Transcript Identifying - Biopharmaceutical Network

Drug Supply Management
for Adaptive Trials
Micheline Marshall (Pfizer)
Nancy Burnham (GlaxoSmithKline)
NitinPatel (Cytel Inc)
March 12, 2010
On behalf of the Adaptive Design Working Group Mfg. & Drug Supply Work Stream
Introduction

Management of the drug supply is often more complicated than
traditional studies in adaptive clinical trials.

This is particularly true for those studies where modifications
involve:



Number of trial subjects may be increased or decreased

Dosages may be dropped or added

Allocation to dose groups is modified
Complications are:

The amount of drug that will be required is unknown

May need to mask modifications to dose in order to minimize bias

Need to ensure availability of drug to sites at adaptive decision points
Efficiency of drug supply is necessary for the success of most
adaptive trials
2
Introduction, cont’d.

Currently, most companies do not have systems, tools, and
processes designed to manage the unique demands of
adaptive trials.

As a consequence, most companies either:


Outsource all or part of trial operations to an expensive vendor
Devise complicated work-around

Adaptive clinical trials are still new within the industry with
many companies just becoming involved in this initative.

The PhRMA Adaptive Designs Working Group formed the
Manufacturing & Clinical Trial Supply Substream to


Share experience across the industry
Communicate good practice and guidance (manuscript near completion)
3
Adaptive Design Working Group –
Manufacturing and Clinical Trial Supplies Substream

Abbott: Ajay Shah, Dave Shah

ClinPhone: Graham Nicholl

Cytel: Judith Quinlan*, Nitin Patel

Tessella: Tom Parke

Lilly: LB Wong

Merck: Weili He

Pfizer: Micheline Marshall

GlaxoSmithKline: Nancy Burnham*, Angelique Ditoro
*Co-chairs
Presentation authors
4
Topics

Randomization / Interactive Response System

Communication Strategy

Supply chain strategy

Simulation Tools

Finance

Key Points
5
Interactive Response Systems

IVRS/IWRS – Interactive Voice/Web Response
System:

Automated centralized system

Allows sites to randomize subjects via the telephone or
web

Assigns clinical supplies at each dispensing visit

Controls site and/or local depot clinical supplies
inventory based on real time utilization
6
Value of IRS in Adaptive Trials


Central tool to update changes to randomization:

Immediate links new randomization schedule to package
numbers

Flexible and efficient inventory management to meet the
needs of adaptive changes to treatment allocation
Robust audit trail to reflect modifications

Date / time stamps
7
Supply Chain Strategy:
Cross Functional Collaboration

To achieve an effective drug supply outcome, it is important the
entire project team discusses the following:

Identify the adaptive modifications in the protocol that would ultimately
affect the supply demand.

Commit to timelines around decision.

Explore likely scenarios of dosing configurations as the study
progresses.

Agree to how many capsules/tablets/injections are acceptable for the
patient per dose.

Define the dose-strengths to be manufactured.

Agree on dosing strategy for ongoing subjects assigned to dropped
doses.

Evaluate different packaging options to minimize overage, maintain
flexibility and achieve appropriate level of masking.
8
Clinical Supply Challenges

Responding to decision points:

Are the right doses available at sites?

How are dose modifications communicated to supplies group?

When is adjusted inventory required at sites?

Has sufficient time been factored to complete inventory
adjustments?




Changes to the randomization
Updating the IRS databases
Adjust inventories to sites
Factor in the “knowns”:

Min/max number of treatment types to be assigned

Min/max subjects per treatment regimen

Is the ratio of comparator subjects known
9
Distribution Logistics

A good strategy can cope with global distribution
as changes to treatment allocation evolve.

