Effects of dynamic lighting on circadian phase, self-reported sleep and performance during a 45-day space analog mission with chronic variable sleep deficiency

Abstract

Spaceflight exposes crewmembers to circadian misalignment and sleep loss, which impair cognition and increase the risk of errors and accidents. We compared the effects of an experimental dynamic lighting schedule (DLS) with a standard static lighting schedule (SLS) on circadian phase, self-reported sleep and cognition during a 45-day simulated space mission. Sixteen participants [mean age (±SD) 37.4 ± 6.7 years; 5F; n=8/lighting condition] were studied in 4-person teams at the NASA Human Exploration Research Analog. Participants were scheduled to sleep 8 h/night on two weekend nights, 5 h/night on five weekday nights, repeated for six 7-day cycles, with scheduled waketime fixed at 7am. Compared to the SLS where illuminance and spectrum remained constant during wake (~4000K), DLS increased the illuminance and short-wavelength (blue) content of white light (~6000K) approximately 3-fold in the main workspace (Level 1), until 3 hours before bedtime when illuminance was reduced by ~96% and the blue content also reduced throughout (~4000K × 2 h, ~3000K × 1 h) until bedtime. The average (±SE) urinary 6-sulphatoxymelatonin acrophase time was significantly later in the SLS (6.22 ± 0.34 h) compared to the DLS (4.76 ± 0.53 h) and more variable in SLS compared to DLS (37.2 ± 3.6 min versus 28.2 ± 2.4 min, respectively, p=0.04). Compared to DLS, self-reported sleep was more frequently misaligned relative to circadian phase in SLS RR: 6.75, 95% CI 1.55–29.36, p=0.01), but neither self-reported sleep duration nor latency to sleep were different between lighting conditions. Accuracy in the abstract matching and matrix reasoning tests were significantly better in DLS compared to SLS (FDR-adjusted p≤0.04). Overall, DLS alleviated the drift in circadian phase typically observed in space analog studies and reduced the prevalence of self-reported sleep episodes occurring at an adverse circadian phase. Our results support incorporating DLS in future missions, which may facilitate appropriate circadian alignment and reduce the risk of sleep disruption.

INTRODUCTION

Spaceflight often exposes crews to unusual sleep-wake and work schedules that lead to misalignment of the circadian pacemaker, resulting in poor sleep, impaired alertness and increased risk of fatigue-related accidents. Untreated circadian misalignment results in sleep and wake occurring at the incorrect circadian phase, which reduces sleep quality and quantity and impairs alertness, reaction time and cognition (e.g., [1–4]). Even without circadian misalignment, sleep duration is usually poor during spaceflight, averaging only ~6 hours per night [5], and prescription hypnotics and caffeine are commonly used to address insomnia and daytime sleepiness, respectively [4–7]. In such a high-risk environment as space exploration, the risk of sleepiness-related performance decrements and accidents must be minimized.

Light is a potential countermeasure for both circadian misalignment and sleepiness associated with spaceflight [1, 8]. In addition to supporting vision, light exposure induces many additional ‘non-visual’ responses in humans including resetting the circadian pacemaker and improving alertness and performance [9]. These responses are primarily mediated by a novel photoreceptor system located in a small subset of intrinsically-photosensitive retinal ganglion cells (ipRGCs), which are functionally and anatomically distinct from the rod (night vision) and cone (color vision) photoreceptor systems. The cells contain the light-sensitive photopigment melanopsin, which is most sensitive to short-wavelength (blue) light (λ_max ~480 nm) [9]. Both short-wavelength monochromatic light and short-wavelength-enriched white appearing (polychromatic) light has been shown to preferentially reduce subjective sleepiness and improve performance under both laboratory (e.g., [10–14]) and real world conditions (e.g., [15–20]). Removal of short-wavelength light before sleep has the opposite effect, reducing the alerting response, preserving melatonin levels and avoiding circadian resetting (e.g., [21–24]). Additionally, blue light exposure induces larger phase shifts of the circadian clock compared to other wavelengths (e.g., [25, 26]). The timing and magnitude of these light-induced responses may be optimized to facilitate circadian adaptation, promote sleep and alertness at appropriate times and improve cognition.

A unique opportunity has arisen with the replacement of the current fluorescent lighting aboard the International Space Station (ISS). NASA has installed a new solid state lighting system with three pre-determined settings to address different operational needs: 1) white light for general vision; 2) blue-enriched white light to enhance high circadian adaptation and alertness; 3) blue-depleted white light to minimize alertness prior to sleep [8]. A Dynamic Lighting Schedule (DLS) has been developed, which determines when each of these three settings are used to optimize lighting to maintain visual function and as a countermeasure to facilitate circadian adaptation, improve alertness and performance and enhance sleep. The aim of the current study was to test the hypotheses that circadian phase alignment, sleep and cognition during a 45-day simulated space mission in the NASA Human Exploration Research Analog (HERA) would be better under a DLS as compared to a standard Static Lighting Schedule (SLS). This experiment was part of a series of ground-based laboratory and analog studies, and an ISS flight study, to assess the role of DLS to improve circadian alignment in preparation for long-duration space missions, most of which exceed the 45 days studied here [4, 27], and will become longer as manned missions beyond low-Earth orbit (e.g., to Mars) are developed.

MATERIALS AND METHODS

Environment

HERA is a three-story, four-port cylindrical habitat unit at Johnson Space Center which includes core, loft, simulated airlock and hygiene modules. This facility provides a high-fidelity simulation of the isolation, confinement, and remote conditions of space exploration. In addition to the physical simulation, HERA can also simulate communication methods in addition to ‘surveillance video and audio system, flight-like timeline and procedures similar to space missions (www.nasa.gov/analogs/hera).

