Sequence - correct ?

[Pepter]

What you think about this sequence (for pesticides analyzes)
Is it correct ? Any errors or idea to improve it) How does it look like
in your laboratory ?

1. Blank
2. 100ng/ml Calibration (all in matrix)
3. 50ng/ml Calibration
4. 10ng/ml Calibration
5. 5ng/ml Calibration
6. 1ng/ml Calibration
7. Recovery sample 10ng/ml
8. sample_1
9. sample_2
10.sample_3
11.sample_4
12.sample_5
13. 10ng/ml Calibration
14. sample_6
15. sample_7
16. (double sample sample_3)
17. sample_8
18. sample_9
19. sample_10
20. 10ng/ml Calibration

i run 5 point calibration on start than only one level (10ng/ml)
every 5 samples (is it correct or i shoudl close the sequence
with 5 levels of calibration ?).

Sorry for my english ... very bad.
And thanks for help.

[Don_Hilton]

I typically run calibration standards from low to high. Any carryover has a greater effect when you run high to low. The frequency of check samples depends a bit on the history of the method and how much data you can afford to lose when you discover the calibration has wandered away.

The repetition of a single sample (or all samples) depends on the application and the measured robustness of the method. In a QC method where a failure means a bad ship/no ship decision, I prefer two replicates of each sample. In a populaiton study with thousands of subjects, a single replicate from each may have to do as a matter of cost and available instrument time. And, if a single sample has been incorrectly prepared, that single bad result will not have a noticible effect on the entire study - as the result of interst is the description of the population as a whole. (Samples with extreme values may be confirmed, as they may incorrectly extend the range of results from the population if they are in error.)

So, like my lawyer says - it all depends.

With the exception of the order of calibration standards, I have no problem with the sequence you propose, as long as it meets the needs of the user of the results.

[lmh]

As DonHilton said, it all depends; there's only one thing I'd add. Sometimes I run my calibration curves in a randomised order (this means that if there is a non-random change in instrument sensitivity, it becomes a random error in the calibration curve rather than a systematic trend, which is harder to spot and more likely to create inaccurate results). Other times I run them as Don Hilton suggested, smallest to largest. I do the randomised version on the good autosampler that doesn't suffer from carry-over, and the small-to-large on a bad autosampler which does suffer from carry-over. In this case I also run a blank after the highest calibration curve point (1) to assess carry-over, and (2) to insulate it from the first real sample, which I'd rather not cause to become a false positive.

If you ever have large numbers of samples, you might consider bracketting them; split them into groups, and put calibration curves at the start and end, and between each group; this way if something goes wrong you can use all samples up to the last good calibration curve, and gradual changes in sensitivity can be dealt with by calibrating each sample using the average of the standards before and after its batch.

[unmgvar]

why have you decided that the 10ng is the standard to use as the check standard over the work?
i would had a few injection as well of the same conc. at the start of the work to check repeatability of the sampler. I would use the concentration that is to be most expected in your samples. and if we do not know then go for the most probable

[Pepter]

Hi.
Thanks for helping me.

Re Don_Hilton:

Will change the order from lowest to higher with blank sample after the highest one.

Re Unmgvar:
10ng/ml is my LOQ. Thats why i choose it for recovery and for bracketing whole sequence
to be sure thats from beginning to end i can see LOQ (am i right ?) or i should add also LOD ?.

Re Imh:
I didnt think about randomize ... good idea will try it.
About the bracketing with full calibration curve... time is a problem. My boss said that
i should do from 20 to 30 samples a day. I check up some applications from Agilent
(we will probably buy Agilent 1290 with 6460 detector) here is link to methods ->

http://www.chem.agilent.com/Library/app ... 4253EN.pdf

Some of them looks like science-fiction (224 or even 300 pesticides in less than 25 minutes)
Continuing...30 samples a day with recovery/blank/calibration curve on start mid and end/
laboratory control sample will take much time.

For first i want to be sure that iam giving good results second thing
i need to do as many samples as possible.

I dont have much practice with big multiresidue methods up to now i was
working on DPL and set up small method for 20 pesticides

[Don_Hilton]

The autosampler may be able to handle the 20 to 30 samples per day - how about the data review! If you need to check for 250 pesticides in each run to be sure found peaks are correctly assigned and integrated and peaks not found are indeed not present, you have a lot of work on each sample.

