probablity value in xmatching

[newBB]

where i can find about how many units must xmatch for a patient wiht mulitple antibodies . and how to explain the 3 cells postiive and 3 cells negative . is there any good website to read about these .

thank you

[Malcolm Needs ☆]

where i can find about how many units must xmatch for a patient wiht mulitple antibodies . and how to explain the 3 cells postiive and 3 cells negative . is there any good website to read about these .

thank you

Sorry newBB, but I don't think that you will find this information anywhere.

The reaons I say this is because each patient is different, and each condition is different. For example, you may have two patients, both of which are pregnant with multiple antibodies, and both requiring a Caesarian Section. But with one it may be her first pregnancy, and two units should suffice (and that is "just in case"), but with the other, with identical antibodies, this may be her fourth or fifth pregnancy, with previous Caesarian Sections and may have Trial of Scar and, possibly, fibroids, and the Obstetrician may want much more blood cover.

[L106]

I think newBB means how do you figure out how many donor units you are going to have to test to find the number of units that the doctor ordered. Is this your question, newBB? If so, I'll go through an example.

Let's say your patient is A Pos and has Anti-c, Anti-Jka, and Anti-K and the surgeon wants 4 donor units crossmatched for surgery. Look up what percentage of the population lacks the corresponding antigens, then multiply those together.

Approx 80% of the population is c Neg (80% = .80 mathematically.)

Let's say that 26% of the population is Jka Neg (26% = .26 mathematically.)

About 90% of the population is K Neg (90%=.90 mathematically.)

So, .80 X .26 X .90 = 0.187 Now, 0.187 (mathematically) equates to 18.7%, so if you tested 100 donor units, 18.7 of them should be compatible with your patient.

18.7 divided by 100 = 4 divided by "X" . In this example, X equals 21.4, so you would probably need to test 21 donor units to find 4 good ones for your pt.

(If you want to take a simpler approach, when you figure out that 18.7% of the donor units should be compatible with your patient, you could round it off to 20%. That means 1 out of 5 donor units should be good. So, if you need 4 good donor units, multiply 4 X 5, which is 20, so you'll probably have to test about 20 donor units to find 4 good one for your patient.)

Now.....Other factors do come into play, such variations in antigen frequencies amoung the different races. It's also helpful to remember that you could test both A Pos and O Pos donor units to find units for this patient. Also, virtually zero percent of Rh Neg donor units will be compatible with your patient, so don't even try them. Rh Neg units are usually rr (dce/dce), and therefore c Pos.

I hope this is of some help to you (and hope it hasn't confused you.)

Donna

Edited April 2, 2011 by L106

[Malcolm Needs ☆]

I think you are probably correct in your assumption about the question, Donna.

[SMW]

I think the example should probably state approximately 20% of the population is c-negative (0.20)?

[Malcolm Needs ☆]

I think the example should probably state approximately 20% of the population is c-negative (0.20)?

It depends very much where you are in the world.

In Asia, for example, the percentage of people who are c negative is much, much higher than 20%.

[adiescast]

Where can one find good statistics for antigen incidence in different world populations? I have seen statistics for what we have here in the U.S. (which doesn't even include all of the different ethnicities we have here - just European based white and West African based black, with an occasional nod to a generic "oriental").

[Malcolm Needs ☆]

Authur Mourant's books, with his (I think) Polish co-authors), neither of which (the books or the authors) I can remember off the top of my head.

I will post tomorrow with the proper citations (given time - I have a couple of telecons tomorrow).

[L106]

I think the example should probably state approximately 20% of the population is c-negative (0.20)?

A big WHOOPS!!! Shame on me! Thank you, SMW, for pointing out my error. (How embarassing!)

The correct calculation that I had intended should have been:

.20 (c-) X .26 (Jka-) X .90 (K-) = .047, meaning that approximately 5% of the donor population should be compatible.

Donna

[Avid123]

And you have to account for the Donor center "cherry picking" their antigen negative needs.

[Malcolm Needs ☆]

And you have to account for the Donor center "cherry picking" their antigen negative needs.

Don't have a clue what you mean!!!!!!!!!!!!!!!!!!!!!!

:haha::haha:

[L106]

I think what Avid123 means is:

Let's say you are trying to find c Negative donor units. 20% of the donor population should be c Negative. However, once in a while we run into a situation where we type 50 donor units for the c antigen, and instead of finding the expected 10 c Negative donor units, we find only one c Negative unit. (ie: The odds are way off.)

