Participate in the interactive webinar now: http://bit.ly/CPVWebinar Product life cycle consists of 3 phases: Process Design, Process Performance Qualification and the last and the lengthiest Continued Process Verification (CPV). As more and more biomanufacturing processes enter commercial phases, the critical need to understand how to efficiently perform CPV programs arises. Explore our webinar library: www.emdmillipore.com/webinars

1. The life science business of Merck KGaA, Darmstadt,
Germany operates as MilliporeSigma in the U.S. and
Canada.
How to develop a regulatory-compliant
Continued Process Verification (CPV)
and Process Monitoring of bioprocesses
Anshuman Bansal
December 5, 2019

2. The life science business
of Merck KGaA, Darmstadt,
Germany operates as
MilliporeSigma in the U.S.
and Canada

3. Agenda
1 Why
2
3
Monitoring Strategy
4
Statistical Process Control (CPV Basics)
Examples5
Other Visual Monitoring/Investigational Tools

4. WHY

5. Bioprocess data complexity
› Multi unit operations involving splits and combinations
› Uncertainty due to biological processes (inherent variability)
› Dynamic relationships shifted in time
› Changing correlation structures
CC1-Lot1 P1-LotA
P2-LotA F1-Lot1 F2-Lot1
Cell Culture Purification Filtration
CC1-Lot2
CC1-Lot3
CC2-Lot1
CC2-Lot2
P1-LotB
P1-LotC
P2-LotB F1-Lot2 F2-Lot2
*Reference: A. Bansal, J. Hans, A. Rathore; Knowledge Management and Process Monitoring of Pharmaceutical Processes in the Quality by Design Paradigm; Measurement,
Monitoring, Modelling and Control of Bioprocesses; Advances in Biochemical Engineering/Biotechnology Volume 132, 2013, pp 217-247

6. Butterfly Effect: Unknown variability at process stages accumulates into product inconsistency
Seed Train
Cell Culture
Clarification
Capture
Purification
ProbabilityofError
Process Step/time
Formulation
Heterogeneity
of biotech
products
Variability
in process
and
analytical
methods
Variability in
quality of
raw
materials
High variability in
product quality
Complexities
associated with
biotech processes
and products
Limited
understanding
of biotech
processes

7. Regulatory Expectations (FDA, EMA)
Quality and regulatory expectations for
process monitoring
Pharmaceutical companies should plan and
execute a system for the monitoring of
process performance and product quality to
ensure a state of control is maintained
-ICH Q10
An ongoing program to collect and analyze product and process
data that relate to product quality must be established. The data
collected should include relevant process trends and quality of
incoming materials or components, in-process material, and
finished products.
-FDA Process Validation Guidance, Jan 2011

8. Regulatory Expectations (FDA)
FDA Expectations for CPV
Must have a system for detecting unplanned departures from the process
• Evaluate the performance of the process
• Identify Problems
• Determine if corrective action is necessary
• Anticipate and Prevent problems to ensure control
An ongoing program for collecting and analysing product and process data that relate
to product quality
• Procedures for data collection and trending
• Data collected should verify the quality attributes
• Intra-batch and inter-batch variation
• Data should be collected to evaluate process stability and capability
• Data should be statistically trended
• It is recommended that a statistician or person with adequate statistical training develop the data collection plans
and methods for analysis
Source: FDA Guidance for Industry: Process Validation: General Principles and Practices (Jan 2011)

9. Examples of some of the 483s issued by FDA

10. Monitoring Strategy
What to routinely monitor and what to
archive for investigational analysis?

12. Bench-top
Instrument Data
Quality Control
Results
Batch Record
Data
Real-time
Sensor Data
% DO
pH
Temperature
%CO2
N2
RPM
Air FLow
O2 Flow
CO2 Flow
Media Quantity
Media Blend Ratio
Harvest Age
Harvest Volume
Feed Age
Temperature shift Age
Feed Flow
Feed Quantity
Generation No.
Hold Times
Integrated Viable Cell Density
Inoculation Density
Titer
RVLP
Mycoplasma
MMV
Bioburden
Endotoxin
Cell Count
%Viability
pH(offline)
Glucose
Glutamate
Glutamine
Lactate
Ammonia
HCO3
Osmolality
Galactose
pCO2
pO2
Sodium
Potassium
Example: Mammalian Cell Culture

