Chromatography Forum

Hi!!

I've very confused!!
I'm analysing the results of a validation of a new method. Everything looks like really good, but when I integrate the results of my "quality control samples" it's horrible!!!!

The calibration range is from 0.02 to 20microg/mL. I've adjusted them to a 1/W2 regression, and I have a r2=0.99. The precision and accuracy of the calibration samples are really good.
But when I "calculate" the concentration of "known" samples it is horrible!!!!!

Please, some idea???
I'm new on it, and I'm really confused!!

Thank you very much!!!

When you say horrible does that mean that they are all off by a factor, as in you forgot a dilution factor somewhere, or are they all over the place, as in the method doesn't work the way you think it should?
If the calibration looks good but the resulting model doesn't apply to samples then something is different. It can be as simple as the solvent from the prep isn't what's in the standards to something more serious.

Is the matrix nasty? Is there a derivatization? Is there water in the final extract? If you spike blank matrix does it show the same problem?

Work on isolating where the problem lies.
Good luck!

Is it possible for you to obtain Blank Matrix or Placebo or whatever and prepare you own samples with analyte for comparison--using the same solvent(s) as for the standards you use?

As Steve points out...a model is only a model. If the model doesn't represent the "real samples" well enough, no matter how sophisticated the model is, the sample data will not necessarily be what they ought.

Hi,

In fact, I think that there was a misunderstanding.
The “known” samples I talks were samples prepared by myself from blank matrix. Maybe I did some mistake, of course….. I’m going to repeat the analysis, preparing again the “Known” samples.
But my doubt was, considering that everything was prepared correctly, could be possible that an “adjustment” of my data was really good, but not able to calculate my samples?
I don’t think so!! For me it sounds really strange!!! But I’m not sure about it….

Thank you

If you ran your calibration using analyte dissolved in clean solvent then there are all sorts of reasons why the results from real samples are not what you expect - the recovery of analyte form the samples may be too low, there may be other substances that elute at the same time as your analyte and make its peaks look bigger (or smaller) etc etc.

Calibrations run in clean solvent tell you only how the instrument behaves, thay don't tell you anything about the sample preparation.

Peter

Yes, I know.
But in fact, I’m not talking about samples. I’m talking about “quality controls’ prepared by me.
And the calibration curve is prepared not in a solvent, but in a blank matrix.
Otherwise, I’ve checked matrix effects, and there isn’t.

So……

In the first post you complain about the results from:

my "quality control samples"

so are these samples not really samples ? Just labelling them "quality control" does not change their analytical properties.

You say that you ran the calibration on spiked matrix - in other words that you added known amounts of analyte to sample matrix that did not originally contain the analyte, measured peak area and plotted that against the quantity of analyte that you added.

You then added known amounts of analyte to another set of blank sample matrix, labelled them "quality control", analysed them using the same method as you used for the calibration standards, measured the peak area and then calculated the quantity from the calibration equation. The known quantity that you added was different from the result.

Most likely there is human error at some stage.

The second possibility is that the method is very sensitive to something that you are not controlling rigorously enough.

Peter

I’ve repeated the preparation of my quality samples different times. It’s always the same…. At level 0.02 and 0.06µg/mL, the accuracy is really bad (for the specified amounts indicated, I obtain calculated amounts with a % of difference of 35%, but today of 8000%!!!!!!!!!)

Which can be the problem?
Maybe my method isn’t sensitive enough to analyse these concentrations?
But the lowest calibrator (0.02µg/mL), processed as a quality control is ok??

Please, some help?

If I undrstand this correctly, when you process "samples" that are prepared the same way as your calibration standards against your calibration curve the variability is too high for the lowest two points. You say that the precision and accuracy of the calibration samples are good, have you refit the lowest points against the curve? With calibration range over a factor of 1000 a COD of 0.99 still allows quite a bit of variability in those lowest points. This is why regulatory methods usually have a check of the sytem sensitivity.
If that is what is going on, your method isn't sensitive enough to quantitate at the 0.02 and 0.06µg/mL levels.

Thank you very much.
Yes, this is exactly what I means.
But when I refit the lowest points against the curve, they are correct....

What do you mean with This is why regulatory methods usually have a check of the sytem sensitivity?

For the type of analyses I run there are required MDL or LLOQ checks. Method dection limit determination statistically derives a minimum amount that will give a measurable signal change, or smallest amount measurable. Lower limit of quantitation check analyzes a blank matrix spiked at the lowest standard level.

Hi,

I've checked everything, but I'm still having problems with the back calculation of my "known samples"....
I've checked that it isn't a problem of stability or human error in the preparation of them....

So..... is that I have to find the "good" model to explain my data?
How can I get it?

You're running a QC sample at 0.02µg/mL? Then your lowest standard should be less than that in order to properly calculate it.

Try narrowing your standard range too. 20-0.02 µg/mL is rather wide considering your QC is at 0.06 and 0.02.

You are taking into account any dilution/multiplication factors if any, yes?