Effect of Dexmedetomidine on Mortality and Ventilator-Free Days in Patients Requiring Mechanical Ventilation With Sepsis: A Randomized Clinical Trial

Kawazoe Y. JAMA. Published online March 21, 2017

Clinical Question

  • In ventilated patients with sepsis, does a sedation strategy with dexmedetomidine compared with no dexmedetomidine improve mortality and number of ventilator-free days?


  • Randomised control trial
    • Permuted block randomisation – stratified by centre, presence of emergency surgery, COPD, soft tissue infection
    • Block size of 4 (physicians were not notified of this during study)
  • Open-label (Physicians non-blinded)
  • Blinding of statistician performing statistical analysis
  • Consecutive patients
  • Superiority trial
  • Intention to treat analysis
  • Primary outcomes changed during study period
    • Initially set as mortality and duration of mechanical ventilation
    • Changed in 05/2015 to mortality and 28 day ventilator free days as duration of ventilation highly influenced by mortality
  • Pre-determined weaning criteria
  • Power calculation
    • Based on results of MENDS trial with a predicted 60% mortality in the intervention group vs. 80% based in control group: 172 patients to achieve 80% power with a 2-sided α level of 0.05 for 28-day survival
    • Aimed to recruit 200 patients to allow for 15% drop-out/withdrawal rate
  • Statistical tests
    • Cox-proportional hazards model to estimate hazards ratio and 95% confidence intervals
    • Chi-squared or Fisher exact tests for other clinical outcomes between groups with categorical variables; t-tests or Wilcoxon rank sum tests for continuous variables


  • Eight ICUs in Japan
  • Data collected: February 2013 – January 2016


  • Inclusion criteria:
    • ≥20 years old
    • Patients with sepsis (defined as SIRS due to infection as per 1992 ACCP guidelines)
    • Mechanically ventilated for >24hrs (NIV or IPPV)
    • Acute pancreatitis included as ‘treatments are similar to those of septic patients’
  • Exclusion criteria:
    • Severe chronic liver disease (Child-Pugh Class B or C)
    • Acute myocardial infarction or severe heart failure (NYHA Class IV)
    • History of drug dependency, alcoholism, psychological illness, cognitive dysfunction
    • Pregnant or lactating
  • Participant numbers: 203 patients assessed for eligibility, one refusal and one excluded – 201 patients randomised and included in primary analysis
  • Comparing baseline characteristic of dexmedetomidine vs. control group:
    • Mean age: 68 vs. 69 years
    • Emergency surgery: 37% vs. 36%
    • Renal replacement therapy: 22% vs. 20%
    • Median APACHE II score: 23 vs. 22
    • Median SOFA score: 8 vs. 9
    • Site of infection
      • Abdomen 39% vs. 35%
      • Thorax 39% vs. 33%
      • Pancreatitis: 3% vs. 9%
      • Skin and soft tissue: 6% vs. 7%
    • No significant difference between groups in co-morbidities (chronic haemodialysis, chronic respiratory disease, chronic heart failure, liver insufficiency)


  • Patients received dexmedetomidine and analgesia continuously, and other sedatives added as needed
    • Dexmedetomidine titrated 0.1-0.7μg/kg/h
    • Fentanyl 0-5μg/kg/h as needed
    • Minimum propofol/midazolam as needed


  • Conventional propofol/midazolam based sedation without Dexmedetomidine
    • Fentanyl 0-5μg/kg/h as needed
    • Propofol titrated 0-3mg/kg/h
    • Midazolam titrated 0-0.15mg/kg/h

Treatments common to both intervention and control groups

  • Target sedation depth of RASS score 0 during day and -2 during night
  • Analgesia targets: Visual analogue scale goal ≤2, under deep sedation Behavioural Pain Score ≤4
  • Sedation maintained during mechanical ventilation or as required
  • Sedation protocols in compliance with ‘Clinical Practice Guidelines for the Sustained Use of Sedatives and Analgesics in the Critically Ill Adult’
  • Treatment protocol for sepsis based on national guidelines


