Prospectively Assessed Summer Mood Status in Major Depression, Recurrent with Seasonal Pattern: Evidence for SAD’s Construct Validity

Abstract

Objective:

Seasonal patterns are often undetectable in population-based depression studies, calling into question the existence of winter seasonal affective disorder (SAD). If SAD has construct validity, individuals with SAD should show spontaneous depression remission in the summer. Data are sparse on prospectively assessed summer mood status in confirmed SAD patients.

Method:

We conducted prospective summer followup of community adults who, the winter before, were diagnosed with Major Depression, Recurrent with Seasonal Pattern on the Structured Clinical Interview for DSM-IV Axis I Disorders, developed a current SAD episode on the Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version (SIGH-SAD), and enrolled in a clinical trial comparing group cognitive-behavioral therapy for SAD and light therapy. In July/August after treatment, 143/153 (93.5%) participants provided data on the SIGH-SAD, the Beck Depression Inventory-Second Edition, and the Longitudinal Interval Followup Evaluation (LIFE).

Results:

Summer mean depression scores were in the normal range, with the substantial majority in remission across different measures. On the LIFE, 113/143 (79.0%) experienced complete summer remission, 19/143 (13.3%) experienced partial summer remission, and 11/143 (7.7%) had major depression in the summer. Depression scores were significantly lower at summer than post-treatment in both treatments, indicating incomplete treatment response.

Limitations:

This was a single-site study with a relatively homogeneous sample.

Conclusions:

Supporting construct validity for SAD, the substantial majority experienced complete summer remission, with a minority in partial remission and a very small minority in episode. Both treatments left residual symptoms at treatment endpoint compared to summer.

Introduction

Winter seasonal affective disorder (SAD) is characterized as a seasonally recurrent depression subtype, involving major depressive episodes in the fall/winter that remit in spring/summer (Rosenthal et al., 1984). SAD is not an official diagnostic label. However, the phenomenon of SAD is captured in the seasonal pattern course specifier for a mood disorder diagnosis using the American Psychiatric Association’s Diagnostic and Statistical Manual (since DSM-III-R) or the World Health Organization’s International Classification of Diseases (since ICD-11).

Data are mixed on whether seasonal patterns are detectable in depression surveillance studies, including in population-based studies with large, representative samples. A review of this literature (Øverland et al., 2020) included 10 studies of depression prevalence, with survey locations in the United States, Canada, and Europe. Inclusion criteria required the use of interviews/measures that did not explicitly inquire about seasonal changes, thereby minimizing confirmation bias as a potential confounder, but study designs varied from cross-sectional to repeated-measures. Five of 10 studies observed higher depression prevalence in winter than in summer. Of note, the study with the largest sample size (over 500,000 when pooling data across two Canadian national surveys) was among the five positive studies (Patten et al., 2017). Patten et al. (2017) found significant differences in month-specific proportions of major depression based on the International Diagnostic Interview, with the distribution displaying a parabolic shape when months were coded from 0 (January) to 11 (December). The proportions of respondents in a major depressive episode were highest in December, January, and February and were lowest in June, July, and August—a pattern that persisted across stratification by age, sex, and latitude (Patten et al., 2017). Of the remaining five studies reviewed by Øverland et al., three found no evidence of seasonal variation in depression prevalence and two reported seasonal patterns outside of winter (e.g., highest depression prevalence in spring or summer).

A study of over 2,000 U.S. children and adolescents, aged 6 to 18 years, from a nationally representative population survey (Nillni et al., 2009) is relevant but was presumably excluded from Øverland et al.’s (2020) review because they excluded studies focused on children. Nillni et al. (2009) found no differences between assessments conducted in September-February versus March-August in depression prevalence, defined as the proportion scoring above the clinical cutpoint on the withdrawn/depressed syndrome scale of the parent-completed Child Behavior Checklist, corresponding to ≥93^rd percentile based on norms. In another study not reviewed by Øverland et al. (2020), Lukmanji et al. (2019) examined depression prevalence using a cutpoint of 5+ on the Patient Health Questionnaire-9 (PHQ-9) in a Canadian national survey of approximately 8,000 youths (aged 12–24) and 45,000 adults aged 25+. Lukmaji et al. (2019) replicated Patten et al.’s (2017) finding of higher major depression prevalence in winter than in summer in youths with the same parabolic shape, but not in adults where the distribution was quite flat. This is consistent with Øverland et al. (2020)’s review in that both included studies that used Patient Health Questionnaire also failed to support a winter elevation in depression prevalence in U.S. adults (Doganer et al., 2015; Traffanstedt, Mehta, & LoBello, 2016).

