Effects of Light Therapy on Mood and Insulin Sensitivity in Patients With

Diabetes Care Volume 42, April 2019 529

Annelies Brouwer,1 Daniel H. van Raalte,2 Effects of Light Therapy on Mood Hoang-Ton Nguyen,3 Femke Rutters,4 CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL CLIN Peter M. van de Ven,4 Petra J.M. Elders,5 and Insulin Sensitivity in Patients Annette C. Moll,3 Eus J.W. Van Someren,1,6,7 Frank J. Snoek,8 With Type 2 Diabetes and Aartjan T.F. Beekman,1 and Depression: Results From a Marijke A. Bremmer1 Randomized Placebo-Controlled Trial Diabetes Care 2019;42:529–538 | https://doi.org/10.2337/dc18-1732

1Amsterdam UMC, Vrije Universiteit, and GGZ inGeest, Department of Psychiatry, Amsterdam Public Health research institute, Amsterdam, the OBJECTIVE Netherlands 2Amsterdam UMC, Vrije Universiteit, Depart- Depression is common in patients with type 2 diabetes and adversely affects ment of Internal Medicine, Diabetes Center, quality of life and diabetes outcomes. We assessed whether light therapy, an Amsterdam, the Netherlands 3 antidepressant, improves mood and insulin sensitivity in patients with depression and Amsterdam UMC, Vrije Universiteit, Depart- ment of Ophthalmology, Amsterdam, the Nether- type 2 diabetes. lands 4Amsterdam UMC, Vrije Universiteit, Depart- RESEARCH DESIGN AND METHODS ment of Epidemiology and Biostatistics, Am- This randomized, double-blind, placebo-controlled trial included 83 patients with sterdam Public Health research institute, depression and type 2 diabetes. The intervention comprised 4 weeks of light Amsterdam, the Netherlands 5Amsterdam UMC, Vrije Universiteit, Depart- therapy (10,000 lux) or placebo light therapy daily at home. Primary outcomes ment of General Practice and Elderly Care included depressive symptoms (Inventory of Depressive Symptomatology [IDS]) Medicine, Amsterdam Public Health research and insulin sensitivity (M-value derived from the results of a hyperinsulinemic- institute, Amsterdam, the Netherlands 6 euglycemic clamp). Secondary outcomes were related psychological and gluco- Netherlands Institute for Neuroscience, Depart- ment of Sleep and Cognition, Amsterdam, the metabolic measures. Netherlands 7Amsterdam UMC, Vrije Universiteit, Depart- RESULTS ment of Integrative Neurophysiology, Centre Intention-to-treat analysis showed that light therapy was not superior to placebo in for Neurogenomics and Cognitive Research, Neu- reducing depressive symptoms (23.9 IDS points [95% CI 29.0 to 1.2]; P = 0.248) and roscience Campus Amsterdam, Amsterdam, the 2 P Netherlands had no effect on insulin sensitivity (0.15 mg/kg*min [95% CI 0.41 to 0.70]; = 8Amsterdam UMC, Vrije Universiteit and Univer- 0.608). Analyses incorporating only those participants who accurately adhered to sity of Amsterdam, Department of Medical Psy- the light therapy protocol (n = 51) provided similar results, but did suggest positive chology, Amsterdam Public Health research effects of light therapy on depression response rates (‡50% reduction in IDS points) institute, Amsterdam, the Netherlands (26% more response; P = 0.031). Prespeciﬁed analysis showed effect moderation Corresponding author: Annelies Brouwer, [email protected] by baseline insulin sensitivity (P = 0.009) and use of glucose-lowering medication P Received 14 August 2018 and accepted 19 Jan- ( = 0.023). Light therapy did not affect depressive symptoms in participants with uary 2019 higher insulin sensitivity or those who use only oral glucose-lowering medication or Clinical trial reg. no. NTR4942, www.trialregister none at all, but it did produce a relevant effect in participants with lower insulin .nl/trial/4802 sensitivity (212.9 IDS points [95% CI 221.6 to 24.2]; P = 0.017) and a trend This article contains Supplementary Data online toward effectiveness in those using insulin (212.2 IDS points [95% CI 221.3 to at http://care.diabetesjournals.org/lookup/suppl/ 23.1]; P = 0.094). Light therapy was well tolerated. doi:10.2337/dc18-1732/-/DC1. © 2019 by the American Diabetes Association. CONCLUSIONS Readers may use this article as long as the work is Although this trial is essentially inconclusive, secondary analyses indicate that light properly cited, the use is educational and not for proﬁt, and the work is not altered. More infor- therapy might be a promising treatment for depression among a subgroup of highly mation is available at http://www.diabetesjournals insulin-resistant individuals with type 2 diabetes. .org/content/license. 530 Light Therapy for Mood and Insulin Sensitivity Diabetes Care Volume 42, April 2019