Participants and protocol

Twenty astronaut-aged participants (4 per mission; mean age ± SD = 38.7 ± 8.2 years; 7 females) lived and worked in the HERA environment during five, 45-day missions completed between May 2017 and June 2018. HERA Campaign 4 anticipated 4 missions per year (one per quarter) with 4 participants per mission. Mission 2 in the original four-mission series was aborted on Mission Day (MD) 23 due to the approach of Hurricane Harvey and therefore Mission 5 was subsequently added. The lighting intervention was not implemented for Mission 3 until MD11, and so it was decided in consultation with NASA that Mission 5 would replace Mission 3 in the light randomization (DLS). Data from Mission 3 were excluded. This resulted in an analytic dataset from 16 astronaut-aged participants (8 per lighting condition; mean age ± SD = 37.4 ± 6.7 years; 5 females)

All participants underwent medical, psychological and behavioral screening, including assessment for color vision deficiency, prior to the study. Participant recruitment, screening and study oversight was provided by NASA personnel; further information is provided in [28]. For 14 days pre-mission, participants underwent mission training. All participants provided written, informed consent and the study was approved by the NASA Institutional Review Board (NASA IRB Pro2328) and Partners Human Research Committee (2017-P-000059).

The missions consisted of six 7-day cycles of chronic variable sleep deficiency with 2 nights of 8 hours of sleep (weekend; 23:00–07:00) followed by 5 nights of 5 hours of sleep, (weekday; 02:00–07:00) to end at the same time (Figure 1). Meals, work duties, exercise, cognitive testing, free time and sleep were scheduled for each day of the mission. Each participant had individual sleeping quarters. Caffeine use was only permitted from 7:45am to 2pm and limited to 2 cups. Napping outside of scheduled sleep episodes was prohibited on all mission days.

Figure 1.

Representative study protocol plotted in raster format. (A) participant 9332 from Mission 1 under the standard lighting schedule and (B) participant 8351 from Mission 5 under the dynamic lighting schedule. The black bars represent the scheduled sleep episode. The white bars represent wake under standard lighting. The blue, yellow and orange bars represent when the lighting was reduced in illuminance and short-wavelength content (6500K, 4100K and 2700K, respectively). The estimated aMT6s acrophase on scheduled sampling days are indicated by the red triangles and the interpolated daily acrophase of aMT6s is indicated by the red line. The timing of the cognitive battery is indicated by black circle marked with an x.

Lighting countermeasure

The initial mission was randomized to either the SLS or DLS. The order of the subsequent interventions was determined by NASA based on operational constraints. Missions 1 and 2 utilized the SLS and missions 3–5 utilized the DLS.

The SLS consisted of fixed spectrum standard lamps switched on during scheduled wake and off during scheduled sleep (see Supplemental Table 1 for details on lamp types and specifications). The DLS consisted of higher irradiance blue enriched white light during the day, changing to dimmer blue-reduced white light starting 3 hours before sleep with dim blue-depleted white light 1-hour pre-sleep (Figure 1; Table 1; Supplemental Table 1; Supplemental Figure 1).

Table 1.