I would like to know how long it takes to review all of the peak assignment and integration before accepting the calibration curve. Do you plan to generate a new calibration curve daily?

Good luck - and let us know how it goes.

[Don_Hilton]

And, LMH made mention of a blank after the high standard, which brings to mind: if you are running standards in matrix, a 0 level standard is a good idea to ensure background levels of analyte in the matrix are correctly addressed as well.

And if you are runnign trace level pesticides and not building your calibration in matrix beware.

Give a bit more detail about what you are doing and some of us may remember a few more things! The 0 level standard came to mind because of your mention of pesticides. As I was downstairs making cofee before heading to work, I remembered what I had neglected in setting up to do some pesticide work, once upon a time.

[Pepter]

Re Don_Hilton
I hope i will be able to set up about 100 pesticides for first method (250 on lc-ms/ms in routine analyzes is something hard to imagine...at least for me -> human). About the peaks review ... cant tell anything as far i dont sit and scan 20 samples x 100 pesticides... learning now how to optimize MassHunter Quantitative for future work as far as i can see this application is very helpful and yes i plan to make calibration curve daily.


Re Imh:
I still thinking about this 5 point calibration curve at beginning and only one calibration level 10ng/ml as the mid* and bracketting sample*.
I want to use 2 internal standards, one before extraction and second (TPP) added to vials. Wont it be enough to carry* MS/MS sensitivity change ?(still have in mind that TPP wont be like hybrid and wont react on every change on machine like the 100 pestcides i want to analyze).

p.s.
Calibration will be in matrix.
p.s.1
Don Hilton i hope i didnt insult you. If i do iam sorry.
Once again thanks for giving me advice and reading my horrible english.

[unmgvar]

logic would have it that if 10 is your LOQ then it should be the lowest point of your curve
and the other points be for information
i do not see how the lower points can be part of the curve is they fall below LOQ
you cannot make amount results best on them

[Pepter]

Re Unmgvar:
My bad than. So i need to add more points in curve above LOQ.
I think i need one sample that will inform me about the sensitivity of my machine (for fast check...eye shoot) for example matrix-spike sample
on LOD level that wont be included in calibration curve (is it good idea ?).

About the LOQ level.
For example i have 100 pesticides ... part of them will give better sensitivity another worse. Can i unify LOQ for all off them to one
concentration 10ng/g ?(lowest mrl for pesticides in food) even if part of analyze pesticides give lower LOQ or is it nonsense what i write ?

[Don_Hilton]

I have seen numbers below the LOQ used. In staistical studies, in which population trends are described, there is a problem in relating non-reportable levels with other values for a particular subject. Statisiticians will fill the non detect wtih a non-zero number to avoid the loss of a subject in the study. In this case it may be better to give a number that is mostly noise rather than use a particular selected value - thus for the purpose of the study, numbers below the LOQ are used.

I have a strong prefernce, given my background, to like analytical results to be reported only above the LOQ. But in a multiresidue analysis, I will included levels below the LOQ for many analytes because I purchase the stock standard with everything at a particular concentration and I need my lowest calibrator to fall low enough to catch the bottom end of the range for the analyte with the lowest LOQ. (And I do not include points below the LOQ as in the curves for the other analytes.) Thus my curve will run something like 0.1, 0.3, 1, 3, 10... As this is a bit closer to even spacing on a log scale, but still convienient to prepare.

[lmh]

I'm perhaps not following this in the detail I should, so ignore anything too silly, but:
(1) you can, if you want, quote a single LOQ that is the highest you get for any of your pesticides, and apply it to all, but you are doing yourself an injustice. It'd be better to accept that some things are easier to measure than others. It'd be better to have individual LOQ's, or at least group the pesticides into groups with broadly similar LOQ's. If you're lucky, the pesticides you need to detect at the lowest concentration might even have the lowest LOQs!
(2) Related to that, there is no particular reason why you should have the same concentrations of each pesticide in each calibration level (except convenience).
(3) Related to both of the above, you need, of course, to have a calibration curve that spans the range of concentrations you generally see in your samples, and these concentrations will vary with pesticide. Your LOQ must be low enough to quantify any level of pesticide that is relevant to environmental/food safety (unless it's sufficient to detect presence/absence, in which case your LOD must be low enough to detect the relevant threshold concentration).