Feasibility and preliminary efficacy of a bright light intervention in ovarian and endometrial cancer survivors

Abstract

Background:

Cancer-related sleep disturbance is common and can adversely affect physical and mental health. Bright light (BL) therapy is a novel intervention that targets sleep by promoting circadian regulation. Emerging evidence suggests BL can improve sleep disturbance, symptom burden, and health-related quality of life in cancer and other populations; however, this research is limited. The present two-phase pilot study assessed the feasibility and preliminary intended effects of BL therapy on sleep in ovarian and endometrial cancer survivors, and explored biologic and chronobiologic factors that may underlie intervention effects.

Methods:

In phase I, focus groups were conducted with 12 survivors and 9 gynecologic oncology clinicians to evaluate and gather feedback about the proposed study. In phase II, a pilot randomized controlled trial was conducted with 18 ovarian or endometrial cancer survivors who were randomized to 1:1 to receive 45 minutes of BL or dim light (DL) for 4 weeks. Participants wore wrist actigraphs, completed sleep diaries and self-report questionnaires, and provided blood, saliva, and urine samples at baseline (T1), post-intervention (T2), and 3-month follow-up (T3).

Results:

Study procedures were modified according to focus group results. Enrollment, retention, and adherence were all ≥ 80%. Mixed-model ANOVAs demonstrated that the number of nighttime awakenings per actigraphy, and sleep quality, cognitive functioning, and depression per self-report trended towards improvements in the BL condition compared to the DL condition. These variables improved from T1 to T2 before returning to baseline at T3. Effect sizes were generally medium to large.

Conclusions:

Study findings suggest that BL therapy is feasible among ovarian and endometrial cancer survivors. It may be an effective, non-pharmacological approach to reduce sleep disturbance and symptom burden in this population.

Introduction

Cancer-related sleep disturbance (e.g., difficulty initiating or maintaining sleep, early awakenings, non-restorative sleep) can adversely affect physical and mental health in cancer survivors, as well as severely interfere with activities of daily life (1–3). Rates of sleep disturbance among cancer survivors are higher than in the general population (2). Sleep disturbance can occur at any point in the continuum of care, and can persist for many years post-treatment (4) and into long-term survivorship. Thus, sleep disturbance represents a serious clinical problem in cancer patients, and has been associated with poorer health-related quality of life (HRQOL) (5, 6), worsening physical and psychological symptoms including greater fatigue and depression (2, 7–9), and impaired cognitive function and performance status (10, 11).

Sleep disturbance is particularly prevalent within the context of gynecologic cancer. Approximately two-thirds of gynecologic cancer survivors report sleep disturbance before adjuvant treatment (12), more than three-quarters do so during chemotherapy (2), and approximately one-third endorse ongoing sleep difficulties following completion of primary cancer treatment (12). Moreover, sleep disturbance is related to greater depression, greater anxiety, and poorer HRQOL in this population (13–16). Similarly, longitudinal improvements in depressive symptoms are associated with improvements in global sleep (13), indicating that depression can confer increased risk for poor sleep quality.

Although pharmacological agents are common treatments for sleep disturbances in cancer survivors (4, 17), they can have undesirable side effects (e.g., nighttime confusion, daytime restlessness (18)), and can exacerbate other symptoms (e.g., medication for improving nighttime sleep disturbance can exacerbate daytime fatigue). Additionally, there is limited research evaluating the efficacy and safety of pharmacologic interventions specifically in cancer survivors (4). Recently, there has been increasing focus on non-pharmacological treatments for sleep disturbance. One promising non-pharmacological intervention is systematic light exposure using bright light (BL), which targets sleep by promoting circadian regulation. Systematic exposure to BL has been shown to improve sleep disturbance, mood, depression, appetite, disease severity, and alertness among healthy and older adults (19–21), patients with sleep disorders (22), and patients with neurologic disorders (e.g., Parkinson’s (23), Alzheimer’s (24), dementia (25)). With regard to cancer populations, preliminary studies have found that systematic BL exposure can prevent increases in fatigue and deterioration of circadian rhythms during chemotherapy (26–28), and reduce cancer-related fatigue following completion of primary treatment (29, 30). However, while studies have shown promising effects of BL therapy in cancer, sleep disturbance has generally been evaluated as a secondary rather than primary outcome (31), and the long-term nature of sleep disturbance has not been well explored. In particular, there is notably little research focused on ovarian and endometrial cancer survivors. Two recent studies evaluating the effects of BL on fatigue in mixed cancer survivors did include gynecologic cancers; however, only ten gynecologic cancer survivors were enrolled in one study (29) and only three in the other (30). As such, the feasibility of this intervention specifically among survivors of ovarian and endometrial cancers remains uncertain.

The present study aimed to address the limited knowledge of sleep disturbance and BL therapy among ovarian and endometrial cancer survivors by conducting a two-phase pilot study. The initial, formative phase consisted of focus groups conducted with both relevant health care professionals and eligible participants to evaluate the feasibility of the proposed study. The primary focus of this phase was to determine whether ovarian and endometrial cancer survivors, despite experiencing elevated symptom burden and diminished HRQOL, would tolerate undergoing various study procedures including: blood draw, actigraphy, urine collection, saliva collection, psychosocial assessments, randomization, and follow-up. Based on the feedback provided by these focus groups, we refined the study procedures prior to conducting a pilot randomized controlled trial (RCT) to quantitatively assess the feasibility and preliminary efficacy of BL versus a dim light (DL) comparison condition to improve sleep disturbance, HRQOL, and symptom burden in patients with ovarian and endometrial cancers. We hypothesized that participants randomized to the BL intervention would report improved sleep quality, greater HRQOL, and less symptom burden after the course of exposure relative to participants in DL comparison group. Additionally, as an exploratory aim, we evaluated intervention-associated changes in cortisol, melatonin, and inflammatory cytokines during the course of the study to explore potential underlying biological and chronobiological mechanisms for the effects of BL therapy.

