TRANSCEND: Using caBIG to Support Adaptive Clinical Trials

1
TRANSCEND Using caBIG to Support Adaptive
Clinical Trials

• Michael Hogarth,MDProfessor, Pathology and
  Laboratory MedicineProfessor, Internal
  MedicineUC Davis School of Medicine
• http//www.hogarth.org

TRANslational Informatics System to Coordinate
Emerging Biomarkers, Novel Agents, and Clinical
Data
2
The drug pipeline
3
From Discovery to Therapy
4
Problems

• The drug discovery pipeline is inefficient and
  very costly
• Averages 10-15yrs to complete
• Clinical trials information infrastructure (the
  pipeline) is paper-based and disjointed -- making
  it difficult to be more efficient.
• Biomarkers are showing promise in informing
  treatment choices, but validation of biomarkers
  in the clinical trial process has proven to be
  difficult - the biomarker barrier

5
How do we do clinical trials today?
develop protocols
archive records
the science
create forms
enroll and care for patients
chart clinical care
store records
6
Adaptive Clinical Trials

• A clinical trial design that attempts to reduce
  cost and determine efficacy faster
• Describes class of trial designs where data is
  used to modify dosing or other parameters --
  group sequential, staged protocols, Bayesian
  designs
• Bayesian Design
• efficacy is a probability, and the probability is
  re-calculated with new information on response to
  therapy in the trial

7
I-SPY 2 A Bayesian Adaptive Trial

• A neoadjuvant Phase 2 trial in women with large
  primary cancers of the breast (gt3.0cm).
• Compare efficacy of novel drugs in combination
  with standard adjuvant chemotherapy
• The goal is to identify improved neoadjuvant
  treatment regimens for patient subsets on the
  basis of molecular characteristics (biomarker
  signatures)
• Regimens showing a high Bayesian predictive
  probability of being more effective graduate from
  the trial - with their biomarker signatures
• Regimens with low probability of being effective,
  are dropped

8
I-SPY 2
9
Informatics Aspects of I-SPY 2 A Bayesian
Adaptive Trial

• Manage information across multiple sites
• Data gathering must be closely monitored as the
  trial depends on rapid eligibility determination
  and therapeutic intervention -- do not want to
  delay standard therapy
• Combining evaluation of drugs and biomarkers
  together -- biomarker data can be of multiple
  types (arrays, imaging volume, numeric scales,
  etc..)
• Scientists need access to data early and in an
  integrated fashion (one stop shopping)
• Randomization as a service (automated - but with
  review)

10
TRANSCEND Objectives

• Develop an information management infrastructure
  to support adaptive clinical trials like I-SPY 2
• Demonstrate integration of a clinical system
  (electronic health record system) with a clinical
  research infrastructure
• Provide a demonstration of caBIG infrastructure
  in use in a large multi-center trial
• Support patient-centric interactions (pt calendar)

11
Why integrate clinical information systems and
clinical trials data capture?

• Routine clinical care and clinical trial care are
  often conducted
• In the same physical space...
• By the same people...
• Engaging the same patients...
• ...but with separate information systems that do
  not share data, data elements, or common
  information models...

12
Functional RequirementsWhat does TRANSCEND need
to do?

• Manage the patient registration lifecycle
• Manage eligibility determination
• Randomize patients
• Track study participants
• Manage bio-specimens
• Capture clinical data at the point of care and
  render CRFs using automated methods
• Provide traditional web-based CRFs
• Manage patient and treatment planning calendars
• Initiate the adverse event lifecycle
• Storage and retrieval of trial data for each
  participant or in aggregate

13
Functional RequirementsWhat will do it in
TRANSCEND v1.0?

• Manage the patient registration lifecycle -
  Tolven eCHR
• Manage eligibility determination - Tolven eCHR
• Randomize patients - MD Anderson Randomization
  Engine
• Track study participants - Tolven eCHR
• Manage bio-specimens - caTISSUE
• Capture clinical data at the point of care and
  render CRFs using automated methods - Tolven
  eCHR
• Provide traditional web-based CRFs - Tolven eCHR
• Manage patient and treatment planning calendars -
  Tolven eCHR
• Initiate the adverse event lifecycle - Tolven
  eCHR --gt caAERS
• Storage and retrieval of trial data for each
  participant or in aggregate - caINTEGRATOR

14
TRANslational informatics System to Coordinate
Emerging biomarkers, Novel agents, and clinical
Data (TRANSCEND)
15
TRANSCEND Status

• 21 of 21 CRFs designed, reviewed, and currently
  being programmed into the Tolven system
• 5 clinical data capture workflows undergoing
  final reviewed prior to being implemented in the
  Tolven system.
• 80 of 1,200 data elements encoded
• 5 of 6 systems already installed and running in
  UCSF Cancer Center data center -- in anticipation
  of final application deployment.

16
What is new or different about TRANSCEND?

• Randomization web service
• Using a clinical information system rather than
  CTMS to collect patient data for CRFs
• Integration of caTISSUE with a clinical
  information system in the context of a trial
• Use caINTEGRATOR v2.0 as a scientist portal to
  study data

17
TRANSCEND Challenges

• Sept 2009 hard date -- system must be
  developed, tested, deployed, and ready for use
• We are integrating some components in novel
  configurations for the first time
• 1,400 data elements, 500 mapped to caDSR
• I-SPY 2 requirements can change, so we must be
  adaptive in our development (agile)

18
TRANSCEND Informatics Strategy

• Manage risk
• try to reduce dependencies
• whenever possible, use tested/final versions of
  software rather future versions
• Install these systems in a data center up-front
  rather than later
• Manage integration scope
• adopt a enterprise bus architecture - caXchange
• avoid point-to-point interfaces between systems
• Engage users early and use a continuous feedback
  for user-center design components (UI, workflows)

19
TRANSCEND Lessons - so far

• Do not underestimate the number of data elements
  you might want/need to code
• Identifying and managing data elements is a
  significant undertaking
• Install systems early to deal with logistical
  issues far in advance of needing the systems
• Involve actual users in UI/workflow reviews
• Integrating with clinical systems demands an
  enterprise approach to information exchange
  (caXchange, HL-7)
• Move towards enterprise bus as an integration
  rather than API-based point-to-point integration

20
Acknowledgements

• Laura Esserman (UCSF)
• Meg Young, Kathy Hajopolous, Sarah Davis, Sorena
  Nadaf (UCSF)
• Nancy Roche (SAIC-F)
• John Koisch
• Subha Madhavan (Georgetown)
• Davera Gabriel, RN (UC Davis)
• Dan Milgram (CCS)
• Christo Andonyadis (NCI) and the NCI CAT Team
• George Komatsoulis and John Speakman (NCI)
• Funding for TRANSCEND provided by the National
  Cancer Institute (NCI)

21
Questions?