High Impact Model
Role of Models in Pharmaceutical Development: A Regulatory Perspective Sharmista Chatterjee, Ph.D. Division Director (Acting) Office of Process and Facilities (OPF) Office of Pharmaceutical Quality (OPQ), FDA
AAPS Annual Meeting October 28, 2015 Disclaimer: This presentation reflects the views of the author and should not be construed to represent FDA’s views or policies.
Examples of Models in a QbD Approach Product profile
• In vivo/In vitro correlations
CQAs
Risk assessment
Design space Control strategy
Continual Improvement
• Conceptual models, Failure Modes & Effects Analysis, DoEs for Parameter Identification, Models for sensitivity analysis
• DoEs for Range Determination, Kinetic Models, Fluid Mechanic Models, MV based models to define design space • Scale-up Correlation, Process Control models Chemometric Models for Spectral Data • Statistical Process Control / MV based models for process monitoring
2
ICH Points to Consider Document
http://www.ich.org/products/guidelines/quality/article/quality-guidelines.html
3
Categorization of Models • Multiple options for categorization • Models can be categorized by BOTH mode of implementation and contribution to quality – Dependent on mode of model implementation – Dependent on the model’s impact on overall control strategy
4
Categorization Dependent on Mode of Implementation
Models for supporting process design
Models for supporting analytical procedures
- Prediction from these models used for supporting process development e.g. models to optimize formulation, model for scale up, design space models
- Includes calibration models associated with various PAT (Process Analytical Technology) based methods e.g. NIR
Models for process monitoring and controlTwo broad categories are: - Univariate or multivariate statistical process control models - Feed forward or feed back models 5
Categorization Dependent on Impact to Quality High Impact Models
Prediction from the model is the sole indicator of quality of the product, e.g. calibration model associated with a NIR method that supports RTRT
Medium Impact Models
Important for assuring quality of the product but are not the sole indicators of quality, e.g. model to define a design space
Low Impact Models Typically used to support process development efforts, e.g. formulation optimization model 6
Examples of High Impact Models Model used as surrogate for traditional release test (e.g. dissolution)
Calibration model associated with NIR method
NIR used for RTRT (Real Time Release Testing) of tablet assay and content uniformity
7
Example: Deciding Impact of Models 0.04 0.035
E(t) (1/s)
0.03 0.025 0.02 0.015 0.01 0.005 0
0
50
100
150
Time (Sec)
Continuous Twin Screw Granulator
Residence Time Distribution (RTD) Model
Low Impact: RTD model used during development to decide on sampling strategy High Impact: RTD model used to make release decisions during routine production Source of Schematics: Fernando Muzzio, Rutgers University
8
Common Regulatory Concerns • Do all models need to be validated? NO • Do all models need to be maintained? NO • Do details for all models need to be in the submission? NO Amount of effort for model validation, maintenance and documentation should be commensurate with the model’s impact 9
Documentation Considerations • Documentation in regulatory filings is dependent on the intended use of the model and the risk associated with it – High, medium, low impact models
• Refer Section 5 of ICH Q8,9, 10 Implementation, Points to Consider document for details • Model maintenance details typically documented within the firm’s Quality System – For high impact models a summary can be provided in the submission – High level approach of model maintenance plans can be considered as an ‘Established Condition’ for an application (per FDA draft ‘Established Conditions’ guidance) t 10
Considerations for Submissions** • Low Impact Model – Discussion of how model predictions were used to make decisions during process development
• Medium Impact Model - Model assumptions, tabular or graphical summary of model inputs and outputs, relevant model equations, statistical analysis where appropriate, comparison of model prediction with measured data, discussion of how elements in the control strategy may help to mitigate uncertainty in the model.
• High Impact Model – Model assumptions, appropriateness of sample size, data pretreatment, justification of variable selection, tabular or graphical summary of model inputs and outputs, model equations, statistical analysis of data showing fit and prediction ability, rationale for setting of model acceptance criteria, model verification (internal and external), discussion of approaches for model maintenance and update 11 **Ref: ICH Points to Consider document
Regulatory Notification of Post Approval Changes to Models • Selection of reporting category for post approval change depends on potential for the change to have an adverse effect on product quality (CFR 314.70) • Can leverage comparability protocols for alternate reporting categories Types of Changes and Reporting Categories * Potential impact of change on control strategy
Potential impact of failure on product quality
Low
Medium
High
Low
Minor Change (Annual report)
Minor Change (Annual report)
Moderate Change (CBE 30)
High
Minor Change (Annual report)
Moderate Change (CBE 30)
Major Change (PAS)
12 *Ref: Adapted from Draft FDA NIR Guidance
12
Common Question Related to Model Maintenance
Does one need to provide all model maintenance information in a regulatory filing? • Details of model maintenance information is typically documented within the site’s Quality System • For high impact models (e.g., models to support RTRT) to facilitate evaluation of overall control strategy, consider providing:
- high level summary of model maintenance plans, that includes (but is
not
limited to): - triggers of model update - criteria for recalibration - level of revalidation
ypically documented within the site’s Quality System • http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_9_10_QAs/PtC/Quality_IWG_PtCR2_6dec2011.pdf
13 13
Conclusion • Models can support pharmaceutical development as well implementation of modern pharmaceutical manufacturing concepts such as design space, RTRT, continual process monitoring, process control • FDA strongly recommends discussing any ‘novel’ high impact model approaches during End-of-Phase II meetings and Pre NDA meetings – Data for submissions – Approaches for model validation – Overall plans for model maintenance
• For further details refer to the ICH Points To Consider document on ‘Models in the QbD Paradigm’ 14
Thank you! Questions, comments, concerns: [email protected] 15
AAPS Annual Meeting October 28, 2015 Disclaimer: This presentation reflects the views of the author and should not be construed to represent FDA’s views or policies.