One of five patients with type 2 diabetes So far, no experimental evidence is created blocks of six participants, with suffers from clinical depression (1). The available on the effects of light therapy randomly varying sequences within prevalence of depression among patients on either mood or glycemic control in blocks. Participants who agreed to par- with type 2 diabetes is double that people with type 2 diabetes. To our ticipate in the hyperinsulinemic-euglycemic among the general population, and this knowledge, we are the first to conduct clamp (HEC) procedure were randomized comorbidity is recognized as a global a randomized, placebo-controlled trial to separately. Upon approval by the ethics public health challenge (2,3). Previous test effects of light therapy on mood and committee, participants were informed research suggests a bidirectional cause- insulin sensitivity in people with depres- that the study investigated differences and-effect relationship between type 2 sion and type 2 diabetes. between white-yellow and green light diabetes and depression (4,5). In addition, therapy, but they were unaware that one shared underlying pathophysiological mech- RESEARCH DESIGN AND METHODS of the conditions was a placebo. Partic- anisms, such as a disturbance of the sleep- Design ipants were told that we could not ensure wake cycle, have been described (4,5). We performed a randomized, double- the effects of either intervention under Unfortunately, depression in people blind, placebo-controlled, parallel-arm study. The research assistants who de- with type 2 diabetes is associated with a trial to compare the effects of light livered the lamps to the participants’ poor prognosis and poor treatment re- therapy with those of a placebo. A de- homes were aware of the treatment al- sults. The depression seems to be more scription of the methods was published location but were not involved in any persistent and recurrent (6), and some at the start of the trial (18), and the trial other study activities; they received ex- antidepressant drugs may even worsen has been registered in the Netherlands plicit instructions regarding their inter- glycemic control (7). Furthermore, pa- Trial Register (NTR4942). Participants actions with the participants in order to tients with depression and type 2 di- were assessed at baseline (week 0), prevent disclosure of treatment alloca- abetes exhibit a higher risk for diabetes after 4 weeks of receiving light therapy or tion. All other study personnel, including complications and mortality (8). Clearly, placebo (week 4), and after 4 weeks of the research assistants and physicians in- more efficacious depression treatments are follow-up (week 8). Some measures were volved in obtaining measurements, were needed for patients with type 2 diabetes. obtained at weekly intervals during the fully blinded to the conditions. Light therapy, in which patients are intervention (weeks 1, 2, and 3). Study exposed to bright light early in the morn- procedures were performed at two loca- Intervention ing for several days, is a patient-friendly tions. Participant enrollment started in Participants were randomly assigned to and low-cost treatment, with early onset 2014 and ended in 2017. Participants gave receive active (broad-spectrum, white- of action and low risk for adverse effects informed consent before participating. The yellow light, 10,000 lux) or placebo (9). It is a first-line treatment for seasonal study was executed in accordance with the (monochromatic green light [545 nm], depression and has recently been proven Declaration of Helsinki and with approval 470 lux) light therapy at home, scheduled to be successful in treating nonseasonal from the medical ethics committee of Am- for 30 min every morning over 4 weeks. depression as well, even in several difficult- sterdam University Medical Centers (Am- CE-certified Diamond-5 SAD Lightbox to-treat patient groups (10–12). Light sterdam UMC), location Vrije Universiteit. light therapy lamps (SAD Lightbox Com- therapy is traditionally assumed to act pany, Berkshire, U.K.), adapted with light by entraining the sleep-wake cycle via Participants filters, were used. Estimates of irradiance ocular stimulation of the brain’s supra- Participants were recruited through ad- are provided in Supplementary Table 1. chiasmatic nucleus (the biological clock) vertisements, referrals from clinicians, Timing of light therapy was set in accor- (9), thereby improving sleep and circa- and databases of patients who con- dance with results from the Morningness- dian rhythmicity (13). Yet, recent findings sented to be informed about research Eveningness Questionnaire (MEQ) (19), also point to effects of light on mood via studies open for participation. Eligibility but each participant agreed to the time other cerebral pathways (14). The brain was determined by a telephone screening in order to ensure the therapy could biological clock plays a main role in and a screening visit. Included patients be incorporated in the participant’s regulating glucose metabolism, and 1) were 18 years or older, 2) had a docu- daily life. Research assistants gave in- sleep disturbances have been associated mented physician-based diagnosis of structions on how to apply the therapy. with higher incidence rates of type 2 type 2 diabetes and used glucose- Therapy compliance was monitored diabetes, hyperglycemia, and insulin lowering medication or had HbA1c weekly by telephone and reinforced resistance (15). .42 mmol/mol (5.9%), and 3) had a when necessary. On the basis of these findings, we major depressive episode according to hypothesized that light therapy may be DSM-IV criteria. Participants were excluded Primary Outcome Measures an effective antidepressant in people with if they had a medical condition or recently The prespecified primary outcome depression and type 2 diabetes, and that experienced a medical event that could measures were changes in depressive it might concurrently improve insulin sen- potentially compromise the effects symptoms and insulin sensitivity from sitivity. This hypothesis is supported by or safety of light therapy (for specifics, baseline (week 0) to the end of the in- two case reports that describe an increase see Supplementary Data, Information 1). tervention period (week 4). in insulin sensitivity and hypoglycemia, Depressive symptoms were assessed and less need for insulin, in two patients Randomization and Masking by using the self-report version of the with diabetes who received light therapy Participants were randomized to groups Inventory of Depressive Symptomatol- to treat seasonal depression (16,17). by using a computer-generated table, which ogy (IDS). A score $14 indicates clinically care.diabetesjournals.org Brouwer and Associates 531

