Biomarkers and Surrogate
Endpoints in Drug Development
Technical and Regulatory Considerations
Gracie Lieberman,
Genentech
2006 FDA/Industry Workshop
Content
Changing landscape
Efficacy surrogate endpoints used for
approval – case studies
– Herceptin
– Iressa
Surrogate endpoints in proof of concept
trials
– Selecting sub-population
– Selecting dose
Role of mechanism-based biomarkers
– PET studies in cancer
Changing landscape
In the past 15-20 years:
– Better molecular understanding of diseases
– Earlier, more precise diagnosis
– New targeted, improved therapies
Impact on clinical trials
– Assessment of improvements in clinically meaningful endpoints
require enrolling many patients and then following them for a
long time.
Emerging need
– Develop strategies for reducing the time and cost of drug
development.
– Use of surrogate endpoints either in proof-of-concept or labelenabling trials.
Defined in clinical practice and used by clinicians to monitor and
treat patients
New mechanism-driven biomarkers
Endpoint considerations
Study defined endpoints supporting product
labeling
– Demonstrate clinically meaningful benefit
– Relevant to a specific indication and study population
– Reliable and reproducible
Study defined endpoints supporting early
decisions
– Correlated with clinically meaningful outcomes
– Relevant to a specific indication and study population
– Sensitive to small sample sizes
Pharmacodynamic markers
Surrogate endpoints
Case Studies
Herceptin – PFS as an endpoint
Iressa – Risks of accelerated approval
Herceptin in MBC
Herceptin a is recombinant DNA-derived
humanized monoclonal antibody that
targets HER2, the protein product of cerbB-2.
In September 1998 Herceptin was approved
for:
– First line treatment in combination with paclitaxel
in MBC patients whose tumors overexpress HER2.
The primary endpoint used was not overall
survival (OS) but progression free survival
(PFS).
PFS as a primary endpoint
How to assure objectivity and minimize bias
– Randomized, placebo control study
Weekly placebo infusions for months
Impact on enrollment
– Strict schedule of efficacy assessment
– Independent review of radiographic images
Process for collecting images
Assessment of skin lesions required distribution of
cameras to sites
– Strictly define rules for missing data
Independent review of images not available
PFS as a primary endpoint
The challenge continues
– 21% enrolled in the first year
– Protocol amendment to remove placebo
Impact on primary endpoint (PFS)
– Invoke real time independent review of
radiographic images
Primary endpoint had to be confirmed by the review
committee
Patients and investigators compliance
– Turn-around review time was critical
– Offer cross-over to control patients
PFS as a primary endpoint
Conclusions
– In September 1998 Herceptin was approved
based PFS
– Survival as a secondary endpoint was
statistically significant
– 65% of control patients crosses-over to
receive Herceptin
– Two years later the survival benefit
continued to be present
Sub-group analysis impacted by crossover
Case Studies
Herceptin – PFS as an endpoint
Iressa – Risks of accelerated approval
Iressa in NSCLC
Iressa a quinazoline-based small molecule,
an EGFR TK inhibitor
In phase I objective responses observed in
NSCLC
Two phase II monotherapy trials
– Response rate (RR) as primary endpoint
– Two dose groups
May 2003 - Accelerate approval for 3rd line
monotherapy use based on RR
Accelerated approval - risks
Need to conduct large, confirmatory trials
– What if negative?
Despite meaningful responses in recurrent
NSCLC patients, Phase III trials failed to show
any significant clinical benefit
Approval was revoked in June 2005
– The medicine should be used only in cancer patients who have
already taken the medicine and whose doctor believes it is helping
them. New patients should not be given Iressa because in a large
study Iressa did not make people live longer.
What went wrong?
– Patient selection ?
– Dose/schedule ?
Can this be avoided?
