I’m posting this message from Patty Erickson at Vanderbilt for assistance from any of our RAQA gurus participating in the blog:
Okay, I’m just not a chemist. I admit it. On my last CAP inspection I was cited for not performing AMR validations on my ELISA assays for anti-cardiolipin and beta-2 glycoprotein antibodies because I report these numerically for patient samples, as opposed to neg/pos. Here’s the question. I have struggled with finding appropriate material to challenge the analytic range for each isotype and after speaking with a clinical chemist, it was suggested that I could use calibrators from a previous lot number kit (there are three different calibrators for each kit type) to challenge the AMR for a new lot number. That is what I am doing. But, since these kits are intended as semi-quantitative assays, the calibrators with assigned values from one lot number kit might not be expected to perform identically with a different lot number of the same type kit. What is my “target” for each calibrator, and what is my acceptance criteria ? Lastly, in general, what is the “industry standard” for deeming an analytic range “validated” and if I fail this criteria, exactly what does one do for reporting off that particular lot number? Have I confused anyone yet, besides myself? I am trying to satisfy those inspectors, but I am also trying to observe good laboratory practice. If anyone has specific information on how they go about doing this validation for ELISA assays, I would be so very, very appreciative!
Patty, I’ve sent this to a couple of quality assurance experts I’m acquainted with and will hope for an authoritative answer. I may contribute one comment, that when validating from lot to lot, I’ve included low, normal, and high patient specimens in the protocol, as well as controls from the kits. By including low and high you test the analytical range, and if the two sets of data reproduce within your chosen range, the new lot can be placed into service without adjusting the reference interval.
Experts, can you help?