A systematic review and meta-analysis on light therapy for sleep disorders in shift workers

Abstract

Sleep disorders constitute a significant disruption for shift workers. Beyond medical interventions, phototherapy is recognized as an effective approach to significantly alleviate sleep disorders, particularly among individuals engaged in shift work. However, the effective dose and efficacy evaluation of phototherapy have not yet been determined. This study conducted a systematic review across five databases from January 1, 1990, to December 31, 2023. A total of 11 articles were selected for meta-analysis using a random-effects model. The results showed that light therapy significantly improved the total sleep time (TST) (MD = 32.54, p < 0.00001) and sleep efficiency (SE) (MD = 2.91, p = 0.007) of shift workers compared to the control group. Subgroup analysis and regression analysis implied that medium illuminance (900–6000 lx) for a long treatment duration (≥ 1 h) during night was more effective in extending total sleep time, whereas higher-illuminance and increasing dose (lx*h) of light therapy was more beneficial for SE. In summary, light therapy has a degree of efficacy in increasing the overall sleep duration and efficiency for shift workers, the findings of the current study contribute reference and evidence for dose setting and experimental design of phototherapy on shift workers’ sleep in clinical and research.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-83789-3.

Introduction

The rapid pace of modern society and specific job requirements mean certain sectors (like policing¹, nursing², oil³, firefighting⁴ and internet services⁵) necessitate round-the-clock shifts. According to statistics, 20% of the global workforce is on some form of shift schedule, including evening, night, rotating, and erratic shifts⁶. Such patterns disrupt natural sleep-wake cycle and circadian rhythm, causing a mismatch with societal norms, known as social jetlag, adversely affecting sleep quality, mood and fatigue^7–10. Shift workers, compared to daytime workers, are more susceptible to a host of issues such as sleep disturbances, insomnia¹¹, depression¹², and may be at increased risk for obesity, cancer, cardiovascular diseases and premature aging, all of which are linked to poor sleep^13–15. Due to miss out on rest and recover during the body’s natural time, night shift workers often face daytime disturbances like light, noise and family activities when they try to sleep after shift¹⁶. Long-term sleep loss and poor sleep quality can diminish work performance and focus, endangering occupational safety and increasing the risk of accidents^17,18.

Sleep disorders have emerged as a significant concern for shift workers, necessitating the implementation of strategies to enhance their sleep quality and mitigate potential health effects^19,20. Extensive evidence from previous researches suggests that light therapy (LT), characterized by its non-invasive and non-pharmacological nature, serves as an effective intervention for ameliorating sleep disturbances in this demographic^21,22. Specifically, nocturnal light exposure has been demonstrated to modulate circadian rhythms²³, postpone sleep onset and augment daytime sleepiness²⁴. For individuals engaged in rotating shift work, exposure to morning light has been shown to reset circadian rhythms, enhance sleep quality during non-working days, and subsequently improve cognitive functioning during next shifts^16,25.

Light therapy predominantly exerts its effects by administering a prescribed dosage of light stimulation. The light enters the eyes and subsequently influence the neural activity within the suprachiasmatic nucleus of the hypothalamus via both visual and non-visual pathways²⁶. Such modulation regulates the synthesis of melatonin (a sleep-related hormone), thereby resets the circadian rhythm, which rectifies disrupted sleep-wake cycles and improve sleep-related issues²⁷. Despite a number of studies examining the impact of light therapy on the sleep patterns of shift workers, there exists variability in the phototherapy protocols and their reported efficacy. To utilize phototherapy more effectively for addressing sleep disturbances in shift workers, it is crucial to delve into the effective dosage and efficacy assessment indicators to further optimize phototherapy strategies and carry out clinical treatment. Researches indicate that light intensity is not the sole determinant of therapeutic efficacy; other factors, including spectral composition, duration of exposure, and timing of treatment, also play significant roles^28,29. A subgroup analysis within a meta-analysis revealed that night light exposure to moderate-intensity light (1000–5000 lx, g = 0.632) for a short duration (≤ 1 h, g = 0.504) mitigates sleepiness among shift workers, whereas high-intensity light (> 5000 lx, g = 2.676) can delay the circadian rhythm and enhance sleep quality³⁰. Furthermore, another meta-analysis on light therapy for insomnia demonstrated varying impacts on sleep-wake cycles depending on the timing of treatment, and sleep measures confirmed that light therapy significantly reduced the wake after sleep onset (SMD = -0.61, p = 0.017)²⁹. Additionally, some studies suggest that a minimum of 30 min of light therapy can yield therapeutic benefits, although too long treatment duration (exceeding six hours) may cause adverse effects such as headaches and ocular discomfort^31,32.

