Statistically Sound Specification Setting – A Practical Approach Consistent with ICH Q6B and Q14
Key Learning Outcomes
- Understand the three types of statistical intervals: confidence, prediction, and tolerance, and why tolerance intervals with defined coverage and confidence are suitable tools for establishing specification acceptance criteria.
- Learn how batch selection, sample size, and distributional assumptions directly influence the validity of statistically derived acceptance criteria, and why ignoring non-normal distributions can lead to unreasonable specifications.
- Discover how Design of Experiments (DoE), process characterisation, and statistical process control methods connect manufacturing process understanding with defensible, fit-for-purpose specification limits.
- Explore how stability data, internal release limits, and analytical method validation all contribute to a comprehensive, statistically grounded framework for setting specifications in line with ICH Q6B and ICH Q14.
Event Overview
This session, Dr. Eric Rozet offers a practical, statistician’s perspective on one of the most important decisions in drug product development: setting acceptance criteria for specifications. Using the ICH Q6B framework, the presentation clarifies common misconceptions and provides scientists with the statistical reasoning needed to establish limits that are meaningful, defensible, and truly connected to patient safety and manufacturing realities.
The session begins by explaining why acceptance criteria must be based on relevant batch data, especially clinical and process consistency batches, and emphasises the importance of the number and representativeness of those batches. The presentation then explores the three types of statistical intervals, covering confidence intervals, prediction intervals, and tolerance intervals with clear visual illustrations to demonstrate why only tolerance intervals are suitable for setting specifications, and what can go wrong when the others are misused.
A significant part of the session focuses on common pitfalls in statistical modelling: the risk of assuming normality when measuring biologics, the distortions caused by small sample sizes, and the simple log-transformation method that addresses many distributional issues. The session then expands to cover the entire range of statistical tools supporting specification development, Design of Experiments and design space mapping from ICH Q8, statistical process control and control charts for process monitoring, and stability modelling for establishing release limits versus shelf-life.
Eric concludes with a forward-looking view on Bayesian statistical methods, which provide a principled way to incorporate prior process knowledge and avoid the “scientific amnesia” of treating each specification exercise as if it begins from zero, with a long-term aim of linking manufacturing parameters directly to patient outcomes.
Who Should Attend?
Anyone involved in drug product development, quality strategy, or regulatory submissions looking to strengthen the statistical basis of their specification-setting approach.
- Analytical Scientists and Method Development Chemists
- Quality Control and Quality Assurance Professionals
- Regulatory Affairs and CMC Submission Specialists
- Biopharmaceutical and Small Molecule Development Teams
- Validation and Compliance Scientists
- Statisticians and Data Scientists supporting pharmaceutical development
- R&D and Technical Operations Managers overseeing analytical and manufacturing functions
Unlock Additional Educational Resources
Register today for Eric’s presentation and gain access to exclusive bonus content, such as the insightful panel discussions on “Overview of ICH Q6B guideline and current practice on specification settings” and “ICH Q6B & challenges in setting patient-centric specifications”.
Dr Eric RozetLead StatisticianGSK (Belgium)Eric Rozet is the Lead Statistician at GSK (Belgium). He has over 20 years of experience in discovery, pre-clinical, and non-clinical statistics, with a particular focus on statistical aspects related to (bio)assays and (bio)processes, including development, optimisation, validation, transfers, manufacturing, and stability. He trains statisticians and analysts in the Bio-Pharmaceutical Industry on topics such as optimisation, validation, robustness, and the transfer of analytical methods and processes. Eric also trains statisticians in Bayesian modelling and Designs of Experiments. He is the author of more than 100 articles and book chapters in applied statistics and regularly gives conferences on these subjects. Eric holds a B.Sc. degree in Bio-engineering, a Master's degree in Biostatistics, and a PhD in Pharmaceutical Sciences.
| Presenter |
|
Brought to you by:
