Transcript Rosario García Campelo, MD

Clinical Trial Endpoint Selection in
Oncology: What Can Make a Difference?
Rosario García Campelo, MD
1
Clarify expectations…
What do patients - and the world - consider
MEANINGFUL CLINICAL BENEFIT?
What do patients and physicians want?
To be cured…
To relieve a person of
the symptoms of a
disease or condition
To eliminate a disease
or condition with
medical treatment
Benefit in Oncology
Prolong survival
Improve QoL
The magnitude of survival benefit depends on many factors…
Estimate of 75% survival
rate at 10 y
STAGE IV GERM-CELL
TUMORS
Verheven et al. EJC 2007
Less than 10% survival rate
at 30 months
Schiller et al. NEJM 2002
STAGE IV NSCLC
A drug that cures…
• Effective
• Non-toxic
• At a manageable financial cost
ANYTHING LESS IS A COMPROMISE.
HOW MUCH ARE WE WILLING TO COMPROMISE?
The appearance of progress
Defining the value of a new drug
VALUE
COST
How to define clinical benefit?
STAGE IV PANCREATIC CANCER
STAGE IV NSCLC
Median survival
11.3 vs 10.1 M
P=0.044
HR 0.87
Median survival
6.2 vs 5.9 M
P=0.038
HR 0.82
Pirker et al. Lancet 2009
Moore M J et al. JCO 2007;25:1960-1966
STAGE IV NSCLC
STAGE IV NSCLC
Median PFS
6.7 vs 6.1
P=0.003
HR 0.75
Reck et al. J Clin Oncol 2009
Median PFS
10.2 vs 6.6
P<0.001
HR 0.58
First task…
• Define patient population to treat
–
–
–
–
–
Type of cancer
Subtype of cancer
Adjuvant setting
Metastatic disease
Line of therapy
• Determine current expectations of both, clinicians and patients
– What is current OS for each cancer/setting
– Determine what would be a clinically meaningful increase in the primary endpoint chosen: OS,
PFS, RR????
Who decides what is important?
• Patients
• Clinicians
• Regulatory agencies
• Industry
What is a clinical trial?
• A clinical trial is a research study conducted in patients, with
patients and for patients, to answer specific questions about
their treatment, diagnostic or follow-up
• Clinical trials are used to determine whether new biomedical or
behavioral interventions are both safe for patients and effective
at treating their disease
(National Institutes of Health’s (NIH) definition http://grants.nih.gov/grants/glossary.htm#C)
• Carefully conducted clinical trials are the fastest and safest way
to find treatments that work in people
Historical clinical trials design
PHASE I: maximum-tolerated dose assessment
Find a dose that kills the tumor rather than the patient…
PHASE II: response signal
Treat patients but expect no response…
PHASE III: comparison to the standard or added to the standard
Randomize and hold your breath
From G. Blackledge
Just to remember…
• Pvalue:
– Is a function of the observed sample results (a statistic) that is used
for testing a statistical hypothesis
– It gives no indication about the clinical importance of the observed
association
• CLINICAL BENEFIT
– Clinically meaningful outcome to be offered to patients in OS and/or
quality of life
– Activity is different to benefit…
• ENDPOINT
– A measure to clarify potential efficacy or safety
– OS, PFS, TTF, TTP, ORR, DCR
The true clinical efficacy measures:
Robust, clear…
What makes a good endpoint?
CHARACTERISTIC
MEANING
Relevant
Clinically important/ useful
Quantifiable
Measured on a appropiate scale
Valid
Measures the intended effect
Objective
Interpreted effect yields consistent measurements
Reliable
Same effect yields consistent measuremnts
Sensitive
Respond to small changes in effect
Specific
Unaffected by extraneous influences
Precise
Small variability
Other
Tradition, cost, time…
Endpoints – FDA view
Regulatory
Evidence
Surrogate
Endpoint
Advantages
Disadvantages
RR/CRR
1-arm possible, smaller, shorter,
attributable to drug
No direct measure of benefit/no comprehensive measure of
drug activity/only subset of benefiting patients
DFS
Smaller, shorter
Not stat. validated as surrogate for OS/not precise, open to
bias/many definitions
PFS
Smaller, shorter, SD included,
crossover/other Tx not affecting,
objective & quantitative
Not stat. validated as surrogate for OS/not precise, open to
bias/many definitions/frequent assessment /need to balance
timing x arms
Symptoms
Patient perspective of direct
clinical benefit
Blinding hard, missing data, clinically relevant effect, validated
tools lacking
OS
Direct measure of benefit, easy,
precise
Large studies, crossover/follow-up Tx affects, noncancer deaths
Clinical benefit
Guidelines for endpoint selection
The decision should always be related to:
• The patient subpopulation of interest
• The stage of disease (depends on the type of cancer)
• The characteristics of the treatment (toxicity, efficacy)
• The aims of trial (superiority/noninferiority/safety)
• The other treatments already available to that group of patients
• Ethics
• Practical feasibility (costs, logistics…)
OS as the gold standard for demonstrating clinical benefit
ACCURATE
NOT PRONE TO
INVESTIGATOR OR
ASSESSMENT BIAS
OBJECTIVE
READILY
COMPARED
ACROSS DISEASES
UNAMBIGUOUS
CLINICALLY
SIGNIFICANT
EASILY MEASURED
Cheema PK, Curr Oncol 2013
Advanced NSCLC:
Why overall survival?
