The Association Between Delayed Sleep-Wake Phase Disorder and Depression Among Young Individuals: A Systematic Review and Meta-Analysis: Association entre le syndrome de retard de phase et la dépression parmi les jeunes : revue systématique et méta-analyse

Abstract

Objectives

Delayed sleep-wake phase disorder (DSWPD) most commonly affects young individuals (adolescents and young adults), but it is often undetected in clinical practice. Despite several reports suggesting a link between DSWPD and depression, no systematic review has investigated this association. The aim of this systematic review was to determine whether DSWPD is associated with depression among young individuals.

Methods

MEDLINE, EMBASE, PsycINFO, and CINAHL Plus were searched up to 29 July 2024. Primary studies investigating DSWPD and depression among young individuals were eligible. Methodological quality and risk of bias was assessed with the National Institute of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. Common-effect models were conducted to examine the relationship between DSWPD status (categorical variable: yes or no) and depression severity (continuous variable). PROSPERO ID: CRD42023458889.

Results

Sixteen studies were included with 766 participants being evaluated against the diagnostic criteria for DSWPD from the International Classification of Sleep Disorders. Thirteen out of 15 studies demonstrated that young individuals with DSWPD had a significantly greater severity of depressive symptoms than young individuals without DSWPD. NIH quality assessment scores ranged between 5 and 9 (out of a total of 11). DSWPD status had a significantly large effect on depression severity in the common-effect model (N: 16 estimates, 693 participants, Cohen's d = 0.92, 95% confidence interval (95% CI) [0.76-1.08]). The subgroup analysis also demonstrated significant findings with the common-effect model that only utilized data from studies that controlled for psychiatric disorders (N: 12 estimates, 535 participants, Cohen's d = 0.88, 95% CI [0.70-1.06]).

Conclusions

DSWPD is associated with a greater severity of depressive symptoms among young individuals. Although more research is required to understand this association, it may be useful to consider the presence of DSWPD when managing young individuals who present with persistent sleep disturbances (e.g., sleep-onset insomnia) and depressive symptoms.

Introduction

Young individuals (i.e., adolescents and young adults) with depression are at a high risk of experiencing adverse psychosocial, behavioral, and vocational outcomes.^1,2 Various biopsychosocial factors can contribute to the development of depression among young individuals, including medical conditions like sleep disorders. ³ In particular, young individuals may be at a higher risk of developing a delayed sleep-wake phase disorder (DSWPD), due to their melatonin secretion occurring later at night compared to children and older adults. ⁴ Indeed, survey data suggests that DSWPD may affect 3.3 to 17.9% of young individuals,^5–8 which is far greater than the prevalence rate of 0.13 to 0.17% seen in the general population.^9,10

It is possible that the sleep disturbances produced by DSWPD may contribute to the development of depression in young individuals. The hallmark feature of DSWPD is the persistence of a major delay in sleep-wake times (e.g., >2 h) with an inability to fall asleep and wake up at socially desired times.^11,12 If young individuals with DSWPD must adhere to a sleep-wake schedule that does not match their endogenous circadian rhythm, this can predispose them to sleep-onset insomnia^11,12 (i.e., difficulty falling asleep), poor sleep quality, ¹² and excessive daytime sleepiness.^11,12 Each of these sleep disturbances have been individually linked to the development of depression,^3,13 and they may mediate the positive association between having a DSWPD and experiencing depression that has been reported in several observational studies.^14–26

Despite these reasons, no systematic review has investigated the link between DSWPD and depression among young individuals. Clarifying the presence and magnitude of this association can help identify whether DSWPD plays a prominent role in the development and presentation of depression in young individuals. Moreover, such work can provide useful knowledge to clinicians as they manage young individuals experiencing depression. Therefore, the aim of this systematic review was to determine whether DSWPD is associated with depression among young individuals. We hypothesized that young individuals with DSWPD would be more likely to experience depression than young individuals without DSWPD.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Review and Meta-analysis 2020 statement ²⁷ (Supplemental Table S1) and the Meta-analysis of Observational Studies in Epidemiology reporting guidelines ²⁸ (Supplemental Table S2). The protocol for this systematic review was registered with PROSPERO on 10 September 2023 (ID: CRD42023458889).

Search Strategy

MEDLINE, EMBASE, PsycINFO, and CINAHL Plus were searched from their respective inceptions until 29 July 2024 for articles on DSWPD. The specific search string employed in each electronic database consisted of key terms related to DSWPD: delayed sleep phase, delayed sleep wake phase, and delayed sleep wake cycle (for the specific search strings employed in each electronic database, please see Supplemental Table S3). A University of Toronto librarian (E.K.) was consulted during the development of the search strategy.

Selection Criteria

Study Design and Samples

Primary research studies comparing depression between young individuals with DSWPD to young individuals without DSWPD were eligible. Studies utilizing either a case-series or case-report study design were excluded, as they would not have a comparison group of non-DSWPD cases. To ensure that our findings were generalizable to young individuals in the general population, we also excluded reports in which study samples consisted of participants with a specific psychiatric (e.g., attention deficit hyperactivity disorder [ADHD]) or medical (e.g., diabetes mellitus) disorder. Moreover, studies composed of perinatal persons were excluded. We also excluded articles that we were unable to translate into English text, as all reviewers can only read English.