significant symptoms of depression. Re- (22,23), we expected moderate effect candidate effect moderators tested in- sponse to the intervention was defined as sizes for depressive symptoms and large cluded age, sex, antidepressant medica- a reduction of IDS score by $50%. De- effect sizes for insulin sensitivity. The tion, and glucose-lowering medication. In pression remission was definedasanIDS planned sample size of 84 yields 80% addition, we performed prespecified me- score #13. Relapse was definedasanIDS power for detecting a moderate effect diation analyses (the Sobel test) to eval- score $26. size (Cohen’s d of at least 0.50 through- uate whether the effects of light therapy Insulin sensitivity was evaluated using out follow-up) for depressive symptoms, on depressive symptoms and insulin sen- an HEC, the gold standard for quantifying assuming four repeated measurements sitivity were mediated by changes in whole-body insulin sensitivity (20). This per participant and a maximum within- sleep (subjective insomnia symptoms, procedure quantifies the rate of exoge- subject correlation of 0.6. The planned objective sleep duration, sleep efficiency, nous glucose infusion required to main- sample size of 54 for the HEC subsample and mid-sleep time). tain the blood glucose concentration at yields 80% power to detect a large effect Significance was set at P= 0.05. We did not euglycemic levels (goal 5.0 mmol/L) in size (Cohen’s d of at least 0.78) for a sin- power the study for post hoc per-protocol response to a fixed increase in blood in- gle follow-up measurement at 4 weeks. effect moderation analyses or mediation sulin concentration through administra- analysis, and therefore those should be tion of insulin (i.e., NovoRapid 100 IU/mL; Statistical Analyses regarded as exploratory. Statistics were Novo Nordisk, Bagsvaerd, Denmark)) at The balance of randomization was tested analyzed by using IBM SPSS Statistics 22. 40 mU/m2*min. After 90 min, a steady with the independent samples t test state is reached, which allows the cor- and x2 test. Primary analyses of out- RESULTS rected M-value to be calculated; this come measures included all participants Sample Selection value, expressed as the amount of glu- (intention-to-treat); secondary analyses In total, 155 participants were screened, of cose infusion (mg/kg*min), reflects mus- included analyses in which only those whom 43 did not meet inclusion criteria, cle insulin sensitivity. Lower M-values participants who accurately adhered to 16 were excluded, and 13 declined to par- represent lower insulin sensitivity. the light therapy protocol were incorpo- ticipate (Supplementary Data, Information 1). rated (per-protocol). Differences between Thus we included 83 adults with depression Secondary Outcome Measures conditions are expressed as the mean and type 2 diabetes, of whom 43 were Secondary outcome measures included absolute difference at week 4, with the allocated to receive light therapy and 40 to anxiety symptoms (measured by using 95% CI and the standardized difference or receive placebo. A subgroup of 60 par- the Beck Anxiety Inventory [BAI]); dia- effect size (Cohen’s d [the mean difference ticipants took part in the HEC procedure betes distress (measured by using the divided by the pooled SD] or relative risk). (Supplementary Data, Information 1). Problem Areas in Diabetes [PAID] ques- Linear mixed model (LMM) (continuous One participant from the light therapy tionnaire; a score $33 indicates high outcomes) and generalized estimating group was excluded from the analyses diabetes distress); self-reported insom- equation (GEE) (dichotomous outcomes) because their depression remitted (IDS nia symptoms (based on the Insomnia analyses included all measures from week score #13) between the screening and Severity Index [ISI]; a score $10 indi- 1 to week 4. To correct for differences at baseline measurements. Three patients cates clinically relevant insomnia); and baseline, baseline values were incorpo- discontinued the protocol after the base- objective sleep duration, sleep efficiency, rated into the model as covariates. Effects line measurements: one for unknown and mid-sleep time, derived from acti- at follow-up (week 8) were evaluated in reasons (placebo group) and two for graphy data through the use of a validated separate analyses. For HbA1c,anindexof medical reasons unrelated to the inter- algorithm. Glucometabolic secondary mean blood glucose concentrations vention (light therapy group). All partic- outcome measures included HbA1c con- within the past 2–3 months, primary ipants who received the intervention centration, fasting blood glucose (FBG) analyses included data from week 4 completed the study protocol. Eight pa- concentration (among the HEC subsam- and week 8. tients discontinued the HEC protocol: ple), self-reported hypoglycemic events, Effect moderation was analyzed in or- two because of difficult venous access, and body weight. Expectations regard- der to identify subgroups that benefitted one because of discomfort during the ing treatment were assessed by using a most from light therapy. For this pur- procedure at baseline, three because of four-item questionnaire. pose we used linear regression with unforeseen absence of the HEC-supervising depressive symptoms or the insulin research-physician, and two due to dis- Side Effects and Adverse Events sensitivity change score as dependent continuation of the general protocol. Side effects and adverse events were in- variables and the condition, the moder- Outcome measures were obtained from ventoried weekly using a side effect ques- ator, and their two-way interaction as 79 