Methods

This pilot study involved two distinct procedural phases. The formative, qualitative phase (Phase I) consisted of focus groups comprised of a) relevant health professionals (to promote provider buy-in and protect against inconsistency between study methods and current clinical practice) and b) cancer survivors eligible to participate in the later pilot RCT (to determine its feasibility). The subsequent pilot RCT phase (Phase II) evaluated the tolerability of BL treatment versus DL comparison experimental conditions and assessed whether randomization to the treatment condition favorably impacted study outcomes relative to participation in the comparison condition. Across both study phases, we received institutional review board approval prior to enrollment, and all participants provided informed consent prior to participation.

Phase I – Qualitative Study Phase

Phase I Participants

The qualitative study phase included both gynecologic oncology clinicians as well as gynecologic cancer survivors. Eligible clinicians were all health care providers working at the Robert H. Lurie Comprehensive Cancer Center who were knowledgeable about and experienced in the clinical management of ovarian and endometrial cancers. We recruited these cancer clinicians through email solicitations from study team members introduced by department chiefs and practice managers. Focus groups with clinicians were held in October 2015.

Survivors who participated in the qualitative study phase included women with a histological diagnosis of any non-metastatic stage of primary ovarian epithelial cancer, primary peritoneal cancer, or cancer of the fallopian tube who were recruited from the Robert H. Lurie Comprehensive Cancer Center. Focus groups with survivors were held between April 2016 and August 2016. Eligible women: 1) were between the ages of 18 and 75 years, 2) had no history of chemotherapy or had completed primary chemotherapy at least 30 days prior to study enrollment, 3) endorsed current sleep complaints by reporting sleep disturbance when asked “would you mind telling me a little bit about your sleeping habits and quality of sleep,” and also either scoring > 5 on the Pittsburgh Sleep Quality Index score, demonstrating sleep efficiency < 80% on at least five out of seven days of screening wrist actigraphy, or demonstrating total sleep time ≤ 6.5 hours on at least five out of seven days of screening wrist actigraphy, 4) had a usual bedtime between 9:00pm and 3:00am, and 5) were fluent in spoken English. Women were excluded if they: 1) reported inpatient psychiatric treatment for mental illness within the past six months or were displaying overt signs of severe psychopathology or substance abuse at the time of screening, 2) had a diagnosis of narrow angle glaucoma or retinal disorders or demonstrated symptoms indicative of these diagnoses during eligibility screening, 3) had a diagnosis of sleep apnea or another sleep disorder besides insomnia, 4) reported using hypnotic medications (excluding benzodiazepines prescribed for purposes other than sleep disturbance), atypical anti-psychotics, photosensitizing medications, or Trazodone, 5) worked nights, early mornings, or swing shifts, 6) frequently traveled across trans-meridian time zones, or 6) received an Eastern Cooperative Oncology Group Performance Status (32) score > 1 or a score < 24 on the Mini Mental Status Examination (33) at the time of screening. Women who took part in the qualitative study phase were not considered eligible for the subsequent pilot RCT.

Phase I Procedures

We conducted two formative focus groups with clinicians, each consisting of a single session facilitated by two doctoral-level clinical psychologists (SFG and HLM) who had prior experience conducting focus groups and who followed a semi-structured guide. This guide contained open-ended questions, each with a series of follow-up probes, to assess providers’ opinions and observations about the sleep quality of ovarian cancer patients and survivors, about sleep treatments for this patient population, and about their impressions and opinions of planned study procedures. The focus groups were held in-person and each lasted approximately one hour.

We later conducted two usability focus groups with ovarian cancer survivors, each of which met for two sessions. These focus groups were facilitated by the same two doctoral-level clinical psychologists (SFG and HLM) who again followed semi-structured focus group guides containing open-ended questions and follow-up probes. During the first focus group meeting, this guide contained questions related to survivors’ sleep behaviors and beliefs, and their initial impressions regarding the study protocol based on a description thereof. At the conclusion of the first session, participants were asked to complete the planned four-week procedures (including sleep log, actigraphy, light intervention, and questionnaires) before returning to participate in the second focus group session. The semi-structured guide for the second session of the survivor focus groups contained open-ended questions regarding the study protocol. All survivor focus groups were held in-person and each lasted between approximately one and a half and two hours.

Finally, we conducted two individual semi-structured interviews with endometrial cancer survivors prior to initiation of Phase II to determine if it would be feasible to expand eligibility to include survivors of endometrial as well as ovarian cancer. These interviews were conducted by research staff (CLB), under the supervision of the two doctoral-level clinical psychologists who conducted the previous focus groups (SFG and HLM). For these interviews, we provided the endometrial cancer survivors with information regarding all study procedures. These participants did not complete the planned four-week procedures. A semi-structured interview guide was then followed, which contained open-ended questions assessing survivors’ reactions to the proposed assessment and experimental design based on a description thereof. These individual interviews were held in-person and each lasted approximately one hour. Phase I participants were reimbursed $100 for their time.

