Transcript bEWOC

Application of a Bayesian
strategy for monitoring
multiple outcomes (safety
and efficacy) in early
oncology clinical trials
Application of a Bayesian strategy for monitoring multiple outcomes
in early oncology clinical trials
|
1
Phase I clinical trials in oncology
●
Recommend a dose for further clinical development
●
Design:
● Patients included in successive cohorts (usually n=3 in each
cohort)
● All patients within the same cohort receive the same dose
• First cohort receive the lowest dose
• Primary endpoint: Dose-Limiting Toxicity
• After completion of each cohort, decision is made on predefined
algorithm to:
• Escalate the dose
• Stay at the same dose
• De-escalate the dose
• Stop the study
2
Several designs
●
●
●
Up-and-Down designs
● « 3+3 »
● Accelerated Titration Design
Model-based dose-response designs
● CRM: Continual Reassessment Method
● bCRM: CRM applied on two binary outcomes (safety and efficacy)
● Designs derived from the CRM : Escalation With Overdose Control
(EWOC)
bEWOC: Escalation With Overdose Control for bivariate outcome
● Dose-response relationships
● Plane (Probability of Activity , Probability of Toxicity)
● Design
3
Up-and-Down design
3+3 design
Dose level (i)
Enter 3 patients
0/3 DLT
≥ 2/3 DLT
1/3 DLT
Add 3 patients
1/6 DLT
Escalate to
dose level (i+1)
≥ 2/6 DLT
Dose level (i-1)
is the MTD
NOM DE LA
|
4
Model-based dose-response relationship
100%
a=0;b=1
logit[(d)] = a + exp(b) log(d/d*)
Probability of DLT
75%
a = -2 ; b = 1
50%
a = -1 ; b = 0
Target level
Target range
25%
0%
0
50
150
100
200
250
Dose
NOM DE LA
|
5
New drugs in development
●
●
Cytotoxic drugs
● Assumption: Monotonous Dose-Efficacy relationship
● Dose level recommended: Maximum Tolerated Dose (MTD)
• A dose with a probability of DLT closest to a target proportion
Molecularly Targeted Agents (MTA)
● Non-monotonous Dose-Efficacy relationship
● Two endpoints: Toxicity and Efficacy (binary outcomes)
Cytotoxic profile
MTA profile
NOM DE LA
|
6
EWOC for bivariate outcome
Dose-response models
●
Safety model
logit( P
i
) = a1  exp( b1 ). log( di / d )
*
1
DLT
Slope > 0 => Monotonous dose-toxicity relationship
● Efficacy model
logit( P
i
RESP
) = a 2  b 2 . log( d i / d )
*
1
  2 . log( d i / d 2 )
* 2
Non-monotonous dose-efficacy relationship
PDLT : Probability of DLT
d1* , d2* : Reference doses
PRESP : Probability of Tumor Response
7
Bayesian estimation
●
Gaussian a priori distributions for all parameters
●
A posteriori distributions obtained by MCMC algorithm
●
●
●
3 independent chains
Convergence checked by Brooks-Gelman-Rubin criterion
Software tools
● R software version 2.12.2
● BRugs package version 0.7.1
● OpenBUGS software version 3.2.1
8
Dose-response curves and related credibility
interval
DLT Probability mean estimate
Tolerability threshold
9
Dose-response curves and related credibility
interval
Response Probability mean estimate
Interest threshold
9
Dose-response curves and related credibility
interval
9
EWOC for bivariate binary outcome
Binary endpoints
● DLT
● Tumor Response
●
Plane (IP(Resp) , IP(DLT))
Not safe
Not active
Looking for doses that are in
the Targeted Area:
Active but
not safe
Over-toxicity
e.g. doses such:
IP(DLT)<0.35
IP(Resp)>0.5
Safe but
not active
Useless
Moderate
●
Targeted
Active
and safe
Probability of Response
IP(resp | di)=0.48
IP(DLT | di)=0.18
EWOC for bivariate outcome
Predictions for dose level escalation decision
Dose 300mg/kg
400mg/kg
300mg/kg
250mg/kg
200mg/kg
150mg/kg
11
Probability to be in each area
125 mg/kg
150 mg/kg
200 mg/kg
250 mg/kg
300 mg/kg
400 mg/kg
12
Conclusion
●
Advantages
● Assess both Toxicity and
Efficacy
● Take into account uncertainty
and control the over-dosing
●
● Efficacy model more flexible for
Molecularly Targeted Agents
● Decision and communication
tool (clinician team)
Limits / Next steps
●
●
●
●
Limited data / Use PK sampling, PD-biomarkers, mechanistic modelling
Better assessment of patient variability / Hierarchical models
No Time-to-Event / Different kinds of toxicities (Acute and Cumulative)
Different schedules / PK model
13
References
●
●
●
●
●
[1] Booth C. M., Calvert A. H., Giaccone G.,Lobbe-Zoo M. W., Seymour L. K.,
Eisenhauer E. A. Endpoints and other considerations in phase I studies of targeted
anticancer therapy: Recommendations from the task force on methodology for the
development of innovative cancer therapies. European Journal of Cancer 2008, 44,
19-24.
[2] O'Quigley J., Pepe M., Fisher L .Continual reassessment method: a practical
design for phase 1 clinical trials in cancer. Biometrics 1990, 46, 33-48.
[3] Neuenschwander B., Branson M., Gsponer T. Critical aspects of the bayesian
approach to phase I cancer trials. Statistics in Medicine 2008, 27, 2420-2439.
[4] Thall P. F., Cook J. D., Estey E. H. Adaptive dose selection using efficacytoxicity trade-off: Illustrations and practical considerations. Journal of
Biopharmaceutical Statistics 2006, 16, 623-638.
[5] Whitehead J., Hampson L., Zhou Y., Ledent E., Pereira A .A Bayesian
approach for dose-escalation in a phase I clinical trial incorporating
pharmacodynamic endpoints. Journal of Biopharmaceutical Statistics 2007, 6, 44.
14
Merci