13. Upstream Process (Production Bioreactor)
Cell growth
Culture Environment
Product Formation
Contamination Control
Cell Density
Viabilities
Harvest (Viab, Age)
CGN
Glucose
Lactate
~Oxygen Flow
~DO
~pH
~Temp
Titer (Daily, Harvest)
CGN
Glycans
IVCC
Peak (VCD, Viab, Age)
Harvest (VCD)
GN
Feed (VCD, Viab, Age)
TempShift (VCD, Viab, Age)
Media (pH, Osmo, Endo)
~Air Flow
~Off Gas (OUR)
~CO2 Flow
Metabolites trend
Specific Productivity
Passage No
Events (Feed, Antifoam, Base etc)
Media Lots (Culture, Feed)
Total Base
Total Antifoam
Total Feed
Total Flow (O2, CO2, Air)
Specific metabolite consumption
pH Probe (ID, cycles)
DO Probe (ID, cycles)
CO2 Probe (ID, cycles)
~Agitation Speed
~Reactor Volume
~Overlay (Air flow)
~Pressure
Sterile Filters
Media Hold
~SIP trends (temp, pr)
CIP pre/post (pH, Conductivity)
Deviations, Root Cause, Change Control, MFR, Experiment ID,
Equipment ID, Process Time, Change Over
The Elements InvestigationSecondary MonitoringPrimary Monitoring
Example: Mammalian Cell Culture

14. Routinely Monitor
% DO
pH
Temperature
Harvest Age
Harvest Volume
Integrated Viable Cell Density
Titer
Bioburden
Endotoxin
Cell Count
%Viability
pH(offline)
Glucose
Lactate
Osmolality
Collect and Archive
%CO2
N2
RPM
Air FLow
O2 Flow
CO2 Flow
Feed Age
Temperature shift Age
Feed Flow
Feed Quantity
Generation No.
Hold Times
Inoculation Density
RVLP
Mycoplasma
Glutamate
Glutamine
Ammonia
HCO3
Galactose
pCO2
pO2
Sodium
Potassium
Media Quantity
Media Blend Ratio
Example: Mammalian Cell Culture

15. Poll question 1

16. Statistical
process
control (SPC)

17. Parameter Type Abbreviation Description Routine Monitoring
Critical Process
Parameter
CPP
A performance or input parameter that
directly impacts product identity, purity,
quality or safety.
Must
Key Process
Parameter
KPP
A performance or input parameter that
directly impacts CCPs or used to
measure the consistency of the process
step
Must
Monitored
Parameter
MP
A performance or input parameter that
may or may not impact KPPs and is used
to measure the consistency of the
process step or routinely trended for
troubleshooting purposes
Not All, Case By Case
Basis
What to SPC?

18. Limit Name Abbreviations Description Limits Source
Applicable to
Parameter type
Specification
Limits
USL, LSL
These limits are defined based on process
characterization limits. Any excursion from
these limits will cause OOS and batch
rejection.
Process
Characterization,
Process
Development
CPP
Action Limits UAL, LAL
These limits are process validation ranges.
Any excursion from these limits will cause
major process deviation or discrepancy.
Process Validation CPP, KPP
Alert Limits
or
Statistical
Control Limits
UCL, LCL
These are monitoring ranges derived from
historical runs for out of trend detection and
measurement of process consistency.
Statistical: Process
History >15
commercial batches
CPP, KPP, MP
Target CL
The target (or centerline) is derived again
from historical runs as a measure to keep
the process consistent and proactively alert
if process is deviating from set target.
Statistical: Process
History >15
commercial batches
All
What SPC limits to apply?