Number of countries

Number of depots

Number of sites

All will affect the dilution of supplies
10
Actual Study Examples
GSK
Pfizer
11
Example: 4 week dosing to prevent thrombosis
after hip replacement (GlaxoSmithKline)
1 tablet from
each bottle,
twice per day
Bottle A
Bottle B
Bottle C
Bottle D
Placebo
Placebo
Placebo
Placebo
Placebo
125 mg BID
Placebo
Placebo
Placebo
Odiparcil
250 mg BID
Placebo
Placebo
Odiparcil
Odiparcil
375 mg BID
Placebo
Odiparcil
Odiparcil
Odiparcil
500 mg BID
Odiparcil
Odiparcil
Odiparcil
Odiparcil
12
Example: Acute Treatment of Migraines
(GSK with Tessella tech support)

Seven doses:


5, 15, 30, 60, 120, 180 mg
Placebo

Subjects dosed at onset of migraine

Endpoint – response after 2 hours

Randomization updated after every patient
response


Continual reassessment method
Increase chances patients receive best dose
13
Migraine example cont’d
14
Migraine Example, cont’d

3 pack types with dosages in different order

Subjects randomized to a pack type

Upon migraine onset subjects call IVRS and told which
tablet to take

After 2 hours subject calls IVRS and keys in response

Randomization system updated for next subject
15
Forecasting Tool
ex: Pfizer (Wyeth) Diabetes
Trial

Tessella custom developed tool:

Tracked treatment inventories at site.

Provided predicted requirements based on 99% and 95%
certainty of randomization revision.

Predictions only based on patients within 4 days of end
of screening period to prevent calculating demand on
dropped patients.

Provided “pick list” of supplies required by site to
accommodate the updated codes.

Information provided to Clinical Supplies one week prior
to having codes loaded into IVRS.
16
17
18
Long Term Considerations

Process flows must be developed to detail the
communication between all involved parties.

IT must be involved to ensure smooth integration between
systems.

Blinding techniques must be developed to automatically
mask pack code gaps due to adding new treatments mid
study.

Build strategy to add new treatment arms mid-study
without alerting the site personnel.

Develop an on-demand strategy to maximize supply
flexibility and minimize supply overages.
19
Masking Pack Codes
Planned into code generation:
Treatment Arm A (known and assigned at study start)
1001
…
1005
5600
Treatment Arm B (known and assigned at study start)
…
1002
1008
5611
Treatment Arm C (dose unknown and added mid study)
1003
1006
???
Treatment Arm D (dose unknown and never utilized)
1004
1007
20
Drug Supply Modeling and Simulation - 1

Computer simulation models of the supply chain
can provide reliable answers to important
questions about drug supply at the planning
stage

Examples:

What quantities should be manufactured and packed?

What is the effect of multiple manufacturing and packing
campaigns?

What is the effect of adding more countries and sites?

How much API can we save by combining pack types to
make up different dosages?
21
Drug Supply Modeling and Simulation - 2

Drug supply simulation models can also enable
efficient operations during the conduct of a trial

Examples:




Adjusting trigger levels and resupply quantities to reflect
actual enrollment rates at sites
Predicting future enrollment at sites based on actual
enrollment to-date
Predicting requirements for different pack types at sites
taking into account new enrollments as well as revisits
and drop-out rates
Suggesting transfer of stocks from low recruiting sites to
faster recruiting sites
22
Leveraging statistical simulations in drug
supply simulations

Design of almost all adaptive trials requires running
simulations to determine the statistical operating
characteristics for a set of likely scenarios. These simulations
compute randomization information for each subject in the
trial that reflects the adaptive aspects of the design. This
randomization information is saved in a file for use by the
drug supply simulation software.

Drug supply simulation software takes various inputs from
the user that describe site accrual rates, dosages, packtypes and supply chain parameters. It then reads the files of
randomization information created by the adaptive design
simulations to compute the behavior of the drug supply
chain for the adaptive design.
23
Case Study: Estimating drug
requirement at start of study

Bayesian Dose Adaptive Trial Design

Placebo-controlled, double-blinded, phase II dosefinding trial with 7 active doses. Sample size is 120
subjects.

