October 18, 2019
MPOG Retreat
Pragmatic cluster-randomized trials in
anesthesia: opportunities for MPOG
Jessica Spence, MD, FRCPC
@jess_spence13
Disclosures
1. I have received career and training support from CIHR
and the Michael G DeGroote Foundation.
2. I hold active research funding from the McMaster
Surgical Association (MSA), Physicians’ Services
Incorporated (PSI) and CIHR.
Objectives
1. To define pragmatic cluster-randomized trials and
discuss when this design is appropriate to answer
clinical questions in anesthesiology.
2. To discuss the steps involved in designing and
conducting a pragmatic cluster-randomized trial in
anesthesiology, using examples from the
Benzodiazepine-free cardiac anesthesia for the
reduction of postoperative delirium (B-Free) trial.
Pragmatic vs Explanatory Trials
• Pragmatic = Effectiveness
• Explanatory = Efficacy
• Efficacy trials determine impact of an intervention in selected
populations under optimal conditions.
• Effectiveness trials determine impact of an intervention in every day
clinical practice.
• Wide variability in routine practice in anesthesiology, e.g.
withholding/giving home medications, FiO2, fluid administration, etc.
• Pragmatic trials are required to answer these questions
Pragmatic vs Explanatory Continuum
rwe-navigator.eu
PRECIS-2 wheel
6
Cluster-randomized trials
 Used to evaluate broad-based approaches applied at the level of
a system
 Traditional RCTs randomize individual patients
 In cluster RCTs, the unit of randomization is a group of patients
 Examples of clusters: hospitals, clinics, ICUs, schools
 Examples of interventions: crystalloid used, policy-based care,
small tubes for blood testing, educational approach
Why use cluster-randomization?
 Intervention is applied at the level of the cluster
 The research question seeks to determine the impact of
the intervention at the level of the cluster
 Takes into account knowledge translation, patient selection
Cluster randomized
Randomized cluster crossover
Stepped wedge design
Ethics of cluster-randomized trials
• Often not possible for a patient to choose to avoid or consent to
an intervention being applied at the level of the health system
• As a result, cluster randomized trials present unique challenges
to researchers and ethics boards
• Modifications to traditional consent processes may be required
• Emerging area; important to have bioethicist involved in trial
development
Ethics of cluster-randomized trials
 For waiver of individual patient consent FDA/TCPS2 requires:
 Altered consent required to answer question
 Research involves minimal risk
 Lack of a priori consent will not adversely affect participant welfare
 Information provided to participants when possible
 Benefits outweigh risks of not obtaining a priori consent
Statistical Considerations for Cluster Trials
 Intracluster correlation (ICC) = measure of variation of outcomes
between centres
 High ICC = need a larger number of clusters
 In general, number of clusters more important than number of
patients
 In cluster crossover and stepped wedge trials, also need to take into
account the interperiod correlation (IPC)
 Statistical representation of period effects
 Complex; key to have a biostatistician on your team
So… you want to design a
cluster-randomized trial…
Step 1: Review the literature
Delirium in the CVICU
 Delirium is a serious problem for patients!
 Incidence 15-30% after cardiac surgery
 Relative risk 8.3 compared to noncardiac surgery
 Associated with increased cognitive decline, functional decline,
institutional discharge, and death
 Modifiable risks: polypharmacy, psychoactive medications,
physical restraints
 Benzodiazepines recurring theme
Arguments For and Against Benzodiazepines
 PRO: minimal effect on hemodynamics, anxiolytic, possible
protection against intraoperative awareness
 Level of evidence: Very low quality; 291 patients
 CON: delirium, cognitive decline
 Level of evidence: Low quality; 1007 patients
Step 2: Establish equipoise
19
Equipoise in benzodiazepine practice
Step 3: Develop your question
21
What kind of trial?
• Cardiac surgery provided in specialized institutions using standardized
policies to optimize outcomes.
• In the majority of cases, anesthesiologists choose whether or not
benzodiazepines based on personal preference, rather than because of
patient considerations
• The most logical way to test impact of limited approach to benzodiazepine
use during cardiac surgery is to standardize care at the level of an
institution.