• Most relevant to patient
• Most objective endpoint
– Clear and accurate endpoint
– Usually easy to get this information
• Relevant for other reasons
– It is easier to decide upon a treatment when its impact on survival
is known
– Reimbursement issues
– Assessment in earlier disease stages
– Impact on drug development in general
A trial that looks for OS…
• Takes too long
• Requires large sample sizes
• Is complex
• Influenced by post-trial therapy
• Is more expensive
• But… the cost of a trial is a small fraction of the costs to society for not
having a reliable answer to a question, since today’s treatments can be
extremely expensive
• Just to remember:
– RCT only require large sample sizes when one is looking for small treatment effects
– Larger trials can stop early with formal internal monitoring if the treatment effect is
surprisingly large
Korn E. J Clin Oncol 2012
Main consensus ASCO: “OS endpoint is important”
• OS is a feasible endpoint, though the OS benefit may be clouded by subsequent
therapies
• Clinical trials should aim to improve OS by >25%
ASCO stresses the use of the OS endpoint.
Agents not expected to provide such OS gain would no
longer be needed in clinical practice.
http://www.asco.org/practice-research/clinically-meaningful-outcomes
Ellis et al. J Clin Oncol 2014
Let´s talk about surrogate markers…
…a measurement used in trials as an early and
adequate substitute for OS, like ORR or PFS
Introduction to surrogate endpoint
• Why do we use a surrogate endpoint?
– Can be measured earlier
– Convenient or less invasive
– Can be measured more frequently
– Can accelerate the approval process
• Advantages:
– May reduce the size of clinical trials
– May shorten the duration of clinical trials
– May reduce the cost of clinical trials
Let´s talk about surrogate markers…
…a measurement used in chemo trials as an adequate substitute
for OS, like ORR or PFS
…An endpoint that is a surrogate for OS would be helpful in
addressing the limitations of OS but must first be validated by
satisfying statistical criteria
…A growing belief in the oncology community that delaying
progression in metastatic disease is a worthy goal,
even if OS is not improved
Use of PFS in the contemporary era
200
11
150
100
50
35
2
0
47
107
167
137
0
1975-1984
1985-1994
No. PUBLISHED RCT
1995-2004
2005-2009
PFS AS PRIMARY ENDPOINT
Booth C. J Clin Oncol 2012
Can PFS be considered a strong and robust surrogate marker?
• Is a treatment that improves PFS really an advance for patients even if OS is not
improved?
• How does PFS relate to patient benefit?
• Or is it only lowering the bar to declare efficacy and accelerate drug approval?
Buyse M. Stat Methods Med Res 2008
Booth C. J Clin Oncol 2012
PFS as an endpoint has potential advantages
• Progression events occur early and more frequently than death
events
• Less influenced by competing causes of death
• PFS is not influenced by post-progression therapy
• Faster time to drug approval
• Faster receipt by patients of novel therapies
• Less cost to manufacturer or sponsor
Saad, Katz. Ann Oncol 2009; Burzykowski et al. JCO 2008
Shi et al. Int J Clin Oncol 2009
But also many disadvantages…
ASSESSMENT BIAS
The investigator may declare a control-arm case a progression to get a patient on an active therapy
ASAP
EVOLUTION-TIME BIAS
A suspected progression may be formally evaluated at a later time when patients are in one arm
rather than the other arm
THE EXACT TIME OF PROGRESSION IS UNKNOWN
It requires frequent radiologic assessment, which introduces imprecision
ATTRITION BIAS
More patients may withdraw from one arm than from the other arm
Phase III trials comparing TKI vs conventional CT in
EGFR mutation–positive NSCLC
Study
Control arm
Ethnicity
Rx evaluation
Independent review
PFS
LUX-lung 3
(n=345)
Afatinib vs
Cis/pem
Asian (71%)
Caucasian
6w
Yes
13.6 vs 6.9
11.1 vs 6.9
LUX-LUNG 6
(n=364)
Afatinib vs
Cis/Gem
Asian
6w
Yes
11 vs 5.6
EURTAC
(n=174)
Erlotinib vs
CT
Caucasian
6w
Yes
10.4 vs 5.4
OPTIMAL
(N=165)
Erlotinib vs
Carb/Gem
Asian
6w
No
13.1 vs 4.6
WJTOG
(n=172)
Gefitinib vs
Cis/Doc
Asian
8w
No
9.6 vs 6.3
NEJ002
(n=230)
Gefitinib vs
Carb/pacl
Asian
8w
Yes
10.8 vs 5.4
IPASS
(n=261)
Gefitinib vs
Carb/pacl
Asian
6w
No
8.4 vs 6.7
SIGNAL
(n=42)
Gefitinib vs
Cis/Gem
Asian
9w
Yes
9.5 vs 6.3
Afatinib
Erlotinib
Gefitinib
What is the true defitinion of meaningful clinical benefit?
