Validation with Designed

Experiments
Lynn Torbeck
Thank You
Thanks to the local WCC-PDA
organization and committee for inviting
me here to discuss one of my favorite
topics.
A special thanks to Joerg Herrmann and
Ron Branning of Genentech for their
assistance in making the arrangements.
Statistics Anyone?
Course in statistics?
Statistics Anyone?
Course in statistics?
Course in Designed Experiments?
Statistics Anyone?
Course in statistics?
Course in Designed Experiments?
Any statisticians in the audience?
Statistics Anyone?
Course in statistics?
Course in Designed Experiments?
Any statisticians in the audience?
Anybody want to be a statistician?
Statistics Anyone?
Course in statistics?
Course in Designed Experiments?
Any statisticians in the audience?
Anybody want to be a statistician?
It is a great career.
What is this Talk About?
What is the problem?
 How to do insanely great validations.
Why is that a problem?
 Validation can be costly and lengthy
What is the solution?
 Many, not all, validations are best done
using designed experiments.
Has it Been that Long?
I gave my first talk on using DOE for
validation in 1978 in Muncie IN.
Director of Validation, WW, at G. D.
Searle in Skokie, IL.
Have been teaching DOE since 1988
with an emphasis on validation.
Book editor for Validation by DOE.
The Source of the Nile
21 CFR 2110(a).
“Such control procedures shall be
established to monitor the output and
to validate the performance of those
manufacturing processes that may be
responsible for causing variability.”
O’No, Not again
1977, “Establishing documented evidence that
a system will do what it purports to do.”
1987, “Process validation is establishing
documented evidence which provides a high
degree of assurance that a specific process
will consistently produce a product meeting
its predetermined specifications and quality
attributes. FDA
Genentech Case Study
“Validation of Large-scale
Chromatographic Processes,”
Timothy Breece, Ellen Gilkerson, and
Charles Schmelzer
BioPharm, May 2002, P 16
BioPharm, July 2002, p 35
Genentech Case Study
“Validating that performance [of a
chromatographic process] involves
demonstrating that the
chromatographic media can be reused,
cleaned, and sanitize and that the
process, as defined, is robust.”
Plastic Body Parts
FDA wanted validation of an old process
Identified 5 factors.
Low spec limit and high spec limit.
Need three lots.
3 * 25 = 96 lots.
Looking at all possible combinations is
not necessary.
Plastic Body Parts
Using a designed experiment:
Use a 25-1 of 16 conditions or runs.
Randomly assign 4 runs to each of 4
batches.
Vary the 5 factors over, but within, the
allowable specification.
Can sell the batches when done if they
meet the quality criteria.
Step Back for the Big Picture
Flowchart the Process.
Identify critical factors and levels
Goal of validation?
 Fully describe the process with equations.
 Show the process is robust empirically.
 Do three lots at target.
Snow Country Company
A MN company developed a very
complex manufacturing process. Only
two in the world.
Did a series of designed experiments to
define the process empirically.
The report is held in a bank vault.
Used when there are problems.
Who is Ed?
Edmund M. Fry
Director of Drug Quality Compliance
FDA
Later the President of the PDA for years
Now in private industry
Ed’s Lost in the Mists of Time
One of the most important papers ever
written on validation.
“The FDA’s ViewPoint”
Drug Cosmetic Industry
137 (1), pp 46-51
1985
What did Ed say?
“The regulations require validation of
those processes responsible for causing
variabilitys in characteristics of in-
process materials and finished product.”
“The regulations imply that not
everything that takes place in a
pharmaceutical manufacturing plant
causes variability.”
Ed Really Said That ?????
“Therefore some things don’t have to
be validated.”
“We never intended to require that