Traffanstedt et al. (2016) argue that a failure to detect seasonal patterns in their study indicates that SAD does not exist and that the seasonal course specifier should be dropped from diagnostic systems. They wrote, “The idea of seasonal depression may be strongly rooted in folk psychology, but it is not supported by objective data” (Traffanstedt et al., 2016, p. 1). We do not question Traffanstedt et al.’s (2016) rigorous methods or the validity of their findings, which are commensurate with the broader literature reviewed above, showing that seasonal patterns in depression prevalence surveillance are only detected in about half of studies (Øverland et al., 2020). However, we view their conclusions as overgeneralized, given that the data are more mixed than they acknowledge, the cross-sectional nature of their data, and the low base rate of seasonal depression in the general population when best methods (i.e., clinical diagnostic interviews) are used. For example, the United States National Comorbidity Study found lifetime prevalence rates of only 0.4% and 1.0% for major and minor depression with a seasonal pattern, respectively (Blazer, Kessler, & Swartz, 1998). Therefore, a failure to detect season, photoperiod, or latitude effects on depression prevalence in a population representative dataset, such as the 2006 Behavioral Risk Factor Surveillance System used by Traffanstedt et al. (2016), is not unexpected and does not, by itself, invalidate the construct of SAD.

Several formal responses to Traffanstedt et al. (2016) have been published (e.g., Young, 2017). Winkler et al. (2017) wrote, “One remarkable aspect of…Traffanstedt et al. is the dismissal of dozens of studies from multiple research centers documenting the existence of thousands of patients who meet criteria for SAD” (p. 55). We agree that this is a valid criticism but note that very few studies have prospectively followed carefully diagnosed SAD patients in the summer to determine if spontaneous remission of fall/winter depression occurs. Such studies are needed to help inform SAD’s construct validity.

As most SAD studies do not include a prospective summer followup, summertime mood status in SAD is not well characterized. Although DSM-5 criteria for the Seasonal Pattern Specifier require full remission in the non-affected (summer) months, available prospective studies support a subgroup of patients with fall/winter depression and incomplete summer remission (Danilenko & Putilov, 1996; Lam et al., 2001), which is associated with less improvement in symptoms after a trial of light therapy (Lingjærde & Regine Føreland, 1999). In addition, some people with SAD may later develop a nonseasonal course of depression (Lam et al., 2001). Indeed, the Seasonal Pattern Specifier requires seasonal major depressive episodes in each of the last two years, and overwhelmingly more seasonal than nonseasonal episodes across the lifetime course of the mood disorder at the time of diagnosis (American Psychiatric Association, 2013). Therefore, these criteria allow for nonseasonal depressive episodes to occur rarely, for people to be occasionally subsyndromal or asymptomatic in the typically affected season, and for people to transition out of the seasonal course specifier as they sustain recovery or develop a nonseasonal course. Accordingly, in any prospective summer followup of individuals diagnosed with Major Depression, Recurrent with Seasonal Pattern the winter before, the substantial majority would be expected to experience complete remission, with a minority in partial remission and a very small minority in episode.

Contributing to the debate on SAD’s construct validity, this study reports the largest prospective summer followup of individuals with SAD to date. Participants were community adults identified as SAD cases using accepted methods. Specifically, they were diagnosed with Major Depression, Recurrent with Seasonal Pattern on the Structured Clinical Interview for DSM-IV Axis I Disorders (including ensuring that the pattern is not better explained by seasonally-varying psychosocial stressors and tasks) and met criteria for current SAD episode on the Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version (SIGH-SAD). Participants were enrolled in a randomized clinical trial comparing SAD-tailored group cognitive-behavioral therapy (CBT-SAD) to light therapy (LT) during the fall/winter months and returned for prospective followup the next summer (in July or August). If SAD has construct validity, most of these individuals should show spontaneous depression remission the next summer.