Lighting specifications by location and condition

SLS			CCT K Mean ± SD	Irradiance W/m-2 Mean ± SD	Photon density log, cm-2, s Mean ± SD	Photopic lux Mean ± SD	S-cone-opic a-opic EDI (lux) Mean ± SD	M-cone-opic a-opic EDI (lux) Mean ± SD	L-cone-opic a-opic EDI (lux) Mean ± SD	Rhodopic a-opic EDI (lux) Mean ± SD	Melanopic a-opic EDI (lux) Mean ± SD	Melanopic DER Mean ± SD
Level	Height, plane	Location
Level 1	72”, horiz	1–6	4260 ± 355	1.47 ± 1.88	18.27 ± 0.60	467 ± 614	266 ± 333	409 ± 533	466 ± 612	316 ± 400	284 ± 354	0.71 ± 0.10
	72”, horiz	2–5	4427 ± 301	0.28 ± 0.10	17.89 ± 0.14	79 ± 19	52 ± 8	72 ± 16	79 ± 19	62 ± 12	60 ± 11	0.76 ± 0.04
	52”, vert × 4	1,2,3,5	4814 ± 754	0.57 ± 0.31	18.03 ± 0.24	122 ± 97	82 ± 37	110 ± 81	122 ± 97	94 ± 55	89 ± 46	0.77 ± 0.12
	72”, vert × 4	4,6	4423 ± 466	0.53 ± 0.34	18.08 ± 0.30	171 ± 124	104 ± 62	152 ± 106	170 ± 122	117 ± 69	106 ± 57	0.68 ± 0.14
Level 2	72”, horiz	7,8,10	3761 ± 110	0.36 ± 0.19	17.95 ± 0.25	108 ± 60	61 ± 31	93 ± 51	109 ± 60	69 ± 37	62 ± 33	0.58 ± 0.02
	52”, vert × 4	7,9	3791 ± 45	0.46 ± 0.01	18.08 ± 0.04	143 ± 2	80 ± 0	123 ± 2	142 ± 2	92 ± 1	82 ± 1	0.58 ± 0.00
	72”, vert × 4	8,10	3780 ± 57	0.39 ± 0.03	18.02 ± 0.02	120 ± 11	67 ± 6	103 ± 10	120 ± 11	77 ± 7	69 ± 6	0.57 ± 0.01
Bunk	Mid, vert	11	2721	0.8	18.38	259	66	201	263	124	101	0.39
DLS			CCT K Mean ± SD	Irradiance uW/m-2 Mean ± SD	Photon density log, cm-2, s Mean ± SD	Photopic lux Mean ± SD	S-cone-opic a-opic EDI (lux) Mean ± SD	M-cone-opic a-opic EDI (lux) Mean ± SD	L-cone-opic a-opic EDI (lux) Mean ± SD	Rhodopic a-opic EDI (lux) Mean ± SD	Melanopic a-opic EDI (lux) Mean ± SD	Melanopic DER Mean ± SD
Level	Height, plane	Location
Level 1-Day	72”, horiz	1–6	6352 ± 379	3.97 ± 2.80	18.93 ± 0.34	1210 ± 870	1150 ± 807	1180 ± 845	1198 ± 857	1099 ± 775	1079 ± 748	0.91 ± 0.10
	52”, vert × 4	1,2,3,5	6297 ± 453	0.87 ± 0.51	18.29 ± 0.26	264 ± 164	243 ± 131	257 ± 154	262 ± 161	239 ± 130	233 ± 119	0.92 ± 0.09
	72”, vert × 4	4,6	6221 ± 219	0.87 ± 0.36	18.34 ± 0.19	271 ± 123	247 ± 93	262 ± 116	267 ± 118	237 ± 89	229 ± 76	0.87 ± 0.11
Level 1 - Evening/Night	72”, horiz	2–5	2986 ± 131	0.13 ± 0.14	17.41 ± 0.40	44 ± 48	15 ± 15	35 ± 38	44 ± 49	25 ± 27	21 ± 22	0.48 ± 0.02
	52”, vert × 4	2–5	3204 ± 157	0.04 ± 0.01	17.01 ± 0.12	13 ± 2	8 ± 5	11 ± 2	13 ± 2	9 ± 2	8 ± 3	0.60 ± 0.13
	72”, vert × 4	4	3343	0.04	17.00	12	6	10	12	8	7	0.55
Level 2 - Day	72”, horiz	7,8,10	5405 ± 121	0.34 ± 0.13	17.95 ± 0.16	111 ± 44	92 ± 38	103 ± 40	108 ± 42	84 ± 32	76 ± 28	0.69 ± 0.01
	52”, vert × 4	7,9	5448 ± 39	0.25 ± 0.01	17.83 ± 0.02	80 ± 3	67 ± 2	75 ± 3	79 ± 3	63 ± 2	58 ± 2	0.72 ± 0.00
	72”, vert × 4	8,10	5421 ± 42	0.37 ± 0.04	17.98 ± 0.05	119 ± 14	97 ± 11	111 ± 13	116 ± 13	93 ± 10	86 ± 8	0.72 ± 0.01
Level 2 - Evening	72”, horiz	7,8,10	3551.00	0.01 ± 0.01	16.54 ± 0.31	4 ± 3	2 ± 1	4 ± 2	4 ± 3	3 ± 2	2 ± 1	0.59 ± 0.03
	52”, vert × 4	7,9	4430	0.01 ± 0.00	16.55 ± 0.04	4 ± 0	3 ± 0	3 ± 0	4 ± 0	3 ± 0	3 ± 0	0.79 ± 0.05
	72”, vert × 4	8,10	3570	0.01 ± 0.00	16.50 ± 0.12	3 ± 1	2 ± 0	3 ± 1	3 ± 1	2 ± 0	2 ± 0	0.60 ± 0.02
Level 2 - Night	72”, horiz	7,8,10		0.01 ± 0.00	16.41 ± 0.25	3 ± 1	2 ± 1	2 ± 1	3 ± 1	2 ± 1	2 ± 1	0.57 ± 0.09
	52”, vert × 4	7,9	4836	0.01 ± 0.00	16.40 ± 0.03	3 ± 0	3 ± 0	2 ± 0	3 ± 0	2 ± 0	2 ± 0	0.90 ± 0.06
	72”, vert × 4	8,10		0.01 ± 0.00	16.29 ± 0.08	2 ± 0	1 ± 0	2 ± 0	2 ± 0	1 ± 0	1 ± 0	0.54 ± 0.03
Bunk	Mid, vert	11	2540	0.20	17.77	63	12	47	64	27	21	0.34

Values were derived from the CIE S 026 α-opic Toolbox – v1.049 - 2020/03/26. SLS – Standard Lighting Schedule; DLS – Dynamic Lighting Schedule; EDI - Equivalent Daylight (D65) Illuminance; DER - Daylight Equivalent Ratio; CCT – Correlated Color Temperature; horiz – horizontal plane; vert – vertical plane; ×4 – refers to an average of measures taken in 4 directions in the vertical plane 90° apart.

^*Location 6 not measured.

n/a – not available. Direct comparisons between lighting conditions can be made using measures at the same level, height and plane, and locations (e.g., Daytime, Level 1 52”, vertical ×4, 1, 2, 3, 5 between SLS and DLS). Evening lighting started 3 hours before scheduled bedtime; Night lighting started 1 h before scheduled bedtime or overnight as a nightlight (see Supplemental Table 1).

Light illuminance (photopic lux) and spectra were measured before each condition across 11 locations using a calibrated spectrometer (Spectromaster C-7000, Sekonic, Tokyo, Japan), 6 on Level 1, 4 on Level 2 plus a representative crew sleeping bunk (all bunks were identical) (Supplemental Figure 1). Light was measured in the vertical plane at a height of 54” (4 measures 90 degrees apart, approximate eye level), and in the horizontal plane at a height of 72”, except in the bunk where a vertical measure was taken in the middle of the bunk, facing the downlight. Light measures were taken for the fixed light setting used in the SLS and for the three light settings used in the DLS, where appropriate (not all locations used all light settings and was based on when the space was used).