everything that takes place in a
manufacturing operation is subject to a
validation study.”
WOW! The FDA said that !
How Ed said to do it.
“Experiments are conducted to assure
that factors that would cause variability,
are under control, and will result in an
output that meets the specifications
within the limits of the ranges that you
had previously established.”
Sounds like the plastic body parts
experiment ?
Was Ed a Sailor?
“We are saying a process should not be
operated under … conditions which
have not been included in validation
studies.
“To put it another way, don’t operate
a process in uncharted waters.”
Charting Ed’s Water
As Breece, Gilkerson, & Schmelzer said,
we need to show the process is robust.
Validation is often, not always, a
demonstration project.
Can we show that varying critical
factors over their ranges doesn’t cause
a significant change in the product?
What is Old is New Again
Statistical Manual of the Association of
Official Analytical Chemists, AOAC, 1975
W.J. Youden and E. H. Steiner.
P 33, “Ruggedness test for procedures.”
“The Design of Optimum Multifactorial
Experiments,” Biometrika, 33, 305
(1946).
What Youden do?
He used experiments called Plackett –
Burman designs.
AKA resolution III designs.
Vary many factors each at two levels in
very few runs.
“… if the program is carefully laid out, a
surprisingly small amount of work
suffices.”
Say it Again (and again … )
“Ruggedness and Robustness with
Designed Experiments”
Lynn Torbeck
In a special supplement to
Pharmaceutical Technology
March, 1996
A Early Paper by Friends
“Prospective Process Validation”
Allen Y. Chao, F. St. John Forbes,
Reginald F. Johnson and Paul von
Doehren
In the book Pharmaceutical Process
Validation, 1 st Ed, 1984, Marcel Dekker
A Quick Example, 27-4
This is the Design Table. It is shorthand notation for how to conduct the experiment.
A B C D E F G Response
- - - + + + -
+ - - - - + +
- + - - + - +
+ + - + - - -
- - + + - - +
+ - + - + - -
- + + - - + -
+ + + + + + +
With Data
Temp Pressure Time Speed Hold Reactor Feed Impurity
-1 -1 -1 1 1 1 -1 0.95
1 -1 -1 -1 -1 1 1 1.67
-1 1 -1 -1 1 -1 1 0.99
1 1 -1 1 -1 -1 -1 1.59
-1 -1 1 1 -1 -1 1 0.47
1 -1 1 -1 1 -1 -1 0.95
-1 1 1 -1 -1 1 -1 0.39
1 1 1 1 1 1 1 0.99
Calculate the Effects
Temp Pressure Time Speed Hold Reactor Feed
-0.95 -0.95 -0.95 0.95 0.95 0.95 -0.95
1.67 -1.67 -1.67 -1.67 -1.67 1.67 1.67
-0.99 0.99 -0.99 -0.99 0.99 -0.99 0.99
1.59 1.59 -1.59 1.59 -1.59 -1.59 -1.59
-0.47 -0.47 0.47 0.47 -0.47 -0.47 0.47
0.95 -0.95 0.95 -0.95 0.95 -0.95 -0.95
-0.39 0.39 0.39 -0.39 -0.39 0.39 -0.39
0.99 0.99 0.99 0.99 0.99 0.99 0.99
Sum 2.40 -0.08 -2.40 -0.02 -0.24 0.00 0.24
Effect 0.60 -0.02 -0.60 -0.01 -0.06 0.00 0.06
Interpreting the Effects
Normal Plot

1.5

0.5
M Value s

0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
-0.5

-1

-1.5
Effects
A Robust Process
Normal Plot of Effects

1.5

0.5
Normal Scores

0
-0.15 -0.1 -0.05 0 0.05 0.1 0.15
-0.5

-1

-1.5
Effects
OK, What Will the FDA Say ?
Depends on who you get, but:
FDA requires the well worn “double
blind clinical trials” which are designed
experiments, usually done by the
textbook.
Most animal trials are DOE.
FDA pushing DOE for stability studies.
You're Still a Skeptic ?
Just the reason we wrote a paper:
“Designed Experiments-A Vital Role in
Validation”
Lynn Torbeck and Ron Branning
Pharmaceutical Technology
June 1996
What to do Next?
Read the paper by Breece, Gilkerson, &
Schmelzer.
Read the paper by Torbeck and
Branning.
Talk to your local statisticians.
Find and read a book on DOE.
One of the Best
Experimental Design in Biotechnology
by Perry D. Haaland
Marcel Dekker
1989
What I Think I Said
Validation is not three runs at target.
Need to demonstrate robustness for
critical factor ranges.
How the data are collected is as
important as the data themselves.
DOE has a long history of use.
DOE is the most efficient way.
Thank You !
Any Questions?