Phase II – Pilot RCT

Phase II Participants

Survivors who participated in the subsequent pilot RCT included women with a primary diagnosis of ovarian cancer, as with the qualitative study phase, or women with a histological diagnosis of any non-metastatic stage of primary endometrial cancer. Women who participated in the pilot RCT were enrolled between February 2017 and October 2017. Inclusion and exclusion criteria were identical to those in the qualitative study phase (Phase I).

Phase II Procedures

Interested participants completed a telephone pre-screening interview and an in-person eligibility screening appointment. Eligible participants who enrolled in the pilot RCT were individually randomized (1:1) to a BL intervention condition or a DL comparison condition by research staff. In addition to the in-person eligibility screening appointment, participants completed in-person appointments at pre-intervention (T1), post-intervention (T2), and three-months post-intervention (T3) during which they completed a battery of psychosocial assessments administered via Research Electronic Data Capture (REDCap), a secure, HIPAA-compliant, web-based application designed to support data capture for research studies (34). In addition, participants provided peripheral blood samples in-person at these appointments, and provided saliva and urine samples they had collected at home prior to their appointments. Each RCT participant also wore a wrist actigraph and completed a daily sleep diary for one week following the eligibility screening appointment, throughout the duration of the intervention period, and for one week prior to the T3 appointment, in order to provide objective sleep data. Finally, we obtained clinical information from participants’ medical records at each assessment time point, and participants reported their sociodemographic information at baseline. Phase II participants were compensated $100 for each in-person assessment completed. Participants were also given the option of receiving personalized feedback about their sleep patterns and habits following completion of study participation, provided by a doctoral-level psychologist with specialized training in behavioral sleep medicine, as supervised by a licensed clinical psychologist with certification in behavioral sleep medicine.

Light Treatment

Light was administered via Re-Timer™ glasses pre-set to emit either usual green (BL) or red (DL) for 45 minutes every morning after awakening for four weeks (i.e., 28 days). The Re-Timer™ glasses are a small (7.9”x5.5”x2.2”), lightweight (2.64 oz.) device that delivers light from below to replicate the natural pathway of light to the eyes. It emits light at a higher wavelength than the UV range and has a UV filter for added safety. It uses four light-emitting diodes (506 lux lm/m2 at highest setting) that pulse at 50–166 hertz. We instructed and encouraged participants to wear the glasses for a single continuous 45-minute time period each morning, but also gave participants the option of wearing the glasses for multiple shorter time periods totaling 45 minutes if necessary. Participants were asked to use the ReTimer™ glasses for 45 minutes with the hope that they would do so for a minimum of 30 minutes (the intended dose). Participants were instructed to be sure to complete light exposure prior to noon and to remain indoors while using the device. We clarified that participants could continue wearing their own eyeglasses for reading or for daily activities while using the ReTimer™ glasses as long as their eyeglasses did not have photochromatic or tinted lenses. We informed participants that they could move around freely while wearing the glasses, but encouraged them to not engage in any activities that could be dangerous when wearing the glasses (e.g., driving, using machinery, using sharp tools or knives, climbing stairs, running or jogging, or walking on extremely uneven surfaces). On the consent form participants were informed that they would be randomly assigned light glasses with either a red light or a green light. No information regarding anticipated effects of light color was included.

Measures

Feasibility and Adherence.

Study feasibility was assessed through study recruitment and retention rates. A 30% recruitment rate and a 70% retention rate were deemed acceptable based on prior studies conducted with gynecologic cancer survivors (35, 36), and prior studies of systematic light exposure in cancer populations (26, 29, 31). Intervention adherence was evaluated according to daily logs on which participants recorded the date, time of day, and total time that they wore the ReTimer™ glasses during the four-week intervention period.

Preliminary Efficacy.

Objective Sleep Disturbance was assessed with wrist actigraphy (Actiwatch Spectrum Plus, Phillips Respironics, Bend, OR, USA), which recorded sleep/wake activity. Outcomes were computed according 30-second epochs and included time in bed (total time spent in bed from lights out to final time out of bed), sleep onset latency (time to fall asleep/time from lights out to first appearance of sleep), number of nighttime awakenings, wake after sleep onset (total time spent awake from sleep onset to final arising time), total sleep time (total time spent sleeping during time in bed), and sleep efficiency (the ratio of total sleep time to total time spent in bed). The recorded actigraphy data were scored using the Actiware® sleep and activity monitoring software program (version 6.0.9, by Mini Mitter/Philips/Respironics). Daily diary data were used to manually edit the actigraphy data and facilitate scoring, as recommended by the Society of Behavioral Sleep Medicine (37). An established scoring method was used in which rest intervals were manually set by a trained technician using event markers and/or sleep log entries as available. When neither an event marker nor sleep log was available, or both were clearly inaccurate, the technician set the rest interval based on a reduction in activity and light levels. All technician-scored rest intervals were reviewed by at least one other scorer. The Actiware® software program then calculated sleep and wake within these defined rest intervals. See Reid et al. (38) for more information regarding this strategy for scoring actigraphy data.

Subjective Sleep Disturbance was measured by The Pittsburgh Sleep Quality Index (PSQI (39)), a 19-item self-report assessment of sleep quality over the past month. Each item is weighted on a 0–3 interval scale, and items yield seven component scores that can be summed to yield one global sleep quality score ranging from 0 to 21. Higher scores indicate worse sleep, with scores > 5 indicating clinically poor sleep (39).

Quality of Life was measured with the 27-item Functional Assessment of Cancer Therapy-General (FACT-G) scale (40) scale, which yields subscales reflecting Physical, Social, Emotional, and Functional Well-Being. Participants indicate the extent to which each item has applied to them over the past seven days using a five-point scale ranging from not at all to very much. Higher scores indicate better quality of life.

