Personalized Medicine: The Value of Targeted Therapies Disclaimer

Disclaimer
Personalized Medicine: The Value of
Targeted Therapies
Past regulatory decisions related to targeted therapies or
adding genetic tests to labels are not necessarily
precedents for future decisions
IPSOR 15th Annual International Meeting
Atlanta, Georgia
May 17, 2010
Lawrence J. Lesko, Ph.D., F.C.P.
Director, Office of Clinical Pharmacology
Food and Drug Administration
Silver Spring, Maryland
Science, guidance and regulatory policies are evolving
rapidly in different Centers and the future is unlikely to be
identical to the past.
Physicians Have Long Recognized That Each
Patient Is Different From Every Other Patient
Outline of Presentation
1. Variability in drug response
2. Concept of targeted therapies
3. Examples of personalized medicine
4. Evidence and clinical utility
5. Four Barriers to overcome
Most Medicines Represent a SemiSemiEmpirical Approach to Therapy
Action
Sir William Osler (1849 – 1919)
The Johns Hopkins University
The Father of Modern Medicine
Variability in response produced by the same drug in different
patients. It was acceptable (until now) because there were no
ways to conduct individual or subgroup analysis – disease risk,
dose-response, demographics, tolerance to treatments – and this
led to remarkable uncertainty in response
Medicines Are Approved Based on
Empirical Evidence of Efficacy from RCTs
Efficacy expressed as average drug response. Average effects
within a RCT population may be quite different from effects
expressed by a given individual.
One Size Fits All
Observation
“If it were not for the great variability among
individuals, medicine might have well been
a science and not an art”
Trial and
Error
Response
Medicines attempt to achieve a positive benefit/risk
(clinical utility) in large patient populations
independent of target status
Source: Steven Paul (Lilly); Brian Spear (Abbott); Barbara Evans (U of Houston)
1
Limitations of RCTs in Populations: Not
Always the Gold Standard of Evidence
Comparability (Internal Validation)
Validation)
– Assumes a superior treatment (vs placebo) is the best
choice for any given individual in the population
– It could be the wrong
g drug
g for manyy people
p p because of
adverse events
– It could be the best drug for many people who are not at
risk for the adverse event
– Relatively low average responses to many drugs means
nonresponders receive treatments with no value

Generalizability (External Validation)
– Most RCT populations are not representative of patients
who will consume the drug (inclusion/exclusion criteria)
Large Benefit with little harm (10%)
Benefit
Mixed Benefit and Harm (30%). Small
benefit for most.
Magnitude

Every Disease Is Comprised of Biological
Subgroups That Differ In Response to Drugs
Neither harm or benefit --Nonresponders(50%)
Harm Without Benefit (10%)
Harm
Frequency of various responses in the RCT treated population
Adapted from presentation by Dr. Barbara Evans, ASCPT Annual Meeting (2010)
Targeted Therapies Render Value When
There Is Variability and Uncertainty
Targeted Therapy: Most Often DrugDrugDiagnostic Pair
Definition: Targeted Therapy
One Size Fits All
Medicine acting on a cellular target that participates
pathogenesis
g
of a diseases
in the p
Observation
Targets can be polymorphic, have variable
expression or be subject to somatic mutations that
affect drug response
Assays are used to interrogate the status of the
target which provides significantly more predictive
value for a medicine’s response
Targeted Therapies Depend on
Stratification of Diseases
1949
2009
• Leukemia: an umbrella term
for “diseases of the blood”
• Five yyear survival of 0%
• Leukemia: 38 subtypes have
been identified
• Five year survival of 70%
Trial and
Error
Action
Response
Targeted Therapy
Observation
Test
Action
Predictable
Response
Definition of Personalized Medicine
No effective
treatments;
one size fits all
Targeted
