You have two different types of endpoints. One is the lower limit of detection; which is the lowest amount of antigen you can reliably distinguish from a blanc. The lower limit of detection is defined as blanc + 2 times the Standard Devation of that blanc (measure your blanc 10times).

The other limit is the lower limit of quantification, which you can define through the same experiment as above (measure 10 times blanc), but now add 10 times the Standard Deviation.

A generally accepted method is also to dilute your analyte and see at which point the duplicate CV is no longer acceptable (say beneath 15%). The concentration at which you can no longer reliably quantify (CV becomes >15%) is your limit of quantification. Tricky thing here is that you can choose your own CV that you find acceptable.

Thanks Wesley, I know that... although there are papers that define more strictly the cutoff calculation. (check: http://www.ncbi.nlm.nih.gov/pubmed/?term=Frey+cutoff+elisa)

my question is whether once  the blanc + 2 times the Standard Devation is defined, do I need to fit my data to a function and interpolate that value in the "Y" axis (absorbance) to obtain the X value (serum dilution). Thanks!

There's nothing in that file concerning my question. Thanks.

Hello Beatriz,

Are you attempting to target the titer more precisely, without doing many, many dilutions (such as 1:1000, 1500, 2000, 2500, 3000, ......)?  I would think that using regression of your data would work for that.  As you probably know, titer as used to determine strength of antibody response depends entirely on the other assay parameters, and comparing pre-immune to immune is usually evaluated for a ###-fold increase.  That said, if you are trying to more precisely determine the presence of antibody, then I would not see a reason to not use regression analysis if you have a solid titration data set.

Yes, that was the question. Not always have we used the same dilutions to test different sera, and I needed a way to make them comparable. I will do regression analysis. Thanks!!

Titration depends largely on the technique used. Stepwise dilutions have to be tested; the easiest way is to perform serial log2 dilutions; each new dilution starts from the previous one (not from the original undiluted sample). The last dilution needs to reach the signal equivalent to the background. Th every last dilution that gives twice the signal of the background indicates the titer. For more precision you can read a method I published back in 1986 (Zrein et al.).

I do not have access to those publications, would you be so kind as to send them to me?

Thanks!

I can't get it from online sources; try via your institutional access:

J Immunol Methods. 1986 Mar 13;87(2):229-37.

Quantitation of rheumatoid factors by enzyme immunoassay using biotinylated human IgG. Zrein M, De Marcillac G, Van Regenmortel MH.

Hello Beatriz Apellaniz,

Did you find the solution? Did you use a program described in Zrein's article? I also need to determine ELISA end point titers for human sera and I can't find a program or the way to calculate it. 

Hello Anna,

This is one of those topics which everyone under the sun has an opinion about.  There is no distinct method for calculating the EPT.  Since the ELISA was widely used to ascertain Ab presence, people have published ideas about calculating the titer.  I know you probably know this already, but the titer is simply expressing the point at which a dilution no longer has enough Ab to produce a response against antigen.  But there isn't a standard amount of antigen (too many antigens to have positive controls for each), and so you are basically trying to quantify the Ab response with respect to the available antigen.  Mainly to compare to other antibody-antigen responses, right?  The Frey paper from the 90's is probably as good as any.

I've always considered that when you picture Ab-Ag responses, as a reversible equilibrium reaction, then what do we say is a no response?  When the Ab response is not strong enough to have a signal generated in an assay?  Sounds kind of arbitrary to me, and unless the coating of the wells was the same, and the conjugation or secondary were the same, from one assay to the one three weeks from now, after bleed #2 or whatever, then the end-point is arbitrary at best.

I personally think that a four-point logistic regression of multiple dilutions (8 as a starting point) allows one to see the point at which the regression plot swings out of the linear range.  I think it is clear that is where the dilution is hampering the signal generation.  But again, the other parameters of the assay greatly affect where that point will be.

Good luck with your project.  Onnea!

Dear Anna and Stephen,

The Frey paper is the one that I found when I was struggling with this. I finally decided to follow their advices to calculate the cut-off and then used the fitting to a four-point logistic, using some of the negative control wells. I used for this Sigmaplot, but I guess you could do the same with any other program you might have available, Origin for example. 

My opinion, similar to Stephens, is that it is a bit arbitrary. However, if you use it to compare sera that bind to the same peptide/protein plated in the wells, it should be fine. (you should also titrate the amount of peptide/protein plated, to check that it is saturated or maintain always the same concentration).

Good luck!

Dear Stephen and Beatriz,

Thank you for reply. I know that 4PL curve is the most common curve recommended for ELISA and I have tried it. The results look great if at the last dilution OD value was about 0,250-0,300. But in some samples which were taken after immunization, the antibody titer is very high and even at the last dilution OD value is 0,600 or higher. In this case I can't determine the titer at OD 0,200 (estimated cut-off) with 4PL. Others curve doesn't seem to fit to the curve pattern. Do you have any ideas to solve the problem? Unfortunately all the tests have done and I can't include more dilutions.. 

This conversation was over 3 yrs ago but I believe the calculation of end point titers from ELISA is always a current issue. The best answer given was by Maan Zrein's 3 yrs ago :) This is briefly how I do: First you need to plot Abs (Y axis). vs Log2 (log base 2) of your actual dilutions (X-axis) in graph. Using the Abs. vs Log2 in the linear range, find the point that the linear line crosses the cut-off line (2X Abs of negative control) at X-axis. That is the Log2 of your end point titre. For instance, the linear line crosses the cut-off line on X axis at 13.5. Then your end point Ab titer for that sample is= 2^13.5 = 11,585.