Measures of the Spectral Power Distribution (SPD) of the lights were used to calculate International Commission on Illumination (CIE) standard international (SI) units for ipRGC-influenced responses to light (see Footnote¹; [CIE, 2018]). including the melanopic Equivalent Daylight (D65) Illuminance (EDI), a measure of light quantity, and melanopic Daylight Equivalent Ratio (DER) which provides an overall indication of the relative melanopic strength of the spectra.

Table 1 shows the average illuminance and for each location and condition. For example, on average, the visual (photopic) illuminance at 72” in the horizontal plane on Level 1 (Locations 1–6) increased from 467 to 1210 lux during the day in the DLS but reduced to 44 lux in the evening. The DLS increased melanopic EDI nearly 4-fold during the day and reduced it by over 90% in the evening. The melanopic DER changed from 0.71 to 0.91 during the day and from 0.71 to 0.48 in the evening between the SLS and DLS conditions, respectively. On Level 2 (Locations 7–10), the photopic lux was similar during the day for both conditions (108 vs. 111 lux) but reduced from 111 lux to 4 lux in the evening during the DLS, with similar changes in the melanopic EDI (Table 1). The DER value for Level 2 increased from 0.58 for the SLS to 0.69 for the DLS during the day but was similar in the evening (0.55–0.59). The major change in the bunk lighting was a large reduction in illuminance (including for melanopic EDI) but there was no change in the light spectrum. Representative spectral power distributions are shown in Supplemental Figure 1.

Circadian phase

Circadian phase was assessed using urinary 6-sulphatoxymelatonin (aMT6s). Participants collected 4-hourly urine samples (5–8 hours overnight) for 48 hours twice pre-mission (starting 10 and 5 days before mission) and mid-week during each week of the mission (starting on mission days 5, 12, 19, 26, 33, 43). Sample collection time and volume were measured and a 5 mL sample was collected and stored (−20°C) until assay. Fifty-three urine samples out of 2231 (2.3%) were excluded from the analysis due to missing information (e.g., sample volume, timing) or values being outside of normative.

Urinary aMT6s concentrations were measured by radioimmunoassay using the methods of Aldhous and Arendt (1988). Antiserum was supplied by Stockgrand, Ltd., and assays conducted by Surrey Assays, University of Surrey (Guildford, UK). Urine samples were assayed in duplicate and the limit of detection was 0.2 ± 0.1 ng/mL. The intraassay coefficients of variation were 6.7%, 7.8%, and 6.1% at 3.3 ng/mL, 15.6 ng/mL and 28.3 ng/mL (n= 20 each), respectively. The interassay coefficients of variation were 9.4%, 4.8% and 6.7% at 2.9, 12.5, and 22.6 ng/mL (n=42 each), respectively. Cosinor regression analysis was used to determine the acrophase of the aMT6s rhythm using $y=m+A\left(\cos \left(\frac{2\pi \left(x-\phi \right)}{\tau }\right)\right)$, where, $m$ is mesor, $A$ is amplitude, $\mathrm{\tau }$ is period (24h), $t$ is time, and $\mathrm{\phi }$ is acrophase. Using the concentration of aMT6s and duration of the specific sample period, we estimated the amount of aMT6s per hour and plotted at the midpoint of the collection times.

Sleep

Daily electronic sleep diaries were completed by each participant every morning, post-sleep, including for 14 days pre-mission. Outcome variables included self-reported sleep duration, sleep onset latency (SOL). For each night of sleep data from each individual available, a phase angle difference in hours was calculated between the self-reported bedtime (reported time of lights switched off) and estimated aMT6s acrophase for that night for that individual.

Cognitive performance

Cognitive performance was assessed using the Cognition test battery for spaceflight [30] which was administered every other day in the evening (~19:00 h) on days with 5 hours of sleep (Figure 1). The battery consisted of 10 individual tests including 1) Motor Praxis test (MPT), 2) Visual Object Learning (VOLT); 3) Fractal 2-Back (N-BACK); 4) Abstract Matching (AM); 5) Line Orientation (LOT); 6) Emotional Recognition (ERT); 7) Matrix Reasoning (MRT); 8) Digit Symbol Substitution (DSST); 9) Balloon Analog Risk (BART); and 10) Psychomotor Vigilance (PVT, 3-minute duration). The cognitive domains assessed by these tests were sensory motor speed, spatial learning and memory, working memory, abstraction and concept formation, spatial orientation, emotion recognition, abstract reasoning, complex scanning and visual tracking, risk decision making, vigilant attention. Outcome measures for each of the 10 tests included metrics of response time and accuracy [30–32].

Surveys

Participants completed a weekly Light Affects and Acceptability Survey (LAAS) created by the investigators (GCB, SWL) but was not validated prior to use and therefore results should be interpreted with caution. This survey asked participants to rate the severity of several symptoms related to physical and emotional well-being, daytime sleepiness, and the overall acceptability of the lighting environment on work and other activities of daily living. Participants answered nine questions regarding the impact of the sleep schedule and 11 questions regarding the impact of the light on physical and mental outcomes using a 4-point Likert scale (0=absent, 3=severe) and nine questions on the impact of the lighting environment on work and other activities using a 5-point Likert scale (0=strongly disagree; 4=strongly agree) both for HERA Level 1 and Level 2. Participants also were de-briefed about the sleep schedule and lighting conditions post-mission.

