Chromatograms not considered in Liquid Chromatography

[mkumar75]

In Regular Testsing at QC end on HPLC & GC , Almost how much Observations are concerned to " Not considered Chromatogram".

Generally in routine testing , sometime inbetween analysis some known or hidden facts occurred due to which analysis may be in question mark.
On the same time you know the facts that in analysis of sample there is no issue , the product is well within the acceptance criteria , though some of the work which you should not considered is in between the whole set of work.

My concern is how frequently analysts or the supervisors observed this and how they proceed with that ?

Thanks......Manoj

[Don_Hilton]

I am not sure I entirely understand the question you ask, but from what I understand of it I will pass on my observations:

The analyst may ignore an analytical result showing a product to be out of specifiction - and may well have the opportunity to begin a new career -- with another employer.

The QC tests are based on the premise: Prove the materials in question are within specification. If the material is not proven to be within specification, the only way to pass the material is to reanalyze the sample or resample and reanalyze. And, if you have a passing result and a failing result from two different analyses of the same sample, you need good justification for ignoring the result that indicates the sample fails.

Clear proof that materials pass QC tests protect the company. I've seen where thousands of pounds of product were brought back and burried in a landfill because the lack of passing QC information was ignored - but the customer did notice, harming brand image as well as resulting in losses beginning with the wasted cost of running defective product through the manufactruing process and ending with costs of pulling product from the market.

If you have data that fails to support that materials fail to pass QC standards, document what you have, including "hidden knowledge." If a manager above you decides that QC standards be ignored or that there are overriding issues - let them make the call.

If you work in an environment in which you are pressured to pass materials which do not pass QC tests, look for another job. I can guarantee you that when one of these decisions turns out to damage the company, you will be looking for another job anyhow. And, it is easier to find another job if you are employed.

[mkumar75]

Thanks Mr. Don_Hilton

The question actually was not concerned with OOS or OOT.
The querry is about the Injection by Chromatograph , which sometime behaved as abnormal.

Like out of 20 injection in whole of the analysis , one injection of reproducibility has not met. ( though the other injection from same vial are well with in accetptance criteria )

Another good example out of 6 sample of the analyte , during one of the sample (gradient run ) , one spike has observed , where there was no peak elution to be supposed . The same was reinjected and found OK.

My querry is how to address these type of issues in whole set of analysis.

[HPLCaddict]

The phenomena you are describing actually ARE OOT (and might be OOS, too)!

Of course, with HPLC you may always have extra peaks/spikes/humps/whatever, whose origin is not clear and which dissappear on the next injection. However, just omitting these can be dangerous (in a regulated environment it definitely IS dangerous). Who judges if these peaks/spikes can be neglected and why?
There should be a proper evaluation of these phenomena. Reinjection would be the first thing to do, in some cases preparation and analysis of a new sample might be advisable. Comparison with other chromatograms of the same sequence in order to prove that this single run was "out of order" can also be a part of this evaluation.
In any case, the whole evaluation has to be properly DOCUMENTED in order to please any auditor!

[Don_Hilton]

If one injection for reproducability fails, reproducability fails to be demonstrated. If there was a specification for reproducability set, the specification was to demonstrate a particular quality or characteristic either of the material or of the analysis process. And if it is an issue with the analysis process, the material may be fine - or it may not, but the anaylisis process is faulty and the result can not be trusted to demonstrate that the material being tested is within specification.

The specifics for handling exceptions has to come from those who set the specifications for the materials and those who developed the analytical method. The analyst must be sufficiently well trained to know how to properly qualify the analytical techniques for the work to be done, properly operate the equipment, and review the data. And, if the instrument is not working properly, must be trained in maintenance appropriate for the lab or must call for appropriate assistance.

Chromatographic issues within various runs may be the result of issues that are within the scope of the analyst to deal with and reruns of the samples may be acceptable (with limits set forth in the operating procedures). On the other hand, chromatographic issues may be an indication of problems beyond the ability of the analyst or an indication of improper operation of the equipment, either of which needs to be addressed by someone beyond the analyst.

