The impact of bright artificial white and ‘blue-enriched’ light on sleep and circadian phase during the polar winter

Authors


Josephine Arendt, Centre for Chronobiology, Faculty of Health and Medical Sciences, University of Surrey, Guildford, Surrey GU2 7XH, UK. Tel.: 01483689712; fax: 01483689712; e-mail: j.arendt@surrey.ac.uk or (preferred) arendtjo@aol.com

Summary

Delayed sleep phase (and sometimes free-run) is common in the Antarctic winter (no natural sunlight) and optimizing the artificial light conditions is desirable. This project evaluated sleep when using 17 000 K blue-enriched lamps compared with standard white lamps (5000 K) for personal and communal illumination. Base personnel, 10 males, five females, 32.5 ± 8 years took part in the study. From 24 March to 21 September 2006 light exposure alternated between 4–5-week periods of standard white (5000 K) and blue-enriched lamps (17 000 K), with a 3-week control before and after extra light. Sleep and light exposure were assessed by actigraphy and sleep diaries. General health (RAND 36-item questionnaire) and circadian phase (urinary 6-sulphatoxymelatonin rhythm) were evaluated at the end of each light condition. Direct comparison (rmanova) of blue-enriched light with white light showed that sleep onset was earlier by 19 min (= 0.022), and sleep latency tended to be shorter by 4 min (= 0.065) with blue-enriched light. Analysing all light conditions, control, blue and white, again provided evidence for greater efficiency of blue-enriched light compared with white (< 0.05), but with the best sleep timing, duration, efficiency and quality in control natural light conditions. Circadian phase was earlier on average in midwinter blue compared with midwinter white light by 45 min (< 0.05). Light condition had no influence on general health. We conclude that the use of blue-enriched light had some beneficial effects, notably earlier sleep, compared with standard white light during the polar winter.

Introduction

The lack of natural sunlight during the Polar winter enables long-term experiments with artificial light during the winter months without the confounding influence of exposure to natural light, a situation very rarely encountered elsewhere. Moreover, Antarctic base personnel provide exceptional material for human studies, being robust, healthy and young, living in the same environment for long periods of time and in general eating the same food. Thus, many of the factors that render clinical research difficult are eliminated. The major disadvantage is the small number of volunteers available.

Some problems reported from Antarctic and other polar regions include delayed circadian phase, delayed sleep timing and sometimes free-run during the period when the sun is below the horizon (sundown) in winter (Broadway et al., 1987; Hansen et al., 1987; Kennaway and Van Dorp, 1991; Lingjaerde et al., 1985; Palinkas et al., 1996; Yoneyama et al., 1999). Early work has shown clearly that provision of extra bright white light as a skeleton photoperiod, during the winter, can restore summer circadian phase position (Broadway et al., 1987). In 2003 at Halley base, Coat’s Land, Antarctica (75°S) it was possible to increase environmental light intensity compared with 2002 using extra light equipment, a gift from Philips Bright Light Devices (Philips Lighting B.V., Eindhoven, The Netherlands). Personal light exposure (lux 24 h−1, mean ± SD) measured using ActiwatchL (AWL) monitors (Cambridge Neurotechnology Ltd, Cambridge, UK) increased from 30 ± 11 to 64 ± 21 lux (average) and 572 ± 276 to 1039 ± 281 lux (maximum), but winter delays in sleep timing persisted (Francis et al., 2008). In 2003, sleep efficiency was poor but slightly increased (72.9.0 ± 4.16% to 74.06 ± 3.38%, < 0.05) using blue-enriched white light (prototype lamps, 10 000 K) compared with standard white light (5000 K), with comparable personal light exposure measured as lux (Francis et al., 2008). We concluded that increased personal light exposure should be investigated for more robust effects.

