Exogenous melatonin in the treatment of pain: a systematic review and meta-analysis

Melatonin is an important hormone for regulating mammalian circadian biology and cellular homeostasis. Recent evidence has shown that melatonin exerts anti-nociception effects in both animals and humans. However, according to clinical trials, the anti-nociception effects of melatonin are still controversial. The aim of this meta-analysis was to investigate the anti-nociception effects of melatonin premedication. The primary outcome was the effects of melatonin on pain intensity. The secondary outcomes included the number of patients with analgesic requirements, total analgesic consumption, and brain-derived neurotrophic factor (BDNF) levels. In total, 19 studies were included in the current meta-analysis. The pooling data show that melatonin significantly decreased the pain intensity, as evidenced by the pain scores. Moreover, melatonin administration also reduced the proportion of patients with analgesic requirements and BDNF levels. However, the effects of melatonin on total analgesic consumption still require further confirmation. Collectively, the current meta-analysis supports the use of melatonin for anti-nociception.

Interestingly, pain perception, especially heat and cold pain tolerance, is observed to vary diurnally, which may result from variations in melatonin levels [21]. In particular, acute pain stimuli influence the salivary melatonin levels [22]. Reduced endogenous melatonin exacerbates nerve injury-induced neuropathic pain [23], whereas melatonin administration significantly attenuates sleep deprivation-induced neuropathic pain [24]. Recent experimental evidence has also shown that melatonin could significantly alleviate pain behaviors under other conditions [19,[25][26][27].

Data sources and searches
Two authors (Chaojuan Zhu and Yonghong Duan) independently searched the electronic databases, including MEDLINE (1990 to July 2017), EMBASE (1990 to July 2017), Scopus (1988 to July 2017), PsycINFO (1990 to July 2017), and the Cochrane Library (Issue 5 of 12, July 2017). Searches were limited to humans but were not limited by published language and publication type. References of published original articles, reviews and contacted experts were further searched. Melatonin, N-acetyl-5-methoxy tryptamine, pain, ache, dolor, pain management and analgesia were used as search terms.
[((((pain) OR ache) OR dolor) OR pain management) and ((melatonin) OR N-acetyl-5-methoxy tryptamine)] were used as search phrases. The full study flow diagram is shown in Figure 1.
RCTs enrolled participants with any type of pain, including inflammatory pain, operation-associated pain, experimental pain and procedural pain. Melatonin could be administered orally or intravenously, either solely or in combination with other drugs. Both trials comparing melatonin with placebo or with positive control were included. Trial protocols without results, animal experiments, and studies without control groups were excluded.

Data extraction
Two authors (Chaojuan Zhu and Xu Yunyun) independently extracted data, including study characteristics and the main outcomes. When discrepancies appeared, consensus was achieved through discussion with a third author (Yonghong Duan). Characteristics of patients, such as age, condition, number of patients and intervention, were collected.
The primary outcome was pain intensity, as evidenced by pain scores. Based on the results reported in RCTs, the Visual Analog Scale (VAS) was the most used scale in the included trials. Additionally, children's anxiety and pain scales and premature infant pain profiles were also pooled. Secondary outcomes were the number of patients with analgesic requirements, total analgesic consumption, and brain-derived neurotrophic factor (BDNF) levels.
When only standard error (SE) was reported, SE was transformed to standard deviation (SD) using the formula: SD=SE×√n. When the 95% confidence interval (95% CI) was recorded, SD=√n×(upper limit-lower limit)/3.92. When only the median and interquartile range (IQR) were recorded, they were used to estimate the mean and SD: mean≈median, SD≈Norm IQR = (P75-P25)×0.7413 according to the Cochrane Handbook for Systematic Reviews of Interventions [48]. If the mentioned data were unavailable in Oncotarget 100584 www.impactjournals.com/oncotarget the text, we attempted to contact the corresponding authors to obtain the related information. If the present data in the figures were not available in the text and it was impossible to retrieve them from the corresponding authors, ImageJ (National Institutes of Health, Bethesda, MD) was applied to measure the values in the figures.

Quality assessment
Two authors (Chaojuan Zhu and Xu Yunyun) independently assessed the quality of the included trials according to the Jadad scales (5 items) [49]. Randomization, blinding, and withdrawals and dropouts were assessed. Table 1 presents the assessments of all included studies. The quality of the trial was described as high (score 5), moderate (score 4), or low (scores 1-3). The trials with low quality were not relied upon in the data pooling. Cohen's kappa (κ) was utilized to measure the inter-rater agreement.