[unmgvar]

for reproducability, if you have other standards along the work and they are ok then you can use a statistical tool to show that the 1 with a bad result is an outlayer and then you will not use it
if that happens too often on several instruments you will need to show then why it happens too often and will not be allowed to overlook it.
one possibility is your vials. for example, the septa maybe. the location of the needle, maybe not in the center anymore

for the second example, again you can show that after a re-injection the given sample was actually clean
but this means you might be having a carry over problem, how do you deal with that in the future to prevent the incident from happening again?
depending on the instrument there are various possible solutions.
on one audit we caught a lab that had several issues on one application and we saw that they were weighing samples in a very narrow range, so they could re-inject from another vial and say that the problem in the problem vials had disappeared. we found it because of the simple statistical probability. if you do it too often then in the end it will be found.

so you can have for a given problem a very simple way to fix it in the beginning but remember that re-occurrence of the same problem too often will require you to deal with it in a preventing manner and not only to report it and say that " we know the problem and we live with it".

[lmh]

If you run 6 samples to test reproducibility and one is very wrong, no matter how you identify it as an outlier, it suggests that if you run 6 unknowns, one of them will also be very wrong. That is a big problem.

I personally only reject individual runs if I can see that something went wrong before I look at the peak. For example, if the vial is full of broken glass and a crushed vial insert, I have good reason to believe that this injection is unreliable (and a documentable reason too!), independent of the fact that the peak has the wrong area.

Years ago, I met a lab who (not using a chromatographic method; this was another analytical technique altogether) routinely measured each sample in 6 replicates. I only found out by chatting with their technician after he'd joined a new group, and we were comparing notes (we'd both used this technique). The technique was difficult, so they allowed for the fact that one or two, even three replicates would fail. I asked what they meant by fail. He explained that some replicates would give results too far from the mean. I asked what they meant by "too far". He explained that his boss meant more than a couple of standard deviations compared to the other points. It never occurred to him (or the boss) that they were defining the standard deviation of their measurements, not measuring it! It's also no surprise that their data were better than everyone else's.

[unmgvar]

lmh, you can do this if you have far more than 6 standards in the sequence. so if 1 out of 6 is wrong but after that standards 7, 8, 9, 10 are also good and so by the dixon-q test you have more point to claim your case

now this holds only for one point, because if you need to take out 2 points to be in the RSD of the SOP, then the work is cancelled because then you cannot say that you had a single event,
and in french there is a say "never 2 without 3", so the 3 third is probably a sample based on murphy's law

http://www.chem.uoa.gr/applets/AppletQt ... Qtest2.htm

[lmh]

unmgvar, I think you are missing my point. I'm not doubting the statistical validity of your demonstration that there is an outlier, or your approach to removing an outlier from a set of standards. I'm merely pointing out that the presence of an outlier in one set of runs (the standards) implies that outliers are likely in another set of nearly-identical runs made at the same time (the samples).

If you habitually have 1 outlier in each 10 standards, you will probably have 1 outlier in each 10 samples, but because the samples contain unknown amounts of analyte, there is no way to tell. There isn't a statistical test for outliers in a set of unreplicated, unrelated figures. On finding unavoidable outliers in standards, we really have to consider duplicating/replicating the sample measurements.

An consequence: if we calculate the expected error of a method using a set of measurements, this error is only valid so long as future measurements are processed in the same way. That means that the precision of a method calculated using 10 standards screened by a (reliable, valid) test for outliers is not the same as the precision of the same method used on unscreened measurements of samples; it's only valid if you replicate the sample measurements and remove outliers in the same way.

[unmgvar]

lmh, i agree with your point, and this is why our SOP requires standards to be run on a very narrow range so has to be very frequent and to make a large "population", that will be statistically large and also to increase the possibility of nailing a problem that is not a once occurrence.
depending on the length of the sample run, we can have a standard between each sample to at most between every 6 runs if the run time is really short. so in many cases we run more standards than samples.
for some applications like you say, we do have duplicates (preparations not just injections) for the assays and the impurities, especially the new products where there is less history of production and QC. for compounds with APIs that are sensitive we always do duplicates for the impurities.