participants (50 participants in the HEC tionnaire and telephonic controls. Visual predictors. A candidate moderator was subsample). A flowchart of the inclusion acuity, contrast sensitivity, and retinopathy declared an effect moderator when the process is shown in Supplementary Fig. 1. grade were analyzed at the screening visit two-way interaction was significant. Pre- and at follow-up (week 8). specified candidate effect-moderators Protocol Deviations included depressive symptoms at base- We did not foresee 13 participants Sample Size line, depression symptoms profile, insu- changing their glucose-lowering medica- On the basis of previous studies of light lin sensitivity, HbA1c, circadian rhythm tion during the protocol. One participant therapy (10–12,21) and of the effects type, insomnia severity, seasonality of used an oral corticosteroid during the of sleep restriction on insulin sensitivity depression, and time of year. Other protocol. Glucometabolic measures that 532 Light Therapy for Mood and Insulin Sensitivity Diabetes Care Volume 42, April 2019

could have been affected by medication Table 1—Baseline characteristics changes were excluded from analyses be- Light therapy Placebo fore treatment allocation was disclosed. (n = 42) (n = 40) Exclusion was equally distributed over the General characteristics light therapy and placebo conditions. Age, years (mean 6 SD) 60.1 6 9.8 62.9 6 10.7 Per-protocol analyses included 51 of Sex, n 82 participants (35 of 58 in the HEC Male 21 24 Female 21 16 subsample). Accurate treatment adher- Ethnicity, n ence was defined as $21 days of light European 38 36 therapy within 120 min of the time Suriname or Dutch Antilles 2 2 South Asian 1 2 recommended by the MEQ. Unknown 1 0 Comorbid diseases in number of medications (mean 6 SD) 9.0 6 5.4 9.3 6 4.4 Sample Characteristics Polypharmacy ($5 medications), % 74 83 2 Randomization was balanced regarding BMI, kg/m (mean 6 SD) 33.1 6 5.4 31.9 6 5.7 the demographic and clinical character- Depression-related characteristics Major depressive disorder type istics of the total sample and of the HEC Single episode 22 22 subsample (Table 1 and Supplementary Recurrent 12 16 Table 2). As BMI was almost significantly Unknown 8 2 fi different (P = 0.088) for the HEC sub- Time since rst episode of major depressive disorder, years (mean 6 SD) 21.7 6 14.6 23.7 6 18.7 group, we additionally analyzed the ef- Depression severity by IDS score classification, n fects of light therapy on insulin sensitivity Mild 8 13 adjusted for BMI. Moderate 21 17 Severe 9 5 Very severe 4 5 Depression Antidepressant medication, n After 4 weeks, light therapy had not None 19 21 statistically significantly reduced depres- Selective serotonin reuptake inhibitor or similar 18 14 Tricyclic antidepressant 5 5 sive symptoms (between-arm differ- Comorbid anxiety disorder, n ence: 23.9 IDS points, or 29.2% [95% No 14 14 CI 29.0 to 1.2 IDS points]; P = 0.248, Yes 28 26 LMM; Cohen’s d = 0.35) (Fig. 1 and Table Benzodiazepine use, n No 33 32 2). When considering response and re- Yes 9 8 mission rates, light therapy did not result Diabetes-related characteristics in more response or remission than Diabetes distress, PAID score (mean 6 SD) 31.4 6 25.0 27.6 6 23.0 that achieved with placebo at week 4 Duration of diabetes, years (mean 6 SD) 10.6 6 6.4 12.7 6 7.9 6 6 6 (between-arm difference: 16% more HbA1c, mmol/mol (mean SD) 54.9 12.2 55.2 14.2 HbA1c, % (mean 6 SD) 7.2 6 1.1 7.2 6 1.3 response (relative risk 1.55 [95% CI FBG, mmol/L (mean 6 SD)* 7.8 6 1.9 7.3 6 2.2 0.84–2.86]; P =0.242,GEE),0%more Insulin sensitivity, M-value as mg/kg*min (mean 6 SD)* 2.03 6 1.02 1.86 6 1.22 remission (relative risk 1.00 [95% CI Glucose-lowering medication, n – None 3 3 0.49 2.03]; P = 0.430, GEE) (Table 2). Oral 26 23 Similar results were obtained 4 weeks Insulin 13 14 after the intervention period was dis- Sleep, chronobiological, and seasonal characteristics continued (week 8) (between-arm dif- Obstructive sleep apnea risk per BQ, n ference: 22.2 IDS points, or 24.1% Low 5 5 High 32 29 [95% CI 27.3 to 2.8 IDS points]; P = Unknown 5 6 0.483, LMM), 8% more response [P = Insomnia symptoms, ISI score (mean 6 SD) 13.9 6 6.3 13.6 6 7.6 0.492, GEE], 5% more remission [P = 0.624, Sleep duration, h† (mean 6 SD) 6.7 6 1.7 6.8 6 1.2 Sleep efficiency, % of time asleep while intending GEE]) (Fig. 1 and Supplementary Table 3). to sleep† (mean 6 SD) 84 6 9866 7 Baseline characteristics did not differ Mid-sleep time, time† (mean 6 SD) 04:04 6 1:27 04:19 6 1:46 between those who responded and Circadian rhythm type, MEQ score those who did not respond to light (mean 6 SD) 50.9 6 12.2 51.5 6 11.7 Seasonality of symptoms, GSS score (mean 6 SD) 6.4 6 5.4 7.4 6 4.8 therapy (Supplementary Table 4). Time of year in protocol, n Autumn 7 10 Insulin Sensitivity Winter 8 12 Spring 18 13 Light therapy had no effect on insulin Summer 9 5 sensitivity (between-arm difference in BQ, Berlin questionnaire; GSS, Global Seasonality Scale. *HEC subsample: light therapy, n = 31; M-value at week 4: 0.15 mg/kg*min placebo, n = 27. PAID scores $33 indicate high diabetes distress. ISI scores $10 indicate clinically [95% CI 20.41 to 0.70]; P = 0.608, relevant insomnia.MEQscores#41 indicate“eveningtypes,” scores42–58 indicate “intermediate LMM; Cohen’s d = 0.16) (Fig. 1 and Table types,” scores $59 indicate “morning types.” GSS scores $11 indicate a seasonal pattern of 2). Analysis in which BMI was incorporated symptoms. Test statistics/P values are shown in Supplementary Table 1. †Calculated from actigraphy-derived data. into the model as a covariate yielded care.diabetesjournals.org Brouwer and Associates 533

