PDA OQ Failure - what would you do?

[juddc]

HI Folks,

I have an interesting situation that cropped up today and I'd be interested to hear how some of you in more regulated labs would handle it.

I was performing a routine qualification of some instrumentation in our QC lab and I obtained an unexpected OOS result. The subject is a Waters 2996 PDA, which is a converted 996 and is quite old (1993 or so), but has always run very well. It was last PM'd in September and it passed all OQ tests with ease at the time.

The test that it failed today was a wavelength accuracy test, run with an aqueous solution of erbuim perchlorate which was supplied by the manufacturer. It failed at one wavelength out of three checked and only by a whisker. Specification was for lambda max at 255 +/- 1.5 nm and the result was 253.4 nm. All other wavelengths were easily within specifications (<1.0-nm error). The lamp has about 500 hours on it, is not an aftermarket unit, instrument settings are unchanged from the last time the test was run, lamp energy is fine ~28,000 @ 235 nm, and all internal diagnostics pass easily; Balmer lines are at 486.4-nm and 656.1-nm, respectively - well within 1.0-nm specification for error. (Recalibrating had no net effect, btw...)

Checking the same OQ solution on a brand new PDA (a 2998 installed on 12-23-2011 with 3 hours on the lamp) revealed a lambda max of 254.5-nm, so I don't think the solution is the issue.

The old beast passed a linearity test easily (% RSD for Abs / conc from ~0.5 to ~2.5 AU was ~2%; specification is <5%) and the baseline is nice and quiet, too. Aside from this, it appears to be running quite nicely.

As I see it it's a failure and a real one, however small, and I have to deal with it, or at least make recommendations to QC management. (I'm in R&D, but do this work for QC / Production because I know how.)

Absent an obvious issue with the detector itself - and I can't see one - I'm tempted to get another ampule or two of erbuim perchlorate and rerun the test to confirm or refute the result. If it fails again, I suppose I could call Waters - the machine is under contract - but I HATE calling service without at least having a solid clue as to where the problem might lie.

If you were in my position, what would you do?

Cheers!

[DR]

If it was among 1st things checked, I'd rerun. You could have had an air bubble or not have had 100% test solution in cell or some other similar transient issue. Most OOS investigations include an early "see if oos result is reproducible" step.

[unmgvar]

well you can perform a corrective operation that you think would allow to fix the issue

for example you can decide to wash the flow cell.
i would wash with water, a weak acid (1-5% acetic acid), water, ACN.
each time at least 30 minutes to an hour.
then retest.

always follow the same logic.
if it fails, then perform a corrective procedure that you think will fix the problem. retest. if not good yet, go the next possible corrective action, retest, and so on. until the problem is fixed.

[Peter Apps]

Is the OOS repeatable ? If you run the same test again, without changing anything, do you get the same result ? If not, then you had a result in one of the tails of the distribution of measurement results - this can be difficult to sell to QC unless you do something "corrective" in between taking replicate measurements to make it look as if you have fixed a problem - QC like nothing more than to see corrective actions, even for problems that did not exist, they hate problems that look as if they have gone away on their own, even if there was no problem in the first place.

If the OOS is repeatable then there might be a real problem with the instrument, Another possibility is that the signal is getting towards the top of the detector linear range, which flattens the top of the scan peak, and makes it more difficult for the software to find the real lambda max. What happens if you scan a more dilute solution ?

Peter

[Bull76]

Hi,

Have you re-ran the OQ? Make sure you flush the flowcell well before performing the test. Usually Methanol is sufficient especially since the energy seems okay. Then have High grade water flowing or using a syringe flush through the flowcell. Start your test and then after a minute when the baseline is stable inject 2 to 3ml of the erbium solution and allow it sit. Review the spectrum and extract your data and move the time cursor across the spectrum and observe the extracted wavelengths to see if 3 wavelengths of interest are present.

[juddc]

I'm on hold for a few days as I have to get more of the test solution. Indeed, I would have run the test a second time prior to posting if I had sufficient material. Thanks for the comments. I'll clean the flow cell, retest, and see where that gets me. FWIW, the cell was flushed with MeOH then water prior to being checked and the spectrum was quite consistent across the peak. The detector had actually been in use until about an hour before I ran the check and the data looked fine in all aspects.

We shall see...