Definition of Young Individuals

Young individuals were defined as those who were 10 to 35 years old. This definition was informed by the American Psychological Association's (APA) definition of adolescence as approximately 10 to 19 years old ²⁹ and young adulthood as roughly 20 to 35 years old. ³⁰ In determining these age ranges, the APA has considered multiple aspects of development during this time of life including biological, cognitive, social, and personality factors.^29,30 Studies in which the mean age of either the DSWPD or non-DSWPD group was ≤9 or ≥36 years old were excluded. This exclusion criterion is similar to one employed in a previous systematic review examining the link between disturbed sleep and depression among children and youth (i.e., the authors excluded studies with samples having a mean age outside of 5 to 24 years old). ³¹

Assessment of DSWPD

Studies in which a trained member of the research team that assessed participants with the diagnostic criteria for DSWPD from one of the editions of either the International Classification of Sleep Disorders (ICSD) ¹¹ or the Diagnostic and Statistical Manual of Mental Disorders (DSM) ¹² were considered eligible. Please refer to Supplemental Tables 4-5 for the diagnostic criteria of DSWPD based on the ICSD 3rd edition and the DSM 5th edition text-revision, respectively, for more detail on the definition of DSWPD.

Studies using any other method (e.g., self-report or operationalizing survey data) were excluded, as they may not be able to utilize measures (e.g., wrist actigraphy) to demonstrate persistent major delays in the sleep-wake cycle of participants or to rule out other primary sleep disorders that share clinical features with DSWPD (e.g., chronic insomnia). Additionally, these studies may not be able to distinguish DSWPD from simply having an eveningness chronotype (i.e., a preference towards being more active at night as well as sleeping later at night and waking up later in the morning). ³² Although individuals with DSWPD typically demonstrate an eveningness chronotype, ¹¹ it can also occur among individuals without this sleep-wake disorder. ³²

Definition and Assessment of Depression

Studies examining depression (the outcome) either as a syndrome or severity of symptoms were considered eligible. The following definitions were used to define depression:

1. A syndrome of depression that is assessed either by using a structured diagnostic interview based on the DSM or International Classification of Diseases, or by using a symptom measure (e.g., Montgomery Asberg Depression Rating Scale (MADRS ³³ )) with a reliable and valid cut-off score;

   OR

2. As severity of depressive symptoms assessed by either a clinician-rated (e.g., MADRS ³³ ) or a self-reported (e.g., Beck Depression Inventory (BDI ³⁴ )) symptom measure.

Study Selection and Data Extraction

Four reviewers (M.H.D, V.K., J.M., and V.K.T.) independently screened titles/abstracts of the retrieved articles and then independently assessed their full-texts to determine if the study met the eligibility criteria. Two reviewers (M.H.D. and J.M.) then independently extracted the following study characteristics: author; country of study; study design; number of DSWPD and non-DSPWD cases; age and sex of DSWPD and non-DSWPD cases; diagnostic manual used to assess DSWPD; definition and assessment of depression; and results related to depression. For more detail on the study selection and data extraction process, please refer to the Supplemental Materials.

Assessment of Methodological Quality and Risk of Bias

Two reviewers (M.H.D. and J.M.) independently assessed the methodological quality and risk of bias of included studies using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. ³⁵ The NIH Quality Assessment Tool uses 14 items to evaluate the internal validity of cohort and cross-sectional studies by assessing their risk of selection, measurement (exposure and outcome), and confounding biases. ³⁵ For each item, the reviewer selected one of three responses: (1) Yes; (2) No; or (3) Other (either cannot determine, not reported, or not applicable (NA)). ³⁵ More Yes responses indicate better methodological quality and a stronger internal validity of the study. ³⁵

All included studies were given a response of NA to item 8 (for exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome), item 10 (was the exposure[s] assessed more than once over time?) and item 13 (was loss to follow-up after baseline 20% or less?). Items 8 and 10 received a response of NA because DSWPD does not vary in amount/level (i.e., it is defined as a categorical variable: yes or no) and does not need to be assessed again over time since it is a chronic disorder that persists for months to years.^11,12 A response of NA was given for item 13 because the link between DSWPD and depression was only examined during the baseline assessment of participants and not afterwards in any of the studies that were included in this review. Therefore, the included studies could only receive a maximum of 11 Yes responses.

To determine the risk of a confounding bias, we consulted literature on circadian rhythm and depression to identify for theoretical confounds. The following variables were considered to be relevant confounds in the association between DSWPD (a circadian rhythm disorder) and depression: (1) age^4,36; (2) sex^37,38; (3) comorbid psychiatric disorder ³⁹ (e.g., ADHD); (4) shift work^40,41; and (5) obesity.^42,43 If studies controlled for these confounds only through their study design (matching and/or exclusion criteria), we would still provide a Yes response for item 14 (which only considers statistical adjustment to control for confounds).

If there were any discrepancies in NIH quality assessment scores, the two reviewers discussed the issue until a consensus was reached. If not reached, the senior reviewer (V.B.) arbitrated the dispute.

Meta-Analysis

To investigate the effect of DSWPD on depression severity in young individuals, we conducted common-effect models using the metafor package in R (v.4.4.0). The independent variable was DSWPD status (categorical variable: yes or no) and the dependent variable was depression severity (continuous variable). Effect sizes were reported as Cohen's d, with 95% confidence intervals (95% CI) as its measure of uncertainty. Cohen's d > 0.2 is considered small, while >0.5 is moderate, and >0.8 is large. ⁴⁴ Forest plots were generated to visualize individual and pooled effect sizes.

Heterogeneity was assessed with I² statistic and funnel plots were visually assessed to evaluate for publication biases. We classified the degree of heterogeneity with the following I² ranges: (1) 0-40% as not important; (2) 30-60% as moderate; (3) 50-90% as substantial; and (4) 75-100% as considerable. ⁴⁴ If a moderate or higher degree of heterogeneity was present, the methodology of each study involved in the model was individually assessed to determine whether features in their study design may have contributed to the heterogeneity seen in our model (i.e., methodological heterogeneity). ⁴⁴ If so, these studies were removed from the model as a part of a sensitivity analysis. This method of identifying and managing heterogeneity is one of the strategies recommended by the Cochrane Collaboration. ⁴⁴

For our primary analysis, we used data from all of the included studies to model the association between DSWPD status and depression severity. We also conducted a subgroup analysis in which the common-effect model only utilized data from studies that controlled for psychiatric disorders. Meta-regression of data was not possible, as sample sizes of studies included in the models were not large enough to conduct this type of statistical analysis. Statistical significance was set at α = 0.05.

