Practical Considerations for Implementation of Real Time Release testing (RTRt)

Graham D. Cook Ph.D.,Senior Director, Global Quality OperationsHavant, UKgraham.cook@pfizer.com

CPAC Satellite MeetingRome, March 25, 2010

1

Overview

What needs to be considered when implementing Real Time Release Testing?

Case Study 1 – Product X

Case Study 2 – Product Y

Comparison of some aspects of Product X and Y

Conclusions

2

Real Time Release Testing

‘The ability to evaluate and ensure the quality of in-process and/or final product based on process data, which typically include a valid combination of measured material attributes and process controls.’ (ICH Q8(R2))

33

REAL TIME RELEASE TESTING

SystemsScience

People

ICH Q10 Pharmaceutical Quality System

ICH Q8 Pharmaceutical Development

ICH Q9 Quality Risk Management

Real Time Release Testing - Essential Elements

44

ScienceProcess data and Control strategy?Analytical methods and specifications?

Sampling and StatisticsAcceptance criteria

PeopleOrganization and Training?

Pharmaceutical Quality SystemQuality risk management?Disaster recovery?Model maintenance?Handling of outliers?Data management? Batch disposition?

Regulatory Interactions

Considerations for Implementing Real Time Release Testing

Topic discussed in Case StudyProduct X Product Y

Note: All above factors considered for Product X and Product Y

5

Overview

What needs to be considered when implementing Real Time Release Testing?

Case Study 1 – Product X

Case Study 2 – Product Y

Comparison of some aspects of Product X and Y

Conclusions

666

Product X - Product Description

Dosage FormBCS Class 3 compound (High Water Solubility, Moderate Permeability)Monolithic Extended Release Tablets (multiple strengths)Relatively high dose compound (high drug load)Level A IVIVC has been established for all dose strengths

Polymer is the primary driver of clinical performanceVariation in process parameters did not impact the dissolution performance

Potential to apply Real Time Release TestingProcess understanding at pilot-scale translated well to commercial manufacturing scaleRobust control strategy

7

Product X - Manufacturing Process

Robust control strategy → Increased assurance of quality → RTRt

PAT1 PAT

1

PAT:1 = Blend Uniformity2 = Granule particle size3 = Weight, Hardness, Potency, Drug concentration, Identity,

Rate–controlling polymer concentration

PAT2 PAT

3

Holistic Control Strategy e.g.:Content Uniformity = Blend uniformity + Drug concentration

+ Weight control

88

8

Development of Sampling Plans

PAT AnalyzersMeasure condition of the process material in real timeCollect more information about the batch

When and where do we sample?Use of prior knowledge of product and risk assessment in design of plan

e.g. consider potential for segregation (failure mode) in a compression mixStatistical rationale for sampling plan

Performance of plan relative to assurance of qualityRelationship between Real Time Release Testing and Pharmacopoeial TestingPlacement of PAT device in manufacturing equipment

Ensure analyzer measurements are representative of the batch

9

Operating Characteristic CurvesPlot the probability of accepting a batch against the quality level

What is the probability of rejecting a good batch? (Producer risk)What is the probability of releasing a bad batch? (Consumer risk)

Operating Characteristic curves enableDifferent batch acceptance sampling plans to be investigatedDesign of a batch sampling plan with the desired performance

N = batch size

n = sample size

c = critical acceptance number

AQL = Acceptable Quality Level

LQL = Limiting Quality Level

Steepness of the curve indicates discriminating ability of the plan

1010

Operating Characteristic Curves for UDU Test

The USP OC curve shows 90% probability of passing a batch with 2% product outside 85-115% of Label ClaimThe Alternative Plan OC curve shows 0% probability of passing a batch with 2% outside 85-115% of Label Claim The Alternative Plan uses a larger sample size and provides a better assurance of quality than USP

Probability to passAlternative Plan90% probability of passing bx by USPProbability to passUSP

% Product outside 85‐115% of Label Claim

0 1 2 3 4 5 6 7

Note: UDU = Uniformity of Dosage Units (Content Uniformity)‘Coverage’ = % Product meeting 85-115% of Label Claim ‘Coverage’ term used in paper by Sandell, D. et al. (see reference slide 24 )

If the Alternative Plan is chosen for sampling using a PAT analyzer what is the fall-back if the analyzer fails?

