Chromatography Forum

I use to calculate LOD=3x(Signal to noise) and LOQ=10x(Signal to noise).
Whai is the meaning of S/N parameter in Agilent 1200 Software?
How far from the peak of analyte you chose a base line for noise measurement?

How far from the peak of analyte you chose a base line for noise measurement?

As close as possible.

That said, in my (not so humble) opinion, S/N is adequate for *estimating* LOD or LOQ, but a poor way to *calculate*, for just the reason to which you allude: where do you measure noise? how long a stretch of baseline is required? what happens if you have other peaks near the analyte?

There are far better approaches:
- statistical (based on the standard error of the y-intercept)
- CV vs A plots (especially for LOQ, since this allows the precision requirement to be accounted for).

Thnks,
I just do no understand about CV vs A. What are those?

CV is "coefficient of variation", also sometimes referred to as "RSD" ("relative standard deviation"). A is analyte amount (or peak area or height). In a CV vs. A plot, you run replicates at various levels across the linear range. A log-log plot of CV vs A will typically be horizontal at relatively high amounts and then curve upward to the left for smaller amounts. Where that plot crosses your target CV tells you the LOQ.

CV is "coefficient of variation", also sometimes referred to as "RSD" ("relative standard deviation"). A is analyte amount (or peak area or height). In a CV vs. A plot, you run replicates at various levels across the linear range. A log-log plot of CV vs A will typically be horizontal at relatively high amounts and then curve upward to the left for smaller amounts. Where that plot crosses your target CV tells you the LOQ.

Hi Tom,

i am doing a linearity experiment, i draw the linearity graph between area and concentration in ppm and i calculate the LOD and LOQ from the linearity curve. My question is in what units i get the LOD and LOQ values?If i want to change the units from ppm to % how do i do that.Could you please explain.
Thanks
suri

LOD / LOQ will have the same units as your x-axis (ppm, %, mg, ng, . . . whatever you used).

% = parts per hundred.
ppm = parts per million

so, 1% = 10,000 ppm

% = parts per hundred.
ppm = parts per million

so, 1% = 10,000 ppm

I'd be careful about that. I've often seen "ppm" used as mass per volume concentration unit. So 1 ppm = 1µg/L (with the background that 1L of water should weigh 1kg = 1,000,000µg). Although this is true only for purely aqueous solutions, it's often used for solutions including organics of any kind.

"%" on the other hand is often related to the "normal"/100%/whatever-you-call-it concentration of the sample.
So, if your sample is e.g. 0.5mg/mL (=100%) = 500µg/mL, 1% would be 5µg/mL. 5µg/mL = 5,000µg/L would tranlsate into 5,000ppm!

Good point, but I wanted to avoid the whole debate about whether "ppm" or "%" make sense (the inherent ambiguity makes them very poor choices!).

Good point on your side, Tom !

That said, in my (not so humble) opinion, S/N is adequate for *estimating* LOD or LOQ, but a poor way to *calculate*, for just the reason to which you allude: where do you measure noise? how long a stretch of baseline is required? what happens if you have other peaks near the analyte?

There are far better approaches:
- statistical (based on the standard error of the y-intercept)
- CV vs A plots (especially for LOQ, since this allows the precision requirement to be accounted for).

This is a debate I'm currently having here, "which is better?" or rather, "which is preferred by the FDA?". Or even, "do they have a preference?".

In a recent validation I calculated the LOQ experimentally using S/N as 1ppm. When I calculated the LOQ statistically based on the linearity data (as per the ICH, 10*sigma/slope), I got a value of 31ppm. I was hoping the two LOQ values would be closer together so I could say that it didn't matter which method was used, but the discrepancy between the two doesn't please me.

LOD / LOQ will have the same units as your x-axis (ppm, %, mg, ng, . . . whatever you used).

% = parts per hundred.
ppm = parts per million

so, 1% = 10,000 ppm

Thanks very much Tom,

i am just estimating the linearity and LOd and LOQ by just using the standards.I am not using any spiked test samples. which way would be most appropriate?

Thanks

Spiked matrix blank would be the most appropriate (more representative of real samples).