Here are my two cents.
if you are using an ELISA based kit, usually the standards are made in surrogate matrix. For example, the recombinant protein will be spiked into a buffer from the kit at the specified concentrations. Then, the kit will usually suggest that you dilute to a specific dilution, such as 1:100, for the samples.
I think it is helpful to think about the goal of a parallelism assessment. You want to ensure that when you are assessing different dilutions of your sample-presumably in matrix-that the behavior as the sample is diluted will match the behavior of the curve. Otherwise, if you dilute samples to different dilutions you would be getting inconsistent results. This consideration is similar to how you would approach parallelism for standard PK assays. Also, in the case of biomarkers, observed parallelism against the buffer curve across multiple samples suggests that you are not seeing matrix interferences that are interfering with the sample quantitation. As you are dealing with the endogenous analyte, this may be more relevant to understanding matrix effects than the spiking of recombinant proteins, in my opinion. Also, with samples that have high endogenous levels of biomarker, spiking them with appreciable amounts of recombinant protein analyte may not be feasible, and so this test is an important one to understanding matrix effects in a biomarker validation, in my personal opinion.
You can select a range of dilution factors for your samples in the parallelism test, both less dilute and more dilute than the recommended kit dilution. You may find that at small/low dilutions (I.e., 1:2) the behavior of the sample differs from the behavior at other dilutions, and based on this you may want to restrict the range at which you can analyze samples. But you aren't necessarily restricted to the dilution the kit recommends. That is why the independent validation of the kit is so critical-you can't be sure of the criteria used to set the recommended dilution. Usually I assume that the kit dilution is set based on the average concentration of the analyte in samples, such that most samples will fall in the mid range of the curve. I will say, from a practical perspective, usually the concentrations of analyte in your natural samples are going to constrain what dilutions you can test. For ELISA assays in particular, the dynamic range is small. So you probably are going to run up against one end of the curve or the other as you dilute. For samples with low concentrations of analyte relative to the curve, this can be a problem. You might consider spike recovery experiments to assess matrix interference in such samples and you may not be able to get a good data set of dilutions before you hit the LLOQ of the curve.
This also requires some care when you set up your sample analysis. Make sure method instructions are very clear as to what samples get a dilution and what do not. These assays can be very confusing since they may not have the same setup as a standard PK-especially if the calibration curve is prepared "at concentration" and then added to the well, which is common. It is also worth noting that in some kits, you add the samples to the plate and there is an additional "dilution" from adding a kit reagent to the well-that would be applied evenly to all samples. So you might have an initial sample prep THEN a step in the well that the kit calls an MRD that is applied to all samples on the plate.
Finally, you have to be careful when reporting results to make sure that you are accounting for any additional dilutions that the sample went through as compared to the calibration curve when back calculating the concentration. This can be very confusing, to be honest, so I highly recommend drawing it out and establishing a standard approach for how the dilutions will be described as well as entered into the LIMS. It is common that standards will be prepared fresh at the final concentration and then the sample will get a dilution factor, and so any dilution that is applied to the samples but not the standard should be accounted for when you calculate and report the final concentration of the sample.
QCs can be done in many ways but I would strongly recommend some endogenous QCs in addition to recombinant protein from the kit. Otherwise, when you have a lot change, it will be difficult to discern. Recombinant protein QCs can be prepped like the standards. Endogenous QCs can be run at different dilutions and you can establish a concentration empirically-you establish your own theoretical in A & P so you don't have to worry about dilution factors etc.
Biomarker tests are done in different ways for fit for purpose approaches so I welcome other answers here and there may be differences in approach.
------------------------------
Catherine Vrentas
[email protected]Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
------------------------------
Original Message:
Sent: 10-07-2024 13:35
From: Anonymous Member
Subject: Parallelism in Biomarker Assay
This message was posted by a user wishing to remain anonymous
A recent discussion came up on biomarker parallelism and was wondering what has been the standard practice in the industry. Do you perform parallelism with neat sample even though the kit recommends a DF or has an MRD?
For example, if the kit recommends a 100 fold dilution, do you perform parallelism at DF=1, 2, so on or at DF=100, 200, ..? In another example, if the STDs, QCs and samples have a MRD of 5, do you perform parallelism at DF=1, 2, so on or DF=5, 10,..?
Thank you in advance for your response!