Abstract
Background
This study is a network intervention analysis comparing the symptomatic response of two first-line treatments for winter seasonal affective disorder (SAD): light therapy (LT) and group cognitive-behavioral therapy for SAD (CBT-SAD).
Methods
Scores on the Beck Depression Inventory-II at mid- and post-treatment on individual items and on four previously-identified symptom clusters were examined in SAD patients randomly assigned to receive LT (n = 88) or CBT-SAD (n = 75).
Results
No treatment differences in either individual symptoms or symptom cluster scores were observed. In both mid- and post-treatment networks, symptoms were primarily associated with each other within their respective, cognitive, vegetative, and affective domains (regardless of treatment). For clusters, Loss of Vitality, Dysregulation, and Negative Cognition were associated with each other (regardless of treatment). At post-treatment only, Increased Appetite and Weight was (relatively weakly) associated with Loss of Vitality.
Conclusions
The lack of treatment differences indicates that these two mechanistically different treatments do not differ in effect on symptoms or symptom clusters. Future work should examine the association of treatment with symptom types over shorter time intervals.
Introduction
Winter seasonal affective disorder (SAD) is a form of major depression characterized by a seasonal course with recurrences in the fall and remissions in the spring (Rosenthal et al., 1984). It is estimated that 1 – 9% of adults in the general population experience SAD, with greater risk associated with higher latitude (Rosen et al., 1990) and shorter days (photoperiod). Symptoms of SAD are broadly characterized as vegetative (e.g., fatigue, increased sleep, increased appetite) or cognitive/affective (e.g., depressed mood, anhedonia, negative thought content). The interplay of cognitive and vegetative symptoms is central to the dual vulnerability model of SAD etiology (Young et al., 1991). The model proposes that individuals with SAD have both a biological vulnerability to experience vegetative symptoms in response to short photoperiods and a psychological vulnerability to develop cognitive/affective symptoms in response to vegetative symptoms. Thus, in the course of an episode vegetative symptoms should precede cognitive/affective symptoms, which has been found to be the case (McCarthy et al., 2002; Whitcomb-Smith et al., 2014; Young et al., 1991).
Meta-analyses support light therapy (LT) as an effective treatment for an acute SAD episode (Golden et al., 2005; Terman et al., 1989). LT is assumed to target and correct a physiological mechanism of SAD, circadian misalignment, by simulating an early dawn with timed daily exposure to bright artificial light (Lewy et al., 1988). Based on the phase-shift theory of SAD (Lewy et al., 1998), morning LT should advance previously phase-delayed circadian rhythms, contributing to alignment of “biological” and actual night. Therefore, LT should exert the most direct and immediate effects on the vegetative symptoms.
A specialized form of group cognitive-behavioral therapy designed for SAD (CBT-SAD; Rohan, 2008) is another empirically-validated SAD treatment and, unlike LT, is designed to target maladaptive thoughts and behaviors to improve coping with seasonal changes in the winter. Randomized clinical trials have found large and comparable improvements with both CBT-SAD and LT, with no differences in depression severities or remission rates at treatment endpoint (Rohan et al., 2015). Based on survival analyses, Meyerhoff et al. (2018) reported no difference between the treatments in time to remission for the majority of symptoms. However, early insomnia and hypersomnia remitted relatively faster with LT, which is consistent with the assumption that LT works primarily through circadian realignment. CBT-SAD has shown superior durability two winters after treatment, with less severe depression symptoms and a lower risk for recurrence (Rohan et al., 2016).
In order to examine the relative impacts of CBT-SAD and LT on the symptoms of SAD and how those symptoms then influence each other over time in treatment, a network intervention analysis (NIA; Blanken et al., 2019) was conducted. In a network framework, a disorder, such as SAD, is conceptualized as a network of interacting symptoms (Cramer et al., 2010). NIA expands on standard symptom network model by adding a binary treatment contrast variable (Blanken et al., 2019) so that direct effects of treatment on each individual symptom can be evaluated (controlling for all other symptom).
In the only published study investigating a network structure for SAD, Smetter and colleagues (2021) conducted a network analysis of baseline depression symptoms on the Beck Depression Inventory-Second Edition (BDI-II; Beck et al., 1996) using Rohan et al.’s (2015, 2016) sample of 177 currently depressed adults with SAD. Both the network analysis and a factor analysis indicated four symptom clusters. Vegetative symptoms were represented in “Increased Appetite/Weight” and “Loss of Vitality” clusters, cognitive/affective symptoms were in a “Negative Cognition” cluster, and a novel “Dysregulation” cluster crossed domains (see details below). Examination of the clusters with a Bayesian directed acyclic graph (Pearl, 2009; Rohrer, 2018) found that the two vegetative clusters influenced the cognitive/affective clusters, both directly and indirectly, consistent with the dual vulnerability model.
No prior work has examined SAD networks longitudinally or over the course of treatment. The results of Smetter et al. (2021) (with the same data as the present study) suggest that SAD symptomatology may be meaningfully understood at the cluster level as well as individual symptom level. Therefore, the present study examined the differential effects of LT and CBT-SAD both on individual symptoms and on the four symptom clusters (i.e., Negative Cognitions, Loss of Vitality, Increased Appetite/Weight, and Dysregulation) at mid-treatment and post-treatment.
Our first hypothesis was that LT would have greater direct effects than CBT-SAD on vegetative symptoms (e.g. Hypersomnia and Increased Appetite) and clusters (Loss of Vitality and Increased Appetite/Weight). We expected that LT would have an indirect effect on the cognitive and affective symptoms (e.g., Pessimism and Sadness) as well as the Negative Cognition cluster through its beneficial physiological effects on vegetative symptom clusters. Our second hypothesis was that CBT-SAD would have greater direct effects than LT on cognitive and affective symptoms as well as the Negative Cognition and Dysregulation clusters. CBT-SAD was hypothesized to have weaker direct connections with vegetative symptoms and clusters, as CBT-SAD targets cognitive and affective symptoms with vegetative symptoms influenced indirectly.
