Objective Measures of Sleep Duration and Continuity in Major Depressive Disorder with Comorbid Hypersomnolence: A Primary Investigation with Contiguous Systematic Review and Meta-Analysis

SUMMARY

Hypersomnolence plays an important role in the presentation, treatment, and course of mood disorders. However, there has been relatively little research that examines objective measures of sleep duration and continuity in patients with depression and hypersomnolence, despite the use of these factors in sleep medicine nosological systems. This study compared total sleep time and efficiency measured by naturalistic actigraphic recordings followed by ad libitum polysomnography (without prescribed wake time) in twenty-two patients with major depressive disorder and co-occurring hypersomnolence (MDD-HYP) against age- and sex-matched healthy sleeper controls (HC). MDD-HYP demonstrated significantly longer sleep duration compared to HC quantified by sleep diaries, actigraphy, and ad libitum polysomnography. No between-group differences in sleep efficiency, latency to sleep, or wake after sleep onset were observed when assessed using objective measures. To further contextualize these findings within the broader scientific literature, a systematic review was performed to identify other comparable investigations. Meta-analysis of pooled data demonstrated patients with mood disorders and co-occurring hypersomnolence have significantly greater sleep duration and similar sleep efficiency compared to healthy controls when assessed using ad libitum polysomnography. These results suggest current sleep medicine nosology that distinguishes hypersomnia associated with psychiatric disorders primarily as a construct characterized by low sleep efficiency and increased time in bed may not be accurate. Future studies that establish the biological bases hypersomnolence in mood disorders, as well as clarify the accuracy of nosological thresholds to define excessive sleep duration, are needed to refine the diagnosis and treatment of these disorders.

INTRODUCTION

Sleep disturbance is a critical factor in the assessment, treatment, and course of mood disorders. A sizeable literature has established the bidirectional relationship between insomnia, defined as a difficulty initiating and/or maintaining sleep despite adequate opportunity, and depression (Krystal, 2012). However, much less attention has been devoted to the role of hypersomnolence, defined as excessive daytime sleepiness (EDS) and/or excessive sleep duration, in psychiatric disorders, despite a considerable proportion of patients with mood disorders experiencing these symptoms (Kaplan and Harvey, 2009). This area of research is also particularly salient to the clinical practice of sleep medicine, since patients with central nervous system (CNS) hypersomnias, such as narcolepsy and idiopathic hypersomnia, experience elevated rates of depressive symptoms, and practitioners are frequently tasked with segregating hypersomnolence associated with a psychiatric disorder from primary sleep pathologies (Dauvilliers et al., 2013).

The current nosological classification system in sleep medicine, the International Classification of Sleep Disorders (ICSD-3) emphasizes that insomnia disorder is comorbid with, rather than secondary to, psychiatric illness (American Academy of Sleep Medicine, 2014). This shift from prior classification schema reflects the now well-established bidirectional relationship between insomnia and psychiatric conditions (Krystal, 2012). This concept of comorbidity between sleep and psychiatric disorders is mirrored in the principal psychiatric nosology, the Diagnostic and Statistical Manual, recently updated to its fifth edition (DSM-5)(American Psychiatric Association, 2013). However, it is noteworthy that a conceptual difference still exists between the DSM-5 and ICSD-3 regarding hypersomnolence in psychiatric disorders, with the former considering hypersomnolence disorder as potentially comorbid with psychiatric conditions, while the latter delineates differences between hypersomnolence associated with psychiatric disorders and other CNS causes of hypersomnolence such as idiopathic hypersomnia (American Psychiatric Association, 2013; American Academy of Sleep Medicine, 2014).

The multiple sleep latency test (MSLT) is centrally important to ICSD-3 nosology in the discrimination between causes of hypersomnolence, with a mean sleep latency below 8 minutes a diagnostic criterion for the diagnosis of both narcolepsy and idiopathic hypersomnia (IH) (American Academy of Sleep Medicine, 2014). However, there are several limitations of the MSLT both in segregating narcolepsy from IH, and IH from psychiatric hypersomnolence. First, the test-retest reliability of the MSLT in patients with narcolepsy without cataplexy and idiopathic hypersomnia may be limited, with nearly 40% of patients demonstrating a change in mean sleep latency crossing the conventional threshold that defines excessive sleepiness (Trotti et al., 2013). Second, roughly 25% of patients with psychiatric hypersomnolence demonstrate mean sleep latency below 8 minutes, similar to the proportion of individuals in the general population who demonstrate high sleep propensity on the MSLT (Mignot et al., 2006; Plante, 2016). Finally, many patients with IH, particularly those with long sleep duration, do not demonstrate pathologic sleep latencies on the MSLT, with estimates of mean sleep latency in large studies of IH bordering the current pathological cutpoint of 8 minutes, further limiting the use of this test solely as a means of segregating CNS hypersomnias (American Academy of Sleep Medicine, 2014; Vernet and Arnulf, 2009).

