GCMS precision and accurary problem

[Edde]

Since IS response can fluctuate as much as 50%, I modified the sample prep procedure as follows:
1. added IS to the precipitating agent, methanol, to minimize pipetting error;
2. a fixed 500 uL supernatant aspirated for blow dry and subsequent derivatization.

During QC run, I found that for replicates 1-4, the IS responses are quite constant. But from replicate 5 onwards, the IS responses fluctuate a lot. Is it possible that the IS (1,2 butanediol) accumulates in the liner and be released after a few run?

The ethylene glycol, diethylene glycol, triethylene glycol responses are not as fluctuated.

[Don_Hilton]

Measure recovery of analytes and internal standards by comparision between samples and standards. To do this, you need to add a recovery standard to the samples after separation from the precipitate and without mixing the precipitate back in.

I doubt that butane diol is hanging around in the inlet liner between injections. You can, however have a contition where you lose butane diol on a clean inlet liner, but as you gather polar materials on active sites, you lose less on each injection. This is observed by an increase in repsonse in a diol after several injections, particularly of samples with other matrix components present. One solution to the problem is to "condition" the GC by making several injections of a sample from the previous day's work before beginning a day's sample sequence. The objective is to keep the stuff that is masking active sites at a steady state.

Start by looking at recovery.

[Edde]

I am not sure about the method to extract partitioned analyte from ppt. Do you mean after separating the supernatant, I should add methanol to the ppt again, mix well, centrifuge and obtain supernatant and pool to previously collected supernatant.

Forgive me for my ignorance. I thought even if recovery is poor, the matrix matched calibration curve should be able to conpensate.

The random imprecision in IS signal is the source of analyte inaccuracy and imprecision. As you mentioned earlier, the lipids may form active sites in liner. I read from paper that sulphuric acid or NH2 SPE column can be used to remove lipids. The latter may affect analyte recovery and the former may affect pH and column lifespan. How can I get rid of the dirty matrix that affects precision?

Appreciate your further comment.

[Don_Hilton]

If you have low recovery, small changes have large effects. Matrix matching of standards helps. But understand where the losses are. Variation in losses results in variation in results. Also, look at the recovery of your internal standard. While it is selected so that losses are similar to those of analyates, you do not want to lose any, if you can help it.

For measureing recovery from in the precipitation step, you can extract the precipitate, add a recovery standard and inject. I would suggest take off the supertantant, add a recovery standard so you can estimate how much you take off. Wash, centerfuge, separate and spike to see how much is left. You can expect a small portion to be carried into this next faction, as the methanol can not be transferred quantitatively, but you can estimate the fraction of methanol moved in the first separation and you would expect a proportionate quantity of analyte to move with it. If you have a disporoportionant quantity of analyte in the second transfer, you have binding of analytes to the precipitate and need to address that (here is a case where recovery will hurt you. variing quantities of precipitate will result in varuing quantities of bound material - high recovery makes this a non-issue.) The sum of material in the two portions should add up to close to 100% of material present. If not, it has been lost somewhere. For an internal standard, this is not good. Try for recoveries of 95% or better.

The use of a sulphuric acid column for removing lipids will be a problem in looking for dios. A sulphuric acid column will dehydrate a diol - and destroy it. Bad for recovery.

The amino column might work. Clean up technique would depend on anticpated sample load. If it just a few samples, I would look at using a GPC column and use it for a preperative cut to eliminate the lipids. If you have large numbers of samles, there would not be sufficient capacity.

[Edde]

Many thanks to you.