How do you do method development?

[bartjoosen]

I most of the times start with a quick literature search (5-10 minutes), to have an idea about the component.
Then I setup a system with 3 pH's and organic gradients (in case of acids/base), if it's a neutral component, water-organic should suffice.

With this gradients, I import in Drylab and look for the next step.


Bart

[Uwe Neue]

Bart, you are the first one who is playing with some of the variables pretty much up front. Are you using only one organic modifier or are you using both methanol and acetonitrile? This would make 6 gradient experiments per column to start. Do you use more than one column for this?

[bartjoosen]

What I initially test depends on the time:
If I setup such thing just before weekend, I will try 3-4 pH buffers, 2 organic modifiers, 2 gradients and 2 temperatures at 4-6 columns.

If I setup in the morning I will try 3-4 pH, 1 gradient and 1 temperature, on the most promising columns first. I can review my runs during the day, so I can quickly adjust some things.

The weekend setup is more a "shoot as much as possible, maybe I hit" approach.

Bart

[jzt]

my work focuses on impurity separation for drug substances, in the clinical phases.

I never started with a literature search. all the compounds I worked with are new chemical entities. At the beginning there’s very little information: chemical structure of the main compound, calculated pka, some limited solubility data, and the synthetic scheme to make this compound. My goal is to separate as many impurities as possible, most of them the structures (and chemical or physical properties) are unknown.

I usually start with column screening using 6 to 9 columns. These include a C8, a CN, a phenyl, a polar embedded, and several C18 columns. The column dimension used to be 4.6 x 150 mm, 3 um or 5 um depending on availability. Now I start with 4.6 x 50 mm columns with sub-2 micron particles, if available. Since the impurities often have closely related structure, sometimes we rely on the slight difference in selectivity from different brands of C18 to provide the desired separation. Not all C18 columns are created equal. our methods specify the exact column by its part number. we don't spend the time to find a "equivalent column".

I will normally set up two gradient times at 15 and 45 min (from 5% to 100% B); two temperatures, 30 oC or 50 oC (or 40 oC); both acetonitrile and methanol as the mobile phase B (with or without an additive such as TFA). This is done overnight or over the weekend. The fun starts when I have to go through all the chromatograms. I will compare number of impurities observed, resolution of critical pairs, peak shape of major peaks, and how peaks moved with regard to gradient or mobile phase changes. I never know how many peaks I should see, just try to pull out as many as I can. Sometimes I used DryLab or LC Simulator. I don’t believe in the so called “automated method developmentâ€

[Uwe Neue]

JZT: So what do you do to satisfy you that your method is complete? Or even, how do you judge that one of your many runs is better than another?

I can see one approach where one is simply counting the number of peaks... More peaks is better, of course. Plus one can do purity profiles (either classically with a PDA or with MS). That is a lot of work. Do you do that?

[koen_shimadzu]

Uwe,

Personally I really like this topic which you have started. Most probably the answers not really surprisingly.

In my opinion there are several philosophies to approach MDV:
o People that rely on there experience.
This are people like you. You know all the fundamentals of chromatography very well and can match the conditions with the compounds/sample of interest.
o People with average knowledge, but a limited time
These people want to have a starting point and from here optimize there method further. In these cases literature search is important.
o People that apply systematic approaches.
In most cases this takes more time. They take all factors in account and work to the method which they desire ( e.g. fast, robust). In most cases a system with column selection and software like DryLab is used.
o The traditional trial and error approach.
Normally this starts with an enthusiastic analyst which thinks in the beginning of the week that development will take him two days. In the end of the week he did over thousand experiments, but still haven’t found the right conditions.

Of course in reality it’s not as black and white as above described. There is always a certain mix of approaches.

I personally stick to the Drylab model. In this way you really can visualize what is going on.

The story of your epoxy’s I also encounter frequently. In the beginning when I joined Shimadzu I was also astound why they didn’t adapted my method. But I see the chromatographic market as a peak with a tailing factor of 5.0 There are labs which search continuous for ways how to make there labs more efficient. (Fast analysis, efficient MDV etc.) There will always be labs that will work in 10 years time still with there 25cm 10um column. Why? There method is running fine, the receive every month their money and this is how it’s already going 25 years.