Results

Results of the Search Strategy

A total of 2272 records were identified from EMBASE, MEDLINE, PsycINFO, and CINAHL Plus (Figure 1). Following removal of duplicates, titles and abstracts of the remaining 1,191 records were screened to determine whether they compared DSWPD cases to non-DSWPD cases. Of these, 97 reports appeared to do so and their full-texts were subsequently assessed against our eligibility criteria. Overall, 81 reports were excluded (Supplemental Table S6), and 16 studies were included in our review (N = 16) (Supplemental Table S7).^{14–26,45–47}

Figure 1.

PRISMA flow chart.

Abbreviations: DSWPD, delayed sleep-wake phase disorder; ICSD, International Classification of Sleep Disorders; DSM, Diagnostic Statistical Manual of Mental Disorders; PRISMA, Preferred Reporting Items for Systematic Review and Meta-analysis.

Study Design and Sample Characteristics

Among the 16 included studies, there were 766 participants (415 with DSWPD and 351 without DSWPD) (Table 1). Five studies were conducted in Europe (United Kingdom and Norway),^{17,22,23,25,46} while 11 were conducted in either the United States of America,^15,26,45,47 Australia,^18–20,24 or Japan.^14,16,21 All 16 studies used a cross-sectional study design to examine the association between DSWPD and depression.^{14–26,45–47} Sample sizes of DSWPD groups ranged from 8 ⁴⁷ to 60, ¹⁶ while sample sizes for non-DSWPD groups ranged from 8 ⁴⁷ to 64. ¹⁶ The mean age in DSWPD groups ranged from 15.7 ²⁰ to 35.0 ⁴⁵ years and 15.9 ²⁰ to 34.2 ⁴⁵ years in non-DSWPD groups. Although, 15 out of 16 studies were in the age range of young adulthood (20 to 35 years old).^{14–19,21–26,45–47} Lastly, only six studies reported that ≥ 50% of participants in both groups were female.^{20,22,23,25,26,47} One study had no female participants, as they were excluded from their study design. ²⁴

Table 1.

Study Characteristics of Studies Eligible for Qualitative Synthesis (N = 16 Studies).

