I think it comes down to the underlying biology of the biomarker being measured. Acknowledging that you may not know a lot of the biology if it's a newer biomarker, but these fundamentals are what guide your validation strategy. Knowing the biological variability of the biomarker such as the diurnal variations, differences in different disease states, races, sexes, and age groups can help shape the scope of what needs to be built into your validation. This will help you determine what is thought of as a "meaningful change" in your patient groups.
From an assay perspective, you build your assay to be able to measure those meaningful changes and validate the parameters to give you confidence in your measurements. Analytical assay performance doesn't need to be analyzed in large detail by statisticians. Typically, these can be monitored via QC performance (both endogenous and Non) to make sure they fall within the expected ranges and are not subject to any trends. Visually inspecting the Levy-Jennings plots and applying understanding from Westgard rule principles allows you to monitor the analytical performance. Then applying your understanding of the biological changes that could occur to your biomarker, you try to interpret your clinical results. However, biologically speaking, generating enough data around a biomarker to justify significant changes will need to be analyzed by a statistician to make sure your population sizes are large enough to draw those conclusions.
Working within a CRO, we work with our sponsor scientists to understand the biomarker of interest and thoroughly discuss what context you want to use the data, ie what changes are you looking for and what would be significant. From there, we work to design the appropriate assay and validate it to be able to make reliable measurements. We monitor the assay performance to make sure it's running acceptably according to our agreed upon context. We then work with the sponsor in helping interpret the data as well as to give insights into the potential limitations of interpretation based off of the assay. This gives confidence in the biological interpretations.
I hope that helps a bit. I may have gone a bit in my explanation.
Thanks
------------------------------
Robert Neely PhD
Scientific Director-Translational Sciences
Immunologix Laboratories
Mt Laurel NJ
[email protected]Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
------------------------------
Original Message:
Sent: 08-28-2023 01:08
From: Anonymous Member
Subject: Determining fit-for-purpose nature based on the COU of a biomarker assay
This message was posted by a user wishing to remain anonymous
Hello,
While I understand the concepts of fit-for-purpose validation in the context of biomarkers and the surrounding literature, I'm facing challenges in practically ensuring that a specific assay meets this requirement. I'm particularly seeking concrete examples of how a biomarker assay can effectively detect meaningful changes for the biomarker in question.
If we have a grasp of analytical, within-subject, and between-subject variability, as well as knowledge of the expected magnitude of change, how can we regularly utilize this data to ensure a fit for purpose assay? Do bioanalytical scientists rely primarily on visually inspecting data, or do they employ routine statistical analyses to determine an assay's detection capabilities? If so, which are appropriate? Is there common reliance on statisticians for data analysis? This becomes more pertinent when collaborating with Contract Research Organizations (CROs) which may have their own validation procedures – do CROs independently assess assay capabilities, or are sponsors providing support in this aspect?
I'm eager to learn from those with hands-on experience and would greatly appreciate the sharing of specific examples.
Thank you.