Sleep inertia drives the association of evening chronotype with psychiatric disorders: epidemiological and genetic evidence

Abstract

Evening chronotypes (a.k.a. "night-owls") are held to be at greater risk for psychiatric disorders. This is postulated to be due to delayed circadian timing increasing the likelihood of circadian misalignment in an early-oriented society. Circadian misalignment is known to heighten sleep inertia, the difficulty transitioning from sleep to wake characterized by low arousal and cognitive impairment, and evening chronotypes experience greater sleep inertia. Therefore, difficulty awakening may explain the relationship between evening chronotype and psychiatric disorders by acting as a biomarker of circadian misalignment. In analyzing the longitudinal incidence of psychiatric disorders in the UK Biobank (n = 496,820), we found that evening chronotype predicted increased incidence of major depressive disorder, schizophrenia, generalized anxiety disorder and bipolar disorder. Crucially, this effect was dependent on sleep inertia, which was a much stronger predictor of these disorders, such that evening types without sleep inertia were at no higher risk as compared to morning types. Longitudinal analyses of suicide and depressed mood (CES-D score) in the Older Finnish Twin Cohort (n = 23,854) replicated this pattern of results. Twin and genome-wide association analyses of difficulty awakening identified the trait to be heritable (Twin H ² = 0.40; SNP h ² = 0.08), enriched for circadian rhythms genes and have substantial shared genetic architecture with chronotype. Marginal and conditional Mendelian randomization analyses mirrored the epidemiological results, such that the causal effect of evening chronotype on psychiatric disorders was driven by shared genetic architecture with difficulty awakening. In contrast, difficult awakening was strongly causally associated with psychiatric disorders independently of chronotype. Psychiatric disorders were only weakly reverse causally linked to difficult awakening. Collectively, these results challenge the notion that evening chronotype is a risk factor for psychiatric disorders per se, suggesting instead that evening types are at greater risk for psychiatric disorders due to circadian misalignment, for which sleep inertia may be acting as a biomarker.

Figure 1.

Cumulative hazard plots for the marginal and conditional longitudinal associations of difficult awakening and evening chronotype with (A-D) major depressive disorder, (E-H) schizophrenia, (I-L) generalized anxiety disorder and (M-P) bipolar disorder. The cumulative incidence ± the 95% CI for difficult awakening vs. easy awakening and late chronotype vs. early chronotype in marginal and conditional models are presented. Model 2 is presented adjusted for age, sex, ethnicity, season, employment, and physical activity. Cumulative incidence for the marginal models are displayed in blue (difficult awakening) and red (late chronotype) and cumulative incidence for the conditional model of difficult awakening and late chronotype after mutual adjustment is displayed in green. In marginal models, difficulty awakening and chronotype are included separately, whereas in conditional models difficulty awakening and chronotype are included together Note: flattening at the time of maximum follow-up is due to the UKB recruitment period occurring over four years.

Figure 2.

Coefficients plot for the longitudinal marginal and conditional association of difficult awakening and late chronotype with (A) suicide and (B) Center for Epidemiologic Studies Depression (CES-D) score in the Older Finnish Older Twin Cohort study (n = 23,095). Coefficients represent hazard ratios ± SEM for (A) suicide and unstandardized b coefficients ± SEM for (B) CES-D score.

Figure 3.

Manhattan plot for the genome-wide association study of difficulty awakening in the UK Biobank cohort. Peaks highlighted in blue represent genomic risk loci passing the genome-wide significance threshold, p < 5×10⁻⁸.

Figure 4.

GSMR Mendelian Randomization effect size plots for the causal effects of difficult awakening and chronotype in marginal and conditional (mtCOJO adjusted) models on (A) major depressive disorder, (B) suicide attempt, (C) schizophrenia, (D) PTSD, (E) anxiety disorder and (F) bipolar disorder. Coefficients represent the MR effect (b_xy) on the logit scale and 95% CIs for difficult awakening and late chronotype GWAS on each respective psychiatric outcome GWAS. Coefficients for the unadjusted MR are displayed in blue (difficult awakening) and red (evening chronotype) and coefficients for the conditional MR effect of difficult awakening and evening chronotype after mutual mtCOJO adjustment are displayed in green. Note: MR effect sizes and 95% CIs for difficult awakening are scaled to the MR effect of evening chronotype on sleep-midpoint to enable comparison of effect sizes. Unscaled values are presented alongside scaled values in Supplementary Table 19.