In light of these observations, the ICSD-3, unlike prior editions (American Academy of Sleep Medicine, 2005), also allows objective measures of sleep duration, such as extended polysomnographic recordings and/or actigraphy, to be utilized to diagnose idiopathic hypersomnia (American Academy of Sleep Medicine, 2014). Additionally, the ICSD-3 suggests that findings on these measures are typically different between IH and hypersomnolence associated with psychiatric disorders. The ICSD-3 considers hypersomnolence associated with psychiatric disorders largely as an insomnia variant, with polysomnographic recordings typically demonstrating prolonged time in bed with fragmented sleep and low sleep efficiency, while prolonged continuous recordings demonstrate increased time spent in bed both day and night, a behavior referred to as clinophilia (American Academy of Sleep Medicine, 2014). These findings purportedly contrast with the frequently high sleep efficiency and total sleep times that may be observed in patients with idiopathic hypersomnia (American Academy of Sleep Medicine, 2014).

To help clarify the nature of hypersomnolence occurring in mood disorders, as well as the validity of current ICSD-3 nosological descriptions, the aims of this study were two-fold. First, we sought to evaluate sleep continuity and duration in patients with major depressive disorder (MDD) who also had significant complaints of co-occurring hypersomnolence using actigraphic and extended polysomnographic recordings. Second, we sought to systematically review the prior literature to determine whether current constructs in the ICSD-3 that describe hypersomnolence associated with mood disorders reflect the empiric evidence regarding sleep continuity and duration in these patients.

METHODS

Participants

Twenty-two consecutively recruited, unmedicated participants with unipolar Major Depressive Disorder and comorbid hypersomnolence (MDD-HYP) were recruited from the greater Madison, Wisconsin area as a part of a larger study evaluating MDD and sleep-related comorbidities. Potential candidates were initially screened via telephone, followed by in-person assessment to verify eligibility. In-person assessment included the Structured Clinical Interview for DSM-IV Axis I disorders (SCID) (First et al., 2002), semi-structured medical and sleep history, and physical exam performed by a physician board-certified in Psychiatry and Sleep Medicine (DTP). MDD was diagnosed via the SCID. Hypersomnolence was defined based on the operationalized criteria proposed by Ohayon and colleagues (Ohayon et al., 2012), that were ultimately adopted as the diagnostic criteria for hypersomnolence disorder in the DSM-5, with only minor modifications (American Psychiatric Association, 2013). These criteria were derived from a large-scale population-based assessment of excessive sleepiness in the general population (Ohayon et al., 2012). Specifically, participants had to report daytime sleepiness with at least one of the following: (1) recurrent periods of an irrepressible need to sleep within the same day, (2) recurrent naps within the same day, (3) a prolonged (>9 hours) main sleep episode each day that is nonrestorative (unrefreshing), or (4) unusual difficulty being fully awake accompanied by the feeling of being disorientated or confused. Additionally, hypersomnolence had to occur at least 3 times per week for at least three months despite a primary sleep period lasting at least 7 hours. Hypersomnolence symptoms also had to be accompanied by significant distress or impairment in cognitive, social, occupational, or other important areas of functioning. Finally, hypersomnolence could not be accounted for by another sleep or medical disorder. Age and sex-matched unmedicated healthy sleeper controls, who reported average sleep duration of 7–9 hours per night with no sleep-related complaints, were also recruited. Healthy sleepers were assessed with the SCID, and had no current or prior history of Axis I disorder.

Exclusionary criteria for all participants included the following: evidence of a clinically significant sleep or medical disorder that would cause hypersomnolence (e.g. obstructive sleep apnea, restless legs syndrome, periodic limb movement disorder, delayed sleep phase disorder, narcolepsy with cataplexy, etc.; confirmed by history and polysomnography, where applicable), history of significant head trauma or loss of consciousness > 30 minutes; regular psychotropic medication use within two weeks of initial visit; smoking of more than 15 cigarettes per day; >3 caffeinated beverages per day; or significant neurologic or medical illness; and imminent risk of self-harm or suicide. Women who were pregnant, breastfeeding, <6 months post-partum, or planning to become pregnant during the study were also excluded. Participants were also excluded if they met DSM-IV criteria for alcohol or substance abuse/dependence within the preceding 6 months.

In addition to clinical assessment, participants also completed the Beck Depression Inventory (BDI-II)(Beck et al., 1996), Epworth Sleepiness Scale (ESS)(Johns, 1991), Pittsburgh Sleep Quality Index (PSQI)(Buysse et al., 1989), and Hypersomnia Severity Index (HSI) (Kaplan et al., 2015a). All participants provided informed consent and were instructed to maintain their usual sleep-wake schedules for the duration of their time in the study. This study was approved by the Institutional Review Board of the University of Wisconsin-Madison, and all participants provided written informed consent.

