umass lowell | umass lowell - systems technology in … a1 braatz lowell talk-v2... ·...

Richard D. Braatz

Novartis-MIT Center for Continuous Manufacturing

Center for Biomedical Innovation

Department of Chemical Engineering

Systems Technology in Pharmaceutical

and Biologics QbD Implementation

2 | Novartis/MIT, All Rights Reserved

Background on QbD approach

Process development involves:

• Define the target product profile

• Identify the critical quality attributes (CQAs)

• Select an appropriate manufacturing strategy

• Implement a control strategy

This talk is on control systems technology for integrated pharmaceutical and biologics manufacturing


Why Integrated Manufacturing?

Reduce contact between

biology/chemistry & personnel

Continuous operation has

the potential to

• Increase product quality

• Increase yields

• Enable new drug product

formulations (e.g., thin films)

• Reduce scale-up risks

• Reduce footprint


Outline

Control Systems for Integrated Continuous Operations

Design Spaces vs. Feedback Control

Application to Biologics Manufacturing


Integrated Control Strategy

for Continuous Manufacturing

Tight integration of continuous operations can result in disturbances propagating downstream, unless their effects are suppressed by an integrated control strategy

The strategy must optimize the overall plant operation instead of only isolated units (i.e., need plantwide control)

D2

D1

R2

R1LT LT

M1 M2 S1

C1 C2

W1

P1

P2

P3

P4

P5

P9

P7 P8

P10LT

LT LT

C3 C4

P18 P19

P6

LT

P11

P13

P14

P17

P20

P12

P15

P16

LT LT

P22

P24

S3

S4 W2

M3 M4

M5

S5S6 E1 CS

CAT

A

B

S1

S2

S3

S1

PU1

S1

PU2

PU4

PU3

S1

S1

S1

C

D

E

S1

EX1

EX2

S2

FP

s1

s2

s3

s4

s5

s6

s7

s10

s9

s8

s11 s12s13

s14s15

s16

s18

s19

s20

s17

s21

s22

s23

s24

s25

s26s27 s28 s29

s30

s31

s32

s33

s35

s34

s36

s37

s38

s40 s41

s43

s45

s46

s47s39

s44

Q

LT

P21

P21S1

CT FT

FT

CT

DT

CT

CT

CT

FT

DT CT

STTCFT CT

s42

FT


Plantwide Control of Continuous Manufacturing

Challenges

• Many connected unit operations

• Very fast to slow processes

• Multi-purpose plants with short development time

• Alignment with regulatory requirements (e.g., design space)

Approach adapted from the chemical industry

• Employs systematic and modular design of plantwide control strategies for continuous manufacturing facilities

• Experimentally demonstrated on continuous pilot plant

R. Lakerveld, B. Benyahia, R.D. Braatz, & P.I. Barton, Model-based design of a plant-wide control

strategy for a continuous pharmaceutical plant, AIChE Journal, 59, 3671-3685, 2013

D2

D1

R2

R1LT LT

M1 M2 S1

C1 C2

W1

P1

P2

P3

P4

P5

P9

P7 P8

P10LT

LT LT

C3 C4

P18 P19

P6

LT

P11

P13

P14

P17

P20

P12

P15

P16

LT LT

P22

P24

S3

S4 W2

M3 M4

M5

S5S6 E1 CS

CAT

A

B

S1

S2

S3

S1

PU1

S1

PU2

PU4

PU3

S1

S1

S1

C

D

E

S1

EX1

EX2

S2

FP

s1

s2

s3

s4

s5

s6

s7

s10

s9

s8

s11 s12s13

s14s15

s16

s18

s19

s20

s17

s21

s22

s23

s24

s25

s26s27 s28 s29

s30

s31

s32

s33

s35

s34

s36

s37

s38

s40 s41

s43

s45

s46

s47s39

s44

Q

LT

P21

P21S1

CT FT

FT

CT

DT

CT

CT

CT

FT

DT CT

STTCFT CT

s42

FT


A continuous pilot plant


D2

D1

R2

R1LC LC

M1 M2 S1

C1 C2

W1

LC LC

C3 C4

LC

LCLC

S3

S4W2

M3 M4

M5

S5S6 E1 CS

CAT

A

B

S2

S1

S3

S1

PU1

PU2

PU4

PU3

S1

S1

S1

C

D

E

S1

EX1

EX2

FP

LC

S1

FCsp

FT

FCsp

spsp

sp

FCsp

CCFT

CTFC

FCsp

TC

sp

spsp

FCsp

DCsp

sp

S1

RC CCsp

DC

LC

sp

1

CATA B I

1 2 BPI C I P

2I E API


• First-principles dynamic models were built for each unit operation (UO) as they were developed

