Detect and Control Variation

An investigation into the sources of variation
Karen S Ginsbury
PCI Pharmaceutical Consulting Israel Ltd
For IVT Valdiation Week Philadelphia
October 2015
Of Deviations
"Every defect is a treasure, if the
company can uncover its cause
and work to prevent it across the
corporation”
- Kilchiro Toyoda, founder of Toyota
Oops… A treasure?
Or Poorly Controlled Variation
• Toyota has two suppliers for the accelerator
pedal: CTS and Japanese supplier Denso
• Toyota spokesman Mike Michaels suggested
that Denso parts are not implicated, but also that
Toyota will not be able to single-source
replacements from Denso
• That suggests that CTS's and Denso's designs
for the same part are slightly different and can't
be used interchangeably
ICH Q10 - Objectives (of a )
Pharmaceutical Quality System
1.5.2 Establish and Maintain a State of Control
• To develop and use effective monitoring and
control systems for process performance and
product quality, thereby providing assurance of
continued suitability and capability of
processes
• Quality risk management can be useful in
identifying the monitoring and control systems
Cause and Effect
• All manufacturing processes are designed
to follow a series of steps which are
interlinked and interdependent
(cause and effect)
• Variation exists in all processes
• Understanding and reducing variation will
always improve your process
W Edwards Deming
1900 - 1993
• “We have learned to live in a world of
mistakes and defective products as if
they were necessary to life
• If I had to reduce my message for
management to just a few words, I’d
say it all had to do with reducing
variation”
What is variation ?
• Write down a definition
• Think about what CAUSES variation
What is variation ?
• What did you write?
What did you write?
• Inconsistencies – inconsistent procedures/
quality of raw materials / attributes of raw
materials are not reproducible (change from
batch to batch / from supplier to supplier)
• Differences in results (outcome) – differences
in the inputs: man, machines, materials,
methods, measurements, miscellaneous or
management or mother nature…other
What did you write?
• The differences in the outcome when you
try to produce the same thing or an
identical thing
• Deviation of the outcome from a
specification…/ control limits / or a or the
requirement(s)
• Slight deviations from an (original)
template or from intentions
Quality Target Product Profile
• A quality target product profile is a
prospective and dynamic summary of the
quality characteristics of a drug product
that ideally will be achieved to ensure that
the desired quality, and hence the safety
and efficacy, of a drug product is realised
• The target product profile forms the basis
of design for the development of the
product
ICH Q11 Drug Substance
Quality Link to Drug Product
• The Quality Target Product Profile
(QTPP), potential CQAs of the drug
product and previous experience from
related products can help identify potential
CQAs of the drug substance
• Knowledge and understanding of the
CQAs will evolve during the course of
development [and over the commercial
part of the lifecycle of the product]
Typical QTPP
Quality Attribute
Target
Comments
Target Population
Diabetics
Home use
Route of Administration
Parenteral, I/M
Probably require lyophilized form for stability
Dosage Form
Injectable
Strength
100 – 200 IU/ml
To be studied may change considerably
Packaging
Pre-filled Syringe
Two compartments to allow mixing immediately prior to
use
Stability
2 years at room temperature
Unlikely to be achievable
Consider 18 months at 2 - 8°C
Pharmacokinetics
Immediate release
Microbiology
Sterile, Endotoxin free
Appearance
White to off-white powder and clear
solvent
Assay
80 – 110% of label claim
Impurities
Individual NMT 0.1%
Total: NMT 0.5%
Version No: 01
Approved by: _______________
Issue Date: 010111
Sales &Marketing
____________
Quality
____________ ____________
Production
Engineering
_______________
R&D
Now write instructions for a
gowning SOP
• You can do it with your neighbor – BUT
THINK about the process before you start
writing the instructions and let us know
how you developed the instructions
• It is for tablet manufacturing
Gowning
•
•
•
•
•
Boil the water
Add the teabag
Let it seep
Remove the teabag
Add sugar or honey
Here is George Orwell’s SOP
for making a cup of tea
Read on… Reducing Variation
Coefficient of Variation
• In probability theory and statistics the
coefficient of variation (CV) is a
normalized measure of dispersion of a
probability distribution
• It is defined as the ratio of the standard
deviation to the mean
Variability
• The amount of spread in collected data / a
group of scores
• Usually defined in statistical terms
(e.g. standard deviation of the mean value
or process control limits / process
capability)
• Variability in a process causes variation of
the output
19
Case Study:
• The batch record for manufacture of crude
API, step 3.6 is written as follows
Do NOT exceed three hours
reflux!!!"
• Your comments please?
Common Cause or Special
Cause?
• No report was available
• The only supporting work, an experiment:
“XXX formation under drastic conditions”
• The only “data” recorded
• yield 3.77g (75%)
• HPLC analysis 26.06mg in 25ml
• Assay 71%
• Yields reduced
• Quality the same
Quality is…
• Meeting all the requirements all the time
• Which means – you have to DEFINE the
requirements
• All the requirements means:
– The specifications
– Customer requirements
– Regulatory requirements
– EHS requirements
– Etc.
