Purity testing by HPLC

[jdlh199]

I have some potential work lined up confirming a compound that should be 100% pure. The CofA gives a value of 99% to 101%. The client wants these values confirmed by an independent lab. They even provided a method for us!

My question is really about precision. Surely there is no way we can be expected to get that kind of precision/accuracy with a HPLC-UV method?

[ym3142]

did you try their method?

[jdlh199]

Yes I had a quick play. However we're still waiting for the independent reference standard, so I simply used the clients sample for both calibration, QC's, and unknowns, but prepping 3 different stock solutions. This was just a test run to get an idea of what to expect. Here's the back calculated concentrations in ug/mL. Calibration curve data represents the mean of 2 curves, one at the start and one at the end of the run.

std10 - 9.98
std20 - 20.27
std50 - 49.56
std100 - 99.18
std200 - 201.00

QC10 - 10.6
QC30 - 30.6
QC120 - 126
QC180 - 182.3

Unknown x 10,000 - 103.68 (=102.76%)
dup - 99.70 (=98.81%)

This data looks great, QC data is <5% error. But if the client needs the sample confirming from 99-101%, surely this is not good enough?

Along with the reference standard we'll be getting an internal standard. Maybe that will make a difference.

[DR]

You're only going to be as good as the standard material on this one, probably worse...

All you can really do is give the client your assay result based on their purity assessment of the provided standard.
Assuming you'll do multiple sample preps (and probably injections), be sure to include a confidence interval with your mean result.

[Mark Tracy]

There is a difference between purity and assay. I think we are talking assay here. Even at 99-101% assay it can still contain 0.5% of intolerable impurity.

The only way to get your precision good enough for this requirement is through numerous replicates, both in the calibrators and in the samples. The confidence interval decreases by sqrt(N). If your system delivers 5% for a single determination, and you want 1%, then you need (5/1)^2 = 25 replicates. Either that, or you spend extra care to reduce that original 5%.

I hope you are charging by the hour...

[Dan]

It is not all that unusual to have an assay specification limit of 98-102% or 99-101% for drug substances (or raw materials); you can find those limits in USP mongraphs.

So, yes, an HPLC method can be that good. You wouldn't set those limits otherwise.

The questions is: Is the method provided to you that good? Can your client provide validation details to you for their method? If their method validation shows that the method can acheive a +/- 1% accuracy (i.e. 99-101% recovery), then you are good to go. If the method isn't good enough, then you'll need another one.

Regards,
Dan