQuillin, Concato, Buysse, Gaziano, Gottlieb, Polimanti, Stein, Bramon and Gelernter2023). The PGS for eveningness was derived from a PGS originally developed for morningness (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019). The entire genome was used, meaning all reliably imputed SNPs that intersected with each trait’s GWAS summary statistics were included. The Illumina Global Screening Array was used to genotype all participants. Samples where the inferred sex was mismatched with the recorded sex were excluded, as were samples with poor genotype quality. Imputation was performed using a Finnish-specific reference panel SISu version 3 (Pärn et al., Reference Pärnn.d.) and only reliably imputed variants (INFO score > 0.8) were kept. Ancestral outliers (N = 71, as defined by being >5 standard deviations from European samples in principal components 1 and 2), were not excluded. The LASER suit was used to infer ancestry (Taliun et al., Reference Taliun, Chothani, Schönherr, Forer, Boehnke, Abecasis and Wang2017).
Measures
Questionnaire-based measures for sleep, well-being, and subjective cognitive functioning
The sleep questions assessed total sleep duration, from which long and short SD were calculated, difficulties initiating sleep (DIS), early morning awakenings (EMAs), fatigue (FAT), and poor sleep quality (poor SQ) (for questions and response categories, see Table 1) (Cederlöf et al., Reference Cederlöf, Holm, Lähteenvuo, Haaki, Hietala, Häkkinen, Isometsä, Jukuri, Kajanne, Kampman, Kieseppä, Lahdensuo, Lönnqvist, Männynsalo, Niemi-Pynttäri, Suokas, Suvisaari, Tiihonen, Turunen and Paunio2022). The questions were based on Finnish general population studies (Aromaa & Koskinen, Reference Aromaa and Koskinen2004; Heistaro, Reference Heistaro2008; Partinen & Gislason, Reference Partinen and Gislason1995), and for subjective well-being, the measures included subjective health (Heistaro, Reference Heistaro2008), health-related quality of life (assessed with EQ-5D-3L, hereafter referred to as EQ5D) (Devlin & Krabbe, Reference Devlin and Krabbe2013; Saarni et al., Reference Saarni, Härkänen, Sintonen, Suvisaari, Koskinen, Aromaa and Lönnqvist2006), psychological distress (assessed with the Mental Health Inventory–5 [MHI-5], according to RAND scoring instructions) (Aalto, Aro, Aro, Mähönen, & Aro, Reference Aalto, Aro, Aro, Mähönen and Aro1995; Berwick et al., Reference Berwick, Murphy, Goldman, Ware, Barsky and Weinstein1991). Questions of subjective cognitive functioning included questions on concentration, learning, and memory (Heistaro, Reference Heistaro2008; Koponen, Borodulin, Lundqvist, Sääksjärvi, & Koskinen, Reference Koponen, Borodulin, Lundqvist, Sääksjärvi and Koskinen2018). Poor subjective cognitive functioning was defined as giving the response poor or very poor to any of the three questions, a dummy variable previously used in Finnish population-based studies (Koponen et al., Reference Koponen, Borodulin, Lundqvist, Sääksjärvi and Koskinen2018).
Questions used in the study
Assessment of objective cognitive functioning
Two tests were used: the Reaction Time (RTI) and the Paired Associative Learning (PAL) task, both from the Cambridge Neuropsychological Test Automated Battery (CANTAB). Five-choice serial reaction time task (5-CRTT) measured RTI to an unpredictable stimulus and hence general alertness and processing speed, while PAL measured visual and episodic learning and memory (CANTAB).
Our outcome for the RTI test was the median five-choice reaction time (RTIFMDRT), that is the median latency of response. The variable was calculated for correct responses where the stimulus could appear in any of the five locations. In the PAL test, our outcome was the Total Errors Adjusted (PALTEA), which reflects how quickly the participant learns when the participant has multiple attempts at each problem.
Interview
Marital status (marriage, common-law marriage, registered partnership), living status (unsupported versus supported housing), participation in the workforce (full- or part-time work, or student status), and education (register data obtained when interview data was unavailable) were asked in the interview.
Clozapine use was applied as an indicator of disease severity, as it is typically used in treatment-resistant psychotic disorders (Elkis & Buckley, Reference Elkis and Buckley2016). Information on medications was retrieved from a question in the interview: ‘What medications do you use regularly?’ The names of the medicines were checked on the label or the prescription form unless the interviewee remembered them.