Although researchers acknowledge the promising potential of non-pharmacological light therapy in enhancing the sleep quality of shift workers, the evidence regarding its efficacy is still unclear. Considering the complex role of various light parameters on sleep disorder, this investigation is grounded in a comprehensive review of the extant literature. It endeavors to elucidate the key light parameters pertinent to the phototherapeutic intervention for sleep disorders among shift workers, as well as the sleep indicators for evaluating treatment efficacy, aiming to provide suggestions for optimizing light therapy strategies and clinical efficacy evaluation.

Methods

Search strategy

This systematic review was registered with PROSPERO under code number CRD42024518298. The literature published from January 1, 1990, to December 31, 2023 were systematically searched based on five electronic databases, namely Web of science, PubMed, Cochrane, Scopus and Embase. The search terms using MeSH and Entry terms includes (“light therapy”, “phototherapy”, “photo-radiation therapy”, or “light”) and (“shift work”, “shift worker”, “sleep”, “sleep-wake schedule”, “sleep disorder”, “sleep wake schedule disorder”, or “circadian rhythm sleep disorders”), which were applied to title, keywords, abstract and full field. The detailed search strategy is shown in Table S1. This systematic review was conducted in compliance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The obtained literatures were managed and further screened using Endnote 21.

Study screening

The included literature complied with the following criteria: (A) The subjects are shift workers or participated in simulated shift work; (B) The study design have to adopt the randomized controlled trial or randomized crossover trial; (C) The intervention group uses light therapy as the only treatment method, while the control group does not receive light therapy or receives low doses of light; (D) Sleep outcome measures include relevant indicators that reflect sleep quality (e.g. total sleep time, wake after sleep onset, sleep efficiency) ; (E) Studies include complete data of sleep indicators to calculate effect sizes. The PRISMA flow chart with inclusion and exclusion protocols was shown in Fig. 1.

Fig. 1.

The flow chart of study selection process.

Risk-of-bias assessment

The trial quality was assessed by two independent reviewers based on the Cochrane “revised tool to assess the risk of bias of randomized studies” (RoB 2.0) for randomized controlled trials or cross-over trials. Using “low risk”, “high risk” or “unclear” qualified each criterion including selection bias, performance bias, detection bias, attrition bias, and reporting bias. The final score was decided by discussion.

Data extraction

The experimental details and outcome data were extracted from articles that met the criteria: sample size, age, gender, shift work period, phototherapy information (illuminance, CCT, treatment period and duration), details of control group, sleep measures (sleep latency (SL), total sleep time (TST), sleep efficiency (SE) and wake after sleep onset (WASO)).

Data analysis

Mean differences (MD) and 95% confidence intervals (95% CI) were calculated based on the mean, standard deviation, sample size, and weights of the intervention and control groups to quantify effect size. The ReviewManager (RevMan) was applied to calculate and analyze data. Assuming that shift workers come from different populations and accounting for the uncertainty of variation and estimation in treatment effects, the random effects model is used. The P and I²values were used to assess heterogeneity. Egger’s test was utilized to assess publication bias. Furthermore, a sensitivity analysis was conducted to assess the robustness of the model by evaluating the heterogeneity and effect size before and after the exclusion of specific studies. Light parameters, such as light intensity, duration, and treatment time, were categorized and examined through subgroup analyses to investigate the essential factors influencing the impact of light therapy on sleep.

