recovery after spiked addition

[Keanu]

When trying to validate an HPLC- method, we stumbled across the following problem: We wanted to determine the recovery of an impurity after spiked addition. The concentration of this impurity in our sample is <LOD (0.025%), but when we spike with 0.3%, 0.5% and 0.8%, we recover 0.4%, 0.6% and 0.9% which corresponds to recoveries of 136%, 123% and 115%. A UV-correction factor is included in the calculation.

The absolute error is relatively constant (+0.1%) but when we inject the unspiked sample there is absolutely none of this impurity in it. Also the blanks before and after does not show any carry over.

The peak area of the main peak in the unspiked sample is about 45000000 mVs, in the spiked experiments only about 40000000 mVs (!). The peak area of the impurity in the spiked sample is about 49000 mVs (0.3%), 70000 (0.5%) and 102000 (0.8%) while when I inject the impurity alone at the given concentrations, I get peak areas of 38000, 64000 and 105000 mVs.
Shouldn't the peak area stay the same at certain concentration, no matter what else is injected?
The problem is reproducible with different columns, eluents and sample solutions.
The product is stable under the HPLC conditions.
The RRT of the impurity is 0.94, the peaks are baseline separated. Is it possible that although baseline separation, the main peak may influence the impurity peak?

Do you have any idea what has happened here?

[tom jupille]

As with a lot of these questions, you really haven't provided enough information (e.g., isocratic or gradient, what was your plate count, column dimensions, flow rate, actual retention times, do your calibrators have both the impurity and the main peak, etc.).

That said, I'll go out on a limb and speculate. I'm assuming that your calibrators (standards) have the impurity only and I'm assuming by "baseline" resolution, you mean the textbook definition of "Rs ≥ 1.5". That definition was originally called "99% baseline resolution", because it is the value at which you have just less than 1% overlap between two equal-size Gaussian peaks (a 6-σ separation between the peak centers). Basically, you have 0.5% of the second peak area extending to the left of the valley between the two peaks. As the first peak gets smaller, the valley shifts to the left (toward the smaller peak). The usual result is to underestimate the smaller peak (rather than overestimating as you are doing, but (and it's a big "but") your actual results are going to be very sensitive to the way in which your data system is handling the baseline and the cutoff point between the peaks.

I'd start by taking a hard look at the printout of the chromatogram from the data system to see how it's handling the baseline and cutoff point. If you're lucky, you may be able to tweak the settings to give you better results. The next thing to look at would be including the main compound in your calibrators.

If your calibrators did include the main peak, then you have to look at other possible differences between the calibrators and the spiked samples (other matrix components? separation done under exactly the same conditions, including things like equilibration time?, adequate buffer? etc.)

[JM]

Hi keanu,
The increase in area and % in spiked samples could be due to presense of these impurities already in your sample ( may be less than LOD but they are there and may compansate till LOD) , you have to factor in the already present amount of these impurities in sample.

Your data also indicate it as the increase in highest at lower conc. and subsided as you increase the spick conc. This kind of higher or lower recoveries are common and allowed in case of impurities.

hope this will help.

JM