An interesting discussion. I think it is worthwhile to recall how the 1% failure rule came to be in the guidance in the first place. Prior to that, sponsors tended to set the LPC rather arbitrarily, and sometimes far above the cutpoint. It might not appropriately monitor the low end of the assay range. Also, a control that never fails is not a useful control. The purpose of any control is to ensure that the assay is performing as it has historically, or how it performed during validation, and to flag us if it isn't.
But there are two types of acceptance criteria for PC. One is that it must be positive; ie, above cut point. In that case, a criteria that the LPC falls below cut point 1% of the time is perfectly reasonable. The other type of acceptance criteria would be to set a 2-sided range, based on either 95th or 99th percentile, similar to how controls in clinical diagnostics are treated. In this case, the LPC will automatically fail 5% or 1% of the time, respectively, regardless of how close to cut point it is.
In our lab we use a combination of these approaches. We set the LPC to fall below cp 1% of the time, but also set up acceptance ranges for both LPC and HPC based on 99th percentile of validation data. Presuming a normal distribution, it should fail low 0.5% of the time and fail high 0.5% of the time*. In some cases the screening LPC ends up being somewhat higher than originally calculated. This could happen due to simple "rounding up", or because it would not confirm otherwise. In this situation, the LPC is still subject to the 99th percentile range, and will fail 1% of the time. For this reason, I think it is acceptable to have a single LPC that is above the calculated screening LPC level, provided it still represents the lower end of the assay range.
*if the low end of the acceptance range is less than cp, then the low end automatically becomes cp and you may have a slightly higher total fail rate. However, in our practice 3 of 4 PCs must pass in order for the run to pass. The probability of two PC falling outside of the range is extremely low absent some kind of problem.
John Kamerud
Pfizer
Original Message:
Sent: 5/1/2023 7:56:00 AM
From: Ritankar Majumdar
Subject: RE: Setting up LPC at a level with less than 1% failure rate
While I agree with @Heather Myler on the course of action and the intent of the ADAH paper (AAPS J. 2021 Dec 1;24(1):4.), please allow me to be the Devil's advocate.
The LPC is set to target 1% failure, which is (t0.01, df) for the one-sided 99% confidence level in the T-distribution. The sensitivity is calculated during LPC determination at 95%. The sensitivity is also the PC conc at which the ADA can be detected reliably and "specifically" in a sample. Hence wouldn't the LPC be associated with specificity and hence the "confirm" level of the assay?
Given that two levels of LPC will add to the logistic and operational burden (as well as lifecycle mgmt), if the confirm LPC sensitivity is within 100 ng/ml and within the expected error rate of the screen LPC, would it make sense to have one LPC level for both confirm and screen rather than two levels?
Lastly, in smaller studies, we seldom encounter batches of > 100 runs; in those cases, the expected failure rate is never achieved; hence for smaller studies the two LPC level is moot.
Just my $0.02.
Ritankar.
------------------------------
Ritankar Majumdar
Team Lead, Lead Scientist
LabCorp Drug Development
[email protected]
Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
------------------------------
Original Message:
Sent: 04-28-2023 07:28
From: Heather Myler
Subject: Setting up LPC at a level with less than 1% failure rate
We have frequently used both 2 different LPC levels for screen and confirm and a single LPC when they are relatively close in concentration where one might be set higher than the 1% failure target. You can reference the ADAH paper, Myler H, Pedras-Vasconcelos J, et al.,. Anti-drug Antibody Validation Testing and Reporting Harmonization. AAPS J. 2021 Dec 1;24(1):4. doi: 10.1208/s12248-021-00649-y. PMID: 34853961; PMCID: PMC8816448, for discussion on this.
What you will want to ensure is that the LPC is appropriately set to monitor your positive study results at low titers. You may need to reset your LPC after you have study data to support this.
------------------------------
Heather Myler Ph.D.
Senior Director
Takeda
Cambridge MA
[email protected]
Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
Original Message:
Sent: 04-27-2023 16:06
From: Anonymous Member
Subject: Setting up LPC at a level with less than 1% failure rate
This message was posted by a user wishing to remain anonymous
Section VII, subsection B of the FDA immunogenicity guidance recommends that the LPC should be set statistically set at a level where it should fail 1% of the time.
However, many times we notice that the screening LPC does not work in the confirmatory assay and the options are to go with a single higher LPC for both the assays (Section IV subsection C.1 Assay sensitivity) or have separate LPCs.
So, there is precedence for going with an LPC that would fail less than 1% of the time in the screening assay.
So, what are the implications of having an LPC with a lower rate of failure than 1%, both assay wise, and regulatory wise?