GCMS limit of detection

[djaydearz]

Hello All.

I am currently determing a limit of detection for a GCMS.

I have calculated my LOD by doing the following

LOB = meanblank + 1.645(SDblank)
LoD (y response) = LoB + 1.645(SD low concentration sample)

Ran a 5 point calibration, rearranging the y = mx +c to to give me LOD (x concentration)

I get a value of 0.06mg/ml......however the concentrations i ran for the LOD (y response) above was 0.02mg/ml (the GCMS could detect a peak)....so surely the LOD would be closer or lower than 0.02mg/ml?


Sorry if i have this completly wrong, bit new to statistics.

[Peter Apps]

Over what range did you calibrate ? If the calibration spans a range a lot higher than the concentrations that you are interested in (in this case the LOD in concentration terms) then the calibration equation as a whole probably does not hold for the low concentration end. If you reduce the range of the calibration, and make the lowest point where you think that your LOD is, the two estimates should be closer - but there is likely to be more noise and so the linear fit might well be worse.

Also; you are double counting the baseline noise in your second line - the SD of the peak area includes the baseline noise, which is also in the LOB term.

Peter

[mckrause]

Do you have detectable analyte in your blank? If not, how did you determine the SD of your blank?

In general the statistically calculated DL should be less than the concentration of the standards used. This is not necessarily true, as the statistics factor in the variability in your analytical technique. You can make multiple injections of a low standard (surely you mean ug/mL, not mg/ml), get a quantifiable peak for each injection and still calculate a "detection limit" significantly above the concentration of the standard. This is the major issue with statistically-derived limits - they fail the simple logic test.

[djaydearz]

Thanks for the replies

No analyte in the blanks, i just intergrated some noise and used the response of that.

mg/ml is correct 0.02mg/ml.
I used a 5 point calibration from 0.1mg/ml - 1mg/ml, so i can reduce this and see what results i gain

thanks

[Camisotro]

Yes the mg/mL range seems a bit concentrated for GC-MS doesn't it?

If you want to visualize how dilute of a standard you can inject while actually seeing a coherent peak, you could try some weaker serial dilutions. If you can dilute to a level with no discernible peak or a peak that is on the order of magnitude of the noise, then you should be searching for your LOD using a standard a bit stronger than that level.

I would also be inclined, when making a direct comparison between a peak and a blank that is just noise, to look at height more than area. Integrating noise is not really a very useful endeavour.