What happened is that our blood supplier recently had requests from two other hospitals for several c Negative donor units, which they found and sent to those hospitals. All the c Positive donor units are now available to be sent out for general inventory. Thus, all the units we received happened to be all those c Positive donor units. (Lucky us!)

Donna

[Malcolm Needs ☆]

I was joking Donna - I know what we're like!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

[George Neill]

DiaMed AG in Switzerland (now marketed under Bio-Rad) have a nice system called VIP which is primarily designed as an aid to antibody identification when using DiaMed panels. However also included are some teaching aids and one is a system where you tick the boxes corresponding to the antibodies already determined in the patient sample, (very useful for patients with multiple antibodies) fill in the box with the number of units requested, then click on the calculate box and the system will then display the number of units you would be required to type from the blood bank to achieve that number of antigen negative units. I think the antigen frequencies used for the calculations will be based on European figures.

[Dr. Pepper]

Getting back to the "rule of three", three positive reactions with antigen positive cells and three negative reactions with antigen negative cells to have a valid antibody ID - this comes from an application of Fisher's Exact Method for calculating probability. It's a commonly accepted statistical convention that you should try to get to a 95% confidence level that your conclusion, in this case an antibody ID, is correct, and that the probability (p) of an incorrect conclusion is one chance in 20 or less. A p value of less than 1/20 (i.e. 1/15) allows for too much chance of random association. The AABB Tech Manuals of several editions ago had a nice section on this. Here's the math:

A = # positive reactions seen with cells positive for antigen

B = # positive reactions seen with cells negative for antigen

C = # negative reactions seen with cells positive for antigen

D = # negative reactions seen with cells negative for antigen

N = A+B+C+D

p = probability

! = factorial, the product of all the whole numbers from 1 to the number involved

p = (A+!x(C+D)!x(A+C)!x(B+D)!

N! x A! x B! x C! x D!

While not too many of us actually crunch the numbers, those old Tech Manuals thoughtfully did:

# cells tested Pos Rx Neg Rx p

6 4 2 1/15

6 3 3 1/20

7 5 2 1/21

7 4 3 1/35

8 7 1 1/8

8 6 2 1/26

8 5 3 1/56

8 4 4 1/70

9 8 1 1/9

9 7 2 1/36

9 6 3 1/84

9 5 4 1/123

10 9 1 1/10

10 8 2 1/45

10 7 3 1/120

10 6 4 1/210

10 5 5 1/252

You can switch the numbers in the Pos Rx and Neg Rx columns by the way and the math comes out the same. You can see with 6 cells tested, 4 and 2 gives you a p = 1/15, not good enough. 3 and 3 is p = 1/20, which is OK. The more cells you test, the easier it is to get a p greater than 1 in 20.

These also illustrate a nice practical point: try not to base your ID on the basis of just one cell. You can see with 7 and 1, 8 and 1 etc. your p is 1/8, 1/9 etc., not close to the desired 1/20 or greater. When will this actually happen? How about identifying anti-Cw with just one panel cell reacting, or anti-e with just one cell not reacting. What if the Cw(+) cell was also positive for some other, unlisted, low frequency antigen? What if the e(-) cell never had serum added to it? You should find a second Cw(+) cell or e(-) cell to test to confirm your ID. (More cells e- will also enable you to eliminate more of other potential antibodies.)

Another point is that if you have multiple antibodies, you can't really count the cells that have multiple antigens on them because you can't tell what antibodies are causing those cells to react. If you think you have anti-D+C, you could be basing that conclusion on just one D+C- cell and one D-C+ cell that are positive. I think it's best to find one more each of those cells to confirm your ID.

This is an ideal world where you have plenty of cells of the desired phenotypes to test, which may not always be the case. We are always limited by what we have to work with, but that's what you should try for. Sorry to drone on here, but you asked.

[Dr. Pepper]

[ATTACH]495[/ATTACH]Literally. Let me try to add an attachment of the p value chart that's easier to read than what got posted.

s method.doc

[AMcCord]

And you have to account for the Donor center "cherry picking" their antigen negative needs.

So true!

[Malcolm Needs ☆]

So true!

As if we would dare!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

[galvania]

[ATTACH]495[/ATTACH]Literally. Let me try to add an attachment of the p value chart that's easier to read than what got posted.

Thank you Dr. Pepper. In all the many many years I have been working in transfusion, that is the first time I have ever seen that explained so neatly.