19. How to estimate statistical control limits?
Distribution Sample Graph Description and Examples Typical Limits applied
Normal
Most of the process parameters will
follow this distribution of a
normal/Gaussian bell shaped curve.
Non-Normal
(Beta or
Gamma)
Some parameters will not follow a normal
distribution pattern and follow a skewed
distribution. For example most of the data
related to process impurities will be
skewed towards the lower bound
(approaching a value of 0). Some
parameters like cell viabilities would be
skewed towards the upper bound
(approaching a value of 100%).
CL Average
UCL Average + 3 SD
LCL Average - 3 SD
CL Median
UCL 99.865th Percentile
LCL 0.135th Percentile

20. How to identify a trend? (Example : Nelsons’ Rules)Table 6 [2]
Rule Description Chart Example Problem Indicated
Rule 1
One point is more
than 3 standard
deviations from the
mean.
One sample (two shown
in this case) is grossly
out of control.
Rule 2
Nine (or more)
points in a row are
on the same side of
the mean.
Some
prolonged bias exists.
Rule 3
Six (or more) points
in a row are
continually
increasing (or
decreasing).
A trend exists.
Rule 4
Fourteen (or more)
points in a row
alternate in
direction,
increasing then
decreasing.
This much oscillation is
beyond noise.
This is directional and
the position of the mean
and size of the standard
deviation do not affect
this rule.

21. Determining Process Capability and Process Performance
A. Estimating Process Capability for Normally Distributed Data
Process capability (Ppk, Cpk) for a normally distributed monitoring process parameter will be calculated using the following:
Ppk = 𝒎𝒊𝒏{
𝑼𝑺𝑳−𝑨𝒗𝒈
𝟑𝝈
,
𝑨𝒗𝒈−𝑳𝑺𝑳
𝟑𝝈
}, Cpk = 𝒎𝒊𝒏{
𝑼𝑺𝑳−𝑨𝒗𝒈
𝟑𝝈 𝑴𝑹
,
𝑨𝒗𝒈−𝑳𝑺𝑳
𝟑𝝈 𝑴𝑹
}
Where
USL = Upper Specification Limit (for CPP) or Upper Action Limit (for KPP)
LSL = Lower Specification Limit (for CPP) or Upper Action Limit (for KPP)
Avg = Average or mean of the population under analysis
𝜎 = Standard deviation of the population under analysis
𝜎 𝑀𝑅 = Moving Range Standard Deviation
B. Estimating Process Capability for Non-normally Distributed Data
Since the average and standard deviations will not represent the non-normally distributed data correctly, process capability (Cpk) cannot be estimated
for non-normal data. Instead Process performance (Ppk) will be evaluated based on all of data points in terms of percentile ranges. For a non-normally
distributed monitoring process parameter Ppk will be calculated using the following:
Ppk = 𝒎𝒊𝒏{
𝑼𝑺𝑳−𝑿 𝟎.𝟓𝟎
𝑿 𝟎.𝟗𝟗𝟖𝟔𝟓− 𝑿 𝟎.𝟓𝟎
,
𝑿 𝟎.𝟓𝟎−𝑳𝑺𝑳
𝑿 𝟎.𝟓𝟎−𝑿 𝟎.𝟎𝟎𝟏𝟑𝟓
}
Where
USL = Upper Specification Limit (for CPP) or Upper Action Limit (for KPP)
LSL = Lower Specification Limit (for CPP) or Upper Action Limit (for KPP)
X0.50 = Median of the population under analysis
X0.99865 = 99.865th Percentile of the population under analysis
X0.00135 = 0.135th Percentile of the population under analysis

