Benefits of Applying QbD Concepts to Analytical Methods for Therapeutic Proteins Case Studies Annick GERVAIS, PhD Analytical Sciences Biologicals, UCB CASSS AT Europe, Vienna, 17 March 2016
QbD for Analytical Methods Why? 2 Analytical methods are key elements of the Control Strategy (ICH Q10) to determine and measure (Critical) Quality Attributes. Understand performance QbD will bring the systematic methodology to ensure the right method at the right time Sources of Variability QbD of method Robust Life Cycle Management
QbD for Analytical Methods What does this mean? 3 Process Target Product Profile // Analytical method Analytical Target Profile Analytical Method LifeCycle Management 1 Critical Quality Attributes Risk Assessment Critical Method Attributes Risk Assessment Stage 1: Method Design & Understanding Design Space Method Operable Design Region (MODR) Control Strategy Continued Process Verification Control Strategy Continued Method Verification Stage 2: Method Performance Qualification Stage 3: Continued Method Verification ICHQ8 Pharmaceutical Development ICHQ9 Quality Risk Management 1 P. Nethercote et al. Pharm. Tech (2010), 34, 2; USP Stimuli to the Revision Process on Lifecycle Management of Analytical Procedures: Method Development, Procedure Performance Qualification and Procedure Performance Verification (2013)
Analytical QbD in Practice 4 Analytical Target profile Method Performance Acceptance Criteria Critical method attributes Risk assessment MODR* Control Strategy DoE Predictive rather than descriptive approach Continued method verification Using trending tools & predictive tools *MODR = Method Operable Design Region
Stage 1 Method Design & Understanding
Analytical Target Profile 6 Definition The ATP defines the objective of the test & quality requirements for the reportable result. It is a prospective summary of the required characteristics of the reportable result that needs to be achieved to ensure the data is fit for purpose. Example : CEX-HPLC method for charge variants of product X The method must be able to determine the relative quantity of monomer peak and charge variants (acidic species (APG) & basic species (BPG)) in DS and DP samples. The method must be: - Specific, no interfering peak from buffers / matrix observed at the retention time of the isoforms - Accuracy profile: acceptance limit 30% at 5% risk for monomer and 50% at 5% risk for APG & BPG. - QL of APG & BPG must be at least 5% - Prepared sample must be at least stable for 72 hours at 5±3 C - Stability-indicating
Technology Selection 7 The method performance requirements defined in the ATP will guide the technology selection. It is key to consider also business drivers: - Cost - Analysis time - Supply continuity - Applicability in different QC labs and different regions
Critical Method Attributes 8 Identification of the critical method parameters: Start from prior knowledge on similar methods Use Ishikawa tools to classify the method parameters C N Classification of Attributes can be Controlled Noise cannot be controlled/ predicted Autosampler temperature Degassing of solution HPLC vials/caps Pipette technique Samples Method Run time Flow rate Syringe draw rate Solvent composition Buffer/Samplepreparation (dilution,...) Eppendorf tubes Solvent (salts, Water,...) Pipettes tips Column temperature Column conditioning Injector volume Shutdown method Column rinsing Column storage + injection number Filter HPLC material (fittings, tubing,...) Magnet stirrer Weighing materials Column / guard column Control/ref samples Calibration solutionsfor ph Gradient mode (comp, slope,...) Detection wavelength Sampling rate Sequence Reagents (mobile phase) General lab glassware Software use Prepared sample stability Manpower Data handling Pipette technique /Lab Handling Equipment preparation (rinsing step,...) Integration (manual/automatic) Control chart Measuring Cell T C Processing Method Instrument use (column installation,...) Method use Manual Integration Sample acceptance critera Calculation Automatic Integration SST (control sample + blank) Power Grid Environment Vibrations Detector Balance Magnetic stirrer Purified water system Fridge/Freezer Pipettes Vacuum filtration system Humidity HPLC autosampler HPLC injector Temperature Light Void volume Degasser ph meter Vortex mixer Ultrasonic bath Timer Software (comparability) X Instrument qualification HPLC pump pressure and flow rate capacity Column oven Experimentally defined Quality of analytical method data Material Measurement Instrument