• Models were validated and then placed into a plant-wide simulation

• Plant simulation used to design UO & plantwide control strategy


Model-based Design of a Plant-wide Control Strategy

2. Optimizing control

objectives Level 1 – Total I/O

Level 4 – Detailed

Control tasks

(CQA, CPP)1

Level 2 – Intermediates

1. Stabilizing control

objectives (local)

Structure of plant-wide

control strategy

Level 3 – Recycles

Control tasks

(CQA, CPP)2

Control tasks

(CQA, CPP)3

Control tasks classified into

optimizing and stabilizing

Hierarchical decomposition

• Reduces complexity

• Exploits separation of time scales


Parametric Sensitivities Used to Evaluate Relationships Between CPPs and CQAs

0 0 0, , ,ft t t x t t x - State variables

- Input variables

- Output variables (CQAs)

for all feedback control loops

0

1( ) ( ) ( ) ,

( ) ( )

t

MV p D

I

SP CV

du t K t d

dt

t y y t

( ) ( ), ( ), , ,

( ) ( ), ( ), ,

( ), ( ), , 0,

dx t f x t u t p t

dt

y t g x t u t p t

h x t u t p t

( )

( )

( )

x t

u t

y t

p - Parameters (PPs)

Use sensitivities (Si,j) to identify

causal relations CPPs-CQAs:

• Direction and order of magnitude

• Guide selection of automated control loops

• Determined from process simulation (could use DOE)

,i

i j

j

yS

p


Example Sensitivity Results: Level 1: Total Inputs & Outputs


Also use sensitivities to evaluate dynamic I/O relationships, to assess controllability and disturbance propagation

0

0.2

0.4

0.6

0.8

1

0 50 100 150

No

rmal

ize

d S

en

siti

vity

Time / h

API dosage

Impurity content

Model-based sensitivity of two final product quality variables

with respect to feed flow rate of reactant


D2

D1

R2

R1LC LC

M1 M2 S1

C1 C2

W1

LC LC

C3 C4

LC

LCLC

S3

S4W2

M3 M4

M5

S5S6 E1 CS

CAT

A

B

S2

S1

S3

S1

PU1

PU2

PU4

PU3

S1

S1

S1

C

D

E

S1

EX1

EX2

FP

LC

S1

FCsp

FT

FCsp

spsp

sp

FCsp

CCFT

CTFC

FCsp

TC

sp

spsp

FCsp

DCsp

sp

S1

RC CCsp

DC

LC

sp

1

CATA B I

1 2 BPI C I P

2I E API


Met all purity

specs in Summer

2012


Outline





Design Space vs. Feedback Control (both are consistent with quality-by-design principles)

Design-space methods:

• Control strategy based on operation

within a fixed parameter space

• Applicable to each continuous

process unit operation

• More complicated to apply to an

entire continuous pharmaceutical

manufacturing plant

Feedback methods:

• Control strategy based on

feedback to a “parameter space”

• Easier to scale up

• Design space does not need to

be exhaustively validated a priori

• Necessary for continuous

manufacturing


Outline




Product manufacturing:

Allston, MA USA

Product QC: Haverhill, UK

Fill/finish: Waterford, IE

Cerezyme patients distributed worldwide

Manufacturing biologic drugs today

17

Towards Biomanufacturing on Demand (BioMOD)

Design Requirements Patient

18

BioMOD capabilities

• Enable flexible methodologies for genetic engineering/modification of microbial strains to synthesize multiple and wide-ranging protein-based therapeutics

• Develop flexible & portable device platforms for manufacturing multiple biologics with high purity, efficacy, and potency, at the point-of-care, in short timeframes (<24 hours), when specific needs arise

• Include end-to-end manufacturing chain (including downstream processing) within a microfluidics-based platform

• Focus on currently approved therapeutics by FDA (i.e. no drug discovery)

Column Tanks

1 2

Downstream

Affinity Chromatography

Sterile Media

Yeast Inoculum

Crude Product Hold Tanks

Upstream

1 2 1 2

Polishing Membrane #1 Tanks

Polishing Membrane #1


Waste Waste Waste


UF/DF Membrane

Waste

Concentration

Identity

Purity

pH, DO, T

Potency

Safety

Sterility

Analytics

Purified Product for Quality Testing

Multivariate

Model

OK for Release

Not OK

☐

Final

Product

On-line Reactor Control

Perfusion

Fill

Elute

Real-Time Process

Data

Integrated and Scalable Cyto-Technology

(InSCyT) biomanufacturing platform

19

Rationale for Pichia pastoris

as microbial host for biosimilar products

Advantages from a regulatory perspective

• Many products on market or in late-stage development

(including one Phase I target)

• Reduced risk for viral contamination in InSCyT process

• Human-like post-translational modifications (folding, glycosylation, etc.)

