N/A or 0

[crimelist]

Hi!
I'm writing a new procedure with regard to the evaluation of areas and peaks.

Now, I have to evaluate a case.
When I find no peak (but the method that I use does not describe LOQ or LOD)

How would you write the result?

Write ZERO?
Or N/A
Or N/D

Thanks

[tkubowicz]

Hello

If you have method with calibration and you have no peaks for region when you should have peaks, you can write "no peaks detected".

Regards

Tomasz Kubowicz

[Peter Apps]

Not so simple; if you set the minimum area in the integration method low enough you will always find a peak - it's called baseline noise. Detecting a peak that is significantly above baseline noise depends on what signal:noise ratio you want for your LOD and LOQ.

Peter

[tkubowicz]

Hello

Not so simple; if you set the minimum area in the integration method low enough you will always find a peak - it's called baseline noise. Detecting a peak that is significantly above baseline noise depends on what signal:noise ratio you want for your LOD and LOQ.

Peter

Of course it is true and I agree.
However question was about method without LOD/LOQ.
Proper procedure should contain lowest concentration (LOD or LOQ) and if calculated sample concentration is below LOD/LOQ report could be setup in the way that it says "below LOD/LOQ"

Regards

Tomasz Kubowicz

[crimelist]

Both right!
But the method it's not mine. It is a customer. That is not going to change it. So I don't have any noise value or LOQ.

[Peter Apps]

And on the other hand if you set the minimum area high enough and increase the maximum of the Y axis to its maximum you will see no peak even if there are high concentrations of analyte present.

The point is that not finding a peak means nothing unless the conditions for finding a peak are defined - and that means defining the LOD.

Peter

[crimelist]

Considering that I have not the LOD, I can force method and apply it?
There is a law that allows me to calculate LOD when I don't have it?

[Consumer Products Guy]

BDL

[lmh]

Report as "not detected". It shouldn't be reported as zero, because zero is a real number, implying a real measurement, rather than a failure to find something (a peak!) to measure.

There are precedents in commercial software. Shimadzu's software, for example, reports the area, height and retention time of peaks that weren't found as "-----" rather than zero.

Of course anyone reading "not detected" should ask themselves how much stuff would have to be present before you would detect it - but you're right that if you have been told not to attempt to determine the LOD, then you're not under an obligation to answer this (very important!) question.

[bunnahabhain]

Just a thought: If I define a "minimum peak area" for integration, and subsequently the sofware does not find a peak, can I then use the minimum area value for calculations? This would result in a value < xxx.

[lmh]

My feeling is that your LOD becomes whichever is higher: the concentration corresponding to the minimum peak area, and the true LOD as dictated by the noise/variability of a very small peak.

[Peter Apps]

Just a thought: If I define a "minimum peak area" for integration, and subsequently the sofware does not find a peak, can I then use the minimum area value for calculations? This would result in a value < xxx.

I don't think that you can. You will not have calibrated down to the level where the peak is too small to find (would you ?? it would be strange to run a calibration and then set minimum area above the peak area for the lowest standard) so if you do a calculation based on the minimum area setting you are extrapolating from the calibration.

There is also the danger that the software misses the peak because it has moved a bit, it has something else on top or next to it, or the slope parameters were set wrong.

Peter

[bunnahabhain]

I don't think that you can. You will not have calibrated down to the level where the peak is too small to find (would you ?? it would be strange to run a calibration and then set minimum area above the peak area for the lowest standard)

No, of course not.

so if you do a calculation based on the minimum area setting you are extrapolating from the calibration.

This is true. I very often set a minimum area to prevent the software from intepreting noise as microscopic small peaks (I know you can do this also with minimum width etc, but minimum area is very convenient). I don't feel uncomfortable with extrapolating from the calibration. Very often you don't need to calibrate close to the LOQ. Then why define an LOQ which is absurdly low?
Calculation example: My minimum detectable area is 0.01, my smallest calibration standard has 15, highest standard 300. I tell my chromatography software to ignore everythin smaller than 1.5. Works great. I know every chromatogram where the software does not find a peak is at least 10fold lower in concentration than my concentration range of interest. If I then define the concentration 10fold smaller than the lowest calibration standard, this is not really the limit of what is possible, but safly (IMHO) defines the useable range.

There is also the danger that the software misses the peak because it has moved a bit, it has something else on top or next to it, or the slope parameters were set wrong.

This has nothing to do with minimum area setting. This is another problem I have to deal with quite frequently...

[James_Ball]

For me I would say it depends on the end use of the data.

You could set a Minimum Reporting Limit(MRL) which is usually equal to the lowest calibration standard and if you get either no peak or a very small peak below that level report as <MRL with the MRL stated in the report.

Also in the new EPA methods for drinking water there is a procedure for determining and verifying MRL which validates the ability of the method to reliably report a value at the MRL. Anything that calculates below this value would be reported as either <MRL or a number qualified as "Estimated" since it would fall into the range between being detectable and being quantified reliably.

If there is only the option of N.D. or 0, then I would choose N.D. for not detected instead of 0.

[lmh]

You'll notice that I referred to LOD, not LOQ or minimum reporting limit! In saying that the LOD becomes the higher of the two values [higher of: (1) the concentration corresponding to the minimum peak area setting, and (2) the true LOD calculated from noise or variability of peaks at low abundance, according to one of the usual methods ] I was implying that the true LOD had indeed been determined. You can't just estimate the concentration corresponding to the minimum peak area setting and treat that as the LOD because it could be less than the LOD.

On reflection I can't see why you would set the minimum peak area above the true LOD because it's a bit of a shady thing to do. It implies that you want to report that something is absent when in fact it can be detected. I also use the minimum peak area setting only as a way to avoid the integrator finding noise-peaks. If you do change the minimum peak area setting to above the LOD, after determining the LOD, then you absolutely must not continue to use the LOD because you are now definitely rejecting peaks that the LOD indicates were detectable.

Obviously any value between the LOD and the LOQ can only be reported as an estimate with worse precision than has been deemed acceptable, and if the LOQ is below the bottom of the calibration curve, then any values between the LOQ and the lowest calibration standard are estimates based on extrapolation and should be annotated as such. I'm writing as someone in a research environment where it's my job to provide the best estimate I can, and annotate any dangers associated with it, but I don't have the duty to reject a value merely because it's an estimate, and I don't have external auditors to worry about. The typical question I get is "OK, so there wasn't enough to measure in experiment 1, but can you estimate how much more concentrated I need to make the sample in experiment 2?", and for this, I think, extrapolations and estimates are sensible.

There is a relevance of the minimum peak area setting to situations where you have variation in retention times. If you have seriously varying retention times (which you shouldn't) and you're selecting the peak that is closest to the correct retention time, then if the integrator is picking up silly little noise peaks, there is a risk that one of these will be closer to the expected retention time than the true peak. One option to get round this is to set a higher minimum peak area. But we're in the territory of making the best of a bad run here, very much the sort of stuff that is just about OK in a research lab where you're trying to get what you can out of the only 5uL of sample that exists while you wait 3 months for the next experiment.