University of Groningen Rhythm & Blues Knapen, Stefan Erik

University of Groningen

Rhythm & Blues Knapen, Stefan Erik

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Knapen, S. E. (2019). Rhythm & Blues: Chronobiology in the pathophysiology and treatment of mood disorders. Rijksuniversiteit Groningen.

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license. More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne- amendment.

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 07-10-2021 Chronobiology in the pathophysiology and treatment of mood disorders

Stefan E. Knapen Colophon The research in this thesis was supported by grants from the NIMH (R01 MH090553), foundation de Drie Lichten and the Ubbo Emmius Fonds Talent Grant.

Publications of this thesis was generously supported by the University of Groningen, University Medical Center Groningen, the Research School for Behavioral and Cognitive Neuroscience, de Nederlandse vereniging voor Slaap – Waak Onderzoek and de vakgroep Neurologie Reinier de Graaf Gasthuis.

Cover design & lay-out: Ellen Beck, www.ellenbeck.nl Printed by: Ridderprint, the Netherlands

ISBN printed version: 978-94-034-1536-9 ISBN digital version: 978-94-034-1537-6

Rhythm & Blues

Chronobiology in the pathophysiology and treatment of mood disorders

Proefschrift

ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen op gezag van de rector magnificus prof. dr. E. Sterken en volgens besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op

woensdag 17 april 2019 om 16.15 uur

door

Stefan Erik Knapen

geboren op 29 november 1990 te Apeldoorn

Promotor Prof. dr. R.A. Schoevers

Copromotor Dr. R.F. Riemersma-van der Lek

Beoordelingscommissie Prof. dr. E. van Someren Prof. dr. A. Wirz-Justice Prof. dr. W.A. Nolen

Rhythm & Blues Chronobiology in the pathophysiology and treatment of mood disorders

Stefan E. Knapen Table of contents

Chapter 1: Introduction 9 Mood disorders Rhythm and the circadian timing system Vulnerability to mood disorders Link between mood disorder and rhythm problems Treatment of mood disorder through rhythm Thesis outline References

Chapter 2: Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data 19A Abstract Introduction Material and methods Results Discussion References

Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety 34 Abstract Introduction Material and methods Results Discussion Supplemental material References

Chapter 4: Letter to the editor: Chronotype not associated with non-remission, but with current state? 48 References Chapter 5: Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls 52 Abstract Introduction Material and methods Results Discussion References

Chapter 6: Coping with a life event in bipolar disorder – ambulatory measurement, signalling and early treatment 65 Summary Background Case presentation Treatment Outcome and follow-up Discussion References

Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder 71 Abstract Introduction Material and methods Results Discussion References

Chapter 8: Fractal biomarker of activity in patients with bipolar disorder 94 Abstract Introduction Methods Results Discussion References Chapter 9: The duration of light treatment and therapy outcome in Seasonal Affective Disorder 107 Abstract Introduction Material and Methods Results Discussion References

Chapter 10: The relation between chronotype and treatment outcome with light therapy on a fixed time schedule 116 Abstract Introduction Methods Results Discussion References

Chapter 11: General discussion 124 Part 1: Vulnerability factors for developing mood disorders Part 2: Rest-activity rhythms and mood – actigraphy Part 3: Therapeutic options with chronobiological mechanisms General discussion Conclusions References

Chapter 12: Addenda 139 Samenvatting Dankwoord Publications Curriculum Vitae Chapter 1 Introduction

Mood disorders are among of the leading causes for disability in the western world, with an estimated lifetime risk of about 28% (1,2). Within the spectrum of mood disorders two main disorders are major depressive disorder (MDD) and bipolar disorder (BD) (3). Since as far as a century back chronobiology, the field of biology that studies periodic (daily, monthly, yearly) phenomena, has played an important role in studying psychiatric diseases and most prominently mood disorders (27). In this thesis, a number of chronobiological mechanisms relevant for mood disorders will be studied.

Mood disorders MDD is the most prevalent mood disorder, with a projected lifetime risk of 23.2% (2). The two core symptoms of MDD are; a depressed mood and loss of interest or pleasure in activities (3). Other symptoms include weight loss or gain, loss of energy and either insomnia or hypersomnia (box 1). A subtype of MDD is seasonal affective disorder (SAD). Patients with SAD suffer from recurring depressive episodes in a seasonal manner (4). BD has a lifetime risk of 5.2%. Patients with BD experience manic and depressive symptoms in an episodic manner (box 2). There are two subtypes of BD, bipolar disorder type I and bipolar type II. Patients with bipolar disorder type I experience at least one manic episode, defined as a period of at least one week with an elevated mood and other symptoms including racing thoughts, increased goal-oriented activities and a decreased need of sleep. Patients with bipolar disorder type II experience the same symptoms, although the symptoms are less severe, not causing significant problems in their daily life. Even under treatment 60% of BD patients experience a relapse within 2 year, and patients have residual symptoms for about a third of their lifetime (5–9). Box 1: Symptoms major Box 2: Symptoms manic episode A. Distinct period of abnormally and per- depressive disorder sistently elevated, expansive, or irritable 1. Depressed mood most of the day, nearly mood and abnormally and persistently every day. increased activity or energy, lasting at least 2. Markedly diminished interest or pleasure one week and present most of the day, nearly in all, or almost all, activities most of the every day. day, nearly every day. B. During the period of mood disturbance at 3. Significant unintentional weight loss or least three of the following symptoms: gain, or decrease / increase in appetite. 1. Inflated self-esteem or grandiosity. 4. Insomnia or hypersomnia nearly every day. 2. Decreased need for sleep (feeling rested 5. Psychomotor agitation or retardation after a short sleep period. nearly every day. 3. More talkative than usual. 6. Fatigue or loss of energy nearly every day. 4. Flight of ideas or racing thoughts. 7. Feelings of worthlessness or excessive or 5. Distractibility. inappropriate guilt nearly every day. 6. Increase in goal-directed behavior 8. Diminished ability to think or concen- 7. Excessive involvement in activities that have trate, or indecisiveness nearly every day. a high potential for painful consequences. 9. Recurrent thoughts of death, recurrent C. The mood disturbance is sufficiently severe suicidal ideation or a suicide attempt or to cause impairment in social or occupational specific plan for committing suicide. functioning.

10 Rhythm and the circadian timing system In mood disorders a number of chronobiological phenomena are clear, such as variation of mood within the day (diurnal variation), where depressed patients may experience worse mood in the morning or in the evening (28,29). Furthermore, a seasonal pattern is one of the key factors of seasonal affective disorder (4). Lastly the prominent place of sleep problems such as early morning awakening, in- and hypersomnia, in the diagnostic criteria shows the link between chronobiology and mood disorders. In both MDD and BD sleeping difficulties is a diagnostic criterion. In MDD, 60-84% of the patients experience insomnia, problems of initiating and maintaining sleep (10). In BD, one of the key components of a manic episode is the decreased need for sleep and during a depressive episode many patients experience insomnia (11). Further- more, aside from the actual sleep duration, the timing of sleep and the daily activity (the rest-acitivity rhythm) is implicated in these mood disorders. Especially in BD, disturbances in this rest-activity rhythm are thought to be an underlying trait factor, present in every state of the disease.

The human rest-activity rhythm is regulated by an intrinsic clock, that steers the intrinsic circadian rhythm. The intrinsic rhythm, the rhythm that would occur if a person would have no timing cues, is 24.2 hours (14,15). The intrinsic clock is entrained daily by these outside timing cues called Zeitgebers (synchronizers). The adaptation of the internal circadian (circa: around, dies: day) rhythm in people is regulated by a set of neurons located in the suprachiasmatic nucleus (SCN) in the anterior part of the hypothalamus (16). The SCN contains a cell-autonomous transcription-translation loop of a core set of genes called the clock genes. Among these genes are Clock, Bmal1, Per1, Per2, Cry1 and Cry2. The SCN regulates the rhythm in downstream targets, including brain regions that are important for the regulation of sleep, such as the ventrolateral preoptic area (VLPO) and the lateral hypothalamus (17,18). The most important Zeitgeber is light through the retina. This light has an effect on the SCN through retinal ganglion cells and causes a shift in the circadian rhythm (19,20). The direction of the shift depends on the time of day, with evening light causing a phase delay, while morning light causes a phase advance (21,22). Our internal clock is aligned with the outside world through this phase shifting effect of light. However, our internal clock can also be out of sync with the outside world, being misaligned with the outside world. This misalignment of the circadian rhythm is associated with a number of health problems, such as obesity and an increased cardiovascular risk profile (23,24). This misalignment not only includes the misalignment with the outside world and the internal clock, but also the misalignment of the central clock (the SCN) and the timing in other organs in the human body, such as the liver and the muscles, although this is outside of the scope of this thesis (16).

Vulnerability to mood disorders Whether a person develops a mood disorder is the result of a multifactorial process (11,30). Finding underlying vulnerability factors for the development of mood disorders is important to increase our understanding of the pathophysiological mechanisms of the disease.

11 Chapter 1: Introduction

Genetic vulnerability One of the underlying factors could include abnormalities in the molecular clock. The circadian timing system is regulated by a transcriptional-translational feedback loop constituted by clock genes (16,31). Abnormalities in these clock genes have been implicated in mood disorders. NPAS2 has been implicated in SAD, and there might be an association with haplotypes on the core clock genes PER3 and ARNTL and BD (32–34). Another indication that the core clock genes might play a role in the BD is the fact that Clock knock-out mice display a behavioral profile similar to a manic episode, including hyperactivity, decreased sleep and an increase in reward-seeking behavior (35). Treating these knock-out mice with lithium, the first treatment of choice for BD, seems to restore this behavior in the direction of the wild-type mice. Despite these findings, a specific genetic marker has not been found in mood disorders, and studying the genetic vulnerability alone might not be sufficient. Finding a link between genetic characteristics and psychiatric outcome might provide an interesting biomarker for clinical practice (11,30).

Chronotype and mood disorders Another possible vulnerability factor is the person’s preference for morning or evening activities, called a person’s chronotype. A morning type (or ‘lark’) has a preference for early awakening, morning activities and an early bedtime, while the evening chronotype (‘owls’) have a preference for later activities. Throughout the population there is a great variety within chronotype, the evening chronotype is associated with a younger age and morning types are more common in older people and in women (36,37). Chro- notype might be a non-invasive marker of internal phase, where people with a later chronotype also have a slightly longer running internal clock (internal phase longer than 24.2 hours) (14,38). Although evening types prefer their timing of activities later on the day, our society typically demands earlier activities. This misalignment results in social jetlag (39). Just as a jetlag caused by travelling across time zones, the internal phase is out of sync with the external world, and just like a jetlag from travel this can cause significant health risks. Social jetlag has been associated with obesity and an adverse cardiovascular risk profile (23,24,40). Most often people with an evening chronotype have more social jetlag. The evening chronotype is associated with MDD in a number of studies (41–47). The evening chronotype has also been proposed as a vulnerability factor for developing depression (41,48). One of the proposed mechanisms for this risk is, just as in cardiovascular risk, the social jetlag associated with the evening type.

Link between mood disorder and rhythm problems A non-invasive method to study the circadian rest-activity rhythm is the use of actigraphy. With actigraphy people wear a wristband on their non-dominant wrist which measures movement over the day (49,50). This wristband records activity continuous-

12 Chapter 1: Introduction

ly and stores it on a minute-by-minute basis. Participants continue their day-to-day life, resulting in representative measures of their daily rest-activity patterns. Sleep and circadian rhythm variables can be calculated from this wrist activity data. To assess the stability and variability of the rest-activity rhythm van Someren et al. developed different non-parametric variables (51). These variables include intradaily variability, interdaily stability, activity measures during the most active 10 and most inactive 5 hours of the day and the amplitude of the rhythm (relative amplitude). These non-parametric measures provide a sensitive method applicable to wrist-activity data to describe circadian rhythmicity (51). To get a representative value for these measures, a couple of days of actigraphy is necessary, typically around 10 days or more (52). Sleep variables are assessed on a day-to-day basis and can also be derived from actigraphy data. The activity during the night, combined with a self-reported sleep log including bed- and get-up time, is used as an approximation of sleep and wake during the night (50). Although complete certainty of the sleep/wake state of a participant cannot be achieved, actigraphy is validated with the golden standard to assess sleep/wake states, polysomnography (53).

Patients with BD experience sleep and rest-activity disturbances during mood episodes. In manic episodes they experience later bedtimes, a more fragmented rhythm and a shorter sleep duration, whereas they experience less sleep efficiency, insomnia and lower activity levels during depressive episodes (11,54). However, these problems are not only implicated during mood episodes, but also arise in patients without any mood problems, i.e. patients in a euthymic phase (55–57). As these studies are typically performed in small sample sizes and with a short duration of actigraphy measurement, the question whether these symptoms are stable trait features or more a state feature of bipolar disorder, fluctuating with the episodic states, remains unsolved. If these rhythm disturbances are a trait-feature, it might be used as a biomarker, which could assist the diagnostic process to get to the diagnosis of BD. As the time from the first onset of symptoms to first treatment of BD is around 10 years, there is a dire need for diagnostic markers in BD (58).

Although the question whether rest-activity disturbances are a trait feature, or a state feature, is of interest for both diagnostic and therapeutic reasons, the interplay between sleep, rest-activity and mood disturbances is of interest for patients already diagnosed with, and thus suffering from, the disorder. Studies conducted as early as the late 70s and early 80s imply there might be a direct link between disruptions in sleep, the rhythm and the development of mood problems (59,60). When patients are asked what preceded their onset of a mood episode, 77% of the patients reported sleep disturbances as an early symptom of a manic episode, making it the most clear prodrome (early symptom of a disorder) for mania (61). For a depressed episode, 24% of the patients reported sleep disturbances as a prodrome. In a study using subjective sleep measures, in the form of a sleep diary, a mood change was preceded by a change in sleep and or bedrest duration (62). The current technological abilities make it easier to study a patient almost in real time and in their own setting. Ecological Momentary Assessment (EMA), a method of studying patient behavior in real time and in their own personal life, provides a unique opportunity to study the link between sleep and mood

13 problems (63). Monitoring patients with BD with fine grained ecological measurements is of great potential, especially as this can provide information on which aspects of the sleep and rest-activity rhythm precede (and might predict) mood relapses, and can also provide information when these problems may result in a full relapse. As preventing a relapse is the key goal of most interventions for BD early signals of an imminent mood episode can aid both patients and clinicians.

Fractal markers from rest-activity data in bipolar disorder From activity data, different diagnostic markers can be derived. A relative new method to study actigraphy data is studying the fractal pattern of the activity rhythm. Many physiological signals, such as motor activity, show similar temporal structures on different time scales (64). The patterns are called fractal fluctuations and are found in many places in nature, and also in healthy human physiology (65). The loss of this fractal patterns is associated with the loss of a physiological mechanism. An example is the loss of a fractal pattern in rodent motor activity on larger time scales caused by the lesion of the SCN, the center in the brain responsibly for the circadian rhythm in the body (66,67). As BD is also associated with problems in the circadian rhythm, fractal patterns, or the loss thereof, might function as a marker for BD.

Treatment of mood disorder through rhythm interventions Circadian disturbances are not only interesting as a diagnostic marker or an episode predictor, they are also an important target point for treating mood disorders. Among different therapies, one of the more effective, with few side effects is light therapy for seasonal affective disorder (4). Since the discovery of light therapy to relieve symptoms of depression in patients with a seasonal depression, a number of studies have shown its effectiveness and it is currently the treatment of choice for SAD in the Netherlands (12,68,69). Although it is the treatment of choice, there is no consensus on the duration of the light treatment. Earlier studies show that only one week of light therapy is enough to relieve and prevent depressive symptoms during the rest of the winter season (70). Other protocols typically use 8 weeks, or even advise to use light therapy until the be- ginning of the spring (71). Alongside the duration of the light therapy, the timing of light therapy is topic of discussion. Especially as one of the hypotheses of the treatment effect is dependent on the phase shifting properties of light in the morning (72–74). Current protocols advise to time the light therapy to a person’s chronotype, with earlier light for morning chronotypes and later light for patients with evening chronotypes (71). Other studies argue against this phase shift hypothesis and suggest that, although light in the morning is needed, timing exactly to a chronotype might not be necessary (17,75,76).

14 Thesis outline Although sleep and rest-activity disturbances are primarily found during symptomatic states of mood disorders, these disturbances can also pose a vulnerability to the development of the disease. Furthermore, the interplay between rest-activity disturbances and the development and the course of mood disorders is of interest. In this thesis different studies are described that have been conducted to investigate genetic risk factors, disturbances in the rest-activity rhythm, and the treatment of mood disorders, all with the aim to better understand the relation between rhythm and blues.

The first part of this thesis (chapter 2-4) will focus on vulnerability factors for mood disorders within chronobiology by studying epidemiological data from the Netherlands Study of Depression and Anxiety (NESDA). First, we study the relation between circadian genes, chronotype and mood disorders, to see if there might be an underlying genetic predisposition for mood disorders within circadian genes (chapter 2). Further- more, we aim to answer the question if this genetic vulnerability is mediated through the chronotype of the patients. Next, we study whether social jetlag, associated with the evening chronotype, is actually related to MDD, as it is often hypothesized (chapter 3). In chapter 4 we discuss the problems with studying chronotype in patients who show different behavior in different mood states.

The second part of this thesis (chapter 5-8) focuses specifically on BD and actigraphy measures within this disease. In this part, the link between rest-activity disturbances and mood within the different stages of the mood disorder will be studied. We study circadian measures in euthymic patients, to see if patients with BD show circadian rhythm disturbances in the euthymic phase of the disease compared to controls (chapter 5). Next, we study if rest-activity patterns can function as an early warning signal of imminent mood changes in a single subject (chapter 6). In chapter 7 we study the temporal relation between sleep and circadian disturbances and mood in the onset of a mood episode. In chapter 8 we show another measure we can obtain from actigraphy data, fractal patterns, to see if they could function as a marker for BD.

The third and last part of this thesis focuses on the treatment side of mood disorders. In a combined sample of different studies studying light therapy effects within SAD we looked at the duration of light therapy (chapter 9) and the timing of light therapy (chapter 10). In the last chapter of this thesis I will discuss our findings in a broader context (chapter 11).

15 begeleiding van volwassen patiënten met een References depressieve stoornis. Utrecht: Trimbos-institu- 1. Murray CJL, Vos T, Lozano R, Naghavi M, ut; 2013. Flaxman AD, Michaud C, et al. Disability-ad- 13. McMenamin TM. A time to work: recent trends justed life years (DALYs) for 291 diseases and in shift work and flexible schedules. Mon Lab injuries in 21 regions, 1990-2010: A systematic Rev. 2007;130:3. analysis for the Global Burden of Disease Study 14. Duffy JF, Rimmer DW, Czeisler CA. Asso- 2010. Lancet. 2012;380(9859):2197–223. ciation of intrinsic circadian period with 2. Kessler RC, Berglund P, Demler O, Jin R, Meri- morningness-eveningness, usual wake time, kangas KR, Walters EE. Lifetime Prevalence and and circadian phase. Behav Neurosci. 2001 Age-of-Onset Distributions of. Arch Gen Psychi- Aug;115(4):895–9. atry. 2005;62(June):593–602. DOI: 10.1001/ 15. Czeisler CA, Duffy JF, Shanahan TL, Brown archpsyc.62.6.593 EN, Mitchell JF, Rimmer DW, et al. Stability , 3. Association AP. Diagnostic and statistical man- Precision , and Near – 24-Hour Period of the ual of mental disorders (DSM-5®). American Human Circadian Pacemaker. Science (80- ). Psychiatric Pub; 2013. 1999;284(JUNE):2177–82. 4. Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Good- 16. Albrecht U. Timing to perfection: the biology of win FK, Davenport Y, et al. Seasonal affective central and peripheral circadian clocks. Neuron. disorder: a description of the syndrome and 2012 Apr 26;74(2):246–60. DOI: 10.1016/j. preliminary findings with light therapy. Arch neuron.2012.04.006 Gen Psychiatry. 1984;41(1):72. 17. LeGates T a, Fernandez DC, Hattar S. 5. Geddes JR, Miklowitz DJ. Bipolar Disorder Light as a central modulator of circadian 3 Treatment of bipolar disorder. Lancet. rhythms, sleep and affect. Nat Rev Neurosci. 2013;381(9878):1672–82. DOI: 10.1016/ 2014;15(June):443–54. DOI: 10.1038/nrn3743 S0140-6736(13)60857-0 18. Chou TC, Bjorkum AA, Gaus SE, Lu J, Scammell 6. Phillips ML, Kupfer DJ. Bipolar Disorder 2 Bipo- TE, Saper CB. Afferents to the ventrolateral pre- lar disorder diagnosis : challenges and future optic nucleus. J Neurosci. 2002;22(3):977–90. directions. Lancet. 2013;381(9878):1663–71. 19. Czeisler CA, Shanahan TL, Klerman EB, Martens DOI: 10.1016/S0140-6736(13)60989-7 H, Brotman DJ, Emens JS, et al. Suppression 7. Gitlin MJ, Swendsen J, Heller TL, Hammen C. of Melatonin Secretion in Some Blind Pa- Relapse and impairment in bipolar disorder. Am tients by Exposure to Bright Light. N Engl J Psychiatry. 1995 Nov;152(11):1635–40. J Med. 1995;332(1):6–11. DOI: 10.1056/ 8. Perlis RH, Ostacher MJ, Patel JK, Marangell LB, NEJM199501053320102 Zhang H, Wisniewski SR, et al. Predictors of re- 20. Berson DM, Dunn FA, Takao M. Phototransduc- currence in bipolar disorder: primary outcomes tion by retinal ganglion cells that set the circadi- from the Systematic Treatment Enhancement an clock. Science. 2002 Feb;295(5557):1070–3. Program for Bipolar Disorder (STEP-BD). Am J 21. Gordijn MC, Beersma DG, Korte HJ, van den Psychiatry. 2006 Feb;163(2):217–24. Hoofdakker RH. Effects of light exposure and 9. Judd LL, Akiskal HS, Schettler PJ, Endicott J, Ma- sleep displacement on dim light melatonin ser J, Solomon DA, et al. The long-term natural onset. J Sleep Res. 1999 Sep;8(3):163–74. history of the weekly symptomatic status of 22. Khalsa SBS, Jewett ME, Cajochen C, Czeisler C bipolar I disorder. Arch Gen Psychiatry. 2002 a. A phase response curve to single bright light Jun;59(6):530–7. pulses in human subjects. J Physiol. 2003 Jun 10. Harvey AG. Sleep and circadian func- 15;549(Pt 3):945–52. DOI: 10.1113/jphysi- tioning: critical mechanisms in the mood ol.2003.040477 disorders? Annu Rev Clin Psychol. 2011 23. Rutters F, Lemmens SG, Adam TC, Brem- Jan;7:297–319. DOI: 10.1146/annurev-clinp- mer M a, Elders PJ, Nijpels G, et al. Is social sy-032210-104550 jetlag associated with an adverse endocrine, 11. Harvey AG. Sleep and circadian rhythms behavioral, and cardiovascular risk profile? in bipolar disorder: seeking synchrony, J Biol Rhythms. 2014;29(5):377–83. DOI: harmony, and regulation. Am J Psychiatry. 10.1177/0748730414550199 2008 Jul;165(7):820–9. DOI: 10.1176/appi. 24. Scheer FAJL, Hilton MF, Mantzoros CS, Shea ajp.2008.08010098 SA. Adverse metabolic and cardiovascular 12. Spijker J, Bockting C, Meeuwissen J, van Vliet consequences of circadian misalignment. Proc I, Emmelkamp P, Hermens M, et al. Multidis- Natl Acad Sci U S A. 2009;106(11):4453–8. DOI: ciplinaire richtlijn Depressie (Derde revisie). 10.1073/pnas.0808180106 Richtlijn voor de diagnostiek, behandeling en 25. Malhi GS, Kuiper S. Chronobiology of mood disorders. Acta Psychiatr Scand.

16 2013;128(S444):2–15. cial time. Chronobiol Int. 2006;23(1–2):497–509. 26. Wirz-Justice A. Chronobiology and mood disor- 40. Roenneberg T, Allebrandt K V, Merrow M, ders. Dialogues Clin Neurosci. 2003;5(4):315–25. Vetter C. Social jetlag and obesity. Curr Biol. 27. Wirz-Justice A. Chronobiology comes of age. 2012 May 22;22(10):939–43. DOI: 10.1016/j. Acta Psychiatr Scand. 2017;136(6):531–3. cub.2012.03.038 28. Murray G. Diurnal mood variation in depres- 41. Drennan MD, Klauber MR, Kripke DF, Goyette sion: a signal of disturbed circadian function? J LM. The effects of depression and age on the Affect Disord. 2007 Sep;102(1–3):47–53. DOI: Horne-Ostberg morningness-eveningness 10.1016/j.jad.2006.12.001 score. J Affect Disord. 1991 Oct;23(2):93–8. 29. Gordijn MC, Beersma DG, Bouhuys a L, Reinink 42. Levandovski R, Dantas G, Fernandes LC, Caumo E, Van den Hoofdakker RH. A longitudinal W, Torres I, Roenneberg T, et al. Depression study of diurnal mood variation in depression; Scores Associate With Chronotype and Social characteristics and significance. J Affect Disord. Jetlag in a Rural Population. Chronobiol Int. 1994 Aug;31(4):261–73. 2011;28(9):771–8. 30. Gottesman I, Gould T. The endophenotype con- 43. Chelminski I, Ferraro FR, Petros T V, Plaud JJ. An cept in psychiatry: etymology and strategic in- analysis of the “‘ eveningness – morningness ’” tentions. Am J Psychiatry. 2003;(April):636–45. dimension in “‘ depressive ’” college students. J 31. McClung C a. Circadian genes, rhythms and the Affect Disord. 1999;52:19–29. biology of mood disorders. Pharmacol Ther. 44. Kitamura S, Hida A, Watanabe M, Enomoto M, 2007 May;114(2):222–32. DOI: 10.1016/j. Moriguchi Y, Kamei Y, et al. EVENING PREF- pharmthera.2007.02.003 ERENCE IS RELATED TO THE INCIDENCE OF 32. Johansson C, Willeit M, Smedh C, Ekholm J, Pau- DEPRESSIVE STATES INDEPENDENT OF SLEEP- nio T, Kieseppä T, et al. Circadian clock-related WAKE CONDITIONS. 2010;528(April 2016). polymorphisms in seasonal affective disorder 45. Hidalgo MP, Caumo W, Posser M, Coccaro and their relevance to diurnal preference. Neu- SB, Camozzato AL, Chaves MLF. Relationship ropsychopharmacology. 2003 Apr;28(4):734–9. between depressive mood and chronotype in DOI: 10.1038/sj.npp.1300121 healthy subjects. 2009;63:283–90. 33. Nievergelt CM, Kripke DF, Barrett TB, Burg E, 46. Merikanto I, Kronholm E, Peltonen M, Laati- Remick RA, Sadovnick AD, et al. Suggestive Ev- kainen T, Vartiainen E, Partonen T. Circadian idence for Association of the Circadian Genes preference links to depression in general adult PERIOD3 and ARNTL With Bipolar Disorder. population. J Affect Disord. 2015;188:143–8. 2006;241(February 2005):234–41. DOI: 10.1016/j.jad.2015.08.061 34. Takahashi JS, Hong H-K, Ko CH, McDearmon EL. 47. Antypa N, Vogelzangs N, Meesters Y, Schoevers The genetics of mammalian circadian order RA, Penninx BWJH. Chronotype associations and disorder: implications for physiology and with depression and anxiety in a large cohort disease. Nat Rev Genet. 2008;9(10):764–75. study. Depress Anxiety. 2016 Jan;33(1):75–83. 35. Roybal K, Theobold D, Graham A, DiNieri JA, 48. Chan JWY, Lam SP, Li SX, Yu MWM, Chan NY, Russo SJ, Krishnan V, et al. Mania-like behavior Zhang J, et al. Eveningness and insomnia: inde- induced by disruption of CLOCK. Proc Natl Acad pendent risk factors of nonremission in major Sci. 2007;104(15):6406–11. DOI: 10.1073/ depressive disorder. Sleep. 2014;37(5):911–7. pnas.0609625104 DOI: 10.5665/sleep.3658 36. Roenneberg T, Kuehnle T, Juda M, Kantermann 49. Hofstra W a., de Weerd AW. How to assess circa- T, Allebrandt K, Gordijn M, et al. Epidemiolo- dian rhythm in humans: A review of literature. gy of the human circadian clock. Sleep Med Epilepsy Behav. 2008;13:438–44. Rev. 2007 Dec;11(6):429–38. DOI: 10.1016/j. 50. Ancoli-Israel S, Cole R, Alessi C, Chambers M, smrv.2007.07.005 Moorcroft W, Pollak CP. The role of actigraphy 37. Roenneberg T, Wirz-Justice A, Merrow in the study of sleep and circadian rhythms. M. Life between Clocks: Daily Temporal Sleep. 2003 May 1;26(3):342–92. Patterns of Human Chronotypes. J Biol 51. Van Someren EJ, Swaab DF, Colenda CC, Cohen Rhythms. 2003 Feb 1;18(1):80–90. DOI: W, McCall W V, Rosenquist PB. Bright light 10.1177/0748730402239679 therapy: improved sensitivity to its effects on 38. Brown S a, Kunz D, Dumas A, Westermark PO, rest-activity rhythms in Alzheimer patients by Vanselow K, Tilmann-Wahnschaffe A, et al. Mo- application of nonparametric methods. Chrono- lecular insights into human daily behavior. Proc biol Int. 1999;16(4):505–18. Natl Acad Sci U S A. 2008 Feb 5;105(5):1602–7. 52. VAN Someren EJW. Improving actigraphic sleep DOI: 10.1073/pnas.0707772105 estimates in insomnia and dementia: how many 39. Wittmann M, Dinich J, Merrow M, Roenneberg T. nights? J Sleep Res. 2007 Sep;16(3):269–75. Social jetlag: misalignment of biological and so- DOI: 10.1111/j.1365-2869.2007.00592.x

17 53. Kaplan K a, Talbot LS, Gruber J, Harvey AG. nov PC, Peng C, Stanley HE. Fractal dynamics in Evaluating sleep in bipolar disorder: com- physiology : Alterations with disease and aging. parison between actigraphy, polysomnog- 2002;99. raphy, and sleep diary. Bipolar Disord. 2012 66. Hu K, Harper DG, Shea SA, Stopa EG, Scheer Dec;14(8):870–9. DOI: 10.1111/bdi.12021 FAJL. Noninvasive fractal biomarker of clock 54. Gershon A, Ram N, Johnson SL, Harvey AG, neurotransmitter disturbance in humans with Zeitzer JM. Daily Actigraphy Profiles Distin- dementia. Sci Rep. 2013;3:1–7. guish Depressive and Interepisode States in 67. Hu K, Scheer FAJL, Ivanov PC, Buijs RM, Shea SA. Bipolar Disorder. 2016; The suprachiasmatic nucleus functions beyond 55. Mckenna BS, Drummond SPA, Eyler LT. Associ- circadian rhythm generation. Neuroscience. ations between circadian activity rhythms and 2007;149(3):508–17. functional brain abnormalities among euthymic 68. Nussbaumer B, Ca F, Lc M, Jh S, Bn G, Greenblatt bipolar patients : A preliminary study. J Affect A, et al. Light therapy for preventing seasonal Disord. 2014;164:101–6. DOI: 10.1016/j. affective disorder. Cochrane Database Syst Rev. jad.2014.04.034 2015;(11). 56. Sh J, Tai S, Evershed K, Knowles R, Early BR. Ear- 69. Golden RN, Gaynes BN, Ekstrom RD, Hamer ly detection of bipolar disorder : a pilot familial RM, Jacobsen FM, Suppes T, et al. The efficacy high-risk study of parents with bipolar disorder of light therapy in the treatment of mood dis- and their adolescent children. 2006;(5):362–72. orders: a review and meta-analysis of the evi- 57. Geoffroy PA, Boudebesse C, Bellivier F, Lajnef dence. Am J Psychiatry. 2005;162(4):656–62. M, Henry C, Leboyer M, et al. Sleep in remitted 70. Meesters Y, Jansen JH, Beersma DG, Bouhuys a bipolar disorder: A naturalistic case-control L, van den Hoofdakker RH. Early light treatment study using actigraphy. J Affect Disord. 2014 can prevent an emerging winter depression Apr;158:1–7. DOI: 10.1016/j.jad.2014.01.012 from developing into a full-blown depression. J 58. Suppes T, Leverich GS, Jr PEK, Nolen WA, Den- Affect Disord. 1993 Sep;29(1):41–7. icoff KD, Altshuler LL, et al. The Stanley Foun- 71. Terman M, Terman JS. Light therapy for sea- dation Bipolar Treatment Outcome Network II . sonal and nonseasonal depression: efficacy, Demographics and illness characteristics of the protocol, safety, and side effects. CNS Spectr. first 261 patients. 2001;67:45–59. 2005 Aug;10(8):647–63. 59. Wehr T, Wirz-Justice A, Goodwin F. Phase 72. Lewy AJ, Sack RL, Singer CM, White DM. The Advance of the Circadian Sleep-Wake Cy- phase shift hypothesis for bright light’s cle as an Antidepressant. Science (80- ). therapeutic mechanism of action: theoretical 1979;206(4419):710–3. considerations and experimental evidence. 60. Wehr TA, Goodwin FK, Wirz-Justice A, Breit- Psychopharmacol Bull. 1987;23(3):349–53. maier J, Craig C. 48-hour sleep-wake cycles in 73. Murray G, Michalak EE, Levitt AJ, Levitan RD, manic-depressive illness: naturalistic observa- Enns MW, Morehouse R, et al. O sweet spot tions and sleep deprivation experiments. Arch where art thou? Light treatment of Seasonal Gen Psychiatry. 1982;39(5):559–65. Affective Disorder and the circadian time of 61. Jackson A, Cavanagh J, Scott J. A systematic sleep. J Affect Disord. 2006 Feb;90(2–3):227– review of manic and depressive prodromes. J 31. DOI: 10.1016/j.jad.2005.10.010 Affect Disord. 2003 May;74(3):209–17. DOI: 74. Terman JS, Terman M, Lo ES, Cooper TB. Circa- 10.1016/S0165-0327(02)00266-5 dian time of morning light administration and 62. Bauer M, Grof P, Rasgon N, Bschor T, Glenn T, therapeutic response in winter depression. Whybrow PC. Temporal relation between sleep Arch Gen Psychiatry. 2001 Jan;58(1):69–75. and mood in patients with bipolar disorder. 75. Koorengevel KM, Beersma DGM, den Boer JA, Bipolar Disord. 2006 Apr;8(2):160–7. DOI: van den Hoofdakker RH. A Forced Desynchrony 10.1111/j.1399-5618.2006.00294.x Study of Circadian Pacemaker Characteristics 63. Bos FM, Schoevers R a., aan het Rot M. Expe- in Seasonal Affective Disorder. J Biol Rhythms. rience sampling and ecological momentary 2002 Oct 1;17(5):463–75. assessment studies in psychopharmacology: 76. Burgess HJ, Fogg LF, Young MA, Eastman A systematic review. Eur Neuropsychophar- CI. Bright Light Therapy for Winter Depres- macol. 2015;1–12. DOI: 10.1016/j.euroneu- sion — Is Phase Advancing Beneficial ? ro.2015.08.008 2004;21(312):759–75. 64. Hu K, Ivanov PC, Chen Z, Hilton MF, Stanley HE, Shea SA. Non-random fluctuations and multi- scale dynamics regulation of human activity. Phys A Stat Mech its Appl. 2004;337(1–2):307–18. 65. Goldberger AL, Amaral LAN, Hausdorff JM, Iva-

18 Chapter 2 Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data

S.E. Knapen1,4, F.J. Bosker2, I.M. Nolte3, P. Terpstra3, R.F. Riemersma-van der Lek4, N. Antypa5, C.A. Hartman1,2,4, H. Snieder3, B.W.J. Penninx6, R.A. Schoevers1,2

1. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Gronin- gen, the Netherlands 2. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Groningen, the Netherlands 3. Department of Epidemiology, University of Groningen, University Medical Center Groningen, the Netherlands 4. Interdisciplinary Center Psychopathology and Emotion Regulation (ICPE), University Medical Center Groningen, University of Groningen, Groningen, the Netherlands 5. Department of Clinical Psychology, Institute of Psychology, Leiden University, Leiden, the Netherlands 6. Department of Psychiatry, EMGO Institute for Health and Care Research and Neuro- science Campus Amsterdam, the Netherlands

In preparation Abstract Objective To identify genes for major depression disorder (MDD) by investigating associations of genetic markers in 338 circadian genes with chronotype and mood disorder, in the Netherlands Study of Depression and Anxiety (NESDA), using GAIN-NESDA-NTR GWAS data encompassing 1352 cases and 1649 controls.

Methods The genetic markers were first tested for association with chronotype, and second for association with mood disorder. For markers significantly associated with mood disorder, it was tested whether chronotype was mediating this association using the Sobel test and similarly the mediation effect of mood disorder on the significant associations between marker and chronotype was assessed. The identified genes including four previously identified candidate genes for MDD TNF, NPY, C5orf20 (DCANP1) and SLC6A2 (3) were then subjected to an over-representation analysis to investigate their relation with biochemical pathways and disease processes.

Results and Discussion We found 13 genetic markers in six genes from the circadian gene set to be associated with chronotype, remaining significant after correction for multiple testing. In a similar fashion we found 59 genetic markers in 18 genes from the circadian gene set to be associated with mood disorder, also remaining significant after correction for multiple testing. A subsequent analysis showed that none of these associations were mediated by the other phenotype. Over-representation analyses failed to identify gene clusters indicative of specific neuronal processes, but it yielded several significant functional clusters involved in metabolic disorders.

20 Introduction According to the Global Burden of Disease Study 2010 mood disorders are expected to become one of the leading causes for disability in the western world (1). Mood disorders have a highly heterogeneous character with both genetic and environmental factors playing a role. For example, the genetic component of major depressive disorder (MDD) as estimated from twin studies is approximately 30%, but the genetic profile (including gene-gene interactions and gene-environment interactions) for MDD is hith- erto largely unknown. Genome-wide association studies (GWAS) for MDD have been largely unsuccessful (2–4), although a recent GWAS identified significant associations with 44 loci (5).

Previously we have used genome-wide association data (GWAS) data from the Neth- erlands Study of Depression and Anxiety (NESDA) study (6) to explore reported candidate gene and single-nucleotide polymorphism (SNP) associations in MDD (7,8). These studies could only replicate 2 out of 92 (2%) and 9 out of 185 (5%) of the SNPs, respectively. Replication rates improved to 7% and 20%, respectively, when all genetic markers in the genes were analyzed. The poor replication of candidate genes is most likely attributable to the heterogeneity of MDD, which is also reflected by the many pathophysiological hypotheses that have been raised in the past such as malfunction- ing mono-amine, stress and immune systems. Another major hypothesis of MDD con- cerns adaptive/restorative processes such as neuroplasticity and neurogenesis. The latter processes may be particularly important during sleep which is often disturbed in MDD. All physiological functions including sleep, activity, appetite and secretion of hormones are controlled by the circadian rhythm which is regulated by the suprachiasmatic nucleus (SCN) (9). The SCN consists of a network of transcriptional-translational feedback loops that gives a rhythmic expression pattern of clock genes (9,10). This physiological process is essential for mental well-being in both humans and animals (11–14). Mood may particularly vary with changes and disruptions of the biological clock (11,14–17). Furthermore, it is clear that restoring circadian rhythms has a beneficial effect on depressive symptoms. For example, the efficacy of light therapy for both seasonal affective disorder (SAD) and non-seasonal depression might suggest that restoring circadian rhythms is relevant for the treatment of mood disorders (18–20). A non-invasive method to study the circadian rhythm in individuals is studying the chronotype of a person (21). Chronotype consists of a spectrum with people being a morning type, preferring daytime activities and people being an evening type, preferring nighttime activities. Chronotype has been linked to mood disorders (22–26) and is a heritable trait (27).

The aim of the present study is to identify genes for mood disorders by investigating the association between chronotype and mood disorder with genetic markers from circadian rhythm related genes in the GAIN-NESDA-NTR GWAS sample (6). In addition in case of a significant association with mood disorder it was tested whether this effect was mediated by chronotype and vice versa, whether the associations with chronotype were mediated by mood disorder.

21 Chapter 2: Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data

Material and methods Population Subjects were derived from the Netherlands Study of Depression and Anxiety (NES- DA), an ongoing cohort study (n = 2981 at baseline, age ranging from 18 to 65 years) including 2329 persons with a lifetime diagnosis of a depressive and/or anxiety disorder, as well as 652 healthy controls. Participants were recruited from the community (19%), general practice (54%) and secondary mental health care (27%). All ethical committees of the participating universities approved the NESDA research protocol and all participants provided written informed consent. For a detailed description of NESDA see Penninx et al. 2008 (28). Participants who participated at the 2-year follow-up assessment were included in this study (n = 2596).

Mood disorder Mood disorder diagnosis was determined at baseline and 2-year follow-up using the Composite International Diagnostic Interview (29). Subjects with a lifetime diagnosis of major depressive disorder, bipolar disorder or dysthymia were classified as the mood disorder group (n=1352). Subjects from NESDA without lifetime diagnosis of major depressive disorder, bipolar disorder or dysthymia and a low depressive score (<15) on the Inventory of Depressive Symptomatology (Self-Rating) were classified as subjects without mood disorder (n=1649) (30).

Chronotype Chronotype data of both cases and controls were available in the form of the Munich Chronotype Questionnaire (MCTQ). The MCTQ is a self-report questionnaire consisting of questions regarding the timing of sleep on free and workdays (31–34). From the MCTQ the midpoint in time between falling asleep and waking up at free days can be calculated, the Midpoint of Sleep on Free days (MSF). The MSF needs to be corrected for oversleep on free days in later chronotypes, as subjects with a later chronotype often sleep longer on free days due to accumulated sleep debt during work days. This is done by subtracting half of the difference between sleep duration on free days and workdays and results in the Mid Sleep on Free days, sleep-debt correct (MSFsc). The MSFsc is a measure of the chronotype of a person, validated with other chronotype questionnaires and internal phase markers.

Genetic data DNA was isolated using the GENTRA Puregene kit following the manufacturer’s protocols (6). Genotyping was done with the Affymetrix Perlegen 5.0 (N=1803) and Affymetrix 6.0 (N=2372) arrays (1610 samples were genotyped on both). Genotype calling was performed with the APT-Genotyper software. Sample and SNP QC was done first within, and then between platforms using the PLINK software (35). With the LiftOver tool (“http://genome.sph.umich.edu/wiki/LiftOver”) the positions of the SNPs were converted to build 37 (HG19) of the Human reference genome for each platform. Strands were aligned using

22 Chapter 2: Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data

the 1000 Genomes phase1 release v3 global reference panel. Genetic markers were excluded if they had ambiguous locations, mismatching alleles with the 1000 Genomes reference set or the allele frequencies differed more than 0.20 with those from the reference. Per platform genetic markers were excluded if the minor allele frequency (MAF) <1%, the Hardy–Weinberg Equilibrium (HWE) p-value < 0.00001 or the call rate <95%. Samples were removed when their reported sex did not match with their genotypes, call rate <90% or the F inbreeding value was >0.10 or <−0.10. Subsequently, the data of the individual’s arrays were merged into a single dataset. The HWE, MAF and the reference allele frequency difference filters were next re-applied in the combined data. C/G and A/T SNPs with a MAF between 0.35 and 0.50 were removed to avoid incorrect strand align- ment. Phasing of all samples and imputation of missing calls of genotyped markers was done with MACH (36). The phased data were next imputed with MINIMAC (37). To avoid issues arising from merging genotype data from different platforms re-imputed calls for all genotyped markers were taken (mean concordance between re-imputed markers and the original genotypes was 0.9868). The resulting genome-wide data set consisted of 31,316,056 imputed genetic markers with a mean imputation quality R2 of 0.38.

Translational genetic data In the supplement of Menger et al. (38) rat genes are listed that could be involved in circadian rhythms. For this study we selected the rat genes mentioned in their suppl. tables S1-44 and legends of Fig1 and Fig2 resulting in 336 unique Rat Locus Link-IDs (LLIDs), Using the “Biomart” tool at the Ensembl website (www.ensembl.org) we found the human homologues of these rat LLID’s. Fourteen of these RAT LLIDs map to 2 HU- MAN Ensembl-IDs. The total of unique HUMAN Ensembl-IDs is potentially 350. Two of the RAT LLIDs could not be found in Ensembl at all. Eight of the RAT LLIDs could not be found with a reliable HUMAN homolog. Seven of the human gene equivalents could not be found on our Illumina gene expression array, leaving us with a set of 333 human circadian genes to test for expression differences between our experimental conditions. Besides these 333 genes, we also selected 5 serotonergic genes previously reported to be involved in circadian rhythm, namely SERT, HTR1A, HTR1B, HTR2C and HTR7 (39–45) leaving us with a set of 338 circadian and serotonergic candidate genes to be investigated in the GAIN-NESDA-NTR GWAS sample.

