MACQUARIE NEUROSURGERY JOURNAL CLUB
14/04/16
14/04/16
Dr Michael Mulcahy
MACQUARIE NEUROSURGERY
JOURNAL CLUB
Andrews PJ, Sinclair HL, Rodriguez A, Harris BA, Battison CG, Rhodes JK, Murray GD;
Eurotherm3235 Trial Collaborators.
Hypothermia for Intracranial Hypertension after Traumatic
Brain Injury
The New England Journal of Medicine, Dec 2015, volume 373, issue 25, pp2403-12.
Authors
• Chief investigator: Professor Peter Andrews, Western General Hospital,
Edinburgh
• 6 co-authors and 154 Eurotherm3235 collaborators
• Funded by National Institute for Health Research Health Technology
Assessment program (UK Department of Health)
• Pilot phase funded by The European Society
of Intensive Care Medicine
• Trial sponsors: University of Edinburgh and
NHS Lothian
Disclosures
•International committee of medical journal editors (ISMJE) disclosure form
•P.Andrews reported lecturing fees from Bard and Integra LifeSciences
•J.Rhodes reported lecturing fees from Bard
Study Relevance
• TBI is underrepresented in medical research compared with other health
problems
• There is little evidence supporting the common interventions to treat
intracranial hypertension
• General observation of ICP reduction associated with hypothermia
• Trend towards benefit for hypothermia in low quality trials, but…
Study Relevance
Study Relevance
•Meta-analyses have been inconsistent due to variability in the RCTs
•Examples:
- presence of intracranial hypertension
- ICP or CT as inclusion criteria
- duration of cooling
- rate of re-warming
- use of barbiturates
Originality
•Raised ICP only
•Duration of hypothermia
•Rapid induction of hypothermia
•Slow rewarming
Study Hypothesis
• Patients treated with therapeutic hypothermia (32-35o
C) have reduced morbidity
and mortality rates compared to those receiving standard care alone after TBI.
• Primary outcome: GOSE score at 6 months
• Secondary outcomes:
• 6 month mortality rate
• ICP control
• Incidence of pneumonia
• Length of stay in ICU and hospital
• MOHS
• Correlation between MOHS at discharge and GOSE at 6 months
Study Protocol
Study Design
• RCT
• Interventional
• Pragmatic
• Multi-centre (25 UK centres, 39 elsewhere)
• Open-label
Study Design
Study Design
Hypothermia:
• initiated with 20-30ml/kg refrigerated 0.9% saline given intravenously
• maintained using cooling systems available
• maintained for at least 48hours
• Depth (from 32-35 degrees) is guided by ICP
• If stage 3 treatments required, hypothermia is terminated
Follow up:
• Participants are sent the GOSE questionnaire by post at 6 months
• Letter is also sent to the person who gave consent for the trial
Study Design
Inclusion criteria:
• ? to 65
• primary closed TBI
• raised ICP >20mmHg for >5min after first line treatments with no obvious
reversible cause
• <72 hours from initial head injury
• cooling devices or techniques available for >48hours
• core temperature >36o
C at time of randomisation
• abnormal CT
Study Design
Inclusion criteria:
• ? to 65
• primary closed TBI
• raised ICP >20mmHg for >5min after first line treatments with no obvious
reversible cause
• <72 hours from initial head injury
• cooling devices or techniques available for >48hours
• core temperature >36o
C at time of randomisation
• abnormal CT
Study Design
Exclusion criteria
• already receiving therapeutic hypothermia
• administration of barbiturate prior to randomisation
• unlikely to survive for the next 24 hours
• temperature <34o
C
• pregnancy
Population
Internal Validity
Sample size/power
• Aimed to detect an absolute decrease in poor outcome of 7%
• 80% power
• Two groups of 815 would detect a decrease in poor outcome from 60% to
53%
• Aimed for two groups of 900 (loss to follow up)
• However…
Internal Validity
Sample size/power
• After the pilot, this was revised to a total sample size of 600
