Help me to calculate LOD, LOQ, please!

[TOT]

Hi all,
I am trying to calculate LOD, LOQ for my method. I can't apply S/N ratio because I can't get the random noise (my method is for a drug in plasma after a pre-column derivatization so the baseline includes a lot of peaks). I found different ways to calculate, for example: mean of the blank + 3xstandard deviation of the blank (s) for LOD and mean of the blank +10xs for LOQ or LOD=3.3xs/slope, LOQ=10xs/slope or LOQ is the concentration at which the RSD is <20%... I really fell confusing, which way I should follow and if I found the LOQ is 5ng/ml from the calculation but I couldn't get the detect at 5ng/ml so could I still report my LOQ as 5ng/ml?
All of your suggestions are highly appreciated! Thanks in advance!

[LY]

Hi!
Did you try LOD/LOQ calculation based on the Standard Deviation of the Response and Slope of the Linear Curves?:

Prepare a stock solution of an analyte with known concentration.
Prepare a series of standard stock dilutions.
Analyze each solution in triplicate, beginning with most diluted and ending with stock solution of analyte.
Perform the least squares linear regression analisis using hte concentration of the analyte as independent variable and its peak responses as the depended variable.
The LOD/LOQ may be expressed as:
LOD= 3.3 d/S and LOQ= 10 d/S where:
d=the standard deviation of y-intercept of the regression line.
S=the average slope of regression lines

[TOT]

Thank you LY for your suggestion. However, could you please show me how to calculate the standard deviation of y-intercept.? I used Excel to do regression anlaysis and couldn't find the standard deviation, it only shows the standard error. ONe more thing is different method to calculate LOD. LOQ give different results so how I have to explain about this? And do I need to see the response at the concentration of LOD? Thanks a bunch!

[lcollard]

Another way to proceed is to make a serie of dilution of known concentration and proceed so.

You make a zoom of the interest zone where you can find your compound.
You calculate N (noise height in mm) H (compound height in mm).

LOD = 2*N/H*C
LOQ = 5*LOD

Where C is the concentration in mg/l

Make sure that the dilution is good enough to see your peak and the noise of the baseline in a good range.

[Lime]

Hi,

I use statgraf for regression analyses . It gives you y-intercept and std.dev. There are many programs like this you can find from internet and download. And.. to validate LOD&LOQ values ? you may prepare three samples includes active subs. at LOD and LOQ conc. Check S/N
( for LOD : min. 3:1 , LOQ : min 10:1 ) and recovery (only for LOQ , 75%-125 % ) after that you can rely on the LOD&LOQ values you found.

Lime

[bert]

In my opinion, you should not determine the LOD/LOQ for a plasma analysis using dilutions of a stock. Prepare calibration samples in the matrix of interest (blank plasma) to perform the calculations!

regards Bert

[LY]

=STDEV(value1,value2,...)

Value1, value2, ... are 1 to 30 values corresponding to a sample of a population. You can also use a single array or a reference to an array instead of arguments separated by commas.

[tom_mizukami]

Hi Tot,

The standard error of a best fit parameter is the same thing as the standard deviation. If you use the LINEST function se b is the standard deviation of the intercept. SEM or the standard error of the mean is a very different thing.

Tom

[bert]

Tot,

maybe it is also helpfull to look at the guidelines for bioanalytical method validation. I know that the FDA as well as the EMEA published these on their websites.

Good luck!

[michaelcarolus]

Hi Tot

We use the following method, amongst others, to determine LOQ/LOD.

Lowest concentration for which RSD is less than 5.0%
This method involves choosing the Quantitation Limit (QL) as the lowest concentration for which the relative standard deviation (RSD) of multiple injections is less than 5.0%. By convention, the Detection Limit (DL) value is taken as 0.3 X QL.

We find this method to be the best for us because we are testing the method at the LOQ.

Hope this helps.

Mike

[TOT]

Thanks all for help! At last I think I will choose 1 suitable method and report the LOD/LOQ with my calculating way, it should be acceptable, shouldn't it?
Dear michaelcarolus, do u think that 5% for the RSD of multiple injections for LOQ is too low? I read somewhere they allow even up to 20%. Anyway, now I think that it really depends on our purpose of our method so there is not only 1 way to calculate these values, am I right?
Thanks all again and best wishes for your researchs!

[michaelcarolus]

Hi Tot

I think it will depend on the type of sample and/or technique you are using. Look at your validation acceptance criteria for accuracy, precision etc. If these limits are quite wide then you might set an RSD limit of 10 or 20%. I work in the pharmaceutical industry and we have found that the using a 5% RSD limit works well for use.

Hope this helps
Mike

[syx]

Thank you LY for your suggestion. However, could you please show me how to calculate the standard deviation of y-intercept.? ... with excel

First, u must calculate the residual stdev, s(r) with formula:
=steyx(known_y's,known_x's)

and then calculate the stdev (error) of the intercept, sc with formula:
=(steyx(known_y's,known_x's))*sqrt((sum(x^2))/(n*(sum((x-x_av)^2)))

n = number of calib points
x_av = means of x values

For complete formula, u can find and download the article "Preparation of Calibration Curve" from vam's website:
http://www.vam.org.uk/
or send me an e-mail to <syx_gf@yahoo.com> and i will send u the article.

Regards,
SYX