According to 21 CFR, Part 820, medical device manufacturers are required to validate as well as monitor and control parameters for their processes. The guideline on Quality Management Systems does not specify how this is accomplished; only that “a process is established that can consistently conform to requirements” and “studies are conducted demonstrating” this. Thorough process development, optimization and control using appropriate statistical methods and tools is recommended for demonstrating that your process is both stable and capable. This session will demonstrate ways to efficiently and effectively apply recommended statistical methods and tools to process validation—with no statistical expertise needed. Using realistic process data, participants will learn how to apply tools, interpret results and draw meaningful conclusions throughout Installation Qualification (IQ), Operational Qualification (OQ) and Performance Qualification (PQ).
4. 4
Why are you here?
According to the Quality System Regulation (QSR), “Where appropriate,
each manufacturer shall establish and maintain procedures for
identifying valid statistical techniques required for establishing,
controlling, verifying the acceptability of process capability and
product characteristics.” Although there are many statistical methods
that may be applied to satisfy this portion of the QSR, there are some
commonly accepted methods that all companies can and should be
using to develop acceptance criteria, to ensure accurate and precise
measurement systems, to fully characterize manufacturing processes,
to monitor and control process results and to select an appropriate
number of samples.
6. 6
Statistical Techniques
“Valid in-process specifications for such characteristics shall be consistent
with drug product final specifications and shall be derived from previous
acceptable process average and process variability estimates where
possible and determined by the application of suitable statistical procedures
where appropriate.”
21 CFR 211.110 (b)
“Where appropriate, each manufacturer shall establish and maintain
procedures for identifying valid statistical techniques required for
establishing, controlling, and verifying the acceptability of process capability
and product characteristics.”
21 CFR 820.250 (a)
7. 7
GHTF Process Validation Guidance for Medical
Device Manufacturers
0. Introduction
1. Purpose and scope
2. Definitions
3. Processes that should be validated
4. Statistical methods and tools for process validation – Appendix A
5. Conduct of a validation – Getting started, Protocol Development, IQ,
OQ, PQ, Final report
6. Maintain a state of validation – Monitor and Control and Revalidation
7. Use of historical data in process validation
8. Summary of activities
Annexes:
A. Statistical methods and tools for process validation
B. Example validation
8. 8
Statistical Methods and Tools for Process Validation
Listed in GHTF Guidance, Annex A
Acceptance Sampling Plan
Analysis of Means
Analysis of Variance
Capability Study
Challenge Test
Component Swapping Study
Control Chart
Design of Experiments
Dual Response Approach to Robust Design
Failure Modes and Effects Analysis
Fault Tree Analysis
Gauge R&R Study
Mistake Proofing Methods
Multi-variable Control Chart
Response Surface Study
Robust Design Methods
Robust Tolerance Analysis
Screening Experiment
Taguchi Methods
Tolerance Analysis
Variance Components Analysis
9. 9
Applying Statistical Methods Throughout
Process Validation
Installation
Qualification
• Sample size calculations
• Hypothesis testing
• Data intervals
• MSA
Operational
Qualification
Performance
Qualification
• Ishikawa diagram
• FMEA
• DOE
• RSM
• SPC
• Process capability
• Robust Design Methods
• SPC
• Process capability
• FMEA
10. 10
Applying Statistical Methods Throughout
Process Validation
Installation
Qualification
• Sample size calculations
• Hypothesis testing
• Data intervals
• MSA
Operational
Qualification
Performance
Qualification
• Ishikawa diagram
• FMEA
• DOE
• RSM
• SPC
• Process capability
• Robust Design Methods
• SPC
• Process capability
• FMEA
12. 12
Installation Qualification
“Installation Qualification (IQ): establishing by objective
evidence that all key aspects of the process equipment and
ancillary system installation adhere to the manufacturer’s
approved specification and that the recommendations of the
supplier of the equipment are suitably considered.”
