Chromatography Forum

Neither USP 621 nor in Ph.Eur. 2.2.46 clearly indicates if reference standards (for quantitation purposes) have to be injected every HPLC sequence. May I use the same calibration curve over a predefined period (e.g. 1 week) if data shows that the calibration curve do not significantly change during this time?

I would expect to see a check standard run in every sequence to demonstrate that the calibration curve is still valid.
A lamp starting to fail or cell window becoming dirty could invalidate the curve. Also this will help if you have an anomolous result.

In the USP <621> it mentions under the system suitability section it talks about % RSD of standard injections _ 3 to 6 injections. We typically perform a minimum of 6 injections of each level of standard. Note different pharmacopeias have slightly different requirements - hence we try to match the requirements of all by trying to match the requiremnts of the strictest. However I must admit it's not very clear in the text - could defintely do with improving to make it unequivocal in plain English!

maybe not in that sense but both FDA and nowdays both USP and Ph Eur mention bracketing or SSTs cover the whole analysis, even though they do not say you have to repeat all SSTs in end.


We had an intresting discussion in a recent FDA audit, one of the auditors was about to give us a 483 for not doing all 5-6 RSD injections prior to sample injections (we do bracketing), a common 483 from that auditor we learnt later, but the auditor had to back down after counsulting with office as USP recently changed wordings which we argued. So sometimes it works your way.

we run 5 or 6(Depending on %RSD requirement) precision injections at the start of each run, and then do a set of 2 bracketing standard injections every 12 injections. The 2 and and 12 seem like arbitrary numbers to me. The bracketing standards must be within a set % difference from the precision injections and the bracketting are used for quantitation.

Looking at the USP it seems to me that once system suitability is established, the system is considored suitable until conditions change.

So, as long as you are doing the same run(Standards and mobile phase) and you don't stop your system, I would think any length of run or multiple sequences would be OK per USP.

OK isn't always commendable.

It is always better to do GOOD SCIENCE than to do the minimum of legal requirements.

I know cost is a factor.

But legal costs can exceed lab costs by SO MUCH and they have penal incentives as well.

I would not feel confident unless I could demonstrate my work was accurate for the lot, batch, sample, whatever, I had analyzed.

I would always run stds before and after a sample, batch, or lot so show if for no other reason my instrument continued to work and to measure as I had intended and from which results I gave a report.

Results that are given without a demonstration of confidence are worthless, no matter how much money was saved in producing them by minimal means.

If you can't prove it to yourself, how can you prove the results to someone else. Be your own devil's advocate. And be sharp about it, too.

Let's say I run stds and then run 1000 samples followed by another set of stds. Well, if the stds match then I feel good about the 1000 samples. But WHAT IF THEY DON'T MATCH? This is where you need to use good science and good thinking on your part. I would always try to insert stds every new shift of work at a minimum (8hours) or after each lot-batch-sample was tested so I would know that FOR THAT PART of the work, the WORK WAS DEMONSTRATED TO BE ACCURATE.

Let your conscience and the science be your guide. Don't overdo it, but don't be a slacker, either.

best wishes,

Rod

Enlarging on Rod's comments.

If a QC measure fails (in this case the standards), all the results since the last pass have to be discarded. At the minimum this would mean re-running the samples after the QC was brought back into spec. In an entirely plausible scenario it could mean product being put on hold, or recalled. The longer the interval between QC checks, the more samples have to be rerun after a QC fail, and the further along in the system any product is that relied on the discarded results. In any scenario that involves a product recall I can absolutely guarantee that you would not be getting a bonus for saving lab costs by running QC too infrequently.

Peter