• Using a dichotomous analysis of GOSE, this would give 81% power to
detect an absolute reduction of 12% (60 to 48%)
• Using an ordinal analysis of GOSE scores with covariate adjustment leads
to a trial of 600 being equivalent to 1000 that assessed a binary outcome
• This will give 80% power at the 5% significance (two sided) to detect an
absolute reduction in poor outcome of 9% (from 60% to 51%)
Internal Validity
Randomisation
• Used either a central internet based randomisation service, or a telephone
randomisation service
• Minimisation, stratified by:
• trial centre
• age < or > 45
• post-resuscitation GCS motor component 1-2 or 3-6
• time from injury < or > 12 hours
• pupils: both reacting or 1 or neither reacting
Internal Validity
Blinding
• Investigators were not blinded
• Primary outcome data assessment was blinded
Internal Validity
Statistical Analysis
• SAS software, version 9.3
• Intention-to-treat analysis
• Ordinal logistic regression to analyse the 6 month GOSE score
• Covariates: age; post-resuscitation GCS; time from injury (<12 or >12);
pupillary response
• The eight points of the GOSE were collapsed to 6 (death was pooled with
a vegetative state and lower severe disability)
• Strict p<0.01 was used for covariate analysis
Results
Results (Primary Outcome)
• At 6 months, GOSE scores were unfavourable for the hypothermia group
• The trial was stopped early by the independent steering committee
Results
Results
Adverse events and mortality
•33 events in hypothermia group, 10 events in control group
Results
Adverse events
•33 events in hypothermia group, 10 events in control group
Unadjusted hazard ratio for death at 6 months: OR 1.45 (1.01 to 2.10) p=0.047Unadjusted hazard ratio for death at 6 months: OR 1.45 (1.01 to 2.10) p=0.047
Results
Intracranial pressure
• Mean daily ICP similar
• First occurrence of failure to
control ICP: 57 cases in
hypothermia group, 84 in control
group
• Therefore, more frequent use of
stage 3 treatments: 102 of 189
(54.0%) v. 84 of 192 (43.8%)
• No difference in decompressive
craniotomy
External Validity/generalisability
• Are the results of this trial important?
• Not representative of the TBI population.
• Feasible intervention, but utilised appropriately
• TBI up to 10 days?
• Trial stopped early
• Intention to treat
• 6 month follow up
2498 screened
1742 raised ICP
387 included
Discussion/Conclusion
• Hypothermia did not result in better outcomes
• The trial was stopped early due to safety concerns
• Hypothermia plus standard care successfully controlled ICP. This was not
statistically significant when compared to standard care alone.
• Unrepresentative population
• Difficult to isolate the impact of one intervention in a pragmatic trial
• Other stage 2 interventions were not measured, and benefits or harms of these
are unknown
• Hypothermia may have an effect on patients with higher ICP
• Barbiturates may be a confounding factor
• Non-blinding may lead to more adverse event reporting in the intervention group

Jc eurotherm3235 ppt

  • 1.
    MACQUARIE NEUROSURGERY JOURNALCLUB 14/04/16 14/04/16 Dr Michael Mulcahy
  • 2.
    MACQUARIE NEUROSURGERY JOURNAL CLUB AndrewsPJ, Sinclair HL, Rodriguez A, Harris BA, Battison CG, Rhodes JK, Murray GD; Eurotherm3235 Trial Collaborators. Hypothermia for Intracranial Hypertension after Traumatic Brain Injury The New England Journal of Medicine, Dec 2015, volume 373, issue 25, pp2403-12.
  • 3.
    Authors • Chief investigator:Professor Peter Andrews, Western General Hospital, Edinburgh • 6 co-authors and 154 Eurotherm3235 collaborators • Funded by National Institute for Health Research Health Technology Assessment program (UK Department of Health) • Pilot phase funded by The European Society of Intensive Care Medicine • Trial sponsors: University of Edinburgh and NHS Lothian
  • 4.