Statistical analysis Phenotype data were prepared using SPSS version 22. Group differences were assessed by analysis of variance (ANOVA) for continuous data and chi-square analyses on categorical data. Genotype-phenotype association analysis was performed with SNPTEST version 2.5 using an additive SNP model (46). Genetic markers in the circadian genes were first tested for association with chronotype, and second for association with mood disorder. As chronotype differs between men and women and over age groups, sex and age at the moment of chronotype assessment are included as covariates. Furthermore chronotype as measured with the MCTQ can be influenced by external timing factors, such as the presence of children in the household and whether a subject is currently employed. These variables were therefore also included as covariates in the analyses.

23 In addition three principle components to control for population stratification and four dummy variables to correct for different genotyping platform were added. Genetic markers were regarded as significant if the p-value was <0.00015 to correct for multiple testing.

For genetic markers significantly associated with mood disorder, it was tested whether chronotype was mediating this association using the Sobel’s test with 10,000 boot- strap permutations to determine the p-value. Similarly the mediation effect of mood disorder on the significant associations between marker and chronotype was assessed. For these analyses the residuals were first computed for chronotype, corrected for the same covariates as used in the previous analyses.

Bioinformatics analysis To investigate the relation of significant genes with biochemical pathways and disease processes they were subjected to both Genetrail-2V1.5 gene set enrichment (GSEA) and overrepresentation analyses (ORA) (47). To this end the default categories, GO - Bi- ological Process, GO - Cellular Component, GO - Molecular Function, KEGG – Pathways and Wiki Pathways were used. In the Genetrail ORA we used false discovery rate (FDR) adjustment (48) and a significance level of 0.05, with 2 as minimum size of category and 700 as maximum size of category.

Results We found 13 genetic markers in 6 genes from the circadian gene set to be associated with chronotype, remaining significant after correction for multiple testing. In a similar fashion we found 59 markers in 18 genes from the circadian gene set to be associated with mood disorder, also remaining significant after correction for multiple testing. A subsequent analysis with Sobel’s test showed that two of the associations with chronotype might be mediated by the mood disorder, but these did not remain significant after multiple testing correction for the six genes (table 1). None of the effects of the genetic markers associated with mood disorder were mediated by chronotype.

In particular many genes associated with mood disorder appeared to be involved in neuronal signaling and/or plasticity (table 1). When subjecting the genes found to be significantly associated with circadian rhythm to gene ontology (GO) programs the involvement in circadian rhythmicity was confirmed. Other significant processes/functions were circadian gene expression, transmembrane helix (receptor, transporter and channel), synapse, corticosteroid receptor and behavior. However, GO analyses of the genes significantly associated with either chronotype or mood disorder failed to identify gene clusters indicative of specific neuronal processes. Yet, an ORA of the significantly associated circadian genes supplemented with previously identified genes for MDD TNF, NPY, C5orf20 (DCANP1) and SLC6A2 (7) yielded several significant functional clusters pointing at a genetic connection between depression and metabolic syndrome (table 2).

24 Table 1: Circadian genes significantly associated with mood disorder or chronotype (p<0.00015 after Bon- ferroni correction) and previous results from literature: 1) number of significant genetic markers in the gene in this study, 2) protein function, 3) reported connection with mood disorder, 4) reported connection with hypotheses of depression, 5) references.

ĐŝƌĐĂĚŝĂŶŐĞŶĞ EƌŽĨ ďŝŽůŽŐŝĐĂůƉƌŽĐĞƐƐͬĨƵŶĐƚŝŽŶ ŵŽŽĚ ƐƚƌĞƐƐ ŝŶĨůĂŵͲ ŶĞƵƌŽͲ ŶĞƵƌŽͲ ŵĞƚĂͲ ƌĞĨ ŵĂƌŬĞƌƐ ĚŝƐ͘ ŵĂƚŝŽŶ ƚƌĂŶƐ ŐĞŶĞƐŝƐͬ ďŽůŝĐ ŵŝƐƐŝŽŶ ƉůĂƐƚŝĐŝƚǇ ϭ Ϯ ϯ ϰ ϰ ϰ ϰ ϰ ϱ DŽŽĚĚŝƐŽƌĚĞƌ /ϯ ϭ EƉƌŽƚĞŝŶďŝŶĚŝŶŐŝŶŚŝďŝƚŽƌ н н н ;ϰϵͿ W'D Ϯ ƐĞƌƵŵ ƉƌŽƚĞŝŶ ĂŶĚ ƌĞĚ ĐĞůů н ;ϱϬͿ ĞŶǌǇŵĞ &DKϭ ϭ ĨůĂǀŝŶ ĐŽŶƚĂŝŶŝŶŐ ŵŽŶŽ ŽǆǇŐĞŶĂƐĞϭ ,<Ϯ Ϯ ŐůǇĐŽůǇƐŝƐ н K>ϯϭ ϭ н н ;ϱϭͿ &'&ϭ ϭ ŐƌŽǁƚŚĨĂĐƚŽƌ н н ;ϱϮͿ s'& ϲ ŐƌŽǁƚŚĨĂĐƚŽƌ н ;ϱϯ͕ϱϰ Ϳ WdWZEϮ ϰ ǀŝƚĂŵŝŶ ƌĞŐƵůĂƚĞĚ ƉƌŽƚĞŝŶͬ н н ;ϱϱͿ ŝŶƐƵůŝŶƐĞĐƌĞƚŝŽŶĂŶĚƉƌŽĐĞƐƐŝŶŐ Zϭ Ϯ ƉƌĞƐǇŶĂƉƚŝĐ ĂŶĚ ŶĞƵƌŽŶĂů н ;ϱϲͿ ƐŝŐŶĂůŝŶŐ Eϭ ϭ >ͲƚǇƉĞ ǀŽůƚĂŐĞͲŐĂƚĞĚ ĐĂůĐŝƵŵ н н н ;ϱϳ͕ϱϴ ĐŚĂŶŶĞů Ϳ ZEd>Ϯ ϭ ĐůŽĐŬŐĞŶĞ н ;ϱϵͿ EZ'Ϯ ϭϮ ĂƐƚƌŽĐǇƚĞ ƉƌŽůŝĨĞƌĂƚŝŽŶ͕ н н н ĚŝĨĨĞƌĞŶƚŝĂƚŝŽŶ͕ ƚƌĂŶƐŵĞŵďƌĂŶĞ ;ϲϬ– ƚƌĂŶƐƉŽƌƚ ϲϮͿ /d' ϰ ĂƵƚŽŝŵŵƵŶĞĚŝĂďĞƚĞƐ н ;ϲϯͿ &^E ϭ ĨĂƚƚǇĂĐŝĚƐǇŶƚŚĂƐĞ н н н ;ϲϰ͕ϲϱ Ϳ >'Wϭ ϵ ƉƌŽƚĞŝŶ ŝŶƚĞƌĂĐƚŝŶŐ ǁŝƚŚ ƉŽƐƚͲ н н н ;ϲϲ͕ϲϳ ƐǇŶĂƉƚŝĐĚĞŶƐŝƚǇ Ϳ ^zdϯ Ϯ ƉƌĞƐǇŶĂƉƚŝĐ ĐĂůĐŝƵŵ н ƐĞŶƐŽƌͬƐǇŶĂƉƚŝĐ ǀĞƐŝĐůĞ ;ϲϴ͕ϲϵ ĨƵƐŝŽŶͬŶĞƵƌŽƚƌĂŶƐŵŝƚƚĞƌƌĞůĞĂƐĞ Ϳ Kyd ϴ ƉŝƚƵŝƚĂƌǇŶĞƵƌŽƉĞƉƚŝĚĞŽǆǇƚŽĐŝŶ н н н ;ϳϬ͕ϳϭ Ϳ 'WZϱϬ ϭ yͲůŝŶŬĞĚ ŽƌƉŚĂŶ ' ƉƌŽƚĞŝŶͲ н ;ϳϮ– ĐŽƵƉůĞĚƌĞĐĞƉƚŽƌ ϳϰͿ ŚƌŽŶŽƚǇƉĞ ϭ >ͲƚǇƉĞ ǀŽůƚĂŐĞͲĚĞƉĞŶĚĞŶƚ н н ;ϳϱ͕ϳϲ Ύ EϮϭ ĐĂůĐŝƵŵ ĐŚĂŶŶĞůͬŵĞŵďƌĂŶĞ Ϳ ĚĞƉŽůĂƌŝǌĂƚŝŽŶ Ϯ ϭ ĐĂƌďŽŶŝĐĂŶŚǇĚƌĂƐĞϮ н н ;ϳϳ– ϳϵͿ KD ϰ ŽƐƚĞŽŵŽĚƵůŝŶ WDWϮϮ ϭ ƉĞƌŝƉŚĞƌĂůŵǇĞůŝŶƉƌŽƚĞŝŶͲϮϮ н ;ϴϬ͕ϴϭ Ϳ ϰϰΎ ϱ ĐĞůůͲƐƵƌĨĂĐĞ ŐůǇĐŽƉƌŽƚĞŝŶ н н ŝŶǀŽůǀĞĚŝŶĐĞůůͲĐĞůůŝŶƚĞƌĂĐƚŝŽŶƐ͕ ;ϴϮ– ĐĞůůĂĚŚĞƐŝŽŶĂŶĚŵŝŐƌĂƚŝŽŶ ϴϱͿ WZ< ϭ ƐĞƌŝŶĞͬƚŚƌĞŽŶŝŶĞŬŝŶĂƐĞ͕ н н ;ϴϲ– ϴϴͿ *Genetic markers within CACNA2D1 (rs149907348) and CD44 (11:35218084:TA_T) showed suggestive mediation of the effect of the marker on chronotype by mood disorder (p=0.047 and p=0.041, respectively).

25 Chapter 2: Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data

Table 2: Over-representation analysis of genes significantly associated with (A) chronotype or (B) mood disorder + four previously identified MDD genes NPY, TNF, C5orf20 and SLC6A2

ͲŚƌŽŶŽƚǇƉĞ 'K–ĞůůƵůĂƌŽŵƉŽŶĞŶƚ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ tŝŬŝWĂƚŚǁĂǇƐ ŶƌŝĐŚĞĚ DŽŶŽĂŵŝŶĞƚƌĂŶƐƉŽƌƚ ƉсϬ͘Ϭϭϱϰ ^>ϲϮ͕dE& <''WĂƚŚǁĂǇƐ ŶƌŝĐŚĞĚ ĚŝƉŽĐǇƚŽŬŝŶĞƐŝŐŶĂůŝŶŐƉĂƚŚǁĂǇ ƉсϬ͘Ϭϰϰϵ EWz͕dE& ŶƌŝĐŚĞĚ ŝůĂƚĞĚĐĂƌĚŝŽŵǇŽƉĂƚŚǇ ƉсϬ͘Ϭϰϰϵ EϮϭ͕dE& ŶƌŝĐŚĞĚ ,ǇƉĞƌƚƌŽƉŚŝĐĐĂƌĚŝŽŵǇŽƉĂƚŚǇ;,DͿ ƉсϬ͘Ϭϰϰϵ EϮϭ͕dE& 'K–ŝŽůŽŐŝĐĂůWƌŽĐĞƐƐ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ 'K–DŽůĞĐƵůĂƌ&ƵŶĐƚŝŽŶ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ ZĞĂĐƚŽŵĞͲWĂƚŚǁĂǇƐ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ –DŽŽĚĚŝƐŽƌĚĞƌнƉƌĞǀŝŽƵƐůǇŝĚĞŶƚŝĨŝĞĚDŐĞŶĞƐ 'K–ĞůůƵůĂƌŽŵƉŽŶĞŶƚ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ tŝŬŝWĂƚŚǁĂǇƐ ŶƌŝĐŚĞĚ &ŽĐĂůĚŚĞƐŝŽŶ ƉсϬ͘Ϭϯϳϲ K>ϯϭ͕/d'͕s'& ŶƌŝĐŚĞĚ /ŶƚĞŐƌĂƚĞĚWĂŶĐƌĞĂƚŝĐĂŶĐĞƌWĂƚŚǁĂǇ ƉсϬ͘Ϭϯϳϲ &'&ϭ͕dE&͕s'& ŶƌŝĐŚĞĚ DŽŶŽĂŵŝŶĞdƌĂŶƐƉŽƌƚ ƉсϬ͘Ϭϯϳϲ ^>ϲϮ͕dE& <''WĂƚŚǁĂǇƐ ŶƌŝĐŚĞĚ dǇƉĞ//ĚŝĂďĞƚĞƐŵĞůůŝƚƵƐ ƉсϬ͘ϬϬϭϭ Eϭ͕,<Ϯ͕dE& ŶƌŝĐŚĞĚ ŵŝŶŽ ƐƵŐĂƌ ĂŶĚ ŶƵĐůĞŽƚŝĚĞ ƐƵŐĂƌ ƉсϬ͘ϬϯϴϬ ,<Ϯ͕W'Dϭ ŵĞƚĂďŽůŝƐŵ ŶƌŝĐŚĞĚ 'ĂůĂĐƚŽƐĞŵĞƚĂďŽůŝƐŵ ƉсϬ͘ϬϯϴϬ ,<Ϯ͕W'Dϭ ŶƌŝĐŚĞĚ DW<ƐŝŐŶĂůŝŶŐƉĂƚŚǁĂǇ ƉсϬ͘ϬϯϴϬ Eϭ͕&'&ϭ͕dE& ŶƌŝĐŚĞĚ WƌŽƚĞŽŐůǇĐĂŶƐŝŶĐĂŶĐĞƌ ƉсϬ͘ϬϯϴϬ &'&ϭ͕dE&͕s'& ŶƌŝĐŚĞĚ dǇƉĞ/ĚŝĂďĞƚĞƐŵĞůůŝƚƵƐ ƉсϬ͘ϬϯϴϬ WdWZEϮ͕dE& ŶƌŝĐŚĞĚ 'ůǇĐŽůǇƐŝƐͬ'ůƵĐŽŶĞŽŐĞŶĞƐŝƐ ƉсϬ͘ϬϰϯϬ ,<Ϯ͕W'Dϭ ŶƌŝĐŚĞĚ W/ϯ<ͲŬƚƐŝŐŶĂůŝŶŐƉĂƚŚǁĂǇ ƉсϬ͘ϬϰϯϬ K>ϯϭ͕&'&ϭ͕s'& ŶƌŝĐŚĞĚ ^ƚĂƌĐŚĂŶĚƐƵĐƌŽƐĞŵĞƚĂďŽůŝƐŵ ƉсϬ͘ϬϰϯϬ ,<Ϯ͕W'Dϭ ŶƌŝĐŚĞĚ ŵdKZƐŝŐŶĂůŝŶŐƉĂƚŚǁĂǇ ƉсϬ͘ϬϰϯϬ dE&͕s'& 'K–ŝŽůŽŐŝĐĂůWƌŽĐĞƐƐ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ 'K–DŽůĞĐƵůĂƌ&ƵŶĐƚŝŽŶ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ ZĞĂĐƚŽŵĞͲWĂƚŚǁĂǇƐ EŽƐŝŐŶŝĨŝĐĂŶƚĐĂƚĞŐŽƌŝĞƐŚĂǀĞďĞĞŶĨŽƵŶĚ

26 Chapter 2: Association of circadian genes with chronotype and mood disorder, an analysis of epidemiological and translational data

Discussion This study was set out to investigate associations of genes involved in circadian rhythm with chronotype and mood disorder using GWAS data from the GAIN-NESDA-NTR study in order to identify genes for MDD through two endophenotypes. When analyzing the genetic markers from the circadian genes 13 markers in six genes were found to be to be significantly associated with chronotype after Bonferroni correction (see table 1). In a similar fashion 59 markers in 18 genes were found to be associated with mood disorder, also remaining significant after correction for multiple testing. A subsequent analysis with Sobel’s test showed that only genetic markers within CACNA2D1 (rs149907348) and CD44 (11:35218084:TA_T) showed suggestive mediation of the effect of the marker on chronotype by mood disorder (p=0.047 and p=0.041, respectively).

The 338 candidate genes from the present study were in large part the human homologues of reported rat circadian genes (38). We have used these human homologues before in a gene expression study of seasonal affective disorder (n=15) but were not able to find significant associations, likely due to the small sample size (89).

Many of the genes associated with mood disorder appeared to be involved in neuronal signaling and/or plasticity. However, over-representation analysis failed to identify gene clusters indicative of specific neuronal processes. The functional categories for the rat circadian genes in Menger et al. (38) seem to be constructed by the authors from the individual gene descriptions and are difficult to compare with established GO programs such as Genetrail. Accordingly we also included five serotonergic genes (SERT, HT1A, HTR1B, HTR2C and HTR7) clearly involved in circadian rhythm (39–45).

Circadian genes that were significantly associated with chronotype or mood disorder were subjected to a GSEA and ORA. Both types of GO analyses confirmed the involvement in circadian rhythmicity. Other significant processes/functions were circadian gene expression, transmembrane helix (receptor, transporter and channel), synapse, corticosteroid receptor and behavior. Because the current and previous candidate gene study (7) were performed on the same GAIN-NESDA-NTR GWAS data set we have merged the positive genes from both studies and subjected them again to a GSEA and ORA. The GSEA did not yield significant results but the perhaps somewhat less stringent ORA yielded several significant and possibly relevant functional clusters. The first one, driv- en by TNF and the noradrenaline transporter, suggests the involvement of monoamine neurotransmission, which is in line with current hypotheses of MDD (90). The second one points at the adipocytokine pathway (NPY, TNF), which plays a role in inflammatory processes, but also in metabolic syndrome (91,92). The third cluster presents a combination of genes from the current and previous study pointing at sugar metabolism (HK2, PGM1), diabetes I (PTPRN2, TNF) and II (CACNA1C, HK2, TNF), dilated cardiomyopathy (CACNA2D1,TNF) and hypertrophic cardiomyopathy (CACNA2D1,TNF). The results from the ORA suggest an association of depression with metabolic syndrome (diabetes II, cardiovascular disease and adipocytokine pathways). A meta-analysis by Pan et al. (93) reported indeed a strong bidirectional association between depressive disorders and metabolic syndrome. In both disorders dysregulation of the immune system

27 and disruptions of the cytokine network have been observed. One of the suggested mechanisms underlying both disorders might be HPA-axis hyperactivity triggered by pro-inflammatory cytokines. Another mechanism that has been put forward involves adipocytokines, such as leptin, adiponectin and resistin. Low levels of leptin have been associated with depression in humans as well as depressive-like behavior in animals (94). This has led to the formulation of a “leptin hypothesis of depression” with both leptin insufficiency and resistance as important factors (94). The MAPK, PI3K-Akt and mTOR pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG) as identified in the ORA converge to T-cell receptor signaling (95) and metabolic programming (96). Moreover, focal adhesion as identified in the Wiki pathways analysis might also hint at a role of the immune system, in which T-lymphocytes migrate along the connec- tive endothelium following cellular signals to damaged biological tissue (97). It is also important to note that the clock gene ARNTL2, while not being identified in the ORA analysis, has been implicated in both diabetes I and II (98,99).

The present study emphasizes the potential of bioinformatics to integrate previously gathered preclinical information into the analysis of gene associations with mental and somatic diseases. Globally the incidence of both mood disorder and metabolic syndrome related diseases such as obesity, diabetes and cardio-vascular disease is rapidly increasing. Each condition separately, and their combination even more, is responsible for an impressive disease burden affecting the lives of many. It is important to note that the present study was set out to investigate association of circadian genes with depression through the endophenotypes chronotype and mood disorder, and definitely not aimed at investigating a genetic association between depression and metabolic syndrome. Thus the outcome of the ORA came rather unexpected. However, the many hints at sugar metabolism in the KEGG pathway analysis are in support of current ideas that the globally increased sugar intake constitutes a risk factor for developing both metabolic syndrome and mood disorder (100). Yet, it is also clear that the present results have to be replicated by other studies.

28 matic nucleus organization. Cell Tissue Res. References 2002;309(1):89–98. 1. Whiteford HA, Degenhardt L, Rehm J, Baxter 11. Barnard AR, Nolan PM. When Clocks Go AJ, Ferrari AJ, Erskine HE, et al. Global bur- Bad: Neurobehavioural Consequences of den of disease attributable to mental and Disrupted Circadian Timing. PLoS Genet. substance use disorders: findings from the 2008;4(5):e1000040. DOI: 10.1371/journal. Global Burden of Disease Study 2010. Lancet. pgen.1000040 2013;382(9904):1575–86. DOI: 10.1016/ 12. Bunney W. Molecular Clock Genes in Man and s0140-6736(13)61611-6 Lower Animals Possible Implications for Circa- 2. Ripke S, Wray NR, Lewis CM, Hamilton SP, dian Abnormalities in Depression. Neuropsy- Weissman MM, Breen G, et al. A mega-anal- chopharmacology. 2000;22(4):335–45. DOI: ysis of genome-wide association studies for 10.1016/s0893-133x(99)00145-1 major depressive disorder. Mol Psychiatry. 13. McClung CA. Circadian rhythms and mood 2012;18(4):497–511. DOI: 10.1038/mp.2012.21 regulation: Insights from pre-clinical models. Eur 3. Okbay A, Baselmans BML, De Neve J-E, Turley P, Neuropsychopharmacol. 2011;21:S683–93. DOI: Nivard MG, Fontana MA, et al. Genetic variants 10.1016/j.euroneuro.2011.07.008 associated with subjective well-being, de- 14. McClung C a. Circadian genes, rhythms and the pressive symptoms and neuroticism identified biology of mood disorders. Pharmacol Ther. through genome-wide analyses. Nat Genet. 2007 May;114(2):222–32. DOI: 10.1016/j. 2016;48(6):624–33. DOI: 10.1038/ng.3552 pharmthera.2007.02.003 4. Direk N, Williams S, Smith JA, Ripke S, Air T, Am- 15. Monteleone P, Maj M. The circadian basis of are AT, et al. An Analysis of Two Genome-wide mood disorders: Recent developments and Association Meta-analyses Identifies a New treatment implications. Eur Neuropsychophar- Locus for Broad Depression Phenotype. Biol macol. 2008;18(10):701–11. DOI: 10.1016/j. Psychiatry. 2017;82(5):322–9. DOI: 10.1016/j. euroneuro.2008.06.007 biopsych.2016.11.013 16. Hasler BP, Buysse DJ, Kupfer DJ, Germain A. 5. Wray NR, Sullivan PF. Genome-wide association Phase relationships between core body tem- analyses identify 44 risk variants and refine the perature, melatonin, and sleep are associated genetic architecture of major depression [Inter- with depression severity: Further evidence for net]. Cold Spring Harbor Laboratory; 2017. DOI: circadian misalignment in non-seasonal depres- 10.1101/167577 sion. Psychiatry Res. 2010;178(1):205–7. DOI: 6. Boomsma DI, Willemsen G, Sullivan PF, Heutink 10.1016/j.psychres.2010.04.027 P, Meijer P, Sondervan D, et al. Genome-wide 17. Boivin DB, Czeisler CA, Dijk D-J, Duffy JF, Folkard association of major depression: description of S, Minors DS, et al. Complex interaction of the samples for the GAIN Major Depressive Disorder sleep-wake cycle and circadian phase modu- Study: NTR and NESDA biobank projects. Eur J lates mood in healthy subjects. Arch Gen Psychia- Hum Genet. 2008;16(3):335–42. DOI: 10.1038/ try. 1997;54(2):145–52. sj.ejhg.5201979 18. Benedetti F, Dallaspezia S, Fulgosi MC, Barbini 7. Bosker FJ, Hartman C a, Nolte IM, Prins BP, B, Colombo C, Smeraldi E. Phase advance is an Terpstra P, Posthuma D, et al. Poor replication of actimetric correlate of antidepressant response to candidate genes for major depressive disor- sleep deprivation and light therapy in bipolar de- der using genome-wide association data. Mol pression. Chronobiol Int. 2007 Jan;24(5):921–37. Psychiatry. 2011;16(5):516–32. DOI: 10.1038/ DOI: 10.1080/07420520701649455 mp.2010.38 19. Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Goodwin 8. Luo X, Stavrakakis N, Penninx BW, Bosker FJ, FK, Davenport Y, et al. Seasonal affective disorder: Nolen WA, Boomsma DI, et al. Does refining the a description of the syndrome and preliminary phenotype improve replication rates? A review findings with light therapy. Arch Gen Psychiatry. and replication of candidate gene studies on 1984;41(1):72. Major Depressive Disorder and Chronic Major 20. Terman M. Evolving applications of light therapy. Depressive Disorder. Am J Med Genet Part B Sleep Med Rev. 2007;11(6):497–507. DOI: Neuropsychiatr Genet. 2015;171(2):215–36. 10.1016/j.smrv.2007.06.003 DOI: 10.1002/ajmg.b.32396 21. Wirz-Justice A. How to measure circadian rhythms 9. Albrecht U. Timing to perfection: the biology of in humans. Medicographia. 2007;29:84–90. central and peripheral circadian clocks. Neuron. 22. Ahn YM, Chang J, Joo YH, Kim SC, Lee KY, Kim YS. 2012 Apr 26;74(2):246–60. DOI: 10.1016/j. Chronotype distribution in bipolar I disorder and neuron.2012.04.006 schizophrenia in a Korean sample. Bipolar Disord. 10. Moore RY, Speh JC, Leak RK. Suprachias- 2008 Mar;10(2):271–5. 23. Mansour H a, Wood J, Chowdari K V, Dayal M,

29 Thase ME, Kupfer DJ, et al. Circadian phase vari- tion-Based Linkage Analyses. Am J Hum Genet. ation in bipolar I disorder. Chronobiol Int. 2005 2007;81(3):559–75. DOI: 10.1086/519795 Jan;22(3):571–84. DOI: 10.1081/CBI-200062413 36. Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR. MaCH: 24. Hidalgo MP, Caumo W, Posser M, Coccaro SB, using sequence and genotype data to estimate Camozzato AL, Chaves MLF. Relationship between haplotypes and unobserved genotypes. Genet depressive mood and chronotype in healthy Epidemiol. 2010;34(8):816–34. DOI: 10.1002/ subjects. 2009;63:283–90. gepi.20533 25. Kitamura S, Hida A, Watanabe M, Enomoto M, 37. Howie B, Fuchsberger C, Stephens M, Marchini J, Aritake-Okada S, Moriguchi Y, et al. Evening Pref- Abecasis GR. Fast and accurate genotype imputa- erence Is Related To the Incidence of Depressive tion in genome-wide association studies through States Independent of Sleep-Wake Conditions. pre-phasing. Nat Genet. 2012;44(8):955–9. DOI: Chronobiol Int. 2010;27(9–10):1797–812. DOI: 10.1038/ng.2354 10.3109/07420528.2010.516705 38. Menger GJ, Lu K, Thomas T, Cassone VM, Earnest 26. Antypa N, Vogelzangs N, Meesters Y, Schoevers DJ. Circadian profiling of the transcriptome in RA, Penninx BWJH. Chronotype associations with immortalized rat SCN cells. Physiol Genomics. depression and anxiety in a large cohort study. 2005 May 11;21(3):370–81. DOI: 10.1152/physi- Depress Anxiety. 2016 Jan;33(1):75–83. olgenomics.00224.2004 27. Barclay NL, Eley TC, Maughan B, Rowe R, Gregory 39. Cuesta M, Mendoza J, Clesse D, Pévet P, Challet E. AM. Associations between diurnal preference, Serotonergic activation potentiates light resetting sleep quality and externalizing behaviours: of the main circadian clock and alters clock a behavioural genetic analysis. Psychol gene expression in a diurnal rodent. Exp Neurol. Med. 2010;41(5):1029–40. DOI: 10.1017/ 2008;210(2):501–13. DOI: 10.1016/j.expneu- s0033291710001741 rol.2007.11.026 28. Penninx BWJH, Beekman ATF, Smit JH, Zitman 40. Sterniczuk R, Stepkowski A, Jones M, Antle MC. FG, Nolen WA, Spinhoven P, et al. The Nether- Enhancement of photic shifts with the 5-HT1A lands Study of Depression and Anxiety (NESDA): mixed agonist/antagonist NAN-190: Intra-su- rationale, objectives and methods. Int J Methods prachiasmatic nucleus pathway. Neuroscience. Psychiatr Res. 2008;17(3):121–40. 2008;153(3):571–80. DOI: 10.1016/j.neurosci- 29. Wittchen HU. Reliability and validity studies of ence.2008.02.062 the WHO-Composite International Diagnostic 41. Pickard GE, Rea MA. TFMPP, a 5HT1B receptor Interview (CIDI): A critical review. J Psychiatr Res. agonist, inhibits light-induced phase shifts of the 1994;28(1):57–84. circadian activity rhythm and c-Fos expression 30. Rush AJ, Carmody T, Reimitz P. The Inventory in the mouse suprachiasmatic nucleus. Neurosci of Depressive Symptomatology (IDS): Clinician Lett. 1997;231(2):95–8. DOI: 10.1016/s0304- (IDS‐C) and Self‐Report (IDS‐SR) ratings of de- 3940(97)00534-x pressive symptoms. Int J Methods Psychiatr Res. 42. Varcoe TJ, Kennaway DJ. Activation of 5-HT2C 2000;9(2):45–59. receptors acutely induces Per1 gene expression in 31. Zavada A, Gordijn M, Beersma D, Daan S, Roen- the rat SCN in vitro. Brain Res. 2008;1209:19–28. neberg T. Comparison of the Munich Chronotype DOI: 10.1016/j.brainres.2008.02.091 Questionnaire with the Horne‐Östberg’s Morning- 43. Westrich L, Sprouse J, Sánchez C. The effects of ness‐Eveningness score. Chronobiol Int. 2005 Mar combining serotonin reuptake inhibition and 1;22(2):267–78. DOI: 10.1081/CBI-200053536 5-HT7 receptor blockade on circadian rhythm 32. Kantermann T, Sung H, Burgess HJ. Comparing regulation in rodents. Physiol Behav. 2013;110– the Morningness-Eveningness Questionnaire and 111:42–50. DOI: 10.1016/j.physbeh.2012.12.009 Munich ChronoType Questionnaire to the Dim 44. Cuesta M, Clesse D, Pévet P, Challet E. New light Light Melatonin Onset. 2015;30(5):449–53. on the serotonergic paradox in the rat circadian 33. Kantermann T, Eastman CI. Circadian phase, circa- system. J Neurochem. 2009;110(1):231–43. DOI: dian period and chronotype are reproducible over 10.1111/j.1471-4159.2009.06128.x months. Chronobiol Int. 2017;35(2):280–8. DOI: 45. Pickard GE, Rea MA. Serotonergic innervation 10.1080/07420528.2017.1400979 of the hypothalamic suprachiasmatic nucleus 34. Roenneberg T, Wirz-Justice A, Merrow and photic regulation of circadian rhythms. Biol M. Life between Clocks: Daily Tempo- Cell. 1997;89(8):513–23. DOI: 10.1016/s0248- ral Patterns of Human Chronotypes. J Biol 4900(98)80007-5 Rhythms. 2003 Feb 1;18(1):80–90. DOI: 46. Marchini J, Howie B, Myers S, McVean G, Donnelly 10.1177/0748730402239679 P. A new multipoint method for genome-wide 35. Purcell S, Neale B, Todd-Brown K, Thomas L, association studies by imputation of genotypes. Ferreira MAR, Bender D, et al. PLINK: A Tool Set Nat Genet. 2007;39(7):906–13. for Whole-Genome Association and Popula- 47. Backes C, Keller A, Kuentzer J, Kneissl B, Com-

30 tesse N, Elnakady YA, et al. GeneTrail--advanced ysis. Am J Med Genet Part B Neuropsychiatr gene set enrichment analysis. Nucleic Acids Res. Genet. 2016;171(6):896–903. DOI: 10.1002/ 2007;35(Web Server):W186–92. DOI: 10.1093/ ajmg.b.32466 nar/gkm323 58. De Jesús-Cortés H, Rajadhyaksha AM, 48. Benjamini Y, Yekutieli D. Quantitative trait loci Pieper AA. Cacna1c: Protecting young analysis using the false discovery rate. Genetics. hippocampal neurons in the adult brain. 2005;171(2):783–90. Neurogenesis. 2016;3(1):e1231160. DOI: 49. Weder N, Zhang H, Jensen K, Yang BZ, Simen A, 10.1080/23262133.2016.1231160 Jackowski A, et al. Child Abuse, Depression, and 59. Geoffroy PA, Lajnef M, Bellivier F, Jamain S, Gard Methylation in Genes Involved With Stress, Neural S, Kahn J-P, et al. Genetic association study of Plasticity, and Brain Circuitry. J Am Acad Child circadian genes with seasonal pattern in bipolar Adolesc Psychiatry. 2014;53(4):417–424.e5. DOI: disorders. Sci Rep. 2015;5(1). DOI: 10.1038/ 10.1016/j.jaac.2013.12.025 srep10232 50. Beckman G, Beckman L, Cedergren B, Perris 60. Ichikawa T, Nakahata S, Tamura T, Manachai C, Strandman E. Serum Protein and Red Cell N, Morishita K. The loss of NDRG2 expression Enzyme Polymorphisms in Affective Dis- improves depressive behavior through in- orders. Hum Hered. 1978;28(1):41–7. DOI: creased phosphorylation of GSK3β. Cell Signal. 10.1159/000152929 2015;27(10):2087–98. DOI: 10.1016/j.cell- 51. Stankiewicz AM, Goscik J, Majewska A, Swiergiel sig.2015.07.012 AH, Juszczak GR. The Effect of Acute and Chronic 61. Araya-Callís C, Hiemke C, Abumaria N, Flugge Social Stress on the Hippocampal Transcriptome G. Chronic psychosocial stress and citalopram in Mice. PLoS One. 2015;10(11):e0142195. DOI: modulate the expression of the glial proteins 10.1371/journal.pone.0142195 GFAP and NDRG2 in the hippocampus. Psycho- 52. Lee H, Raiker SJ, Venkatesh K, Geary R, Robak LA, pharmacology (Berl). 2012;224(1):209–22. DOI: Zhang Y, et al. Synaptic Function for the Nogo-66 10.1007/s00213-012-2741-x Receptor NgR1: Regulation of Dendritic Spine 62. NICHOLS NR. Ndrg2, a Novel Gene Regulated Morphology and Activity-Dependent Synaptic by Adrenal Steroids and Antidepressants, Is Strength. J Neurosci. 2008;28(11):2753–65. DOI: Highly Expressed in Astrocytes. Ann N Y Acad 10.1523/jneurosci.5586-07.2008 Sci. 2003;1007(1):349–56. DOI: 10.1196/an- 53. Tseng P-T, Cheng Y-S, Chen Y-W, Wu C-K, Lin P-Y. nals.1286.034 Increased levels of vascular endothelial growth 63. Barrie ES, Lodder M, Weinreb PH, Buss J, Rajab A, factor in patients with major depressive disorder: Adin C, et al. Role of ITGAE in the development A meta-analysis. Eur Neuropsychopharmacol. of autoimmune diabetes in non-obese diabetic 2015;25(10):1622–30. DOI: 10.1016/j.euroneu- mice. J Endocrinol. 2015;224(3):235–43. ro.2015.06.001 64. Tsuboi H, Sakakibara H, Yamakawa-Kobayashi 54. Berent D, Macander M, Szemraj J, Orzechowska A, K, Tatsumi A, Inamori T, Hamamoto R, et al. Val- Galecki P. Vascular endothelial growth factor 1483Ile polymorphism in the fatty acid synthase A gene expression level is higher in patients gene was associated with depressive symptoms with major depressive disorder and not under the influence of psychological stress. affected by cigarette smoking, hyperlipidemia J Affect Disord. 2011;134(1–3):448–52. DOI: or treatment with statins. Acta Neurobiol Exp. 10.1016/j.jad.2011.05.010 2014;74(1):82–90. 65. Liang S, Byers DM, Irwin LN. Chronic Mild Stress- 55. Yang B-Z, Han S, Kranzler HR, Farrer LA, ors and Diet Affect Gene Expression Differ- Gelernter J. A Genomewide Linkage Scan of ently in Male and Female Rats. J Mol Neurosci. Cocaine Dependence and Major Depressive 2007;33(2):189–200. DOI: 10.1007/s12031- Episode in Two Populations. Neuropsychophar- 007-0064-x macology. 2011;36(12):2422–30. DOI: 10.1038/ 66. Mathias SR, Knowles EEM, Kent JW, Mckay DR, npp.2011.122 Curran JE, de Almeida MAA, et al. Recurrent major 56. Lenihan JA, Saha O, Heimer-McGinn V, Cryan JF, depression and right hippocampal volume: a bi- Feng G, Young PW. Decreased Anxiety-Related variate linkage and association study. Hum Brain Behaviour but Apparently Unperturbed NUMB Mapp. 2016;37(1):191–202. Function in Ligand of NUMB Protein-X (LNX) 67. Grados MA, Specht MW, Sung H-M, Fortune D. 1/2 Double Knockout Mice. Mol Neurobiol. Glutamate drugs and pharmacogenetics of OCD: 2016;54(10):8090–109. DOI: 10.1007/s12035- a pathway-based exploratory approach. Expert 016-0261-0 Opin Drug Discov. 2013;8(12):1515–27. DOI: 57. Rao S, Yao Y, Zheng C, Ryan J, Mao C, Zhang F, 10.1517/17460441.2013.845553 et al. Common variants inCACNA1Cand MDD 68. Brunger AT. Structure of Proteins Involved in susceptibility: A comprehensive meta-anal- Synaptic Vesicle Fusion in Neurons. Annu Rev

31 Biophys Biomol Struct. 2001;30(1):157–71. DOI: 79. Torella D, Ellison GM, Torella M, Vicinanza 10.1146/annurev.biophys.30.1.157 C, Aquila I, Iaconetti C, et al. Carbonic Anhy- 69. Marqueze B, Boudier JA, Mizuta M, Inagaki drase Activation Is Associated With Worsened N, Seino S, Seagar M. Cellular localization of Pathological Remodeling in Human Ischemic synaptotagmin I, II, and III mRNAs in the central Diabetic Cardiomyopathy. J Am Heart Assoc. nervous system and pituitary and adrenal glands 2014;3(2):e000434–e000434. DOI: 10.1161/ of the rat. J Neurosci. 1995;15(7):4906–17. DOI: jaha.113.000434 10.1523/jneurosci.15-07-04906.1995 80. Kéri S, Szabó C, Kelemen O. Blood biomark- 70. Mottolese R, Redoute J, Costes N, Le Bars D, ers of depression track clinical changes Sirigu A. Switching brain serotonin with oxytocin. during cognitive-behavioral therapy. J Affect Proc Natl Acad Sci. 2014;111(23):8637–42. DOI: Disord. 2014;164:118–22. DOI: 10.1016/j. 10.1073/pnas.1319810111 jad.2014.04.030 71. Sasaki T, Hashimoto K, Oda Y, Ishima T, Yakita 81. Aston C, Jiang L, Sokolov BP. Transcriptional M, Kurata T, et al. Increased Serum Levels of profiling reveals evidence for signaling and oligo- Oxytocin in “Treatment Resistant Depression dendroglial abnormalities in the temporal cortex in Adolescents (TRDIA)” Group. PLoS One. from patients with major depressive disorder. 2016;11(8):e0160767. DOI: 10.1371/journal. Mol Psychiatry. 2005;10(3):309. pone.0160767 82. Ventorp F, Barzilay R, Erhardt S, Samuelsson 72. Ryan J, Carrière I, Ritchie K, Ancelin M-L. Involve- M, Träskman-Bendz L, Janelidze S, et al. The ment of GPR50 polymorphisms in depression: CD44 ligand hyaluronic acid is elevated in the independent replication in a prospective elderly cerebrospinal fluid of suicide attempters and is cohort. Brain Behav. 2015;5(3):n/a-n/a. DOI: associated with increased blood–brain barrier 10.1002/brb3.313 permeability. J Affect Disord. 2016;193:349–54. 73. Delavest M, Even C, Benjemaa N, Poirier M-F, DOI: 10.1016/j.jad.2015.12.069 Jockers R, Krebs M-O. Association of the intronic 83. Patouraux S, Rousseau D, Bonnafous S, Leb- rs2072621 polymorphism of the X-linked GPR50 eaupin C, Luci C, Canivet CM, et al. CD44 is a gene with affective disorder with seasonal key player in non-alcoholic steatohepatitis. J pattern. Eur Psychiatry. 2012;27(5):369–71. DOI: Hepatol. 2017;67(2):328–38. DOI: 10.1016/j. 10.1016/j.eurpsy.2011.02.011 jhep.2017.03.003 74. MacIntyre DJ, McGhee KA, MacLean AW, Afzal 84. Lin D, Chun T-H, Kang L. Adipose extracellular M, Briffa K, Henry B, et al. Association of GPR50, matrix remodelling in obesity and insulin resis- an X-linked orphan G protein-coupled recep- tance. Biochem Pharmacol. 2016;119:8–16. DOI: tor, and affective disorder in an independent 10.1016/j.bcp.2016.05.005 sample of the Scottish population. Neurosci 85. Liu LF, Kodama K, Wei K, Tolentino LL, Choi Lett. 2010;475(3):169–73. DOI: 10.1016/j. O, Engleman EG, et al. The receptor CD44 is neulet.2010.03.072 associated with systemic insulin resistance and 75. Davies A, Kadurin I, Alvarez-Laviada A, Douglas L, proinflammatory macrophages in human adipose Nieto-Rostro M, Bauer CS, et al. The α2δ subunits tissue. Diabetologia. 2015;58(7):1579–86. DOI: of voltage-gated calcium channels form GPI-an- 10.1007/s00125-015-3603-y chored proteins, a posttranslational modifica- 86. Shi Y, Yuan Y, Xu Z, Pu M, Wang C, Zhang Y, et tion essential for function. Proc Natl Acad Sci. al. Genetic variation in the calcium/calmod- 2010;107(4):1654–9. ulin-dependent protein kinase (CaMK) pathway 76. Erk S, Meyer-Lindenberg A, Schnell K, Opitz von is associated with antidepressant response in Boberfeld C, Esslinger C, Kirsch P, et al. Brain females. J Affect Disord. 2012;136(3):558–66. Function in Carriers of a Genome-wide Support- DOI: 10.1016/j.jad.2011.10.030 ed Bipolar Disorder Variant. Arch Gen Psychiatry. 87. Li Y-F, Sun H-X, Wu G-D, Du W-N, Zuo J, Shen Y, et 2010;67(8):803. DOI: 10.1001/archgenpsychia- al. Protein kinase C/ζ (PRKCZ) Gene is associated try.2010.94 with type 2 diabetes in Han population of North 77. Yasukawa Z, Sato C, Kitajima K. Identification of China and analysis of its haplotypes. World J an Inflammation-inducible Serum Protein Recog- Gastroenterol. 2003;9(9):2078. nized by Anti-disialic Acid Antibodies as Carbonic 88. Campbell CSG, Caperuto LC, Hirata AE, Araujo Anhydrase II. J Biochem. 2007;141(3):429–41. EP, Velloso LA, Saad MJ, et al. The phospha- DOI: 10.1093/jb/mvm047 tidylinositol/AKT/atypical PKC pathway is 78. Knudsen JF, Carlsson U, Hammarström P, Sokol involved in the improved insulin sensitivity GH, Cantilena LR. The Cyclooxygenase-2 Inhibitor by DHEA in muscle and liver of rats in vivo. Celecoxib Is a Potent Inhibitor of Human Carbonic Life Sci. 2004;76(1):57–70. DOI: 10.1016/j. Anhydrase II. Inflammation. 2004;28(5):285–90. lfs.2004.06.017 DOI: 10.1007/s10753-004-6052-1 89. Bosker FJ, Terpstra P, Gladkevich A V., Janneke

32 Dijck-Brouwer D a., te Meerman G, Nolen W a., 99. Yamaguchi M, Uemura H, Arisawa K, Katsuu- et al. Changes in winter depression phenotype ra-Kamano S, Hamajima N, Hishida A, et al. As- correlate with white blood cell gene expres- sociation between brain-muscle-ARNT-like pro- sion profiles: A combined metagene and gene tein-2 (BMAL2) gene polymorphism and type 2 ontology approach. Prog Neuro-Psychophar- diabetes mellitus in obese Japanese individ- macology Biol Psychiatry. 2015;58:8–14. DOI: uals: A cross-sectional analysis of the Japan 10.1016/j.pnpbp.2014.10.015 Multi-institutional Collaborative Cohort Study. 90. Schildkraut JJ. The catecholamine hypothesis Diabetes Res Clin Pract. 2015;110(3):301–8. of affective disorders: a review of supporting 100. Hryhorczuk C, Sharma S, Fulton SE. evidence. Am J Psychiatry. 1965;122(5):509–22. Metabolic disturbances connecting obesity 91. Elks CM, Francis J. Central Adiposity, Systemic and depression. Front Neurosci. 2013;7. DOI: Inflammation, and the Metabolic Syndrome. 10.3389/fnins.2013.00177 Curr Hypertens Rep. 2010;12(2):99–104. DOI: 10.1007/s11906-010-0096-4 92. López-Jaramillo P, Gómez-Arbeláez D, López- López J, López-López C, Martínez-Ortega J, Gómez-Rodríguez A, et al. The role of leptin/ adiponectin ratio in metabolic syndrome and diabetes. Horm Mol Biol Clin Investig. 2014;18(1):37–45. 93. Pan A, Keum N, Okereke OI, Sun Q, Kivimaki M, Rubin RR, et al. Bidirectional Association Be- tween Depression and Metabolic Syndrome. Di- abetes Care. 2012 May 1;35(5):1171 LP-1180. 94. Lu X-Y. The leptin hypothesis of depression: a potential link between mood disorders and obesity? Curr Opin Pharmacol. 2007;7(6):648– 52. DOI: 10.1016/j.coph.2007.10.010 95. Yamagishi M, Watanabe T. New paradigm of T cell signalling: learning from malignancies. J Clin Cell Immunol. 2012;12. 96. Zeng H, Yang K, Cloer C, Neale G, Vogel P, Chi H. mTORC1 couples immune signals and metabolic programming to establish Treg-cell function. Nature. 2013;499(7459):485–90. DOI: 10.1038/nature12297 97. Yuan SY, Shen Q, Rigor RR, Wu MH. Neutro- phil transmigration, focal adhesion kinase and endothelial barrier function. Micro- vasc Res. 2012;83(1):82–8. DOI: 10.1016/j. mvr.2011.06.015 98. Hung M-S, Avner P, Rogner UC. Identification of the transcription factor ARNTL2 as a candidate gene for the type 1 diabetes locus Idd6. Hum Mol Genet. 2006;15(18):2732–42. DOI: 10.1093/hmg/ddl209

33 Chapter 3 Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

S.E. Knapen1*, R.F. Riemersma-van der Lek2, N. Antypa3, Y. Meesters2, B.W.J.H. Penninx4 and R.A. Schoevers1

1. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Inter- disciplinary Center for Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands; 2. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Groningen, the Netherlands; 3. Department of Clinical Psychology, Institute of Psychology, Leiden University, Leiden, the Netherlands; 4. Department of Psychiatry, Amsterdam Public Health and Amsterdam Neuroscience Research Institutes, Amsterdam, the Netherlands

Chronobiology International 2018;35(1), 1-7. Abstract Social jetlag, the misalignment between the internal clock and the socially required timing of activities, is highly prevalent, especially in people with an evening chronotype and is hypothesized to be related to the link between the evening chronotype and major depressive disorder. Although social jetlag has been linked to depressive symptoms in non-clinical samples, it has never been studied in patients with major depressive disorder (MDD). This study aimed to study social jetlag in patients with major depressive disorder and healthy controls, and to further examine the link between social jetlag and depressive symptomatology. Patients with a diagnosis of MDD (n = 1084) and healthy controls (n=385), assessed in a clinical interview, were selected from the Netherlands Study of Depression and Anxiety. Social jetlag was derived from the Munich Chronotype Questionnaire, by calculating the absolute difference between the midsleep on free days and midsleep on work days. Depression severity was measured with the Inventory of Depressive Symptomatology. It was found that patients with MDD did not show more social jetlag compared to healthy controls, neither in a model without medication use (β = 0.06, 95% CI: -0.03 – 0.15, p = 0.17) nor in a model where medication use is accounted for. There was no direct association between the amount of social jetlag and depressive symptoms, neither in the full sample, nor in the patient group nor the healthy control group. This first study on social jetlag in a clinical sample showed no differences in social jetlag between patients with MDD and healthy controls.

