Post on 27-Mar-2015
transcript
Flexible designs for pivotal clinical trials
Vlad Dragalin, RSU-SDS-BDS-GSK
FDA/Industry Workshop
Session: Flexible Designs – Are We Ready Yet?
Washington, D.C., September 14-16, 2005
22
Declaimer The views expressed in this presentation are
not necessarily those of PhRMA
The views expressed in this presentation are not necessarily those of GlaxoSmithKline
The views expressed in this presentation are not necessarily my
33
Acknowledgment Judith Quinlan for inviting me to work on this trial
GSK Clinical Team for compound SBx for giving me the opportunity to design this trial
Compound SBx
44
Outline
Overview of adaptive designs
GSK experience
Details of Design
Points to Consider
55
What are Adaptive Designs?
Adaptive Design
• uses accumulating data to decide on how to modify aspects of the study
• without undermining the validity and integrity of the trial
PROTOCOL
AMENDMENTS
66
General Structure of Adaptive Designs An adaptive design requires the trial to be conducted in several
stages with access to the accumulated data
An adaptive design may have one or more rules:
Allocation Rule: how subjects will be allocated to available arms Sampling Rule: how many subjects will be sampled at the next stage Stopping Rule: when to stop the trial (for efficacy, for harm, for futility) Decision Rule: the final decision and interim decisions pertaining to
design change not covered by the previous three rules
At any stage, the data may be analyzed and subsequent stages redesigned taking into account all available data
77
Adaptive Design Process
DecisionRule
New Patient
SamplingRule
StoppingRule
AllocationRule
Data
88
GSK Experience PoC Study in Neuropathic Pain:
Three-stage adaptive design: p-values combination test Allocation Rule: drop the “loser” Stopping Rule for efficacy/futility Sampling Rule: timing of the 2nd/3rd stage depends on data
99
Background
Compound SBx lead indication in Psychiatry (anxiety & depression) secondary indications in Neuropathic pain, RLS & FMS
Objectives : To establish superiority of SBx dose(s) versus placebo
Confirm efficacy (and durability of response) 8 week treatment, but expect treatment effect at 2 weeks correlation between early and late treatment effects
Establish safety profile Establish dose-response
Strategic Aim: pivotal quality to potentially support registration
1010
Study Designs Last thing we want is to get to the end only to discover
no doses are effective OR we missed obtaining a significant result because our original assumptions
were too optimistic
Standard Dose Ranging Design known entity, but lacks flexibility
Adaptive Design Potential savings in terms of both resource and time if there are clear signs
that SBx does not work Allows for addition of more patients to a promising dose
Protects against underestimate of variance Potential to get to decision quicker, e.g. 5 - 9 months Full data package on doses of interest Statistical validity maintained
1111
Team Concerns Regulatory acceptance as a pivotal quality study
statistical rigor is maintained Thought EU feedback may delay study start up
concerns proving unfounded: design accepted by EMEA CPMP after one F2F meeting
Patients may be enrolled before you can adapt the study performed simulations use electronic data capture
GSK inexperience (e.g. protocol development, electronic data capture) may delay study start
Unequal information on all doses
1212
Details of the DesignPrimary Endpoint:
Primary Goal:
Target Difference:
STDeviation:
Type I error:
Power:
Traditional Dsgn:
Adaptive Dsgn:
Efficacy Bndry:
Futility Bndry:
Inflation Factor:
PI-NRS change from Baseline at 8thW of treatment
Comparison of three SBx doses (LD, MD, HD) with Plb
1.3 units
2.1 units
= 0.05 (adjustment for multiplicity = 0.05/3 = 0.017)
90%
4 parallel groups - 72 patients/per arm (total 288)
3 stage inverse-normal combination test
O’Brien-Fleming type
nominal levels: (0.0006, 0.014, 0.0235)
nominal levels: (0.5, 0.5)
1.025 - maximum 75 patients/arm
1313
1 2 3 4 115 6 7 8 9 10 Month
Enrollment Period
1st Stage Data
Plb
MDLD
HD
3rd IA2nd IA1st IA
Randomization
CROCRO
0 8w2w
1st Stage
Decisions:
• Stop arm for futility• 49.5%
• Stop arm for efficacy• 2.9%
• Stop the study for futility• Stop arm(s) for safety • Determine Randomization• Determine timing of 2nd IA(based on 80% Cond. Power)
Timing by 80% CP
GSK GSK Steering Steering
CommitteeCommittee
GSK GSK Steering Steering
CommitteeCommittee
1414
1 2 3 4 115 6 7 8 9 10 Month
Enrollment Period
1st Stage Data
PlbMDHD
3rd IA2nd IA1st IA
Randomization
CROCRO
0 8w2w
2nd Stage
2nd Stage Data
GSK GSK Steering Steering
CommitteeCommittee
Decisions:
• Stop arm for futility• 13.1%
• Stop arm for efficacy• 47.6%
• Stop the study for futility• Stop arm(s) for safety • Determine Randomization• Determine timing of 3rd IA(based on 80% Cond. Power)
1515
1 2 3 4 115 6 7 8 9 10 Month
Enrollment Period
1st Stage Data
PlbMD
3rd IA2nd IA1st IA
Randomization
CROCRO
0 8w2w
3rd Stage
2nd Stage Data
GSK GSK Steering Steering
CommitteeCommittee
3rd Stage Data
Final Analysis
• Overall p-value • Estimate of TRT eff.• Confidence Interval
1616
Interim Analyses: Data
For each armat each stage
1717
Interim Analyses: Test
Null Hypothesis:
Estimate of mean 8thW endpoint:
Standardized Test Statistics:
Reduced Variance:
1818
Multiplicity Adjustment Due to interim analyses
O’Brien-Fleming stopping for superiority nominal levels: (0.0006, 0.014, 0.0235)
Stopping for futility nominal levels: (0.5, 0.5)
Due to multiple comparisons Holm procedure at each stage
Due to adaptive design Inverse normal p-value combination rule
1919
Design PropertiesDifference -0.2 0.0 0.2 0.5 1.0 1.3 1.5
Power 0.001 0.008 0.033 0.171 0.678 0.898 0.936
AvSS 42.7 51.2 59.4 68.4 66.3 56.7 50.5
AvSS 2ndStage
29.9 29.8 29.6 29.0 26.4 23.2 20.7
AvSS 3rdStage
29.9 29.7 29.2 28.0 23.0 18.5 15.4
Pr. Reach2nd Stage
0.368 0.505 0.629 0.796 0.949 0.971 0.967
Pr. Reach3rd Stage
0.224 0.374 0.541 0.728 0.704 0.495 0.359
Difference -0.2 0.0 0.2 0.5 1.0 1.3 1.5
Power 0.001 0.008 0.033 0.171 0.678 0.898 0.936
AvSS 42.7 51.2 59.4 68.4 66.3 56.7 50.5
AvSS 2ndStage
29.9 29.8 29.6 29.0 26.4 23.2 20.7
AvSS 3rdStage
29.9 29.7 29.2 28.0 23.0 18.5 15.4
Pr. Reach2nd Stage
0.368 0.505 0.629 0.796 0.949 0.971 0.967
Pr. Reach3rd Stage
0.224 0.374 0.541 0.728 0.704 0.495 0.359
Parallel 4 arm Dsgn: 72 per arm
2020
Points to Consider
Logistics issues pertaining to traditional group-sequential designs also pertain to adaptive designs
Establish an IDMC (charter, contracts) for pivotal trials
Have adaptation performed by an independent third party with no conflict of interest issues
During interim adaptation, unblind only data that are necessary to be unblinded
Patient recruitment is not interrupted
2121
Points to Consider Adaptation entails careful planning at the protocol design stage
Every detail of the statistical design and analysis that can be fixed in advance is described in the study protocol:
number of interim analyses information rates stopping guidelines tests
Depending on the information rates, the interim analysis is scheduled. The time of the IA is unknown to the investigators.
On IA a snapshot of the DB is made (“soft close”). All available data are used for IA.