Symptom Burden.

Fatigue was measured by the 13-item Functional Assessment of Chronic Illness Therapy (FACIT) – Fatigue scale (41), which assesses fatigue during usual daily activities over the past seven days. Participants indicate level of fatigue on a four-point scale ranging from not at all to very much. Higher scores indicate less fatigue and better quality of life.

Cognitive Function was measured by the 37-item FACT-Cognitive Function scale (42), which yields four subscales that assess impairment of cognitive abilities, deficits observed by others, changes in cognitive function over time, and its impact on quality of life. For the Perceived Cognitive Impairments and Comments From Others subscales, participants indicate the extent to which each item occurred over the past seven days using a five-point scale ranging from never to several times a day. For the Perceived Cognitive Abilities and Impact on Quality of Life subscales, participants indicate the extent to which each item has applied to them over the past seven days using a five-point scale ranging from not at all to very much. Higher scores indicate better cognitive functioning and quality of life.

Depression was assessed using the Patient Reported Outcome Measurement Information System (PROMIS) Depression Item Bank computer adaptive test (CAT) (43). PROMIS assessments are t-scored so that a mean score of 50 with a standard deviation of 10 represents the average U.S. population score. Higher scores represent greater depression.

Pain Interference, or the degree to which pain limits or interferes with an individual’s ability to engage in typical activities, was assessed using the Patient Reported Outcome Measurement Information System (PROMIS) Pain Interference Item Bank CAT (44). Higher scores represent more interference from pain.

Cancer-Specific Stress was assessed by the 22-item Impact of Events Scale – Revised (IES-R) (45). Participants rate distress caused by intrusive thoughts, avoidance, and hyperarousal over the past seven days on a five-point response scale ranging from not at all to extremely. The measure yields three subscales and one total score, with higher scores reflecting greater distress.

Perceived Stress was assessed by the 14-item Perceived Stress Scale (PSS) (46). Participants rate stress-related thoughts and feelings over the past month on a scale ranging from never to very often. Positively-worded items are reverse coded prior to scoring, and items are summed to yield a single total score with higher scores indicating greater stress.

Stress.

Salivary cortisol was collected on two consecutive days following a standard procedure (47, 48). Saliva collection kits were provided so participants could collect cortisol four times per day (i.e., upon waking, 30 minutes after waking, at 6:00pm, and at 9:00pm) on each of the two days immediately prior to each in-person appointment. Four samples were collected to enable evaluation of cortisol patterns over the course of the day without excessively burdening respondents. Participants were instructed not to drink alcohol or exercise vigorously on collection days. Participants then brought the saliva samples to the study clinic on the day of their appointment. Cortisol was measured by a well-established enzyme-linked immunosorbent assay (ELISA) kit, following the manufacture’s protocol. For the present analysis, diurnal cortisol output was measured by computing the area under the curve (AUC) of cortisol concentrations with respect to zero, averaged over the two days of collection (49). Data were log transformed prior to analysis.

Circadian markers.

Urinary 6-sulfatoxymelatonin was measured as a marker of circadian rhythm. Prior to each in-person appointment, participants were asked to collect all overnight urine and to bring the urine samples to the study clinic on the day of their appointment. Urine was then assayed for 6-sulfatoxymelatonin (aMT6-s), the primary urinary metabolite of melatonin (50–52). aMT6-s and creatinine (Cr) were determined ELISA (53). Concentration of aMT6-s was expressed as μg/g Cr (54).

Inflammatory markers.

Pro- (IL-1β, IL-6, TNF-α) and anti- (IL-8) inflammatory cytokines from serum and lipopolysaccharide-stimulated peripheral blood mononuclear cells were measured via ELISA kits, following the manufacturer’s protocols (53). Venous blood was collected at each assessment time point by a licensed phlebotomist in one 8.0ml tube containing sodium heparin with a polymer gel (for lymphocytes), one 10.0ml uncoated tube, one 8.5 ml tube containing spray-coated silica and a polymer gel (for serum), one 4.0ml tube containing spray-coated K2EDTA (for whole blood hematology), and one 2.5ml PAXgene blood RNA tube. A total of 33.0ml (about two tablespoons) of blood was collected from each participant at each time point. Data were log transformed prior to analysis.

Analytic Plan

Phase I.

For the first, qualitative study phase, focus groups and interviews were audio recorded and transcribed semi-verbatim, excluding identifying information, by a professional transcription company. Transcripts were then organized, coded, and analyzed systematically using NVivo qualitative data analysis software (QSR International Pty Ltd. Version 11, 2015). Using a constant comparative approach (55), trained coders independently generated a preliminary list of emergent content themes. Next, coders met to discuss initial thoughts, insights, and observations for the development of initial coding categories. Disagreements were resolved by group discussion until consensus was reached. Through a systematic analysis process, consensus was reached and themes were continually refined by collapsing redundant and removing irrelevant themes. Once no new categories emerged, a coding dictionary was developed for the remaining analysis.

Phase II.

For pilot RCT participants, chi-square analyses and t-tests were used to examine baseline differences between the study conditions. Descriptive statistics were calculated to evaluate study feasibility and adherence. To assess preliminary efficacy, mixed model ANOVAs evaluated significant mean differences in primary study outcomes over time between the BL and DL conditions. Study outcomes were entered in separate models as dependent variables, and a dummy coded study condition variable (BL vs. DL) was entered as the independent variable. Given the preliminary nature of these analyses and the small sample size, both results significant at the p < 0.05 level and non-significant trends at the p < .10 level were reported. Covariates were not included in the models. For all interaction effects, partial eta squared values were calculated and interpreted with cutoffs of 0.01, 0.06, and 0.14 indicating small, medium, and large effects, respectively (56). For all interaction effects and main effects, Cohen’s ds were calculated and interpreted with cutoffs of 0.20, 0.50, and 0.80 indicating small, medium, and large effects, respectively (57).