22. Significance of Process Capability
Ideally : (USL-LSL) >> 6σ

23. Goal of CPV
Process Evolution with time

24. ControlLimits
Specificationor
ActionLimits
A Process Control Chart
2
3
3
2
Process Capability
Trend rules violation
Centerline

25. Setting up a SPC based monitoring program
Identify CPP, KPP and MP
from performance and
operating parameters
from process
characterization/validatio
n documents
1
Evaluate the root cause
of each rule violation and
is impact on product and
process. Initiate
appropriate CAPA if
needed.
8
Publish quarterly process
summary reports (can
feed APRs). Will have
recommendations for
process improvements
and limit updates
9
Identify parameters that need
to be trended periodically
2 Get validation ranges and
specification limits (if any) for
these parameters respectively
3 Establish a technique for
statistical control limits and
frequency to update the
control limits
4
Monitor parameters against
set limits
7 Publish an in-process control
and monitoring document for
each commercial process
6
Establish trending rules
(Nelson Rules)
5
Update control limits in the
system upon every limit
update via a change control
11Update
Control Limit
10

26. CPV related procedures and documentation
.
CPVDocumentation
1
2
3

27. Statistical Process Control – SPC Phase
SPC Phase
Gather Data for
future
trending.
Monitor only
with action or
spec limits.
Set Alert limits
(UCL, LCL and
Target) based
on stats of
gathered
history (and
excluding any
special cause
outliers).
Setting Alert Limits
After setting Alert Limits (UCL, LCL and Target)
from data gathering phase, routine process
monitoring is done in this SPC phase and all future
batches are compared against these set limits.
Preliminary Process Monitoring – PPM Phase
(Data gathering Phase)
How it shows up in real life
PPM Phase

28. Poll question 2

29. Other Process
Monitoring and
Troubleshooting Aids
Investigation Analysis Toolset

30. Timelines
Readymade
process
execution
times
Lot Genealogy
On the fly generate
full genealogy from
vial (cell line) to
vial (drug product)
Profiles
Overlay
batch
profiles
Correlations
Find
correlations
across
Parameters
SQC
Control
charting for
batch trends
and six sigma
analysis
Multivariate
Understand
interactions and
discover new
correlations in
ever dynamic
relationships
Compare groups
Compare groups of
data within batch or
across batches to
examine differences
Records
Access to raw
execution
records data
for context
and reference
When
What
How
Why
Is there
precedence?
What’s
the
benchmark?
How far
from
target?
TheQuestions
TheAnswers
Investigation Toolset

31. SS0101
Spinner Stage
SF0101
Seed Fermentation
PF0101
Fermentation
DS0101-1
HIC-Chromatography
SF0102
Seed Fermentation
DS0101-2
HIC-Chromatography
P0101
BDS
SS0102
Spinner Stage
RM0301
RM0302
RM0303
RM0102
RM0101
RM0103
RM0201
RM0102
RM0101
RM0101
RM0101
RM0401
RM0402
RM0501
RM0502
RM0601
RM0602
RM0603
RM0701
RM0702
RM0703
Lot Genealogy

32. Intra-batch variations

33. Intra-batch variations

34. Correlations: Upstream vs Downstream

35. Grouping batches for comparability

36. Process equipment differences

37. Example stories
How the toolset is applied for
troubleshooting

38. YieldVariability
Note: All Charts except indicated were generated in ProcessPad

39. Roundinganddistribution
Analyst Used JMP
for this analysis
* Analyst Used JMP
for this analysis
Note: All Charts except indicated were generated in ProcessPad

40. InstrumentMaintenance
Note: All Charts except indicated were generated in ProcessPad

41. Poll question 3

42. T
Thank you!

43. The vibrant M, and Millipore are trademarks of Merck KGaA, Darmstadt, Germany or its affiliates. All other trademarks are the property of their respective
owners. Detailed information on trademarks is available via publicly accessible resources.
© 2019 Merck KGaA, Darmstadt, Germany and/or its affiliates. All Rights Reserved.