Examples of Models in a QbD Approach Product profile
• In vivo/In vitro correlations
CQAs
Risk assessment
Design space Control strategy
Continual Improvement
• Conceptual models, Failure Modes & Effects Analysis, DoEs for Parameter Identification, Models for sensitivity analysis
• DoEs for Range Determination, Kinetic Models, Fluid Mechanic Models, MV based models to define design space • Scale-up Correlation, Process Control models Chemometric Models for Spectral Data • Statistical Process Control / MV based models for process monitoring
2
ICH Points to Consider Document
http://www.ich.org/products/guidelines/quality/article/quality-guidelines.html
3
Categorization of Models • Multiple options for categorization • Models can be categorized by BOTH mode of implementation and contribution to quality – Dependent on mode of model implementation – Dependent on the model’s impact on overall control strategy
4
Categorization Dependent on Mode of Implementation
Models for supporting process design
Models for supporting analytical procedures
- Prediction from these models used for supporting process development e.g. models to optimize formulation, model for scale up, design space models
- Includes calibration models associated with various PAT (Process Analytical Technology) based methods e.g. NIR
Models for process monitoring and controlTwo broad categories are: - Univariate or multivariate statistical process control models - Feed forward or feed back models 5
Categorization Dependent on Impact to Quality High Impact Models
Prediction from the model is the sole indicator of quality of the product, e.g. calibration model associated with a NIR method that supports RTRT
Medium Impact Models
Important for assuring quality of the product but are not the sole indicators of quality, e.g. model to define a design space
Low Impact Models Typically used to support process development efforts, e.g. formulation optimization model 6
Examples of High Impact Models Model used as surrogate for traditional release test (e.g. dissolution)
Calibration model associated with NIR method
NIR used for RTRT (Real Time Release Testing) of tablet assay and content uniformity
7
Example: Deciding Impact of Models 0.04 0.035
E(t) (1/s)
0.03 0.025 0.02 0.015 0.01 0.005 0
0
50
100
150
Time (Sec)
Continuous Twin Screw Granulator
Residence Time Distribution (RTD) Model
Low Impact: RTD model used during development to decide on sampling strategy High Impact: RTD model used to make release decisions during routine production Source of Schematics: Fernando Muzzio, Rutgers University
8
Common Regulatory Concerns • Do all models need to be validated? NO • Do all models need to be maintained? NO • Do details for all models need to be in the submission? NO Amount of effort for model validation, maintenance and documentation should be commensurate with the model’s impact 9
Documentation Considerations • Documentation in regulatory filings is dependent on the intended use of the model and the risk associated with it – High, medium, low impact models
• Refer Section 5 of ICH Q8,9, 10 Implementation, Points to Consider document for details • Model maintenance details typically documented within the firm’s Quality System – For high impact models a summary can be provided in the submission – High level approach of model maintenance plans can be considered as an ‘Established Condition’ for an application (per FDA draft ‘Established Conditions’ guidance) t 10
Considerations for Submissions** • Low Impact Model – Discussion of how model predictions were used to make decisions during process development
• Medium Impact Model - Model assumptions, tabular or graphical summary of model inputs and outputs, relevant model equations, statistical analysis where appropriate, comparison of model prediction with measured data, discussion of how elements in the control strategy may help to mitigate uncertainty in the model.
• High Impact Model – Model assumptions, appropriateness of sample size, data pretreatment, justification of variable selection, tabular or graphical summary of model inputs and outputs, model equations, statistical analysis of data showing fit and prediction ability, rationale for setting of model acceptance criteria, model verification (internal and external), discussion of approaches for model maintenance and update 11 **Ref: ICH Points to Consider document
Regulatory Notification of Post Approval Changes to Models • Selection of reporting category for post approval change depends on potential for the change to have an adverse effect on product quality (CFR 314.70) • Can leverage comparability protocols for alternate reporting categories Types of Changes and Reporting Categories * Potential impact of change on control strategy
Potential impact of failure on product quality
Low
Medium
High
Low
Minor Change (Annual report)
Minor Change (Annual report)
Moderate Change (CBE 30)
High
Minor Change (Annual report)
Moderate Change (CBE 30)
Major Change (PAS)
12 *Ref: Adapted from Draft FDA NIR Guidance
12
Common Question Related to Model Maintenance
Does one need to provide all model maintenance information in a regulatory filing? • Details of model maintenance information is typically documented within the site’s Quality System • For high impact models (e.g., models to support RTRT) to facilitate evaluation of overall control strategy, consider providing:
- high level summary of model maintenance plans, that includes (but is
not
limited to): - triggers of model update - criteria for recalibration - level of revalidation
ypically documented within the site’s Quality System • http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_9_10_QAs/PtC/Quality_IWG_PtCR2_6dec2011.pdf
13 13
Conclusion • Models can support pharmaceutical development as well implementation of modern pharmaceutical manufacturing concepts such as design space, RTRT, continual process monitoring, process control • FDA strongly recommends discussing any ‘novel’ high impact model approaches during End-of-Phase II meetings and Pre NDA meetings – Data for submissions – Approaches for model validation – Overall plans for model maintenance
• For further details refer to the ICH Points To Consider document on ‘Models in the QbD Paradigm’ 14
Thank you! Questions, comments, concerns: [email protected] 15