Study (origin)	Study design	DSWPD and non-DSWPD cases	Age (mean ± SD); female (%)	Diagnostic criteria for DSWPD (ICSD or DSM)	Definition and assessment of depression (syndrome or severity)	Results for depression	Relevant confounds controlled for (method of control)
Hirose et al. 14 (Japan)	Cross-sectional	17 DSWPD cases 17 non-DSWPD cases	DSWPD cases 23.8 ± 11.0 years; female (35.3%) Non-DSWPD cases 23.6 ± 10.9 years; female (35.3%)	ICSD-2	Severity: BDI score	DSWPD cases reported a significantly higher BDI score than non-DSWPD cases (8.6 ± 7.1 vs 3.8 ± 2.7, p = .01)	Age and sex (matched) Comorbid psychiatric disorder (excluded)
Joo et al. 15 (USA)	Cross-sectional	42 DSWPD cases 26 non-DSWPD cases	DSWPD cases 34.5 ± 10.8 years; female (52.4%) Non-DSWPD cases 33.4 ± 13.1 years; female (46.2%)	ICSD-3	Severity: BDI score	DSWPD cases significantly reported a higher BDI score than non-DSWPD cases (9.6 ± 7.3 vs 4.1 ± 4.6, p < .001)	Age and sex (matched)
Kanda et al. 16 (Japan)	Cross-sectional	60 DSWPD cases 64 non-DSWPD cases	DSWPD cases 28.0 ± 9.4 years; female (33.3%) Non-DSWPD cases 30.3 ± 3.4 years; female (29.7%)	ICSD-3	Severity: PHQ-9 score	DSWPD cases significantly reported a higher PHQ-9 score than non-DSWPD cases (18.1 ± 6.1 vs 12.6 ± 4.8, p < .001)	Age (matched) Shift work and comorbid psychiatric disorder (excluded)
MacMahon et al. 17 (UK)	Cross-sectional	22 DSWPD cases 20 non-DSWPD cases	DSWPD cases 21.8 ± 2.2 years; female (45.5%) Non-DSWPD cases 28.2 ± 10.1 years; female (55.0%)	Both DSM-IV and ICSD-R	Severity: BDI-SF score	DSWPD cases significantly reported a higher BDI-SF score than non-DSWPD cases in Tukey post-hoc test (4.0 ± 3.6 vs 0.5 ± 0.7, p < .05)	None
Marchetti et al. 46 (UK)	Cross-sectional	30 DSWPD cases 30 non-DSWPD cases	DSWPD cases 22.7 ± 3.56 years; female (49.0%) Non-DSWPD cases 23.2 ± 1.69 years; female (50.0%)	Both DSM-IV and ICSD-R	Severity: BDI-SF score	No statistically significant difference between DSWPD cases and non-DSWPD cases in BDI-SF score (2.3 ± 3.3 vs 2.9 ± 2.4, n.s.)	Comorbid psychiatric disorder (excluded participants scoring above cut-off markers for depression based on the BDI SF-4)
Micic et al. 18 (Australia)	Cross-sectional	26 DSWPD cases 18 non-DSWPD cases	DSWPD cases 21.9 ± 5.0 years; female (34.6%) Non-DSWPD cases 23.7 ± 5.1 years; female (44.4%)	ICSD-3	Severity: DASS-21 depression subscale score	DSWPD cases reported a significantly higher DASS-21 depression subscale score than non-DSWPD cases (16.77 ± 11.50 vs 4.78 ± 4.66, p < .001)	Comorbid psychiatric disorder, shift work, and obesity (excluded)
Micic et al. 19 (Australia)	Cross-sectional	16 DSWPD cases 14 non-DSWPD cases	DSWPD cases 21.1 ± 2.8 years; females (43.8%) Non-DSWPD cases 23.4 ± 5.9 years; females (50.0%)	ICSD-3	Severity: DASS-21 depression subscale score	DSWPD cases that are circadian entrained reported a significantly higher DASS-21 depression subscale score than non-DSWPD cases (11.6 ± 9.0 vs 4.5 ± 4.5, p < .05)	None
Reid et al. 45 (USA)	Cross-sectional	48 DSWPD cases 25 non-DSWPD cases	DSWPD cases 35.0 ± 11.4 years; female (56.3%) Non-DSWPD cases 34.2 ± 11.8 years; female (48.0%)	Both DSM-IV TR and ICSD-2	Syndrome: SCID Axis-I depressive disorder diagnosis	No significant difference in the proportion of current SCID Axis-I MDD (8.3% vs 16.0%, no p-value reported) or dysthymic disorder (4.2% vs 0.0%, no p-value reported) diagnoses between DSWPD cases and non-DSWPD cases	None
Richardson and Gradisar 20 (Australia)	Cross-sectional	33 DSWPD cases 40 non-DSWPD cases	DSWPD cases 15.7 ± 2.4 years; female (64%) Non-DSWPD cases 15.9 ± 2.4 years; female (75%)	ICSD-3	Severity: SMFQ score	Two-way ANOVA analysis demonstrated there was a significant difference in SMFQ scores between DSWPD cases with depressive disorder vs DSWPD cases without depressive disorder vs than non-DSWPD cases without depressive disorder (F = 56.9, p < .001); DSWPD cases without self-reported depressive disorder reported a significantly higher SMFQ score than non-DSWPD cases without depressive disorder in post-hoc test (7.6 ± 5.4 vs 3.0 ± 3.0, p ≤ .05); DSWPD cases with self-reported depressive disorder than non-DSWPD cases without depressive disorder in post-hoc test (15.4 ± 6.0 vs 3.0 ± 3.0, p ≤ .05)	Comorbid psychiatric disorder (ANOVA analysis—controlled for self-reported doctor diagnosed depressive disorder)
Shirayama et al. 21 (Japan)	Cross-sectional	22 DSWPD cases 20 non-DSWPD cases	DSWPD cases Females: 29.2 ± 5.1 years (10 females) Males: 25.5 ± 5.5 years (12 males) Non-DSWPD cases Females: 29.9 ± 3.3 years (10 females) Males: 28.0 ± 6.0 years (10 males)	ICSD	Severity: MMPI depression subscale score	Two-way ANOVA analysis demonstrated that DSWPD cases had significantly higher MMPI depression score than non-DSWPD cases (F = 16.806, p = .0002); Females: DSWPD cases scored significantly higher than non-DSWPD cases (27.9 ± 5.5 vs 20.5 ± 4.6); Males: DSWPD cases scored significantly higher than non-DWPD cases (26.1 ± 9.5 vs 17.1 ± 3.8)	Sex (ANOVA analysis) Comorbid psychiatric disorder (excluded)
Solheim et al. 22 (Norway)	Cross-sectional	20 DSWPD cases 16 non-DSWPD cases	DSWPD cases 24.8 ± 3.0 years; female (55.0%) Non-DSWPD cases 24.4 ± 3.4 years; female (75.0%)	Both DSM-IV and ICSD-3	Severity: BDI score	DSWPD cases reported significantly higher BDI score than non-DSWPD cases (10.8 ± 9.4 vs 3.4 ± 3.1, p = .006)	Age (matched) Comorbid psychiatric disorder (excluded)
Solheim et al. 23 (Norway)	Cross-sectional	9 DSWPD cases 9 non-DSWPD cases	DSWPD cases 22.5 ± 2.2 years; female (55.6%) Non-DSWPD cases 23.3 ± 2.4 years; females (55.6%)	DSM-IV and ICSD-2	Severity: BDI score	DSWPD cases reported a significantly higher BDI score than non-DSWPD cases (14.7 ± 11.7 vs 2.8 ± 2.9, p = .01)	Age and sex (matched)
Watson et al. 24 (Australia)	Cross-sectional	10 DSWPD cases 11 non-DSWPD cases	DSWPD cases 20.8 ± 2.4 years; female (0%) Non-DSWPD cases 22.4 ± 3.3 years; female (0%)	ICSD-3	Severity: BDI II score	DSWPD cases significantly reported a higher BDI II score than non-DSWPD cases (6.00 ± 3.00 vs 1.09 ± 1.14, p < .001)	Sex, comorbid psychiatric disorder, shift work, and obesity (excluded)
Wilhelmsen-Langeland et al. 25 (Norway)	Cross-sectional	40 DSWPD cases 21 non-DSWPD cases	DSWPD cases 20.7 ± 3.1 years; female (70.0%) Non-DSWPD cases 21.1 ± 2.2 years; female (71.4%)	ICSD-2	Severity: HADS depression subscale score	DSWPD cases significantly reported a higher HADS depression subscale score than non-DSWPD cases in unadjusted model (2.9 ± 2.3 vs 1.5 ± 1.8) even after controlling for vocational status (2.8 ± 0.3 vs 1.6 ± 0.5) in ANCOVA model (p = .0430)	Comorbid psychiatric disorder (moderate to severe psychological disorder) and shift work (excluded)
Wilson et al. 26 (USA)	Cross-sectional	12 DSWPD cases 12 non-DSWPD cases	DSWPD cases 31.1 ± 12.6 years; female (58.3%) Non-DSWPD cases 33.6 ± 15.5 years; female (50.0%)	ICSD-2	Severity: BDI score	DSWPD cases significantly reported a higher BDI score than non-DSWPD cases (8.00 ± 6.09 vs 1.83 ± 3.83, p = .007)	Age (matched) Comorbid psychiatric disorder and shift work (excluded)
Wyatt et al. 47 (USA)	Cross-sectional	8 DSWPD cases 8 non-DSWPD cases	DSWPD cases 22.9 ± 2.7 years; female (62.5%) Non-DSWPD cases 23.0 ± 2.6 years; female (62.5%)	ICSD-R	Severity: BDI-II score	There was no statistically significant difference between DSWPD cases and non-DSWPD cases for the BDI-II score (8.1 ± 11.7 vs 3.5 ± 2.2, n.s.)	Age and sex (matched) Comorbid psychiatric disorder and shift work (excluded)

