How do you do method development?

[Uwe Neue]

I would like to get some inputs on the way everybody is going about the development of a method. I am mostly interested in the early steps (not validation etc.).

Where do your start? (e.g. search the literature, ask on the Forum, what else?)

How do you select the column and the mobile phase?

What do you do next? Do you optimize a method? Optimize for what?

I am sure that there will be many different answers, and I am curious about all the ways that people do this.

[DR]

I'm in Pharma and we start most of our dosage forms w/ our own APIs, so there are usually good methods available as a starting point for dosage form versions. If there are no "in-house" methods, I check the compendia. If there's nothing promising there, I get out the ACN, TFA, Water & Drylab.






(Can I put "Uwe Neue once asked me how I develop LC methods." on my résumé now?)

[Consumer Products Guy]

Uwe - we first look at the structure of the analyte and gather any information we can on it: does it have UV or fluorescence, is it small enough to be volatile, or made volatile through derivatization? We check its solubility in different solvents such as methanol, ACN, DMF which are amenable to HPLC or GC. Sometimes the analyte supplier can provide their test procedures but most times these are not appropriate (Note: this week a supplier provided [1] a packed-column GC procedure using internal standard, indicating that they likely didn't have an autosampler and [2] an HPLC procedure using 10 micron 30 cm semi-obsolete column, also using internal standard indicating lack of autosampler.) Oftentimes we'll do a literature search, why re-invent the wheel, but mostly stuff in our field is not published, kept trade secret or as proprietary. I might do a "Google" search with "HPLC" or "GC" but usually don't find too much. For columns, oftentimes we'll start with columns we have on hand, such as a modern Type B C18 such as 3 micron Betasil C18. For mobile phases, we usually start with acidified water and either methanol or ACN. We can use other detectors like RI, ELSD, or conductivity, depending upon the analyte's structure. After we get good retention, peak shape, run time, etc. for the analyte itself, we then try a placebo product and a lab-made product to see if there are matrix components that interfere and determine resolution of the analyte from any close-eluting peaks. Then we'll tweak the chromatography if necessary. I have searched the Forum here a few times, but our stuff is rarely encountered by others here; so I rarely ask the Forum here for method development advice (maybe I should use that feature more often). If we do find a supplier or literature priocedure that looks good, we'll buy that column and set up exactly like written, and stay with that if it works; in any case, at least that would provide a starting point with a brand new column so we could then assess any potential improvements. After that, if it's a procedure that requires validation (a regulated product) we begin validation protocols, which take maybe four weeks, purchasing then using columns of three different lots, etc.

[Bruce Hamilton]

I consults for industry, mainly bioactives, and have developed about 50 HPLC methods in the last year, but only on HPLC-UV, as i don't have MS.

Assuming I have been given very clear instructions ( eg an achiral method, and they want to see x, y, z at < 0.1% ). If not, the very first step is to discuss what is possible with the client, and ensure we both know what the desired outcome is.

First - I look at it's structure ( if available ), then for chemical and regulatory names, and any closely related relatives that may be more common. All of those may be used in the searches.

Second - search Martindale ( tells me if it's in a pharmacopoeia ), Merck Index ( tells me if it's common ), Pubmed, then the Web. Pubmed and Web searches are " HPLC xxx"- inclusing any closely related compounds. If the chromatographic information isn't sufficient in any suitable free abstracts, I may interloan some articles.

If the molecule belongs to a family, such as Vitamins or Lipids, I would also consult the standard chromatography texts ( such as Christie for lipids ). Same applies for standard topics, eg Water ( AWWA), Food ( AOAC, Food chemicals Codex ), Oils ( AOCS ), Petroleum ( IP, ASTM ), etc etc

If if all that fails, I use the library, and search using pay-to-use services - such as Scifinder, CAS, etc. I don't want to reinvent the wheel, so efficient searching is important.

I should note that I usually don't use equipment manufacturer's applications, as generally it's best to use compendial methods if possible, and many of those methods don't relate to separation of likely precusors, degradation products, or related substances. I may search some chemical catalogues ( aldrich etc.), just toe ensure I have not missed an obvious name.

If I don't know the compound, then, assuming it is soluble in 1:1 acetonirtrile : water, I go straight to 250 x 4.6 5um Luna C18(2) HPLC gradient with 90:10 to 10:90 with H2O + 0.1% TFA, and CH3CN + 0.1% TFA, column at 20C, detector at 230, 250, 280nm.

If it's not soluble in water:acetonitrile, but is soluble in 1:1 nC6:EtOH, then I will most likely try a similar sized cyano column with normal phase solvents, with or without TFA or HOAc

I then look at peak retention, and shape, and may select either an alternative column and/or modify the mobile phase. Target is have main peak in 5-15 minute retention window for 250x4.6 5um, and the method devlopment and analysis completed in a morning, but usually takes closer to a day.

