Tumor volume and Clinical outcome – Is

there a role for re-use of high quality

quantitative image databases ?

Gudrun Zahlmann
Imaging Biomarker Roundtable Ad hoc
committee on ‘Open Image Archives’
Imaging biomarkers for lung cancer
• Diameter measurements of lung lesions – qualified biomarker.

• Volumetric assessments , where possible:

– Evidence that volume change is more sensitive than diameter change
– No evidence so far that this is beneficial in terms of clinical development time.
– Not accepted as surrogate endpoint due to missing confirmation of relationship
to clinical outcome

• RECIST 1.1 based on sum of largest diameters:

– Originally coming from phase II assessments;
– in version 1.1. some of the problems for other phases as well clarified
– is currently the standard accepted by regulatory bodies as surrogate endpoint
in controlled studies

What about volume measurements based clinical outcome

assessment in clinical trials ?
FDA view – April 2010 workshop: J.
Woodcock
• Imaging as a Biomarker:
– Concept: Biomarker qualification is a conclusion that within the
stated context of use, the results of biomarker measurements
(e.g. imaging) can be relied upon to have a stated interpretation
and utility

– Context of use must be clearly specified.

– Regulatory implication: industry relies upon the use of the

biomarker in the qualified manner in IND, NDA and BLA
submissions, without a need to resubmit data for the relevant
CDER review group to consider and reconfirm biomarker usage
FDA qualification process – April workshop: J.
Woodcock
• Process in General:
– A framework for interaction between CDER and sponsors so that
CDER can provide guidance towards compiling comprehensive
evidence to support qualification of the selected biomarker(s)

– A mechanism enabling CDER to have a well-organized, multi-

discipline, CDER-unified, formal review of the data supporting a
biomarker, eventually leading to a CDER decision on
qualification

– Enables a scientifically well-supported statement by CDER of

qualification providing confidence that the evaluation has been
comprehensive and the conclusions can be relied upon
FDA qualification process – April workshop: J.
Woodcock
• Process in reality:
– Formalized process still under development

– Sponsors brings imaging biomarker concept / information package to

CDER

– Information package reviewed

– Advice given on how to further progress development for intended use

• Consultation and advice
• Continued until development is complete

– Full detailed CDER review and decision on qualification

– Formal statement of qualification if appropriate

Possible ways forward

• To develop Volumetric measurements as novel biomarker for clinical

outcome and qualify it as a novel approach on its own

• To develop Volumetric measurements as replacement of diameter

measurements in RECIST

• To develop a combination of 2D – 3D measurements as biomarker

What is needed ?
• Clear outline of the question that we
want to answer – imaging biomarker
that can be used as a surrogate for drug
related tumor response in lung cancer,
esp. NSCLC that is more sensitive,
accurate than SLD.
• Reliable clinical and study related
imaging procedure.
• Reliable imaging analysis.
• Reliable data management (do not
corrupt).
• Proven Relationship of the imaging
biomarker to clinical outcome.
• Understanding what imaging procedure is
relevant for the question that we want to
answer – high resolution CT.
• Definition what do we mean by high resolution:
is it 0.75 or 5 mm? Other scan parameters?
• Selection of relevant scanners.
• Standardized study protocol for all relevant CT
scanners.
• Site qualification according to the protocol.
• Standardized phantom scans across scanners.
• Ongoing quality inspection on a per scan basis.
• ….
 Imaging procedure – ongoing consensus
activities
LIDC RIDER Quantitative imaging (data collection Phantom design
and QI-clin decision tools to measure
2005 - 2010 response to therapy) Longitudinal repeat
studies
Implementation model (phantom data
collection, volunteer data collection,
patient retrospective data)

Compare tool performance

QIN Quantitative imaging (data collection Broad consensus for QI (15
-clin decision tools to measure responseUS centers by Sept 2010,
2010 - to therapy) CCs, CR UK Imaging
centers)
Implementation model (clinical trials in
cancer centers and other oncology trial Working groups (QC
groups methods, data collection –
imaging protocols,
Retrospective – prospective data archive-informatics
collections web assessable solution, benchmarking
L. Clarke presentation April workshop clinical decision tools)
LIDC / QIN
- what we can learn
This is a serious effort:
Spiral CT Scanning, Dose, and Quality Control Committee
Inclusion/Exclusion Criteria Committee
Spatial Ground Truth Issues Committee
Pathologic Ground Truth Committee
Database Committee
Common Metrics; Software Tools Committee
Past Funding Opportunities Committee
Internet Availability Committee
Industry Liaison Committee
FDA Liaison Committee
ACRIN Liaison Committee
Publications Committee
Statistics Committee
NIH Group
It takes time and resources to build it-
It is still very US centric and would need ways to be multiplied internationally.
Quality assurance / quality control
Phantom imaging

picture placeholder FDA phantom initiative: Excellent for scanner

validation in an ongoing initiative. Will this
become a standard for clinical trial related CT
scanner and SW assessment?

picture placeholder Per scan phantom imaging: Ongoing

investigation used in Abigail. Modelling and
possible correction for the individual image data
set per time point.
Imaging process description for clinical
trial applications

• QIBA profiles
– Process standardization
– Medical context
– Multi stage process
– Filling a gap

• Still in early stages and not widely accepted / used

• Stakeholders engagement might assure broad use in the future.

Translate guidelines and templates into
real life trial

• Example Abigail
– Take evolving QIBA profiles into consideration.
– QA/QC based on per scan phantoms plus additional procedures according to
SOPs.
– Multinational multi-centre is a challenge - But it is doable.
– Qualified project management.
– Qualified image analysis site for RECIST (diameters) and volumes.
– Experience in qualification process with FDA.

We can build evidence and databases that can be used to discuss with
regulatory bodies relationship between volumetric measurements and
clinical outcome. There is first experience in FDA – sponsor
relationship.
Imaging Biomarker Roundtable

• Ad hoc committee:
– Open Imaging archives of high quality imaging data
– Use cases
• Image processing for quantitative lesion asessments –
‚algorithm development‘
• Biomarker qualification of volumetric measurements in
RECIST
• ….
– Develop use cases and work with stakeholders to translate it
into something that can be realized.
Can we use trial databases for clinical
outcome relationship

• Definition who is the sponsor of a database – public and companies

– Legal and regulatory implications

• Definition of data (image and results) quality as a full concept.

• Definition of image management.

• Definition of image interpretation.

– What is a signal based on a thorough variability / error assessment taking into account the missing
ground truth.
– Currently there is no fully automatic solution.
– Consistent image processing given the variability of the existing algorithms.

• Definition of meta data is different for different use cases – experience so far for research data bases
mainly used for algorithm development. We need to agree on meta data for biomarker qualification.
Can we use trial databases for clinical
outcome relationship

• We need to have clarification on legal requirements and need to translate this into
action on an international basis.

• There must be a clearly identifiable benefit for trial sponsors to contribute to such an
activity – faster (?) qualification based on shared effort.

• We need to have a standardized procedure to generate images across studies across

programs – work in progress

• We need to have a clear quality understanding – work in progress

• We need to have an understanding of necessary meta-data that make a database fit

for purpose - first experiences, needs further consideration

• We need guidance from regulatory bodies whether this is a worthwhile effort – first
experiences on a per sponsor basis but need to learn from previous e.g. RECIST
experiences

• We need an independent group that is driving this activity that is accepted by

regulatory bodies.