How do you guarantee that a sample had an ok extration?

[biotechno]

Hi all,

We are at the moment working in a quality control procedure to ensure that every sample had a good enough extration that guarantees detection of a compound at the reported limit of detection.

How do you guys do that?

Thank you all for your cooperation.

[Peter Apps]

At the risk of sounding like a smart *rs - you only want to detect it when it is really there. Your problem is actually to show that you only get a negative when the compound is really absent (i.e. below the LOD).

Specificity of response i.e. not mistaking other compounds for the compound of interest depends on having a clean blank.

Detecting what is there is an issue of recovery - usually addressed by adding internal standards, although these have shortcomings when the target is strongly bound to, or incorporated into the sample.

Peter

[Don_Hilton]

This depends on the matrix. The typical strategy for a matrix that is easy to extract is to spike the analyte of interest into the matrix, mix well, perhaps evaporate the solvent used in spiking, and extract. If you can not obtain matrix that is analyte free, then measure spiked and unspiked matrix and subtract the response for the analyte in the unspiked sample from the response in the spiked sample.

I will often dothis kind of study by adding a recovery standard after the extraction and separation of the extraction solution from the matrix (works like an internal standard, but is not subject to the extraction process).

One can also extract for varying periods of time and look at the time trend. If one is extracting from a solid material where the analyte is in the solid and not on particle surfaces, one can try varying the particle size of the ground material to see if that changes the rate of extraction. One can also extract, remove the solvent and extract again repeatedlyl to estimate how complete the extraction is.

Thsi should give you a set of conditions that will work for all samples in the same matrix. If you must confirm extraction in all samples with variable matrix, the best may be to extract, remove the solvent and extract again. Analyze both extracts - but be sure that you have studied the extraction so that you know that the first extraction should be exhaustive for the given matrix, particle size, etc.

[biotechno]

Thank you for your replies, but to be honest i'm a litle lost.

We work with internal standards, and we were asked to have a criteria to whats the minimum amount of internal standard we accept in each sample so we guarantee that if a sample has the coumpound at the LOD level we will with no doubt see it.

Basicaly, whats the criteria to say the extration worked out like usual.

We have positive and negative controls, but they (external audit) want us to control each sample extration. To see the internal standard at the usual level in each sample is not enough. They want a criteria of whats the minimum amount of IS we accept at each sample.

[Peter Apps]

Thank you for your replies, but to be honest i'm a litle lost.

We work with internal standards, and we were asked to have a criteria to whats the minimum amount of internal standard we accept in each sample so we guarantee that if a sample has the coumpound at the LOD level we will with no doubt see it. The usual definition of LOD is a signal that is three times noise; therefore it is fundamentally impossible to detect a compound that is at the LOD "with no doubt" - the statistics of result distribution mean that some samples with analyte above the LOD will generate a result below the LOD. In practise LODs need to be far enough below the "action level" that the chance of a false negative leading to no action when action was required is acceptably low

Basicaly, whats the criteria to say the extration worked out like usual.

We have positive and negative controls, but they (external audit) what us to control each sample extration. To see the internal standard at the usual level in each sample is not enough. They want a criteria of whats the minimum amount of IS we accept at each sample. If this is really what they want, and you are determined to give it to them, then you have to spike each sample with an isotope labelled analogue of the analyte at the LOD and show that you can detect that

Peter

[biotechno]

Thank you for your replies, but to be honest i'm a litle lost.

We work with internal standards, and we were asked to have a criteria to whats the minimum amount of internal standard we accept in each sample so we guarantee that if a sample has the coumpound at the LOD level we will with no doubt see it. The usual definition of LOD is a signal that is three times noise; therefore it is fundamentally impossible to detect a compound that is at the LOD "with no doubt" - the statistics of result distribution mean that some samples with analyte above the LOD will generate a result below the LOD. In practise LODs need to be far enough below the "action level" that the chance of a false negative leading to no action when action was required is acceptably low

Yes, we allow 1 in 20 samples to fail the criteria (5%).

Basicaly, whats the criteria to say the extration worked out like usual.

We have positive and negative controls, but they (external audit) what us to control each sample extration. To see the internal standard at the usual level in each sample is not enough. They want a criteria of whats the minimum amount of IS we accept at each sample. If this is really what they want, and you are determined to give it to them, then you have to spike each sample with an isotope labelled analogue of the analyte at the LOD and show that you can detect that

Wow. Well, they are not that demanding, nor do we have all those isotope analogue to use, BUT they would like us to (some how) verify that the extration worked out fine using the IS.

We have developed this procedure:

1) Determine the S/N of each compound at the fortified sample of the batch.
2) Calculate how many times lower the S/N of each compound would be if it was fortified at the LOD level.
3) Calculate that same amount of times lower the IS would be.
4) Calculate how much area of IS does that S/N represents.
5) Verify that all samples have IS areas above that minimum calculated.

