Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Feb 27.
Published in final edited form as: J Psychiatr Res. 2022 Oct 10;156:114–121. doi: 10.1016/j.jpsychires.2022.10.019

Sleep and circadian rhythm profiles in seasonal depression

Delainey L Wescott 1, Meredith L Wallace 2,3, Brant P Hasler 2, Alison M Klevens 1, Peter L Franzen 2, Martica H Hall 2, Kathryn A Roecklein 1,4,*
PMCID: PMC12937005  NIHMSID: NIHMS1842790  PMID: 36244199

Abstract

Sleep and circadian rhythm disruptions are symptoms of, and hypothesized underlying mechanisms in, seasonal depression. Discrepant observational findings and mixed responses to sleep/circadian-based treatments suggest heterogenous sleep and circadian disruptions in seasonal depression, despite these disruptions historically conceptualized as delayed circadian phase and hypersomnia. This study used a data-driven cluster analysis to characterize sleep/circadian profiles in seasonal depression to identify treatment targets for future interventions.

Biobehavioral measures of sleep and circadian rhythms were assessed during the winter in individuals with Seasonal Affective Disorder (SAD), subsyndromal-SAD (S-SAD), or nonseasonal, never depressed controls (total sample N=103). The following variables were used in the cluster analysis: circadian phase (from dim light melatonin onset), midsleep timing, total sleep time, sleep efficiency, regularity of midsleep timing, and nap duration (all from wrist actigraphy). Sleep and circadian variables were compared across clusters and controls.

Despite limited sleep/circadian differences between diagnostic groups, there were two reliable (Jaccard Coefficients >0.75) sleep/circadian profiles in SAD/S-SAD individuals: a ‘Disrupted sleep’ cluster, characterized by irregular and fragmented sleep and an ‘Advanced’ cluster, characterized by early sleep and circadian timing and longer total sleep times (>7.5 hours). Clusters did not differ by depression severity. Midsleep correlated with DLMO (r=0.56), irregularity (r=0.3), and total sleep time (r=−0.27).

Sleep and circadian disruptions in seasonal depression are not uniformly characterized by hypersomnia and circadian phase delay. Presence of distinct sleep and circadian subgroups in seasonal depression may predict successful treatment response. Prospective assessment and tailoring of individual sleep and circadian disruptions may reduce treatment failures.

Introduction

Seasonal depression comprises 10–20% of individuals with depression and is estimated to affect 5% of US adults (Magnusson & Partonen, 2005) and between 0.4 and 16% of adults worldwide (Freed et al., 2010). Empirically-supported treatments for seasonal depression, including light therapy, cognitive-behavioral therapy (CBT-SAD), and antidepressant medications, are not effective for all patients (Roecklein et al., 2013a). Heterogeneous symptom patterns, particularly sleep and circadian disturbances, may explain the 50% of non-responders to light therapy and 30% non-responders to CBT-SAD (Rohan et al., 2016; Terman et al., 1989; Wescott et al., 2020). For instance, realigning circadian rhythms is hypothesized to be an antidepressant mechanism of morning light therapy and may be most effective for individuals with delayed circadian phase (Lewy et al., 2009). Identifying factors that may predict who responds to light therapy compared to alternative treatments such as CBT-SAD could both enhance our understanding of the pathophysiology and reduce treatment failures. Cluster-analytic techniques are a first step towards improving successful treatment response. Cluster analyses enable researchers to holistically examine patterns of sleep and circadian disturbances rather than examining individual sleep and circadian characteristics in isolation. Identifying such patterns may be important for identifying treatment targets, considering sleep and circadian clusters may be stronger predictors than the individual measures they comprise (Wallace et al., 2019; Wallace et al., 2022). The current study aimed to characterize key patterns of sleep and circadian disturbances in seasonal depression to inform future interventions.

Individuals with seasonal depression have been typecast as having delayed circadian timing despite heterogenous findings. Discrepant findings may be related to how delayed timing has been operationalized – a circadian delay relative to sleep timing, a circadian delay relative to clock time, or a delay of the sleep/wake cycle relative to clock time. The most prominent circadian hypothesis in seasonal depression, the phase-shift hypothesis, postulates that seasonal depression results from a circadian delay relative to the sleep/wake cycle during the winter (Lewy et al., 2006). This timing metric is often measured as the interval or phase angle between internal circadian timing and the sleep/wake cycle, using a marker of circadian phase (i.e., melatonin secretion, core body temperature, etc.) and the timing of sleep (e.g., sleep midpoint), respectively. Although delayed circadian timing relative to sleep timing was presumed to be the typical presentation of seasonal depression, advanced timing relative to sleep timing has been observed in about one third of participants (N = 68; Lewy et al., 2006), while one quarter to one half exhibit circadian timing similar to controls (Burgess et al., 2004; Eastman et al., 1993). Using the timing of the sleep/wake cycle, delayed sleep timing was found in a small seasonal depression sample (n=17; Winkler et al., 2005), whereas a larger sample did not find a conclusive delay in sleep timing (n=64; Wescott et al., 2021). A third study found the majority of individuals with seasonal depression had delayed timing using actigraphic rest-activity rhythms (RARs), 24-hour periods of activity and rest (Teicher, 1997); yet, a significant minority (29%) had advanced sleep/wake timing. Our prior work using RARs found an early evening settling during the winter (down-mesor; Smagula et al., 2018). These discrepancies suggest there may be multiple sleep and circadian subgroups within seasonal depression depending on how timing is operationalized, each with different phase relationships, degrees of alignment (Roecklein, Wong, et al., 2013) or lack of sleep and circadian timing disruptions altogether.

While self-reported hypersomnia is presumed to be a common symptom of seasonal depression (Kaplan & Harvey, 2009), actigraphic findings do not uniformly support hypersomnia (Winkler et al., 2005; Wescott et al., 2021). While we recently showed that individuals with seasonal depression have a winter increase of 0.6 hours in total sleep time (7.6 hours total) compared to non-depressed controls, the group average was below the clinical cut-off for hypersomnia (>9–10 hours), and not all participants showed this increase in sleep. Further, later sleep midpoints, daytime naps, and fatigue all contributed to endorsing hypersomnia (Wescott et al., 2021), highlighting the various sleep disruptions present in seasonal depression conflated with self-reported hypersomnia. In fact, prior work has shown sleepiness (e.g., daytime naps) and lengthened sleep duration to be distinct components of hypersomnia (Kaplan et al., 2015), thus there might be separate subgroups of hypersomnolent presentations in seasonal depression. Findings of insomnia (Borisenkov et al., 2015) and co-occurring insomnia and hypersomnia in seasonal depression suggests additional heterogeneity in sleep presentations. Combined self-reports of co-occurring insomnia and hypersomnia (47%, n=51) in a seasonal sample might reflect oscillating nights of sleep length and/or fragmented sleep (Roecklein, Carney, et al., 2013). After a night of difficulty initiating or maintaining sleep, individuals might compensate by sleeping in the next morning or taking daytime naps (Bond & Wooten, 1996). This sleep extension could lead to reports of hypersomnia and may precipitate future nights of insomnia. Notably, this combined presentation of sleep-onset insomnia and hypersomnia could be a behavioral reflection of delayed circadian phase.

