Which endpoint to choose ? (in phase II sarcoma clinical trials)

download report

Transcript Which endpoint to choose ? (in phase II sarcoma clinical trials)

Which endpoint to choose ?
(in phase II sarcoma clinical trials)
(and also in other sarcoma trials!)
George D. Demetri, MD
Dana-Farber / Harvard Cancer Center, Boston Mass. USA
for
Robert Maki, MD PhD in absentia
Mount Sinai School of Medicine
New York, NY, USA
Tell me! What is the answer?
Well, what is the question?
Example: Is this randomized study design
definitive proof of activity?
Ovarian cancer xenografts in mice treated with
cyclopamine or saline
n=4
n=4
{probably not good enough for FDA approval}
McCann CK et al. PLoS One 2011; 6(11): e28077
4
Example: Is this randomized study design
definitive proof of activity?
Citrus fruit dietary supplementation in Sailors with Scurvy!
Intervention
100%
n=8
Active
Disease
n=4
{probably not good enough for FDA approval}
Lind 1747 (published in summary, in posteriori)
5
And so it is with phase II designs.
• It depends on the question…
• Your choice of endpoint is the most
important decision you make in the design
of a (phase II) clinical trial
Phase I / II study general principles
• Treat a group of patients, typically 20-100, to obtain safety data
• Identify activity (or not)
–
–
–
–
Got a great biomarker? YIPPEE!!!! (e.g. viral load in HIV)
Radiological change typically used in solid tumors
Radiological, hematological, molecular parameters for heme malignancies
Perhaps the most common stage of drug development abandonment
• Proceed to phase III if sufficient activity
– Response rate often highest in phase II, lower in single center phase III,
lowest in cooperative group phase III
• Bias is inherent in a group of highly-selected patients
– Often no comparison group
– Short term therapy: for metastatic disease, most patients often off treatment
within 6-8 weeks
– ( How does one identify toxicity in patients treated longer? )
7
What is your goal?
• Is this a proof-of-concept study?
– First phase I or phase II in humans
– Often single agent, single arm study
– Phase IB or Phase IIA
8
Phase II Trial in “Cancer”
1972
Doxorubicin
Activity
Noted in
Sarcomas
Phase II Trial in “Cancer”
1972
Doxorubicin
Activity
Noted in
Sarcomas
Phase II in “Cancer”
1972
Doxorubicin
and DTIC
Activity
Noted in
Sarcomas
What is your goal?
• Is this a proof-of-concept study?
– First phase I or phase II in humans
– Often single agent, single arm study
– Phase IB or Phase IIA
• Are you trying to rule in or rule out activity
for further development?
–
–
–
–
Probably most reliable with a randomized study
Can compare against other therapy or placebo
Relaxed type I and type II errors (vs. phase III)
Phase IIB
12
Endpoint options
• “Progression-free survival” (PFS): commonly used
– PFS: time from treatment initiation to tumor progression or death from any
cause, with censoring of patients who are lost to follow-up
• “Time to tumor progression” (TTP): used much less often
– TTP: the only event of interest is disease progression
• Response rate (WHO, RECIST, modified RECIST, Choi…)
• Biomarker
– Disease marker definitely tied to outcomes [e.g. viral load in HIV]
– Tumor marker [e.g. PSA]
– Imaging [e.g. PET SUVmax]
• Patient-reported outcomes
• Aim: To test impact of study intervention on
“how a patient feels, functions or survives”
13
How do I assess thee?
Let me count the ways (…well, two)
Landmark analysis
Time-to-event analysis
Time to event data vs. landmark analysis
More data
p>0.05
Data are lost
p<0.05
p>0.05
15
Which type of “phase II” study?
• One stage
• 2, 3, … n stage
• Continuous monitoring
__________________________________
•
•
•
•
Response adaptive randomization
Phase I-II
Phase II-III “seamless”
Randomized discontinuation
16
A brief history of clinical trials
Early Clinical Trial
Phase II Investigational device study
Shut down by IRB for:
• Consent from spouse not sufficient
• Failure to file annual report
• History of basal cell cancer 3 yrs ago
made patient ineligible per entry
criteria.
Inscription:
Meester snyt die keye ras
Myne name Is Lubbert Das
Master, cut away the stone;
My name is Lubbert Das
H Bosch. The Cure of Folly. ~ 1494
18
One-stage design
n patients
Does this drug work?
(Does your dog hunt? Art Skarin)
19
2-stage design
n patients
m patients
Preliminary stage
Follow-up stage
Rejects completely inactive treatments more rapidly
Gehan EA et al. J Chronic Dis 1961; 13: 346
20
“Optimized” 2-stage design: Simon
n patients
m patients
• Binary outcome (Response or not)
• Sample size minimized for given type I, type II errors
• Study terminated only for early lack of activity
• Very commonly used in oncology
Simon R. 1989; Control Clin Trials 10: 1
21
Off to the races!
n patients
m patients
p patients
3 stages
n patients
Analyze after every patient:
continuous monitoring
Enign LG et al. Stats Med 1994;13: 1727
Thall PF and Sung HG 1998; 17: 1563
22
To randomize & randomize not
• Using RECIST tumor response
– Monotherapy?
– Combination therapy?
“Mít a nemít”, 1944
– In each case randomization can help, but with IIA designs single
arm studies are often used
• PFS
– Not universally validated as a validated (earlier, more direct)
endpoint of clinical benefit compared to overall survival
– Usually requires randomization vs another therapy
– Key issues: time-based restaging (lead time bias) and variations in
underlying disease biology (well diff liposarc…..wild type GIST…)
23
Randomizations: graphically
A
R
B
Other Rx
Other
Rx
R
B
A
Simple
B
A
Other Rx
A
1-way crossover
B
2-way crossover
R
A
A
B
Other
Rx
R
nil
Randomized discontinuation
24
Phase II: What to do?
• Try to randomize
• Sometimes not practical to randomize
– Very rare sarcoma subtype
– Can’t accrue sufficient number to show a small difference
– Hey, it’s O.K….
Prospective data are better than retrospective data
• Bayesian designs: useful to choose therapies for further
study (I-SPY 2, BATTLE, etc.)
– May be difficult to use for regulatory approval based on very
real concerns regarding patient referral bias over time
– Randomization as a function of drug
– Randomization as a function of biomarker
• Weigh your options, call a friend or twelve…
25
Phase III: Be Happy?
• Try to avoid expensive phase III failures!
• Set a high bar in Phase II before going to Phase III
in order to minimize risk of phase III failure
– Most signals of activity are less obvious in phase III trials
than they were in phase II.
• Decide what is truly clinically important for a patient
– And then be honest about whether you think you can
actually achieve it based on available data…
• Do you really need a phase III trial to prove benefit to
patients?
– If you think “no”, you had better have darned great data!
26
Conclusion :
Choose your trial design with care
27