0.44]; in hypoglycemic events: 0.51 day/ week [95% CI 20.00 to 1.02] at week 4[P = 0.506, GEE; Cohen’s d = 0.42] and 0.10 day/week [95% CI 20.25 to 0.46] at week 8) (Fig. 1, Table 2, and Supplementary Table 3). Body weight did not change in response to light therapy (between-arm difference at week 4: 20.23 kg [95% CI 20.73 to 0.27] [P = 0.420, LMM; Cohen’s d = 0.23]; between- arm difference at week 8: 20.66 kg [95% CI 21.40 to 0.07] [P = 0.069, LMM]) (Table 2 and Supplementary Table 3). Light therapy had no effect on subjective insomnia symptoms, objective sleep duration, sleep efficiency, or mid-sleep time at week 4 (between- arm differences in ISI points: 20.2 points [95% CI 21.9 to 2.2]; P = 0.783, LMM; Cohen’s d = 0.04; in sleep duration: 213 min [95% CI 253 to 27]; P = 0.329, LMM; Cohen’s d = 0.16); in sleep efficiency: 20.8% [95% CI 24.3 to 2.8]; P =0.130, LMM; Cohen’s d =0.38;andinmid-sleep time: 22 min [95% CI 227 to 24]; P = 0.580, LMM; Cohen’s d = 0.03]) or at week 8 (ISI points: 20.2 points [95% CI 22.1 to 2.5]; P = 0.951, LMM]; sleep duration: 26min[95%CI27to59];P=0.266,LMM; sleep efficiency: 0.7% [95% CI 23.4 to 4.9]; P = 0.804, LMM; mid-sleep time: 8 min [95% CI 13–30]; P = 0.577, LMM) (Table 2 and Supplementary Table 3).

Expectancy Expectancy scores did not differ between conditions or between responders (reduction of IDS score by $50%) and nonres- ponders. Expectancy scores were not correlated with the percentage reduction Figure 1—Effects of light therapy on depression, insulin sensitivity, and secondary outcome in depressive symptoms (IDS points) (Sup- measures. Data points represent the mean values for the light therapy (blue squares) and placebo plementary Table 5). (green triangles) arms and the mean difference between conditions (light therapy minus placebo; baseline corrected) (red circles). The error bars represent the 95% CI. Week 0 represents the baseline; week 4, the end of the intervention period; week 8, follow-up. The intervention period is marked Treatment Adherence in gray. Treatment adherence did not differ between patients receiving the light similar results (P = 0.546, LMM). Mean CI 210.5 to 1.6]; P = 0.142, LMM; Cohen’s therapy and those receiving placebo blood glucose concentration during the d = 0.33; between-arm difference at week 8: (Supplementary Table 6). On average, steady state was 5.0 mmol/L (SD 0.4 21.0 BAI points [95% CI 24.5 to 2.6]; P = participants were exposed to light ther- mmol/L); mean plasma insulin concen- 0.503, LMM; and 21.9 PAID points [95% apyorplacebofor24days,atameantime tration was 434 pmol/L (SD 81 pmol/L) CI 28.2 to 4.4]; P = 0.574, LMM) (Fig. 2, of 7:56 A.M., which was, on average, (subsample; n 5 10 HEC procedures). Table 1, and Supplementary Table 3). 84 min later than the time recommended Light therapy did not affect HbA1c, FBG by the MEQ. Participants generally Secondary Outcome Measures concentration, or the number of self- kept to the time that they agreed to Differences in anxiety symptoms and reported hypoglycemic events (be- during the screening visit; only one par- diabetes distress between conditions tween-arm differences in HbA1c: 2.0 ticipant postponed the time of their were not signiﬁcant (between-arm dif- mmol/mol [95% CI 20.5 to 4.4] at therapy. One participant receiving light ference at week 4: 21.3 BAI points [95% week 8 [P = 0.116, LMM; Cohen’s d = therapy decided to stop the intervention CI 25.1 to 2.4]; P = 0.848, LMM; Cohen’s 0.42]; in FBG: 0.8 mmol/L (95% CI 20.2 to after 2 weeks because of headaches d = 0.16; and 24.4 PAID points [95% 1.8) at week 4 [P = 0.181, LMM; Cohen’s d = related to the intervention. 3 ih hrp o odadIslnSensitivity Insulin and Mood for Therapy Light 534