Results

Phase I Results

Two focus groups were conducted with clinicians, one with six nurses and one with three physicians, which each met once. Four primary themes emerged regarding provider perceptions of patients’ sleep: 1) pharmacologic sleep treatment, 2) effects of cancer treatment, 3) effects of lifestyle, and 4) effects of mental health. See Table 1 for examples of quotations for each theme. In addition, across both focus groups, clinicians generally agreed that they would recommend BL as an intervention to their patients, and expressed beliefs that patients would be willing to try BL and would be compliant to intervention procedures.

Table 1.

Illustrative quotes for themes from Phase I focus groups with oncology clinicians

Theme	Prompt	Quote	Clinician
Pharmacologic sleep treatment	“What have patients reported to you regarding their sleep practices?	“We have a lot of people on Ativan after chemo for sleep. We know that’s not good.”	Nurse
		“…are frequently self-medicating or some only partially successfully using Ativan typically or various prescription or over-the-counter sleep aids”	Physician
Effects of cancer treatment	“How much do your patients talk to you about their sleep?	“All the chemo patients I revisit [talk about their sleep]	Nurse
		“I would say mainly patients on chemo are talking about sleep”	Physician
Effects of lifestyle	“What factors do you think most contribute to your patients’ sleep quality?	“The exercisers do better”	Nurse
Effects of mental health	“What problems, if any, do you notice are associated with poor sleep quality in your patients?”	“…that subset of patients that have that higher level of anxiety that it spills over into their sleep disturbances”	Physician

We subsequently conducted two focus groups with six ovarian cancer survivors each (N=12). Each of the two groups met twice (before and after attempting portions of the four-week RCT intervention protocol). Focus group participants’ mean age was 55.3 (SD = 11.1; range=35–72) years. Their cancer staging was: Stage I (n = 3, 25%); Stage II (n =3, 25%); Stage III (n = 4, 33%); missing (n = 2; 17%). Emergent themes from the first focus group sessions included a majority of participants reporting: 1) prior use of pharmacological treatments for sleep but a concurrent dissatisfaction with that use, 2) prior use of non-pharmacological interventions, with limited success, 3) interest in behavioral and integrative medicine treatments, but 4) minimal knowledge of light therapy or other evidence-based interventions. The most commonly identified non-pharmacological interventions participants had previously attempted were changing sleep environments, eliminating caffeine, and taking supplements (e.g., melatonin); less commonly mentioned were exercise and relaxation. The majority of participants reported having little to no discussions with their health care teams, and being amenable to attempting both light therapy and tracking their sleep. In the second focus group sessions, respondents generally provided favorable responses to the proposed study protocol, and indicated that four weeks would be a reasonable time commitment for study participation. Specifically, emergent themes included: 1) general dislike of the first actigraphy watch initially piloted in Phase I compared to the second, which was used in Phase II (e.g., the feel on the wrist as well as difficulty using the event marker), 2) challenges to adherence posed by interferences or variability in daily routines, 3) procedures (e.g., sleep logs) resulting in more focus on sleep (described as both helpful at times and slightly unwelcome at others), and 4) minor issues with fit of the light glasses. As a result of this feedback, we made minor adjustments to the RCT protocol: improved orientation and instructions regarding the light glasses (e.g., how to adjust fit and work use into daily routines) and selection of the ActiWatch model (which was used and generally received favorably by participants in the second focus group cohort) for the RCT over another actigraphy watch (which was used and generally received poorly by participants in the first focus group cohort).

Phase II Results

Phase II sample characteristics are presented in Table 2. Participants were generally middle-aged, married, and highly-educated. The mean age was 57.1 years (standard deviation [SD] = 10.0) and two-thirds were married or in an equivalent intimate relationship. The entire sample had a high school diploma, and nearly three-quarters (72.2%) had completed a Bachelor’s degree or beyond. The majority of the sample (61.1%) was working, and the entire sample self-identified as Non-Hispanic White. The majority of the sample had a prior diagnosis of endometrial (77.8%) cancer, while the remainder either had a diagnosis of ovarian (16.7%) or ovarian and fallopian tube (5.6%) cancer. All participants had undergone hysterectomy, 38.9% had previously undergone radiation therapy, and 22.2% had previously undergone chemotherapy. The majority of the sample (88.9%) had no major medical comorbidities, though 28% of the sample had a BMI ≥ 30. At baseline, participants in the DL condition demonstrated higher scores on the Sleep Medication component of the PSQI (t(8) = 2.31, p = 0.05; BL = 0.00, DL = 0.67), better sleep efficiency (t(10.5) = 2.37, p = 0.04; BL = 86.06%, DL = 91.38%), and lower levels of IL-6 (t(15) = −2.40, p = 0.03; BL = 3.27, DL = 3.20), compared to participants in the BL condition. There were no other statistically significant differences across the BL and DL conditions at baseline.

Table 2.