Methods
Sample
Participants were volunteers recruited primarily from the northwestern region of Vermont. For inclusion in the parent study (Rohan et al., 2015), participants (a) were ≥ 18 years old; (b) met DSM-IV criteria for Major Depression, Recurrent with Seasonal Pattern on the Structured Clinical Interview for DSM-IV Axis I Disorders-Clinician Version (First et al., 1996); (c) were currently in depressive episode based on total score ≥ 20 and atypical subscale score ≥ 5 on the Structured Interview for the Hamilton Depression Rating Scale-Seasonal Affective Disorder Version (SIGH-SAD; Williams et al., 1992); and (d) had stable antidepressant medication usage (if applicable) for at least four weeks prior to study entry with no plans to change. Participants were excluded if they (a) had current or previous treatment with light therapy or CBT for SAD; (b) met criteria for any comorbid Axis I disorder whose treatment would take priority over SAD (e.g., substance use disorder in the previous year, bipolar disorder, or any psychotic disorder); (c) had laboratory values for TSH or T4 outside of the normal range; (d) had acute and serious suicidal intent; or (e) planned to be absent for more than a week through the winter until the end of March. For more information about recruitment and demographic characteristics of the sample, please refer to the parent study (Rohan et al., 2015).
The current analysis included data from participants who completed treatment and thus completed the Beck Depression Inventory-Second Edition (BDI-II; Beck et al., 1996) at pre-treatment, mid-treatment (3 weeks), and post-treatment (6 weeks). Participants in the present study consisted of 74 of 88 (84.1%) initially randomized to CBT-SAD and 85 of 89 (95.5%) initially randomized to LT.
Procedure
Participants were recruited annually September to mid-January over six years. After initial diagnostic screening procedures, participants were administered the SIGH-SAD every 2 weeks. If/when potential participants met the SIGH-SAD threshold for a SAD episode, they underwent medical screening) and, if cleared to participate, completed a pre-treatment battery of surveys and were then randomized to 6 weeks of either CBT-SAD or LT. Randomization procedures were stratified by sex, presence/absence of a comorbid Axis I diagnosis, and antidepressant medication status. In order to avoid confounding spontaneous remission in spring with treatment effects, treatment started no later than the first week of February.
Treatments
Light Therapy (LT).
Participants randomized to LT were given a light box that produced 10,000 lux of cool-white fluorescent light delivered through an ultraviolet filter (Rohan et al., 2015). Participants were informed about how to use the device, the treatment rationale, and potential side effects. The starting dose was 30 minutes per day immediately after waking up. Daily duration was increased in 15-minute increments each week if participants did not have a reduction in SIGH-SAD scores per previous standards (Terman et al., 1990). Upon significant side effects, daily duration was decreased in 15 minute/day decrements each week until side effects subsided. The principal investigator (K.J.R.), along with a chronobiological psychiatrist with expertise in LT, monitored the LT in the study and made weekly dosing adjustments per the algorithm. For more information please refer to the parent study (Rohan et al., 2015).
Cognitive-Behavioral Therapy for SAD (CBT-SAD).
The six-week CBT intervention was an adaptation of traditional cognitive therapy for depression (Beck et al., 1979) to target SAD (Rohan, 2008). Sessions were delivered in a condensed schedule of twice-weekly 90-minute group sessions in order to be feasible before natural remission in the spring. CBT-SAD utilizes (a) behavioral activation, (b) tracking negative automatic thoughts in thought diaries, (c) identifying underlying core beliefs, (d) challenging and restructuring negative thoughts and core beliefs through Socratic questioning, and (e) maintaining gains and to prevent recurrence of depression in future winters. Typical cognitions negative thoughts concerning winter, light availability, and inclement weather conditions, as well as common depressive cognitions. CBT-SAD groups were facilitated by one of three licensed psychologists, with a clinical graduate student co-facilitator present in each. Integrity of treatment, both between therapists and across sessions, was assessed through an adapted form of the National Institute of Mental Health’s Collaborative Study Psychotherapy Rating Scale (Hollon, 1984; Hollon et al., 1984). CBT-SAD sessions were easily distinguished from LT orientation sessions, and CBT-SAD therapists showed good adherence to the manual content (Rohan et al., 2015).
Outcome Measures
Participants self-rated their depressive symptom severity using the BDI-II (Beck et al., 1996) at baseline, mid-point of treatment (3 weeks), and post-treatment (6 weeks). The BDI-II consists of 21 items assessing cognitive, affective, physiological, and behavioral symptoms of depression over the past 2 weeks. For applicability to SAD patients, the BDI-II was modified to include an additional item for weight gain, rated on the same four-point (0–3) Likert scale as the original 21 items. This scale has shown high reliability and internal consistency (Beck et al., 1996) and was used in the network and factor analyses of data from this sample at baseline (Smetter et al., 2021). Analyses employed both individual symptoms (BDI-II items) and the four clusters observed by Smetter et al. (2021) with this same dataset at baseline: Negative Cognition, Dysregulation, Increased Appetite/Weight, and Loss of Vitality. Symptom cluster variables were determined by summing item scores within each cluster without weighting in order to enhance generalizability (Einhorn & Hogarth, 1975).1 The authors decided to use the BDI-II rather than the SIGH-SAD because all BDI-II items have the same response scale whereas SIGH-SAD items do not (e.g., items are scored 0–2, 0–3, or 0–4, depending on the item). In addition, the Hamilton Depression Rating Scale (Hamilton, 1960), which constitutes the 21 of the 29 total SIGH-SAD items, has shown poor content validity and inconsistent replication of constructs across studies (Bagby et al., 2004).