    Disclosures •International committee ofmedical journal editors (ISMJE) disclosure form •P.Andrews reported lecturing fees from Bard and Integra LifeSciences •J.Rhodes reported lecturing fees from Bard
  • 5.
    Study Relevance • TBIis underrepresented in medical research compared with other health problems • There is little evidence supporting the common interventions to treat intracranial hypertension • General observation of ICP reduction associated with hypothermia • Trend towards benefit for hypothermia in low quality trials, but…
  • 6.
  • 7.
    Study Relevance •Meta-analyses havebeen inconsistent due to variability in the RCTs •Examples: - presence of intracranial hypertension - ICP or CT as inclusion criteria - duration of cooling - rate of re-warming - use of barbiturates
  • 8.
    Originality •Raised ICP only •Durationof hypothermia •Rapid induction of hypothermia •Slow rewarming
  • 9.
    Study Hypothesis • Patientstreated with therapeutic hypothermia (32-35o C) have reduced morbidity and mortality rates compared to those receiving standard care alone after TBI. • Primary outcome: GOSE score at 6 months • Secondary outcomes: • 6 month mortality rate • ICP control • Incidence of pneumonia • Length of stay in ICU and hospital • MOHS • Correlation between MOHS at discharge and GOSE at 6 months
  • 10.
  • 11.
    Study Design • RCT •Interventional • Pragmatic • Multi-centre (25 UK centres, 39 elsewhere) • Open-label
  • 12.
  • 13.
    Study Design Hypothermia: • initiatedwith 20-30ml/kg refrigerated 0.9% saline given intravenously • maintained using cooling systems available • maintained for at least 48hours • Depth (from 32-35 degrees) is guided by ICP • If stage 3 treatments required, hypothermia is terminated Follow up: • Participants are sent the GOSE questionnaire by post at 6 months • Letter is also sent to the person who gave consent for the trial
  • 14.
    Study Design Inclusion criteria: •? to 65 • primary closed TBI • raised ICP >20mmHg for >5min after first line treatments with no obvious reversible cause • <72 hours from initial head injury • cooling devices or techniques available for >48hours • core temperature >36o C at time of randomisation • abnormal CT
  • 15.
    Study Design Inclusion criteria: •? to 65 • primary closed TBI • raised ICP >20mmHg for >5min after first line treatments with no obvious reversible cause • <72 hours from initial head injury • cooling devices or techniques available for >48hours • core temperature >36o C at time of randomisation • abnormal CT
  • 16.
    Study Design Exclusion criteria •already receiving therapeutic hypothermia • administration of barbiturate prior to randomisation • unlikely to survive for the next 24 hours • temperature <34o C • pregnancy
  • 17.
  • 18.
    Internal Validity Sample size/power •Aimed to detect an absolute decrease in poor outcome of 7% • 80% power • Two groups of 815 would detect a decrease in poor outcome from 60% to 53% • Aimed for two groups of 900 (loss to follow up) • However…
  • 19.
    Internal Validity Sample size/power •After the pilot, this was revised to a total sample size of 600 • Using a dichotomous analysis of GOSE, this would give 81% power to detect an absolute reduction of 12% (60 to 48%) • Using an ordinal analysis of GOSE scores with covariate adjustment leads to a trial of 600 being equivalent to 1000 that assessed a binary outcome • This will give 80% power at the 5% significance (two sided) to detect an absolute reduction in poor outcome of 9% (from 60% to 51%)
  • 20.
    Internal Validity Randomisation • Usedeither a central internet based randomisation service, or a telephone randomisation service • Minimisation, stratified by: • trial centre • age < or > 45 • post-resuscitation GCS motor component 1-2 or 3-6 • time from injury < or > 12 hours • pupils: both reacting or 1 or neither reacting
  • 21.
