dynochem_webinar_novartis_flav_susanne_11jun2014
DESCRIPTION
"Application of QbD on a Continuous Process" Development supported by process simulation. Flavien Susanne, Chemical Engineer, NOVARTISTRANSCRIPT
Development supported by process simulation
Flavien Susanne, Chemical Engineer
DynoChem Webinar 11th of June 2014
Application of QbD on a Continuous Process
Outline
Flavien Susanne / Dynochem Webinar 11th of June 2014
2
QbD The Right Time to Change
Why QbD for us in continuous process?
Our journey towards QbD...
Enabling process understanding for development and manufacturing
Build process robustness and reliability
Case study: Application to a multi-steps continuous process for PhII
Continuous process exist for century as
reported in The book De Re Metallica
published in 1556 representing 3 CSTRs with
gravity transfer system
Let’s move forward...
QbD and Design Space
Flavien Susanne / Dynochem Webinar 11th of June 2014 3
A Methodology for Good Development Practice
Process step
Process step
Process step
Process step
Continuous Process
Input
materials
Input
Process
Parameters
Monitoring of
the parameters
Product
pass or fail
Design Space
Process control
PAT
ICH concept of space design: multi-dimensional combination and interaction of input variables (e.g. material attributes) and process parameters that have been demonstrated to provide assurance of quality.
Goals: Deliver a robust and cost effective continuous process.
How can we achieve this? By reducing our product variability and increase our process acceptance.
Main Current Focus
Flavien Susanne / Dynochem Webinar 11th of June 2014 4
Risk Assessment and Space Design Activities
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Target the product profile
Determine critical quality attibues (CQAs)
Link process parameters to CQAs and perform risk assessment
Develop a Design Space
Design and implement a control strategy
Manage product lifecycle
Target the product profile
Determine critical quality attibues (CQAs)
Link process parameters to CQAs and perform risk assessment
Develop a Design Space
Design an implement a control strategy
Manage product lifecycle
Main Current Focus
Flavien Susanne / Dynochem Webinar 11th of June 2014 5
Risk Assessment and Space Design Activities
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Activities discussed in this talk
Continuous Synthesis
Flavien Susanne / Dynochem Webinar 11th of June 2014 6
Process Description
This synthesis consists of three consecutive transformations to be conducted as a continuous process. Such Lithium exchange chemistry is known as being highly scale dependant. Continuous processing was seen as a solution to improve scale up reliability.
reactions
side reactions
Impurity 1
Sum mass loss
Impurity 2
Critical Quality Attribute
Flavien Susanne / Dynochem Webinar 11th of June 2014 7
Objectives
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Definition: A Critical Quality Attribute (CQA) is a physical, chemical, biological, or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality. CQAs are generally associated with the drug substance, excipients, intermediates, and drug product.
High level of purity to meet specification of the product
Purity greater than 98% after full synthesis, work up and crystallisation. All impurity levels have to be below 0.1%
High yielding Yield greater than 60% for the complete sequence
High reliability of the process Minimise or suppress effect of scale up Minimise variability to less than 5-10%.
Enabling continuous processing has the purpose of minimising process variability and maximising process control. Significant yield and purity increase is seen as a bonus.
Q8(R1) Pharmaceutical Development Revision 1 :
http://www.fda.gov/downloads/RegulatoryInformation/Guidances/ucm128005.pdf
Critical Quality Attribute
Flavien Susanne / Dynochem Webinar 11th of June 2014 8
What Do We Want to Control
The Impurity 1 level is critical to quality output. This impurity is generated by reprotonation of
Compound 4 and decompostion of Compound 5.
Mass loss is a significant issue to the overall yield and needs to be controlled.
One CQA may
dominate the picture Reactions
Side reactions
Impurity 1
Sum mass loss
Impurity 2
One KQA may
dominate the picture
Standard process development is conducted through a laboratory optimization exercise and process validation is done with Lab statistical (DoE) models.
Complex continuous processes are scale dependent. Therefore, the risk assessment needs to include the non-linear scale-up effect. DoE are rarely appropriate.
Risk Assessment in a QbD Approach
Flavien Susanne / Dynochem Webinar 11th of June 2014 9
Mechanistic Thinking & Understanding
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Mechanistic
Thinking
Statistical
DoE
Critical Process
Parameters
Design
Space
Continuous
Process
Mechanistic
Model
Process
Understanding
For complex processing, mechanistic
modeling are put in place for process
development as well as process
validation.