Statistical Analyses
Prior to conducting the treatment network analyses, a preliminary network comparison test was used to examine evidence of networknon-equivalency at baseline, which could complicate interpretation of treatment-related changes. This test utilized the R package Network Comparison Test (Borkulo et al., 2017), which compares the network structures of the two treatment groups based on 1) overall network structure, 2) overall connectivity, 3) differences in individual edges between nodes. This analysis was performed only on the baseline networks to ensure that there was no difference in symptom or symptom cluster networks between the groups at baseline. Individual symptom magnitude differences between treatment groups at baseline were also examined and reported.
In a typical psychopathology network model, each symptom is a “node,” with the unique association between each node pair (i.e., partial correlation controlling for all other symptoms) designated as “edges.” In the present study, in order to assess the relative impact of treatments on symptoms and symptom clusters, a network intervention analysis (NIA; Blanken et al., 2019) was conducted using as nodes, in four separate analyses, the individual symptoms and the symptom cluster scores at mid-treatment and at post-treatment2. By including treatment allocation as a node, each treatment-symptom edge represents the direct differential effect of treatment (0 =LT; 1 = CBT-SAD) on the symptom or symptom cluster, controlling for all other symptoms or symptom clusters. Thus, a positive treatment-symptom edge indicates that CBT-SAD had a stronger effect on the symptom/symptom cluster than LT, whereas a negative edge indicates that LT had a stronger effect on the symptom/symptom cluster. An edge being absent from the network (i.e., set to zero by network procedures) indicates a lack of evidence for differential treatment being non-zero (Blanken et al., 2021).
NIA was implemented using the R package Mixed Graphical Model (mgm; Haslbeck & Waldorp, 2020; (version 1.2–7), which treats each symptom or symptom cluster score as a continuous variable and the treatment variable as binary. The least absolute shrinkage and selection operator (LASSO; Epskamp & Fried, 2018) reduces likely false positive edges, setting them to zero. This package has previously been used in NIA studies (Blanken et al., 2019, 2021; Mullarkey et al., 2019).
Results
Preliminary Analyses
The sample (N = 159) was predominantly white (n = 147, 92.5%), female (n = 134, 84.3%), and fully employed (n = 117, 73.6%). The two treatment groups were similar in demographic characteristics, as well as BDI-II depression severity at baseline, mid-treatment, and post-treatment (Table 1).
Table 1.
Demographic characteristics and baseline depression severity (N = 159)
| CBT-SADb (n = 74) | LTc (n = 85) | Total | ||
|---|---|---|---|---|
| BDI-IIa | Mean (SD) | Mean (SD) | Mean (SD) | Statistic |
| Baseline | 22.8 (9.3) | 24.0 (8.8) | 23.5 (9.0) | t(157) = 0.85, p = 0.40 |
| Mid-Treatment | 13.6 (6.9) | 12.4 (7.4) | 13.0 (7.1) | t(157) = −0.99, p = 0.33 |
| Post-Treatment | 7.3 (5.5) | 7.2 (6.0) | 7.3 (5.8) | t(157) = −0.06, p = 0.95 |
| CBT-SAD (n = 74) | LT (n = 85) | Total | ||
| Variable | n (%) | n (%) | n (%) | Statistic |
| Sex | ||||
| Male | 10 (13.5) | 15 (17.6) | 25 (15.7) | X(1) = 0.51, p = 0.48 |
| Female | 64 (86.5) | 70 (82.4) | 134 (84.3) | |
| Age (SD) | 46.5 (12.5) | 43.9 (12.9) | 45.1 (12.7) | t(157) = −1.25, p = 0.21 |
| Race | ||||
| White | 71 (95.9) | 76 (89.4) | 147 (92.5) | X(5) = 5.44, p = 0.37 |
| Asian | 0 (0.0) | 1 (1.2) | 1 (0.6) | |
| African American | 1 (1.4) | 1 (1.2) | 2 (1.3) | |
| American Indian | 0 (0.0) | 5 (5.9) | 5 (3.1) | |
| Hispanic | 1 (1.4) | 1 (1.2) | 2 (1.3) | |
| Marital status | ||||
| Married | 49 (66.2) | 57 (67.1) | 106 (66.7) | |
| Divorced | 6 (8.1) | 12 (14.1) | 18 (11.3) | X(3) = 3.14, p = 0.37 |
| Widowed | 5 (6.8) | 2 (2.4) | 7 (4.4) | |
| Single | 14 (18.9) | 14 (16.5) | 28 (17.6) | |
| Highest education level | ||||
| High School Degree | 7 (9.5) | 2 (2.4) | 9 (5.7) | |
| Some College | 16 (21.6) | 18 (21.2) | 34 (21.4) | X(4) = 8.17, p = 0.09 |
| College Degree | 28 (37.8) | 28 (32.9) | 56 (35.2) | |
| Some Graduate | 5 (6.8) | 16 (18.8) | 21 (13.2) | |
| Graduate Degree | 18 (24.3) | 21 (24.7) | 39 (24.5) |
Beck Depression Inventory II
Cognitive Behavioral Therapy for Seasonal Affective Disorder
Light therapy
There was no evidence of non-equivalency of the baseline symptom networks between the two treatment groups for either individual symptoms or symptom clusters. As per the three domains tested in Network Comparison Test package (Borkulo et al. 2017), there was no difference between LT and CBT-SAD in terms of network structure for individual symptoms (M <0.01, p = >.99) or symptom clusters (M = 0.26, p = 0.39), overall network connectivity of individual symptoms(LT: <0.01, CBT-SAD: <0.01, S = <0.01, p >.99) or clusters (LT: 1.40, CBT-SAD: 1.30, S = 0.10, p = 0.57), or edge weights of pairs of symptom clusters (Table 2). Additionally, there were no significant group differences in the mean magnitudes of individual items at baseline (Table 3). Symptom cluster scores at each timepoint are presented in Figure 1.
