Effect of Administration of Ramelteon, a Melatonin Receptor Agonist, on the Duration of Stay in ICU

Nishikimi. Critical Care Medicine 2018; published online doi: 10.1097/CCM.0000000000003132

Clinical Question

  • In critically ill adults, does the administration of ramelteon compared to placebo reduce the length of intensive care stay?


  • Melatonin is a hormone excreted by the pineal gland with regulatory effects on circadian rhythm and possibly antioxidant and anti-inflammatory effects
  • Exogenous melatonin has been shown to improve sleep quality and reduce delirium, but little research has been conducted specific to Intensive Care Unit patients


  • Single centre, randomised, controlled trial
  • Recruitment by consequetive sampling of admissions to 10 bed emergency medical Intensive Care Unit (ICU)
  • Randomisation in 1:1 ratio prepared by “an outsider”, stratified by age, APACHE 2 score and intubation status in block sizes of four
  • Allocation concealment is not described
  • Patients, clinicians and statisticians were all blinded to the allocation
  • Intention-to-treat statistical analysis
  • Primary outcome was length of stay (LOS) in ICU
    • Due to non-normal (right skew) distribution of LOS data, the authors chose to apply log-transformation to approximate a normal distribution
    • Once transformed, Student’s t test was used to compare two means
  • Power calculation performed a priori
    • Clinically important reduction in length of stay (LOS) defined as one day
    • Power 80% (20% probability of false negative conclusion)
    • Significance level 0.05 (5% probability of false positive conclusion)
    • Required sample size of 182 patients


  • Single centre in Nagoya, Japan
  • 10 bed Emergency and Medical ICU
    • Surgical ICU separate unit
  • May 2015 – April 2017


  • Inclusion: Adults (≥20 years); able to recieve medication enterally (oral or via nasogastric tube) during first 48 hours of admission
  • Exclusion: Withheld consent; allergy to ramelteon; prior administration of ramelteon or fluvoxamine (potential drug interaction)
  • 98 patients screened; 92 randomised (6 excluded due to lack of consent); 88 included in modified intention-to-treat (4 discharged from ICU without receiving any study drug [2 in each group])
  • Baseline demographics were similar – authors highlighted some imbalances but since P-values were all > 0.05 it is unlikely these differences were due to an improper randomisation process (Ramelteon group vs Control group; * = intentionally balanced through stratification of randomisation)
    • Median age*: 68 years vs 68 years
    • Mean APACHE 2*: 23.98 vs 23.95
    • Mechanical ventilation*: 40.0% vs 46.5%
    • Sex: ♂ 73.3% | ♀ 26.7% vs ♂ 55.8% | ♀ 44.2% (P = 0.12)
    • Existing dementia: 13.3% vs 2.3% (P = 0.11)
    • Diagnosis of sepsis: 26.7% vs 20.9% (P = 0.62)
    • Heavy alcohol use:6.7% vs 2.3% (P = 0.62)
    • Pre-admission psychiatric or sleeping medication: 5.0% vs 7.2% (P = 0.57)


  • Ramelteon
    • 8 mg administered at 8pm daily until ICU discharge


  • Placebo
    • 1 g lactose powder indistinguishable from study drug administered at 8pm daily until ICU discharge

Management common to both groups

  • Bedside nurses and clinicians were unaware of allocation and encouraged to treat all patients equally
  • Delirium prevention was standardised
    • Avoidance of unnecessary mobilisation
    • Limitations to family visits (11am–12pm; 3pm–4pm)
    • Low lighting and sound level after 10pm
  • Treatment of delirium was standardised
    • Primarilily using non-pharmacological approach
    • Early mobilisation
    • Comfortable and sleep-conducive environment
    • Treatment of pain, dyspnoea and anxiety
    • Risperidone (1 mg) or dexmedetomidine (up to 0.7 ug/kg/hr) at clinicians’ discretion
    • Haloperidol 5 mg IV as last resort
  • Decision to discharge from ICU was made by a multidisciplinary group at a twice daily meeting based upon:
    • No / minimal sedation required
    • Oxygen saturation > 90% requiring less than FiO2 0.5
    • No mechanical ventilation
    • No / minimal inotropic or vasopressor support
    • Urine output > 0.5 ml/kg/hr
    • No need for renal replacement therapy
  • Delirium assessed using CAM-ICU
  • Sedation level assessed using Richmond Agitation Sedation Scale (RASS)


