Chromatography Forum

Hello,

I want to get some expert's opinion about how you usually design a calibration curve in a HPLC test. I find limited information about how to do it correctly. AOAC said equal distance design (add different amount of stock solution not weigh standard separately) is better because a true serial design will bias the weight comes from the point with higher concentrations.

For example, which one do you use and why you use it?

A true serial dilution design (made by diluting the last level solution):
1ug/ml, 2ug/ml, 4ug/ml, 8ug/ml, 16ug/ml

An equal distance design (made by adding different amount of stock):
1ug/ml, 5ug/ml, 9ug/ml, 13ug/ml, 17ug/ml

A mixed type of design (made by adding different amount of stock):
1ug/ml, 3ug/ml, 5ug/ml, 10ug/ml, 20ug/ml

I appreciate any inputs. Thank you guys!

I don't think there's a need for these awkward numbers of 9/13/17 just to be 4 apart. I use a mixed type of design from your example. Never a true serial.

The matter of serial dilutions has been discussed in a number of other topics, for instance

viewtopic.php?f=19&t=32911&hilit=serial+dilution
viewtopic.php?f=1&t=43341&hilit=serial+dilution

Check out these two articles in the "Statistics in Analytical Chemistry" series (actually, the whole series is worthwhile):

http://www.americanlaboratory.com/913-T ... on-design/

http://www.americanlaboratory.com/913-T ... on-design/

Your calibration curve really depends upon the concentration of your samples. Almost all of the stock statistical solution used in the chromatography software packages will heavily weight the calibration curve to the high end standards (there are options that will make it better, but most people never use them). Thus, if you are measuring at or near the bottom of your curve then you want to weight your standards towards the bottom; for example:

1 ug/L
2 ug/L
5 ug/L
10 ug/L
20 ug/L

This will typically give you better precision at the bottom end of your curve.

If you are doing QC measurements of an API, for example, where you know the concentration of your analyte then you would make your mid-point standard equivalent to the concentration of your sample. This will typically provide you with the best precision at that calibration point.

Many of the regulatory agencies will try to force you to use a linear estimate for the curve modeling. This very heavily weights the curve toward the top end standard, so you're normally better off working with a limited calibration range, such as a decade calibration curve.

We have done both serial dilutions and dilutions directly from stock. As long as your pipetting technique is good I don't really see much difference except, as James notes, with highly volatile compounds. You check your pipetting/diluting skills by running a secondary check standard (typically called the initial calibration verification or ICV standard) that is independently prepared from a different stock solution.

Thank you for the suggestions!

Your calibration curve really depends upon the concentration of your samples. Almost all of the stock statistical solution used in the chromatography software packages will heavily weight the calibration curve to the high end standards (there are options that will make it better, but most people never use them). Thus, if you are measuring at or near the bottom of your curve then you want to weight your standards towards the bottom; for example:

1 ug/L
2 ug/L
5 ug/L
10 ug/L
20 ug/L

This will typically give you better precision at the bottom end of your curve.

If you are doing QC measurements of an API, for example, where you know the concentration of your analyte then you would make your mid-point standard equivalent to the concentration of your sample. This will typically provide you with the best precision at that calibration point.

Many of the regulatory agencies will try to force you to use a linear estimate for the curve modeling. This very heavily weights the curve toward the top end standard, so you're normally better off working with a limited calibration range, such as a decade calibration curve.

We have done both serial dilutions and dilutions directly from stock. As long as your pipetting technique is good I don't really see much difference except, as James notes, with highly volatile compounds. You check your pipetting/diluting skills by running a secondary check standard (typically called the initial calibration verification or ICV standard) that is independently prepared from a different stock solution.

Thank you Tom!

Check out these two articles in the "Statistics in Analytical Chemistry" series (actually, the whole series is worthwhile):

http://www.americanlaboratory.com/913-T ... on-design/

http://www.americanlaboratory.com/913-T ... on-design/

Thank you!

I don't think there's a need for these awkward numbers of 9/13/17 just to be 4 apart. I use a mixed type of design from your example. Never a true serial.

The matter of serial dilutions has been discussed in a number of other topics, for instance

viewtopic.php?f=19&t=32911&hilit=serial+dilution
viewtopic.php?f=1&t=43341&hilit=serial+dilution