Abbreviations: DSWPD, delayed sleep wake phase disorder; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, fourth edition; DSM-IV TR, Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision; ICSD, International Classification of Sleep Disorders; ICSD-R, International Classification of Sleep Disorders, Revised; ICSD-2, International Classification of Sleep Disorders, Second edition; ICSD-3, International Classification of Sleep Disorders, Third edition; HADS, Hospital Anxiety and Depression Scale; BDI, Beck Depression Inventory; BDI II, Beck Depression Inventory II; BDI-SF, Beck Depression Inventory Short Form; PHQ-9, Patient Health Questionnaire-9; DASS-21, Depression, Anxiety and Stress Scale 21 items; SCID, Structured Clinical Interview for the DSM-IV; SMFQ, Short Mood and Feelings Questionnaire; MMPI, Minnesota Multiphasic Personality Inventory; ANOVA, analysis of variance; ANCOVA, analysis of covariance; NIH, National Institute of Health.

Assessment of DSWPD

All 16 studies utilized an edition of the ICSD to diagnose participants with DSWPD.^{14–26,45–47} The 3rd edition of the ICSD was most commonly used (six studies)^{15,16,18–20,24} followed by the 2nd edition (three studies)^14,25,26 and 1st edition (two studies).^21,47 Five studies identified DSWPD cases with both the ICSD and DSM.^{17,22,23,45,46} No study solely used the DSM to assess DSWPD status among participants.

Assessment of Depression

Only one study explored depression as a syndrome, in which the Structured Clinical Interview for DSM-IV was used to diagnose participants with either a major depressive disorder, dysthymic disorder, or depressive disorder not otherwise specified. ⁴⁵

Depression was examined as severity of symptoms in 15 studies and each of these used a self-reported symptom measure to assess depression.^{14–26,46,47} The most common symptom measure used was the BDI (five studies),^{14,15,22,23,26} followed by two different iterations of it: BDI-II (two studies)^24,47 and BDI-Short form (two studies)^17,46 Other symptom measures included the Depression, Anxiety, and Stress Scale-21 (2 studies),^18,19 Short Mood and Feelings Questionnaire (one study), ²⁰ Patient Health Questionnaire-9 (one study), ¹⁶ Hospital Anxiety Depression Scale (one study), ²⁵ and the Minnesota Multiphasic Personality Inventory (one study). ²¹

Results Related to Depression

As for the one study examining depression as a syndrome, no significant associations were detected. ⁴⁵ However, 13 out of 15 studies showed that the DSWPD group had a significantly greater severity of depressive symptoms than the non-DSWPD group.^14–26

Methodological Quality Scores and Risk of Bias

Methodological quality scores ranged from 5 to 9 (out of total of 11), with nine out of 16 studies scoring 7 out of 11^{14,15,17–20,25,26,46} (Table 2). Of the 16 studies included in our review, 11 of these posed a risk of a selection bias, given that they were unable to report a participation rate ≥ 50%.^{15,16,18,19,21–23,25,26,45,47} Moreover, seven out of 16 studies did not indicate whether DSWPD and non-DSWPD groups were sampled from the same population or underwent the same inclusion/exclusion criteria.^{14,16,17,20,23,24,46} Lastly, 12 studies did not provide any justification for their sample size nor any calculations to determine the statistical power of their study sample.^{14–16,18–21,23–26,46}

Table 2.

National Institute of Health Quality Assessment Tool for Cohort and Cross-Sectional Studies Scores for Studies Eligible for Qualitative Synthesis (N = 16 Studies).

Author	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9	Item 10	Item 11	Item 12	Item 13	Item 14	Overall score (out of a total of 11)
Hirose et al. 14								N/A		N/A			N/A		7
Joo et al. 15			N/R					N/A		N/A			N/A		7
Kanda et al. 16			N/R					N/A		N/A			N/A		6
MacMahon et al. 17								N/A		N/A			N/A		7
Marchetti et al. 46								N/A		N/A			N/A		7
Micic et al. 18								N/A		N/A			N/A		7
Micic et al. 19			N/R					N/A		N/A			N/A		7
Reid et al. 45			N/R					N/A		N/A			N/A		8
Richardson and Gradisar 20								N/A		N/A			N/A		7
Shirayama et al. 21			N/R					N/A		N/A			N/A		6
Solheim et al. 22			N/R					N/A		N/A			N/A		8
Solheim et al. 23			N/R	C/D				N/A		N/A			N/A		5
Watson et al. 24								N/A		N/A			N/A		8
Wilhelmsen-Langeland et al. 25			N/R					N/A		N/A		C/D	N/A		7
Wilson et al. 26			N/R					N/A		N/A		C/D	N/A		7
Wyatt et al. 47			N/R					N/A		N/A			N/A		9

Abbreviations: Green, yes response; red, no response: N/R, not reported; C/D, cannot determine; N/A, not applicable.