Table 2—Effects of light therapy on depression, insulin sensitivity, and secondary outcome measures Intention-to-treat analysis Per-protocol analysis Cohen’s d or Cohen’s d or relative Test statistic, relative risk Test statistic, Difference risk (95% CI) LMM/GEE W1-W4 (95% CI) LMM/GEE W1-W4 Condition DW0-W4 at W4* at W4 (P value) DW0-W4 Difference at W4* at W4 (P value) Primary outcomes Depressive Light therapy 212.9 23.9 (29.0 to 1.2) 0.35 F = 1.357 (0.248) 212.9 25.9 (213.0 to 1.2) 0.47 F = 0.809 (0.373) symptoms, Placebo 29.0 27.0 IDS score Insulin sensitivity, Light therapy 0.22 0.15 (20.41 to 0.70) 0.16 F = 0.266 (0.608) 0.17 20.05 (20.86 to 0.76) 0.05 F = 0.004 (0.952) M-value as Placebo 0.08 0.22 mg/kg*min Secondary outcomes Depression response, Light therapy 44% 16% 1.55 (0.84–2.86) Wald x2 = 1.366 (0.242) 43% 26% 2.46(0.92–6.62) Wald x2 = 4.643 (0.031) % of participants Placebo 28% 17% Depression remission, Light therapy 28% 0% 1.00 (0.49–2.03) Wald x2 = 0.622 (0.430) 25% 21% 0.96(0.37–2.46) Wald x2 = 0.195 (0.659) % of participants Placebo 28% 26% Anxiety symptoms, Light therapy 26.6 21.3 (25.1 to 2.4) 0.16 F = 0.037 (0.848) 27.1 22.1 (26.9 to 2.6) 0.25 F = 0.035 (0.853) BAI score Placebo 25.3 25.0 Diabetes distress, Light therapy 28.2 24.4 (210.5 to 1.6) 0.33 F = 2.203 (0.142) 28.8 25.3 (212.8 to 2.2) 0.40 F = 2.619 (0.112) PAID score Placebo 23.8 23.5 FBG, mmol/L Light therapy 0.8 0.8 (20.2 to 1.8) 0.44 F = 1.842 (0.181) 0.4 0.3 (20.7 to 1.3) 0.22 F = 1.092 (0.306) Placebo 0.1 0.1

HbA1c, mmol/mol† Light therapy 1.4 2.0 (20.5 to 4.4) 0.42 F = 2.534 (0.116) 2.1 2.4 (20.6 to 5.4) 0.53 F = 1.117 (0.297) Placebo 20.5 20.3 Hypoglycemic events, Light therapy 0.12 0.51 (20.00 to 1.02) 0.42 Wald x2 = 0.443 (0.506) 0.08 0.50 (20.12 to 1.12) 0.55 Wald x2 = 0.513 (0.474) days/week Placebo 20.39 20.42 Body weight, kg Light therapy 0.02 20.23 (20.73 to 0.27) 0.23 F = 0.658 (0.420) 0.09 20.24 (20.86 to 0.39) 0.24 F = 0.464 (0.500) Placebo 0.25 0.33

Insomnia, ISI score Light therapy 22.5 20.2 (21.9 to 2.2) 0.04 F = 0.077 (0.783) 22.9 21.2 (21.0 to 3.4) 0.31 F = 1.746 (0.193) Care Diabetes Placebo 22.6 21.7 Sleep duration,‡ min Light therapy 219 213 (253 to 27) 0.16 F = 0.969 (0.329) 29 16 (224 to 56) 0.28 F = 0.342 (0.563) Placebo 27 225 Sleep efﬁciency,‡ % Light therapy 20.6 20.8 (24.3 to 2.8) 0.38 F = 2.354 (0.130) 0.8 20.7 (25.4 to 4.1) 0.10 F = 1.512 (0.228) oue4,Arl2019 April 42, Volume Placebo 0.2 1.5 Mid-sleep time,‡ min Light therapy 215 22(227 to 24) 0.03 F = 0.309 (0.580) 226 211 (250 to 28) 0.19 F = 0.048 (0.827) Placebo 214 215

NA, not available; W, week. *Data are the mean difference (95% CI) or the percentage difference. †HbA1c: Statistical testing included data from W4 and W8; Cohen’s d was calculated on the basis of the mean difference between W0 and W8. ‡Calculated from actigraphy-derived data. care.diabetesjournals.org Brouwer and Associates 535