FinnGen cohort
The FinnGen cohort data was used for analyzing the distribution of PGSs across the diagnostic groups and the general population. FinnGen is a public-private partnership, which combines genome information and digital healthcare data (Kurki et al., Reference Kurki, Karjalainen, Palta, Sipilä, Kristiansson, Donner, Reeve, Laivuori, Aavikko, Kaunisto, Loukola, Lahtela, Mattsson, Laiho, Parolo, Pietro, Arto, Kanai, Mars, Rämö and Palotie2023). In this study, we used FinnGen Release 8, which comprises 356,077 Finnish individuals. We categorized the sample, using the same ICD-codes as in SUPER, into people with (1) schizophrenia (N = 6,280), (2) schizoaffective, bipolar disorder, or psychotic depression (N = 8,177), and (3) no psychiatric disorder (N = 251,638). Exactly 4,708 patients with schizophrenia and 3,511 patients with affective psychotic disorders were also included in the SUPER study. PGSs in FinnGen were calculated with the PRS-CS method (Ge et al., Reference Ge, Chen, Ni, Feng and Smoller2019). A total of 3,511 patients were found both in the FinnGen and the SUPER sample, but only 3,078 patients in SUPER (before applying inclusion criteria) had a diagnosis of schizoaffective disorder, bipolar disorder, or psychotic depression. This is likely due to diagnoses used in drug purchases and reimbursements being available in FinnGen, while only Finnish Care Register for Health Care (HILMO) diagnoses are available in the SUPER study. HILMO recognizes mostly only diagnoses made in the public health care system of Finland. An additional difference between the samples is that SUPER recruited patients with at least one psychotic episode in their medical history, and this does not apply to patients with bipolar disorder in the FinnGen study. Hence, this group from the FinnGen sample is referred to as ‘affective psychotic and bipolar disorders’.
Statistical analysis
For analyzing associations between PGSs and outcomes, we conducted regression analyses in SPSS version 29.0, with age, gender, diagnostic group (schizophrenia and affective psychotic disorders), the first four principal components to control for population stratification, and one PGS value as independent variables. Logistic regression was conducted for the dichotomized dependent variables of sleep traits, subjective cognitive functioning, clozapine use, and work status, and linear regression was performed for MHI-5, EQ5D, RT, PAL, and days in involuntary hospitalization. We used a Bonferroni–Holm correction (0.05/14 [number of outcomes] = 0.0036 as the lowest significant threshold) to correct for multiple analyses (Aickin & Gensler, Reference Aickin and Gensler1996). In supplemental analyses, we conducted the same analyses separately in the two diagnostic groups. Nagelkerke’s R 2 was used to assess the proportion of variance explained by PGSSD and PGSINS (Nagelkerke, Reference Nagelkerke1991).
To analyze the distribution of the PGSs in the FinnGen cohort, we conducted binomial regression analyses in R version 4.1.3. The first four principal components and gender were included as covariates. The PGS was the independent variable, while the diagnosis was the dependent variable. A Bonferroni–Holm correction (0.05/6 [amount of PGSs] = 0.0083) was used to correct for multiple analyses. The regression analyses were conducted separately for the two diagnostic groups, and people with no psychiatric disorders were the reference group. We replicated the full FinnGen sample analyses, with the two subsamples of (1) patients who were a part of both FinnGen and the SUPER study, and (2) the patients in FinnGen who were not in the SUPER study.
Results
Demographics
The largest diagnostic group was patients with schizophrenia (demographics in Table 2). Patients with schizophrenia had on average a longer time since the first diagnosis of psychosis (22.0 vs. 12.9 years), lower level of education, and lower participation in the workforce than patients with affective psychotic disorders.
Demographics of the SUPER study
Note: SZ, schizophrenia; SZ-A, schizoaffective disorder; BD, bipolar disorder; Ps-DEP, psychotic depression.
