Transcript SAPA - Biopharmaceutical Network

Planning and Execution of LearnPhase Clinical Trials
Parvin Fardipour
7/21/2015
Introduction
The planning and execution of response adaptive
randomization trials involves complexities beyond
those encountered in the planning and execution of
more traditional trial designs
 These complexities involve:

Availability and flow of information required to support adaptive decision
making
Implementation of changes in the randomization scheme
Composition and role of data monitoring committees (DMCs)
- Who can see the data, when and to what effect?
Drug supply
7/21/2015
2
Decision to Use a Response-Adaptive
Randomization Design


While learn-phase trials often have multiple clinical and scientific
objectives, the primary goal(s) of the trial are used to guide the
adaptive randomization
An adaptive randomization design is most likely to be
successfully implemented when the clinical team support the
approach but their enthusiasm is based on realistic expectations
 Unlikely to be successful when the design is being used in a late attempt to “rescue”
the development program for a compound with inadequate efficacy or an
undesirable side effect profile

It is difficult to design and implement a response-adaptive
randomization design as quickly as a traditional “off the shelf”
design
 Additional planning, validation, and implementation time may be poorly received in
settings in which the key objective is to launch a trial quickly, rather than to launch a
more flexible and potentially informative design
7/21/2015
3
PLANNING PHASE

Selection of Endpoints
Identification of efficacy and safety endpoints for the trial, as applicable to
the specific situation
Several issues merit consideration
- The endpoints that drive the adaptive allocation rule must be available in a timely
manner to allow modification of the randomization probabilities in real time
- Ideally, the primary endpoint used in a learn phase adaptive randomization trial
would be identical to the endpoint likely to be used in a subsequent confirmatory
trial
7/21/2015
4
PLANNING PHASE

Trial Objectives
The statistical design of a learn phase trial begins by asking the question
- What is it we are trying to learn in this trial? – research question
- We must also decide how much certainty is needed when answering the research question(s)?
In a dose finding study, the learning objectives are defined in terms of
understanding some aspect of the dose-response relationship
- for example,
- What is the minimum effective dose (MED)?
- What is the maximum tolerated dose (MTD)?
Stopping rules may written in terms of having answered the key research
question(s) with the targeted degree of certainty; for example:
- Stop for futility if, for every dose, the probability of a clinically significant advantage over placebo is <0.1
- Stop for success if we have identified at least one dose where the probability of a clinically significant
advantage over placebo is 0.95 and we have identified the MED with a probability of at least 0.6
7/21/2015
5
PLANNING PHASE

Development of an Adaptive Design
 The adaptive design generally utilizes
-
A dose-response model (e.g. NDLM, Logistic)
And
Longitudinal modeling to incorporate information from enrolled subjects that have not
completed the protocol
 Treatment arms in the study should also be discussed with special consideration to
logistical issues
-
How many treatment arms are required and what is the rationale?
PK and PD modeling
Feasibility of doses
 Enrolment profile in the study is crucial for understanding information accrual and
how the adaptations will be made during the study
 It is important to discuss with the clinical team about the constraints to consider in
the design such as the timelines (e.g., patent expirations, competitor projects),
budget (e.g., number of centers, patients), clinical (e.g., highest number of tablets
taken at any one time), drug formulations, maximum number of subjects
7/21/2015
6
PLANNING PHASE



Proposed adaptive designs are evaluated by extensive simulation
to assess operating characteristic of design across various
response pattern scenarios (e.g. flat dose response, sigmoid dose
response, etc)
The adaptive design is implemented as a software package that
runs quickly enough to allow thousands of simulations to be run
The implementation software is configured to read in the trial data
contained in the response and current randomization files and to
produce the details of the model fit, the predictive probabilities for
each stopping criterion, and the revised randomization scheme
7/21/2015
7
PLANNING PHASE

Documentation
Even thought we are discussing a learn trial, we work under the
assumption that the trial will become part of a future regulatory dossier;
thus, all documentation should be in place prior to initiation of the trial and
in a format that will facilitate regulatory review
These documents define the foundation for the execution phase and
should be approved by sponsor personnel, DMC if applicable and then
signed off, and stored in a submission-ready format
Required documentation that needs to be retained includes:
- trial protocol, the original randomization scheme, the DMC charter, the adaptive
design simulation report, SAP, etc
7/21/2015
8
PLANNING PHASE

