1. Using Shiny Tools to Establish a Stability Thermal Budget Contingency
Aimee Buesgen, Research Scientist
Eli Lilly and Company
May 24, 2017

2. Outline Introduction Stability Thermal Budget Statistical Simulation
Shiny Analytics Tool
Conclusion
05/24/2017

3. Introduction
Supply chain for cold chain products is designed and qualified to minimize temperature excursions during production, distribution, and consumer use
Allowances for unplanned temperature excursions enhance the patient experience and ensure a continued supply
05/24/2017

4. Introduction
With the current focus on clinical relevance during specification negotiations for new products, a previous strategy for establishing allowances for unplanned temperature excursions have become inadequate, leading to the use of statistical simulations
05/24/2017

5. Motivation Patients Regulatory Product Use Business Patients
Need to know that their medicine is acceptable to use.
Regulatory
Increasing regulatory pressure on specification setting. Previous decision-making strategy of stacking every worst case event is no longer feasible.
Product Use
Every product has a registered shelf-life period during which it is acceptable to use if it is stored at label storage conditions.
It is impractical to keep products at recommended storage conditions at ALL times during production, distribution and patient use.
Business
Any single batch does not experience every worst case scenario throughout manufacture/distribution, so a risk-based decision strategy is possible.
05/24/2017

6. Outline Introduction Stability Thermal Budget Statistical Simulation
Shiny Analytics Tool
Conclusion
05/24/2017

7. Stability Thermal Budget
Definition: The amount of time a drug product in primary container can spend outside of registered label conditions without risk to patient (i.e., to ensure specifications are met throughout shelf-life period and any applicable patient-use/ in-use period)
See PDA TR 53 for examples
05/24/2017

8. Stability Thermal Budget Concept
Note: This is an example and is not representative of an actual allocation amount
Assumption: 2-8°C shipping
Units: Thermal budget is expressed as time units for simplicity but is actually based on product change
STORAGE TIME AT 2-8°C
PATIENT USE / IN USE TIME
DISTRIBUTION TOR
UNPLANNED TIME/TEMPERATURE EXCURSIONS
MANUFACTURING TOR
Fixed
Variable
TOR = Time Out of Refrigeration
05/24/2017

9. Stability Budget “Units”
4/14/2018
Temperature outside of registered label conditions is as important as Time
Time Out of Refrigeration (TOR)
Time at a given Temperature
Change in drug product during exposure
05/24/2017

10. Outline Introduction Stability Thermal Budget Statistical Simulation
Shiny Analytics Tool
Conclusion
05/24/2017

11. Product Change Travel to Assembly Assembly Travel to Packaging
End of Shelf-
Life Criteria
Batch Age
Expiry
Long-Term Stability
In-Use or Patient-Use Stability
Available Space for Unplanned Excursions
Analytical Property
Travel to Assembly
Assembly
Travel to Packaging
Packaging
Distribution
and Supply Chain
Batch Release
05/24/2017

12. Manufacturing/Release
Worst Case Stacking
Contingency
No room for unplanned excursions
In-use/Patient-use
Long-term Stability
Distribution
Packaging
Travel to Packaging
Assembly
Manufacturing/Release
Jeopardized patient experience
Jeopardized supply
Travel to Assembly
05/24/2017

13. Simulation of Many Batches at Expiry
Batch Release
Distribution
Distribution at Expiry
05/24/2017

14. Simulation Components
Batch Release
Travel to Assembly
Assembly
Travel to Packaging
Packaging
Distribution
Long-Term Stability
In/Patient-Use Stability
Obtain Distributions at Expiry
Objective: To predict a distribution of batch results at end of the product shelf-life period
05/24/2017

15. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

16. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

17. Batch Release
Batch release data already account for a range of experienced Manufacturing TOR
Process average
Process standard deviation
Simulate a distribution of batch
release results
Option to truncate distribution
Batch Release
Distribution
05/24/2017

18. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

19. Travel to Assembly
Product travels from the manufacturing site to another site for assembly
2 hours at 25C is allowed for loading at manufacturing site
2 hours at 25 C is allowed for unloading at assembly site
(similar for Travel to Packaging)
05/24/2017

20. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

21. Packaging Operations
Site SOP provides maximum TOR for packaging operations, e.g., 48 hours
Very few pallets experience 48 hours TOR
Use actual data to determine probabilities
(similar for Assembly Operations)
05/24/2017

22. Probability Mass Function
Packaging TOR (hours)
per pallet
Probability
2 < X ≤ 8
0.7
8 < X ≤ 12
0.235
12 < X ≤ 24
0.05
24 < X ≤ 36
0.01
36 < X ≤ 48
0.005
The probabilities sum to 1 and all packaging events must be captured in one of the bins.
05/24/2017

23. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

24. Distribution Operations
Product travels the supply chain through various distribution nodes
x hours at y˚C is allowed at each distribution node for loading/unloading, with time and temperature dependent upon node
05/24/2017

25. Components of Statistical Simulation
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

26. Shelf-Life Period
Change is calculated based on shelf-life period and rate of change at long-term stability storage condition (2-8˚C)
05/24/2017

