LOD/LOQ determination with carryover

[domino1]

Hi all,

For our early phase development HPLC methods, we normally use the main compound peak to estimate the LOD/LOQ of our method for impurities/related species. There are assumptions there, I know, but pre-validation stage, we are willing to make those assumptions.

However, this strategy kind of falls apart when you have carryover of the main peak or a small mobile phase artifact at that RT since you never have a S/N = 0 point. Sometimes the carryover can be greater than a S/N=10!!

Has anyone found a good way to determine method LOD/LOQ in situations like this? The best idea I could come up with is to use (if available) an impurity authentic that has a similar response factor as the main compound and elutes in a "clean" portion of the chromatogram.

Any suggestion are welcome!!

Domino

[Noser222]

There's a couple of other ways to do it without having to measure signal to noise.

1) RSD of response. Use a cutoff like +/- 20% for LOD. Inject lower concentrations until you reach this point.
2) Standard Deviation of Y-intercept. Multiply this by 3 to get LOD and 10 to get LOQ.

If you are having a lot of carryover, however, you need to try your best to fix that. Are you using a wash vial between injections?

[domino1]

Thanks for the response.

The problem I see with using method #1 is that there is always that peak present. It will also influence the RSD calculation (to an extent). I would like to calculate RSD of just the active peak not the active + carryover/buffer artifact peak (since that will be bigger and could actually make the LOD/LOQ - as based on just the ug of injected compound - compute out lower than without the artifact peak present).

Same applies to method #2. The carryover/buffer artifact peak could similarly influence the precision of the intercept (which should come out to the be the value of the carryover/artifact).

I am having my peeps look into whether it is actually true carryoever or an artifact. It is small (0.2% of the nominal load) so it does not really influence anything other than trying to get at LOD/LOQ!

What do people think about using either of these methods after subtracting out the contribution of the artifact peak? It still does not account for the variabiltiy of actually integrating the peak if it were found without a co-eluting artifact peak (since it would be smaller) - so I think I am biasing the result toward a lower calcualted LOD/LOQ (since it was easier to consistently integrate the combined peak).

Thanks again!!

Domino.

[danko]

Hi Domino 1,

Try injecting several blanks (5 – 10 whatever it takes) until you lose the carry-over peak.
Then inject your diluted standard or whatever you’d use for determination of LOD and LOQ.

Best Regards

[unmgvar]

Hi Domino1,

see if you are dealing with a carry over.
do a little test, inject a linear curve of your compound from the high conc. to a lower one, use big steps. if a carry over is present then you should clearly see a very bad linear curve.
if carry over is the case, then tell us what kind of instrumentation you are using, probably someone here will have a solution for you to make it go away.

[Noser222]

I would say that if you cannot get rid of this carryover, system peak, or whatever it is, then you have to consider it part of the noise. When you are at low concentrations, approaching the 0.2% you mentioned, the variation of that peak will contribute variation to your analyte peak. Obviously, it will have hardly an effect at all when you are at high concentrationns, but the point of the validation of LOD/LOQ is to show how low you can go and still get reliable data, so you have to consider all sources of "noise".

Just curious as to what makes you think this may be a system peak or buffer peak and not carryover. Do you have a PDA? What's your column and mobile phase?

[domino1]

To answer the system peak vs carryover peak...

I just "inherited" this method so I am looking over the work that was previously done on it. The spectra of the peak does look suspiciously like the main peak (although it is pretty ragged due to the low amount).

They looked into washing with 100% ACN between runs, multiple blank injections, changing source of water, TFA, ACN etc and were never able to get rid of the peak (always run on a newer model Alliance). It goes down to a point and then levels off to a constant amount. I am going to try my favorite trick of running a doubled up gradient (to take the injector loop out of the equation) to see if it also shows up in the second half of the chromatogram. Then I throw in a longer equilibration time in the middle which can "concentrate" a mobile phase artifact sometimes. So, I might expect to see a larger peak in the second half of the chromatogram if I double the re-equlibration. If it is gone in the second half of the chromatogram, then my injector loop may be the issue and it is indeed carryover.

The fact that it never goes to zero makes me think it is mobile phase related, however I have a hard time believing that they were not able to get it to go away after changing every single reagent. Plus the spectra point to carryover.

We are going to repeat some of the earlier work (and maybe run on an Agilent) just to verify that there is no way to get rid of this pesky peak before I completely throw up my hands.

Thanks again for everyone's thoughts. It is amusing to me that there are other "geeks" out there who are thinking about HPLC on the weekend.



Domino

[Mark Tracy]

Noser222 is quite right, the artifact must be considered part of the noise.

[p.mcconville]

In your last posting you mention that the assay is running on an Alliance system. Have you replaced the lower needle wash frit? The lower frit retainer is inside the sample compartment on the top-right side, approximately 4 inches within the cavity. A contaminated needle wash frit will present as “ghostingâ€

[domino1]

That is something I have not investigated before! Thanks for the idea.

Domino.

[p.mcconville]

Hello domino1,
Did replacing the needle wash frit help?

[GoldLeader]

Noser222,

You mentioned that you must consider this ghost peak as part of the "noise." I agree with you but I am trying to figure out how you would measure that. I would normally pick a section of the chromatogram with no peaks eluting as the "noise interval" for my software to calculate, however, the ghost peak that I also want to include is not in an area of the chromatogram without peaks (i.e., coelutes with the peak I'm trying to quantify).

Would the following be a reasonable approach or am I breaking some fundamental LOQ rule:

1. Inject the "blank" containing the ghost peak.
2. Measure S/N of ghost peak (say I get a number like 9.2)
3. Inject sample at theoretical LOQ and measure S/N (sample S/N = 23.5)
4. Therefore, measured S/N for theoretical LOQ sample is 23.5 - 9.2 = 14.3), which passes our criteria of NLT 10.

I'm in a regulated industry so I would have to defend this at some point. Do I have a case?

Thanks,
Jeff

[Noser222]

I don't know that doing a subtraction to get to signal-to-noise is all that meaningful.
No matter which approach you take, you're going to have to defend why you have carryover or coelution with a ghost peak.
I suppose I would run a standard curve and inject concentrations as low as possible until the response is not significantly different than the y-intercept (which should be very close to the response of your ghost peak). You can confirm it both ways, by determining the confidence limits of the intercept and to where the RSD of your response minus the intercept is > 20%. This will be a deviation from your normal validation procedure, but maybe all that you can do.
Hope that helps...