How do we determine the
Requirements
• We ask a lot of questions
• We get answers from subject matter
experts – people who are expert in the
particular requirement we are trying to
define
Uncertainty = Risk
• No indication of what constitutes a normal yield
• No indication as to normal HPLC and assay
results
• No indication as to how quality was assessed
• In fact, the instruction arose from a FAILED
BATCH which was ASSUMED to have been
caused by over refluxing
• = SPECIAL CAUSE
Variation
How did we go from here…
To Here
ICH Q11 Drug Substance
Quality Link to Drug Product
• The intended quality of the drug substance
should be determined through
consideration of its use in the drug product
as well as from knowledge and
understanding of its physical, chemical,
biological, and microbiological properties
or characteristics, which can influence the
development of the drug product (e.g., the
solubility of the drug substance can affect
the choice of dosage form)
The Relationship
Uncertainty
Risk
Variability
29
Product Control Strategy (ICH Q10)
• A planned set of controls, derived from current
product and process understanding, that
assures process performance and product
quality
• The controls can include parameters and
attributes related to drug substance and drug
product materials and components, facility and
equipment operating conditions, in-process
controls, finished product specifications, and the
associated methods and frequency of monitoring
and control
Critical Process Parameters vs Critical
Quality Attributes
Critical Process Parameter Critical Quality Attribute
• A physical, chemical,
• A measured variable that has
biological, or microbiological
a known effect upon a product
property or characteristic
quality attribute
• that should be within an
• A process parameter that
appropriate limit, range, or
must be controlled within a
distribution
specified range to assure
• to ensure the desired product
product quality
quality
31
Output (Y) is a CQA affected by the
Variable Inputs (X)
I Chart
115
UCL=111.55
Individual Value
110
105
_
X=99.63
100
95
90
LCL=87.71
60
62
64
66
68
70
72
Observation
74
76
78
80
People
I Chart
115
UCL=112.65
Inputs to the process
control variability
of the Output
110
Individual Value
105
100
_
X=97.94
95
90
85
LCL=83.23
80
40
44
46
48
50
52
Observation
54
56
58
60
Equipment
I Chart
115
UCL=112.65
110
Individual Value
105
100
_
X=97.94
95
90
85
LCL=83.23
y = ƒ(x)
I Chart
80
40
42
44
46
48
50
52
Observation
54
56
58
60
115
Measurement
UCL=116.68
Individual Value
115
110
105
_
X=102.37
100
95
UCL=114.17
110
y
Individual Value
I Chart
120
105
_
X=99.95
100
95
90
LCL=88.05
20
22
24
26
28
30
32
Observation
34
36
38
40
90
Process
11
21
31
41
51
61
Observation
71
81
91
UCL=111.55
110
Individual Value
LCL=85.72
85
1
I Chart
115
105
_
X=99.63
100
95
90
OUTPUT
LCL=87.71
60
62
64
66
68
70
72
Observation
74
76
78
80
Materials
I Chart
UCL=111.17
110
105
Individual Value
I
N
P
U
T
S
(X)
42
_
X=98.76
100
95
90
LCL=86.35
85
80
82
84
86
88
90
92
Observation
94
96
98
100
Environment
Adapted from slide by Moheb Naser, FDA
32
Learning about the process
• Means understanding the process
• Which means performing small scale
studies on different portions of the process
so as to learn which parameters are
critical
• So-called “failed” experiments are often
those that teach the most about the
product and process
Classic / “Empiric” Development
and Change Control
QbD / Managed Changes in R&D
PDCA leverage QSR
Control Strategy
•
•
•
•
•
Analytical test method
Specification
Manufacturing instructions
Qualification and Validation
Embed the good
and
• CHANGE the bad in a controlled manner
What is Knowledge
• Knowledge comes from Information and
experience
• Where does information come from?
• Information comes from data
• Convert data to information by analysis
• Convert information to knowledge by
applying the information to our process
and product (and then communication and
training and documentation)
Transferring Knowledge to Stakeholders
• Don’t be miserly with information
• Identify key players:
– Production employees / operators
– Analytical staff
– Quality personnel
(will be monitoring the process)
• Walk them through the development report
and tech transfer documentation
Capture Information in Batch Record
• Emphasize critical data
• Use the design space to ensure that the process
is rugged
• The more critical the parameter the smaller the
range allowed
(e.g. viral inactivation could be 600.5C)
• The less critical, the wider the range
(e.g. heating to get a chemical into solution
could be “heat to a maximum of 45 C for
10 - 20 minutes)
Yield – is it a Quality Metric?