But i'm not sure this is a good aproach, that's why i started this forum by simply ask how other labs do it.

[Peter Apps]

If the auditors are happy with your approach, then use it. It involves some extra spreadsheet calculations, but no extra analyses

Peter

[biotechno]

If the auditors are happy with your approach, then use it. It involves some extra spreadsheet calculations, but no extra analyses

Peter

Huumm...

Dificult to fullfil this criteria with some compounds.

I'm not sure there is a 1:1 relation between concentration and S/N.
Not sure we can say that S/N is 10x lower when we drop to concentrations 10x.

I was hopping someone had better aproach to this problem.

[Don_Hilton]

If you look at some of the US EPA methods, they add a recovery standard at the end and compute the recovery of the internal standard. The recovery of the internal standard must be within a particular range and, if I recall correctly, the area for that peak must fall withing a percentage of the area in a standard solution. A low level standard at the level where you want to assure detection should be part of the check because even though peak areas may be the same, adjustments in detector voltage may change the noise level.

[Peter Apps]

If the auditors are happy with your approach, then use it. It involves some extra spreadsheet calculations, but no extra analyses

Peter

Huumm...

Dificult to fullfil this criteria with some compounds.

I'm not sure there is a 1:1 relation between concentration and S/N.
Not sure we can say that S/N is 10x lower when we drop to concentrations 10x.

I was hopping someone had better aproach to this problem.

Your concerns are valid - extrapolating downwards to the LOD is almost certainly fraught with problems.

The underlying problem here is that the auditors are asking you to do something that is fundamentally impossible using any of the usual QC checks for recovery. By definition the LOD is below the range of the calibration (whose lower end is the LOQ) in the area where statistical variation is too large for the relationship of peak area to quantity to be reliable. This is why I thnk that the only way to satisfy their requirement is to add an isotope labelled standard to every sample. You then eliminate sample to sample variation which is the major cause of noise.

Peter

[carras]

Hi, biotechno
We add IS before extracting the samples and the rest of the workup (such as SPE). For the IS peak we monitor one ion (or MRM transition) whose abundance is close to or lower than the least intense ion (or MRM transition) of the all the analytes. This ion/transition should have a signal/noise ratio higher than 3 in every sample. If that is the case and there are no peaks for an analyte then the sample is negative for that analyte. The rationale being that if an analyte is really present in a sample we should be able to see its peaks in all the ion/transition chromatograms since their S/N ratios are higher. The trick I guess is choosing an IS that really behaves as your analyte(s) during sample workup. This approach was OKed by our external auditors under ISO 17025.

Hope it helps

[James_Ball]

The biggest problem we have encountered is that every auditor interprets the methods and regulations differently.

We always use a monitoring compound during extraction at the least, just to assess recovery, along with running at least 1 in 20 samples as a matrix spike and matrix spike duplicate to show recovery of the actual analytes and the reproducibility of the extractions. For a further QC you could be extracting the internal standards also and use a fortifying compound to assess the recovery of the extracted internal standard.

If you are reporting an actual LOD then when you have any recovery of monitoring compounds less than 100% theoretically you would not be able to meet that LOD, but have to adjust upward to account for the loss in the extraction process. What I think works better is to determine the LOD of the instrument without extraction then set a MDL(Method detection limit) that takes into account the normal extraction efficiency. If you can detect 10ppb on the instrument, and you normally recover 70% of a spiked analyte then your MDL could not be lower than 13ppb, since once you lose 30% if you tried to see 10ppb you would actually have only 7ppb present to inject into an instrument that can only detect if above 10ppb. The problem I have seen though is that many regulating agencies believe that if you can detect 10ppb in a clean solvent matrix of a standard, then you should always be able to report a detection limit of 10ppb even when the sample is something like tar or crude oil.

[biotechno]

Thank you all for your repplies. Its a nice help.

I think the biggest problem here is that diferent samples have diferent LOD.
The LOD determined is the evaluation of the average performance of the method. So we can not guarantee that LOD at every sample.

And i dont believe that's what our auditor realy mean.
What he wants is for us to verify that there wasnt any problem with out extration. That our method is preforming as usual. Only that, nothing more.

So we can never measure that with a S/N, because we receive some unusual dirty samples sometimes.

I like that option of setting a average performance for the IS and stablish a variation based on what we have seen on the validation of the method.

But to be honest, in my experience, what would realy give me peace of mind is the intensity of the peak it self, not the area. That way a dirty sample would show a higher background but should show me the same peak intensity (much less area but same intensite at the peak). But this would never pass in my QC team (lol).