Development and Implementation of Continuous Manufacturing...
Transcript of Development and Implementation of Continuous Manufacturing...
Development and Implementation of Continuous Manufacturing Processes for API
Paul C. Collins, Ph.D., Senior DirectorSmall Molecule Design and DevelopmentEli Lilly and Company
Outline
♦ Introduction to CM API• Why do we do CM?• Why is it better for enhancing product quality?• Examples from Lilly CM API journey
♦ Future state• Possibilities• Things to ponder
What is Continuous Processing?
Continuous manufacturing (CM) is a method to manufacture, produce or process without interruption.
Constant material flow in and out of the process
While batch operations are dynamic processes (changing over time), continuous processes rely upon steady state operation (no change over time)
Smaller amounts of material “at risk” at any time is central to the concept
3
CM History in Lilly
2 Hybrid Processes in Manufacturing at
Kinsale
Continuous Unit Operation (Mfg)
IE2 SVC Facility, KinsaleUnder construction
Small Volume Continuous (SVC) Mànufacturing in Fumehood at Kinsale
Coil inside
1990’s 2017+201420132006-10
More Potent, Lower Volume Medicines
Platform & Technology
Development in R&D
Ozonolysis Quench
PumpPump
Tank 1 Tank 2
Hybrid CM Multiple Steps in CM
Realized Benefits of Continuous at Lilly
Enabling Chemistry-Extreme conditions-New options on old reactions-New SRs-Expanded toolbox
Safety
-Smaller
volumes
-↑ heat transfer
-No headspace
Quality-Natural QbDlinkage- PAT integration-Steady State control-Less at risk-↓Operational complexity
Going Green!-Less waste
-Lower energy
-Neat reactions
Cost Savings-Capital avoidance-New capabilities-Cytotoxic & HiPo capacity-↑Throughput
Early proofs of concept in Lilly Manufacturing
Project AAPI cost reduction of >50%
Robust Pd removalGreen chemistry reduces waste
Registration Stability120 kgs
Project BEliminated $20M spend on H2 bunker
Safer alternative to batchGreen chemistry
Registration Stability1800 kgs
Project CDevelopment time reduced
Tech transfer req’d half the timeGreen chemistry reduces waste
Pre-registration stability150 kgs
CM Opportunities
Pro
cess
Lilly’s Future Medicines are evolving towards more potent,
lower volume productsCurrently >70% post FHD portfolio has
projected volume of <1.5 MT/year
Forward Looking Strategy for CM
500 to 1000 mg dose 2 to 20 mg doses
Bus
ines
s
MINIMIZING VOLUMES including NEAT reactions.Containment of hazardous reagents3-15 kg/day
MAX THROUGHPUT FOR SVC
SCALEDEVELOPMENT ≡ MANUFACTURING
Leaner and faster tech transfers
Small Volume Continuous (SVC) Manufacturing Opportunities
CSTR 30 L
0.8 L PFR
300 Gal Batch reactor
Greater SECURITY OF SUPPLY due to shorter manufacturing lead times
REDUCED Processing & Turnaround TIMESDedicated/disposable equipment sets
Reduced Environmental ImpactLower volumes & number of isolations
Skids are part of a platform to support chemical unit operations in any product
• Modular to be combined into unit operations (e.g. CSTR/mixer skids combined for a counter-current extraction).
• Flexible and adaptable – simple skids with standard components (where possible)
• Plug into Distributed Control System (DCS)• Working with high end equipment e.g. gear
pumps & data management system.
SVC Facility for GMP API ManufactureFuture State
CSTR Skid
Plug Flow Reactor Skid
Feed System Skid
SVC Facility Virtual Tour
Reshaping our Manufacturing Footprint
MATERIALSBatch Definition SpecificationsMaterial Tracking / Genealogy incl. Deviation boundaries
EQUIPMENTReactors RTDCommercial skidsSingle use technology
PROCESS DESIGN including IMPURITY REJECTION
Rejection connected to unit operation
RISK ASSESSMENT
PRODUCT COLLECTION & MATERIAL DIVERSION
STATE of CONTROLincluding PROCESS MONITORING & DATA MANAGEMENTincl. Sampling | IPC | PAT | Offline testing | Modelling | MVA | Real-time analytics | Automation | Feedback Control
Product Control
Strategy*
Why CM is better for Product Quality -Control Strategy Elements
‘’Divert & Surge Strategies’’
Manufacturing process produces products of intended quality in a reproducible way
Failure modes
*Note: Elements not equally weighted
Control Strategy DiagramCase Study
Surge locations whereby material was isolated and provided complete decoupling of steps.