therapies,
e.g., Gleevec
for CML where
5 year survival
is 90%
When a biomarker for target status is used to define
a subgroup of patients who have a different
benefit/risk ratio than an unselected population
Most biomarkers are mechanistic and not merely
associations which increases their credibility
Personalized medicine is not new, and it doesn’t
matter if the test is DNA-based or not
Personalized medicine relies on targeted therapies
Adapted from a Presentation by Mara Aspinall on “Future Directions in Personalized Medicine” (2007)
2
Many Types and Applications of Targeted
Therapies in Clinical Decision Making
Representative categories of diagnostic tests that
facilitate personalized medicine:
-------
Diagnosis (LQTS
(
and
d at-riskk d
drug avoidance)
d
)
Prognosis (Oncotype DxR and chemotherapy),
Improve benefit (HER2 and trastuzumab),
Predict lack of benefit (KRAS and EGFRI)
Predict risk (HLA-B*5701 and abacavir)
Estimate dose (CYP2C9/VKORC1 and warfarin)
Targeted Therapies Are Not New and Don’t
Have to Be Genetic
Efficacy (Benefit)
Tamoxifen ((1977):
) onlyy effective in treating
g estrogen
g
receptor-positive breast cancers – test for tumor’s
hormone receptor status before treatment
Imatinib (2001): only effective in treating Philadelphia
chromosome-positive chronic myelogenous leukemia -test for abnormal chromosome in patient’s bone marrow
Examples of Targeted Therapies That
Prevent or Reduce Risk
Example of Targeted Therapies That
Optimize Doses
Safety (Risk)
Increase Efficacy, Improve Safety (B/R Ratio)
Isoniazid (1952): avoid in patients with genetic G6PD
deficiency to prevent hemolytic anemia and jaundice –
test for G6PD activity in blood sample
Fluorouracil (1970): avoid or use at profoundly lower
doses in patients with genetic DPD deficiency to
prevent neutropenia and other toxicities – full sequence
PCR test of the DPYD gene from a blood sample
Abacavir (2008): avoid in patients who carry the HLAB*5701 allele to prevent serious or fatal HSR -- test for
HLA allele using PCR-amplified genomic DNA from blood
Warfarin (2008): adjust doses from between 0.5 mg to 7
mg daily in patients with different combinations of 2C9
and VKORC1 gene variants – 5 FDA approved tests
CYP2C9-VKORC1 Genotypes Used to Stratify
CYP2C9Patients into Different Maintenance Dose Ranges
Comparison of Language Related to Tests
and/or Testing: Review of Labels
Available
LOW
MEDIUM
HIGH
Central Theorem of Label Updates: clarity of presenting genetic information in
labels depends on how well genetic associations have been studied; insufficient
clinical evidence about implications of gene sequence variations will lead to
ambiguous or uninformative label updates
http://www.accessdata.fda.gov/drugsatfda_docs/label/2010/009218s108lbl.pdf
Considered
Recommended
Required
Warfarin
N
Y
N
N
Clopidegrel
Y
N
N
N
Panitumumab
N
N
N
N
Cetuximab
N
N
N
N
Carbamazepine
N
N
Y
N
Abacavir
N
N
Y
N
Irinotecan
N
N
N
N
6-MP
Y
N
Y
N
• Prior clearance/approval of a genetic test, although preferred, is not a
requirement for inclusion of test in drug label if risk/benefit is significantly improved
• Issues can arise when the intended use of an in vitro diagnostic is different than
the intended use of the in vitro diagnostic in a drug label
• Where specific language around testing is absent, it is self-evident from other
label information that testing should be done for appropriate use of the drug
3
Personalized Medicine: When We Can Identify
Right Drug for the Right People
Green = Benefit
Red = Harm
Yellow = Mixed Benefit and Risk
Silver = Neither Benefit or Risk
Regulatory Framework for Evidence
Beyond Guidances for Industry
1. Labeling Guidances (Jan 2006 – Mar 2009)
Based on Physicians Labeling Rule
Large Benefit with little harm (10%)
http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulato
ryInformation/Guidances/ucm079645.pdf (page 38)
Benefit
Magnitude
Mixed Benefit and Harm (30%).
(30%) Small
benefit for most.
