LOD Formula

[vdnsrinivas]

Hello,

I want to know the exact formula for LOD calculation:

LOD=3XSD/Slope??? or
LOD=3.3X SD/Slope (as per ICH)

which one is the exact formula?

[Don_Hilton]

There are a number of ways of calculating LOD. I used the method publised in EPA methods for a number of years, but by policy of my current employer, I need to use a different method.

First question is do you work in a regulated environment or have a management policy you need to follow? If so, your answer is given by the regulations or policy.

[vdnsrinivas]

Its a regulated one!

[chromatographer1]

To calculate a LOD is to make an educated guess.

The best way is to determine the actual LOD. A bit more work and it has to be done and reported according to the parameters specified. How much signal to noise, etc.

In your case, follow the regulations which specify the equation to be used.

best wishes,

Rod

[chromatonewbie]

I found the EPA method 8000A Determinative Chromatographic Separations very helpful for thinking about the LOD.

As I was using GCMS I chose retention time, identifier ion ratios, signal to noise and peak symmetry, with +/- tolerances for each parameter to define my LOD.

I think as long as you explain what you have chosen and why, it will be OK.

[sdegrace]

Srinivas, first of all, can I safely assume that this is for a GC method? Also, is this for a drug product? There's a number of different approaches you can take, but I can tell you what I have been able to do for some drug products.

If both of these assumptions are true, then I would not try to calculate LOD and LOQ based on a statistical analysis of the calibration curve. I tend to save that for methods like UV assays that don't have noise you can measure as conveniently. Instead I would use the signal to noise approach, allowed by the ICH guidelines and very suitable for chromatographic data.

I would use your chromatography software to obtain a measurement of the signal to noise ratio (s/n) using peak to peak noise (easy to do if you're using Empower, and I've also done this analysis in ChemStation and Galaxie, but it was not as pleasant) based on a noise measurement close to the peak of interest.

What I would do first is make up a low standard, inject it a few times, and use it to get estimates of the s/n as per the formulas:

LOQ = 10 x Conc. Std. / signal-to-noise
LOD = 3 x Conc. Std. / signal-to-noise

Then I would make standards at close to these values, maybe a little higher, and call them my targets for LOQ and LOD. I would inject each at least six times and declare these concentrations my nominal LOQ and LOD if the measured average s/n were NMT 10 and NMT 3 respectively, and for LOQ at least, the %RSD NMT 10%.

Hope that gives you some useful ideas, anyway.

Stephen

[vdnsrinivas]

Thanks Stephen

[thohry]

I am not sure of calculating the LOD of additional method. For example, in analyzing some compounds in beer, we can not make a real blank to prepare standards. We have to add standards into the sample and meassure. But the LOD is not as simple as normal methods like external standards ..
Anyone have any ideas on this ?

[Alexandre]

For some sample/analyte combinations e.g. food composition chemistry there is always an issue of LOD determination as analytes are always present, e.g. vitamins, fatty acids etc.
We determine DLs for these cases by method of standard addition.
Happy experimenting!

[Don_Hilton]

The limit of detection is the lowest value you can reliably differentiate from 0. On cases where you have background levels of analyte, the best you can do is find a sample that has as low a level as you can find and do the statistics and show that at the level in that sample, when replicate analyses are performed, the result obtained is distinguishable from 0. If the compound you are examining is at a level close to indistinguishable from 0, you can use the variability statistics to compute a LOD. If the background level is well above indistinguishable from 0, the best you can do is document the LOD is less than the number you are looking at.

If you work in a regulated environment, this may be spelled out for you. Otherwise you can look at existing methods and/or a statistics book for methods of computation.

Some people will use the concentration for a chromatographic peak wiht signal to noise of 3:1 as LOD. Others use multiple samples read from a calibration curve and compute the standard deviation of the results and compute a confidence interval. Remember that if you attempt to give a LOD level below the lowest concentration you actually measure, you are making an assumption about linearity - which may not be true.