Results

Study characteristics

Figure 1 exhibited a flow diagram with inclusion and exclusion protocols. 11 studies of 5994 studies identified through 5 electronic databases were included in this meta-analysis based on multiple screening criteria^{1–3,8,16–19,22,27,28}. As shown in Table 1, the characteristics of the included studies briefly describing Basic information, sample size, intervention, shift or simulated shift period, measurement tools and results. This study comprised a cohort of 195 individuals, with an age range of 18–60 years. The participants were engaged in rotational shift work or assigned to a simulated shift schedule, which invariably incorporated night shifts. A subset of the investigations incorporated a hybrid shift paradigm, alternating between day and night work periods. The experimental group was subjected to a range of interventions, including exposure to bright light, single light and natural daylight, in contrast to the control group, which received either a placebo, dim light, standard indoor illumination, or no active treatment. The duration of the intervention was identical in both experimental and control groups. Post-intervention sleep assessments were conducted subsequent to nocturnal shift work. Except for one study that used Karolinska sleep diary, the remaining studies predominantly employed actigraphy or polysomnography for sleep measurement. The preponderance of research favored light intensities exceeding 2500 lx, while a study that opted for a less intense illumination selected a blue-enriched or red-enriched light, regarding it to be as efficacious as greater intensities of standard white light. The Egger’s test was conducted to check the publication bias, no significant publication bias was found for the sleep latency (p = 0.348), total sleep time (p = 0.507), sleep efficiency (p = 0.214) and wake after sleep onset (p = 0.432).

Table 1.

Characteristics of the included studies. Experiment group (EG), Control group (CG), Bright light (BL), sunlight (SL), Blue intervention (BI), Red intervention (RI), room light (RL), White light (WL), no light (NL), actiwatch (ACT), Karolinska sleep diary (SD).

Study	Sample size	Age M ± SD	Shift work period	Intervention, duration	Control	Treatment period	Days	Tool	Measures
Baehr 1999	EG:8 CG:8	23.8 ± 4.6	01:30 − 09:30	BL:5000 lx, 40 min * 6	WL:< 500 lx	BL:01:30 − 07:30	BL:3	ACT	TST
Bjorvatn 1999	EG:6 CG:6	38.9 ± NA	19:00–07:00	BL:10,000 lx, 30 min	RL:200–300 lx	BL:03:30 − 5:30	BL:4	SD	SL, TST
Bjorvatn 2007	EG:17 CG:17	42 ± NA	18:30 − 06:30, 06:30 − 18:30	BL:10,000 lx, 30 min	placebo	BL:00:00–05:00	BL:4	ACT	SL, TST, SE
Bivion 2012	EG:9 CG:8	NA (25–53 years)	00:00–08:00	BL:3243 ± 2274 lx, 6 h	RL:111 lx	BL:0:00–06:00	BL:3	Nightcap, PSG	TST, SE
Campbell 1995	EG:13 CG:13	49.1 ± 6.4	24:00–08:00	BL:>5000 lx, 4 h WL:1000 lx, 385 min	RL:< 100 lx	BL:00:00–04:00 WL:00:00–08:00	BL:1 WL:2	EEG	SL, TST, SE, WASO
Dawson 1991	EG:6 CG:7	21.2 ± 3.1	23:00–08:00	BL:6000 lx, 4 h	RL:100–200 lx	BL:00:00–04:00	BL:3	EEG	SL, TST, SE, WASO
Lowden 2004	EG:9 CG:9	36.2 ± 3	24:00–06:30	BL:2500 lx, 20.0 ± 0.48 min and 20.9 ± 0.78 min	RL:300 lx	BL:24:00–06:00	BL:5	ACT	SE
Martin 2021	EG(BI):15 EG(RI):15 CG:15	26.6 ± 2.5	00:00–07:00	BI: NA, ~ 66.3 lx, 5 h RI: NA, ~ 66.3 lx, 5 h	NL	BI:00:00–05:00 RI:00:00–05:00	BI:4 RI:4	ACT	SL, TST, SE, WASO
Sunde 2019	EG:36 CG:36	NA (19–30 years)	23:00–07:00	BL:900 lx, 6 h	RL:90 lx	BL:23:00–5:00	BL:3	ACT	TST
Thorne 2010	EG:14 CG:14	46 ± 11(8 subjects), 49 ± 7(6 subjects)	19:00–07:00(eight), 18:00–06:00(six)	WL:3000 lx, 1 h	NL	WL:13:00–14:00	WL:4	ACT	SL, TST, SE, WASO
Yoon 2002	EG:12 CG:12	NA (21–24 years)	22:00–08:00	BL:4000–6000 lx, 4 h SL:10,000 lx,1 h	RL:100–500 lx	1:00–5:00, 08:30 − 09:30	BL:3 SL:3	ACT	SL, TST, SE

Risk of bias

As shown in Table 2, the assessment results of each bias risk item showed that all included studies were not assessed as low risk in all items. One study had a high risk of deviations from intended intervention, four had a high risk of missing outcome data due to subject loss, equipment failure, and data loss, and one had a high risk of measurement of the outcome. The remaining studies showed low risk or some concerns in the items. Therefore, the overall assessment results showed that 6 studies had high risk, 2 had some concerns, and 3 were low risk.