Phase II sample characteristics

Variable	Full Sample (N = 18)	BL (n = 9)	DL (n = 9)
Age	57.11 (10.02), 37–70	53.89 (11.2), 37–67	60.33 (7.94), 51–70
BMI	28.23 (5.43), 21.71–42.57	27.30 (6.00), 21.71–36.49	29.17 (4.96), 23.84–42.57
≥ College education	13 (72.2%)	6 (66.7%)	7 (77.8%)
Married/equivalent	12 (66.7%)	5 (55.6%)	7 (77.8%)
White	18 (100.0%)	9 (100.0%)	9 (100.0%)
Hispanic/Latina	0 (0.0%)	0 (0.0%)	0 (0.0%)
Employed	10 (55.6%)	6 (66.7%)	4 (44.4%)

^amean (SD), range; ^bn (%). BL = bright light; DL = dim light; BMI = body mass index.

Feasibility

The study CONSORT flow diagram is shown in Figure 1. In-person or telephone contact was made with 211 women, 63 of whom (29.9%) completed telephone screening. The top two reasons for refusal were lack of interest and lack of sleep disturbance. Of those who completed telephone screening, 32 women (50.8%) were deemed preliminarily eligible and 25 (39.7%) completed in-person screening. Of those who completed in-person screening, 22 (88.0%) were deemed fully eligible and enrolled. One participant withdrew after consent but before randomization, leaving a randomized sample of 21. Two participants in the DL condition and one in the BL condition withdrew during the intervention. All participants who completed the intervention were retained at T3. Thus, the retention rate at three-month follow-up was 90% (9 of 10 participants) for BL and 81.8% (9 of 11 participants) for DL.

Figure 1.

CONSORT diagram

Adherence

Participants randomized to the BL condition used the ReTimer™ glasses for ≥ 30 minutes prior to noon for an average of 24.78 days (88.5%), for an average of 42.91 (SD = 3.52) minutes per day. Participants randomized to the DL condition used the glasses for an average of 23.33 days (83.3%), for an average of 39.68 (SD = 7.10) minutes per day. These differences were not statistically significant.

Preliminary Efficacy

Estimated marginal means and associated standard errors for all preliminary efficacy outcome measures as a function of treatment condition and time are presented in Table 3.

Table 3.

Means and standard errors for preliminary efficacy outcomes as a function of treatment condition and time

Outcome	Time 1 M (SE)		Time 2 M (SE)		Time 3 M (SE)
	BL	DL	BL	DL	BL	DL
Objective Sleep Disturbance
Time in bed, min	484.91 (19.01)	498.19 (20.16)	480.42 (17.88)	507.90 (18.96)	497.17 (15.47)	503.26 (16.41)
Sleep onset latency, min	8.66 (1.80)	5.01 (1.91)	13.40 (2.95)	6.34 (3.13)	12.96 (3.26)	7.46 (3.46)
Number of awakenings, n	43.37 (4.31)	31.70 (4.57)	40.86 (3.79)	35.74 (4.02)	43.46 (5.00)	37.96 (5.30)
Wake after sleep onset, min	60.68 (7.75)	38.89 (8.22)	55.60 (7.72)	45.94 (8.18)	56.87 (8.42)	46.02 (8.93)
Total sleep time, min	415.55 (15.33)	454.89 (16.29)	411.41 (16.24)	455.61 (17.23)	427.33 (13.57)	449.76 (14.40)
Sleep efficiency, %	86.06 (1.59)	91.28 (1.69)	85.85 (1.92)	89.71 (2.04)	86.30 (1.96)	89.41 (2.08)
Subjective Sleep Disturbance
PSQI Global	7.80 (0.99)	10.00 (0.99)	5.20 (0.47)	9.00 (0.47)	7.00 (0.89)	7.00 (0.89)
Quality of Life
FACT-G PWB	25.33 (0.61)	24.56 (0.61)	27.00 (0.45)	25.28 (0.45)	26.22 (0.75)	24.78 (0.75)
FACT-G SWB	19.55 (1.60)	23.49 (1.70)	22.98 (1.47)	23.35 (1.56)	24.76 (1.84)	24.38 (1.95)
FACT-G EWB	22.00 (0.62)	21.22 (0.62)	22.11 (0.53)	21.36 (0.53)	21.33 (0.53)	21.56 (0.53)
FACT-G FWB	19.78 (1.37)	21.61 (1.37)	22.29 (0.97)	22.22 (0.97)	21.40 (1.30)	20.78 (1.30)
Symptom Burden
FACIT-Fatigue	41.67 (1.75)	42.89 (1.75)	47.11 (1.16)	44.94 (1.16)	45.78 (0.99)	43.67 (0.99)
FACT-Cognitive Function
PCI	63.14 (2.40)	63.56 (2.40)	64.44 (2.33)	65.73 (2.33)	64.89 (2.24)	65.89 (2.24)
PCI-QOL	14.11 (0.68)	14.78 (0.68)	15.22 (0.42)	15.56 (0.42)	15.56 (0.29)	15.44 (0.29)
CFO	15.89 (0.08)	16.00 (0.08)	15.67 (0.12)	16.00 (0.12)	15.89 (0.11)	15.89 (0.11)
PCA	21.13 (1.78)	22.82 (1.78)	22.04 (1.51)	24.63 (1.51)	21.78 (1.86)	24.89 (1.86)
PROMIS Depression	49.49 (1.78)	48.86 (1.78)	43.28 (2.27)	46.24 (2.27)	47.31 (1.66)	48.16 (1.66)
PROMIS Pain Interfer	43.06 (2.79)	47.96 (2.63)	43.06 (2.71)	47.49 (2.55)	43.08 (3.11)	47.21 (2.93)
IES-R	9.00 (3.14)	8.44 (3.14)	2.78 (2.20)	8.56 (2.20)	4.11 (2.93)	9.67 (2.93)
PSS	18.78 (2.59)	16.67 (2.59)	14.22 (2.34)	16.56 (2.34)	15.44 (1.85)	16.89 (1.85)
Stress
Salivary Cortisola	0.73 (0.06)	0.74 (0.06)	0.77 (0.08)	0.77 (0.08)	0.73 (0.06)	0.78 (0.05)
Circadian Marker
Urinary aMT6-s	24.96 (4.81)	13.80 (5.10)	25.85 (5.24)	13.98 (5.55)	25.05 (4.60)	12.79 (4.88)
Inflammatory Markers
IL-1βa	2.87 (0.12)	2.92 (0.11)	3.05 (0.10)	2.94 (0.09)	2.85 (0.32)	2.58 (0.30)
IL-6a	3.27 (0.02)	3.20 (0.02)	3.32 (0.05)	3.28 (0.04)	3.24 (0.27)	2.83 (0.25)
TNF-αa	2.64 (0.13)	2.85 (0.12)	2.94 (0.16)	2.82 (0.15)	2.56 (0.23)	2.47 (0.22)
IL-8a	3.30 (0.08)	3.26 (0.08)	3.28 (0.07)	3.34 (0.07)	3.23 (0.28)	2.86 (0.28)

