Melatonin treatment has consistent but transient beneficial effects on sleep measures and pain in patients with severe chronic pain: the DREAM-CP randomised controlled trial

Abstract

Background: Sleep disturbance is a major issue for patients with chronic pain. Melatonin has been shown to improve symptoms of fibromyalgia, but its efficacy in other chronic non-malignant pain conditions is not fully known. Hence, we determined the effect of melatonin in patients with severe noncancer chronic pain.

Methods: This was a randomised double-blinded crossover trial of modified-release melatonin as Circadin™ compared with placebo. Sixty male and female subjects with chronic severe pain were randomised to receive either 2 mg of Circadin™ or placebo before sleep for 6 weeks, followed by a >4 week washout, then crossing over to the other treatment. Sleep disturbance, quality, and latency were measured using three different validated sleep assessment tools. The primary outcome measure was self-reported sleep disturbance after 6 weeks of treatment. Adverse events were also recorded.

Results: Sleep disturbance after 6 weeks was not significantly altered by melatonin treatment, but differences between melatonin and placebo treatment periods after 3 weeks were seen: sleep disturbance (P=0.014), latency (P=0.04), overall sleep quality (P=0.004), and effect of pain on sleep (P=0.032). Pain intensity scores improved during both treatment periods (both P<0.001). There were no differences in adverse events between treatment periods.

Conclusions: Circadin™ treatment did not improve sleep disturbance in patients with severe chronic pain compared with placebo at 6 weeks, but there were consistent improvements in aspects of sleep in the shorter term. Given its favourable safety profile, it could be beneficial for some patients with chronic pain.

Clinical trial registration: ISRCTN12861060.

Keywords: chronic pain; clinical trial; crossover; melatonin; sleep.

Fig 1

Schematic view of the trial design showing the ‘journey’ that each participant followed. For Group A, treatment allocation 1 was melatonin and treatment allocation 2 was placebo. For Group B, treatment allocation 1 was placebo and treatment allocation 2 was melatonin.

Fig 2

CONSORT diagram for randomised crossover trials.

Fig 3

Verran and Snyder–Halpern (VSH) sleep disturbance scores (intention-to-treat analysis). There was a significant difference between melatonin and placebo treatment periods at Visit 2 (3 weeks, P=0.014) but not at Visit 3 (6 weeks). There was also a significant difference between Visit 1 (baseline) and Visit 2 (3 weeks, P<0.001) and between Visit 2 and Visit 3 (P=0.004) during melatonin but not placebo treatment periods. Box and whisker plots show median, inter-quartile, and full range with individual data points overlaid. Linear mixed effects models for treatment (P=0.13), sequence (P=0.99), and period (P=0.66) were not significant. P-values shown are Bonferroni corrected as appropriate.

Fig 4

(a) Pain and Sleep Questionnaire three-item index (PSQ-3) (intention-to-treat analysis). There was a significant difference in PSQ-3 between melatonin and placebo treatment periods at Visit 2 (3 weeks, P=0.032) but not at Visit 3 (6 weeks). There was also a significant difference in PSQ-3 between Visit 1 (baseline) and Visit 2 (P<0.001) and between Visit 2 and Visit 3 (P<0.001) during melatonin but not placebo treatment periods. (b) Pittsburgh Sleep Quality Index (PSQI) global sleep quality score (intention-to-treat analysis). There was a significant difference in PSQI between melatonin and placebo treatment periods at Visit 2 (3 weeks, P=0.004) but not at Visit 3 (6 weeks). There also was a significant difference in PSQI between Visit 1 (baseline) and Visit 2 (P<0.001) and between Visit 2 and Visit 3 (P<0.001) during melatonin but not placebo treatment periods. Box and whisker plots show median, inter-quartile, and full range with individual data points overlaid. Linear mixed effects models on PSQ-3 were significant for treatment (15.3; 95% confidence interval [CI] 1.1–29.5; P=0.034), but not for sequence (P=0.51) and period (P=0.38). Linear mixed effects models on PSQI were significant for treatment (0.74; 95% CI 0.16–1.32; P=0.012), but not for sequence (P=0.35) and period (P=0.53). P-values shown are Bonferroni corrected as appropriate.