Table 2.

Risk-of-bias assessment: judgments about each risk-of-bias items of included studies. x-high risk, --some concerns, √-low risk, / -not applicable.

Study	Research design	Domain 1	Domain S	Domain 2	Domain 3	Domain 4	Domain 5	Overall
Baehr 1999	RCT	-	/	-	√	x	√	x
Bjorvatn 1999	Crossover	-	√	-	√	-	√	-
Bjorvatn 2007	RCT	√	/	√	x	√	√	x
Bivion 2012	RCT	-	/	-	x	√	√	x
Campbell 1995	RCT	-	/	-	x	-	√	x
Dawson 1991	RCT	-	/	-	√	√	√	-
Lowden 2004	Crossover	-	√	x	√	√	√	x
Martin 2021	Crossover	-	√	√	√	√	√	√
Sunde 2019	Crossover	-	√	√	√	√	√	√
Thorne 2010	Crossover	-	√	√	x	-	√	x
Yoon 2002	Crossover	√	√	-	√	√	√	√

Domains 1: Bias arising from the randomization process.

Domains S: Bias arising from period and carryover effects.

Domains 2: Bias due to deviations from intended intervention.

Domains 3: Bias due to missing outcome data.

Domains 4: Bias in measurement of the outcome.

Domains 5: Bias in selection of the reported result.

Meta-analyses of sleep and light therapy

Due to the differences between subjective and objective sleep, ten articles including objective sleep (a total of 11 articles) were included for meta-analysis and subgroup analysis.

Sleep latency

As shown in Fig. 2, no significant differences were observed in sleep latency between the LT and control group (MD = 0.72, p = 0.55, 95% CI = -1.65 to 3.09, I² = 43%). The effects of LT on reducing sleep latency were observed in three studied but it was associated with an increase or no change in sleep latency in five other studies.

Fig. 2.

Forest plot: sleep latency, light therapy (LT) group vs. control group.

Total sleep time

As shown in Fig. 3, there are a significant improvement in total sleep time in LT group compared with control group (MD = 32.54, p < 0.00001, 95% CI = 21.67 to 43.41, I² = 57%). Total sleep time was increased after LT in nine studies but not in one study. A sensitivity analysis indicated that the overall effect remained when each individual study was sequentially omitted (all p < 0.00001).

Fig. 3.

Forest plot: total sleep time, light therapy (LT) group vs. control group.

Sleep efficiency

As shown in Fig. 4, the outcome of sleep efficiency presented a significant improvement in the LT group vs. the control group (MD = 2.91, p = 0.007, 95% CI = 0.80 to 5.02, I² = 82%). Five of these studies indicated an associated between LT and enhancement of sleep efficiency, whereas three other studies demonstrated opposite results. The result of sensitivity analysis suggests a stable overall effect when each individual study was sequentially omitted (p = 0.03 − 0.004).

Fig. 4.

Forest plot: sleep efficiency, light therapy (LT) group vs. control group.

Wake after sleep onset

As shown in Fig. 5, no significant differences were observed in wake after sleep onset when comparing LT group with control group (MD = -9.09, p = 0.26, 95% CI = -24.99 to 6.80, I² = 89%). Two studies present that LT decrease WASO but not in three studies.

Fig. 5.

Forest plot: wake after sleep onset, light therapy (LT) group vs. control group.

Circadian rhythm phase

As shown in Fig. S1, there are a significant delay in circadian rhythm phase in LT group compared with control group (MD = 1.72, p = 0.001, 95% CI = 0.66 to 2.77, I² = 89%). All studies indicated the delay in circadian rhythm phase after light therapy.

Subgroup analysis of light therapy and research design

The effects of light therapy on sleep

Table 3 delineates the array of phototherapy parameters and values as documented in the studies, including illuminance, duration of treatment, total treatment time and treatment period. In addition, the values were calculated to investigate the possible interactive effects of illumination and total time.