Data analysis

Data are presented as mean ± SE unless otherwise noted. Data distribution was checked for normality using the Shapiro-Wilkinson test and visual inspection of QQ plots for regression analyses. Longitudinal data for circadian phase were analysed using linear mixed models (LMM) with lighting condition (DLS vs. SLS, as categorical) and mission days (mission days -15, -10, 5, 12, 19, 26, 33 and 43, as categorical) and their interaction as fixed effects and participant-level random effects. If a main effect of lighting condition was found, then post-hoc comparisons with the Tukey correction was carried out to assess differences between lighting conditions at scheduled sampling time-points. One out of the 146 48-h profiles analyzed could not be modelled using Cosinor regression (amplitude p = 0.06) and was removed from the analysis. Optimal timing of circadian phase was defined as the urinary aMT6s acrophase occurring between 2:45 am and 5:39 am (± 1.0 SD for the mean in sighted individuals) [33]. Daily estimated acrophases for each individual were dichotomized as being optimal or not based on this range and then summed to calculate a total count of optimal acrophases per participant. The number of optimal acrophases were compared between lighting conditions using generalized estimating equations (GEE) with a Poisson distribution and robust standard errors.

Self-reported sleep duration and latency to sleep were analyzed using generalized linear mixed models (GLMM) with lognormal and normal distributions, respectively, with lighting condition, week (as categorical), and their interaction as fixed effects and participant-level random effects. Analyses were stratified by weekday vs. weekend since the time-in-bed was different by design resulting in bimodal data distribution within each week. Additionally, the effect of lighting condition on these sleep outcomes was assessed stratified by whether the sleep episodes were aligned or misaligned relative to circadian phase. Models were not adjusted for the interaction between lighting condition and week for these two-level stratified analyses due to insufficient data in all required combinations.

Longitudinal cognition data were averaged within an individual across their mission for each of the 10 cognitive tests. Then compared between conditions using GLMM with a lognormal distribution, with lighting condition as the main effect. Statistical significance values were adjusted for α-inflation due to multiple comparisons using the false discovery rate (FDR) method [34]. Analyses were carried out using SAS 9.4 (SAS, Cary, NC, USA).

Responses to the LAAS were analyzed for the effects of lighting condition (DLS vs. SLS) and mission week using a cumulative link mixed model (function clmm, package “ordinal”, RStudio version 1.2.5033). For analysis, symptom responses to Q1 and Q2 were dichotomized into “absent” vs. “present” (slight, moderate, or severe) and responses to Q3 and Q4 were dichotomized into “disagree” (strongly disagree, somewhat disagree, or neutral) vs. “agree” (somewhat agree, strongly agree). Fisher’s exact test was applied to the responses to each question across all participants and all weeks by mission type (SLS vs. DLS) using the function fishertest in MATLAB (version R2015b; The MathWorks, Inc., Natick, MA, USA). No post-hoc comparisons were completed for the secondary outcome measures (sleep, cognition, and LAAS).

RESULTS

Circadian phase

Figure 1 shows examples of aMT6 output measured during the 48-h serial urine collection periods for two participants: one participant from Mission 1 under SLS (Figure 1A, B) and one participant from Mission 5 under DLS (Figure 1C, D). The average (±SE) pre-mission aMT6s acrophase times did not differ between the DLS (3.69 ± 0.42 h) and SLS (4.24 ± 0.33 h) (LMM F_1,14 = 1.03, p = 0.33, Figure 2A). The mean acrophase time was significantly later in the SLS (6.22 ± 0.34 h) compared to the DLS (4.76 ± 0.53 h) across the 45-day mission (LMM F_1,57 = 15.46, p = 0.0002, Cohen’s d = 1.88, Figure 2A). There was a trend for mission day (LMM F_5,57 = 2.33, p=0.054) and no interaction between lighting condition and mission day (LMM F_5,57 = 1.05, p=0.40) on acrophase times (Figure 2B). Urinary aMT6s acrophase for each individual participant appeared to be more variable in the SLS than in the DLS (Figure 2C). Group differences in acrophase variability was, therefore, estimated by comparing between lighting conditions the group-mean standard deviation in daily acrophase estimates. The variability in acrophases across the 45-day mission was significantly higher in the SLS (37.2 ± 3.6 min) than in the DLS (28.2 ± 2.4 min, t-test t(14)=2.21, p=0.04, Cohen’s d = 1.11). Amongst the 83 acrophase times estimated during the mission, the proportion of times that were optimal (i.e., between 2.75 h and 5.65 h [± 1.0 SD for the mean in sighted individuals] was higher in the DLS (45/48) compared to the SLS condition (13/35). The rate ratio (RR) of optimal urinary aMT6s acrophase was 3.5 times higher in the DLS compared to SLS [RR: 3.46 (95% CI: 1.43 – 8.36), GEE χ²=10.1, p=0.002].

Figure 2.

Effect of lighting schedule on circadian phase. Group average urinary 6-sulphatoxymelatonin (aMT6s) acrophase by lighting condition pre-, mid-, and end of mission (A) and across mission days (B). Data are expressed as the mean ± SEM. Urinary aMT6s acrophase times for each individual, separated by lighting condition (SLS or DLS) and mission (M) (C). Vertical dotted line (B, C) represents scheduled wake time. Vertical dashed lines (B, C) represent the interval within which the occurrence of aMT6s peak time was considered optimal (±1.0 SD from the mean in entrained individuals [33]). Collection started 15 days prior to the start of the mission (MD-15). Mission 2 was terminated on Mission Day = 22 due to Hurricane Harvey.