I hope this helps,

Bruce Hamilton

[Uwe Neue]

Thanks for your inputs, and I am hoping to get many more...

To Consumer Products Guy, DR and Bruce: What is a good run time for your method? Is 15 minutes OK, and 1 hour too long? If you got a reasonable separation, do you optimize further (for example for more uniform peak spacing), or is this a waste of time?

[Consumer Products Guy]

Uwe: great question. For our R&D lab, run time really doesn't make too much difference, as we have our autosamplers all humming overnight, and results are complete by the next morning. In reality, most of our R&D HPLC run times are less than 10 minutes; most are isocratic, with our gradient runs including organic clean-up and re-equilibration taking about 25 minutes. Yes, we would optimize to make a 10-minute run 6 minutes if there were no downside to such and could be readily done; oftentimes our products may contain fragrances which might interfere if we speed things up too much. We don't use any software like DryLab to help us in our test procedure development either. Now, for our production facilities, run time is much more important, as the product is sampled before it gets packaged or QC and Manufacturing will squawk, so we try to make sure those assays are less than 10 minutes each if possible ("you want it quick or correct?"). But we may have a whole bunch of samples at one time in R&D (e.g. stability studies) where Manufacturing has one batch at a time and must turn over the product quickly. We do try to use the same procedures, just that we do a lot more stuff here at R&D than Manufacturing QC does. We primarily use narrow-bore HPLC columns at all locations to save solvent costs and disposal costs.

[Bruce Hamilton]

In most industries, HPLCs are used in three stages.

1. Raw materials - Did I get what I might get around to paying for?.

I assume stock control is OK ( big assumption, "we need release tomorrow, and the sample arrive in the morning" and generally the sample will actually arrive at 1600, if lucky... ). The HPLC method is only a small part of the quality checks required to release the product, so run times usually aren't too important provided stock control is good.

These are usually well defined, often have to have gradient method because the precusors and contaminants can be quite diverse, For these gradients ( which can cover quite a range of mobile phase polarity ), I like to have a cycle time of 45 mins +-15 mins, and normally default use is a C18 column, but often special-purpose columns and conditions are used. They are usually optimied for resolution and sensitivity.

2. In process controls - is it ready for the next step yet?.

These are inevitably time-constrained. 10 mins cycle time is desired, and usually the precusors and products are fairly similar, so even isocratic can be used. Generally very shallow and short gradients, just to keep the cycle time down - I'm a big fan of the old wives' tale - maxiumum 10%/min composition change.

The method must be availble on demand, so cycle time is critical, and customers may only want to see the changes in one peak, and ignore the rest. Typically, I use a 150mm x 4.6 with guard, and target retention of 6-8 minutes, with last peak at 9.5 minutes. If people want more resolution of minor peaks, a 250mm column is used, and they accept the additional time.

One critical choice about these analyses is the HPLC column robustness, in general, a C8 column will croak if fed with junk, whereas a C18 may survive. I pay quite a bit of attention to column durability, because you really want to minimise any sample preparation for IPC. Because of this, sometimes a 4.6mm ID is preferred, just because the guard copes with more junk before dying. Often the junk isn't visible at the wavelength used, or even visible at all - I may choose a non-optimum wavelength to lose interfering junk, rather than optimise the method.

In some IPC cases, I have to consider the sample matrix, and choose column/mobile phases that are compatible. The simplest and fastest may be the best, even if guards only last a few runs. Internal standards are seldom used, as sample baselines often look like the Himalayas

3. Final Product

Have we met or exceeded the product specification?.

Often this will also be the raw material for client, so the methods tend to be tailored to that. Long gradient or isocratic ( isocratic run time is 3x actual peak retention time ), sometimes with differing methods looking for precusors, related substances, degradation products, but often a wide gradient of 45+-15mins keeps everybody happy.

These are similar to compendial methods, with all the associated nausea. I tend to be able to choose any column that provides reasonable separation, and column robustment and method sensitivity ( <0.1% ) usually aren't issues,

These are well defined products, so behaviour in acid/base mobile phases, temperature etc, are all checked out, as well as ensuring all peaks of interest ( usually to 0.05% ) are resolved. I don't want samples to degrade during testing. If all peaks aren't resolved, then another method may be added ( eg related substances ) to address the need. These will have the most method refinement, even without any cGMP stuff.

Same sort of instrument timeline as for raw materials ( 45+-15 minutes ), as some other tests usually are required on such samples. These methods are the most optimised for resolution, accuracy, and repeatability.

Obviously, if labs are instrument limited, then faster methods are only advantageous if similar systems and columns are used, as you don't want to keep swapping. However the ability to not croak with real life samples is important, as is minimal sample preparation.