  • Primary outcome: duration of intensive care length of stay was shorter in the Ramelteon group compared to the Placebo group
    • Ramelteon group: median 4.56 days (range 2.10 – 7.07)
    • Placebo group: median 5.86 days (range 2.97 – 14.16)
    • Median difference: 1.3 days
    • P-value (after log-transformation) = 0.082
    • After a pre-specified multivariate analysis, administration of Ramelteon was independently associated with a statistically significant reduction in length of stay
      • P-value = 0.028
  • Secondary outcome:
    • Delirium occurred less frequently in the Ramelteon group
      • Ramelteon 24.4% vs Placebo 46.5%
      • OR: 2.69 (95% CI 1.09 to 6.65; P = 0.044)
      • ARR: 22.07% (95% CI 2.58% to 41.56%)
      • NNT: 5
      • Fragility Index: 1
    • Duration of delirium was shorter in the Ramelteon group
      • Mean 0.78 days vs 1.40 days (P = 0.048)
    • In-ICU mortality did not differ between the groups
      • Ramelteon 6.7% vs Placebo 7.5%
      • ARR: 0.31% (95% CI -10.23% to 10.85%; P = 1.0)
    • Quality of sleep was better in the Ramelteon group
      • Fewer awakenings per night (0.8 awakenings vs 1.31 awakenings; P = 0.045)
      • More nights without awakenings (51% vs 30%; P = 0.048)
      • No difference in mean hours of sleep (7.29 hrs vs 6.78 hrs; P = 0.252)

Authors’ Conclusions

  • Administration of Ramelteon to ICU patients was associated with a “tendency toward a decreased duration of ICU stay”, and a decrease in the incidence and duration of delirium that reached statistical significance


  • Sensible theory and testable hypothesis
  • Appropriate outcome measure relevant to hospital resource use and patients, investigating the overall multi-modal effect of melatonin-receptor agonism
  • Balanced baseline variables suggest the randomisation process was unbiased, despite the concerns about sequence generation and allocation concealment described as weaknesses below
  • Good blinding of all relevant individuals
  • Primary outcome was guided by criteria and multidisciplinary input to reduce subjectivity
  • Statistical manipulation and analysis was appropriate to the data type
  • Clinically important difference was defined a priori


  • Massive under-recruitment has reduced the statistical power of the conclusion of this trial
    • A false negative conclusion may have been drawn
    • The manuscript reports that fewer patients than expected met the inclusion criteria
    • The number of patients screened and the reasons for not meeting the inclusion criteria are not provided
  • Method for construction of randomisation sequence is not clearly defined
    • Was it computer generated or was a simpler but possibly biased method used?
  • Concealment of allocation not described
    • Was the randomisation sequence hidden adequately prior to patient allocation?
  • Maintenance of allocation not reported
    • Did all patients receive intended drug and were they otherwise treated the same?
  • Single centre and medical only ICU design may limit the external generalisability beyond the scope of medical emergencies in Japan

The Bottom Line

  • This is a promising trial that adds to the growing evidence of benefit from administration of melatonin and melatonin-receptor agonists
  • Despite concern about some of the methods, the overall conduct of this trial is good and I believe the internal validity is good enough to accept the results as accurate
  • However, the small sample size and single centre design limits the generalisability and I shall await larger trials before recommending melatonin or ramelteon to all critically ill patients

External Links


Summary author: Duncan Chambler
Summary date: 11 April 2018
Peer-review editor: Adrian Wong

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.