Self-reported sleep

Amongst the 588 sleep episodes analysed, the proportion of sleep episodes that were aligned with optimal aMT6s acrophase was higher in the DLS (93%, 318/342) compared to the SLS (34%, 84/246). Across the 45-day mission, the odds of having optimally aligned sleep was significantly higher in the DLS compared to the SLS [(RR: 6.75, 95% CI 1.55–29.36), GEE χ²=6.10, p=0.0135].

Average sleep duration on weeknights (with 5 hours of time-in-bed scheduled) in the DLS (301.28 ± 2.12 min) or the SLS conditions (290.88 ± 2.12 min) were not different (GLMM F_1,372 = 2.25, p=0.1341, Figure 3A). Sleep duration on weekend nights (8 hours of time-in-bed scheduled) was also not different between DLS and SLS (450.0 ± 6.0 min vs. 440.4 ± 4.8 min, respectively, GLMM F_1,160 = 0.51, p=0.48, Figure 3A). Average sleep onset latency (SOL) on weeknights in the DLS (3.77 ± 1.65 min) or the SLS conditions (7.80 ± 1.96 min) were not different (GLMM F_1,372 = 2.16, p=0.14, Figure 3B). SOL on weekend nights was also not different between DLS and SLS (8.97 ± 2.55 min vs. 14.01 ± 2.47 min, respectively, GLMM F_1,160 = 1.67, p=0.20, Figure 3B).

Figure 3.

Self-reported sleep duration and sleep onset latency across the ~6-week mission. Data are shown as mean ± SE. Number of sleep records analyzed per condition – Weekday: SLS = 168, DLS: 232; Weekend: SLS = 78, DLS: 110. SLS: Standard Lighting Schedule, DLS: Dynamic Lighting Schedule.

Cognition

Cognitive batteries were scheduled to be taken at clock time 19.00 h and were taken approximately at that time based on other operational demands. The mean (±SD) test times under the DLS (19.9 ± 1.1 h, range 18.5 – 21.5 h) was not different from SLS (19.9 ± 1.2 h, range 18.4 – 21.7 h) (t-test t(14)=−0.05, p=0.96). Mean (±SE) reaction time and accuracy for each cognitive test is shown in Supplemental Figure 2. Type III p-values from statistical tests comparing DLS and SLS for the different cognitive before and after FDR adjustment are presented in Supplemental Tables 2 and 3. Among the 10 cognitive tests, reaction time on the MP test trended to be faster in the DLS compared to the SLS (p=0.033, FDR-adjusted p=0.066, Cohen’s d = 1.15), but otherwise no statistically significant differences were observed between the lighting conditions for any of the other tests. In contrast, accuracy on the AM and MRT were significantly better in the DLS compared to SLS [p=0.020 and p=0.004 (FDR-adjusted p=0.039 and p=0.008), Cohen’s d = 1.33 and 1.79, respectively], and trended to be better in the MP test in the DLS compared to the SLS (p=0.078, FDR-adjusted p=0.078, Cohen’s d = 0.96), but otherwise no statistically significant differences were observed between the lighting conditions for any of the other tests.

Surveys

The results of the LASS questionnaires by mission type are reported in Supplemental Table 4. There were no significant differences between the missions in terms of physical or mental symptoms related to the sleep schedule or the lighting environment. On HERA Level 1, the light was significantly (p<0.001) more likely to be rated as uncomfortably dim, poorly distributed, and cause deep shadows in SLS compared to DLS. On HERA Level 2, the SLS light was significantly more likely to be rated as uncomfortably bright or causing problematic reflections compared to DLS, whereas it was significantly more likely to be rated as uncomfortably dim, and causing deep shadows in DLS compared to SLS.

DISCUSSION

Circadian disruption and sleep loss are inherent in spaceflight and space analog missions. A combination of exposure to atypical light-dark patterns, unusual day-lengths, night work and chronic sleep loss causes deterioration of alertness, performance safety and in long-term health. The HERA Campaign is considered a high-fidelity analog of space missions, mimicking the isolation and communication delays, team dynamics under stressful conditions, high workloads and schedules typical of spaceflight. While sleep has often been assessed during these missions, circadian phase assessment using validated measures has rarely been performed. In the current study, we assessed the timing of urinary 6-sulphatoxymelatonin (aMT6s) weekly, a gold-standard marker of circadian phase for field studies [35, 36]. Under typical lighting conditions, all participants showed evidence of circadian misalignment relative to their sleep-wake cycle, with more than half of their sleep episodes (and therefore wake episodes) occurring at an adverse circadian phase. In contrast, under the dynamic lighting condition, only 2 participants showed evidence of circadian misalignment with less than 10% of sleep and wake episodes occurring at the wrong circadian phase.

Light is the primary environmental time cue (or Zeitgeber) for resetting the circadian pacemaker [37]. In addition, light is a direct stimulant that can enhance subjective and objective markers of alertness [38]. The magnitude of these responses depends on the timing, illuminance, pattern, duration, and spectrum of light, as well as light history [39]. In the current study, we examined the effects of light spectra and illuminance on circadian entrainment by employing a DLS comparing it to a SLS typically used in spaceflight and analog missions. The DLS used multiple light sources from various manufacturers of different sizes and power positioned at different locations throughout HERA, some for general ambient lighting and others being task specific (see Supplemental Table 1). Collectively, the lighting was designed to enhance short-wavelength light to increase exposure to melanopic lux during the day or reduce exposure to melanopic lux in the evening. We sourced lamp replacements that met the lighting specifications required as closely as possible (Supplemental Table 1). Our results show that this approach did enhance exposure to melanopic lux during the day through a combination of both improved spectra and increase in illuminance, and decreased it in the evening primarily through reductions in overall illuminance (Table 1, Supplemental Figure 1). Whether the benefits observed during the DLS were due to changes in illuminance, spectra or both cannot be determined from the ‘living lab’ study environment. In the future, given the developments in LED technology, it should be possible to provide more controllable light sources throughout any environment to create larger changes in both illuminance and spectra between day and evening for analog and space habitats.

