Upload
dangdat
View
218
Download
4
Embed Size (px)
Citation preview
Applying Quality by Design to Generic Drug Manufacturing
Bikash Chatterjee President & CTO Pharmatech Associates
1
2
Agenda• What is QbD?• Why it has become important • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required statistical processes• Practical application of the ideas-Case Study• Review of past records to determine CPP-Case Study• Development of acceptable operation range• Benefits• Cost savings
3
QbD’s Proposition• QbD concerns the making of drug substances and drug products• QbD is the new pharmaceutical quality system that:
• Replaces current GMP concepts• Does not depend on the trial and error approach of drug
substance and drug product development & production• Is a systemic, knowledge and risk-based quality methodology • Complies with the general purpose of product quality: the
product is suitable for use • Patient driven philosophy• A quality system customized for pharmaceuticals
•QbD is GMP for the 21st century
4
What is Quality by Design (QbD)?
• First introduced in 1985 by Dr. Juran
• Juran said most quality problems are designed into the process. A clear plan is needed to identify and eliminate these issues
• No single definition…
5
a systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science and quality risk
ICH Q8 Definition of QbD
6
Another Way of Thinking about QbD
Once a system has been tested to the extent that the test results are predictable, further testing can be replaced by establishing that the system was operating within a defined design space.
7
Understanding what factors have an impact on variation in your process and also on your product’s performance; then establishing a control plan tomonitor and maintain product quality
My definition of QbD
8
Elements of Quality by Design (QbD)
•• •
• ••
GMPsGMPs
EUUSFDAPIC/S
ICHQ8,Q9Q10Q11
GMPsEU
USFDAPIC/S
Risk
QualityTargetProductProfileCTPP
CriticalQualityAttributesCQAs
QbD
CriticalProcessParametersCPPs
ControlStrategy
DesignSpace
9
Quality by Design Stages
QbD
Quality PlanningQuality Target Product Profile (QTPP)Quality Critical Attributes (QCAs)
Quality ControlCritical Process Parameters (CPPs)Control Strategy
Quality ImprovementProcess ControlProcess Monitoring
Quality Planning
Quality Improvement
Quality Control
10
What is Quality by Design (QbD)?
Pharma s version of Juran s Model
ProcessControl Features
Product Development
QTPPCQAsInputs
Process
Outputs
QbD
Implem
entation
11
Sources of VariationManagement Man Method
Cause Cause Cause
Cause
Cause
Cause
Cause
Cause
Cause
Effect(Y)
Cause
Cause
Cause
Cause Cause
Cause
Cause Cause Cause
Measurement Machine Material
12
Agenda• What is QbD?• Why QbD has become important • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required statistical processes• Practical application of the ideas-Case Study• Review of past records to determine CPP- Case Study• Development of acceptable operation range• Benefits• Cost savings
13
Business Dynamics
14
Why Has QbD Become Important?• Business Drivers
o New market opportunities
o Improved market competitiveness
o Improved profitability
o Reduced product risk exposure
15
QbD
Three Areas of Improvement...
• Better
• Faster
• Cheaper
• Quality
• Time/Flow
• Waste/Costs
16
Drive Financial Performance
Increase Revenue: Grow the Business• Improve customer satisfaction, sales, throughput, and
competitive position
Decrease the Cost of Goods Sold• Reduce process variation and defects, improve yield
• Identify and eliminate root causes of problems
• Develop systems robust to problems
• Reduce unnecessary costs and excessive cycle time
17
QbD is a Better Business Model• R&D drives new innovative products• Do we really need QbD? The conservative criticism
• “All the billions of dollars poured into research and development in the U.S. won’t mean a thing. We must streamline and strengthen the regulatory science”
• Areas cited where this is being accomplished include FDA’s partnership with ICH around Quality by Design (QbD)
New FDA commissioner Margaret Hamburg’s keynote address at Regulatory Affairs Professionals Society annual conference in Philadelphia, September 2009
Conclusion: QbD is a way to innovate within the pharmaceutical industry
18
Regulatory Drivers for QbD
Escalating and non-uniform compliance expectations:
- ASEAN Harmonization Activities
- ICH, PIC/S, EU, CFDA (China), MHLW (Japan), CDSCO (India), MOH (Malaysia), FDA Thailand, NA-DFC (Indonesia)
19
US/EU/PIC/S QbD Regulatory Timeline
ASEAN Harmonization Milestones
• A-CTD Implemented• A-CTR & technical guidelines established (maintenance and
enhancement of common interpretation ongoing)• Post-Market Alert System established• GMP Inspection MRA finalized• Training identified• Pan-ASEAN registration
1999 2002 2005 2006 2009
PPWG IWG GMPMRATF
BA/BETF
A-CTDImplementation
20
21
Regulatory Drivers-ICH Q8, 9, 10, 11
ICH Q8, Q9, Q10 & Q11are designed as separate but linked in a series of documents exploring pharmaceutical products lifecycle (www.ich.org)
• ICH Q8 - Pharmaceutical Development • ICH Q9 - Quality Risk Management • ICH Q10 - Pharmaceutical Quality System • ICH Q11 - Development and Manufacture of Drug
Substances
22
Agenda• What is QbD?• Why it has become important• What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Practical application of the ideas-Case Study• Review of past records to determine CPP-Case Study • Required statistical processes• Development of acceptable operation range• Benefits• Cost savings
23
Is QbD a Shift in Quality Philosophy?You can’t test quality into drug products” has
been heard for decades – so what s new?• Quality is based on process and product
understanding, not just test results• It’s a shift in culture: incorporates quality principles
and strong compliance function• Incorporates risk assessment and management• Refocuses attention and resources on what’s
important to the customer, i.e. the patients, health professionals, payors and distribution chain
24
QbD is a Commitment to Improve• Continuous improvement is a key element of QbD
- G. Taguchi on Robust Design: Design changes during manufacture can result in the last product produced being different from the first product
• However, in pharmaceutical manufacturing, we want improvement that improves consistency–patients and physicians must count on each batch of drug working just like the batches that came before
25
QbD for Generic Drugs
In generic pharmaceutical manufacturing, there are additional constraints:
• Fixed bioequivalence targets
• Regulatory requirements to duplicate formulation of innovator drug
• Lack of access to innovator development data
26
The Changing Regulatory Compliance Environment
Quality by Design
• Adequate resources for quality: number, qualifications, etc.