Table 2.
Comparing edge strengths of baseline symptom networks between those who received CBT-SAD and LT.
| Cluster 1a | Cluster 2 | CBT-SAD Edge Strength | LT Edge Strength | p Value |
|---|---|---|---|---|
| Negative Cognition | Loss of Vitality | 0.44 | 0.42 | 0.81 |
| Negative Cognition | Dysregulation | 0.19 | 0.22 | 0.90 |
| Negative Cognition | Increased Appetite and Weight | 0.00 | 0.00 | 0.45 |
| Loss of Vitality | Dysregulation | 0.37 | 0.23 | 0.38 |
| Loss of Vitality | Increased Appetite and Weight | 0.30 | 0.00 | 0.08 |
| Dysregulation | Increased Appetite and Weight | 0.00 | 0.15 | 0.31 |
Designation as cluster 1 and cluster 2 is arbitrary.
Table 3.
Comparison of individual item means on the Beck Depression Inventory-II between the two treatment groups at baseline
| CBT-SADa
(n = 74) |
LTb
(n = 85) |
|||
|---|---|---|---|---|
| Symptom | M(SD) | M(SD) | t(157) | p |
| Sadness | 0.9(0.4) | 0.9(0.6) | 0.42 | .68 |
| Pessimism | 0.9(0.6) | 0.9(0.6) | −0.89 | .19 |
| Past Failures | 1.0(0.9) | 1.1(0.9) | 0.37 | .71 |
| Loss of Pleasure | 1.4(0.7) | 1.3(0.5) | −0.11 | .92 |
| Guilty Feelings | 0.7(0.7) | 0.9(0.6) | 1.65 | .17 |
| Punishment Feelings | 0.3(0.8) | 0.5(0.9) | 1.55 | .12 |
| Self-Dislike | 1.3(0.8) | 1.3(0.8) | 0.05 | .96 |
| Self-Criticism | 1.0(0.9) | 1.1(0.8) | 0.23 | .82 |
| Suicidal Thoughts | 0.3(0.4) | 0.3(0.4) | 0.46 | .65 |
| Crying | 1.1(1.0) | 1.0(1.1) | −0.33 | .74 |
| Agitation | 0.8(0.8) | 0.8(0.7) | −0.50 | .62 |
| Loss of Interest | 1.4(0.8) | 1.3(0.7) | −0.40 | .69 |
| Indecisiveness | 1.2(0.9) | 1.2(0.8) | 0.52 | .60 |
| Worthlessness | 0.8(0.8) | 0.9(0.8) | 0.57 | .57 |
| Loss of Energy | 1.5(0.6) | 1.6(0.7) | 0.75 | .45 |
| Hypersomnia | 0.9(0.8) | 1.0(0.9) | 0.73 | .47 |
| Irritability | 1.1(0.8) | 1.3(0.8) | 1.45 | .15 |
| Increased Appetite | 1.0(0.9) | 0.9(1.0) | −0.37 | .71 |
| Concentration | 1.3(0.7) | 1.4(0.6) | 0.70 | .49 |
| Tiredness/Fatigue | 1.5(0.7) | 1.5(0.7) | 0.05 | .96 |
| Loss of Interest in Sex | 1.0(0.9) | 1.3(1.0) | 1.93 | .06 |
| Increased Weight | 0.5(0.7) | 0.6(0.9) | 0.73 | .47 |
Cognitive Behavioral Therapy for Seasonal Affective Disorder
Light therapy
Figure 1. Mean symptom cluster scores across time in two treatment groups.
Note. CBT-SAD = cognitive-behavioral therapy tailored for SAD. LT = light therapy.
Mid-Treatment Network.
For the individual symptoms, the treatment node had no positive or negative edges with any of the symptoms, indicating no difference between the two treatments in the magnitude of individual item scores by mid-treatment. Among the symptom-to-symptom associations, there were positive associations between items assessing similar constructs, such as “Loss of Interest” and “Loss of Pleasure” (Figure 2). The edges between symptoms had a sizeable range of strength, with associations between similar constructs being stronger (e.g., Tiredness/Fatigue and Loss of Energy; 0.34) and the few edges across differing constructs being weaker (e.g., Loss of Interest and Sadness; 0.07). Further details of the partial correlations between symptoms are presented in Table 4.
Figure 2.
Network of treatment with individual symptoms at mid-treatment
Table 4.