Sleep Diaries and Actigraphy

During the period between in-person screening and in-laboratory polysomnography, participants completed daily sleep diaries supplemented by wrist-worn actigraphy (Actiwatch 2, Philips Respironics). Sleep logs assessed self-reported sleep-wake variables including time in and out of bed, estimated sleep onset latency, number and duration of nocturnal awakenings, time of final awakening from nighttime sleep, as well as number and estimated duration of daytime naps. Actiwatch data were analyzed offline utilizing the medium threshold with five-minute immobility time for sleep onset/offset, as these settings have been shown to have the strongest correlation with polysomnographic variables (Chae et al., 2009). The onset and offset of rest intervals were determined using rest interval markers (participant-delivered) in the actogram. Participants were instructed to activate a rest interval marker when they were attempting to fall asleep and when they were getting out of bed (both for nocturnal sleep and nap episodes). If a rest interval marker was not available, then the beginning and end of the interval was estimated using sleep diaries and visual inspection of the actogram. Participants were asked to document approximate times and duration of daytime naps on sleep logs, which were subsequently used to identify nap episodes; thus there was no minimum inactivity duration used to define naps. Individual days in which the participant did not wear or improperly utilized actigraphy, or failed to complete a sleep diary, were excluded from analyses. Nocturnal total sleep time (TST), time in bed (TIB), wake after sleep onset (WASO), sleep onset latency (SOL), and sleep efficiency (SE) were derived from sleep diary and actigraphic data, as well as frequency and duration of naps.

Ad Libitum Polysomnography

Participants returned to the sleep center approximately two weeks after eligibility screening (minimum 1 week, maximum 4 weeks) for ad libitum overnight polysomnography. Subjects were allowed to go to bed as close to their typical bedtime as feasible. Importantly, all participants were minimally disturbed and allowed to sleep ad libitum, and thus were not awoken at a pre-specified time the following morning. Polysomnographic data were collected using an integrated recording system utilizing a 256-channel EEG net (Electrical Geodesics, Eugene, OR) along with other standard recording sensors including electrooculogram (EOG), sub-mental electromyogram (EMG), electrocardiogram (ECG), bilateral tibial EMG, respiratory inductance plethysmography, pulse oximetry, and a position sensor (Alice® Sleepware; Philips Respironics, Murrysville, PA). A registered sleep technologist staged all sleep recordings according to standard criteria based on 6 EEG channels at approximate 10–20 locations (F3, F4, C3, C4, O1, and O2) referenced to the mastoids, electrooculogram, and sub-mental electromyogram (Berry et al., 2012).

Statistics

Demographic, polysomnographic, actigraphic, and sleep diary data were compared between groups using two-tailed, unpaired Student’s T-tests. Bland-Altman analysis was utilized to compare total sleep time and sleep efficiency measured subjectively (sleep log) and objectively (actigraphy) prior to polysomnography. Alpha equaled 0.05 for statistical significance.

Systematic Review and Meta-Analysis

To contextualize our findings within the previously published literature, we performed a systematic review and meta-analysis. Relevant studies were identified through searches of PubMed and Psychinfo performed on April 13, 2016 using the following search terms: (psychiatr* OR mood OR depress* OR dysthym* OR bipolar) AND (hypersom* OR sleepiness OR atypical) AND (polysomnogra* OR polygrap* OR EEG OR electroenceph* OR actigra*). Waterfall and ancestral searches were also performed. There were no limitations on year of publication or language of article, with translations performed on an as needed basis utilizing native-language speakers or Google Translate (http://translate.google.com). For inclusion, studies had to utilize an objective measure of sleep duration/continuity (polysomnography or actigraphy), evaluate mood-disordered subjects with hypersomnolence, and include a healthy comparison group. Study quality was assessed (unblinded) using the Methodological Index for Non-Randomized Studies (MINORS) rating scale (Slim et al., 2003).

The primary variable of interest assessing sleep duration was total sleep time, measured by ad libitum polysomnography (or actigraphy). Since it was anticipated a priori there would be relatively few studies with ad libitum recordings, the primary sleep continuity variable considered was sleep efficiency, derived by polysomnographic recordings that utilized either fixed or flexible sleep-wake times. Meta-analysis was performed using random-effects model (DeSimonian-Laird), utilizing OpenMetaAnalyst, an open-source, cross-platform software (http://www.cebm.brown.edu/open_meta/) (Wallace et al., 2012). Heterogeneity among studies was evaluated using I² with cutoffs of 0%, 25%, 50%, and 75% defining no, low, moderate, and high heterogeneity (Higgins and Thompson, 2002, Higgins et al., 2003).

RESULTS

Participants

Demographic data for participants is detailed in Table 1. As expected, MDD-HYP participants had significantly greater symptoms of depression, hypersomnolence, and sleep disturbance as compared to age and sex-matched healthy controls (HC).

Table 1.

