Here's a mystery I hope you can help me solve...

[AnaM]

I have an unusual case that occurred in my lab and I am hoping someone out there can help point me in the right direction:

I tested 10 tablets for CU by HPLC and got results that were 10-20% higher than expected. System suit passed and the bracketing standards throughout the run had 0.5% RSD. The retention times are right on and the peak shape is good. I tested an additional 20 and they all came in at the expected value. I re-injected the solutions from the original 10 tablets on another instrument and all the results were as expected. All solutions were re-injected on two different instruments using different standard preps for comparison -always the same thing, the results are as expected and I cannot recreate the original high results.

Long story short, based on the investigation, it was not the sample. Does anyone know of any way an error could occur on the instrument that would affect the sample but not the standard injection? FYI - The injections volumes were programmed correctly in the run. Also, there was a crack discovered in the sample loop after the fact, but I am skeptical this happened during the run since this should have affected the standard and the sample equally and the RSD and RT's of the drift checks should have failed.

[Consumer Products Guy]

Is the peak 10-20% larger for the samples which gave results of high values? If not, maybe a sample amount number got entered incorrectly, but entered correctly on the second instrument.

Not that I've ever done this....

[DR]

Gradient or isocratic?
Inj. volume?
Single point standard or curve?
If curve, are you varying conc. or inj. vol.?

[AnaM]

Gradient or isocratic?
Inj. volume?
Single point standard or curve?
If curve, are you varying conc. or inj. vol.?

Isocratic
10 uL
Single point standard

[AnaM]

Is the peak 10-20% larger for the samples which gave results of high values? If not, maybe a sample amount number got entered incorrectly, but entered correctly on the second instrument.

Not that I've ever done this....

The peak areas of the samples are 10-20% higher than the PA of standard

[lmh]

how sure are you of the standard as used on the first run that gave the faulty results? Was this standard used in any subsequent runs?

[AnaM]

how sure are you of the standard as used on the first run that gave the faulty results? Was this standard used in any subsequent runs?

Yes, we repeatedly re-injected the original solutions (sample and standard) against a freshly prepared standard and the results were the same calculated against either standard prep. Never obtained the high result again.

[Vlad Orlovsky]

we had a case like this when a standard was a different salt f the basic drug. The difference came from counter-ion.

[Peter Apps]

Are your bracketing standards made up in the same solvent as the samples ?

Was the potential leak detected and fixed before or after you reran the series on the original instrument ?

At what stage(s) in all of this did you run the multi-point calibration ? Are the calibration standards in the same solvent as the bracketing standards and samples ?

Peter

[DR]

we had a case like this when a standard was a different salt f the basic drug. The difference came from counter-ion.

...pKa of analyte? pH of mobile phase?

[tristanewalters]

Honestly, it sounds like a preparation error to me. If the solution wasn't mixed effectively initially and placed in the autosampler to be ran, it is definitely feasible that you could get a high result. With the time elapsed between your initial testing and your retesting, it had more time to equilibrate in the solution which could be why you saw normal results after testing again.

[mattmullaney]

Hello Everyone,

Sometimes for cases like this, I find it helpful to make a list of potential causes: (in no particular order)

What Affects Peak Area in LC? Many of these were already mentioned above--this is only to put them in one place.

1. Injection Volume. If an injected solution on the autosampler does not experience change of any sort, injecting a greater volume will afford a larger peak and vice versa. [In this case, we believe this Did Not Occur, see above.]

2. Eluent Flow Rate. With a concentration-dependent detector such as UV, the slower the eluent flow rate, the longer the injected and chromatographed solute resides in the detector cell and the larger the apparent peak area will become. [Now, a cracked flow cell Would Alter Flow a Bit in the proper direction...but this Doesn't Seem Likely as the analyte retention time would change as well- increasing - and for everything injected afterward as well unless the flow somehow "fixed itself"? Just not likely to me, and we also think this Did Not Occur, see above.]

3. Detector Wavelength. This fellow hasn't been broached...was the measurement wavelength the same for all injections in the first LC run? Is the measurement wavelength "on-the-shoulder" of the analyte's UV spectrum as it is chromatographed? [If the wavelength varied for some reason (or the analyte absorbance, more below...) the analyte retention time would remain unchanged.]

4. Column Heater Temperature. This can affect Eluent pH--see below. [Not terribly likely, but a potential source of error.]

These Four Items Above represent the most likely Instrument-Related Culprits...at least the ones I think about. Now for Non-Instrument Culprits...

5. Eluent pH. If there is a buffer, is it of adequate concentration? Is the eluent pH constant (ought to be in an isocratic run...)? Is there are large discrepancy in pH between the injected sample and the eluent in pH (often, there can be...)? [pH may affect peak area/height, but also retention time, I don't know how likely this is, particularly as the observed behavior was transient and mid-sample sequence.]

6. Sample Adsorption to Stainless Steel Surfaces. Yes, this is Also An Unlikely Gremlin...but if the loop is SS and was a bit "gunked-up" from the previous system suitability injections, and the samples from the initial lot were, say, just a bit different in composition than everything else after it...perhaps "gunk" could have been loosened from the loop in the sample injections? [Not terribly likely, though stranger things have happened. Usually this phenomenon works the other way, lowering peak areas per analyte and requiring "priming injections" to achieve repeatable peak areas/heights.]

7. Nature of sample solvent...could it have been "different enough" in pH, organic composition, whatever, in that first lot as compared to the others tested later on to behave differently once, but not again later on? [Not terribly likely, but as in 6, stranger things have happened.]

Pretty much everyone else responding hit on all of these points--I owe apologies for "tail-gating." My small contributions are measurement wavelength, column heater temp and sample loop adsorption...as (un)likely as they are.

My feeling...whatever happened, it most likely was something unrelated to the instrument, per se. Environment, preparation of sample (after Tristanewalters), pH/pKa conflict, again sample/eluent prep (after DR and Peter Epps--was the eluent pumped from one vessel or multiple vessels, ie. "Dial-a-Mix"...could be pump behavior, just thought of that in the latter case.)

[Csaba]

Besides all error sources mentioned above, you may have heterogenous standard (but it is more common with heterogenous samples). You might have not mixed the standard enough when starting the experiment and a small residual particle dissolved during the sampling in the needle, resulting in higher concentration.