Altogether these discrepancies suggest multiple sleep and circadian presentations in seasonal depression, each with distinct treatment approaches. Chronotherapies (e.g., light therapy, melatonin, blue light blockers) would ostensibly be more effective for participants with circadian delays, whereas later sleep/rest cycle timing might benefit from more behavioral interventions, such as sleep advancing or sleep stabilizing from cognitive-behavioral therapy for insomnia (CBT-I; Murray et al., 2017). Additionally, CBT-I may be effective for insomnia/fragmented sleep, but may be relatively ineffective for hypersomnia without modifications. Characterizing distinct sleep and circadian profiles in seasonal depression could inform sleep and circadian interventions (Wescott et al., 2020).

Using cluster analyses to examine heterogeneous subgroups in psychiatric disorders has been at the forefront of the Research Domain Criteria (RDoC) framework implemented by the National Institutes of Mental Health (NIMH) in 2009. Sleep and circadian disruptions are ideal candidate markers for stratifying heterogeneity because they may be transdiagnostic symptoms, pathophysiological mechanisms, and modifiable treatment targets in psychiatric disorders. In unipolar depression, Robillard et al. (2018) identified two profiles characterized by delayed and conventional timing using circadian markers of dim light melatonin onset (DLMO) and core body temperature (N=50). The delayed cluster showed higher depressive symptom severity. In contrast, Carpenter et al. (2017) found three distinct sleep profiles (delayed sleep, disrupted sleep, and long sleep; N=50) that did not differ on depressive symptoms or overall functioning. This analytic approach has been underutilized in seasonal depression and may elucidate heterogeneous sleep and circadian characteristics.

Here, we extended prior analyses in nonseasonal unipolar depression by similarly implementing a cluster analytic technique in seasonal depression. We empirically identified sleep and circadian profiles using a data-driven cluster-analytic approach in a sample of individuals with varying degrees of seasonal depression. We chose our sleep and circadian measures a priori, with each variable capturing a different domain of sleep and circadian health (e.g., circadian timing, sleep timing, duration, efficiency, regularity, alertness; Buysse, 2014). Next, we compared the sleep and circadian characteristics of the resulting clusters from the SAD/S-SAD sample to the sleep and circadian characteristics of a sample of nonseasonal, never depressed control participants. We hypothesized distinct sleep and circadian profiles in seasonal depression that would differ by depression severity and differ from controls on sleep and circadian features. Potential subgroups may include hypersomnia, circadian delay, circadian advance or early settlers, and/or fragmented, non-restorative sleep more typical of nonseasonal depression (Wescott et al., 2020). Meaningfully parsing heterogeneity is a necessary first step towards identifying relevant pathophysiology and/or treatment targets in order to reduce treatment failures in seasonal depression.

Materials and methods

Participants

Participants aged 18–65 were recruited through the university-based Pitt+Me® Research Participant Registry in Pittsburgh, Pennsylvania (latitude 40°26′N) from 2013 to 2020. In Pittsburgh, the average photoperiod is 9.5 hours during winter and 15 hours during summer. The University of Pittsburgh Institutional Review Board approved all study procedures prior to participant recruitment, and the research was conducted in accordance with the Helsinki Declaration (1989). All study procedures were explained to participants prior to providing and documenting informed consent. Individuals with psychotic disorders, bipolar disorder, sleep-disordered breathing, narcolepsy, substance-induced mood disorder, or shift-workers were excluded. Participants completed all assessments during the winter months (December 21st – March 21st).

Clinical assessments

The Structured Clinical Interview for DSM-5 including the seasonal pattern specifier was used to assess seasonal depression and to screen for select comorbid Axis I disorders (SCID-I/P; First, 2015). The Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version (SIGH-SAD) was used as a measure of depression severity (Williams, Link, Rosenthal, Amira, & Terman, 1992). The Modified Seasonal Pattern Assessment Questionnaire (M-SPAQ) was used to measure the degree of individual variation in mood and behavior across the seasons. Previously-established criteria were used as inclusion criteria (Kasper et al., 1989), including the global seasonality scale (GSS), a sum of self-reported change in sleep, appetite, mood, energy level, weight and social behavior across the seasons.

Diagnostic criteria

Inclusion criteria for the SAD group included a GSS > 11, endorsing at least a “moderate” problem with seasonal changes, and “feeling worst” during January and/or February. Individuals with SAD had a history of SAD on the SCID, a current Major Depressive Episode, and met SIGH-SAD criteria for a current SAD episode (Terman et al., 1990). Individuals in the S-SAD group had a GSS of 8 or 9, endorsed at least mild problems with the seasons or a GSS > 10, and had a winter pattern of feeling worst during January and/or February. Individuals with S-SAD did not meet criteria for SAD or MDD on the SCID, nor meet the SIGH-SAD criteria for a current SAD episode. Control participants did not have a current or past major depressive episode and did not meet the SIGH-SAD criteria for a current SAD episode.

Circadian assessment

Dim Light Melatonin Onset (DLMO) is a stable and reliable biomarker of circadian phase (Dijk & Duffy, 2020) and was collected on Friday evenings in a supervised laboratory setting for all participants. Participants refrained from consuming bananas, chocolate, or aspirin/ibuprofen during their day of testing and were encouraged to limit caffeine to the morning hours. Participants were sedentary in dim light (<25 lux) for a 6-hour period starting 5 hours before and ending 1 hour after their self-reported habitual bedtime. Participants were provided with approved snacks and were not permitted to eat 10 minutes before sample collection. Saliva samples were collected every 30 minutes, stored at −20 degrees Fahrenheit, and assayed for melatonin by SolidPhase, Inc. (Portland, ME; sensitivity = 0.2 pg/mL). DLMO was determined both by linear interpolation between saliva samples using a fixed threshold of 3 pg/mL and a relative threshold of 2 standard deviations above the average of 3 baseline values (Benloucif et al., 2008). The fixed threshold has been shown to have greater consistency but reduced precision relative to the relative threshold (Burgess et al., 2010); however, the relative threshold has produced spurious results in other samples (Crowley et al., 2015). Sample size did not differ across thresholds in the current sample. We ultimately chose the fixed threshold given greater consistency across raters (DLW and AMK with final consensus with KAR). Samples were deemed unusable if they did not cross the 3pg/mL threshold.