Step 1 Rxn Step 1 Extraction Step 1
DistillationStep 2 SNAR
Step 2 Crystallisation
Step 2 Filtration/Wash/
Dissolution
Step 3 Deprotection
Step 4Distillation
Step 4 Crystallisati
on
Nitrile SM, THFMeOH, AcOHHydrazine
DMSOWaterNa2CO3 DMSO
Pyrazine SMNEM
MeOHMTBE
Formic AcidWater
Lactic AcidWaterTHF
AcONa, Hydrazine, DeBoc, Water
MeOH, THFToluene, Water
MeOH, MTBE, DMSO, NEM*HClRegioisomersBis-PyrazinePyrazine SM & deg
HydrazoneDes-AminoNitrile SM
API*HClRegioisomersBis-PyrazineNEM*HCl API*HCl
CO2C4H8
Formic AcidWater
Formic Acid, Lactic Acid, Water, THF, t-Bu Amide, Hydrolysis imp., T-Bu Amine
API
API*HClt-Bu AmideHydrolysis Imp.t-Bu Amine
API*HClt-Bu AmideHydrolysis Imp.t-Bu Amine
MeOH
.
Surge after crystallization was a small CSTR with small tau, to dampen the 60 minute intermittent flow cycle time from the dissolve off filter.
Plug Flow Reactors (PFRs)Cleaner Reaction Profile | ‘Consumable Reactors’
Step 1 – 316L Stainless Steel PFR• 130 °C, 1 h V/q, 500 psi• Assay yields of 97–99%• Reduced hydrazine eq. versus
batch.• Faster reaction
Step 2 & Step 3 PFA PFRs• Step 2 - SnAR: ~85 °C at 3 h V/q• Step 3 - Deboc: 20–50 °C at 4 h V/q
− Mild conditions for deboc reaction. Gas-LiqPFR used to allow for gas sweeping & volatile removal
− Single use equipment
Counter-Current Extraction in Step 1Efficiently removed Hydrazine from 6000 ppm to <2 ppm
Counter-Current Extractions & DistillationImpurity Rejection
Distillation in Step 3• Distillation rate could not be replicated in
batch mode! Batch distillation times were longer which led to elevated impurity formation.
• Effective in formic acid removal to <1 equiv.
Continuous Crystallization Impurity Rejection & Safety
Automated Parallel Filters for filtration, washing, drying, and dissolution - “Isolation-Free Isolation”• Properly sized filters
achieve very consistent filtration rates and residual solvent levels, which is important for subsequent process step.
• No handling of cytotoxic product with OEL = 1 µg/m3
Mixed suspension mixed product removal (MSMPR) vessels• Crystallisation represented the
most significant element of the impurity control strategy
• The purity of the crystallized solids was in excess of 99.8 area%.
Online UPLC used in Steps 1, 2, 3:
Process MonitoringPAT data for Process Monitoring
Step 1
‘’Step 2’’
Step 2 Step 3
Offline HPLC confirmation for Step 3
PATrol Unit
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
95
95.5
96
96.5
97
97.5
98
98.5
99
99.5
100
Days of Sampling4 8 12 16 20 24
Product Starting Material Cis Isomer
Are
a %
Pro
duct
Area %
Starting M
aterial, Cis
Isomer
In-process Monitoring: Manufacturing
Distillation End Point (Step 3)Parametric Analysis for Process Monitoring
• First distillation performance evaluated offline by 1H NMR spectroscopy. Parametric analysis used thereafter.• Distillation performance assessed by distillate temperature (temperature, pressure, time, and mass flow)• Confidence in distillation performance and removal of formic acid.
Step 3 Distillation Temperature
Process Monitoring
Monitoring models for parameter and PAT analysis can provide traffic light signals to operators/ tech support. Multivariate capability. Feedback loop controls.
MODELS
DCSincl. ALARMS
MANUAL DATA RECORDING
SOPH
ISTI
CAT
ION
Periodic checks by operators/tech support. Data can also be entered manually into a control chart system. Resource and time intensive –not desirable!
Real time automated data analysis. No MSPC and relies on set limits and simple calculations.
PAT central to Process Monitoring
Step 2 EventIncrease in Key Impurity
Day 6 of 7, the Step 2 SNAr reaction performance changed as per online UPLC.
Area% Impurity over time
STARTING MATERIAL
PRODUCT
KEY IMPURITY Area
%
Time (h)
Root Cause:Reagent lost causing HClsalt formation that could not be rejected downstream.
