Adjunctive Bright Light Therapy for Bipolar Depression: A Randomized Double-Blind Placebo-Controlled Trial

We included patients 18–75 years old with bipolar I or II disorder (confirmed on the Structured Clinical Interview for DSM-IV Axis I Disorders [SCID]) (17), a current moderate or severe episode of major depression (a score ≥20 on the Structured Interview Guide for the Hamilton Depression Rating Scale With Atypical Depression Supplement [SIGH-ADS] and a score ≥1 on SIGH-ADS item H1 or H2) (18), and no hypomania or mixed symptoms (a score ≤4 on the Mania Rating Scale [MRS]) (19). Eligible patients received antimanic medication at stable dosages for at least 4 weeks (20), within the therapeutic dosage range and plasma concentration (3); if applicable, patients received adjunctive antidepressant medication at stable dosages for at least 8 weeks and low-dose sleep aids for mild insomnia. Patients continued the same medications and dosages during the study. We included pregnant women because light therapy presents no additional risk and nonpharmacological treatments are limited (15). We excluded patients who had any manic, hypomanic, or mixed episodes in the past 6 months; rapid cycling in the past year; active suicidal ideation or attempted suicide; acute or chronic psychosis; obsessive-compulsive disorder; uncontrolled thyroid disease; any substance use disorder in the past 6 months; a positive urine drug screen; chronic eye diseases; or treatment with photosensitizing drugs (phenothiazine, antimalarial drugs, propranolol, melatonin, Hypericum, stimulants, or chronic treatment with nonsteroidal anti-inflammatory drugs).

At the eligibility visit, we administered the SCID and recorded age, race, sex, marital status, highest level of education, employment, age at onset, number of mood episodes, hospitalizations, medical history, medications and dosages, substance use, body mass index (BMI), and time of year of enrollment. We assessed clinical symptoms with the SIGH-ADS, which provides a benchmark for depression severity, including atypical features (18); the MRS, an 11-item measure derived from the Schedule for Affective Disorders and Schizophrenia for hypomania/mania levels (19); and the Clinical Global Impressions Scale for Bipolar Disorder (CGI-BD). We used the Brief Psychiatric Rating Scale (BPRS) to assess for psychosis; scores ≥31 correspond with active symptoms (21). To monitor side effects, we used the Systematic Assessment for Treatment Emergent Effects (10) and the Beck Scale for Suicide Ideation (22). We evaluated psychosocial functioning with the Global Assessment of Functioning Scale (GAF) and the Social Problems Questionnaire. Response to light therapy is associated with sleep quality and circadian and seasonality traits, which we assessed with the Pittsburgh Sleep Quality Index (23), the Morningness-Eveningness Questionnaire (24), and the Personal Inventory for Depression and Seasonal Affective Disorder (10), respectively. Expectations can moderate outcomes, especially in clinical trials involving light therapy, which test a visible intervention (10). Therefore, we evaluated expectancy on a brief multipoint rating scale at baseline and at the final or endpoint visit.

Eligible patients were randomly assigned in a 1:1 ratio to active bright white or inactive dim red light and stratified by antidepressant use with a block design (six patients per block). Within each block, patients were independently assigned to each group in equal numbers according to different randomization sequences.

Personnel conducting the visits and performing the clinical ratings were blind to the treatment condition to avoid allegiance effects, and they worked separately from the nonblinded personnel, who dispensed the study light boxes (concealed in numbered sealed containers, away from the blinded team), provided patient training on use of the light box, and regularly checked operation of the units. Ratings of side effects were separated from mood ratings to prevent contamination of the outcome measures. The active and placebo light boxes appeared identical when not illuminated. We implemented the same dosing protocol for both groups (25). Patients agreed not to search for information on light box design. We disclosed the treatment condition to patients only at the final visit and to blinded personnel and statisticians only after the analyses were completed. The study design elements were incorporated to mitigate unbalanced treatment expectations; therefore, we expected the observed treatment response to be related to the active light condition rather than to nonspecific effects.

At baseline, eligible patients were randomly assigned to a 7,000-lux, 4,000-K, broad-spectrum white fluorescent (Carex Day-Light Classic) or a 50-lux red light unit. The unit conforms to stringent standards, including illumination of a broad visual field, lighting from above to avoid glare, and maximal ultraviolet filtration. Participants were provided with standardized instructions on the appropriate use of their light box: optimal placement of the unit on its desk stand 12 inches from the eyes, and facing the light box without directly staring at it (to minimize discomfort) during the daily sessions of light therapy. Patients agreed to use the light box daily at home or work.