Probability Risk Assessment Use of FMEA 9 Risk = Relevance x Probability Relevance 2 4 6 8 10 2 4 8 12 16 20 4 8 16 24 32 40 6 12 24 36 48 60 8 16 32 48 64 80 10 20 40 60 80 100 Risk value Table (relevance x probability) Effect Value Mitigation Color Low x 12 Optional Green Medium 12 < x < 40 Recommended to mitigate if possible Yellow High 40 Must mitigate Red
Risk Assessment Use of FMEA 10 Risk = Relevance x Probability Relevance Table Score Effect Description 2 Negligible Very low possibility of an impact on the quality of analytical method data 4 Minor 6 Moderate 8 Significant 10 Severe Slight possibility of an impact on the quality of analytical method data Possible impact on the quality of analytical method data Likely impact on the quality of analytical method data Strong likelihood of an impact on the quality of analytical method data Probability Table Score Effect Description Not certain that this will ever happen. Chances that it occurs one day are zero or close 2 to zero. Very unlikely For example: 1x per 1000 reportable result or 0,1% chance that it happens. 4 Unlikely Not certain that this will ever happen. Chances that it occurs one day are very low. For example: 1x per 100 reportable result or 1% chance that it happens. 6 Possible Not certain that this will ever happen. Chances that it occurs one day however are real. For example: 1X per 50 reportable result or 2% chance that it happens. 8 Likely It is certain that this is happening. Estimated frequency of its occurrence are estimated For example: 1x per 20 reportable result. or 5% chance that it happens. 10 Very likely It is certain that this is happening. Estimated frequency of its occurrence are estimated For example: 1x per 5 reportable result or 20% chance that it happens.
Example of CEX-HPLC method for Charge Variants 11 Mitigation plan Initial scoring Scoring after mitigation Category Method Attributes Potential Failure mode Instrument Environment Manpower Measurement Material Method Probabilit Risk Relevance y scoring Potential impact on method Classifica Relevan Probabi after after after performance tion ce lity Risk scoring Mitigation mitigation mitigation mitigation 8 4 32 8 4 32 8 4 32 8 4 32 6 6 36 6 6 8 4 32 8 2 16 8 4 32 8 2 16 8 4 32 8 2 16 8 4 32 8 2 16 8 4 32 8 2 16 10 6 60 10 4 40 8 4 32 8 2 16 10 2 20 10 2 20 4 2 8 2 2 4 6 2 12 6 2 12 8 2 16 8 2 16 6 2 12 6 2 12 8 8 64 8 4 32 10 6 60 10 4 40 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 8 6 48 8 2 16 0 0 0 0 6 4 24 0 8 6 48 6 2 12 10 6 60 10 4 40 0 0 8 4 32 8 2 16 8 6 48 0 8 6 48 0 0 0 0 0 0 0 8 4 32 8 2 16 8 6 48 0 8 4 32 0 6 4 24 0 6 4 24 0 8 4 32 8 2 16 8 4 32 6 2 12 8 4 32 0 10 10 100 0 10 10 100 0 0 6 6 36 0 6 6 36 0 10 10 100 0 6 6 36 6 4 24 6 6 36 0 8 6 48 0 6 4 24 0 8 4 32 8 2 16 8 4 32 8 2 16 8 4 32 8 2 16 6 6 36 6 4 24 For method attributes that can be experimentally defined (X), definition of MODR using Designs of Experiments (DoE) Iterative process
Critical Method Attributes - MODR 12 Design of experiments (DOE) is a test or series of tests in which purposeful changes are made to the input variables of a process so that we may observe and identify corresponding changes in the output response from Douglas Montgomery Introduction to statistical quality control ATTRIBUTE X Amount of enzyme DigestionTemperature Digestion duration Analytical method RESPONSE %Peak Area Experiments/assays variability.. NOISE (N) Use of DoE to: Evaluate effect of the most influencial parameters Identify the interactions between parameters Optimise the best operating condition settings
Critical Method Attributes - MODR 13 Screening designs Influent factors determination/ranking Eg Plackett & Burman Screening design Main influent factors determination Factorial designs Factors effects/interaction characterization Optimisation design Main factor & interactions Response surface designs Prediction in a domain Robustness design Small variations Eg Central Composite Design
Screening Design 14 Example of RP-HPLC method for a product related impurity Factors (chromatographic conditions): %TFA in mobile phase A %TFA in mobile phase B % ACN in mobile phase B % IPA in mobile phase B Flow rate Wavelength Column temperature Responses: %product related impurity Model: Plackett & Burman (only main effects) 12 runs Run %TFA in A %TFA in B %ACN %IPA Flow rate 1 1-1 -1 1 1-1 1 l Colum n T 2 1 1 1-1 -1-1 -1 3 1 1-1 1-1 -1 1 4 1-1 1-1 1 1-1 5-1 -1-1 1 1-1 -1 6 1 1 1-1 1 1 1 7 1-1 -1 1-1 1-1 8-1 1-1 1-1 1-1 9-1 1 1-1 1-1 -1 10-1 -1 1-1 -1-1 1 11-1 -1 1-1 -1 1 1 12-1 1-1 1 1 1 1
