Transcript Norbert Benda

Federal Institute for Drugs
and Medical Devices
The use of modelling and simulation in
drug approval: A regulatory view
Norbert Benda
Federal Institute for Drugs and Medical Devices
Bonn
Disclaimer:
Views expressed in this presentation are the author's personal views and not necessarily the views of BfArM
The BfArM is a Federal Institute within the portfolio of the Federal Ministry of Health (BMG)
Overview
 Principles in drug approval
 Challenges
 Modelling ?
 Simulation ?
 Problems
 Longitudinal analysis
 Small population dilemma
 Conclusions
2/20 N Benda: M&S in Drug Approval
Federal Institute for Drugs
and Medical Devices
General principles in drug approval
Federal Institute for Drugs
and Medical Devices
1. Demonstrate efficacy
2. Show favourable benefit risk
3. Additional requirements
 Additional claims to be demonstrated after general
efficacy (1) has been shown
 Homogeneity
 Subgroups to be excluded / justified
 Relevant dose / regimen
3/20 N Benda: M&S in Drug Approval
Statistical principles in drug approval
Federal Institute for Drugs
and Medical Devices
 Independent confirmatory conclusion
 no use of other information
 type-1 error control limiting false positive approvals
 Internal validity
 blinded randomized comparison
 assumption based
 External validity
 relevant population to study
 random sampling, etc
4/20 N Benda: M&S in Drug Approval
Federal Institute for Drugs
and Medical Devices
Areas that may challenge approval principles
 Paediatrics
 Orphan drugs
 Integrated benefit risk assessments
 Dose adjustments (body weight, renal impairement,
etc.)
 Individualized dosages / therapies
5/20 N Benda: M&S in Drug Approval
Federal Institute for Drugs
and Medical Devices
Example: Limitations in paediatric drug approvals
 Sample size
 small
 Treatment control
 placebo unethical / impossible
 Endpoints
 different from adults / between age groups
 Dosages
 age / weight dependent
6/20 N Benda: M&S in Drug Approval
General use of M&S
 Prediction
 dose response
 dose adjustment
 impact of important covariates
 identification of subgroups of concern
 Optimization of development program
 identification of optimal / valid methods
 informed decision making
 accelerating drug development
7/20 N Benda: M&S in Drug Approval
Federal Institute for Drugs
and Medical Devices
Impact of M&S on the regulatory review
Federal Institute for Drugs
and Medical Devices
 Low impact
 internal decision making (hypothesis generation, learning)
 more efficient determination of dose regimen for phase III
 optimise clinical trial design
 Medium impact
 identify safe and efficacious exposure range
 dose levels not tested in Phase II to be included in Phase III
 inferences to inform SPC (e.g. posology with altered exposure)
 High impact
 extrapolation of efficacy / safety from limited data (e.g. paediatrics)
 Model-based inference as evidence in lieu of pivotal clinical data
8/20 N Benda: M&S in Drug Approval
Model based inference
Models = assumptions
 Models with increasing complexity
 random sampling from relevant population
 variance homogeneity
 proportional hazard
 generalisability of treatment differences (scale)
 longitudinal model for the treatment effect
 PK models / population PK models
 PK / PD models
 models on PK – PD – clinical endpoints
9/20 N Benda: M&S in Drug Approval
Federal Institute for Drugs
and Medical Devices
Modelling
Federal Institute for Drugs
and Medical Devices
Modelling = Model building + model based inference
 Model building aspects
 biological plausibility
 extrapolation from
• animal models
• healthy volunteers
• adults
 interpretational ease
 robustness
 evidence based
• derived from / supported by data
10/20 N Benda: M&S in Drug Approval
Problems with modelling
Federal Institute for Drugs
and Medical Devices
 Model selection bias
 if model selection and inference based on same data
 Ignored pathway
 Dose  PK  PD  clinical endpoint ?