Item 1: Was the research question or objective in the paper clearly stated? Item 2: Was the study population clearly specified and defined? Item 3: Was the participation rate of eligible persons at least 50%? Item 4: Were all the subjects selected or recruited from the same or similar populations (including the same time period)? Were inclusion and exclusion criteria for participants in the study prespecified and applied uniformly to all participants? Item 5: Was a sample size justification, power description, or variance and effect estimates provided? Item 6: For the analyses in this paper, were the exposure(s) of interest measured prior to the outcome(s) being measured? Item 7: Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed? Item 8: For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., categories of exposure or exposed measured as continuous variable) Item 9: Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? Item 10: Was the exposure (s) assessed more than once over time? Item 11: Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? Item 12: Were the outcome assessors blinded to the exposure status of participants? Item 13: Was loss to follow-up after baseline 20% or less? Item 14: Were key potential confounding variables measured and adjusted statistically for their impact on the relationship between exposure(s) and outcome(s)? Studies that controlled for confounding variables through study design techniques (matching and exclusion criterion) were also given a score of Yes for item 14.

Overall score is the total number of yes responses.

For measurement biases, 12 studies posed a risk of an outcome bias, as their participants may have been aware of their DSWPD status prior to completing their self-reported depression measures.^{14–17,19–23,25,26,46} Lastly, only three out of 16 studies posed a risk of a confounding bias, because they did not control for any of the relevant confounds identified for this review either in their study design or statistical analyses.^17,19,45 Among the 13 studies that did control for confounds, 11 of these adjusted for ≥2 confounding variables.^{14–16,18,21–26,47} The most common confound controlled for was comorbid psychiatric disorder (11 studies),^{14–16,18,20–22,24–26,46,47} followed by age (seven studies)^{14–16,22,23,26,47} sex (six studies),^{14,15,21,23,24,47} shift work (six studies),^{16,18,24–26,47} and obesity (two studies).^18,24 Ten studies excluded participants with psychiatric disorders from their study samples^{14,16,18,21,22,24–26,46,47} and one study adjusted for the presence of psychiatric disorders in their statistical model. ²⁰

Statistical Analysis

Primary Analysis

Among the studies included in our review, 15 reports provided 16 estimates to model the association between DSWPD status and depression severity (Supplemental Table S8). The common-effect model demonstrated that DSWPD status had a significantly large effect on depression severity with moderate to substantial heterogeneity (N: 16 estimates, 693 participants (DSWPD: 367 participants; non-DSWPD: 326 participants), Cohen's d = 0.92, 95% CI [0.76-1.08], I²= 53%) (Figure 2(a)). The funnel plot for the model was symmetrical for the most part with 2 outliers (Figure 2(b)).

Figure 2.

Meta-analysis and funnel plot of DSWPD status and depression severity among young individuals (adolescents and young adults). (a) Forest plot with individual and pooled effect sizes; (b) Funnel plot to assess for publication bias.

Abbreviations: DSWPD, delayed sleep-wake phase disorder; BDI, Beck Depression Inventory; BDI II, Beck Depression Inventory II; PHQ-9, Patient Health Questionnaire-9; BDI-SF, Beck Depression Inventory Short Form; DASS-21, Depression, Anxiety and Stress Scale 21 items; SMFQ, Short Mood and Feelings Questionnaire; MMPI, Minnesota Multiphasic Personality Inventory; HADS, Hospital Anxiety and Depression Scale.

Note. I² statistic ranges to assess the degree of heterogeneity: 0-40% as not important; 30-60% as moderate heterogeneity; 50-90% as substantial heterogeneity; and 75-100% as considerable heterogeneity.

Two studies were suspected to largely contribute to the heterogeneity in the model, since both of them used predefined cut-off values for their depression measures to exclude participants from their study samples.^24,46 As a part of a sensitivity analysis, we removed these two studies from the model. DSWPD status continued to have a significantly large effect on depression severity but now without any heterogeneity in the model (N: 14 estimates, 612 participants (DSWPD: 327 participants; non-DSWPD: 285 participants), Cohen's d = 1.02, 95% CI [0.85-1.19], I²= 0%) (Figure 3(a)). The funnel plot was more symmetrical with no outliers detected (Figure 3(b)).

Figure 3.

Sensitivity analysis of DSWPD status and depression severity in young individuals (adolescents and young adults). (a) Forest plot with individual and pooled effect sizes; (b) Funnel plot to assess for publication bias.

Abbreviations: BDI, Beck Depression Inventory; BDI II, Beck Depression Inventory II; PHQ-9, Patient Health Questionnaire-9; DASS-21, Depression, Anxiety and Stress Scale 21 items; SMFQ, Short Mood and Feelings Questionnaire; MMPI, Minnesota Multiphasic Personality Inventory; HADS, Hospital Anxiety and Depression Scale.

Note. Two studies were removed from the model since both of them used predefined cut-off values for their depression measures to exclude participants from their study samples,^24,46 while the other studies in the model did not.

I² statistic ranges to assess the degree of heterogeneity: 0-40% as not important; 30-60% as moderate heterogeneity; 50-90% as substantial heterogeneity; and 75-100% as considerable heterogeneity.

Subgroup Analysis

DSWPD status also had a significantly large effect on depression severity in the common-effect model that only utilized data from studies that controlled for psychiatric disorders, but with moderate to substantial heterogeneity (N: 12 estimates, 535 participants (DSWPD: 278 participants; non-DSWPD: 257 participants), Cohen's d = 0.88, 95% CI [0.70-1.06], I²= 62%) (Supplemental Figure S1(a)). The funnel plot for the model was symmetrical for the most part with two outliers (Supplemental Figure S1(b)).