Actigraphy

Participants wore an Actiwatch Spectrum (Philips Respironics, Bend, OR, USA) on their non-dominant wrist for 5–14 days. Actigraphy data were sampled in 30-second epochs. The Actiware software program (Philips Respironics) used a threshold sensitivity value of 40 counts per epoch, and rest intervals were manually set using sleep diary information, event markers, and/or a cutoff of activity below 40 counts. The beginning of the rest interval was determined by the timing of when the participant clicked the event marker to indicate attempt to initiate sleep. If the event marker was not pressed, rest onset was determined as the point at which activity was less than 40 counts for at least 10 consecutive epochs (5 min). The end of the rest interval was determined by the timing of event markers if the participant clicked the marker to indicate sleep offset and if the marker appeared within 30 min after activity onset. In cases with no marker indicated sleep offset, sleep offset was manually set as the last 0 that is surrounded by 20 epochs (10 min) or activity of 40 counts or less.

Actigraphy variables included total sleep time, sleep efficiency, sleep onset, sleep midpoint (halfway point between sleep onset and sleep offset), SD of sleep midpoint, wake after sleep onset (WASO), sleep onset latency (SOL), and nap duration. Sleep midpoint was used in conjunction with DLMO to estimate circadian alignment based on the DLMO-sleep midpoint phase angle.

We calculated additional non-parametric actigraphic variables (Van Someren et al., 1999) using the package ‘nparACT’ (Blume et al., 2016). We compared resulting clusters on intradaily variability (IV), interdaily stability (IS), onset of the least active 5 hours (L5), and onset of the most active 10 hours (M10). Greater fragmented rhythms within days are reflected by greater IV values and greater variability across days are indicated by higher IS values.

Sleep Diaries

Participants completed daily Pittsburgh sleep diaries (Monk et al., 1994) for 5–14 days. We compared resulting clusters on: total sleep time, the duration between sleep onset and sleep offset after subtracting SOL and WASO; sleep efficiency, the ratio of total sleep time to time spent in bed; standard deviation (SD) of sleep midpoint, the SD of the timing midpoint between sleep onset and sleep offset; sleep onset, the timing of attempt to initiate sleep; and naps.

Statistical analysis

All statistical procedures were conducted using R Studio 1.1.463 (R Core Team, 2017). Key variables for each sleep/circadian dimension of interest (circadian timing, sleep/wake timing, duration, continuity, variability, sleepiness) were included to minimize correlations between clustering variables. Circadian timing (DLMO) and circadian alignment relative to sleep timing (DLMO-midsleep) were significantly correlated (r=0.69) and were separated into two cluster analyses. Clustering variables included DLMO or DLMO-midsleep, and sleep variables assessed by actigraphy: sleep midpoint, standard deviation of sleep midpoint (variability), sleep efficiency, total sleep time, and nap duration. Prior to performing cluster analyses, we assessed the clustering tendency of the data in two ways. First, by using the Hopkins statistic (Lawson & Jurs, 1990), which tests the null hypothesis that the data is uniformly distributed (i.e., no underlying clusters). The Hopkins statistic was >0.5, which indicated there was an underlying cluster structure to our data. Second, we used a visual assessment of cluster tendency (VAT; Bezdek & Hathaway, 2002) to visually assess the dissimilarity matrix of the data, which also indicated clustering tendency was evident.

Cluster analyses were conducted using a k-means approach with the Euclidean distance metric as a measure of similarity. Due to the sensitivity of this approach to large ranges and variances, we standardized each indicator. The number of clusters (k) was determined using the Nbclust package in R (Charrad et al., 2014), which simultaneously uses information from 30 different indices to determine the optimal number of clusters. See supplemental figures for additional clustering information. Resulting clusters were compared on demographics, sleep and circadian variables, depression severity, and diagnostic group. We compared resulting clusters to a sample of nonseasonal, never depressed controls. A Bonferroni correction was applied to cluster contrasts to control for type 1 error. Clusters were compared using one-way analysis of variance with student’s t-tests for post-hoc contrasts or Krustal-Wallis chi-square test with Dunn’s test for post-hoc contrasts if normality and homogeneity of variance assumptions were not met. Stability of the resulting clusters was determined using Jaccard Coefficient (JC; Hennig, 2007), with >0.75 indicating stable cluster solutions and <0.60 suggesting unreliable, spurious clusters. Unreliable cluster solutions were not reported.

Results

Sample description

There were 37 individuals with SAD, 21 individuals with S-SAD, and 45 nonseasonal controls with no history of depression (N=103). Neither age nor sex differed between the diagnostic groups in the full sample. As expected, seasonal depression severity measured by the SIGH-SAD was significantly higher in the SAD group compared to S-SAD and controls, and higher in S-SAD compared to controls (Table 1). Notably, sleep diary reported efficiency differed between the groups; individuals with SAD and S-SAD reported ~5% less sleep efficiency than controls.

Table 1.

Demographic, sleep, and circadian measures across diagnostic groups: controls (C), Seasonal Affective Disorder (SAD), Subsyndromal Seasonal Affective Disorder (S-SAD).

Controls SAD S-SAD Full Sample Test statistics Pair-wise comparisons
N 45 37 21 103
Age 33.5 (12.0) 38.8 (11.8) 43.1 (13.6) 37.3 (12.7) F(2, 99)=4.6* C < S-SAD
Sex (%female) 34 (76%) 33 (89%) 17 (81 %) 84 (82%) FE3
Race (%white) 30 (67%) 31 (84%) 19 (90%) 80 (78%) F(2, 100)=1.3
Ethnicity (%Non-Hispanic/Latino) 43 (96%) 33 (89%) 21 (100%) 97 (94%) F(2, 100)=1.6
SIGH-SAD1 3.7 (4.9) 26.8 (7.6) 15.1 (5.4) 14.6 (12.0) χ2=67.6*** C < S-SAD < SAD
Actigraphy & DLMO
DLMO2 (HH:MM) 9:36 PM (1.4) 8:54 PM (1.5) 9:24 PM (1.2) 9:18 PM (1.4) F(2, 100)=2.5
DLMO-midsleep (hours) 5.8 (1.3) 6.3 (1.0) 6.1 (1.4) 6.1 (1.2) F(2, 100)=1.7
Total sleep time (hours) 7.0 (0.9) 7.3 (0.9) 7.1 (0.9) 7.1 (0.9) F(2, 100)=1.3
Efficiency (%) 89.0 (3.8) 87.3 (5.2) 85.8 (7.3) 87.7 (5.2) χ2=5.9
SD of midsleep (hours) 0.9 (0.4) 1.0 (0.4) 0.9 (0.3) 0.9 (0.4) F(2, 100)=0.8
Onset (HH:MM) 11:36 PM (1.6) 11:12 PM (1.1) 11:36 PM (1.4) 11:30 (1.4) F(2, 100)=0.9
Midpoint (HH:MM) 3:24 AM (1.4) 3:18 AM (0.9) 3:30 AM (1.2) 3.4 AM (1.2) F(2, 100)=0.4
Naps (minutes) 2.9 (6.2) 6.3 (12.9) 2.1 (6.5) 3.4 (9.3) χ2=4.2
WASO4 (minutes) 39.7 (15.8) 43.3 (23.6) 42.1 (19.3) 41.5 (19.5) F(2, 100)=0.7
SOL5 (minutes) 5.6 (4.3) 11.0 (13.5) 13.5 (9.3) 9.3 (9.9) χ2=14.2*** C < S-SAD
Intradaily variabilty 0.9 (0.2) 0.8 (0.2) 0.9 (0.2) 0.9 (0.2) F(2, 97)=1.5
Interdaily stability 0.4 (0.2) 0.4 (0.2) 0.4 (0.2) 0.4 (0.2) F(2, 97)=1.3
L5 onset (HH:MM)6 11:30 PM (3.2) 11:42 PM (1.8) 10:42 PM (4.2) 11:24 PM (3.0) F(2, 85)=0.7
M10 onset (HH:MM)7 9:12 AM (3.9) 9:15 AM (3.8) 8:20 AM (3.6) 9:03 AM (3.7) F(2, 85)=0.4
Intradaily variabilty 0.9 (0.2) 0.8 (0.2) 0.9 (0.2) 0.9 (0.2) F(2, 97)=1.5
Sleep diaries
Total sleep time (hours) 7.4 (0.8) 7.5 (0.9) 7.4 (0.9) 7.4 (0.9) F(2, 86)=0.2
Efficiency (%) 93.1% (0.1) 88.0% (0.1) 87.7% (0.1) 90.0% (0.1) χ2=14.3*** SAD, S-SAD < C
SD of midsleep (hours) 0.8 (0.4) 0.9 (0.5) 0.8 (0.6) 0.8 (0.5) F(2, 86)=0.7
Onset (HH:MM) 11:43 PM (1.5) 11:21 PM (1.1) 11:49 PM (1.2) 11:37 (1.3) χ2=1.7
Naps (minutes) 9.9 (12.9) 16.2 (34.1) 6.6 (20.0) 11.3 (23.8) χ2=2.7
Open in a new tab
1

Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version total score

2

Dim light melatonin onset

3

Fisher’s Exact Test

4

Wake after sleep onset

5

Sleep onset latency

6

Onset of least active 5 hours

7

Onset of most active 10 hours

Note:

*

= p<.05

**

= p<.01

***

= p<.001

Cluster analysis

There were 13 unusable DLMOs (5 SAD, 2 S-SAD, 6 Controls) potentially due to DLMOs outside of the testing window and/or high or low secretion. Unusable low secretors did not differ from usable DLMOs in sleep timing (F(1, 129)=0.9; p>0.05) or weeknight/weekend differences in sleep timing (i.e., social jetlag; F(1, 127) =0.7; p>0.05). Individuals with S-SAD had longer actigraphic sleep onset latencies than controls. There were no other differences in circadian or actigraphy assessed sleep variables across diagnostic groups (Table 1).

Clustering results for SAD and S-SAD participants including circadian timing (DLMO) suggested two clusters that did not differ by diagnostic group or depression severity (Table 2). The ‘Disrupted sleep’ cluster (n=25; JC=0.78) was characterized by low sleep efficiency and irregular sleep (higher SD in midsleep). The ‘Advanced’ cluster (n=33; JC=0.84) was characterized by early sleep and circadian timing and longer total sleep times (see Figure 1).

Table 2.

Comparisons on demographic, clinical, sleep-wake, and circadian variables across cluster groups and controls (C).

Cluster 1 ‘Disrupted sleep’ Cluster 2 ‘Advanced’ Controls Test statistics Significant pair-wise comparisons
N 25 33 45
Demographics
Age 38.9 (12.7) 41.3 (12.5) 33.5 (12.0) F(2, 99)=4.1* C < Advanced
Sex (%female) 21 (84%) 29 (88%) 34 (76%) FE3
Race (%White) 19 (76%) 31 (94%) 30 (67%) F(2, 100)=2.1
Ethnicity (%Non-Hispanic/Latino) 22 (88%) 32 (97%) 43 (96%) F(2, 100)=1.2
Group (%SAD vs. S-SAD) 17 (68%) 20 (61%)
SIGH-SAD1 23.3 (8.5) 22.1 (9.3) 3.7 (4.9) χ2=58.9*** C < Both clusters
Clustering variables
DLMO2 (HH:MM) 10:06 PM (0.9) 8:18 PM (1.1) 9:36 PM (1.4) F(2, 100)=19.7*** Advanced < C, Disrupted sleep
Total sleep time (hours) 6.8 (0.9) 7.6 (0.8) 7.0 (0.9) F(2, 100)=5.5** C, Disrupted sleep < Advanced
Efficiency (%) 84.4 (7.0) 88.5 (4.5) 89.0 (3.8) χ2=14.2*** Disrupted sleep < C, Advanced
SD of midsleep (hours) 1.2 (0.4) 0.9 (0.4) 0.9 (0.4) χ2=16.9*** C, Advanced < Disrupted sleep
Midpoint (HH:MM) 4:00 AM (1.0) 2:48 AM (0.7) 3:24 AM (1.4) χ2=17.1*** Advanced < Disrupted sleep
Naps (minutes) 6.3 (14.9) 3.6 (7.1) 2.9 (6.2) χ2=0.1
Sleep & circadian variables
Onset (HH:MM) 12:12 AM (1.1) 10:42 PM (0.9) 11:36 PM (1.6) χ2=22.8*** Advanced < C, Disrupted sleep
WASO4 (minutes) 46.1 (20.6) 40.4 (23.0) 39.7 (15.8) F(2, 100)=0.9
SOL5 (minutes) 12.9 (15.2) 11.1 (9.2) 5.9 (9.2) χ2=10.8** C < Both clusters
Intradaily variability (IV) 0.9 (0.2) 0.8 (0.2) 0.9 (0.2) F(2, 97)=0.8
Interdaily stability (IS) 0.4 (0.2) 0.5 (0.2) 0.4 (0.2) F(2, 97)=2.8
L56 11:56 PM (3.3) 10:51 PM (2.5) 11:30 PM (3.2) F(2, 85)=0.9
M107 9:50 AM (4.2) 8:10 AM (3.1) 9:12 AM (3.9) F(2, 85)=1.3
DLMO-midsleep (hours) 5.9 (1.2) 6.6 (1.1) 5.8 (1.3) F(2, 100)=3.9* C < Advanced
Sleep diaries
Total sleep time (hours) 7.2 (0.8) 7.7 (1.0) 7.4 (0.8) F(2, 86)=2.4
Efficiency (%) 84.8 (0.1) 90.1 (0.1) 93.1 (0.1) χ2=19.4*** Disrupted sleep < C
SD of midsleep (hours) 1.1 (0.6) 0.7 (0.4) 0.8 (0.4) χ2=5.7
Onset (HH:MM) 12:11 AM (1.3) 11:04 PM (0.8) 11:44 PM (1.5) χ2=10.9** Advanced < Disrupted sleep
Naps (minutes) 21.9 (42.8) 5.4 (8.3) 9.9 (12.9) χ2=2.7
Open in a new tab
1

Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version total score

2

Dim light melatonin onset

3

Fisher’s Exact Test

4

Wake after sleep onset

5

Sleep onset latency

6

Onset of least active 5 hours

7

Onset of most active 10 hours

Note:

*

= p<.05

**

= p<.01

***

= p<.001

Figure 1.