35 Introduction People with an evening chronotype have a preference for evening activities, a late onset of sleep and like to sleep in longer in the morning. Although they prefer a later timing of their life, society often demands an earlier timing of activities. The large differences that evening chronotypes experience in their timing of sleep during work days and free days leads to a sleep debt during work days and more sleep during free days, as a compensation (1). This difference in timing is termed social jetlag which stands for the misalignment between the internal clock and the socially required timing of activities (1). Social jetlag (SJL) is quantified as the difference between midpoint of the sleep on free days and the midpoint of the sleep on work days. Social jetlag is highly prevalent in the general population, with a peak during the end of adolescence (2). Misalignment of the clock is linked with obesity, an increased cardiovascular risk profile (3,4) and there is a relationship between social jetlag and depressive symptoms, as a previous study shows (5). Levandovski et al. showed a relation between social jetlag and depressive symptoms assessed by the Beck Depression Inventory (BDI) in an adult rural population (5,6). This was not replicated by De Souza et al. (2014) as they did not find a link between social jetlag and depressive symptoms on the BDI in adolescents (7). Sheaves et al. looked at students and observed no relationship between social jetlag and a higher risk for psychiatric symptoms, including depressive symptoms (8). A study conducted by Borisenkov et al. in 2015 in young adults showed that females with seasonal affective disorder, winter type (SAD), a sub-form of MDD, have more social jetlag compared to females without SAD. This was not the case in male subjects (9). Although social jetlag has not been studied in clinical samples, some studies have looked at late chronotype, the evening type. So- cial jetlag is most often brought up in relation to the late chronotype since people with the evening chronotype show more differences in sleep timing between work- and free days (1). These studies linked the evening chronotype to depressive symptoms and to the diagnosis of major depressive disorder (10–15). As social jetlag has been suggested as the explanation for the mood symptoms in subjects with an evening chronotype this would be an interesting concept to study in a clinical sample (5,16). If there is an association between social jetlag and depressive symptoms in patients with MDD as well, this might suggest an influence of the circadian misalignment on depressive symptomatology caused by the evening chronotype.

This study investigates the relation between clinically diagnosed major depressive disorder and social jetlag in a large cohort study. Since the evening chronotype is typically found more often among patients in a current state of depression, the patient group will be split between patients with a current episode and patients who are remitted to test whether this shows differences (17). The effect of antidepressants will be studied as well. We hypothesize that subjects in the MDD group will experience more social jetlag than healthy controls and that social jetlag is associated with higher depression scores.

36 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

Material and methods Population The subjects were derived from the Netherlands Study of Depression and Anxiety (NES- DA), an ongoing cohort study (N=2981 at baseline, age 18–65 years) including 2329 persons with a lifetime diagnosis of a depressive and/or anxiety disorder, as well as 652 healthy controls. Participants were recruited from the community (19%), general practice (54%) and secondary mental health care (27%). The research protocol was approved by the ethical committee of the participating universities and all the participants provided written informed consent. For a detailed description about the NESDA study, see Penninx et al. (18). Data from participants who participated in the 2-year follow-up assessments were included in this study (n = 2596). Psychopathology was determined using the Composite International Diagnostic Interview (CIDI) (19) at baseline and at the 2-year follow-up. Other measures were all derived from the 2-year follow-up. The subjects included in this analysis were control subjects and patients with a lifetime diagnosis of MDD according to the CIDI, independent of comorbid anxiety disorders. Patients without a depressive episode in the preceding month at the 2-year follow-up, but previously diagnosed with MDD, were indicated as ‘remitted’, whereas patients with a depressive episode in the preceding month at the 2-year follow-up were indicated as currently depressed. Depression severity was measured with the Inventory of Depressive Symptoms (Self-Rating) (IDS-SR) (20). All subjects with a group status and a valid value of social jetlag were used for the analyses.

Sleep timing parameters and social jetlag Sleep timing parameters were derived from the Munich ChronoType Questionnaire (MCTQ) (21). SJL was calculated as the absolute difference between the midsleep on free days and midsleep on work days (2).

Covariates SJL differs across age groups and differs significantly between men and women (2). To adjust for differences in groups based on age and sex, these were included as covariates. As social jetlag is the difference between work and free days, external influences on the internal clock should be taken into account as well (10). Therefore, the analysis included the presence of children in the household as well as whether subjects had a current status of employment. Furthermore, as social jetlag is related to sleep duration the analyses were adjusted for average sleep duration (2). Average sleep duration was calculated by taking the sleep duration on work days times 5, adding the sleep duration on free days times 2 and dividing it by 7. To adjust for medication use, another model was tested in which antidepressant use and benzodiazepine use were included as covariates. Antide- pressants used in the analysis were selective serotonin reuptake inhibitors (ATC code N06AB), tricyclic antidepressants (ATC code N06AA) and other antidepressants (ATC code N06AF and N06AX). Benzodiazepines consisted of the ATC codes N05BA, N05CF, N05CD and NO3AE. Both medication groups are combined as antidepressant medication use.

37 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

Statistics Data was prepared using SPSS version 22 syntax, analyses were done with R (22). To assess group differences analysis of variance (ANOVA) was used on continuous data, chi-square analyses on categorical data. Multiple regression with SJL as outcome variable and MDD diagnosis (yes or no) as predictor, with sex, age, employment status and the presence of children in the household as covariates, was used to inspect the effect of MDD diagnosis on social jetlag. A second model included antidepressant medication use (antidepressants and benzodiazepines) as a covariate. An extra analysis was run with a sample where MDD diagnosis is split in remitted MDD and current episode of MDD. To investigate the association between social jetlag and depression symptoms a multiple regression was conducted in the full sample, the MDD group and the control group with IDS-SR score as outcome variable, SJL as predictor and sex, age, employment status and presence of children in the household as covariates. To better understand the differences in social jetlag, additional analyses were run to study the differences in sleep timing parameters. Multiple regression analysis was performed with group status (healthy control or MDD diagnosis) as predictor, with sex, age, employment status and children in the household as covariates.

Results Sample characteristics Of the 2596 participants a valid psychiatric status and social jetlag could be derived for 1469 subjects (figure 1). The included sample had the same number of female subjects (66.3% in the included sample, 65.8% in the excluded sample, p = 0.837), was slightly younger than the excluded sample (mean age 41.7 years versus 47.1 years in the excluded sample, p < 0.001) and in the included sample the percentage of patients with a depression diagnosis in a current state of depression was lower compared to the excluded sample (35% vs 58%, p < 0.001). The included sample contained 385 control subjects and 1084 patients with MDD. Of the MDD patients, 874 were remitted, 210

Figure 1. Flowchart with in- and exclusion criteria

38 patients experienced a current depressive episode. Sample demographics were similar across groups, depressive patients were slightly more often female and MDD patients had more depressive symptoms. MDD patients in a depressive episode were less likely to be currently employed (table 1). Sleep timing characteristics and social jetlag between the groups are described in table 2.

Table 1. Socio-demographic and clinical characteristics according to diagnosis group (n=1469).

ŽŶƚƌŽůƐ DƉĂƚŝĞŶƚƐ ƉͲǀĂůƵĞ ;ŶсϯϴϱͿ ;ŶсϭϬϴϰͿ ^ŽĐŝŽĚĞŵŽŐƌĂƉŚŝĐ ŐĞ;ŵĞĂŶ;ƐĚͿͿ ϰϬ͘ϳ;ϭϯ͘ϲͿ ϰϮ͘ϭ;ϭϭ͘ϳͿ Ϭ͘Ϭϰϴ &ĞŵĂůĞƐ;йͿ Ϯϯϳ;ϲϭ͘ϲͿ ϳϯϳ;ϲϴͿ Ϭ͘ϬϮϲ ŚŝůĚƌĞŶŝŶŚŽƵƐĞŚŽůĚ͕ǇĞƐ;йͿ ϭϯϮ;ϯϰ͘ϯͿ ϯϵϱ;ϯϲ͘ϰͿ Ϭ͘ϰϴϳ ƵƌƌĞŶƚĞŵƉůŽǇŵĞŶƚ͕ǇĞƐ;йͿ ϯϯϵ;ϴϴ͘ϭͿ ϵϬϵ;ϴϯ͘ϵͿ Ϭ͘Ϭϱϴ ŝƐĞĂƐĞĐŚĂƌĂĐƚĞƌŝƐƚŝĐƐ /^ƐĐŽƌĞ;ŵĞĂŶ;ƐĚͿͿ ϱ͘ϴ;ϱ͘ϭͿ ϭϳ͘ϭϯ;ϭϭ͘ϭͿ фϬ͘ϬϬϭ ŶƚŝĚĞƉƌĞƐƐĂŶƚ ŵĞĚŝĐĂƚŝŽŶ ƵƐĞ͕ ϱ;ϭ͘ϯͿ ϯϳϭ;ϯϰ͘ϮͿ фϬ͘ϬϬϭ ǇĞƐ;йͿ

Table 2. Sleep parameters between the different groups (n=1469). Times are shown in continuous variables, meaning decimals are parts of an hour.

ŽŶƚƌŽůƐ ZĞŵŝƚƚĞĚD ;ŶсϯϴϱͿ ;ŶсϴϳϰͿ tŽƌŬĚĂǇƐ ^ůĞĞƉŽŶƐĞƚ;ŵĞĂŶ;ƐĚͿͿ ͲϬ͘ϲϰ;Ϭ͘ϴϳͿ ͲϬ͘ϱϭ;ϭ͘ϬͿ ^ůĞĞƉŽĨĨƐĞƚ;ŵĞĂŶ;ƐĚͿͿ ϳ͘Ϭϯ;Ϭ͘ϵϲͿ ϳ͘ϭϭ;ϭ͘ϬϮͿ ^ůĞĞƉĚƵƌĂƚŝŽŶ;ŵĞĂŶ;ƐĚͿͿ ϳ͘ϲϳ;Ϭ͘ϵϭͿ ϳ͘ϲϭ;ϭ͘ϬϱͿ DŝĚƐůĞĞƉ;ŵĞĂŶ;ƐĚͿͿ ϯ͘ϭϵ;Ϭ͘ϴͿ ϯ͘ϯ;Ϭ͘ϴϳͿ &ƌĞĞĚĂǇƐ ^ůĞĞƉŽŶƐĞƚ;ŵĞĂŶ;ƐĚͿͿ Ϭ͘Ϭϰ;ϭ͘ϬϯͿ Ϭ͘Ϯϭ;ϭ͘ϭϭͿ ^ůĞĞƉŽĨĨƐĞƚ;ŵĞĂŶ;ƐĚͿͿ ϴ͘ϭϵ;ϭ͘ϰϭͿ ϴ͘ϭϵ;ϭ͘ϰϳͿ ^ůĞĞƉĚƵƌĂƚŝŽŶ;ŵĞĂŶ;ƐĚͿͿ ϴ͘ϭϱ;ϭ͘ϮϳͿ ϳ͘ϵϴ;ϭ͘ϰϯͿ DŝĚƐůĞĞƉ;ŵĞĂŶ;ƐĚͿͿ ϰ͘ϭϭ;ϭ͘ϬϲͿ ϰ͘ϮϬ;ϭ͘ϬϵͿ ^ŽĐŝĂů:ĞƚůĂŐ;ŵĞĂŶ;ƐĚͿͿ Ϭ͘ϵϲ;Ϭ͘ϴϯͿ Ϭ͘ϵϴ;Ϭ͘ϴϭͿ

39 SJL and MDD diagnosis Multiple regression analysis was performed to test if diagnostic group (control or MDD diagnosis) predicted the continuous outcome social jetlag, while controlling for sex, age, employment status and presence of children in the household. Interaction effects of sex, children in household and employment status were tested and were not significant. Patients with MDD did not have more social jetlag (β = 0.06, 95% CI: -0.03 – 0.15, p = 0.17) compared to the control group (table 3, model 1). Antidepressant medication use was tested as an interaction term, but showed no significant interaction effect. Anti- depressant medication use was added as a covariate, and did not affect the differences between the patient group and the control group regards the amount of social jetlag they experience (table 3, model 2). However, a statistically significant association was found between antidepressant use and social jetlag (β = 0.16, 95% CI: 0.06 – 0.25, p = 0.002).

Table 3. Multiple regression analyses with social jetlag as outcome and MDD diagnosis as predictor. Model 1 shows the relation without medication effect, model 2 shows the relation when medication is added as a covariate. DŽĚĞůϭ DŽĚĞůϮ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ

^ŽĐŝŽĚĞŵŽŐƌĂƉŚŝĐ β β ŐĞ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ;ͲϬ͘Ϭϯ–ͲϬ͘ϬϭϵͿ ;ͲϬ͘ϬϯͲͲϬ͘ϬϮͿ

^Ğǆ;ĨĞŵĂůĞͿ ͲϬ͘Ϭϵ Ϭ͘Ϭϰ ͲϬ͘Ϭϵ Ϭ͘Ϭϰϴ ;ͲϬ͘ϭϳͲͲϬ͘ϬϬϯͿ ;ͲϬ͘ϭϳ–ͲϬ͘ϬϬͿ ŚŝůĚƌĞŶŝŶŚŽƵƐĞŚŽůĚ ͲϬ͘ϭϭ Ϭ͘Ϭϭ ͲϬ͘ϭϭ Ϭ͘Ϭϭ ;ͲϬ͘ϭϵͲͲϬ͘ϬϮͿ ;ͲϬ͘ϮϬͲͲϬ͘ϬϯͿ

ƵƌƌĞŶƚĞŵƉůŽǇŵĞŶƚ Ϭ͘Ϭϴ Ϭ͘ϭϰ Ϭ͘Ϭϵ Ϭ͘ϭϭ ;ͲϬ͘Ϭϯ–Ϭ͘ϮϬͿ ;ͲϬ͘ϬϮ–Ϭ͘ϮϬͿ ǀĞƌĂŐĞƐůĞĞƉĚƵƌĂƚŝŽŶ;ŚŽƵƌƐͿ ͲϬ͘ϬϮ Ϭ͘ϰϱ ͲϬ͘ϬϮ Ϭ͘ϯϬ ;ͲϬ͘Ϭϲ–Ϭ͘ϬϯͿ ;ͲϬ͘Ϭϳ–Ϭ͘ϬϮͿ ŝƐĞĂƐĞĐŚĂƌĂĐƚĞƌŝƐƚŝĐƐ ZĞĨĞƌĞŶĐĞ;ŶŽDͿ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ DĚŝĂŐŶŽƐŝƐ Ϭ͘Ϭϲ Ϭ͘ϭϳ Ϭ͘Ϭϭ Ϭ͘ϴϭ ;ͲϬ͘Ϭϯ–Ϭ͘ϭϱͿ ;ͲϬ͘Ϭϴ–Ϭ͘ϭϭͿ ŶƚŝĚĞƉƌĞƐƐĂŶƚŵĞĚŝĐĂƚŝŽŶƵƐĞ Ϭ͘ϭϲ Ϭ͘ϬϬϮ ;ǇĞƐͿ ;Ϭ͘Ϭϲ–Ϭ͘ϮϱͿ

40 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

As a sensitivity analysis, this procedure was repeated in a subsample consisting of only those subjects with a current job or children in the household (n = 1305), as these subjects tend to have social jetlag. This yielded similar results, as patients with MDD did not show more social jetlag compared to controls (table S1).

When the sample was split based on current episode of MDD or not, patients with a current episode of MDD showed slightly more social jetlag (β = 0.13, 95% CI: 0.0008 – 0.26), p = 0.049) and when different types of medication use were added as covariates, no groups differences were found (table S2, model 1 and 2).

Sleep timing parameters Table S3 and S4 show the estimates and p-values for group status (full models are available upon request). Patients with MDD showed later sleep onset on both work days and free days (table S3) and showed later midsleep on both work and free days (table S4).

SJL and depression severity Multiple regression analysis was performed to test if social jetlag predicted the continuous depression severity score (IDS-SR), while controlling for sex, age, employment status and presence of children in the household in the full sample, in patients with a lifetime MDD diagnosis and in control subjects. The regression model showed that social jetlag provided no explanation for depression severity in the full sample (β = -0.09, 95% CI: -0.82 – 0.64, p = 0.80), nor in the MDD subsample (β = -0.33, 95% CI: -1.18 – 0.52, p = 0.45) nor in the control subsample (β = -0.13, 95% CI: -0.81 – 0.54, p = 0.69). As the IDS-SR includes 4 questions related to sleep and sleep timing, a sensitivity analysis was performed to see if there was an effect of these questions on the data (23). Removing these questions from the total IDS-SR score yielded similar results.

Discussion This is the first study investigating social jetlag in clinically diagnosed patients with MDD. It shows that patients do not experience more social jetlag compared to healthy controls. When antidepressant use and benzodiazepine use were added to the model, there were also no differences between the groups, although antidepressant medication use was significant associated with social jetlag in the model. Patients using antidepressant medication could form a subgroup of patients with a more severe type of depression. This would indicate that patients suffering from a more severe depression experience more social jetlag, although we could not find an association between severity and the IDS-SR score. These results are not in line with an earlier study conducted by Levandovski et al. in 2011, which found a direct relation between the amount of social jetlag and the severity of mood symptoms in a non-clinical sample (5). As there was no formal depression diagnosis in this study, these subjects may have been on the milder end of the mood disorder spectrum. This is illustrated by the fact that depression scores in that study were also relatively low (95% of the sample had a BDI score below 17). We tried to replicate this finding in our study in subjects with milder symp-

41 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

toms, by studying the possibility of a relation between social jetlag and depression severity in our healthy controls. However, no association was found here either. This is in line with other studies that did not show an association between social jetlag and depressive symptoms in younger samples (7,8,13).

As this is the first study analyzing a clinical rather than a community cohort study, this is also the first time the effect of antidepressants is specifically taken into account. When the use of antidepressants was added to the model, no differences between the groups were found, indicating the initial, small, differences found in the split sample are associated with the use of antidepressants. Although antidepressant use is not related to chronotype, the adverse effects of antidepressants on sleep might explain this association (10,24). This possible influence may also be related to the timing of medication, as antidepressant intake in the evening might have an excitatory effect and lead to later bedtime, resulting in more social jetlag. As timing of medication intake was not measured in our sample, there was no way to study the effect of the moment of intake. Another possibility might be that patients who need antidepressants are suffering from a more severe, or different type of, depression. In the current sample this possible difference in depression severity was not reflected in IDS score, most likely as they already are on treatment.

This cross-sectional study extends the prior studies by studying an adult sample with clinically diagnosed MDD, and from this we can conclude the effect of social jetlag on depression might not be as relevant as thought. The lack of a difference between healthy controls and patients with MDD, the small difference in social jetlag between the diagnostic states in the split sample (current episode vs. control subjects), which is influenced by medication use, and the finding of no association between social jetlag and depression severity suggests there is no direct effect of social jetlag on depressive symptoms, although there might still be an indirect effect of chronotype (10). A possible explanation for this effect might be the altered functioning of the internal clock in a current state of depression (25). Another mechanism for the later chronotype is an increase in sleep latency, as depressed patients tend to have a later sleep onset (26). When the sleep timing parameters derived from the MCTQ are studied, this is confirmed, as patients with MDD have a later sleep onset compared to healthy controls.

The large sample size and the well-defined clinical diagnosis of MDD are strengths of the current study. The fact that we were able to adjust for sociodemographic variables and medication use is another strength of this study. A limitation of the study is the fact that social jetlag is based on a self-reported questionnaire asking sleep timing data of the past weeks. The actual circadian misalignment remains unknown in this study. A possible method would be to study the daily patterns, calculating the daily circadian misalignment, which can easily be done using smart devices (27–29). Furthermore, as this is a cross-sectional analysis, nothing can be concluded on whether social jetlag has a predictive value in developing major depressive disorder or not. Lastly, the percentage of subjects in a current state of a depression was higher in the excluded sample. This might suggest subjects not in a current state are more likely to fill in the MCTQ, although the number of subjects with a depressive diagnosis excluded due to missing MCTQ data is small.

42 In conclusion, there are no differences in patients with MDD and controls when account- ing for antidepressant use, and there is no link between social jetlag and depressive symptomatology, implicating the link between circadian misalignment and depressive symptomatology might be more complicated than initially thought.

Differences between sleep timing parameters are studied using multiple regression, with group status (control, remitted MDD or current episode MDD) as a predictor, with sex, age, employment status and children in the household as covariates.

Table S3 & S4 shows the estimates and p-values for group status. Full models are available upon request.

Supplemental material

Table S1. Multiple regression analyses with social jetlag as outcome and MDD diagnosis as predictor in the subsample of people with current employment and children in the household (n = 1305). Model 1 shows the relation without medication effect, model 2 shows the relation when medication is added as a covariate.

DŽĚĞůϭ DŽĚĞůϮ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ

^ŽĐŝŽĚĞŵŽŐƌĂƉŚŝĐ β β ŐĞ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ;ͲϬ͘Ϭϯ–ͲϬ͘ϬϮͿ ;ͲϬ͘ϬϯͲͲϬ͘ϬϮͿ ^Ğǆ;ĨĞŵĂůĞͿ ͲϬ͘ϭϭ Ϭ͘ϬϮ ͲϬ͘ϭϬ Ϭ͘ϬϮ ;ͲϬ͘ϮϬͲͲϬ͘ϬϮͿ ;ͲϬ͘ϭϵͲͲϬ͘ϬϭͿ ǀĞƌĂŐĞƐůĞĞƉĚƵƌĂƚŝŽŶ ͲϬ͘ϬϬϭ Ϭ͘ϵϳ ͲϬ͘ϬϬϲ Ϭ͘ϳϳ ;ŚŽƵƌƐͿ ;ͲϬ͘Ϭϱ–Ϭ͘ϬϰͿ ;ͲϬ͘Ϭϱ–Ϭ͘ϬϰͿ ŝƐĞĂƐĞĐŚĂƌĂĐƚĞƌŝƐƚŝĐƐ ZĞĨĞƌĞŶĐĞ;ŶŽDͿ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ Ŷсϯϰϰ DĚŝĂŐŶŽƐŝƐ Ϭ͘Ϭϵ Ϭ͘Ϭϳ Ϭ͘Ϭϰ Ϭ͘ϰϰ Ŷсϵϲϭ ;ͲϬ͘Ϭϭ–Ϭ͘ϭϴͿ ;ͲϬ͘Ϭϲ–Ϭ͘ϭϰͿ ŶƚŝĚĞƉƌĞƐƐĂŶƚŵĞĚŝĐĂƚŝŽŶ Ϭ͘ϭϰ Ϭ͘ϬϬϲ ƵƐĞ;ǇĞƐͿ ;Ϭ͘Ϭϰ–Ϭ͘ϮϰͿ

43 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

Table S2. Multiple regression analyses with social jetlag as outcome and MDD diagnosis as predictor for the splitted sample. Model 1 shows the relation without medication effect, model 2 shows the relation when medication is added as a covariate. DŽĚĞůϭ DŽĚĞůϮ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ ;ϵϱй/Ϳ ƉͲǀĂůƵĞ

^ŽĐŝŽĚĞŵŽŐƌĂƉŚŝĐ β β ŐĞ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ͲϬ͘ϬϮ фϬ͘ϬϬϭ ;ͲϬ͘ϬϯͲͲϬ͘ϬϮͿ ;ͲϬ͘ϬϯͲͲϬ͘ϬϮͿ ^Ğǆ;ĨĞŵĂůĞͿ ͲϬ͘Ϭϵ Ϭ͘Ϭϰϴ ͲϬ͘Ϭϴ Ϭ͘Ϭϱ ;ͲϬ͘ϭϳͲͲϬ͘ϬϬϬϴͿ ;ͲϬ͘ϭϳͲϬ͘ϬϬϭͿ ŚŝůĚƌĞŶŝŶŚŽƵƐĞŚŽůĚ ͲϬ͘ϭϭ Ϭ͘Ϭϭ ͲϬ͘ϭϭ Ϭ͘Ϭϭ ;ͲϬ͘ϭϵͲͲϬ͘ϬϮͿ ;ͲϬ͘ϭϵͲͲϬ͘ϬϯͿ ƵƌƌĞŶƚĞŵƉůŽǇŵĞŶƚ Ϭ͘Ϭϵ Ϭ͘ϭϮ Ϭ͘ϭϬ Ϭ͘Ϭϵ ;ͲϬ͘ϬϮ–Ϭ͘ϮϬͿ ;ͲϬ͘ϬϮ–Ϭ͘ϮϭͿ ǀĞƌĂŐĞƐůĞĞƉĚƵƌĂƚŝŽŶ ͲϬ͘Ϭϭϳ Ϭ͘ϰϱ ͲϬ͘ϬϮ Ϭ͘ϯϭ ;ŚŽƵƌƐͿ ;ͲϬ͘Ϭϲ–Ϭ͘ϬϮͿ ;ͲϬ͘Ϭϳ–Ϭ͘ϬϮͿ ŝƐĞĂƐĞĐŚĂƌĂĐƚĞƌŝƐƚŝĐƐ ZĞĨĞƌĞŶĐĞ;ŶŽDͿ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ DƌĞŵŝƚƚĞĚ Ϭ͘Ϭϱ Ϭ͘ϯϯ Ϭ͘ϬϬϮ Ϭ͘ϵϳ ;ͲϬ͘Ϭϱ–Ϭ͘ϭϰͿ ;ͲϬ͘Ϭϵ–Ϭ͘ϭϬͿ DĐƵƌƌĞŶƚĞƉŝƐŽĚĞ Ϭ͘ϭϯ Ϭ͘Ϭϰϵ Ϭ͘Ϭϲ Ϭ͘ϯϱ ;Ϭ͘ϬϬϬϴ–Ϭ͘ϮϲͿ ;ͲϬ͘Ϭϳ–Ϭ͘ϮϬͿ ŶƚŝĚĞƉƌĞƐƐĂŶƚŵĞĚŝĐĂƚŝŽŶ Ϭ͘ϭϱ Ϭ͘ϬϬϮ ƵƐĞ;ǇĞƐͿ ;Ϭ͘Ϭϱ–Ϭ͘ϮϰͿ

44 Chapter 3: Social jetlag and depression status: results from the Netherlands Study of Depression and Anxiety

Table S3. Multiple regression model studying reported sleep timing parameters between the groups.

^ůĞĞƉŽŶƐĞƚ ^ůĞĞƉŽŶƐĞƚ ^ůĞĞƉŽĨĨƐĞƚ ^ůĞĞƉŽĨĨƐĞƚ ǁŽƌŬĚĂǇƐ ĨƌĞĞĚĂǇƐ ǁŽƌŬĚĂǇƐ ĨƌĞĞĚĂǇƐ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ

β/Ϳ β/Ϳ β/Ϳ β/Ϳ ZĞĨĞƌĞŶĐĞ;ŶŽ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ DͿ DĚŝĂŐŶŽƐŝƐ Ϭ͘ϭϰ Ϭ͘Ϭϭ Ϭ͘Ϯϭ Ϭ͘ϬϬϭ Ϭ͘Ϭϳ Ϭ͘Ϯϰ Ϭ͘Ϭϳ Ϭ͘ϯϴ ;Ϭ͘Ϭϯ– ;Ϭ͘Ϭϴ– ;ͲϬ͘Ϭϰ– ;ͲϬ͘Ϭϴ– Ϭ͘ϮϱͿ Ϭ͘ϯϯͿ Ϭ͘ϭϴͿ Ϭ͘ϮϯͿ

Table S4. Multiple regression model studying derived sleep timing parameters between the groups.

^ůĞĞƉĚƵƌĂƚŝŽŶ ^ůĞĞƉĚƵƌĂƚŝŽŶ DŝĚƐůĞĞƉ DŝĚƐůĞĞƉ ǁŽƌŬĚĂǇƐ ĨƌĞĞĚĂǇƐ ǁŽƌŬĚĂǇƐ ĨƌĞĞĚĂǇƐ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ ;ϵϱй ƉͲǀĂůƵĞ

β/Ϳ β/Ϳ β/Ϳ β/Ϳ ZĞĨĞƌĞŶĐĞ;ŶŽ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ ƌĞĨ DͿ DĚŝĂŐŶŽƐŝƐ ͲϬ͘Ϭϳ Ϭ͘ϮϬϯ ͲϬ͘ϭϰ Ϭ͘Ϭϴ Ϭ͘ϭϬ Ϭ͘Ϭϰ Ϭ͘ϭϰ Ϭ͘ϬϮ ;ͲϬ͘ϭϵ– ;ͲϬ͘Ϯϵ– ;Ϭ͘ϬϬϳ ;Ϭ͘ϬϮ– Ϭ͘ϬϰͿ Ϭ͘ϬϭͿ –Ϭ͘ϮϬͿ Ϭ͘ϮϲͿ

45 tween depressive mood and chronotype in References healthy subjects. 2009;63:283–90. 1. Wittmann M, Dinich J, Merrow M, Roenneberg 14. Kitamura S, Hida A, Watanabe M, Enomoto M, Ar- T. Social jetlag: misalignment of biological itake-Okada S, Moriguchi Y, et al. Evening Pref- and social time. Chronobiol Int. 2006;23(1– erence Is Related To the Incidence of Depressive 2):497–509. States Independent of Sleep-Wake Conditions. 2. Roenneberg T, Allebrandt K V, Merrow M, Vet- Chronobiol Int. 2010;27(9–10):1797–812. DOI: ter C. Social jetlag and obesity. Curr Biol. 10.3109/07420528.2010.516705 2012 May 22;22(10):939–43. DOI: 10.1016/j. 15. Merikanto I, Kronholm E, Peltonen M, Laatikain- cub.2012.03.038 en T, Vartiainen E, Partonen T. Circadian prefer- 3. Rutters F, Lemmens SG, Adam TC, Bremmer ence links to depression in general adult pop- M a, Elders PJ, Nijpels G, et al. Is social jet- ulation. J Affect Disord. 2015;188:143–8. DOI: lag associated with an adverse endocrine, 10.1016/j.jad.2015.08.061 behavioral, and cardiovascular risk profile? 16. Muller MJ, Kundermann B, Cabanel N. Evening- J Biol Rhythms. 2014;29(5):377–83. DOI: ness and poor sleep quality independently con- 10.1177/0748730414550199 tribute to self-reported depression severity in 4. Scheer FAJL, Hilton MF, Mantzoros CS, Shea SA. psychiatric inpatients with affective disorder. Adverse metabolic and cardiovascular conse- Nord J Psychiatry. 2015;9488(December):1–6. quences of circadian misalignment. Proc Natl 17. Knapen S, Druiven S, Meesters Y, Riese H. Chro- Acad Sci U S A. 2009;106(11):4453–8. DOI: notype not associated with nonremission, but 10.1073/pnas.0808180106 with current state? Sleep. 2017;40(6). 5. Levandovski R, Dantas G, Fernandes LC, Cau- 18. Penninx BWJH, Beekman ATF, Smit JH, Zitman mo W, Torres I, Roenneberg T, et al. Depression FG, Nolen WA, Spinhoven P, et al. The Nether- Scores Associate With Chronotype and Social lands Study of Depression and Anxiety (NESDA): Jetlag in a Rural Population. Chronobiol Int. rationale, objectives and methods. Int J Meth- 2011;28(9):771–8. ods Psychiatr Res. 2008;17(3):121–40. 6. Beck AT, Steer RA, Brown GK. Beck depression in- 19. Wittchen HU. Reliability and validity studies of ventory-II. San Antonio, TX. 1996;72498–8204. the WHO-Composite International Diagnostic 7. de Souza CM, Hidalgo MPL. Midpoint of sleep on Interview (CIDI): A critical review. J Psychiatr school days is associated with depression among Res. 1994;28(1):57–84. adolescents. Chronobiol Int. 2014;31(2):199– 20. Rush AJ, Gullion CM, Basco MR, Jarrett RB, Trive- 205. DOI: 10.3109/07420528.2013.838575 di MH. The Inventory of Depressive Symptom- 8. Sheaves B, Porcheret K, Tsanas A, Espie CA, Fos- atology (IDS): psychometric properties. Psychol ter RG, Freeman D, et al. Insomnia, Nightmares, Med. 1996;26(3):477–86. and Chronotype as Markers of Risk for Severe 21. Roenneberg T, Kuehnle T, Juda M, Kantermann Mental Illness: Results from a Student Popula- T, Allebrandt K, Gordijn M, et al. Epidemiolo- tion. Sleep. 2016;39(1):173–81. DOI: 10.5665/ gy of the human circadian clock. Sleep Med sleep.5342 Rev. 2007 Dec;11(6):429–38. DOI: 10.1016/j. 9. Borisenkov MF, Petrova NB, Timonin VD, Fradko- smrv.2007.07.005 va LI, Kolomeichuk SN, Kosova AL, et al. Sleep 22. R Core Team. R: A Language and Environment characteristics, chronotype and winter depres- for Statistical Computing [Internet]. Vienna, sion in 10-20-year-olds in northern European Austria; 2018. Russia. J Sleep Res. 2015;24(3):288–95. 23. Wardenaar KJ, Van Veen T, Giltay EJ, Den Hol- 10. Antypa N, Vogelzangs N, Meesters Y, Schoevers lander-Gijsman M, Penninx BWJH, Zitman FG. RA, Penninx BWJH. Chronotype associations The structure and dimensionality of the In- with depression and anxiety in a large cohort ventory of Depressive Symptomatology Self study. Depress Anxiety. 2016 Jan;33(1):75–83. Report (IDS-SR) in patients with depressive 11. Chelminski I, Ferraro FR, Petros T V, Plaud JJ. An disorders and healthy controls. J Affect Dis- analysis of the “‘ eveningness – morningness ’” ord. 2010;125(1–3):146–54. DOI: 10.1016/j. dimension in “‘ depressive ’” college students. J jad.2009.12.020 Affect Disord. 1999;52:19–29. 24. Lam RW. Sleep disturbances and depression: a 12. Drennan MD, Klauber MR, Kripke DF, Goyette LM. challenge for antidepressants. Int Clin Psycho- The effects of depression and age on the Horne- pharmacol. 2006;21:S25–9. Ostberg morningness-eveningness score. J Af- 25. Li JZ, Bunney BG, Meng F, Hagenauer MH, fect Disord. 1991 Oct;23(2):93–8. Walsh DM, Vawter MP. Circadian patterns of 13. Hidalgo MP, Caumo W, Posser M, Coccaro SB, gene expression in the human brain and dis- Camozzato AL, Chaves MLF. Relationship be- ruption in major depressive disorder. Pnas. 2013;110(24):9950–5.

46 26. Mendlewicz J. Sleep disturbances: core symptoms of major depressive disorder rather than associated or comorbid disorders. World J Biol Psychiatry. 2009;10(4):269–75. 27. Fischer D, Vetter C, Roenneberg T. A novel method to visualise and quantify circadian misalignment. Nat Publ Gr. 2016;1–10. DOI: 10.1038/ srep38601 28. Knapen SE, Lek RFR Der, Haarman BCM, Schoev- ers RA. Coping with a life event in bipolar disorder : ambulatory measurement , signalling and early treatment. 2016;1–3. 29. Aan het Rot M, Hogenelst K, Schoevers RA. Mood disorders in everyday life: A systematic review of experience sampling and ecological momentary assessment studies. Clin Psy- chol Rev. 2012;32(6):510–23. DOI: 10.1016/j. cpr.2012.05.007

47 Chapter 4 Letter to the editor: Chronotype not associated with non-remission, but with current state?

S.E. Knapen1,2, S.J.M. Druiven1, Y. Meesters2 and H. Riese1

1. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Interdisciplinary Center for Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands; 2. University of Groningen, University Medical Center Groningen, Department of Psy- chiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Gronin- gen, the Netherlands;

Sleep 2017; 40(6) With great interest we read the paper by Chan and colleagues.(1) In their paper they showed the importance of studying chronotype, whether people are morning or evening persons, in patients diagnosed with major depressive disorder (MDD). The number of citations to our esteemed colleagues’ paper (20 in 31 months according to a Scopus search in January 2017) shows the impact of the paper. However, when studying the citing papers we noticed that over half of the citing papers misinterpreted the results and conclusions of the paper by Chan and colleagues.(1) We want to point out two terms, which were used in the paper, that may be related to the incorrectly interpreted causal- ity of the reported findings. Although incorrect citations are more common in research, these used terms might have unintentionally triggered some of the confusion.

First, Chan et al.(1) described a longitudinal cohort of patients diagnosed with current status of MDD (current episode in the past month) at inclusion in 2006 (T0). At T0, 419 MDD patients were included. This cohort was followed-up in 2010 (T1, n=371) and in 2011 (T2, n=253). At T2, diagnosis of MDD was confirmed, and severity of depression, eveningness and sleep-wake habits were assessed with valid instruments. For their publication data collected at T2 were tested (p911) to assess group differences in chronotype. This makes the study a cross-sectional study, as they indeed stated in their Dis- cussion (p916). The statistical analysis used to test the relationship between eveningness and MDD is a logistic regression analysis. The independent variable, eveningness, is described as a ‘risk-factor’ for depressive symptomatology (p915). This term implies a causal relationship, while this cannot be studied in their cross-sectional study. Mis- interpretation of this cross-sectional analysis in this longitudinal cohort is likely when the term risk factor is used, as we saw in citing papers.(2,3)

Second, another misinterpretation of the citing papers might be the result of the cat- egorisation of the MDD patients.(4,5) Prior analysis, patients were categorized into a ‘remission’ group and a ‘nonremission’ group. The patients in this latter group have a cut-off score of 8 or higher on the Hamilton Rating Scale for Depression and a current status of MDD according to the 1-month prevalence on the Mini-International Neuro- psychiatric Interview at T2 (note: independent of their depressive status at T1).(6,7) We agree with the definition of the remission group as all patients were depressed at T0 and remitted at T2. However, for the nonremission group the definition is lacking precision. It is unknown from the collected data whether this group also included patients who were in remission at T1 and subsequently diagnosed with MDD at T2 and thus should be defined as suffering from a relapse. A more straightforward, or pure- ly descriptive way of labelling this group is ‘not-in-remission’ or, even more precisely ‘current depressive status according to the 1-month prevalence data’. Patients fulfilling the criteria of this latter definition are reported to be more evening type in the literature before.(8) To define a better non-remission group, Chan and colleagues could have selected patients with a persisting depressive status (current depressive episode at T0, T1 and T2) as their non-remission group. This could decrease the sample size and negatively affect the power of the statistical analysis, although it likely creates more homogeneous groups, which may increase power. Albeit a subtle difference, we would like to point out that providing specific labels could make an important difference for the interpretability of the conclusions by the readers. In this case, it is important as the

49 term nonremission may unintentionally suggest that the authors studied the longitudinal course of depression.

The combination of the two terms (risk factor and nonremission) may unintentionally have implied a causal relationship between eveningness and the course of depression, something that should not be concluded from Chan and colleagues’ paper. We are grateful Chan et al. increased the knowledge of the field by studying the influence of sleep problems on the association between eveningness and depression. However, whether there is a causal relationship between chronotype and the course of depression remains unknown and a relevant subject for future research.

50 References 1. Chan JWY, Lam SP, Li SX, Yu MWM, Chan NY, Zhang J, et al. Eveningness and insomnia: independent risk factors of nonremission in major depressive disorder. Sleep. 2014;37(5):911–7. DOI: 10.5665/sleep.3658 2. Merikanto I, Kronholm E, Peltonen M, Laatikain- en T, Vartiainen E, Partonen T. Circadian preference links to depression in general adult population. J Affect Disord. 2015;188:143–8. DOI: 10.1016/j.jad.2015.08.061 3. Bei B, Ong JC, Rajaratnam SMW, Manber R. Chro- notype and improved sleep efficiency independently predict depressive symptom reduction after group cognitive behavioral therapy for insomnia. J Clin Sleep Med. 2015;11(9):1021–7. 4. Kripke DF, Elliott JA, Welsh DK, Youngstedt SD. Photoperiodic and circadian bifurcation theories of depression and mania. F1000Research. 2015;4(May):107. DOI: 10.12688/f1000re- search.6444.1 5. Melo MCA, Abreu RLC, Linhares Neto VB, de Bruin PFC, de Bruin VMS. Chronotype and circadian rhythm in bipolar disorder: a systematic review. Sleep Med Rev. 2016; DOI: 10.1016/j. smrv.2016.06.007 6. Williams JBW, Link MJ, Rosenthal NE, Terman M. Structured Interview Guide for the Hamilton Depression Rating Scale—Seasonal Affective Disorder Version (SIGH-SAD). New York, New York State Psychiatr Inst. 1988; 7. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-Internation- al Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59:22–33. 8. Antypa N, Vogelzangs N, Meesters Y, Schoevers RA, Penninx BWJH. Chronotype associations with depression and anxiety in a large cohort study. Depress Anxiety. 2016 Jan;33(1):75–83.

51 Chapter 5 Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls

S.E. Knapen1*, S. Verkooijen2, R.F. Riemersma-van der Lek1, R.S. Kahn2,3, R.A. Ophoff2,4 , M.P.M. Boks2, R.A. Schoevers1

1. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Interdisciplinary Center Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands; 2. Brain Center Rudolf Magnus, University Medical Center Utrecht, Department of Psychiatry, Utrecht, the Netherlands. 3. Icahn School of Medicine at Mount Sinai, New York, USA 4. Center for Neurobehavioral Genetics, University of California Los Angeles, Los Angeles, CA, USA

In preparation Chapter 5: Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls

Abstract Background Patients with bipolar disorder suffer from circadian rhythm disturbances, suggesting that changes in the circadian timing system may be an underlying pathophysiological mechanism of this disorder. It is unclear whether these disturbances are limited to manic and depressive episodes. We therefore used actigraphy to study the circadian rest-activity cycle of patients in the euthymic phase of their illness and compared them to unaffected siblings and healthy controls.

Methods Patients with bipolar disorder type I, siblings and healthy controls enrolled in a 14-day actigraphy protocol as part of the Dutch Bipolar Cohort study (DBC). From the actigraphy data circadian variables were calculated and compared between the groups using multiple regression. In addition, chronotype (from the Munich Chronotype Question- naire) was compared between the groups.

Results 107 patients, 72 siblings and 78 healthy controls were included. There were no differences in the non-parametric circadian variables between the groups (all 7 variables p > 0.007), neither was there a difference in chronotype.

Conclusions In this large sample of patients with bipolar disorder, we can conclude that neither euthymic bipolar patients, nor unaffected siblings, have more circadian rhythm disturbances than healthy controls. This shows patients outside mood episodes are able to maintain a stable rhythm and does not support reports suggesting that altered circadian rhythmicity is a trait characteristic of bipolar disorder.

