How to improve limit of quantitation?

[zheyin]

When I joined, the method has been developed and been used for 2-3 years.
And now they started doing validation, and just found out the LQ is 4% (which is too high). Ideal one is 0.5-1%.

We increased the 100% API concentration 2x and injection volume 2x, the peak shapes were poor (bad tailing).

What else can I do to improve LQ (instead of developing a new method)?

I am thinking about changing column (same chemistry, but different dimensions: smaller ID and length and dp)
changing wavelength to have lower noise (MP is buffer-MeOH isocratic)

Thanks.

[Consumer Products Guy]

What exactly do you mean by LQ (limit of quantitation) 4% being too high? We don't include such limit in our validations, and not sure I've seen this in ICH/cGMP.

Do you mean that your limit of quantitation is 4% of your target peak, and that's considered bad? We typically validate for 80% to 120% of target, and outside of that is not relevant. LD and LQ studies are done her, but just to detail what those levels are?

[zheyin]

What exactly do you mean by LQ (limit of quantitation) 4% being too high? We don't include such limit in our validations, and not sure I've seen this in ICH/cGMP.

Do you mean that your limit of quantitation is 4% of your target peak, and that's considered bad? We typically validate for 80% to 120% of target, and outside of that is not relevant. LD and LQ studies are done her, but just to detail what those levels are?

LQ is defined as 10x S/N ratio of each of the 3 API's. And I found out it was 4% for API2 and API3.
The current method for 100% concentration is 0.25 mg/mL each for injection. So 4%'s actual concentration is 0.04x0.25 mg/mL.
During validation, the "Detection limit" part requires to find out the LQ (the validation protocol is already there when I joined).
But upper management sees 4% too high to be acceptable for FDA...

The validation protocol also has linearity part: from LQ to 130% of each API.

[tom jupille]

What is the purpose of the assay? Validation requirements for potency or content uniformity (as CPG pointed out) are typically from 80 -120% or 70 - 130% of the stated value. LOD and LOQ in that situation are informational only. Get the SOP changed.

Increasing the sample size may help the low end, but (as you found) is likely to take you above the linear range at the high end. The same is true for going to a narrower/shorter column.

The other approach is to deal with the noise: good degassing, new lamp, better detector, time constant (or data system filtering settings) can all help.

[prepcolumn]

Zheyin

As far as understand the method you mentioned is not an impurity method. In that case -as Tom stated-you don't need to get levels like 4-5% unless your method is for delayed or sustained release tablet dissolution method (or maybe first time point of a dissoluion profile is below 5%)

If you have 3 actives a more suitable wavelength or using more than one vavelength in same method may help.

[zheyin]

Zheyin

As far as understand the method you mentioned is not an impurity method. In that case -as Tom stated-you don't need to get levels like 4-5% unless your method is for delayed or sustained release tablet dissolution method (or maybe first time point of a dissoluion profile is below 5%)

If you have 3 actives a more suitable wavelength or using more than one vavelength in same method may help.

How do you define a potency method or an impurity method?
They want to see all the impurities (also has a stress test in validation protocol)

[zheyin]

What is the purpose of the assay? Validation requirements for potency or content uniformity (as CPG pointed out) are typically from 80 -120% or 70 - 130% of the stated value. LOD and LOQ in that situation are informational only. Get the SOP changed.

Increasing the sample size may help the low end, but (as you found) is likely to take you above the linear range at the high end. The same is true for going to a narrower/shorter column.

The other approach is to deal with the noise: good degassing, new lamp, better detector, time constant (or data system filtering settings) can all help.

I am really not sure, they want to see the impurities, and they also use the method for stability test of the products.
What does an impurity method do?

[HPLCaddict]

An impurity method is supposed to quantify the impurities in the product - you're usually aiming at very low concentrations. A potency method is used to quantify the API - you're aiming at something close to 100%.
What do you mean by "they want to see the impurities". Do you actually want to QUANTIFY them (then we're talking 'bout an impurity method). Do you just want to see them once in the validation (aka checking specificity)?
The necessary LOQ is NOT defined by what your bosses think is suitable (might be hard to tell them) but by the SPECIFICATIONS of the product. If you're running a potency method, as already stated you're in the range of 70 - 120%, so a LOQ of 4% is perfectly fine. If you're quantifying impurites with a specification of NMT 0.2%, you'll be in trouble with a LOQ of 4% .

[prepcolumn]

I completely agree with HPLCaddict..

[tom jupille]

As with clothing, " one size fits all" in methods all to often means "one size fits none". You would be better off having separate, optimized methods for major components and for impurities.

[zheyin]

It is a stability-indicating method
we do need to have a much better DL.
Ccan someone tell me about changing the parameters in the UV detector?
response time
slit
...

[zheyin]

It is a stability-indicating method
we do need to have a much better DL.
Ccan someone tell me about changing the parameters in the UV detector?
response time
slit
...

UPDATE:
response time was changed 10x greater, so that we have much less data points and the noise became less.

slit was changed from 4 nm to 8 nm, and noise became less.

this way, the s/n was improved.

But I am not sure if slit is proper for the analysis because of no guidelines.

[tom jupille]

wider slit = more light = less noise
wider slit = wider bandpass = worse spectral resolution

Since the goal is quantitation, not getting a UV spectrum, and since UV spectra in solution are fairly featureless, there is generally no reason to prefer a narrower slit. Assuming your wavelength is set to an absorbance maximum, at some point increasing the bandpass will let in more off-maximum light and the signal will start to decrease. That's the point at which you want to stop! So, to the extent that there is a guideline, it is "use the slit width that gives the best S/N ratio".

[zheyin]

wider slit = more light = less noise
wider slit = wider bandpass = worse spectral resolution

Since the goal is quantitation, not getting a UV spectrum, and since UV spectra in solution are fairly featureless, there is generally no reason to prefer a narrower slit. Assuming your wavelength is set to an absorbance maximum, at some point increasing the bandpass will let in more off-maximum light and the signal will start to decrease. That's the point at which you want to stop! So, to the extent that there is a guideline, it is "use the slit width that gives the best S/N ratio".

The response time has to stop at 10x higher, because higher response time will exceed the peak width of the major peaks and data points become not sufficient based on the guidelines from ChemStation.
That is why I also widened the slit from 4 nm to 8 nm.
But now I encounted another issue, after changing the slit several times, my chromatograms showed some wavy patterns,
it could show up in the middle of the run.
I doubt it is the detector malfunction.

[tom jupille]

But now I encounted another issue, after changing the slit several times, my chromatograms showed some wavy patterns,
it could show up in the middle of the run.
I doubt it is the detector malfunction.

It's easy to tell: just set the slit width back to its initial value and see if the problem goes away. That said, I agree that it is unlikely to be a detector issue; more likely to be air bubbles or a sticking check valve. The intermittent nature really makes air bubbles the most likely suspect.