Doing this work for people at the tail of the peak is nice for you, but finally wasted money.

[jzt]

Uwe,

To determine if I have a reasonable method, the first thing is to count the number of peaks. However, I don't know how many peaks to expect, have no idea what compounds they are either. If I am lucky, I can get the synthetic starting materials and intermediates to check the separation of known compounds. I don't use MS during the initial column screening. I do use the UV peak purity test but it is not always right. the UV spectra of the structurally related impurities are often very similar. peak area is an important parameter in tracking those peaks. peak shape also gives some indication if there's any coelution. since I use different columns and different mobile phases (each at two different gradient times), I will compare chromatograms from different conditions and try to decide which peak correlate to which. It gets confusing sometimes, especially when the elution order of some impurities witched under different conditions. Experience is very important here.

when do I stop? first stop is when the samples start to pour in. I will use whatever condition I believe is the best to get the results out. based on what I see in these new samples, I may have to go back to modify my method to separate additional impurities. the synthetic process is also evolving during this time. LC-MS is often used to determine impurity structures so the chemists can understand the reacions better. after seeing samples from several batches, I have a good idea which synthetic impurities are here to stay and which one can be removed during purification. I will finalize the method with the most representative sample (separating the most number of peaks). This takes weeks to months, and a good working relationship with the chemists is critical.

Once the synthetic impurities are taken care of, it's time to check out the degradants through "forced degradation" studies. again I don't know how many peaks to expect. at this time I usually do it with LC-UV-MS. peak purity is a major parameter, especially for the main peak. Watch for any new peaks in the chromatogram, as well as any significant peak area change in existing peaks from the unstressed sample. If there's any indication of a degradant coeluting with a known peak, I have to modify the method again.

finally, the most labor intensive part: method validation. the specificity is pretty much proven during method development. we just have to make many solutions to check out linearity/range, sensitivity, precision, accuracy (spike recovery), solution stability, robustness, etc. then try it with different columns, instruments, analysts, and for different days. Appropriate system suitability parameters will be determined. The method is written into an official document. Now we are ready to test the clinical materials, and then the stability samples.

I hope I’ve provided sufficient justification for my paycheck.

[Cathy]

Thanks Uwe for starting this topic. I learn a lot from those discussions.

I also focus on impurity method development for both drug substance and drug product. My appoach is very similiar to what JZT is doing. But I normally do some forced degradation at very beginning to make up some degradation impurities. I use LC-MS to check the impurity profile for reasonable method when I scouting columns.

I have qustions about peak purity. normally, the main peak is overscale, so the peak purity check is not accurate for it any more, how do you overcome this problem? do you dilute the sample to make it on scale? In that case, how to prove there is no small impurity coelute with main peak?

Another question for JZT, how do you decide system suitability parameters for method? I also find your sample test sequence is very different from us. Could you please explain a little bit more? What we are doing for a impurity test is: 2 blank, 2 std a, 1 std b, 1 sample, 1 std a, 2 std b. the six total std will serve as system suitability check.

[jzt]

Hi Cathy,

I'd like to do the forced degradation earlier if I have the material. Normally we only get 10 mg to 50 mg samples at early stage, and sometimes for multiple analysis. we use one solution for several weeks during method development, and keep an eye on how peaks are coming and going. this gives us an idea on solution stability in that diluent. the forced degradation will start when we have more material, and we do calculate the recovery against the unstressed sample.

Peak purity by UV is not very reliable for the main peak. I'd rather compare the peak shape and number of peaks in chromatograms obtained under different conditions. LC-MS is another way to go, but you may also miss a small impurity that does not have good ionization efficiency, especially in the presence of the main compound. In our reports, we don't say "there's no coelution", rather make a statement such as "there's no evidence of inteference to the main peak". I don't think you can absolutely prove peak purity, unless you use truely orthogonal techniques and can correlate each and every impurity observed with the two techniques.

We use the same method for chemical assay and impurity determination. that's why the sequence is rather long. the blank after 1st standard B is to check potential carry over, as well as the consistency of any system peaks (compare the 2nd blank to the blank after std B). The assay result is calculated based on bracket standards of both A and B. we do duplicate preparations for each sample. I am suprised to learn that you don't check for sensitivity in an impurity method. what's your impurity reporting limit? 0.05%? do you have to prove your quantitation limit is alway below your reporting limit? we know that the actual quantitation limit for a method varies on different days, different instrument, etc.