• Self-assessments play key role
• Continuous analysis & improvement
• Change management based on good science
• Focus on what’s important (risk management)
Current Regulatory Situation: US/EU
• Little guidance on adequate resources or qualifications
• Self-assessments not trusted• Annual product reviews instead
of continuous analysis• Formidable barriers to change,
including intimidating enforcement emphasis
• Seldom admit that anything is not important; test everything
27
Quality by Design (QbD) CharacteristicsBasics: • Uses systemic (multivariate statistics) development and
manufacturing by use of prior knowledge• Risk assessment guided design and process control• Applies to the total life cycle of a product (continuous
improvement)
Implications:• Quality back to the roots: product suited for its purpose• Quality is dynamic: continuous improvement• Quality must be built in• Quality means first time right
28
The QbD Development Model is Different
•
Patient Idea Design Space Control Strategy Risk Assessment Product Life Cycle
Idea Development Preclinical & Licensing Manufacturing Marketing/
Clinical Testing Sales
Traditional
QdB In the QbD Development Concept The Chain is Reversed
29
QbD Will Require Enhanced Supplier Management
• Why?You will need to measure and control the important characteristics of your raw materials and API
• Clearly defined supplier quality and supply agreements are necessary
30
Agenda• What is QbD?• Why it has become important • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required statistical processes• Practical application of the ideas• Review of past records to determine CPP • Development of acceptable operation range• Benefits• Cost savings
31
Building a QbD Organization
• Starts in product development
• Multidisciplinary team representing the product development lifecycle
• Presents opportunities to build in existing commercial experience into the product and process design phase
• Presents the opportunity to not repeat mistakes in formulation and product design
32
Team Structure
QbDCore Team
R&D & Marketing
Corporateand Mfg.
Engineering
Technical Services
Regulatory and QA
Compliance
Facilities/GC
Validation
Oversight Committee
33
QbD Core Team• Program Manager• Decision makers from all
six areas• Clear mandate to deliver
product. In the US FDA market measured by being the First to File
QbDCore Team
34
Team Chartering Process
Define and Identify:• Success metrics for the project• Timeline• Budgetary and cost tracking assumptions• Key stakeholders• Project champion and project milestones• Extended Chartering to discussion of communication, review
and issue resolution mechanism• Also established initial team rules: what behaviors would be
encouraged and what would not be encouraged
35
Managing Team Dynamics
TeamCharter
Structure
Systems
Staff
Strategy
Skills Style
Knowledge Management
ProductSelection
Development And
Characterization
Site Selection/Process Design/Tech Transfer
Reg.Filing
Process Understanding Process Predictability Measurement
ContinuousMonitoring
Phase 1 Phase 2 Phase 3 Phase 4 Phase 5
Key Activity
• QTPP• Strategic
Analysis• Site
Capability Analysis
• ProjectTimeline
• Risk Analysis
KeyActivity
• Platform Knowledge
• Identify CPP• MSA• CMA Risk
Analysis• Process Risk
Analysis• Commercial
Factors
Go/N
oGo
Go/N
oGo
Go/N
oGo
KeyActivity
• Site Suitability• Mapping CPP• MSA• Process Risk
Analysis• Confirmation
Process • DOE • Process
Validation
Key Activity• Filing
Prep.
Key Activity
• Metrics Review
A Generic Drug QbD Framework
Go/N
oGo
36
37
Agenda• What is QbD?• Why it has become important • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required statistical processes• Practical application of the ideas-Case Study• Review of past records to determine CPP-Case Study• Development of acceptable operation range• Benefits• Cost savings
Reducing Variation by Robust Design (QbD)
By Robust Process Design ...
By tighter controls of the inputs ...
Input
Proc
ess
"Y"
Traditional Method of Reducing Variation Alternate Method of Reducing Variation
•TransferRelationship
•$ $ $
38
39
Effect of Design on the Product Development Life Cycle
Design Produce/Build Deliver Service Support
Cost &Time vs. Impact
Potential is PositiveImpact >Cost and Time
Impact< Cost &Time
40
Scope of Recent Guidances
ProductDesign ManufacturingProcess
Design
ProcessMonitoring
/ContinuousVerification
ICH Q8/Q8(R) - Pharmaceutical DevelopmentICH Q11 Development and Mfg. of Drug Substances
PAT Guidance
ICH Q9 – Quality Risk Management
FDA Guidance on Quality Systems (9/06)FDA Process Validation GuidanceICH Q10 – Pharmaceutical Quality Systems
41
ICH Q8- Pharmaceutical Development
• Introduces the concept of pharmaceutical Quality by Design
• Defines QbD as:
A systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science andquality risk management.
