While assessing intra-assay coefficient of variation (CV), it is generally advised that CVs should be calculated from the calculated concentrations rather than the raw optical densities.
If the raw optical densities of my samples are just in the middle of calibration curve (see attached illustration - situation A), the curve is quite steep here, and low variation of absorbance reflects low variation of concentrations, and reported intra-assay CV is satisfactory.
But in my case, all my samples are on the lower edge of calibration curve (see attached illustration - situation B). Although I have pretty low variations in raw optical densities of the duplicates (CVs about 5%), variations of the calculated concentrations are much worse (CVs about 20%), due to gentle slope of my calibration curve in this area.
If I report such high CV, it will be perceived as unacceptable, however, my pipetting technique is satisfactory.
What should I do? How should I report my intra-assay CV in order to make it misinformative for reviewers and readers of the article?
Hi Michal,
In my opinion, you cannot express the CV from the DO values because it's inaccurate. The fact that you do work in a standard curve zone with a low slope is itself a part of the measurement imprecision too and should be reflected in the results! It's not only about pippetting imprecision. The actual results are the concentrations so the imprecision expression should relate to the concentrations. You can check it in the different guidelines from FDA or EMA about ligand-binding assays such ELISAs, which are enclosed to this message.
Also my advice would be : (i) to less dilute your samples (if it's diluted) or to concentrate it by any preliminar pre-analytical step if possible; (ii) to increase the number of replicates (3 instead of 2 for instance) so that you may level down your CVs.
I know ELISA can be frustrating sometimes and you can feel like being in a cul-de-sac but an inaccurate expression of CV is not helpful in my opinion.
Good luck,
Cécile
Can you use a larger amount of sample in order to get results more central to the standard curve, or find some other means to improve the sensitivity of the ELISA? For example, you might use a more sensitive method of detection, such as chemiluminescence.
Thank you, Adam.
I have already done my best to increase sensitivity of the assay by adding larger amount of sample, and have currently little access to alternative methods.
Do you think this large intra-assay CV can be acceptable, if explanation provided?
In general, this high intra-assay CV reflects not pipetting imprecision, but the final performance of this particular method for such low concentrations.
Hi Michal,
In my opinion, you cannot express the CV from the DO values because it's inaccurate. The fact that you do work in a standard curve zone with a low slope is itself a part of the measurement imprecision too and should be reflected in the results! It's not only about pippetting imprecision. The actual results are the concentrations so the imprecision expression should relate to the concentrations. You can check it in the different guidelines from FDA or EMA about ligand-binding assays such ELISAs, which are enclosed to this message.
Also my advice would be : (i) to less dilute your samples (if it's diluted) or to concentrate it by any preliminar pre-analytical step if possible; (ii) to increase the number of replicates (3 instead of 2 for instance) so that you may level down your CVs.
I know ELISA can be frustrating sometimes and you can feel like being in a cul-de-sac but an inaccurate expression of CV is not helpful in my opinion.
Good luck,
Cécile
Just to reword and add something to Cécile's excellent advice:
The high CV in the report is not misinformative. It is providing factual information; the accuracy of those results is indeed lower, because they are derived from a part of the curve with a shallower slope. If, as you mention, you cannot use larger samples and do not have access to alternative methods, then you have to explain that in your report, and point out that pipetting technique is OK (as proven by the low CV of the raw OD duplicates). Whether that is acceptable or not depends on your purposes and any regulatory requirements you need to fulfill.
Hope it helps!
Thank you Adam, Cecile and Alejandro for your valuable answers and suggestions!
I have overcome my frustration about poor CVs and will follow your suggestions to report them sincerely and with appropriate discussion.
I found a method to experimentally determine the real LLOQ by making a scatter plot of the %CV as a function of concentration for all tested samples and identifying the lowest mean level above which the %CV < 20% for the greater majority of the samples (Andreasson et al. 2015 Article A Practical Guide to Immunoassay Method Validation
). This method confirms that my LLOQ may be underestimated and requires recalculation.Thank you once more!
- Badges
- Science method
- Similar topics
- Medicine
- Immunology
- Immunology Techniques
- Immunoassay
- ELISA