    Internal Validity Blinding • Investigatorswere not blinded • Primary outcome data assessment was blinded
  • 22.
    Internal Validity Statistical Analysis •SAS software, version 9.3 • Intention-to-treat analysis • Ordinal logistic regression to analyse the 6 month GOSE score • Covariates: age; post-resuscitation GCS; time from injury (<12 or >12); pupillary response • The eight points of the GOSE were collapsed to 6 (death was pooled with a vegetative state and lower severe disability) • Strict p<0.01 was used for covariate analysis
  • 23.
  • 24.
    Results (Primary Outcome) •At 6 months, GOSE scores were unfavourable for the hypothermia group • The trial was stopped early by the independent steering committee
  • 25.
  • 26.
    Results Adverse events andmortality •33 events in hypothermia group, 10 events in control group
  • 27.
    Results Adverse events •33 eventsin hypothermia group, 10 events in control group Unadjusted hazard ratio for death at 6 months: OR 1.45 (1.01 to 2.10) p=0.047Unadjusted hazard ratio for death at 6 months: OR 1.45 (1.01 to 2.10) p=0.047
  • 28.
    Results Intracranial pressure • Meandaily ICP similar • First occurrence of failure to control ICP: 57 cases in hypothermia group, 84 in control group • Therefore, more frequent use of stage 3 treatments: 102 of 189 (54.0%) v. 84 of 192 (43.8%) • No difference in decompressive craniotomy
  • 29.
    External Validity/generalisability • Arethe results of this trial important? • Not representative of the TBI population. • Feasible intervention, but utilised appropriately • TBI up to 10 days? • Trial stopped early • Intention to treat • 6 month follow up 2498 screened 1742 raised ICP 387 included
  • 30.
    Discussion/Conclusion • Hypothermia didnot result in better outcomes • The trial was stopped early due to safety concerns • Hypothermia plus standard care successfully controlled ICP. This was not statistically significant when compared to standard care alone. • Unrepresentative population • Difficult to isolate the impact of one intervention in a pragmatic trial • Other stage 2 interventions were not measured, and benefits or harms of these are unknown • Hypothermia may have an effect on patients with higher ICP • Barbiturates may be a confounding factor • Non-blinding may lead to more adverse event reporting in the intervention group

Editor's Notes

  • #4 UoE provided research governance
  • #5 no other potential conflict of interest reported
  • #6 Due to heterogeneity of patients, long approval times, cost, low screen-recruitment rate Brain Trauma Foundation guidelines on stages of intervention ICP not necessarily raised
  • #7 More rigorous randomised trials have shown unfavourable outcome NEJM 2001, no difference Lancet 2011, terminated due to futility
  • #8 2009 Cochrane review: 23 RCTs. Only meta-analysis that looked only at ICH trials. Low quality trials tending to show benefit from hypothermia. In 9 trials considered good quality there was no decrease in the likelihood of death with hypothermia, but this finding was not statistically significant.
  • #9 Trend in this meta-analyses towards improved outcome when hypothermia &amp;gt;48hrs Hypothermia within 30 minutes Re-warming rate 1°C/4 hours (0.25°C per hour) until core temperature is ≥ 36°C Also some data that slower rewarming and rapid induction improve outcome
  • #10 Extended Glasgow Outcome Scale (more sensitive to differences in poor categories, which is typical TBI patient) Modified Oxford Handicap Scale at one month (or at discharge/death)
  • #11 Published in 2011. Pilot published 2013. Pilot ran from Jan 2009 to Aug 2011. “Trial was conducted and reported with fidelity to the protocol” Along with this were guidelines on induction and maintenance of hypothermia, re-warming, and detection and treatment of shivering.
