validation troubleshooting, please

[ym3142]

one method deveopled 11/2004 is validated now in our lab with Agilent Zorbax 300SB C8 250mmx4.6mm column(we are working on small molecules, MW<1000), isocratic mobile: buffer(NH4H2PO4 47mM, pH adjusted to 2.9~3.1 with H3PO4):MeOH = 85 : 15; 60 C degree; Flow 1.2ml; injection 5 ul(0.025mg/ml). Two critic peak seapration changed from previous resoluiton 2.9 to now 2.3. And this later resolution was out of a new column.
The only thing I can think of that is different between the developing and validation runs is that current new column has been stored long than those last year. Any one of you has this kind of experience?
What else could be different?
We did check the resolutions under pH=3.6 and 2.8, the former 1.6; later 2.4.
Any suggestions?
Thanks

[tom jupille]

First thing that comes to mind is, what aspect of resolution changed? are both peaks at the correct retention time, but wider, or is the peak width the same but one or both peaks have moved?

If the former, column aging is certainly a possibility, as is column-to-column variation (are both columns from the same lot?), but I'd look first at "pilot error": are all the fittings properly assembled, is the tubing the right diameter, etc.

If the latter, mobile phase problems would have been the first suspect, but you've eliminated that with your pH tests, so yes, column deterioration or lott-to-lot variation is a possibility. I'm assuming that you're using the same instrument. If not, check the temperature (not the setting, the actual temperature of the column) and make sure that both instruments are set up to preheat the mobile phase the same way.

Next question is how much Rs do you need? For regulatory purposes, Rs > 2.0 would meet FDA and ICH recommendations. If the results are otherwise acceptable, I wouldn't be overly concerned about the change. On the other hand, if you really need the higher Rs (e.g., peak shape or size issues), then you will have to chase the cause of the problem down.

And yes, we have seen column changes on storage. If your separation is extremely sensitive to column chemistry, this could be an issue. It also raises the question of lot-to-lot variation, which should be addressed as part of the robustness validation.

[ym3142]

Tom:
Thank you so much for your suggestions and also for all others you wrote and I read.

First thing that comes to mind is, what aspect of resolution changed? are both peaks at the correct retention time, but wider, or is the peak width the same but one or both peaks have moved?

Former RTs were 4.6 and 6.0; current were 4.3 and 4.6. Peak shape looked OK.

If the former, column aging is certainly a possibility, as is column-to-column variation (are both columns from the same lot?),

I have tried two new columns. ANd our lab does not have the column lot recorded. People are saying we do not want a method working with one column but not with another one from different lots or batches.

but I'd look first at "pilot error": are all the fittings properly assembled, is the tubing the right diameter, etc.

Unlikely, since all the machines are being used every hour by dozens of chemists. For my projects, two of us haved used four machines with four columns.

If the latter, mobile phase problems would have been the first suspect, but you've eliminated that with your pH tests, so yes, column deterioration or lott-to-lot variation is a possibility. I'm assuming that you're using the same instrument. If not, check the temperature (not the setting, the actual temperature of the column) and make sure that both instruments are set up to preheat the mobile phase the same way.

As I said , we are using different machines. The temperature may be an issue. On one of robustness run, complet overlapping was found at both 55 and 65 degree, which might be not valid results since it does not make much sense to me. Because a robustness is a series runs the LC may not have gotten enough time to change the temperture. I (maybe We) do not have idea how long it takes the LC to increase from 60 to 65 or decrease from 60 to 55. What do you think ?

It was said the highest limit for LC is 65 degree for our Waters systems.

Next question is how much Rs do you need? For regulatory purposes, Rs > 2.0 would meet FDA and ICH recommendations. If the results are otherwise acceptable, I wouldn't be overly concerned about the change. On the other hand, if you really need the higher Rs (e.g., peak shape or size issues), then you will have to chase the cause of the problem down.

Rs >2.0 is good enough for us. The reasons for me to investigate this is that 1) why we fail to repeat; 2) Is it possible that this time it changed from Rs 2.9 to 2.1 and next time it changed from 2.1 to 1.6; 3) we have a specififity to take care of.

Thank again and have a nice weekend,

[tom jupille]

Okay, either I'm misinterpreting something or there is a typo in your numbers.

In the original post, you said that resolution decreased from 2.9 to 2.3 (this is a bit more than a 20% decrease). This doesn't fit with the retention time numbers in your last post, where the retention time difference decreased from 1.4 minutes (6.0-4.6) to 0.3 minutes (4.6-4.3).

That question aside, you have the following variables to look at:
- column lot
- column storage time
- instrument configuration
- instrument equilibration time.
- mobile phase preparation.

If this were my problem, I'd "divide and conquer":

1. start with a new column of an identifiable lot number on one instrument, and track the results over several batches of mobile phase to convince yourself that the results are consistent with respect to the mobile phase.

2. come up with a way of measuring mobile phase temperature in real time. The best would be a tee with a thermocouple. I'd try to get two sets of measurements: just upstream of the column and just downstream of the column. Check what the real temperature is (not just your set point), and check that your incoming mobile phase has been heated to the column temperature before it gets to the column.

3. Once you're getting consistent results, move the column and temperature sensor to the second instrument. Repeat with the other instruments.

If you get instrument-to-instrument variations (which is what I suspect is the most likely cause of the problem), then find and fix the cause. You may have to set each instrument to a different temperature in order to get the same on-column temperature, and you may have to add preheater tubing to warm up the mobile phase. Check out the posts in this thread for some additional input and comments that may be relevant:
http://www.sepsci.com/chromforum/viewtopic.php?t=2568

4. Once you can get consistent results from instrument to instrument, get a new column from a different batch and check that. Hopefully, the results will match. If not, you will either have to "cherry pick" column batches or else re-work the method to minimize the variation (either a different brand of column or change conditions to be more robust). Assuming you can establish that the lot-to-lot variation is acceptable, then I wouldn't worry about the issue of using a year-old column, just buy a new column next year. The price of a new column is small compared to the cost of running the lifetime test!

[ym3142]

Tom:
First of all, I like to thank you. These are great guidances.

Secondly, I am sorry. As a typo, one of retention was 5.0 instend of 6.0min. (4.6-5.0).

I still failed to achieve the 2.9 resolution when repeating the method. The possible reason should be the combination of the following: 1) pH, a. the buffer pH was found to increase, such as from 3.0 to 3.6, after adjusted to 3.0 by H3PO4 and stored for some time; b, pH effected peak tailing, at 45 C, pH= 3 tailing = 8.8~9.2; pH=2.5 tailing 0.97; pH 2.1 tailing 1.03
2) temperature, a, the upper limit for these LC is 65C, and only calibrated to 45C; b, the higher temperature the less fronting;
3) maybe a different column properties.

Tom, as you may know, people are always pursuing high productivity, I may not have the luxury to try out all your advices. But they are good guidance and I will try whenever possible. Thanks again,