The ongoing building of a new Base, with a requirement for lighting specifications, and the recent availability of 17 000 K blue-enriched white light equipment (ActiViva Active, Philips Bright Light Devices) prompted a further study of sleep and circadian phase at Halley. This project has compared sleep (by actigraphy) during sequential 4–5-week periods of increased exposure to standard white light followed by 4–5-week periods of ActiViva blue-enriched light, from the autumn to the spring equinoxes in 2006, with 3-week control periods (no light treatment) before and after the treatment sequence.

Materials and Methods

The protocol for this study was essentially the same as that used in a previous study (Francis et al., 2008), and will be summarized here.

Subjects, samples and data collection

Ethical permission for this study was given by the University of Surrey Advisory Committee on Ethics. All participants gave informed consent and the Base doctor attested to their fitness to take part. Personnel over-wintering at Halley Bay in 2006 (10 men and five women aged 32.5 ± 8 years, mean ± SD) took part in the study. Subjects (= 12) wore wrist activity and light monitors (AWL, recording in 30-s epochs, Cambridge Neurotechnology Ltd) continuously, except when showering. The monitors were worn on the non-dominant wrist outside the clothing, when indoors, and under the clothing when outdoors, from 3 March to 12 October. All subjects (= 15) kept sleep diaries daily (bedtime, desired sleep start time, latency to sleep, number and duration of wake ups during sleep, wake up time, get up time, sleep quality and alertness; Visual Analogue Scales, VAS).

At the end of each light condition, C1, W1, B1, W2, B2, W3, B3, C2 (Table 1) the participants completed the RAND SF-36 health questionnaire (Hays et al., 1993). Only data relating to the specific days (i.e. last day of each light condition, including the control light periods) were used for analysis. Ten variables were modelled: each of the individual scores and the PCS (combined physical health score) and MCS (combined mental health score). The PCS and MCS were derived using the population norms and factor score coefficients from the Oxford Healthy Life Survey [OHLS, Health Services Research Unit (HSRU), University of Oxford].

Table 1.   Light treatment
Light treatment datesConditionNameAverage lux per 24 h ± SDMaximum lux per 24 h ± SD
  1. Lux is given as the average or maximum of all 30-s epochs.

  2. From 11 May to 11 August, the sun is below the horizon.

3 March–23 MarchControl 1 no extra lightC172 ± 413697 ± 1637
24 March–20 April5000 K standard white lightW150 ± 152206 ± 746
21 April–18 May17 000 K blue-enriched lightB140 ± 151812 ± 652
19 May–16 June5000 K standard white lightW254 ± 191631 ± 487
17 June–13 July17 000 K blue-enriched lightB266 ± 252068 ± 485
14 July–17 August5000 K standard white lightW352 ± 171840 ± 727
18 August–21 September17 000 K blue-enriched lightB342 ± 152235 ± 1152
22 September–12 OctoberControl 2 no extra lightC251 ± 174094 ± 2309

The RAND assesses general health and provides eight composite variables: energy/fatigue; general health; physical functioning; social functioning; role limitations caused by physical and emotional problems; emotional wellbeing; and pain. It is sensitive to changes in health among general populations (Hemingway et al., 1997). Subjects also completed the Horne–Ostberg questionnaire (Horne and Ostberg, 1976) for diurnal preference three times, during C1, W2 and C2.

In addition, at the end of each light condition they collected sequential urine samples for 48 h, every 3–4 h (longer over the sleep period) for assessment of circadian phase using the urinary metabolite of melatonin, 6-sulphatoxymelatonin (aMT6s). The total volume of each urine sample was measured and an aliquot frozen for transport to the University of Surrey for analysis of aMT6s by RIA (Aldhous and Arendt, 1988).