Figure 1.

Open in a new tab

Scatterplot matrix of clustering variables for the ‘Disrupted sleep’ and the ‘Advanced’ cluster.

When the two clusters were compared to a sample of nonseasonal, never depressed controls, the controls were younger than the ‘Advanced’ cluster (F(2, 99)=4.1; p=0.02). The ‘Advanced’ cluster (7.6 hours) had significantly longer total sleep times than the control group (7 hours) and the ‘Disrupted sleep’ cluster (6.8 hours). The ‘Advanced’ cluster had significantly earlier circadian timing (DLMO; 8:18pm) and sleep onset (10:42pm) than the control group (DLMO=9:36pm; sleep onset=11:36pm) and the ‘Disrupted sleep’ cluster (DLMO=10:06pm; sleep onset=12:12am). The ‘Advanced’ cluster had a longer DLMO-midsleep interval (6.6 hours) compared to the control group (5.8 hours). The ‘Disrupted sleep’ cluster had significantly lower sleep efficiencies (84%) compared to the control group (89%) and the ‘Advanced’ cluster (89%) and reported lower sleep efficiency on sleep diaries (85%) compared to controls (93%). The ‘Disrupted sleep’ cluster also evidenced more irregular sleep. The two clusters did not differ on SOL, WASO, or non-parametric measures of sleep including IS, IV, L5, and M10. See Table 2 and Figure 2.

Figure 2.

Figure 2.

Open in a new tab

Average sleep and circadian timing for clusters and controls. Colored rectangles represent average total sleep time; large triangles represent average DLMO for each cluster while smaller triangles represent each individual’s DLMO timing within the cluster. Large circles represent average sleep onset, midpoint, and offset for each cluster. Smaller circles represent each individual’s sleep midpoint timing within the cluster. Error bars represent standard deviation. Figure inspired by Carpenter et al. (2017).

Including circadian alignment (DLMO-midsleep) instead of circadian timing (DLMO) in the cluster analysis also yielded a similar two-cluster solution (‘Advanced’, n=35; ‘Disrupted sleep’, n=23). However, the resulting clusters were unstable (JC=0.6 and 0.2), potentially due to reduced separation between clusters (see supplemental materials), suggesting this clustering solution may be a spurious finding.

Discussion

Distinct sleep and circadian profiles in seasonal depression were identified, each with unique and potentially-modifiable treatment targets. These distinct profiles are especially notable considering (1) there were no sleep and circadian differences between diagnostic groups (controls, SAD, S-SAD) and (2) this is the largest sleep and circadian study to date in a seasonal depression sample. The current study identified a ‘Disrupted sleep’ cluster, characterized by irregular sleep and low sleep efficiencies, and an ‘Advanced’ cluster, characterized by longer total sleep times and early sleep and circadian timing. Contrary to our hypothesis, clusters did not differ on depression severity or stratify by diagnostic group. Each cluster comprises a unique constellation of sleep and circadian characteristics with associated clinical implications, which may potentially explain mixed findings and incomplete treatment response in previous samples.

A surprising finding was that over half (57%) of this well-characterized seasonal sample was ‘Advanced’, despite the fact that circadian delays relative to sleep have previously been considered the most prominent seasonal depression phenotype (Lewy et al., 2009). The ‘Advanced’ cluster had significantly earlier circadian timing, with average DLMOs (8:18pm) over an hour earlier than both controls (9:36pm) and the ‘Disrupted sleep’ cluster (10:06pm). It is possible individuals with later circadian phase in our sample had low melatonin secretions that were not captured by our protocol (last DLMO collected at 1am) or were acutely suppressed by light in the 10 – 25 lux range (Phillips et al., 2019). However, these unusable low secretors did not differ from participants with usable DLMOs on measures of sleep timing or social jetlag, although circadian timing could still be delayed in those participants. Importantly, prior work measuring circadian timing in seasonal depression has not consistently supported a delayed circadian phase presentation, at least in part due to varying circadian markers (DLMO, 24-hour melatonin, core body temperature), circadian timing measures (circadian timing vs. circadian timing relative to sleep/wake timing), and study design protocols. While individuals with seasonal depression were phase-delayed compared to control participants in a constant routine protocol using DLMO and core body temperature (n = 6; Avery et al., 1997; Dahl et al., 1993), in a forced desynchrony study, no differences in the period or phase of melatonin secretion or core body temperature were found in 14 individuals with SAD (N = 14; Koorengevel et al. 2002). In larger samples, Checkley et al. (1993) found no differences in 24-hour melatonin rhythms (N = 40), and Burgess et al. (2004) and Eastman et al. (1993) found that 46% (N = 26) and 22% (n = 22) of individuals with seasonal depression did not differ from control participants on circadian rhythms of core body temperature relative to sleep timing. It is most likely that circadian alignment varies across individuals with seasonal depression and that prior samples had varying percentages of one phenotype or the other, a sampling issue that is exacerbated in smaller samples. Of note, prior work in seasonal depression suggests altered retinal sensitivity (e.g., the post-illumination pupil response; PIPR; Roecklein et al., 2021) as well as disruptions in neural pathways of photic-mood regulation (Maruani & Geoffroy, 2022). Future research should examine whether alterations in retinal-neural pathways contribute to earlier circadian profiles.

In the ‘Advanced’ cluster, earlier sleep onset times are consistent with early settling, or an earlier decrease in physical activity at the end of the active period (Smagula et al., 2018), which could lead to longer total sleep times. Early settling may be a function of fatigue, lethargy, and/or behavioral disengagement during a depressive episode. Individuals experiencing fatigue or disengagement might discontinue physical activity in the evening and retire to bed early, which could contribute to longer total sleep times. While the mean actigraphic total sleep time in the ‘Advanced’ cluster (7.6 hours) is below the hypersomnia threshold (>9–10 hours; APA, 2013), this cluster slept 35 min longer than controls (7 hours), a statistically and clinically significant difference. Hypersomnia in seasonal depression may be better conceptualized as longer total sleep times potentially related to earlier sleep and circadian timing.

Sleep disturbances beyond hypersomnia are often overlooked in seasonal depression. The ‘Disrupted sleep’ cluster evidenced fragmented and irregular sleep, supporting prior sleep fragmentation symptoms reported in seasonal samples (Borisenkov et al., 2015; Roecklein, Carney, et al., 2013; Winkler et al., 2005). The presence of an ‘Disrupted sleep’ cluster with total sleep times similar to controls could have obscured the existence of longer total sleep times in seasonal depression in prior studies (Anderson et al., 1994; Shapiro et al., 1994; Winkler et al., 2005; Wescott et al., 2021). While the ‘Disrupted sleep’ cluster had later sleep and circadian times than the ‘Advanced’ cluster, this cluster did not significantly differ from controls, suggesting that this ‘Disrupted sleep’ cluster does not have significant delays in sleep or circadian timing.