Partial correlations between treatment and individual symptoms at mid-treatment
| Node | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Treatment Groupa | - | ||||||||||||||||||||||
| 2. Sadness | .00 | - | |||||||||||||||||||||
| 3. Pessimism | .00 | .00 | - | ||||||||||||||||||||
| 4. Past Failures | .00 | .00 | .25 | - | |||||||||||||||||||
| 5. Loss of Pleasure | .00 | .10 | .00 | .00 | - | ||||||||||||||||||
| 6. Guilty Feelings | .00 | .00 | .00 | .11 | .00 | - | |||||||||||||||||
| 7. Punishment Feelings | .00 | .00 | .00 | .09 | .00 | .00 | - | ||||||||||||||||
| 8. Self-Dislike | .00 | .00 | .11 | .00 | .11 | .00 | .14 | - | |||||||||||||||
| 9. Self-Criticism | .00 | .00 | .00 | .00 | .07 | .17 | .00 | .16 | - | ||||||||||||||
| 10. Suicidal Thoughts | .00 | .00 | .06 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||||||||||
| 11. Crying | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||||||||
| 12. Agitation | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||||||||
| 13. Loss of Interest | .00 | .07 | .00 | .00 | .25 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||||||
| 14. Indecisiveness | .00 | .00 | .00 | .08 | .00 | .00 | .00 | .00 | .16 | .00 | .00 | .10 | .00 | - | |||||||||
| 15. Worthlessness | .00 | .14 | .00 | .26 | .00 | .00 | .00 | .00 | .17 | .16 | .00 | .00 | .00 | .00 | - | ||||||||
| 16. Loss of Energy | .00 | .00 | .00 | .00 | .07 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .08 | .09 | .00 | - | |||||||
| 17. Hypersomnia | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||
| 18. Irritability | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .10 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||
| 19. Increased Appetite | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||
| 20. Concentration | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .32 | .00 | .00 | .00 | .00 | .00 | - | |||
| 21. Tiredness/Fatigue | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .08 | .00 | .00 | .00 | .10 | .00 | .00 | .34 | .16 | .00 | .00 | .00 | - | ||
| 22. Loss of Interest in Sex | .00 | .00 | .00 | .00 | .13 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .09 | .00 | .00 | .00 | .00 | .00 | - | |
| 23. Increased Weight | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .16 | .00 | .00 | .00 | - |
| M | 0.5 | 0.5 | 0.5 | 0.5 | 0.8 | 0.3 | 0.1 | 0.7 | 0.6 | 0.1 | 0.4 | 0.4 | 0.7 | 0.7 | 0.4 | 1.0 | 0.6 | 0.6 | 0.5 | 0.9 | 1.0 | 0.6 | 0.2 |
| SD | 0.5 | 0.5 | 0.5 | 0.7 | 0.6 | 0.5 | 0.4 | 0.8 | 0.7 | 0.3 | 0.7 | 0.6 | 0.6 | 0.7 | 0.6 | 0.6 | 0.7 | 0.6 | 0.7 | 0.6 | 0.6 | 0.7 | 0.5 |
0 = Light therapy, 1 = CBT-SAD.
Note. Correlations of .00 were set to zero by network regularization process
For the symptom clusters, there also were no edges between the treatment node and any of the clusters (Figure 3). Among the cluster-to-cluster associations, positive edges connected Negative Cognition with Loss of Vitality and Dysregulation, and Loss of Vitality with Dysregulation (Table 5). These edges were similar in strength (0.27 – 0.37). Notably, Increased Appetite and Weight at mid-treatment was not associated with any other symptom cluster.
Figure 3.
Network of treatment with raw cluster scores at mid-treatment
Table 5.
Partial correlations between treatment and symptom clusters at mid-treatment
| Node | Treatment | Negative Cognition | Loss of Vitality | Dysregulation | Increased Appetite and Weight |
|---|---|---|---|---|---|
| Treatment | - | ||||
| Negative Cognition | .00 | - | |||
| Loss of Vitality | .00 | .36 | - | ||
| Dysregulation | .00 | .27 | .37 | - | |
| Increased Appetite and Weight | .00 | .00 | .00 | .00 | - |
Note. Correlations of .00 were set to zero by network regularization process
Post-Treatment Network.
For the individual symptoms, again the treatment node had no positive or negative edges with any of the symptoms, indicating no difference between the two treatments in the magnitude of individual item scores by post-treatment. Among the symptom-to-symptom associations, there were positive associations between items assessing similar constructs. This network had more cross-construct symptom connections in comparison to those at mid-treatment, although these tended to be weaker than those within-construct, such as the edge between Sadness and Loss of Energy (0.12). Specifically, we see that the items “Loss of Pleasure” and “Loss of Energy” are positively associated, indicating stronger associations across symptom domains (Figure 4). Further details of the partial correlations between symptoms are presented in Table 6.
Figure 4.
Network of treatment with individual symptoms at mid-treatment
Table 6.
Partial correlations between treatment and individual symptoms at post-treatment
| Node | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Treatment Groupa | - | ||||||||||||||||||||||
| 2. Sadness | .00 | - | |||||||||||||||||||||
| 3. Pessimism | .00 | .14 | - | ||||||||||||||||||||
| 4. Past Failures | .00 | .00 | .12 | - | |||||||||||||||||||
| 5. Loss of Pleasure | .00 | .00 | .00 | .00 | - | ||||||||||||||||||
| 6. Guilty Feelings | .00 | .00 | .00 | .11 | .00 | - | |||||||||||||||||
| 7. Punishment Feelings | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||||||||||||
| 8. Self-Dislike | .00 | .00 | .00 | .00 | .00 | .08 | .00 | - | |||||||||||||||
| 9. Self-Criticism | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .06 | - | ||||||||||||||
| 10. Suicidal Thoughts | .00 | .00 | .20 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||||||||||
| 11. Crying | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||||||||
| 12. Agitation | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||||||||
| 13. Loss of Interest | .00 | .00 | .00 | .06 | .32 | .00 | .00 | .11 | .00 | .00 | .00 | .00 | - | ||||||||||
| 14. Indecisiveness | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||||||
| 15. Worthlessness | .00 | .00 | .18 | .00 | .00 | .00 | .00 | .22 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||||
| 16. Loss of Energy | .00 | .12 | .00 | .00 | .22 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .12 | .00 | - | |||||||
| 17. Hypersomnia | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||||
| 18. Irritability | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |||||
| 19. Increased Appetite | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | ||||
| 20. Concentration | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .10 | .41 | .00 | .00 | .00 | .00 | .00 | - | |||
| 21. Tiredness/Fatigue | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .10 | .00 | .00 | .00 | .14 | .00 | .00 | .35 | .16 | .00 | .00 | .00 | - | ||
| 22. Loss of Interest in Sex | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .15 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - | |
| 23. Increased Weight | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | .00 | - |
| M | 0.5 | 0.2 | 0.3 | 0.3 | 0.5 | 0.2 | 0.0 | 0.4 | 0.3 | 0.0 | 0.2 | 0.2 | 0.5 | 0.4 | 0.1 | 0.6 | 0.4 | 0.3 | 0.4 | 0.5 | 0.6 | 0.4 | 0.1 |
| SD | 0.5 | 0.4 | 0.5 | 0.5 | 0.5 | 0.4 | 0.3 | 0.6 | 0.5 | 0.2 | 0.5 | 0.4 | 0.5 | 0.6 | 0.3 | 0.6 | 0.6 | 0.5 | 0.5 | 0.6 | 0.6 | 0.7 | 0.4 |
0 = Light therapy, 1 = CBT-SAD.