- Must maintain same level of quality by using the same sampling plan with the replacement analytical technology

11

Product X - Sampling

Location of PAT Analyzer in BlenderRationale based on prior knowledge (including publications)Blender design: no ‘hot’ spots therefore any location suitableAnalyser placed in bottom of blender

Sampling plan for TabletsTablets samples taken uniformly across batchX tablets taken every Y minutes during compression run

Drug Content of TabletsDrug concentration by NIR and Weight Uniformity determined on different tablet samples

Low risk because high drug loadingCheck using Monte Carlo simulation

12

Monte Carlo SimulationsMonte Carlo Simulation*

A technique that converts uncertainties in input variables of a model into probability distributions. By combining the distributions and randomly selecting values from them, it recalculates the simulated model many times and brings out the probability of the outputAllows several inputs to be used at the same time to create the probability distribution of one or more outputsThe output is generated as a range instead of a fixed value and shows how likely the output value is to occur in the range

Assessment of Drug Content of TabletsDrug Concentration (NIR) and Tablet Weights are independent probability distributionsWhat is probability of releasing a batch of marginal quality?Use Monte Carlo simulation to explore the variation in:

Drug Concentration (NIR mean; NIR std.dev.)Tablet Weight (Weight mean; Weight std. dev.; difference in weights from double-sided tablet press)

*Sanford Bolton, Charles Bon: Pharmaceutical Statistics - Practical and Clinical Applications, Fourth Ed., Marcel Dekker

13

Using Monte Carlo Simulations to Assess RiskAssessment of risk of releasing a marginal quality batch for chosen Coverage level:Combination of Drug Concentration (NIR mean) and Tablet Weight (WT mean) parameters

Indicates potential scenarios where there is a high probability of releasing a batch of marginal quality:cases with probability < 6%cases with probability 6 - 8%cases with probability 8 - 10%cases with probability > 10%

+■

Monte Carlo simulations:• 1400 cases, 5000 simulations run• Worst-case scenarios run

Very low number of high probability cases found – equates to very low probability of releasing marginal quality batch

Simulations help provide an assessment of the risk for the chosen coverage level and sampling plan

1414

RTRt - OrganizationMulti-disciplinary / cross-functional teams are key to RTRt

New skill sets may be needed

Real Time Release testing (RTRt)

Technology

Regulatory Affairs

QualityOperations

Statistics

FormulationDevelopment

Chemometrics

AnalyticalDevelopment/PAT

Operations

1515

15

Impact on Quality Systems

Some considerationsWhat happens if parameters are outside -

Normal Operating Ranges (NORs)? Design space?

What do we do if a PAT measurement system stops functioning?How do we handle atypical results/outliers?

Cannot ‘test into compliance’ using alternative methods (ICH IWG Q&A)

What do we do if the chemometric model is no longer appropriate?What is the impact on the batch release process?

What are alternative sampling plans and measurement systems to enable disposition decisions?

16

Quality Risk Management

Quality Risk ManagementEnabler for implementation of Real Time Release testing

Incorporate QRM into proceduresFacilitates uniform implementation of QRM across the entire organizationConsistent use of the same language (terminology) and processEstablish criteria for re-evaluation of risks and mitigation plans

Triggered by time, event or new knowledge

Training program Different levels

Awareness trainingUser training - participant, facilitator, team leader

Facilitates selection of correct QRA approach and toolCulture - proactive approach to quality

1717

Decision Tree for Equipment Failure

* Note that it may be impractical to stop some unit operations (e.g. wet granulation) during processing

Does PAT pass System Suitability?

NO

Proceed with unit operation/manufacturing

process

YES

YES

NO

Is PAT functional during

the Process?

YES

Can the instrument be repaired in a suitable time frame?

Can the instrument be replaced with a spare instrument?

Generate Event Report Form, Fix/replace

instrument

NO

Are there alternative controls to ensure/control process variability?

Are there measurements downstreamwhich could be used to correlate data?

Do we need further sampling?

Generate Event Report Form, Capture Process/

Action Items

Revert to Testing using Regulatory Analytical Procedure

Generate Investigation/ERF to identify root cause

Capture Process/Action Items

YES

Stop Process*:Evaluate Instrument

Define response in advance = ‘proactive quality’, rather than after problem has occurred = ‘reactive quality’

Extract of Decision Tree for PAT equipment failure

18

Quality Systems and Processes-Handling of Outliers

Atypical Results/OutliersBad data or valuable information?‘Statistical’ processes subject to random variability

Pre-defined mechanisms or systems should be developed = proactive approach to quality

Process to define outliers or invalid dataDefinition of deviations that require investigationSpecifications must be carefully defined

Consideration of Impact on Product QualityUse science- and risk-based approaches Use an holistic assessment of all product and process measurements to assess qualityFrequency of outliers may be importantExtent of process knowledge and process/product history (e.g. changes)Potential impact to patient safety and efficacy

19

Product X – Regulatory Interactions

FDAFDA CMC Pilot

February 2008 NDA ApprovedComparability Protocol for RTRt approved as part of the original application

July 2007 Pre-Operational Visit to manufacturing facilityOctober 2009 Prior Approval Supplement for RTRt submittedNovember 2009 PAI for RTRt

Pre-Operational VisitFDA personnel

Center (Reviewers (2), Compliance officer (1))Field (District (2), Patriot team (1))

Focus on PAT systems and Quality System includingProcedures for chemometric model maintenance Chemometric modelsEquipment failure recovery systemsSampling Plans Risk Assessments

(PAI similar to POV)

20

Overview

What needs to be considered when implementing Real Time Release Testing?