• Yes it is
• Yield tells you about the performance of
your process
Yield
• If your yield is out of control
• Your process is out of control
Yield
• The amount obtained from an undertaking
• To give forth from its own substance by a
natural process or in return for cultivation or
labour
• To produce as profit
• Capacity for producing - especially with
qualifying word referring to the amount or
quality of the produce
(Source: Shorter Oxford English Dictionary 1964 edition)
43
Types of Yield
• Theoretical:
– The quantity that would be produced at any
phase in production, based on the quantity of
raw materials to be used, in the absence of any
loss or error in actual production
• Expected:
– The quantity or percentage of the theoretical
yield anticipated based on previous
manufacturing data
Yield Calculations
• Actual Yield
– The quantity that is actually produced at any
phase of manufacture
(based on material actually manufactured)
• Reconciled Yield
– Takes into consideration samples, rejected
units and material that is in hand and can be
weighed or measured
• Unknown Losses:
– can’t be accounted for
– although the company is paying for them
Special Cause, Common Cause
and Tampering
Visual Inspection
What is going on?
Visual Inspection Yields 2008
106
96
Yield %
Yield
Average
UCL
LCL
86
76
Batch Number
TAMPERING
• The operator gets feedback from the
supervisor
• “The number of rejects in visual inspection
from particles is very high…again”
• The operator adjusts the capping machine
• The problem is resolved….
• Until the next time
Machinery Replaced
Visual Inspection Yields 2009
96
Yield %
94
Yield
Average
UCL
LCL
92
90
88
86
1
3
5
7
9
11
13
Batch Number
15
17
19
Common Cause (Process) –
Case Study
• API – X
• Hundreds of batches manufactured over
many years
• Analysis of yields shows….what?
Process Yields 2009
109
104
Yield %
Yield
LCL
99
UCL
LSL
94
USL
89
84
1
4
7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67
Batch Number
Out of Control Process?
•
•
•
•
To tamper or not to tamper?
Investigate?
…and discover…
Every x batches the operators collect material
that has amassed on the filter, re-dissolve it in
process solvent and throw it back into the
process
• A trend is seen – after throwing it back in – a
peak in yield followed by a drop and then
several points show…what? Apparent increase
but not necessarily related to same common
cause
Process Improvement
• Place a change request to deal with filter blockage
• Move the location of the filter such that process flow
has an additional solvent wash BEFORE the filter
• Process flow moves through filter without build-up
on filter and blockage
• Operators don’t have to stop process, remove
material
re-dissolve and add back to process resulting in
uneven yields
• Process more Efficient
Problem Solved? Continual Improvement…
Corrected Process Yields 2009
Yield %
109
104
Yield
99
LCL
UCL
94
LSL
USL
89
84
1
4
7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61
Batch Number
Where to go
•
•
•
•
•
Investigate remaining trends
Correct root cause
Reduce variation
Increase process understanding
Reduce risk of batch failure – by
increasing process capability
Ishikawa (Fishbone)
Cause and Effect
What is the Problem
•
•
•
•
•
Variation in resin output
Why?
When (did it happen) ?
Who (was involved in the manufacture)?
What (happened in the previous step)?
Why ?
• "A relentless barrage of 'why’s' is the best
way to prepare your mind to pierce the
clouded veil of thinking caused by the
status quo. Use it often."
• ~ Shigeo Shingo
Cause and Effect
Process Outcome as a tool for
improvement
• Actual yield = amount available for delivery to
customer
• Reconciled yield:
•
•
•
•
The white beads: can their % be increased?
Was it improved
Better / worse than previous data period
Where were the most losses
• Do we have red beads in the incoming materials.....
• Is there ageing machinery / malfunctions?
• Is it common cause or special cause
• AVOID tampering
60
Process Validation Guide
• Control all operations
• Validate critical manufacturing operations
• Formal design / development prior to
validation
• Change / continuous improvement
encouraged
• Fuller process understanding leads
to improved problem solving
In Conclusion…
• We don’t have to live with excessive
variation
• Variability causes uncertainty
• Uncertainty increases the risk of failure
• Process yields are a tool for quality
improvement
• Don’t shoot the messenger
• Find and fix the common cause
62/27
Thank You For
Your Attention
Any Questions...
Write down a definition of
variation
• A deviation from requirements?
• A deviation outside the approved limits we
would hope would not be part of the inherent
variation of a process and we surely have a
responsibility to make our processes
sufficiently robust to PREVENT them from
varying OUTSIDE the approved specification
limits and any regulator is entitled to cite you
if your process regularly does go outside the
specifications
Variation – we are still trying to
define it
• Natural randomness within a sample set?
Which represents the natural randomness
within the entire population
• Range of values (that might be obtained)
for a given parameter or quality attribute
over repeated measurements or data
points – observed routinely as the process
is run time and time again
Variation - Definition
• An instance of change; the rate or
magnitude of change (the process of
varying or being varied)
• An activity that varies from a norm or
standard
• (a repetition of a musical theme in which it
is modified or embellished)
• Version: something a little different from
others of the same type