GHTF Guidance on Process Validation
13. 13
Installation Qualification
“Each medical device manufacturer is ultimately
responsible for evaluating, challenging, and testing the
equipment and deciding whether the equipment is
suitable for use in the manufacture of a specific
device(s).”
GHTF Guidance on Process Validation
14. 14
IQ for Heat Sealer
A new heat sealer will be installed, checked, and calibrated. This
installation qualification will ensure the average exhaust of
pressurized air in the clean room does not exceed the
requirements of 14 psi. Also, the heat sealer contains a device
which measures seal strength. As part of the IQ, ensure the
device provides accurate and precise measurements of seal
strength.
Confidence interval for the true mean (one-sided)
One-sample t-test (one-tailed)
Sample size considerations
Two-sample t-test and Equivalence (Comparability)
15. 15
Point Estimators
You are using a sample from a larger population to estimate the
mean, variance, and standard deviation; you use these
estimators to describe your sample.
Because you are estimating the true (population) parameter
with a single value, this is called a point estimator.
If instead you used a range of values to estimate the true
parameter, this range is called an interval estimator.
17. 17
Hypothesis Testing
The null hypothesis (H0) is the statement about what you
assume about the population parameter.
Usually, this is a statement that there is no difference.
The alternate hypothesis (Ha) is the statement about what you
prove about the population parameter.
Usually, this is a statement that there is a difference.
18. 18
Hypothesis Testing
Does the seal strength equal 6.5?
Is the seal strength the same for Supplier A and B?
Is the seal strength the same for each size pouch (small, medium,
large)?
Does the supplier effect depend on the size of the pouch?
Do different levels of time, temperature, pressure, and rate affect
the seal strength?
19. 19
One-sample t-test
H0: µ > 14 the true mean is greater than 14
Ha: µ < 14 the true mean is less than 14
α = 0.05 95% confidence
t-stat =
p-value =
20. 20
Types of Errors
Did you make the right decision?
The probability of a Type I error is α.
The probability of a Type II error is β.
The power of the test is 1- β.
True
Conclude H0 Ha
H0
CORRECT
Type II error
Ha Type I error
CORRECT
21. 21
Power
Power is the ability to detect differences that actually exist.
Power depends on:
• Sample size (n)
• α
• Difference to detect (δ) or effect size
• Standard deviation (σ)
5.5 6.5
22. 22
Using an alpha level of 0.05, a standard
deviation of 1.0, a difference to detect of 0.5,
and a power of (at least) 80%, determine an
appropriate sample size.
Power and Sample Size
22
23. 23
One-sample t-test
Using the randomly generated data, determine if
the true average exhaust of pressurized air in the
clean room is less than the requirement of 14 psi.
H0: µ > 14 the true mean is greater than 14
Ha: µ < 14 the true mean is less than 14
α = 0.05 95% confidence
t-stat =
p-value =
One-sided (95%) confidence interval:
(µ < )
Conclusion:
24. 24
Using the randomly generated data, determine
if the true average exhaust of pressurized air in
the clean room is less than the requirement of
14 psi.
One sample t-test
24
25. 25
Two-Sample t Test
H0: µA = µB The means are equal.
Ha: µA ≠ µB The means are different.
α = 0.05 95% confidence
t stat =
p-value =
t Test
Reactor B-Reactor A
Assuming equal variances
Difference
Std Err Dif
Upper CL Dif
Lower CL Dif
Confidence
-16.272
1.858
-12.465
-20.079
0.95
t Ratio
DF
Prob > |t|
Prob > t
Prob < t
-8.75575
28
<.0001*
1.0000
<.0001*
26. 26
Two-Sample t Test
H0: µA = µB The means are equal.
Ha: µA ≠ µB The means are different.