Sleep disturbances may be further characterized by examining differences in sleep architecture using polysomnology (PSG) and/or sleep diaries. While actigraphy provides an ecologically-valid measure of sleep-wake behavior (Sadeh, 2011), it cannot assess structural changes in sleep architecture. A recent meta-analysis investigating PSG sleep abnormalities in seasonal depression suggests increased REM sleep during episodes and decreased REM latency during remission, but no changes in slow-wave sleep (Bertrand et al., 2021). Further, sleep diaries capture the subjective component of sleep. Future work should consider including PSG and sleep diarymarkers when characterizing sleep and circadian profiles in seasonal depression.

Each distinct subgroup was characterized by potentially-modifiable treatment targets. Participants in the ‘Disrupted sleep’ cluster might benefit from CBT-I, which could improve sleep efficiency and stabilize irregular sleep. Participants in the ‘Advanced’ cluster might respond better to behavioral activation, a component of CBT-SAD (Rohan et al., 2016), which may increase physical and social activity in the evening to ward off early settling. Morning bright light therapy may be effective for participants in the ‘Disrupted sleep’ cluster, given prior work showing morning bright light therapy consolidates the sleep-wake cycle in seasonal depression for participants with fragmented sleep (Winkler et al., 2005). The ‘Advanced’ participants may be the non-responders to morning light (Burgess et al., 2004), and this subgroup of participants may benefit from the ‘direct’ alerting and mood improving benefits of light therapy (LeGates et al., 2012). Prior work comparing light therapy to CBT-SAD in seasonal depression found that individuals endorsing sleep-onset insomnia or hypersomnia remit faster with light therapy compared to CBT-SAD (Meyerhoff et al., 2018). Ultimately at post-treatment, both treatment groups showed similar treatment responses. While the current observational study was not able to directly test whether distinct sleep and circadian profiles predict successful response to either light therapy or CBT-SAD, our findings provide preliminary evidence challenging the dogma that all individuals with seasonal depression show similar sleep and circadian disruptions and thus respond to similar treatments. Identifying distinct sleep and circadian clusters may inform future work using a precision medicine approach to identify appropriate treatment targets. Ultimately, our findings challenge the current perspective of uniform sleep and circadian disruptions in seasonal depression. Individuals with seasonal depression may benefit from assessing transdiagnostic sleep and circadian dysfunctions when evaluating treatment options.

Strengths & Limitations

This study benefitted from multiple measures of sleep and circadian disruptions to elucidate data-driven subgroups. The focus on sleep and circadian disruptions helped identify constellations of symptoms to characterize subgroups in seasonal depression. While our use of the Jaccard Coefficient to assess cluster stability acts supported internal validity of the resulting clusters, our findings should be considered preliminary before replicated in larger samples.

The cross-sectional design limits any causal interpretations of whether these sleep and circadian disruptions are symptoms of seasonal depressive episodes or potential contributors to onset and recurrence of seasonal depression. Due to the data-driven nature of cluster analyses and smaller sample size, these findings would be bolstered by verification in a separate dataset. Additionally, the validity and reliability of actigraphic measures (range 5–14 days in the current sample) may be reduced in participants with less than 7 nights (Fischer et al., 2021). The number of actigraphic measurements varied according to participants’ availability for DLMO scheduling. Since DLMO collection only occurred on Friday evenings, sometimes participants were not available for DLMO until two weeks after their initial assessment. There were 13 unusable DLMOs where onset did not occur during the testing window. Additionally, our use of salivary melatonin as a marker of circadian phase differs from previous work in seasonal depression using plasma melatonin (Lewy et al., 2006) or core body temperature (Avery et al., 1997; Burgess et al., 2004; Dahl et al., 1993; Eastman et al., 1993), although prior work suggest both measures are comparable for measuring DLMO (Leibenluft et al., 1996; Nowak et al., 1987; Voultsios et al., 1997).

Conclusion

Main findings.

The data-driven clustering analysis unveiled distinct presentations of sleep and circadian disruptions in seasonal depression. Although findings should be viewed as preliminary until replicated, our data suggests two reliable sleep and circadian profiles: ‘Disrupted sleep’ – irregular and fragmented sleep and ‘Advanced’ – early sleep and circadian timing and relatively longer total sleep times (> 7.5 hours). The presence of divergent subgroups in seasonal depression may explain null and heterogenous findings of prior work.

Clinical implications.

Treating sleep and circadian disruptions in seasonal depression may benefit from an individually-tailored, precision medicine approach. Changing the perspective of sleep and circadian disruptions in seasonal depression from uniformly hypersomnia and phase delay to a more accurate heterogenous presentation will be more effective when identifying the most promising interventions. Identifying the key drivers of sleep-related pathophysiology in seasonal depression may minimize time to remission and reduce recurrence rates.

Future directions.

Replication of the current findings is critical. While the current study focused on seasonal depression, sleep and circadian disruptions are transdiagnostic. Targeting specific sleep-wake and biological rhythm profiles could aid our understanding of the etiology of mood dysregulation if tested prospectively. Matching transdiagnostic sleep and circadian processes with targeted treatment approaches may improve treatment response.

Supplementary Material

Supplement

Funding:

This research was supported by the National Institute of Health (R03MH096119; R01MH103313; T32 HL082610).