53 Chapter 5: Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls

Introduction Bipolar disorder is a severe psychiatric disorder characterized by mood fluctuations occurring in episodes, affecting around 1% of the population (1). Patients suffering from a mood episode experience problems in the regulation of the daily 24 hour rhythm, the circadian rhythm (2). Problems with the circadian rhythm are not only reported during an episode, but also in the euthymic phase of the disorder, when patients have few or no mood symptoms. The circadian timing system may be involved in bipolar disorder regardless of disease state, although the pathophysiological mechanism is still unclear (2,3). A minimally invasive method to assess the circadian rhythm in people is the use of actigraphy, with study participants wearing a wristband on their non-dominant wrist that measures movement over the day (4,5). It allows participants to go on with their day-to-day life, and yields a representative measure of daily activity and rest patterns. To assess the stability of the circadian activity rhythm, different non-parametric variables have been developed by van Someren et al. (6). These variables include intradaily variability (IV), interdaily stability (IS), activity (and its start time) during the most active 10 (M10) and most inactive 5 hours of the day (L5) and the amplitude of the rhythm (relative amplitude, RA).

Previous studies suggest that patients with bipolar disorder in a euthymic phase may differ from control participants on these measures (7–13). Castro et al. showed that currently unaffected people who are at risk for psychosis or bipolar disorder show higher interdaily variability, less interdaily stability, more L5 activity and less M10 activity (10). Rock et al. showed people with a bipolar phenotype, as assessed with the Mood Disorder Questionnaire, have more L5 activity and a lower relative amplitude (8). This lower RA was also shown in individuals at risk for hypomania (11) and at risk for bipolar disorder in an Australian sample (13). A less robust rhythm was found in (formally diagnosed) euthymic patients compared to controls as well by McKenna et al. (9). Jones et al. showed higher intradaily variability and less interdaily stability in euthymic patients in a 7-day actigraphy protocol (7). Geoffroy et al. studied these variables during 21 days and found that only the interdaily stability differed between euthymic patients with bipolar disorder type I and II and controls (14). Pagani et al. studied 26 pedigrees of bipolar disorder patients, resulting in 558 individuals, and showed that L5 and amplitude were related to the phenotype of bipolar disorder. Furthermore they showed a high heritability in non-parametric circadian variables, in particular the IS, IV, RA, and onset of L5 and M10 (15). This suggests that circadian disturbances might be a heritable trait of bipolar disorder and might represent a circadian endophenotype (13,16). An- other circadian measure can be an individual’s chronotype, the person’s preference of timing of activities in the morning (morning types), or later on the day (evening types) (17). Different studies show the evening chronotype is associated with bipolar disorder, both during an episode and independent of episodes (18,19). This includes one large study with over 250 bipolar disorder (type I and II) patients who are followed for over 2 years, which showed the evening chronotype was present independent of disease state (19). However, other studies, with similar sample sizes, have shown that patients with bipolar disorder have a similar chronotype during the euthymic phase compared to controls (20).

54 Although circadian rhythmicity seems a promising endophenotype, there are still un- resolved issues. The few studies that studied a clinical sample had low numbers of participants (of about 30 patients), used an actigraphy period too short for a reliable calculation of the non-parametric circadian variables, or combined bipolar disorder type I and type II (7,9,14). Furthermore, there are only a few studies which look at circadian rhythm variables in non-affected family members. The current study aims to examine circadian rhythm disturbances in a well-defined sample of euthymic bipolar type 1 patients, unaffected siblings and healthy controls. Previously we showed that euthymic patients had no differences in sleep variables compared to healthy controls (21). In this study, non-parametric circadian variables and chronotype will be studied comparing euthymic patients, unaffected siblings and healthy controls.

Material and methods Population Participants were derived from the Dutch Bipolar Cohort (DBC) study, a collaboration between the University Medical Center Utrecht (UMCU), University Medical Center Gron- ingen (UMCG), various other mental health care providers in the Netherlands and the University of California Los Angeles (UCLA). The DBC study is developed to investigate genetic and (endo)phenotypic vulnerability factors for bipolar disorder. The medical ethical committee of the UMCU, UCLA and the UMCG approved the DBC study and the additional actigraphy study and both studies were in accordance to the Declaration of Helsinki. Informed consent was obtained from all participants prior to participation. All participants in the DBC were informed of the actigraphy protocol through a newsletter and were approached to participate in this additional protocol. Inclusion criteria for all participants were a minimum age of 18 years, at least three grandparents of Dutch de- scent, no major somatic illness (such as sleep apnea or Parkinsons disease) and no current pregnancy. Patients were required to have bipolar type I disorder, which was confirmed using the Structural Interview for DSM-IV (SCID-I) (22) and could not be admitted to a hospital. Although none of the participants reported being in a current mood episode, 16 patients scored above the cut-off score for depressive symptoms (> 26 on the Inventory of Depressive Symptomatology – Self-Rating, IDS-SR (23)) and 4 patients scored above the cut-off score for manic symptoms (> 5 on the Altman Self-Rating Mania scale, ASRM (24)). Siblings and control participants with a diagnosis of bipolar disorder or a psychotic disorder were excluded. Control participants with a first or second degree relative with such a diagnosis were excluded as well. Both the siblings and controls were assessed using the Mini-International Neuropsychiatric Interview (MINI) (25).

A total of 466 eligible candidates were approached for participation via telephone, post or e-mail. 106 participants did not respond to the invitation, 57 refused to participate and 17 did not show up for their appointment. In 11 participants Actiwatch data were not available due to hardware problems. Four participants were excluded due to a somatic illness (admitted for cancer treatment and sleep apnea). Three individuals were excluded because of a current depressive episode and six were excluded because of technical problems in the storage of actiwatch data. One healthy control is excluded

55 for not meeting the inclusion criteria of not having a sibling with bipolar disorder. Two participants were excluded for not having enough available data points (<10 days of data). This resulted in the analyses of 107 patients, 72 siblings and 78 controls.

Actigraphy recordings Rest-activity patterns were recorded with an Actiwatch (Actiwatch 2, Philips Respiron- ics). The Actiwatch is a small wristband that measures activity and light input and stores it in a minute-to-minute basis (1 minute epochs). All participants wore the Actiwatch for a period of 14 consecutive days on their non-dominant wrist and were instructed to only remove it when exposed to water for long periods of time. Furthermore, participants kept a sleep diary with bed times, nap times and off-wrist periods. All Actiwatch- es were subjected to two calibration protocols prior to initial data collection and after every battery service (for details, see Pagani et al. 2016) (15).

Circadian variables Actiwatch data were exported from the Philips Respironics Actiware software and im- ported in R (26). All starting points of the data were automatically defined and all data within 14 days after that starting point were taken into account. Non-wear, as reported in the sleep diary, was excluded from the analysis. When more data were available, activity data after the 14-day period were excluded. If less than 14 days were available, the data were cut off after the end of the activity. Full days were included for the analysis, if this was not possible the last day with at least 22 hours of data was used. If fewer hours were available the data was skipped to the previous day. The script for the analysis can be retrieved from https://github.com/compsy/ACTman/ (27). Circadian variables were computed according to the original formulas which can be found in the paper by Van Someren (6). These variables are called non-parametric variables as the calculation is a non-parametric calculation.

Chronotype Chronotype was measured with a Dutch translation of the Munich Chronotype Ques- tionnaire (MCTQ) (28). This self-report questionnaire is composed of 11 questions regarding sleep times on workdays and free days and commonly used to assess chronotype. Chronotype is defined as the midpoint of sleep on free days, corrected for oversleep on free days (MSFsc) (29). This midpoint shows good correlation with internal phase markers (30,31).

Statistical analysis All statistical analyses were carried out using the R statistical package (26). Demo- graphic data, group differences on continuous variables were analyzed using analysis of variance (ANOVA) with Bonferroni post hoc tests. Group differences on categorical variables were analyzed using chi-square tests. To analyze differences in circadian variables and chronotype between patients, siblings and controls three separate lin-

56 Chapter 5: Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls

ear regression models were used per variable. One for the comparison of patients and siblings, one for the comparison of patients and controls and one for the comparison of siblings and controls. To control for the confounding effect of sex and age on the variables these measures were added to the model. An extra analysis was run only with subjects who did not experience any mood symptoms (depressive or manic) above the cut-off scores on the validated questionnaire to have a more strictly euthymic sample. As a sensitivity analysis the analyses were repeated in a sub-sample of patients and controls all without important external timing cues such as children in the household or being currently employed.

Results Participants A total of 107 patients, 72 siblings and 78 control participants were included in the study (table 1). Patients and siblings were older compared to controls. Siblings more often had children in their household than patients and controls. There were no differences in employment status between the groups. Patients experienced more depressive symptoms. Compared to the overall Dutch Bipolar Cohort, patients included in the actigraphy study had the same age and the same number of females compared to the overall sample. The group included in the actigraphy study was more frequently employed (66% in the actigraphy sample, compared to 46% in the overall sample).

Table 1. Sample characteristics showing differences between the groups.

'ƌŽƵƉƐ WĂƚŝĞŶƚƐ ^ŝďůŝŶŐƐ ŽŶƚƌŽůƐ ƉͲǀĂůƵĞ ŶсϭϬϳ ŶсϳϮ Ŷсϳϴ ŐĞ;ц^Ϳ ϱϬ;ϭϮͿ ϱϰ;ϭϮͿ ϰϳ;ϭϲͿ Ϭ͘ϬϬϯ &ĞŵĂůĞ;йͿ ϲϬ;ϱϲйͿ ϰϰ;ϲϭйͿ ϰϬ;ϱϭйͿ Ϭ͘ϰϴϬ ŚŝůĚƌĞŶ ŝŶ ϮϬ;ϭϵйͿ ϯϬ;ϰϮйͿ Ϯϭ;ϮϳйͿ Ϭ͘ϬϬϯ ŚŽƵƐĞŚŽůĚ͕ ǇĞƐ;йͿ ƵƌƌĞŶƚůǇ ϲϴ;ϲϱйͿ ϱϭ;ϳϰйͿ ϰϱ;ϱϵйͿ Ϭ͘ϭϳϰ ĞŵƉůŽǇĞĚ͕ ǇĞƐ;йͿ /^ ƐĐŽƌĞ ϭϱ͘Ϯ;ϭϭ͘ϭͿ ϳ;ϲ͘ϲͿ ϱ͘ϵ;ϰ͘ϵͿ фϬ͘ϬϬϭ ;ц^Ϳ ^ZD ƐĐŽƌĞ ϭ͘ϵ;ϭ͘ϵͿ ϭ͘Ϯ;ϭ͘ϰͿ ϭ͘ϲ;Ϯ͘ϮͿ Ϭ͘Ϭϲϯ ;ц^Ϳ Circadian variables Table 2 shows the circadian parameters of the three groups and table 3 the summary of the regression models. No differences between the non-parametric circadian variables were found between patients and controls after Bonferroni correction for multiple testing (p < 0.007). The only difference surviving Bonferroni correction was the start time

57 Chapter 5: Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls

of the 10 most active hours, where siblings showed an earlier start time compared to patients and compared to controls. There were no differences in chronotype between the groups.

Table 2. Circadian variables across groups, without analyzing differences. Mean ± SD.

WĂƚŝĞŶƚƐ ^ŝďůŝŶŐƐ ŽŶƚƌŽůƐ /ŶƚƌĂĚĂŝůǇǀĂƌŝĂďŝůŝƚǇ;/sͿ Ϭ͘ϴϭ;Ϭ͘ϭϵͿ Ϭ͘ϴϮ;Ϭ͘ϭϴͿ Ϭ͘ϳϳ;Ϭ͘ϭϵͿ /ŶƚĞƌĚĂŝůǇƐƚĂďŝůŝƚǇ;/^Ϳ Ϭ͘ϱϮ;Ϭ͘ϭϭͿ Ϭ͘ϱϰ;Ϭ͘ϭϭͿ Ϭ͘ϱϭ;Ϭ͘ϭϮͿ ZĞůĂƚŝǀĞŵƉůŝƚƵĚĞ;ZͿ Ϭ͘ϵϯ;Ϭ͘ϬϰͿ Ϭ͘ϵϰ;Ϭ͘ϬϯͿ Ϭ͘ϵϰ;Ϭ͘ϬϰͿ ĐƚŝǀŝƚǇ ŝŶ ůĞĂƐƚ ĂĐƚŝǀĞ ϱ ϭϭ͘ϵ;ϴ͘ϳϯͿ ϭϭ͘ϲϲ;ϲ͘ϱϭͿ ϭϬ͘ϴϱ;ϲ͘ϭϴͿ ŚŽƵƌƐ;>ϱͿ KŶƐĞƚŽĨ>ϱ ϭ͗Ϭϵ;ϭ͗ϮϬͿ Ϭ͗ϰϵ;ϭ͗ϬϭͿ ϭ͗ϭϲ;ϭ͗ϭϰͿ ĐƚŝǀŝƚǇ ŝŶ ŵŽƐƚ ĂĐƚŝǀĞ ϭϬ ϯϰϳ͘ϬϮ;ϭϮϲ͘ϭϱͿ ϯϲϭ͘ϲϯ;ϴϰ͘ϭϴͿ ϯϴϳ͘ϯϳ;ϭϬϭ͘ϳϴͿ ŚŽƵƌƐ;DϭϬͿ KŶƐĞƚŽĨDϭϬ ϵ͗ϰϮ;ϭ͗ϮϱͿ ϴ͗ϱϲ;Ϭ͗ϱϴͿ ϵ͗ϱϲ;ϭ͗ϯϳͿ D^&ƐĐ ϯ͘ϵϯ;Ϭ͘ϵϲͿ ϯ͘ϲϬ;Ϭ͘ϲϲͿ ϯ͘ϵϲ;ϭ͘ϬϬͿ We added IDS-SR scores and ASRM scores to the model to test if there was an effect of mood symptoms and no different results were found (supplemental table S1). A separate analysis was run only with subjects (91 patients, 70 unaffected siblings and 77 healthy controls) without any current mood symptoms based on the IDS-SR and ASRM (see supplemental table S2 for characteristics) and no different results were found (supplemental table S3).

Table 3. Results of multiple regression model.

WĂƚŝĞŶƚƐǀƐĐŽŶƚƌŽůƐ WĂƚŝĞŶƚƐǀƐƐŝďůŝŶŐƐ ^ŝďůŝŶŐƐǀƐĐŽŶƚƌŽůƐ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ /ŶƚƌĂĚĂŝůǇ ǀĂƌŝĂďŝůŝƚǇ Ϭ͘Ϭϰ Ϭ͘Ϯϭϭ ͲϬ͘ϬϬϵ Ϭ͘ϳϲϬ Ϭ͘Ϭϱϱ Ϭ͘ϬϴϮ β β β ;/sͿ /ŶƚĞƌĚĂŝůǇ ƐƚĂďŝůŝƚǇ ͲϬ͘ϬϬϯ Ϭ͘ϴϮϮ ͲϬ͘ϬϭϬ Ϭ͘ϱϰϳ Ϭ͘ϬϬϳ Ϭ͘ϳϭϮ ;/^Ϳ ZĞůĂƚŝǀĞ ŵƉůŝƚƵĚĞ ͲϬ͘Ϭϭ Ϭ͘ϬϳϮ ͲϬ͘ϬϬϱ Ϭ͘ϯϱϵ ͲϬ͘ϬϬϱ Ϭ͘ϯϯϰ ;ZͿ ĐƚŝǀŝƚǇ ŝŶ ůĞĂƐƚ ϭ͘ϯϯ Ϭ͘Ϯϰϵ Ϭ͘Ϭϯ Ϭ͘ϵϴϯ ϭ͘ϬϬϵ Ϭ͘ϯϰϳ ĂĐƚŝǀĞϱŚŽƵƌƐ;>ϱͿ

KŶƐĞƚŽĨ>ϱ ͲϬ͘ϬϮ Ϭ͘ϵϬϭ Ϭ͘ϯϭϵ Ϭ͘Ϭϴϴ ͲͲϬ͘ϯϯϮ Ϭ͘Ϭϴϯ ĐƚŝǀŝƚǇ ŝŶ ŵŽƐƚ Ͳϯϯ͘ϳ Ϭ͘Ϭϰϲ ͲϮϮ͘ϭϯ Ϭ͘ϭϵϬ ͲϮϰ͘ϲϳϲ Ϭ͘ϭϭ ĂĐƚŝǀĞ ϭϬ ŚŽƵƌƐ ;DϭϬͿ KŶƐĞƚŽĨDϭϬ ͲϬ͘Ϭϰ Ϭ͘ϴϱϱ Ϭ͘ϲϭϱ Ϭ͘ϬϬϭΎ ͲϬ͘ϲϰϳ Ϭ͘ϬϬϮΎ ŚƌŽŶŽƚǇƉĞ;D^&ƐĐͿ Ϭ͘Ϭϵ Ϭ͘ϱϮϱ Ϭ͘ϮϮϳ Ϭ͘ϬϴϮ ͲϬ͘ϭϰϴ Ϭ͘Ϯϳϰ

* Significant group differences after Bonferroni correction (p < 0.007).

58 Sensitivity analysis To test whether no differences were found due to an effect of external timing cues, an extra analysis was conducted as a sensitivity analysis. Patients and control participants without external timing influences, such as children in the household or being employed, were compared. This was studied in 24 patients and 27 controls. Patients with bipolar disorder showed more intradaily variability (β = 0.11, p = 0.018), which did not remain significant after correcting for multiple testing (Bonferroni p < 0.007).

Discussion This study showed that there were no differences in circadian characteristics or chronotype in euthymic patients with bipolar disorder compared to healthy controls in a two week actigraphy protocol. In all analyses, the only difference we found was between siblings and patients and controls, where unaffected siblings had an earlier start time of their most active 10 hours.

These findings suggest that patients are still able to maintain a regular rest-activity rhythm in the euthymic phase of the disease, just as they are able to maintain healthy sleep characteristics, as was shown previously in this sample (21). It might be that patients in this study have a fairly stable rest-activity schedule, as an important part of the treatment of bipolar disorder is restoring and promoting behavioral rhythmicity and a steady sleep schedule (32). This lack of a difference in patients compared to controls, and the fact that siblings do not show any differences compared to controls show circadian rhythm instability is not a stable trait feature. It could be a state feature, something which has been suggested for chronotype as well (33). If circadian disturbances are indeed a state feature, they might be unsuitable to be used as a endophenotype to study bipolar disorder. Although possibly unsuitable as an endophenotype, circadian disturbances are interesting to study in relation to the development of a mood episode. Future research could focus on how transitions to mood episodes relate to changes in circadian rhythm. Earlier work suggests a causal relation between circadian rhythm problems and the onset of a mood episode (34). Circadian rhythm disturbances might function as a prodrome for the onset of a mood episode, and we have recently shown that using an actiwatch can be of help to signal and prevent an upcoming mood episode (35).

The current study is the largest actigraphy study to date in patients with bipolar disorder that examined circadian variables. It included a homogeneous sample of bipolar type I patients which were screened for current mood symptoms. Furthermore, we used a 14 day actigraphy protocol which has been shown to be a good, reliable, period to assess these variables (36). Some limitations should be taken into account. As mentioned above, the sample might suffer from an inevitable sample selection bias, as the participants included in the study had already participated in a demanding baseline interview. There might be a selection of more motivated participants with a particular interest and capability of completing this study. Furthermore, it should be noted that the use of medication is not taken into account in the current study. All patients with bipolar disorder are usually advised to continue their medication, even after being

59 functionally recovered (32). The lack of a difference found might be caused by the sta- bilizing effect of a mood stabilizer on the rest-activity rhythm, one of the hypothesized mechanisms of lithium (37). As patients may use a range of different types of mood sta- bilizers with different neurophysiological mechanisms, taking every medication separately into account is a problem in all studies looking at bipolar disorder and prevented us from analyzing this effect.

From this study, we conclude patients do not show more circadian rhythm problems compared to healthy controls in the euthymic phase, demonstrating that patients are able to maintain a stable circadian rhythm, when they are outside of a mood episode. This suggests circadian rest-activity rhythm problems are a state-phenomenon, instead of a trait phenomenon, although medication use might have a masking effect. These findings create the opportunity to examine the relation between the circadian disturbances and the changes in disease state within bipolar disorder.

60 2008 Mar;10(2):256–65. DOI: 10.1111/j.1399- References 5618.2007.00505.x 1. Ferrari AJ, Stockings E, Khoo JP, Erskine HE, 13. Bullock B, Murray G. Reduced Amplitude of Degenhardt L, Vos T, et al. The prevalence and the 24 Hour Activity Rhythm : A Biomarker of burden of bipolar disorder: findings from the Vulnerability to Bipolar Disorder ? 2016; Global Burden of Disease Study 2013. Bipolar 14. Geoffroy PA, Boudebesse C, Bellivier F, Lajnef Disord. 2016;18(5):440–50. M, Henry C, Leboyer M, et al. Sleep in remitted 2. Harvey AG. Sleep and circadian rhythms bipolar disorder: A naturalistic case-control in bipolar disorder: seeking synchrony, study using actigraphy. J Affect Disord. 2014 harmony, and regulation. Am J Psychiatry. Apr;158:1–7. DOI: 10.1016/j.jad.2014.01.012 2008 Jul;165(7):820–9. DOI: 10.1176/appi. 15. Pagani L, St. Clair PA, Teshiba TM, Service SK, ajp.2008.08010098 Fears SC, Araya C, et al. Genetic contribu- 3. McClung C a. How might circadian rhythms tions to circadian activity rhythm and sleep control mood? Let me count the ways... Biol pattern phenotypes in pedigrees segregat- Psychiatry. 2013 Aug 15;74(4):242–9. DOI: ing for severe bipolar disorder. Proc Natl 10.1016/j.biopsych.2013.02.019 Acad Sci. 2015;201513525. DOI: 10.1073/ 4. Ancoli-Israel S, Cole R, Alessi C, Chambers M, pnas.1513525113 Moorcroft W, Pollak CP. The role of actigraphy 16. Hasler G, Drevets WC, Gould TD, Gottesman II, in the study of sleep and circadian rhythms. Manji HK. Toward constructing an endopheno- Sleep. 2003 May 1;26(3):342–92. type strategy for bipolar disorders. Biol Psychi- 5. Hofstra W a., de Weerd AW. How to assess circa- atry. 2006 Jul 15;60(2):93–105. DOI: 10.1016/j. dian rhythm in humans: A review of literature. biopsych.2005.11.006 Epilepsy Behav. 2008;13:438–44. 17. Roenneberg T. What is chronotype? Sleep 6. Van Someren EJ, Swaab DF, Colenda CC, Cohen Biol Rhythms. 2012 Apr 2;10(2):75–6. DOI: W, McCall W V, Rosenquist PB. Bright light 10.1111/j.1479-8425.2012.00541.x therapy: improved sensitivity to its effects on 18. Wood J, Birmaher B, Axelson D, Ehmann M, rest-activity rhythms in Alzheimer patients by Kalas C, Monk K, et al. Replicable differences application of nonparametric methods. Chrono- in preferred circadian phase between bipolar biol Int. 1999;16(4):505–18. disorder patients and control individuals. 7. Jones SH, Hare DJ, Evershed K. Actigraphic Psychiatry Res. 2009;166(2–3):201–9. DOI: assessment of circadian activity and sleep 10.1016/j.psychres.2008.03.003 patterns in bipolar disorder. Bipolar Disord. 19. Seleem M a, Merranko J a, Goldstein TR, 2005 Apr;7(2):176–86. DOI: 10.1111/j.1399- Goldstein BI, Axelson D a, Brent D a, et al. The 5618.2005.00187.x longitudinal course of sleep timing and circadi- 8. Rock P, Goodwin G, Harmer C, Wulff K. an preferences in adults with bipolar disorder. Daily rest-activity patterns in the bipolar Bipolar Disord. 2014 Dec 19;(7):1–11. DOI: phenotype: A controlled actigraphy study. 10.1111/bdi.12286 Chronobiol Int. 2014 Mar;31(2):290–6. DOI: 20. Saunders EF, Novick DM, Fernandez-Mendoza J, 10.3109/07420528.2013.843542 Kamali M, Ryan KA, Langenecker SA, et al. Sleep 9. Mckenna BS, Drummond SPA, Eyler LT. Associ- quality during euthymia in bipolar disorder: ations between circadian activity rhythms and the role of clinical features, personality traits, functional brain abnormalities among euthymic and stressful life events. Int J Bipolar Disord. bipolar patients : A preliminary study. J Affect 2013;1:16. Disord. 2014;164:101–6. DOI: 10.1016/j. 21. Verkooijen S, van Bergen AH, Knapen SE, Vreek- jad.2014.04.034 er A, Abramovic L, Pagani L, et al. An actigraphy 10. Castro J, Zanini M, Brandão S, Morgadinho F, study investigating sleep in bipolar I patients, Coelho S, Bressan R, et al. Circadian rest – ac- unaffected siblings and controls. J Affect Dis- tivity rhythm in individuals at risk for psychosis ord. 2017;208:248–54. and bipolar disorder. 2015;168:50–5. 22. First MB, Gibbon M. The Structured Clinical 11. Ankers D, Jones SH. Objective Assessment Interview for DSM-IV Axis I Disorders (SCID-I) of Circadian Activity and Sleep Patterns in and the Structured Clinical Interview for DSM- Individuals at Behavioral Risk of Hypomania. IV Axis II Disorders (SCID-II). 2004; 2009;65(10):1071–86. 23. Rush AJ, Gullion CM, Basco MR, Jarrett RB, Trive- 12. Salvatore P, Ghidini S, Zita G, De Panfilis C, di MH. The Inventory of Depressive Symptom- Lambertino S, Maggini C, et al. Circadian activ- atology (IDS): psychometric properties. Psychol ity rhythm abnormalities in ill and recovered Med. 1996;26(3):477–86. bipolar I disorder patients. Bipolar Disord. 24. Altman EG, Hedeker D, Peterson JL, Davis JM. The Altman Self-Rating Mania Scale.

61 1997;(1991). 25. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59:22–33. 26. Team RC. R: A language and environment for statistical computing. R Foundation for Statis- tical Computing, Vienna, Austria. 2013. ISBN 3-900051-07-0; 2016. 27. Kunkels YK, Knapen SE, Riese H, Emerencia AC. Actigraphy Manager [ACTman] [Internet]. 28. Roenneberg T, Wirz-Justice A, Merrow M. Life between Clocks: Daily Tempo- ral Patterns of Human Chronotypes. J Biol Rhythms. 2003 Feb 1;18(1):80–90. DOI: 10.1177/0748730402239679 29. Roenneberg T, Allebrandt K V, Merrow M, Vetter C. Social jetlag and obesity. Curr Biol. 2012 May 22;22(10):939–43. DOI: 10.1016/j. cub.2012.03.038 30. Zavada A, Gordijn M, Beersma D, Daan S, Roen- neberg T. Comparison of the Munich Chrono- type Questionnaire with the Horne‐Östberg’s Morningness‐Eveningness score. Chronobiol Int. 2005 Mar 1;22(2):267–78. DOI: 10.1081/CBI- 200053536 31. Kantermann T, Sung H, Burgess HJ. Comparing the Morningness-Eveningness Questionnaire and Munich ChronoType Questionnaire to the Dim Light Melatonin Onset. 2015;30(5):449–53. 32. Hirschfeld RM a. Guideline Watch : Prac- tice Guideline for the Treatment of Patients With Bipolar Disorder , 2Nd Edition. Practice. 2005;(November):1–9. DOI: 10.1176/appi. books.9780890423363.148430 33. Knapen S, Druiven S, Meesters Y, Riese H. Chrono- type not associated with nonremission, but with current state? Sleep. 2017;40(6). 34. Jackson A, Cavanagh J, Scott J. A systematic review of manic and depressive prodromes. J Affect Disord. 2003 May;74(3):209–17. DOI: 10.1016/ S0165-0327(02)00266-5 35. Knapen SE, Lek RFR Der, Haarman BCM, Schoevers RA. Coping with a life event in bipolar disorder : ambulatory measurement , signalling and early treatment. 2016;1–3. 36. VAN Someren EJW. Improving actigraphic sleep estimates in insomnia and dementia: how many nights? J Sleep Res. 2007 Sep;16(3):269–75. DOI: 10.1111/j.1365-2869.2007.00592.x 37. Moreira J, Geoffroy PA. Lithium and bipolar disorder: Impacts from molecular to behavioural circadian rhythms. Chro- nobiol Int. 2016;528(March):1–23. DOI: 10.3109/07420528.2016.1151026

62 Supplementary tables

Table S1. Results of multiple regression model including IDS-SR score.

WĂƚŝĞŶƚƐǀƐĐŽŶƚƌŽůƐ WĂƚŝĞŶƚƐǀƐƐŝďůŝŶŐƐ ^ŝďůŝŶŐƐǀƐĐŽŶƚƌŽůƐ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ /ŶƚƌĂĚĂŝůǇ Ϭ͘Ϭϰ Ϭ͘ϭϵϴ ͲϬ͘ϬϬϵ Ϭ͘ϳϳϵ Ϭ͘Ϭϱϯ Ϭ͘Ϭϵϲ β β β ǀĂƌŝĂďŝůŝƚǇ;/sͿ /ŶƚĞƌĚĂŝůǇ ƐƚĂďŝůŝƚǇ Ϭ͘ϬϬϵ Ϭ͘ϲϮϳ ͲϬ͘ϬϬϱ Ϭ͘ϳϲϯ Ϭ͘ϬϬϲ Ϭ͘ϳϰϰ ;/^Ϳ ZĞůĂƚŝǀĞ ŵƉůŝƚƵĚĞ ͲϬ͘Ϭϭ Ϭ͘Ϭϲϱ ͲϬ͘ϬϬϳ Ϭ͘ϮϮϰ ͲϬ͘ϬϬϰ Ϭ͘ϰϯϰ ;ZͿ ĐƚŝǀŝƚǇ ŝŶ ůĞĂƐƚ Ϯ͘ϭϯ Ϭ͘ϭϬϭ Ϭ͘ϴϴ Ϭ͘ϱϬϮ Ϭ͘ϵϰϱ Ϭ͘ϯϴϰ ĂĐƚŝǀĞϱŚŽƵƌƐ;>ϱͿ KŶƐĞƚŽĨ>ϱ ͲϬ͘Ϭϵ Ϭ͘ϲϳϬ Ϭ͘ϯϮϮ Ϭ͘ϭϭϮ ͲͲϬ͘ϯϭϴ Ϭ͘ϭϬϬ ĐƚŝǀŝƚǇ ŝŶ ŵŽƐƚ ͲϮϬ͘ϯϱ Ϭ͘Ϯϳϲ Ͳϲ͘ϳϲ Ϭ͘ϳϬϴ ͲϮϰ͘ϵϬ Ϭ͘ϭϬϳ ĂĐƚŝǀĞ ϭϬ ŚŽƵƌƐ ;DϭϬͿ KŶƐĞƚŽĨDϭϬ ͲϬ͘ϭϴ Ϭ͘ϰϭϲ Ϭ͘ϱϲ Ϭ͘ϬϬϳ ͲϬ͘ϲϯϮ Ϭ͘ϬϬϮ ŚƌŽŶŽƚǇƉĞ Ϭ͘ϭϬ Ϭ͘ϱϭϰ Ϭ͘Ϯϯ Ϭ͘ϭϬϭ ͲϬ͘ϭϰϬ Ϭ͘ϯϬϭ ;D^&ƐĐͿ

Table S2. Sample characteristics subjects without mood symptoms (ASRM < 6, IDS-SR < 26).

'ƌŽƵƉƐ WĂƚŝĞŶƚƐ ^ŝďůŝŶŐƐ ŽŶƚƌŽůƐ ƉͲǀĂůƵĞ Ŷсϵϭ ŶсϳϬ Ŷсϳϳ ŐĞ;ц^Ϳ ϱϬ;ϭϭͿ ϱϱ;ϭϮͿ ϰϳ;ϭϲͿ Ϭ͘ϬϬϭ &ĞŵĂůĞ;йͿ ϱϬ;ϱϱйͿ ϰϰ;ϲϯйͿ ϰϬ;ϱϭйͿ Ϭ͘ϯϮϮ ŚŝůĚƌĞŶ ŝŶ ϭϵ;ϮϭйͿ ϯϬ;ϰϯйͿ Ϯϭ;ϮϳйͿ Ϭ͘ϬϬϵ ŚŽƵƐĞŚŽůĚ͕ ǇĞƐ;йͿ ƵƌƌĞŶƚůǇ ϱϲ;ϲϯйͿ ϱϬ;ϳϱйͿ ϰϱ;ϲϬйͿ Ϭ͘ϭϱϱ ĞŵƉůŽǇĞĚ͕ ǇĞƐ;йͿ /^ ƐĐŽƌĞ ϭϭ͘ϲ;ϳ͘ϭͿ ϲ͘ϯ;ϱ͘ϯͿ ϱ͘ϵ;ϰ͘ϵͿ фϬ͘ϬϬϭ ;ц^Ϳ ^ZD ƐĐŽƌĞ ϭ͘ϴ;ϭ͘ϵͿ ϭ͘Ϯ;ϭ͘ϰͿ ϭ͘ϲ;Ϯ͘ϮͿ Ϭ͘ϭϰϭ ;ц^Ϳ

Table S3. Results of multiple regression model in subjects without mood problems (ASRM < 6, IDS-SR < 26).

WĂƚŝĞŶƚƐǀƐĐŽŶƚƌŽůƐ WĂƚŝĞŶƚƐǀƐƐŝďůŝŶŐƐ ^ŝďůŝŶŐƐǀƐĐŽŶƚƌŽůƐ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ ƉͲǀĂůƵĞ /ŶƚƌĂĚĂŝůǇ ǀĂƌŝĂďŝůŝƚǇ Ϭ͘Ϭϰ Ϭ͘ϮϬϭ ͲϬ͘ϬϬϰ Ϭ͘ϴϴϵ Ϭ͘ϬϱϬ Ϭ͘ϭϮϴ β β β ;/sͿ /ŶƚĞƌĚĂŝůǇ ƐƚĂďŝůŝƚǇ ͲϬ͘ϬϬϭ Ϭ͘ϵϰϯ ͲϬ͘ϬϬϯ Ϭ͘ϴϰϮ Ϭ͘ϬϬϭ Ϭ͘ϵϰϭ ;/^Ϳ ZĞůĂƚŝǀĞ ŵƉůŝƚƵĚĞ ͲϬ͘Ϭϭ Ϭ͘Ϭϲϲ ͲϬ͘ϬϬϲ Ϭ͘Ϯϵϭ ͲϬ͘ϬϬϲ Ϭ͘ϯϭϳ ;ZͿ ĐƚŝǀŝƚǇ ŝŶ ůĞĂƐƚ ϭ͘ϳϳ Ϭ͘ϭϳϮ Ϭ͘ϯϴ Ϭ͘ϳϴϬ ϭ͘ϭϰ Ϭ͘ϯϭϯ ĂĐƚŝǀĞϱŚŽƵƌƐ;>ϱͿ

KŶƐĞƚŽĨ>ϱ Ϭ͘Ϭϱ Ϭ͘ϴϬϯ Ϭ͘ϯϱϴ Ϭ͘Ϭϲϳ ͲϬ͘Ϯϳϭ Ϭ͘ϭϳϳ ĐƚŝǀŝƚǇ ŝŶ ŵŽƐƚ ͲϯϬ͘ϴϰ Ϭ͘Ϭϵϴ ͲϮϭ͘ϭϵ Ϭ͘Ϯϱϱ ͲϮϯ͘ϯϳ Ϭ͘ϭϰϰ ĂĐƚŝǀĞ ϭϬ ŚŽƵƌƐ ;DϭϬͿ KŶƐĞƚŽĨDϭϬ Ϭ͘Ϭϭ Ϭ͘ϵϱϳ Ϭ͘ϳϬ Ϭ͘ϬϬϭΎ ͲϬ͘ϱϴϵ Ϭ͘ϬϬϱΎ ŚƌŽŶŽƚǇƉĞ;D^&ƐĐͿ Ϭ͘ϭϬ Ϭ͘ϱϮϵ Ϭ͘ϮϲϬ Ϭ͘Ϭϲϭ ͲϬ͘ϭϱϵ Ϭ͘ϮϲϬ

* Significant group differences after Bonferroni correction (p < 0.007).

64 Chapter 6 Coping with a life event in bipolar disorder – ambulatory measurement, signalling and early treatment

S.E. Knapen1, R.F. Riemersma-van der Lek2, B.C.M. Haarman2, R.A. Schoevers1

BMJ Case Reports 2016 Summary Disruption of the biological rhythm in patients with bipolar disorder is a known risk factor for a switch in mood. This case study describes how modern techniques using ambulatory assessment of sleep parameters can help in signalling a mood switch and start early treatment. We studied a 40-year old female with bipolar disorder experienc- ing a life event while wearing an actigraph to measure sleep-wake parameters. The night after the life event the subject had later sleep time and shorter sleep duration. Adequate response of both the subject and the treating psychiatrist resulted in two normal nights with the use of 1 mg lorazepam, possibly preventing further mood disturbances. Ambulatory assessment of the biological rhythm can function as an add- on to regular signalling plans for prevention of episodes in patients with bipolar disorder. More research should be conducted to validate clinical applicability, proper protocols and to understand underlying mechanisms.

66 Background Modern techniques such as ambulatory assessment of physiological parameters may be helpful to monitor and signal potential elicitors of new episodes in patients with a mental disorder. In bipolar disorder disruptions of the biorhythm such as a change in sleep are known risk factors for a switch in mood(1). We describe a case where ambulatory measurement could have functioned as a signal for early treatment.

Case presentation A 40-year-old woman is in treatment for bipolar disorder type I since 8 years. The bipolar disorder was diagnosed by a psychiatrist from the department of psychiatry of the University Medical Center Groningen (UMCG) according to the Diagnostic and Statistical Manual of Mental Disorders IV criteria. Since diagnosis and treatment she was admitted once to the psychiatric department for a depressive episode. She visited her treating psychiatrist around every six weeks. Furthermore she experienced subclinical mood swings once or twice a year, particularly after stressful life events. These subclinical mood swings persisted after different psychological treatments (cognitive behavioural therapy, interpersonal psychotherapy and individual conversations). She was treated with lithium, 600mg a day, with a blood level of 0.56 mmol/l, after recent dosage low- ering because of a developing renal insufficiency, a known side effect of lithium treatment. As the renal insufficiency continued after dosage decrease of lithium the patient switched to lamotrigine as mood stabilizer. This medication change happened after the presented events.

The patient participated in the study “Circadian timing systems in bipolar disorder” in the Netherlands from July 15th 2014 till July 22nd 2014 (2). The aim of the study was to identify variations in circadian rest-activity rhythms in-vivo. As part of the study she wore an actigraph (the Actiwatch 2, Philips Respironics, validated for sleep analysis), a small wristband that can measure activity and light and calculates sleep timing, duration and quality of sleep (3,4). During the study the patient had no insight in her own actigraphy data. During this period she experienced a serious life event, the loss of a friend by an accident.

The news of the loss came to her in the afternoon. The following night she immediately had a change in her sleep pattern, went to bed approximately 2 hours later than usual and slept around 1,5 hours less than usual.

Treatment The day after the life event, she contacted her psychiatrist to discuss the event and its emotional impact, because she knew from earlier experience that this might destabi- lize her. The psychiatrist prescribed 1mg lorazepam for two consecutive nights, know- ing that sleep loss can provoke a mood episode. Using lorazepam restored a stable rhythm, with regular in bed and out of bed times and having a night of about 7,5 hours.

67 Chapter 6: Coping with a life event in bipolar disorder – ambulatory measurement, signalling and early treatment

Figure 1 Actogram of 8 days. Black bars represent activity per minute. The dark blue overlay indicates sleep time, while light blue overlay shows time in bed, without being asleep. The red blocks under the activity represent time awake. Red blocks during a sleep period indicate moments of being awake. Single asterisk marks means a life event, and double asterisks mark the intake of 1 mg lorazepam. On the third day there was a life event, coped with two nights of lorazepam use.

68 Chapter 6: Coping with a life event in bipolar disorder – ambulatory measurement, signalling and early treatment

Outcome and follow-up After two nights of lorazepam use, she slept at her regular time, going to bed at 23:14 and waking up at 8:10, without the use of lorazepam. Although she was sad, her mood remained euthymic afterwards.

See figure 1 for a visualization of the rest-activity rhythm in an actogram.

Discussion A major goal of bipolar disorder treatment is the prevention of manic or depressive episodes by early signalling and treatment (5). In this patient, an episode was probably prevented by adequate signalling of an increased risk and the use of a benzodiazepine. Two factors were important to achieve this, the insight of the patient into her disease and the adequate use of medication to maintain a regular sleeping pattern. Although there were no formal psychotherapeutic interventions from the psychiatrist, getting a brief supportive reaction from her psychiatrist could have been helpful as well. A strong, direct effect on sleep with the same strength as one might expect from lorazepam is unlikely.

As part of the study, sleep and activity were also measured by an actiwatch during this period. Apart from recognizing the potential risk of this emotional event herself, the objective measure of sleep could also have been helpful to alert the patient. A monitoring system could inform both patient and treating health personnel. For many years, the use of early alarm-systems, such as a signalling plan, has been advocated and tested, but modern technology (‘wearable technology’) now makes objective measurements very accessible to patients and practitioners(6,7).

It is clear that there is an interplay between sleep and the onset of mood episodes in bipolar disorder. The onset of mania is often preceded by changes in sleep patterns, and the antidepressant effect of sleep deprivation on relieving depression suggests a similar mechanism (8). With this in mind, continuous sleep measurement in patients with bipolar disorder could help to prevent full-blown episodes by early signalling of changes in these patterns. Future studies are needed to see if this would outperform patients’ own early recognition of an upcoming mood episode and be helpful to prevent new episodes. Current commercially available actigraphs may show the result of sleep-wake measurement on a person’s smartphone, but lack the clinical validity needed to apply them to patients (9). Further research is therefore needed, both in the form of longitudinal studies to determine what early changes in physiological patterns precede new episodes, and in the form of randomized controlled trials examining the benefit of feedback of such changes to patient and clinician in order to intervene early.

69 References 1. Jackson A, Cavanagh J, Scott J. A systematic review of manic and depressive prodromes. J Affect Disord. 2003 May;74(3):209–17. DOI: 10.1016/S0165-0327(02)00266-5 2. Verkooijen S, van Bergen AH, Knapen SE, Vreek- er A, Abramovic L, Pagani L, et al. An actigraphy study investigating sleep in bipolar I patients, unaffected siblings and controls. J Affect Dis- ord. 2017;208(April 2016):248–54. 3. Kaplan K a, Talbot LS, Gruber J, Harvey AG. Evaluating sleep in bipolar disorder: comparison between actigraphy, polysomnography, and sleep diary. Bipolar Disord. 2012 Dec;14(8):870–9. DOI: 10.1111/bdi.12021 4. Kushida C a., Chang a., Gadkary C, Guillemi- nault C, Carrillo O, Dement WC. Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disor- dered patients. Sleep Med. 2001;2(5):389–96. 5. NICE. Bipolar disorder: the assessment and management of bipolar disorder in adults, children and young people in primary and secondary care. NICE guideline (CG185) [Internet]. 2014. 6. Perry A, Tarrier N, Morriss R, McCarthy E, Limb K. Randomised controlled trial of efficacy of teaching patients with bipolar disorder to identify early symptoms of relapse and obtain treatment. BMJ. 1999 Jan 16;318(7177):149 LP-153. 7. Bos FM, Schoevers R a., aan het Rot M. Expe- rience sampling and ecological momentary assessment studies in psychopharmacology: A systematic review. Eur Neuropsychophar- macol. 2015;1–12. DOI: 10.1016/j.euroneuro.2015.08.008 8. Wehr T, Wirz-Justice A, Goodwin F. Phase Advance of the Circadian Sleep-Wake Cy- cle as an Antidepressant. Science (80- ). 1979;206(4419):710–3. 9. Kooiman TJM, Dontje ML, Sprenger SR, Krijnen WP, van der Schans CP, de Groot M. Reliability and validity of ten consumer activity trackers. BMC Sports Sci Med Rehabil. 2015;7(1):24. DOI: 10.1186/s13102-015-0018-5

70 Chapter 7 The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

S.E. Knapen1,2*, E. Snippe1, A.C. Smit1, R.A. Schoevers2, M. Wichers1, R.F. Riemersma-van der Lek2

1. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Interdisciplinary Center Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands 2. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Research School of Behavioural and Cognitive Neurosciences (BCN. Groningen, the Netherlands

In preparation Abstract Introduction Patients with bipolar disorder suffer from sleep disturbances both during, as well as prior to a mood episode. Still, sleep disturbances implicated in the transition to a mood episode have not been formally assessed using objective sleep measurements. This study aims to understand the temporal relation between sleep disturbances and mood symptoms in the transition to a mood episode.

Methods Thirteen patients with bipolar disorder were studied for 180 days with continuous actigraphy and a daily mood diary and a weekly validated mood questionnaire, of which eight were suitable for analysis. Sleep measures were calculated from actigraphy. From the daily measures mean shifts and sudden peaks were obtained and plotted together to understand the temporal relation. To find the direct interplay between mood and sleep items, the week around the start of a mood episode is specifically studied.

