Today I came across an interesting problem. We are validating a dual active product. The chromatography is reverse phase with UV detection at 240nm. We noticed that the hydrchlorothiazide data was biased a few percent on the one hplc system. The only known difference is one system has a waters dual wavelength detector(one with bias) and the other has an ABI detector. The other active results looked fine on both systems. On closer inspection, it is noted that the hydochlorothiazide resonse is three times greater on the waters system than the ABI for both standards and samples. I am at a loss.

Is it possible that the Waters detector has a wider bandwidth resulting in selectivity problem?

Have you done an OQ on both detectors to look at linearity and wavelength accuracy?

Good call Anon 10:10. Looks like the good results deluded us. The wavelength accuracy was way off. Looks like a power outage put out the detector. This brings up an interesting question. What daily(i.e. practical) method could be used to assess this component of the system. If this was not a dual active we might not have seen this affect since the one active yielded good system suitability data and results that were consistent with the entire data set. I don't like when the result tips off the issue.

Anon 5:44

All I can say is you picked up that something was wrong based on the results you obtained. Intuition, unfortunatley, I believe is important in an analyst.

You run sys suit and that should give you an indication of a problem. You calibrate your instruments reguarly to insure instrument accuracy. But, unless you have an ability to discern what this data represents you will miss something.

Sorry for the rambling.

Anon 10:10

Since you noticed the response was different on the bad detector, you could calculate response factors from your standard. We do this for an external standard method and the results are consistent over time. Don't know how it will work on your system and actives. The response factors may be different for different systems, so you would need to determine "typical" results for all systems you use with this method. Obviously, the accuracy of the detector would need to be established before your first response factor determination.

I'm not near my systems to check the manuals, but can you simply run a holmium filter check? This doesn't take long on our HP's, and I think the manual hints at periodic (weekly?) checks.

We had a similar problem years ago with an HP FLD system. The system ran the optical bench up and down every time the method was used--every analysis, and eventually the stepper motor missed steps. At first 1 in 20, and then 1 in 10 samples. The optics still caught the peak, but the response was low--off the peak apex, similar to the bandwidth problem mentioned. It took about a month, 2 chemists and 3 HP engineers to isolate the problem. It took less than an hour to fix it.

I'll bet most systems do it this way. If so, it's not if, but when, they will fail.

Our Wavelength Accuracy and Absorbance Qualification Kit is intended for the validation and regular control of also wavelength accuracy and wavelength precision, and based among others on the NIST SRM 2034; “Holmium oxide Solution Wavelength Standard from 240 nm to 650 nm”