References

  1. Anderson JL, Rosen LN, Mendelson WB, Jacobsen FM, Skwerer RG, Joseph-Vanderpool JR, Duncan CC, Wehr TA, & Rosenthal NE (1994). Sleep in fall/winter seasonal affective disorder: Effects of light and changing seasons. Journal of Psychosomatic Research, 38(4), 323–337. 10.1016/0022-3999(94)90037-X [DOI] [PubMed] [Google Scholar]
  2. Avery DH, Dahl K, Savage MV, Brengelmann GL, Larsen LH, Kenny MA, Eder DN, Vitiello MV, & Prinz PN (1997). Circadian temperature and cortisol rhythms during a constant routine are phase-delayed in hypersomnic winter depression. Biological Psychiatry, 41(11), 1109–1123. 10.1016/S0006-3223(96)00210-7 [DOI] [PubMed] [Google Scholar]
  3. Benloucif S, Burgess HJ, Klerman EB, Lewy AJ, Middleton B, Murphy PJ, Parry BL, & Revell VL (2008). Measuring Melatonin in Humans. Journal of Clinical Sleep Medicine, 04(01), 66–69. [PMC free article] [PubMed] [Google Scholar]
  4. Bertrand L, d’Ortho M-P, Reynaud E, Lejoyeux M, Bourgin P, & Geoffroy PA (2021). Polysomnography in seasonal affective disorder: A systematic review and meta-analysis. Journal of Affective Disorders, 292, 405–415. 10.1016/j.jad.2021.05.080 [DOI] [PubMed] [Google Scholar]
  5. Blume C, Santhi N, & Schabus M (2016). ‘nparACT’ package for R: A free software tool for the non-parametric analysis of actigraphy data. MethodsX, 3, 430–435. 10.1016/j.mex.2016.05.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bond T, & Wooten V (1996). The etiology and management of insomnia. Virginia Medical Quarterly: VMQ, 123(4), 254–255. [PubMed] [Google Scholar]
  7. Borisenkov MF, Petrova NB, Timonin VD, Fradkova LI, Kolomeichuk SN, Kosova AL, & Kasyanova ON (2015). Sleep characteristics, chronotype and winter depression in 10–20-year-olds in northern European Russia. Journal of Sleep Research, 24(3), 288–295. 10.1111/jsr.12266 [DOI] [PubMed] [Google Scholar]
  8. Burgess HJ, Fogg LF, Young MA, & Eastman CI (2004). Bright Light Therapy for Winter Depression—Is Phase Advancing Beneficial? Chronobiology International, 21(4–5), 759–775. 10.1081/CBI-200025979 [DOI] [PubMed] [Google Scholar]
  9. Burgess HJ, Revell VL, Molina TA, & Eastman CI (2010). Human Phase Response Curves to Three Days of Daily Melatonin: 0.5 mg Versus 3.0 mg. The Journal of Clinical Endocrinology and Metabolism, 95(7), 3325–3331. 10.1210/jc.2009-2590 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Buysse DJ (2014). Sleep health: Can we define it? Does it matter? Sleep, 37(1), 9–17. 10.5665/sleep.3298 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Carpenter JS, Robillard R, Hermens DF, Naismith SL, Gordon C, Scott EM, & Hickie IB (2017). Sleep-wake profiles and circadian rhythms of core temperature and melatonin in young people with affective disorders. Journal of Psychiatric Research, 94, 131–138. 10.1016/j.jpsychires.2017.07.007 [DOI] [PubMed] [Google Scholar]
  12. Crowley SJ, Cain SW, Burns AC, Acebo C, & Carskadon MA (2015). Increased Sensitivity of the Circadian System to Light in Early/Mid-Puberty. The Journal of Clinical Endocrinology & Metabolism, 100(11), 4067–4073. 10.1210/jc.2015-2775 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Dahl K, Avery DH, Lewy AJ, Savage MV, Brengelmann GL, Larsen LH, Vitiello MV, & Prinz PN (1993). Dim light melatonin onset and circadian temperature during a constant routine in hypersomnic winter depression. Acta Psychiatrica Scandinavica, 88(1), 60–66. 10.1111/j.1600-0447.1993.tb03414.x [DOI] [PubMed] [Google Scholar]
  14. Dijk D-J, & Duffy JF (2020). Novel Approaches for Assessing Circadian Rhythmicity in Humans: A Review. Journal of Biological Rhythms, 35(5), 421–438. 10.1177/0748730420940483 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Eastman CI, Gallo LC, Lahmeyer HW, & Fogg LF (1993). The circadian rhythm of temperature during light treatment for winter depression. Biological Psychiatry, 34(4), 210–220. 10.1016/0006-3223(93)90074-N [DOI] [PubMed] [Google Scholar]
  16. First MB (2015). Structured Clinical Interview for the DSM (SCID). In The Encyclopedia of Clinical Psychology (pp. 1–6). American Cancer Society. 10.1002/9781118625392.wbecp351 [DOI] [Google Scholar]
  17. Fischer D, Klerman EB, & Phillips AJK (2021). Measuring sleep regularity: Theoretical properties and practical usage of existing metrics. Sleep, 44(10), zsab103. 10.1093/sleep/zsab103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Freed MC, Osborn RL, & Rohan KJ (2010). Estimating the Disease Burden of Seasonal Affective Disorder. In Preedy VR & Watson RR (Eds.), Handbook of Disease Burdens and Quality of Life Measures (pp. 1549–1568). Springer. 10.1007/978-0-387-78665-0_90 [DOI] [Google Scholar]
  19. Hennig C (2007). Cluster-wise assessment of cluster stability. Computational Statistics & Data Analysis, 52(1), 258–271. 10.1016/j.csda.2006.11.025 [DOI] [Google Scholar]
  20. Kaplan KA, & Harvey AG (2009). Hypersomnia across mood disorders: A review and synthesis. Sleep Medicine Reviews, 13(4), 275–285. 10.1016/j.smrv.2008.09.001 [DOI] [PubMed] [Google Scholar]
  21. Kaplan KA, McGlinchey EL, Soehner A, Gershon A, Talbot LS, Eidelman P, Gruber J, & Harvey AG (2015). Hypersomnia Subtypes, Sleep and Relapse in Bipolar Disorder. Psychological Medicine, 45(8), 1751–1763. 10.1017/S0033291714002918 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. LeGates TA, Altimus CM, Wang H, Lee H-K, Yang S, Zhao H, Kirkwood A, Weber ET, & Hattar S (2012). Aberrant light directly impairs mood and learning through melanopsin-expressing neurons. Nature, 491(7425), 594–598. 10.1038/nature11673 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Leibenluft E, Feldman-Naim S, Turner EH, Schwartz PJ, & Wehr TA (1996). Salivary and plasma measures of dim light melatonin onset (DLMO) in patients with rapid cycling bipolar disorder. Biological Psychiatry, 40(8), 731–735. 10.1016/0006-3223(95)00488-2 [DOI] [PubMed] [Google Scholar]
  24. Lewy AJ, Emens JS, Songer JB, Sims N, Laurie AL, Fiala SC, & Buti AL (2009). Winter Depression: Integrating mood, circadian rhythms, and the sleep/wake and light/dark cycles into a bio-psycho-social-environmental model. Sleep Medicine Clinics, 4(2), 285–299. 10.1016/j.jsmc.2009.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lewy AJ, Lefler BJ, Emens JS, & Bauer VK (2006). The circadian basis of winter depression. Proceedings of the National Academy of Sciences of the United States of America, 103(19), 7414–7419. 10.1073/pnas.0602425103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Magnusson A, & Partonen T (2005). The Diagnosis, Symptomatology, and Epidemiology of Seasonal Affective Disorder. CNS Spectrums, 10(08), 625–634. 10.1017/S1092852900019593 [DOI] [PubMed] [Google Scholar]