Ten cohorts of 12 subjects each enrolled during the
study, with first cohort having 4 subjects on placebo, 2
on dose 4 and 1 each on all other doses

Subject responses in each cohort are used to update
randomization ratios for active doses for the next
cohort

Single packs for each dose (placebo or active dose),
requiring 120 kits for 120 subjects
24
Overage Estimate (no simulation)

Allocation of active doses in first cohort is known but for the
remaining nine cohorts it depends on the responses
observed. A conservative approach is to provide 72 kits
(worst case = 9 cohorts x 8 subjects per cohort) for each
dose of the study drug. Total number of kits required is 552
(40+2+1x6+72x7); overage is 360% [(552-120)/120]

If we consider specifics of the adaptive design and likely
scenarios considered in the design we can simulate the
chances of extreme requirement outcomes and arrive at a
better balance between risk of randomization failure (stockout) and overage.
25
Adaptive Design
The design was constructed by simulating the statistical
operating characteristics using three likely dose response
curves
25
Fla t Re s p o n s e :S c e n a r io A
Me d iu m Re s p o n s e :S c e n a r io B
Hig h Re s p o n s e : S c e n a r io C
20
M ean R esp o n se

15
10
5
0
0
1
2
3
4
D ose
5
6
7
26
Sim#
Subject#
Cohort#
Dose
Response
Adaptive Design Simulation:
Randomization Output File from Scenario C
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
6
7
5
3
4
4
0
0
1
0
0
1
0
0
1
10
32.9
18.5
26.1
2.13
20.6
12.4
-8.5
-0.4
17.6
-5.7
0.61
7.99
9.06
20
9.75
27
Overage Estimate using Simulation

Drug supply simulations were run using the
randomization files from design simulations for each of
the three likely dose response curves.

Running 500 drug supply simulations showed that
there would be no stock out under any scenario if 216
kits were prepared.

Thus the estimated requirement of 552 (360%
overage) without simulations is reduced to 216 kits
(80% overage) if a stock-out probability of less than
1/500 is acceptable

We can run more simulations if we wish to estimate
drug requirement for a lower stock-out probability
28
Evaluating Financial Benefits

Beyond the study level consider a financial comparison of
development strategies

Consider financial implications of different strategies in a
clinical development program in which adaptive designs are
used instead of, or in addition to traditional designs
29
Financial Considerations:
Increased Costs

Drug supply costs for an adaptive design is apt to be higher
due increased overage.

Randomization costs are also likely to be higher for designs
that require frequent changes in randomization ratios.

Continual Reassessment Models may also require expensive
outsourcing of randomization to a vendor with a more
sophisticated system.
 E.g.: GSK migraine study
30
Managing Increased Costs

Overage is highest when the maximum for each dose is
drug packaged in advance to accommodate every possible
scenario at every site.

Overage can be reduced by an adaptive strategy that
adjusts while the trial is ongoing to accommodate
algorithmic modifications suggested by the accruing
clinical data.

The best approach for a given trial will be driven by amount
of drug available, the number of regions, and recruitment
rate.
31
Managing Increased Costs, cont’d

Consider redistributing supplies from slow recruiting sites to
fast recruiting sites, closing slow recruiting sites.

Consider the feasibility of slowing or suspending
recruitment at decision points so that sites can be
resupplied.

If at trial is long enough, there may be opportunity for
multiple manufacturing and/or packaging campaigns to
preserve material that isn’t needed.

If a trial allows an adaptive increase in sample size think
about opening new centers. Look for opportunities to pool
supplies across trials.
32
Weighing Increased Costs Against
Benefits

At the study level, increased drug supply costs maybe
offset by the benefits of stopping the trial early and reduce
the number of subjects and other operational costs

At the program level, increased drug supply costs maybe
offset against improvements in development strategies
 Combining objectives
 PoM and POC
 POC and dose ranging
 Improved dose selection for phase III
 Reduced late phase attrition
 Lower down-stream costs and higher revenues.
33
Financial Implications
Example: Wyeth Trial

Additional supplies were manufactured, packaged and
stored at regional warehouses to accommodate evolving
supply demands

Overall cost of drug supply for this study:
 Cost of adaptive design:
$422,000
 Number of patient kits packaged:1440
 Cost of traditional design:
$201,000
 Number of patient kits packaged: 686

But, savings to Clinical for closing study 2 months earlier
and 180 less patients
$1.5 million
34
Key Points:



Efficient trial supply for an adaptive clinical trial
requires:

X-functional discussions

Modeling and Simulation

Supply Forecasting
There’s no one-size fits all packaging and distribution
strategy:

Useful tools: IVRS, forecasting tools

Flexibility in packaging

Imaginative thinking
Potential greater cost of adaptive trial should be
considered in the context of benefit to overall asset
development strategy
35