Eugm 2012 mehta - future plans for east - 2012 eugm
1. Future Plans for EastFuture Plans for East
Cyrus Mehta
President, Cytel IncPresident, Cytel Inc
2. Where are we going with East?
Four New InitiativesFour New Initiatives
• Enhance the simulations capabilities by
permitting external calls to R and SAS
• Conditional simulation of the remainder of the
trial given the interim data
• Multi arm group sequential designsg p q g
• Population enrichment designs
3. External calls to R for Adaptive Decisions
• East already simulates patient arrivals, patientEast already simulates patient arrivals, patient
responses and and drop outs
• Adaptive decision rules are currently based onAdaptive decision rules are currently based on
conditional power
• Permit user specified decision rules:Permit user specified decision rules:
– Bayesian criteria for SSR
– More general functions for SSRg
Implement through calls to R or SAS at each
interim analysis for each simulated trialy
5. Conditional Simulation of Remainder of Trial
• At an interim analysis both DMC and SponsorAt an interim analysis both DMC and Sponsor
are very interested in the question:
“What is the chance that this trial will succeed”What is the chance that this trial will succeed
• Conditional power has some limitations
i l di l f h ld h– no visual display of what could happen
– does not give a sense of variability
– does not estimate final treatment effect
8. Multi Arm Group Sequential Designs
• Compare D dose groups to a common controlCo pa e dose g oups to a co o co t o
without dose response assumptions
• Standard Approachs include:pp
– Single stage design with closed testing (MCP in East)
– Two stage adaptive design with early stopping or
l i 1 bi i i htreatment selection at stage 1, combining stages with
pre specified weights, and closed testing (Posch et al)
• Limitations:Limitations:
– correlation between test statistics not exploited
– Not easy to generalize to multiple stagesy g p g
9. Generalization of Two Arm GSD
• Monitor the D test statistics over K looks:Monitor the D test statistics over K looks:
– C1, C2, … CK are the K efficacy boundaries
Stop and claim efficacy if one of the test statistics– Stop and claim efficacy if one of the test statistics
crosses an efficacy boundary
– Incorporate non binding futility boundaries– Incorporate non binding futility boundaries
• Find C1, C2, … CK such that the probability that
the max of D multivariate normal statisticsthe max. of D multivariate normal statistics
exceeds one of the Cj’s is
10. Example: Four Arms and Three Stages
Look Cumulative Stopping Boundaries
Stopping Boundaries of (-4) Spending Function at =0.05
Look Cumulative
Sample Size/Arm
Stopping Boundaries
4 Arm Design 2 Arm Design
1 25 3.39 3.011
2 50 2.89 2.547
3 75 2.77 1.999
Alternative Hypothesis Power
0.4, 0.4, 0.4 67.6%
0, 0.4, 0.4 58.7%, ,
0, 0.2, 0.4 44.6%
0, 0, 0.4 42.3%
Power of corresponding two-arm
design with 75 patients/arm is 68%
11. Population Enrichment Designs for Oncology
• Failure rate for late stage oncology trials is almostFailure rate for late stage oncology trials is almost
60% (Kola and Landis, 2004)
• Two recent scientific developments can improve thisp p
track record
– development of molecularly targeted agentsp y g g
– statistical methodology of adaptive trial design
applied to time to event data
• Fact: Some subgroups benefit differentially from
others when treated with the targeted agent
12. Oncology Products Approved in the USA
for Selected Patient Population
Compound/Target Indication (prevalence target)Compound/Target Indication (prevalence target)
Crizotinib (Xalkori®)/ ALK
rearrangement
•Non small cell lung cancer with ALK rearrangements
(5%)
Vem rafenib (Zelboraf®)/ BRAFVemurafenib (Zelboraf®)/ BRAF
mutation •Advanced melanoma with mutant BRAF (30 40%)
Trastuzumab (Herceptin®);
Lapatinib (Tykerb®/ Her2
•Her2 expressing breast cancer (25%)
•Her2 expressing metastatic gastric cancer (20 30%)p ( y / •Her2 expressing metastatic gastric cancer (20 30%)
Aromatase inhibitors (letrozole,
exemestane) •ER(+) breast cancer (60 70%)
Rit i b (Rit ®)/ CD20 ( ) ll l h ( )Rituximab (Rituxan®)/ CD20 •CD20(+) B cell lymphomas (90%+)
Cetuximab (Erbitux®);
Panitumumab (Vectibix®) /
EGFR
•Advanced Head/neck cancer (~100%)
•EGFR(+) metastatic colorectal cancer (60 80%)
WT l l ( )
EGFR
•KRASWT metastatic colorectal cancer (60%)
DIA Adaptive Design Scientific
Working Group
14. Statistical Method
• and are null hypotheses for the subH Hand are null hypotheses for the sub
population and the full population, respectively
• Specify a two stage combination test for
SH FH
H H• Specify a two stage combination test for ,
and
D id h i i l i h h
F SH H
SH FH
• Decide at the interim analysis whether to
proceed with the full population or the sub
l ipopulation
• Perform a test of both and if youF SH H SH
enrich
15. Software Challenges for Event Driven Trials
• Provide simulation tools to generate patientProvide simulation tools to generate patient
arrivals with different arrival rates, dropout
rates and hazard rates for the two subgroupsrates and hazard rates for the two subgroups
• Provide ability for user to specify flexible
decision rules for population enrichmentdecision rules for population enrichment
• Provide visulation tools for determining when
f h i i l i ( li dto perform the interim analysis (generalized e
charts)
17. Population enrichment designs can have
a very large pay offa very large pay off
• Design Parameters:Design Parameters:
– 4.5 months median PFS on control arm
Suppose HR = 0 55 for Bio+ and HR=1 for Bio– Suppose HR = 0.55 for Bio+ and HR=1 for Bio
– 40% of population is Bio+. Thus overall HR = 0.8
C id t i l ith 370 ti t 300 t• Consider a trial with 370 patients; 300 events:
– Non adaptive design has 50% overall power
– Adaptive design: has 78% overall power and 94%
power conditional on enriching
18. Thank you for participating
• Lots of good discussion• Lots of good discussion
• Many ideas for new software• Many ideas for new software
• Cytel is well on its way.Cytel is well on its way.
Looking forward to the next 25
years of growth