The DLS successfully improved entrainment of endogenous circadian phase relative to the sleep-wake cycle. Under the SLS, the average shift in the timing of the aMT6s peak was 1.46 h greater than under DLS. While the absolute differences in phase do not appear substantial, they are remarkable in that they occurred under highly restricted conditions with the same photoperiod in both groups, i.e., only the properties of light during the scheduled wake/light episodes changed, not the timing. Circadian phase varied less under the DLS and therefore could be considered more strongly entrained to the light-dark cycle. The definition of ‘entrainment’ quality requires further clarification, however. Circadian entrainment usually refers to the synchronization of an internal circadian rhythm with an environmental time cue, or T-cycle. In this study, the external T-cycle is 24 h as the scheduled sleep-wake and dark-light cycle have a 24-hour period. An internal rhythm could therefore be considered ‘entrained’ if it also exhibited a period of 24 hours. Both conditions therefore entrained the melatonin rhythm by this definition, as the periods of the rhythms were 24 hours. The alignment of the entrained phase relative to self-reported sleep (phase angle) varied, however. Phase angle refers to the relative timing of two defined phases, in this case the aMT6s peak and scheduled sleep end time. Given that the imposed sleep-wake (and therefore dark-light) cycle and fixed 7 am wake time limited the potential variability in phase angle, we considered an optimal phase angle one where the aMT6s peak time occurred between 2.75 and 5.65 h (±1.0 SD from the mean in entrained individuals [33]), and suboptimal outside of this range. This is a narrower definition than used previously; for example, during spaceflight [4], ‘in phase’ was defined as when the modelled core body temperature minimum (which occurs at a similar time to aMT6s peak) occurred within the sleep episode, and ‘out of phase’ outside this time. Even using this broader definition, however, (‘in phase’ defined as when the aMT6s peak occurred within the sleep episode), 26% of sleep episodes (65/246) during the SLS were misaligned despite the strict sleep schedule, similar to that found on ISS, but only 1 sleep episode (1/342) occurred during the DLS. In the DLS group, the timing of circadian phase varied less, demonstrating a more stable phase angle of entrainment with the fixed sleep-wake schedule (Figure 2).

We did not observe major changes in self-reported sleep in the current mission as, under the chronic variable sleep deficient schedule imposed, participants maximized the sleep opportunities and fell asleep relatively quickly, limiting the degree to which self-reported sleep duration and latency could differ between lighting conditions. Although HERA is a high-fidelity analog of spaceflight in many respects, the 8 h weekend sleep episodes do not reflect the flight experiences, as it is rare that individuals sleep nearly 8 hours in space. In an analysis of 21 astronauts over 925 mission days, only 16.5% of sleep episodes were reported to be 8 hours or longer given operational and personal demands [5], and were likely more intermittent than the multiple consecutive 8-hour nights scheduled weekly in the current mission [4]. Future analog studies assessing the impact of sleep on cognition should limit sleep opportunities to match the sleep obtained during spaceflight – on average only 6 hours per night, and a ‘real-world’ chronic variable schedule employed, given the exacerbated negative effects of such schedules compared to stable chronic sleep loss [40].

We also did not observe major effects of the light intervention on performance, measured once in the evening on selected days. Only the Abstract Matching (AM) and Matrix Reasoning (MRT) tests showed slightly higher accuracy during the DLS overall. The known acute effects of light (e.g., [13]) or prior light history (e.g., [41]) on cognitive performance could not be assessed given the sparsity of testing. The lack of difference may have been due to several factors including differences between individuals in the circadian time of testing, caffeine use (see below) or inter-individual differences in practice effects [30–32]. The tests were scheduled in the evening which meant that over half of the tests fell within the wake maintenance zone (WMZ), including 43.2% for the SLS and 62.6% for the DLS. The WMZ is a time where the circadian system is promoting alertness to offset the homeostatic increase in sleep propensity [42] and occurs during a ~3 hour window ending at melatonin onset (or ~7 h before the aMT6s peak; approximately 6–9pm, depending on individual circadian phase). During the WMZ, performance and alertness are temporarily elevated and testing at this time may therefore lead to an overestimation of overall alertness state, particularly if performance is only measured at a single timepoint and cannot be placed within the context of the daily time course and a known circadian phase. While there were insufficient data to test the influence of the WMZ, it is likely that the circadian phase of testing influenced performance and future missions should place the test later in the evening [32] or have multiple tests per day [28] to understand fully the degree of impairment. In addition, future missions should measure circadian phase as part of the NASA Human Factors and Behavioral Performance (HFBP) Element Standard Measures assessment to allow the performance results to be interpreted with respect to circadian phase.