• This can be done by assessing change in institutional policy as it may
occur in every day clinical practice.
Does an institutional policy of limited
intraoperative benzodiazepine use reduce
incidence of delirium after cardiac surgery?
Study question
Study Interventions
Limited Benzodiazepine Policy
• No routine administration of any
benzodiazepines
• Accepted administration of
benzodiazepines according to
anesthesiologist discretion
• (Expected administration of
benzodiazepines to 10% of patients)
Liberal Benzodiazepine Policy
• Administration of 0.03 mg/kg
Midazolam equivalent
• Accepted avoidance of
benzodiazepine administration
according to anesthesiologist
discretion
• (Expected avoidance of benzodiazepines
to 10% of patients)
Site Considerations
 Hospitals randomized to a policy for 4 weeks, then crossover
 Require cardiac anesthesia group to agree to policy
implementation
 All other activities are per standard of care
Study Outcomes
Primary Outcome
Incidence of in-hospital
delirium
Secondary Outcomes
ICU LOS
Hospital LOS
In-hospital mortality
Study question
Step 4: Consider ethical issues
28
ETHICAL ISSUES
 The intervention being studied within B-Free is being applied at
the level of the hospital.
 Clinicians determine whether or not it filters down to individual
patients
 Opted to use a waived consent model
 Engaged a clinical bioethicist and intensivist to be a member of
our steering committee
ETHICAL ISSUES
 For waiver of individual patient consent FDA/TCPS2 requires:
 Altered consent required to answer question
 Can’t understand impact of broad-based policy change without it
 Research involves minimal risk
 Clinical equipoise; both approaches used
 Lack of a priori consent will not adversely affect participant welfare
 Information provided to participants when possible
 Letter of information provided; participants may withdraw data
 Benefits outweigh risks of not obtaining a priori consent
ETHICAL ISSUES
 Central ethics approval obtained in Ontario, BC, Alberta,
Québec
 Site level approval obtained in Winnipeg, Halifax
 Site level provisional approval obtained in Saskatchewan
= Ethics approval at 19/20 sites
Step 5: Calculate your sample
size
32
B-Free Sample Size
 Population: All cardiac surgery patients
 Treatment effect: 15% relative risk reduction
 Sample Size
 16 sites with average case volume 1,000 patients/year
 ~ 16,000 patients
 Study Duration: 12 x 4 week crossover periods = ~ 1year
80% power to detect 15% relative risk reduction
ICC = 0.02 and IPC = 0.5*ICC
Step 6: Decide how you will
collect data
Data Collection
 As much as possible through hospital administrative
electronic records
 Cumulative data transfers every 2 months
 No data manipulation required
 Minimal data entry (may vary by site)
 Delirium scales
 Perioperative medications
Step 7: Establish feasibility.
Conduct a pilot study!
Pilot Trial Objectives
Evaluate Feasibility
1. ≥ 80% adherence to each treatment policy
2. Complete outcomes collection using nurse-administered
delirium scales
1 delirium scale in 95% of patients
1 delirium scale per ICU day in 90% of patients
3. Awareness screening in 1 site
Incidence not greater than upper 95% CI (i.e. 2%) during limited benzodiazepine period
Pilot Trial Methods
 2 centres (Hamilton and Winnipeg)
 4 x 4 week crossover periods
 Randomized to first period then alternate
1. 88.6% adherence to limited benzo policy; 92.3% adherence
to liberal benzo policy
Feasibility Outcomes
2. 96% with one delirium scale; 92% with one delirium
scale per ICU day
3. 1/521 (0.2%) cases of awareness
-managed during limited policy but received benzodiazepine
Other Outcomes
40
 Both REBs approved alterations to individual consent
 Enrolled 800 patients over 8 periods
Full trial is feasible
Can engage cardiac anesthesia providers
Acceptable outcomes collection using nurse-administered CAMs
No increased risk of intraoperative awareness
Conclusions
Step 8: Collaborate!
Anesthesia research in Canada
43
Step 9: Get funded!
 CIHR funded; Ranked 1
st
out of 43 applications
 Entire trial to cost $469,000
Funding Plan
Step 10: Get it done!
46
Questions?
Proposed Study Timeline