Demographic and Characteristic Data

	MDD-HYP	HC	p*
	N=22	N=22
Age	28.1 (5.8)	28.4 (5.6)	0.88
Sex (F/M)	18/4	18/4	-
BMI (kg/m2)	24.5 (4.9)	25.2 (3.9)	0.65
BDI	22.3 (7.4)	1.1 (1.7)	<0.0001
ESS	12.4 (2.6)	4.6 (2.3)	<0.0001
HSI	21.0 (4.2)	3.4 (2.2)	<0.0001
PSQI	5.7 (2.0)	1.8 (1.4)	<0.0001

MDD-HYP, major depressive disorder with comorbid hypersomnolence; HC, healthy control; BDI, Beck Depression Inventory; ESS, Epworth Sleepiness Scale; HSI, Hypersomnia Severity Index; PSQI, Pittsburgh Sleep Quality Index. Values are displayed as mean (standard deviation).

^*p-value derived using 2-tailed, independent samples t-tests.

Significant items marked in bold.

Actigraphy and Sleep Diaries

Using sleep diaries, MDD-HYP participants subjectively reported both increased average nocturnal TST and TIB relative to HC, by approximately 42 and 52 minutes respectfully (Table 2). Similarly, they reported higher proportions of days with nocturnal TST and TIB above 9 hours (Table 2). Notably, MDD-HYP also reported significantly longer SOL and lower SE than controls, however, average values for both groups fell into ranges that were not suggestive of clinically significant insomnia (Table 2). MDD-HYP participants also had significantly greater average daily nap time than HC, with a trend towards a higher proportion of days with at least one nap (Table 2).

Table 2.

Sleep Diary Data.

	MDD-HYP	HC	p*
Nocturnal Sleep
TST (min)	498.0 (42.6)	456.0 (41.9)	0.002
TIB (min)	527.4 (43.9)	475.2 (42.7)	0.0003
WASO (min)	11.7 (7.9)	7.5 (9.8)	0.12
SE (%)	94.0 (2.5)	96.0 (2.3)	0.01
SOL (min)	18.0 (9.4)	11.3 (5.4)	0.006
TST>9 hours (%)	34.7 (24.0)	12.8 (12.4)	0.0006
TIB>9 hours (%)	43.4 (25.1)	20.9 (17.5)	0.001
Naps
% days with naps	19.8 (18.2)	11.6 (10.6)	0.07
Nap time/day (min)	14.8 (13.1)	6.8 (8.6)	0.02

MDD-HYP, major depressive disorder with comorbid hypersomnolence; HC, healthy control; TST, total sleep time; TIB, time in bed, WASO, wake after sleep onset; SE, sleep efficiency (TST/TIB); SOL, sleep onset latency; TST>9 hours, percentage of days nocturnal TST exceeded 9 hours; TIB>9 hours, percentage of days nocturnal TIB exceeded 9 hours;% days with naps, total number of days with naps/days of observation; Nap time/day, cumulative amount of nap time/days of observation. Values are displayed as mean (standard deviation).

^*p-value derived using 2-tailed, independent samples t-tests.

Significant items marked in bold.

Two MDD-HYP participants were not included in analysis of actigraphic data due to device failure. Among the 42 participants with intact actigraphic recordings, both groups had similar numbers of days recorded prior to polysomnography (PSG) (MDD-HYP: 14.3±5.5 vs. HC: 13.8±5.2, p=0.73). Similar to sleep diary data, MDD-HYP participants had significantly greater average nocturnal TST and TIB, and greater proportions of nights with TST and TIB greater than 9 hours, relative to HC (Table 3). Unlike sleep diaries, there were no significant differences in any sleep continuity measures including SE, SOL, or WASO when measured using actigraphy (Table 3). MDD-HYP also had significantly more nap time per day than HC (Table 3).

Table 3.

Actigraphic Data.

	MDD-HYP	HC	p*
Nocturnal Sleep
TST (min)	457.8 (42.8)	422.0 (34.7)	0.005
TIB (min)	513.3 (46.0)	470.6 (35.3)	0.002
WASO (min)	40.7 (11.5)	36.4 (9.7)	0.21
SE (%)	89.3 (2.2)	89.7 (2.4)	0.59
SOL (min)	8.1 (4.3)	6.2 (3.7)	0.15
TST>9 hours (% of days)	14.3 (17.1)	5.1 (6.4)	0.03
TIB>9 hours (% of days)	35.8 (26.0)	13.3 (10.1)	0.001
Naps
% days with naps	15.6 (12.4)	10.0 (10.2)	0.12
Nap time/day (min)	11.4 (9.8)	5.7 (7.8)	0.04

MDD-HYP, major depressive disorder with comorbid hypersomnolence; HC, healthy control; TST, total sleep time; TIB, time in bed, WASO, wake after sleep onset; SE, sleep efficiency (TST/TIB); SOL, sleep onset latency; TST>9 hours, percentage of days nocturnal TST exceeded 9 hours; TIB>9 hours, percentage of days nocturnal TIB exceeded 9 hours;% days with naps, total number of days with with naps/days of observation; Nap time/day, cumulative amount of nap total sleep time/days of observation. Values are displayed as mean (standard deviation).

^*p-value derived using 2-tailed, independent samples t-tests.

