Injection sequence

[biochemist]

One analysis we do has an run time of 3 hours (20+ peaks).
The analysis has to comply to GMP requirements; however it is from early in the production proceess, not a drug product or API (it's fermentation broth).

Could someone help guide me with the required injection sequence this analysis needs? How many injections of the standard, how many levels, how many replicates of each sample, and how many replicates of each standard level?

With such long run times many possibilities don't seem to make sense anymore..

Thanks

[Johnny Rod]

This really depends on the precision you need, and the precision of the method. If a rough answer is good enough and your method has reasonable precision, then a single injection will suffice. Are you usign UV detection? Do you really need to run a standard for every analysis or would one a day give you good enough accuracy?

On a side note, why is the run so long? Could you speed it up using a monolithic column or something like that?

[Alexandre]

Your problem is quite challenging. Can you move to UPLC or fast GC?

If not, the rule of thumb - recalibrate your system when it drifts more than 20%.
ISTD is good to include too, use bracket calibration of you have response drift.

1 point is OK as long as it is near expected concentration, if not – min 3 to cover expected range. If you have time inj samples in duplicates. Check carryover too.

But for precise characterization, you really need to do more, and this is not practical with 3 hrs runs. You really need UPLC or fast GC.

[biochemist]

The accuracy and precision requirement is not that strict, because the fermentation itself is a biological process with a lot of variability. There is an ISTD included. We can't switch to GC or UPLC, but will certainly check with the supplier if there's a new monolithic column for this analysis.Thanks for the responses and all the suggestions!

The analyte range is wide, because the components in the fermentation broth get depleted over time. So then, with 3 levels of standards, would the inj sequence be:

STD LVL 1 replicate 1 injection 1
STD LVL 2 replicate 1 injection 1
STD LVL 3 replicate 1 injection 1
STD LVL 2 replicate 1 injection 2
STD LVL 2 replicate 1 injection 3
SMPL 1
..... (more samples)
STD LVL 2 replicate 1 injection 4
STD LVL 2 replicate 1 injection 5
STD LVL 2 replicate 1 injection 6

This is already a 27-hour analysis for 1 sample and it doesn't include preparation replicates, or re-injecting the STDs to account for drift

Is the "6 injections for system suitability" really a requirement in every case?

[Alexandre]

Too much... Learn your system. Determine stability interval. Are responses stable within 24 hrs, 48, 72, a week?

As accuracy and precision are not critical use only 1 point cal at upper linearity range (need to be established).
Inject your samples one by one or in dpl if you can afford time.
Finish samples within stability interval, then inj std at the end, if drifted 20 % use bracket calibration.

Regarding depletion you need to fix/stop fermentation of the samples before analysis.

[biochemist]

Thanks.. I would still very much like to reduce the total run time, in a way which is acceptable in cGMP.

1. Could I exclude or reduce the number of injections required for system suitability, for precision (6)? Under what circumstances is this allowed?

2. The samples don't degrade, they are frozen .. it's the concentration of the analytes in the samples, which could be between 1% to 90% of the calibration curve range, depending on the manufacturing process. So am I correct that I don't have to put in several levels of the standard, as long as I prove during validation that the calibration curve is linear?

3. Which is preferable, a replicate injection, or a replicate preparation?

4. Since the FDA's requirements for 6 replicate injections originate from the same vial, should they have reduced/no weighting in generating a calibration curve?

Sorry for so many questions, but the regulatory requirements have always seemed somewhat arbitrary to me, so I'd very much appreciate everyone's opinions.

Also forgot : THe sequence should include the blank. So it could be up to 30 h now

[Peter Apps]

Having been in a similar situation - we got a 4 hour method given to us for a plant extract I would strongly suggest that you invest time in getting the separation optimised for time - even if you have to run separate methods to get all of the peaks it could still work out a lot quicker that grinding them all out over 3 hours. At the rate you are going now the bugs are growing quicker than you can run the samples.

Peter

[Alexandre]

I would accept a periodical demonstration of SS using your samples, and for regular routine run SS performance solution. Prepare some model mix of same/similar analytes for quick SS check.

No multi level calibration is needed if there is no stat significant difference in slope/intercept.

Repeat of the preparation is better as you want to be sure that the fermentation process is OK, not the injector.

As you are not quantitating final intravenous preparation but only intermediate slurry and have such technically difficult system the exceptions can be made, but they have to be reasoned, the risk assessed and documented. Ask your QC dept. to contact local FDA office for help.

Blank is needed to demonstrate that the system is clean and that that there is no carryover, at least periodically.