In most methods, samples would be dissolved in mobile phase, and playing with column temperature, mobile pahse pH, etc is constrained unless I know the sample is stable. Rejecting product because of an analytical artifact isn't considered as agood deed by most managers

Note that places that setup to run studies, eg clinical trials, or specific analyses ( environmental ) will only want methods that look for one or two peaks, and they want even faster cycle times than IPC. That's usually 3-5 minutes, and with MS. Method development may involve an Internal Standard for those.

I hope this is helpful,

Bruce Hamilton

[Noser222]

We work mostly with carbohydrate polymers, so we usually look for a size-exclusion column that fits our needs, or if we're doing monosaccharides composition we have most often used the polystyrene-divinylbenzene sulfate in the H+ form with dilute acid mobile phase or an amino phase and start with 80% acetonitrile.

On the rare occasion I do reverse phase, if there is no literature method, I usually start with a generic MS-friendly mobile phase of 0.1% acetic or formic acid with a 95% water to 95% methanol or acetonitrile gradient.

[Uwe Neue]

Bruce,

Thank you for your very good and differentiated answer. I understand that different environments require different run times.

I encountered a situation in my early days of HPLC, where I was participating in a Round Robin study on an essay for some epoxy materials, don't recall the details. The orginal method took about one hour, and the chromatogram was full of peaks. After looking at the chromatograms, I asked the marketing guy who gave me the stuff, if it was OK by him, if I did the method in 10 minutes. He said that this would be OK, probably thinking that I was crazy, because the chromatogram was complicated. I came back a few days later, indeed with a chromatogram with a 10 minute run time, and all peaks resolved. I never forget his face (~30 years later). He looked at me as if I were a Martian. He took the method to the committee that wanted it tested, but I never heard about it again. I actually doubt that they would have adopted my method, and it is even possible that he never submitted my method.

Anyway, it may be easier today with automated methods to accept long methods. In the times of manual operation, a method that took much longer than smoking a cigarette was no good, at least as far as I was concerned...

[Uwe Neue]

Several of you have mentioned that they start with literature methods. I have to admit that I never did, when I did this stuff. I always started from scratch, figuring that it will take longer to research a method than to get some basic information about what I will be up against. When collecting the chromatograms, I got enough information to take the method to the next step.
Another reason for starting from scratch was that literature method rarely matched what I needed to do. There may be an impurity profile, but I was interested in the drug and its metabolites in urine, or vice versa. Or, I looked at a suitable method, but it had problems. Of course, in my case, I was always interested in developing a method based on products from my company, which is of course not a constraint encountered by most people.

[Bruce Hamilton]

Several of you have mentioned that they start with literature methods. I have to admit that I never did, when I did this stuff. I always started from scratch, figuring that it will take longer to research a method than to get some basic information about what I will be up against.

I think that's probably because you understand chromatography far, far, more than I. The reason for using literature methods these days is because the research is now so quick. I can search online and find a free abstract or even full paper in 10 mins.

That information usually gives me an idea about the mobile phase, column, and detector wavelength that works for the principal peak. Many of the more recent methods published are stability-indicating, which using means even more useful information.

Getting the full paper itself takes a 15 minutes visit to the library to download if it's in a popular journal. Papers, also, importantly will often mention issues with stability or difficult separations of analytes. which also helps.

Today, I came in to find the overnight run showed the novel product I'm currently analysing degraded in 1:1 acetonitrile/water by 5% over a two hour period. Oops, more work to do.

Bruce Hamilton

[Noser222]

I should add something to mine about looking up a published method before starting anything. If I have a column that I think will do the job, I'll probably play with a few injections while I'm digging up information.
Since I'm not yet 30, it's hard for me to imagine the days when you actually had to leave the office to get some info., but since I'm typing this from our facility in Costa Rica, it's easy to remember 3 years ago when we had to use <<gasp>> dial-up internet here.

[AA]

First, buy this book:
Practical HPLC Method Development, 2nd Edition
by Lloyd R. Snyder, Joseph J. Kirkland, Joseph L. Glajch
Second, read that book cover to cover.
Third, adapt and modify the techniques in the book, the suggestions listed in this forum, the info for the scientific journals (some papers are good, some are not).
Fourth, trust some of you gut instincts.
Fifth, experiment, experiment, experiment changing one thing at a time until you beging to understand what is happening during the seperation.

At the end of it all, you should have some kind of a method.

[DR]

I don't optimize peak spacing or anything like that. I set up a method to run as fast as I can while still maintain a critical peak pair resolution of ~2.0 with a minimum k of ~1.5. These targets are a little high, but I have to remember that others may have to run my methods on systems with more dwell volume, extracolumn volumes etc. than I have.

For assay methods, I frequently get into the 5-10 minute range. For stability indicating methods, I'll tolerate run times near an hour, but I certainly prefer 20-30 minutes. For a method with a RT>60 minutes, it's time to start looking at using a second method to get at those late eluters. This rarely happens.

[bgiles]

Hire a contract lab.