Significant items marked in bold.

Bland-Altman analysis demonstrated both groups equally overestimated TST on sleep diaries relative to actigraphy (mean difference: MDD-HYP= 37.3 minutes vs. HC= 34.0 minutes, p=0.71) (Figure 1). Similarly, there was no significant between-group difference regarding mean difference of sleep efficiency, with both groups overestimating sleep efficiency on sleep logs relative to actigraphy (mean difference: MDD-HYP= 4.9% vs. HC= 6.3%, p=0.11).

Figure 1.

Bland-Altman plot for actigraphic (ACT) and sleep diary (SD) estimated A) total sleep time (TST) and B) sleep efficiency (SE) for MDD-HYP (blue triangles) and HC (red circles) groups. Black line estimates mean (solid) and 95% confidence interval (dashed) of difference for all participants; blue and red lines for MDD-HYP and HC groups, respectively.

Ad Libitum Polysomnography

Results of ad libitum polysomnography (PSG) are detailed in Table 4. Similar to actigraphic recordings, MDD-HYP participants exhibited significantly longer TST and TIB than HC, without significant differences in SOL, WASO, or SE. Sleep staging demonstrated that increased sleep time was the result of increases in total quantities of N1 and N2 sleep, without significant between-group differences in proportions of sleep stages for NREM and REM sleep (Table 4). Eleven MDD-HYP participants slept longer than 9 hours, while only 2 HC slept this duration during ad libitum PSG. Of these MDD-HYP participants, four slept longer than 10 hours, with two longer than 11 hours, while no HC slept longer than 9.26 hours on ad libitum PSG.

Table 4.

Polysomnographic Data.

	MDD-HYP	HC	p*
TST (min)	526.9 (98.4)	447.6 (63.0)	0.003
TIB (min)	602.7 (100.2)	511.9 (62.6)	0.001
WASO (min)	61.6 (36.1)	51.8 (29.8)	0.33
SE (%)	87.3 (6.5)	87.3 (7.0)	0.99
SOL (min)	14.8 (13.3)	13.8 (9.6)	0.77
N1 (min)	36.9 (19.6)	24.1 (12.3)	0.01
N2 (min)	308.4 (65.4)	260.1 (56.9)	0.01
N3 (min)	78.5 (31.9)	71.1 (31.5)	0.45
N1 (%)	7.0 (3.5)	5.3 (2.5)	0.08
N2 (%)	58.6 (6.9)	57.4 (8.5)	0.62
N3 (%)	15.2 (6.4)	16.3 (7.7)	0.61
REM (min)	104.6 (41.3)	95.1 (33.5)	0.41
REM (%)	19.2 (5.8)	21.0 (5.9)	0.32
REML (min)	125.5 (79.1)	125.0 (49.6)	0.98
AHI (#/hr)	0.88 (1.5)	0.63 (0.9)	0.51
RDI (#/hr)	2.7 (2.2)	2.5 (2.5)	0.85
PLMI (#/hr)	4.7 (4.7)	6.5 (11.3)	0.49
PLMAI (#/hr)	1.51 (1.3)	1.55 (1.8)	0.92
AI (#/hr)	11.9 (4.4)	11.4 (4.7)	0.70

MDD-HYP, major depressive disorder with comorbid hypersomnolence; HC, healthy control; TST, total sleep time; TIB, time in bed, WASO, wake after sleep onset; SE, sleep efficiency (TST/TIB); SOL, sleep onset latency; N1/2/3, NREM stage 1/2/3 (min &% of TST); REM, stage REM (min &% of TST); REML, REM latency (time from sleep onset to first REM sleep epoch);AHI, apnea-hypopnea index (apneas+hypopneas/hour); RDI, respiratory disturbance index (apneas+hypopneas+respiratory effort related arousals/hour); PLMI, periodic limb movement index; PLMAI, periodic limb movement arousal index; AI, spontaneous arousal index. Values are displayed as mean (standard deviation).

^*p-value derived using 2-tailed, independent samples t-tests.

Significant items marked in bold.

Meta-Analysis

The preferred reporting items for systematic reviews (PRISMA) flow diagram is presented in Figure 2 (Moher et al., 2009). After removal of duplicates, database and other searches identified 4,917 records, which were subsequently screened. Reasons full-text articles were excluded are detailed in Figure 2. Seven studies (in addition to the current study) were identified via systematic review for inclusion in meta-analyses (Dolenc et al., 1996; Hawkins et al., 1985; Hiyama, 1982; Plante et al., 2012; Quitkin et al., 1985; Thase et al., 1989; Vgontzas et al., 2000) (Table 5). Three of these studies had reports of ad libitum total sleep time, while seven either reported sleep efficiency or provided sufficient data to derive this variable. One ad libitum study did not report time in bed (Dolenc et al., 1996), and thus sleep efficiency could not be derived. Two studies from the same research group with sleep efficiency data were identified that utilized an identical control group (Hiyama, 1982, Shimizu et al., 1979), and to minimize bias, the most recent (and largest) study (Hiyama, 1982) was selected for inclusion. No studies that used actigraphy were identified that met inclusion/exclusion criteria.