Demonstrating clinical benefit with moleculartargeted agents is more complex than with
conventional cytotoxic agents
– Selection of sub-population: who is most likely to
benefit
– Identification of optimal biological dose
Answers before phase III – is this achievable
Proof-of-concept trials
– Is PFS a sufficient endpoint
Sub-population selection
Complex process
Tissue samples required
– Blood/serum feasible
– Tumor samples are challenging
Missing data
Archival samples not always relevant
Randomized, controlled studies required
– Stratification by biomarker for sub-population selection
At randomization or during analysis
– Not possible to distinguish between a prognostic and
predictive biomarker without a proper control
Biomarker based population selection
PFS with no control arm
Proportion progression-free
1.0
0.8
0.6
Median PFS
pHER2 positive
n=8 20.9 weeks
pHER2 negative
n=20 5.8 weeks
pHER2 unknown
n=27 9.1 weeks
All subjects
n=55 6.6 weeks
All patients treated with pertuzumab
0.4
0.2
0
0
100
200
300
400
Time (days)
pHER2+ tumors trend toward longer PFS
Treatment effect or natural course of disease?
Gordon et al. J Clin Oncol. 2005;23:16S (abstract 5051).
Gordon et al. J Clin Oncol. In press.
Dose/schedule selection
Complex process
– May be indication specific
– May be regimen specific
Typical trial
– Randomized
– 3 arms
Control/lower dose/higher dose
30-40 subjects per arm
– PFS as primary endpoint
How is dose selected
– Better efficacy compared to control - winner
Time to event endpoints
Optimal vs. sub-optimal dose
Probability that the observed HR  0.75
True HR
Number of events
40
60
80
100
200
0.67 0.64
0.67
0.70
0.72
0.80
0.80 0.42
0.40
0.39
0.37
0.33
0.90 0.28
0.24
0.21
0.18
0.02
We need to do better
Mechanism-based biomarkers
Demonstrating clinical benefit with moleculartargeted agents is more complex than with
conventional cytotoxic agents
– Escalating clinical trials costs and large numbers of
patients required for currently used clinical endpoints
mandate becoming more efficient in determining how
well new agents can address unmet medical needs.
– That efficiency can be achieved by validating
correlations between specific biological mechanisms
of disease and clinical outcomes.
– Easier said than done!
Mechanism-based biomarkers
Technological advances provide great
opportunity for the development of biomarkers
– Molecular and cellular techniques
Tissue samples
– Tumor/blood/surrogate
– Imaging technologies
Current pre-clinical models still have limit ability
to predict clinical effects
– Biomarkers need to be co-developed with the novel
agent
In early phases – no clinical data
This will benefit second generation of the new agents or new
indications
– Systematic way of analyzing and interpreting data
Mechanism-based biomarkers
PET imaging
Use of surrogate endpoints in cancer
prevention
PET studies
Speed development
Provide information about the activity of
molecular pathways
Determine if new agents are hitting the target
Measure treatment effect
Tissue samples are not required
FDG PET in NSCLC
Wolfgang Weber et al. J Clin Oncol. 2003;21:2651
FDG PET in Lymphoma
L. Kostakoglu, J Nuc Med 43:1018 2002
Challenges
How to define metabolic response
– Change in standard uptake values (SUV) that based
on re-tests can be reliably detected
Arbitrary cut-off
– Optimized thresholds correlated with outcomes
Based on analysis
– What adjustments made for minimum p-values
Not applicable to other treatments or indications
Use of core labs in multi-center trials
Not ready as a surrogate efficacy outcome for
combination trials
Not all lesions are PET avid
Cancer Prevention
Preventing heart diseases
– Lowering cholesterol / blood pressure
Surrogate biomarker endpoints for cancer
prevention trials
– Establishment of long term safety and efficacy
for preventive drugs is critical
– Process for accelerated approval based on
biomarkers will be needed
Colorectal adenomas
– Current development of mechanism-driven
biomarkers is critical for future cancer
prevention trials.
Questions?
Thank you!