Transcript Gobburu

Impact of Exploratory Analysis
on Drug Approval
Joga Gobburu
Pharmacometrics
Office Clinical Pharmacology, CDER, FDA
[email protected]
Take Home Message
• Exploratory (e.g., pharmacometric) analyses
are often used to make regulatory decisions
– Decisions are not entirely driven by the pre-specified
statistical analysis
• Need for change
– Integrate strengths of both approaches
• Think “How exploratory analyses can help drug
development?”
– Opportunities for collaboration between
pharmacometricians and statisticians are
abundant
• Think about “How can I facilitate this collaboration?”
2
Pharmacometrics
(or Quantitative Experimental Medicine?)
• Science that deals with quantifying disease and
pharmacology
• Applications
– Benefit/Risk, dose individualization, trial design
• Diverse expertise
– Clinical pharmacologists, Pharmacometricians,
Clinicians, Statisticians, Bioengineers
• Tools
– Linear/Nonlinear Mixed effects models, Longitudinal
data analysis, Biological models, Stochastic simulations
3
Impact of Exploratory Analyses
2000-2004
Pivotal: Regulatory decision will not be the same without PM review
Supportive: Regulatory decision is supported by PM review
Impact
Approval
Labeling
Pivotal
54%
57%
Supportive
46%
30%
0
14%
No Contribution
Bhattaram et al. AAPS Journal. 2005; 7(3): Article 51. DOI: 10.1208/aapsj070351
4
Impact of Exploratory Analyses
2005-2006
Pivotal: Regulatory decision will not be the same without PM review
Supportive: Regulatory decision is supported by PM review
Impact →
Discipline
PM Reviewer
Approval
Labeling
95%
100%
DCP Reviewer
95%
100%
DCP TL
90%
94%
Medical Reviewer
90%@
90%@
DCP=Division of Clinical Pharmacology
@=survey pending in 1 case
5
NDA Case Study
•
•
Drug is proposed for a ‘rare’ debilitating,
fatal disease with no approved treatment.
One trial successful and other failed
– Failure likely due to trial execution errors
•
Potential miscommunication about dose timing
– Primary variable: Change in symptom score
•
Key question
– Is there adequate evidence for the
effectiveness?
6
Equivocal Evidence of Effectiveness
Pivotal Studies
DB#1
Dbl-blind (DB)
Randomized
PBO Controlled
Dose Titration
N=75
P<0.051
(withdrawal)
DB#2
Dbl-blind (DB)
Randomized
PBO Controlled
Dose Withdrawal
N=30
P>0.051
1change
Agency at this point can ask for more
evidence (one or more studies)
OR
Investigate further across the clinical
trial database whether there is a
consistent signal of effectiveness or not
in score at the end of study
7
Equivocal Evidence of Effectiveness
Pivotal + Other Studies
DB#1
Dbl-blind (DB)
Randomized
PBO Controlled
Dose Titration
N=75
P<0.05
(withdrawal)
OL-1
Open label (OL)
Withdrawal
Dose Titration
N=75
DB#2
Dbl-blind (DB)
Randomized
PBO Controlled
Dose Withdrawal
N=30
P>0.05
OL-2
Open label (OL)
Continue ‘old’ dose
N=30
8
Significant Dose-Response
Relationship – DB1, OL1
Parameter
Mean
(Confidence Interval)
Slope of doseresponse,
% per mg
4.3*
(3.7, 4.6)
Within-Patient
Variability
26%
(23%, 29%)
Between-Patient
Variability (CI)
56%
(46%, 66%)
* p<0.001
Linear mixed effects model employed
Estimate of dose-response slope is similar for individual and combined
9
analyses. Results from combined shown here.