Case Study 1 – Product X

Case Study 2 – Product Y

Comparison of some aspects of Product X and Y

Conclusions

212121

Product Y - Product Background

Dosage FormBCS Class 1 compound (High Water Solubility, High Permeability)Immediate Release TabletsPotent, low dose compound, low drug load

Launched from small-scale containment manufacturing facility

Potential for Real Time Release TestingIncreased demand – need to increase efficiency

More information in real-timeRobust control strategy

22

API & Excipients

Dispensing Blend BlendSieving Granulation

MagStearate

MagStearate

PAT1

PAT2

BlendTablettingCoating PAT3

PAT4

PAT2

PAT2

PAT:1 = Identification2 = Blend Uniformity3 = Weight, Hardness, Disintegration, Potency, Content Uniformity 4 = Water Determination

Product Y - Manufacturing Process

23

Control StrategyMethods for conventional QC Release

TestingMethods for Real Time Release

Testing

Appearance (Visual) Appearance (Visual)

Identity (TLC / HPLC) Identity (NIR)

Impurities (HPLC) Not Tested routinely - due to process capability and understanding

Assay / CU (NIR + Weight)

Dissolution (Dissolution Tester) Disintegration (Disintegration Tester)

Water Determination (Karl Fischer) Water Determination (NIR)

Microbial Quality (Microbiology) Not Tested routinely - due to process capability and understanding

Assay / CU (HPLC)

Note: Real Time Release Testing does not mean less testing!

24

Specifications for ‘Large n’ UDU testing

Specifications based on zero tolerance* problematic when testing large samples (i.e. >30)‘Large n Counting Test’**

Analyze X samplesAssess number of samples (n) outside of limitsCompare n to defined criteria (c) for allowed number of samples outside of limitsBatch is in control when n < c

*Zero tolerance specifications (e.g. no tablet outside of 75 – 125% of target) discourage intensive sampling to gather process information because the larger the sample size, the greater the chance of an OOS result

**Sandell, D., Vukovinsky, K., Diener, M., Hofer, J., Pazdan, J. and Timmermans, J (2006), ‘Development of a content uniformity test suitable for large sample sizes’, Drug Information Journal 40 (3), 337 - 344.

100%75% 125%

Failing Tablets

25

Product Y - Sampling

Sampling plan for TabletsOC curves used to assess performance of sampling plans‘Large n Counting Test’ adoptedTablets samples taken using stratified sampling plan across batchTablets samples taken at minimum of P points during compression runMinimum of Q tablet samples taken

Drug Content of TabletsTablet is assayedNIR for drug concentration and weight measured on the same tablet

26

31302928272625242322212019181716151413121110987654321

6,5

6,0

5,5

5,0

4,5

4,0

3,5

Location Number

CU S

ingl

e V

alue

s

3,75

6,25

4,25

5,75

Boxplot of CU in batch 810467400XXXXXXXXX

Product Y – Sampling for Content Uniformity

125%

115%

473523118864c

1000750500250200183150100n

‘Large n Counting Test’ Criteria

n = number of tablets sampledc = acceptable number of tablets outside 85% to 115%

‘Large n counting test’:

-- accounts for possibility of tablets outside 85% to 115% when collecting large samples

27

Chemometric Model Maintenance and Update

Periodic Evaluation of Applicability of NIR Model used for UDUNIR model evaluated with every batch (system suitability)Annual evaluation of model

Updating the ModelMechanism - Internal change control procedures used to identify material or process changes with potential to impact the NIR modelNIR measurements may not be valid

Operation outside NIR model limitsNew variability in materials and/or process Fall-back: reference HPLC method used for batch release

Assessment for model update:No impact:

– New variability is unrelated to active (i.e. not significant in the regression coefficients) or is compensated for by model parameters.

– Update model (without changes) to include robustness to this new variability Impact:

– Update the model e.g. include new variability within the model or optimize calibration model parameters

28

RTRt - Training

Quality by Design TrainingGeneral conceptsSite-wide training Culture change?