“A mechanistic approach is justified whenever a basic
understanding of the system is essential to progress.”
George Box, Statistics for Experimenters, Wiley, 1978.
Standard process development is conducted through a laboratory optimization exercise and process validation is done with Lab statistical (DoE) models.
Complex continuous processes are scale dependent. Therefore, the risk assessment needs to include the non-linear scale-up effect. DoE are rarely appropriate.
.
Risk Assessment in a QbD Approach
Flavien Susanne / Dynochem Webinar 11th of June 2014 10
Mechanistic Thinking & Understanding
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Mechanistic
Thinking
Statistical
DoE
Critical Process
Parameters
Design
Space
Continuous
Process
Mechanistic
Model
Process
Understanding
For complex processing, mechanistic
modeling are put in place for process
development as well as process
validation.
“A mechanistic approach is justified whenever a basic
understanding of the system is essential to progress.”
George Box, Statistics for Experimenters, Wiley, 1978.
Define an unifying reaction mechanism for the main transformations and the side product formations. The overall mechanism is postulated initially from literature searches and experimentally validated.
Methodology to Process Simulation
11
Process Understanding: Establish the Overall Process View
Reactions
Side reactions
Impurity 1
Sum mass loss
Impurity 2
Flavien Susanne / Dynochem Webinar 11th of June 2014
Bulk liquid
Feed zone
Feed zone
Bulk liquid
Bulk liquid
UA
UA
UA
Solid
Solid
Solid
Solubility
kL
Solubility
kL
Solubility
kL
kinetic and physical rates
kinetic and physical rates
kinetic and physical rates
Graphical representation of a
Continuous Stirred Tank Reactor
What are the phases? Liquid and solid formation.
What are the components? Two solvents, starting material,
product and impurities.
What are the rates? Physical rates with mixing and mass
transfer. Kinetic rates for product and imputies formations.
Process Scheme
Flavien Susanne / Dynochem Webinar 11th of June 2014 12
Process Variables
In our continuous process, what are the process variables?
Which ones are the Critical Process Parameters?
Bulk liquid
Bulk liquid
Feed zone
Feed zone
UA
SolidSolubility
kL
kinetic and physical rates
kinetic and physical rates
Bulk liquid
Feed zone
Feed zone
Bulk liquid
Bulk liquid
UA
UA
UA
Solid
Solid
Solid
Solubility
kL
Solubility
kL
Solubility
kL
kinetic and physical rates
kinetic and physical rates
kinetic and physical rates
Flow rate and
residence time
Feedstock composition - Stoichimetric ratio
- Dilution
Heat transfer Heating or cooling capacity
- Continuous reactor type
- Property of the media
Must consider
kinetic
In our continuous process, what are the process variables?
Which ones are the Critical Process Parameters?
Variation of concentrations and stoichiometry From substoichiometric to large excess Extremely important to include time as a variable of the process (residence time).
Process Development Methodology
13
Process Variables
process temperature
Variation of temperature From -50°C to rt
process temperature
Flavien Susanne / Dynochem Webinar 11th of June 2014
Established platform enabling fast and reliable continuous process development
Process Development Methodology
14
Experimental Aspect
Process understanding &
development platform
Dedicated laboratory and infrastructure
batch and continuous equipments
Specific tools for process understanding FT-IR, FBRM, calorimetry
Process simulation tools DynoChem kinetic modeling
Reaction followed by FT-IR, Heat Flow, off line
LC and NMR
Mass balance needs to be greater than 95%
Mettler Toledo RC1 batch experiments for
energy balance, calculation and reaction
profiling.
Flow experiments for really short residence
time end points.