Results In two of the eight patients, the mean shift in mood symptoms was preceded by either a mean shift or a sudden peak in sleep symptoms. Closer analysis showed that the start of a depressive episode preceded sleep disturbances by 1.5 days (for sleep mean shifts) and 4.0 days (for sudden peaks in sleep variables), while sleep disturbances (both mean shifts and sudden peaks) preceded the start of a manic episode by respectively 2.4 days and 0.3 days.

Discussion With a novel method, we showed that in almost all patients sleep disturbances are happening at the same time as the start of a mood episode. Studying the exact order with this method we conclude that the start of a depressive episode precedes sleep disturbances, while the start of a manic episode is preceded by sleep disturbances.

72 Introduction According to the Global Burden of Disease, bipolar disorder is one of the top causes of disability worldwide (1). Bipolar disorder is a chronic psychiatric disorder characterized by episodes of mood symptoms, ranging from depressive episodes to manic episodes (2). The course of the disease differs for every patient and the main treatment focus is preventing new mood episodes, especially as patients with more mood episodes experience a worse course of the disorder with more harmful consequences of the disease in daily life (3,4). Psycho-education, alongside pharmacological treatments, aims to decrease the number and the severity of episodes through finding early markers of an upcoming mood change. For instance, an action plan, such as the wellness recovery action plan, enables a patient to acknowledge changes in their own behavior and mood (5,6). However, patients and clinicians lack objective markers to signal upcoming mood episodes. A possible objectively assessable marker of an upcoming manic or depressive episode might be disturbances of the sleep/wake rhythm.

Sleep alterations are clearly implicated in bipolar disorder, being one of the diagnostic criteria for both depressive and manic episodes (2,7). Sleep disturbances are also shown in the euthymic phase of the disease, the period where patients show no severe mood symptoms (8). Furthermore, Takaesu et al. showed in a sample of 104 patients with bipolar disorder that patients with circadian rhythm sleep wake disorders had a shorter time to relapse (9). A more direct relationship between sleep and mood symptoms has been shown in patients with bipolar disorder as well. When patients are asked what preceded their onset of a mood episode, a median of 77% of the patients reported sleep disturbances as an early symptom of a manic episode, making it the most prominent prodrome (early symptom of a disorder) for manic problems (10). In contrast, 24% of the patients reported sleep disturbances as a prodrome for a depressed episode. In a study using sleep measures, obtained from a sleep diary, a mood change was preceded by a change in sleep and or bedrest duration the night before the change in mood (11). Objectively sleep has been implied as a potential marker for change in mood in patients with bipolar disorder, however this was in the late 70s and early 80s of the previous century (12,13). These studies show sleep disturbances and mood changes happen at about the same time, although their relation, and the exact order, in the transition to a mood episode is still unknown. To examine whether sleep disturbances precede the start of a depressive or manic episode, objective and non-invasive measurement of sleep and frequent mood assessment in daily life are needed. Furthermore, in order to be able to register a mood episode, patients with bipolar disorder should be monitored for a long term.

A non-invasive way to study sleep objectively is using a wrist-worn device, a so called actiwatch, measuring activity over the day and night (14,15). Actigraphy is validated with the golden standard of sleep, polysomnography, in bipolar disorder and is proven to be a cost-effective method to assess sleep in patients in their everyday life (16). We showed in a case report that sleep as measured with actigraphy, combined with the patients’ own experiences, could be useful in preventing clinical relapses, as a clear later sleep onset, and a shorter sleep duration was seen (17). This method of studying patient behaviour in real time and in their own setting, is called ecological momentary

73 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

assessment (EMA) and provides a unique opportunity to study the link between sleep and mood problems when combined with a mood diary (18). To study the temporal order of sleep and mood disturbances in patients with bipolar disorder we studied 13 patients for 180 days, using both subjective mood ratings and objective sleep measures. We will study both sudden peaks in sleep measures (for example, one extreme short sleep duration) and mean shifts in both sleep and mood items. We expect to find more abrupt changes in sleep preceding a manic episode and no abrupt changes in sleep preceding a depressive episode (11).

Material and methods Patients Patients with bipolar disorder were recruited from the outpatient clinic of the Univer- sity Center Psychiatry of the University Medical Center Groningen. The treating psychiatrist informed the patient about the study and asked if he or she could be approached for the study. If so, patients were called by the investigator and after a phone screening to exclude for obstacles regarding wearing an actigraph or filling in a mood diary daily for 180 days, they were invited for the baseline interview in the hospital. Furthermore, patients were recruited from the Dutch patient society (Vereniging voor Manisch De- pressieven en Betrokkenen).

Inclusion criteria were bipolar disorder type I, and the motivation to participate in such a long-term protocol. Exclusion criteria were somatic diseases which could hamper actigraphy measures, and somatic sleep disorders (such as sleep apnea). Bipolar disorder type I and comorbid psychiatric disorders were confirmed using the Mini-Internation- al Neuropsychiatric Interview (MINI) (19). Chronotype was assessed using the Munich Chronotype Questionnaire (20). All patients gave written informed consent and the medical ethical committee of the University Medical Center Groningen waived the ne- cessity to review the study due to its non-invasive nature and as it is an extension of care as usual. All procedures were in accordance with the declaration of Helsinki. Fif- teen patients provided informed consent for the study, of which one (participant #14) dropped out during the study for personal reasons. Data from one participant (participant #5) showed to be largely incomplete after the study and were excluded from the analysis as well. In total data from 13 patients were suitable prior to the analyses. Data acquisition Sociodemographic variables were assessed using a questionnaire, where health problems, recent doctor visits and medication use were noted as well. Patients were measured for 180 consecutive days. All daily and weekly data were assessed online, with the patient receiving a text or an email notification that the diary is available to fill in. Sleep Patients were instructed to wear the Motionwatch 8 (CamNtech), a light weight, water-proof accelerometer, continuously, only taking it off when absolutely necessary, for instance during a sauna visit. Patients completed an electronic sleep diary daily in the morning re-

74 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

porting when they turned the lights off the evening before, when they tried to fall asleep and how many minutes it took to fall asleep. They also reported at what time they got up and how they rated their own night. Sleep variables were calculated from the actigraphy data, using our own developed script in R. The script for the analysis can be retrieved from https://github.com/compsy/ACTman/ (21,22). The functionality of the script is to select a rest period where sleep is possible using the sleep log. From the activity during that period likely wake bouts are calculated. After the rest period was defined, the sleep algorithm was used to calculate sleep duration, timing of sleep onset, timing of sleep offset, sleep onset latency, sleep efficiency (i.e. minutes asleep divided by minutes in bed), wake after sleep onset (WASO). As a marker of circadian misalignment the composite phase delay is computed, which is the distance of the midpoint of sleep of the current night to the past night and to the ideal midpoint of sleep based on the chronotype of a subject (23). Mood In the evening patients filled in daily the electronic Life-Chart (LC-self), reporting their mood on a VAS scale from ‘depressed’ (0) to ‘manic’ (100) (24). On the LC-self they also reported medication use, alcohol and drug use and reported important life events. Fur- thermore, they filled in a set of questions daily on a VAS scale ranging from ‘Not at all’ to ‘Totally’. Such questions included “I feel down”, “I feel excited” and “I feel satisfied”, a full list of items can be found in table S1. These questions were selected by three people and aimed to reflect day to day fluctuations in bipolar disorder. Ten questions were mandatory to fill in, covering both an elevated and a decreased mood. Patients could select three additional ‘manic’ items, and two additional ‘depressed’ items, and they could provide one question themselves as well. Manic and depressive episodes Manic and depressive episodes were defined using validated questionnaires. On a weekly basis patients completed the Inventory of Depressive Symptomatology – Self Rating (IDS-SR) and the Altman Self Rating Scale for Manic symptoms (ASRM) (25,26). For a manic episode, patients had to score above 5 points on the Altman Self Rating scale for Manic symptoms (ASRM) on two consecutive weeks, making sure there was at least one full week of manic symptoms (26). Furthermore, patients had to report on the Life-Chart above the midline at least 75% of the time. For a depressed episode, patients had to score above 25 points on the IDS-SR on at least 3 consecutive weeks, resulting in at least two full consecutive weeks of symptoms. They also had to report their daily mood below the midline at least 75% of the time. Selection of the episodes was performed by two independent raters (first and last author) and discrepancies were discussed. Data analysis Finding disruptions in the symptoms Sudden peaks in sleep and mood were defined as extreme outliers and quantified using a method known as a process control chart (27), which has been used before to detect special-cause variation in health care management (28). First, a stable period before the manic or depressive episode was selected in order to define the outliers. Stable periods were defined as five consecutive weeks of no manic or depressive symptoms and a lower variability on the Life-Chart compared to the full 180 days. This five-week

75 period was selected to reflect 35 days, which is similar to a previously found ideal period to detect special-cause variation (29). The stable periods were selected as early as possible in the data, and had to end at least two weeks before the onset of an episode. All selections were done by two independent raters, and discrepancies were discussed. When conforming to all the rules was not possible, for instance due to too many mood symptoms in the stable period, the most ideal stable period was selected. Of the 11 patients with a mood episode, three experienced a mood episode, following the definition above, during the selected ‘stable’ period, and were therefor excluded from the analysis (two depressed, one manic). This resulted in eight patients which were suitable for analysis. Using data of the stable period, we computed the mean of each variable of interest and included upper and lower limits which consist of three times the standard deviation of that variable during the stable period. Whenever a variable scored out of these limits, this was marked as an outlier (fig 1A).

To detect mean shifts in the sleep and mood variables, changepoint analysis was performed using the strucchange package in R. This package determines the earliest moment on which a significant structural change in the time series can be detected (30). For every variables (both mood and sleep items) the change point and confidence intervals around the change point were estimated using the sequential F-test from the strucchange package in R (30,31), requiring a minimal segment size of 7 days (fig 1B). The period from the start of the stable date up until the end of the episode period was used for the changepoint analysis.

Figure 1. Example of detection of mean shifts and extreme values based on raw data

Note: Panel A shows the outlier analysis, where the dashed lines show the upper and lower limit as calculated by 3 times the standard deviation of the variable in the stable period (marked as yellow). Point above or below those limits are marked red and transferred to the final plot (C). Panel B shows the change point

76 detection, where the black dashed horizontal lines show the mean of that the variable during that period, the vertical black dashed lines and the red and blue dot show the change points, and the horizontal blue and red lines represent the confidence intervals. Change points and confidence intervals are transferred to the image in panel C, which shows an example of how the CPs (points with confidence intervals) and outliers (triangles) are plotted together. The size of the triangle is based on how much the extreme value deviates from the mean.

Testing the temporal order All outliers from the process control chart method and change points (CPs) from the change points analysis were exported and plotted over time for the analysis (fig 1C). This process was repeated for all variables, resulting in one large CP/outlier plot (figure 2). First, the timing of the start of an episode was determined based on the change points in the mood variables. For every patient, we selected all CPs in the transition period preceding a mood episode (i.e., the period between the stable period and the mood episode) and counted the number of CPs that were a maximum of 7 days apart and of which the confidence intervals were overlapping. If multiple CPs on the same items were found, the cluster of mood items with the largest number of CPs in the same period closest to the episode, was defined to be the start of the mood episode.

Second, the occurrence of CPs and outliers in sleep items 7 days around the start of the mood episode were studied. If the CPs (with confidence interval) or outliers in sleep were not overlapping with the start of the mood episode (i.e. falling within the confidence interval of one of the CPs in the mood cluster) change in these variables was reported to occur as ‘before’ the start in the mood episode. If the sleep items were overlapping, it was reported as occurring ‘simultaneously’ with the start of a mood episode. To further understand the direct interplay between mood CPs and sleep variables, the distances in days between the average sleep CPs, sleep outliers and average mood CPs indicating the start of a mood episode were calculated and analyzed between type of episode between the patients. See figure 2 for a graphic explanation of this process.

Results Group characteristics Eight patients were suitable for analyses, aged between 31 and 67 years, of which seven women and one man. Further characteristics of the patients can be found in table 1. The weekly scores, together with the highlighted selection of stable and episode periods for all individuals can be found in the supplementary information (fig S1-S8). Of the eight patients with a mood episode, three experienced a manic episode and five a depressive episode. Figure 1 shows the data of one participant to illustrate how we approached the analysis (plots of the other patients are included in the supplementary information).

77 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

Figure 2. The selection of mean shifts and extreme values in one patient.

Note: First, the cluster of mood variables which constitute the switch to a mood episode were selected (asterisks). Next, the outer confidence interval of these selected variables was used to determine whether sleep symptoms preceded or followed on the mood symptoms (dotted line). In this case, the sleep disturbances (both changepoint and outliers) were all within in the confidence interval of the mean shifts in mood (red oval).

78 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

Table 1. Characteristics of the study patients.

WĂƌƚŝĐŝƉĂŶƚ ^Ğǆ ŐĞ DŽŽĚ KƚŚĞƌ DŽŽĚĞƉŝƐŽĚĞ ^ĞǀĞƌŝƚǇ ƉŝƐŽĚĞ ƐƚĂďŝůŝǌĞƌ ƉƐǇĐŚŽƚƌŽƉŝĐƐ Ύ;ƚǇƉĞͿ ƐĐŽƌĞ;/^Ͳ ĚƵƌĂƚŝŽŶŝŶ ^Zͬ^ZD͕ ǁĞĞŬƐ ŵĞĂŶц^Ϳ

ϭ & ϰϮ >ŝƚŚŝƵŵ KůĂŶǌĂƉŝŶĞ͕ ĞƉƌĞƐƐĞĚ ϯϬ͘ϱцϮ͘ϲ ϰ ĞƐĐŝƚĂůŽƉƌĂŵ Ϯ & ϯϲ >ŝƚŚŝƵŵ KůĂŶǌĂƉŝŶĞ͕ DĂŶŝĐ ϵцϭ ϯ ĨůƵŽǆĞƚŝŶĞ ϯ & ϱϰ ĂƌďĂŵĂǌĞƉŝŶĞ ĞƉƌĞƐƐĞĚ ϯϯ͘ϳцϯ͘ϱ ϯ ϰ & ϲϳ sĂůƉƌŽĂƚĞĂĐŝĚ DĂŶŝĐ ϭϭ͘ϯцϰ͘ϲ ϯ ϴ D ϯϭ >ŝƚŚŝƵŵ DĂŶŝĐ ϭϬ͘ϱцϬ͘ϳϭ Ϯ ϵ & ϰϲ >ŝƚŚŝƵŵ ĞƉƌĞƐƐĞĚ ϯϬ͘ϯцϯ͘ϳ ϰ ϭϭ & ϰϱ >ĂŵŽƚƌŝŐŝŶĞ ĞƉƌĞƐƐĞĚ Ϯϵ͘ϴцϮ͘ϵ ϰ ϭϱ & ϯϱ >ŝƚŚŝƵŵ sĞŶůĂĨĂǆŝŶĞ ĞƉƌĞƐƐĞĚ ϯϰ͘ϯцϴ͘ϱ ϯ Table 2. The temporal order of mean shifts and extreme values in mood and sleep.

/ dǇƉĞŽĨ DĞĂŶƐŚŝĨƚƐ ŝƐƚĂŶĐĞ ^ƵĚĚĞŶ ŝƐƚĂŶĐĞŵŽŽĚ ĞƉŝƐŽĚĞ ŝŶƐůĞĞƉ ŵŽŽĚŵĞĂŶ ƉĞĂŬƐŝŶ ŵĞĂŶƐŚŝĨƚĂŶĚ ŵĞĂƐƵƌĞƐ ƐŚŝĨƚĂŶĚ ƐůĞĞƉ ƐůĞĞƉƐƵĚĚĞŶ ƐůĞĞƉŵĞĂŶ ŵĞĂƐƵƌĞƐ ƉĞĂŬƐǁŝƚŚŝŶϳ ƐŚŝĨƚƐ;ŝŶ ĚĂǇƐ ĚĂǇƐͿ ϬϬϭ с Ͳϭ͘ϰ →ͬс ϯ͘ϳ ϬϬϮ D с ϭ͘ϵ с Ϯ͘ϰ ϬϬϯ с ϲ͘ϯ с ϰ͘ϯ ϬϬϰ D E E →Ύͬс Ͳϭ͘ϳ ϬϬϴ D с Ͳϲ͘ϳ с Ͳϭ͘ϳ

ϬϬϵ с ϭ͘Ϯ с E Ϭϭϭ → E E E Ϭϭϱ с ͲϬ͘ϭ →Ύͬс E

*points were found more than 7 days away from the change in mood, but were within 14 days of the change. D = depressive, M = manic, NA = not available. If the sleep measure preceded the mood changepoint an arrow is shown in the table, if the symptoms were found simultaneously an equal to sign is shown. The distance is calculated by looking at the distance of the points only in the week before and after the mean mood changepoint. If the distance in days is negative, the sleep measure preceded the mood mean shift, when positive it followed the mood mean shift.

79 Do sleep disturbances occur before a mean shift in mood? Table 2 shows whether sleep mean shifts and sudden peaks preceded or occurred si- multaneous with the mood change point, based on the individual plots including the mean shifts and extreme values (see supplementary information). We found in one subject (#011) that a mood mean shift was preceded by a mean shift in sleep variables (sleep offset and sleep duration) and in one subject (#001) that the mood mean shift was preceded by sleep sudden peaks (sleep efficiency and minutes awake after sleep onset) within 7 days. When the analyzed period was expanded to 14 days, the mood mean shift was preceded in three patients by a sudden peak in sleep variables. In other patients both mean shifts and sudden peaks in sleep variables were seen within the confidence intervals of the start of the episode.

What is the amount of time between sleep disturbances and the start of a mood episode? Distances are calculated from the mood mean shift to both the sleep mean shift and the sleep sudden peak within a week of the mood mean shift and studied in all patients together. When we look at the calculated average differences in time between the sleep mean shift and mood mean shift for both types of episodes together, the sleep mean shift followed the mood mean shift by 0.2 days. Sleep sudden peaks followed the mood mean shift after on average 1.4 days. In the transition to depressive episodes the sleep mean shift followed the mood mean shift with 1.5 days and the sleep sudden peaks followed the mood mean shift by 4.0 days, in all depressive episodes analyzed. In the transition to a manic episode the sleep mean shift preceded the mood mean shift by 2.4 days and sleep sudden peaks preceded the mood mean shift by 0.3 days, in all manic episodes analyzed. Note that not all episodes follow the same order as the calculated average for all patients (table 2).

Zooming in on patient #011: what changes besides sleep disturbances? We further studied the patients that showed a change in sleep before change in mood to examine whether there were other changes happening around the days that the sleep disturbances started. In patient #011, who showed a sleep mean shift before the mood mean shift, the weekly questionnaires show a peak in manic symptoms two weeks before the onset of the depressive episode. This can be traced back as well in the plot with the CPs and outliers, where the mean shifts linked to the depressive episode were clearly preceded by a lot of mean shifts in the opposite direction. For example, the score on “I feel down” first decreased, and at the start of the depressive episode it increased. The opposite happened with “I feel excited” and “I feel satisfied” with an increase first and later a decrease indicating the onset of the depressive episode. This shows that the depressive episode were preceded by manic symptoms. This also explains the decrease in sleep duration and earlier sleep offset which preceded the depressive episode, it could be that this was related to the manic symptoms the patient experienced.

80 Zooming in on patient #001: what happened besides sudden peaks in sleep variables? In patient #001, a peak in manic symptoms was found before the onset of the depressive episode. However, not sufficient mood mean shifts were found in the CP/outlier plot. Two mean shifts which were found in that plot showed that the patient had an increase in irritated (“I feel irritated”) and stressed feelings (“I feel stressed”). This might have resulted in the sudden peak in minutes awake after sleep onset and therefore the sleep efficiency. Studying the self-reported events which the patient wrote down in the Life Chart showed no particular activities or life events on the day of those mean shifts.

Discussion In this paper, we set out to study the temporal order between sleep disturbances and mood symptoms in the transition from euthymia to a mood episode in patients with bipolar disorder. In the eight patients with a mood episode which were suitable for analysis, we found that only two out of eight patients showed sleep disturbances occurring significantly earlier than the change in mood symptoms. The other six patients showed sleep disturbances occurring simultaneously with the mean shift in mood symptoms. However, when examining the mean time in days between mean shifts and sudden peaks in mood and sleep, we found that sleep disturbances in the transition to a depressive episode were following the changes in mood (on average 1.5 days later for mean shifts and 4.0 days later for sudden peaks), while sleep disturbances in transition to a manic episode preceded the change in mood (on average 2.4 days for mean shifts and 0.3 days for sudden peaks).

When studying the transition to manic episodes specifically, it seems that especially in the sudden peaks in the sleep timing parameters are involved (sleep onset, offset and duration). We only found this using the less restrictive method, as the disturbances are very close to the mood change, which results in overlapping confidence intervals. This is in line with the literature, where only one shorter night of sleep can result in more manic symptoms (11,12). Specifically the study by Bauer et al., showed that sleep duration was significantly decreased according to the sleep diary the night prior to a mood shift in the manic direction (11).

Studying the transition to a depressive episode specifically we see that the sleep disturbances are in the same period and sometimes later than the change in mood. This is in line with what can be expected from the literature, as a much smaller percentage of the patients report sleep disturbances as a possible prodrome for a depressive episode (10).

To study the temporal relation, we developed a new method based on existing methods in statistical process control. We combined the process control chart, developed to find outliers, or sudden peaks, in a process, with a change point analysis, which is developed to find changes in the distribution, such as mean shifts in a variable. To our knowledge, we are the first to use a process control chart method to select sudden peaks of interest in studying sleep and mood. One earlier study by Rosenfield et al. showed that change point analysis might be useful to detect cardio-respiratory changes prior to panic attacks,

81 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

showing it might be useful in psychiatry (32). In the current study, both methods are combined in one large plot, in which the temporal relation was studied. Different techniques can be used to obtain mean shifts in the data. Some methods not only use a change distribution of the mean score, but also in variance. We tested the strucchange breakpoints method against a more robust analysis with the ECP package in R, and change points were found on the same dates (analysis not included) (30,31,33). As strucchange is currently the only method which also reports confidence intervals, this method was chosen for the analysis, as this resulted in more strict results. With this method, we show that it is possible to study the association between mood symptoms and sleep disturbances on an individual level. When we zoomed in on individual patients, it seems that information is lost when looking solely at the days on which the change in mood happens. This discrepancy shows our method might be an interesting add-on for the current clinical practice, where patients keep track of their symptoms together with their treating physician.

Our results indicate sleep disturbances might precede the transition to a mood episode, specifically manic episodes. This indicates that sleep disturbances might be a useful early warning signal to prevent further mood disturbances (17). A next step would be to look at the sleep variables from our analysis to see if we could actually predict a mood episode, without having any prior knowledge on when the mood episode would happen. This is especially interesting since commercial activity trackers become more advanced and more readily available, there is a great opportunity in analyzing ambulatory rest-activity rhythms and their relation to mood (34). Specifically in patients with bipolar disorder this is promising, as patients show an interest in a wide range of applications for self-management of their disease (35).

This study is, to our knowledge, the first studying objectively assessed sleep variables specifically in the transition to a mood episode. Although patients have been studied in the clinic for a longer time, studying patients in their own daily life with objective measures, without any restrictions, has never been done before. However, there are some limitations to this paper. Most notably is the small sample size with only eight patients suitable for analysis. This resulted in only 5 depressive episodes and 3 manic episodes to be analyzed. Although there a no other studies with this type of data, the generalizability of our findings is hampered by this number. Another limitation is the fact that we mainly studied sleep variables, while it is also possible to calculate circadian rhythm variables from actigraphy data. As changes in the rhythm might be an interesting early warning signal itself as well, this might be a promising extension of this current work (9,36). Although circadian rhythm regulation might be interesting to study, the methods we used are specifically suitable for variables assessed on a daily basis and the circadian rhythm variables are typically calculated over a period of at least a week (37). In order to still retain some circadian rhythmicity variables in the data, a variable which can be calculated on a daily basis, the composite phase delay was computed and used in the analyses as well (38).

In conclusion, we showed a novel method to study the temporal order of changes in symptomatology related to mood episodes and showed that patients suffer from sleep disturbances in the two to three days prior to the mood change in the transition to a manic episode, while this is not the case in the transition to a depressive episode.

82 cle as an Antidepressant. Science (80- ). References 1979;206(4419):710–3. 1. Murray CJL, Vos T, Lozano R, Naghavi M, 13. Wehr TA, Goodwin FK, Wirz-Justice A, Breitmai- Flaxman AD, Michaud C, et al. Disability-ad- er J, Craig C. 48-hour sleep-wake cycles in man- justed life years (DALYs) for 291 diseases and ic-depressive illness: naturalistic observations injuries in 21 regions, 1990-2010: A systematic and sleep deprivation experiments. Arch Gen analysis for the Global Burden of Disease Study Psychiatry. 1982;39(5):559–65. 2010. Lancet. 2012;380(9859):2197–223. 14. Hofstra W a., de Weerd AW. How to assess circa- 2. Association AP. Diagnostic and statistical man- dian rhythm in humans: A review of literature. ual of mental disorders (DSM-5®). American Epilepsy Behav. 2008;13:438–44. Psychiatric Pub; 2013. 15. Ancoli-Israel S, Cole R, Alessi C, Chambers M, 3. Peters A, Sylvia LG, Da Silva Magalhães P V., Moorcroft W, Pollak CP. The role of actigraphy Miklowitz DJ, Frank E, Otto MW, et al. Age at on- in the study of sleep and circadian rhythms. set, course of illness and response to psycho- Sleep. 2003 May 1;26(3):342–92. therapy in bipolar disorder: Results from the 16. Kaplan K a, Talbot LS, Gruber J, Harvey AG. Systematic Treatment Enhancement Program Evaluating sleep in bipolar disorder: com- for Bipolar Disorder (STEP-BD). Psychol Med. parison between actigraphy, polysomnog- 2014;44(16):3455–67. raphy, and sleep diary. Bipolar Disord. 2012 4. Bond K, Anderson IM. Psychoeducation for re- Dec;14(8):870–9. DOI: 10.1111/bdi.12021 lapse prevention in bipolar disorder: A system- 17. Knapen SE, Lek RFR Der, Haarman BCM, Scho- atic review of efficacy in randomized controlled evers RA. Coping with a life event in bipolar trials. Bipolar Disord. 2015;17(4):349–62. disorder : ambulatory measurement , signalling 5. Perry A, Tarrier N, Morriss R, McCarthy E, Limb and early treatment. 2016;1–3. K. Randomised controlled trial of efficacy of 18. Bos FM, Schoevers R a., aan het Rot M. Expe- teaching patients with bipolar disorder to rience sampling and ecological momentary identify early symptoms of relapse and obtain assessment studies in psychopharmacology: treatment. BMJ Br Med J. 1999;318(Janu- A systematic review. Eur Neuropsychophar- ary):149–53. macol. 2015;1–12. DOI: 10.1016/j.euroneu- 6. Copeland M. Guide to Developing a WRAP– ro.2015.08.008 Wellness Recovery Action Plan. 2000. 19. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim 7. Harvey AG. Sleep and circadian rhythms P, Janavs J, Weiller E, et al. The Mini-Interna- in bipolar disorder: seeking synchrony, tional Neuropsychiatric Interview (MINI): the harmony, and regulation. Am J Psychiatry. development and validation of a structured 2008 Jul;165(7):820–9. DOI: 10.1176/appi. diagnostic psychiatric interview for DSM-IV and ajp.2008.08010098 ICD-10. J Clin Psychiatry. 1998;59:22–33. 8. Geoffroy P a, Scott J, Boudebesse C, Lajnef M, 20. Roenneberg T, Wirz-Justice A, Merrow Henry C, Leboyer M, et al. Sleep in patients M. Life between Clocks: Daily Temporal with remitted bipolar disorders: a meta-analy- Patterns of Human Chronotypes. J Biol sis of actigraphy studies. Acta Psychiatr Scand. Rhythms. 2003 Feb 1;18(1):80–90. DOI: 2014 Nov 28;1–11. DOI: 10.1111/acps.12367 10.1177/0748730402239679 9. Takaesu Y, Inoue Y, Ono K, Murakoshi A, Fu- 21. Kunkels YK, Knapen SE, Riese H, Emerencia AC. tenma K, Komada Y, et al. Circadian Rhythm Actigraphy Manager [ACTman] [Internet]. Sleep-Wake Disorders Predict Shorter Time to 22. R Core Team. R: A Language and Environment Relapse of Mood Episodes in Euthymic Patients for Statistical Computing [Internet]. Vienna, With Bipolar Disorder: A Prospective 48-Week Austria; 2018. Study. J Clin Psychiatry. 2017 Dec;79(1). 23. Fischer D, Vetter C, Roenneberg T. A novel 10. Jackson A, Cavanagh J, Scott J. A systematic method to visualise and quantify circadian review of manic and depressive prodromes. J misalignment. Nat Publ Gr. 2016;1–10. DOI: Affect Disord. 2003 May;74(3):209–17. DOI: 10.1038/srep38601 10.1016/S0165-0327(02)00266-5 24. Born C, Amann BL, Grunze H, Post RM, Schärer 11. Bauer M, Grof P, Rasgon N, Bschor T, Glenn T, L. Saving time and money: a validation of the Whybrow PC. Temporal relation between sleep self ratings on the prospective NIMH Life-Chart and mood in patients with bipolar disorder. Method (NIMH-LCM). BMC Psychiatry. 2014 Bipolar Disord. 2006 Apr;8(2):160–7. DOI: Jan;14:130. DOI: 10.1186/1471-244X-14-130 10.1111/j.1399-5618.2006.00294.x 25. Rush AJ, Gullion CM, Basco MR, Jarrett RB, Trive- 12. Wehr T, Wirz-Justice A, Goodwin F. Phase di MH. The Inventory of Depressive Symptom- Advance of the Circadian Sleep-Wake Cy- atology (IDS): psychometric properties. Psychol Med. 1996;26(3):477–86.

83 26. Altman EG, Hedeker D, Peterson JL, Davis JM. The Altman Self-Rating Mania Scale. 1997;(1991). 27. Montgomery DC. Introduction to statistical quality control. John Wiley & Sons (New York); 2009. 28. Benneyan, JC.,Lloyd, RC., Plsek, PE. Statis- tical process control as a tool for research and healthcare improvement. Qual Saf Heal Care. 2003;12(6):458–64. DOI: 10.1136/ qhc.12.6.458 29. Wiemken TL, Furmanek SP, Carrico RM, Mattingly WA, Persaud AK, Guinn BE, et al. Process control charts in infection prevention: Make it simple to make it happen. Am J Infect Control. 2017;45(3):216–21. DOI: 10.1016/j. ajic.2016.09.021 30. Zeileis A, Leisch F, Hornik K, Kleiber C. strucchange: An R Package for Testing for Structural Change in Linear Regression Models. J Stat Softw. 2002 Jan 10;1(2). 31. Zeileis A, Kleiber C, Kramer W, Hornik K. Testing and dating of structural changes in practice. 2003;44:109–23. 32. Rosenfield D, Zhou E, Wilhelm FH, Conrad A, Roth WT, Meuret AE. Change point analysis for longitudinal physiological data : Detection of cardio-respiratory changes preceding panic attacks. Biol Psychol. 2010;84(1):112–20. DOI: 10.1016/j.biopsycho.2010.01.020 33. James NA, Matteson DS. ecp: An R package for nonparametric multiple change point analysis of multivariate data. arXiv Prepr arX- iv13093295. 2013; 34. Blaauw FJ, Schenk HM, Jeronimus BF, van der Krieke L, de Jonge P, Aiello M, et al. Let ’ s get Physiqual – a platform to integrate sensor technology with ecological momentary assessments. J Biomed Inform. 2015;63:1–9. DOI: http://dx.doi.org/10.1016/j.jbi.2016.08.001 35. Nicholas J, Boydell K, Christensen H. Beyond symptom monitoring: Consumer needs for bipolar disorder self-management using smart- phones. Eur Psychiatry. 2017;44:210–6. DOI: 10.1016/j.eurpsy.2017.05.023 36. Novak D, Albert F, Spaniel F. Analysis of actigraph parameters for relapse prediction in bipolar disorder: A feasibility study. Conf Proc . Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Annu Conf. 2014 Aug;2014:4972– 5. DOI: 10.1109/EMBC.2014.6944740 37. VAN Someren EJW. Improving actigraphic sleep estimates in insomnia and dementia: how many nights? J Sleep Res. 2007 Sep;16(3):269–75. DOI: 10.1111/j.1365-2869.2007.00592.x 38. Fischer D, Vetter C, Roenneberg T. A novel method to visualise and quantify circadian misalignment. Nat Publ Gr. 2016;1–10. DOI: 10.1038/srep38601

84 Supplementals

Table S1. VAS items, both obligatory to report and other possible items patients could choose.

/ƚĞŵ /ĨĞĞůĞǆĐŝƚĞĚ /ĨĞĞůŝƌƌŝƚĂƚĞĚ /ĨĞĞůƐĂƚŝƐĨŝĞĚ /ĨĞĞůĚŽǁŶ /ĨĞĞůƐƚƌĞƐƐĞĚ /ĂŵĨƵůůŽĨĞŶĞƌŐǇ /ĨĞĞůĂŐŝƚĂƚĞĚ /ĨĞĞůƚŝƌĞĚ /ĂŵĂďůĞƚŽĐŽŶĐĞŶƚƌĂƚĞ /ŚĂǀĞƚƌƵƐƚŝŶŽƚŚĞƌƐ DĂŶŝĐŝƚĞŵƐΎ DǇƚŚŽƵŐŚƚƐĂƌĞŐŽŝŶŐĨĂƐƚ /ŚĂǀĞĂůŽƚŽĨƚŚŽƵŐŚƚƐŝŶŵǇŚĞĂĚ /ĐĂŶŚĂŶĚůĞĂŶǇƚŚŝŶŐ /ĨĞĞůĞǆƚƌĞŵĞůǇǁĞůů /ŵĂĚĞĂůŽƚŽĨĂƉƉŽŝŶƚŵĞŶƚƐƚŽĚĂǇ /ƚĂůŬĂůŽƚƚŽĚĂǇ /ĨĞĞůŝŵƉƵůƐŝǀĞ ĞƉƌĞƐƐŝǀĞŝƚĞŵƐΎΎ /ĨĞĞůŝŶƐĞĐƵƌĞ /ĨĞĞůůŽŶĞůǇ /ĨĞĞůĂƉĂƚŚĞƚŝĐ /ĨĞĞůĂŶǆŝŽƵƐ /ĨĞĞůŐƵŝůƚ /ƐƉĞŶĚĂůŽƚŽĨƚŝŵĞŝŶďĞĚƚŽĚĂǇ ǀĞƌǇƚŚŝŶŐƚĂŬĞƐĞĨĨŽƌƚƚŽĚĂǇ /ǁŽƌƌǇĂůŽƚ ǆƚƌĂΎΎΎ “Own factor” * Patients were allowed to pick maximum three items ** Patients were allowed to pick two items *** Patients were allowed to set their own item

85 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

Figure S1. Weekly scores and CP / outlier plot of subject #001

86 Figure S2. Weekly scores and CP / outlier plot of subject #002

87 Figure S3. Weekly scores and CP / outlier plot of subject #003

88 Figure S4. Weekly scores and CP / outlier plot of subject #004

89 Chapter 7: The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder

Figure S5. Weekly scores and CP / outlier plot of subject #008

Figure S6. Weekly scores and CP / outlier plot of subject #009

91 Figure S7. Weekly scores and CP / outlier plot of subject #011

92 Figure S8. Weekly scores and CP / outlier plot of subject #015

93 Chapter 8 Fractal biomarker of activity in patients with bipolar disorder

S.E. Knapen1,2*, P. Li2, R.F. Riemersma- van der Lek1, S. Verkooijen3, M.P.M. Boks3, R.A. Schoevers1, K. Hu2,#, F.A.J.L. Scheer2,#

1. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Interdisciplinary Center Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands 2. Division of Sleep and Circadian Disorders, Brigham and Women’s Hospital; and Division of Sleep Medicine, Harvard Medical School, Boston, Massachusetts, United States. 3. Brain Center Rudolf Magnus, University Medical Center Utrecht, Department of Psychiatry, Utrecht, the Netherlands.

# shared senior authorship Under review Abstract Many physiological signals in healthy systems show fractal fluctuations characterized by similar temporal structures at different time scales. The loss of the fractal pattern is associated with pathological conditions. To test whether fractal patterns in motor activity are altered in patients with bipolar disorder, we first analyzed 14-day actigraphy data collected from 106 patients with bipolar disorder type I in euthymic state, 73 unaffected siblings, and 76 controls. Compared to controls, bipolar patients showed altered fractal patterns with excessive regularity in fluctuations at small time-scales

(<90 min), quantified by a larger scaling exponent α( 1). The value of α1>1 indicates a more rigid motor control system. In the siblings, α1 lay between that of patients and controls. Further examinations revealed that group differences in α1 were only observed in females (not in males). Sex also affected the group differences in fractal patterns at larger time scales, i.e., female patients and siblings had more random activity fluctuations at >2h as quantified by a smaller scaling exponentα ( 2<1) as compared to female controls; and male patients showed an increased α2 as compared to male controls. Second, to examine the link between altered fractal patterns and symptoms, we analyzed 180-day synchronized actigraphy and mood symptom data of 14 bipolar patients. Interestingly, the depression score during one week was associated with a lower

α1 during the subsequent week. Our results show sex- and scale-dependent alterations in fractal activity regulation in patients with bipolar disorder. The mechanisms underlying the alterations are yet to be determined.

95 Introduction Bipolar disorder is one of the leading causes of disability worldwide (1). Patients with bipolar disorder experience characteristic episodes of mood symptoms, ranging from depressive episodes to manic episodes (2). With a delay of 5-10 years between illness onset and diagnosis, finding diagnostic markers to allow earlier therapeutic interventions is one of the key challenges in bipolar disorder (3,4). Ideally, these diagnostic markers are non-invasive, reliable and cost-efficient.

Many physiological signals, such as motor activity, show complex fluctuations with similar temporal structures at different time scales (5). These patterns, called fractal fluctuations, are robust in healthy human biology, but are altered under physiological disturbances and in diseases (6). For instance, fractal patterns are perturbed with ageing and in Alzheimer’s disease (7,8). It is believed that fractal patterns represent the integrity and adaptability of biological systems, i.e., maintaining the internal stability while being able to respond to changes from external influences. Thus, fractal patterns are accepted as a biomarker for healthy biology.

Recent studies have provided evidence for the role of the circadian control system in fractal regulation. An example is the breakdown of fractal patterns in rodent motor activity following the lesioning of the suprachiasmatic nucleus (SCN)—the master circadian clock located in the hypothalamus responsible for the coordination of circadian rhythms in various physiological processes (9,10). Importantly, fractal activity patterns appear to be more sensitive to the neuronal changes in the SCN as compared to the traditional circadian measures estimated by daily rhythms (7). As the link between the dysfunctioning circadian timing system and bipolar disorder has been shown, fractal patterns may serve as a promising route for a diagnostic biomarker of bipolar disorder (11). In the current study, we test the hypothesis that bipolar patients have altered fractal regulation compared to healthy controls, and that unaffected siblings of the bipolar patients share a milder alteration. Furthermore, to study whether this is a state or trait effect, we examined the relationship between fractal regulation and daily mood in a small group of patients followed over 180 days.

Methods Participants Two different samples were used for the analyses.

Cross-sectional sample For the analysis in the euthymic sample, participants were included from the Dutch Bipolar Cohort (DBC) study, a collaboration between the University Medical Center Utrecht, University Medical Center Groningen, various other mental health care providers in the Netherlands and the University of California, Los Angeles. The DBC study is developed to investigate genetic and (endo)phenotypic vulnerability factors for bipolar

96 Chapter 8: Fractal biomarker of activity in patients with bipolar disorder

disorder. The medical ethical committees of the three University Medical centers approved the study and their follow-up studies and the studies were in accordance to the Declaration of Helsinki. The DBC study consisted of a baseline study and several follow-up studies, including an actigraphy study to investigate the circadian rhythm and sleep disturbances in patients with bipolar disorder. All participants had a minimum age of 18, and did not suffer from major self-reported somatic illness or pregnancy. Inclusion criteria for patients was a bipolar type I diagnosis, which was verified using the Structural Clinical Interview for DSM-IV (SCID-I) (12). For further selection procedures and dropout rates, see Verkooijen et al.(13) Participants with valid actigraphy and sleep diary data for at least 8 days were included in the analysis. Non-wear periods, that were documented in the sleep diary, were excluded from the analysis. For the analysis data from 106 patients, 73 unaffected siblings and 76 control participants were included. Medication use of the patients was checked and mood stabilizer (lithium, carbamazepine, lamotrigine or valproate acid) was noted. Longitudinal sample Participants for the longitudinal sample were from the Sleep-Wake patterns In The CHange of mood in Bipolar Disorder (SWITCH-BD) study. This protocol was designed to study the temporal relation between sleep, circadian rhythm and mood changes in a naturalistic setting. Patients were recruited from the outpatient clinic of the University Center for Psychi- atry of the University Medical Center Groningen and through a newsletter in the patient society for manic depressive illness. Patients came to the hospital 5 times, the first time for a baseline interview with the Mini-International Neuropsychiatric Interview (MINI) (14) and to receive instructions about the actiwatch and the online diaries. The other times were to read out the actiwatch, change the battery and to make sure everything was clear for the participants. Patients were followed for 180 days with actigraphy, a sleep diary filled out in the morning and a mood diary filled out in the evening, consisting of question on affect, agitation and energy on a visual analog scale. Furthermore patients received two validated questionnaires every week to fill out to assess mood status: the Inventory for Depressive Symptomatology (Self-Report) (IDS-SR) (15) and the Altman Self-Rating Scale for Mania (ASRM) (16). Ten patients were included over a winter period (starting in September/Octo- ber) and 5 patients were included over a summer period (starting in April-June). One patient dropped out due to the large time burden of the study. To study differences between a mood episode period and a euthymic period, mood episodes were selected. Mood episodes were assessed using the validated weekly questionnaires combined with the Lifechart method (LC-self) (15–17). For a manic episode, a patient had to score above 5 on the ASRM on two consecutive weeks, making sure there was at least one full week of manic symptoms (16). Furthermore, patients had to report their daily mood above the midline on the life chart for at least 75% of the time. For a depressed episode, a patient had to score above 26 on the IDS-SR on at least 3 consecutive weeks, resulting in at least 2 consecutive weeks of symptoms. This is one week longer than the strict DSM-IV diagnosis (2 weeks for a depression and 1 week for a mania), to be sure at least two (or one for manic) full weeks the symptoms were present. They also had to report their daily mood below the midline on the life chart for at least 75% of the time. Stable euthymic episodes were defined as 5 consecutive weeks of no manic or depressive symptoms and a lower variability on the life chart compared to the full 180 days, as seen by the standard deviation. Episode according to these criteria were selected by two independent raters and any discrepancies were discussed.

97 Chapter 8: Fractal biomarker of activity in patients with bipolar disorder

Data acquisition Motor activity data were collected using triaxial actigraphy, using Actiwatch 2 (Philips Respironics) for the euthymic sample and the Motionwatch 8 (CamNTech) for the longitudinal sample. Activity counts were stored every minute. Sleep data was collected by use of a sleep diary. Fractal analysis As only data of periods of wakefulness were used for the fractal analysis all sleep data was excluded. This was done with a rigorous method, as not all diaries were filled in complete- ly. For every participant, both in the cross-sectional and longitudinal study, the mean of the bedtime and the get-up time was calculated. For the bedtime, subject specifically, two times the standard deviation of the mean was subtracted to minimize the inclusion of any sleep in the analysis. Similarly, for the get-up time two times the standard deviation of the mean was added to the mean got-up time to minimize the inclusion of sleep.

To assess fractal patterns in motor activity, we performed detrended fluctuation analysis (DFA) to examine the temporal correlations of activity fluctuations at multiple time scales (18,19). This method involves the following four steps: i) integrating the activity counts recording after removing the global mean, i.e., cumulative summation from the first epoch all the way to the end; ii) dividing the integrated signal into non-overlapping windows of the same window size n (i.e., time scale); iii) removing the trend estimated using polynomial functions in the integrated signal within each window to obtain residuals; and iv) calculating the root mean square of residuals from all windows which was named as the fluctuation amplitude F(n). The last three steps are repeated for range of different time-scales.

To reliably estimate F(n) at a specific time scale n, at least 6 windows of size n without gaps are required. Otherwise, the iteration stops and F(n) will not be calculated at that time scale and larger. The 2nd order polynomial functions were used to extract the trend within each window (19). A power-law form of F(n), i.e., F(n)~nα, indicates a fractal structure in the fluctuations. The parameterα , called the scaling exponent, quantifies the temporal correlation as follows: if α = 0.5, there is no correlation in the fluctuations (“white noise”); if α > 0.5, there are positive correlations, where large values are more likely to be followed by large values (and vice versa for small values). If α is greater than 1 and becomes closer to 1.5, it indicates that the control system becomes more rigid or excessive regular. Note that α=1 indicates the most complex fluctuation patterns (i.e., not too regular while not being random). The α values that are close to 1.0 have been observed in many physiological outputs under healthy conditions (5,20–22).