42
ICH Q8 Concept of QbD
Process Understanding
• Process Parameters• Process Controls
43
Designing a Robust Process
Problems detected after they
occur, throughproduct testing and
inspection
Reproducible process within
narrow operatingranges
Robust & reproducible
process
Low High
Low
High
PROCESS UNDERSTANDINGPR
OCE
SSCO
NTR
OL
High potential for failures
44
Role of Quality Risk Management inDevelopment & Manufacturing
ManufacturingImplementation
Process Scale-up & Tech Transfer
Risk Management
ProcessDevelopment
ProductDevelopment
Product qualitycontrol strategy
RiskControl
RiskAssessment
Processdesign space
ProcessUnderstanding
Excipient & drug substance
design space
Product/priorKnowledge
RiskAssessment
Continualimprovement
ProcessHistory
RiskReview
44
45
Define desired product performance
upfront;identify product CQAs
Design formulation and process to meet product CQAs
Understand impact of material attributes and process parameters on
product CQAs
Identify and control sources of variability
in material and process
Continually monitor and update
process to assure consistent quality
Risk assessment and risk control
Product & process design and development
Qualityby
Design
The FDA QbD Model
46
Process Step Analysis
For product and process:
- Risk assessment- Design of experiments- Design space definition- Control strategy- Batch release
CRM DOEs Design Space
Control Strategy
Batch Release
47
Adding the QbD Framework
CMA DOEs Design Space
Control Strategy
Batch Release
QTPP/ CQAs CPPs
Quality Risk Management
48
Quality Target Product Profile (QTPP)
“A prospective summary of the quality characteristics of a drug product that ideally will be achieved to ensure the desired quality, taking into account safety and efficacy of the drug product”
• Defines the product development requirements. Used to be called the product Requirement Specification (PRS)
• ICH Q8 Definition is:
49
QTTP- PRS Example
Dosage form and strengthImmediate release tablet taken orally containing 30 mg of active ingredient
Specifications to assure safety and efficacy during shelf-life
Assay, Uniformity of Dosage Unit (content uniformity) and dissolution
Description and hardness Robust tablet able to withstand transport and handling.
Appearance
Film-coated tablet with a suitable size to aid patient acceptability and compliance
Total tablet weight containing 30 mg of active ingredient is 100 mg with a diameter of 6 mm
50
QTTP- Safety and Efficacy ExampleTablet Product Requirements Critical to Quality
Attributes (CQA)Dose 30 mg Identity, assay and
CUMarketing Taste masking, coated
tablet, suitable for global market
Size, Appearance, Potency
Safety- Purity Impurities and degradation products meet ICH guideline
API impurities and degradation products <1%, residual Solvents
Efficacy-API PSD* Drug bioavailable with PSD that meet mfg needs
Dissolution >60% 1 hour per USP 711
Shelf Life 2 years and meets ICH guidelines
Primary packaging oxygen barrier required for shelf life
*PSD: Particle Size Distribution
51
Risk Analysis
• It doesn’t need to be complex• High, medium and low risk ratings are
acceptable• Anything with a high rating should be
justified• Apply the risk analysis to the product
design (formulation) and the process design activity at the outset
52
Example Product Risk AnalysisCQA Microcrystalline
cellulosePovidone Mg. Stearate API
Appearance Low Low Low LowAssay Low Low Low HighContentUniformity Low Low Medium High
Dissolution Low Medium Medium HighHardness Medium Low Low LowJustification PSD critical to
solubility of drug. Low loaded dose can affect CU
53
Process Unit Operation Risk Assessment
CQA Process StepsGranulation Drying Milling Blending Compression Coating
Appearance Low Low Low Low Medium HighAssay Low Low Low Medium Low Low Impurity Low Low Low Low Low Low BlendUniformity
Low Low Medium High High Low
Drug Release Low Low Low Medium Medium HighParticle Size Distribution
Medium Low High Low Low Low
Justificationsfor High Rating
N/A N/A
Milling screen size and speed can affect the PSD and therefore the powder flow and tablet fill weight control
Blending can affect blend uniformity, assay, and drug release profile
Compression can affect drug uniformity in the tablet based upon particle size variability and flow
The final appearance and drug release rate are affected by the coating quality and reproducibility
54
Risk Analysis
• Important to go back to the risk assessments at the end of the process development activity and finalize the risk assessment based upon real data
55
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical processes• Development of acceptable operation range• Practical application of the ideas-Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
56
The QbD Framework for Process Space
57
Knowledge Space
The potential range of limits for all parameters controlled or measured during the process characterization process
58
Knowledge Space Determination
• Determine which parameters have an impact on the products performance
• Requires establishing a range for each parameter to evaluate for each unit operation
• Pharma has historically used One-Factor-At-A-Time (OFAT) to do this but this is not adequate today
59
Design of Experiments (DOEs)• An approach which allows us to
understand the contribution to variation of a parameter(s) upon a known response variable
• Establish a mathematical model which describes the impact of each variable controlled on the dependent variable of interest
• OFAT studies cannot do this
60
DOE vs. OFAT• DOEs allow you to understand the
process behavior in a very few studies
• DOEs allows the experimenter to apply statistics to back-up their conclusion
• The only way to have confidence your conclusion is correct
61
Experimental VariabilityAny experiment is likely to involve three kinds of variability:• Planned, systematic variability This type of variability we want
since it includes the differences due to the treatments
• Chance-like variability This type of variability our probability models allow us to live with. We can estimate the size of this variability if we plan our experiment correctly
• Unplanned, systematic variability This type of variability threatens disaster! We deal with this variability in two ways, by randomization and by blocking. Randomization turns unplanned, systematic variation into planned, chance-like variation, while blocking turns unplanned, systematic variation into planned, systematic variation
The management of these three sources of variation is the essence of experimental design.