Furthermore, additional studies are required specifically designed to elucidate the interaction between sleep restriction and circadian misalignment when sleep is chronically limited [43]. Prior studies with forced-desynchrony protocols with varying wake durations indicate that circadian misalignment impacts sleep and neurobehavioral performance under condition of both high and low homeostatic sleep pressure [44], and therefore the “override” of chronic sleep loss versus circadian misalignment on sleep observed here is not held up in more appropriate study designs. Comparable analyses are not possible in our current dataset due to the limited sample size, the study design with imposed sleep opportunities, the pseudo-naturalistic design of the study, which included caffeine consumption and exercise, and the limited range of combinations between circadian phase and wake duration, however.

The sanctioning of caffeine use in the mission likely also masked the impact of sleep loss and circadian phase on performance and safety. Caffeine is a powerful performance-enhancing stimulant that overrides much of the neurobehavioral degradation that would typically be observed with limited sleep and circadian misalignment [45], and effects sleep and metabolism (e.g., [46, 47]). While caffeine is available to astronauts aboard ISS (and until 2pm herein; 2 cups maximum), there are no data on what would happen to performance and safety during space missions if caffeine were not available, for example through a damage to stores or simply running out on a long-duration mission. Analog missions present an opportunity to measure the real impact of sleep loss and circadian misalignment without the masking effects of caffeine that would represent vital information to inform development of non-pharmacological countermeasures: We strongly encourage future analog missions to study this effect.

We did not objectively assess the impact of lighting conditions on visual acuity or color rendering in the current study. Self-reported assessment of lighting conditions impacting work (Supplemental Table 4) indicated that there was a trend (p=0.08) for reflections from the light fixtures hindering work more in DLS compared to the SLS on Level 1, whereas on Level 2, there was significantly less interference due to reflections from the light fixtures (p=0.01) and brightness on work (p<0.001) under DLS compared to SLS. Taken together, these results suggest that impact of lighting conditions on task performance may have differed based on tasks performed since different tasks were performed on the different levels of HERA and further direct objective assessments are needed.

Until recently, little specific consideration has been applied to lighting during space missions and analogs, with light of modest illuminance and non-specialized spectra being used. As is the case for most earth-based occupational settings, white fluorescent light is generally predominant with fixed spectra that are suboptimal in the short-wavelength part of the spectrum (i.e., with a CCT in the 3000–4000K range). The current study adds to a series of space analog studies where changing the lighting spectrum and/or illuminance have shown to be an effective countermeasure to improve circadian entrainment. Under operational conditions, we have shown that timed exposure to blue light was able to help synchronize flight controllers to a Martian ‘day’ (sol) of 24.65 hours during the Mars Phoenix Lander mission, as measured by aMT6s rhythms [1], and a combination of blue-enriched white light and exercise during breaks attenuate performance impairment in flight controllers working overnight at NASA Johnston Space Center Mission Control [48]. Furthermore, studies in Antarctica, considered a high-fidelity analog of space missions, have demonstrated benefits of blue-enriched white light on circadian phase, sleep, and performance [49, 50]. White-appearing light, rather than more limited spectra, is generally required in operational environments in order to support adequate vision and color rendering functions [51].

When adequate lighting interventions have not been implemented during space flight or analog missions, circadian misalignment is highly prevalent (e.g., [3, 4, 52]). Under the fluorescent lighting initially installed on the ISS, it is estimated that approximately 20% of sleep episodes occurred at an adverse circadian phase, reducing sleep by ~1 hour per night, which represents a substantial proportion relative to a mean of only 6 hours [4], and pushing crewmembers into a category of insufficient sleep that is of concern clinically (e.g., [53]) and has substantial safety and performance risks (e.g., [54, 55]). During the Mars 520-day mission, despite the 6 participants being scheduled to a 24-h earth sol, substantial disruption in sleep-wake timing was observed, including one participant who exhibited a non-24-hour rest-activity rhythm [3]. Similarly, in the Mars 105-day mission, participants exhibited clear misalignment of their aMT6s rhythms from the sleep-wake cycle, again despite being scheduled to a 24-hour day [2]. Entrainment to a Martian sol represents an even greater challenge [56, 57] and while possible with access to appropriately timed very bright light [58], more work is required to understand how to optimize the illuminance and spectra of the currently available lighting aboard ISS or analogs to facilitate entrainment to unusual day lengths. To this end, NASA has embarked on a lighting replacement program for ISS that exchanges the older General Luminaire Assemblies (GLA) that contain fluorescent lamps with a Solid State Lighting Assembly (SSLA) containing programmable Light Emitting Diode (LED) lamps [59]. Assessments of the impact on circadian phase, sleep and performance are ongoing in both analog ISS flight studies.

Importantly, sleep is one of the many physiological processes impacted by circadian alignment. Recent evidence indicates that circadian misalignment is associated with various adverse health impacts ranging from mood to metabolic dysfunction. These are of particular concern for long-duration space flight when disruption to circadian rhythms in metabolism, immune function, reproductive function, mood and many others may be of greater concern than sleep. Therefore, attenuating circadian misalignment is important independent of changes seen in sleep.

In summary, the DLS led to more stable entrainment of the circadian system, leading to a significant reduction in the number of self-reported sleep episodes occurring at an adverse circadian time and subsequently better self-reported sleep, albeit with a small effect size, during the weekday restricted sleep nights. Cognitive performance was also modestly better in the dynamic lighting condition although whether this was through the direct alerting effects of light or via improvements in sleep is not known. Multiple missions, including the current one, demonstrate that simply relying on standard lighting is insufficient to maintain appropriate circadian entrainment and results in impaired sleep and performance. Proactive implementation of DLS is warranted in all future space missions.

Data Availability.

The datasets generated and analysed during the current study are available by request from the NASA Life Sciences Data Archive: https://lsda.jsc.nasa.gov. The LAAS will be made available to investigators upon request.