Beyond characterizing summer mood status in SAD and informing construct validity, clinical trials that include a summer followup of SAD patients treated the previous winter can examine the “completeness” of acute treatment response by comparing mood status at treatment endpoint to summer. In a sample of 15 adults with SAD, Postolache et al. (1998) observed significantly higher depression scores after two weeks of LT relative to the next summer. This preliminary study supports clinical observations that individuals with SAD feel their very best in the summer, and even efficacious treatments leave residual symptoms.

The current study has two aims. First, it presents mood outcomes at the prospective summer followup after participating in a randomized clinical trial comparing CBT-SAD and LT, overall and by treatment condition. This objective contributes towards improved description of the summer mood status among individuals with SAD in the broader field and towards the debate on SAD’s construct validity by testing whether carefully diagnosed SAD cases show the expected pattern of summer remission. We report depression scores, remission status per SIGH-SAD and Beck Depression Inventory-Second Edition (BDI-II) criteria, and current depressive episode status per DSM-5 and SIGH-SAD criteria. If SAD has construct validity, mean depression scores should be relatively low in the summer, with the substantial majority in complete remission and a slight minority in episode. We test for, but did not expect to find, significant differences between CBT-SAD and LT on summer outcomes. As a second aim, this study examines the “completeness” of acute response to CBT-SAD vs. LT by comparing mood status at treatment endpoint to summer. Similar to Postolache et al.’s (1998) finding with LT, we expected lower depression severity at summer relative to post-treatment, indicative of residual symptoms within each treatment.

Methods

Design Overview

This study was a randomized clinical trial conducted at the University of Vermont and was approved by the University’s institutional review board. Participants (N=177) were community adults, aged 18 or older, randomized to six weeks of LT or CBT-SAD and assessed weekly during treatment, at treatment endpoint, the next summer in July/August, and at followups one and two winters later. The informed consent document stated the timing of the followup visits and described the procedures that would occur at each. The summer assessment was added to the design after the first cohort of 24 subjects, leaving 153 randomized subjects eligible to provide summer data. Prior publications detail the study protocol (Rohan et al., 2013), participant demographics and baseline characteristics and treatment fidelity (Rohan et al., 2015), and participant (CONSORT) flow and treatment outcomes at post-treatment (Rohan et al., 2015) and at the two winter followups (Rohan, Meyerhoff et al., 2016). Participant demographics and baseline characteristics for the sample that provided summer followup data are provided in Table 1. Participants were predominantly women, White, non-Latina/o, and college-educated. Study participants met current criteria for: 1) Major Depression, Recurrent with Seasonal Pattern on the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID; First et al., 1995) and 2) a current depressive episode on the SIGH-SAD. After the SCID, the SIGH-SAD was administered to SCID-eligible individuals. If SIGH-SAD depression criteria were not met initially, it was repeated every-other week through the end of January to detect possible depression onset. Individuals were excluded who endorsed prior treatment with CBT-SAD or LT, planned or current LT or psychotherapy for depression, met criteria for a DSM-IV disorder requiring more immediate treatment, expressed current and serious suicidal intent, reported new antidepressant use or changed antidepressant dose in the past month, or tested positive for hypothyroidism per medical workup.

Table 1.