Screening Design Example of RP-HPLC method for a product related impurity 15 Prediction intervals chromatographic conditions Prediction intervals - sample preparation
Optimisation Design 16 Example of size variant method (SE-HPLC) Factors: NaCl in mobile phase [250 mm 350 mm] NaPO4 in mobile phase [90 mm 110 mm] ph of mobile phase [6.8 7.2] Column temperature [27 C 33 C] Responses: %HMWS % main peak ("monomer") %LMWS Model: Central Composite Design with 3 central points Full Factorial Design with replicate points 22 runs
Stage 2 Method Performance Qualification
Variabiliy Variabiliy Mean Mean Method Validation 18 From Descriptive to Predictive Approach Method Driven classical validation Data Driven Total Error % Bias< 10% % CV< 10% % Bias< 10% Will the method provide good results? % CV< 10% «Good» methods do NOT necessarily provide «good» results «Good» results can only be obtained by «good» methods What is important is the result, not the assay!
Method Validation 19 From Descriptive to Predictive Approach Total Error µ T x i - µ T = Systematic Error + Random Error = Bias + Standard Deviation = Trueness + Precision = Measurement Error = Accuracy 1 1 Accuracy = the closeness of agreement between an individual result found and the true value
Method Validation 20 From descriptive to Predictive Approach β-expectation tolerance limits 1 Acceptance Limits Relative bias The method is considered accurate within the range for which the accuracy profile is within the predefined acceptance limits. This Total Error Approach gives the guarantee that each future measurement of unknown samples is included within the tolerance limits with a given risk level (usually 5%) 1 The β-expectation tolerance interval is the interval wherein each future measurement will fall with a defined probability β. It represents the location where β% of the future results are expected to lie.
Method Validation by Total Error Approach 21 Example 1 : Validation of a RP-HPLC method for product related species Reportable result: %area of product related species X Risk = 5% Acceptance limits = 35% Expected %product related species X Expected %product related species X Use of E-Noval software - Arlenda
Method Validation by Total Error Approach 22 Example 2 : Validation of HCP ELISA assay Risk = 5% Acceptance limits = 30% LQL = 10.6ng/mL Use of E-Noval software - Arlenda
Stage 3 Continued Method Verification
Analytical Control Charts 24 Control strategy includes the use of control charts as follows: Use a control sample in each analytical run Report the parameters of interest measured on the control sample: Reportable result from the method Resolution, etc.. Trend these parameters using control charts Benefits of this control strategy: Determine if results performed on a routine basis are/remain acceptable for the intended purposes of the method. Allow anticipating drifts in the analytical methods. Allow comparing the performance of a method over time and also between laboratories/testing sites.
Analytical Control Charts 25 Example of Exponential Weighted Moving Average (EWMA) charts. EWMA upper limit UCL EWMA line LCL Date of analysis (chronological order) EWMA lower limit Trending Rules : Value outside of [LCL; UCL]: invalid analysis EWMA line crossing EWMA limits: out of trend LCL, UCL temporary fixed after 10 runs, and permanently fixed after 30 runs.
Conclusion
Conclusions 27 QbD for analytical methods is a systematic methodology based on three stages: Stage 1 method design and understanding: ATP, risk assessment, DoE to define the MODR for critical method attributes Stage 2 method performance qualification: Use predictive rather than descriptive approach: total error approach Stage 3 method continuous verification Use of analytical control charts for method performance trending It is an iterative process Applying QbD to analytical methods shows clear benefits in terms of: Method understanding Method robustness Ensuring to produce consistent and reliable data throughout the method lifecycle
THANKS TO 28 Method Development Team Aurélie DELANGLE Christophe BEAUFAYS Cyrille CHÉRY Grégory SCHITTEKATTE Jérémie CUISENAIRE Julie BRAUN Marc JACQUEMIN Marc SPELEERS Sandrine VAN LEUGENHAEGHE All other team members Statistician Team Bianca TEODORESCU Dimitris GAYRAUD Anastasia KOKOREVA Carl JONE Lance SMALLSHAW Chinedu MADICHIE