Transcript file - PREPARE Education Virtual Learning Centre

Fusing ICU Care & Clinical Research in
Severe Community-Acquired Pneumonia
Jean-Daniel Chiche, MD PhD
MICU & Dept ‘Infection, Immunity & Inflammation’
Hôpital Cochin & Institut Cochin, Paris-F
Evidence-based medicine
“Doctors must base what
they do on randomized
clinical trials (RCTs)”
Archie Cochrane
How we know something works ...
Patients with disease X
The good news …

We now test many ideas with RCTs
 37,000 started in 2010 …
 All FDA drug and device approvals

We now conduct RCTs very well
 Methodologic conduct
 Ethical oversight
 Reporting
Good News, But…

RCTs are too narrow
 Cherry-picked population; not real life

RCTs are too broad
 No data on treatment effects across patients

No comparative effectiveness
 Rx A vs. B is not very helpful
 What about A vs. B, vs. C, vs. D,…


Depending on whether I give E or F …
I just want the answer
 Don’t want my patient to be a guinea pig…
Parallel universes …
Clinical care
Clinical research
Enter the era of ‘Big Data’

Integration of ‘deep’ personalized data




Causal inferences on optimal care
Broad – ‘real-world’ practice
Narrow – ‘personal’ estimates
Comparative – considers all options


Vanderbilt-IBM ‘BioVU’ initiative
‘Live’ presentation of information at time of clinical
decision-making
 ‘Just-in-time’ cohort study in EHR
 No guinea pigs…
PRECISION MEDICINE IN THE ERA OF BIG DAT
Molecular Systems Biology (2012) 8: 612
Feature
‘Big Data’
Analytics
Leverage the EHR
✔
Low incremental costs
✔
Real-world ‘effectiveness’
✔
Consider multiple therapies
✔
‘Personalized’ estimates
✔
Offer ‘live’ tailored options
✔
Robust causal inference
✗
Point-of-care (POC) Clinical Trials
A clinical moment in the EHR ‘alerts’ the clinical trial
machinery
 VA EHR system

 In-patient diabetics with poor glucose control
 When physician placed insulin order in CPOE system, an
opportunity to randomize…

Sliding scale vs weight-based algorithm
• Fiore et al. Clinical Trials 2011

Targeting the large ‘pragmatic’ trial arena
 2 thiazide diuretics in >13k high BP patients (NCT02185417)
 2 aspirin doses in 20k CVD patients (ADAPTABLE) (PCORI)
‘Big Data’
Analytics
POC
Trials
Leverage the EHR
✔
✔
Low incremental costs
✔
✔
Real-world ‘effectiveness’
✔
✔
Consider multiple therapies
✔
✗/✔
‘Personalized’ estimates
✔
Offer ‘live’ tailored options
✔
Robust causal inference
✗
✗
✗
Feature
✔
Platform Trials

Adaptive trials






Focus on disease, not a particular Rx
Multiple interventions (arms)
‘Perpetual’ enrollment
Often based on Bayes’ theorem
Tailor choices over time
Focus on pre-approval space
 Emphasis on efficiency with (very) small sample sizes
 Different therapies ‘graduate’ to next phase while trial
continues
Global adaptive trials: the principles

Simulation & in-trial adaptation

Enroll patients in one of several arms
 Randomization based on probability of success
 Closer to QI – broader participation?

Thresholds for dropping arms
 Assigns pts to arms that appear more likely to work
 Sample size a function of required certainty, not a pre-determined
number! Number of treatments, randomization scheme changed
during trial by the results of the trial itself

Selective randomization for …
 Different types of patients or clinical settings
 Different use of co-interventions
The Traditional RCT…
Patients with disease X
At the start,
50% chance
that A > B
The Traditional RCT…
Patients with disease X
At the end,99% sure that A > B
What about in the middle?
A planned trial of A vs. B in 400 patients
After 40 enrolled ….
40
No. of
patients
Dead
20
Alive
A


B
The probability that A > B = 78%
Start randomizing MORE patients to A than B …
After 80 patients …
Dead
40
No. of
patients
Alive
20
A
B
Now, the probability that A > B = 99.9%
 Stop the trial!

Response-adaptive randomization
A
B
Randomization rule
Statistical model
Response-adaptive randomization
Odds weighted
towards best
RX
A
B
Randomization rule
Statistical model
Response-adaptive randomization
New arms
activated
C
A
B
D
Randomization rule
Statistical model
Response-adaptive randomization
Or dropped…
C
A
D
Randomization rule
Statistical model
Response-adaptive randomization
Different weights
for different
patient groups
C
A
D
Randomization rule
Statistical model
‘Big Data’
Analytics
POC
Trials
Platform
Trials
Leverage the EHR
✔
✔
Low incremental costs
✔
✔
Real-world ‘effectiveness’
✔
✔
✗
✗
✗
Consider multiple therapies
✔
✗/✔
✔
‘Personalized’ estimates
✔
✗
✔
Offer ‘live’ tailored options
✔
✗
✔
Robust causal inference
✗
✔
✔
Feature
‘Big Data’
Analytics
POC
Trials
Platform
Trials
???
Leverage the EHR
✔
✔
✔
Low incremental costs
✔
✔
Real-world ‘effectiveness’
✔
✔
✗
✗
✗
Consider multiple therapies
✔
✗/✔
✔
✔
‘Personalized’ estimates
✔
✔
✔
Offer ‘live’ tailored options
✔
✔
✔
Robust causal inference
✗
✗
✗
✔
✔
✔
Feature
✔
✔
A novel blend of ‘POC’ + platform designs