Taken from In Introduction to the Design and Analysis of Experiments, George Cobb (1998)
62
Things to Consider
• Sample size• Sampling Plan• Additional characterization tests, e.g,
powder flowability, bulk/tapped density, PSD- d10, d50 an d90, intermediate dissolution time points
• Acceptance criteria. Specifications are not always fully descriptive
63
High Level Map of Experiments
Screening Designs
CharacterizationStudies
OptimizationStudies
One Factor at a timeFractional Factorials
Full Factorials
Response Surface Methods
64
Knowledge Space Output
• Will reduce the number of variables that matter in terms of product performance
• Will define the broad limits of the knowledge space which will be used to drive the Design Space
65
Example DOE - Compression
• Examine the impact of Turret Speed (rpm) and Compression Force (N) on Tablet Hardness and Tablet Dissolution
• Turret Speed: 15-30 rom• Compression Force : 10-20 kN• So we have 2 factors each with 2
levels
66
Example DOE - CompressionTurret Speed Compression Force Tablet Hardness 4 Hr Dissolution
(rpm) (kN) (kP) (%)30 20 11 7615 20 13 7930 20 12 7830 10 11 7215 10 10 7630 10 9 7415 10 10 7715 20 15 71
Does Turret Speed matter?Does Compression Forces matter?
67
Example DOE- Compression ANOVADo these variables have an impact on tablet hardness? = 0.05
Estimated Effects and Coefficients for Tablet Hardness (coded units)
Term Effect Coef SE Coef T PConstant 11.3750 0.3750 30.33 0.000Turret Speed -1.2500 -0.6250 0.3750 -1.67 0.171Compression Force 2.7500 1.3750 0.3750 3.67 0.021Turret Speed*Compression Force
-1.2500 -0.6250 0.3750 -1.67 0.171
S = 1.06066 PRESS = 18R-Sq = 82.61% R-Sq(pred) = 30.43% R-Sq(adj) = 69.57%
Yes!
68
Example DOE- Compression ANOVADo these variables have an impact on tablet dissolution? = 0.05
Estimated Effects and Coefficients for 4 Hr Dissolution (coded units)
Term Effect Coef SE Coef T PConstant 75.3750 1.068 70.58 0.000Turret Speed -0.7500 -0.3750 1.068 -0.35 0.743Compression Force 1.2500 0.6250 1.068 0.59 0.590Turret Speed*Compression Force
2.7500 1.3750 1.068 1.29 0.267
S = 3.02076 PRESS = 146R-Sq = 34.68% R-Sq(pred) = 0.00% R-Sq(adj) = 0.00%
69
What if I am Not Strong In Statistics?
The Pareto Chart provides the same answer
70
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical processes• Development of acceptable operation range• Practical application of the ideas- Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
71
Design Space
72
Look Only at the Parameters that Affect the Drugs Performance
• In our example Compression force affected tablet hardness which was a drug release criteria
• Narrow the range to be evaluated and this becomes your new Design Space limits for this variable, e.g. conform the contribution from 12-18 kN
73
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical Processes• Development of acceptable operation range• Practical application of the ideas- Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
74
Statistical Testing
• The purpose of applying statistical tests is to compensate for the fact that we cannot test every unit we make
• So we make a guess ,i.e. a hypothesis of whether, within a predefined level of error, our decisions are correct
75
What is a Test of Hypothesis?
• A statistical test designed to answer a question, or allow one to choose between two or more alternatives:
• Is material A better than material B?
• Does the new process have a larger yield over the our older process?
• Does this lot meet our specifications?
• Tests of hypothesis provide a structure for learning
• Properly handle uncertainty
• Minimize subjectivity
• Question assumptions
• Prevent the omission of important information
• Manage the risk of decision errors
Hypothesis testing concepts allow us to.....?
