Study Protocol
This is a multi-center, randomized, double-blind, parallel-group clinical trial of light therapy for participants with SAD (winter type). Participants were seen at a Baseline Visit, a Randomization Visit, and after 1, 2, 3 and 4 weeks of treatment. Participants who appeared to meet the inclusion criteria and not meet exclusion criteria at the Baseline Visit were invited to return in 1 week for a Randomization Visit. At this visit participants who continued to meet study criteria were issued either an active light treatment device or a placebo inactivated ion generator. Participants were seen at weekly intervals during 4 weeks of treatment. Participants were enrolled between October 1 and March 1 to reduce confounding effects of natural remission as expected in the spring.
Severity of depressive symptoms was rated at each visit using a 24-item SIGH-SAD, a scripted version of the Hamilton Depression Rating Scale [15] modified to reflect better the atypical symptomatology of SAD [16]. This version of the SIGH-SAD consists of the HDRS 17-item scale plus the first 7 atypical items (i.e., excluding Reverse Diurnal Variation). At the Randomization Visit and the subsequent 4 visits, SIGH-SAD ratings were carried out by a clinician blinded to the assigned treatment device. The blinded clinician also completed a systematic inquiry about any adverse events. A separate unblinded clinician dispensed and demonstrated the treatment device at the Randomization Visit, and was available at subsequent visits if required.
The study was conducted at 5 sites, in New Haven (USA), Vancouver, Montreal and Ottawa (Canada), and Groningen (The Netherlands). The research protocol was approved by applicable institutional review boards and met standards established by the Helsinki Declaration, and participants signed appropriate consent forms. The trial was registered at the U.S. National Institutes of Health clinical trials database [17].
Participants
Participants were recruited through media advertisements or professional referrals, screened by experienced interviewers by telephone, and if appropriate invited for a Baseline Visit. At this visit, participants received a full psychiatric evaluation, physical exam, urine toxicology for commonly abused substances, and urinary beta-HCG for female participants. Participants were required to be between ages 18 and 65, to have a DSM-IV diagnosis of SAD (major depressive episode, with seasonal pattern, winter type [18] and to have a SIGH-SAD score of 20 or greater. Diagnosis was established with the Structured Clinical Interview for DSM-IV (SCID) [19]. Participants also completed the Morningness-Eveningness Questionnaire (MEQ), a measure of preference for activity in the early or late part of the day [20].
Participants were told that the study involved treatment with either a new light treatment device or a negative ion generator, that both types of treatment were experimental, and that the study was placebo-controlled. In particular, participants were told that one half of the devices in the study were modified in such a way that the investigators did not expect the device to be efficacious. In order to demonstrate informed consent, participants had to demonstrate understanding that if they participate they have a one in two chance of being assigned to treatment expected to be inactive for 4 weeks.
Exclusion criteria were: significant medical illness, any retinal disease or medical disorder associated with retinal disease; pregnancy; use of photosensitizing medications, mood-altering medications, light therapy or other treatment for SAD within 1 week of the Baseline Visit (except within 4 weeks in the case of pharmacological antidepressant agents); initiation of psychotherapy within 3 months of the Baseline Visit, except where terminated by the participant prior to this visit; current organic mental disorder, panic disorder, anorexia or bulimia nervosa, obsessive-compulsive disorder or posttraumatic stress disorder; a history of any psychotic disorder or bipolar I disorder (history of manic episode); a history of substance use disorder not in full remission for at least one year; unstable sleep or mood patterns (such as severe premenstrual syndrome); previous unsuccessful trial of light therapy with an accepted device for at least 2 weeks; inability to provide informed consent; poor likelihood of complying reliably with study requirements; suicidal risk or other factor making trial participation clinically inappropriate. Participants were required to have a habitual sleep onset time before 1 A.M., and a habitual sleep end time before 9 A.M., prior to entry in the trial. Participants were required to agree to avoid other treatments for SAD or excluded medications, alteration of daily schedule to change light exposure, or travel to sunny destinations, to maintain a stable sleep schedule, and if female and potentially fertile to use an appropriate form of contraception during the trial.