Flavien Susanne / Dynochem Webinar 11th of June 2014
Mechanistic modeling
Established platform enabling fast and reliable continuous process development
Process Development Methodology
15
Experimental Aspect
Process understanding &
development platform
Dedicated laboratory and infrastructure
batch and continuous equipments
Specific tools for process understanding FT-IR, FBRM, calorimetry
Process simulation tools DynoChem kinetic modeling
Flavien Susanne / Dynochem Webinar 11th of June 2014
Starting material 1 + BuLi > intermediate + butane
intermediate + BuLi > Compound 4 + butane
Aryl lithium + BuLi = Compound 4- Li
+ + butane
Compound 4 + Starting material 1 > Impurity 1
Compound 4 + intermediate > Impurity 1
Compound 4 > Impurity 2
Established platform enabling fast and reliable continuous process development
Process Development Methodology
16
Experimental Aspect
Process understanding &
development platform
Dedicated laboratory and infrastructure
batch and continuous equipments
Specific tools for process understanding FT-IR, FBRM, calorimetry
Process simulation tools DynoChem kinetic modeling
Flavien Susanne / Dynochem Webinar 11th of June 2014
Starting material 1 + BuLi > intermediate + butane
intermediate + BuLi > Compound 4 + butane
Aryl lithium + BuLi = Compound 4- Li
+ + butane
Compound 4 + Starting material 1 > Impurity 1
Compound 4 + intermediate > Impurity 1
Compound 4 > Impurity 2
Process Development Methodology
Flavien Susanne / Dynochem Webinar 11th of June 2014 17
Fitting of Kinetic Parameters
From the series of experiments, the kinetic parameters of the synthesis (main and side reactions) were fitted using DynoChem Algorithm (Simplex & Levenberg)
Main reactions defined as differential rate equations
-rSM = k1.exp[Ea/R*(1/T-1/T
0)]*CSM0
*(1-XSM)*CBuLi*(1-XBuLi)
T = T0 – DHrx/CpsolX*X
-rint = k1.exp[Ea/R*(1/T-1/T
0)]*Cint0
*(1-Xint)*CBuLi*(1-Xint)
T = T0 – DHrx/CpsolX*X
Energy balance is critical
in a continuous process
Fast reactions. Only relative
ratio compared to mixing and
impurity formation is required
Process Development Methodology
Flavien Susanne / Dynochem Webinar 11th of June 2014 18
Fitting of Kinetic Parameters
From the series of experiments, the kinetic parameters of the synthesis (main and side reactions) were fitted using DynoChem Algorithm (Simplex & Levenberg).
Side reaction and decomposition defined as differential rate equations.
-rprod = k1.exp[Ea/R*(1/T-1/T
0)]*[Cprod0
*(1-Xprod)]2*CBuLi*(1-XBuLi)
-rprod = k1.exp[Ea/R*(1/T-1/T
0)]*Cprod0
*(1-Xprod)
▼Experimental values
-- model prediction
Process Development Methodology
Flavien Susanne / Dynochem Webinar 11th of June 2014 19
Fitting of Kinetic Parameters
From the series of experiments, the kinetic parameters of the synthesis (main and side reactions) were fitted using DynoChem Algorithm (Simplex & Levenberg).
Side reaction and decomposition defined as differential rate equations.
-rSM = k1.exp[Ea/R*(1/T-1/T
0)]*CSM0
*(1-XSM)*CBuLi*(1-XBuLi)
T = T0 – DHrx/CpsolX*X
-rint = k1.exp[Ea/R*(1/T-1/T
0)]*Cint0
*(1-Xint)*CBuLi*(1-Xint)
T = T0 – DHrx/CpsolX*X
-rprod = k1.exp[Ea/R*(1/T-1/T
0)]*[Cprod0
*(1-Xprod)]2*CBuLi*(1-XBuLi)
-rprod = k1.exp[Ea/R*(1/T-1/T
0)]*Cprod0
*(1-Xprod)
Acceptable prediction
of the chemistry
Design Space
Flavien Susanne / Dynochem Webinar 11th of June 2014 20
How to use Process Simulation
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Design space definition: A design space can be defined in terms of ranges of input variables or parameters, or through more complex mathematical relationships. It is possible to define a design space as a time dependent function.
Design space will be defined using a mechanistic/kinetic model.
The design space is generated through process understanding
Q8(R1) Pharmaceutical Development Revision 1 :
http://www.fda.gov/downloads/RegulatoryInformation/Guidances/ucm128005.pdf
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 21
Process Operation Step 1
The process temperature is a CPP, it is directly affected by:
The heat balance, energy input (DHr) – energy output (heat transfer)
Process flow rate, energy provided by unit of time
Characteristic of the reactor, CSTR Vs PFR
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 22
Process Operation Step 1
Impurity 1 below acceptance level Impurity 3 below acceptance level
The process flow rate is a CPP, it is directly affected by: Relative stoichiometry
Residence time
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 23
Process Operation Step 1
Yield above acceptance level
Overlapping multiple response surface
to construct design space
Process design
space
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 24
Process Operation Step 1
Process
design space
To account for the process simulation uncertainty (Sum of Square = deviation between
prediction and experimental values) as well as the response error.