Baseline characteristics of participants who provided data at summer followup (N=143)

	Total Sample (N=143)		CBT-SADa (n=71)		Light Therapy (n=72)			Analysis
	Mean	SD	Mean	SD	Mean	SD	t	df	p
Age (years)	45.3	13.0	46.6	13.0	44.1	13.0	1.18	141	0.24
	No.	%	No.	%	No.	%	χ2	df	p
Sex							0.17	1	0.68
Male	24	16 .8	11	15 .5	13	18 .1
Female	119	83.2	60	84.5	59	81.9
Race							6.36	5	0.27
White	130	90 .9	66	93 .0	64	88 .9
Asian	1	0.7	1	1.4	0	0.0
African American	2	1.4	1	1.4	1	1.4
Hispanic	3	2.1	2	2.8	1	1.4
American Indian	5	3.5	0	0.0	5	6.9
Other	2	1.4	1	1.4	1	1.4
Marital status							1.46	3	0.69
Married or living with partn	88	61 .5	41	57 .8	47	65 .3
Divorced	19	13.3	9	12.7	10	13.9
Widowed	5	3.5	3	4.2	2	2.8
Single	31	21.7	18	25.4	13	18.1
Highest education level							4.17	4	0.38
High school diploma	10	7. 0	7	9. 9	3	4. 2
Some college	33	23.1	18	25.4	15	20.8
College degree	50	35.0	26	36.6	24	33.3
Some graduate school	17	11.9	6	8.5	11	15.3
Graduate degree	33	23.1	14	19.7	19	26.4
Employment status							1.67	3	0.64
Employed full-time	108	76 .6	56	80 .0	52	73 .2
Employed part-time	5	3.6	2	2.9	3	4.2
Retired	7	5.0	4	5.7	3	4.2
Unemployed	21	14.9	8	11.4	13	18.3
Comorbid diagnosis							0.02	1	0.90
Yes	41	28.7	20	28 .2	21	29 .2
No	102	71.3	51	71.8	51	70.8
Antidepressant medication status							0.003	1	0.96
Yes	38	26.6	19	26 .8	19	26 .4
No	105	73.4	52	73.2	53	73.6

^aCBT-SAD = SAD-tailored cognitive-behavioral therapy.

The parent study’s primary efficacy results are published. To summarize, both treatments were associated with large symptom improvements during treatment, with no differences between conditions at treatment endpoint in depression scores or the proportions in remission (47.6% in CBT-SAD and 47.2% in LT; Rohan et al., 2015). The treatments did not differ at followup one winter later; however, two winters later, CBT-SAD was associated with less severe symptoms and fewer depression recurrences than LT (27.3% in CBT-SAD vs. 45.6% in LT; Rohan, Meyerhoff et al., 2016).

Treatments

Light Therapy (LT).

LT was administered via a standard light box that emits 10,000 lux of cool-white fluorescent light through an ultraviolet filter (SunRay^® by SunBox Company, Gaithersburg, MD). LT was initiated at 30-minutes per day immediately upon waking, and titrated weekly using a treatment algorithm that accounted for treatment response and side effects (Rohan et al., 2013). Duration adjustments were made in consultation with the study psychiatrist (T.T.P.). Final LT doses are published (Rohan et al., 2015).

CBT-SAD.

CBT-SAD (Rohan, 2008) was delivered via twelve 90-minute group therapy sessions held twice a week for six weeks. Content included standard elements of CBT for depression (i.e., psychoeducation, behavioral activation, cognitive restructuring, and relapse prevention) and was tailored specifically to address coping with the winter season and reduced daylight. Groups were facilitated by the study principal investigator (K.J.R.) or one of two doctoral-level psychologists in the community with a graduate student co-facilitator.

Summer Outcome Measures

Structured Interview Guide for the Hamilton Rating Scale for Depression-Seasonal Affective Disorder Version (SIGH-SAD).

The SIGH-SAD (Williams et al., 1992) is a semi-structured interview that assesses the presence and severity of depression symptoms and includes the 21-item Hamilton Depression Rating Scale (HAM-D) and the 8-item subscale of atypical symptoms. SIGH-SAD outcomes at summer include total SIGH-SAD score, depression remission, and depression recurrence. Based on Terman et al. (1990), depression remission on the SIGH-SAD was classified as either: 1) at least 50% improvement of SIGH-SAD from pre-treatment to summer followup, with a HAM-D score ≤ 7 and atypical score ≤ 7, or 2) HAM-D score ≤ 2 and atypical score ≤ 10 at summer followup. Threshold for depressive episode (at study enrollment as well as at followup) was total SIGH-SAD score ≥ 20, including atypical subscale score ≥ 5 (Terman et al. 1990). The study followed our protocol for SIGH-SAD rater training and item scoring rules as articulated in Rohan, Rough et al. (2016). Inter-rater reliability between the initial blinded interviewer and a second blinded rater of the audio-recording was high (ICCs > .96 at pre-treatment and summer followup; see Rohan, Rough et al., 2016).