Figure 2.

Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) flow diagram for systematic review and meta-analysis.

Table 5.

Table of Evidence. Total sleep time (minutes) and sleep efficiency (%) reported ± standard deviation.

Study	Study design (MINORS score)	Hypersomnolent Psychiatric Group (N, age, sex)	Healthy Controls (N, age, sex)	Ad Libitum TST (min; Mood disorder vs. HC)	Nocturnal Sleep Efficiency (%; Mood disorder vs. HC)	Notes
Dolenc et al. 1996	Case-control, dysthymic patients with a complaint of excessive night and daytime sleepiness (15)	12; median age 37y (range 23– 57); 7M:5F	12; median age 36y (range 15– 64); 5M:7F	553±83.1 vs. 524±94.4	NR	Ad libitum night held on second night in laboratory
Hawkins et al. 1985	Case-control, hypersomnolent unipolar MDD (18)	14; mean age 22.3±2.5y; 4M:10F	14; mean age 21.3±2.3y; 5M:9F	615.1±100.5 vs. 526.7±79.3	94.2±21.9 vs. 95.6±20.5	Reported ad libitum sleep data was 5th night in the research laboratory
Hiyama, et al. 1982	Case-control, 24-hr polygraphic recordings in hypersomnolent MDD (active and illness-free periods) versus controls (18)	10; mean age 32.7y (range 25– 25); 4M:6F	9; mean age 25.4y (range 23–37); 5M:4F	NR	72.4±17.1 vs. 81.6±8.0	Wake time from nocturnal sleep was 08:00 during 24-hr recording; overall 24-hr sleep duration increased in hypersomnolent MDD patients
Plante et al. 2012	Case-control, MDD with and without hypersomnia versus controls (16)	7; mean age 22.4±2.1; 3M;4F	7; mean age 22.0±1.3y; 3M;4F	NR	82.8±10.2 vs. 89.7±7.6)	Hypersomnia derived post-hoc from questionnaires; sleep not ad libitum
Quitkin et al, 1985	Case-control, Atypical MDD versus controls and endogenous depression (20)	26; mean age 37.7±10.6y; 9M;17F	21; mean age 35.1±13.0y; 10M:11F	385.8±56.0 vs. 372.4±39.3	94.0±5.3 vs. 92.2±6.9	Data averaged from 1–3 nights in the sleep laboratory; patients requested lights out and allowed to awaken spontaneously(Puig-Antich, et al. 1982)
Thase et al, 1989	Case-control, sleep EEG in patients with anergic bipolar depression versus controls (18)	26; 37.2±10.1y; 9M:17F	26; NR (“age and sex- matched)	NR	89.6±7.8 vs. 90.4±6.9	Two-night protocol in the sleep lab; patients had anergia, psychomotor retardation, and hypersomnia (≥1 hour more sleep than usual) or significant weight gain
Vgontzas et al. 2000	Consecutive patients with a chief complaint of EDS segregated into primary hypersomnia or psychiatric hypersomnia versus controls (18)	23; 39.9±11.5y (range 19–74); NR	50; 43.2±12.7; NR	NR	71.7±13.9 vs. 83.6±11.3	12/23 patients with psychiatric hypersomnia had mood disorder; 15/59 patients with primary hypersomnia had comorbid mood disorder (not included in sleep efficiency estimation)

EEG=electroencephalogram; F=female; HC=healthy control; M-male; MDD=major depressive disorder; NR=not reported; TST=total sleep time.

Quantitative meta-analysis of sleep duration on ad libitum polysomnography demonstrated hypersomnolent patients with mood disorders slept significantly longer than healthy controls (mean difference 48.5 minutes, 95% CI 7.3–89.7, p=0.02) (Figure 3). Overall heterogeneity was moderate among studies evaluating sleep duration (I²=62.7%). Pooled estimates of sleep efficiency did not demonstrate significant differences in sleep continuity between hypersomnolent participants and controls (mean difference −3.0%, 95% CI −6.9–0.8, p=0.12) (Figure 4). Overall heterogeneity was also moderate among studies evaluating sleep efficiency (I²=64.1%).

Figure 3.

Forest plot of mean difference of total sleep time (minutes) measured by ad libitum polysomnography in mood-disordered patients with hypersomnolence compared to healthy controls.

Figure 4.

Forest plot of mean difference of sleep efficiency (percent) measured by polysomnography in mood-disordered patients with hypersomnolence compared to healthy controls.

