Dr Christopher Burgess, Burgess Analytical Consultancy
Dr Joachim Ermer, Sanofi-Aventis
Statistical calculations and tools are applied extensively in pharmaceutical analysis including ;
- Procedure development and validation
- Transfer of analytical procedures
- Setting or verification of specification limits
- Data evaluation, comparison and trending
The ICH Q10 Guideline “Pharmaceutical Quality System”, the FDA Guidances on Process Validation and Methods Validation require monitoring of “process performance and product quality” and “Trend analysis on method performance” throughout the product lifecycle. Hence the appropriate use of statistical trending and evaluation tools has become mandatory.
Consequently, a thorough understanding of statistical fundamentals is essential in order to be able to select parameters and test methods that are ‘fit for purpose’.
Do you speak statistics?
In addition, such an understanding facilitates the communication with other technical and regulatory functions applying statistical tools in order to ensure an overall consistent approach.
The course will provide the participants with recommendations, tools and examples to apply scientifically and pragmatically sound statistical principles to their day-to-day business as well as to meet future challenges described above.
The relevance of such statistical tools is also increasingly recognised by the Compendia, as reflected, for example, in the USP General Information Chapter <1010> “Interpretation and treatment of analytical data” and the recently introduced <1033> “Biological assay validation” together with USP Medicines Compendium, <10> “Assessing Validation Parameters for Reference and Acceptable Procedures”.
Statistical tools are needed, for example, to evaluate:
- Distribution of data and its parameters
- How to detect outliers and trends?
- How to establish the total variability of the method?
- How to identify method parameters that must be controlled?
Method performance and specification limits
- Which accuracy and precision is needed to achieve an acceptable risk of OOS results?
- Scientifically based justification and optimisation of the reportable result (single or average?)
- What are the requirements for impurity methods?
- Comparison of methods and data
- What are the requirements for calibration models?
- How to optimise the number of calibration replicates on a scientific basis?
A brief discussion of supporting software tools (e.g. Excel, Minitab, JMP) to facilitate the generation of statistical information in a consistent manner will be undertaken.
One of the main features of this new course is the balance of presentations and more than five hours of practical exercise workshops which will allow participants to gain ‘hands on’ practical experience in applying the statistical methods described. By means of statistical simulation tools, the participants will gain intuitive understanding of the consequences of appropriate and inappropriate performance parameters, for example the relationship between precision and OOS results.
For this reason, the course is limited to 30 participants so that individual attention and support can be given. In order to fully benefit from the workshops, attendees should preferably bring a notebook with Excel® 2007 or later.
This best practice oriented course is designed for analytical laboratory managers and their colleagues charged with the day to day management and evaluation of laboratory data throughout the lifecycle, i.e. in method development, validation, transfer, specification setting, batch release and stability, continuous performance verification and change control.
QA, manufacturing and regulatory affairs professionals will benefit from participation by gaining a clear understanding of the statistical fundamentals which are important to implement scientifically sound and pragmatic tools to conform to GMP and regulatory requirements for example Product Quality Review.
Analytical Procedure Lifecycle Management (USP & ICH initiatives)
(Normal) Distribution of Data and its Parameters
- Principles of APLM
- Proposed USP <1220>
- Risk based approach
- Target Measurement Uncertainty
- Decision rules
Calculation and Evaluation of Precision Levels
- Data shape and its importance
- Characterisation of distributions (Location and Dispersion)
- Probability considerations; all measurements are subject to error
- Populations and samples
- Confidence intervals
- What is an outlier?
- Error of the error
Trending of Data
- System precision, repeatability, intermediate precision, reproducibility
- ANOVA: Identification of relevant variance components from injection, measurement, sample preparation, intermediate conditions
- Total variability: precision of the reportable result and its optimisation
- Optimisation of single-point calibration
- Relationship between precision and probability of OOS results
- Practically relevant acceptance criteria for precision
Monte Carlo simulation of Analytical Procedures
- Why trend?
- Evaluation; do we expect a trend or not?
- Statistical Process Control principles
- Types of Control charts and their application
- Application to stability testing
Comparison of Data & Accuracy
- Principles of Monte Carlo simulation
- Understanding variance contributions and how they combine
- Measurement uncertainty
- Application to analytical procedures
- Examples of unit and complete procedures using Companion by Minitab
Calibration Models, Linear and non-Linear
- Significance (F- and t-test) and equivalence tests
- Statistical significance and practical relevance
- Differences caused by random variability: observed and true bias
- Applications in transfer and cross-validation
Performance Requirements for Impurity Procedures
- What is a calibration model?
- What is the difference between linear and non-linear models?
- The principle of least squares and why it is important
- Applying the principles to linear and non-linear models
Summary Workshop & Discussion: Appropriate Choice of Tests/Calculations
- Concentration dependence of precision (Horwitz relation)
- Detection and Quantitation Limits
- Practical objectives and data sets are provided
- The participants will discuss and define appropriate tests and parameters to be calculated
- The participants are given the calculation results and are asked to make an evaluation
- The defined tests and results are discussed in the audience
Understanding the Variability (Statistical Simulations)
Range of expected data
Variability of standard deviations
Number of data and reliability of calculated standard deviations
Optimisation of Variability
Statistically based format of the reportable result (single or average)
Number of determinations for various levels
Probability of results outside established limits
Control Charts & Trending
Interactive workshop based on supplied real data sets for interpretation
Use of Minitab for control charting
Team working on evaluation and interpretation of trend data
Comparison of Data (Statistical Simulations)
Significance and equivalence tests: influence of number of data and series
Differences between means and variability
Linearity (Statistical Simulations)
Regression range and evaluation of the intercept
Basics to consider for calculation from linearity
How to determine appropriately from precision