Beck Depression Inventory-Second Edition (BDI-II).

The BDI-II (Beck et al., 1996) is a 21-item self-report depression symptom measure. BDI-II outcomes at summer include total score and remission status, defined as BDI-II score ≤ 8.

Longitudinal Interval Follow-up Evaluation (LIFE).

The LIFE (Keller et al., 1987), a semi-structured interview that tracks the longitudinal course of psychiatric disorders, was used to assess presence and severity of major depression symptoms in the summer. A LIFE trainer from M.B. Keller’s group provided onsite training to all raters at study outset, which was recorded and reviewed annually during data collection. Using standard LIFE interview probes, trained, blinded clinical psychology graduate student interviewers collected weekly psychiatric status ratings (PSR) for DSM-IV major depression on a scale from 1 (usual self; no residual depressive symptoms) to 6 (definite criteria and extreme impairment) for the interim between treatment endpoint and the summer assessment. The LIFE was also used to identify summer remission status. Complete summer remission was defined as negative for both of the following between June 1 and interview date: 1) ratings of 3 (i.e., obvious evidence of depression with moderate impairment, but considerably less than in full major depression) or higher for four or more consecutive or non-consecutive weeks, and 2) any two-week period or longer during which DSM-IV major depressive episode criteria were met. We allowed for up to three weeks of subsyndromal depressive symptoms in summer because individuals with SAD are vulnerable to depression, in general (Lam et al. 2001), and may experience situationally depressed mood, even in the summer. Subjects without a complete remission were classified as either partial remission (met only the first criterion) or major depressive episode (met the second criterion with or without satisfying the first).

Data Analysis

Although no differences were expected, we compared CBT-SAD and LT on dichotomous summer outcomes (depression episode status and remission status) using Pearson Chi-Square tests and on continuous summer outcomes (depression scores) using Wilcoxon rank sum tests because the data were not normally distributed. Post-treatment and summer SIGH-SAD and BDI-II scores were analyzed using mixed effects linear regression models to test the Treatment × Time interaction and the main effects of Treatment and Time. The mixed effects linear regression analyses included all available data on subjects in the intent-to-treat sample, using a likelihood-based approach for estimation of missing observations. To obtain valid statistical tests, SIGH-SAD and BDI-II scores were square root transformed to meet the normality assumption for these analyses. All statistical analyses were conducted using SAS 9.4 software (SAS Institute, Inc., Cary, NC, USA).

Results

Summer Outcomes

Descriptive statistics for summer SIGH-SAD and BDI-II outcomes, overall and within each treatment condition, and statistics comparing the treatments are presented in Table 2. Mean SIGH-SAD and BDI-II scores were in the normal range, the majority were in remission on each measure, and a slight minority met SIGH-SAD depression criteria. On the SIGH-SAD, 31/143 (21.7%) were neither in remission nor in episode. As expected, CBT-SAD and LT did not differ on any summer outcome (all U-statistics and Χ² tests were ns). The majority (113/143, 79.0%) of the sample met criteria for complete summer remission on the LIFE. Of the remaining 30 subjects, 19/143 (13.3%) had a partial summer remission and 11/143 (7.7%) had a summer major depressive episode. Ten of the 11 subjects with a major depressive episode in the summer failed both criteria for complete summer remission (i.e., PSR ratings ≥3 for any four or more weeks, as well as the ≥ 2 consecutive weeks of PSR ratings ≥ 5 required for major depressive episode).

Table 2.

