Light Therapy on Sleep Phase-Shift in Shift Workers: A Meta-Analysis

Sunday, 30 July 2017

Min-Huey Chung, PhD, RN
School of Nursing, College of Nursing, Taipei Medical University, Taipei, Taiwan
Calvin Lam, MSSc
School of Nursing, College of Nursing, Taipei Medical University, Taiwan, Taiwan

Dose-response of Light Therapy on Sleep Phase-shift in Shift Workers: Meta-Analysis


The treatment effects of Light therapy on sleep were widely studied. However, there was limited meta-analysis had investigated the dose-response of light therapy and for shift workers. We are conducting a meta-analysis by including eligible randomized controlled trials (RCTs) to investigate whether there is dose-response of light therapy on sleep phase shift of shift workers.


We searched for relevant articles in 10 databases, included PubMed, Medline, Cochrane, Embase, LILACS, Open Grey, PsycINFO, PubPsych, SCOPUS, and CINAHL. Search keywords included “light therapy”, “shift work”, and “randomization”. Two reviewers independently screened the articles and extracted the data. No additional data were found even after contacting the authors of the eligible studies. We finally included 9 RCTs from 1819 records. Analyses were performed using the random-effects model.


Outcomes were measured by melatonin concentration and phase shift in hour. Light therapy exerted a significant effect on sleep phase shift, with a large pooled effect size (Hedges’s g = 1.476, p < 0.001). However, There was significant high heterogeneity (Q = 55.56, p < 0.001, I2 = 85.6) but low publication bias (Begg’s test, p = 0.076; Egger’s test, p = 0.081, no missing study in the funnel plot). There were significantly different effect sizes between the studies with the outcomes of melatonin concentration (k = 4, g = 1.201, p = 0.07) and phase shift in hour (k = 5, g = 1.723, p = 0.004). However, no significant treatment dose-response on the sleep phase shift.


The large effect size suggests that Light therapy is effective in sleep phase shift for shift workers. However, the high heterogeneity, inconsistent effect sizes between different measurements, and non-significant treatment dose-response turn the current findings inconclusive. Further investigation on the possible moderators, such as demographic data, and potential research bias, such as methodological quality, are needed.