Documentation
All external parties that are involved in the execution of the trial must be
identified and appropriate confidentiality agreements, conflict of interest
disclosures, and contracts must be produced and approved by relevant
authorities
The flow of clinical data and other information during the trial needs to be
planned and managed
- How are the reports generated by the adaptive design and the ISC shared with
the DMC?
- How are the DMC’s recommendations communicated to the sponsor’s steering
committee?
- What are the formats of the input file to the adaptive design software and the
output file that must be loaded into the sponsor’s randomization system (e.g.,
IVRS)?
- Has the process of loading the adaptive randomization file into the sponsor’s
randomization system been tested and validated?
7/21/2015
9
PLANNING PHASE

Validation Process
 The adaptive design software is validated by a combination of
independent source code review, analysis of simulation results and “back-
to-back testing”
 Back-to-back testing is performed by independently creating a second
version of the program (or crucial parts of the model)
- This process helps ensure the detection of programming errors and similar
problems. The full validation process is documented in the validation report for
the adaptive design
 All the reports produced by the ISC must be validated and a validation
and quality control report is produced
 The complete data flow required for trial execution must be tested using
test data
7/21/2015
10
PLANNING PHASE

Validation Process
To load the adaptive randomization file output into the randomization
system, a thorough validation of all the components must be performed
The update to the randomization system must be synchronized on the
database and drug supply system to ensure that the next subject will be
allocated to the new scheme
It is advised that upload of the adaptive randomization file into the
randomization system is automated to avoid manual errors
7/21/2015
11
PLANNING PHASE

The Response File
 The response file is an input to the adaptive trial design software
 The response file contains randomization identifiers, measurement time
points if necessary, and the measured endpoint values for each relevant
patient visit reported to date

The Randomization File
 The randomization file is one output from the adaptive trial design
software
 The randomization file contains a new randomization allocation with
treatment assignments for subsequent patients (until replaced by the next
update)
 In our trial, there is normally a burn-in period of balanced treatment
assignment to drive the adaptation
 The randomization file is imported into the clinical database, as well as the
randomization and drug supply systems
7/21/2015
12
PLANNING PHASE

Independent Statistical Center (ISC)
 Charged with developing and executing programs for the creation of
tables, listings, and graphs
 Execute the adaptive trial design software
 The data are transferred to the ISC via a secure connection and at
frequent enough intervals to support the adaptive algorithm and allow
production of the interim reports required by the DMC and specified in the
SAP.
 The results from the adaptive design are typically the estimated doseresponse curve for the key variable(s), the predicted probabilities of
meeting critical criteria (e.g., being the minimal effective dose), and also
the new randomization allocation schema
7/21/2015
13
PLANNING PHASE

Development of Interim Reports
DMC will sometimes have responsibility to monitor the study adaptation
process, as well as having more general oversight for safety within the trial
This may lead to different types of interim analyses that must be addressed
in the DMC charter (e.g. weekly interim and full interim)
Different reports will be generated to support each type of interim analysis.
The type of reports required for each interims is documented in the SAP
7/21/2015
14
BUSINESS CASE




it is important that a business model be developed to facilitate
quantitative comparison of design options
During the requirement phase prior to the start of the trial, a key
responsibility of the team is to identify the alternative (traditional)
design if an adaptive design is not utilized
The alternative design will be used for comparison with the
proposed adaptive design so that the business case justifying the
latter can be developed
During the design phase, the business model is developed and
applied in conjunction with the assessment of operating
characteristics for various designs (including, in particular, the
traditional alternative)
7/21/2015
15
BUSINESS CASE

The key elements of the business case
are:
What is the probability of the trial meeting its objectives
(e.g., establishing proof of concept or identifying the
minimally effective dose)?
What is the probability of success in phase III (i.e., the
probability of moving from phase III to registration)?
What is the time to market?
What is the development cost based on number of
subjects and duration of the trial?
7/21/2015
16
BUSINESS CASE

In addition to the simulation work providing estimates of the expected
number of subjects and the expected duration of the trial, the
development cost estimate requires assumptions about various incurred
costs such as monitoring, data management, drug supply, cost per
patients and other generic costs such as investigator’s meeting, etc