  27. Maruani J, & Geoffroy PA (2022). Multi-Level Processes and Retina–Brain Pathways of Photic Regulation of Mood. Journal of Clinical Medicine, 11(2), 448. 10.3390/jcm11020448 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Meyerhoff J, Young MA, & Rohan KJ (2018). Patterns of depressive symptom remission during the treatment of seasonal affective disorder with cognitive-behavioral therapy or light therapy. Depression and Anxiety, 35(5), 457–467. 10.1002/da.22739 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Monk TH, Reynolds CF, Kupfer DJ, Buysse DJ, Coble PA, Hayes AJ, Machen MA, Petrie SR, & Ritenour AM (1994). The Pittsburgh Sleep Diary. Journal of Sleep Research, 3(2), 111–120. 10.1111/j.1365-2869.1994.tb00114.x [DOI] [PubMed] [Google Scholar]
  30. Murray JM, Sletten TL, Magee M, Gordon C, Lovato N, Bartlett DJ, Kennaway DJ, Lack LC, Grunstein RR, Lockley SW, & Rajaratnam SMW (2017). Prevalence of Circadian Misalignment and Its Association With Depressive Symptoms in Delayed Sleep Phase Disorder. Sleep, 40(1). 10.1093/sleep/zsw002 [DOI] [PubMed] [Google Scholar]
  31. Nowak R, McMillen IC, Redman J, & Short RV (1987). The correlation between serum and salivary melatonin concentrations and urinary 6-hydroxymelatonin sulphate excretion rates: Two non-invasive techniques for monitoring human circadian rhythmicity. Clinical Endocrinology, 27(4), 445–452. 10.1111/j.1365-2265.1987.tb01172.x [DOI] [PubMed] [Google Scholar]
  32. Phillips AJK, Vidafar P, Burns AC, McGlashan EM, Anderson C, Rajaratnam SMW, Lockley SW, & Cain SW (2019). High sensitivity and interindividual variability in the response of the human circadian system to evening light. Proceedings of the National Academy of Sciences, 201901824. 10.1073/pnas.1901824116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Robillard R, Carpenter JS, Rogers NL, Fares S, Grierson AB, Hermens DF, Naismith SL, Mullin SJ, Feilds K-L, Glozier N, Scott EM, & Hickie IB (2018). Circadian rhythms and psychiatric profiles in young adults with unipolar depressive disorders. Translational Psychiatry, 8(1), 213. 10.1038/s41398-018-0255-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Roecklein KA, Carney CE, Wong PM, Steiner JL, Hasler BP, & Franzen PL (2013). The role of beliefs and attitudes about sleep in seasonal and nonseasonal mood disorder, and nondepressed controls. Journal of Affective Disorders, 150(2), 466–473. 10.1016/j.jad.2013.04.041 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Roecklein KA, Franzen PL, Wescott DL, Hasler BP, Miller MA, Donofry SD, DuPont CM, Gratzmiller SM, Drexler SP, Wood-Vasey WM, & Gamlin PD (2021). Melanopsin-driven pupil response in summer and winter in unipolar seasonal affective disorder. Journal of Affective Disorders, 291, 93–101. 10.1016/j.jad.2021.04.084 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Roecklein KA, Wong PM, Miller MA, Donofry SD, Kamarck ML, & Brainard GC (2013). Melanopsin, photosensitive ganglion cells, and seasonal affective disorder. Neuroscience & Biobehavioral Reviews, 37(3), 229–239. 10.1016/j.neubiorev.2012.12.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Rohan KJ, Meyerhoff J, Ho S-Y, Evans M, Postolache TT, & Vacek PM (2016). Outcomes One and Two Winters Following Cognitive-Behavioral Therapy or Light Therapy for Seasonal Affective Disorder. The American Journal of Psychiatry, 173(3), 244–251. 10.1176/appi.ajp.2015.15060773 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Sadeh A (2011). The role and validity of actigraphy in sleep medicine: An update. Sleep Medicine Reviews, 15(4), 259–267. 10.1016/j.smrv.2010.10.001 [DOI] [PubMed] [Google Scholar]
  39. Shapiro CM, Devins GM, Feldman B, & Levitt AJ (1994). Is hypersomnolence a feature of seasonal affective disorder? Journal of Psychosomatic Research, 38, 49–54. 10.1016/0022-3999(94)90135-X [DOI] [PubMed] [Google Scholar]
  40. Smagula SF, DuPont CM, Miller MA, Krafty RT, Hasler BP, Franzen PL, & Roecklein KA (2018). Rest-activity rhythms characteristics and seasonal changes in seasonal affective disorder. Chronobiology International, 1–7. 10.1080/07420528.2018.1496094 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Teicher MH (1997). Circadian Rest-Activity Disturbances in Seasonal Affective Disorder. Archives of General Psychiatry, 54(2), 124. 10.1001/archpsyc.1997.01830140034007 [DOI] [PubMed] [Google Scholar]
  42. Terman M, Terman JS, Quitkin FM, McGrath PJ, Stewart JW, & Rafferty B (1989). Light therapy for seasonal affective disorder. A review of efficacy. Neuropsychopharmacology: Official Publication of the American College of Neuropsychopharmacology, 2(1), 1–22. [DOI] [PubMed] [Google Scholar]
  43. Terman M, Terman JS, & Rafferty B (1990). Experimental design and measures of success in the treatment of winter depression by bright light. Psychopharmacology Bulletin, 26(4), 505–510. [PubMed] [Google Scholar]
  44. Van Someren EJ, Swaab DF, Colenda CC, Cohen W, McCall WV, & Rosenquist PB (1999). Bright light therapy: Improved sensitivity to its effects on rest-activity rhythms in Alzheimer patients by application of nonparametric methods. Chronobiology International, 16(4), 505–518. 10.3109/07420529908998724 [DOI] [PubMed] [Google Scholar]
  45. Voultsios A, Kennaway DJ, & Dawson D (1997). Salivary Melatonin as a Circadian Phase Marker: Validation and Comparison to Plasma Melatonin. Journal of Biological Rhythms, 12(5), 457–466. 10.1177/074873049701200507 [DOI] [PubMed] [Google Scholar]
  46. Wallace ML, Lee S, Stone KL, Hall MH, Smagula SF, Redline S, Ensrud K, Ancoli-Israel S, & Buysse DJ (2022). Actigraphy-derived sleep health profiles and mortality in older men and women. Sleep, 45(4), zsac015. 10.1093/sleep/zsac015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Wallace M, Lee S, Hall M, Stone K, Langsetmo L, Redline S, Schousboe J, Ensrud K, LeBlanc E, & Buysse D (2019). Heightened sleep propensity: A novel and high-risk sleep health phenotype in older adults. Sleep Health, 5(6), 630–638. 10.1016/j.sleh.2019.08.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Wescott DL, Franzen PL, Hasler BP, Miller MA, Soehner AM, Smagula SF, Wallace ML, Hall MH, & Roecklein KA (undefined/ed). Elusive hypersomnolence in seasonal affective disorder: Actigraphic and self-reported sleep in and out of depressive episodes. Psychological Medicine, 1–10. 10.1017/S003329172100283X [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Wescott DL, Soehner AM, & Roecklein KA (2020). Sleep in seasonal affective disorder. Current Opinion in Psychology, 34, 7–11. 10.1016/j.copsyc.2019.08.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Winkler D, Pjrek E, Praschak-Rieder N, Willeit M, Pezawas L, Konstantinidis A, Stastny J, & Kasper S (2005). Actigraphy in Patients with Seasonal Affective Disorder and Healthy Control Subjects Treated with Light Therapy. Biological Psychiatry, 58(4), 331–336. 10.1016/j.biopsych.2005.01.031 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement
NIHMS1842790-supplement-Supplement.docx (598.6KB, docx)

RESOURCES