What is the true defitinion of PFS?
EGFR MUT PATIENT: Slow progression to erlotinib, excellent symptoms control
What we know of PFS as a tool..
And this is evidence…
• Published literature on advanced breast, prostate, and NSCLC have not supported the
surrogacy of PFS for OS
• Colorectal cancer and ovarian cancers seem to be the only tumor types for which data
supporting the surrogacy of PFS for OS have been demonstrated…and this is not
completely true
Booth C. J Clin Oncol 2012
Surrogacy of PFS in advanced NSCLC?
• IPD-based analysis
• 5 RCT DOC vs VNB
• PFS - OS: some correlation
(τ=0.59; 95% CI 0.58 to 0.61)
• Great magnitude of PFS benefit
(at least HR <0.5) needed to predict
any significant OS effect (HR <1.0)
• Most trials have usually failed to
show such PFS
• PFS is unlikely to be an acceptable
endpoint
Laporte S et al. BMJ Open 2013
However…
• Although I do believe it is time to significantly raise the bar…superstars in oncology are
the exception, but progress
in oncology is incremental
• I do believe we know much more about cancer than some
years ago:
– Tumors can now be defined by molecular drivers
• New challenges posed by the new classes of agents
• New trial designs and validated endpoints are needed to
test them
What is the true definition of clinical benefit?
Shaw A, et al. N Engl J Med 2013;368:2385–2389
Times have changed...
EMPIRICAL
ONCOLOGY
1980
1990
MOLECULAR
ONCOLOGY
2010
New challenges for new therapies
• INMUNOTHERAPY
– Pseudo progression effect: Go vs no-go decision
– Survival benefit in the absence of significant responses, or very significant PFS
• PERSONALIZED THERAPIES
– Intra-patient heterogeneity
– How to deal with
• Brain mets
• Single site progression
• Benefit beyond progression
• WILL WE BE ABLE TO DEFINE THE BEST SEQUENCING?
If the design of trials is evolving, should trial
endpoints evolve in parallel?
New trends of drug development
Phase I
Phase II
Phase I -II
Phase III
Phase III
Novel clinical trial designs
• PHASE I:
– Limited number of doses, drug combinations, and selected
populations of patients
• PHASE II:
– Randomized designs
• PHASE III:
– Smaller and smarter
• Adaptative designs (adaptative enrichment or adaptative
estratification design)
• Basket designs: multiple diseases for a given marker
• Umbrella designs: multiple markers for a given disease
KEYNOTE-001 phase Ib trial(NSCLC cohorts):
Study design
2mg/kg q3w
• Measurable disease
• ECOG PS 0–1
• Known PD-L1 status (PD-L1+ defined as ≥1%
expression)
• EGFR mutations or ALK gene rearrangements
permitted in previously treated patients (with PD
on relevant EGFR TKI) but not for treatment-naïve
patients
Pembrolizumab in treatment-naïve
patients with PD-L1+ NSCLC
(n=45)
R
10mg/kg q3w
1:1‡
10mg/kg q2w
Currently only investigator-assessed ORR
by irRC is
available for this cohort
2mg/kg q3w§
(n=45)
• Disease progression on most recent prior
systemic therapy
• No systemic steroid therapy, active autoimmune
disease or active brain metastases
(n=345*)
(n=307 for expansion cohorts)
10mg/kg q2w; PD-L1+
(n=58)
≥2L
Pembrolizumab in previously treated
patients with PD-L1+ or PD-L1NSCLC
≥3L
R
3:2
10mg/kg q3w; PD-L1+
(n=86)
10mg/kg q2w; PD-L1(n=40)
10mg/kg q3w; PD-L1+
(n=33)
Primary endpoints
• DLTs
• Biomarker expression
• AEs
• ORR
Secondary endpoints
• Pharmacokinetics
• PFS
• OS
• DoR
irRC: immune-related response criteria.
