Evaluate column variability as part of method development?

[DanielWH]

I watched the latest Chromacademy Tutorial on mixed mode chromatography (if you haven't seen it, I would say it's pretty good and it's free this month). Some of the columns they described seemed fairly elaborate and it got me to wondering whether these new types of columns have more variability than traditional C18 columns. I imagine they do, but I could be off base. (By column variability, I'm not talking about between columns from different manufacturers but within columns that have the same manufacturer and part number.) I posted a similar question on their forum, but it doesn't seem very active over there, so I thought I'd see if anyone here had thoughts about it.

When developing a method, do you evaluate column-to-column variability by, for example, requesting several columns from the manufacturer that were packed with different lots of packing material or at least packed on different days? Or do you think it's generally not an issue? I know column manufacturers like to talk about the umpteen different ways their columns are checked, but no doubt there's some variability in the manufacturing process.

What are typical sources of chromatographic variability in the column and what are some good strategies to build ruggedness into the method to control for slight (or not so slight) changes in selectivity that you might see from columns which are nominally the same but packed at different times possibly with different lots of packing material?

[Gerhard Kratz]

It's recommended to use minimum 3 columns from different lots for validation. All manufacturers have a very high standard for QC of packing materials and the packing process itself.
Some column manufacturers offer so called Validation kits with 3 to 4 columns, same kind of packing material and same dimensions, but different lots.
Older packing materials were manufacturered in a different way as the new developed phases and cannot compared with new column materials in terms of varability from batch to batch.
When you get a new column please do the same test as mentioned on the COA. You should get the same result and can compare with the previous purchased column. If you see significant variability please check with the tech support of the manufacturer. If that test is ok next test should be with a standard solution of your target compounds to check number of theoretical plates, resolution and retention time. Keep this chromatogramm and compare also with the previous purchased column.
Any time you do such testing procedure please use ONLY fresh prepared buffer solutions, fresh opend bottle of ACN etc. and check pH of the buffer with a fresh calibrated pH meter!

[Peter Apps]

I suppose that to some extent this depends on the scope of the method validation. At one extreme would be a method that has to run hundreds of mucky aggressive samples, where column changes would be frequent, or one that has to run on multiple instruments, each with its own column. In these cases you have to show that the method works on different columns. At the other extreme would be a small (20 - 30 sample) batch of something nice and clean that runs in a "soft" mobile phase, and where the method is going to be used once on this batch. In such a case it would not make sense to demonstrate robustness to column changes - but strictly the method itself would have to specify that particular individual column was to be used.

Peter

[danko]

Mixed mode stationary phases provide more “strings to play on” but at the same time require somewhat deeper understanding of the various parameters at the chromatographers’ disposition. And variability is almost a natural consequence of the variety of the parameters one should understand and control.
So, in this particular case the other factors – beside the different columns - are as important. F. eks. salt/buffer concentration , pH, temperature, content of the organic modifier in the mobile phase etc. should be tested and valid ranges for these should be established before everything else.
Column variability could follow the testing of the parameters mentioned above.

Best Regards

[DanielWH]

If we wait to test for column variability until we have established valid ranges for method conditions, don't we run the risk of finding out that we did method development on an atypical column?

For example, I had a column which I had done some development work on until I had a decent resolution for a critical pair. To check that I hadn't done anything to the column I took that column off the system and put on a brand new column of the same lot. System, mobile phase, and samples remained the same. New column gave similar resolution (about 0.1 or 0.2 less, if I remember right). But then I put on another new column with a different packing lot and the resolution had dropped by 0.7! The separation between the critical pair was not acceptable. A fourth new column with yet another packing lot gave resolution which dropped by 0.5 from the original column – also not acceptable. Original column back on and I'm at the same decent resolution again. In fact, all of the resolutions were reproducible for each column, across injections and different samples. Further method development allowed me to bring the resolution differences between the three different lots of packing material to about 0.2 or 0.3.

That was one of the experiences that led me to think about column variability as almost one of the first things I'm concerned about, rather than taking column reproducibility for granted. I think this is even more true if you use modeling software like Drylab or Fusion AE.

[DanielWH]

Gerhard – when you say variability of older packing materials cannot be compared with new column materials, you mean we should not have the same expectations, right? Should we have higher expectations for better reproducibility of the newer materials or older materials?

I was thinking that the newer packing materials are more elaborate, more challenging to make, and therefore may be subject to more variability. But of course manufacturing technology has also advanced and column manufacturers might have better control over their production, so maybe not.

Also – I know you're right that we should do those QC tests, but it's hard to find the time!

Older packing materials were manufacturered in a different way as the new developed phases and cannot compared with new column materials in terms of varability from batch to batch.

[Peter Apps]

Also – I know you're right that we should do those QC tests, but it's hard to find the time!

[/quote]

There's an old saying; "If you can't find the time to do it right, you will have to find the time to do it again"

In the scenario that you discuss above - would it have been better to have done a quick and dirty validation on only one column, and then have been caught out in the middle of production QC (or whatever) when you changed to a new column ?

