Thursday, February 10, 2011

EMQN part 2

So not long ago I told you about how in our annual QC test my lab got an answer that was not the one posted by the agency. And now I have a response from them. It was indeed a trap.
At the end of that post I’m sure I lost you all in an explanation of how it could be a trap, but there’s a way to verify that and the original instructions explicitly said that this verification test had been done, and did not show the unlucky alternative mutation. So we didn’t repeat it.
But ahaaaa. Those original instructions I printed on Nov 8, the day we received the samples. An updated version of the instructions was posted on the QC website on the 16th, changing the key phrase to something ambiguous. The instructions are a 42-page PDF. Okay, only 15 of them are useful once you know how to use their website. But still. Am I going to notice the difference of part of one sentence on one page that says almost the same thing as before? I didn't.
Should I have?
Aw, geeze. Ideally, I should be perfect. This is a goal.
However they decide to score this year’s scheme, they did raise a valid point.
When you use quantitative techniques and find a deletion of a single exon, you should use a different test to back up the conclusion (in addition to repeating the first experiment on a new sample of course). This is because a small mutation could land right on the probe you’re using quantitatively, and prevent it from giving you the signal it should. That’s what happened here. Labs using the MLPA kit from a certain manufacturer should find a diminished signal for exon 11 of MLH1. Labs using other methods should find no difference at all and report the absence of large deletion/duplication mutation detectable by the method used (that there is indeed a point mutation in the sample is not their problem, because the method is not meant to detect them).
In practice, say I use MLPA and find a deletion of exon 11 in the patient. I test another sample from her, same thing. As long as I use the same test, and the makers of the kit don’t change that probe, any DNA with this mutation will give the same result. So I can follow the mutation in the family and give correct advice about digestive cancer risk. Of course it is better to have correctly identified the mutation. With the right mutation in hand I can use any appropriate method and detect it.
The big problem is when a relative in another city has their test done at their local genetics lab. I tell them what I found, and they go look for it using their own method. Their probe is in a different part of exon 11, and naturally it gives a normal result no matter if the person carries the mutation or not.
So for the QC program, the labs that used a different technique and saw nothing, they just say they found nothing within the limits of the test, and they get full marks.
Labs that used MLPA and report the definite deletion of exon 11 get the mutation wrong and get dinged for that.
Labs that used MLPA and report that the proposed deletion of exon 11 needs to be confirmed by a second technique get full marks. This was the point of their trick test. Only: the original instructions specified that a second technique had been done (and for that test to be normal and the MLPA to give a deletion means there is a deletion). Had they not said that, I would have worded my report to allow for this possibility.
The screw-up is in the wording of the original instructions. They said that sequencing gave a normal result, and this is not true. Sequencing is the very test you want to do to uncover this mutation. In fact, we went back and sequenced the case, and yep, there it is. Clear as day.
What I want to say is, these QC tests should be hard. They should give us unusual cases. They should try to trick us. But they can’t give us false information to work from.


Titus said...

Ahaaaa! Knew it.

Bagman and Butler said...

I love to read posts that I can't understand. I thought Exon was an oil company.