Light treatment

Each subject was provided with a light box (Philips Bright Light Devices, HF3305) for personal use in their bedrooms. The light boxes were placed within 1.5 m of the bed. The subjects were requested to use the lights in the morning on wake up and to avoid bright light in the evening. A further 23 light boxes were placed in the communal areas together with the workshops and laboratories. They were turned on in the morning (approximately 08:00 hours) and off in the evening (from approximately 18:00 hours). The same light boxes were used for both types of light exposure, by changing the bulbs at the specified intervals. From 24 March to 21 September 2006 light exposure alternated between 4–5-week periods of standard white (5000 K) and blue-enriched light ActiViva Active lamps, 17 000 K (Philips Lighting B.V.,). The light treatment sequence is shown in Table 1, together with the group average and maximum lux per 30-s epoch measured for each condition.

Actigraphy data analysis

The AWL data were edited to remove the week of night shift and the two subsequent weeks (to be analysed separately), days missing more than 1 h of data, and the period 16 June to 24 June 2006 (solstice week) in view of abnormal activity. Sleep parameters were derived from actigraphy using subjects’ sleep logs and the manufacturer’s software to give bedtime, sleep start time (sleep onset), sleep latency, actual sleep time, sleep efficiency, fragmentation index and sleep end time (wake up). Sleep diary data only (sleep start time, wake up and sleep quality by VAS) were used for the three subjects for whom AWLs were not available. Individual means were derived for each sleep parameter and light exposure for each light condition. Group means were derived for light exposure during the designated period of sundown (11 May–11 August).

For the purposes of comparison with previous years’ light exposure, previously published data from 2002 and 2003 were used (Francis et al., 2008). Average and maximum daily light exposures, during the period of sundown (11 May–11 August), were compared by one-way anova (factor year) with Tukey post hoc tests (Instat).

For daily sleep parameters during the control and light treatment periods, modelling was performed using rmanova (SAS version 9.1; SAS Institute, Cary, NC, USA). Initially a direct comparison of white versus blue light periods was made, excluding control periods, the factors being light type, sundown (11 May–11 August) and observation day number (to allow for acclimatization). Then a full model was applied including control periods, with factors control, white light, blue light, sundown, observation day number, gender, age, maximum lux and average lux.

Circadian phase by aMT6s

The subjects’ circadian status was assessed via the urinary melatonin metabolite aMT6s, measured by radioimmunoassay (Aldhous and Arendt, 1988) the intra-assay coefficients of variation at 4.4, 13.8 and 26.5 ng mL−1 were 9.8, 8.1 and 7.8%, inter-assay coefficients of variation were 12.5, 11.8 and 11.6%. All samples from one individual were measured in the same assay. The aMT6s acrophase time (phi) was calculated by cosinor analysis (Dr D. S. Minors, University of Manchester, UK) using 48-h windows. aMT6s data with a non-significant (> 0.05) or <50% fit to the cosine curve were rejected. As for sleep, data from periods of night shift and for 2 weeks post-night shift were not used in this analysis.

Data are expressed as means ± SD unless otherwise stated.

Results

Compliance

Actigraphy and sleep diaries

Compliance was good, 15 of 16 Base personnel took part, with the average number of days’ AWL data from each subject being 164 ± 14 (88% of theoretical maximum).

Urine sampling

Urine collection was optional. Three subjects did not take part, two subjects’ data provided no, or very few, significant cosinor fits, two subjects provided very few and/or mislabelled samples. Eight subjects provided one or more complete collections with sufficient significant cosinor fits for each of the midwinter light treatment periods, B1, B2, and W2, W3.

RAND questionnaire

Two subjects provided no questionnaire data, the remaining 13 subjects provided 89.5% of the theoretical maximum. Data were missing for various reasons, including day sleeps after night shift and periods away from the Base on field trips after sun-up.

Light exposure

Average and maximum light exposure levels during sundown at Halley for the years 2002 (no extra light), 2003 (alternating periods of standard white, 5000 K, and prototype blue-enriched light, 10 000 K) and 2006 (alternating periods of standard white, 5000 K, and blue-enriched, ActiViva light, 17 000 K) are shown in Table 2. Maximum light exposure was increased during sundown in 2006 (11 May–11 August 2006) compared with 2003 and 2002; however, average light exposure showed no change during sundown compared with 2003 or 2002.

