Bias in Recovery for Drug Substance

[GoldLeader]

All,

Please help.
System Parameters:
HPLC = Shimadzu LC2010
Mobile Phase = PO4 Buffer, pH2.5:ACN::75:25
Column = C18
Flow = 1 mL/min
Detector Wavelength = 210 nm
Sample Diluent = Mobile Phase

I've prepared samples of a drug substance from 50 - 150% spec.
Recoveries are as follows:
50%Prep = 106% Rec
75% Prep = 102% Rec
100% Prep = 100% Rec
125% Prep = 94% Rec

Standards are prepared to 100% spec.

What could cause such a bias in my recovery? My acceptance criteria is 98-102%, so this method would not be acceptable. What should I look at to fix this?

Thanks,

Jeff

[JA]

My guesses: A co-eluting substance with differing response/concentration ratio or a sample reactive with itself

[danko]

Standards are prepared to 100% spec.

If I understand your set up correctly, you don’t verify the linearity over the range you analyse your samples.
It’s a good study to conduct at least when developing and validating a method of analysis. It might not be the linearity the problem lies in, but if you generate a calibration curve you’ll be able to determine the slope of it and if it doesn’t intercept zero (or approximately) then your problem could be “carry overâ€

[LC_labrat]

What could cause such a bias in my recovery?

Are the samples prepared by separate weighings or serial dilution from a stock solution? What is the RSD of multiple injections from the same sample? Have you looked at the peak purity?

[AdrianF]

You use the term 'recovery' which implies an extraction. Is that the case?

If so could you give more details.

[GoldLeader]

All,
Thanks for your replies. Some more info:

-Different sample levels are from separate weights, not serial dilution

-When I say "recovery," what I mean is "preparing a sample at a certain level (no extraction step) and measuring against bracketing standards"

-Standard and sample are prepared from the same material

-RSD for mulitple preps at the same level is <2%

-We are evaluating linearity for this method, RSquare is good (0.999), but I am concerned about the slope. Also, y-intercept is not zero.

-Response on the UV dtector ranges from about 750 to 1200 mV (I'm told this type of detector is linear to about 2.5 mV...)

Any more insight?
Thanks.

[AdrianF]

I suspect the problem is the linearity of the detector. At 210nm you will probably have a considerable background absorbance. It is best not to go above 1A. (Look for a previous thread - AdrianF)

The 2.5AU includes the signal from the background.

I suggest you reduce all your concentrations by half or more and then you will have no problem.

Having spent a lot of my time in trace analysis when we would consider a max absorbance of 0.01 a luxury I think methods often put unecessary amounts on columns.

[Shanextan]

I think the sample has solubility problem.--Shane

[DR]

In addition to the above mentioned ideas, I see three things that are worth considering:

1) PO4 buffer at pH 2.5? This is a problem as you are outside the effective buffering range of PO4. Can you switch to Formic or Tfifluoroacetic acid?

2) You do not indicate what concentration your 100% sample represents in terms of mg. There are some compounds that can swamp a detector before the detector ever reaches its maximum output (eg: paracetamol), or your compound could be altering the pH of the system at higher concentrations (due to PO4's lack of buffering capacity at your pH).

3) Speaking of maxima - Does your data system's input match your detector's output? If your detector is trying to send a 1.2V signal to a A/D that is set to accept up to 1.0V, you're going to have issues.

[danko]

Phosphate buffers just fine at pH 2.5 (pKa = 2.1)

Also, I don’t think we are dealing with an overload of any kind in this situation. The reason for me being sceptic towards the overload hypothesis is that the correlation coefficient is 0.999 which means the calibration is linear (as opposed to a bending cirve) but the intercept is way above zero. That would result in the observed values.

Best Regards

[AdrianF]

I tried plotting the data as stated and found a correlation of 0.995 not 0.999 and it was a definite curve - especialy considering that the line should go though zero.

I hope you will do the experiment with lower concentrations - I am confident that you will get a much better result.

Also try measuring the absorbance of the mobile phase.

[danko]

I think GoldLeader referred to the calibration curve generated using standards and not the recovery samples. So I assumed that the standard curve demonstrated a 0.999 corr.

But that makes me think of another issue: Were the recovery samples tested within the calibrated range? In other words: Was the lowest standard beneath the 50% Prep. and the highest standard above the 125% Prep?

Best Regards

[shaun78]

The problem here is that 100% is spot on.

If we were maxing the AU of the detector, the recoveries should fall out of line at higher concentrations (which they did, the most out of line was 125%, but the lower concentrations would fail too).

If there were a consistent sample bias, then nothing should have passed, yet two of the levels did.

The buffer might well be causing the problems as you are indeed a hair out of the buffer range for phosphate. TFA would be a better option (hit about 0.1% to get a pH of about 2.5) and could help to give you more consistent chromatography.

If these standards were made by simply adding raw analyte to the matrix, then I would be left thinking:

1. Amount weighed out on balance is too small for the balance to accurately measure.

2. Samples were inaccurately prepared.

Did you perform any characterization work on the method before you started with the validation protocol? Generally, once I have something developed I "evaluate" the method for a series of weeks. My evaluation is running through the unwritten validation protocol. Once I am sure the method performs to our standards, I write the protocol and run it again.

May seem a little wasteful, but when it saves you from running into somehting like a failure in a validation run it is more than worth the invested time...

Best of luck

Shaun

[AdrianF]

Goldleader - could you clarify your procedure.

Are you using a single point(100%) as calibration, with the other concentrations compared to the single standard.

[AdrianF]

Have you managed to solve the problem yet?