Previous human studies showed that degraded fractal patterns, as occurred with aging and in dementia, lead to distinguished behaviors of F(n) over two distinct time scale regions with the boundary at ~1.5-2 hours (i.e., differentα values) (7–9). We thus calculated the α value for each participant at two non-overlapping time scale regions, i.e., α1 at 1.5-90 min and α2 at 2 and up to 10 hours with the transition region from 90 min to 2 hours excluded.

98 To ensure good signal quality, all actigraphic recordings were checked with the assis- tance of a self-designed MATLAB GUI program (Ver. R2016b, the MathWorks Inc., Natick, MA, USA). The most common types of quality issues were: i) isolated spikes with an amplitude 10 standard deviations (SD) beyond the individual global mean level; and ii) sequences of zeros with duration > 60 minutes during the daytime (likely occurred when participants took the device off). The episodes with those issues were marked as gaps (5,19) and were skipped when performing the DFA in order to avoid any potential effects of interpolating missing data and/or manipulating the signal (e.g., stitching the rest data after removing the missing data) on F(n) (7). Because of gaps, it is possible that the maximum time scale cannot reach 10 hours. To assure a reasonable fit, no α2 value was calculated if the maximum time scale was smaller than 6 hours. Furthermore, we excluded α (usually α2) values for which the goodness of the log-log fit between time scale n and F(n) was smaller than 0.8.

Statistical analyses Group differences in continuous variables were tested using analysis of variance (ANO- VA) with Bonferroni-corrected post hoc tests. Group differences in categorical variables were tested using chi-square tests. To determine group differences in scaling exponents, analysis of variance (ANOVA) and post-hoc Student T-tests were used, with post-hoc tests to establish final group differences. State differences in the longitudinal study were tested using multiple regressions with participant as a random effect to test within-subject differences. To assess the lag relation between mood ratings and scaling components, linear regression was used using the scores of previous days as the dependent variable.

Results Cross-sectional sample The characteristics, mood symptoms, and medication use of the 255 participants in the cross-sectional analysis are provided in table 1. Age differed between the groups, with the siblings being somewhat older than the patients and controls (post-hoc analysis, Stu- dent’s t-test, sibling vs controls, t(1,252) = 3.38, p < 0.001, sibling vs patient, t(1,252) = 2.00, p = 0.046). Furthermore, scores on the manic and depressive scales were higher and psychotropic medication use more frequent in patients compared to the other two groups.

Fractal patterns in cross-sectional group

Patients with bipolar disorder showed a higher α1 (β = 0.009, p = 0.028, table 2) in an unadjusted model compared to healthy controls (figure 1). When the model was adjusted for age and sex, patients still showed a higher α1 (β = 0.009, p = 0.04, table 2). Post-hoc tests showed that in the adjusted model the difference was significant in female patients (t(1,246) = 3.26, p = 0.001), but not in male patients (figure 2). In the unadjusted and adjusted model, patients and siblings do not show a difference compared to the control group in α2 although there was a significant sex interaction effect. When the sample is split in males and females, it is shown that male patients have a

99 higher α2 compared to controls (post-hoc Student’s t-test, t(1,102) = 2.22, p = 0.028) and in females both patients (post-hoc Student’s t-test, t(1,141) = -2.78, p = 0.006) and siblings (t(1,141) = -2.66, p = 0.009) show a lower α2 compared to controls (figure 2).

To test if the difference in α1 is caused by depressive or manic symptoms in the patient group, depressive and manic symptoms as measured with the IDS-SR and ASRM were added to the model independently, and the scores had no significant effect on α1 or α2 and showed no different results.

Post-hoc sex difference explorations To understand the unexpected sex effect within the patient group, post-hoc tests were conducted to explore differences between men and women in the patient group. There were no differences in medication use between men and women, and no association between medication use and either α1 or α2. For daily rhythm variables, of which the methods are described in another paper, male patients showed a lower interdaily stability compared to female patients. This is unlikely an explanation for the larger α2 in male patients, as a lower α2 is linked to an increase in circadian disturbances, which would mean a decrease in stability. Disease characteristics between male and female patients were studied as well. Although male patients had a later age of onset, this was not related to α2.

Fractal patterns in the longitudinal group For the longitudinal analysis, 14 patients with bipolar disorder type 1 were included, 11 women and 3 men, mean age ± SD 44.7 years ± 10.7. No differences in the scaling components assessed on a weekly basis were found between stable, depressive or manic states. There was no relation between the scaling components (both on time scales < 2 hours and > 2 hours) and the weekly mood scores (ASRM and IDS-SR) within the same week. However, when a lag effect was tested, a significant negative correlation was found between the IDS score during one week and α1 during the next week, indicating a higher IDS-SR score resulted in a lower α1 during the next week (mixed model, β = -0.0007, p = 0.047). To understand the finer-grained temporal dynamics of this relationship better, the day-to-day scores were analyzed and a relation was found between the depression score (measured by the visual analog scale from the question

“I am feeling down”) and α1 computed on a daily basis with a lag of 5 and 7 days, indicating that there is approximately a one week delay between changes in depressive symptoms and the change in α1 (figure 3). Discussion This is, to our knowledge, the first study showing altered fractal activity patterns in patients with bipolar disorder. The effect of the disease was pronounced at lower time scales, i.e., more excessive regular activity patterns at< 1.5 hours (larger α1, >1) in patients with bipolar disorder. Interestingly, this effect was sex-dependent, where female bipolar patients showed a higher α1 compared to female controls while this was not the case in men. The sex difference was more distinct for fractal activity patterns at larger

100 Chapter 8: Fractal biomarker of activity in patients with bipolar disorder

time scales (> 2 hours, α2). Specifically, we show that female patients show a lowerα 2 (<1) compared to controls, while male patients show a larger α2. The female unaffected siblings also showed similar smaller α2 as female patients, which was not expected and requires further studies to investigate the underlying mechanism. One possibility is that there is a genetic effect on α2 in people affected by (patients), or vulnerable to (i.e., siblings), bipolar disorder. As both α2 and bipolar disorder have been associated with disturbances in circadian regulation, and α2 specifically with lesioning of the suprachiasmatic nucleus, we hypothesize both bipolar disorder and fractal regulation have a shared underlying vulnerability within the circadian regulation systems (9,10,23). As no differences were found between a stable period and a mood episode in the longitudinal study, we confirm that a fractal pattern might be a trait feature, present independent of major mood episodes. When within analyses are conducted to see the week-to-week fluctuations, we did find a negative lag effect between depression score and the scaling component on a later time point. We confirmed this finding using a higher time resolution, where we look at day-to-day fluctuations of mood and α1, showing that the depressed mood is associated with a lower α1 5 and 7 days later. Note that we also used a different method to quantify depressed mood, showing this is a robust finding. This negative correlation appears to contradict our observation of larger α1 in bipolar patients and a previous observation of larger α1 in patients with major depression by another group (24). One implication of this finding may be that the group effect is not immediately translated to within-individual patient variations across time. A hypothesis for the negative lag correlation between the depressed symptoms and α1 is that the patients in the longitudinal sample might respond to the depressive symptoms with coping behavior, activating themselves and implement- ing learned techniques from psychotherapy. This is possible because the patients in this sample are familiar with their disease status and they had the motivation to participate in such a long-term study. This selection bias is likely in intensive studies like these (25).

The difference between men and women in the scaling components, especially in patients, has never been shown before and has not been systematically studied. Additional analysis provided no support for differences in medication use, interdaily stability, or disease characteristics as explanation for the observed sex differences. This link between α and sex in patients with bipolar disorder should be further explored. The sex difference in α2 is interesting, because of the critical role of the central circadian clock in α2 and because sex differences in fundamental properties of the human circadian system and in human clock gene expression have been demonstrated (26,27). Furthermore, the specific relation with α1 and the patient group is something which needs to be replicated before drawing strong conclusions.

Fractal regulation is believed to reflect system integrity and adaptability (i.e., the ability to respond to external changes while maintaining certain stability for orchestrated internal physiological functions). The balance between regularity and flexibility can be estimated by the scaling exponent (i.e., α) derived from the detrended fluctuation analysis (6,28). When α is larger than 1 and increases toward 1.5 as observed in female patients, the fluctuations become overly regular, suggesting that the system might be less responsive to external changes. This loss of responsiveness has previously been shown in depressed patients in network structures of depressive symptoms, where the resilience

101 Chapter 8: Fractal biomarker of activity in patients with bipolar disorder

of patients to counter an external influence to their own symptom network structure is decreased (29,30). This rigidity, or loss of responsiveness, in both the fractal patterns and network structures shows this might be a core characteristic within mood disorders. Fu- ture studies should aim to link the responsiveness in both symptom networks and fractal motor control to see if they might be a suitable biomarker of mood disorders.

The current study has some limitations. In the cross-sectional study, although we took medication effects into account, controlling for all different medications was impossible to do, as patients use all different types of medications, with different mechanisms of action. The longitudinal part is limited by the low number of patients and there might be a selection bias of well-functioning patients with bipolar disorder, as they were all able to complete an intense and long protocol (25). Furthermore, there were mainly depressive symptoms in this group, and the low prevalence of manic symptoms in this group leaves the question on how manic symptoms and fractal patterns are related partially unanswered.

In summary, we showed in two separate studies, looking cross-sectional as well as longitudinal patterns, that there is a relationship between fractal regulation of motor activity and bipolar disorder. We showed that fractal activity patterns at lower time scales (<~2h) are less complex in bipolar patients, supporting the possibility that fractal measures may serve as a possible biomarker for bipolar disorder. Note that the patterns appear to be a stable trait feature, as no differences were found in mood episodes. Fractal activity patterns on larger timescales might be regulated with a heritable component as both female patients and unaffected siblings show a lower α2 compared to controls. Future work should focus on replicating these findings in other samples, ideally with a larger sample size. The temporal relation between depressive symptoms and the scaling exponent should be studied in a larger sample and if possible in a less well-functioning sample. These results indicate that, when replicated, fractal patterns are an interesting, cost-effective and non-invasive biomarker.

Table 1. Cross-sectional sample characteristics. WĂƚŝĞŶƚ ^ŝďůŝŶŐ ŽŶƚƌŽů Ɖ EсϭϬϲ Eсϳϯ Eсϳϲ ^Ğǆ;ĨĞŵĂůĞ͕йͿ ϲϮ;ϱϵйͿ ϰϰ;ϲϬйͿ ϰϭ;ϱϰйͿ Ϭ͘ϳϭϴ

ŐĞ͕ǇĞĂƌƐ;ƐĚͿ ϱϬ͘ϯ;ϭϭ͘ϯͿ ϱϰ͘ϯ;ϭϮͿ ϰϳ;ϭϲ͘ϯͿ Ϭ͘ϬϬϰ

ŵƉůŽǇŵĞŶƚ;ǇĞƐ͕йͿ ϳϬ;ϲϳйͿ ϱϮ;ϳϰйͿ ϰϯ;ϱϴйͿ Ϭ͘ϭϭϴ

ŚŝůĚƌĞŶ;ǇĞƐ͕йͿ Ϯϱ;ϮϰйͿ ϯϬ;ϰϭйͿ Ϯϭ;ϮϴйͿ Ϭ͘Ϭϯϳ

^ZDƐĐŽƌĞ;ƐĚͿ ϭ͘ϵ;ϭ͘ϴͿ ϭ͘Ϯ;ϭ͘ϰͿ ϭ͘ϲ;Ϯ͘ϮͿ Ϭ͘Ϭϰϲ

/^Ͳ^ZƐĐŽƌĞ;ƐĚͿ ϭϱ;ϭϬ͘ϴͿ ϳ;ϲ͘ϲͿ ϱ͘ϵ;ϰ͘ϵͿ фϬ͘ϬϬϭ

DŽŽĚƐƚĂďŝůŝǌĞƌ;ǇĞƐ͕йͿ ϳϱ;ϳϭйͿ ϭ;ϭйͿ ϭ;ϭйͿ фϬ͘ϬϬϭ

ŝƌĐĂĚŝĂŶŵĞĚŝĐĂƚŝŽŶ;ǇĞƐ͕йͿ ϳϯ;ϲϵйͿ Ϯ;ϯйͿ ϭ;ϭйͿ фϬ͘ϬϬϭ

102 Table 2. Models for bipolar genetics analysis – α1 and α2

ϭ Ϯ ĂƐŝĐ ĚũƵƐƚĞĚ ĂƐŝĐŵŽĚĞů ĚũƵƐƚĞĚ ŵŽĚĞů ŵŽĚĞů ŵŽĚĞů ƐƚŝŵĂƚĞ Ɖ ƐƚŝŵĂƚĞ Ɖ ƐƚŝŵĂƚĞ Ɖ ƐƚŝŵĂƚĞ Ɖ /ŶƚĞƌĐĞƉƚ ϭ͘ϬϮϭϮϰ Ϭ͘ϵϲϴϯϳ Ϭ͘ϴϬϯϳϴ Ϭ͘ϳϵϵϬϯ 'ƌŽƵƉ͗ĐŽŶƚƌŽů ZĞĨ ZĞĨ ZĞĨ ZĞĨ 'ƌŽƵƉ͗ƉĂƚŝĞŶƚ Ϭ͘ϬϬϵϰ Ϭ͘ϬϮϳϳ Ϭ͘ϬϬϴϲ Ϭ͘Ϭϯϵϱ Ϭ͘ϬϬϭϴϵϵ Ϭ͘ϴϲϭϱ Ϭ͘ϬϬϳϵϬϲ Ϭ͘ϰϲϬϲ 'ƌŽƵƉ͗ƐŝďůŝŶŐ ͲϬ͘ϬϬϭϮϳϮ Ϭ͘ϳϴϱ ͲϬ͘ϬϬϱϲϯ Ϭ͘ϮϮϴϰ ͲϬ͘ϬϭϭϲϮ Ϭ͘ϯϯϬϭ ͲϬ͘Ϭϭϯϯ Ϭ͘Ϯϲϱϵ ^Ğǆ;ŵĂůĞͿ ͲϬ͘ϬϬϵϵϴ Ϭ͘ϬϬϭϲ Ϭ͘ϬϮϴϯ Ϭ͘ϬϬϬϱ ŐĞ Ϭ͘ϬϬϭϬϭ фϬ͘ϬϬϬϭ Ϭ͘ϬϬϬϭϱ Ϭ͘ϴϬϳϳ 'ƌŽƵƉ͗WĂƚŝĞŶƚΎĂŐĞ ͲϬ͘ϬϬϬϮϯϱ Ϭ͘ϰϵϰϱ ͲϬ͘ϬϬϭϱϳ Ϭ͘ϬϳϳϮ 'ƌŽƵƉ͗^ŝďůŝŶŐΎĂŐĞ Ϭ͘ϬϬϬϭϭϯ Ϭ͘ϳϳϱϴ Ϭ͘ϬϬϭϴϳ Ϭ͘Ϭϰϳϳ

'ƌŽƵƉ͗WĂƚŝĞŶƚΎƐĞǆ ͲϬ͘ϬϬϱϲϮ Ϭ͘ϭϴϮ Ϭ͘ϬϮϱϮϬ Ϭ͘ϬϮϬϯ 'ƌŽƵƉ͗^ŝďůŝŶŐΎƐĞǆ Ϭ͘ϬϬϭϲϬ Ϭ͘ϳϮϱϱ Ϭ͘Ϭϭϰϳϵ Ϭ͘ϮϬϵϱ

Figure 1. Left panels show 14 day actigraphy recording of an example control participant (A) and an example patient participant (B). The right panel shows the scale invariance of these participants (C). The scaling component is the coefficient of the regression line. The line before the dotted vertical line isα 1 as it is from lower timescales, while the regression line on the right from the dotted line is α2.

103 A: all subjects B: women C: men * * 1.04 1.04 1.04 ..1 ..1 ..1 1.02 1.02 1.02

1.00 1.00 1.00

Controle Sibling Patient Controle Sibling Patient Controle Sibling Patient

D: all subjects E: women F: men 0.90 0.90 * 0.90 * 0.85 0.85 * 0.85 ..2 ..2 ..2 0.80 0.80 0.80

0.75 0.75 0.75 Controle Sibling Patient Controle Sibling Patient Controle Sibling Patient

Figure 2. Scaling components differences between groups and between men and women.

sig no yes

0.00000

−0.00005 AS score)

−0.00010

−0.00015 Estimate from lag analysis (in V

−0.00020

** −6 −4 −2 0 Lag (in days)

Figure 3. Lag effect of “I’m feeling down” onα 1. Note that the effect is significant on lag -5 and -7. As the “I’m feeling down” question is answered on a visual analog scale with a range from 0 to 100 the estimates are very small.

104 Chapter 8: Fractal biomarker of activity in patients with bipolar disorder

A, Abramovic L, Pagani L, et al. An actigraphy References study investigating sleep in bipolar I patients, 1. Murray CJL, Vos T, Lozano R, Naghavi M, Flaxman unaffected siblings and controls. J Affect Disord. AD, Michaud C, et al. Disability-adjusted life 2017;208(April 2016):248–54. years (DALYs) for 291 diseases and injuries in 14. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, 21 regions, 1990-2010: A systematic analysis Janavs J, Weiller E, et al. The Mini-International for the Global Burden of Disease Study 2010. Neuropsychiatric Interview (MINI): the develop- Lancet. 2012;380(9859):2197–223. ment and validation of a structured diagnostic 2. Association AP. Diagnostic and statistical manual psychiatric interview for DSM-IV and ICD-10. J of mental disorders (DSM-5®). American Psychi- Clin Psychiatry. 1998;59:22–33. atric Pub; 2013. 15. Rush AJ, Carmody T, Reimitz P. The Inventory 3. Baldessarini RJ, Tondo L, Baethge CJ, Lepri of Depressive Symptomatology (IDS): Clinician B, Bratti IM. Effects of treatment latency on (IDS‐C) and Self‐Report (IDS‐SR) ratings of de- response to maintenance treatment in man- pressive symptoms. Int J Methods Psychiatr Res. ic-depressive disorders. Bipolar Disord. 2007 2000;9(2):45–59. Jun;9(4):386–93. 16. Altman EG, Hedeker D, Peterson JL, Davis JM. The 4. Phillips ML, Kupfer DJ. Bipolar Disorder 2 Bipolar Altman Self-Rating Mania Scale. 1997;(1991). disorder diagnosis : challenges and future direc- 17. Born C, Amann BL, Grunze H, Post RM, Schärer tions. Lancet. 2013;381(9878):1663–71. DOI: L. Saving time and money: a validation of the 10.1016/S0140-6736(13)60989-7 self ratings on the prospective NIMH Life-Chart 5. Hu K, Ivanov PC, Chen Z, Hilton MF, Stanley HE, Method (NIMH-LCM). BMC Psychiatry. 2014 Shea SA. Non-random fluctuations and multi- Jan;14:130. DOI: 10.1186/1471-244X-14-130 scale dynamics regulation of human activity. Phys 18. Peng CK, Buldyrev S V., Havlin S, Simons M, A Stat Mech its Appl. 2004;337(1–2):307–18. Stanley HE, Goldberger AL. Mosaic organiza- 6. Goldberger AL, Amaral LAN, Hausdorff JM, Ivanov tion of DNA nucleotides. Phys Rev E Stat Phys PC, Peng C, Stanley HE. Fractal dynamics in Plasmas Fluids Relat Interdiscip Topics. 1994 physiology : Alterations with disease and aging. Feb;49(2):1685–9. 2002;99. 19. Hu K, Ivanov PC, Chen Z, Carpena P, Stanley HE. 7. Hu K, Someren EJW Van, Shea SA, Scheer FAJL. Effect of trends on detrended fluctuation analy- Reduction of scale invariance of activity fluc- sis. Phys Rev E. 2001 Jul;64(1):11114. tuations with aging and Alzheimer ’ s disease : 20. Peng CK, Havlin S, Hausdorff JM, Mietus JE, Stan- Involvement of the circadian pacemaker. ley HE, Goldberger AL. Fractal mechanisms and 2009;106(8). heart rate dynamics. Long-range correlations 8. Hu K, Riemersma-van der Lek RF, Patxot M, and their breakdown with disease. J Electrocar- Li P, Shea SA, Scheer FA, et al. Progression of diol. 1995;28 Suppl:59–65. Dementia Assessed by Temporal Correlations of 21. Peng CK, Mietus JE, Liu Y, Lee C, Hausdorff JM, Physical Activity: Results From a 3.5-Year, Lon- Stanley HE, et al. Quantifying fractal dynamics of gitudinal Randomized Controlled Trial. Sci Rep. human respiration: age and gender effects. Ann 2016;6(May):27742. DOI: 10.1038/srep27742 Biomed Eng. 2002 May;30(5):683–92. 9. Hu K, Harper DG, Shea SA, Stopa EG, Scheer 22. Hausdorff JM, Mitchell SL, Firtion R, Peng CK, FAJL. Noninvasive fractal biomarker of clock Cudkowicz ME, Wei JY, et al. Altered fractal neurotransmitter disturbance in humans with dynamics of gait: reduced stride-interval cor- dementia. Sci Rep. 2013;3:1–7. relations with aging and Huntington’s disease. 10. Hu K, Scheer FAJL, Ivanov PC, Buijs RM, Shea SA. J Appl Physiol (Bethesda, Md 1985). 1997 The suprachiasmatic nucleus functions beyond Jan;82(1):262–9. circadian rhythm generation. Neuroscience. 23. Bradley AJ, Webb-Mitchell R, Hazu A, Slater 2007;149(3):508–17. N, Middleton B, Gallagher P, et al. Sleep and 11. Harvey AG. Sleep and circadian rhythms in circadian rhythm disturbance in bipolar disor- bipolar disorder: seeking synchrony, har- der. Psychol Med. 2017;1–12. DOI: 10.1017/ mony, and regulation. Am J Psychiatry. S0033291717000186 2008 Jul;165(7):820–9. DOI: 10.1176/appi. 24. Aybek S, Ionescu A, Berney A, Chocron O, Amin- ajp.2008.08010098 ian K, Vingerhoets FJG. Fractal temporal organ- 12. First MB, Gibbon M. The Structured Clinical Inter- isation of motricity is altered in major depres- view for DSM-IV Axis I Disorders (SCID-I) and the sion. Psychiatry Res. 2012;200(2–3):288–93. Structured Clinical Interview for DSM-IV Axis II 25. Bos FM, Schoevers R a., aan het Rot M. Expe- Disorders (SCID-II). 2004; rience sampling and ecological momentary 13. Verkooijen S, van Bergen AH, Knapen SE, Vreeker assessment studies in psychopharmacology: A systematic review. Eur Neuropsychophar-

105 macol. 2015;1–12. DOI: 10.1016/j.euroneuro.2015.08.008 26. Lim ASP, Myers AJ, Yu L, Buchman AS, Duffy JF, De Jager PL, et al. Sex difference in daily rhythms of clock gene expression in the aged human cerebral cortex. J Biol Rhythms. 2013;28(2):117–29. DOI: 10.1177/0748730413478552 27. Duffy JF, Cain SW, Chang A-M, Phillips AJK, Munch MY, Gronfier C, et al. Sex difference in the near-24-hour intrinsic period of the human circadian timing system. Proc Natl Acad Sci. 2011;108(Supplement_3):15602–8. DOI: 10.1073/pnas.1010666108 28. Pittman‐Polletta BR, Scheer FAJL, Butler MP, Shea SA, Hu K. The role of the circadian system in fractal neurophysiological control. Biol Rev. 2013;88(4):873–94. 29. Van Borkulo C, Boschloo L, Borsboom D, Penninx BWJH, Lourens JW, Schoevers RA. Association of symptom network structure with the course of longitudinal depression. JAMA Psychiatry. 2015;72(12):1219–26. 30. Wichers M. The dynamic nature of depression: a new micro-level perspective of mental disorder that meets current challenges. Psychol Med. 2014 May;44(7):1349–60. DOI: 10.1017/ S0033291713001979

106 Chapter 9 The duration of light treatment and therapy outcome in seasonal affective disorder

S.E. Knapen1,2, M. van de Werken1, M.C.M. Gordijn1,3, Y. Meesters2

1. University of Groningen, Department of Chronobiology, Centre for Life Sciences, Groningen, The Netherlands 2. University of Groningen, University Medical Center Groningen, Department of Psychiatry, University Medical Center Groningen, Groningen, The Netherlands 3. Chrono@Work B.V. Groningen, The Netherlands

Journal of Affective Disorders (2014);166, 343-346. Chapter 9: The duration of light treatment and therapy outcome in seasonal affective disorder

Abstract Background Seasonal affective disorder (SAD) is characterized by recurrent episodes of major depression with a seasonal pattern, treated with light therapy (LT). Duration of light therapy differs. This study investigates retrospectively whether a single week of LT is as effective as two weeks, whether males and females respond differently, and whether there is an effect of expectations as assessed before treatment.

Methods 83 women, and 25 men received either one-week (n=42) or two weeks (n=68) of LT were included in three studies. Before LT, patients’ expectations on therapy response were assessed.

Results Depression severity was similar in both groups before treatment (F(1,106)=0.19, ns) and decreased significantly during treatment (main effect “time” F(2,105)=176.7, p<0.001). The speed of therapy response differs significantly in treatment duration, in favor of 1 week (F(2,105) = 3.2, P = 0.046). A significant positive correlation between expectations and therapy response was found in women (ρ=0.243, p=0.027) and not in men (ρ=0.154, ns). When expectation was added as a covariate in the repeated-measures analysis it shows a positive effect of the level of expectation on the speed of therapy response (F(2,104) = 4.1, p = 0.018).

Conclusions There is no difference between 1 and 2 weeks of LT in overall therapy outcome, but the speed of therapy response differed between 1 week LT and 2 weeks LT. Together with the significant correlation between expectations and therapy response in women, we hypothesize that expectations play a role in the speed of therapy response.

108 Chapter 9: The duration of light treatment and therapy outcome in seasonal affective disorder

Introduction Seasonal affective disorder (SAD) is a mood disorder characterized by recurrent episodes of major depression with a seasonal pattern (1). SAD has a prevalence of 2 to 10% in Europe and North America (2). Light therapy (LT) is the treatment of first choice for winter type SAD in the Netherlands (3). The effectiveness of LT is well established; response rates are high with minor adverse events (4,5). However, there is no consensus on the duration of treatment required to be effective; treatment duration ranges from 3 days to 8 weeks (5–8). Levitt and Levitan indicate that a shorter duration of LT (2 weeks) can be as effective as a longer duration (5 weeks), suggesting a faster response rate in the group receiving shorter LT duration (9). Prior to the observed faster response, the expectations of the two patient groups regarding the speed of the response might have differed and this difference might have played a role in the faster response rate in the group that received 2 weeks of LT. This fits with previous findings that a positive expectation about response rate at the start of a therapy is related to therapy outcome (10). Since there are indications outside the field of light treatment that expectations may differ between men and women, we included sex as an independent parameter into our analysis (11).

In a database of studies with either 1 week or 2 weeks of light therapy we retrospectively analysed the relationship between expectations of patients on therapy response with therapy response itself and the relationship with treatment duration and also to sex differences in expectations related to outcome.

Material and Methods Study design and participants For the current analysis we combined data obtained from three different studies, performed over a time span of 7 years (2005-2012). The studies were all performed in the SAD outpatient clinic of the University Medical Center Groningen (UMCG), The Nether- lands. In two studies patients were treated with 2 weeks of light therapy (LT), in one study patients were treated with 1 week of LT. Patients received LT on five workdays each week. The choices for either 2 weeks or 1 week of treatment were made prior to the start of the separate studies, hence the choice between one week or two weeks of LT was made based on the research protocol of that specific study.

The first study compared blue-enriched light (for either 30 or 20 min) to standard full spectrum (30 min) over a period of two weeks (12). The second study compared low-intensity blue-enriched light to standard light treatment over a period of two weeks (13). The third study compared low-intensity narrow band blue light to standard light treatment over a period of one week (14). For specifications of the different light treatments see table 1.

109 Table 1. Characteristics of patients, light treatment and results. * All light conditions except the LED Blue light condition: full spectrum light, without UV. ^ƚƵĚǇ ^ƚƵĚǇϭ ^ƚƵĚǇϮ ^ƚƵĚǇϯ WĂƌƚŝĐŝƉĂŶƚƐ ϱϮ ϭϰ ϰϮ dŚĞƌĂƉǇĚƵƌĂƚŝŽŶ ϮǁĞĞŬƐ ϮǁĞĞŬƐ ϭǁĞĞŬ ^ĞǆŶ;йͿ DĂůĞ ϭϯ;ϮϱͿ ϯ;ϮϭͿ ϵ;ϮϭͿ &ĞŵĂůĞ ϯϵ;ϳϱͿ ϭϭ;ϳϵͿ ϯϯ;ϳϵͿ ŐĞŵĞĂŶ;ц^Ϳ ϯϳ͘ϲ;цϭϭ͘ϰͿ ϯϴ͘ϰ;цϭϮ͘ϲͿ ϯϳ͘ϯ;цϭϯ͘ϭͿ

ĂƐĞůŝŶĞ^/',Ͳ^ƐĐŽƌĞ Ϯϲцϲ Ϯϰцϴ Ϯϱцϱ ;ŵĞĂŶц^Ϳ WƌŽƉŽƌƚŝŽŶĂůƌĞĚƵĐƚŝŽŶ ϲϲ͘ϯцϯϰ͘ϰ ϲϭ͘ϮцϮϴ ϳϬ͘ϮцϮϱ ^/',Ͳ^ƐĐŽƌĞ;ŵĞĂŶц ^Ϳ ǆƉĞĐƚĂƚŝŽŶ;ŵĞĂŶц^Ϳ ϵ͘ϮцϮ͘ϭ ϭϮ͘Ϯцϭ͘ϭ ϭϭ͘ϯцϮ͘ϳ DY;ŵĞĂŶц^Ϳ ϱϮцϭϭ ϱϮцϴ ϱϭцϯ >ŝŐŚƚƐƉĞĐŝĨŝĐĂƚŝŽŶΎ ^ƚĂŶĚĂƌĚ ϱϬϬϬΣ< ϱϬϬϬΣ< ϱϬϬϬΣ< ;ϭϬϬϬϬůƵǆͿ ;ϭϬϬϬϬůƵǆͿ ;ϭϬϬϬϬůƵǆͿ ǆƉĞƌŝŵĞŶƚĂů ϭϳϬϬϬ< ϭϳϬϬϬ< >ůƵĞůŝŐŚƚ ;ϭϬϬϬϬůƵǆͿ ;ϳϱϬůƵǆͿ ϰϳϬŶŵ ;ϭϬϬůƵǆͿ zĞĂƌƐŽĨƐƚƵĚǇ ϮϬϬϱͬϮϬϬϲ ϮϬϬϴͬϮϬϬϵ ϮϬϭϬͬϮϬϭϭ

A total number of 120 patient cases were retrospectively selected based on the following criteria: all subjects met the criteria of major depressive disorder with a seasonal (winter) pattern according to the DSM-IV-TR and did not suffer from other DSM-IV classified psychiatric disorders as assessed by the Mini-International Neuropsychiatric Interview(15,16). Patients who did not fill out all questionnaires were excluded (n=12). The remaining group of 108 subjects consisted of 83 women and 25 men, mean age ± SD 37.6 ± 12 years.

Procedures The Structured Interview Guide for the Hamilton Depression Rating Scale-Seasonal Af- fective Disorder 24 items version (SIGH-SAD) (17) was used to assess severity of depression. SIGH-SAD ratings were obtained prior to the start of LT, immediately after the last LT day, 1 week after the last LT day and for the two-week protocol also halfway the LT period. These studies measured depression score one week after the last LT session as depression score tend to decrease even after the end of treatment.(18) Proportional

110 improvement scores on the SIGH-SAD were calculated for both conditions. In all three studies no significant differences between light conditions were observed (see for more details the relevant papers(12–14)): study 1, main effect “condition” F(2,49) = 0.73, ns; study 2, main effect “condition” F(1,20) = 0.012, ns; study 3, 67% recovery for standard treatment and 63% recovery for experimental treatment, ns. For the current analysis we pooled the data of all three studies and all different light conditions.

At baseline, patients filled out a questionnaire about their expectations, consisting of three questions: whether patients believed they would benefit from the therapy, if they thought it was a suitable treatment and whether they would recommend it to a friend with SAD. Answers were given on a 5-point Likert scale. These questions were asked for both the standard treatment and for the experimental treatment (the minimum score was 3 and the maximum score was 15). Significant differences were found between -ex pectations ratings of the different types of treatment (expectation score ± SD; standard treatment: 10.9 ± 2.2, experimental treatment: 10.1 ± 2.5, p < 0.05). We decided to use the expectation ratings in accordance to the type of light patients received, as we want to link the therapy expectations to the therapy they received.

Statistical analysis The two groups with either one or two weeks of light therapy duration were compared with Chi-square for dichotomous variables ‘group’ and ‘sex’. One-Way ANOVA was used to test for differences in age or baseline depression score between the two groups. SIGH-SAD results were compared with a repeated measures ANOVA. Within-subject factor was the depression severity score on timepoints D1, D8 and D15, between-subjects was ‘group’ (1 week or 2 weeks LT) and covariate was the rating they gave concerning their expectations of the treatment. Final depression scores were calculated by the proportional difference between D1 and the last time point (D15 for 1-week LT and D22 for 2-weeks LT). All correlations were analyzed using Spearman (rank) correlation statistics; expectation scores are correlated to percentage depression score reduction.

Results There were no differences in demographics between the two groups (one-week LT and two-weeks LT) (table 1). Not in sex ratio (f/m 9/33, 16/50, χ2 = 0.114, ns), nor in age (mean ± SD, 37.3±13.1y; 37.7±11.6y, F(1,106) = 0.027, ns), and not in baseline depression score (SIGH-SAD score; F(1,106) = 0.19, ns).There was also no significant difference in the proportional reduction in SIGH-SAD score between the two groups; 70.2%±25.0; 65.2%±33.0, F(1,106) = 0.71, ns.

Speed of therapy response Depression score decreased significantly over time from day 1 to day 15 in both groups (1 and 2 weeks LT) (SIGH-SAD, main effect “time” F(2,105) = 176.7, p < 0.001). Although there is no significant difference in final therapy outcome (mean proportional reduction ± SD, 1 week LT; 70 ± 25, 2 week LT; 65 ± 33, p > 0.05) between the conditions, there

111 Chapter 9: The duration of light treatment and therapy outcome in seasonal affective disorder

is a significant interaction effect of time and group (interaction effect “time*group” F(2,105) = 3.2, p = 0.046). Patients in the 2 week LT group showed a slower decrease in depression score compared to patients in the 1 week LT group (figure 1).

Figure 1. Pattern of depression ratings over time for 1 week and 2 weeks separately.

Figure 2. Differences in expectations and therapy outcome in male subjects and female subjects. To make the graph visually understandable the expectation scores are divided in groups, where 4-6 is lowest and 13-15 is highest.

Role of expectation and sex in therapy response Expectation ratings did not differ significantly between males and females (expectation score ± SD; males, 10 ± 2.8, females, 10.5 ± 2.4, ns). Looking at the relationship between therapy outcome and expectations, no significant correlation was found in males (Spearman correlation, ρ = -0.154, ns), but a significant positive correlation was found in female subjects (ρ = 0.243, p = 0.027) (figure 2).

112 Chapter 9: The duration of light treatment and therapy outcome in seasonal affective disorder

When added to the repeated measures statistics expectations was found to have a significant positive effect (interaction expectation*time F(2,104) = 4.0, p = 0.02) and there still a significant effect of treatment duration was found (interaction group*time F(2,104) = 4.1, p = 0.018).

Adding sex as an interaction variable shows there is no significant effect between treatment duration and sex with expectation as a covariate (interaction group*sex*time F(2,102) = 1.7, p = 0.18).

Discussion This study showed that there is no significant difference in final depression score between SAD patients receiving either one week or two weeks of LT in the setup of our clinical treatment. This could suggest that one week of LT is of sufficient duration for people with seasonal affective disorder. This conclusion is in line with results of earlier studies; studies in our own clinic show that early treatment, consisting of 5 days light treatment is able to prevent relapses for the entire winter (19) and 1-2 weeks of light treatment is recommended by Partonen and Magnusson (20). More studies comparing the effect of light treatment with different durations during different periods of the depressive episode may shed light on the question whether the speed of the effect and recovery differ depending on treatment timing in relation to the duration of the depressive episode.

Although there is no difference in the final depression score between one week of LT and two weeks of LT, there is a difference in the speed of the reduction of the depression score over time. Subjects with one week of LT have a faster decline in depression score compared to subjects with two weeks of LT. Similar to the experiment of Levitt and Levitan treatment duration was known to the subjects prior to the start, and it was told to be effective to treat the symptoms. This knowledge obviously resulted in the same overall expectation of the therapy response prior to the start, but might have induced the difference in the speed of the effect during the treatment. Patients with two weeks of LT expected that another week of LT was necessary, while the group that received one week of LT thought that one week was enough. Although the expectation could account for the slower decrease, other unknown factors may play a role as well.

Further analysis showed that it is only in women that the expectation on therapy response shows a relation with the actual therapy response. If a woman has a higher expectation of the therapy results, the therapy outcome will be better, while this effect is not shown in men. Taking expectations as a covariate in the speed of therapy response shows this has a significant effect. These results are in line with a study by Rutherford et al. in 2012 in which baseline expectation scores correlated with lower final depression severity score in patients with major depressive disorder (21). Outside the field of psychiatry Yee et al. in 2008 examined patient expectations before spinal surgery (22). They showed a significant positive correlation between patients’ expectations and postoperative improvements in the physical domain.

113 Hypothesizing on a causal explanation of the correlation between expectations and therapy response is intriguing. Scott et al. showed in a functional MRI study that the basis of an expectation effect could be in the brain reward system (23). They linked the increased expectations of a monetary reward to increased dopamine release in the nucleus accumbens, a central component of the brain reward system.

Limitations A limitation of this study is that it is a retrospective analysis; we combine data from separate studies to test a hypothesis the studies were not designed for. These studies tested different types of light. Although within each study there were no significant differences found between the light conditions, the data remain confounded by these experiments and probably by other uncontrolled variables as year and weather conditions.

If confirmed in a prospective study, LT for SAD could be of a short duration and should be accompanied by the message that this short treatment duration is highly effective to retrieve the best result.

114 Seasonal Affective Disorder. J Affect Disord. References 2012 Jan;136(1–2):72–80. DOI: 10.1016/j. 1. Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Good- jad.2011.08.016 win FK, Davenport Y, et al. Seasonal affective 13. Meesters Y, Dekker V, Schlangen LJM, Bos disorder: a description of the syndrome and EH, Ruiter MJ. Low-intensity blue-enriched preliminary findings with light therapy. Arch white light (750 lux) and standard bright Gen Psychiatry. 1984;41(1):72. light (10,000 lux) are equally effective in 2. Mersch PPA, Middendorp HM, Bouhuys AL, treating SAD. A randomized controlled study. Beersma DGM, van den Hoofdakker RH. BMC Psychiatry. 2011 Jan;11(1):17. DOI: Seasonal affective disorder and latitude: 10.1186/1471-244X-11-17 a review of the literature. J Affect Disord. 14. Meesters Y, Duijzer WH. The effects of low 1999 Apr;53(1):35–48. DOI: 10.1016/S0165- intensity monochromatic blue light treatment 0327(98)00097-4 compared to standard light treatment in SAD. 3. Spijker J, Bockting C, Meeuwissen J, van Vliet SLTBR Abstr 2352. 2011; I, Emmelkamp P, Hermens M, et al. Multidis- 15. American Psychiatric Association. Diagnostic ciplinaire richtlijn Depressie (Derde revisie). and statistical manual of mental disorders: Richtlijn voor de diagnostiek, behandeling en DSM-IV-TR®. American Psychiatric Pub; 2000. begeleiding van volwassen patiënten met een 16. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim depressieve stoornis. Utrecht: Trimbos-institu- P, Janavs J, Weiller E, et al. The Mini-Interna- ut; 2013. tional Neuropsychiatric Interview (MINI): the 4. Golden RN, Gaynes BN, Ekstrom RD, Hamer development and validation of a structured RM, Jacobsen FM, Suppes T, et al. The efficacy diagnostic psychiatric interview for DSM-IV and of light therapy in the treatment of mood disor- ICD-10. J Clin Psychiatry. 1998;59:22–33. ders: a review and meta-analysis of the evi- 17. Williams JBW, Link MJ, Rosenthal NE, Terman dence. Am J Psychiatry. 2005;162(4):656–62. M. Structured Interview Guide for the Hamilton 5. Lam RW, Levitt AJ, Levitan RD, Enns MW, More- Depression Rating Scale—Seasonal Affective house R, Michalak EE, et al. The Can-SAD study: Disorder Version (SIGH-SAD). New York, New a randomized controlled trial of the effective- York State Psychiatr Inst. 1988; ness of light therapy and fluoxetine in patients 18. Meesters Y. The timing of light therapy and re- with winter seasonal affective disorder. Am sponse assessment in winter depression. Acta J Psychiatry. 2006 May;163(5):805–12. DOI: Neuropsychiatr. 1995;7(2):61–3. 10.1176/appi.ajp.163.5.805 19. Meesters Y, Jansen JH, Beersma DG, Bouhuys a 6. Eastman CI, Young MA, Fogg LF, Liu L, Meaden L, van den Hoofdakker RH. Early light treatment PM. Bright light treatment of winter depres- can prevent an emerging winter depression sion: a placebo-controlled trial. Arch Gen from developing into a full-blown depression. J Psychiatry. 1998;55(10):883. Affect Disord. 1993 Sep;29(1):41–7. 7. Meesters Y, Jansen JH, Beersma DG, Bouhuys 20. Partonen T, Magnusson A. Practice guidelines. a L, van den Hoofdakker RH. An attempt to In: Partonen T, Pandi-Perumal SR, editors. prevent winter depression by light exposure Seasonal Affective Disorders: practice and at the end of September. Biol Psychiatry. 1994 research. Second edi. Oxford University Press; Feb 15;35(4):284–6. 2009. p. 315–20. 8. Terman M, Terman JS. Light therapy for sea- 21. Rutherford BR, Marcus SM, Wang P, Sneed JR, sonal and nonseasonal depression: efficacy, Pelton G, Devanand D, et al. A randomized, protocol, safety, and side effects. CNS Spectr. prospective pilot study of patient expec- 2005 Aug;10(8):647–63. tancy and antidepressant outcome. Psy- 9. Levitt AJ, Levitan R. Length of light treatment chol Med. 2012 Sep 13;1–8. DOI: 10.1017/ trial: does it influence outcome. Chronobiol Int. S0033291712001882 2003;20:1213–4. 22. Yee A, Adjei N, Do J, Ford M, Finkelstein J. Do 10. Eastman CI. What the placebo literature can tell patient expectations of spinal surgery relate us about light therapy for SAD. Psychopharma- to functional outcome? Clin Orthop Relat Res. col Bull. 1990; 2008 May;466(5):1154–61. DOI: 10.1007/ 11. Robinson ME, Riley JL, Myers CD, Papas RK, s11999-008-0194-7 Wise E a, Waxenberg LB, et al. Gender role 23. Scott DJ, Stohler CS, Egnatuk CM, Wang H, expectations of pain: relationship to sex differ- Koeppe R a, Zubieta J-K. Individual differences ences in pain. J Pain Off J Am Pain Soc. 2001 in reward responding explain placebo-in- Oct;2(5):251–7. DOI: 10.1054/jpai.2001.24551 duced expectations and effects. Neuron. 2007 12. Gordijn MCM, ’t Mannetje D, Meesters Y. The Jul 19;55(2):325–36. DOI: 10.1016/j.neu- effects of blue-enriched light treatment ron.2007.06.028 compared to standard light treatment in

115 Chapter 10 The relation between chronotype and treatment outcome with light therapy on a fixed time schedule

Stefan E. Knapen, BSc1, Marijke C.M. Gordijn, PhD2,3, Ybe Meesters, PhD4

1. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Research School of Behavioural and Cognitive Neurosciences (BCN), Interdisciplinary Center for Psychopathology and Emotion regulation (ICPE). Groningen, the Netherlands 2. Department of Chronobiology, GeLifes, University of Groningen, Groningen, The Netherlands 3. Chrono@Work B.V. Groningen, The Netherlands 4. University of Groningen, University Medical Center Groningen, Department of Psychiatry, Groningen, The Netherlands

Journal of Affective Disorders (2016); 202, 87-90. Abstract Background Seasonal affective disorder (SAD) is characterized by recurrent episodes of major depression in a seasonal pattern. The therapy of choice is light therapy (LT). It is suggested that LT should be administered relative to the chronotype of the patient, with the optimal timing earlier for morning than for evening types. This study aims to retrospectively investigate the relation between chronotype and the effect of LT on a fixed time in the morning in a population of SAD patients.

Methods Data from four different studies conducted at the University Center of Psychiatry in Groningen, the Netherlands was used. Data from 132 patients was used (103 women). Depression score was determined by a structured interview (SIGH-SAD) prior to LT and after LT. Prior to LT morningness/eveningness preference of the patient was determined by the ‘Morningness/Eveningness Questionnaire’ (MEQ). All patients received LT at 8:00 AM at the clinic, independent of chronotype.