Adverse Effects Four participants were hospitalized for several days because of serious adverse events not related to the study: non- specific chest pain (n = 2), gastrointestinal bleeding from known esophageal varices (n = 1), and renal colic (n = 1). Adverse events that occurred in relation to the study protocol, such as headache and ocular complaints (n = 25), seemed to be equally common among patients in both Figure 2—Effects of light therapy on depression according to insulin sensitivity (left) and glucose- groups (Supplementary Table 10, with ac- loweringmedicationuse (right)at baseline. Thebar plots showthe changein depressivesymptoms companying information). Self-reported from baseline (week 0 [W0]) to week 4 (W4) in the groups receiving light therapy (blue striped bars) adverse effects did not differ between or placebo (green dashed bars); error bars indicate the 95% CI. In the left panel, subgroups are sorted according to median value (lower #1.75, higher .1.75 M-value mg/kg*min). those receiving light therapy and those receiving placebo (Supplementary Table 10, with accompanying information). Study Effects on Sleep and Circadian d=1.19),whereasnoeffectwasobserved Light therapy did not affect best- Rhythmicity in those with higher insulin sensitivity corrected visual acuity (between-arm Analyses showed that in the 1st week of (subgroups sorted according to median difference: right eye, 20.08 [95% CI the intervention period all participants value) (Fig. 2 and Supplementary Table 20.10 to 0.02]; P = 0.122; left eye, 20.08 got out of bed earlier, both those re- 9). Subgroup analysis based on glucose- [95% CI 20.19 to 0.03]; P = 0.156) or 2 ceiving the light therapy ( 36 min [95% lowering medication showed a trend binocular contrast sensitivity (between- 2 2 , CI 49 to 24 min]; P 0.001) and toward positive effects of light therapy arm difference 20.03 [95% CI 20.10 to 2 those receiving the placebo ( 16 min on depressive symptoms in those using 0.03]; P = 0.263). One participant who re- 2 2 [95% CI 29 to 2]; P = 0.025), probably insulin (212.2 IDS points [95% CI 221.3 ceived light therapy demonstrated progres- so they could follow the intervention at to 23.1]; P = 0.094, LMM; Cohen’s d = sion of diabetic retinopathy (European the agreed-upon time. The intervention 1.14), whereas no effect was observed in Diabetes [EURODIAB] Prospective Compli- period (irrespective of condition) did those who did not use glucose-lowering cations Study stage 1–2) during the protocol. notresultinsignificant changes in medicationorwhousedanoralform sleep duration, sleep efficiency, or (Supplementary Table 9). Subgroups CONCLUSIONS mid-sleep time. Changes in these mea- based on insulin sensitivity and glucose- This is, to our knowledge, the first con- sures were not associated with changes lowering medication partially consisted trolled experiment to investigate effects in depressive symptoms (Supplementary of similar participants: 55% of the of light therapy on mood and insulin Table 7). participants using insulin were also sensitivity in patients with type 2 dia- in the lower sensitivity subgroup. Com- betes and comorbid depression. The Per-Protocol Analyses bined modeling of both interaction execution of the study seemed to be Per-protocol analyses provided results sim- effects suggests that the interaction successful, given the balanced random- ilar to those of the intention-to-treat anal- effect of glucose-lowering medication ization and patient expectations, few yses but suggest larger effects of light (P = 0.094) is partly explained by dropouts, and acceptable treatment therapy on depressive symptoms (25.9 the effect of insulin sensitivity (P = adherence. Intention-to-treat analysis IDS points, or 214.3% [95% CI 213.0 to 1.2 0.014). showed no superiority of light therapy IDS points]; Cohen’s d = 0.47) (Table 2). In Other effect moderators we tested over placebo in reducing depression this subgroup of participants, light therapy did not moderate the effect of light symptoms (IDS points). Nor did light resulted in 26% more response ($50% therapy on depressive symptoms or therapy affect insulin sensitivity. Similar reduction in depressive symptoms [IDS insulin sensitivity (Supplementary results were obtained for the secondary points]) (relative risk 2.46 [95% CI 0.92– Table 8). outcomes of anxiety symptoms, diabe- 6.62]; P = 0.031, GEE) (Table 2). tes distress, insomnia symptoms, objective Effect Mediation Analyses sleep duration, sleep efficiency, mid-sleep Effect Moderation Analyses Any effects of light therapy on depres- time,FBG,HbA1c, hypoglycemic events, Glucose-lowering drugs (insulin vs. not sive symptoms and insulin sensitivity and body weight. Per-protocol analysis, insulin; P = 0.023) and baseline insulin were not mediated by changes in sub- however, suggested positive effects of light sensitivity (P = 0.009) moderated the jective insomnia symptoms, objective therapy on depressive symptoms, under- effect of light therapy on depressive sleep duration, objective sleep effi- scoring the importance of adherence to symptoms (Supplementary Table 8). Sub- ciency, or mid-sleep time (depressive light therapy. Moreover, baseline insulin group analysis showed positive effects of symptoms: P =0.861,P = 0.590, P = sensitivity moderated the effect of light light therapy on depressive symptoms 0.794, P = 0.903, respectively; insulin therapy on depressive symptoms, with in those with lower insulin sensitivity sensitivity: P = 0.926, P =0.961,P = clinically relevant effects in favor of light (212.9 IDS points [95% CI 221.6 0.860, P = 0.558, respectively [all values therapy in patients with highly insulin- to 24.2]; P = 0.017, LMM; Cohen’s calculated with the Sobel test]). resistant type 2 diabetes. 536 Light Therapy for Mood and Insulin Sensitivity Diabetes Care Volume 42, April 2019

Although previous research has con- to improvement in depressive symptoms effects turned out to be most distinct firmed the effectiveness of light therapy (26–28). This is further supported by the after 4 weeks. Also, the exclusion of for depression in a variety of patient antidepressant effects of medications several glucometabolic measurements groups, we could not establish such an that improve insulin sensitivity (28). This from analyses because of changes in effect in this sample of patients with mechanism of antidepressant action could glucose-lowering medication, albeit type 2 diabetes. This may be due to a lack be more pronounced in patients with equally for both conditions, potentially of power to affirm the observed small marked insulin-resistant type 2 diabetes (26). biases the glucometabolic results. Fur- standardized effect size, as the study Contrarytotwo casereports(16,17),our thermore, energy expenditure, caloric was designed to detect a moderate ef- study showed no changes in insulin sen- intake, timing of food intake and physical fect size. The absolute effect size, sitivity in response to light therapy. These activity (in general or relative to the howeverd9.2% higher reduction of de- negative findings may be explained by a intervention), exposure to environmen- pressive symptoms with light therapy lack of an effect of light therapy on circa- tal light, and circadian timing of measure- than with placebodwas comparable dian and sleep measures, as we hy- ments were not taken into account, to effect sizes found in previous studies pothesized that these would mediate although these factors may have influ- of the effects of light therapy and other improvements in insulin sensitivity. The enced findings. Last, we did not assess antidepressant treatments (10–12,21). majority of the participants in this study other factors that hypothetically might This discrepancy between the standard- are at risk for sleep apnea, whereas im- have mediated improvements in insulin ized and absolute effect sizes suggests provements in sleep and thereby in insulin sensitivity in response to light therapy, more response variability in our study. sensitivity as effects of light therapy may such as melatonin and cortisol concen- Response variability may be due to var- notbepossibleinthesepatients(29).Also, trations, autonomic nervous system iation in treatment adherence and the baseline circadian and sleep measures may function (15), and changes in health heterogeneity of our sample, which is have been disordered in different ways for behaviors. The observed placebo effect supported by both the per-protocol and different individuals, so that unidirectional regarding depressive symptoms is proba- the effect moderation analyses. Yet, our effects of an intervention, such as dem- bly attributable to general study-related negative findings might also represent onstrated by a sleep-extension study in factors, which is suggested by the mag- the genuine effect of light therapy in this chronically sleep-restricted healthy volun- nitude of the placebo effect being com- specific population, as depression in people teers, could not be expected (30). Last, parable to that found in previous studies with type 2 diabetes seems to be more effects of sleep restoration on insulin of light therapy and oral antidepressant persistent and recurrent than in those sensitivity may not be as big as effects drugs (21,39). It seems unlikely that gen- without type 2 diabetes (6). Also, improving of sleep restriction, or they may be small eral effects of our study design on sleep or depression and sleep in patients with sleep whensleep is only minimally disordered or circadian rhythmicity contributed to a pla- apnea, which is prevalent among those in chronically disturbed (31,32). Moderate to cebo effect, as the study did not result in our study sample, may be difficult. small effects on insulin sensitivity would general changes in sleep duration, sleep Prespecified moderation analyses havegone unnoticed in this study, as it was efficiency, or mid-sleep time, and changes showed significant and clinically relevant powered to detect large effects of light in these measures were not associated effects of light therapy on depressive therapy on insulin sensitivity. with changes in depressive symptoms. symptoms in patients with marked in- Light therapy did not affect anxiety Potential opportunities to increase the sulin resistance, whereas no effect of symptoms and diabetes distress, two effectiveness of light therapy in future light therapy was observed in more in- measures that to our knowledge have studies are related to the timing of the sulin-sensitive patientsdan intriguing not been studied previously. Nor did light therapy, treatment adherence, and the finding, as this subgroup may represent therapy affect glycemic control. On the specific light therapy protocol. For in- patients with more advanced disease. basis of recent publications, we may have stance, the timing of light therapy, when Previous studies have demonstrated anticipated glycemic control to deterio- based on the onset of melatonin con- larger effects of light therapy in those rate as a result of the acute effects of light centration under dim light conditionsd with a stronger appetite and a higher on both preprandial and postprandial which is the most accurate marker BMI, factors that are associated with glucose concentrations (33–35). Yet, for assessing an individual’scircadian insulin resistance (24,25). A similar re- 30 min of light daily may not be enough to rhythmdmay improve treatment results lationship between antidepressant ef- change HbA1c. Although previous reports (19). Also, compliance can be optimized fect and insulin resistance at baseline have suggested an increase in the num- with, for example, daily monitoring by was demonstrated in a study that in- ber of hypoglycemic events (16,17) and a using wearable light-measuring devices. vestigated the mood-altering effects of decrease in body weight (36–38) with Furthermore, larger effects could be pioglitazone (26). Depressive symptoms light therapy, we could not replicate those achieved with prolonged daily exposure are associated with insulin resistance, findings. to light, as the effect of light therapy and insulin resistance seems to im- Limitations of our study include the seems to increase with the intensity and prove as depressive symptoms improve inabilitytoobtaintherequirednumberof duration of daily exposure to light (9,40) (26,27). This has led to the hypothe- complete data sets and the selection and with increased duration of the ther- sis that improvement of insulin resis- of repeated measures a priori, which apy, as indicated by the increasing effect tance, or of related factors such as might have contributed to a further de- of light therapy over time in our study. chronic low-grade inflammation, may crease in statistical power for the analy- Although this is essentially an incon- be the underlying mechanism leading ses regarding depressive symptoms, as clusive trial, we did find evidence to care.diabetesjournals.org Brouwer and Associates 537