Table 3.

The parameters and dose of light therapy of included studies.

Study	Intervention		Total time (h)	lx*h (per 1000)	Treatment period
	Illuminance (lx)	Duration of treatment (h/day)
Baehr 1999	5000	4	12	60	Night
Bjorvatn 2007	10,000	0.5	2	20	Mixed
Boivin 2012	3243 ± 2274	6	18	NA	Night
Campbell 1995	> 5000, 1000	3.5	16.8	NA	Night
Dawson 1991	6000	4	12	72	Night
Lowden 2004	2500	NA	NA	NA	Night
Martin 2021 BI	~ 66.3	5	20	1.3	Night
Martin 2021 RI	~ 66.3	5	20	1.3	Night
Sunde 2019	900	6	18	16.2	Night
Thorne 2010	3000	1	4	12	Night
Yoon 2002	4000–6000, 10,000	5	15	NA	Day

Table 4 displays the outcomes of the subgroup analysis, indicating that the impact on total sleep time and sleep efficiency varies across different subgroups. For total sleep time, bright light with a medium illuminance has a significant effect: 900–6000 lx (MD = 37.10, p < 0.00001, 95% CI = 22.95–51.25, I² = 75%). Daily exposure to the medium and long treatment duration can significantly improve TST: 1–4 h (MD = 48.68, p < 0.00001, 95% CI = 29.07–68.30, I² = 32%) and > 4 h (MD = 24.79, p < 0.00001, 95% CI = 19.72–29.85, I² = 0%). The significant effect was observed in medium and long treatment duration on total: 5–15 h (MD = 58.77, p < 0.00001, 95% CI = 43.34–74.20, I² = 0%) and > 15 h (MD = 24.69, p < 0.00001, 95% CI = 19.61–29.77, I² = 0%). And treatment conducted during nighttime were notably more effective compared to those during daytime or mixed period: night (MD = 31.57, p < 0.00001, 95% CI = 20.09–43.05, I² = 61%). These findings suggest light exposure with a medium illuminance and longer duration for achieving a favorable efficacy of LT on TST, whereas a longer total treatment duration and intervention during nighttime was more effective. For sleep efficiency, Medium and high intensity light has a significant effect on SE: 900–6000 lx (MD = 4.35, p = 0.01, 95% CI = 0.94–7.76, I² = 90%) and > 6000 lx (MD = 3.26, p = 0.05, 95% CI = -0.07-6.58, I² = 0%). The medium treatment duration on total showed a significant effect: 5–15 h (MD = 7.81, p = 0.05, 95% CI = 0.16–15.45, I² = 86%). And The effect of light exposure is more significant at night: night (MD = 2.87, p = 0.57, 95% CI = 5.17, I² = 87%). These findings implied that the significant effect of LT on SE was associated with medium and high illuminance, and that a medium total treatment duration along with intervention during nighttime was more efficacious.

Table 4.

Subgroup analysis of light therapy on total sleep time and sleep efficiency.

Sleep measure	TST				SE
Subgroup	k	MD (95% CI)	p (Z)	I2	k	MD (95% CI)	p (Z)	I2
Illuminance (lx)
< 900	2	17.65 (-3.21, 38.50)	0.10	0%	2	-0.24 (-2.07, 1.58)	0.8	0%
900-6000	6	37.10 (22.95, 51.25)	<0.00001	75%	5	4.35 (0.94, 7.76)	0.01	90%
>6000	2	21.99 (-11.02, 55.00)	0.19	0%	2	3.26 (-0.07, 6.58)	0.05	0%
Duration of treatment (h/day)
<1	1	16.00 (-25.36, 57.36)	0.45	/	1	2.00 (-3.72, 7.72)	0.49	/
1-4	5	48.68 (29.07, 68.30)	<0.00001	32%	3	6.16 (-1.17, 13.49)	0.1	71%
>4	4	24.79 (19.72, 29.85)	<0.00001	0%	4	1.93 (-0.75, 4.61)	0.16	77%
Total time (h)
<5	2	38.77 (-3.71, 81.25)	0.12	58%	2	2.77 (-1.48, 7.01)	0.20	0%
5-15	3	58.77 (43.34, 74.20)	<0.00001	0%	2	7.81 (0.16, 15.45)	0.05	86%
>15	5	24.69 (19.61, 29.77)	<0.00001	0%	4	1.33 (-1.72, 4.37)	0.39	77%
Treatment period
Night	8	31.57 (20.09, 43.05)	<0.00001	61%	7	2.87 (0.57, 5.17)	0.01	87%
Day or mixed	2	38.77 (-3.71, 81.25)	0.07	58%	2	2.77 (-1.48, 7.01)	0.2	0%