Quality SystemQuality System Elements revised

DeviationsChange ManagementValidationKnowledge Management

Training on revised SOPs

29

Control Charts

Integrating PAT into Facility Data Management

10X Unit

NIR

Report

MES-Systems

30

Batch Disposition

Batch dispositionBatch disposition process and decision to accept or reject the batch is still needed when RTRt is used (Ref. ICH IWG Q&A)

Batch disposition decision may be facilitated byUse of electronic batch records and exception reporting systems (flag deviations) to provide information in real timeProduct/process monitoring systems (e.g. Statistical Process Control) can provide real time assessment of process performance

Certificate of AnalysisValues from online testing reportedTests not routinely performed: ‘Would comply if tested’

Impact on Qualified Person / Quality Release personRelease process essentially unchanged with RTRtNeed to understand systems supporting the disposition decision

31

Product Y – Regulatory Interactions

EMEAJuly 2006 - MAA Approved via Centralised procedureJuly 2009 - Meeting with EMEA PAT TeamJuly 2009 - Type II Variation for RTRt submittedSeptember 2009 - Questions and ResponsesOctober 2009 - Final Assessment reportNovember 2009 - Approval

FDANovember 2005 - SubmissionMay 2006 - FDA CMC Pilot NDA Approved September 2009 - Prior Approval Supplement for RTRt submittedOctober 2009 - Information RequestJanuary 2010 - User Fee goal date

32

Overview

What needs to be considered when implementing Real Time Release Testing?

Case Study 1 – Product X

Case Study 2 – Product Y

Comparison of some aspects of Product X and Y

Conclusions

33

Comparison of projects – Control Strategy Examples

Product X Product Y

Identification NIR testing of Tablets NIR test of API input into a closed manufacturing system

Related Substances / Impurities

Product/process knowledge shows low risk of generation during manufacture – ‘Would pass if tested’

Product/process knowledge shows low risk of generation during manufacture – ‘Would pass if tested’

Bioavailability / Dissolution

Level A IVIVC – specification and content of rate-controlling polymer

Disintegration used as surrogate for dissolution

34

Comparison of Projects – Common Features

ScienceExtensive use of PAT systems for online monitoring and controlOperational Characteristic curves used to assess relationship between online testing and conventional end-product testing

Sampling plans different

Quality SystemsEvolution, not revolution, of existing quality systemsUse of Quality Risk Management

RegulatoryProducts approved with more conventional end-product testing focussed control strategiesReal Time Release Testing introduced via Variation to Marketing Authorisation

35

35

Challenges and Opportunities -1

ResourcesCross-functional / multi-disciplinary team necessary for successful implementation of Real Time Release testingIncreased need for statistics, control engineering etc. skills for successful PAT implementationCulture/mindset challenges (proactive versus reactive quality) Initial capital commitment is needed for PAT equipment

Quality Systems DevelopmentRobust change management systems necessary Quality risk management

e.g. Need systems in place for PAT equipment failureImpact to QP/ Q release person

Understand control strategy, quality systems etc.

36

Challenges and Opportunities -2

Regulatory challenges Acceptance of alternative sampling plans to pharmacopoeial plansGlobal acceptance of RTRt and harmonization?

BenefitsLower manufacturing costs

Improved yieldsFewer deviations and/or rejectsReduced QC resources

Faster cycle timesIncreased assurance of quality for our patients

37

Overview

What needs to be considered when implementing Real Time Release Testing?

Case Study 1 – Product X

Case Study 2 – Product Y

Comparison of some aspects of Product X and Y

Conclusions

38

Conclusions - 1

Multiple approaches to achieve RTRt are possibleRTRt does not mean less testingRTRt may be applied to new or existing products

RTRt and Quality by Design (QbD)RTRt is a possible outcome of QbD developmentQbD may not be necessary for RTRt

RTRt implementation is supported byProduct and process understandingQuality Risk ManagementA robust control strategy (including PAT and appropriate sampling plan)Science and risk-based quality systems aligned with ICH Q10Discussions with the Regulatory Agencies prior to submissions

39

Conclusions - 2

RTRt and Quality SystemsEvolution, not revolution, of quality systems to enable RTRtPre-define responses to deviations and failures

RTRt benefits may includeLower manufacturing costs and cycle timesImproved yields through less wasteIncreased assurance of quality for our patients

40

40

Acknowledgements

Ferdinando AspesiChunsheng CaiCarlos Conde-ReyesPlinio Delos-SantosParimal DesaiNirdosh JagotaCarl Longfellow Steve SimmonsShailesh SinghMerlin Utter T.G. Venkateshwaran Dom Ventura

John GroskophThomas KatzschnerRoger NosalKarl RedlMark SmithJohn O’Sullivan