Material TrackingDeviation Management
Drum 12 Collection08:35 to 20:30
12 h
Drum 13 Collection20:30 to 09:55
13.5 h
Drum 14 Collection09:55 to 00:01
14 h
Drum 15 collection00:01 to 21:16
21.25 h
τ ~ 3 h
τ ~ 2 x 1h
Cycle~ 2 x 1h
Surge time adjustable0 to 12 h
τ ~ 4h
PATr
ol
22:10Rxn stopped
02:10 Base equiv. corrected & rxn restarted
02:19Atypical
material starts collecting
19:25Start rxn &
resume typical material.
15:23Pause upstream unit op
to pump down CSTR surge to almost empty.*
Trend detected by PAT and not by parametric data.*If the level in the CSTR surge had not been pumped down then Drum #15 would have also had elevated impurity.
Divert StationsDivert & Surge Strategies for SVC
“The extent of material to be isolated and rejected depends on the duration, frequency, and severity of the disturbance and the mixing patterns of the system” FDA J. Pharma Innov. (2015) 10:191-199
DIVERT at source
Non-Conforming Material
SURGE Collect downstream
• Process Risk Assessment is key for determining appropriate strategy.• Process design, equipment robustness will be influencing factors as well as
the magnitude and duration of the disturbance.• Process monitoring, PAT, Modeling, key for managing these scenarios
effectively.
Strategy at Lilly
Data ManagementThe journey…
• PAT will be a key component of manufacturing control systems.
• Identified synTQ software platform to develop, deploy and manage data.
• PAT will communicate directly to the DCS.• Future potential: Advanced control e.g. MVA, MSPC,
Computational models for decision making, live batch tracking.
2 PATrols collect samples every
hour
Manual check by Chemist
Results exported and trended (manual)
Trended data shared with team
Repeat for duration of campaign2014 Proof of Concept:
Future CM Processes
• Approach not sustainable long term
Current control strategy status
♦ Control Strategy is a combination of Level 1 and 2 controls today
Lilly manufacturing:• 2 registration stability campaigns in 2013• 2 clinical trial campaigns in 2014• 1 validation campaign in 2015• SVC on line in 2017
External manufacturing:• 2 clinical trial campaigns in 2013• 2 validation campaigns in 2015• 3 fully continuous multi-step process in SVC format
Regulatory activity:• 1 Type C meeting with FDA in 2013• Several End of Phase 2 (EOP2) with FDA meetings since 2013• 1 NDA submission and 1 planned NDA submission with FDA for 2015• HPRA visit in 2016
Accomplishments to Date
Possible Future
♦ Demonstrated on-line monitoring gives much better view of “state of control” – robustness and productivity• In process testing done offline in traditional labs not
needed (already possible now)♦ A set of approved unit operation modules could be
deployed in an approved manufacturing site in only days/weeks – supply chain flexibility and robustness• Digitized workflows check the CM API process train is
compliant for manufacture. • An example – deployment of a second manufacturing
train in a second location would be analogous to a system suitability test to use in a second location.
• “Scale up” exits the pharmaceutical vernacular…
Possible Future
♦ Miniaturization of API manufacture allows for dedicated processing equipment – cross contamination concerns eliminated
♦ Deviations and associated cost will be reduced over ten fold – robustness and control
♦ Traditional inspection approaches could be done digitally – overall pharma oversight becomes less costly, yet more rigorous• Continuous process control data• Continuous in-process analytical data
Missing elements and gaps to full implementation
♦ Regulatory• Approaches to control strategy, technology transfer
(including site to site), validation, filing parameters (such as design space) were created on a batch paradigm
• CM API changes the meaning associated with each of these
– Equipment design and automated control more important than range definition/procedural control
– Scale and equipment differences associated with traditional batch development and manufacture are largely eliminated
• Confusion and energy barrier associated with how to fit square pegs into round holes will keep some companies out of CM API
• Geographic flexibility, response to supply chain disruptions are very possible with miniaturized, modular CM API – but regulatory flexibility does not currently match
CM API – A few questions we might ponder…♦ What is the goal of CM API from the regulatory
perspective?♦ Why would you have a control strategy if you
could actually control?♦ What was the original reason for things like
validation and why do we need them in CM?
♦ Our regulatory perspectives need to evolve to fit the capabilities now available for full implementation and utilization
Backups
Batch and Continuous Comparison
Bridge = Continuous Operation
Car Ferry = Batch Operation
3/15/2017
Batch vs Continuous Processes
3/15/2017
30 L CSTR
2 L PFR
Vessels and auxiliary equipment are smaller than batch vessels for the same throughput!
1200L Batch Tank