Outcomes at Summer Followup, Overall and within each Treatment Condition

	Overall		Within Treatment Condition
	(N=143)		CBT-SADa (n=71)		Light Therapy (n=72)		Statistical Analysis
	M	SD	M	SD	M	SD	U		p
Depression Scores
SIGH-SADb	7.92	7.01	7.97	7.52	7.86	6.51	5096.5		0.95
HAM-Dc	5.73	4.71	5.80	4.80	5.67	4.66	5205.5		0.71
Atypicald	2.18	2.99	2.17	3.15	2.19	2.85	5044		0.78
BDI-IIe	4.12	5.58	3.91	6.63	4.33	4.35	4642		0.06
	No./Total N	%	No./n	%	No./n	%	X2	df	p
Remission Status
SIGH-SADf	103/143	72.0	53/71	74.7	50/72	69.4	0.48	1	0.49
BDI-IIg	123/143	86.0	64/71	90.1	59/72	81.9	2.00	1	0.16
Depressive Episodeh	9/143	6.3	6/71	8.5	3/72	4.2	1.11	1	0.33

^aCBT-SAD = SAD-tailored cognitive-behavioral therapy.

^bSIGH-SAD = Structured Interview Guide for the Hamilton Rating Scale for Depression—Seasonal Affective Disorder Version.

^cHAM-D = 21-item Hamilton Rating Scale for Depression.

^dAtypical = 8-item, atypical subscale score of the SIGH-SAD.

^eBDI-II = Beck Depression Inventory—2^nd Edition.

^f≥ 50% improvement in summer SIGH-SAD relative to pre-treatment+ HAM-D ≤ 7 + Atypical ≤ 7 OR HAM-D ≤ 2 + Atypical ≤ 10.

^gBDI-II ≤ 8.

^hTotal SIGH-SAD score > 20 + HAM-D score > 10 + Atypical score > 5.

Completeness of Treatment Response Relative to Summer Mood Status

Comparison of mean SIGH-SAD score at post-treatment vs. summer by treatment condition yielded a significant main effect of time [F(1,140)=50.76, p<0.0001], with lower score means at summer than post-treatment in both treatments (Cohen’s d=0.62). The Treatment × Time interaction was ns [F(1,140)=0.63, p=0.43]. The component subscales of the SIGH-SAD, the HAM-D and atypical scales, showed the same pattern of results. For the BDI-II, there was a significant Treatment × Time interaction [F(1,139)= 4.18, p=0.04] and a significant Time main effect [F(1,139)=62.76,p<0.0001]. The time effect was significant within each treatment, with lower BDI-II score means at summer than post-treatment [F(1,139)=49.03,p<0.0001 for CBT-SAD and F(1,139)=17.45,p<0.0001 for LT]. However, the treatments did not differ significantly on mean BDI-II at post-treatment [F(1,139)=1.46, p=0.23] or summer [F(1,139)=1.40), p=0.24]. The significant interaction was due to the fact that that mean BDI-II scores in CBT-SAD were slightly higher than in LT at post-treatment and slightly lower than in LT at summer. Cohen’s ds for the magnitude of post-treatment vs. summer difference in mean BDI-II score were 0.69 in CBT-SAD and 0.58 in LT. (See Figure 1).

Figure 1.

Depression scores at post-treatment and summer followup in cognitive-behavioral therapy for seasonal affective disorder (CBT-SAD) and light therapy (LT). These values represent Least Squares Means ± SE. Structured Clinical Interview Guide for the Hamilton Rating Scale for Depression-Seasonal Affective Disorder Version (SIGH-SAD) on the top panel. Beck Depression Inventory, 2^nd edition (BDI-II) on the bottom panel. CBT-SAD = cognitive-behavioral therapy tailored for SAD. LT = light therapy.¹

Discussion

The presumed predictable course of SAD affords the opportunity to evaluate whether identified SAD cases demonstrate spontaneous remission in the summertime. This study reports outcomes at a prospective summer assessment following diagnosis and treatment in a randomized clinical trial for SAD with CBT-SAD or LT. To our knowledge, this is the largest prospective summer assessment of SAD patients to date. Our results support construct validity for SAD in that mean depression scores for the sample overall were in the normal range across measures, the majority scored in the remitted range at time of summer assessment on the SIGH-SAD (72%) and BDI-II (86%), and very few met SIGH-SAD threshold for a current depressive episode (9 participants, 6.3%). Regarding course of DSM-IV major depression symptoms in the summer on the LIFE, the majority (79%) had a full remission, 13.3% had a partial remission (i.e., ≥ 4 consecutive or nonconsecutive weeks of PSR ratings ≥ 3), and 7.7% met major depressive episode criteria (i.e., ≥ 2 consecutive weeks of PSR ratings ≥ 5).