REMAP
 Randomized
 Embedded
 Multifactorial
 Adaptive
 Platform trial
REMAP
✔
✔
✔
✔
✔
✔
✔
WP 5/Practice C
AD-SCAP study
WP leaders: Marc Bonten & Jean-Daniel Chiche
PREPARE is funded by the European Commission under grant number 602525
CAP is a fascinating disease
CAP patients
Hospital
ICU
Moderate
& severe
ARDS
Design full study (AD-SCAP)


Perpetual trial -> Research platform
Start with 3 selected interventions:
 Antibiotics, steroids & ventilatory strategy



High equipoise
Low ethical barriers
Low costs (incl. no additional diagnostics)
Response adaptive randomization (RAR)
 Future options: selection of new suitable treatments:

 Diagnostics
 Antivirals
Bug package

Comparison of two guideline recommended
antibiotic regimes
 3G cephalosporin + macrolide
 Respiratory quinolones
Administration: IV
 Dose & duration according to (local) guidelines

 If macrolides then > 3 days for immunomodulatory effect

Exclusion criteria:
 known allergies to study drugs
Adjunctive package
Treatment with hydrocortisone vs. no
hydrocortisone
 Administration: IV
 Dose & duration: Hydrocortisone 100 mg TID
for 7 days
 Exclusion criteria:

 Contra-indication to randomization
 E.g.: Patient receives systemic corticosteroids for
an indication other than severe CAP
Ventilation: who should be enrolled?
Centre selection
 As per WP5 main protocol
 Expert centres for « Ultraprotective arm »

Inclusion criteria
 As per WP5 main protocol
 Plus specific criteria



Intubated
Within 48 hours of admission to ICU
PaO2/FiO2 ratio <200 on a PEEP of 5 cmH2O
ICU
Moderate
& severe
ARDS
Randomisation
I
II
ABthx Steroids
III
Vent
72h

PaO2/FiO2<200
with PEEP=5cmH2O
Y
Non eligible
N
Assess PEEP response
(PEEP 15, VT 6 mL/kg)
PaO2/FiO2 ≥20%
& PaCO2 stable (15% margin) or 
& Crs stable (15% margin) or 
B
N
Y
A
D
C
PEEP responders
In selected centres
« ARMA »
PEEP-FiO2 algorithm
VT 6 mL/kg
MaxPEEP for VT 6 mL/kg
&Pplat 27cmH2O
while ∆P<12 cmH2O
Max PEEP for
VT 4 mL/kg
&Pplat 24cmH2O
with ECCO2R
Minimal distension arm
Maximal recruitment arm
Ultraprotective arm
PEEP non-responders
Register your site for AD-SCAP

AD-SCAP is conducted by the PREPARE
consortium, a platform foR European
Preparedness Against (Re-)emerging Epidemics

Interested? Please contact
 [email protected] or [email protected]
 More information:
http://www.esicm.org/research/adscap/prepare
PREPARE Ad-SCAP: Advantages

Gets closer to individualized treatment decisions …

For example, should my patient receive IV steroids?
 Depends on






Whether shock is present
How sick (hypoxic) the patient is
Whether underlying cause is viral or not
Whether an anti-viral is being administered
Whether other strategies are being used that may minimize lung injury
and inflammation (protective vs. ultraprotective ventilation)
Separate probability estimate for each consideration …
 Trial enrolls until a predefined level of certainty
 As soon as one question hits threshold, answer is announced
PREPARE Ad-SCAP

Can be fully embedded
 ICU admission orders


Approved in Netherlands with delayed consent
Multifactorial
 24 regimens (3 x 2 x 4)
 4 subgroups (shock and severe hypoxia, Y/N)
 96 estimates of treatment effect
Antimicrobial
Immunomodulation
Ventilation
Factor 1
A-1
B-1
C-1
Factor 2
A-2
B-2
C-2
Factor 3
A-3
Factor 4
C-3
C-4
Run the trial ‘in silico’ ahead of time …

Monte-Carlo simulations
 Run 1,000s of times under different scenarios
Scenario #1: 2 of 8 regimens are best
‘True’
mortality
80 fewer deaths;
higher power
Average results
from 1,000s of
simulations
Scenario #2: 1 regimen is best
‘True’
mortality
Average results
from 1,000s of
simulations
94 fewer deaths;
higher power
PREPARE Ad-SCAP

Funding
 EU FP7 PREPARE WP 5 program (25M euro)
 Australian NHMRC ‘OPTIMISE’ program ($6M)

Simultaneously test
 Different anti-microbial strategies
 Different host immunomodulation strategies
 Different ventilation strategies

Separate RAR & stopping rules for multiple subgroups
 Patients preferentially assigned to best performing arm



Allocation is random, but NOT 50:50
Odds of assignment proportional to odds of success
Not a guinea pig!
This all looks very nice, but …








EHR data quality
Institutional commitment
Ethics
Statistics and design
Reporting and dissemination of results
Funding
Oversight
Integration with other clinical research programs
Conclusions

RCTs remain our strongest ‘truth’ finder
 But, current RCT enterprise lets us down

‘Big Data’ should not be cast as an
alternative to the RCT
 This is a false choice

Instead, the digital age enables novel RCTs
designs
 Smarter and safer
Self-learning healthcare is …
fused care and research
Ackowledgments







Derek Angus, UPMC
Scott Berry, Berry Consultants
Roger Lewis, Harbor UCLA Med Center
ESICM Trials Group
PREPARE – Ad-SCAP team Utrecht
Combacte & Capnetz networks
PREPARE Consortium