76
77
Ho: Parameter or Measure = a value, or is trueHA: Parameter or Measure {< or > } a value, false
= a low probability typically of 1%, 5%, or 10%
• The hypothesis of equality,or that condition that is considered true is typically called the Null Hypothesis
• The hypothesis of non-equality is called the Alternate Hypothesis
• All hypothesis' include a level of significance, , which is the risk of incorrectly rejecting a true Null Hypothesis
Hypothesis Test Configuration
78
Fundamentals of Hypothesis Testing
• Based on existing knowledge, we form a hypothesis to explain something about the unknown observation
• Frequently, the hypothesis is the opposite of what we hope to show
• Collect data to evaluate the null hypothesis • Assume the null hypothesis is true (favored hypothesis)
• Seek compelling evidence in the data to support or contradict that hypothesis
• If the null hypothesis is contradicted we reject the null hypothesis and accept the alternative hypothesis
79
Hypothesis and Decision Risk• When testing a hypothesis, we do so with a known
degree of risk and confidence• We must specify in advance:
• Magnitude of acceptable decision risk • Test sensitivity
• These provide the necessary information to determine an appropriate sample size
• Consider practical limitations of cost, time, and available resources to arrive at a rational sampling plan
• We can never acheive absolute certainty
TheTruth
The Decision Errors
Your Decision
Ho is True
Ho is False
Type IError
Risk)
Type II Error
Risk)
Correctdecision
Correctdecision
Reject HoDo not reject Ho
Keller and Warrack, Statistics for Management and Economics
80
Alarm’s Decision
Nothing In Bag
TheTruth
Nothing In Bag
Weapon In Bag
Type IError
Risk)
Type II Error
Risk)
Correct
Correct
Weapon In Bag
Consequences: _____________
Consequences:__________________________
Example: Airport Security
81
82
• Sampling from a distribution must be representative or independent• Random sampling is the key assumption• Often Normality is the key assumption• The random sampling assumption is also
known as the statistical independence assumption
• A plot of the data in time order should not show any trends
• Check by finding out how the samples were chosen and tested
Hypothesis Testing – Assumptions
83
Hypothesis Testing – Common Tests• 1 sample t-test (compares sample mean to a value)
• 2 sample t-test (compares one sample mean to another)
• 1 way analysis of variance (ANOVA) (compares more than two sample means)
• Correlation and Regression Analysis (compares paired data to a linear model)
• Design of Experiments(compares the effects of factors on the process output)
• Chi square test for independence (compares multiple proportions)
84
Hypothesis Testing –Procedure
The Test is on Populations, NOT Samples…
1. Write the null hypothesis
2. Write the alternate hypothesis
3. Decide on the p value
4. Choose hypothesis test
5. Gather evidence and test/conduct analysis
6. Reject H0 /not reject H0 and draw conclusion
H0 : x Sample A = x Sample B (e.g. new way is the same as the old way)
HA : There is a difference between Samples A and B
p = .05 (typical for characterization projects)
Choose the correct test, given the type of X and Ydata (in this example, a t-test)
Collect data, run analysis, get p value
If p >.05 conclude that your data does not show a significant difference between samplesIf p<.05 conclude the samples are different
Steps •Example (2 Samples)
• Key Question: Do you have sufficient evidence to reject the Ho ?
• The p-value is the most common way to evaluate the results of your test
• Common ways to remember what the p-value means:
If p is low, Ho must go!
or
If the p is high keep the guy!
Making Decisions with Hypothesis Tests
85
86
For most cases we will use .05
How Low Must the p-value Be?
• We would like there to be less than a 10% chance of falsely rejecting Ho ( = .10)
• 5% is much more comfortable ( = .05)
• 1% feels very good ( = .01)
• This alpha level is based on our assumption of “no difference” and a reference distribution of some sort
• But, it depends on interests and consequences
P-value is required to reject Ho
87
Data Types
• Discrete• Counts of discrete events (1, 2, 3, …
defects)• Qualitative descriptions
• Good / Bad • Supplier 1, Supplier 2, …• Method A, Method B, …
• Continuous• Decimal sub-divisions are meaningful
• Time, money, etc.
88
Hypothesis Testing Roadmap
• Different statistical tools apply to different types of input and output data combinations
• Minitab supports all these combinations• Structured approach to choosing the right
analysis method
“If the only tool you have is a hammer...every problem looks like a nail” - Abraham Maslow
Testing Rubric
X DataSingle X Multiple Xs
Sing
le Y
X DataDiscrete Continuous
Y D
ata D
iscr
ete
Con
tinuo
us
•Chi-Square•Logistic
•Regression
•ANOVA
•T-test •Regression
X DataDiscrete Continuous
Y D
ata D
iscr
ete
Con
tinuo
us
•Multiple
•Regression
•Logistic•Regression
•Multiple
•Medians Tests
•2, 3, 4+ way•ANOVA
•Logistic•Regression
•Multiple
89
90
Sampling Strategy
91
Sampling Strategy
• Commonly Overlooked in terms of its importance in establishing process understanding
• If you do not have confidence your sample is representative of your true process population you cannot be sure your conclusions are correct
92
Terms: Additional Definitions
• Sample: A subset of a population. For us this is the data we collect
• Population: Not the same as “sample”, rather this is the data we would have collected if you had repeated the experiment an infinite number of times
93
Terms: Additional Definitions• Acceptance Sampling:
Form of inspection applied to lots or batches of items before or after a process, to judge conformance with predetermined standards
• Sampling Plans: Plans that specify lot size, sample size, number of samples, and acceptance/rejection criteria• Single-sampling• Double-sampling• Multiple-sampling
94
Sampling Myths• Sampling plans can make a bad process better
• A stringent sampling plan ensures only good product goes out the door
• My sampling plan justifies my quality decision for my production lots
Sampling plans have no impact on process capability and are not a surrogate for process improvement
No. The only way to ensure 100% good product goes out the door isto make 100% good product
No. Your sampling plan can only extrapolate the behavior of the population from which it was sampled. You must use scientific inference to apply this to other lots
95
When to Use Sampling
• Product Development: Demonstrating product performance
• Process Development: Understanding process behavior• Process Optimization: Improving process behavior• Quality Control: Verifying incoming raw materials,
API, components and product release testing
• Stability: Product Expiration testing• In-Process Testing: Establishing an effective control
strategy
96
Operating Characteristic Curves
97
Acceptance Sampling
• Modern acceptance sampling involves a system of principles and methods. Their purpose is to develop decision rules to accept or reject product based on sample data. Factors are: • The quality requirements of the product in the
marketplace • The capability of the process • The cost and logistics of sample taking
98
Probability of Acceptance (Pa)• The primary characteristics when evaluating a
sampling plan is to understand what the probability of accepting a lot is as the percentage defects in the population changes
• The behavior the sampling plan is defined by its Operating Characteristic (OC) Curve
• All OC Curves have certain properties in common:• At 0% defective the probability of acceptance is 1• At 100% defective, the probability of acceptance is 0• As the percent defective is increased the OC curve
decreases
Operating Characteristic Curve
00.10.20.30.40.50.60.70.80.9
1
0 .05 .10 .15 .20 .25
Pro
babi
lity
of a
ccep
ting
lot
Lot quality (fraction defective)
3%
99
Decision Criteria
0
1.00
Pro
babi
lity
of a
ccep
ting
lot
Lot quality (fraction defective)
“Good”
“Bad”
Ideal
Not verydiscriminating
100
101
Sampling Terms• Acceptance Quality Level (AQL)
the percentage of defects at which consumers are willing to accept lots as “good”
• Lot Tolerance Percent Defective (LTPD)the upper limit on the percentage of defects that a consumer is willing to accept
• Consumer’s Riskthe probability that a lot contained defectives exceeding the LTPD will be accepted
• Producer’s Riskthe probability that a lot containing the acceptable quality level will be rejected
102
OC Definitions on the Curve
•Pro
babi
lity
of A
ccep
ting
Lot
Lot Quality (Fraction Defective)
100%
75%
50%
25%
.03 .06 .09
= 0.0590%
= 0.10
AQL
LTPD
Indifferent
Good Bad
Producer Risk
Consumer Risk
OC Curves can be summarized by two points:
AQL: Percent defectivewith a 95% chanceof acceptance
LTPD: Percent defectivewith a 10% chance of acceptance
103
OC CurvesP
roba
bilit
y of
Acc
epti
ng L
ot
Lot Quality (Fraction Defective)
100%
75%
50%
25%
.03 .06 .09
OC Curves come in various shapes depending on the sample size and risk of and errors
This curve is more discriminating
This curve is less discriminating
104
Pro
babi
lity
of A
ccep
ting
Lot
Lot Quality (Fraction Defective)
100%
75%
50%
25%
.03 .06 .09
This curve distinguishes perfectly between good and bad lots.
The Perfect OC Curve
What would allow you to achieve a curve like this?
105
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical processes• Development of acceptable operation range• Practical application of the ideas- Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
106
Acceptance Criteria
• At a minimum all process testing must meet specifications
• The specifications should be derived from the product requirements and the process’ capability
• Ideally you can steer and predict the process’ performance
107
Establishing Acceptance Criteria
• Confidence Intervals: Determines the probability that the confidence interval produced will contain the true parameter value. Common choices for the confidence level Care 0.90, 0.95, and 0.99. These levels correspond to percentages of the area of the normal density curve
• Because the normal curve is symmetric, half of the area is in the left tail of the curve, and the other half of the area is in the right tail of the curve. For a 95% confidence interval, the area in each tail is equal to 0.05/2 = 0.025.
Measures our degree of uncertainty in the population mean
108
Establishing Acceptance Criteria
• Prediction Intervals: Determines the probability interval that a single value will fall. Tends to be larger than confidence intervals.
Measures our degree of uncertainty and the variability in the distribution of the population mean is affected by sampling error.
109
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical processes• Development of acceptable operation range• Practical application of the ideas- Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
110
CASE STUDY
111
Case Study Framework
• This case study is a real process that has been qualified to US, EU and PIC/S standards
• Applies the principles of QbD to demonstrate process understanding and process control
112
Pharmatech’s Technology Transfer Roadmap
Point...Point...Point...Point...
