FDA Validation for Endogenous Compound in Urine

[jdlh199]

Here's a question that I've been thinking of for a while now.

I need to validate a method that determines compound X in human urine. However, this compound is endogenous to urine and is present in a wide range of concentrations between "blank" urine lots.

How do I approach the validation of this method to satisfy the FDA's GLP bioanalytical method validation guidance?

In other words, how do I prepare calibration curves and QC? Usually I would use a "blank matrix". But since I cannot obtain a blank matrix what do I do?

Would it be acceptable to have calibration curves in solvent and use a bulk urine matrix (with a low analyte concentration) for QCs. I could add a QC0 to the method and perform blank subtraction to calculate QC recoveries?

Anyone have experience with this or can suggest a better approach?

[g_wallace]

Is it possible to "strip" the urine of the analyte to produce blank urine ?
We're using a similar approach for a method now.

[jdlh199]

No this is not possible.

My latest thought is using blank subtraction for both calibration standards and QCs and using a different lot# (with different endogenous concentrations) for each?

[g_wallace]

It's a difficult one.
What we would probably do is analyse six different lots to determine the endogenous range and set our LLOQ accordingly. Then mix the six lots to use as our "blank" urine and use standard addition for our curve and QCs.

[Bruce Hamilton]

If you are using MS detection, maybe you could purchase and use a deuterated form of the compound for your validation.

There are several custom synthesis places that make deuterated forms of drugs and metabolites to order. I regularly perform analytical work for one such firm that has a worldwide reputation in such synthesis.
http://www.bdg.co.nz/

Bruce Hamilton

[jdlh199]

We did look at deuterated forms, unfortunately we were quoted $25,000 for 500 mg from one supplier, and $15,000 from another.

So we gave up on that idea since the standard is $50 for 1g!

I'm going to approach our client with blank subtraction for now with different lots for STDs and QCs (add QC0 to QC levels).

Thanks anyway!

[oscarBAL]

Hello JNH.

I do not have experience on this but has you tried standart aditions?
This is moer work; but in this case; "you don't need a blank" because you prepare your STD's adding aliquotes of the Xcompund to your sample.

Let me know if this work for you; beside I dont know if Standart Adition is the correct term in english.

Oscar.

[AICMM]

I am really interested in OscarBal's approach since I am faced with the same problem with a petroleum application (meoh in crude.) I think mine is more straightforward but I would be very interested to hear views on this approach to solving this problem.

Best regards.

[Bruce Hamilton]

Standard additions techniques were fairly common for detection of trace elements in used lubricating oils when AAS was used for analysis..

However, there are quite a few traps for the unwary with standard additions, and something like trace MeOH in crude presents several of those.

I think ASTM was working on a GC method to cover relatively high concentrations ( 0.001 - 0.1% ? ) in crude, because it is used to recover some crudes ( gas hydrates ? ). At those concentrations, they probably added an internal standard as well, and they probably wouldn't need to use Standard Additions, although the combination of IS with SA would improve precision, and they could also use CD3OH as IS to mimic CH3OH, if MS was used.

The main assumption for std additions is that the increase in concentration from the added MeOH will behave in the same way as the MeOH already present. For trace components in a diverse environment , such as yucky crude oils, that is often not true.

The MeOH may be associated with entrained polar material, and once that environment is saturated with MeOH, the rest of the MeOH would be dissolved in the HCs until solubility is reached. Evidence for such mechanisms can be poor recoveries at low concentrations of the spike, especially if trials have shown the recoveries are good using clean ( distilled ) hydrocarbon fractions.

In general, it's preferred to remove all the analyte from a sample and spike that, however with something like methanol, the act of removing it may change the crude composition. The issue can also appear with headspace analysis, as volatilily can also be affected.

Assuming it's a full boiling range crude, the method of extraction using a hot, polar, solvent could also extract other material that has affinity with methanol, so future use as a zero blank is compromised.

I haven't looked at the literature, but I suspect there are several reasonable methods out there for 10 ppm or more, but it depends how low you need to go. I'd look to the API for information and contacts in the industry, as there may be specialist detectors already used.

Please keep having fun,

Bruce Hamilton

[danko]

Hi jdlh199,

You don’t describe your method of analysis, so I’ll just make some assumptions.
The first assumption is: It’s some kind of chromatography.
The second: You use a C18 column or any kind of reversed stationary phase.
The third: Your strong eluent is some organic solvent (ACN for instance).

If the assumptions above are roughly accurate, then try the following suggestion:

1. Collect some urine (it shouldn’t be a problem).
2. Run it through the column you use for the analysis (preferably a discarded one).
3. Use the collected effluent as a solvent for your standards, controls and furthermore as a blank.
4. Under the validation you could dissolve a known quantity of the analyte in this effluent (stripped urine) as well as in untreated urine and see if you can find the mass balance. In this case you’ll need to inject treated as well as untreated urine as blanks.

Hope you get some ideas you can build up on.

Best Regards

[lmh]

I'd have used standard addition too, but just in case anyone reading this has had to deal with the situation of "measuring" something where no standard is available at all (!), plant biologists used to have a good trick of doing mixing experiments. Here, true quantification is obviously out of the question, but it is important to know whether the signal in sample "A" is lower than that in sample "B" because "A" contains something that messes up extraction and measurement, or because "A" contains less analyte. Therefore one mixes sample A and B, and looks to see whether there is a linear relationship of analyte to mix, linearly interpolating from A to B. If A messes things up, it also messes up B, and the results come out low.

In effect, we use the samples as standards.

Good luck with your X in urine.