Sleep-related measures
Using polygenic scores for sleep traits, diurnal preference (eveningness), and schizophrenia, we investigated how the polygenic burden was associated with sleep problems within the psychotic disorder spectrum (Figure 1 and Supplementary Table 1). PGSINS, PGSSD, and PGSSS showed the strongest impact on sleep. After Bonferroni–Holm correction, high PGSINS was significantly associated with DIS (OR = 1.18, 95 % CI = 1.12–1.24, p = 7.12 × 10−11), EMAs (OR = 1.11, CI = 1.06–1.17, p = 6.60 × 10−6), poor SQ (OR = 1.22, CI = 1.15–1.30, p = 8.68 × 10−10), and FAT (OR = 1.08, CI = 1.03–1.13, p = .002), and similar findings were observed with PGSSS (DIS, OR = 1.18, CI = 1.05–1.17, p = 7.20 × 10−5; EMAs, OR = 1.08, CI = 1.03–1.13, p = .001; short SD, OR = 1.18, CI = 1.09–1.27, p = 6.15 × 10−5; poor SQ, OR = 1.18, CI = 1.11–1.26, p = 4.53 × 10−4). High PGSSD was associated with less EMAs (OR = 0.92, CI = 0.87–0.96, p = 4.41 × 10−4), better SQ (OR = 0.87, CI = 0.82–0.93, p = 3.87 × 10−5), and with longer SD (short SD, OR = 0.86, CI = 0.79–0.93, p = 1.83 × 10−4; long SD, OR = 1.09, CI = 1.03–1.14, p = .001). PGSLS was only associated with more EMAs (OR = 1.07, CI = 1.02–1.12, p = .003) and PGSME was not significantly associated with any of the sleep outcomes. High PGSSZ was associated with less EMAs (OR = 0.90, CI = 0.86–0.95, p = 2.34 × 10−4) and better sleep quality (poor SQ, OR = 0.87, CI = 0.81–0.94, p = 2.88 × 10−4). In a linear regression analysis of sleep duration, PGSSD explained 0.3% of the variation in phenotypic sleep duration, while 0.9% of poor SQ variance, as measured by Nagelkerke R 2, was explained by PGSINS.
Results from logistic and linear regression analyses for the polygenic scores. For congruency, the outcomes of subjective health, MHI-5, and EQ5D were inversed for this figure. Darker gray indicates positive coefficients (representing a worse outcome), while lighter gray indicates negative coefficients. * = significant after Bonferroni–Holm correction. All results significant before correction (p < 0.05) are colored. Note: PGSs in rows: INS, ‘insomnia’; SD, ‘sleep duration’; ME, ‘morning-eveningness (chronotype)’; SZ, ‘schizophrenia’; MHI-5, ‘Mental Health Inventory -5’; EQ5D, ‘Health-related quality of life’.
Well-being and subjective cognitive functioning
PGSINS was the only PGS with significant associations (Figure 1 and Supplementary Tables 2 and 3), and the score was associated with all non-sleep subjective measures, including poorer subjective cognitive functioning (OR = 1.14, CI = 1.09–1.20, p = 1.86 × 10–7), poorer subjective health, β = −0.06, CI = −0.08, −0.03, p = 5.23 × 10−7), worse EQ5D (β = −0.07, CI = −0.09, -0.05, p = 1.89 × 10−9) and MHI-5 (β = −0.06, CI = −0.08, −0.02, p = 4.05 × 10−8).
Cognitive tests and measures for disease severity
Mainly PGSSZ had significant associations with objective outcomes (Figure 1 and Supplementary Table 4), including worse results in PAL (β = 0.07, CI = 0.04, 0.09, p = 1.60 × 10–6) and RTI (β = 0.05, CI = 0.02, 0.07, p = .001), and more frequent clozapine use (OR = 1.15, CI = 1.08–1.22, p = 6.13 × 10−7). PGSINS, and PGSSS were associated with a lower risk of unemployment (OR = 0.89, CI = 0.83–0.96, p = .003, and OR = 0.88, CI = 0.82–0.95, p = .002, respectively).
In analyses with diagnostic splits (Supplementary Table 7), the association between high PGSSZ and worse performance in PAL and RTI remained significant only in patients with affective psychotic disorders.
Distribution of PGSs across the diagnostic spectrum and the general population
We investigated the difference in the distribution of PGSs in diagnostic groups (schizophrenia and affective psychotic and bipolar disorders) compared with people with no psychiatric disorders (Figure 2 for boxplot, Table 3). When analyzing the full FinnGen sample, patients with schizophrenia had a very strong association with PGSSZ, and associations with PGSSD (OR = 1.09, CI = 1.06–1.11, p = 1 × 10−10) and PGSLS (OR = 1.09, CI = 1.06–1.12, p = 2 × 10−11), and associations with PGSME (OR = 1.06, CI = 1.04–1.09, p = 2 × 10−6), and PGSSS (OR = 0.96, CI = 0.92–0.98, p = .002). For patients with affective bipolar and psychotic disorders, there were associations with PGSSZ, PGSINS (OR = 1.08, CI = 1.05–1.10, p = 3 × 10−12), PGSLS (OR = 1.12, CI = 1.09–1.14, p < 2 × 10−16), and an association with PGSSS (OR = 1.05, CI = 1.03–1.07, 1 × 10−5).