suggest that light therapy could be a 3. Fisher EB, Chan JC, Nan H, Sartorius N, insulin secretion and resistance. Am J Physiol promising treatment for depression in Oldenburg B. Co-occurrence of diabetes and 1979;237:E214–E223 a subgroup of insulin-using and highly depression: conceptual considerations for an 21. Lieverse R, Van Someren EJ, Nielen MM, emerging global health challenge. J Affect Disord Uitdehaag BM, Smit JH, Hoogendijk WJ. Bright insulin-resistant individuals with type 2 2012;142(Suppl.):S56–S66 light treatment in elderly patients with non- diabetes. Light therapy is a well-tolerated 4. Tabak´ AG, Akbaraly TN, Batty GD, Kivimaki¨ M. seasonal major depressive disorder: a random- intervention, also in patients with type 2 Depression and type 2 diabetes: a causal asso- ized placebo-controlled trial. Arch Gen Psychiatry diabetes. The findings of this trial warrant ciation? Lancet Diabetes Endocrinol 2014;2:236– 2011;68:61–70 245 22. Donga E, van Dijk M, van Dijk JG, et al. Partial further study of the antidepressant and 5. Moulton CD, Pickup JC, Ismail K. The link sleep restriction decreases insulin sensitivity in glucometabolic effects of light therapy in between depression and diabetes: the search type 1 diabetes. Diabetes Care 2010;33:1573– people with highly insulin-resistant type 2 for shared mechanisms. Lancet Diabetes Endo- 1577 diabetes, and encourage research into crinol 2015;3:461–471 23. DongaE,vanDijkM,vanDijkJG,etal.Asingle the mechanisms by which insulin resistance 6. de Groot M, Crick KA, Long M, Saha C, night of partial sleep deprivation induces insulin Shubrook JH. Lifetime duration of depressive resistance in multiple metabolic pathways in moderates the antidepressant effect of disorders in patients with type 2 diabetes. Di- healthy subjects. J Clin Endocrinol Metab light therapy. abetes Care 2016;39:2174–2181 2010;95:2963–2968 7. Deuschle M. Effects of antidepressants on 24. Levitan RD, Levitt AJ, Michalak EE, et al. glucose metabolism and diabetes mellitus type 2 in Appetitive symptoms differentially predict treat- adults. Curr Opin Psychiatry 2013;26:60–65 mentresponseto fluoxetine,light,and placeboin Acknowledgments. The authors thank Michaela 8. Petrak F, Baumeister H, Skinner TC, Brown A, nonseasonal major depression. J Clin Psychiatry Diamant (deceased), Professor of Diabetology Holt RIG. Depression and diabetes: treatment 2018;79:17m11856 at the VU University Medical Center, Amsterdam, and health-care delivery. Lancet Diabetes Endo- 25. Dimitrova TD, Reeves GM, Snitker S, et al. the Netherlands; the research assistants; and crinol 2015;3:472–485 Prediction of outcome of bright light treatment the participants for their contributions to this 9. Wirz-Justice A, Benedetti F, Terman M. Chro- in patients with seasonal affective disorder: study. notherapeutics for Affective Disorders: A Clini- discarding the early response, confirming a Funding. This study was supported by a European cian’s Manual for Light and Wake Therapy. 2nd higher atypical balance, and uncovering a Foundation for the Study of Diabetes (EFSD)/Lilly rev. ed. Basel, Switzerland, Karger, 2013 higher body mass index at baseline as predictors Mental Health award 2012. 10. Tuunainen A, Kripke DF, Endo T. Light ther- of endpoint outcome. J Affect Disord 2017;222: fl Duality of Interest. No potential con icts of apy for non-seasonal depression. Cochrane Da- 126–132 interest relevant to this article were reported. tabase Syst Rev 2004;(2):CD004050 26. Lin KW, Wroolie TE, Robakis T, Rasgon NL. Author Contributions. A.B. performed statis- 11. Perera S, Eisen R, Bhatt M, et al. Light therapy Adjuvant pioglitazone for unremitted depres- tical analyses, wrote the manuscript, set up and for non-seasonal depression: systematic review sion: clinical correlates of treatment response. coordinated the trial, and performed study pro- and meta-analysis. BJPsych Open 2016;2:116– Psychiatry Res 2015;230:846–852 cedures. D.H.v.R. wrote the manuscript, de- 126 27. Kan C, Silva N, Golden SH, et al. A systematic signed and interpreted the diabetes-related 12. Martensson˚ B, Pettersson A, Berglund L, review and meta-analysis of the association outcome measures, and coordinated practical Ekselius L. Bright white light therapy in depres- between depression and insulin resistance. Di- research assistance. H.-T.N. and A.C.M. wrote the sion: a critical review of the evidence. J Affect abetes Care 2013;36:480–489 manuscript, designed and interpreted the oph- Disord 2015;182:1–7 28. Moulton CD, Hopkins CWP, Ismail K, Stahl D. thalmological measures, and graded retinopa- 13. van Maanen A, Meijer AM, van der Heijden Repositioning of diabetes treatments for de- thy. F.R. and P.J.M.E. wrote the manuscript, KB, Oort FJ. The effects of light therapy on sleep pressive symptoms: a systematic review and recruited participants, and coordinated practical problems: a systematic review and meta- meta-analysis of clinical trials. Psychoneuroen- research assistance. P.M.v.d.V. wrote the man- analysis. Sleep Med Rev 2016;29:52–62 docrinology 2018;94:91–103 uscript and designed statistical analyses. E.J.W.V.S. 14. LeGates TA, Altimus CM, Wang H, et al. 29. Tasali E, Leproult R, Ehrmann DA, Van Cauter E. wrote themanuscriptanddesignedandinterpreted Aberrant light directly impairs mood and Slow-wave sleep and the risk of type 2 diabetes in the trial conditions and sleep measures. F.J.S. and learning through melanopsin-expressing neu- humans. Proc Natl Acad Sci U S A 2008;105:1044– A.T.F.B. conceived of the study, designed the study, rons. Nature 2012;491:594–598 1049 interpreted the results, and wrote the manu- 15. Stenvers DJ, Scheer F, Schrauwen P, la 30. Leproult R, Deliens G, Gilson M, Peigneux script. M.A.B. conceived of the study, set up the Fleur SE, Kalsbeek A. Circadian clocks and P. Beneficial impact of sleep extension on study design, supervised the coordination of the insulin resistance. Nat Rev Endocrinol 2018; fasting insulin sensitivity in adults with habit- trial, wrote the manuscript, and is the principal 15:75–89 ual sleep restriction. Sleep (Basel) 2015;38: investigator of this study. A.B. and M.A.B. are the 16. Allen NH, Kerr D, Smythe PJ, Martin N, Osola 707–715 guarantors of this work and, as such, had full ac- K, Thompson C. Insulin sensitivity after photo- 31. Broussard JL, Wroblewski K, Kilkus JM, Tasali cess to all the data in the study and take therapy for seasonal affective disorder. Lancet E. Two nights of recovery sleep reverses the responsibility for the integrity of the data and 1992;339:1065–1066 effects of short-term sleep restriction on diabe- the accuracy of the data analysis. 17. Nieuwenhuis RF, Spooren PF, Tilanus JJ. tes risk. Diabetes Care 2016;39:e40–e41 Prior Presentation. This study was presented Less need for insulin, a surprising effect of 32. Zielinski MR, Kline CE, Kripke DF, Bogan RK, as a poster at the Society for Light Treatment phototherapy in insulin-dependent diabetes Youngstedt SD. No effect of 8-week time in bed and Biological Rhythms (SLTBR) 30th Annual mellitus. Tijdschr Psychiatr 2009;51:693–697 restriction on glucose tolerance in older long – Meeting, Groningen, the Netherlands, 21 24 [in Dutch] sleepers. J Sleep Res 2008;17:412–419 June 2018; and was presented at Chronotherapie 18. Brouwer A, van Raalte DH, Diamant M, et al. 33. Versteeg RI, Stenvers DJ, Visintainer D, et al. Netwerk Nederland (CNN) Annual Symposium, Light therapy for better mood and insulin Acute effects of morning light on plasma glucose Amsterdam, the Netherlands, 18 May 2018. sensitivity in patients with major depression and triglycerides in healthy men and men with and type 2 diabetes: a randomised, double- type 2 diabetes. J Biol Rhythms 2017;32:130–142 References blind, parallel-arm trial. BMC Psychiatry 2015; 34. Cheung IN, Zee PC, Shalman D, Malkani RG, 1. Roy T, Lloyd CE. Epidemiology of depression 15:169 Kang J, Reid KJ. Morning and evening blue- and diabetes: a systematic review. J Affect Disord 19. Terman JS, Terman M, Lo ES, Cooper TB. enriched light exposure alters metabolic function 2012;142(Suppl.):S8–S21 Circadian time of morning light administration in normal weight Adults. PLoS One 2016;11: 2. Snoek FJ, Bremmer MA, Hermanns N. Con- and therapeutic response in winter depression. e0155601 structs of depression and distress in diabetes: Arch Gen Psychiatry 2001;58:69–75 35. Opperhuizen AL, Stenvers DJ, Jansen RD, time for an appraisal. Lancet Diabetes Endocrinol 20. DeFronzo RA, Tobin JD, Andres R. Glucose Foppen E, Fliers E, Kalsbeek A. Light at night 2015;3:450–460 clamp technique: a method for quantifying acutely impairs glucose tolerance in a time-, 538 Light Therapy for Mood and Insulin Sensitivity Diabetes Care Volume 42, April 2019

intensity- and wavelength-dependent manner in and obese individuals. Obesity (Silver Spring) efﬁcacy of treatments and controls for depres- rats. Diabetologia 2017;60:1333–1343 2007;15:1749–1757 sion. PLoS One 2012;7:e41778 36. Danilenko KV, Mustaﬁna SV, Pechenkina EA. 38. Figueiro MG, Plitnick B, Rea MS. Light mod- 40. Virk G, Reeves G, Rosenthal NE, Sher L, Bright light for weight loss: results of a controlled ulates leptin and ghrelin in sleep-restricted Postolache TT. Short exposure to light treatment crossover trial. Obes Facts 2013;6:28–38 adults. Int J Endocrinol 2012;2012:530726 improves depression scores in patients with 37. Dunai A, Novak M, Chung SA, et al. Moderate 39. Khan A, Faucett J, Lichtenberg P, Kirsch I, seasonal affective disorder: a brief report. Int exercise and bright light treatment in overweight Brown WA. A systematic review of comparative J Disabil Hum Dev 2009;8:283–286