Table 2.   Personal light exposure in lux: average or maximum of all 30-s epochs (2006) or 60-s epochs (2003, 2002) during sundown (11 May–11 August) for 2002 (no extra light), 2003 (5000 K standard white and 10 000 K blue-enriched) and 2006 (5000 K standard white and 17 000 K blue-enriched light)
Year200620032002Main effect of yearTukey post hoccomparisons
  1. 2002 and 2003 data are from Francis et al. (2008). = days.

Maximum light exposure per 24 h lux ± SD1997 ± 9221039 ± 478572 ± 306< 0.0012006 > 2003, < 0.01
2006 > 2002, < 0.001
Average light exposure per 24 h lux ± SD53 ± 2864 ± 4230 ± 18NS 
No. subjects12107  
Average N per subject738578  

Recorded light exposure during periods B1, W2, B2 and W3 was very similar (Table 1). However, the maximum but not average exposure to blue light just after the winter solstice, 28 June–11 July 2006, was slightly but significantly higher (maximum 2168 ± 405 lux, average 68.7 ± 24 lux) than the equivalent white period just before the winter solstice, 2 June–15 June 2006 (maximum 1699 ± 347 lux, average 58.4 ± 18 lux; < 0.01, maximum, average N per subject = 10). Most subjects showed two or three peaks of light exposure (analysed on an hourly basis) during the day, associated with mealtimes.

Sleep

Relationship to light condition

The values obtained for the different sleep parameters during each light condition are shown in Fig. 1 and Table 3, with the significant main effects and post hoc comparisons given in Tables 4 and 5.

Figure 1.

 Sleep characteristics derived from actigraphy during the different light conditions, mean + SD. The sun was below the horizon for all of W2 and B2, and part of B1 and W3, these values are plotted separately as B1a, B1b, W3a and W3b.

Table 3.   Sleep characteristics from 3 March to 12 October
Dates3 March–23 March 200624 March–20 April 200621 April–18 May 200619 May–16 June 200617 June–13 July 200614 July–17 August 200618 August–21 September 200622 September–12 October 2006
Treatment periodC1W1B1W2B2W3B3C2
  1. The sun is below the horizon from 11 May to 11 August. See Tables 4 and 5 for significant differences between treatment periods. Sleep data are given as mean (SD).

  2. C = no extra light, W = standard white light, B = ActiViva blue-enriched light.

No. subjects (minimum)13 (11)15 (12)15 (12)15 (12)14 (11)15 (12)15 (12)14 (11)
Actual sleep time, decimal h6.92 (0.61)6.58 (0.58)6.55 (0.56)6.62 (0.63)6.34 (0.87)6.15 (0.90)6.59 (0.67)6.77 (0.49)
Fragmentation index31.12 (7.91)33.34 (8.71)32.51 (10.53)31.51 (8.84)32.09 (12.64)31.39 (9.44)32.45 (9.28)30.26 (8.25)
Sleep efficiency %82.33 (5.10)81.20 (5.50)80.45 (5.13)78.77 (5.63)78.09 (7.18)78.03 (6.54)78.50 (5.08)79.16 (5.19)
Sleep end, decimal h8.48 (0.70)8.50 (0.89)8.60 (0.89)8.98 (1.03)8.77 (0.98)8.59 (0.74)8.64 (0.85)8.52 (0.89)
Sleep latency, min18.21 (10.91)19.00 (12.84)22.33 (15.94)31.30 (22.35)33.00 (29.85)36.95 (23.12)27.19 (19.29)24.32 (14.71)
Sleep start, decimal h1.09 (0.81)1.23 (0.89)1.29 (0.88)1.73 (1.14)1.70 (1.24)1.66 (1.11)1.25 (1.08)0.92 (1.02)
Sleep quality VAS58.63 (10.75)58.72 (10.36)57.37 (15.90)59.28 (15.14)61.97 (19.46)59.60 (19.96)62.20 (19.16)64.91 (15.97)
Alertness VAS54.89 (12.39)55.64 (12.75)54.14 (16.99)56.92 (18.23)60.27 (20.35)58.29 (20.61)62.94 (19.84)64.06 (19.22)
Table 4.   Sleep: statistical differences and trends by condition
Sleep variableWhite mean (SD)Blue mean (SD)Main effectDifference (blue − white)Comment
  1. Direct comparison of blue light with standard white, excluding control. Factors: light type, sundown, observation day number.

