Outcomes One and Two Winters Following Cognitive-Behavioral Therapy or Light Therapy for Seasonal Affective Disorder

This trial was conducted at the Mood and Seasonality Laboratory at the University of Vermont and was approved by the university’s institutional review board. The enrolled sample of patients (N=177) was randomly assigned to 6 weeks of CBT-SAD or light therapy and prospectively tracked through two new winters following treatment endpoint. Previous reports detail our full protocol (9) and report baseline characteristics, treatment integrity, and acute treatment outcomes (8). Participants were aged 18 or older and met DSM-IV-TR criteria for major depression, recurrent, with seasonal pattern on the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) and a current SAD episode on the Structured Interview Guide for the Hamilton Depression Rating Scale-Seasonal Affective Disorder Version (SIGH-SAD) (10) (same criteria as for recurrence, as described below). Exclusion criteria were kept to a minimum to maximize external validity. Potential participants were screened out for current light therapy or psychotherapy for depression, prior light therapy or CBT for SAD, a comorbid axis I disorder primary to SAD requiring immediate treatment, acute and serious suicidal intent, initiation of a new antidepressant medication in the past month or plans to change the dose of a current antidepressant, or positive laboratory findings for hypothyroidism at medical workup.

The study was powered to detect clinically meaningful differences between treatments on SAD recurrences (primary outcome) following treatment of the index episode. With these sample sizes (CBT-SAD, N=88; light therapy, N=89), there was 80% power to detect differences between treatments of 0.21 in recurrence proportions and ≥3.6 points in SIGH-SAD scores at follow-up.

The first week of September, letters were mailed to participants treated the winter before, prompting resumption of study treatment. For light therapy-treated participants, the letter encouraged reinitiating daily light therapy upon onset of the first depressive symptom and provided two options: borrowing a study light box for the duration of the winter or purchasing a unit. The letter provided contact information for manufacturers, with a list of specifications to match our devices (i.e., full-sized units emitting 10,000-lux cool-white light through ultraviolet filter). For CBT-SAD-treated participants, the letter encouraged use of the skills learned in CBT-SAD on their own (without a therapist). Both letters stated that if the recommended strategy proved insufficient, participants should pursue formal treatment, and contact information for local mental health centers and treatment providers was included. These letters were intended to promote fidelity with study treatment over follow-up, while addressing ethical concerns about proscribing additional treatment, if needed.

The 29-item SIGH-SAD (10) includes the 21-item Hamilton Depression Rating Scale (HAM-D) and the 8-item atypical symptoms subscale. The primary outcome was SIGH-SAD recurrence status as assessed at the next- and second-winter follow-ups, indicated by a total SIGH-SAD score ≥20, HAM-D score ≥10, and atypical score ≥5. Other SIGH-SAD-derived outcomes at the next- and second-winter follow-ups included continuous depression scores (total score, as well as HAM-D and atypical scores) and remission status. Remission status was classified as either ≥50% improvement in the SIGH-SAD score from pretreatment to follow-up plus a follow-up HAM-D score ≤7 plus a follow-up atypical score ≤7 or a follow-up HAM-D score ≤2 plus a follow-up atypical score ≤10. A second blind rater rated audio recordings of the SIGH-SADs. Intraclass correlations for interrater reliability were 0.965 at the next winter and 0.967 at the second winter.

The Beck Depression Inventory-Second Edition (BDI-II) (11), a 21-item self-report measure of depressive symptom severity, was also administered at the next winter and second winter. BDI-II outcomes at follow-up included total scores and a cutoff score ≤8 as a secondary marker of remission, consistent with our prior trials.

The primary analysis was an intent-to-treat analysis based on multiple imputation of missing next-winter SIGH-SAD scores, which were then used to classify depression recurrence status for individuals who dropped out during the treatment phase, withdrew from protocol, or were subsequently lost to follow-up. The fully conditional specification regression method was used to obtain imputed values based on age, sex, baseline comorbid diagnosis status, and depression scores at other time points. Separate regressions were used to impute values for the CBT-SAD and light therapy conditions because of differing effects of the predictors in the two treatments. Imputed values included a random component reflecting the residual distribution for the dependent variable, and 10 data sets with differing imputed values were generated. The difference between the CBT-SAD and light therapy groups in the proportions of participants with a recurrence and in remission the next winter was estimated for each of the 10 imputed data sets, and the estimates were combined using the inference methods for multiple imputation described by Little and Rubin (12). SAS PROC MI and PROC MIANALYZE were used to carry out the imputation and analysis. Sensitivity analysis was conducted to examine the robustness of the results under alternative imputation methods, including using the best- and worst-case scenarios for each treatment and using a logistic regression in the multiple imputation analysis to directly impute recurrence (rather than linear regression to impute SIGH-SAD scores). The BDI-II outcomes were analyzed in the same manner, as were outcomes for the second-winter follow-up. Analyses based on available data, without imputation, were also performed using logistic and linear regression analyses for dichotomous and continuous outcomes, respectively. In addition to prospectively assessed SIGH-SAD recurrence status, analyses using available data were performed for fulfilling DSM-IV-TR major depression criteria for the interims assessed by the October and December tracking telephone calls. For dichotomous outcomes that differed by treatment at next- and second-winter follow-ups, logistic regressions were performed to assess the effect of treatment group after adjustment for ongoing treatment(s) reported at that time point using data without imputation. We considered any treatment(s), in general, and any new treatment, psychotherapy, light therapy, and antidepressant medications, specifically. When coded for analysis, light therapy-treated patients reporting light therapy were counted as light therapy but not as any new treatment, and psychotherapy with a therapist among CBT-SAD patients was considered as both psychotherapy and any new treatment. For continuous outcomes that differed by treatment, linear regressions were used to assess the effect of treatment on depression scores after adjustment for any, new, and each ongoing treatment.