Boxplot showing the distribution of polygenic scores across different diagnostic groups.
Results from linear regression analyses of PGSs in the diagnostic groups, compared with people with no psychiatric disorder
A separate analysis of (1) SUPER patients and (2) patients only in FinnGen (Supplementary Table 8), found similar patterns of associations in both data sets for PGSSZ. In SUPER, PGSSD was associated with affective psychotic disorders, as did PGSME with marginal statistical evidence. These associations were not found in Finngen, or in the complete data set. PGSSZ was higher in the SUPER sample than in the FinnGen sample.
Discussion
To the best of our knowledge, this is the first systematic study on polygenic influences for sleep traits in psychotic disorders. We found that sleep PGSs, derived from the general population, are significantly associated with sleep traits also in this population with psychiatric comorbidity and high rates of sleep-impacting medications (Cederlöf et al., Reference Cederlöf, Holm, Taipale, Tiihonen, Tanskanen, Lähteenvuo, Lahdensuo, Kampman, Wegelius, Isometsä, Kieseppä, Palotie, Suvisaari and Paunio2024). Additionally, we identified varied distributions of sleep PGSs across non-affective and affective psychotic disorders and individuals with no psychiatric disorder. A high genetic risk for insomnia was consistently associated with worse well-being, while that for schizophrenia had an impact on objective measures, including poorer performance in cognitive tests and more clozapine use.
Different types of sleep problems have been previously linked to psychotic disorders in clinical studies, where they are often considered secondary to the disorder or medication (Krystal, Reference Krystal2012), albeit our findings in non-medicated patients show that they also present more frequently with both insomnia and hypersomnia symptoms than the general population (Cederlöf et al., Reference Cederlöf, Holm, Lähteenvuo, Haaki, Hietala, Häkkinen, Isometsä, Jukuri, Kajanne, Kampman, Kieseppä, Lahdensuo, Lönnqvist, Männynsalo, Niemi-Pynttäri, Suokas, Suvisaari, Tiihonen, Turunen and Paunio2022, Reference Cederlöf, Holm, Taipale, Tiihonen, Tanskanen, Lähteenvuo, Lahdensuo, Kampman, Wegelius, Isometsä, Kieseppä, Palotie, Suvisaari and Paunio2024). Our present finding on the association of sleep PGSs to sleep traits in psychotic disorders shows how intrinsic genetic vulnerability for insomnia independently links to sleep problems. On the other hand, the lack of association of the sleep PGSs to any of the objective measures was striking. For example, while PGSINS was associated significantly with subjective experience of impaired cognitive functioning, it was not correlated with the objective measures of cognition (vigilance or verbal learning). Contrary to this, PGSSZ was not linked to subjective but only to worse objective measures on cognitive functions, potentially linked to sleep spindle and slow-wave abnormalities observed in earlier studies of patients with schizophrenia (Ferrarelli, Reference Ferrarelli2021). Furthermore, PGSSZ correlated negatively with subjective experiences of sleep problems. Thus, according to these findings, sleep symptoms seem to form a separate domain of symptoms with distinctive etiological influences among patients with psychotic disorders.
In our attempt to find differences in genetic influences on non-affective and affective psychosis, we found a consistent pattern of deviating risks in both datasets, the SUPER sample and the FinnGen data. As expected, PGSSZ is associated robustly with schizophrenia, showing a > 2-fold risk compared with the general population (Allardyce et al., Reference Allardyce, Leonenko, Hamshere, Pardiñas, Forty, Knott, Gordon-Smith, Porteous, Haywood, Di Florio, Jones, McIntosh, Owen, Holmans, Walters, Craddock, Jones, O’Donovan and Escott-Price2018; Sullivan et al., Reference Sullivan, Agrawal, Bulik, Andreassen, Børglum, Breen, Cichon, Edenberg, Faraone, Gelernter, Mathews, Nievergelt, Smoller, O’Donovan, Daly, Gill, Kelsoe, Koenen, Levinson and Sklar2018). It was also elevated in affective psychotic and bipolar disorders but with a lower OR. PGSINS was significantly higher among patients with affective psychotic disorders, but not among those with schizophrenia, as compared with the general population. This finding supports the previously well-known role of insomnia as a risk factor for affective disorders (Hertenstein, Benz, Schneider, & Baglioni, Reference Hertenstein, Benz, Schneider and Baglioni2023).