Efficiency %78.84 (10.83)78.86 (9.76)NS0.02Lower over time
Sleep start, decimal h1.65 (1.63)1.34 (1.54)= 0.02218.6 minBlue earlier, later with sundown
Sleep end, decimal h8.72 (1.81)8.65 (1.63)NS−4.2 minLater with sundown
Sleep latency, min31.20 (42.43)27.30 (37.97)= 0.0653.9 minBlue shorter, longer with time and sundown
Fragmentation index32.17 (14.86)32.25 (15.17)NS0.08 
Actual sleep time, decimal h6.40 (1.46)6.52 (1.38)NS7.2 minBlue longer, shorter with time
Sleep quality (VAS)58.97 (21.65)60.57 (20.50)NS1.6Higher with time
Alertness (VAS)56.65 (23.41)59.42 (22.87)NS2.8Higher with time
Table 5.   Sleep: statistical differences and trends by condition
Sleep variableControl mean (SD)White mean (SD)Blue mean (SD)Main effectPost hoc comparisons
  1. Full model factors: control, white light, blue light, sundown, gender, age, maximum lux, average lux and observation day number. Eight observation days, 1, 8, 22, 50, 78, 134, 169, 204 omitted due to missing maximum and average lux values.

Efficiency %81.38 (8.61)78.78 (10.90)78.84 (9.80)0.036C > W, B
Sleep start, decimal h0.94 (1.27)1.65 (1.63)1.33 (1.53)0.034W > B > C
Sleep end, decimal h8.44 (1.22)8.69 (1.80)8.63 (1.61)NS 
Sleep latency, min19.45 (21.34)31.47 (42.94)27.29 (37.96)NS 
Fragmentation index30.20 (12.18)32.20 (14.94)32.34 (15.31)NS 
Actual sleep time, decimal h6.88 (1.13)6.37 (1.43)6.51 (1.36)0.0103C > W, B
Sleep quality (VAS)61.90 (18.31)58.93 (21.65)60.73 (20.45)0.042C > B > W
Alertness (VAS)59.28 (20.80)56.52 (23.50)59.62 (22.84)NS 

Direct comparison of blue-enriched and white light conditions

The direct comparison of blue-enriched light with standard white light excluding control periods elicited the following differences: for sleep onset the participants slept significantly earlier during blue light periods (= 0.022); sleep latency tended to be shorter with blue light (= 0.065). Sleep end time, actual sleep time, sleep efficiency, fragmentation, sleep quality and alertness were not significantly different between conditions.

Comparison of control, white and blue-enriched light conditions using the full model

Sleep efficiency.  Sleep efficiency was reasonably good in all conditions, with no differences between conditions. Overall efficiency was 79.36 ± 5.3% for all white periods, 79.02 ± 4.86% for all blue periods, and 80.90 ± 4.89% for control periods. There were no differences between blue-enriched and white light conditions.

Sleep start time.  There was a delay in sleep start time in midwinter, compared with control, this being greater for white light than for blue (18.6 min, = 0.034). The maximum delays from the autumn equinox to midwinter were non-significantly reduced compared with the results obtained in 2003 (66 ± 55 min, 2003; 38 ± 47 min, 2006).