Table S2 shows the results of subgroup analysis of rhythm phase shift, light therapy with medium illumination, daily exposure to the medium treatment duration and medium total time have a significantly effect: 900–6000 lx (MD = 2.78, p = 0.001, 95% CI = 1.12–4.43, I² = 89%), 1–4 h (MD = 3.50, p < 0.00001, 95% CI = 2.68–4.32, I² = 0%) and 4–15 h (MD = 3.46, p < 0.00001, 95% CI = 2.56–4.35, I² = 0%).

Figure 6 depicts the dose-response relationships between total sleep time and sleep efficiency with respect to illuminance and lx*h. Studies lacking explicit dosage information have been omitted from this analysis. Specifically, Fig. 6a,b,c excludes the data from the studies conducted by Boivin, Campbell and Yoon, while Fig. 6d do not incorporate the study by Boivin, Campbell, Lowden and Yoon. For TST, we observed a significant dose-response relationship with illuminance (Fig. 6a), which could be optimally modeled using a quadratic function (R² = 0.822, p = 0.014). Conversely, the relationship between TST and lx*h did not demonstrate a significant dose-response (Fig. 6c). Regarding SE, no significant dose-response relationship was found between SE and illuminance (Fig. 6b). However, there was a significant dose-response relationship between SE and lx*h (Fig. 6d), it can be best fitted by linear regressions (R² = 0.92, p = 0.006). The findings suggest that there are nonlinear effects on the illuminance and improvement of TST, indicating that a moderate intensity of illumination can achieve the optimal therapeutic benefits for TST. While a linear relationship was observed between lx*h and SE, with SE showing gradual improvement as the dose of light exposure increased.

Fig. 6.

the dose-response curve of total sleep time versus (a) illuminance, (c) lx*h, and sleep efficiency versus (b) illuminance, (d) lx*h. The color areas represent 95% confidence intervals.

The effects of research design on sleep

The information on research design of included studies were listed in Table 5, including shift work, light form, research design, shift design and control design. The control group with light exposure was classified as active treatment.

Table 5.

The research design of included studies.

Study	Shift work	Light form	Research design	Shift design	Control design
Baehr 1999	Night shift	Single	RCT	Real	WL
Bjorvatn 2007	Mixed shift	Single	Crossover	Stimulated	NL
Boivin 2012	Night shift	Single	RCT	Real	RL
Campbell 1995	Night shift	Mixed	RCT	Real	RL
Dawson 1991	Night shift	Single	RCT	Real	RL
Martin 2021 BI	Night shift	Single	Crossover	Stimulated	NL
Martin 2021 RI	Night shift	Single	Crossover	Stimulated	NL
Sunde 2019	Night shift	Single	Crossover	Stimulated	RL
Thorne 2010	Mixed shift	Single	Crossover	Stimulated	NL
Yoon 2002	Night shift	Mixed	Crossover	Stimulated	RL

As shown in Table 6. For total sleep time, the significant treatment effect was observed in most of subgroups: night shift (MD = 33.00, p < 0.00001, 95% CI = 21.38–44.62, I² = 68%); single (MD = 35.21, p < 0.00001, 95% CI = 23.73–46.68, I² = 70%); RCT (MD = 39.28, p < 0.00001, 95% CI = 16.17–62.40, I² = 73%); crossover (MD = 23.67, p < 0.00001, 95% CI = 17.74–29.60, I² = 0%); real shift (MD = 23.47, p < 0.00001, 95% CI = 17.59–29.35, I² = 0%); simulated shift (MD = 40.71, p < 0.0001, 95% CI = 21.18–60.24, I² = 76%); active treatment (MD = 38.20, p < 0.00001, 95% CI = 24.35–52.06, I² = 79%) and no treatment (MD = 26.72, p = 0.009, 95% CI = 6.61–46.84, I² = 30%). For sleep efficiency, the significant treatment effect was observed in half of subgroups: night shift (MD = 2.87, p = 0.01, 95% CI = 0.57–5.17, I² = 87%); single (MD = 2.87, p = 0.01, 95% CI = 0.57–5.17, I² = 87%); crossover (MD = 0.89, p < 0.0001, 95% CI = 0.45–1.33, I² = 0%); real shift (MD = 4.47, p = 0.02, 95% CI = 0.68–8.27, I² = 93%); active treatment (MD = 4.35, p = 0.009, 95% CI = 1.08–7.61, I² = 90%). These results of subgroup analysis present that the research design significantly influenced treatment effects.