Patients began with 15-minute sessions of light therapy between 12:00 p.m. and 2:30 p.m. They attended weekly visits for assessment of mood level, side effects, suicidality, sleep quality, and psychosocial functioning. After each visit, the duration of light therapy was increased by 15 minutes to attain a target dose of 60 minutes per daily session by week 4 or until remission. Upward titration was conditional on having manageable side effects (item scores <3 on the Systematic Assessment for Treatment Emergent Effects) and no hypomania (an MRS score ≤4). Patients with worsening depression, severe suicidal ideation, or emergent hypomania (according to DSM-IV criteria and with an MRS score ≥5) were evaluated by the principal investigator (D.K.S.) and a nonblind monitor (R.S.D., K.L.W.), and appropriate clinical management was provided. To monitor adherence, patients recorded their daily light therapy sessions on the corresponding self-report form and called the time-stamped machine (checked daily by nonblind personnel). At every clinic visit, patients brought in their units; the nonblind personnel performed quality checks and downloaded the light sensor and sensitivity data recorded by the logger device (a HOBO U9 light on/off data logger; sensitivity threshold, 10–100 lumens/m²) to confirm appropriate, missed, or ill-timed sessions.

The primary outcome measures were the rate of remission (defined as a SIGH-ADS score ≤8) within the week 4–6 period and the continuous SIGH-ADS scores at the endpoint visit. We assumed that patients who were still actively enrolled at weeks 4–6 had fully (or nearly) completed the dose-titration protocol, and those who withdrew early also provided informative outcome data. Secondary outcome measures included the rate of response (the proportion with reductions ≥50% on SIGH-ADS score), frequency of mood polarity switch, clinical severity (based on the CGI-BD severity score), and the continuous ratings of side effects, suicidal thoughts, anxiety symptoms (using the five-item subscale of the 21-item Hamilton Depression Rating Scale [HAM-D]) (26), atypical features, psychosis (based on the BPRS), sleep quality (based on Pittsburgh Sleep Quality Index), and psychosocial functioning (based on the GAF and the Social Problems Questionnaire). To ascertain adherence, we assessed the total number of visits attended, the early withdrawal rate, the maximum light dose (minutes of daily light therapy exposure) to which the patient titrated, and the actual light dose per daily session reported by the patient at the endpoint visit.

We proposed to enroll two groups of 30 patients for an 80% power to detect group differences in the mean SIGH-ADS score of 2.6 points with a moderate effect size (0.37) and remission rates of 33%−38% (for an odds ratio of 6.9 at an alpha of 0.05, according to the PASS 2002 program [NCSS Statistical Software, Kaysville, Utah]) within the observed range in published trials (65% [SD=24] for active treatment and 29% [SD=12] for placebo treatment) (27). We stopped enrollment at 23 participants per group because funding had ended.

We used a modified intent-to-treat approach whereby all participants were analyzed according to the group to which they were randomized. For the primary outcome measures, we analyzed data from patients who had at least one visit between weeks 4 and 6; for the secondary, longitudinal analyses, we analyzed data from patients who had at least one follow-up visit. We performed unadjusted and adjusted analyses. The adjusted analyses accounted for the potential confounding effects of age, time of enrollment (season), and baseline SIGH-ADS and GAF scores. For continuous outcomes, we performed linear regressions, and for binary outcomes (SIGH-ADS score ≤8, Beck Scale for Suicide Ideation score ≥1, Beck Scale for Suicide Ideation score ≥8, and CGI-BD severity score ≥2), we performed logistic regressions to assess treatment effect. The secondary longitudinal analyses involved a mixed-modeling approach; we assumed fixed effects of intervention group, study week, and group-by-week interaction, and a random patient effect. We conducted the analyses in SAS, version 9.4 (SAS Institute, Cary, N.C.) and R, version 3.2.0 (R Foundation for Statistical Computing, Vienna, Austria). All statistical tests were two-sided with the significance threshold set at 0.05. Because of the sample size and numbers per group, we did not make adjustments for multiple hypothesis tests.

We obtained informed consent from 93 potential participants (see the CONSORT chart in the data supplement that accompanies the online edition of this article). Rapid cycling, substance use disorders or a positive urine screen, mild depression, concurrent hypomania, and psychosis were common reasons for exclusion. Forty-six patients underwent randomized assignment, 23 to bright white light (active treatment) and 23 to dim red light (inactive treatment). The demographic characteristics did not differ significantly between groups (Table 1). Participants were 44.7 years old on average, 40% had completed college, and 67% were women. We enrolled one woman during pregnancy. The sample’s racial distribution was 80% white, 11% black, and 9% multiracial. Two-thirds of eligible patients had bipolar I disorder. Age at onset was early on average (mean=16.0 years, SD=8.7 years). The first episode was usually major depression (56.5%), and many participants had bipolar disorder for decades (the mean duration was 28.5 years [SD=16.7]), complicated by other medical disorders, including obesity (the mean BMI was 30.9 [SD=8.32]).