Regards,

Jessica

[Alfred88]

Dear all:

This is an interesting question. I do method development once in a while, when our company wants to have a new product, or when we need to replace an outdated method (e.g wet chem.) by LC method.

To start, I would do a ‘background check’ to see available literature. Google.com may give some leads. USP/NF may have some info. I also have a copy of the Merck index, of “Stability-Indicating HPLC methods for Drug Analysisâ€

[Consumer Products Guy]

I find the "forced degradation" stuff very difficult to really get a handle on. Conditions others have used are very vague, and we don't get our actives to break down using those. Also, what good is it to know that your active degrades in base, if your product is pH-controlled and would be OOS for pH before that happened? I know that theoretically forced degradation is to show that any degradation products do not interfere with the analytes of interest, but with chromatography any smaller degradation products normally elute sooner.

[Alfred88]

Dear CPG and all:

The purpose of Forced-degradation Study (FDS) is to show that the method is "stability-indicating," and it may not be needed if you don't test stability samples (correct me if I am wrong).

I am not sure if there is a complete, systematic approach to this FDS. Perhaps because compounds differ in their stability (some will degrade even if we don't do anything to them). I have seen some elaborate schemes to degrade compounds by refluxing, or boiling, or autoclaving!

How about this "universal scheme":
1. Adjust pH by using NaOH 0.5 or 1N.
2. Adjust pH by HCl 0.5 or 1N.
3. UV (254nm) treatment for 2 hours, and 24 hours
4. Heat (50-60oC) treatment for 2 hrs
5. Peroxide (3%) treatment

And we don't need to worry about degradation had actually happened (or not). From my past experience, UV treatment would give detectable degradation.

Please comment on this proposed scheme. Thanks.

Alfred.

[Uwe Neue]

I had hoped to get a little bit more feedback. Or do you think that all has been said already?

[Consumer Products Guy]

Uwe - the one I just completed uses the same sample preparation and same autosampler vials and I've optimized to have 10 minute runs, either by isothermal capillary GC-FID, or by isocratic HPLC-UV. This time, because production would also be doing QC tests, there was a little push to hold the assay time to 10 minutes or less. We'll be doing validations for these in parallel, with same solutions, haven't gone that route yet. GC would be simpler for the manufacturing plant that was chosen.

[Alex Buske]

Hallo,

Pretty much has been said already. Everything depends very much on the specific problem in our case.
First thing would be to gather some information. For known substances this means looking into the existing procedures. For new (to me) substances some literature study (or internet) is necessary. that would include Journals and column manufacturers application notes.
When a suitable model solution is availible sample runs are made. That would be for ionisable substances usually 2 or 3 pHs, MeOH and ACN, two temperatures, two gradients each and up to five columns. Often we deal with neutral substances that have some isomers (i. e double bond isomers) as related substances. In that case just water is used and as organic modifiers MeOH, ACN and mixtures of both on a number of different columns (different phase, endcapping, carbon load...). 150x4.6 mm, 3 µm columns are a good starting point.
Evaluation is done by UV, Peak tracking can be laborous, especially when there are two actives with 5-10 impurities each. From suitable combinations drylab models are made and suitable conditions verified.

The third part of method development is the adjustment to the routine procedures. That includes definition of detection together with sensitivity checks and some detailed look into sample preparation. As routine equipment differs from method development equipment, chromatographic adjustment can be necessary.
Then finally validation (if it is still required) that also takes some time.
Main goal in method development is maximum robustness, failed system suitability tests or chromatographic problems usually cost at least one day in routine. Our systems run overnight, so run time isn't that important. The main bottlenecks are often sample preparation and data evaluation and documentation.
As our routine equipment is pretty conventional in its layout and in no way optimised for low volumes I am not aiming for to much speed. Tuning the dead volume on one or two systems (at the cost of robustness or universal use) would offer some room for more speed, but then changing the "horses" would be problematic.
Any more specific questions?

Alex