Table 6.

Subgroup analysis of research design on total sleep time and sleep efficiency.

Sleep measure	TST				SE
Subgroup	k	MD (95% CI)	p (Z)	I2	k	MD (95% CI)	p (Z)	I2
Shift work
Night	8	33.00 (21.38, 44.62)	<0.00001	68%	7	2.87 (0.57, 5.17)	0.01	87%
Mixed	2	38.77 (-3.71, 81.25)	0.07	58%	2	2.77 (-1.48, 7.01)	0.2	0%
Light therapy mode
Single	8	35.21 (23.73, 46.68)	<0.00001	70%	7	2.87 (0.57, 5.17)	0.01	87%
Mixed	2	12.86 (-28.83, 54.54)	0.55	6%	2	2.77 (-1.48, 7.01)	0.2	0%
Research design
RCT	4	39.28 (16.17, 62.40)	<0.00001	73%	3	6.10 (-0.40, 12.59)	0.07	85%
Crossover	6	23.67 (17.59, 29.35)	<0.00001	0%	6	0.89 (0.45, 1.33)	< 0.0001	0%
Shift design
Real shift	6	23.47 (17.59, 29.35)	<0.00001	0%	4	4.47 (0.68, 8.27)	0.02	93%
Simulated shift	4	40.71 (21.18, 60.24)	<0.00001	76%	5	1.02 (-0.81, 2.86)	0.27	19%
Control design
Active	6	38.20 (24.35, 52.05)	<0.00001	79%	5	4.35 (1.08, 7.61)	0.0009	90%
No treatment	4	26.72 (6.61, 46.84)	0.009	30%	4	0.23 (-1.45, 1.90)	0.79	0%

Discussion

This meta-analysis systematically explored the impact of phototherapy on sleep disturbances in shift workers, incorporating findings from 10 studies. A study (Bjorvatn 1999) was not included in our meta-analysis due to only subjective sleep measures. Given the high heterogeneity of the included studies, a random effects model was applied to assess the overall effect of light therapy on four objective sleep indicators: sleep onset latency, total sleep duration, sleep efficiency, and wake after sleep onset. The analysis revealed that, in comparison to the control group, there were significant improvement in total sleep time (32.54 min longer) and sleep efficiency (2.91% increase) within the LT group, while there was no significant effect on the other two sleep parameters. This differential effect could be attributed to the typical sleep pattern of shift workers. Due to social jet lag and multiple interference factors, it is difficult for shift workers to sleep for a long enough time (< 8 h) during the day after night shift. Phototherapy appears to reduce social jet lag by delaying the circadian rhythm, thereby facilitating longer sleep periods.

Based on the results of the subgroup analysis. The groups of longer duration and medium illuminance led to greater improvements in total sleep time, indicating that exposure to medium-intensity light is more beneficial, while low illumination has no significant effect and the opposite effect may occur at high illumination, whereas a longer treatment duration was more effective. In terms of sleep efficiency, the groups of medium and high illuminance demonstrated significant improvements in sleep efficiency, which means that the light exposure with higher illuminance is more beneficial. Previous research indicates that light exposure with a higher proportion of blue components, as indicated by higher Melanopic Effective Daytime Illuminance (EDI) or Circadian Stimulus (CS) values, plays a more significant role in regulating the human circadian rhythm^29,33,34. Although the lack of spectral data in most previous studies on light therapy prevents us from calculating the specific impact of blue light components in our study, it is recommended that future research should consider the influence of blue light on therapeutic outcomes.