This distribution matches our a priori expectations and is consistent with the reformulated dual vulnerability model of SAD, which proposes that SAD involves vulnerabilities to both depression and seasonality (Lam et al., 2001). An underlying vulnerability to depression among individuals with SAD could contribute to occasional depressed moods and/or depressive episodes in the summer months. In addition, there may be a relatively small but distinct group of individuals that experiences winter depression symptoms with incomplete summer remission, termed “SAD with incomplete summer remission” (SAD-ISR; Lam et al., 2001). A followup study of the first 59 individuals with SAD from the NIMH Seasonal Studies Program conducted, on average, 8.8 years following initial SAD diagnosis, reported that 42% of those cases maintained a purely seasonal, and 44% developed a nonseasonal, course of depression (Schwartz et al., 1996). Some participants in our study who did not fully remit may fall into the more complicated category of SAD-ISR or may have developed a nonseasonal course to their disorder.

We used data on prospectively assessed summer mood status to examine the degree of completeness of the acute treatment response. Both CBT-SAD and LT were associated with significantly lower SIGH-SAD and BDI-II scores at summer than post-treatment, indicating residual symptoms at treatment endpoint. This replicates Postolache et al.’s (1998) finding for LT and suggests it generalizes to CBT-SAD and to the BDI-II. The magnitude of the difference between post-treatment and summer mood scores was medium in each treatment. Although LT and CBT-SAD are both efficacious first-line SAD treatments that produce large and comparable improvements in depression symptoms during treatment (Rohan et al., 2015), clinically meaningful residual symptoms remain.

Combination CBT-SAD+LT represents one potential option to achieve post-treatment mood status that does not differ from summer mood status. A previous clinical trial observed lower post-treatment depression scores in combined CBT-SAD+LT relative to solo treatment (Rohan et al., 2007), but was under-powered to detect differences between solo and combined treatment. Further, combination treatment did not differ from solo LT on next winter outcomes after adjusting for retreatment, whereas solo CBT-SAD was superior to solo LT across outcomes the next winter, even after adjusting for retreatment (Rohan et al., 2007). Therefore, even if combined CBT+LT confers a short-term advantage for the completeness of acute treatment response, it might be outweighed by solo CBT-SAD’s advantage over LT in future winters.

As a limitation, this was a secondary analysis of data from a single-site trial not designed to test SAD’s construct validity. The sample was relatively racially and ethnically homogenous, comprised of predominantly White, non-Latina/o individuals. Generalizability of these findings to other locations and to more diverse samples remains unknown. As another limitation, the summer followup consisted of a single assessment in either July or August following treatment. Although this approach limited participant burden and the LIFE was used to capture mood status in the interim between treatment endpoint and summer assessment, the methods relied on retrospective recall to estimate mood over this interval rather than conducting more frequent assessments. In addition, the study lacks a comparison group of untreated people diagnosed with SAD using these methods who were then evaluated in the summer for remission. Relevant to the debate on SAD’s construct validity, we cannot rule-out the possibility that some participants may have unreliably reported on their summer mood in accord with their pre-existing expectations about a relationship between the summer/long photoperiods and euthymic mood or may have been biased by their awareness of SAD diagnosis. This criticism is somewhat mitigated by the prospective summer assessment design and supplementing clinical interview (LIFE) with accepted measures (the SIGH-SAD and BDI-II), thereby not relying exclusively on retrospective recall of mood states that may be more subject to confirmation bias.

Highlights.

• Construct validity for winter seasonal affective disorder (SAD) has been questioned

• Full remission at prospective summer assessment would support construct validity

• 143 confirmed SAD patients were identified in winter and followed the next summer

• 79% showed complete, and 13% showed partial, summer remission; 8% were in episode

• Results support construct validity for SAD