Note. Correlations of .00 were set to zero by network regularization process
Again, the treatment node also had no positive or negative edges to any of the symptom clusters, indicating no difference between the two treatments in the magnitude of cluster scores by post-treatment (Figure 5). As with mid-treatment, there were connections between symptom clusters, with positive edges connecting Negative Cognition with Loss of Vitality and Dysregulation, and Loss of Vitality with Dysregulation (range 0.24 – 0.44; Table 7). Unlike the mid-treatment results, Increased Appetite and Weight had an association with Loss of Vitality (0.13), although weaker than other edges. By the end of treatment, Increased Appetite and Weight cluster was still not associated with Dysregulation and Negative Cognitions.
Figure 5.
Network of treatment with raw cluster scores at post-treatment
Table 7.
Partial correlations between treatment and symptom clusters at post-treatment
| Node | Treatment | Negative Cognition | Loss of Vitality | Dysregulation | Increased Appetite and Weight |
|---|---|---|---|---|---|
| Treatment | - | ||||
| Negative Cognition | .00 | - | |||
| Loss of Vitality | .00 | .28 | - | ||
| Dysregulation | .00 | .24 | .44 | - | |
| Increased Appetite and Weight | .00 | .00 | .13 | .00 | - |
Note. Correlations of .00 were set to zero by network regularization process
Discussion
To our knowledge, this is the first network analysis to examine SAD symptoms longitudinally over the course of treatment. NIA is particularly applicable to examine treatment response differences in the randomized clinical trial comparing two distinct SAD treatments with presumably different targets and mechanisms: physiological in LT and cognitive-behavioral in CBT-SAD. The NIA framework distinguishes the direct effect of treatment on symptoms and symptom clusters from those that are mediated by associations of other symptoms and symptom clusters. The study benefitted from prior network and factor analyses of the pre-treatment data (Smetter et al., 2021) that generated the four symptom domains for the current NIA: Negative Cognition, Loss of Vitality, Increased Weight and Appetite, and Dysregulation. In addition, the absence of treatment group differences in pre-treatment depression severity and network characteristics (at both the individual item and cluster level) indicated that NIA results were not influenced by pre-treatment differences.
The results indicated no differences between the direct effects of LT and CBT-SAD on symptoms or clusters at either mid-or post-treatment. This finding is contrary to our a priori hypotheses that LT would have greater direct effects on vegetative symptoms (i.e., Loss of Energy, Hypersomnia, etc.) or clusters (i.e., Increased Weight and Appetite, Loss of Vitality) and that CBT-SAD would have greater direct effects on the cognitive/affective symptoms (i.e. Sadness, Pessimism, Self-Criticism, etc.) and clusters (i.e., Negative Cognition, Dysregulation). Although the absence of treatment differences in symptom change is unrelated to the magnitudes of the actual symptom changes themselves, this finding is consistent with the finding of the parent study that the treatments were comparable in terms of both global depression severity and the proportions in remission at post-treatment (Rohan et al., 2015).
The clear lack of difference in symptomatic treatment response may seem difficult to reconcile with the assumed distinct cognitive and chronobiological targets and mechanisms in CBT-SAD vs. LT, respectively. Consideration of the time courses of the treatments and the treatment responses may shed some light on this issue, since neither treatment is homogenous across time. In CBT-SAD, by mid-treatment there has been extensive focus on behavioral activation (Rohan, 2008) and increasing engagement in positive activities which may, subsequently, reduce vegetative symptoms (e.g., fatigue, increased sleep, hyperphagia, etc.). In contrast, in the second half of treatment, CBT-SAD focuses extensively on cognitions, with cognitive restructuring of automatic thoughts and core beliefs, which should reduce cognitive/affective symptoms (e.g., sadness, hopelessness, self-criticism, etc.). For LT, light exposure was adjusted based on symptom improvement, so that by the third week participants may have benefitted from increasing light exposure enough so that cognitive/affective symptoms were subsequently affected. Thus, the effects of both CBT-SAD and LT on various types of symptoms may be more intermingled over time than originally thought. It also should be noted that it is possible that both types of treatment impact a common (unknown) antidepressant mechanism and so yield common outcomes (Meyerhoff et al., 2018; Stassen et al., 1993)
Secondarily, results revealed several positive associations between symptoms regardless of treatment received. At both timepoints, individual symptoms were most associated with each other within their respective domains (e.g., Loss of Interest and Loss of Pleasure). However, at post-treatment more associations between cognitive and vegetative symptoms began to emerge, although more weakly. Among the clusters at both mid-and-post treatment Negative Cognition, Loss of Vitality, and Dysregulation were associated with each other. At post-treatment only, Increased Appetite and Weight was (relatively weakly) associated with Loss of Vitality. Consistent with the dual-vulnerability model, it is possible that Loss of Vitality contributes to Increased Appetite/Weight during the second half of treatment, although the causal direction cannot be determined from these data. Unlike associations between treatment and symptoms (which can be causal in only one direction), the current symptom – symptom results do not inform the direction of these connections. Future work examining directional influences can shed light on whether the observed associations are in-line with the dual-vulnerability hypothesis (Young et al., 1991). Regarding potential mechanisms underlying the observed associations, future work could incorporate cognitive and chronobiological vulnerability-related process measures.
