LOD in matrix matched calibration

[ray_chelle8]

hello,

I just want to consult regarding my validation method in multiresidue analysis of pesticides in GCMSMS. I computed my LOD and LOQ based from SD in linear regression. I got 4 ppb for LOD and 12 for LOQ. My problem now is that thesignal to noise ratio is below 10 because of the matrix peak beside the peak of the analyte. The recovery and the RSD passed the criteria of 70-120% and RSD of below 20. should I increase it my LOD/LOQ until I get a SN of 10? is matrix peak considered as baseline noise? Thanks

[Johnny Rod]

If you're looking for LODs/LOQs in MS data then you can forget signal/noise ratios as there can be zero noise in MS, and work only from the SD. It is more referred to as MDQ (minimum detectable quantity) or similar.

[James_Ball]

An obvious peak would not be considered baseline noise in my opinion. If you can not separate the analyte from the matrix peak completely then that would be a reason to question the detection limits, but as long as they can be separated I would just look at the noise in the baseline versus the height of the actual peak.

As said above, for MS it is often difficult to do S/N when noise can be near zero, I normally set a peak area limit of something like 1000 area under the peak( or whatever gives a good peak in your instrument) and back calculate to what concentration should give me that area, then inject that concentration and see if it works.

[Peter Apps]

The most rigorous and robust estimate of LOD and LOQ for a method are the concentrations (or quantities) at which the rsd on repeated measurements is 30 and 10% respectively. This is proof positive of how repeatable the method is at that level. The SD of the calibration can be influenced by very good (or bad) repeatability at levels way above the lower limits, and the signal:noise on a chromatogram gives the most optimistic possible estimate of LOD and LOQ, because it completely ignores all the other sources of variation.

Peter

[Steve Reimer]

For a brief discussion of why S/N is not a good judge of limit of detection and quantitation check out http://www.absciex.com/Documents/Downlo ... 059150.pdf
They cover the affect of filtering and smoothing on S/N calculation.