PointPointPointPoint
ProductDesign
CPPs/RiskAssessment
HistoricalPerformance
EquipmentDesign
CharacterizationStudies
EstablishPAR/NOR
PPQPrerequisites
PPQ
RiskAssessmentVerification
Change Controland Stage 3
Recommendation
Pro
cess
Und
erst
andi
ng
Pro
cess
Rep
rodu
cibi
lity
ContinuousImprovement
RiskAssessmentVerification
Pro
cess
Mon
itorin
g
Pro
cess
Und
erst
andi
ng
Pro
cess
Rep
rodu
cibi
lity
Pro
cess
Mon
itorin
g
113
Case Study Application
114
Lexicon• Critical Process Parameter (CPP): A process parameter
whose variability, within defined limits, has an impact on a critical quality attribute and therefore should be monitored or controlled to ensure the process produces the final drug product quality
• Critical Quality Attribute (CQA): A physical, chemical or microbiological property or characteristic that should be within a predetermined range, range or distribution to ensure the desired final product drug quality
• Critical To Quality Attribute (CTQ): An in-process output parameter that is measured and/or controlled that should be within a predetermined range, range or distribution to ensure the desired final product drug quality
115
Stage 1 Process Understanding
• Product Design• Process Risk Assessment• Equipment/Process Characterization Studies
• Sampling Plans• Sampling Techniques• Method Robustness
• Design Space Establishment• Validation Master Plan
116
Product Design• Why go back to product design?
• Understand what is important: ProductRequirement Specification (PRS)
• Have solid grasp of formulation and product design rationale
• Formulation may provide insight as to which processing steps are critical downstream
• Rationale for product design helps define how the formulation, raw materials and process steps are related to achieving desired product performance
117
Key QTPP PRS SpecificationsKey criteria from the PRS include:• Greater than 50 percent Active
Pharmaceutical Ingredient (API)• Round 200 mg tablet• Coated to mask taste• 12-hour drug release with the following
specifications:• 4 hour dissolution 20-40 percent• 8 hour dissolution 65-85 percent
118
Raw Material and API Considerations
• Consider existing qualified Suppliers when choosing excipients
• Includes a review of products with similar PRS requirements
• Foundation for Knowledge Management effort
• API characterization includes Supply Chain and Quality Engineering feedback from current products
119
Tablet FormulationRaw Material %w/w Function
API 60 Active ingredientMicrocrystalline cellulose 22 Excipient fillerPovidone K 29-32 5 Granulation binderLactose 12 Excipient fillerMg Stearate 1 LubricantPurified water QS SolventCoating Solution Raw Material %w/w FunctionEudragit Coating Solution 12 Controlled release
polymerTriethyl Citrate 1 PlasticiserTalc 1.5 GlidantWater QS Solvent
120
Process Risk Assessment• Helps identify which processing steps could
affect process stability in Stage 2• Process map to capture inputs, outputs,
and control variables• Process FMEA’s to prioritize key process
steps and KPIV’s• Critical to Quality Attributes(CTQs) identified
• Helps focus characterization studies
121
Risk Assessment Process Map
• Identify formulationdriven PRS requirements
• Establish boundaries forthe process step riskassessment
• Conduct risk map• Review development data• Analyze historical
performance to setacceptance criteria
Develop Process MapIdentifyCPP/CTQ/CQAs
Development/HistoricalData Gap Analysis
• Identify controlled/uncontrolled variables
• Establish basicmeasurement approach
• Separate between scaleindependent anddependent variables
122
Process Unit Operation Risk Assessment
CQA Process StepsGranulation Drying Milling Blending Compression Coating
Appearance Low Low Low Low Medium HighAssay Low Low Low Medium Low Low Impurity Low Low Low Low Low Low BlendUniformity
Low Low Medium High High Low
Drug Release Low Low Low Medium Medium HighParticle Size Distribution
Medium Low High Low Low Low
Justificationsfor High Rating
N/A N/A
Milling screen size and speed can affect the PSD and therefore the powder flow and tablet fill weight control
Blending can affect blend uniformity, assay, and drug release profile
Compression can affect drug uniformity in the tablet based upon particle size variability and flow
The final appearance and drug release rate are affected by the coating quality and reproducibility
123
•Target Set Point
•Max Set Point Run(s)
•Min Set Point•Run(s)
•PAR•NOR
•Limit of individual•excursions
•Duration of process
•Variability of actual data•around set point
Relationship Between Proven Acceptable Range and Normal Operating Range
124
Historical Analysis
• The absence of development data establishing the PAR and NOR for the CPP can be ascertained to some extent by evaluating the historical behavior of each parameter along with the corresponding behavior of the CQAs for the unit operation
• Data should be extracted from multiple batch records to determine whether the process is stable within lot and between lots
• The team went back into the batch records of approximately 30 lots across a period of one year to extract the necessary data. This exercise also gave some indication as to whether the parameter was truly a CPP, based upon whether it had an impact on the corresponding CQA for the unit operation
• Evaluate scale independent and scale dependent parameters
125
Control Charts
126
Process Capability Analysis
127
I Chart of PSD
128
Correlation Plot
129
Equipment Design Considerations• Compare underlying equipment design and
configuration differences• Focus on impact of equipment design on scale
dependent parameters• Objective during transfer and scale-up is to
understand where equipment can affect the PAR And NOR for the transferred process
• Also consider final PV considerations such as sampling plans, sampling technique, and method robustness
130
Historical data Review Conclusion
• Dissolution testing of uncoated tablets across the process range were 100% dissolved in 3 hours
• Storage studies determined bulk granulation and uncoated tablets were sensitive to humidity
• Operating characteristic (OC) curves developed for each unit operation to understand the relationship between sampling size and sampling risk (AQL vs. LTPD)
• Highlight sampling challenges prior to design space activity
131
Tech Transfer Equipment Comparison
132
Summary of CPP/CTQ and CQAAssumptionfor Tech Transfer
Unit Operation CPP CTQ CQACompounding Mixing speed Fully Dissolved-
Visual
Water temperature Addition rate Fluid Bed Granulation/Drying
Spray Rate Granulation PSD-d10, d50, d90
Content Uniformity
Inlet Air Humidity Moisture content Potency Atomization
pressure LOD
Bulk/Tapped Bulk Density
Milling Screen size PSD Blending Mixing Speed Content
Uniformity Mixing Time Potency-Assay Compression Pre-compression
force Tablet Thickness Dissolution
profile Compression force Tablet Weight Content
Uniformity Tablet Hardness Potency-Assay Friability Coating Spray Rate Percent Weight
Gain Dissolution Percentage at 4 and 8 hours
Atomization Air Pressure
Appearance Potency-Assay
Inlet Air Temperature
133
Tech Transfer-Sampling Qualification
• Sampling Methodology QualificationGage R&R conducted with sampling equipment for each unit operation. GRR< 20%, Distinct Categories > 5
• Sampling Plan DevelopmentCould use ANSI Z1.4-2008 or Zero-Acceptance Plan. Used Power calculation, e.g. Powered at 80% with 5% as significant difference for a known SD
134
Tech Transfer Characterization Study
• Historical review concluded final product CQA for dissolution is not affected by upstream process before coating
• Confirmation DOEs are required to establish PAR and NOR upstream with a focus on process predictability
• Coating process DOE’s designed to demonstrate comparability, confirm CPP’s, and provide supportive data for PAR and NOR
• Also included commercial studies, e.g. solution hold time, pan load studies, etc.