Results Patients had an average MEQ score of 51.5±8.2. There was no significant relationship between MEQ score and therapy success as measured with the SIGH-SAD (F2,129 = 0.05, ns). When patients were divided by chronotype (ranging from definite morning to moderate evening) no significant relation between MEQ score and therapy success was found (F2,129 = 0.02, ns).

Conclusions The lack of a significant relationship between chronotype, as measured with the MEQ, and therapy success with LT at a fixed timepoint may indicate that the anti-depressive effect of morning light in SAD patients is not explained by a phase shift of the biological clock.

117 Introduction Seasonal affective disorder (SAD) is a mood disorder characterized by recurrent episodes of major depression with a seasonal pattern (1). For winter type SAD, light therapy (LT) is the treatment of choice in the Netherlands (2). Although the effectiveness of LT is well established and the response rates are high the mechanism underlying the effect is still unclear. Since the introduction of SAD and the positive effects of LT in 1984 by Rosenthal and colleagues, various theories have been discussed. One of the most prominent hypotheses explaining the success of light therapy is the phase advancing effect of properly timed morning light (3). SAD patients are suggested to have a delayed circadian phase underlying depressive mood. Morning light is thought to be therapeutic as it causes a corrective phase advance. This hypothesis is later specified to the “phase angle difference” hypothesis, where it is not just the phase delay in SAD patients, but the internal phase delay compared to the mid-point of sleep that is the crucial factor for the therapeutic response (4).

In 2001, Terman et al. showed a correlation between the magnitude of the phase advance with morning light exposure and therapy success in SAD patients (5). Based on this study an optimal timing of light therapy was defined according to an individual’s circadian phase (6). This optimum is found to be 8.5 hours after dim light melatonin onset (DLMO), a circadian phase marker. Although DLMO is a good phase marker, it is hard to obtain in the clinical practice, as the determination is both time consuming and expen- sive. Instead of measuring DLMO, a reasonable approximation of the timing of DLMO can be obtained with collecting a morning-evening score with the morningness-eveningness questionnaire (MEQ) developed by Horne and Östberg (7). The rating of this questionnaire is strongly correlated to circadian phase in SAD (8). By making use of an individual’s MEQ score, a reasonable estimation of the optimal timing of light can be made.

Aims of the study At the University Medical Center Groningen (UMCG), light therapy for SAD patients is scheduled at a fixed clock time; all patients receive light at 8AM. In the current study we aimed to see whether the therapy response at this fixed time point is different for early and late chronotypes. To our knowledge, this is the first study to link chronotype with therapy success on a fixed LT time. We hypothesize that patients with a lower MEQ score, more evening type, show a better therapy outcome than morning types, as LT on 8AM would be more optimal for late types.

Methods Study design and participants For the analysis, data from four different studies conducted between 2005 and 2011 (9–12) are used. The studies have all been performed in the SAD outpatient clinic of the UMCG, the Netherlands. For specifications of the different light treatments see table 1. All studies were approved by the Medical Ethical Committee of the University Medical Center Groningen.

118 Chapter 10: The relation between chronotype and treatment outcome with light therapy on a fixed time schedule

A total number of 132 patients have been retrospectively selected based on the following criteria: all subjects met the criteria of major depressive disorder with a seasonal (winter) pattern according to the DSM-IV-TR and did not suffer from other DSM-IV classified psychiatric disorders as assessed by the Mini-International Neuropsy- chiatric Interview (13). The group consisted of 29 men and 103 women, mean age ± SD 37.4y ± 11.9 (table 1).

Table 1. Characteristics of subjects, light treatment and results.

^ƚƵĚǇ ^ƚƵĚǇϭ ^ƚƵĚǇϮ ^ƚƵĚǇϯ ^ƚƵĚǇϰ

WĂƌƚŝĐŝƉĂŶƚƐ;ŶͿ ϱϰ ϰϮ ϭϳ ϭϵ ŐĞŵĞĂŶŝŶǇĞĂƌƐ;ц^Ϳ ϯϴ͘Ϯ;цϭϭ͘ϳͿ ϯϳ͘ϭ;цϭϯ͘ϭͿ ϯϵ͘Ϯ;цϭϮ͘ϯͿ ϯϯ͘ϵ;цϴ͘ϳͿ

ĂƐĞůŝŶĞ^/',Ͳ^ƐĐŽƌĞ;ŵĞĂŶц^Ϳ Ϯϲ͘ϯ;цϱ͘ϴͿ Ϯϰ͘ϵ;цϱ͘ϭͿ Ϯϰ͘ϳ;цϳ͘ϭͿ ϯϬ͘ϯ;цϲ͘ϯͿ

WƌŽƉŽƌƚŝŽŶĂůƌĞĚƵĐƚŝŽŶ^/',Ͳ^ƐĐŽƌĞ ϲϲ͘Ϯ;цϯϰ͘ϮͿ ϳϭ͘Ϯ;цϮϯ͘ϴͿ ϲϰ͘ϯ;цϮϲ͘ϮͿ ϲϳ͘Ϯ;цϮϳ͘ϰͿ ŝŶй;ŵĞĂŶц^Ϳ

DY;ŵĞĂŶц^Ϳ ϱϭ͘ϵ;цϭϬ͘ϳͿ ϱϬ͘ϳ;цϮ͘ϴͿ ϱϮ͘ϯ;цϳ͘ϭͿ ϱϭ͘ϯ;цϵ͘ϰͿ

>ŝŐŚƚƐƉĞĐŝĨŝĐĂƚŝŽŶĂ ^ƚĂŶĚĂƌĚ ϱϬϬϬΣ< ϱϬϬϬΣ< ϱϬϬϬΣ< ϱϬϬϬΣ< ;ϭϬϬϬϬůƵǆͿ ;ϭϬϬϬϬůƵǆͿ ;ϭϬϬϬϬůƵǆͿ ;ϭϬϬϬϬůƵǆͿ

ǆƉĞƌŝŵĞŶƚĂů ϭϳϬϬϬ< ϭϳϬϬϬ< >ůƵĞůŝŐŚƚ ŶͬĂ ;ϭϬϬϬϬůƵǆͿ ;ϳϱϬůƵǆͿ ϰϳϬŶŵ ;ϭϬϬůƵǆͿ

zĞĂƌƐŽĨƐƚƵĚǇ ϮϬϬϱͬϮϬϬϲ ϮϬϬϴͬϮϬϬϵ ϮϬϭϬͬϮϬϭϭ ϮϬϬϳͬϮϬϬϴ

a All light conditions except the LED Blue light condition: full spectrum light, without UV.

Procedures The Structured Interview Guide for the Hamilton Depression Rating Scale-Seasonal Af- fective Disorder 24 item version (SIGH-SAD) was used to assess severity of depression (14). Prior to LT the morningness-eveningness questionnaire by Horne and Östberg was used to assess morningness-eveningness preference (7).

SIGH-SAD ratings were assessed before LT and one week after the end of LT. In the three studies which compared different methods of light therapy no significant differences between light conditions were observed: study 1, main effect “condition” F(2,49) = 0.73, ns; study 2, main effect “condition” F(1,20) = 0.012, ns; study 3, 67% recovery for standard treatment and 63% recovery for experimental treatment, ns. The fourth study did not use an experimental light therapy, as this study looked at gene expression in SAD patients (12). In study 1 and 2, patients received 2 weeks of LT, while in study 3 and 4 patients received 1 week of LT. No significant effect of treatment duration was found (15). For the current analysis we pooled the data of all studies and all different light conditions.

119 Chapter 10: The relation between chronotype and treatment outcome with light therapy on a fixed time schedule

All patients received light therapy at 8 A.M. at the clinic, independent of their chronotype. Statistical analysis Therapy outcome is defined as the percentage decrease in SIGH-SAD score comparing the rating 1 week after LT with the rating just prior to LT. Responders were calculated as patients with a decrease of at least 50% in their depression score.

Under the hypothesis that there is an optimum in the therapeutic response to light for a certain chronotype, a quadratic curve fitting was used to correlate MEQ score with therapy outcome. In addition, based on the MEQ score the patients have been separat- ed into 5 chronotype groups (1=definitely evening, 2=moderately evening, 3=intermediate, 4=moderately morning, 5=definite morning). Another quadratic curve fit is done, correlating chronotype group and therapy outcome.

Results Patients had an average MEQ score of 51.5 ± 8.2 (range 32 to 70). There were no definite evening chronotypes, moderate evening: 12, intermediate: 95, moderate morning: 23, definite morning: 2. Patients had an average proportional change in depression score of 68% ± 29 (range -36% to 100%, a positive change resembles a reduction in depression score). The percentage responders (decrease in depression score of equal or more than 50%) amounts to 76% (n=100).

The individual therapy outcome data show a large variation. Most of the patients show a response to the therapy, having a therapy outcome between 40-100%. There are 3 patients who do not respond to the therapy (therapy success <10%) and two patients show a worse depression score after light therapy (therapy success < -10%). No significant quadratic correlation between MEQ score and therapy outcome was found (F2,129 = 0.05, ns) (figure 1).

Figure 1. Distribution of proportional depression score reduction over different MEQ scores. Lower MEQ scores reflect more evening type, higher scores more morning type. 120 When patients were binned by chronotype score, no significant relation between chronotype group and therapy outcome was found either (F2,129 = 0.02, ns). The definite morning group shows on average the best therapy outcome, but consisted only of two patients. Furthermore it should be noted there are no definite evening types.

Discussion In this study, no significant relationship was found between MEQ score of SAD patients and therapy success of light therapy on a fixed time schedule. A better therapy outcome for evening types was hypothesized, as the timing of 8 AM is the optimal timing for evening type patients (MEQ score 27-30), based on a previous paper (6). This hypothesis is not supported by the data of this study.

An often cited hypothesis of the pathophysiology of SAD is the phase shift hypothesis (PSH) (3). This hypothesis states that a delayed circadian pacemaker plays a role in the complaints of SAD patients. The PSH suggests the benefit of light therapy in the morning comes from treating the phase delay in patients with SAD, by establishing a phase advance. Looking at the phase response curve (PRC), which shows how phase changes in response to light applied at different times of the day, properly timed morning light has this phase advancing effect (16,17). Therefore patients who are treated with morning light will most likely be phase advanced, if light is properly timed.

Although there are papers with arguments that confirm the phase shift hypothesis, there are also papers who argue against the hypothesis (18). Koorengevel et al. showed through a forced desynchrony protocol that there are no abnormalities of circadian phase in SAD patients compared to control subjects and no differences between summer, when subjects were not depressed, and winter, nor in the patients before LT and after LT (19). Furthermore Wirz-Justice and colleagues argued against the PSH showing there was no beneficiary effect of morning light compared to evening light in a randomized controlled trial (20) nor was there a relationship between circadian phase and a preferential response to morning or evening light. In a recent study morningness-eveningness preference showed no relation with a predisposition of SAD (21).

Our study shows no correlation between chronotype and therapy success with morning light, where timing should have been more optimal for late types and suboptimal for morning types. Although light in the morning is definitely an effective treatment for SAD, the mechanism behind this anti-depressive effect might be different from the phase shifting explanation. Some support for other mechanisms is found; in mice light has been shown to directly affect mood through melanopsin-expressing photorecep- tors, without any changes in the circadian system (18,22). Low intensity blue enriched light and short durations of blue enriched light exposure have been found to be as successful in treating SAD as normal white polychromatic light suggesting the possible involvement of melanopsin cells (9,11).

121 Limitations A limitation of this study is the retrospective design. The data are obtained in four different studies, which tested different methods of light therapy, although these different methods did not result in different therapy outcomes. Although our subjects show a good therapy response, the therapy response may have been even larger when we had optimized for chronotype. This should be tested in a prospective study, where timing of therapy is optimized to the chronotype of the patient. In such a study, chronotype should not only be assessed by the MEQ, the DLMO should be taken into account as well, as this is the golden standard for assessing chronotype. A recent study showed, although questionnaires can be used to predict DLMO, there is variation in the correlation between these variables (23).

Another limitation is the lack of extreme chronotypes, our study mostly consisted of intermediates or moderate morning/evening types. The current timing is optimized for the extreme evening types. As the extreme evening chronotypes are missing we cannot say anything about their therapy success.

Conclusions The current data indicates that SAD patients could be treated with a very good respond- er rate of 78% with morning light therapy on a fixed time point, independent of their chronotype. The results of this study are interesting both from a scientific interest in the pathological background of seasonal affective disorder and the mechanism of light therapy, but is definitely also of interest for the clinical practice. Being able to admin- ister light therapy on a fixed clock time in the morning is more convenient for both patients and the clinicians.

122 et al. Changes in winter depression phenotype References correlate with white blood cell gene expres- 1. Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Good- sion profiles: A combined metagene and gene win FK, Davenport Y, et al. Seasonal affective ontology approach. Prog Neuro-Psychophar- disorder: a description of the syndrome and macology Biol Psychiatry. 2015;58:8–14. DOI: preliminary findings with light therapy. Arch 10.1016/j.pnpbp.2014.10.015 Gen Psychiatry. 1984;41(1):72. 13. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim 2. Spijker J, Bockting C, Meeuwissen J, van Vliet I, P, Janavs J, Weiller E, et al. The Mini-Interna- Emmelkamp P, Hermens M, et al. Multidisciplinaire tional Neuropsychiatric Interview (MINI): the richtlijn Depressie (Derde revisie). Richtlijn voor development and validation of a structured de diagnostiek, behandeling en begeleiding van diagnostic psychiatric interview for DSM-IV and volwassen patiënten met een depressieve stoor- ICD-10. J Clin Psychiatry. 1998;59:22–33. nis. Utrecht: Trimbos-instituut; 2013. 14. Williams JBW, Link MJ, Rosenthal NE, Terman 3. Lewy a. J, Sack RL, Singer CM, Whate DM, M. Structured Interview Guide for the Hamilton Hoban TM. Winter Depression and the Phase- Depression Rating Scale—Seasonal Affective Shift Hypothesis for Bright Light’s Therapeutic Disorder Version (SIGH-SAD). New York, New Effects: History, Theory, and Experimental Evi- York State Psychiatr Inst. 1988; dence. J Biol Rhythms. 1988 Jan 1;3(2):121–34. 15. Knapen SE, van de Werken M, Gordijn MCM, DOI: 10.1177/074873048800300203 Meesters Y. The duration of light treatment and 4. Lewy AJ, Lefler BJ, Emens JS, Bauer VK. The therapy outcome in seasonal affective disorder. circadian basis of winter depression. Proc Natl J Affect Disord. 2014 Jun;166:343–6. DOI: Acad Sci U S A. 2006 May 9;103(19):7414–9. 10.1016/j.jad.2014.05.034 DOI: 10.1073/pnas.0602425103 16. Khalsa SBS, Jewett ME, Cajochen C, Czeisler C 5. Terman JS, Terman M, Lo ES, Cooper TB. Circa- a. A phase response curve to single bright light dian time of morning light administration and pulses in human subjects. J Physiol. 2003 Jun therapeutic response in winter depression. 15;549(Pt 3):945–52. DOI: 10.1113/jphysi- Arch Gen Psychiatry. 2001 Jan;58(1):69–75. ol.2003.040477 6. Terman M, Terman JS. Light therapy for sea- 17. Gordijn MC, Beersma DG, Korte HJ, van den sonal and nonseasonal depression: efficacy, Hoofdakker RH. Effects of light exposure and protocol, safety, and side effects. CNS Spectr. sleep displacement on dim light melatonin 2005 Aug;10(8):647–63. onset. J Sleep Res. 1999 Sep;8(3):163–74. 7. Horne JA, Ostberg O. A self-assessment 18. LeGates T a, Fernandez DC, Hattar S. questionnaire to determine morningness-eve- Light as a central modulator of circadian ningness in human circadian rhythms. Int J rhythms, sleep and affect. Nat Rev Neurosci. Chronobiol. 1976;4(2):97–110. 2014;15(June):443–54. DOI: 10.1038/nrn3743 8. Terman M, Terman J. Morningness-evening- 19. Koorengevel KM, Beersma DGM, den Boer JA, ness, circadian phase and the timing of sleep van den Hoofdakker RH. A Forced Desynchrony in patients with seasonal affective disorder. Study of Circadian Pacemaker Characteristics In: Society of Light Treatment and Biological in Seasonal Affective Disorder. J Biol Rhythms. Rhythms. 2001. p. 13. 2002 Oct 1;17(5):463–75. 9. Gordijn MCM, ’t Mannetje D, Meesters Y. The 20. Wirz-Justice A, Graw P, Krauchi K, Gisin B, effects of blue-enriched light treatment Jochum A, Arendt J, et al. Light Therapy in compared to standard light treatment in Seasonal Affective Disorder Is Independent Seasonal Affective Disorder. J Affect Disord. of Time of Day or Circadian Phase. Arch Gen 2012 Jan;136(1–2):72–80. DOI: 10.1016/j. Psychiatry. 1993 Dec 1;50(12):929–37. jad.2011.08.016 21. Oginska H, Oginska-Bruchal K. Chronotype and 10. Meesters Y, Duijzer WH. The effects of low personality factors of predisposition to season- intensity monochromatic blue light treatment al affective disorder. Chronobiol Int. 2014 Jan compared to standard light treatment in SAD. 7;1–9. DOI: 10.3109/07420528.2013.874355 SLTBR Abstr 2352. 2011; 22. LeGates T a, Altimus CM, Wang H, Lee H-K, 11. Meesters Y, Dekker V, Schlangen LJM, Bos Yang S, Zhao H, et al. Aberrant light di- EH, Ruiter MJ. Low-intensity blue-enriched rectly impairs mood and learning through white light (750 lux) and standard bright melanopsin-expressing neurons. Nature. light (10,000 lux) are equally effective in 2012;491(7425):594–8. DOI: 10.1038/na- treating SAD. A randomized controlled study. ture11673 BMC Psychiatry. 2011 Jan;11(1):17. DOI: 23. Kantermann T, Sung H, Burgess HJ. Comparing 10.1186/1471-244X-11-17 the Morningness-Eveningness Questionnaire 12. Bosker FJ, Terpstra P, Gladkevich A V., Janneke and Munich ChronoType Questionnaire to the Dijck-Brouwer D a., te Meerman G, Nolen W a., Dim Light Melatonin Onset. 2015;30(5):449–53.

123 Chapter 11 General discussion

Chapter 11: General discussion

In this thesis we studied the relationship between chronobiology and mood disorders. The first part describes vulnerability factors for mood disorders, such as genetic factors and chronotype, the second part looks into the direct relation between sleep-wake rhythms and mood using actigraphy and the third part studies therapeutic options in- fluencing chronobiological mechanisms. Different types of studies varying from large cohort studies to small scale, in-depth clinical pilot studies are used to investigate chronobiological mechanisms in mood disorders. In this general discussion, the most important findings will be summarized and possible pitfalls of the work, suggestions for future work and clinical implications will be discussed.

Part 1: Vulnerability factors for developing mood disorders Underlying genetic predisposition for mood disorders within circadian genes. In chapter 2 we looked at a large sample of patients with major depressive disorder in the Netherlands. We studied the relationship between genetic markers in pre-selected genes, chronotype and mood disorders, to see if there is an underlying genetic predisposition for mood disorders, mediated by chronotype. We found that 13 genetic markers within six genes were significantly associated with chronotype, and 59 genetic markers within 18 genes were significantly associated with mood disorders. Of these genes one showed a possible mediating effect of mood on chronotype, which did not survive correction for multiple testing. Looking at specific gene clusters, although none were found to represent any neuronal process, one cluster was found to be associated with metabolic syndrome, which has been associated with MDD before (1). Finding this cluster within circadian genes linking mood disorders to metabolic syndrome is of special interest, as also circadian regulation and metabolic syndrome have been studied before (2). It is important to realize that we did not conduct the analysis with metabolic syndrome in mind, so it is a chance finding which has to be replicated before drawing strong conclusions.

Chronotype is frequently suggested as a vulnerability factor for developing, or causing a worse course, of major depressive disorder (3–5). This is the case for the evening chronotype where people have a preference for a later timing of activities (5). This is because people with an evening chronotype experience more social jetlag, where the preference of timing is misaligned with the socially accepted timing (6–8). An example of this are adolescents who typically have an evening chronotype, while school timing is all directed towards early morning activities (9–11).

Is social jetlag related to major depressive disorder? In chapter 3 we studied whether social jetlag is actually associated with the MDD diagnosis and how it relates to depressive symptoms in patients and healthy people as well. In our study patients with MDD did not show more social jetlag compared to controls,

125 Chapter 11: General discussion

furthermore there was no relation between social jetlag and depression severity. This lack of a group difference shows that although patients are more likely to be evening types, the relation between chronotype and MDD cannot be explained by social jetlag. Another reason why patients in a current depressive episode show a later chronotype is reactive to the depression, instead of a cause of the depression. Patients with MDD often show a diurnal preference, with less depressive symptoms in the evening compared to the morning (12,13). A reaction to this preference might be a shift of activities towards the evening in these patients, when patients experience less mood symptoms. As we evaluated chronotype with a questionnaire concerning their daily rhythm (sleep times on work and free days), this change in behavior might be reflected in the chronotype.

Problems with studying chronotype in patients with different states of the disease In chapter 4 we discussed the difficulties with the interpretation of evening chronotype and depressive episodes. The key question in this chapter was whether the evening type is a state effect of a depressive episode, or if it is a vulnerability factor. Although the recent literature suggests it is a vulnerability factor, we hypothesize it might be a state effect.

Part 2: Rest-activity rhythms and mood – actigraphy The second part of this thesis (chapter 5-8) focused on bipolar disorder and actigraphy measures. Especially the link between rest-activity rhythm disturbances and mood with the different stages of the disease were studied.

Do patients with bipolar disorder show circadian rhythm disturbances in the euthymic phase of the disease compared to controls? In chapter 5 we examined rest-activity measures in euthymic patients with bipolar disorder, by conducting the largest actigraphy study to date in patients with bipolar disorder type I, while studying unaffected siblings and healthy controls as well. In this sample, we looked whether patients with BD in a euthymic phase still experienced circadian rhythm disturbances, by looking at their rest-activity rhythm as measured with actigraphy. Patients did not show less stability between days, variability within days nor a less robust amplitude compared to controls. This is an interesting finding, especially as we know that patients in a current mood episode do show rest-activity rhythm instability (14). Our results show patients are capable of maintaining a healthy circadian rhythm in their euthymic phase. There might be a selection bias in this sample, as all patients were medicated and were already familiar with the disorder for quite some time. As both pharmacological and cognitive behavioral therapy in bipolar disorder aim to stabilize rest-activity rhythms as well as mood, this might be a

126 possible confounder. Previously we showed similar results for sleep variables: euthymic patients showed no differences in sleep variables compared to controls (15). Can rest-activity patterns function as an early warning signal of imminent mood changes in a single subject? Chapter 6 shows an example of how actigraphy may be used for direct clinical benefit in one individual patient, and how rest-activity patterns might be an early warning signal for imminent mood changes. A patient with bipolar disorder experienced a major life event while participating in a two-week actigraphy protocol. We described how her actogram, the graph of her activity, reflected the life event. In the night following the life event she experienced a later sleep onset and a shorter sleep duration. As the patient noticed the change in mood as well, she reacted adequately by taking lorazepam for two nights, prescribed by her psychiatrist, which resulted in two good nights of sleep and no further mood disturbances. This adequate reaction to a disturbed night of sleep might have prevented a relapse into a full mood episode. In this chapter, we discussed how ambulatory assessment with actigraphy, together with the patients’ own experience could be useful in predicting relapses.

The temporal relation between sleep and circadian disturbances and mood in the onset of a mood episode. Based on this experience we conducted a new study, studying 13 patients with bipolar disorder for 180 days with both actigraphy and a sleep and mood diary to test the temporal order between mood and sleep symptoms. In chapter 7 we explored this relation with a novel method, studying both mean shifts in sleep variables (i.e. slowly sleeping less and less) and sudden peaks (such as one night of disrupted sleep) in the onset of a manic or depressive episode. To our knowledge, this is the first study of such scale, and with such frequent measurements looking specifically at this transition in the patients’ own daily life. We found that in two of the eight subjects with a mood transition the mood episode was preceded by sleep disturbances. When using a less restrictive method, we found that specifically in the onset of a manic episode the mood changes were preceded by sleep disturbances, while in the onset of a depressive episode the sleep disturbances came after the onset of mood symptoms.

Can fractal patterns function as a marker for biopolar disorder? In chapter 8 we again used the data from chapter 5 and 7 to look at another measure which can be derived from actigraphy, the fractal pattern. Fractal patterns are patterns which show a similar structure on different scales (16–18). These fractals patterns are also found in human physiology and this fractal pattern is often interpreted as a marker of good homeostatic regulation. The loss of this pattern has been associated with diseases such as Alzheimer’s disease and major depressive disorder (18,19). In chapter 8 we studied the cross-sectional data from chapter 5 and the longitudinal data from chapter 7. All activity data from wakefulness were analyzed for fractal patterns using a Detrended Fluctuation Analysis (DFA). From the DFA a scaling exponent is derived on different time scales. An ideal scaling component has a value of about 1. A value lower

127 than 1 indicates a random pattern, while a value above 1 indicates a smooth pattern.

The scaling component on small time scales (α1, from 5 minutes up to 90 minutes) has been shown to be higher in depressed patients before (20). The scaling component on larger time scales (α2, from 2 to 10 hours) has been shown to go to a random pattern when the suprachiasmatic nucleus (SCN) is lesioned (21,22). In chapter 8 we showed that patients with bipolar disorder have a higher α1 compared to their siblings and healthy controls. Further analysis showed this was only the case in female subjects.

Female patients, and female unaffected siblings, also showed a lower α2 compared to controls, indicating a possible genetic relation between α2 and vulnerability towards bipolar disorder. On the other hand, male patients showed a higher α2 compared to controls, which was unexpected and could not be explained by post-hoc analyses. In the longitudinal sample, no differences in the scaling components on both timescales were found between mood states. A relationship between more depressive symptoms and a lower α1 on a later time point was found in both the weekly and in the daily scores. The higher α1 in female patients, together with the lack of differences between mood episodes, show an interesting possible biomarker for bipolar disorder, especially because the scaling component is derived from a relatively cheap and non-invasive method. Replication is needed, but fractal patterns might be a promising route for biomarkers in bipolar disorder, and possible in more mood disorders.

Part 3: Therapeutic options with chronobiological mechanisms The last part of this thesis is focused on chronobiological treatment options for mood disorders.

Duration and timing of light therapy for seasonal affective disorder In chapter 9 and 10 we retrospectively studied a large sample of patients with seasonal affective disorder, by combining data from earlier studies that investigated different types of light treatment. All data from individual studies were pooled and analyzed. In chapter 9 we focused on the duration of light therapy, as we compared the effect of light therapy given for one week with light therapy during two weeks. There were no differences in therapy outcome between one and two weeks, only the speed of the recovery from depressive symptoms was faster in the group that received one week of therapy. As patients were already familiar with the length of the treatment, we hypothesize that the expectation of another week of light therapy might delay the effect of the treatment in the two-week group. To further understand the effects of expectations, we looked at the relation between expectations of the treatment and actual therapy outcome, and only in women we found a positive relationship between the expectations and outcome. This link between expectation and therapy outcome is an interesting construct to study, especially because this expectation has been suggested to be a mediator in pharmacological treatment of patients with major depressive disorder as well (23,24).

128 Chapter 11: General discussion

In chapter 10 we focused on the timing of light therapy. It has been suggested that for an optimal effect of light therapy patients should be given light therapy at a time based on their own chronotype (25). Patients with an earlier chronotype should be administered light therapy as early as 4:00 am, while later chronotypes should be given light therapy as late as 8:45 am. In our sample from the University Center Psychiatry in Gron- ingen, all patients were administered light at 8 am, independent of their chronotype. This time would be most ideal for later chronotypes according to the protocol from the literature (25). As chronotype was assessed before the start of therapy, we could check whether the evening chronotypes have indeed more therapy success compared to morning types. This was not the case, we found a good therapy outcome in all study participants, independent of chronotype. This indicates that light therapy in the early morning might be sufficient for treating SAD and that the timing the light therapy opti- mally to the patients chronotype is not necessary.

General discussion State or trait? A key question. Rhythm and blues are heavily intertwined and changes in one domain may result in changes in the other (26,27). However, in this thesis we have shown that the associations between rest-activity disturbances and mood disorders are often not as straightforward as we had expected when these studies were initiated. For example, in the relationship between social jetlag and depression it seems that patients with MDD do not necessarily show more social jetlag than healthy controls, as found in chapter 3. In other study which do show a relation between the evening type and major depressive disorder, this result is typically found in patients who are in a current depressive episode (28). This suggests the evening chronotype is more a state feature (depending on a mood episode) than a trait feature (independent of mood episodes). Similarly, in chapter 5 we did not find disturbances in circadian variables derived from actigraphy in the largest study in euthymic patients with bipolar disorder to date. This also suggests that circadian disruption is not a trait feature of the disease, but could still be a state feature.

On the other hand, in chapter 2 we found genetic markers within a set of circadian genes for both chronotype and mood disorders, and no mediating effect of chronotype on the prevalence of mood disorders. As the subjects we tested in chapter 3 and 5 were in remission, a possibility would be that patients simply do not show circadian disruptions anymore as they have less mood symptoms, as they are all under a treatment (29). To gain further understanding of the temporal interplay between rhythm and blues we set out to study the relation between sleep-wake disruptions and mood symptoms in a patient sample in chapter 7, finding that in almost all subjects the sleep disturbances co-occur with a change in mood, which makes the relationship between mood symptoms and sleep symptoms even more clear.

Although we did not find a clear trait feature in the circadian measurements from actigraphy, if we look at fractal patterns in the activity in patients, unaffected siblings and controls, we did find a significant difference in the scaling component on small time-

129 Chapter 11: General discussion

scales. As we did not find any differences within subjects between a stable period and both manic and depressive periods, a state effect of this scaling component is unlikely. Not only the scaling component on smaller time scales, but also the scaling component on larger time scales, which is hypothesized to be regulated by the circadian timing system, is disrupted in female patients and their unaffected siblings. This effect is similar to what is seen in rats with a lesion in their suprachiasmatic nucleus, making the hypothesis of a link between the circadian timing system and this scaling component more likely (21). These results indicate the regulation of this complex system might be compromised, independent of state, specifically in female patients.

Methodological considerations Circadian rhythm or rest-activity rhythm? An important consideration, and for some a limitation, is how we assessed circadian rhythm in this thesis. One might even ask if this thesis studied the circadian rhythm at all (30). By using a questionnaire to assess chronotype and by using actigraphy as a measure of circadian stability, the actual circadian rhythm is not assessed under optimal test conditions; we have only looked at rest-activity rhythms in real life. Ideally a subject is studied in controlled conditions, for instance in a forced desynchrony protocol, where a subject is asked to sleep on an either very short or very long cycle for a longer period (20- or 28-hour day). As the circadian system cannot entrain to these ex- tremes, it remains at its own endogenous period, and runs independently of the sleep- wake cycle. This allows the researcher to study the circadian component of a variable and the sleep component, independent of each other. However, forcing a patient with bipolar disorder out of their daily, Zeitgeber-synchronized, rhythm is highly unethical, because this may trigger a full-blown mood episode, which can take a long period to recover from. Another method to study the circadian rhythm is to measure the internal phase of that rhythm by studying the output of the circadian clock, the pineal hormone melatonin (31,32). Melatonin can be collected from saliva under dim light conditions to prevent light influences on the melatonin levels. By assessing saliva melatonin in the evening, the Dim Light Melatonin Onset (DLMO) can be derived. The DLMO is an interesting and relatively easy phase marker to assess. However, melatonin profiles are costly to analyze and the requirement of dim-light makes studying day-to-day behavior unpractical for regular clinical care. The methods used in this thesis, the chronotype questionnaires and actigraphy have been linked to internal phase in a number of studies. Both chronotype questionnaires, the morningness-eveningness questionnaire and the Munich Chronotype Questionnaire have been validated to each other and to the DLMO, showing good correlations (11,33–35). This shows that although we did not use the best method to assess phase, we did use the most suitable method for our purposes.

As we aimed to study patient in their own daily life, actigraphy is the most suitable method (36,37). It is non-invasive, and also minimally disruptive to the patients’ behavior. Actigraphy is well-validated for assessing sleep parameters and circadian parameters, also specifically in patients with bipolar disorder (37). As outcome measures we used the non-parametric variables, which are capable of capturing more than just the circadian rhythm (38,39). Interdaily stability (IS) measures the synchronization of the

130 Chapter 9: The duration of light treatment and therapy outcome in Seasonal Affective Disorder

rhythm to external Zeitgebers, Intradaily variability (IV) is a measure of fragmentation and the relative amplitude (RA) is the ratio of the activity during the most active 10 hours (M10) and the activity during the least active 5 hours (L5). This ratio is consid- ered the amplitude of the rhythm, which can be seen as the power of the rhythm. The start-times of L5 and M10 are a variable related to the timing of the rhythm. A limitation of using this measurement is the number of days needed for a reliable estimate of these variables, which is at least a week to 10 days (40). In this thesis, we used at least 14 days for the estimates. An opportunity would be to explore the different methods to calculate these variables. For the analyses in this thesis the average activity of an hour is used for the calculation of the variables. A recent paper showed other bin sizes, shorter than the hour used here, may give a more precise method to detect both fragmentation (IV) and synchronization (IS) (41).

Another limitation relevant to the work in this thesis is the lack of light measurements. Although light measurements are possible with actigraphy, they are mostly unreliable and take up a lot of storage on the device. Furthermore, it would require continuous attention from the participants to not put their sleeve over the light sensor, which made it unsuitable for the studies conducted for this thesis.

Objective or subjective sleep measures The work in this thesis is focused exclusively on objective measures of the rest-activity rhythm, either measured with a questionnaire studying the timing of sleep, or with actigraphy using sleep diaries only as a method to validate the sleep periods. No subjective measures of sleep duration and, possibly more importantly, sleep quality have been used to study patients. In the sample of patients with bipolar disorder subjective measures are assessed as well and can be studied at a later stage. However, we chose not to study the subjective measures, as we were trying to find the temporal relation between mood and objective sleep measures, without a masking effect of the interpretation of the patient. This is important, as for instance in an upcoming mania, patients might report a good sleep quality and no fatigue, while differences are definitely found in the objective sleep measures. In contrast, it has been shown that in remitted patients with bipolar disorder objective and subjective sleep and circadian measures correlate very well (42). Whether this is the case for these measures in, or in the transition to, a mood episode as well remains unclear.

Studying a temporal relation Recently, alongside the possibilities of studying patients for a longer period, the possible methodologies to study their behavior has increased as well (43,44). In chapter 7 we used a novel method for the temporal relation which has not been used before. We specifically wanted to see how sleep and mood symptoms were related in the transition to a manic or depressed episode in patients with bipolar disorder. In our method outliers are used to visualize the influence of, for instance, one short night. This method using outliers is more commonly done in statistical processing (45). We expanded

131 the analysis with a changepoint analysis to be able to study the slower changes, for instance a decreasing sleep duration day by day as well (46). A simpler method, such as a cross-correlation function analyzing one sleep variable to one mood variable might have been easier, but a lot of information would be lost with this method (47). Re-analyzing the sample with different methods, which are yet under development or to be developed, is possible and would be interesting to do.

Medication use in patients with mood disorders An important limitation of the work in this thesis is medication use in patients with mood disorder. In the chapter about social jetlag we accounted for antidepressant medication, and this showed that the small differences within social jetlag between patients and controls was partially explained by the fact that patients used medication. In the chapters studying patients with bipolar disorder, taking medication use into account is difficult because patients tend to use different types of medication that may have different mechanisms of action (48). Important to note is that lithium, the first-choice treatment of bipolar disorder, has an effect on the circadian rhythm and is one of the possible working mechanisms (49,50).

Secondary analyses of treatment studies In the last part of the thesis, all patients with seasonal affective disorder (SAD) who have been studied, were treated with light therapy. Data from different treatment studies were combined into one dataset. This means that, although we had a large sample of SAD patients who were assessed with similar instruments and procedures, we performed a secondary analysis using data that had not been collected with the specific aim to study optimal light therapy duration and timing. On top of that, a randomized trial would be best suited to answer the optimal duration and the timing question. For the timing question patients could be randomized into two treatment arms, one with exact timing to the chronotype of the subject and one with light treatment at 8 am. However, the overall good treatment effect, with a response rate of 78%, found in chapter 10 and the lack of differences between chronotypes indicates that finding differences between these groups would require a large sample size. This would also suggest that the clinical relevance of making such a distinction would be limited.

Future directions and future research Ambulatory assessment in real time Although ambulatory assessment and studying actigraphy measures in mood disorders has been around for decades, recent technological advancements have resulted in a number of opportunities for this field (51). For one, all work in this thesis analyzed the data retrospectively, after an event occurred. In other words, we received the actiwatch data after a patient had experienced sleep disturbances and mood symptoms. There is currently still no real-time method to read out the data. However, advancement of commercial products similar to actigraphy suggests that using real-time data for analyses is only a matter of time. Especially in patients with bipolar disorder, 86% of the patients show an interest in using consumer products to aid their self-management strategies (52). A number of requirements however need to be met to make this possible (51). First

132 Chapter 11: General discussion

of all, the device should have a large enough storage to store multiple days. Secondly, a high enough resolution, ideally of at least activity counts every minute (epoch length of a minute), is necessary for reliable calculations. The third requirement is that the device is not only capable of storing this data multiple days, but also able to function for for multiple days without need for recharge. Current devices often have to charge overnight, which would make sleep analysis impossible as the device is not worn overnight. The fourth requirement is that the activity data is freely, and in raw format, available to not only the patient, but also the clinician and researchers. A continuous stream of raw data would allow real time algorithms to search for patterns in these data and study the rest-activity rhythm of the patient at each time point. Currently this is not the case. In chapter 7 there was one participant wearing a commercial activity device alongside the actiwatch, which illustrated another aspect, namely that due to lack of openness from the manufacturer data could not be derived from the device to be studied next to the actiwatch data. The fifth requirement is the validation of these devices. Are they actually able to measure sleep as they promise, which is currently not the case for most devices (53).

Using commercially available devices might be an interesting way to study more patients for a longer time, each with their own device and possibly also with their own personalized diaries. However, important to note is that not only good things come from self-registration of the patient. A recent case-series presented patients with orthosomnia, where the main sleeping problem was the preoccupancy with the feedback from the device, instead of their subjective sleep quality (54). This preoccupancy resulted in more anxiety about their sleeping problems, resulting in symptoms of insomnia, which in these patients were specifically hard to treat.

Integrating ambulatory assessment in clinical practice Although actigraphy is a good example of how ambulatory assessment could assist the clinician treating patients with mood disorders, as we showed in chapter 6, moving from ambulatory assessment for research to clinical implementation is not as easy as it may seem. Significant steps have been taken to capture more personalized and precise patters of symtoms of individual patients over time (44,45), and ambulatory assessment of individual symptom networks have been used to develop tailor made and focused treatments for specific patients (43,44,55). Such personalized diagnosis and treatment could have benefits for clinical practice, but larger trials are needed to show that that this may indeed be of help in the diagnostic, and therapeutic process.

Automation of sleep and circadian measures In multiple chapters of this thesis sleep and circadian measures are calculated using different types of programs to derive these measures. Future work should focus on making these type of analyses more accessible to the general public. Most of all as the required skills for these analyses are not typically the skills a clinician or interested researcher might have. For this thesis we developed our own script (ACTman, available from https://github.com/compsy/ACTman/), which is publicly available to everybody who might be interested in analyzing sleep and circadian rhythmicity. It can be used in the open-source statistical program R (56,57).

133 State or trait Although we did study the state/trait feature quite elaboratively for bipolar disorder, the question remains whether chronotype is a stable trait feature, or more dependent on the mood state. Ideally the same participant would be assessed a number of times over the years, to see how chronotype might change across age, occupation and most interesting, how it might change depending on the state of the disease. The Nether- lands Study of Depression and Anxiety, which was used in chapter 3, might be suitable for such an analysis, as the same patients are followed year after year. Using such an elaborative database might give the final answer on the stability of chronotype over different states of major depressive disorder.

Sleep deprivation as an antidepressant Although this thesis aimed to further our understanding of the chronobiological un- derpinnings of mood disorders, a lot remains to be studied. For example, an important application of chronobiology in mood disorders has not been mentioned at all in this thesis, namely treatment of depression with sleep deprivation (58,59). Sleep deprivation is one of the most effective types of treatment of depression, however the effects are often short in duration, as the depressive symptoms often return with the recovery night. Great progress has been made by adding light therapy to multiple nights of sleep deprivation with recovery nights in between. Work from this thesis may be relevant to apply in chronotherapy. For instance, the fractal patterns of chapter 8, which are related to the dysfunction of the circadian rhythm, would be interesting to study before and after chronotherapy to see if the chronotherapy also has a restorative function on that marker of circadian rhythmicity. Furthermore, the discussion on optimal treatment duration and repeating the therapy multiple times are promising as well, just as we did in chapter 9 and 10 (60).

Clinical implications The results from this thesis show that circadian disruption, sleep problems and mood symptoms are very much intertwined in patients with mood disorders, and most of all, very complex. Specifically, we show in chapter 5 that patients with bipolar disorder in a euthymic period are actually able to show a good rhythm, without more fragmentation of the rhythm than healthy controls, and show a similar synchronization to external Zeitgebers. The results from chapter 7 suggest that in the transition into a mood episode a lot is happening at the same time, but we see that specifically in manic episodes the start of the episode is preceded by sleep disturbances. Although further work is necessary to see if there are actually predicting factors within sleep and circadian variables, the feasibility of such a study as well as these pilot data are very promising. Chapter 8 shows a promising statistically defined possible biomarker, the fractal pattern, although replication is needed. Lastly, in chapter 9 and 10 we showed that patients with seasonal affective disorder can be treated successfully with only one week, instead of two weeks of light therapy, in the morning. The results showing optimal timing to the individual chronotype is not necessary for good response rates are promising, not in the least for the people responsible for the logistics behind planning and administering the light therapy.

134 Conclusions From the previous chapters, we can conclude that circadian rhythm disturbances are most likely a state factor in patients with mood disorders. These disturbances, together with sleep problems play a prominent role in the transition to a mood episode in patients with bipolar disorder. Although more work is needed to fully understand this role, these disturbances might function as a predictor of upcoming transitions, enabling both patient and clinician to intervene in a timely manner. Furthermore, we showed that for a sub diagnosis of major depressive disorder, seasonal affective disorder, light therapy is very effective, and only has to be administered for a short duration, on the same time for all patients, independent of chronotype.

This thesis set out to add to the rich tradition of chronobiology within mood disorders, especially as the methods which are currently available give the opportunity to re-examine certain questions. Although this thesis shines some light on the relation between Rhythm & Blues, there are still many dark spaces to be enlightened.

