Page 1 of 1
Random peak area variability
Posted: Mon Sep 15, 2008 11:12 pm
by hotliner61
I have been running a residual ethanol test in a pharmaceutical API by GC/HS using DMSO as a diluent. This is a rugged and proven method.
The RSD for system suitability is usually very tight. Yesterday as I was reviewing an overnight run I noticed that one standard injection was off by ~ 20 % (higher peak area), whereas all the other standard injections were very reproducible.
I am using a DB-624 30mx0.53mmx3um split injection w/ a 3mL loop.
Excluding analyst error (pipetting, etc.), can anyone come up w/ a reasonable explanation for this random out of trend results?
Posted: Tue Sep 16, 2008 7:32 am
by Peter Apps
Most likely this is some random machine glitch that will never happen again. If you want a more fancy sounding explanantion to get your boss and the QC auditors off your back you can dress it up with some statistics. You have a certain rsd for your analysis, from that you can work out what proportion of standards will be at any given deviation from the mean (in this case 20%), then you look at how many standards you have run (I'd guess several dozen to a few hundred if the method is rugged and proven). If one in however many is less than or equal to the expected number that far from the mean then the method is still behaving in the way that the validation and performance history says that it will. Even if you have not run enough standards to expect a 20% deviation, the single occurrence still does not require a special explanation, statistics says that the deviations can occur at any time. Only if you begin to get multiple deviations at a higher frequency than your rsd predicts should you start start troubleshooting.
The second law of troubleshooting is: A problem that only occurs once is not a problem.
Peter
Posted: Wed Sep 17, 2008 6:08 pm
by mbicking
My good friend, the late Dr. Joe Levy always had a simple explanation - Negative Energy! It was something that hit every lab in random places at random times. As long as you don't violate the "Second Law of Troubleshooting" (or, as I call it, the "Rule of Two"), you are OK.
Or, using the common phrase: "stuff happens!" (editied for content)