Harm
2. Code of Federal Regulations
21 C.F.R. Part 201 -- Labeling
Subpart B
Revised April 2009
Neither harm or benefit --Nonresponders(50%)
Harm Without Benefit (10%)
3. Federal Laws
Title IX of FDA Amendments Act (FDAAA)
Post-Marketing Safety Regulation
September 2007
Frequency of various responses in the RCT treated population
Adapted from presentation by Dr. Barbara Evans, ASCPT Annual Meeting (2010)
Quantity: Depends on Circumstances
 Efficacy = One AWC trial with independent substantiation –
to p
protect against
g
unintended bias, chance findings,
g reliance on
single centers – preapproval efficacy with clinical endpoints
 Dosing = Extrapolate efficacy based on PK-PD relationships
when genetics changes drug exposure and specific effect linked
to clinical efficacy – efficacy established, no additional RCT
 Safety = Reliance on one or more small studies, often
retrospective but statistically persuasive and causal– unethical
to conduct new studies – no additional RCT
Quantity of Evidence : General Principles
Supports decision on acceptable benefit/risk
Flexible depending on context of use
Independent replication (endpoint, effect size)
Consistency across multiple studies (diverse covariates)
Level of understanding of causal pathways
Independent supportive evidence (biomarkers, multiple
endpoints)
Scenario I: Preapproval Genetic Information
in Labels (Efficacy PGx)
NE
EW DRUGS
Evidence: Matter of Regulations, Data and
Judgment
Trastuzumab and HER2+ tumors
Imatinib and Kit+ GIST
Dasatinib and PH+ ALL
Maraviroc and CCR5+ HIV-1
Tetrabenazine and 2D6 (Based on DDI)
l
b and
d EGFR+ tumors*
*
Erlotinib
Nilotinib and UGT hyperbilirubinemia
(preapproval,
(preapproval,
(preapproval,
(preapproval,
(preapproval,
(
l
(preapproval,
(preapproval,
prospective)
prospective)
prospective)
prospective)
retrospective)
retrospective))
retrospective)
* Tumor EGFR protein expression status removed from label in April 2009
Linked co-development provides best opportunity to obtain appropriate
evidence of clinical utility (efficacy) on both test and drug
Strength of evidence comes from prospective hypothesis, RCT and
replication; differentiate prognostic from predictive biomarkers
Some examples of early phase studies providing evidence of clinical
utility (safety) of test to predict side effects and optimize dosing
Sponsor or company assumes primary responsibility for generating the
evidence of clinical utility
Quality of Evidence : General Principles
Studies are adequately designed and conducted
regardless of the purpose and type of study
Completeness
p
of documentation and access to primary
p
y
study data to verify results – consider all data
Flexibility when postapproval studies are not conducted
by original NDA holder – less extensive data gathering
Published studies when persuasive – multiple and various
investigators, detailed methods, appropriate endpoints,
robust results, PIs with history of quality research
4
What Does Clinical Utility Mean?
No consensus definition – it depends on the value
proposition for stakeholders in personalized medicine.
With no agreed
agreed-upon
upon definition it becomes virtually
impossible for stakeholders to agree on evidence
Important to recognize the continuum of clinical utility
with a tipping point of acceptance for each stakeholder
A measure of the impact that genetic information can
have on the assessment of those who will benefit or
those who are at risk
Translational Challenges for Labels and
Label Updates




Labels and label updates are complex and will only
become more so because of the novelty of information
Labels are static; science behind labels is dynamic so
label updates (preventing harm) will be common
Clinical utility of genetic information is central to label
decisions
How evidence is generated and the quantity/quality of
evidence drive labeling decisions and language
Clinical Utility Is Of Most Importance to
Personalized Medicine
Central issue is the quality/quantity of evidence and
how to generate the evidence – no single answer
but depends on context of use for diagnostic test
I continue to be surprised that so many believe that
only RCTs can generate acceptable evidence of
clinical utility – versus observational studies
Clinical utility of a targeted therapy depends on the
clarity with which information is presented – this
depends on how well the test has been studied
Summary: What Needs to Be Done to
Advance Personalized Medicine
1. Technology: problem of too much data all of which is not
clinically relevant, poor understanding of disease biology
and drug pharmacology, moving research technology to
clinic , and making tests clinically useful to physicians
2. Regulatory: beyond discovery, influences what research
needs to be done; has not been forthright and clear
3. Strategic: asking the right questions, designing the right
trials and generating appropriate quantity/quality of
evidence for stakeholders such as FDA
4. Implementation: having a deliberate process between test
developers and end users , and synthesis of knowledge
Thank you for your time
and attention.
attention
[email protected]
5