135 Psychiatry. 2009;10(4):269–75. References 14. Harvey AG. Sleep and circadian rhythms 1. Pan A, Keum N, Okereke OI, Sun Q, Kivimaki M, in bipolar disorder: seeking synchrony, Rubin RR, et al. Bidirectional Association Be- harmony, and regulation. Am J Psychiatry. tween Depression and Metabolic Syndrome. Di- 2008 Jul;165(7):820–9. DOI: 10.1176/appi. abetes Care. 2012 May 1;35(5):1171 LP-1180. ajp.2008.08010098 2. Maury E, Ramsey KM, Bass J. Circadian Rhythms 15. Verkooijen S, van Bergen AH, Knapen SE, Vreek- and Metabolic Syndrome From Experimental er A, Abramovic L, Pagani L, et al. An actigraphy Genetics to Human Disease MS. 2010;1047–61. study investigating sleep in bipolar I patients, 3. Merikanto I, Lahti T, Kronholm E, Peltonen unaffected siblings and controls. J Affect Dis- M, Laatikainen T, Vartiainen E, et al. Eve- ord. 2017;208:248–54. ning types are prone to depression. 16. Hu K, Ivanov PC, Chen Z, Hilton MF, Stanley HE, Chronobiol Int. 2013;30(5):719–25. DOI: Shea SA. Non-random fluctuations and multi- 10.3109/07420528.2013.784770 scale dynamics regulation of human activity. Phys 4. Drennan MD, Klauber MR, Kripke DF, Goyette A Stat Mech its Appl. 2004;337(1–2):307–18. LM. The effects of depression and age on the 17. Hardstone R, Poil SS, Schiavone G, Jansen R, Horne-Ostberg morningness-eveningness Nikulin V V., Mansvelder HD, et al. Detrend- score. J Affect Disord. 1991;23(2):93–8. ed fluctuation analysis: A scale-free view on 5. Chan JWY, Lam SP, Li SX, Yu MWM, Chan NY, neuronal oscillations. Front Physiol. 2012;3 Zhang J, et al. Eveningness and insomnia: inde- NOV(November):1–13. pendent risk factors of nonremission in major 18. Goldberger AL, Amaral LAN, Hausdorff JM, Iva- depressive disorder. Sleep. 2014;37(5):911–7. nov PC, Peng C, Stanley HE. Fractal dynamics in DOI: 10.5665/sleep.3658 physiology : Alterations with disease and aging. 6. Levandovski R, Dantas G, Fernandes LC, Caumo 2002;99. W, Torres I, Roenneberg T, et al. Depression 19. Hu K, Someren EJW Van, Shea SA, Scheer FAJL. Scores Associate With Chronotype and Social Reduction of scale invariance of activity fluc- Jetlag in a Rural Population. Chronobiol Int. tuations with aging and Alzheimer ’ s disease : 2011;28(9):771–8. Involvement of the circadian pacemaker. 7. Wittmann M, Dinich J, Merrow M, Roenneberg T. 2009;106(8). Social jetlag: misalignment of biological and so- 20. Aybek S, Ionescu A, Berney A, Chocron O, Amin- cial time. Chronobiol Int. 2006;23(1–2):497–509. ian K, Vingerhoets FJG. Fractal temporal organi- 8. Wittmann M, Paulus M, Roenneberg T. De- sation of motricity is altered in major depres- creased psychological well-being in late “chro- sion. Psychiatry Res. 2012;200(2–3):288–93. notypes” is mediated by smoking and alcohol 21. Hu K, Scheer FAJL, Ivanov PC, Buijs RM, Shea SA. consumption. Subst Use Misuse. 2010;45(Sep- The suprachiasmatic nucleus functions beyond tember):15–30. circadian rhythm generation. Neuroscience. 9. Vinne V Van Der, Zerbini G, Siersema A, Pieper 2007;149(3):508–17. A, Merrow M, Hut RA, et al. Timing of Examina- 22. Hu K, Harper DG, Shea SA, Stopa EG, Scheer tions Affects School Performance Differently in FAJL. Noninvasive fractal biomarker of clock Early and Late Chronotypes. 2015;30(1):53–60. neurotransmitter disturbance in humans with 10. Urbán R, Magyaródi T, Rigó A. Morningness-eve- dementia. Sci Rep. 2013;3:1–7. ningness, chronotypes and health-impairing 23. Rutherford BR, Marcus SM, Wang P, Sneed JR, behaviors in adolescents. Chronobiol Int. Pelton G, Devanand D, et al. A randomized, 2011;28(3):238–47. prospective pilot study of patient expec- 11. Roenneberg T, Wirz-Justice A, Merrow tancy and antidepressant outcome. Psy- M. Life between Clocks: Daily Temporal chol Med. 2012 Sep 13;1–8. DOI: 10.1017/ Patterns of Human Chronotypes. J Biol S0033291712001882 Rhythms. 2003 Feb 1;18(1):80–90. DOI: 24. Rutherford BR, Wall MM, Brown PJ, Choo 10.1177/0748730402239679 T-H, Wager TD, Peterson BS, et al. Patient 12. Gordijn MC, Beersma DG, Bouhuys a L, Reinink Expectancy as a Mediator of Placebo Effects E, Van den Hoofdakker RH. A longitudinal in Antidepressant Clinical Trials. Am J Psychi- study of diurnal mood variation in depression; atry. 2016;(February):appi.ajp.2016.1. DOI: characteristics and significance. J Affect Disord. 10.1176/appi.ajp.2016.16020225 1994 Aug;31(4):261–73. 25. Terman M, Terman JS. Light therapy for sea- 13. Mendlewicz J. Sleep disturbances: core symp- sonal and nonseasonal depression: efficacy, toms of major depressive disorder rather than protocol, safety, and side effects. CNS Spectr. associated or comorbid disorders. World J Biol 2005 Aug;10(8):647–63.

136 26. Harvey AG. Sleep and circadian func- therapy: improved sensitivity to its effects on tioning: critical mechanisms in the mood rest-activity rhythms in Alzheimer patients by disorders? Annu Rev Clin Psychol. 2011 application of nonparametric methods. Chro- Jan;7:297–319. DOI: 10.1146/annurev-clinp- nobiol Int. 1999;16(4):505–18. sy-032210-104550 39. Gonçalves BSB, Adamowicz T, Louzada FM, 27. Wirz-Justice A. Biological rhythm disturbances Moreno CR, Araujo JF. A fresh look at the use in mood disorders. Int Clin Psychopharmacol. of nonparametric analysis in actimetry. Sleep 2006 Feb;21 Suppl 1:S11-5. DOI: 10.1097/01. Med Rev. 2015;20:84–91. DOI: 10.1016/j. yic.0000195660.37267.cf smrv.2014.06.002 28. Antypa N, Vogelzangs N, Meesters Y, Schoevers 40. VAN Someren EJW. Improving actigraphic sleep RA, Penninx BWJH. Chronotype associations estimates in insomnia and dementia: how many with depression and anxiety in a large cohort nights? J Sleep Res. 2007 Sep;16(3):269–75. study. Depress Anxiety. 2016 Jan;33(1):75–83. DOI: 10.1111/j.1365-2869.2007.00592.x 29. Hirschfeld RM a. Guideline Watch : Practice 41. Gonçalves BSB, Cavalcanti PRA, Tavares GR, Guideline for the Treatment of Patients With Campos TF, Araujo JF. Nonparametric meth- Bipolar Disorder , 2Nd Edition. Practice. ods in actigraphy: An update. Sleep Sci. 2005;(November):1–9. DOI: 10.1176/appi. 2014;7(3):158–64. books.9780890423363.148430 42. Boudebesse C, Geoffroy PA, Bellivier F, Henry 30. Wirz-Justice A. How to measure circadi- C, Folkard S, Leboyer M, et al. Correlations be- an rhythms in humans. Medicographia. tween objective and subjective sleep and circa- 2007;29:84–90. dian markers in remitted patients with bipolar 31. Albrecht U. Timing to perfection: the biology of disorder. Chronobiol Int. 2014;31(5):698–704. central and peripheral circadian clocks. Neuron. DOI: 10.3109/07420528.2014.895742 2012 Apr 26;74(2):246–60. DOI: 10.1016/j. 43. Bos FM, Schoevers R a., aan het Rot M. Expe- neuron.2012.04.006 rience sampling and ecological momentary 32. Dibner C, Schibler U, Albrecht U. The mamma- assessment studies in psychopharmacology: lian circadian timing system: organization and A systematic review. Eur Neuropsychophar- coordination of central and peripheral clocks. macol. 2015;1–12. DOI: 10.1016/j.euroneu- [Internet]. Vol. 72, Annual review of physiology. ro.2015.08.008 2010. 517-49 p. DOI: 10.1146/annurev-physi- 44. Aan het Rot M, Hogenelst K, Schoevers RA. ol-021909-135821 Mood disorders in everyday life: A systematic 33. Zavada A, Gordijn M, Beersma D, Daan S, Roen- review of experience sampling and ecological neberg T. Comparison of the Munich Chrono- momentary assessment studies. Clin Psychol type Questionnaire with the Horne‐Östberg’s Rev. 2012;32(6):510–23. DOI: 10.1016/j. Morningness‐Eveningness score. Chronobiol cpr.2012.05.007 Int. 2005 Mar 1;22(2):267–78. DOI: 10.1081/ 45. Benneyan, JC.,Lloyd, RC., Plsek, PE. Statis- CBI-200053536 tical process control as a tool for research 34. Kantermann T, Sung H, Burgess HJ. Comparing and healthcare improvement. Qual Saf Heal the Morningness-Eveningness Question- Care. 2003;12(6):458–64. DOI: 10.1136/ naire and Munich ChronoType Question- qhc.12.6.458 naire to the Dim Light Melatonin Onset. 46. Zeileis A, Leisch F, Hornik K, Kleiber C. struc- J Biol Rhythms. 2015;30(5):449–53. DOI: change: An R Package for Testing for Structural 10.1177/0748730415597520 Change in Linear Regression Models. J Stat 35. Horne JA, Ostberg O. A self-assessment Softw. 2002 Jan 10;1(2). questionnaire to determine morningness-eve- 47. Bauer M, Grof P, Rasgon N, Bschor T, Glenn T, ningness in human circadian rhythms. Int J Whybrow PC. Temporal relation between sleep Chronobiol. 1976;4(2):97–110. and mood in patients with bipolar disorder. 36. Ancoli-Israel S, Cole R, Alessi C, Chambers M, Bipolar Disord. 2006 Apr;8(2):160–7. DOI: Moorcroft W, Pollak CP. The role of actigraphy 10.1111/j.1399-5618.2006.00294.x in the study of sleep and circadian rhythms. 48. Geddes JR, Miklowitz DJ. Bipolar Disorder Sleep. 2003 May 1;26(3):342–92. 3 Treatment of bipolar disorder. Lancet. 37. Kaplan K a, Talbot LS, Gruber J, Harvey AG. 2013;381(9878):1672–82. DOI: 10.1016/ Evaluating sleep in bipolar disorder: com- S0140-6736(13)60857-0 parison between actigraphy, polysomnog- 49. Moreira J, Geoffroy PA. Lithium and bi- raphy, and sleep diary. Bipolar Disord. 2012 polar disorder: Impacts from molecular Dec;14(8):870–9. DOI: 10.1111/bdi.12021 to behavioural circadian rhythms. Chro- 38. Van Someren EJ, Swaab DF, Colenda CC, Cohen nobiol Int. 2016;528(March):1–23. DOI: W, McCall W V, Rosenquist PB. Bright light 10.3109/07420528.2016.1151026

137 50. Roybal K, Theobold D, Graham A, DiNieri JA, Russo SJ, Krishnan V, et al. Mania-like behavior induced by disruption of CLOCK. Proc Natl Acad Sci. 2007;104(15):6406–11. DOI: 10.1073/ pnas.0609625104 51. Piwek L, Ellis DA, Andrews S, Joinson A. The Rise of Consumer Health Wearables : Promises and Barriers. 2018;1–9. 52. Nicholas J, Boydell K, Christensen H. Beyond symptom monitoring: Consumer needs for bipolar disorder self-management using smart- phones. Eur Psychiatry. 2017;44:210–6. DOI: 10.1016/j.eurpsy.2017.05.023 53. Evenson KR, Goto MM, Furberg RD. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act. 2015;12(1):159. DOI: 10.1186/ s12966-015-0314-1 54. Baron KG, Abbott S, Jao N, Manalo N, Mullen R. Orthosomnia : Are Some Patients Taking the Quantified Self Too Far ? 2017;13(2):1–4. 55. Kroeze R, Van Veen DC, Servaas MN, Bastiaan- sen JA, Oude Voshaar R, Borsboom D, et al. Personalized feedback on symptom dynamics of psychopathology: a proof-of-principle study. J Pers Res. 2016; 56. R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna, Austria; 2018. 57. Kunkels YK, Knapen SE, Riese H, Emerencia AC. Actigraphy Manager [ACTman] [Internet]. 58. Benedetti F, Riccaboni R, Locatelli C, Poletti S, Dallaspezia S, Colombo C. Rapid treatment response of suicidal symptoms to lithium, sleep deprivation, and light therapy (chronothera- peutics) in drug-resistant bipolar depression. J Clin Psychiatry. 2014 Feb;75(2):133–40. DOI: 10.4088/JCP.13m08455 59. Wirz-Justice A. Chronobiology comes of age. Acta Psychiatr Scand. 2017;136(6):531–3. 60. Druiven SJM, Meesters Y, Riese H, Schoevers RA. Repeated sleep deprivation for a more rapid decrease in depressive symptoms in combined chronotherapy? Vol. 137, Acta psychiatri- ca Scandinavica. United States; 2018. p. 442.

138 Chapter 12 Addenda

Nederlandse Samenvatting

Chapter 12: Addenda

Introductie Stemmingsstoornissen komen veel voor en hebben in de westerse wereld een groot aandeel in de zorgvraag binnen de geestelijke gezondheidszorg. Binnen het spectrum van stemmingsstoornissen zijn twee grote groepen te onderscheiden, de depressieve stoornis en de bipolaire stoornis. De depressieve stoornis wordt gekenmerkt door een periode verlaagde stemming en/of verlies van interesse of plezier in activiteiten die normaal wel voor plezier zorgen. De depressieve stoornis (ook wel unipolaire depressie genoemd) maakt het grootste deel uit van de groep van stemmingsstoornissen, een kwart van de westerse bevolking maakt tijdens het leven een depressie door. Een subtype van de depressieve stoornis is de seizoensgebonden depressie. Hierbij keert de depressie jaarlijks terug in hetzelfde seizoen, typisch gezien in de winter, en trekt ook weer bij rondom dezelfde tijd van het jaar, typisch gezien in de lente.

De bipolaire stoornis wordt gekenmerkt door het doormaken van een manische episode, waarbij de stemming juist verhoogd is en patiënten buitengewoon goed gestemd zijn, (te) snel denken, een toename hebben van doelgerichte activiteit en een verminderde hoeveelheid slaap nodig hebben. In sommige gevallen komen bij een manische episode ook psychotische symptomen in de vorm van wanen of hallucinaties voor. Daarnaast maken patiënten met de bipolaire stoornis vaak ook depressieve episoden door. Bij een op de twintig mensen in de bevolking komt de bipolaire stoornis, in meer of minder ern- stige vorm, voor. Patiënten met de bipolaire stoornis krijgen vaak een terugval ondanks behandeling, en kunnen ook buiten de episoden stemmingsklachten houden.

Al vroeg in de vorige eeuw beschreef Kraepelin het verschil in slaapbehoefte tussen een manie en een depressie. Hier werd de link al gelegd tussen stemmingsstoornissen en chronobiologie, het veld binnen de biologie dat zich bezighoudt met periodieke (dagelijkse, maandelijkse of jaarlijkse) fenomenen. Chronobiologie komt terug in stemmingsstoornissen. Zo hebben patiënten met een depressieve stoornis vaak een wisseling van de stemming binnen de dag, met over het algemeen meer stemmingsproblemen in de ochtend. Ook het seizoensgebonden patroon van de seizoensgebonden depressie is een duidelijk periodiek fenomeen. Tenslotte zijn, zoals hierboven genoemd, slaapproblemen in de depressieve stoornis en in de bipolaire stoornis een kernsymptoom. Bij de depressieve stoornis wordt veelal insomnie, het onvermogen tot slapen, gezien. Een kleiner deel van de patienten heeft juist een grotere slaapbehoefte dan normaal. Bij de bipolaire stoornis kan dit ook voorkomen in de depressieve episode, in een manische episode wordt juist het verminderd nodig hebben van slaap gezien als een belangrijk symptoom.

In dit proefschrift is de relatie tussen stemmingsstoornissen en de chronobiologie onderzocht. Het proefschrift bestaat uit drie delen, het eerste deel gaat over kwetsbaarheid voor stemmingsstoornissen, het tweede deel over de directe relatief tussen rust-activiteit ritme en stemming en het derde deel over de therapeutische functie van licht.

Om de kwetsbaarheid voor stemmingsstoornissen te, is gebruik gemaakt van data uit de Nederlandse Studie naar Depressie en Angst (NESDA). Hierin hebben we vooral gekeken naar chronotype en afgeleide maten daarvan. Chronotype geeft aan of een per-

140 Nederlandse Samenvatting

Chapter 12: Addenda

soon een ochtend- of een avondtype is, of er ergens tussenin zit. Het avond chronotype wordt veelal geassocieerd met stemmingsstoornissen. In het tweede deel van dit proefschrift wordt de directe relatie gelegd tussen het rust-activiteit ritme en stemming bij patiënten met de bipolaire stoornis. Het rust-activiteit ritme is vastgelegd met behulp van actigrafie. Met actigrafie kan, vrij nauwkeurig, het activiteitspatroon van een persoon worden gemeten door de bewegingen van de pols te registreren. Met speciale berekeningen op de data verkregen met actigrafie kan je betrouwbaar vaststellen wanneer iemand slaapt en wanneer iemand wakker is. In dit proefschrift is deze methode gebruikt om verschillende stadia van de bipolaire stoornis te onderzoeken. In het derde deel is de therapeutische functie van licht vastgelegd bij een pa- tiëntengroep die veel baat heeft bij lichttherapie, de patiënten met winterdepressie. In een behandelstudie bij patiënten met winterdepressie is gekeken wat de optimale duur en de optimale timing is van lichttherapie voor een goed therapeutisch effect.

Kwetsbaarheid voor stemmingsstoornissen In hoofdstuk 2 onderzochten we de relatie tussen genen, chronotype en stemmingsstoornissen. Er werden vooraf genen geselecteerd die een verband houden met de biologische klok en die werden gerelateerd aan chronotype en daarna aan het aanwezig zijn van stemmingsstoornissen. Hierna werd onderzocht of de aanwezigheid van een stemmingsstoornis van invloed was op de relatie tussen genen en chronotype. Dertien genetische markers, binnen zes genen, zijn significant geassocieerd met chronotype, 59 genetische markers binnen 18 genen zijn significant geassocieerd met stemmingsstoornissen. Eén marker toonde een mogelijk mediatie effect, wat niet bleef staan na correc- tie voor multiple testing. Vervolgens hebben we in de gevonden significante genetische markers onderzocht of er specifieke genenclusters te vinden zijn, die functioneel bij elkaar horen. Geen van de gevonden clusters vertegenwoordigde een neuronaal proces, echter één cluster was geassocieerd met metabool syndroom. Het is interessant dat dit cluster gevonden werd binnen een set circadiane genen die geassocieerd zijn met stemming, omdat circadiane regulatie en metabole stoornissen ook gerelateerd zijn aan elkaar. Er is dus blijkbaar overlap in de kwetsbaarheid voor chronotype en stemmingsstoornissen in een metabool cluster. Aan de andere kant is het belangrijk te benoemen dat dit een toevalsbevinding is, en dat het niet het oorspronkelijke plan was om deze relatie te onderzoeken.

In hoofdstuk 3 onderzochten we de associatie van social jetlag (de mismatch tussen het chronotype en het daadwerkelijke slaap-waak ritme) en de diagnose depressieve stoornis. Social jetlag wordt over het algemeen meer bij mensen gezien met een avond chronotype. Wij vonden dat patiënten met de depressieve stoornis niet meer social jetlag hadden dan controlepersonen. Dit laat zien dat, hoewel patiënten met de depressieve stoornis wel vaker avondtype zijn, de relatie waarschijnlijk niet berust op het ontwikkelen van social jetlag. Het zou ook kunnen zijn dat patiënten als symptoom ten gevolge van een dagelijks ritme in de stemmingssymptomen, met meer depressieve symptomen in de ochtend, uiteindelijk meer avondtype worden.

141 Nederlandse Samenvatting

Chapter 12: Addenda

In hoofdstuk 4 gingen we in op de uitdagingen bij de interpretatie van het avondtype en depressieve episodes. Hierin concludeerde we dat het huidige onderzoek de vraag of avondtype het effect is van de depressieve episode of juist een kwetsbaarheid voor depressie nog niet kan beantwoorden, omdat de studies alleen naar cross-sectionele data hebben gekeken, en nooit longitudinaal.

Rust-activiteit ritme en stemming – actigrafie Met actigrafie keken we in patiënten met de bipolaire stoornis specifiek naar rust-activiteit ritme maten en stemming. In hoofdstuk 5 onderzochten we deze maten in patiënten met de bipolaire stoornis, hun niet-aangedane broers en zussen en controlepersonen. Specifiek werd er gekeken of patiënten in een euthyme periode, de periode buiten de stemmingsepisode, ritme problemen hadden. In dit onderzoek vonden we dat patiënten niet minder stabiliteit of meer variabiliteit vertoonden, ook leek de amplitude van het ritme niet anders dan bij familieleden en controlepersonen. Aangezien patiënten tijdens een stemmingsepisode vaak wel minder stabiliteit, meer variabiliteit en een minder sterk ritme (kleinere amplitude) hebben, is het interessant te weten dat patiënten hun circadiane ritme wel onder controle kunnen krijgen buiten een stemmingsepisode.

In hoofdstuk 6 hebben we een voorbeeld getoond van hoe actigrafie als een waarschu- wingssignaal kan dienen voor een opkomende stemmingsepisode bij een individuele patient. Dit hoofdstuk beschrijft hoe deze patiënt een life event doormaakte met aan- sluitend een afname in slaapduur. Dit werd zichtbaar gemaakt met een actogram, de vi- suele weergave van de actigrafie data. Hier werd adequaat op gereageerd door patiënt en behandelaar, door te starten met slaapmedicatie. Het zichtbare effect van het life event, gecombineerd met het zichtbare effect van de interventie, namelijk langere en betere slaap, roept de vraag op of het op deze manier meten en waarschuwen zinvol in de praktijk toegepast kan worden om tijdig in te grijpen bij een beginnende terugval.

In hoofdstuk 7 hebben we onderzocht wat de volgorde is tussen slaapproblemen en stemmingsproblemen in het ontstaan van een episode. We creëerden een nieuwe methode waarbij we keken naar gemiddelde veranderingen (bijvoorbeeld langzaam steeds minder slapen) en plotselinge veranderingen (bijvoorbeeld één korte nacht slaap) specifiek in de beginperiode van een manie of een depressie. In twee van de acht patiënten werden de stemmingsproblemen voorafgegaan door slaapproblemen. Met een minder restrictieve methode vonden we specifiek dat bij een manische episode slaapproblemen voorafgingen aan stemmingsproblemen, terwijl bij een depressieve episode de slaapproblemen na de stemmingsproblemen kwamen.

In hoofdstuk 8 hebben we de data van hoofdstuk 5 en 7 opnieuw gebruikt om een andere maat te berekenen, het fractale patroon. Fractale patronen zijn patronen die een gelijke structuur vormen op verschillende schalen. Een voorbeeld zijn de vertakkingen van een boom, waarbij de kleinste vertakkingen in de bladeren lijken op het patroon van vertakkingen zoals dat in de hele boom wordt gevonden, alleen op een veel kleiner detailniveau. Fractale patronen zijn ook terug te vinden in de menselijke

142 Nederlandse Samenvatting

Chapter 12: Addenda

fysiologie en worden geïnterpreteerd als een teken van een goede homeostatische regulatie. Het verlies van dit patroon is onder meer gezien bij patienten met de ziek- te van Alzheimer en bij de depressieve stoornis. In hoofdstuk 8 lieten we zien dat patiënten met de bipolaire stoornis ook een verminderd fractaal patroon hebben in vergelijking met gezonde controlepersonen. Specifiek werd dit gevonden in vrouwe- lijke patiënten, maar ook, in mindere mate, bij zussen van patiënten. Dit geeft aan dat er mogelijk een kwetsbaarheid te vinden is in het fractale patroon voor de bipolaire stoornis, en dat het mogelijk als biomarker kan dienen voor de bipolaire stoornis. In de ontwikkeling naar een episode toonden we aan dat patiënten eerst depressieve symptomen ontwikkelen, en later verlies van het fractale patroon.

Therapie met chronobiologische mechanismen In de laatste twee hoofdstukken onderzochten we een therapie met een mogelijk chro- nobiologisch mechanisme binnen winterdepressie, namelijk lichttherapie. Retrospec- tief werd gekeken naar een groep patiënten die aan verschillende studies meededen waarin verschillende types licht werden getest. Aangezien er in de individuele studies geen verschillen tussen de types licht werden gevonden werden alle data bij elkaar genomen. In hoofdstuk 9 keken we naar de duur van lichttherapie, waarbij één week lichttherapie met twee weken lichttherapie werd vergeleken. Er werd geen verschil in uiteindelijk effect gevonden. Wel werd aangetoond dat patiënten die één week therapie kregen sneller van hun depressieve symptomen af waren dan patiënten die twee weken lichttherapie kregen. Aangezien de patiënten op de hoogte waren van de uiteindelijke therapieduur is de hypothese dat de verwachting van een tweede week lichttherapie het effect van de behandeling vertraagd.

In hoofdstuk 10 onderzochten we de timing van lichttherapie. Er wordt gedacht dat voor een optimaal effect lichttherapie specifiek getimed zou moeten worden op het chronotype van een individuele patiënt. Patiënten met een vroeg chronotype zouden dan eerder lichttherapie moeten krijgen dan patiënten met een laat chronotype. In het Universitair Centrum Psychiatrie in Groningen krijgen alle patiënten om 8 uur ’s ochtends lichttherapie. Volgens de protocollen is dit optimaal voor late chronotypes. Het chronotype op basis van een gevalideerde vragenlijst, was bekend in de dataset, en we keken of patiënten met een avondtype het beter deden dan ochtendtypes. Dit was niet het geval, er werd een goede therapie uitkomst gezien in alle patiënten, onafhankelijk van chronotype. Dit laat zien dat lichttherapie in de ochtend voldoende is om winterdepressie te behandelen, en er geen specifiekere timing nodig is.

Methodologische overwegingen In dit proefschrift hebben we getracht het circadiane ritme te onderzoeken in pati- ënten met stemmingsstoornissen. De vraag is of we wel daadwerkelijk het circadiane ritme onderzocht hebben, of dat we alleen hebben gekeken naar het rust-activiteit

143 Nederlandse Samenvatting

Chapter 12: Addenda

ritme. We hebben gebruik gemaakt van gevalideerde vragenlijsten, van activiteit als maat van ritme, maar niet van conventionele circadiane maten. Idealiter wordt de circadiane ritmiek met een melatonine profiel onderzocht, om zo de directe output van de biologische klok te zien. Hoewel een melatonine profiel betrouwbaar en relatief makkelijk verkrijgbaar is via het speeksel, maken de kosten en de noodzaak om het in het donker te verkrijgen het niet bruikbaar om dag tot dag fluctuaties in de klinische praktijk te bestuderen. De gebruikte vragenlijsten zijn echter goed gevalideerd op melatonine profielen. Slaap zou idealiter gemeten worden met polysomnografie, de gouden standaard om slaap te bepalen. Ook dit is niet bruikbaar in de klinische praktijk, vanwege de invloed die het heeft op de nachtrust zelf. Actigrafie is daarom een goed alternatief, en specifiek ook gevalideerd binnen patiënten met de bipolaire stoornis met polysomnografie.

In dit proefschrift zijn alleen objectieve maten van slaap en ritme gebruikt. Hoewel in sommige studies subjectieve maten wel aanwezig waren, bijvoorbeeld in hoofdstuk 7, hebben we gekozen om nu alleen de objectieve slaapmaten te onderzoeken om de volgorde van symptomen te bestuderen. Dit vooral omdat de interpretatie van een patiënt erg relevant kan zijn. Zo kunnen, bij een opkomende manische episode, patiënten een goede slaapkwaliteit en een uitgerust gevoel rapporteren, terwijl er juist grote afwijkingen worden gezien in de objectieve slaapmaten.

Medicatiegebruik in patiënten met stemmingsstoornissen vormt een uitdaging voor ge- degen onderzoek. Zo zagen we dat het verschil in social jetlag tussen patiënten en controlepersonen wegvalt wanneer we corrigeren voor medicatiegebruik. In de studies naar de bipolaire stoornis was het corrigeren voor medicatiegebruik nagenoeg onmogelijk, omdat patiënten veel verschillende types medicatie, met verschillende werkingsme- chanismen, gebruiken. Het blijft wel belangrijk te realiseren dat medicatiegebruik voor dit type onderzoek belangrijk is, zo heeft lithium een duidelijk effect op het circadiane ritme en wordt zelfs gedacht dat het therapeutische effect hiermee samenhangt.

Tenslotte is het belangrijk te realiseren dat in het laatste deel van dit proefschrift, de studies naar winterdepressie, data uit verschillende studies gecombineerd werden. Dit zorgde voor een grote dataset, maar de studies waren niet specifiek opge- zet voor het onderzoeken van onze onderzoeksvragen.

Vervolgonderzoek Dit proefschrift laat zien dat hoewel ambulante metingen, zoals actigrafie en een stemmingsdagboek, al decennia gedaan worden, de technische ontwikkelingen nu veel nieuwe onderzoeksmogelijkheden bieden. De studies in dit proefschrift hebben de data geanalyseerd nadat de patiënt het onderzoeksmateriaal heeft in- geleverd. Met andere woorden, nadat slaap- en stemmingsproblemen zich hebben voorgedaan. De technische ontwikkeling, specifiek ook de commerciële producten die activiteit meten, laat zien dat het een kwestie van tijd is tot we patiënten in ‘real-time’ kunnen analyseren. Bij de bipolaire stoornis wordt ook gezien dat 86% van de patiënten geïnteresseerd is in het gebruik van innovatieve producten voor

144 Nederlandse Samenvatting

Chapter 12: Addenda

zelfmanagement. Hiervoor zijn nog wel een aantal technische ontwikkelingen nodig, met name in opslagcapaciteit en batterij duur van deze meetapparatuur. Het belangrijkste is echter de validatie van de commerciële producten, die gezien de kwaliteit nu vaak nog te wensen over laat. Als deze problemen opgelost zijn, is het wel duidelijk dat hier een grote kans ligt om het beloop van de bipolaire stoornis gunstig te beïnvloeden.

Naast het product om te meten zelf, wordt de analyse van de actigrafie data steeds sneller en gemakkelijker. In dit proefschrift werd de data geanalyseerd met een zelf- gemaakt script in het open-source softwarepakket R, dit script is publiek beschik- baar en kan (en mag) door iedereen gebruikt worden.

Klinische relevantie De verschillende studies in dit proefschrift hebben een duidelijke klinische relevantie. Zo maakt hoofdstuk 3 duidelijk dat er misschien geen focus nodig is op social jetlag bij patiënten met een depressieve stoornis, omdat daar niet zulke grote verschillen worden gezien tussen patiënten en controlepersonen als gedacht. Het gebrek aan verschil tussen patiënten en controlepersonen in hoofdstuk 5 op het gebied van variabiliteit en stabiliteit van het ritme in de bipolaire stoornis, geeft aan dat patiën- ten in staat zijn om een goed ritme te hebben. Dit is goed nieuws voor de patiënt, en een goed doel om na te streven voor de clinicus. De mogelijkheden die hoofdstuk 7 biedt in het zien aankomen van een episode is daarnaast voor de patiënt en de clinicus relevant, hoewel er nog een stap gemaakt moet worden om dit ook in real-time, in een prospectieve studie te zien gebeuren. Dat we een mogelijke biomarker hebben gevonden voor bipolaire stoornis in het fractale patroon kan de diagnostiek naar de bipolaire stoornis versnellen, wat hard nodig is. Het feit dat we in hoofdstuk 9 en 10 vonden dat één week lichttherapie, in de ochtend, onafhankelijk van chronotype voldoende was scheelt overbodige complexiteit in de behandeling van winterdepressie.

Conclusie Uit dit proefschrift kan geconcludeerd worden dat circadiane ritme problemen veelal samen met stemmingsproblemen komen. Deze problemen spelen een prominen- te rol in de ontwikkeling naar een episode in patiënten met bipolaire stoornis. Hoe- wel meer onderzoek nodig is kunnen de problemen in slaap en ritme mogelijk als voorspeller dienen voor een opkomende episode. Dit stelt de patiënt en clinicus in staat om een interventie toe te passen op het juiste moment. Verder werd getoond dat lichttherapie erg effectief is voor winterdepressie en dat een week licht in de ochtend voldoende is om het effect te bereiken.

Het doel van dit proefschrift was om de relatie tussen chronobiologie en stemmingsstoornissen te bestuderen en meer inzicht te geven in de kwetsbaarheid voor en de behandeling van stemmingsstoornissen. Hoewel er nog veel vragen overblijven en er meer vragen bijkomen over de relatie tussen Rhythm en Blues, is er zeker weer nieuw licht op de relatie verschenen.

145 “Hoewel het soms anders voelt als je alleen achter je computer zit te typen, schrijf je een proefschrift niet alleen. Hier wil ik de kans grijpen om een aantal mensen te bedanken.”

146 Dankwoord

Chapter 12: Addenda

Allereerst wil ik alle patiënten die hebben meegedaan aan de verschillende studies bedanken voor de inzet en het secuur registreren van alle data. De motivatie die de patiënten van de SWITCH-BD studie hadden om elke dag, gedurende een halfjaar, een dagboek in te vullen en daarnaast continu een actiwatch te dragen is bewon- deringswaardig en hebben als inspiratie gediend om hard door te werken. Specifiek wil ik ook de patiëntenvereniging Plusminus bedanken voor hun medewerking en specifiek regio Noord.

Aan begeleiding had ik niets te kort.

Beste Rixt, via een onderzoek waar ik als gezond proefpersoon aan mee deed kwam ik met je in aanraking. Na te vertellen dat ik interesse had in slaap is het balletje gaan rollen. Via een omweg bij de biologie ben je mijn copromotor geworden. Ik heb veel van je mogen leren, niet alleen inhoudelijk, maar ook logistiek rondom onderzoek en natuurlijk de eindeloze activistische passie over onderwerpen die niets met het onderzoek te maken hadden waarmee onze meetings steevast begonnen.

Beste Robert, als promotor heb je een bijzondere stijl van leidinggeven, die ik pas snapte toen je me vertelde dat je het boek “Leidinggeven aan professionals: NIET DOEN” had gelezen. Ik heb niet alleen van de gegeven vrijheid veel geleerd, maar natuurlijk ook van de nodige bijsturing. Het geeft een hoop zekerheid om te weten dat je promotor het bijna altijd met je eens is. Je kijk op welk artikel waar heen moet, hoe een beursaanvraag bijgeschaafd moet worden en hoe je een bepaalde boodschap brengt zijn van grote waarde geweest. Bedankt!

Beste Marijke, je liet me de wereld van de chronobiologie en het slaap-waak onderzoek zien. Je enthousiasme over het vak en je waanzinnig scherpe blik ben ik erg gaan waarderen. Ik ben er trots op dat ik mijn stage wetenschap bij jou en Do- mien Beersma heb mogen doen en ik ben blij dat je (samen met Maan natuurlijk) mij als geneeskundestudent hebt willen begeleiden. Ik heb een hoop geleerd, en vooral geleerd hoe prettig het samenwerken met biologen is!

Beste Ybe, bedankt voor het eindeloos meeslepen naar alle SLTBR congressen. “Ja, dan moet je ook wel mee gaan naar alles van het congres, ook het diner.” De lichttherapie studies hebben daarnaast veel deuren geopend, en je gaf een mooi kijkje in hoe dit onderzoeksveld allemaal begonnen was al die tijd geleden. Ik kan me ook nog goed mijn eerste spam-mail herinneren die ik kreeg na mijn eerste publi- catie. Ik vroeg wat ik daar mee moest en je zei: “Ja Stefan, nu begint het, je wordt wereldberoemd!”. Dat klopte, want ik krijg inmiddels meerdere mails per dag!

Beste Fokko, voor het genetica paper ben ik je veel dank verschuldigd. Je kwam met het idee om chronotype aan circadiane genen te linken. Het tomeloze enthousiasme van je werkt aanstekelijk, en de mogelijkheden die je ziet met de data ook. Beste Ilja, bedankt voor de statistische begeleiding, erg prettig om even kort mijn script te kunnen checken als ik weer in paniek was en iets niet snapte in de complexe snptest software.

147 Dankwoord

Chapter 12: Addenda

Beste Harriëtte, bedankt voor de samenwerking in de begeleiding van Stella, en ook bedankt voor de letter die we samen hebben geschreven. Tijdens het schrij- ven raakte ik af en toe wat gefrustreerd, maar dat gaf alleen maar aan hoeveel ik nog te leren had. De frustratie en het leerproces kwam vooral omdat je recht tegenover mij staat in de eindeloze perfectionisme en check-check dubbelcheck. Achteraf kan ik er erg om lachen en het vooral erg waarderen!

Beste Marieke, bedankt dat ik je hulp laat in het promotietraject heb mogen in- schakelen en dat ik volledig in je onderzoeksgroep werd betrokken. Je hulp bij hoofdstuk 7 heeft erg geholpen, vooral ook omdat ik daarbij hulp kreeg van Eve- lien Snippe. Evelien, je energie en inzicht in deze nieuwe methode werkt aanstekelijk en ik ben blij dat we deze methode samen hebben mogen opschrijven!

Supervisors from the Division of Sleep and Circadian Disorders at Harvard, professor Frank Scheer and dr. Kun Hu, thanks for having me over in Boston and supervising the fractal paper. I learned a lot over there, not only in how you work, but especially how both of you think, which was very enlightening. The rest of the Medical Chronobiology Program and Medical Biodynamics Program, thanks for having me there as well!

Geachte leden van de leescommissie, bedankt voor het beoordelen van mijn proefschrift. Dear professor Wirz-Justice, dear Anna, thanks for being in the reading committee of this thesis, and for all the inspiring words at the different SLTBR meetings. Your papers are probably the biggest inspiration for the work done in this thesis. Geachte professor van Someren, beste Eus, bedankt voor het lezen van het proefschrift en bedankt voor het delen van de kennis over actigrafie. Ook bedankt voor het contact met de groep van professor Scheer in Boston. Geachte professor Nolen, bedankt voor de kritische blik.

Co-auteurs van de Bipolar Genetics studie, Sanne Verkooijen en Marco Boks, bedankt voor de samenwerking rondom het project. Ook de co-auteurs op de NESDA papers, professor Brenda Penninx en Niki Antypa, bedankt voor het meeschrijven.

Yoram en Ando, bedankt voor de samenwerking aan het ACTman project. Hoewel het maken van een geautomatiseerd script om actigrafie data te analyseren misschien wel meer tijd heeft gekost dan het zelf berekenen, zal dit in de toekomst van grote waarde zijn!

Alle mede PhD-studenten en andere onderzoekers in het ICPE, bedankt voor de samenwerking en nodige afleiding. Karlijn, het was een goed begin samen, tegelijk bij Robert beginnen. Kamergenootjes, Petra, Ymkje Anna, Annelene, Annelies, Huifang en Astrid bedankt voor de goede tijd. Stella, mooi dat je inmiddels als MD-PhD’er het chronobiologie onderzoek binnen de psychiatrie voort kan zetten!

Nog een speciale dank aan alle ondersteunende personen bij de psychiatrie en het ICPE. Paulien, Margo, Esther, Martha en Jaap ik heb soms het gevoel meer

148 Dankwoord

Chapter 12: Addenda

vragen aan jullie gesteld te hebben dan aan mijn officiële begeleiders. Bedankt voor het aankunnen van een MD-PhD student die zo ontzettend vaak van kliniek naar onderzoek wisselt.

Beste collega’s uit het Reinier de Graaf Gasthuis, afdeling neurologie. Het is mooi om vanuit het onderzoek direct volle bak in de kliniek te mogen beginnen. Collega’s arts-assistenten, bedankt voor de lol die we samen kunnen hebben tussen het har- de werken door. Neurologen, bedankt voor het laten zien van een ontzettend mooi vakgebied en de dagelijkse begeleiding.

Bedankt ook aan de kliertjes uit Ugchelen, voor de nodige relativering over het onderzoek. Specifiek Diederick Oosterhoff, ook al snap je er niet veel van, de support wordt gewaardeerd. Bedankt ook dat je mijn paranimf wil zijn, je ziet vanzelf wat je moet doen.

Dan de vrienden uit Groningen die ik vanaf dag 1 Geneeskunde had, Willem Balder, Thom Bongaerts, Koen De Decker, Paul Schoonbeek en Sjoerd van den Berg, bedankt voor de talloze avonden eten, de wijnwoensdagen en alle vakanties en andere manieren van lol. Willem helemaal bedankt voor de steun tijdens het MD-PhD tra- ject wat we tegelijk deden. Mooi dat je op de dag van de verdediging naast me wil staan als paranimf.

Familie, pap en mam, bedankt dat ik van jullie heb geleerd dat als je meer kan, je er ook meer uit moet halen. Bedankt voor de belangstellende vragen. Ik weet dat jullie trots zijn, en dat motiveert altijd weer. Vera en Lisanne bedankt voor de gezelligheid en geklier altijd als we samen thuis zijn. Leuk dat we met z’n allen op vakantie gaan de dag na mijn promotie. Bedankt ook aan de schoonfamilie voor de ondersteuning, en ook bedankt voor het commentaar dat ik nog altijd niet klaar ben met mijn studie..

Lieve Sandra, bedankt voor de eindeloze ondersteuning, het nalezen en vooral ook bedankt voor het ongenuanceerd zeggen dat dingen beter kunnen. Bedankt voor de steun in elke nieuwe stap die we samen aangaan. Ik ben vooral blij dat ik me met jou nog geen moment verveel, ondanks dat we elkaar inmiddels langer wel dan niet kennen.

149 List of publications

Chapter 12: Addenda

Publications Knapen, S.E., van de Werken, M., Gordijn, M.C.M., & Meesters, Y. (2014). The duration of light treatment and therapy outcome in seasonal affective disorder. Journal of Affective Disorders, 166, 343-346.

Knapen, S.E., Gordijn, M.C.M., & Meesters, Y. (2016). The relation between chronotype and treatment outcome with light therapy on a fixed time schedule. Journal of Affective Disorders, 202, 87-90.

Knapen, S.E., Riemersma-van der Lek, R.F., Haarman, B.C.M., & Schoevers, R.A. (2016). Coping with a life event in bipolar disorder: ambulatory measurement, signalling and early treatment. BMJ Case Reports.

Verkooijen, S., van Bergen, A.H., Knapen, S.E., Vreeker, A., Abramovic, L., Pagani, L., Jung Y, Riemersma-van der Lek R.F., Schoevers, R.A., Takahashi, J.S., Kahn, R.S., Boks, M.P.M., Ophoff, R.A. (2017). An actigraphy study investigating sleep in bipolar I patients, unaffected siblings and controls. Journal of Affective Disor- ders, 208, 248-254.

Knapen, S.E., Druiven, S.J.M., Meesters, Y., & Riese, H. (2017). Chronotype not associated with nonremission, but with current state? Sleep, 40(6).

Knapen, S.E., Riemersma-van der Lek, R.F., Antypa, N., Meesters, Y., Penninx, B.W., & Schoevers, R.A. (2018). Social jetlag and depression status: results obtained from the Netherlands Study of Depression and Anxiety. Chronobiology International, 35(1), 1-7.

van Bergen, A.H., Verkooijen, S., Vreeker, A., Abramovic, L., Hillegers, M.H., Spi- jker, A.T., Hoencamp, E., Regeer, E.J., Knapen, S.E., Riemersma-van der Lek, R.F., Schoevers, R.A., Stevens, A.W., Schulte, P.F.J., Vonk, R., Hoekstra, R., van Beveren, N.J., Kupka, R.W., Sommers, I.E.C., Ophoff, R.A., Kahn, R.S., Boks, M.P.M. (2018). The characteristics of psychotic features in bipolar disorder. Psychological Medicine, 1-13.

Druiven, S.J.M., Knapen, S.E., Penninx, B.W.J.H., Antypa, N., Schoevers, R.A., Riese, H., & Meesters, Y. (2019). Can chronotype function as predictor of a per- sistent course of depressive and anxiety disorder? Journal of Affective Disor- ders, 242, 159-164.

150 List of publications

Chapter 12: Addenda

Planned publications Kunkels, Y.K, Knapen, S.E., Zuidersma, M., Wichers, M., Riese, H., Emerencia, A.C. ACTman: Automated preprocessing and analysis of actigraphy data. (under review)

Druiven, S.J.M., Hovenkamp-Hermelink, J.H.M., Knapen, S.E., Kamphuis, J., Haarman, B.C.M., Penninx, B.W.J.H., Antypa, N.A., Meesters, Y., Schoevers, R.A., Riese, H. Stability of chronotype over a seven-year follow-up period and its association with severity of depressive and anxiety symptoms. (under review)

Knapen, S.E., Li, P., Riemersma-van der Lek, R.F., Verkooijen, S., Boks, M.P.M., Schoe- vers, R.A., Hu, K., Scheer, F.A.J.L. Fractal biomarker of activity in patients with bipolar disorder. (under review)

Knapen, S.E., Snippe, E., Smit, A.C., Schoevers, R.A., Wichers, M., Riemersma-van der Lek, R.F. The temporal order of sleep disturbances and mood changes before the transition to a mood episode in bipolar disorder.

Knapen, S.E., Verkooijen, S., Riemersma-van der Lek, R.F., Kahn, R.S., Ophoff, R.A., Boks, M.P.M., Schoevers, R.A. Circadian rhythm disturbances in bipolar disorder: an actigraphy study in patients, unaffected siblings and healthy controls.

151 Curriculum Vitae

Chapter 12: Addenda

Curriculum Vitae Stefan Erik Knapen was born on November 29th in Apeldoorn, the Netherlands. He graduated cum laude from the Christelijk Lyceum Apeldoorn in 2009, after which he started medical school at the Rijksuniversiteit Groningen. He was a board member of several student organizations, among which the largest (bio)medical student conference, ISCOMS, and the Faculty Council of the Medical Faculty. Stefan also developed an interest in research, resulting in a research clerkship at the Faculty of Science and Engineering studying melatonin patterns in patients with seasonal affective disorder under supervision of prof. dr. Domien Beersma and dr. Marijke Gordijn. Stefan prolonged this clerkship into a MD/PhD program, where he combined the last years of medical school with a PhD program at the Rijksuniversiteit Groningen at the Department of Psychiatry, under the supervision of prof. dr. Robert Schoevers and dr. Rixt Riemersma-van der Lek. In this PhD program he studied chronobiological mechanisms in patients with mood disorders. For a project in this PhD program Stefan was awarded the Ubbo Emmius Fonds Talent Grant and presented his research at multiple international conferences. Stefan visited Harvard Medical School as a research fellow under supervision of prof. dr. Frank Scheer and dr. Kun Hu. Stefan finished medical school in 2018 and started working as a resident not in training in the neurology department at the Reinier de Graaf Gasthuis in Delft. Stefan currently lives in Leiden, together with his girlfriend Sandra Müller and their two cats, Leia and Louie.