Exploratory analyses were performed to examine sources of heterogeneity for sleep duration and efficiency findings. One study accounted for the majority of heterogeneity among studies examining sleep efficiency (Vgontzas et al., 2000), while another study accounted for heterogeneity of sleep duration (Quitkin et al., 1985), as their omission from models decreased I² to 9.1% (low heterogeneity) and 0% (no heterogeneity), respectively. Stepwise omission of all other studies failed to reduce I² below 50% for any model tested. Results of meta-analysis without these studies further confirmed and strengthened findings of increased sleep duration among hypersomnolent patients with mood disorders (mean difference 69.8 minutes, 95% CI 35.3–104.3, p<0.001) without significant differences in sleep efficiency (mean difference −0.4%, 95% CI −2.7–1.9, p=0.74) compared to healthy controls.

DISCUSSION

Our results demonstrate several important findings that are pertinent to clinical, research, and nosological issues in sleep medicine. First, patients with hypersomnolence co-occurring with depression demonstrated significantly greater total sleep duration than healthy sleeper controls when measured objectively using both ad libitum polysomnography and actigraphy. Second, these objective increases in sleep duration occurred in the context of sleep efficiencies that were nearly identical to healthy sleepers. Third, our meta-analysis suggests that our findings are congruent with the prior literature, and suggests the conceptual framework regarding psychiatric hypersomnolence used in current sleep medicine nosology may benefit from reconsideration in light of these findings.

A significant strength of our investigation is the use of ad libitum polysomnography to assess sleep duration in the laboratory. The vast majority of clinical and research polysomnograms are not performed allowing the patient to go to bed and wake on their own accord, most likely due to logistical difficulties encountered when patients sleep well into the day. However, as demonstrated by these results and by prior studies, extended polysomnographic recordings are likely invaluable in quantifying hypersomnolence, particularly when it is comorbid with a mood disorder (Hawkins et al., 1985). In addition, our actigraphic measures substantiate that increased sleep duration in participants with depression and comorbid hypersomnolence relative to controls can also be documented outside of the sleep laboratory, and that results of polysomnography were not likely due to sleep deprivation prior to in-laboratory assessment. Our results are also congruent with prior reports that suggest patients with mood disorders and comorbid hypersomnolence may demonstrate increased sleep duration when measured by actigraphy (Bassetti et al., 2003; Kofmel et al., 2014).

Results of our primary investigation suggest that depressed persons with comorbid hypersomnolence on average do not demonstrate reduced sleep efficiency, as suggested by recent versions of the International Classification of Sleep Disorders (American Academy of Sleep Medicine, 2005; American Academy of Sleep Medicine, 2014). In fact, our pooled meta-analysis suggests that mood-disordered patients with hypersomnolence in aggregate do not have statistically different sleep efficiencies relative to healthy controls. Such findings are congruent with other prior investigations of hypersomnolent subjects with mood disorders that have demonstrated sleep efficiencies above 85% (Kofmel et al., 2014; Kupfer et al., 1972; Nofzinger et al., 1991), a common cut-point used to define sleep disturbance (Buysse et al., 1989). Moreover, studies of seasonal affective disorder, which is frequently associated with hypersomnolence (Rosenthal et al., 1984), have also demonstrated similar sleep efficiencies in these patients compared to healthy persons (Schwartz et al., 2000). Notably, the heterogeneity among studies in our meta-analysis that examined sleep efficiency was attributable to the results of one oft-cited study (Vgontzas et al., 2000). Although speculative, the variability between this study and other investigations included in our meta-analysis may be due to methodological differences relative to other studies, in particular the use of a mixed group of patients with psychiatric hypersomnolence, of which nearly 50% had forms of psychiatric disorders (somatoform, anxiety, and personality) other than depression, which may affect comparability of findings (Vgontzas et al., 2000).

Our results should not be construed to mean that all mood-disordered patients with a complaint of hypersomnolence will be free of sleep initiation and/or maintenance difficulties, as there are certainly individual patients who may present with complaints of daytime sleepiness and/or increased sleep duration (with or without a subjective desire to increase time in bed), who also have concomitant difficulties with sleep initiation and maintenance, as suggested by other prior actigraphic and polysomnographic studies (Kaplan et al., 2015b; Vgontzas et al., 2000). Nor should our results be interpreted as a criticism of the ICSD-3 itself, as we fully recognize the importance of (and complexities surrounding) nosological systems, which are ever-evolving tools that provide clinical guidance, consistency, and effective communication among practitioners and researchers. However, our findings do call into question the utility of using difficulties with sleep maintenance as a supportive finding of hypersomnolence associated with a psychiatric disorder. Thus, we would posit that the available data and literature suggests that any combination of excessive daytime sleepiness, insomnia, and increased sleep time can occur comorbid with mood disorders (Figure 5). In this context, the prevalence of any given combination of sleep complaints will vary depending on whether they are assessed subjectively or objectively, and what cutpoints are used to define pathologies on these measures.

Figure 5.

Heuristic model of types of sleep disturbance that may occur comorbid with mood disorders. Clinical and empiric evidence suggests any combination of excessive daytime sleepiness (EDS), increased sleep time, and insomnia (difficulty initiating and/or maintaining sleep) may occur in these patients, with prevalence of any given complaint or combination thereof dependent on the methodologies and thresholds used to define them. Dotted line denotes clinophilia may occur with or without a complaint of daytime sleepiness.

