Transcript Statistical and Practical Aspects of a Non-Stop Drug

Statistical and Practical Aspects of a
Non-Stop Drug Development Strategy
Karen L. Kesler and Ronald W. Helms
Rho, Inc.
Contact: [email protected]
1
Introduction
•
•
•
•
Approval Process—Phase I, II, & III
Phase II designed to:
– Discover best treatment regimen: route, dose, timing, etc.
– Profile Safety
– Develop information needed to design Phase III trials
Phase III designed to show efficacy
Time from patent to approval >12 years and growing
How can we use new statistical methods to shorten
overall development time?
2
NonStop Concept
•
•
•
•
Design a Phase II study with enough arms to cover
all plausible regimens, plus placebo
– K possible doses  K+1 arms
Conduct a sequence of frequent interim analyses
(using group sequential methods).
At each interim analysis, potentially prune
treatment arms, except control.
Ultimately, at some interim analysis, the treatment
arms are reduced to one active treatment regimen
and the placebo.
3
NonStop Overview
•
•
Specify frequent interim analyses whose primary
objective is to prune (kill) treatment arms as quickly
as possible.
– For either safety or futility.
– The placebo arm is never pruned.
The project’s primary objective is to get to
Phase III ASAP.
4
NonStop Overview
5
4
Boundary for stopping for efficacy
Score statistic
3

2
Treatment pruned: safety
Administrative
Boundary
Boundary for pruning
1
0
0
5
10
15
20
25
30
35
-1
Number of subjects per treatment group
-2
5
NonStop Overview
5
4
3


2



1

0
0
-1
-2
5
10

15
20
25
30
35
6
What makes it work?
•
•
NonStop development is literally non-stop: virtually
all the “wasted” time is eliminated:
– Between sequential Phase II studies, and
– Between the last Phase II and first Phase III study.
Recall our Phase II goals:
–
–
–
–
Profile safety.
Choose the most appropriate outcome.
Choose the most efficacious dose or regimen.
Eliminate ineffective compounds quickly.
7
Issues
•
•
•
•
•
How many treatment arms to start with? Which
doses/regimens to use?
How to select the most appropriate outcome?
What are the challenges associated with this type
of design?
– Centralized randomization
– Quick data capture and management
Who gets to see the interim analysis results?
Will regulatory agencies (like FDA) accept this
design?
8
How many treatment arms?
•
•
•
•
Clinical Decision
– Depends on how much is known about the
compound.
Dose Response
– Need to cover enough of the dose response
curve.
Safety
– Do not want to endanger the subjects.
Balance
9
How many treatment arms?
An Example
•
•
•
•
•
Untested compound, but some literature on the
class of compound.
Suggests highest safe dose is 750 mg/kg.
Investigators interested in 500 mg/kg also.
Recommend adding 250 mg/kg to get doseresponse curve.
Decision: Four arms (Pbo, low, med, high).
10
How many treatment arms?
An Example
•
•
•
•
•
Known compound, interested in combination
therapy for resistant strain.
Current standard of care is 75mg/kg/day.
Want to investigate high doses, possibly up to 125
mg/kg/day.
Above 125 not feasible due to cost constraints
Decision: Start with 75 and 100, develop safety
profile of higher dose and add 125 if determined
to be safe.
11
How to select a primary outcome
•
•
Some etiologies do not have clearly defined
associated outcomes.
– Multiple scales
– Length of follow up
– Choices of statistical model
Typical considerations are
–
–
–
–
Cost
Accuracy
Power (continuous vs. categorical)
Clinical Relevance (surrogate vs. “hard” endpoint)
12
How to select a primary outcome
•
With NonStop design in Phase II, we analyze all of
the outcomes at each interim.
– Allows for complete picture
• Do the various measures agree?
– Can monitor for variability in the outcome
• Is the outcome consistent over time?
– Gives a good idea of collection issues
•
• Can we collect this accurately?
Only works for outcomes with a short follow up
time.
13
How to select a primary outcome
An Example
•
•
•
Infection rate
–
–
–
–
Rate of infection w/in 30 days
Time to first infection
Rate of infection per person-day
Varying definitions of infection
Declare Primary for sample size calculations
Compare interpretations of each type of outcome
14
Other Key Logistics
•
•
•
•
•
Statistical/Clinical Communication
Constant Team Coordination
Centralized Randomization
Efficient Data Management
Planning, Planning, Planning. . .
Let’s examine some of these in detail . . .
15
Centralized Randomization
•
•
•
Maintains balance over many treatment arms (and
any important strata)
Allows for instantaneous curtailment of a pruned
treatment arm
Preserves masking at the site when treatment
arms are pruned
16
Quick Data Capture and Management
•
•
•
Need as up-to-date information as possible to
make decisions
Need to enter and query data as quickly as
possible
Monitoring issues:
– Using monitored vs. unmonitored data
– Timing
17
Regulatory Issues: Maintaining the
Mask
Who gets to see the results of the Phase II interim
analyses?
• This is a controversial issue, both for NonStop and
traditional strategies.
• In this case, the results of interim analyses can be
held as closely as one wishes.
• In a traditional strategy, the results of each Phase
II study – the analog of this study’s interim results are disseminated widely.
18
Regulatory Issues: Maintaining the
Mask
•
Who is allowed to see the interim results?
– Definitely “No”:
• Site Personnel
• Anyone making a determination on patient care
• Anyone making a determination on outcomes
– Definitely “Yes”:
• Statistical team creating the interim reports
• Primary investigator/Clinical Lead making continuation decisions
– Undecided:
• Statistical team performing the final analysis
• Non-clinical personnel: investors, project coordination leads,
19
Regulatory Issues: FDA Acceptance
•
•
•
•
NonStop design suited for Phase II or exploratory
trials, not confirmatory trials.
Bring FDA representatives in early, explain the
design fully and why it would be appropriate.
Willingness to accept innovative statistical
approaches varies from group to group.
We do have one in progress and we are working
with the FDA. . .
20
Summary
•
•
Philosophical shift from the typical hypothesis
testing structure of a confirmatory Phase III design
to an exploratory treatment selecting Phase II
design.
Recall goals:
–
–
–
–
–
Profile Safety.
Select most appropriate outcome,
Select most efficacious (yet safe) dose/regimen,
“Kill” ineffective compounds,
Generally: Get enough information to design a
successful confirmatory trial!
21
Summary
•
•
The NonStop strategy is not a panacea – it’s not
useful for all drug development situations.
Relatively few patients enrolled between taking a
snapshot of the database for an interim analysis
and the meeting to make decisions based on the
interim analysis. Implies:
– Fast data capture, processing.
– Centralized Randomization
– At the start of Phase II, one can specify a set of
treatment regimens that will very likely include the
regimen(s) to be tested in Phase III.
22
Conclusions
•
•
•
A NonStop design can be a challenging, but
efficient and cost-effective design for a Phase II
exploratory study.
With sufficient planning for quick data capture and
management, as well as centralized
randomization, this design can save a lot of time in
the exploratory stage of a drug development plan.
With preplanning and communication, this design
can be accepted by regulatory agencies.
23
A version of this presentation is available
(PDF) at: www.RhoWorld.com
5
4
3


2

1
0
0
5
10
15
20
25
30
35
-1
-2
24