Besides the various direct costs, any indirect costs such as utilized
resources should be included in the equation as well

As with the assessment of operating characteristics, the business case
assessment must be made across a range of response pattern scenarios

The business case is presented to the study team, considering the
various response scenarios and facilitating comparison of the design
alternatives

Once the final design is selected, the business case justification is
incorporated into the portfolio management strategy
7/21/2015
17
Data Monitoring Charter (DMC) Charter




The DMC Charter defines the roles, responsibilities, and activities
of the DMC, the independent statistical center (ISC), the sponsor’s
steering committee, contract research organizations (CROs) and
other participating entities and/or individuals during the conduct
of the planned adaptive clinical trial
Additionally, the Charter describes the purpose and timing of DMC
meetings
The Charter also outlines the procedures for ensuring
confidentiality and appropriate communication between the DMC
and sponsor and the expectations for the Open and Closed
Meeting Summaries that will be prepared by the DMC
The intention is for the charter to contribute to preserving the
integrity and validity of the trial
7/21/2015
18
Data Monitoring Charter (DMC) Charter

The function and composition of the DMC, adjudication
board and executive committee need to be described
There are certain elements of the DMC that need to be considered prior to
composition of the DMC
For example for the learn phase, the DMC could be composed of internal
sponsor’s members unless there are extraordinary safety issues that
require it to be external to the sponsor or to have external DMC members

The structures, members, roles and responsibilities of
all parties are clearly defined
For example, does the DMC take into account important safety data
originating from other trials within the same program?
If the DMC recommendation impacts drug supply, the communication
between the DMC and drug supply must be defined to ensure that the
integrity of the trial is not impacted
7/21/2015
19
Data Monitoring Charter (DMC) Charter

The key part of any DMC charter is to define in detail
the communication process flow and any operational
issues such as conflict of interests, quorum, meetings’
purpose, timings and decision processes

The charter must document what information will be
provided to the DMC members and how the DMC
communicate to the ISC if additional reports are
requested

If an adjudication board is required for the DMC, then
the roles and responsibility and information flow to the
adjudication board must be documented in the charter
7/21/2015
20
Organizational structure
Sponsor Steering Committee
Study Team Lead Members
Blinded Endpoint
Adjudication Committee
Study Team
Blinded
Unblinded
Reports
Adjudicated
endpoints
Independent Statistical Center
Reports
Clinical
Data
Queries and requests for
additional reports
Clinical data
DMC Liaison
Reports
Queries and requests for
additional reports
Data Monitoring Committee
7/21/2015
21
Recommendations
Communication Flow Diagram – Example
Trials
continues
Adaptive
Design Model
Requests Stop
Trial (efficacy,
futility)
YES
YES
Communicate
to Steering
Committee
Committee
Agrees?
DMC
convene a
meeting
YES - for
stopping an arm,
applies only if the
decision is not to
inform the sites
Stop the
Trial or
an arm?
Stop
Randomization
(an arm)
YES - for stopping an arm, applies only
if the decision is to inform the sites
YES
Communicate
to Study
Team
Full Interim
Analysis
Stop Trial
(safety)
Stop
Randomization
(trial or an arm)
YES
Full Interim
Analysis
Stop an arm
(safety)
7/21/2015
22
Communicate
to Sites
ACCEPTANCE PHASE


Once all the programs are validated and quality controlled,
programs are moved into the production area and the acceptance
phase starts
During the acceptance phase, the reports are generated in blinded
format
 The goal is to ensure that the formats of the reports are acceptable to all reviewers
and to add any additional reports that are requested by the DMC.




