How do facilities handle missed antibodies on automated testing platforms? What do you require before returning to using your automated method?

[DaMaam]

I realize that all automated systems will not catch all clinically significant antibodies as no automated system is perfect; however, when a clinically significant antibody is not detected on the antibody screen, I'm curious to know what actions facilities have taken before "trusting" the automated method again.  Of course, any other recommendations on how you might address this type of situation would be most helpful. 

[galvania]

But surely exactly the same can be said of manual techniques?

There are a whole load of issues within this question.

1. Has you instrument been properly validated, do you carry out routine maintenance, are your daily QC in order, are the antibodies you use in your daily QC sufficiently weak to be meaningful, are you using the instrument as per instructions, are the cells OK, is the samplke fresh, etc etc?

2. Did the instrument miss the antibody because it called a clearly positive result negative?  Or did the result really look negative?  If it's the first, then the instrument needs to be checked.  Or was it a positive, but such an atypical reaction that it is understandable that it was called negative?

3. Did the instrument 'miss' the antibody because it was too weak to detect?  Did you repeat with exactly the same reagents manually and get the same result?  If you did, then you can't blame the instrument.  If, on the other hand, you get a good positive manually and neg with the instrument, maybe you had bubbles on the top of your plasma sample and the instrument pipetted bubbles?  Or maybe you didn't mix your cells before putting them on the instrument and it pipetted a cell suspension that was too weak?  Or maybe the cells were cold?

4.  Why do you say it's a clinically significant antibody?  Some systems are more sensitive than others at picking up cold antibodies or anti-xxx-like antibodies than others.  What your system 'missed' might be one of these that was picked up unneccessarily by a different type of instrument

I think that will do to start with.  I get the feeling this will be a very active post!

[tbostock]

In full agreement with everything Anna asks above.

The key is knowing the limitations of each method, and having a few methods that you can use to tease out weak antibodies.

[catchmenow51]

Please understand that NO ONE method is guaranteed to detected 100% of the antibodies 100% of the time. We have 2 automated machines, 2 semi-automated machines and plain old fashion manual tube testing. We have seen, REPEATEDLY, where one method would detect an antibody while the other one would not. If your instrument failed to call a pos. a pos. then yes there is a problem. If it failed to detect the antibody, then as Anna said, you have to evaluate other possible issues and possibly chalk it up to methodology. Good luck. 

[Pony]

If all your controls and daily QC are performing as expected, there is no reason to take the instrument out of service. As everyone has clearly explained above - often the issue is sample-related or method-dependent. You'll drive yourself crazy trying to figure out how to re-validate an instrument if the only thing that demonstrated unacceptable performance was the one sample

[John C. Staley]

I was very excited to read the 4 responses above.     Excellent answers all.  The hardest thing I had to do when bringing in the ABS2000 in 1999 was convince the Transfusion Service QA Director that the instrument was not expected to be 100% Sensitive and 100% Specific any more than any manual method!  That was and is simply unrealistic.  I suggest not doing anything regarding the instrument that you would not do to a manual test / tester.  Strange things happen occasionally, that's just the way it is when you are dealing with biological systems. 

[AMcCord]

Yes Yes...good answers everyone!

[goodchild]

Out of curiousity, what automated analyzer was it and what manual method was able to identify the antibody?