To ensure CQAs within specification.
The design space has been reduced to 50% of its overall space.
“Buffer regions” to account for error
Compound 1 flow rate can deviate by +/-
6% from set point
BuLi 1 flow rate can deviate by +/- 6%
from set point
Temperature can deviate by +/- 5C from
set point
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 25
Process Operation Step1
The quality output from step 1 becomes
one of the variables of step 2
Process
design space
“Buffer regions” to account for error
Compound 1 flow rate can deviate by +/-
6% from set point
BuLi 1 flow rate can deviate by +/- 6%
from set point
Temperature can deviate by +/- 5C from
set point
To account the process simulation uncertainty (Sum of Square = deviation between prediction
and experimental values) as well as the response error.
To ensure CQAs within specification.
The design space has been reduced to 50% of its overall space.
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 26
Process Operation Step 2 and Following Steps
Process
design space
Process
design space
Consecutive design spaces of operation for each step have been defined.
It is critical to ensure that each design space is compatible to the next without restraining too much the potential of operation.
Process
design space
Step A
Step C
Step B
Overall design space of
operation of the process.
Response Surface
Flavien Susanne / Dynochem Webinar 11th of June 2014 27
Process Simulation for Continuous Reactor
As all the reactions have been highly exothermic, a reactor system allowing fast and efficient control of the internal temperature was critical.
Continuous Stirred Tank Reactor was forseen with a split of the feed of BuLi. CSTRs was also accommodating the solid formation by avoiding clogging.
Various combinations of size and materials were insilico calculated to assess the compatibility with the process.
Vessel 1
Process Solvent
Temperature -20
Fouling factor 5000
Agitation properties
Impeller speed 500 rpm
Service Fluid 10 10
Temperature -35
Fouling factor 5000 1
Results summary
hi 2007.68 W/m2 K
hw 1908.71 W/m2 K
ho 1330.72 W/m2 K
U 563.864 W/m2 K
UA 0.746 W/K
duty -0.011 kW
cooling rate -87.699 C/min
Twall -24.21 C
UA intercept 0.13 W/K
UA(v) in straight side 112.77 W/L K
Add to Report
Development of the Scale-up Process
Flavien Susanne / Dynochem Webinar 11th of June 2014 28
Based on the response surface the process conditions were defined Residence time
Process temperature Stoichiometry
Small footprint of 4m2
Process Control
Flavien Susanne / Dynochem Webinar 11th of June 2014 29
Simple Pass/ Fail Strategy
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
The process has been developed to accommodate reasonable process
deviation and still deliver quality matching specification.
The process has been developed to accommodate reasonable process
deviation and still deliver quality matching specification.
As the process is developed in CSTRs, deviation outside the operating range
could be acceptable if they can be buffered by the volume of the vessel.
A deviation below 10% for less than 0.75sec. is still acceptable
Process Control
Flavien Susanne / Dynochem Webinar 11th of June 2014 30
Simple Pass/ Fail Strategy
Product
profile
CQAs
Risk
assessment
Design
space
Control
strategy
Continuous
Improvement
Deviation still below time specification
Conclusion
Flavien Susanne / Dynochem Webinar 11th of June 2014 31
Process development was supported by process simulation and PAT.
The process transfer was enabled using QbD with process simulation.
As the first campaign using such methodology of process simulation and QbD, additional process checks were conducted to validate the prediction during Piloting.
• Compound 6 + 2 steps has been continuously produced during 1 week production.
The average yield to compound 6 (3 continuous chemical transformations) was 85%. The yield on the continuous process was increased by 31%.
Acknowledgement
32
CM Team Fransceco Venturoni
Benjamin Martin
Michel Aubry
Jutta Polenk
Berthold Schenkel
Fabio Lima
Joerg Sedelmeier
Mario Rentsch
Serbi Sevinc
Safety Lab Christoph Heuberger
Michael Schonhardt
LSC Carlo Jungo
Holger Scheidat
CHAD Isabelle Gallou
Thomas Kuhnle
Flavien Susanne / Dynochem Webinar 11th of June 2014