α = 0.05 95% confidence
t stat = -8.756
p-value = <0.0001
t Test
Reactor B-Reactor A
Assuming equal variances
Difference
Std Err Dif
Upper CL Dif
Lower CL Dif
Confidence
-16.272
1.858
-12.465
-20.079
0.95
t Ratio
DF
Prob > |t|
Prob > t
Prob < t
-8.75575
28
<.0001*
1.0000
<.0001*
27. 27
Equivalence Testing
The t test can conclude only that two sample means are
different. It cannot be used to show that the means are
the same.
An equivalence test reverses the null and alternative
hypotheses from the t test. If the result of an equivalence
test is significant, then the conclusion is that the two
means are practically equivalent.
28. 28
Equivalence Testing
H0: |µA − µB| > δ The means differ by more than δ.
HA: |µA − µB| ≤ δ The means differ by at most δ.
α = 0.05 95% confidence
An equivalence test is performed by forming a confidence
interval around the difference in sample means. If this
confidence interval is entirely contained within a user-selected
interval (−δ, δ), then equivalence is concluded.
Check whether the 90% CI formed around xA − xB is
contained within the interval (−δ, δ).
A test size of α constructs a (1 − 2α) confidence interval
because two different comparisons are being performed
(against the lower and upper sides of the CI).
The selection of δ is subjective and depends on subject-
matter expertise.
29. 29
Equivalence Margin
Selection of the equivalence criteria (δ) is the key to the
outcome of similarity.
Reference: Tsong, Yi, and OB CMC Analytical Biosimilar Method Development Team (Meiyu
Shen, Cassie Xiaoyu Dong). 2015. Development of Statistical Approaches for Analytical
Biosimilarity Evaluation [PowerPoint]. DIA/FDA Statistics Forum.
30. 30
Using the randomly generated data, determine
if two products are comparable (practically
equivalent).
Two Sample t-test and
Equivalence
30
32. 32
Operational Qualification
“Operational Qualification (OQ): establishing by
objective evidence process control limits and action levels
which result in product that meets all predetermined
requirements.”
GHTF Guidance on Process Validation
33. 33
Operational Qualification
“In this phase the process parameters should be challenged
to assure that they will result in a product that meets all
defined requirements under all anticipated conditions of
manufacturing, i.e., worst case testing. During routine
production and process control, it is desirable to measure
process parameters and/or product characteristics to allow
for the adjustment of the manufacturing process at various
action level(s) and maintain a state of control. These action
levels should be evaluated, established and documented
during process validation to determine the robustness of the
process and ability to avoid approaching ‘worst case
conditions.’ ”
GHTF Guidance on Process Validation
34. 34
GHTF Process Validation Guidance for Medical Device
Manufacturers
[Considerations include] “Potential failure modes, action
levels and worst case conditions (Failure Modes and Effects
Analysis, Fault Tree Analysis)”
“The use of statistically valid techniques such as screening
experiments to establish key process parameters and
statistically designed experiments to optimize the process
can be used during this phase.”
35. 35
OQ for Heat Sealer
First, determine potential key process parameters. Next, evaluate the
stability of these parameters; determine levels for screening experiments.
Then conduct both a screening experiment to set Installation optimal
conditions and a response surface study to center the process and
determine Installation process capability. Lastly, determine the sensitive
of the process to variations in these key process parameters and
establish process capability (Cpk > 1.0).
Cause-and-effect diagrams and FMEA
SPC
Screening experiment
Response surface study
Process capability
36. 36
Factors using Ishikawa
The first step to establishing key process parameters is to
brainstorm which process parameters (factors ) may
‘cause’ an ‘effect’ on seal strength. A key quality tool to
accomplish this is a cause-and-effect diagram (also
known as a Ishikawa or fishbone diagram).
37. 37
Factors using FMEA
The next step is to prioritize which process
parameters/factors to include in your experiments. Both
Failure Modes and Effects Analysis (FMEA) and Fault
Tree Analysis can be used to accomplish this.