The acceptance phase is used to ensure that the system is
functioning as designed
For the acceptance phase, live data can be used (normally around
5 subject’s data)
Once all reviewers agree with the results, the execution phase
begins
It is critical that adaptive design software is executed during the
acceptance phase to ensure that the results are accurate
7/21/2015
23
EXECUTION PHASE





The execution phase begins once the acceptance phase is
complete
The time-points for transferring data are defined in the DMC
charter
Also, it is defined in the charter and/or SAP what reports are
produced for what group (i.e., what DMC receives, what the
adjudication board receives, etc.)
Data are transferred via secure connection to ISC
ISC produces reports and performs a quality control prior to
releasing these reports to DMC
 In general DMC reports should be produced with unblinded treatment displayed, however,
appropriate data security must be in place
 Hence, it is important that the reports are stored in secure location for DMC to review.
 Patient listings should be limited and should be justified based on DMC purpose
7/21/2015
24
EXECUTION PHASE

DMC meets according to the schedule that is defined in the charter and
meeting minutes are noted and stored in a secure location

For adaptive design studies, there should be time allocated aside for
DMC to deliberate when adaptation is executed

DMC follows the charter when it requires communicating its
recommendation

The adaptive design software output consists of a revised set of
randomization scheme, as well as predictive probabilities for key trial
outcomes

The randomization file is forwarded to the person who is responsible for
loading the new randomization probabilities into the interactive voice
randomization system (IVRS), verifying the new randomization tables,
and informing DMC that a new randomization table has been loaded
7/21/2015
25
EXECUTION PHASE

Additional Reports Requested by the DMC
The chair of the DMC may ask the ISC to produce additional new reports
and/or to prepare an interim report package at unscheduled time points
These requests must be validated and quality controlled prior to release to
the DMC and also documented by the chair or designee in a secure
location
It is understood that the adaptive design software must not be modified
during the trial and hence any additional requests must not interfere with
on-going execution of the adaptive design
7/21/2015
26
EXECUTION PHASE

Availability of Clinical and PK Data
Availability of the clinical data is important in order to ensure that
DMC has sufficient up to date data to make a correct
recommendation
Sites need to be informed that this is an adaptive study and hence
the data need to be entered in the electronic data capture as soon
as it is available and if PK data plays a role in DMC
recommendation, then these data need to be available as soon as
they are collected
It is advised that a mechanism is put in place to monitor the
availability of data versus the expected data, so that the sites are
informed when there are substantial discrepancies
7/21/2015
27
EXECUTION PHASE

Cleaning of Data
The key is to identify the parameters that need to be cleaned as soon as
they are available, for example the end points that are used in the adaptive
design software, entry of randomization numbers into the database, safety
parameters of interests and any other key data that impacts DMC
recommendation

Updating Randomization Scheme
The adaptive design software should update the randomization system
automatically - hence a utility is built for this purpose
The utility should also allow DMC recommendation to be implemented such
as dropping an arm for safety that is not included in the allocation algorithm
7/21/2015
28
EXECUTION PHASE

Impact on Drug supply
With an adaptive design, the drug supply requirements at a given site are
less predictable. Thus, it is a particular challenge to insure that the
required treatment is available at the point of randomization
We need to avoid having to use an alternative treatment because the
assigned treatment is not available (“forced randomization”)

Reproducibility
Each time that the interim reports are produced, the programs, inputs, logs,
output, analysis datasets, seed number that was used to generate the
adaptive randomization number and raw data need to be stored in the
secure location
Everything required to be able to re-run the adaptive design and obtain the
same outputs must be saved in order to ensure the reproducibility of the
results in future
7/21/2015
29
EXECUTION PHASE

Decision Outcome
It is understood that the DMC committee will serve strictly in an advisory
capacity and its recommendations are non-binding
Formal implementation and communication of DMC recommendations will
be managed by the sponsor steering committee (SSC)
It is SSC responsibility to endorse DMC recommendation and
communicate their decision to the appropriate group
7/21/2015
30
CLOSE OUT ACTIVITIES

Once the trial is completed all documents must be
stored in a secure location so that they can be
accessible for any audits
7/21/2015
31
CONCLUSION

Learn phase clinical trials incorporate adaptive randomization
have a number of advantages:
 Potential to accelerate the identification of promising compounds
 Improve the precision of the key dose-response information required for phase III
success
 Allow the efficient elimination of ineffective or poorly-tolerated compounds from
additional testing

Conducting such trials introduces additional logistical complexity
 Requires unblinded trial data be available in a timely fashion while maintaining data
security
 Interfacing of multiple software systems
 Processes to address the information flow between DMC and other parties

In many situations the advantages of these designs justify the
effort to deals with these logistical challenges
7/21/2015
32