In addition, the efficiency of light therapy is more significant at night, which may be related to the fact that nocturnal light exposure delays the circadian rhythm of shift workers. Supporting this, Lam et al. found that high-intensity (> 5000 lx) light exposure is more effective on sleep phase shifting in shift workers³⁰. Moreover, we found significant dose-response relationships between TST and illuminance, as well as SE and lx*h. The inverted U-shaped dose-response curve observed for TST and illuminance suggests the existence of an optimal illuminance level for phototherapy, beyond which the benefits do not further increase or may even decrease. While the linear relationship between SE and lx*h implies that as the dose of phototherapy increases, there is a corresponding improvement in SE. Therefore, this can provide a reference for dose setting of photo-therapy in future studies.

Light therapy also affects workers’ circadian rhythms. Meta-analysis results showed a significant delay in rhythm phase (1.72 h delay) within the LT group, this is consistent with the previous studies of Babkoff³⁵, Horowitz³⁶, Lee³⁷, Nagashima³⁸ and Smith³⁹, which also revealed that bright or blue-rich white light significantly delays circadian rhythms in shift workers. The subgroup analysis showed that the groups of shorter duration and higher illuminance led to an effective circadian phase delay. A meta-analysis of Lam et al.³⁰ showed the same results, but the required treatment duration remained inconclusive. Among the studies that assessed rhythm and sleep concurrently, two did not delay the rhythm phase and enhance sleep, while one study did delay the circadian rhythm but did not lead to improved sleep. In contrast, three studies managed to delay the rhythm phase and also increase total sleep duration. Therefore, future research is needed to investigate the relationship between rhythm delay and sleep enhancement in shift workers undergoing phototherapy.

Shift workers are known to be at a higher risk for sleep disorders compared to their non-shift counterparts. The research of Juliette et al. highlighted that light therapy significantly impacted WASO in individuals suffering from insomnia, with daytime light exposure yielding a greater improvement in SE and TST⁴⁰. The variance observed in research outcomes could stem from the differences in the populations studied, specifically insomniacs versus healthy subjects. Light therapy exhibited enhanced effectiveness in improving sleep among night shift workers (p < 0.05), likely because these individuals tend to have a more stable sleep-wake cycle compared to those working mixed shifts^32,41. Additionally, the effect of phototherapy on TST was not affected by shift design, but light therapy on SE was found to be more effective for real shift workers than those in simulated shift work conditions, possibly due to the former group suffering from more pronounced sleep disturbances and reduced sleep duration under long-term shift conditions. The presence or absence of light treatment in control groups did not alter the effect of phototherapy on TST, not on SE. Therefore, the impact of phototherapy also varied depend on the type of light therapy mode and research design. This provides a reference for the experimental design of future studies.

This meta-analysis acknowledges several limitations. A notable constraint is the small number of participants across studies, which diminishes the statistical power. Most studies of phototherapy lack the key parameter of spectrum, and there are large differences in illumination, treatment duration and treatment period. The results of subgroup analysis and regression models reflect the dose-responses of TST and SE, providing the optimal dose for reference, but more research data are still needed for a more precise dose-response curve. Moreover, the presence of high heterogeneity within some subgroups during the analysis necessitates a cautious interpretation of these findings. The mechanism by which phototherapy improves sleep in shift workers may be related to the adjustment of circadian rhythm, but this meta-analysis did not investigate the influence of LT on circadian rhythm. This gap underscores the need for further research to more precisely define how phototherapy can be utilized to improve the sleep quality of shift workers.

Conclusion

In summary, the findings of this meta-analysis show that light therapy can significantly improve total sleep time and sleep efficiency in shift workers. Exposure to medium illuminance and long treatment duration during nighttime can markedly improve sleep time and a potential saturation of the treatment dose of light therapy on total sleep time, while improvement of SE is associated with gradually increasing doses of phototherapy. It suggests that shift workers may be exposed to bright light during the night shift to delay circadian rhythms, increase daytime sleepiness, which in turn help in improving sleep-related disorders experienced by this population. The strategically timed and adequately intense light therapy could be a valuable tool in the effort to improve the sleep quality of shift workers.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Y.D.L. initiated and guided the research. C.Y.Z. contributed to the data collection and analysis, interpretation, and drafting of the manuscript. N.Y.L. and W.Q.M. contributed to the data analysis, interpretation. Y. H. contributed to the writing and editing.

Data availability

All data generated or analysed during this study are included in this published article.

Declarations

Competing interests

The authors declare no competing interests.