27. Components of Statistical Simulation
4/14/2018
Through Batch Release
Travel to Assembly
Assembly Operations
Travel to Packaging
Packaging Operations
Distribution Operations
Shelf-Life Period
Patient-Use/In-Use Period
Different Times
Different Temperatures
Different Changes
05/24/2017

28. Patient-Use/In-Use Period
Change is calculated based on patient-use or in-use period and rate of change at applicable storage condition (e.g., 25˚C)
05/24/2017

29. Change Estimates
Rate of change at key temperatures is calculated using the Arrhenius Model
Long-term storage condition
Accelerated storage condition
Patient-use/In-use storage condition
05/24/2017

30. Arrhenius Model k = reaction rate coefficient a = frequency factor
Ea = activation energy coefficient
R = universal gas constant
T = temperature (in Kelvin)
We fit this model and assume rate of change (k) is linear over time.
The time periods considered are short, so the linear
assumption is reasonable
05/24/2017

31. Simulation of Total Allowable Change
Simulated Distribution of
Release Values at Expiry
Batch Release
Distribution
End of Shelf-Life Criteria
Percentile
Total Allowable Change (TAC) = Available Space for Unplanned Excursions
05/24/2017

32. Per Excursion Allowable Change (PEACh)
More than a single excursion may occur during distribution and/or consumer use
Assume the worst case number of excursions based on the number of distribution nodes and number of patient excursions allowed
PEACh= TAC / max # of excursions
05/24/2017

33. Threshold Time and Limiting Property
One can solve explicitly for the threshold time (the time above which the excursion has changed the product too much)
Multiple analytical properties are considered, and each has its own PEACh, temperature model, and threshold times
The property with the smallest threshold time at a given temperature is the limiting analytical property for that temperature
05/24/2017

34. Time Temperature Matrix (TTM)
A Time Temperature Matrix defines changes to a drug product for a range of times and temperatures for an analytical property for a single excursion.
It is established using data from stability studies and modeling techniques (e.g., Arrhenius model).
05/24/2017

35. TTM Example
4/14/2018
The product can be stored at 20˚C for up to 24 hours before the change during a single excursion is “too much”.
05/24/2017

36. Outline Introduction Stability Thermal Budget Statistical Simulation
Shiny Analytics Tool
Conclusion
05/24/2017

37. Shiny Analytics Tool Programmed using R/Shiny applications
Simulates distribution of limiting analytical property at end of shelf-life period
Allows for calculation of available space for unplanned excursions (contingency)
Provides Time Temperature Matrix for limiting analytical property
05/24/2017

38. Threshold Time Plot (TTP)
Helps determine limiting analytical property and the temperature range for which that is the case
Temperature (˚C)
Product XYZ Threshold Time Plot
Threshold Time for Single Excursion (hours)
Property A is the limiting analytical property
05/24/2017

39. Total Allowable Change
The Total Allowable Change is calculated for each considered analytical property and is based on the end of shelf-life criterion (RAC)
05/24/2017

40. Specification Change
Suppose the end of shelf-life criterion for Property C changes from NLT 97.0% to NLT 97.5%
What is the impact?
05/24/2017

41. Specification Change The limiting analytical property changes
Temperature (˚C)
Threshold Time for Single Excursion (hours)
Product XYZ Threshold Time Plot
Property C is now the limiting analytical property
05/24/2017

42. Specification Change
There is very little available space for unplanned excursions
TAC reduction of 0.5
From to
05/24/2017

43. Outline Introduction Stability Thermal Budget Statistical Simulation
Shiny Analytics Tool
Conclusion
05/24/2017

44. Conclusion
4/14/2018
Allowances for unplanned temperature excursions enhance the patient experience and ensure a continued supply
Statistical simulation using R/Shiny is a valuable Analytics tool that
Provides flexibility and allows rapid evaluation of “what if” scenarios
Allows for risk-based business decisions
Leverages in-house computing infrastructure
Simulates 50k batches in 2 seconds, 1 million batches in 30 seconds.
05/24/2017

45. Questions
05/24/2017

46. Acknowledgements Rebecca Elliott, Eli Lilly & Company (retired)
Luke Settles, Eli Lilly & Company (summer intern)
05/24/2017

47. References
PDA TR 53. (2011). Guidance for Industry: Stability Testing to Support Distribution of New Drug Products.
van Asselt, E. & Bishara, R. (2015). Establishing and Managing the Drug Product Stability Budget. Pharmaceutical Outsourcing.
05/24/2017

48. Aimee Buesgen buesgen_aimee@lilly.com (317) 433 - 4122
Contact Information
Aimee Buesgen (317)
05/24/2017

49. Speaker Biography B.S. Mathematics, University of Mary Washington
M.S. Mathematics, University of Tennessee
M.S. Statistics, University of Tennessee
Over 16 years experience in pharmaceutical manufacturing with expertise in insulin commercialization and stability data analysis
ASQ Certified Quality Engineer
05/24/2017