Statistical analysis
Review Manager analysis software (RevMan 5.2) was used to analyze the collected data according to the Cochrane Handbook for Systematic Reviews of Interventions [48]. For continuous data measured by different scales, we used weighted mean differences (WMDs) with applicable 95% CIs to measure the mean values or mean changes. WMDs were calculated for pain intensity, total analgesic consumption and BDNF levels. For dichotomous data, the odds ratios (ORs) with applicable 95% CIs were used for the pooling data. ORs were calculated for the number of patients with analgesic requirements. I 2 statistics were used to measure heterogeneity of the RCTs. If the I 2 value was less than 50%, a fixed-effects model was applied. If the I 2 value was 50% or more, a subgroup analysis was conducted. Subgroup analyses for pain intensity primarily included different pain types and compared groups, while analyses for total analgesic consumption included the stage used and compared groups. If a significant heterogeneity could not be identified by subgroup analysis, a random-effects model meta-analysis was used [48]. Sensitivity analyses were conducted to examine the stability of the pooling outcome and to trace the heterogeneity source by excluding the sole study with low quality or high risk of bias. Publication bias was tested using a funnel plot ( Figure 2) and Egger's test [50,51]. Figure 1 present the full study screening process. Ultimately, 19 studies were included in this meta-analysis. Their quality assessment is shown in Table 1. High interrater reliability of initial decisions on the inclusion of studies was indicated (κ = 0.838, 95% CI: 0.53-1). All of the included studies were RCTs. Among them, 18 studies were randomized double-blind trials, 14 studies were placebo-controlled trials, and 17 studies were assessed as high quality. In total, 586 patients were included in the melatonin groups, while 507 patients received control therapy, including placebo and standard therapy. The characteristics of these studies are summarized in Table 2.
The secondary outcomes were the number of patients with analgesic requirements (Figure 4), total analgesic consumption ( Figure 5), and BDNF levels ( Figure 6). As reported in 4 studies, melatonin administration significantly decreased the proportion of patients requiring analgesic drugs (OR= 0.43, 95% CI, 0.24 to 0.79; P = 0.006; I 2 = 0%) (Figure 4). A randomeffects model was used to evaluate the effects of melatonin on requirements with analgesic drugs ( The funnel plot is shown in Figure 2. No significant publication bias was observed (Eggers test: 1.121428, 95% CI: -3.588125 to 5.830981, P = 0.623). No further trials were identified through extensive searches. No significant adverse events were reported in the included studies. Because the safety of exogenous melatonin supplementation has been confirmed, we did not pool the data relevant to this outcome.

DISCUSSION
This pooling study showed that melatonin could significantly decrease the intensity of every type of pain, including operation-associated pain under topical anesthesia, operation-associated pain under general anesthesia, inflammatory pain, procedural pain, and experimental pain. The mechanism for this reduction may depend on decreasing BDNF levels. Furthermore, melatonin also decreased the proportion of patients requiring additional analgesic drugs, but its effects on total analgesic consumption still require further confirmation.
In the excluded studies, the analgesic effects of melatonin were also demonstrated. In critically ill patients, long-term administration of melatonin also significantly improved the patients' pain [52]. Two other studies showed that melatonin significantly reduced the frequency and intensity of nocturnal pain in patients with ulcer-like dyspepsia [53] and the intensity of visceral pain in patients with irritable bowel syndrome [54]. Another trial was designed to investigate the efficacy of melatonin in Intensive Care Unit patients [55]. Accompanying the alleviation of pain, melatonin also significantly improved sedation and anxiety [35,37,42,43]. Similarly, compared with midazolam, melatonin also significantly contributed to sedation induction in children [31]. These studies were excluded primarily due to their unavailable data about the included outcomes, especially the pain scores.
The diurnal variation of pain tolerance may be attributed to melatonin variation [21]. Further evidence has shown that endogenous elevation of melatonin also improves pain tolerance. Probiotic administration increased the rectal distension pain threshold in irritable bowel  syndrome and elevates morning melatonin levels [56]. In particular, the increased morning melatonin levels were correlated with improved bowel habits and irritable bowel syndrome-associated pain (with a regression coefficient of 0.61) [56]. This study is coincident with the observation that direct supplement of melatonin also improved irritable bowel syndrome-associated visceral pain [54,57]. Melatonin is involved in the physiopathology of pain. BDNF is an important mediator and a central modulator of pain [58][59][60]. In neuropathic pain, BDNF release has been observed in the spinal cord, contributing to chronic pain [61]. Further study found that BDNF contributed to hyperpathia through presynaptic GABAergic inhibition [62]. In the colonic mucosa of patients with irritable bowel syndrome, increased BDNF expression was correlated with visceral hyperalgesia and increased abdominal pain scores [63]. The mechanisms for these observations may be involved in elevating tyrosine receptor kinase B expression [63]. BDNF is an important regulator of pain, and the antinociceptive mechanisms of melatonin may be attributed to decreased BDNF levels [34]. Furthermore, production of inflammatory factors is important to induce and maintain pain [64]. Melatonin can also reduce inflammatory cytokine levels, including interleukin (IL)-6, IL-8, IL-10 and IL-12, in newborns undergoing intubation and mechanical ventilation [39]. Another experimental animal study showed that melatonin significantly attenuated inflammation-mediated hyperalgesia in rats [65].
The analgesic effects of melatonin may also be influenced by several factors, such as the duration of administration and gender. In an experimental animal study with Freund's adjuvant-induced inflammatory pain, the different administration durations of melatonin differently influenced the BDNF levels [66]. To be specific, short-term (3 days), but not long-term (8 days) administration of melatonin increased BDNF levels [66]. The analgesic effects of melatonin may be better in females than in males [47]. Seet et al. [47] found insignificant effects of melatonin in patients undergoing elective extraction of all four wisdom teeth, while further subgroup analysis showed a positive effect of melatonin in female, but not male, patients. Sexual dimorphism has also been indicated in other studies [56]. Elevated melatonin levels due to probiotics are observed in male patients, but not in female patients, further alleviating pain. In particular, the anti-nociceptive effects of melatonin in the pediatric population have been explored. However, further confirmation regarding melatonin application in this population is needed from more clinical trials.
In this meta-analysis, we conducted a systematic search for melatonin administration on anti-nociception. The pooling data included all available RCTs, involving 1053 patients. Strong evidence supports the utilization of melatonin on anti-nociception against many types of pain. Several limitations need to be considered. First, profound heterogeneity of the secondary outcome was found, i.e., total analgesic consumption. Although further subgroup analyses were performed to identify the heterogeneity source, no potential methodological and clinical sources of heterogeneity were identified. The considerable heterogeneity was also unaffected by exclusion of individual RCTs. Thus, a random-effects model was used for these data. Furthermore, the magnitude of this outcome