I’ve been trying out a fairly straightforward (?) reversed-phase (C8, 5-micron p-size) gradient method to detect impurities in a medicinal product. The conditions are 40:60 ACN: 0.05M Phosphate buffer increased to 75:25 in 40 mins at a flow rate of 1ml/min with detection at 240nm. The problem I am facing is that when I change over to a new column (even from the same lot/batch) the impurities detected for the same sample are significantly different to that on the previous column (they increase and decrease). In some cases only selective impurities are increased and others remain constant.

I had the impression that 5-micron p-size columns gave very reproducible results between lots and within lots. My question is what could be causing this phenomenon? Has anyone had this problem before?

A couple of things to look at. One is that the column truely is differnent. However, at least for the major brands of columns, the reproducibility for manufacturers tests is very good. George Guiochon (Tennessee) presented a paper at the recent HPLC '99 meeting in Granada about this and has done other work as well, showing that for the columns he tested, things looked very reproducible (e.g., 8 batches of Symmetry). On the other hand, if you have basic solutes, very small changes in surface coverage can make differences column to column. Some ways to address this are to use low pH (pH 2-3) to suppress silanol ionization and use an amine additive (e.g., triethylamine at 25 mM) to suppress unwanted silanol interactions.

The key observation here is that the problem occurs when different columns from the same lot or batch are used. This fact makes it unlikely that the column is in fact the problem. With gradient methods, you need to worry about column conditioning between runs. Is the column regeneration time long enough? You did not mention if retention times change, but your message implies that areas change. You need to clarify this, as it is very important to a final diagnosis of the problem. Another very important piece of information can be found in the history of your observations: results as a function of time and column. If you can construct such a history in convenient form, a diagnosis may be much easier. Also, what is known about the drug product being analyzed? mol. wt., acid, base or neutral, "interesting" functional groups, etc.

Have you tried to subtract a blind gradient chromatogram from all of yours chromatogram. Impurities fra mobil fase A can be released from the collum when the ACN concentration raises. Start with 1 or 2 blind-gradient runs before your sampels, and finish with 1 or 2 blind-gradient runs.

In response to L. Snyder's comments:

First of all, thanks for your comments and help. I believe you are correct about the problem not being the column lot reproducibility. The history of the columns have been monitored and initially one particular column was working well and now it is not. This leads me to think that the problem lies with the mobile phase/sample dissoving solution and it's interaction with the sample.

The compound of interest is weakly acidic. Note that the blank solution is perfect and it is only the sample which suffers from the irreproducible peak area of one particular peak. In general all of the other impurity peaks do not differ significantly. Could this be degradation and what would cause this, bearing in mind the operating conditions I mentioned in my first message?

Thanks again,