Determination of LOD/LOQ

[Remu]

Hye,

I have a problem regarding determination of LOD and LOQ in olive oil. I did the validation for PAH's and couldn't find any blank oil matrix. That's why I had to use "normal oil" and use blank reduction in my results. It's ok with everything else except LOD and LOQ determination. I can't determine them using blank samples so what could be the next best way to determine them?

[tom jupille]

You should be able to determine LOD and LOQ without a blank by using the slope of the calibration curve and the standard error of the intercept.

[Remu]

Thanks for the help, I tried that. Still the results for LOD and LOQ are unrealistic small.(With the blank sample measurements the LOD and LOQ are too high.) I count them against matrix curve which is of course made of using the same oil as the blanks. I suppose I have to take the LOD and LOQ directly from the calibration curve without any calculations. I just wonder why it doesn't work using the linearity curve.

[tom jupille]

Something puzzles me here. What is the distinction are you making between "linearity curve" and "calibration curve"?

[Remu]

A calibration curve is a general method for determining the concentration of a substance in an unknown sample by comparing the unknown to a set of standard samples of known concentration. The calibration curve is a plot of how the instrumental response, the so-called analytical signal, changes with the concentration of the analyte (the substance to be measured). The calibration curve can be either linear or non-linear. I made both matrix calibration curve (a series of standards in oil) and standard calibration curve (a series of standards in solvent). Hope this clarifies

If you can specify how to calculate LOD and LOQ of my PAH oil method having the validation data available I would be very grateful.

[tom jupille]

Okay, now I understand

What you are calling the "matrix" calibration plot is the appropriate one to use. Assuming a normal least-squares fit, LOD can be estimated as 3.3 times the standard error of the y-intercept (you can get that from Excel) divided by the slope of the line. The LOQ is ten times the standard error of the intercept divided by the slope.

In the case of LOQ, it should be verified by running replicates (five is probably enough) at or near the LOQ to verify that the reproducibility is, indeed, adequate.

A better (but more time-consuming) way to determine the LOQ is by the use of a CV vs A plot. See this thread for a discussion of that approach:
viewtopic.php?t=12738