Peter

[HPLCaddict]

Generally speaking you might expect less lot-to-lot variability with modern columns as compared to older ones. Not primarily because modern stationary phases are better than older ones, but because the column manufacturers learned over the years and gained more and more expirience concerning the synthesis of the stationary phase itself and the packing process of the column. GMP, GLP and rigid quality control are a must for the manufacturers today.

[danko]

Hi DanielWH,

Quote:

Further method development allowed me to bring the resolution differences between the three different lots of packing material to about 0.2 or 0.3.

There you go. You optimized the conditions and suddenly – you can use all the columns (of the type) you have.
But you’re right about the risk of trusting one single column not knowing what the next lot will offer. It’s all about the hen and the egg – isn’t it? Which came first? Let me put it this way: If you buy a lot of columns trying every and each of them using less optimal conditions you’ll end up concluding (potentially wrongly) that the column’s not good. The difference is the amount of money you’ve spent buying those columns
Having said all that I agree that some column manufacturers offer more reliable and consistent products than others. I’ve logged (in my head) over time the “good” and the “bad” brands and usually don’t start developing on the latter products.

Best Regards

[Gerhard Kratz]

Hi DanielWH,

I always say that there are no bad columns on the market, it all depends on the application which column will show the best performance.
Old packing materials vs new packing materials, very difficult to say which is better. When you come across a monograph which states that a µBondapack C18 column was used, irregular particles, and not endcapped, and you will try to get the same separation on a new developed C18 column, spherical particles, endcapped, in most cases you will fail.
Different ways to do the synthesis of the silica, new developed silanes to do the bonding, and of course the purity of the raw materials and silanes are responsible for different performance of old and new packing materials and the different brands.
To select a column is always critical. You might prefer the one or the other manufacturer. My recommendation is to check out the hydrophobicity tables published from the manufacturers. That will give you a good idea which columns are similar. That means if you have difficulties to get a good separation it makes no sence to use another column with the same hydrophobicity.
Try to get as much as possible information about your column. Look on the certificate of analysis, check the content of metal impurities.
Good luck

[KM-USA]

We've looked through official documents on test method validation (and so has our QA department), and we don't see anything there about trying out on multiple columns of different lot numbers of packing. We do see some manufacturers actually offer such, and even call them method validation kits.

Anyway, our company has decided that such multi-column studies are a good idea but not documented as "a necessity", and that such testing belongs in pre-validation test method development, so does not show up in our official cGMP protocols or validation reports.

We simply call and ask for columns of different lots of packings. Agilent and Thermo have been able to supply with no problem. Grace Discovery told me a couple years ago that they cannot do this any longer (as they could do when it was Alltech), that the parent company didn't understand "why" it would be necessary, that it was seen as additional cost of preparing and storing batches. I don't believe that we have contacted other column suppliers about "three lots of packings", so I cannot comment on them.

[HippyLabRat]

I think Waters provides validation packages too. They're super friendly about getting you what you need to do the job.

All though in my experience, their Symmetry line could use some better QC. Holy column to column variability...

[Vlad Orlovsky]

Our spec for lot to lot variation is 3%, but we are trying to stay below 2%. Each batch of mixed-mode column is tested on a t least 8 different compounds and 2 different mobile phases before being released. Since mixed-mode chemistry is more complex there ia limited number of cases where the reproducibility will shift to 5%, but there is always an option to reserve a specific batch (we have companies who reserved batches for years). If you develop a method with understanding of column chemistry and follow column care instruction you should not have problem with reproducibility of SIELC columns. If column care instruction states "Do not use alcohols, as you can esterify stationary phase" then don't use alcohols. Mixed-mode interactions are more complicated then single mode, you need to consider more variables, but once you learn how to use then it is an easy task. If you have single ligand design of mixed-mode columns (like we do) and not a physical mixture of two or three silica gels you chances of developing reproducible method increase. We always try to develop method with resolution of at least 3 and K' of at least 2.5.

[DanielWH]

Vlad – what would you say are some of the main sources of column-to-column variation? I certainly don't expect 0% variation, but I am interested in exploring ways to adjust method conditions to compensate for the inevitable. In the example I described above, I attributed the differences between different packing lots to differences in residual charges on the stationary phase, since the buffer concentration was key to minimizing differences between packing lots. But I don't know if that's really true or just a nice story.

In a separate conversation with a column rep about a different issue, the rep said (about a competitor's hybrid phenyl column) that differences in stationary phase coverage on the particle could cause differences in column selectivity. I do not know what changes I could make to a method to minimize those differences, except to switch to a different column.

We've looked through official documents on test method validation (and so has our QA department), and we don't see anything there about trying out on multiple columns of different lot numbers of packing. . .

Anyway, our company has decided that such multi-column studies are a good idea but not documented as "a necessity", and that such testing belongs in pre-validation test method development, so does not show up in our official cGMP protocols or validation reports.

I would consider it part of method robustness/ruggedness, and something that should be built in to the method a la Quality by Design.