Treatment Devices
At the Randomization Visit, eligible participants were issued an active or control treatment device by the unblinded clinician. Assignment to active or control group was determined by telephone call by the unblinded clinician to the trial sponsor, and was balanced in blocks of 4 for each site and gender. The proper use of the device was demonstrated to the participant by the unblinded clinician. After experiencing the assigned device in operation, the participant completed a brief questionnaire about expectations [21]. Participants were given a tape 20 inches in length to indicate the correct distance from the device.
The active treatment consisted of a Litebook treatment device with 60 LEDs (The Litebook Company Ltd., Alberta, Canada). The 60 LEDs employed in this Litebook model contain emitters which have a spectral emission peak at approximately 464 nm and fluorescent phosphors which provide a broader, secondary spectral peak near 564 nm: of the energy emitted over the range 400 to 700 nm, about 48% is emitted over the range 420 to 508, and 37% is emitted over the range 512 to 616 nm. Collectively the emitted light appears white. This device produces approximately 1,350 lux light at 20 inches. Participants assigned to this device were carefully instructed on aligning the device to illuminate maximally the eyes. An evaluation by an independent consultant physicist confirmed that the Litebook device meets the relevant sets of standards for light exposure safety [22–24].
Control treatment consisted of a negative ion generator, modified to emit no negative ions (SphereOne, Inc., Silver Plume, CO) and to generate a faint high-pitched whine, used at the same distance. Participants using the ion generator were instructed to wear a wrist strap connected to the device to maximize the transfer of negative ions, as this intervention has been found to increase expectations regarding efficacy for the device [3].
Participants were instructed to use the device for 30 minutes each morning, as soon as possible upon arising, and to complete treatment before 8 A.M. Participants were asked to maintain as stable a schedule of sleep and treatment as possible during the trial, and were asked to complete a log of the times of the beginning and end of sleep and of treatment. Participants were asked not to disclose to the blinded clinician which treatment device they were assigned. The blinded study clinician was permitted to reduce the duration of treatment to 15 minutes per day until the next study visit in the event of jitteriness or over stimulation, but this reduction was not required for any participant during the trial.
Statistical analysis
SIGH-SAD scores were analyzed in both a last observation carried forward (LOCF) analysis, including all 26 participants who were randomized, and an observed cases (OC) analysis, including all 23 participants who completed the trial. Remission was defined as a SIGH-SAD score less than 9. The a priori endpoint hypothesis was whether the proportion of participants in remission differed between the active and control treatment groups in the LOCF analysis using the Fisher's exact test. In a secondary analysis, end trial SIGH-SAD scores, as %SIGH-SAD scores consisting of final score as percentage of individual score at randomization, were compared between active and control groups by t test. Post hoc comparisons of the proportion of participants in remission and mean %SIGH-SAD score were made at Weeks 1, 2 and 3. Secondary analysis also included a repeated measures ANOVA mixed model with SIGH-SAD as dependent variable and time, treatment, and interaction of time and treatment as fixed effects, including all randomized participants. A variety of models were considered, including participant intercept and slope as random effects, autoregressive time-dependent, compound symmetry or unstructured correlation structures, and possible transformation of time by the log of one plus the week of treatment. The best model was selected by Schwartz Bayesian criterion, but all models indicated a significant interaction of time by treatment. The final model included linear time trend, no random effects, transformed time, and autoregressive correlation structure (SAS PROC MIX procedure, Kenward-Rogers method for degrees of freedom). Statistical assumptions were verified by examination of residuals.
Comparisons between the groups at baseline were made with t-tests in the case of continuous variables and Fisher exact tests in the case of dichotomous variables. Changes in time of sleep or treatment were analyzed with paired t-tests for participants who completed the trial, excluding one participant with incomplete sleep log data. End trial %SIGH-SAD scores were used to assess any relationship between therapeutic response and times of sleep or treatment or other covariate. Statistical analysis was performed with STATVIEW version 5.0.1 and SAS version 9.1.3 (both from SAS Institute, Cary, NC). All results are reported as means ± standard deviations.