This study was limited in several ways. First, the BDI-II was assessed at only three time points (i.e., pre-, mid-, and post-treatment), potentially obscuring effects from week to week during treatment. As a result, either LT or CBT-SAD might influence greater change in specific symptom or symptom clusters with a higher resolution time frame. This may be especially true symptoms at mid-point, where symptoms and symptom clusters may already be engaged by their respective treatment mechanisms. For instance, LT may have direct effects on Loss of Energy at a symptom level, or in the larger cluster of Loss of Vitality in the first weeks of treatment, but not on Pessimism (and subsequent Negative Cognition), with downstream effects starting in the following weeks. Additionally, closer examination beyond mid-treatment may shed light on why Increased Appetite and Weight is not associated with other clusters, and at what point the association occurs. Shorter time intervals between assessments may reveal the points in treatment wherein treatments may act directly on certain symptoms, and at what point the treatments affect other symptom clusters indirectly. This would inform the point in treatment that affects the greatest change in a particular domain of SAD symptoms. Further, the lack of greater time resolution with more time points in this study limited examination of dynamical processes between symptoms as in typical dynamic network analyses (David et al., 2018).
Another limitation is that the unclear wording of the increased weight item of the BDI-II extended edition. Despite the scale’s overall instructions to rate symptoms during the past two weeks, this item refers to “lately.” Thus, respondents may have different interpretations, with some reporting weight gain for the duration of the depressive episode relative to their euthymic weight. A tendency to reference the timeframe of “since the episode began” may be enhanced by the option of a 15-pound weight gain, which is unlikely in two weeks.
Further, this sample was very homogenous in terms of its demographics: primarily white, female, educated, and employed. Thus, the results may not be generalizable to broader SAD populations. Future work should include more diverse samples of individuals with SAD to increase generalizability to a broader SAD population, including those with more varied identities and backgrounds. In addition, future work should use more frequent assessments of symptoms (at least weekly) and examine the directionality of these associations and the mechanisms underlying them.
Footnotes
Negative Cognition: Pessimism, Past Failures, Guilty Feelings, Punishment Feelings, Self-Dislike, Self-Criticism, Suicidal Thoughts, Crying, Worthlessness, Sadness. Dysregulation: Irritability, Concentration, Agitation, Indecisiveness. Increased Appetite/Weight: Weight Gain, Increased Appetite. Loss of Vitality: Loss of Pleasure, Loss of Interest, Loss of Energy, Tiredness/Fatigue, Loss of Interest in Sex, Hypersomnia.
Several methods for measuring change exist. We also ran analyses using raw change scores and residualized change scores (e.g.,Mullarkey et al., 2020). Results were virtually identical to those using raw scores at both mid- and post-treatment and so we report here the raw score results based on their simplicity and ease of interpretation.
References
- Bagby RM, Ryder AG, Schuller DR, & Marshall MB. (2004). The Hamilton Depression Rating Scale: Has the gold standard become a lead weight? The American Journal of Psychiatry, 161(12), 2163–2177. 10.1176/appi.ajp.161.12.2163 [DOI] [PubMed] [Google Scholar]
- Beck AT, Rush AJ, Shaw BF, & et al. (1979). Cognitive therapy of depression. Guilford Press. [Google Scholar]
- Beck AT, Steer RA, & Brown GK. (1996). Beck Depression Inventory (2nd ed.). The Psychological Corporation. [Google Scholar]
- Blanken TF, Jansson-Fröjmark M, Sunnhed R, & Lancee J. (2021). Symptom-specific effects of cognitive therapy and behavior therapy for insomnia: A network intervention analysis. Journal of Consulting and Clinical Psychology. 10.1037/ccp0000625 [DOI] [PubMed] [Google Scholar]
- Blanken TF, Van Der Zweerde T, Van Straten A, Van Someren EJW, Borsboom D, & Lancee J. (2019). Introducing Network Intervention Analysis to Investigate Sequential, Symptom-Specific Treatment Effects: A Demonstration in Co-Occurring Insomnia and Depression. Psychotherapy and Psychosomatics, 88(1), 52–54. 10.1159/000495045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Borkulo CV, Bork RV, Boschloo L, Kossakowski J, Tio P, Schoevers R, Borsboom D, & Waldorp L. (2017). Comparing network structures on three aspects: A permutation test. 10.13140/RG.2.2.29455.38569 [DOI] [PubMed] [Google Scholar]
- Cramer AOJ, Waldorp LJ, van der Maas HLJ, & Borsboom D. (2010). Comorbidity: A network perspective. Behavioral and Brain Sciences, 33(2–3), 137–150. 10.1017/S0140525X09991567 [DOI] [PubMed] [Google Scholar]
- David SJ, Marshall AJ, Evanovich EK, & Mumma GH. (2018). Intraindividual Dynamic Network Analysis – Implications for Clinical Assessment. Journal of Psychopathology and Behavioral Assessment, 40(2), 235–248. 10.1007/s10862-017-9632-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Einhorn HJ, & Hogarth RM. (1975). Unit weighting schemes for decision making. Organizational Behavior and Human Performance, 13(2), 171–192. 10.1016/0030-5073(75)90044-6 [DOI] [Google Scholar]
- Epskamp S, & Fried EI. (2018). A tutorial on regularized partial correlation networks. Psychological Methods, 23(4), 617–634. 10.1037/met0000167 [DOI] [PubMed] [Google Scholar]
- First M, Spitzer R, Gibbon M, & Williams J. (1996). Structured Clinical Interview for DSM-IV Axis I Disorders (Clinician Version). American Psychiatric Press. [Google Scholar]