135
Drug Dissolution Dependence on Coating Weight
136
Validation Master Plan• Summarizes the rationale for Process performance
Qualification• CPPs, CTQs and CQAs• Summarizes the impact of controlled variables• Introduces approach for understanding impact of
uncontrollable parameters• Justifies sampling plan based upon process risk• Defines acceptance criteria based upon product
CQA’s
137
Stage 2- Process Qualification
• Demonstration phase of the PV cycle• Precursors to this stage
• Facility and utilities that support the process must be in state of control
• Process equipment must be qualified (i.e. IQ, OQ, PQs are complete)
• In-process and release methods used for testing must be validated and their accuracy and precision well understood
• Cleaning validation is complete• Essential to have precursors completed to ensure
unknown variability is due to process alone
138
Stage 2 Process Qualification (cont.)• New term: Process Performance Qualification (PPQ)
• Intended to include all known variables from the manufacturing process
• Focused on demonstrating reproducibility. This drives the acceptance criteria
• Cumulative understanding of Stage 1 and Stage 2• No more three lots and we’re done• Performed as many lots needed to demonstrate a clear
understanding of variables and process is in control• Data derived from studies will be used to measure
manufacturing process in Stage 3
139
Establishing Acceptance Criteria• Based upon reproducibility criteria• For example if the Stage 1 performance for the 4 hr.
dissolution was 32% 2% against a specification of 20-40%:• Acceptance criteria could be: 95% confidence
interval applied to a spec of 32 6%• Used a 2 sided t-Test with an = 0.05 (0.025 on
the HA for < comparison)• We used the 6% because it is 3 x std. dev. In a
normal distribution this covers 99.7 of the data variability for a controlled process
140
Why Can’t I Just Compare My Result Against the Acceptance Limits?
• We did not know the true mean and standard deviation of the population That is the premise behind the t-test. If we knew it we would use the z-test
• We only knew the behavior of our sample population and we must infer that the process population behaves the same. That is why we apply the confidence interval to the assessment and apply the alternative hypothesis to test if the variability and mean is within what has historically seen
141
Agenda• What is QbD?• Why it has become important? • What companies need to know, overview• How to set up a team to develop QbD• Process understanding• Knowledge space• Design space• Required Statistical processes• Development of acceptable operation range• Practical application of the ideas- Case Study• Review of past records to determine CPP-Case Study• Benefits• Cost savings
142
Benefits• Improved new product development
capability and flexibility• Reduced quality overhead and reduced
quality issues• Greater productivity and predictability of
the process and overall business operations
• Ability to correct for process drift without impacting quality or yield
• Access to larger profitable pharmaceutical markets
143
Cost of Poor Quality (COPQ)
COPQ is derived from the non-value adding activities of waste in a process and is made up of costs associated with one of the following five categories:
1. Internal failure2. External failure3. Appraisal4. Prevention5. Lost Opportunity
•Reference; Basu and Wright, Quality Beyond Six Sigma 2003
144
COPQ Components
•Reference: Wild 2002
145
Cost Savings Examples
• Generic drug could not be made consistently. Off market for 1 year, Applying QbD principles over 6 weeks restored $200 million revenue stream
• Applying QbD to a platform drug product reduced the number of non-conformance reports by 75%saving nearly $1million/annually
146
Conclusion• The principles of Quality by Design have been
proven in multiple industries including pharmaceutical
• Pursuing Quality by Design does not require additional capital or overhead. Just good science
• The business benefits of improved control and greater productivity provide for amore stable and predictable business operation
147
Questions?
148
Thank You for Your Attention!Bikash Chatterjee, President & CTO
Pharmatech Associates, Inc.22320 Foothill Blvd. #330
Hayward CA 94541510-732-0177
Or visit our website at:www.pharmatechassociates.com