Regarding the three PGSs related to sleep duration, we observed a pattern in which patients with schizophrenia had higher levels of the PGSSD and PGSLS scores (reflecting genetic propensities for long sleep duration), while patients with affective psychotic and bipolar disorders had elevated levels of both PGSLS and PGSSS (reflecting genetic propensities for both short and long sleep duration). The finding on schizophrenia supports the role of genetic influences in our earlier observation of longer sleep duration in psychotic disorders compared with the general population (Cederlöf et al., Reference Cederlöf, Holm, Lähteenvuo, Haaki, Hietala, Häkkinen, Isometsä, Jukuri, Kajanne, Kampman, Kieseppä, Lahdensuo, Lönnqvist, Männynsalo, Niemi-Pynttäri, Suokas, Suvisaari, Tiihonen, Turunen and Paunio2022). Comprehensively, our findings provide evidence of distinct patterns of genetic sleep-related risks in non-affective and affective psychosis.
Previous studies have proposed diurnal preference for eveningness as a risk for compromised mental health, in particular depression (Norbury, Reference Norbury2021), and according to some studies, schizophrenia and bipolar disorder as well (Linke & Jankowski, Reference Linke and Jankowski2021). In this study, PGSME was not associated significantly with any subjective or objective measures among the SUPER patients. In the full FinnGen data, there was, however, a significant and relatively robust correlation to schizophrenia but not to affective psychotic and bipolar disorders. One potential mechanism underlying eveningness-type of diurnal preference is a slower dynamic of sleep pressure build-up and dissipation(Mongrain & Dumont, Reference Mongrain and Dumont2007; Taillard, Philip, Coste, Sagaspe, & Bioulac, Reference Taillard, Philip, Coste, Sagaspe and Bioulac2003). Our findings for PGSSD and PGSME suggest the presence of deviating processes in homeostatic sleep regulation in schizophrenia – a conclusion that would be in line with the previous studies indicating slow-wave sleep deficits in schizophrenia (Ferrarelli, Reference Ferrarelli2021).
Strengths and limitations. Our study comprised a large nationwide sample of patients with psychotic disorders, combining genetic information with rich phenotype data including extensive subjective metrics and cognitive tests. The use of the FinnGen database enabled novel large-sample analyses on sleep PGSs in the different diagnostic groups. There are, however, some limitations that should be considered when interpreting the findings. It is important to note that these findings were derived from a relatively homogenous population, which can be advantageous for some of the exploratory investigations presented in this article; however, this homogeneity may also limit the generalizability of the results (Woodward, Urbanowicz, Naj, & Moore, Reference Woodward, Urbanowicz, Naj and Moore2022). The impact of sleep PGSs was modest on sleep phenotypes, in line with previous studies in the general population and patients with major depressive disorder (Dashti, Jones, et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019; Melhuish Melhuish Beaupre et al., Reference Melhuish Beaupre, Tiwari, Gonçalves, Zai, Marshe, Lewis, Martin, McIntosh, Adams, Baune, Levinson, Boomsma, Penninx, Breen, Hamilton, Awasthi, Ripke, Jones, Jones and Kennedy2021). Second, self-assessment of cognition and sleep is known to be deficient in patients with psychotic disorders, as factors such as depressed mood can exaggerate symptoms (Demant, Vinberg, Kessing, & Miskowiak, Reference Demant, Vinberg, Kessing and Miskowiak2015; Gonzalez, Tamminga, Tohen, & Suppes, Reference Gonzalez, Tamminga, Tohen and Suppes2013; Raffard, Lebrun, Bayard, Macgregor, & Capdevielle, Reference Raffard, Lebrun, Bayard, Macgregor and Capdevielle2020), while simultaneously some patients may have limited awareness of their cognitive deficits (Homayoun, Nadeau-Marcotte, Luck, & Stip, Reference Homayoun, Nadeau-Marcotte, Luck and Stip2011; Prouteau, Roux, Destaillats, & Bergua, Reference Prouteau, Roux, Destaillats and Bergua2017). Moreover, the cognitive measures could also be more extensive and include domains such as executive functions or social cognition (CANTAB). Finally, we did not have objective sleep data in the present study. Such data would be valuable not only in considering the modest effect of sleep PGSs on objective sleep in the general population (Dashti, Jones, et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont, Jeffries, Dashti, Hillsdon, Ruth, Tuke, Yaghootkar, Sharp, Jie, Thompson, Harrison, Dawes, Byrne, Tiemeier and Weedon2019; Foldager et al., Reference Foldager, Peppard, Hagen, Stone, Evans, Tranah, Sørensen, Jennum, Mignot and Schneider2022), but also in further elucidating potential pathogenic sleep-related processes in psychotic disorders.