Comparisons between the treatments on next- and second-winter outcomes are shown in Table 1 and Table 2, separately for the multiple imputation analyses using the intent-to-treat sample and for the secondary analyses using all available data. Table 1 presents the dichotomous outcomes of recurrence and remission, and Table 2 presents the continuous outcomes of SIGH-SAD and BDI-II scores. The main findings are displayed in Figure 2. There were no statistically significant differences between CBT-SAD and light therapy on any of the outcomes at the next-winter follow-up. There was also no significant difference between treatments in recurrence status based on the October and December tracking calls: 9/73 (12.3%) CBT-SAD participants and 16/76 (21.1%) light therapy participants with tracking call data met major depression criteria (p=0.154).

At the second-winter follow-up, CBT-SAD was statistically superior to light therapy on two out of three dichotomous outcomes and on three out of four continuous outcomes, and this pattern was consistent across analyses using multiple imputation and all available data. On the primary outcome (SIGH-SAD recurrences), CBT-SAD was associated with fewer recurrences at the second winter than light therapy (27.3% compared with 45.6% using imputation; 28.0% compared with 46.5% without imputation). CBT-SAD was also associated with more remissions at the second winter than light therapy using BDI-II criteria (68.3% compared with 44.5% using imputation; 62.7% compared with 44.2% without imputation) but not using SIGH-SAD criteria. Depression scores in CBT-SAD were significantly lower than in light therapy at the second winter on both the SIGH-SAD and the BDI-II. Considering the component scales of the SIGH-SAD separately, HAM-D scores were lower in CBT-SAD than in light therapy at the second winter, but the treatments did not differ on atypical subscale scores.

Given that the treatments did not differ at the next-winter follow-up, but differed on the majority of outcomes at the second-winter follow-up, we conducted exploratory analyses to probe this result. Using SIGH-SAD recurrence status, we examined the association between recurrence status at the next winter and the risk of recurrence at the second winter within each treatment group. CBT-SAD participants without recurrence at the next winter were about five times more likely to not have recurrence at the second winter compared with CBT-SAD participants with recurrence at the next winter (87.5% compared with 15.5%; relative risk=5.12). In contrast, light therapy participants without recurrence at the next winter were only about twice as likely to not have recurrence at the second winter compared with light therapy participants with recurrence at the next winter (63.5% compared with 30.0%; relative risk=1.92). The ratio of relative risk (2.67) differed significantly from one (z=2.38, p=0.017), indicating that although next-winter recurrence status was predictive of second-winter recurrence in both treatments, the relationship was stronger in CBT-SAD than in light therapy. McNemar tests indicated that CBT-SAD participants with recurrence in one, but not both, winters were just as likely to have their recurrence in either winter (p=0.804), whereas light therapy participants with recurrence in only one winter were significantly more likely to recur the second winter (p=0.003). These findings suggest greater durability of a treatment effect in CBT-SAD than light therapy.

Descriptive information for ongoing treatment utilization reported at each winter follow-up visit is presented in Table 3. All participants who reported treatment on the October and/or December tracking telephone calls also reported treatment at the next-winter follow-up. Therefore, the more inclusive next-winter follow-up reports are presented here. A larger proportion of light therapy participants compared with CBT-SAD participants reported any treatment at next winter, but this difference was driven by 34.9% of light therapy participants reporting continued light therapy as instructed (compared with only 7.1% of CBT-SAD participants reporting light therapy). Consequently, the treatment groups did not differ in reports of any new treatment at next winter. At the second winter, the treatments did not differ in the proportions reporting any treatment or any new treatment, but more light therapy participants (30.6%) compared with CBT-SAD participants (13.4%) reported using light therapy. The CBT-SAD and light therapy groups did not differ on psychotherapy or antidepressant medication use at either follow-up. Some cross-over was evident with generally more light therapy participants pursuing psychotherapy than CBT-SAD participants pursuing light therapy.

Logistic and linear regression analyses indicated that differences between CBT-SAD and light therapy at the second winter were not attributable to concurrent treatment(s), in general, or to any new treatment or ongoing light therapy, psychotherapy, or antidepressant medication, specifically. After adjustment for concurrent treatment utilization, CBT-SAD participants continued to have fewer SIGH-SAD recurrences and lower SIGH-SAD, BDI-II, and HAM-D scores at the second winter than light therapy participants. The difference between the CBT-SAD and light therapy groups in BDI-II remissions at the second winter persisted with these adjustments, except when adjusted for ongoing light therapy (see Table 4).