38. 38
FMEA and FTA
“An FMEA is a systematic analysis of the potential failure
modes. It includes the identification of possible failure
modes, determination of the potential causes and
consequences and an analysis of the associated
risk…FMEA can be performed on both the product and
the process. Typically, an FMEA is performed at the
component level, starting with potential failures and then
tracing up to the consequences. This is a bottoms up
approach. A variation is a Fault Tree Analysis, which
starts with possible consequences and traces down to the
potential causes.”
GHTF Guidance on Process Validation
39. 39
FMEA
During FMEA brainstorming sessions, the following
ratings for Severity (Sev), Probability of Occurrence
(Occ), and the Probability of Detection (Det) are
determined. The Risk Priority Number (RPN) is
computed as:
RPN = Sev * Occ * Det
Item/Function
Potential Failure
Model
Potential Effect(s) of
Failure Severity
Potential Cause(s) of
Failure Occurrence
Current Design
Controls Detectability RPN
Platen Platen too hot Seal Strength too
low
10
Temp setting too
high 5 1
Platen defective 3 3
Platen too cool Seal Strength too
low
10
Temp setting too
high 5 1
Platen defective 3 3
40. 40
Control Charts
“Control charts are used to detect changes in the process. A sample, typically
consisting of 5 consecutive units, is selected periodically. The average and
range of each sample is calculated and plotted. The plot of the averages is
used to determine if the process average changes. The plot of the ranges is
used to determine if the process variation changes. To aid in determining if a
change has occurred, control limits are calculated and added to the plots. The
control limits represent the maximum amount that the average or range should
vary if the process does not change. A point outside the control limits indicates
the process has changed. When a change is identified by the control chart, an
investigation should be made as to the cause of the change. Control charts help
identify key input variables causing the process to shift and aid in reduction of
the variation. Control charts are also used as part of a capability study to
demonstrate that the process is stable or consistent.”
- GHTF Guidance on Process Validation
45. 45
Design of Experiments (DOE)
“The term designed experiment is a general term that
encompasses screening experiments, response surface
studies, and analysis of variance. In general, a designed
experiment involves purposely changing one or more
inputs and measuring the resulting effect on one or more
outputs.”
GHTF Guidance on Process Validation
46. 46
DOE for 2-Level Process Parameters
DOE allows you to detect the significance of main effects as well as their interactions.
Time Press
1
2
3
4
- 1
+1
- 1
+1
- 1
- 1
+1
+1
+1
- 1
- 1
+1
Time *
Press
Seal
Strength
5.7
6.3
7.0
7.5
+
_
_
Time
+
Press
47. 47
DOE for 2-Level Process Parameters
The benefits of designed experiments increases as the number of key
process parameters are added to the design. Add Temperature to the
design.
+
_ Temp
_
_
+
Time
+
Press
48. 48
DOE for 2-Level Process Parameters
The benefits of designed experiments increases as the number of key
process parameters are added to the design. Add Rate to the design.
+
_
Press
Temp
_
_
+
Time
+
−
Rate
+
+
_
_
_
+
Time
+
49. 49
Screening Experiment
“A screening experiment is a special type of designed
experiment whose primary purpose is to identify the key
input variables. Screen experiments are also referred to as
fractional factorial experiments...”
` GHTF Guidance on Process Validation
50. 50
Screening Experiment
Fractional factorial experiments give up information about some of all
interactions in favor of examining more parameters. For the heat sealer, we
may want to know whether Time, Temperature, Pressure, or Rate has the
largest effect on Seal Strength. A 24 full-factorial design will have 16 runs. A
half-fraction factorial will have 8 runs.
Temp Press
1
2
3
4
5
6
7
8
- 1
+1
+1
- 1
+1
- 1
- 1
+1
- 1
+1
- 1
+1
- 1
+1
- 1
+1
- 1
- 1
+1
+1
- 1
- 1
+1
+1
Time Rate
- 1
- 1
- 1
- 1
+1
+1
+1
+1
51. 51
This demonstration illustrates how to design
and analyze a screening design with Seal
Strength as the response and Time,
Temperature, Pressure, and Rate as the
factors. For Seal Strength, Match a Target of
6.5 (specification is 5.5 – 7.5).