Conclusions
Sleep PGSs derived from the general population influence self-reported sleep also in persons with psychotic disorders. PGSINS is associated with worse subjective well-being, but not with cognitive deficits or other objective measures of disease severity, while PGSSZ is related to worse objective measures, but not to worse subjective measures. Our findings underscore the importance of sleep for the quality of life of patients, and the etiologic heterogeneity between core objective elements of psychotic disorders and the subjective measures related to sleep and well-being.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725000844.
Acknowledgments
We want to thank all study participants in both SUPER and FinnGen for making this study possible. We also want to thank the entire SUPER-Finland study group, including: Aija Kyttälä (Finnish Institute for Health and Welfare), Auli Toivola (Finnish Institute for Health and Welfare), Benjamin Neale (Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard), Huei-yi Shen (Institute for Molecular Medicine Finland), Imre Västrik (Institute for Molecular Medicine Finland), Jouko Lönnqvist (Finnish Institute for Health and Welfare), Jussi Niemi-Pynttäri (Psychiatric and Substance Abuse Services, City of Helsinki), Katja Häkkinen (Niuvanniemi Hospital, University of Eastern Finland), Kimmo Suokas (Faculty of Social Sciences, Tampere University), Mark Daly (Institute for Molecular Medicine Finland), Noora Ristiluoma (Finnish Institute for Health and Welfare), Olli Pietiläinen (Neuroscience Center, HiLIFE, University of Helsinki), Risto Kajanne (Institute for Molecular Medicine Finland), Steven E. Hyman (Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard), Tarjinder Singh (Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard), Teemu Männynsalo (Psychiatric and Substance Abuse Services, City of Helsinki), Tuomas Jukuri (Medical Research Center, University of Oulu and Oulu University Hospital), Willehard Haaki (Faculty of Medicine, University of Turku).
Funding statement
This work was supported by the Stanley Center for Psychiatric Research at Broad Institute. Cederlöf and Paunio were supported by grants from Finska Läkaresällskapet (#9-1600-15 and #8-1353-9), HUS (TYH2021328) and Academy of Finland (#357643). Holm was supported by a grant from the Academy of Finland (#310295). Kämpe was supported by Svenska sällskapet för medicinsk forskning, SSMF (# PD20-0190). The funding organizations had no role in the design or execution of the study; in the collection, management, analysis, or interpretation of data; in the preparation, review, or approval of the manuscript; or in the decision to submit the manuscript for publication.
Competing interests
Regarding conflicts of interest, Markku Lähteenvuo is an owner and board member of Genomi Solutions Ltd. and Nursie Health Ltd. and has received honoraria from Sunovion, Orion Pharma, Janssen-Cilag, Otsuka Pharma, Lundbeck, and Medscape, travel funds from Sunovion, and research grants from the Finnish Medical Foundation, the Emil Aaltonen Foundation, and the Finnish Cultural Foundation. Jari Tiihonen has participated in research projects funded by grants from Janssen-Cilag and Eli Lilly to their employing institution; has received personal fees from the Finnish Medicines Agency (Fimea), European Medicines Agency (EMA), Eli Lilly, Janssen-Cilag, Lundbeck, and Otsuka; is a member of the advisory board for Lundbeck; and has received grants from the Stanley Foundation and the Sigrid Jusélius Foundation. Tiina Paunio has no COI related to the content of the current manuscript. Outside this work, she has received honoraria for lectures given at Idorsia Pharmaceuticals Ltd Masterclass 2023 and Biogen Workshop 2023. None of the other authors report any financial relationships with commercial interests.
Ethical standard
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The authors also assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional guides on the care and use of laboratory animals.
Open access