Sleep end (wake up).  There was no significant delay in wake up during sundown during either white or blue light compared with control. The maximum delay from the autumn equinox to midwinter (from C1 to W2) was reduced compared with the results obtained in 2003 (Francis et al., 2008): 70 ± 58 min, 2006: 30 ± 33 min, < 0.05 (unpaired Student’s t-test).

Latency.  Latency to sleep onset was slightly but not significantly longer with white but not blue light compared with the spring control period.

Sleep fragmentation.  There were no significant differences between light conditions for fragmentation.

Actual sleep time.  A main effect of light type on actual sleep time indicated that sleep was longest during control conditions, compared with blue or white light. During blue light sleep was 7.2 min longer than during white light, but this difference did not reach significance.

Sleep quality and alertness (VAS).  Subjective sleep quality was highest in control conditions, followed by blue then white light, albeit with very small differences. There were no differences in alertness using the full model.

Relationship to light intensity

Sleep parameters were evaluated as a function of light intensity by linear regression (Instat) using the group mean values for each condition. Wake up and sleep start time were earlier with increasing maximum light values (< 0.05). Wake up versus maximum lux, r2 = 0.56, = 7.76, = 0.031, sleep start versus maximum lux, r2 = 0.64, = 10.76, = 0.017.

Questionnaires

RAND questionnaire

The only significant difference in RAND questionnaire data using the full model was that of gender: women had lower aspects of physical (PCS, = 0.029) and mental health (MCS, = 0.001).

Horne–Ostberg questionnaire

The group average Horne–Ostberg score was 48.1 ± 11 (mean ± SD). One subject on one occasion provided a score that was classified as morning preference. All other scores were within the intermediate range. There were no differences in score between the different conditions.

Circadian phase: acrophase (phi) of aMT6s

Large differences in individual acrophase were evident (range 4.10–8.37 h). No difference by anova was found between control, white and blue light (aMT6s phi, all Control 5.76 ± 0.83 h, = 8; all White 5.84 ± 1.17 h, = 18; all Blue 5.61 ± 1.16 h, = 18; mean ± SD). However, restricting the comparison to paired average values from each subject during midwinter periods (B1, B2, and W2, W3) indicated that aMT6s phi was later during white light periods (6.18 ± 1.39 h) than blue (5.43 ± 1.30 h; = 8, = 0.024, two-tailed, paired Student’s t-test).

Discussion

Polar regions provide a useful environment for optimizing the intensity and spectral composition of ambient light. It is unusual to have no access to natural light for extended periods. However, in many circumstances, e.g. high latitudes in winter, short photoperiods together with an interior workplace are associated with a lack of bright natural light. In midwinter in the UK for example (London, 52°N) travel to and from work will often be in artificial light, since from 4 December to 7 January the sun rises after 08:00 hours and sets before 16:00 hours. Workstations may well have no windows, and outdoor activity at lunchtime or during daylight hours may be difficult if not undesirable in poor weather conditions. Thus, data from Polar regions are likely to have applicability in urban environments with winter daylengths shorter than the working day.

Night time sleep and daytime alertness and performance are optimized when appropriate circadian synchrony is maintained with the 24-h day (Akerstedt and Gillberg, 1982). Light is the most important influence on the circadian system (Broadway et al., 1987; Czeisler, 1995), short wavelengths having the most powerful effects on suppression of melatonin, phase shifting and alertness in strictly controlled experimental conditions (Brainard et al., 2001; Lockley et al., 2003; Revell et al., 2005, 2006; Thapan et al., 2001). However, in two further studies blue-enriched polychromatic light (17 000 K, 4000 lux) could not produce larger phase delays or advances than white light (4100 K, 5000 lux) of equal photon density (4.2 × 1015 photons cm−2 s−1; Smith and Eastman, 2009; Smith et al., 2009); neither were 17 000 K lamps superior to 5000 K lamps, both at 10 000 lux, in the treatment of winter depression (Gordijn et al., 2006). The controversy has been summarized by Terman (2009), whereby he questions whether the use of ‘blue-enriched’ light is really a major advance.