  • Primary outcomes: comparing dexmedetomidine vs. control groups – no significant differences
    •  28 day mortality: 19% vs. 28%, p=0.14
      • Fragility index: -4 patients
    • Cumulative incidence of death at 28 days
      • 22.8% vs. 30.8%, p=0.20
      • Hazard ratio 0.69 (95% CI, 0.38-1.22; P=0.2)
    • Median number of 28-day ventilator-free days: 20 vs. 18 days, p=0.2
  • Secondary outcomes: comparing dexmedetomidine vs. control group
    • No significant differences in:
      • Median ICU length of stay (LOS): 7 vs. 8 days, p=0.43
      • Median Hospital LOS: 25.5 vs. 30 days, p = 0.27
      • Renal replacement therapy: 38% vs. 39%, p=0.93
      • Delirium (CAM-ICE positive): 44% vs. 45%, p=0.94
      • Daily SOFA score
      • Ventilator days: 6 vs. 6, p=0.64
      • Rate of adverse events:
        • Bradycardia: 7% vs. 2%, p=0.1
        • Acute coronary syndrome: 1% vs 1%
    • Rate of well-controlled sedation during ICU stay significantly higher in dexmedetomidine group than control group
      • Range: 17%-58% vs 20%-39%, p=0.01
    • Frequency and dose of propofol and midazolam were lower in the dexmedetomidine group than control group, but frequency and dose of fentanyl were not significantly different
  • Subgroup analyses:
    • Patients with APACHE II scores ≥23
      • Significantly lower mortality in dexmedetomidine group
      • Hazard Ratio, 0.39; 95% CI, 0.16-0.91; p=0.03

Authors’ Conclusions

  • In patients requiring mechanical ventilation, the use of dexmedetomidine compared with no dexmedetomidine did not result in significant improvements in mortality or ventilator-free days


  • Robust study design, multicentre trial
  • Attempts to reduce effects of bias from unblinding – use of a prearranged spontaneous breathing trial protocol
  • Whilst not an international study, the choice of pharmacological agents, scoring systems and weaning protocols was largely generalisable to most developed countries
  • Validated techniques for assessing sedation, delirium and pain
  • Registered with US National Institute of Health Clinical Trials Registry


  • Open-label study open to bias – efforts taken to reduce bias but assessment of delirium and sedation may be considered a largely subjective measure; with small block sizes clinicians may have been able to predict which group the patient was likely to be allocated to.
  • Lower dose of dexmedetomidine used in Japan due to limits placed by Japanese medical insurance
  • MENDS trial used to calculate sample size required for this study but important differences exist between DESIRE and MENDS: higher APACHE II scores in MENDS subgroup, lower dosage limit of dexemedetomidine and increased age of patients in DESIRE
    • The study was underpowered to detect an absolute difference in mortality of 9%, which is still clinically significant
  • Study had two primary outcomes. Did not perform power calculation for ventilator free days; and primary outcome changed during study period.
  • Six patients in control group received dexmedetomidine at physicians discretion
  • Measurement of ventilator-free days is arguably not a patient-centered outcome and no long-term patient outcomes recorded
  • Although not statistically different there were some baseline differences between groups. The numbers of patients with pancreatitis were 3% vs. 9%. As pancreatitis is not an infection, including it as part of a sepsis trial may not have been appropriate due to differences in mortality and length of ventilation.
  •  No mention of how many patients received NIV or IPPV in each group

The Bottom Line

  • In ventilated patients with sepsis or pancreatitis, this study fails to demonstrate a significant impact of dexmedetomidine on 28-day survival or ventilator-free days but was likely underpowered for mortality. Further well-powered studies with clear patient-centered outcomes are required before dexmedetomidine should be considered as the gold standard sedative for ventilated patients with sepsis on ICU.

External Links


Summary author: Gary Misselbrook
Summary date: 29th March 2017
Peer-review editor: @davidslessor


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.