Significant and Consistent Drug
Effects Across Studies
6
Mean Symptom Score
5
4
3
2
Placebo (DB1)
Drug (DB1)
Drug (OL1)
Drug (OL2)
1
0
0
2
4
6
10
Months
Drug in OL1 beat Placebo in DB1
Cross-over comparison
6
Mean Symptom Score
5
4
3
2
1
Placebo (DB1)
Drug (OL1)
0
0
2
4
6
11
Months
Value of Exploratory Analysis
• To Patients/FDA
– Availability of drug sooner, especially given no
approved treatments (debilitating disease)
– Efficient solution to challenging patient enrollment
– Fewer review cycles (because of this issue alone)
– Ultimately might lead to lower drug costs
• To Sponsor
– Alleviated the need for additional trial(s) to
demonstrate effectiveness
– Save $$ and time
• Pharmacometrics analyses can and do influence
approval decisions!
12
Why did the sponsor not consider
making a similar case?
Unlikel
y
• Unanticipated concern
• Lack of expertise (both technical, strategic)
• Prescriptive behavior on analysis
• Unclear expectations from FDA
Likely
13
Parkinson’s Disease
Collaboration between Statistics
and Pharmacometrics
Dr. Bhattaram and Dr. Siddiqui are the project
leads with the following team members:
FDA
Statistics, Clinical, Policy Makers
External
Statistician, Disease experts
14
Symptomatic or Protective?
Total UPDRS
20
Placebo
18
Drug A
16
Drug B
14
12
10
0
6
Tim e, m onths
12
15
Symptomatic or Protective?
Total UPDRS
20
Placebo
18
Drug A
16
Drug B
14
12
10
0
6
Tim e, m onths
12
16
Discern Symptomatic vs. Protective
Effects: Delayed Start Design
30
25
UPDRS
20
Protective
15
Key Questions:
-Endpoint ?
-Analysis ?
-Handling missing data?
10
5
Placebo Phase
0
0
Active Phase
20
40
60
Weeks
If drug is protective then patients who received drug longer will have
lower scores compared those who receive drug late.
17
Parkinson’s Disease Database
Data
Source
#Patients Trial Duration
Trial#1
NDA
400
1yr + 3yr follow-up
Trial#2
NIH
400
1yr + follow-up
Trial#3
NDA
900
9mo + follow-up
Trial#4
NDA
200
9mo + follow-up
Trial#5
IND
300
1.5yr
18
Selegiline ( 5 years)
Published Data
Mean (SD) of Total UPDRS scores for patients with Parkinson’s disease
treated with levodopa alone or in combination with selegiline for 5 years
and during the one-month washout period
The vertical line represents 2 months
Eur.J.Neurology, 1999, 6: 539-547
19
Fraction Remaining
Patients with slower progression remain
longer in clinical trials (TEMPO)
20
Value of Collaboration between
Pharmacometrician, Statistician
• Statistician’s Contribution
– Primary statistical analysis
• Drop-outs
– Trial design
– Power calculations
• Pharmacometrician’s/Disease Expert’s
Contribution
– Biological/Mechanistic Interpretation
• Disease Progression
• Drug Effects
• Drop-outs
– Trial design, alternative analysis
21
Value of Exploratory Analyses
• Collected a large database of clinical trials
• Extracted patient population,
placebo/disease progression, drug effect
(not shown) and drop-out information.
• Simulations to answer the key questions
mentioned earlier are in progress
– Directly useful to advice sponsors
• Conference planning is underway
Disease Models Background:
http://www.fda.gov/ohrms/dockets/ac/06/briefing/2006-4248B1-04-FDA-topic%203%20replacement.pdf
22
Take Home Message
• Exploratory (e.g., pharmacometric) analyses
are often used to make regulatory decisions
– Decisions are not entirely driven by the pre-specified
statistical analysis
• Need for change
– Integrate strengths of both approaches
• Think “How exploratory analyses can help drug
development?”
– Opportunities for collaboration between
pharmacometricians and statisticians are
abundant
• Think about “How can I facilitate this collaboration?”
23
24