 Ignored between-study variability
 validation usually within similar settings
 no “long-term validation”
11/20 N Benda: M&S in Drug Approval
Simulations
Federal Institute for Drugs
and Medical Devices
 Simulation = numerical tool
 Complex models / methods require unfeasible high
dimensional numerical integration
• e.g. type-1 error / power calculation under complex assumptions (dropouts, adaptive designs, etc) or model deviations
 Simulation = visualization
 Focus on statistical distributions
• between subjects / within subjects
• considering complex variance structures / non-linear mixed models
 Visualize resulting distribution for specific settings
(treatments, fixed covariates)
12/20 N Benda: M&S in Drug Approval
Simulations
Federal Institute for Drugs
and Medical Devices
 Advantages:
 visualization on distributions / populations
 allow for an population based assessment
 Disadvantages
 often (unconsciously ?) misinterpreted as “new” data
• inference from simulation impossible
 depend on (unverifiable) model assumptions
 incorrect variance modelling may be misleading
13/20 N Benda: M&S in Drug Approval
Longitudinal model-based inference
Federal Institute for Drugs
and Medical Devices
 Repeated Scientific Advice question:
 Pivotal confirmatory Phase III study
 Longitudinal measurements at time t1, t2, ..., tn
 relevant endpoint at tn (end of treatment)
 primary analysis based on tn only or on a longitudinal
model ?
different possibilities
• time dependency functional or categorical ?
• covariance structured or unstructured ?
 Robustness (tn) vs more informative analysis
 “borrowing strength” or
“relying on assumptions difficult to verify” ?
14/20 N Benda: M&S in Drug Approval
Longitudinal model-based inference
Federal Institute for Drugs
and Medical Devices
 Case-by-case decision
 Relevant missing data issue and non-inferiority:
 consider assay sensitivity
 longitudinal analysis / MMRM (Mixed-Effect Model
Repeated Measure) preferred
 justify model (by M&S ?)
 Non-compliance and superiority vs placebo:
 use of measurements under non-compliance / after
discontinuation (retrieved data): “effectiveness”
 longitudinal analysis under compliance: “efficacy”
15/20 N Benda: M&S in Drug Approval
Small population dilemma
Federal Institute for Drugs
and Medical Devices
 Independent confirmation
vs historical information
 Population concerned
vs extrapolation from other population
 Modelling approaches to
 bridge historical information
 extrapolate from other population
 Trade-off
 Robustness and independent confirmation vs
presumably more informative analysis
 Less data available – more assumptions needed
16/20 N Benda: M&S in Drug Approval
Small population proposals
Federal Institute for Drugs
and Medical Devices
 M&S approaches to extrapolate
 Surrogate endpoints (PD) + adult evidence
 Meta-analytic approaches using historical data
 Bayesian: Evidence synthesis
 (Paediatric) subgroup analyses
 rely on transferability of (some) model components
 Increase type-1 error
Relying on more assumptions
False positives  - false negatives 
 missed drug worse than ineffective drug ?
17/20 N Benda: M&S in Drug Approval
Conclusions (1)
Federal Institute for Drugs
and Medical Devices
 Differentiate
 M&S to optimise study design
 M&S to explore and optimise development program
 M&S to predict efficacy and safety
 Differentiate
 M&S / Model building and exploration
 Model-based inference
18/20 N Benda: M&S in Drug Approval
Conclusions (2)
Federal Institute for Drugs
and Medical Devices
 Be honest with simulations
 Numerical tool
 Visualizing tool
 Be honest with modelling
 confirmatory inference independent of model building
 inference is always model-based
• amount and quality of assumptions to be assessed
 simplicity preferred if robustness is of concern
 trade-off between
• precision vs robustness
• false positives vs false negatives
19/20 N Benda: M&S in Drug Approval
Conclusions (3)
Federal Institute for Drugs
and Medical Devices
 Virtues of M&S
 increased understanding of underlying process
 may facilitate focus on distributions
 may optimise development program design
 Independent confirmation
 still required in Phase III in most applications
 low amount of assumptions / simplicity to ensure
robustness
 possible exceptions where false positive decisions are
worse than false negatives
20/20 N Benda: M&S in Drug Approval