  • #12 Pragmatic, focus on outcomes rather than mechanistic pathways. Tests effectiveness in clinical practice. Good at guidance current practice but not necessarily the best at isolating a single treatment effect Possible measurement bias 53% from UK. Most other centres from Europe, 2 from the middle east
  • #13 Point to highlight here is the stage 2 treatments only if needed in hypothermia group. Also, not clear in design how much hypertonic or mannitol given and to who etc. How was ICP measured?
  • #14 Stage 2 treatments added only if needed Lowered by 0.5degree increments until ICP is under 20mmHg Coordinating centre mails the GOSE. GP is phoned prior to this to inform them of trial involvement. If no response, the GOS is sent. If no response they are telephoned.
  • #15 ICP inclusion based on J Neurotrauma guidelines from 2007 CT - haematoma, contusion, swelling, herniation or compressed basal cisterns
  • #16 Changed after pilot study to include older patients and allow for evolving brain swelling
  • #17 Opinion of ICU or neurosurgery consultant
  • #18 Pilot - 16% of screened patients were eligible Pre-trial, they thought they would get 50% inclusion based on previous studies As a result of this data, inclusion criteria was changed so there would be no upper age limit and recruitment was extended to 10 days. Baseline characteristics were similar for the two groups
  • #19 Other TBI studies use 10% which they thought was unrealistic due to heterogeneity of TBI patients.
  • #20 Changed sample size due to homogeneity (all had raised ICP) and successful cooling was achieved. This is the subgroup of patients that may benefit from hypothermia, and in some of the meta-analyses had better outcome.
  • #21 Depending on technologic availability Minimisation to stop imbalance in the subgroups Minimisation factors reasonable, ?age Time from injury, early hypothermia supposedly better
  • #22 Obvious when hypothermia used due to equipment, temp, bloods etc.
  • #23 Primary analysis based on key baseline covariates. So the analysis would not favour intervention that reduced mortality by increasing severe disability.
  • #24 CONSORT diagram Largest proportion excluded due to normal ICP (30%) 7% unlikely to survive 5% already were receiving hypothermia 2498 - 756 = 1742, minus 387 leaves 1355 (22%) included still a large amount excluded leading to possible selection bias.
  • #25 Favourable outcomes (GOSE 5-8, or moderate disability to good recovery) occurred in 25.7% (49 of 191) of the hypothermia group and 36.5% (69 of 189) of the control group. Odds ratio 1.53 (&amp;lt;1 would confer benefit to hypothermia group) The odds ratio was 1.55 without adjustment for the covariates Secondary outcomes showed no significant interaction, except the risk of death…
  • #26 Forest plot of the pre-specified subgroups. No significant interaction effect between the intervention and pre specified subgroups
  • #27 Adverse events collected: bleeding, cardiovascular instability, thermal burns and CPP&amp;lt;50mmHg. Didn’t collect other data as adverse events are ‘expected’. Adverse events weren’t a predefined secondary outcome. Mortality at 6 months was.
  • #28 At 28 days, there were 60 (30.8)% dead in hypothermia, and 42 (21.9%) in control Kaplan-Meier curve: Risk of death was higher in the hypothermia group
  • #29 Analysed post-hoc using a linear model, with study days as repeated measurements with a compound-asymmetry covariance matrix. More frequent in control group 57 of 192 and 84 of 189 (data actually not presented) Whilst barbiturate therapy was used more often (41 v. 20), decompressive craniotomy was not (27 each group) Core temp significantly lower
  • #30 Yes, common condition with poor outcome 2498 - 756 = 1742, minus 387 leaves 1355 (22% or 15% - which is low in a pragmatic study) Hypothermia after stage 2 therapies, rather than before TBI up to 10 days, not representative and may have different effects Potential bias in stopping trial early intention to treat, more reliable as it accounts for real-work drop-out, non-compliance etc. 6 month follow up reasonable, and what GOSE was designed for
  • #31 Especially in critically ill patients with complex medical problems Concerns that hypothermia used too early in patient’s not requiring it. Enough evidence to not use hypothermia in this specific way