- Golden RN, Gaynes BN, Ekstrom RD, Hamer RM, Jacobsen FM, Suppes T, Wisner KL, & Nemeroff CB. (2005). The efficacy of light therapy in the treatment of mood disorders: A review and meta-analysis of the evidence. The American Journal of Psychiatry, 162(4), 656–662. 10.1176/appi.ajp.162.4.656 [DOI] [PubMed] [Google Scholar]
- Hamilton M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery, and Psychiatry, 23, 56–62. 10.1136/jnnp.23.1.56 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Haslbeck JMB, & Waldorp LJ. (2020). Estimating Time-Varying Mixed Graphical Models in High-Dimensional Data. Journal of Statistical Software, 93(8). 10.18637/jss.v093.i08 [DOI] [Google Scholar]
- Hollon SD. (1984). Final report: System for rating psychotherapy audiorecordings. US Department of Health and Human Services. [Google Scholar]
- Hollon SD, Evans M, Elkin E, & Lowery HA. (1984, May). System for rating therapies for depression. Annual Meeting of the American Psychiatric Association, Los Angeles, CA. [Google Scholar]
- Lewy AJ, Sack RL, Singer CM, White DM, & Hoban TM. (1988). Winter depression and the phase-shift hypothesis for bright light’s therapeutic effects: History, theory, and experimental evidence. Journal of Biological Rhythms, 3(2), 121–134. 10.1177/074873048800300203 [DOI] [PubMed] [Google Scholar]
- McCarthy E, Tarrier N, & Gregg L. (2002). The nature and timing of seasonal affective symptoms and the influence of self-esteem and social support: A longitudinal prospective study. Psychological Medicine, 32(8), 1425–1434. 10.1017/s0033291702006621 [DOI] [PubMed] [Google Scholar]
- 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]
- Mullarkey MC, Marchetti I, & Beevers CG. (2019). Using Network Analysis to Identify Central Symptoms of Adolescent Depression. Journal of Clinical Child & Adolescent Psychology, 48(4), 656–668. 10.1080/15374416.2018.1437735 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mullarkey MC, Stein AT, Pearson R, & Beevers CG. (2020). Network analyses reveal which symptoms improve (or not) following an Internet intervention (Deprexis) for depression. Depression and Anxiety, 37(2), 115–124. 10.1002/da.22972 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nussbaumer B, Kaminski-Hartenthaler A, Forneris CA, Morgan LC, Sonis JH, Gaynes BN, Greenblatt A, Wipplinger J, Lux LJ, Winkler D, Van Noord MG, Hofmann J, & Gartlehner G. (2015). Light therapy for preventing seasonal affective disorder. The Cochrane Database of Systematic Reviews, 11, CD011269. 10.1002/14651858.CD011269.pub2 [DOI] [PubMed] [Google Scholar]
- Pearl J. (2009). Causality: Models, reasoning and inference (2nd ed.). Cambridge University Press. [Google Scholar]
- Rohan KJ. (2008). Coping with the seasons: A cognitive-behavioral approach to seasonal affective disorder: therapist guide. Oxford University Press. [Google Scholar]
- Rohan KJ, Mahon JN, Evans M, Ho S-Y, Meyerhoff J, Postolache TT, & Vacek PM. (2015). Randomized Trial of Cognitive-Behavioral Therapy Versus Light Therapy for Seasonal Affective Disorder: Acute Outcomes. The American Journal of Psychiatry, 172(9), 862–869. 10.1176/appi.ajp.2015.14101293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 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]
- Rohrer JM. (2018). Thinking Clearly About Correlations and Causation: Graphical Causal Models for Observational Data. Advances in Methods and Practices in Psychological Science, 1(1), 27–42. 10.1177/2515245917745629 [DOI] [Google Scholar]
- Rosen LN, Targum SD, Terman M, Bryant MJ, Hoffman H, Kasper SF, Hamovit JR, Docherty JP, Welch B, & Rosenthal NE. (1990). Prevalence of seasonal affective disorder at four latitudes. Psychiatry Research, 31(2), 131–144. 10.1016/0165-1781(90)90116-m [DOI] [PubMed] [Google Scholar]
- Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Goodwin FK, Davenport Y, Mueller PS, Newsome DA, & Wehr TA. (1984). Seasonal affective disorder. A description of the syndrome and preliminary findings with light therapy. Archives of General Psychiatry, 41(1), 72–80. 10.1001/archpsyc.1984.01790120076010 [DOI] [PubMed] [Google Scholar]
- Smetter JB, Antler CA, Young MA, & Rohan KJ. (2021). The symptom structure of seasonal affective disorder: Integrating results from factor and network analyses in the context of the Dual Vulnerability Model. 43, 95–107. 10.1007/s10862-020-09861-0 [DOI] [Google Scholar]
- Stassen HH, Delini-Stula A, & Angst J. (1993). Time course of improvement under antidepressant treatment: A survival-analytical approach. European Neuropsychopharmacology: The Journal of the European College of Neuropsychopharmacology, 3(2), 127–135. 10.1016/0924-977x(93)90264-m [DOI] [PubMed] [Google Scholar]
- 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. 10.1016/0893-133x(89)90002-x [DOI] [PubMed] [Google Scholar]
- 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]
- Whitcomb-Smith S, Sigmon ST, Martinson A, Young M, Craner J, & Boulard N. (2014). The Temporal Development of Mood, Cognitive, and Vegetative Symptoms in Recurrent SAD Episodes: A Test of the Dual Vulnerability Hypothesis. Cognitive Therapy and Research, 38(1), 43–54. 10.1007/s10608-013-9577-5 [DOI] [Google Scholar]
- Williams J, Link M, Rosenthal N, Amira L, & Terman M. (1992). Structured Interview Guide for the Hamilton Depression Rating Scale-Seasonal Affective Disorder Version. New York State Psychiatric Institute. [Google Scholar]
- Young MA, Watel LG, Lahmeyer HW, & Eastman CI. (1991). The temporal onset of individual symptoms in winter depression: Differentiating underlying mechanisms. Journal of Affective Disorders, 22(4), 191–197. 10.1016/0165-0327(91)90065-z [DOI] [PubMed] [Google Scholar]