Antibody confirmation testing

[JerryB]

Currently, for antibody confirmation, we use the Fisher exact method (3 homozygous reactive & 3 nonreactive cells). Wth the influx of generalists in our lab, we are forced to revisit this approach. Many of the newbies are advocating confirmation based on 1 homozygous reactive and a few (2) nonreactive cells. I can't find any references for this approach. I'd like to know what approach others are using (referenced if possible) before I make any kind of decision.

[Lcsmrz]

Using Fisher's method, you are getting a statistical probability that the pattern is the antibody specificity. Using less than 3 pos/3 neg just means you are altering your confidence to something other than 95%, maybe to the point that I would have trouble sleeping.

At my facility, anything less than 3 pos/3 neg (with a few rare exceptions) leads to a search for more cells or sending the sample to the reference lab. But more important is eliminating all other significant possibilities from the mix, and we add the patient's antigen typing just to make sure I'm not fooled again. The AHG crossmatch confirms that a low freq isn't hiding in there somewhere.

Panels and proper follow-up is a part of every educational opportunity here, to make sure everyone understands what is expected and why they are doing it. I purposely don't make them easy, to make sure everyone has to think before getting the right answer.

[Mabel Adams]

What do you mean by "antibody confirmation"? If you mean a patient that is known to have anti-Fya returns and you are doing another panel then the generalists are right; if this is the first time an antibody is ID'd then, 3 & 3 or 5 & 2 give you the good statistics for ID.

If the antibody was previously known then you have a different issue about what you require to rule out (as oppososed to ID) additional antibodies--a bigger concern since you will always be giving Fya neg units anyway. I like to run enough cells that react negative so that I can rule out new antibodies with the same sensitivity as I have with my usual antibody screen--i.e. double-dose cells for all the antigens that usually have double-dose expression on my screening cells. C & E in the presence of anti-D are specific exceptions. If very few cells are expected to react negative, obviously judgement comes into play, but usually the generalist have handed it off to a BBer by then.

Sorry I ramble on so.

[jlemmons]

As an instructor in an MLT program, I have taught my students to use 1 homozygous or 3 heterozygous (except on Kells, Kidds, and MNS) to RULE OUT but they must use 3 positive and 3 negative to PROVE the antibody. These are two different things. You must rule out all other antibodies but you must also prove the antibody you believe that you have.

[JerryB]

What do you mean by "antibody confirmation"? If you mean a patient that is known to have anti-Fya returns and you are doing another panel then the generalists are right; if this is the first time an antibody is ID'd then, 3 & 3 or 5 & 2 give you the good statistics for ID.

If the antibody was previously known then you have a different issue about what you require to rule out (as oppososed to ID) additional antibodies--a bigger concern since you will always be giving Fya neg units anyway. I like to run enough cells that react negative so that I can rule out new antibodies with the same sensitivity as I have with my usual antibody screen--i.e. double-dose cells for all the antigens that usually have double-dose expression on my screening cells. C & E in the presence of anti-D are specific exceptions. If very few cells are expected to react negative, obviously judgement comes into play, but usually the generalist have handed it off to a BBer by then.

Sorry I ramble on so.

Mabel, thanks for the excellent info. Per our SOP, we use panels to confirm all "first time" presented postive screens by using Fishers Exact method to confirm an antibody. Upon subsequent presentations (screen still pos), we then rule out all other common rbc antigens on our panel choosing cells neg for the antibody previously presented. One of my chief issues as well is some of these generalist work "irregular tours" providing sole BB coverage-not my reccommendation. They seem to take issue with the initial confirmation process using 3 and 3. They have worked at other places using less cells in the intial confirmation process. I thankfully haven't! My concern is what's best for the patient and as a senior tech, although I want to consider the wishes of these generalist, my first loyalty is towards the patient's in this hospital. If it ain't broke.....

[David Saikin]

I think AABB has rewritten the rule on the number of cells needed to confirm. It is now 2 pos and 2 neg. This was done to accomodate IRLs especially for those abs/ags that are rare (they may have only 1 or 2 cells that can be run). I still use the 3/3 scenario. As Mabel has pointed out, this is only necessary for a first time id . . . Once I know an ab is present, I just run cells to r/o any additional new sensitizations. I also try to run a heterozygous ag+ cell for the ab I know to be present (just to see that it is still reactive).

[Mabel Adams]

I once listened to a teleconference by Teresa (sp) Harris that said that 2 pos and 5 neg gives essentially the same statistical accuracy as 3 pos and 3 neg. Somewhere more recently I heard it said that the same is true for 5 pos and 2 neg. I have not seen actual statistics.

[Lcsmrz]

Per Fisher's Test, 2 pos / 5 neg equals 3 pos / 3 neg in statistical probability. The latter is just easier to remember.

[JerryB]

We have decided to compromise and use 2 and 2 as an acceptable alternative per AABB Technical Manual (16th ed).

[Eagle Eye]

we use 2pos and 2 neg...we need two homozygous to rule out any clinically significant antibodies(mostly RH, Kell, Kid, Duffy, S and s).

[ckcheng]

p values for antibody identification. 3pos3neg is based on the null hypothesis in Fisher's exact test of p value of 0.05.

Harris & Hochman use an alternative method to calculate p value which is on the basis of antigen frequencies. 2pos2neg of different antigens have different p values, but close to 0.05. For further information, please read the article ''Misinterpretation & Misapplication of p values in antibody identification: the lack of value of a p value' written by MH Kanter, G Poole & G Garratty in Transfusion Vol 37 Aug 1997 .

Hope that helps.

CK Cheng, MSc, SBB(ASCP), CQA(ASQ)

Hong Kong

Mar 5, 2009

[NedB]

To go back to the first post in this thread, JerryB asks about using 3 homozygous cells. The Fisher statistic refers to 'heterozygous' or better expressed as single dose antigens. A double dose cell or homozygous cell would be counted as 2 expressions of the antigen. We use the Fisher statistic, but always include a double dose cell. We make exception for K because of its rarity in panel cells. This system has worked well over the years, allowing us to 'prove' such rare antibodies and Anti-Lua, Anti-Kpa and Anti-Kpb.