The delineation of non-narcoleptic disorders of hypersomnolence is a complicated issue, which is in part reflected in considerable differences between the ICSD-3 and DSM-5. Besides the emphasis on objective measures in the former compared to the latter, the quantity of sleep occurring in the context of daytime sleepiness that is considered pathologically excessive varies substantially between these nosologies. The DSM-5 considers greater than 9 hours of nocturnal sleep, when combined with daytime sleepiness, a diagnostic criterion for hypersomnolence disorder, while the ICSD-3 considers 11 hours or more of sleep per 24 hours measured by extended duration recordings (EEG, actigraphy) potentially diagnostic of IH (American Psychiatric Association, 2013; American Academy of Sleep Medicine, 2014). The DSM-5 is congruent with population-based assessments that demonstrate the proportion of individuals with impairment or distress related to diminished quality of wakefulness occurs in both persons who self-report sleeping below 6 and above 9 hours per day (Ohayon et al., 2013). A cutpoint of 11 hours of sleep to define excessive sleep duration is likely to sizably reduce the number of patients identified as having hypersomnolence, however, it is less likely to falsely identify patients without true pathology. Because the population prevalence of DSM-5-defined hypersomnolence disorder is roughly 1.5% (Ohayon et al., 2012), with estimates of the prevalence of IH poorly-defined but generally considered to be far lower, the nosological cutpoint that defines excessive sleep duration is likely to have significant effects on the clinical practice of sleep medicine. In light of this, further research that clarifies the optimal diagnostic thresholds to define CNS hypersomnolence is clearly indicated. However, because hypersomnolence disorders not related to hypocretin deficiency do not have well-established biological causes, such work will be quite challenging and, whenever possible, should emphasize empirical evidence to minimize bias.

There are limitations of our study which merit discussion. First, consistent with DSM standards that do not require objective measures of sleepiness, diagnoses of MDD and hypersomnolence were based on self-report, and we did not obtain MSLT in our participants. Thus, we do not know how many patients in our study may have also met ICSD-3 criteria for IH or narcolepsy, other than the two participants who slept longer than 11 hours on ad libitum overnight polysomnography who might meet current objective ICSD-3 criteria for IH, were it not for co-occurring MDD (American Academy of Sleep Medicine, 2014). However, while MSLT findings could potentially enhance our results, the use of an MSLT cutpoint to include or exclude participants could also significantly bias findings, since roughly 1 in 4 patients with psychiatric hypersomnolence will have a mean sleep latency below 8 minutes (Plante, 2016). Additionally, while it is conceivably possible that some patients included in our study had occult narcolepsy, this is less likely given the universal absence of cataplexy among participants and the absence of SOREMs on PSG. Moreover, had patients with narcolepsy inadvertently been included, this would have likely biased our results towards reductions in sleep efficiency relative to controls, given sleep fragmentation observed in both type 1 and 2 narcolepsy (Baumann et al., 2014, Roth et al., 2013), and thus our results are unlikely to have been substantially affected by this issue. Second, our ad libitum polysomnographic protocol occurred after a naturalistic observation period outside of the laboratory, but did not include repeated in-laboratory nights as utilized in some other investigations (Dolenc et al., 1996; Hawkins et al., 1985; Vernet and Arnulf, 2009), which may limit comparability between studies. However, our protocol had the advantage of more closely mirroring clinical practice and thus the generalizability of findings. Additionally, our protocol obviates potential iatrogenic sleep curtailment that can occur if a patient is awoken at an earlier time than is typical for him/her on the morning prior to ad libitum polysomnography (e.g. for other procedures such as the MSLT) (Dolenc et al., 1996; Vernet and Arnulf, 2009). Third, we did not have a separate group of hypersomnolent patients without comorbid mood disorders against which we could compare MDD-HYP participants. Such data would certainly help determine whether differences in sleep duration and quality exist between hypersomnolent patients with and without mood disorders, which would inform nosology and should be a focus of future research. However, it is noteworthy that quantities of nocturnal sleep observed on actigraphy and ad libitum polysomnography are comparable to those observed in prior studies that have examined IH patients without depression (Ali et al., 2009; Vernet and Arnulf, 2009). Finally, our meta-analysis must be interpreted with some caution given the relatively limited number of studies performed in this area of research, as well as the heterogeneity among studies.

In summary, we have demonstrated that patients with major depression and comorbid hypersomnolence have increased sleep duration relative to healthy sleepers, as well as similar objective measures of sleep efficiency. Meta-analysis also demonstrated that these results are congruent with the prior scientific literature. Thus, our results suggest that current ICSD-3 nosology concerning hypersomnolence associated with psychiatric disorders may require revision in future iterations to more optimally reflect the available scientific evidence. Further research is indicated to delineate the pathophysiology of disorders of hypersomnolence, to more rigorously define diagnostic boundaries as well as enhance our ability as a field to diagnose and manage these challenging disorders.