Strom

Strom: A protocol of no sedation for critically ill patients receiving mechanical ventilation: a randomised trial

Strom et al. Lancet 2010; 375:475-480. DOI:10.1016/S0140- 6736(09)62072-9

Clinical Question

  • In adult patients undergoing mechanical ventilation, does a strategy involving no continuous sedation, compared to a strategy of continuous sedation by infusion with a daily interruption, reduce the number of days of mechanical ventilation in the first 4 weeks?

Design

  • Randomised, controlled clinical trial
  • Adequately concealed allocation by opaque envelopes
  • No attempt at blinding of clinicians, patients, investigators or data analysers due to practical limitations
  • Modified intention-to-treat:
    • Patients excluded if extubated or died within first 48 hours
  • Powered at 80% with accepted alpha error of 5%, to detect an absolute reduction of 1.5 ventilator-days from a baseline of 5–7.5 days (based on data from Kress sedation study), if 100 patients were recruited

Setting

  • Single-centre combined medical–surgical ICU in Denmark with 1:1 nursing ratio and ICM specialist doctor at all times, with existing practice of ‘no sedation’
  • April 2007 to December 2008

Population

  • Inclusion: patients that were expected to need mechanical ventilation for more than 24 hours.
  • Exclusion: under 18 years old, pregnancy, clinical reasons for sedation (e.g. raised intracranial pressure, post-cardiac arrest, status epilepticus), ready for weaning and extubation.
  • 428 patients assessed, of which 140 randomised
    • 67% male
    • 66 years old (median)
    • APACHE II 26 (median)

Intervention

  • No sedation
    • Intravenous morphine 2.5 or 5 mg bolus as needed
    • Additional staff to “verbally comfort and reassure the patient”
    • Haloperidol for delirium
    • Propofol for 6 hours if agitation persists despite these previous measures
    • If 3 periods of propofol required, patient managed as per ‘control group’ but analysed as ‘intervention group’ (that is, intention-to-treat analysis)

Control

  • Titrated sedation to Ramsay score of 3–4
    • Propofol 2% infused as required
    • Changed to midazolam 1 mg/ml after 48 hours of sedation
    • Intravenous morphine 2.5 or 5 mg bolus as needed
  • Daily sedation breaks
    • Nurse administered interruption in sedating drugs
    • Assessed as awake if performing 3 of 4 functions: eyes open, looking at clinician, squeezing hands, sticking out tongue.
    • Sedative infusion restarted at half previous rate and titrated to same Ramsay score
  • Sedation was stopped if FiO2 ≤ 40% and PEEP ≤ 5cmH2O, and restarted if the ventilatory support increased again.
Both groups were mobilised daily. Pressure supported ventilation was the standard. Both groups were weaned and extubated once pre-defined (and fairly standard) criteria were met. Physical restraints were never used.

Outcome

In words

  • The primary outcome, which was the number of days without mechanical ventilation in a 28-day period beginning at intubation, favoured the intervention of ‘no sedation’.
    • The difference between the mean of each group was 4.2 days (95% CI 0.3–8.1).
  • The secondary outcomes either favoured the intervention of ‘no sedation’ (some with statistical significance and some without), or showed no difference, with the exception of the incidence of delirium:
    • ICU length of stay – statistically significant mean difference of 9.7 days
    • Hospital length of stay – statistically significant mean difference of 24 days
    • In hospital mortality – trend toward better outcome with ‘no sedation’ but not statistically significant
    • Brain radio-imaging – no difference in need
    • Accidental extubation – no difference in occurrence (n=7 [13%] vs n=6 [10%], p=0.69)
    • Ventilator associated pneumonia (defined as CXR changes and one of fever,  high or low white cell count, or purulent sputum) – no difference in occurrence
    • Re-intubation – no difference in need
    • Delirium – greater incidence in the ‘no sedation’ group (n=11 [20%] vs n=4 [7%], p=0.04). This may represent difficulty to detect delirium in sedated patients, rather than a true difference in incidence.

In numbers

Primary Outcome
Measure No Sedation Sedation MD 95% CI p
Ventilator-free days
Mean (SD)
13.8 (11.0) 9.6 (10.0) 4.2 days 0.3–8.1 0.0191
‘Ventilator-free days’ = number of days from intubation to day 28 where mechanical ventilation was not requires; MD = mean difference; SD = standard deviation; CI = confidence interval; p = p-value
Secondary Outcomes
Measure No Sedation Sedation HR AD p
ICU LOS
(median days)
13.1 22.8 1.86
(95% CI 1.05–3.23)
9.7 0.0316
Hospital LOS
(median days)
34 58 n/a 24 0.0039
Measure No Sedation Sedation OR ARR NNT p
In-Hospital Mortality 36% 47% 0.65
(95% CI 0.31–1.39)
10.2 10 0.27
LOS = Length of Stay; HR = Hazard Ratio; AD = Absolute Difference; OR = Odds Ratio; ARR = Absolute Rate Reduction; NNT/NNH = Number needed-to-treat

Authors’ Conclusions

  • Results from this single-centre study suggest that a strategy of no sedation is promising, but a multi centre trial is needed to show that the benefits of this strategy can be reproduced in other facilities.

Strengths

  • Authors designed trial to question their own established practice.
  • Control group were treated as per previously published protocol by Kress, allowing direct comparison.

Weaknesses

  • There was a baseline difference in patient gender, with a greater proportion of women being allocated to the control group. Although not reported, this difference is statistically significant (p=0.05 by Fisher’s exact test). Given the adequate concealment of allocation, this is likely to have occurred by an outside chance. In my opinion, it is unlikely that this gender difference will have affected the observed primary or secondary outcomes, but that cannot be said with certainty.
  • Given the intervention of ‘no sedation’ was the institution’s normal practice, and the complete lack of any blinding, there is an inherent bias towards finding a larger difference in favour of the intervention (i.e. a more positive outcome). This also reduces the external validity, as there may be a learning curve associated with using ‘no sedation’.
  • 19% of the randomised patients were excluded from data analysis (modified intention-to-treat analysis) as they were extubated or dead within 48 hours. Whilst this is a common method (see Kress for example), I would like to see mortality and length of stay data including these patients. Unbiased and fair randomisation should minimise any bias from this modified analysis.
  • 10 (18%) patients in the ‘no sedation’ group deviated from protocol and received continuous sedation. This will reduce the observed difference, and highlights the difficulty with applying this ‘no sedation’ strategy despite their expertise and staffing.
  • The outcome measures were reported as corrected for baseline characteristics using multiple linear regression statistics. Although the uncorrected findings were still statistically significant (p=0.035), it may have been better to report all raw data then a selection of corrected data.
  • Non-standard definition of ventilator-associated pneumonia.
  • Additional staffing was required to support 20% of the ‘no sedation’ patients, compared to 5% of the ‘sedation’ patients. This additional resource may outweigh any benefit from a reduced length of stay. The 24 hour availability of an ICM specialist doctor may also reduce the generalisability to units with less staffing.

The Bottom Line

  • This study should make clinicians question the dogma of sedation in ICUs, but clinical practice should not change immediately because of this study alone.
  • A strategy of ‘no sedation’ may be feasible in ICUs, but may require additional staffing.
  • Further multi-centre trials are required with patient-centred outcomes and cost-analyses.

External Links

Metadata

Summary author: @DuncanChambler
Summary date: 10th July 2014
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.