Screening Designs
51
52. 52
Response Surface Study
“A response surface study is a special type of designed
experiment whose purpose is to model the relationship
between the key input variables and the outputs.
Performing a response surface study involved running the
process at different settings for the inputs, called trials,
and measuring the resulting outputs. An equation can be
fit to the data to model the effect of the inputs on the
outputs. This equation can then be used to find optimal
targets...To ensure that only key input variables are
included in the study, a screening experiment is
frequently performed first.”
GHTF Guidance on Process Validation
53. 53
Central Composite Design
A Central Composite Design (CCD) is a widely-used response
surface design.
Adds axial runs to the initial design.
Each factor in the design has 5 levels.
Each (added) experimental run has one factor at its axial value
and all others at 0.
+
_ Temp
_
_
+
Time
+
Press
54. 54
This demonstration illustrates how to design
and analyze a CCD. What are your optimal
settings to match a Seal Strength of 6.5?
Run confirmation runs at the process settings to
determine Installation process capability.
Response Surface Design
54
55. 55
Capability Study
“Capability studies are performed to evaluate the ability of a
process to consistently meet a specification. A capability study is
performed by selecting a small number of units periodically over
time. Each period of time is called a subgroup. For each
subgroup, the average and range is calculated. The averages and
ranges are plotted over time using a control chart to determine if
the process is stable or consistent over time. If so, the samples
are then combined to determine whether the process is
adequately centered and the variation is sufficiently small. This is
accomplished by calculating capability indexes. The most
commonly used capability indices are Cp and Cpk. If acceptable
values are obtained, the process consistently produces product
that meets the specification limits. Capability studies are
frequently used towards the end of validation to demonstrate that
the outputs consistently meet the specifications.”
GHTF Guidance on Process Validation
56. 56
Is the Process Capable?
The most commonly used capability index is Cpk.
Example:
LSL USL
57. 57
Is the Process Capable?
The most commonly used capability index is Cpk.
Example:
LSL USL
58. 58
Process Capability
LSL USL LSL USL
Cpk = 1.0 Cpk = 2.0
When Cpk = 1,
27/10,000 results
will fall outside of
the specification.
When Cpk = 2,
2/1,000,000,000
results will fall outside
of the specification.
59. 59
This demonstration illustrates the use of
confirmation runs to determine if your process
is capable (Cpk > 1.0).
Process Capability
59
60. 60
This demonstration illustrates the use of
confirmation runs to determine if your process
is capable at:
a. Optimal settings
b. All low settings
c. All high settings
Determine the Cpk for each of the three settings.
Process Capability
60
62. 62
Performance Qualification
“Performance Qualification (OQ): establishing by
objective evidence that the process, under anticipated
conditions, consistently produces a product which meets all
predetermined requirements.”
GHTF Guidance on Process Validation
63. 63
Performance Qualification
“In this phase the key objective is to demonstrate the
process will consistently produce acceptable product under
normal operating conditions.’”
GHTF Guidance on Process Validation
64. 64
PQ for Heat Sealer
In IQ, we ensured the heat sealer was installed correctly. In OQ,
we conducted tests to ensure the seal strength would meet the
pre-determined specifications under all manufacturing conditions.
In PQ, we want to demonstrate process consistency under normal
operating conditions. In order to accomplish this, we need to test
seal strength for an extended period of time; we need to
determine if our process is stable and capable. We would also like
to evaluate if our process is centered - how our process average
compares to the target.
SPC
Process Capability
66. Adsurgo provides direct engagement consulting services
and training workshops focused on the use of analytics.
Our passion is for solving interesting, challenging, and
meaningful problems in collaborative, team-based
engagements with our clients.
W. Heath Rushing
Principal Consultant
206-369-5541
Heath.Rushing@adsurgo.com