Note.Cond = Condition; BL = Bright light; DL = Dim light; PSQI = Pittsburgh Sleep Quality Index; FACT = Functional Assessment of Cancer Therapy; PWB = Physical Wellbeing; SWB = Social Wellbeing; EWB = Emotional Wellbeing; FWB = Functional Wellbeing; FACIT = Functional Assessment of Chronic Illness Therapy; FACT = Functional Assessment of Cancer Therapy; PCI = Perceived cognitive impairment; PCI-QOL = Impact of perceived cognitive impairment on quality of life; CFO = Comments from others; PCA = Perceived cognitive abilities; PROMIS = Patient Reported Outcomes Measurement Information System; Pain Interfer = Pain Interference; IES-R = Impact of Events Scale – Revised; PSS = Perceived Stress Scale; aMT6-s = 6-sulfatoxymelatonin .

^aValues were log transformed prior to analysis.

Objective sleep disturbance.

The interaction between study condition and time trended towards significance with a large effect for number of nighttime awakenings (F(2,30) = 2.70, p = 0.08, ƞ_p² = 0.025). Specifically, the number of nighttime awakenings, per actigraphy, increased over time for the DL condition, while it decreased from T1 to T2 before returning to baseline at T3 for the BL condition.

A main effect of condition also trended towards significance with a large effect. The DL group (M = 453.22, SE = 12.85) had longer total sleep time than the BL group (M = 418.10, SE = 12.12) across time points (F(1, 15) = 3.96, p = 0.07, ƞ_p² = 0.25, d = 0.94).

Subjective sleep disturbance.

The interaction between study condition and time trended towards significance with a large effect for the PSQI global score, with those in the BL group improving over time compared to those in the DL group (F(2, 16) = 2.90, p = 0.08, ƞ_p² = 0.27).

The percentage of DL participants with healthy sleep (i.e., PSQI score < 5) was 0% at T1, 11.1% at T2, and 0% at T3. For BL participants, these percentages were 11.1% at T1, 44.4% at T2, and 11.1% at T3. This difference was statistically significant at T2 (χ²(1) = 3.88, p = 0.05).

Quality of Life.

There were no statistically significant or trending group-by-time interactions for the FACT-G.

A main effect of condition trended towards significance with a large effect for the Physical Well-being subscale of the FACT-G. The DL group (M = 24.87, SE = 0.52) had lower scores, indicating worse HRQOL, than the BL group (M = 26.19, SE = 0.52) across time points (F(1, 16) = 3.23, p = 0.09, ƞ_p² = 0.17, d = 0.85). Additionally, a significant main effect of time with a large effect was found for the Physical Well-being subscale of the FACT-G (F(2, 32) = 4.27, p = 0.04, ƞ_p² = 0.21, d_T1-T2 = 1.06; d_T2-T3 = 0.49). Scores improved across groups with a large effect from T1 (M = 24.94, SE = 0.43) to T2 (M = 26.14, SE = 0.32), but then worsened again across groups at T3 (M = 25.50, SE = 0.53) with a medium effect. A main effect of time with a large effect trended towards significance for the Social Well-being subscale of the FACT-G (F(2, 30) = 2.88, p = 0.07, ƞ_p² = 0.16, d_T1-T2 = 0.49; d_T2-T3 = 0.38), with scores improving with medium effects from T1 (M = 21.52, SE = 1.17) to T2 (M = 23.17, SE = 1.07) to T3 (M = 24.57, SE = 1.34) across groups.

Symptom Burden.

The interaction between study condition and time trended towards significance with large effects for the Comments from Others subscale of the FACT-Cog (F(2, 32) = 2.55, p = 0.09, ƞ_p² = 0.14) and the PROMIS Depression CAT (F(2, 32) = 3.08, p = 0.06, ƞ_p² = 0.16). Compared to participants in the DL group, those in the BL group improved over time on both measures.

Significant main effects of time with large effects were found for the FACIT Fatigue (F(2, 33) = 5.01, p = 0.01, ƞ_p² = 0.24, d_T1-T2 = 1.19; d_T2-T3 = 0.57) and the Quality of Life subscale of the FACT-Cog (F(2, 32) = 3.89, p = 0.05, ƞ_p² = 0.20, d_T1-T2 = 0.79; d_T2-T3 = 0.14). For the FACIT-Fatigue, across groups scores improved with a large effect from T1 (M = 42.28, SE = 1.24) to T2 (M = 46.03, SE = 0.82) but then worsened again at T3 (M = 44.72, SE = 0.70) with a medium effect. For the Quality of Life subscale of the FACT-Cog, scores across groups improved from T1 (M = 14.44, SE = 0.48) to T2 (M = 15.39, SE = 0.30) with a medium effect, and continued improving to T3 (M = 15.50, SE = 0.21) with a very small effect.

