1. Modeling and Simulation: Tool for Optimized Drug Development
Martin Roessner
Biostatistics sanofi aventis
Bridgewater, NJ

2. Outline Background Modeling and Simulation (M&S) approach
Clinical Utility Index (CUI)
Example: SERM
Conclusion

3. The industry needs to work smarter not harder
Industry challenge
Drug development process not much changed over the last 25 years
Drug development cost continue to increase ($802 Mill +)
Time to market, attrition rates and the number of late stage failures remain unchanged
The industry needs to radically rethink the drug development process to remain competitive
The industry needs to work smarter not harder

4. Modeling and Simulation is a tool for quantitative decision-making
It is a methodology that uses mathematical/statistical models and simulations in a predictive manner
M&S provides an integrated framework to use this information to optimize the drug development process
Preclinical Information
PK/PD data
Dose response information
Clinical outcome data (safety/efficacy)
Prior information: Historical data, information on related compounds, SBOAs, EPARs, etc.
Marketing and Financial projections

5. Development and broad adoption of M&S will help create value
Implementation of M&S
Development and broad adoption of M&S will help create value
Benefits
Optimized development strategies
Early termination of unpromising compounds
Reduction in late stage attrition
Shorter development time earlier to approval and launch
Increase number of drugs to market
Enhanced labeling
More accurate and dynamic risk assessment along the development

6. Integrated modeling and simulation can be used any time there is an important question impacting project value
“Is it worth developing a new dosage form?”
“What’s the best dose and schedule?”
“Should we continue this development program?”
“What is the optimal patient population for this drug?”
“Is there a clinical trial design that will show PoC and find the best dose?”
“What’s the probability of success in Phase 3?”
“Is this treatment likely to be as good as the competitors?”
“Which indication should we go into first to maximize the value of the program?”
“Should we in-license this compound?”
“What are the most important attributes of a 2nd generation compound?”

7. A modeling approach to decision-making involves integration of information from a number of sources
Clinical and Preclinical Data
Exploratory Data Analysis
Safety Dose-Response Model
Efficacy Dose-Response Model
Simulation
Physician Market Research
Clinical Utility Model
Integration

8. A modeling approach to decision-making involves integration of information from a number of sources
Clinical and Preclinical Data
Exploratory Data Analysis
Safety Dose-Response Model
Efficacy Dose-Response Model
Simulation
Physician Market Research
Clinical Utility Model
Integration

9. Clinical Utility Index (CUI) - a metric for the benefit of treatment to the patient (1)
Every drug has benefits and risks.
The relative importance of these characteristics depend on the disease the drug is intended to treat
They also change with dosage, patient population, etc.
Trade-offs must often be made among the drug effects comprising the product profile, balancing the benefits and risks.

10. Clinical Utility Index (CUI) - a metric for the benefit of treatment to the patient (2)
The CUI quantifies trade-offs by providing a single metric for the multiple dimensions of benefit and risk.
It is…
a systematic approach to understand subjective preferences
a transparent way of weighing tradeoffs
knowledge-driven; available data are used; if not available, rely on expert opinion
closely related to the Target Product Profile
It is not …
an “objective” measure in the sense of a physiological measurement

11. The framework for the CUI is elicited from the project team; when combined with models of response, it provides a relative estimate of the patient benefit
Identify Metrics
Identify Critical
Assign
and Relevant
Treatment
Preference
CUI
Response
CUI Distributions for
Attributes and
Values for each
Framework
Levels for each
Competing Treatments
Relative Weights
Response Level
Attribute 1 A
E(CUI ) A
P(CUI < X)
E(CUI ) B B
Treatment-
Response
CUI
Models
Probability of
Estimated
Here, treatment B is
Individual
Product
expected to be superior to A
Attribute Levels
Profile
Expert Opinion

12. Example
SERM, a Selective Estrogen Receptor Modifier for the Treatment of Osteoporosis in Post-Menopausal Women
Two Phase II studies:
1. Placebo, SERM (2.5mg, 10mg, 50mg) and Raloxifene, n=118
2. Placebo, SERM (0.5mg, 5mg) n=79
Primary efficacy endpoint was % change from baseline U-CTX
Included additional safety and activity endpoints
How does the efficacy, safety and tolerability of SERM compare with its major competitor drug and at which dose
Explorative analysis
Clinical Utility Index (CUI)
Simulation results and sensitivity analysis
Is it worthwhile to continue development

13. Possible responses and their clinical value for each attribute were defined
Preference Ratio
Efficacy on Bone
Worse than Raloxifene
Equivalent to Raloxifene
Better than Raloxifene 1 10 20
Endometrial Proliferation
The same or better than Ralox. 30
Endometrial Lining Thickness
The same or better than Ralox 5
Cardiovascular
Smaller effect on LDL than Ralox
Same or larger effect on LDL vs. Ralox.
Same effect on LDL + effect on HDL
7.5 15
…..
Food Effect on PK
Presence of food effect
Absence of food effect 2

14. Important attributes were ranked and their importance weighted
Rating
Relative Weight
Efficacy on Bone 1
100
0.27
Endometrial Proliferation
Endometrial Lining Thickness 3 50
0.14
Thromboembolic Disease 4 40
0.11
Hot Flashes 5 30
0.08
Breast Tenderness 6 15
0.04
Cardiovascular 7 10
0.03
Muscle Cramps
Atrophic Vaginitis
Food Effect on PK
0.01

15. Dose-Response for Urinary CTX (measure of bone turnover)
Models of dose-response provided estimates of attribute level and uncertainty in these estimates
Dose-Response for Urinary CTX (measure of bone turnover)
Baseline-adjusted week-12
% Difference from Placebo
Clear dose response
Log-Linear model adequately describe available data

16. Major Result: There was no dose for which SERM was expected to be considered equivalent or superior to Raloxifene 60
Based on CUI and simulated drug response
CUI for Raloxifene 50 40
Clinical Utility Index 30 20 10
0.25
0.5 1
2.5 5 10
SERM Dose (mg)

17. What if…….
If SERM did not cause endometrial proliferation, available data support effects of SERM would be similar or better at doses of 1 mg and higher
100
SERM
similar to Raloxifene
i.e. no endometrial
proliferation 80
Clinical Utility Index 60
Raloxifene 40 20
0.25 mg
0.5 mg
1 mg
2.5 mg
5 mg
10 mg
SERM Dose

18. Impact: Further development of SERM was halted, saving $50-100M in development costs
SERM fails to show equivalent clinical utility to Raloxifene at all doses examined
“Based on that simulation, ‘we stopped funding development of the compound,’ says Frank Douglas… the ratio between the therapeutic benefit and the side effect demonstrated that this [compound] was not as beneficial as Evista.’ … Douglas estimates that the … computer model … saved the company $50 million to $100 million, the cost of later-stage clinical trials. ‘We also avoided exposing a lot of women to a drug that ultimately would have failed,’ he adds. ‘And we were able to switch to another project with a greater chance of success.’ “
—Forbes 10/7/02

19. Conclusion Industry needs to operate smarter
M&S provides a framework to optimize drug development at various levels
Clinical Utility Index can be used to assess the potential success of a product in the market

20. Acknowledgement
B. Korsan, K. Dykstra, T.J. Carrothers (Pharsight)