As shown in Table 2, at baseline patients assigned to active treatment and inactive treatment had moderate depression levels (mean SIGH-ADS scores, 24.22 and 27.70, respectively; (Mann-Whitney U=5.68, p=0.017), moderate clinical severity (CGI-BD severity score ≥2), no hypomania or psychosis, increased social problems, and moderate impairment in global functioning (mean GAF score, 57.05 and 53.48, respectively; U=4.83, p=0.028). Despite randomization, the active and inactive treatment groups differed significantly on baseline depression and functioning. The frequency of suicidal thoughts (23.9% for the sample) did not differ significantly between groups. Patients in both conditions had prominent anxiety (mean HAM-D subscale score, 5.6 [SD=2.01]), reduced sleep quality (mean Pittsburgh Sleep Quality Index score, 8.3 [SD=3.4]), an intermediate morning-evening preference (mean Morningness-Eveningness Questionnaire score, 46.9 [SD=9.7]), and moderate levels of seasonal sensitivity (mean Personal Inventory for Depression and Seasonal Affective Disorder, part 2 score, 11.1 [SD=5.8]) and winter exacerbation of atypical neurovegetative symptoms (part 4 score, 6.1 [SD=2.1]). The majority enrolled in the fall (39.1%) or winter (34.8%). Three patients were randomly assigned to active treatment and seven to inactive treatment in February or March (the months that immediately precede remission of seasonal affective disorder). We enrolled 78.3% patients who were receiving combined treatment with a mood stabilizer and an antidepressant and 21.7% who were receiving treatment only with a mood stabilizer. Baseline expectancy was not significantly different between groups at randomization, when the light units were dispensed.

Forty-five patients attended one or more postbaseline visits (median=6 visits), and 40 patients (87%) completed the study (Table 3). The active treatment group, compared with the inactive treatment group, had a significantly higher remission rate at weeks 4–6 (SIGH-ADS scores ≤8, 68.2% and 22.2%, respectively; odds ratio=7.50, 95% CI=1.80, 31.28, p=0.003; adjusted odds ratio=12.64, 95% CI=2.16, 74.08, p=0.004) and a lower mean depression score at the endpoint visit (SIGH-ADS score, mean=9.18 [SD=6.57] compared with mean=14.94 [SD=9.16]; β=−5.76, p=0.026; adjusted β=−5.91, p=0.023) (Table 4, Figure 1). No hypomania or mood polarity switch was observed. One patient enrolled during pregnancy; the analyses with and without her indicated no differences in outcomes.

As shown in Table 4, patients in the active treatment group, compared with those in the inactive treatment group, had significantly better global functioning (GAF score, mean=74.77 [SD=9.70] and mean=67.65 [SD=9.86], respectively; β=7.13, p=0.030; adjusted β=7.61, p=0.042) and a significantly larger percent reduction in SIGH-ADS score (mean=−62.23% [SD=24.45] and mean=−46.21% [SD=29.67], respectively; adjusted β=−21.06, p=0.027). The active treatment group also experienced a higher response rate, lower clinical severity, less anxiety, fewer social problems, fewer neurovegetative symptoms, and better sleep quality compared with the inactive treatment group, but the differences were not significant. We encountered no serious adverse effects. Patients in both conditions experienced infrequent side effect worsening and, rarely, thoughts of suicide (see Tables 3 and 4). On the Systematic Assessment for Treatment Emergent Effects, patients in the active treatment group, compared with those in the inactive treatment group, had significantly less excessive sleep (21.7% compared with 52.2%; χ²=4.57, df=1, p=0.033) and less trouble concentrating (52.2% compared with 78.3%; χ²=3.45, df=1, p=0.063).

The maximum dose of light therapy exposure (46.3 minutes/day [SD=14.9] and 45.0 minutes/day [SD=17.5], for the active and inactive treatment groups, respectively), and the actual light dose per daily session reported by the patient at the endpoint visit (45.3 minutes/day [SD=16.7] and 44.0 minutes/day [SD=17.9], respectively) did not differ significantly between groups (Table 3). Although the median number of postbaseline visits (median=6 [interquartile range=6–6] for the active treatment group, compared with median=6 [interquartile range=5–6] for the inactive treatment group; U=185, p=0.003) differed significantly between groups, the early withdrawal rate (4.3% and 21.7%, respectively; see Table 3) did not differ significantly between groups. The one patient who withdrew from the active treatment group indicated that she was not permitted to use the light box at her new job. In the inactive treatment group, one patient returned the light unit after 3 weeks and did not complete another visit; others withdrew for one or more reasons: stopped using the light box for undefined reason, job, or school (N=4); preferred medication change (N=2); and missed two or more visits (N=1). Expectations at endpoint were not significantly different between groups. The secondary longitudinal analyses were not significant (not reported here).