Chromatography Forum

We have been trying to validate a procedure for an analyte in a powder blend. The HPLC method itself has been confirmed as ready using wet spike methodology however the switch to analyzing the powder blend itself has shown problems in getting appropriate precision (RSD <5).

We are dealing with a very low level analyte in a complex powder blend - 0.0006% w/w

Recognizing that larger sample sizes improves consistency we have gone from 1 gram, to 2 gram, to 5 gram sample prep procedures which involves extraction with a solvent, evaporation, reconstitution in water for injection.

RSD on all of these were not good. We were going to move up to 10 gram sampling when it hit me that I might just be dealing with such a low level of dilution in a powder blend (batches are in hundreds of kilograms) for this particular analyte that in order to achieve appropriate precision levels I will need to sample sizes in the thousands of grams before consistency and a proper RSD can be achieved.

I scanned the literature for typical drug concentrations in tablets and found dosages at the mg level for typical tablet sizes. My 10 gram sample will have an estimated analyte amount of 6 ug.

This question is not about sensitivity of the instrumentation or of the ability to extract out the analyte itself from the other materials in the blend - that is not the problem. The problem is that I think it could be possible that the analyte is so dilute in this powder blend that sample sizes of under 100 grams would not "cast a wide enough web" so to speak to generate consistency between samples and a proper RSD when validating for precision. Am I right? Am I wrong? Am I down some rabbit hole in my thinking that is nonsensical?

When developing a method it is important to factor in all causes of deviation. This may just prove what you are thinking that it is impossible to evenly distribute the analyte within the matrix itself. If spiking aliquots of the matrix shows good reproducibility then you have proven the method itself works, but if the normal matrix has a large variance, then that is beyond the control of the analytical process and must be factored in giving a wider acceptance range.

If it is impossible to widen the acceptance range due to a regulatory restriction, then the process of blending the analyte and matrix must be reevaluated.

Thanks James

I don't think widening the acceptance range would fly but we are going to increase our sampling size to the maximum amount that is still practical and see if that gets us over the bridge.

I've had experience with non-uniform powders and mixtures, like spray-dried powdered laundry detergents and powder autodish detergents. Most often, since sample was available and our dilution solvent was water, we just used large samples and large volumetric flasks and took aliquots for sampling.

With autodish detergents, we found that the chlorine content was higher as the particles were smaller, we took like 40 gram samples for that, and titrated an aliquot.

Yes, we riffled samples, stuff like that too...

100 gram samples took us under 5% RSD - This is still manageable for us. I was afraid we might need to reach kg + sampling before our precision checked out which would have presented some problems and a possible unhappy conversation with product formulators.

Try increasing your injection volume (50 to 250 uL) and using a more sensitive wavelength. A ppm analyte should be no problem for HPLC! I routinely obtained an LOQ (10% RSD) at 50 ng/mL.

There is clearly a problem with heterogeneity of the product - if you have to take 100 g samples to get repeatable results then what is the likely distribution of your analyte in a 10 g dose ?

Peter Apps wrote:
There is clearly a problem with heterogeneity of the product - if you have to take 100 g samples to get repeatable results then what is the likely distribution of your analyte in a 10 g dose ?

Not good enough for any sort of repeatability if we were dealing with doses of that size. But - our "dose" is in the 100's of pounds that is dissolved in a commercial batch of liquid product.

The heterogeneity is still an analytical problem though - with mixing that rough the repeatability itself (what you call precision) is not likely to be repeatable from batch to batch. As long as the analyte is not something that tends to separate out you could ask for a longer mixing in the process, or you could take a whole bucket full, homogenize it, then sub-sample 10g to save solvent and time. Chuck the rest of the bucketful back into the processor.

Peter

I'm hoping with the larger sample size we don't run into the problem. Although your point is quite salient in that if our batch analysis does not meet spec we will be asking for continued mixing - and if that becomes a regular occurrence or in fact a nuisance occurrence the formulators might be back on the hook.

As for taking a bucket full and doing some form of additional mixing out of batch and then testing and reintroducing to the main batch, that would not fly proceduraly as the purpose of the analytical test at that point would be to show that proper mixing of the batch at large (with documented addition of appropriate levels of analyte) had taken place.

If you are checking for proper mixing the simple fact that you have to take such large samples to get repeatable results answers the question - the mixing is poor. To assess mixing you have to determine heterogeneity - that means taking multiple samples from different places in the bulk and you method validation surely needs to show that if you take samples are various times during mixing you can demonstrate heterogeneity the early in the mixing process, and that it goes away as mixing continues.