Statistics for Bioassay

Flexible 2 Day Course

This flexible course covers the fundamental elements of biostatistics as applied to bioassay. It is designed for scientists, and moves through the basics of analysis and assay development, through to assay optimisation, validation and long term management in routine use. The course will provide a solid background knowledge of the statistical methodology used sufficient to understand and resolve issues that can cause problems with real world data analysis.

This module starts with an explanation of what relative potency is, and why it has become accepted by both the industry and Regulators as the standard way of reporting potency. We will introduce the concept of parallelism and how this is applied as well as discussing the issues and what to look out for in applying this approach.

This module moves on to discuss different types of biological responses and the mathematical models that are used to analyse them. Choosing the best model is covered for both continuous responses (with reference to linear, 4 parameter logistic “4PL” and 5 parameter logistic “5PL” models) and binary responses (probit, logit), and touching on the use of survival models. The practice of transforming responses and how to do this to improve model fit and homogeneity will also be discussed

The concept of parallelism introduced in module 1 is discussed in more detail, together with the tests (significance and equivalence) that can be used to test that samples are parallel to the reference. Common problems with these tests (unexpected “fails”) are explained with reference to real data and examples.

The differences between the aims of system suitability criteria (checking that the assay is working) and sample suitability criteria (checking that the test sample is behaving as expected) are explained. You will also be introduced to choosing appropriate criteria and how to use current and historical data to set limits for these criteria.

Starting with an explanation of the different types of outlier, the module goes on to explore why outlier management is a controversial issue, considering practical and regulatory viewpoints. Mathematical ways of detecting potential outliers are described and the module concludes with some practical suggestions for creating an outlier management policy.

Biological assays are very variable and understanding the sources of, and factors contributing to this variability is vital to optimising the assay design, to achieve low failure rates and high throughput. How statisticians use variance components analysis to distinguish between within-plate, between-plate and between-run variability will be explained by reference to real data. The module concludes with a brief look at how this information can also be used to simulate new assay designs to aid optimisation.

Assay validation involves defining and characterising the accuracy, linearity, intermediate precision, repeatability, range and specificity of the assay. These terms are explained in module 7 and module 8 explores how historic data is used to develop an assay validation study plan, and how the results of the study are analysed and presented for regulatory use.

Monitoring the performance of an assay over time is important both from a practical and regulatory point of view. Identification of critical parameters, trend analysis and statistical process control measures will be explained with reference to real data. The relationship of these parameters to batch release specifications and assay management will be discussed.

Transferring an assay to a different laboratory involves many practical issues, and once set up at the new location the assay must be re-validated. Whilst the process is similar to the original validation, careful examination of the available data and transfer planning from a statistical point of view will reduce the experimental work required to a minimum. Similar issues occur with reference bridging, and practical ways of managing this issue in long term assay use will be explored.