How to Conduct the Method Validation with
a New IPT (In-Process Testing) Method
IVT Method Validation, San Francisco, July 30, 2010
Weifeng “Frank” Zhang QA Engineer
BMS
Ernest LeeSenior Manager, Facilities & Engineering
Medarex, a fully owned subsidiary of BMS
Forward Looking Statements
Except for historical information, the matters contained in this slide presentation may constitute forward-looking statements that involve risks and uncertainties, including uncertainties related to product development and clinical trials, unforeseen safety issues resulting from the administration of antibody products in patients, uncertainties related to the need for regulatory and other government approvals, dependence on patents and proprietary technology, the need for additional capital, uncertainty of market acceptance of Medarex’s product candidates, the receipt of future payments, the continuation of business partnerships and other risks detailed from time to time in Medarex’s filings with the Securities and Exchange Commission (SEC).
All forward-looking statements included in this slide presentation are based on information available to us as of July 30, 2010. We do not assume any obligation to update any information contained in these materials. Our actual results may differ materially from the results discussed in the forward-looking statements.
Disclaimer:
All data in this presentation is not from actual
tests. Virtual data is used in this presentation for
confidentiality reasons.
Weifeng “Frank” Zhang
B.Sc. (Materials Eng.), M.Eng. (Chem. Eng.),
12 years in pharmaceutical, biotech
Ernest Lee
AGENDA
Overview of method validation
and IPT
New automated method
Validation approach
Q&A, Open Discussion
OVERVIEW
Method validation elements
– Accuracy
– Specificity
– Precision
– Suitability
– Linearity
– Repeatability
– Limit of Detection
– Limit of Quantitation
– …
Things to Consider in Method Validation
Parameters Applicable to Different Analytical Procedures: From: Method Validation Guidelines, Sep 15, 2005
By: Alex D. Kanarek, Ph.D. BioPharm International
IPT in This Study
Cedex
Automated Method
IPT (In-Process Test)
The most important parameters
Total cell counts
Cedex Description
Photos from: http://www.innovatis.com
IPT in This Study – Cedex Description
Counts cells and determines viability based on the
Trypan Blue Exclusion method
Takes 20 digital images of the fluid as it passes through
a flow cell
Evaluate light level differences between the background
and the discovered cells
Counts – Cells are included or excluded from the count based on
relative size
Viability – Shading differences between the cells are utilized to
assess viability
IPT in This Study – Dynamic Data
In bio-production, every day, every hour,
every minute counts!
Accounting for Human Factors Variance
User-dependent
New Automated Method
Manual vs. Automated (
human factors in here?)Manual method uses a microscope and a hemacytometer
after staining the cells with Trypan blue
After study and practice, develop SOPs to minimum the
human variance in the manual measurement
Challenges
Different testing instruments between two
methods
No “Gold Standard”
Validation Acceptance Criteria
IOQ and PQ for instruments have been
successfully performed
Identify parameters
No significant difference between new
automated method and existing manual method
Validation Approach
Design validation experiments
We will have to consider
all factors!
Accounting for All Variance
Formula:
Where, σ
is total variance,
In manual method, σ
1is sample variance, σ
2is human factor variance;
In automated method, σ
1is sample variance, σ
2is instrument
variance;
We need to make sure sample variance is
minimized between methods!
...
2
2
2
1
2
=
σ
+
σ
+
σ
Sampling Matrix Design
Day #
Sample Taken
BioreactorManual
Cedex
0
Two samples taken at the same time. Time______#1 Operator 1
Cedex log 1
1
Two samples taken at the same time. Time______#1 Operator 2
Cedex log 2
2
Two samples taken at the same time. Time______#1 Operator 3
Cedex log 3
3
Two samples taken at the same time. Time______#1 Operator 1
Cedex log 4
…
…
#1…
…
n
Two samples taken at the same time. Time______#1 Operator 3
Cedex log n
Raw Data
Cedex\Method Validation\Cedex Method
Validation Analysis -- Final.xls
Data Overview
Comparison Between Two Methods
4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 le C e ll C ount s ( x 1 0 6 c e ll/m L ) Manual Method Auotmated Method
Statistical
Since viable cell counts base changes on different days, suggesting the
differences are not directly comparable. So the data was transformed
to percent differences to exhibit constant variance over the entire study.
Viable Cell Count Difference Between Manual and Auotmated Methods -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 3.00 0 1 2 3 4 5 6 7 Days C e ll C o u n t D iffe re n ce (x 1 0 6 cel l/m L ) Day 0 Day 4 Day 7 Day 11 Day 14 Day 18
Viable Cell Count Percent Difference Between Manual and Automated Method -20.00% -10.00% 0.00% 10.00% 20.00% 30.00% 40.00% 0 1 2 3 4 5 6 7 Days P er cen t D if fer en ce Day 0 Day 4 Day 7 Day 11 Day 14 Day 18
Statistical
Since there is no “Gold Standard” in this study, we have to test
whether the two methods are different
null hypothesis: The measurements from two (2) methods have no
significant difference
From testing data of two methods, we can calculate the mean and
standard of the percent differences. Then construct a 95%
confidence interval on the mean
Whether the 95% confidence interval includes zero will decide if the
null hypothesis is rejected
Remember, trade-offs associated with using higher confidence
levels
Statistical
Mean of dataZero Upper 95% confidence level
Lower 95% confidence level
null hypothesis can beacceptedin this case.
Mean of data Zero Upper 95% confidence level
Lower 95% confidence level null hypothesis is rejected in this case
Statistical
Null Hypothesis
Actual Result
Reject Fail To Reject
True
Type I Error
α: Probability of Type I Error (significance level)
True
False True (1-β) Type II Error : Probability of Type II Error
Data Analysis – Step by Step
Step 1 – Calculate mean for each day across all
bioreactor runs
Step 2 – Calculate the percent difference in cell count
Step 3 – Calculate the mean and standard deviation for
the percent difference in cell count
Step 4 – Calculate the upper and lower 95% confidence
intervals
Step 5 – Verify that “0” falls within the upper and
lower 95% confidence intervals
Theory Behind Validation Approach
ANOVA Test
A collection of statistical models. It provides a
statistical test of whether or not the means of
several groups are all equal.
SSTREATMENT SSERROR ANALYSIS OF VARIANCE BIOREACTOR
ANOVA Test
M
H0: μ1= μ2 H1: μ1≠μ2 M = f ( t, ε(t))M :Measurand (Cell Concentrations) (Viable and Non-Viable)
AGITATOR GROWTH PLATEAU DEATH Cavg(r, z, φ, t) (P) (pH, T, PO2) (rpm)
+
SSTOTAL SSTREATMENT SSERROR ANALYSIS OF VARIANCE BIOREACTORANOVA Test
M
M = f ( t, ε(t)) AGITATOR GROWTH PLATEAU DEATH Cavg(r, z, φ, t) (P) (pH, T, PO2) (rpm)+
SSTOTAL XMEAN= 4 SS = ∑(Xi– XMEAN)2MSTREATMENT
MSERROR
MICROSCOPE
ANALYSIS OF VARIANCE
HEMACYTOMETER / HUMAN SYSTEM
AUTOMATED CELL COUNTER SYSTEM LINEARITY REPEATABILITY LIMIT OF DETECTION LIMIT OF QUANTITATION BIOREACTOR
ANOVA Test
M
μP SENSORSAMPLING / DILUTION / DIVISION
H0: μ1= μ2
H1: μ1≠μ2
M = f ( t, ε(t))
M :Measurand (Cell Concentrations) (Viable and Non-Viable)
(TRYPAN BLUE) (TRYPAN BLUE) AGITATOR GROWTH PLATEAU DEATH Cavg(r, z, φ, t) (P) (pH, T, PO2) (rpm) ε0 MST= ∑∑(cij– x)2/(Degrees of Freedom) C11 C12 C13 C21 C22 C23 RANDOMIZATION REPLICATION F < FCRITICAL
Summary
Resolved the challenge of testing in the absence
of a standard
Designed experiments in validation protocol
Used statistical method to perform data analysis
Established acceptance criteria based on 95%
confidence interval
Special Thanks to:
David Lorah
Supervisor, Validation Department
Lance Marquardt
Senior Manager, Biopharm Production Department
John R. Mosack
Senior Director, Clinical Manufacturing & Validation
Interactive Exercise
Q &A
Open item
707 State Road • Princeton, NJ 08540 T: +1-609-430-2880
F: +1-609-430-2850 www.medarex.com