As a part of a sensitivity analysis, we removed the two studies that used pre-defined cut-off values for their depression measures to exclude participants from their study samples^24,46 from the model. DSWPD status continued to have a significantly large effect on depression severity with no heterogeneity detected in the model (N: 10 estimates, 454 participants (DSWPD: 238 participants; non-DSWPD: 216 participants), Cohen's d = 1.01, 95% CI [0.81-1.20], I²= 0%) (Supplemental Figure S2(a)). The funnel plot for the model was more symmetrical with no outliers detected (Supplemental Figure S2(b)).

Discussion

To our knowledge, this is the first systematic review to explore the link between DSWPD and depression among young individuals. Among the included studies, 13 out of 15 studies showed that young individuals with DSWPD had a significantly greater severity of depressive symptoms than young individuals without DSWPD. However, the one study examining depression as a syndrome did not find any such associations between DSWPD status and having a depressive disorder. Furthermore, our common-effect models indicated that DSWPD status had a significantly large effect on depression severity, even in the models that only utilized data from studies that controlled for psychiatric disorders. Overall, these results support our hypothesis, by demonstrating that there is a consistent and strong relationship between having a DSWPD and experiencing a greater severity of depressive symptoms among young individuals.

Although 13 out of 15 studies demonstrated that having a DSWPD is significantly linked to more severe depressive symptoms among young individuals, two studies did not. The lack of significant findings in the first study may be related to the authors excluding participants who scored above a particular cut-off on their depression measure, ⁴⁶ therefore creating a ceiling effect. As a result, this does not allow an appropriate comparison between those with and without DSWPD regarding the severity of depressive symptoms. As for the other study, its small study sample size (16 participants) may not have had sufficient statistical power to detect a significant difference, although the DSWPD group scored higher than the non-DSWPD group ⁴⁷ (Table 1). As for the one study examining depression as a syndrome, the lack of significant associations seen in this study may be related to all participants in the non-DSWPD group having an eveningness chronotype, ⁴⁵ which itself has been shown to be independently linked to depression. ⁴⁸

The temporal relationship between DSWPD and depression is difficult to determine, because all of the studies included in our review examined this association cross-sectionally. Thus, this leads to each included study having the issue of reverse causality (i.e., rather than DSWPD causing depression, depression itself could have caused DSWPD). However, it is unlikely that our findings are explained by depression causing the development of DSWPD. All included studies used the ICSD to determine DSWPD status, which requires mental disorders (e.g., major depressive disorder) to be ruled out as the cause for the major sleep-wake delays seen in DSWPD cases. Other studies also support that DSWPD may precipitate depression. Indeed, one large survey of individuals with circadian rhythm sleep-wake disorders (CRSWD) reported that nearly 60% of participants began to experience depression after the development of their CRSWD. ⁴⁹ Nevertheless, prospective studies are still required to confirm whether DSWPD can produce depression over time and not the reverse.

The depressive symptoms seen in DSWPD may possibly be a specific phenotype of depression secondary to circadian disturbances. One observational study indicated that diurnal variation in mood (worse in the morning and better in the evening) and somatic features of depression (sleep disturbances, fatigue, and psychomotor retardation) are the most reported and persistent type of depressive symptoms seen among young individuals with DSWPD. ⁵⁰ Furthermore, this report described that most of those with DSWPD presented with moderate to severe depression, ⁵⁰ which is consistent with the large effect size noted in our meta-analysis. The collection of depressive symptoms seen among young individuals with DSWPD from this observational study somewhat resembles the presentation of depression in bipolar disorder. ⁵¹ This similarity may be related to the shared pathophysiological mechanisms (e.g., abnormal melatonin secretion profiles) ⁵² between DSWPD and bipolar disorder. Since the included studies only reported on the total score of their depression measures, we are unable to explore whether this phenotype of depression may be more common among those with DSWPD compared to those without DSWPD. Future research examining this association should also consider investigating the type of depressive symptoms that occur, to help clarify whether a specific phenotype of depression develops following the emergence of DSWPD.

It is likely that the association between DSWPD and depression is at least partly mediated by sleep-onset insomnia. This subtype of insomnia is common among individuals with DSWPD,^11,12 when they are supposed to adhere to a sleep-wake schedule that does not match their endogenous circadian rhythm. It is well established from longitudinal studies that sleep-onset insomnia can contribute to the onset of a depressive episode. ³ When present in a young individual who also has an eveningness chronotype (a typical trait of DSWPD), ¹¹ sleep-onset insomnia may increase depression risk by 12-fold. ⁵³ Therefore, future prospective studies examining the link between DSWPD and depression must also utilize validated insomnia (e.g., Insomnia Severity Index ⁵⁴ ) and chronotype (e.g., Morningness-Eveningness Questionnaire ⁵⁵ ; MEQ) measures to clarify the role and magnitude of effect that sleep-onset insomnia has in this association. Additionally, other sleep disturbances produced by DSWPD (i.e., poor sleep quality and excessive daytime sleepiness)^11,12 should also be investigated as mediators, as they have been independently linked to the development of depression. ¹³

In addition to the sleep disturbances produced by DSWPD, dysfunction in certain circadian systems may mediate the association between DSWPD and depression. For example, individuals with depression may be more likely to demonstrate a reduced and delayed phase of melatonin secretion than individuals without depression. ⁵⁶ Moreover, disruption in the rhythm of cortisol secretion may also be linked to the development of depression.^39,56 Abnormalities in the striatal dopaminergic tone may also mediate the link between DSWPD and depression, as this is an etiopathological mechanism that may be shared between persons with disrupted circadian rhythms ⁵⁷ and persons with depression. ⁵⁸

While most studies posed a low risk of a confounding bias, many of them were at risk of different types of selection biases. In particular, most reports were unable to indicate a participation rate ≥ 50%, suggesting that several of the samples in these studies may not be representative of the population of young individuals with DSWPD. This may also explain why most of these studies did not provide any justification for their sample sizes, as it may have been difficult for them to recruit participants. Several reports also posed a risk of an outcome bias, as their participants were likely aware of their DSWPD status prior to completing their self-reported depression measures. Also, none of the included studies removed sleep items from their depression measures, further increasing their risk of an outcome bias. To avoid these outcome biases, future prospective studies should have blinded assessors using clinician-rated measures (e.g., MADRS ³³ ) that have their sleep items removed (e.g., modified-MADRS ⁵⁹ ) in order to rate depression severity among participants.