There is nevertheless recent evidence that blue-enriched light (ActiViva) in the workplace can improve sleep and performance compared with standard white light (Viola et al., 2008), that bluish (6500 K, about 1800 lux) compared with yellowish (2700 K, about 1800 lux) light may have beneficial effects in dementia (van Hoof et al., 2009), and that in some circumstances blue-appearing light (98 lux) may be more efficient than white light enriched with equivalent short-wavelength photons but with sevenfold more lux (700 lux) in the treatment of depression (Anderson et al., 2009). Our previous data (Francis et al., 2008) suggested that ambient blue-enriched white light (10 000 K) might have small objective benefits for sleep during the Polar winter compared with standard white (5000 K) at comparable lux levels (about 1000 lux maximum per 24 h on average for both light conditions). We hypothesized that further increasing the intensity of blue-enriched light would lead to more robust beneficial effects, particularly with reference to reducing or eliminating the delay in winter sleep timing.

The decrements in sleep during the polar winter are evident from the present data, with control sleep being more efficient, starting earlier, lasting longer and being of subjectively better quality. Improvements were seen with extra light, albeit small, but blue-enriched light was superior to white when significant effects were observed. It is of interest in the current study, with maximum exposure increased to about 2000 lux in both light conditions, that the previously observed delay in midwinter wake up time compared with the autumn was abolished by both blue and white light. A strong adherence to the scheduled normal work time by Antarctic base personnel may also have contributed to this observation. The delay in sleep start time was reduced compared with previous data with a clear advantage of blue light. This may relate to the earlier circadian phase with blue light. A slightly higher light intensity (lux) was registered with blue-enriched compared with white light during midwinter, and this may have influenced responses.

The lack of impact on sleep efficiency, with no differences between the blue and white conditions, may relate to the generally good sleep observed during 2006, leaving little room for improvement. In 2003 sleep efficiency was poor overall (73% efficient compared with 80% in the present study). There are large individual differences in sleep characteristics, and conceivably the participants in 2006 were simply better sleepers and/or the conditions on the base (e.g. weather, psychological) were more propitious than those in 2003.

Using the group mean values for each light condition, earlier sleep was associated with maximum (but not average) light intensity irrespective of spectral characteristics: the higher the light exposure the earlier the sleep timing. This suggests that intensity is of primary importance provided that spectral characteristics are suitable, and that brief daily periods of very bright artificial light may be as efficacious as a full daylength of dimmer light. Timing light exposure according to circadian phase to minimize delays in winter would be ideal but, given the range of individual aMT6s acrophases, timing and duration would need to be carefully optimized.

From the point of view of Polar bases and other extreme situations such as space flight, a saving in fuel related to the use of brief light pulses is of considerable interest. With the judicious manipulation of spectral quality and intensity of ambient light it may well be possible to reduce fuel use.

Analysis of the RAND SF-36 health questionnaire provided little evidence of differences between blue-enriched light and standard white light, or indeed compared with control data in these young and healthy participants.

Conclusions

We present evidence that the use of blue-enriched light (17 000 K, ActiViva) in personal light boxes and as extra light in communal areas on an Antarctic base during the winter has small beneficial effects with respect to sleep timing and sleep quality compared with standard white light of comparable lux. Particularly noteworthy is that the delay in wake up over winter was largely abolished by extra light of either spectral composition. An increase in the maximum amount of light exposure per 24 h, irrespective of spectral composition, is probably the most important consideration. Larger numbers of subjects are needed to confirm these observations: the use of multiple Antarctic bases or comparable Arctic environments may address these issues.

Acknowledgements

This study was supported by the British Antarctic Survey via the British Antarctic Survey Medical Unit, Stockgrand Ltd, UK, and Philips Lighting B.V., Eindhoven, The Netherlands (unrestricted research grant).

Ancillary