Stress, Circadian Marker, and Inflammatory Markers.

There were no statistically significant or trending group-by-time interactions or main effects on changes in salivary cortisol, urinary 6-sulfatoxymelatonin, or plasma cytokines.

Discussion

To the best of our knowledge, this is the first study to evaluate the feasibility and preliminary efficacy of a four-week, self-administered BL intervention using Re-Timer™ glasses to improve sleep quality, HRQOL, and symptom burden among ovarian and endometrial cancer survivors. The qualitative results confirmed common complaints of sleep disturbance, interest in non-pharmacological interventions, minimal knowledge of BL therapy, and willingness to use BL therapy in this population. Findings from the pilot RCT indicate the intervention is feasible, as the recruitment and retention rates were over 80%. Treatment adherence was also high in both conditions, with participants using the light device for more than 30 minutes per day (intended dose) for over 80% of the days in the study period. Study results also demonstrate that the BL therapy may have beneficial effects on sleep disturbance and symptom burden among ovarian and endometrial cancer survivors.

Our results demonstrate that the four-week BL intervention improved objective and subjective measures of sleep disturbance. Specifically, from baseline to post-intervention there were trends towards improvements in nighttime awakenings and subjective sleep quality with large effects in the BL condition, although values returned to baseline three months post-intervention. Similarly, cognitive functioning and depressive symptoms improved in the BL condition with large effects following the intervention before returning to baseline levels at the three-month follow-up. This is consistent with previous studies documenting medium to large effect sizes in self-reported sleep quality after exposure to morning BL for four weeks among cancer survivors (31), as well as improvements in depression among terminally ill cancer patients (58). These findings provide further support for the use of BL therapy in cancer survivors, particularly given the large effect sizes of many of the observed changes. The lack of sustained improvements at three-month follow-up is likely due to the discontinuation of the light exposure after 28 days. Although some past studies have found lasting effect of BL exposure even following termination of daily use, effects have generally been evaluated after a shorter follow-up period (30, 31). For our study, participants used the BL intervention on average for 42.91 (SD = 3.52) minutes per day, greater than the intended dose of 30 minutes per day, for four weeks. While other studies with cancer survivors have similarly used a 30-minute four-week intervention and have demonstrated significant improvements in fatigue and sleep efficiency (29, 31), the optimal duration of BL therapy remains unknown and requires further evaluation.

Contrary to our hypotheses, there were no significant group differences over time (significant group x time interactions) on domains of well-being, fatigue, pain, or stress. While the lack of significant findings may reflect a true lack of BL intervention effect, it may also in part be due to the overall low levels of symptom burden reported in both groups for those outcomes. For example, both conditions across the study period had fatigue scores above the clinical cutoff ≤ 33 (41, 59) and low reported levels of cancer-specific stress and general perceived stress. BL therapy may have beneficial effects on fatigue and stress among cancer survivors when symptoms are more prevalent and distressing, and these improvements may in turn positively affect HRQOL. On the other hand, more direct, active participation in behavioral and/or psychosocial interventions that target those specific outcomes and strategies to improve those outcomes (e.g., stress and symptom management) may be needed to improve physical and psychosocial symptoms that are more bothersome or severe. The present study did not find any statistically significant changes in biomarkers of stress, circadian rhythms, or inflammation across the study period for either condition. However, biomarkers can be affected by other individual factors such as weight, exercise, diet, metabolism, and the presence of other diseases (60). Moreover, the effects of cancer and cancer treatment on the body can confound the results. Additional research with larger samples is needed to further examine potential mechanisms of the effects of BL therapy on sleep quality, HRQOL, and symptom burden in this population.

The main limitation of this study is the small sample size. Stronger intervention effects may emerge with a larger sample. During the qualitative phase, a small number of focus groups were conducted and some were attended by relatively few participants. However, as this was a pilot RCT, the main purpose was to assess the feasibility and preliminary efficacy of BL therapy among ovarian and endometrial cancer survivors in order to inform future large-scale trials to more definitively establish the effects of BL therapy on sleep and psychosocial outcomes in this population. Also, this study focused solely on ovarian and endometrial cancer survivors, so results cannot be generalized to other cancers, and differential intervention effects by disease type or other cancer characteristics were not assessed given the small sample size. In addition, the dosing or duration of the intervention may have had an impact on our findings. For example, while improvements were reported in global sleep over time, the average sleep quality in both conditions remained in the disturbed sleep range (PSQI > 5) over the course of the study, and a more intense or longer intervention may be needed to improve sleep quality to normal ranges.

Nonetheless, the present study suggests that BL therapy is a feasible, low-burden, and easily administered intervention, and may be an effective, non-pharmacological approach to reduce sleep disturbance and symptom burden among ovarian and endometrial cancers survivors. This study contributes to the growing empirical evidence supporting the benefits of BL therapy to improve sleep. Future larger-scale studies are needed to determine the optimal dosing and duration of BL therapy, evaluate the underlying mechanisms for the effects of the intervention, and confirm the effectiveness of BL therapy to improve sleep and psychosocial outcomes for ovarian and endometrial cancer survivors.