Although more research is required to further understand this association, it may be useful to consider the presence of DSWPD when managing young individuals who present with persistent sleep disturbances (e.g., sleep-onset insomnia) and depressive symptoms. Unfortunately, nearly 80% of individuals with CRSWDs report being misdiagnosed initially with either a depressive disorder or chronic insomnia. ⁴⁹ Furthermore, these individuals often receive incorrect diagnoses for years prior to receiving the correct diagnosis of a CRSWD. ⁴⁹ In addition to clinical judgement, the use of reliable and valid chronotype measures like the MEQ can support clinicians in identifying possible cases of DSWPD. An MEQ total score ≤ 30 indicates that the individual has an extreme or definitive eveningness chronotype, ⁵⁵ which is a typical feature of DSWPD. ¹¹ Furthermore, using a chronotype measure allows clinicians to determine whether the social and vocational commitments of the individual (e.g., attending school in the early morning) are aligned with their eveningness chronotype. If misaligned, this can contribute to sleep-wake complications (e.g., poor quality of sleep or excessive daytime sleepiness) in the individual.

To differentiate DSWPD from other primary sleep disorders (e.g., chronic insomnia), guidelines recommend using sleep logs and/or wrist actigraphy to identify cases.^60,61 Wrist actigraphy is generally preferred as it may reduce patient burden compared to sleep logs. ⁶¹ In combination with monitoring one's mood, actigraphy may also be useful as a psychoeducation tool.

Once clinically diagnosed, guidelines recommend using measures to assess the rhythm of melatonin secretion (e.g., timing of dim light melatonin onset (DLMO) or urinary 6-sulfatoxymelatonin (aMT6s)) to confirm the presence of a delayed circadian phase ¹¹ and to help determine the time of day to administer evidence-based interventions. ¹² In choosing a measure, aMT6s may be more pragmatic, as DLMO may not be accessible in all healthcare settings and it may be more difficult to use in practice. It is also important to note that the results of these measures may be affected by the presence of comorbid mental disorders (e.g., MDD) ⁶² and the use of certain medications (e.g., serotonergic antidepressants). ⁶³ Phenotyping DSWPD patients with DLMO may also help in assessing their depression risk, since individuals with a circadian DSWPD phenotype (DLMO occurring at or after desired bedtime) may be more likely to present with moderate to severe depression than individuals with a noncircadian DSWPD phenotype (DLMO occurring before desired bedtime). ⁶⁴

As for evidence-based interventions, only strategically timed oral administration of melatonin^60,65 and scheduled light therapy ⁶⁵ have been recommended by guidelines to manage DSWPD. One small randomized controlled trial (RCT) suggests that exogenous melatonin treatment may also be useful in reducing depressive symptoms among DSWPD cases with depression. ⁶⁶ Although more research is required, agomelatine (a melatonergic agonist with antidepressant properties ⁶⁷ ) may be a potential therapeutic option in this patient population, as it has been shown to improve the sleep-wake rhythm among young individuals with DSWPD. ⁶⁸ However, there is a paucity of data to suggest whether its antidepressant effect would also be present in this patient population. Therefore, RCTs comparing agomelatine against placebo and active (e.g., exogenous melatonin treatment) controls among young individuals with DSWPD are needed. Such research will determine whether agomelatine can be a safe and effective therapeutic option to optimize the sleep-wake cycle in DSWPD cases, while also reducing their severity of depression.

In addition to the limitations described above, there are other issues to consider in this review. Firstly, the number of participants in the common-effect models is small and therefore reduces the generalizability of our findings. Furthermore, this small sample limited our ability to conduct meta-regression analyses to identify potential moderators (e.g., BDI measures vs other depression measures). Examination of comorbid mental disorders (e.g., ADHD) as a moderator would also be useful in future studies, as those with DSWPD and comorbid mental disorders may be more likely to present with more severe depressive symptoms than those with only DSWPD.

Additionally, we were unable to comment more on the association between DSWPD and depression as a syndrome (e.g., major depressive disorder), since there was only one study examining this. ⁴⁵ Also, only one study was in the age range of adolescence (10 to 19 years old). ²⁰ Therefore, we were unable to explore whether DSWPD status may have a different magnitude of effect on depression among adolescents compared to young adults (20 to 35 years old). Lastly, we were unable to retrieve the full-text of one article that was suspected of comparing DSWPD cases to non-DSWPD cases (Supplemental Table S6).

Conclusions

This systematic review examined the link between DSWPD and depression among young individuals. Among the included studies, 13 out 15 studies indicated that young individuals with DSWPD had a significantly greater severity of depressive symptoms than young individuals without DSWPD. However, the one study examining depression as a syndrome, did not find any such associations between DSWPD status and having a depressive disorder. Additionally, our common-effect models demonstrated that DSWPD status had a significantly large effect on depression severity, even in the models that only utilized data from studies that controlled for psychiatric disorders. However, the different risks of selection and outcome biases of the included studies as well as the small sample utilized in the common-effect models limits our findings. Future prospective studies using better sampling methods to recruit larger and more representative study samples and utilize clinician-rated depression measures without their sleep items (e.g., modified-MADRS ⁵⁹ ) and investigate for mediators (e.g., sleep-onset insomnia) longitudinally will help clarify the impact that DSWPD has on the severity of depressive symptoms among young individuals.