Technical benefits

• Genetically stable organism

• High density cultivation (culture volume >70% biomass)

• High yields of secreted proteins (up to ~15 g/L)

• Limited host cell protein (HCP) profile (eases burden on downstream)

• Amenable to lyophilization

20

Column Tanks

1 2

Downstream


Sterile Media

Yeast Inoculum


Upstream

1 2 1 2




Waste Waste Waste


UF/DF Membrane

Waste

Concentration

Identity

Purity

pH, DO, T

Potency

Safety

Sterility

Analytics


Multivariate

Model

OK for Release

Not OK

☐

Final

Product


Perfusion

Fill

Elute

Real-Time Process

Data



On-line quality

analytics

21

Recall Quality by Design approach

Process development involves:

• Define the target

product profile

• Identify the

critical quality

attributes (CQAs)

• Select an

appropriate

manufacturing

strategy

• Implement a

control strategy

22

Plant-wide control approach

• Characteristics of InSCyT

– Many connected unit operations

– Many discrete operations

– Multi-product plant

– Alignment with regulatory

requirements (e.g., design space)

• QbD approach adapted from chemical industry

– Employing systematic and modular design of plantwide

control strategies for production-scale manufacturing facilities

– Using numerical algorithms that can handle discrete operations

and multiple products

R. Lakerveld, B. Benyahia, R.D. Braatz, & P.I. Barton, Model-based design of a plant-wide control

strategy for a continuous pharmaceutical plant, AIChE Journal, 59, 3671-3685, 2013

Column Tanks 1 2

Downstream


Sterile Media

Yeast Inoculum


Upstream

1 2 1 2




Waste Waste Waste


UF/DF Membrane

Waste

Concentration

Identity

Purity

pH, DO, T

Potency

Safety

Sterility

Analytics


Multivariate

Model

OK for Release

Not OK

☐

Final Product


Perfusion

Fill

Elute

Real-Time Process

Data

23

Application to biologic drug production

• Build first-principles

dynamic models for

each unit operation (UO)

• Design control system

for each UO to meet

“local” material attributes

• Evaluate performance in

simulations and propose

design modifications as needed

• Implement and verify the control system for each UO

• Design and verify plantwide control system to ensure

that the CQAs are met 24



dynamic models for



for each UO to meet








Column Tanks

1 2

Downstream


Sterile Media

Yeast Inoculum


Upstream

1 2 1 2




Waste Waste Waste


UF/DF Membrane

Waste

Concentration

Identity

Purity

pH, DO, T

Potency

Safety

Sterility

Analytics


Multivariate

Model

OK for Release

Not OK

☐

Final

Product


Perfusion

Fill

Elute

Real-Time Process

Data



26

Local “UO” control for bioreactor unit operations

27

28

Local “UO” Control: Microscale controlled cell culture

Kevin S. Lee et al., Lab on a Chip, 11, 1730-1739 (2011)

OD

600 (

Cell C

on

cen

trati

on

)

POD 3

POD 2

POD 1

POD 0

Outgrowth Transition Production

Time, h

Reproducible microbioreactor cultivations

29

Kevin S. Lee et al., Lab on a Chip, 11, 1730-1739 (2011)

Local “UO” control: Microscale controlled cell culture

30



dynamic models for



for each UO to meet








Acknowledgments (Novartis-MIT)

Paul I. Barton

Bernhardt L. Trout

Allan T. Myerson

Klavs F. Jensen

Timothy F. Jamison

Richard Lakerveld (Delft)

Brahim Benyahia (Loughborough)

32

Acknowledgments (Bioman)

Chris Love

Rajeev Ram

Anthony Sinskey

Timothy Lu

Michael Strano

Jongyoon Han

James Leung

GK Raju

Stacy Springs

Paul Barone

Kerry Love

Philippe Mouriere

Nate Cosper

William Hancock

Lisa Bradbury

Ralf Kuriyel

Russell Newton

Susan Dexter

Steve Cramer

Pankaj Karande

33

umass lowell | umass lowell - systems technology in … a1 braatz lowell talk-v2... ·...

Documents