*An initial cohort of 38 patients was also enrolled; ‡First 11 patients were randomized to
2mg/kg q3w and 10mg/kg q3w, and the remaining 34 patients were randomized to 10mg/kg
q2w and 10mg/kg q3w; §Additional non-randomized cohort not included in the previously
treated population of 217.
Garon et al. ESMO 2014 (abstract LBA-43)
PCD4989g: Phase I study design

Incurable/metastatic solid
tumours

Tumour specimen
available

ECOG PS 0–1
Dose escalation
Expansion
(safety evaluation)
MPDL3280A
0.01–20mg/kg iv q3w
(safety evaluation,
response assessment)
MPDL3280A
10, 15 and 20mg/kg iv q3w
PD
(n=1,242)
NSCLC
Melanoma
Primary endpoints
• Safety
• Tolerability
• MTD and DLTs
• RP2D
RP2D = recommended phase II dose
RCC
Mandatory
serial tumour
biopsies
UBC
Secondary endpoints
• Pharmacokinetics
• ATAs
• Best ORR
• Objective response
• DoR
• PFS
• Potential biomarkers
TNBC
Other tumour
types
Exploratory objectives
• Assess tumour cell
PD-L1 expression (IHC),
evaluate baseline
tumour cell transcript
expression of immune
and checkpoint markers
(PCR)
NCT01375842
Soria, et al. ESMO 2014 (abstract 1322P); Hodi, et al. ECC 2013
Basket trials
Ariel Lopez-Chavez et al. JCO 2015;33:1000-1007
KEY STRENGTHS OF THE BASKET TRIAL DESIGN
 The ability to identify a favorable response to
targeted therapy with a small number of patients
 The ability to validate a clinical target
KEY CONDITIONS FOR BASKET TRIALS SUCCESS
 The tumor must depend on the target pathway
 The targeted therapy must reliably inhibit the
target
Development of reliable endpoints to allow an early recognition of
clinical benefit
Wilson et al. Lancet Oncol 2015
MOLECULAR ONCOLOGY…NEW TRIALS ENDPOINTS
TREATMENTS
ENDPOINTS
Signal transduction
inhibitors
Effect on molecular
target
Gene expression
modulators
Circulating tumor cells
Angiogenesis inhibitor
Circulating tumor DNA
Immunotherapies
Functional imaging
Buyse, M. et al. Nat. Rev. Clin. Oncol 2010
Necessary changes to drug development
• Development and agreement on quality standards for biomarkers
and routines for quality assessment
• Harmonisation of biomarker validation
• Collaboration between drug developers and regulatory agencies
• Agreement on best practices for co-development of diagnostics and
drugs
• Educating citizens and health professionals on the need to collect
biomarkers
Kheif et al. Clin Cancer Res 2013
How PD-L1 is being assessed?
“ME TOO DRUGS…WITH MY OWN BIOMARKER TEST”
HARMONISATION OF BIOMARKER VALIDATION!!!
PATIENT REPORTED OUTCOMES (PRO)
Are PRO measures underappreciated and underused in oncology?
• PRO address more than just symptoms
• Direct measures of patient benefit, they can be considered independent endpoints
• This approach is especially useful in trials of drugs for patients with incurable cancers,
in which one of the main goals is to improve palliation of symptoms
• Only 1/3 of phase 3 breast cancer trials registeries with US NIH have collected or are
presently collecting PRO
Wilson et al. Lancet Oncol 2015
In summary
• A difference is only a difference if it makes a difference…
• Overall survival remains the most relevant endpoint for phase lll trials in patients
with advanced NSCLC
– Easy to assess, accurate, and reliable
• It is time to change the traditional design of clinical trials
• It is time to advance endpoints to parallel changes in trial design and therapeutic
intervention
• Consistency and validation in outcome of the various endpoints is also important
–
–
–
–
–
RR
PFS
QoL
PRO
Biomarkers
McNamara Fallacy, 1960
1. Measure whatever can be easily measured…This is correct
2. Disregard that which cannot be measured easily…This is
artificial and leads to errors
3. Presume that which cannot be measured easily is not
important…This is blindness
4. Presume that which cannot be measured easily does not
exist…This is suicide
Basier MH. BMJ 2009